Intermittent drug techniques for schizophrenia

  • Review
  • Intervention

Authors


Abstract

Background

Antipsychotic medication is considered the mainstay of treatment for schizophrenia and is generally regarded as highly effective, especially in controlling positive symptoms. However, long-term antipsychotic exposure has been associated with a range of adverse effects, including extra-pyramidal symptoms (EPS), neuroleptic malignant syndrome (NMS), tardive dyskinesia and death. Intermittent drug techniques refers to the 'use of medication only during periods of incipient relapse or symptom exacerbation rather than continuously'. The aim is to reduce the risk of typical adverse effects of antipsychotics by 'reducing long-term medication exposure for patients who are receiving maintenance treatment while limiting the risk of relapse', with a further goal of improving social functioning resulting from the reduction of antipsychotic-induced side effects

Objectives

To review the effects of different intermittent drug techniques compared with maintenance treatment in people with schizophrenia or related disorders.

Search methods

We searched The Cochrane Schizophrenia Group Trials Register (April 2012) and supplemented this by contacting relevant study authors, handsearching relevant intermittent drug treatment articles and manually searching reference lists.

Selection criteria

All randomised controlled trials (RCTs) that compared intermittent drug techniques with standard maintenance therapy for people with schizophrenia. Primary outcomes of interest were relapse and hospitalisation.

Data collection and analysis

At least two review authors selected trials, assessed quality and extracted data. We calculated risk ratios (RR) and 95% confidence intervals (CI) of homogeneous dichotomous data and estimated the 95% confidence interval (CI) around this. For non-skewed continuous endpoint data extracted from valid scales, we estimated mean difference (MD) between groups with a 95% CI. Where data displayed heterogeneity, these were analysed using a random-effects model. Skewed data are presented in tables. We assessed overall quality for clinically important outcomes using the GRADE approach.

Main results

Of 241 records retrieved by the search, 17 trials conducted between 1961 and 2011, involving 2252 participants with follow-up from six weeks to two years, were included. Homogenous data demonstrated that instances of relapse were significantly higher in people receiving any intermittent drug treatment in the long term (n = 436, 7 RCTs, RR 2.46, 95% CI 1.70 to 3.54, moderate quality evidence). Intermittent treatment was shown to be more effective than placebo, however, and demonstrated that significantly less people receiving intermittent antipsychotics experienced full relapse by medium term (n = 290, 2 RCTs, RR 0.37, 95% CI 0.24 to 0.58, very low quality evidence). Hospitalisation rates were higher for people receiving any intermittent drug treatment by long term (n = 626, 5 RCTs, RR 1.65, 95% CI 1.33 to 2.06, moderate quality evidence). Results demonstrated little difference in instances of tardive dyskinesia in groups with any intermittent drug technique versus maintenance therapy, with equivocal results (displaying slight heterogeneity) at long term (n = 165, 4 RCTs, RR 1.15, 95% CI 0.58 to 2.30, low quality evidence).

Authors' conclusions

Results of this review support the existing evidence that intermittent antipsychotic treatment is not as effective as continuous, maintained antipsychotic therapy in preventing relapse in people with schizophrenia. More research is needed to assess any potential benefits or harm of intermittent treatment regarding adverse effects typically associated with maintained antipsychotic treatment, as well as any cost-effectiveness of this experimental treatment.

Résumé scientifique

Techniques d'administration intermittente de médicaments contre la schizophrénie

Contexte

Les médicaments antipsychotiques sont considérés comme la base du traitement contre la schizophrénie et sont généralement perçus comme étant extrêmement efficaces, en particulier pour contrôler les symptômes positifs. Cependant, une exposition à long terme aux antipsychotiques a été associée à divers effets néfastes, notamment des symptômes extrapyramidaux (SEP), un syndrome malin des neuroleptiques (SMN), une dyskinésie tardive et le décès. Les techniques d'administration intermittente de médicaments font référence à « l'utilisation de médicaments uniquement pendant les périodes de début de récidive ou d'exacerbation des symptômes plutôt que de manière continue. » L'objectif est de réduire le risque d'effets indésirables classiques des antipsychotiques en « réduisant l'exposition aux médicaments à long terme pour les patients qui reçoivent un traitement d'entretien, tout en limitant le risque de récidive », avec comme but supplémentaire d'améliorer le fonctionnement social par la réduction des effets secondaires induits par les antipsychotiques.

Objectifs

Examiner les effets de différentes techniques d'administration intermittente de médicaments comparées au traitement d'entretien chez les personnes atteintes de schizophrénie ou de troubles connexes.

Stratégie de recherche documentaire

Nous avons effectué des recherches dans le registre des essais du groupe Cochrane sur la schizophrénie (avril 2012) et avons complété ces recherches en contactant les auteurs des études pertinentes, en effectuant des recherches manuelles d'articles pertinents sur les traitements médicamenteux intermittents et des recherches manuelles dans les listes bibliographiques.

Critères de sélection

Tous les essais contrôlés randomisés (ECR) qui comparaient des techniques d'administration intermittente de médicaments à un traitement d'entretien standard pour les personnes atteintes de schizophrénie. Les principaux critères de jugement étaient la récidive et l'hospitalisation.

Recueil et analyse des données

Au moins deux auteurs ont sélectionné les essais, évalué leur qualité et extrait les données. Nous avons calculé les risques relatifs (RR) et les intervalles de confiance (IC) à 95 % des données dichotomiques homogènes et estimé l'intervalle de confiance (IC) autour de cela. Pour les données continues non biaisées issues d’échelles valides, nous avons estimé la différence moyenne (DM) entre les groupes avec un IC à 95 %. Lorsque les données ont montré une hétérogénéité, elles ont été analysées à l'aide d'un modèle à effets aléatoires. Les données biaisées sont présentées dans des tableaux. Nous avons évalué la qualité globale pour les critères de jugement cliniquement importants en utilisant l'approche GRADE.

Résultats principaux

Sur les 241 articles obtenus par les recherches, 17 essais conduits entre 1961 et 2011 et portant sur 2 252 participants avec un suivi de six semaines à deux ans, ont été inclus. Les données homogènes ont démontré que les cas de récidive étaient significativement plus nombreux chez les personnes recevant un traitement médicamenteux intermittent à long terme (n = 436, 7 ECR, RR 2,46, IC à 95 % 1,70 à 3,54, preuves de qualité modérée). Le traitement intermittent s'est toutefois révélé plus efficace que le placebo et a démontré qu'un nombre significativement moins important de personnes recevant des antipsychotiques par intermittence connaissaient une récidive totale à moyen terme (n = 290, 2 ECR, RR 0,37, IC à 95 % 0,24 à 0,58, preuves de qualité très médiocre). Les taux d'hospitalisation ont été plus élevés pour les personnes recevant un traitement médicamenteux intermittent à long terme (n = 626, 5 ECR, RR 1,65, IC à 95 % 1,33 à 2,06, preuves de qualité modérée). Les résultats ont démontré peu de différences dans les cas de dyskinésie tardive dans les groupes avec une technique d'administration intermittente de médicaments versus traitement d'entretien, avec des résultats équivoques (présentant une légère hétérogénéité) à long terme (n = 165, 4 ECR, RR 1,15, IC à 95 % 0,58 à 2,30, preuves de qualité médiocre).

Conclusions des auteurs

Les résultats de cette revue soutiennent les preuves existantes indiquant que le traitement antipsychotique intermittent n'est pas aussi efficace que le traitement continu, permanent, aux antipsychotiques pour prévenir la récidive chez les personnes atteintes de schizophrénie. Des recherches supplémentaires sont nécessaires pour évaluer les bénéfices ou les préjudices potentiels du traitement intermittent concernant les effets indésirables classiquement associés au traitement d'entretien aux antipsychotiques, ainsi que le rapport coût-efficacité de ce traitement expérimental.

Plain language summary

Intermittent drug techniques for schizophrenia

Antipsychotic medication is the main treatment for schizophrenia and helps people cope with positive symptoms such as hearing voices, seeing things and having strange beliefs. However, long-term exposure to these drugs has been associated with serious side effects, such as: weight gain; uncontrollable shaking of the head, body or hands; tremors; muscle stiffness; difficulties with walking and balance; sleepiness or apathy; and even death. Some people stop taking their medication as these side effects limit people’s quality of life. Not taking medication can be a contributory factor that leads to relapse and hospitalisation. Against this backdrop, there is cause to consider the role of intermittently administering antipsychotic medication compared to the continuous use of antipsychotic medication.

Intermittent drug techniques refer to the use of medication only during periods close to relapse of symptoms rather than continuously taking antipsychotic drugs all the time. Intermittent drug techniques include: prodrome-based intervention (which assesses the risk or early stage of relapse); crisis intervention during an acute episode or downturn in mental health; gradually increased drug-free periods; and drug holidays. The aim is to reduce exposure to drugs and decrease side effects.

This review assesses different intermittent drug techniques compared with maintenance treatment in people with schizophrenia or related disorders. Seventeen studies with 2252 participants compared intermittent drug techniques with standard maintenance on medication. Relapse was significantly higher in people receiving intermittent drug treatment. Hospitalisation was higher for people receiving intermittent drug treatment.

Results suggest that intermittent treatment is not as effective as continuous or maintained treatment in preventing relapse. Although information favours maintenance and continuous treatment, this is not always the case in real settings, where people may stop their medication due to debilitating side effects that affect their quality of life. More research is needed to assess any potential benefits or harm of intermittent treatment, particularly regarding the side effects commonly associated with maintained antipsychotic treatment. There was no exploration of economic/money savings, specifically relating to the potential cost-effectiveness of intermittent techniques.

Until further evidence is available concerning the potential benefits or harms of intermittent treatment, managers, psychiatrists and policy makers should consider it an experimental therapy.

This plain language summary has been written by a consumer Ben Gray, Service User and Service User Expert, Rethink Mental Illness.

Résumé simplifié

Techniques d'administration intermittente de médicaments contre la schizophrénie

Les médicaments antipsychotiques représentent le principal traitement contre la schizophrénie et aident les personnes à faire face aux symptômes positifs, tels que le fait d'entendre des voix, de voir des choses et de croire en des choses étranges. Cependant, l'exposition à long terme à ces médicaments a été associée à de graves effets secondaires, tels que : la prise de poids ; une agitation incontrôlable de la tête, du corps ou des mains ; des tremblements ; une raideur des muscles ; des difficultés à marcher et à garder l'équilibre ; une somnolence ou une apathie ; et même le décès. Certaines personnes arrêtent de prendre leurs médicaments, car ces effets secondaires limitent la qualité de vie des patients. L'absence de prise des médicaments peut être un facteur contribuant à une récidive et à une hospitalisation. Dans ce contexte, il y a lieu de prendre en compte le rôle de l'administration intermittente de médicaments antipsychotiques comparé à l'utilisation continue de ces médicaments.

Les techniques d'administration intermittente de médicaments font référence à l'utilisation de médicaments uniquement pendant les périodes de proximité de la réapparition des symptômes plutôt que la prise continue de médicaments antipsychotiques en permanence. Les techniques d'administration intermittente de médicaments comprennent : l'intervention prodromique (qui évalue le risque ou le stade précoce de la récidive) ; l'intervention de crise au cours d'un épisode aigu ou d'une baisse de la santé mentale ; une augmentation progressive des périodes sans médicaments ; et les fenêtres thérapeutiques. L'objectif est de réduire l'exposition aux médicaments et de diminuer les effets secondaires.

Cette revue évalue différentes techniques d'administration intermittente de médicaments comparées au traitement d'entretien chez les personnes atteintes de schizophrénie ou de troubles connexes. Dix-sept études, portant sur 2 252 participants, ont comparé des techniques d'administration intermittente de médicaments au maintien standard du traitement médicamenteux. La récidive a été significativement plus importante chez les personnes recevant un traitement médicamenteux intermittent. L'hospitalisation a été plus importante pour les personnes recevant un traitement médicamenteux intermittent.

Les résultats suggèrent que le traitement intermittent n'est pas aussi efficace que le traitement continu ou permanent pour prévenir la récidive. Bien que les informations soient en faveur du traitement permanent et continu, cela n'est pas toujours le cas dans la réalité, car les personnes peuvent arrêter leur traitement en raison d'effets secondaires invalidants qui affectent leur qualité de vie. Des recherches supplémentaires sont nécessaires pour évaluer les bénéfices ou les préjudices potentiels du traitement intermittent, en particulier concernant les effets secondaires fréquemment associés au traitement antipsychotique permanent. Il n'y a pas eu d'examen des économies pécuniaires/d'argent, en particulier concernant le rapport coût-efficacité potentiel des techniques intermittentes.

Jusqu'à ce que de nouvelles preuves soient disponibles concernant les bénéfices ou les préjudices potentiels du traitement intermittent, les cadres, les psychiatres et les décideurs politiques doivent le considérer comme un traitement expérimental.

Ce résumé en langage simplifié a été rédigé par un consommateur, Ben Gray, Bénéficiaire du service et Expert auprès des bénéficiaires du service, Rethink Mental Illness.

Notes de traduction

Traduit par: French Cochrane Centre 4th September, 2013
Traduction financée par: Pour la France : Minist�re de la Sant�. Pour le Canada : Instituts de recherche en sant� du Canada, minist�re de la Sant� du Qu�bec, Fonds de recherche de Qu�bec-Sant� et Institut national d'excellence en sant� et en services sociaux.

Summary of findings(Explanation)

Summary of findings for the main comparison. ANY INTERMITTENT DRUG TECHNIQUE compared with MAINTENANCE THERAPY for schizophrenia
  1. 1 Assumed risk: calculated from the included studies - presents 3 risks based on the control group risks - 'moderate' risk equates with that of control group (14%).
    2 Imprecision: 'serious' - small sample size - confidence intervals for best estimate of effect include both no effect and appreciable benefit/ harm
    3 Assumed risk: calculated from the included studies - presents 3 risks based on the control group risks - 'moderate' risk equates with that of control group (26.1%).
    4 Indirectness: 'serious' - intermittent drug techniques differ between studies, with some instead examining 'targeted intervention' or 'dose reduction and family treatment'
    5 Denotes endpoint data
    6 Risk of bias: 'serious' - only 30% of included studies adequately described randomisation methods, the remaining 70% provided no description of randomisation methods
    7 Denotes endpoint and change data.
    8 Risk of bias: 'serious' - 100% of included studies provided no description of randomisation methods
    9 Assumed risk: calculated from the included studies - presents 3 risks based on the control group risks - 'moderate' risk equates with that of control group (17.2%).
    10 Assumed risk: calculated from the included studies - presents 3 risks based on the control group risks - 'moderate' risk equates with that of control group (27.9%).
    11 Indirectness: 'serious' - intermittent drug techniques differ between studies, with some instead examining 'targeted intervention' or 'guided discontinuation'

ANY INTERMITTENT DRUG TECHNIQUE compared with MAINTENANCE THERAPY for schizophrenia
Patient or population: patients with schizophrenia
Settings: inpatients; outpatients; multi-centre (Germany, Netherlands, UK, US)
Intervention: ANY INTERMITTENT DRUG TECHNIQUE
Comparison: MAINTENANCE THERAPY
OutcomesIllustrative comparative risks* (95% CI)Relative effect
(95% CI)
No of Participants
(studies)
Quality of the evidence
(GRADE)
Comments
Assumed riskCorresponding risk
MAINTENANCE THERAPY ANY INTERMITTENT DRUG TECHNIQUE
Relapse: long term (+ 26 weeks)
as defined in each study
Follow-up: mean 17 months
Low1 RR 2.46
(1.70 to 3.54)
436
(7 studies)
⊕⊕⊕⊝
moderate 2
 
0 per 1000 0 per 1000
(0 to 0)
Moderate1
150 per 1000 369 per 1000
(255 to 531)
High1
300 per 1000 738 per 1000
(510 to 1000)
Hospitalisation: long term (+ 26 weeks)
as defined in each study
Follow-up: mean 24 months
Low3 RR 1.65
(1.33 to 2.06)
626
(5 studies)
⊕⊕⊕⊝
moderate 4
 
50 per 1000 82 per 1000
(66 to 103)
Moderate3
300 per 1000 495 per 1000
(399 to 618)
High3
500 per 1000 825 per 1000
(665 to 1000)
Global state: average score: long term (+ 26 weeks)
Global Assessment Scale (GAS). Scale from: 0 to 100.
Follow-up: mean 22 months
The mean global state: average score: long term (+ 26 weeks) in the control groups was
59 points 5
The mean global state: average score: long term (+26 weeks) in the intervention groups was
1.32 higher
(2.75 lower to 5.39 higher)
 133
(3 studies)
⊕⊕⊝⊝
low 2,6
 
Mental state: average score: long term (+ 26 weeks)
Brief Psychaitric Rating Scale (BPRS). Scale from: 0 to 126.
Follow-up: mean 21 months
The mean mental state: average score: long term (+ 26 weeks) in the control groups was
11.95 points 7
The mean mental state: average score: long term (+26 weeks) in the intervention groups was
0.1 higher
(0.32 lower to 0.53 higher)
 77
(2 studies)
⊕⊕⊝⊝
low 2,8
 
Adverse effects: tardive dyskinesia: by long term (+26 weeks)
as defined in each study
Follow-up: mean 15 months
Low9 RR 1.15
(0.58 to 2.3)
165
(4 studies)
⊕⊕⊝⊝
low 2,8
 
0 per 1000 0 per 1000
(0 to 0)
Moderate9
200 per 1000 230 per 1000
(116 to 460)
High9
600 per 1000 690 per 1000
(348 to 1000)
Economic outcomes: cost effectiveness
as defined in each study
See commentSee commentNot estimable0
(0)
See commentNo study reported this outcome
Leaving the study early/ loss to follow-up: by long term (+ 26 weeks)
as defined in each study
Follow-up: mean 19 months
Low10 RR 1.63
(1.23 to 2.15)
996
(10 studies)
⊕⊕⊝⊝
low 2,11
 
100 per 1000 163 per 1000
(123 to 215)
Moderate10
400 per 1000 652 per 1000
(492 to 860)
High10
700 per 1000 1000 per 1000
(861 to 1000)
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: Confidence interval; RR: Risk ratio;
GRADE Working Group grades of evidence
High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low quality: We are very uncertain about the estimate.

Summary of findings 2 INTERMITTENT (EARLY-BASED) compared with MAINTENANCE THERAPY for schizophrenia

Summary of findings 2. INTERMITTENT (EARLY-BASED) compared with MAINTENANCE THERAPY for schizophrenia
  1. 1 Assumed risk: calculated from the included studies - presents 3 risks based on the control group risks - 'moderate' risk equates with that of control group (16.7%).
    2 Risk of bias: 'serious' - only 50% of included studies adequately described randomisation methods, the remaining 50% provided no description of randomisation methods
    3 Imprecision: 'serious' - small sample size - confidence intervals for best estimate of effect include both no effect and appreciable benefit/ harm
    4 Assumed risk: calculated from the included studies - presents 3 risks based on the control group risks - 'moderate' risk equates with that of control group (25.9%).
    5 Risk of bias: 'serious' - only 40% of included studies adequately described randomisation methods, the remaining 60% provided no description of randomisation methods
    6 Denotes endpoint data
    7 Denotes change data
    8 Risk of bias: 'serious' - study provided no description of randomisation methods
    9 Mean baseline risk presented for single study.
    10 Assumed risk: calculated from the included studies - presents 3 risks based on the control group risks - 'moderate' risk equates with that of control group (33.9%).
    11 Risk of bias: 'serious' - only 20% of included studies adequately described randomisation methods, the remaining 80% provided no description of randomisation methods

INTERMITTENT (EARLY-BASED) compared with MAINTENANCE THERAPY for schizophrenia
Patient or population: patients with schizophrenia
Settings: outpatients (Germany, UK, US)
Intervention: INTERMITTENT (EARLY-BASED)
Comparison: MAINTENANCE THERAPY
OutcomesIllustrative comparative risks* (95% CI)Relative effect
(95% CI)
No of Participants
(studies)
Quality of the evidence
(GRADE)
Comments
Assumed riskCorresponding risk
MAINTENANCE THERAPY INTERMITTENT (EARLY-BASED)
Relapse: long term (+ 26 weeks)
as defined in each study
Follow-up: mean 24 months
Low1 RR 2.33
(1.32 to 4.12)
155
(2 studies)
⊕⊕⊝⊝
low 2,3
 
0 per 1000 0 per 1000
(0 to 0)
Moderate1
150 per 1000 349 per 1000
(198 to 618)
High1
300 per 1000 756 per 1000
(396 to 1000)
Hospitalisation: long term (+ 26 weeks)
as defined in each study
Follow-up: mean 24 months
Low4 RR 1.16
(1.33 to 2.08)
625
(5 studies)
⊕⊕⊕⊝
moderate 5
 
100 per 1000 166 per 1000
(133 to 208)
Moderate4
300 per 1000 498 per 1000
(399 to 624)
High4
600 per 1000 996 per 1000
(798 to 1000)
Global state: average score: long term (+ 26 weeks)
Global Assessment Scale (GAS). Scale from: 0 to 100.
Follow-up: mean 24 months
The mean global state: average score: long term (+ 26 weeks) in the control groups was
47.5 points 6
The mean global state: average score: long term (+26 weeks) in the intervention groups was
0.99 higher
(4.24 lower to 6.22 higher)
 82
(2 studies)
⊕⊕⊝⊝
low 2,3
 
Mental state: average score: long term (+ 26 weeks)
Brief Psychiatric Rating Scale (BPRS). Scale from: 0 to 126.
Follow-up: mean 24 months
The mean mental state: average score: long term (+ 26 weeks) in the control groups was
1.8 points 7
The mean mental state: average score: long term (+26 weeks) in the intervention groups was
0.1 higher
(0.33 lower to 0.53 higher)
 26
(1 study)
⊕⊝⊝⊝
very low 3,8
 
Adverse effects: tardive dyskinesia: long term (+26 weeks)
as defined in each study
Follow-up: mean 24 months
444 per 1000 9 249 per 1000
(84 to 756)
RR 0.56
(0.19 to 1.7)
30
(1 study)
⊕⊕⊝⊝
low 3,8
 
Economic outcomes: cost effectiveness: long term (+ 26 weeks)
as defined in each study
See commentSee commentNot estimable0
(0)
See commentNo study reported this outcome
Leaving the study early/loss to follow-up: long term (+ 26 weeks)
as defined in each study
Follow-up: mean 24 months
Low RR 1.67
(1.17 to 2.37)
562
(5 studies)
⊕⊕⊝⊝
low 3,11
 
200 per 1000 10 334 per 1000
(234 to 474)
Moderate
400 per 1000 10 668 per 1000
(468 to 948)
High
600 per 1000 10 1000 per 1000
(702 to 1000)
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: Confidence interval; RR: Risk ratio;
GRADE Working Group grades of evidence
High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low quality: We are very uncertain about the estimate.

Summary of findings 3 INTERMITTENT (CRISIS INTERVENTION) compared with MAINTENANCE THERAPY for schizophrenia

Summary of findings 3. INTERMITTENT (CRISIS INTERVENTION) compared with MAINTENANCE THERAPY for schizophrenia
  1. 1 Mean baseline risk presented for single study.
    2 Risk of bias: 'serious' - study provided no description of randomisation methods
    3 Imprecision: small sample size (n = 237) from single study.

INTERMITTENT (CRISIS INTERVENTION) compared with MAINTENANCE THERAPY for schizophrenia
Patient or population: patients with schizophrenia
Settings: outpatients (Germany)
Intervention: INTERMITTENT (CRISIS INTERVENTION)
Comparison: MAINTENANCE THERAPY
OutcomesIllustrative comparative risks* (95% CI)Relative effect
(95% CI)
No of Participants
(studies)
Quality of the evidence
(GRADE)
Comments
Assumed riskCorresponding risk
MAINTENANCE THERAPY INTERMITTENT (CRISIS INTERVENTION)
Relapse: long term (+ 26 weeks)
as defined in each study
See commentSee commentNot estimable0
(0)
See commentNo study reported provided useable data for this outcome by long term
Hospitalisation: long term (+ 26 weeks)
as defined in each study
See commentSee commentNot estimable0
(0)
See commentNo study reported provided useable data for this outcome by long term
Global state: average score: long term (+ 26 weeks)
Global Assessment Scale (GAS). Scale from: 0 to 100.
See commentSee commentNot estimable0
(0)
See commentNo study reported provided useable data for this outcome by long term
Mental state: average score: long term (+ 26 weeks)
Brief Psychiatric Rating Scale (BPRS). Scale from: 0 to 126.
See commentSee commentNot estimable0
(0)
See commentNo study reported provided useable data for this outcome by long term
Adverse effects: tardive dyskinesia: long term (+ 26 weeks)
as defined in each study
See commentSee commentNot estimable0
(0)
See commentNo study reported provided useable data for this outcome by long term
Economic outcomes: cost effectiveness: long term (+ 26 weeks)
as defined in each study
See commentSee commentNot estimable0
(0)
See commentNo study reported this outcome
Leaving the study early/ loss to follow-up: long term (+ 26 weeks)
as defined in each study
Follow-up: mean 24 months
426 per 1000 1 669 per 1000
(524 to 852)
RR 1.57
(1.23 to 2)
237
(1 study)
⊕⊝⊝⊝
very low 2,3
 
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: Confidence interval; RR: Risk ratio;
GRADE Working Group grades of evidence
High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low quality: We are very uncertain about the estimate.

Summary of findings 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) compared with MAINTENANCE THERAPY for schizophrenia

Summary of findings 4. INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) compared with MAINTENANCE THERAPY for schizophrenia
  1. 1 Assumed risk: calculated from the included studies - presents 3 risks based on the control group risks - 'moderate' risk equates with that of control group (12.7%).
    2 Risk of bias: 'serious' - 66% of included studies adequately described randomisation methods, the remaining 34% provided no description of randomisation methods
    3 Imprecision: 'serious' - small sample size - confidence intervals for best estimate of effect include both no effect and appreciable benefit/ harm
    4 Denotes endpoint score.
    5 Risk of bias: 'serious' - study provided no description of randomisation methods
    6 Single included study reported no instances of tardive dyskinesia (0%).
    7 Assumed risk: calculated from the included studies - presents 3 risks based on the control group risks - 'moderate' risk equates with that of control group (17.2%).

INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) compared with MAINTENANCE THERAPY for schizophrenia
Patient or population: patients with schizophrenia
Settings: inpatients; outpatients (Germany, Netherlands)
Intervention: INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS)
Comparison: MAINTENANCE THERAPY
OutcomesIllustrative comparative risks* (95% CI)Relative effect
(95% CI)
No of Participants
(studies)
Quality of the evidence
(GRADE)
Comments
Assumed riskCorresponding risk
MAINTENANCE THERAPY INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS)
Relapse: long term (+ 26 weeks)
as defined in each study
Follow-up: mean 18 months
Low1 RR 2.76
(1.63 to 4.67)
219
(3 studies)
⊕⊕⊝⊝
low 2,3
 
0 per 1000 0 per 1000
(0 to 0)
Moderate1
100 per 1000 276 per 1000
(163 to 467)
High1
300 per 1000 828 per 1000
(489 to 1000)
Hospitalisation: long term (+ 26 weeks)
as defined in each study
See commentSee commentNot estimable0
(0)
See commentNo study reported provided useable data for this outcome by long term
Global state: average score: long term (+26 weeks)
Global Assessment Scale (GAS). Scale from: 0 to 100.
Follow-up: mean 18 months
The mean global state: average score: long term (+26 weeks) in the control groups was
82.17 points 4
The mean global state: average score: long term (+ 26 weeks) in the intervention groups was
1.83 higher
(4.66 lower to 8.32 higher)
 51
(1 study)
⊕⊝⊝⊝
very low 3,5
 
Mental state: average score: long term (+ 26 weeks)
Brief Psychiatric Rating Scale (BPRS). Scale from: 0 to 126.
Follow-up: mean 18 months
The mean mental state: average score: long term (+26 weeks) in the control groups was
22.1 points 4
The mean mental state: average score: long term (+ 26 weeks) in the intervention groups was
0.2 higher
(2.77 lower to 3.17 higher)
 51
(1 study)
⊕⊝⊝⊝
very low 3,5
 
Adverse effects: tardive dyskinesia: long term (+ 26 weeks)
as defined in each study
Follow-up: mean 18 months
See comment6See commentNot estimable85
(1 study)
⊕⊝⊝⊝
very low 3,5
 
Economic outcomes: cost effectiveness: long term (+ 26 weeks)
as defined in each study
See commentSee commentNot estimable0
(0)
See commentNo study reported this outcome
Leaving the study early/ loss to follow-up: long term (+ 26 weeks)
as defined in each study
Follow-up: mean 18 months
Low7 RR 2.12
(0.7 to 6.37)
257
(3 studies)
⊕⊝⊝⊝
very low 2,3
 
0 per 1000 0 per 1000
(0 to 0)
Moderate7
200 per 1000 424 per 1000
(140 to 1000)
High7
500 per 1000 1000 per 1000
(350 to 1000)
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: Confidence interval; RR: Risk ratio;
GRADE Working Group grades of evidence
High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low quality: We are very uncertain about the estimate.

Summary of findings 5 INTERMITTENT (DRUG HOLIDAY) compared with MAINTENANCE THERAPY for schizophrenia

Summary of findings 5. INTERMITTENT (DRUG HOLIDAY) compared with MAINTENANCE THERAPY for schizophrenia
  1. 1 Assumed risk: calculated from the included studies - presents 3 risks based on the control group risks - 'moderate' risk equates with that of control group (12.1%).
    2 Risk of bias: 'serious' - 100% of included studies provided no description of randomisation methods
    3 Imprecision: 'serious' - small sample size - confidence intervals for best estimate of effect include both no effect and appreciable benefit/ harm
    4 Mean baseline risk presented from single study.
    5 No explanation was provided
    6 Assumed risk: calculated from the included studies - presents 3 risks based on the control group risks - 'moderate' risk equates with that of control group (27.3%).

INTERMITTENT (DRUG HOLIDAY) compared with MAINTENANCE THERAPY for schizophrenia
Patient or population: patients with schizophrenia
Settings: inpatients; outpatients (Canada, UK)
Intervention: INTERMITTENT (DRUG HOLIDAY)
Comparison: MAINTENANCE THERAPY
OutcomesIllustrative comparative risks* (95% CI)Relative effect
(95% CI)
No of Participants
(studies)
Quality of the evidence
(GRADE)
Comments
Assumed riskCorresponding risk
MAINTENANCE THERAPY INTERMITTENT (DRUG HOLIDAY)
Relapse: long term (+ 26 weeks)
as defined in each study
Follow-up: mean 9.5 months
Low1 RR 1.70
(0.54 to 5.38)
61
(2 studies)
⊕⊝⊝⊝
very low 2,3
 
0 per 1000 0 per 1000
(0 to 0)
Moderate1
150 per 1000 255 per 1000
(81 to 807)
High1
200 per 1000 340 per 1000
(108 to 1000)
Hospitalisation: medium term (13-25 weeks)
as defined in each study
Follow-up: mean 6 months
222 per 1000 4 58 per 1000
(7 to 476)
RR 0.26
(0.03 to 2.14)
35
(1 study)
⊕⊝⊝⊝
very low 3,5
 
Global state: average score: long term (+ 26 weeks)
Global Assessment Scale (GAS). Scale from: 0 to 100.
See commentSee commentNot estimable0
(0)
See commentNo study reported provided useable data for this outcome by long term
Mental state: average score: long term (+ 26 weeks)
Brief Psychiatric Rating Scale (BPRS). Scale from: 0 to 126.
See commentSee commentNot estimable0
(0)
See commentNo study reported provided useable data for this outcome by long term
Adverse effects: tardive dyskinesia: long term (+ 26 weeks)
as defined in each study
Follow-up: mean 9.5 months
Low6 RR 1.64
(0.82 to 3.3)
50
(2 studies)
⊕⊝⊝⊝
very low 2,3
 
100 per 1000 164 per 1000
(82 to 330)
Moderate6
300 per 1000 492 per 1000
(246 to 990)
High6
600 per 1000 984 per 1000
(492 to 1000)
Economic outcomes: cost effectiveness: long term (+ 26 weeks)
as defined in each study
See commentSee commentNot estimable0
(0)
See commentNo study reported this outcome
Leaving the study early/ loss to follow-up: long term (+ 26 weeks)
as defined in each study
Follow-up: mean 9.5 months
Low6 RR 1.01
(0.46 to 2.2)
62
(2 studies)
⊕⊝⊝⊝
very low 2,3
 
100 per 1000 101 per 1000
(46 to 220)
Moderate6
300 per 1000 303 per 1000
(138 to 660)
High6
400 per 1000 404 per 1000
(184 to 880)
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: Confidence interval; RR: Risk ratio;
GRADE Working Group grades of evidence
High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low quality: We are very uncertain about the estimate.

Summary of findings 6 ANY INTERMITTENT DRUG TECHNIQUE compared with PLACEBO for schizophrenia

Summary of findings 6. ANY INTERMITTENT DRUG TECHNIQUE compared with PLACEBO for schizophrenia
  1. 1 Assumed risk: calculated from the included studies - presents 3 risks based on the control group risks - 'moderate' risk equates with that of control group (46.2%).
    2 Risk of bias: 'serious' - 100% of included studies provided no description of randomisation methods
    3 Imprecision: 'serious' - small sample size - confidence intervals for best estimate of effect include both no effect and appreciable benefit/ harm
    4 Assumed risk: calculated from the included studies - presents 3 risks based on the control group risks - 'moderate' risk equates with that of control group (31.1%).

ANY INTERMITTENT DRUG TECHNIQUE compared with PLACEBO for schizophrenia
Patient or population: patients with schizophrenia
Settings: inpatients
Intervention: ANY INTERMITTENT DRUG TECHNIQUE
Comparison: PLACEBO
OutcomesIllustrative comparative risks* (95% CI)Relative effect
(95% CI)
No of Participants
(studies)
Quality of the evidence
(GRADE)
Comments
Assumed riskCorresponding risk
PLACEBO ANY INTERMITTENT DRUG TECHNIQUE
Relapse: medium term (13-25 weeks)
as defined in each study
Follow-up: mean 16 weeks
Low1 RR 0.36
(0.23 to 0.57)
288
(2 studies)
⊕⊝⊝⊝
very low 2,3
 
200 per 1000 74 per 1000
(48 to 116)
Moderate1
400 per 1000 148 per 1000
(96 to 232)
High1
800 per 1000 296 per 1000
(192 to 464)
Hospitalisation: medium term (13-25 weeks)
as defined in each study
See commentSee commentNot estimable0
(0)
See commentNo study reported provided useable data for this outcome by long term
Global state: average score: medium term (13-25 weeks)
Global Assessment Scale (GAS). Scale from: 0 to 100.
See commentSee commentNot estimable0
(0)
See commentNo study reported provided useable data for this outcome by long term
Mental state: average score: medium term (13-25 weeks)
Brief Psychiatric Rating Scale (BPRS). Scale from: 0 to 126.
See commentSee commentNot estimable0
(0)
See commentNo study reported provided useable data for this outcome by long term
Adverse effects: tardive dyskinesia: medium term (13-25 weeks)
as defined in each study
See commentSee commentNot estimable0
(0)
See commentNo study reported provided useable data for this outcome by long term
Economic outcomes: cost effectiveness: medium term (13-25 weeks)
as defined in each study
See commentSee commentNot estimable0
(0)
See commentNo study reported this outcome
Leaving the study early/ loss to follow-up: medium term (13-25 weeks)
as defined in each study
Follow-up: mean 11 months
Low4 RR 1.97
(1.28 to 3.01)
90
(2 studies)
⊕⊝⊝⊝
very low 2,3
 
100 per 1000 197 per 1000
(128 to 301)
Moderate4
300 per 1000 591 per 1000
(384 to 903)
High4
600 per 1000 1000 per 1000
(768 to 1000)
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: Confidence interval; RR: Risk ratio;
GRADE Working Group grades of evidence
High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low quality: We are very uncertain about the estimate.

Background

Description of the condition

In 1896, Emil Kraepelin isolated the illness of schizophrenia under the name of 'dementia praecox', which was considered to lead to a 'deterioration of the personality' (Kraepelin 1896). It was eventually reframed into a substantially different concept by Swiss psychiatrist Paul Eugen Bleuler, who identified a demonstrable 'splitting of psychic functions' and re-labelled the mental illness as schizophrenia (Bleuler 1908). It is widely accepted that schizophrenia encapsulates a range of clinical manifestations, which are typically categorised into positive and negative symptoms. Positive symptoms are based on active disturbance of cerebral function (including disorders of thought such as disorganised speech/behaviour; thought-flow disturbances; delusions and auditory and visual hallucinations). Negative symptoms reflect a reduction or loss of normal function, including deficits of normal emotional responses or other thought processes (including avolition; depression; blunted affect or emotion and poverty of speech). Schizophrenia is a debilitating mental disorder, with a lifetime prevalence of about 1% (Jablensky 1992). Although schizophrenia can occur as a single episode of illness, up to 41% of those who develop schizophrenia suffer a chronic and often disabling illness with remission and relapses (Prudo 1987).

Before the discovery of antipsychotic medication, treatment for schizophrenia was found in numerous physical approaches aimed at promoting more specific benefits, including the inducement of insulin comas; electroconvulsive therapy (ECT) and prefrontal leucotomy (lobotomy) (Johnstone 1998); treatments that are now considered 'outdated' and have fallen into disuse. The discovery of chlorpromazine in 1952 paved the way for a pioneering approach for the treatment of schizophrenia, and the use of antipsychotic medication subsequently became one of the most comprehensively researched areas in psychiatry (Klein 1969). Antipsychotic medication (classified as first generation (typical) or second generation (atypical)) is now considered the mainstay of treatment for schizophrenia (Dencker 1980) and is generally regarded as highly effective, especially in controlling positive symptoms such as abnormal perceptions (hallucinations), disordered thinking and fixed false beliefs (delusions) (Kane 1998).

Antipsychotic maintenance treatment is the current predominant approach to treatment of schizophrenia, and has been recognised for its efficacy in treating symptoms in previous controlled trials (Davis 1976). A study by Baldessarini and colleagues showed antipsychotic maintenance treatment to be adequate for 50% of patients when administered with low doses of chlorpromazine (50 to 100 mg/day or equivalent) (Baldessarini 1988), and subsequent systematic reviews of randomised controlled trials have shown that relapse rates are lower with slightly higher doses of the same antipsychotic during maintenance treatment (200 to 500 mg) (Barbui 1996). Other studies have shown that antipsychotic maintenance (prophylaxis) treatment may reduce the risk of relapse (Schooler 1993), particularly with people who had recently recovered from an acute episode of schizophrenia (Leff 1971). However, long-term antipsychotic exposure has been associated with a range of adverse effects, including extra-pyramidal symptoms (EPS) (acute dystonia, akathisia, tardive dyskinesia, tardive dystonia and parkinsonism); neuroleptic malignant syndrome (NMS) (muscle rigidity, pyrexia, cognitive changes and autonomic disturbance) (Haddad 2008), and death (Ray 2009). Randomised trials have demonstrated that participants with schizophrenia who had been assigned continued antipsychotic treatment tended to discontinue treatment owing to inefficacy of the treatment or as a result of intolerable side effects (Lieberman 2005) (antipsychotic discontinuation has also proven adverse effects for those with schizophrenia, including anxiety, delirium, motor restlessness, insomnia, nausea and vomiting (Borison 1998; Haddad 2008)). Such adverse effects have been shown to impact negatively on the patient's quality of life and are a frequently cited reason for non-compliance with antipsychotic medication, which could be a contributing factor towards relapse, hospitalisation and persistent psychotic symptoms (Morken 2008).

Against this backdrop, there is cause to consider the role of intermittently administering antipsychotic medication (intermittent drug techniques) versus maintained/continued use of antipsychotic medication (maintenance treatment), in order to ascertain the extent of any beneficial effects for those with schizophrenia.

Description of the intervention

Intermittent drug techniques (or 'intermittent treatment') refers to the 'use of medication only during periods of incipient relapse or symptom exacerbation rather than continuously' (Schooler 2004).The aim is to reduce the risk of typical adverse effects of antipsychotics (including tardive dyskinesia) by 'reducing long-term medication exposure for patients who are receiving maintenance treatment while limiting the risk of relapse', with a further goal of improving social functioning resulting from the reduction of antipsychotic-induced side effects (Schooler 2004).

How the intervention might work

People with chronic schizophrenia very frequently discontinue and re-instigate their own antipsychotic medications, and few continue to take their medication for lengthy periods of time (Lieberman 2005); self-imposed 'drug holidays' are common. It is clear that antipsychotics have proven effects in controlling positive symptoms, but are accompanied with the caveat of causing potential adverse effects; research has turned to assessing the ways in which various 'intermittent' strategies may maximise the benefits and minimise the negative effects of long-term antipsychotic use; these strategies include dose reduction, antipsychotic cessation and intermittent drug techniques, such as drug holidays.

The focus of this review will be on the use of intermittent drug techniques, which include:

  • prodrome-based intervention;

  • crisis intervention;

  • gradually increased drug-free periods; and

  • drug holidays.

Intermittent treatment is a long-term strategy that is used 'only when needed', relying on:

  • the presence of a prodromal period that would allow for intervention with medication in order to prevent relapse; and

  • an effective strategy for clinical monitoring in order to accurately identify prodromal symptoms in time for medication to be introduced (Schooler 2004).

Various methods of detecting early prodromal signs and symptoms have been developed (including the Early Signs Questionnaire, Herz 1980) in order to provide a higher standard of accurate judgement in determining when to introduce/reintroduce medication.

Why it is important to do this review

Co-operation and adherence to medication is a further significant issue in the clinical management of schizophrenia. Early treatment discontinuation in patients with schizophrenia or schizophrenia-like disorders is strikingly common, with estimates of its prevalence in antipsychotic drug trials ranging from 25% to 75% (Nose 2003). Discontinuation of a prescribed antipsychotic drug is associated with symptom exacerbation, relapse, and increased hospitalisation (Perkins 2002). A review that analysed 66 studies found a mean cumulative relapse rate of 53% in patients completely withdrawn from antipsychotic therapy compared to 16% for those maintained on a regimen of antipsychotic therapy over a mean follow-up period of 9.7 months (Gilbert 1995). Evidence also points to the fact that experiencing a relapse of schizophrenia lowers a person's level of social functioning and quality of life (Curson 1985).

There are proven adverse effects associated with continued exposure to antipsychotic medication, but the extent to which maintenance treatment prevents relapse is unclear. The employment of non-continuous antipsychotic treatment has been tested in practice, under the support of literature that implies that patients decompensating off medication are either more responsive to treatment and/or experience more benign decompensations than patients experiencing relapse on medication; either due to the barrier provided by continuous medication or any potential diminutive effects of medication owing to such prolonged, continued use (Carpenter 1983; Gardos 1976).

An intermittent approach depends on the accurate identification of times when medication should be administered, as well as the creation of an effective treatment structure to incorporate ongoing therapeutic patient monitoring and support. The approach further employs the assumption that those people with schizophrenia require antipsychotic medication on an 'only when needed' basis (during symptom exacerbation) and that the development of florid symptoms can be averted through the introduction of medication during the prodromal period (Schooler 2004). With an intermittent technique comes potential economic and practical advantages, and a key question is whether an 'intermittent' intervention would eliminate or merely ameliorate an impending psychotic episode. It is for these reasons that intermittent drug techniques merit a dedicated systematic review; as an approach to using antipsychotic medication, it is fundamental to assess its efficacy as a technique that seeks to maximise the benefits and minimise the problems associated with continued antipsychotic exposure.

Objectives

To review the effects of different intermittent drug techniques compared with maintenance treatment (as defined by the trial authors) in people with schizophrenia or related disorders.

Methods

Criteria for considering studies for this review

Types of studies

All relevant randomised controlled trials. Where a trial was described in such a way as to imply randomisation, we included such trials in a sensitivity analysis (see Sensitivity analysis). Had their inclusion not resulted in a substantive difference, they remained in the analyses. Had their inclusion resulted in statistically significant differences, we would not add the data from these lower quality studies to the results of the better trials, but would have presented such data within a subcategory. We excluded quasi-randomised studies, such as those allocating by alternate days of the week.

Types of participants

We included people with schizophrenia and other types of schizophrenia-like psychoses (schizophreniform and schizoaffective disorders diagnosed by any criteria), irrespective of gender, age or nationality. There is no clear evidence that the schizophrenia-like psychoses are caused by fundamentally different disease processes or require different treatment approaches (Carpenter 1994).

Types of interventions

Any type of intermittent drug technique versus maintenance therapy:

a. Any intermittent drug technique

Including:

1. Prodrome-based/early intervention

Defined as treatment given on the early signs of relapse.

2. Crisis intervention

Defined as treatment given only in case of full relapse and discontinued again after re-stabilisation.

3. Gradually increased drug-free period

Defined as increasing the cessation period of the treatment constantly.

4. Drug holiday

Defined as stopping medication for fixed periods, and then reintroducing it (repeating this more than once).

Compared with:

1. Maintenance therapy

As defined in each study.

2. Placebo
b. Any intermittent drug technique (specific named drug)
1. High dose (as defined by each study)
2. Low or moderate dose (as defined by each study)

Types of outcome measures

We grouped outcomes into the short term (up to 12 weeks), medium term (13 to 25 weeks) and long term (over 26 weeks).

Primary outcomes
1. Relapse (as defined in the individual studies)
2. Hospitalisation
Secondary outcomes
1. Death - suicide and natural causes
2. Global state
2.1 Any clinically important change in global state (as defined by individual studies)
2.2 Average endpoint global state score
2.3 Average change in global state scores
3. Service outcomes
3.1 Time to hospitalisation
4. Mental state (with particular reference to the positive and negative symptoms of schizophrenia)
4.1 No clinically important change in general mental state
4.2 Average endpoint general mental state score
4.3 Average change in general mental state scores
4.4 No clinically important change in specific symptoms (positive symptoms of schizophrenia, negative symptoms of schizophrenia, depression, mania)
4.5 Average endpoint specific symptom score
4.6 Average change in specific symptom scores
5. General functioning
5.1 No clinically important change in general functioning
5.2 Average endpoint general functioning score
5.3 Average change in general functioning scores
5.4 No clinically important change in specific aspects of functioning, such as social or life skills
5.5 Average endpoint specific aspects of functioning, such as social or life skills
5.6 Average change in specific aspects of functioning, such as social or life skills
6. Behaviour
6.1 No clinically important change in general behaviour
6.2 Average endpoint general behaviour score
6.3 Average change in general behaviour scores
6.4 No clinically important change in specific aspects of behaviour
6.5 Average endpoint specific aspects of behaviour
6.6 Average change in specific aspects of behaviour
7. Adverse effects - general and specific
7.1 Clinically important general adverse effects
7.2 Average endpoint general adverse effect score
7.3 Average change in general adverse effect scores
7.4 Clinically important specific adverse effects
7.5 Average endpoint specific adverse effects
7.6 Average change in specific adverse effects
8. Engagement with services
9. Satisfaction with treatment
9.1 Leaving the studies early
9.2 Recipient of care not satisfied with treatment
9.3 Recipient of care average satisfaction score
9.4 Recipient of care average change in satisfaction scores
9.5 Carer not satisfied with treatment
9.6 Carer average satisfaction score
9.7 Carer average change in satisfaction scores
10. Quality of life
10.1 No clinically important change in quality of life
10.2 Average endpoint quality of life score
10.3 Average change in quality of life scores
10.4 No clinically important change in specific aspects of quality of life
10.5 Average endpoint specific aspects of quality of life
10.6 Average change in specific aspects of quality of life
11. Cognitive functioning
11.1 No clinically important change in cognitive functioning
11.2 Average endpoint cognitive functioning score
11.3 Average change in cognitive functioning scores
11.4 No clinically important change in specific aspects of cognitive functioning
11.5 Average endpoint specific aspects of cognitive functioning
11.6 Average change in specific aspects of cognitive functioning
12. Economic outcomes
12.1 Direct costs (as defined by each study)
12.2 Indirect costs (as defined by each study)
12.3 Cost-effectiveness (as defined by each study)
13. Leaving the study early/loss to follow-up
14. 'Summary of findings' table

We used the GRADE approach to interpret findings (Schünemann 2008) and used GRADE profiler (GRADEPRO) to import data from RevMan 5 (Review Manager) to create 'Summary of findings' tables. These tables provide outcome-specific information concerning the overall quality of evidence from each included study in the comparison, the magnitude of effect of the interventions examined, and the sum of available data on all outcomes we rated as important to patient-care and decision making. We selected the following main outcomes for inclusion in the summary of findings table.

  1. Relapse

  2. Hospitalisation

  3. Adverse effects

  4. Mental state

  5. Quality of life

  6. Economic outcomes

  7. Leaving the study early/loss to follow-up

Search methods for identification of studies

Electronic searches

Cochrane Schizophrenia Group Trials Register

An initial search of the Cochrane Schizophrenia Group Trials Register was carried out in 2006 (see Appendix 1) This register is compiled by systematic searches of major databases, handsearches and conference proceedings (see Group Module).

We additionally made a later search of the Cochrane Schizophrenia Group Trials Register in April 2012, using the phrase:

[((intermit* or drug?holiday* or drug?free* or internal?med*) in title, abstract and index fields OR *targeted* in title or *targeted medication* in abstract or *drug administration methods*  in indexing terms REFERENCE ) OR ((intermittent medication or drug-free period) in interventions field in STUDY].

Searching other resources

1. Reference searching

We inspected references of all included studies for further relevant studies and sought additional relevant trials by searching reference lists of included and excluded trials.

2. Requests for additional data

We attempted to contact authors of relevant trials to inquire about other sources of relevant information.

Data collection and analysis

Since the protocol for this review was published the Cochrane Schizophrenia Group has updated its methodology for data collection and analysis. We have updated the relevant sections below to incorporate these new methods. See Appendix 2 for information relating to data collection and analysis specified in the protocol.

Selection of studies

Review author SS independently inspected citations, with help from Kajal Joshi (Acknowledgements) from the searches and identified relevant abstracts. A random 20% sample was independently re-inspected by CEA to ensure reliability. Where disputes arose, the full report was acquired for more detailed scrutiny. Full reports of the abstracts meeting the review criteria were obtained and inspected by KJ and SS. Again, a random 20% of reports was re-inspected by CEA in order to ensure reliable selection. Had it not been possible to resolve disagreement by discussion, we would have attempted to contact the authors of the study for clarification.

Data extraction and management

1. Extraction

Review author SS extracted data with help from Kajal Joshi (Acknowledgements) from all included studies. In addition, to ensure reliability, CEA independently extracted data from a random sample of these studies, comprising 10% of the total. Again, any disagreement was discussed, decisions documented and, if necessary, authors of studies were contacted for clarification. With remaining problems, CEA helped clarify issues and these final decisions were documented. Data presented only in graphs and figures were extracted whenever possible, but included only if two review authors independently had the same result. We attempted to contact authors through an open-ended request in order to obtain missing information or for clarification whenever necessary. If studies were multi-centre, where possible, we extracted data relevant to each component centre separately.

2. Management
2.1 Forms

We extracted data onto standard, simple forms.

2.2 Scale-derived data

We included continuous data from rating scales only if:
a. the psychometric properties of the measuring instrument were described in a peer-reviewed journal (Marshall 2000); and
b. the measuring instrument had not been written or modified by one of the trialists for that particular trial.
Ideally, the measuring instrument was either i. a self-report or ii. completed by an independent rater or relative (not the therapist). We realise that this is not often reported clearly; in Description of studies we have noted if this was the case or not.

2.3 Endpoint versus change data

There are advantages of both endpoint and change data. Change data can remove a component of between-person variability from the analysis. On the other hand, calculation of change needs two assessments (baseline and endpoint) which can be difficult in unstable and difficult to measure conditions such as schizophrenia. We decided primarily to use endpoint data, and only use change data if the former were not available. We combined endpoint and change data in the analysis as we used mean differences (MD) rather than standardised mean differences throughout (Higgins 2011).

2.4 Skewed data

Continuous data on clinical and social outcomes are often not normally distributed. To avoid the pitfall of applying parametric tests to non-parametric data, we to applied the following standards to all data before inclusion: a) standard deviations (SDs) and means are reported in the paper or obtainable from the authors; b) when a scale started from the finite number zero, the SD, when multiplied by two, was less than the mean (as otherwise the mean is unlikely to be an appropriate measure of the centre of the distribution, (Altman 1996); c) if a scale started from a positive value (such as the Positive and Negative Syndrome Scale (PANSS) which can have values from 30 to 210), the calculation described above was modified to take the scale starting point into account. In these cases skew is present if 2 SD > (S-S min), where S is the mean score and S min is the minimum score. Endpoint scores on scales often have a finite start and end point and these rules can be applied. We entered skewed endpoint data from studies of less than 200 participants as 'Other data' within the Data and analyses section rather than into a statistical analysis. Skewed endpoint data pose less of a problem when looking at mean if the sample size is large (over 200), we entered such data into statistical syntheses.

When continuous data are presented on a scale that includes a possibility of negative values (such as change data), it is difficult to tell whether data are skewed or not and we entered skewed change data into statistical analysis.

2.5 Common measure

To facilitate comparison between trials, we had intended to convert variables that can be reported in different metrics, such as days in hospital (mean days per year, per week or per month) to a common metric (e.g. mean days per month). However, no such variables were found.

2.6 Conversion of continuous to binary

Where possible, we made efforts to convert outcome measures to dichotomous data. This can be done by identifying cut-off points on rating scales and dividing participants accordingly into 'clinically improved' or 'not clinically improved'. It is generally assumed that if there is a 50% reduction in a scale-derived score such as the Brief Psychiatric Rating Scale (BPRS, Overall 1962) or the PANSS (Kay 1986), this could be considered as a clinically significant response (Leucht 2005a; Leucht 2005b). If data based on these thresholds were not available, we used the primary cut-off presented by the original authors.

2.7 Direction of graphs

Where possible, we entered data in such a way that the area to the left of the line of no effect indicated a favourable outcome for intermittent drugs. If we had to enter data so the area to the left of the line indicated a favourable outcome for the control group, this was noted in the relevant graphs.

Assessment of risk of bias in included studies

Review authors KSW, NM and SS worked independently to assess risk of bias by using criteria described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) to assess trial quality. This set of criteria is based on evidence of associations between overestimate of effect and high risk of bias of the article such as sequence generation, allocation concealment, blinding, incomplete outcome data and selective reporting. If the raters disagreed, the final rating was made by consensus, with the involvement of another member of the review group. Where inadequate details of randomisation and other characteristics of trials were provided, authors of the studies were contacted in order to obtain further information. Non-concurrence in quality assessment was reported, but if disputes arose as to which category a trial was to be allocated, again, resolution was made by discussion. The level of risk of bias was noted in both the text of the review and in the 'Summary of findings' tables (see below).

Measures of treatment effect

1. Binary data

For binary outcomes, we calculated a standard estimation of the risk ratio (RR) and its 95% confidence interval (CI). It has been shown that RR is more intuitive (Boissel 1999) than odds ratios and that odds ratios tend to be interpreted as RR by clinicians (Deeks 2000). The Number Needed to Treat/Harm (NNT/H) statistic with its confidence intervals is intuitively attractive to clinicians but is problematic both in its accurate calculation in meta-analyses and interpretation (Hutton 2009). For binary data presented in the 'Summary of findings' tables, where possible, we calculated illustrative comparative risks.

2. Continuous data

For continuous outcomes, we estimated mean difference (MD) between groups. We preferred not to calculate effect size measures (standardised mean difference SMD). However, if scales of very considerable similarity were used, we presumed there was a small difference in measurement, and we calculated effect size and transformed the effect back to the units of one or more of the specific instruments.

Unit of analysis issues

1. Cluster trials

Studies increasingly employ cluster randomisation (such as randomisation by clinician or practice) but analysis and pooling of clustered data poses problems. Authors often fail to account for intra-class correlation in clustered studies, leading to a unit of analysis error (Divine 1992) whereby P values are spuriously low, confidence intervals unduly narrow and statistical significance overestimated. This can cause Type I errors (Bland 1997; Gulliford 1999).

If clustering had not been accounted for in primary studies, we would have presented the data in a table, with a (*) symbol to indicate the presence of a probable unit of analysis error. In subsequent versions of this review we will seek to contact first authors of studies to obtain intra-class correlation coefficients (ICCs) of their clustered data and to adjust for this using accepted methods (Gulliford 1999). Where clustering may be incorporated into the analysis of primary studies, we will also present these data as if from a non-cluster randomised study, but adjusted for the clustering effect.

We have sought statistical advice and have been advised that the binary data as presented in a report should be divided by a design effect. This is calculated using the mean number of participants per cluster (m) and the ICC [Design effect = 1 + (m-1)*ICC] (Donner 2002). If the ICC is not reported, it would be assumed to be 0.1 (Ukoumunne 1999).

If cluster studies had been appropriately analysed taking into account ICCs and relevant data documented in the report, we would have synthesised these with other studies using the generic inverse variance technique. However, no cluster trials were identified in this review.

2. Cross-over trials

A major concern of cross-over trials is the carry-over effect. It occurs if an effect (e.g. pharmacological, physiological or psychological) of the treatment in the first phase is carried over to the second phase. As a consequence, on entry to the second phase, the participants can differ systematically from their initial state despite a wash-out phase. For the same reason cross-over trials are not appropriate if the condition of interest is unstable (Elbourne 2002). As both effects are very likely in schizophrenia, we would only have used data of the first phase of cross-over studies. However, no cross-over trials were identified.

3. Studies with multiple treatment groups

Where a study involves more than two treatment arms, if relevant, the additional treatment arms were presented in comparisons. If data were binary, these have simply been added and combined within the two-by-two table. If data were continuous, we combined the data following the formula in section 7.7.3.8  (Combining groups) of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Where the additional treatment arms were not relevant, we did not use these data.

Dealing with missing data

1. Overall loss of credibility

At some degree of loss of follow-up, data must lose credibility (Xia 2009). We chose that, for any particular outcome, should more than 50% of data be unaccounted for, we would not reproduce these data or use them within analyses, (except for the outcome 'leaving the study early') - this was the case for the Pietzcker 1993* study. If, however, more than 50% of those in one arm of a study were lost, but the total loss was less than 50%, we marked such data with (*) to indicate that such a result may well be prone to bias. This was the case in Carpenter 1990*; Herz 1991* and Olson 1962*

2. Binary

In the case where attrition for a binary outcome was between 0% and 50% and where these data were not clearly described, we presented data on a 'once-randomised-always-analyse' basis (an intention-to-treat analysis). Those leaving the study early were all assumed to have the same rates of negative outcome as those who completed, with the exception of the outcome of death and adverse effects. For these outcomes the rate of those who stayed in the study - in that particular arm of the trial - were used for those who did not. We undertook a sensitivity analysis to test how prone the primary outcomes were to change when data only from people who completed the study to that point were compared to the intention-to-treat analysis using the above assumptions.

3. Continuous
3.1 Attrition

In the case where attrition for a continuous outcome was between 0% and 50%, and data only from people who completed the study to that point were reported, we reproduced these.

3.2 Standard deviations

If standard deviations (SDs) were not reported, we first would have tried to obtain the missing values from the authors. If not available, where there were missing measures of variance for continuous data, but an exact standard error (SE) and confidence intervals (CIs) available for group means, and either 'P' value or 't' value available for differences in mean, we would have calculated them according to the rules described in theCochrane Handbook for Systemic reviews of Interventions (Higgins 2011). If only the SE was reported, SDs would have been calculated by the formula SD = SE * square root (n). Chapters 7.7.3 and 16.1.3 of the Cochrane Handbook for Systemic reviews of Interventions (Higgins 2011) present detailed formulae for estimating SDs from P values, t or F values, CIs, ranges or other statistics. If these formulae did not apply, we would have calculated the SDs according to a validated imputation method which is based on the SDs of the other included studies (Furukawa 2006). Although some of these imputation strategies can introduce error, the alternative would be to exclude a given study’s outcome and thus to lose information. We nevertheless would have examined the validity of the imputations in a sensitivity analysis excluding imputed values, however, no imputations were made regarding SDs.

3.3 Last observation carried forward

We anticipated that in some studies the method of last observation carried forward (LOCF) would be employed within the study report. As with all methods of imputation to deal with missing data, LOCF introduces uncertainty about the reliability of the results (Leucht 2007). Therefore, where LOCF data have been used in the trial, if less than 50% of the data have been assumed, we presented and used these data and indicated that they were the product of LOCF assumptions.

Assessment of heterogeneity

1. Clinical heterogeneity

We considered all included studies initially, without seeing comparison data, to judge clinical heterogeneity. We simply inspected all studies for clearly outlying people or situations which we had not predicted would arise. When such situations or participant groups were noted, these were fully discussed.

2. Methodological heterogeneity

We considered all included studies initially, without seeing comparison data, to judge methodological heterogeneity. We simply inspected all studies for clearly outlying methods which we had not predicted would arise. When such methodological outliers arose, these were fully discussed.

3. Statistical
3.1 Visual inspection

We visually inspected graphs to investigate the possibility of statistical heterogeneity.

3.2 Employing the I2 statistic

Heterogeneity between studies was investigated by considering the I2 method alongside the Chi2 'P' value. The I2 provides an estimate of the percentage of inconsistency thought to be due to chance (Higgins 2003). The importance of the observed value of I2 depends on i. magnitude and direction of effects and ii. strength of evidence for heterogeneity (e.g. 'P' value from Chi2 test, or a confidence interval for I2). An I2 estimate greater than or equal to 50% accompanied by a statistically significant Chi2 statistic, was interpreted as evidence of substantial levels of heterogeneity (Section 9.5.2 - Higgins 2011). When substantial levels of heterogeneity were found in the primary outcome, we explored reasons for heterogeneity (Subgroup analysis and investigation of heterogeneity).

Assessment of reporting biases

1. Protocol versus full study

Reporting biases arise when the dissemination of research findings is influenced by the nature and direction of results. These are described in section 10.1 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We intended to locate protocols of included randomised trials. If the protocol had been available, outcomes in the protocol and in the published report would have been compared. If the protocol was not available, outcomes listed in the methods section of the trial report were compared with actually reported results. No protocols were available for any of the included studies.

2. Funnel plot

Reporting biases arise when the dissemination of research findings is influenced by the nature and direction of results (Egger 1997). These are again described in Section 10 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We are aware that funnel plots may be useful in investigating reporting biases but are of limited power to detect small-study effects. In other cases, where funnel plots were possible, we sought statistical advice in their interpretation.

Data synthesis

We understand that there is no closed argument for preference for use of fixed-effect or random-effects models. The random-effects method incorporates an assumption that the different studies are estimating different, yet related, intervention effects. This often seems to be true to us and the random-effects model takes into account differences between studies even if there is no statistically significant heterogeneity. There is, however, a disadvantage to the random-effects model. It puts added weight onto small studies which often are the most biased ones. Depending on the direction of effect, these studies can either inflate or deflate the effect size. We therefore chose to use a fixed-effect method for all analyses and used a random-effects model where heterogeneity were found. For primary outcomes, we also synthesised data using a random-effects model, and where the estimate of the effect was notably changed, this was noted in the conclusion.

Subgroup analysis and investigation of heterogeneity

Subgroup analyses

Data were sub-grouped according to duration of follow-up (short term, medium term or long term). The data for one study were compared in a single subgroup independently of the main comparison. The review authors believed that the inclusion of this data from a three-armed intervention trial was required, as a direct comparison of the two intermittent drug techniques ('early-based' and 'crisis-based').

Investigation of heterogeneity

If inconsistency were high, this was reported. First, we investigated whether data had been entered correctly. Second, if data were correct, the graph was visually inspected and outlying studies were successively removed to see if homogeneity was restored. For this review, we decided that should this occur with data contributing to the summary finding of no more than around 10% of the total weighting, data would be presented. If not, data were pooled and issues were discussed. We know of no supporting research for this 10% cut-off but are investigating use of prediction intervals as an alternative to this unsatisfactory state. When unanticipated clinical or methodological heterogeneity were obvious, we simply stated hypotheses regarding these for future reviews or versions of this review. We did not anticipate undertaking analyses relating to these.

Sensitivity analysis

1. Implication of randomisation

We included trials in a sensitivity analysis if they were described in some way as to imply randomisation. For the primary outcomes, we included these studies and if there was no substantive difference when the implied randomised studies were added to those with a better description of randomisation, then all data were employed from these studies. We undertook a sensitivity analysis excluding these studies from the primary outcomes to assess whether this affected the results.

2. Assumptions for lost binary data

Where assumptions had to be made regarding people lost to follow-up (see Dealing with missing data), we compared the findings of the primary outcomes when we used our assumption compared with completer data only. If there was a substantial difference, we reported results and discussed them but continued to employ our assumption.

Where assumptions would have had to be made regarding missing SDs data (see Dealing with missing data), we would have compared the findings on primary outcomes when we used our assumption compared with completer data only. A sensitivity analysis would have been undertaken to test how prone results were to change when 'completer' data only were compared to the imputed data using the above assumption. However, no SD data were imputed.

3. Risk of bias

We analysed the effects of excluding trials that were judged to be at high risk of bias across one or more of the domains of randomisation (implied as randomised with no further details available) allocation concealment, blinding and outcome reporting for the meta-analysis of the primary outcome. If the exclusion of trials at high risk of bias did not substantially alter the direction of effect or the precision of the effect estimates, then data from these trials were included in the analysis. However, all included studies were rated as 'high' on risk of bias on one or more of the domains, therefore, the effects of excluding each study would have resulted in no data to compare.

4. Imputed values

We had also intended to undertake a sensitivity analysis to assess the effects of including data from trials where we used imputed values for ICC in calculating the design effect in cluster randomised trials. If substantial differences were noted in the direction or precision of effect estimates in any of the sensitivity analyses listed above, we would not have pooled data from the excluded trials with the other trials contributing to the outcome, but instead presented them separately. However, no cluster randomised trials were identified.

5. Fixed and random effects

All data have been synthesised using a fixed-effect model, however, we also synthesised data for the primary outcome using a random-effects model to evaluate whether this altered the significance of the results.

6. Skewed data

Where scale derived data were both skewed and not skewed, all data were synthesised together and then the skewed data removed. This process was described and where clinically meaningful changes resulted, only the non-skewed data were presented in the synthesis. Skewed data were presented in 'other' tables within the Data and analyses section.

Results

Description of studies

Please also see Characteristics of included studies; Characteristics of excluded studies.

Results of the search

The initial search of the Cochrane Schizophrenia Group Trials Register, carried out in 2006 (see Appendix 1), identified 174 studies. We additionally made a later search of the same Trials Register in April 2012 (using a modified search, based on the search strategy from the original review), which identified 56 references. An additional 11 potentially relevant references were identified from the reference lists of published trials and reviews (see Figure 1).

Figure 1.

Study flow diagram.

Included studies

The review includes 17 studies published between 1962 and 2011. All studies were described as randomised; however, only four trials provided adequate descriptions of randomisation (Gaebel 2011; Herz 1991* Schooler 1997; Wunderink 2007), the remaining majority of included studies did not provide any further details beyond describing their trial as 'randomised' (for more details, see Characteristics of included studies and the accompanying 'Risk of bias' tables).

1. Setting

Of the 17 included studies within the meta-analysis, six were large, multi-centre trials involving two 18-hospital collaborative studies (Caffey 1964; Prien 1973), two German multi-centre studies (Gaebel 2011; Pietzcker 1993*), a study conducted at seven district mental health care centres and a University Department of Psychiatry (all in the Netherlands, Wunderink 2007), and one multi-centre trial conducted in the US (Schooler 1997). Two studies were conducted in hostel wards in Scotland (UK) (McCreadie 1980; McCreadie 1982), four studies took place in psychiatric hospitals/institutes in the US (Carpenter 1987; Carpenter 1990*; Herz 1991*; Shenoy 1981), and one in Germany (Wiedemann 2001). One study took place in a London (UK) hospital (Jolley 1989/1990), and a further trial was conducted in a Centre for Addiction and Mental Health, Canada (Remington 2011). It is not clear exactly where the trial was conducted in Blackburn 1961 and Olson 1962*.

2. Length of trials

We included trials that varied in duration; from six weeks (Shenoy 1981), four months (Blackburn 1961; Caffey 1964; Prien 1973), six months (Olson 1962*; Remington 2011) nine and ten months (McCreadie 1980; McCreadie 1982), one year (Gaebel 2011), 18 months (Wiedemann 2001) to two years (Carpenter 1987; Carpenter 1990*; Herz 1991*; Jolley 1989/1990; Pietzcker 1993*; Schooler 1997; Wunderink 2007).

3. Participants

There were a total of n = 2252 participants in the 17 included studies that provided data for the meta-analyses. Participants in seven of the included trials appear to have been inpatients (Blackburn 1961; Caffey 1964; McCreadie 1980; McCreadie 1982; Olson 1962*; Prien 1973; Wiedemann 2001), while in the remaining 10 studies, participants were outpatients (Carpenter 1987; Carpenter 1990*; Gaebel 2011; Herz 1991*; Jolley 1989/1990; Pietzcker 1993*; Remington 2011; Schooler 1997; Shenoy 1981; Wunderink 2007).

All participants had a diagnosis of schizophrenia using different operational diagnostic criteria; RDC criteria 1978 (Carpenter 1987; Carpenter 1990*), ICD-9/10 (ICD-10) (Gaebel 2011; Pietzcker 1993*; Wiedemann 2001), DSM-III/IV (APA 1980; APA 1994) (Herz 1991*; Jolley 1989/1990; Remington 2011; Schooler 1997; Shenoy 1981; Wunderink 2007), 'definite schizophrenia' according to Feighner 1972 criteria (McCreadie 1980; McCreadie 1982) - classification and diagnostic criteria were unclear in four studies (Blackburn 1961; Caffey 1964; Olson 1962*; Prien 1973).

In total, there were more male (n = 1555) than female (n = 540) participants randomised. Ten studies included both male and female participants (Carpenter 1987; Carpenter 1990*; Gaebel 2011; Herz 1991*; Jolley 1989/1990; Pietzcker 1993*; Remington 2011; Schooler 1997; Wiedemann 2001; Wunderink 2007), five studies included male participants only (Blackburn 1961; Caffey 1964; McCreadie 1980; McCreadie 1982; Prien 1973), and two studies did not specify the sex of participants (Olson 1962*; Shenoy 1981).

4. Trial Size

The overall sample size in all the included trials was generally small. The total number of participants in each trial ranged from n = 29 (McCreadie 1982) to n = 375 (Prien 1973). Only seven studies had a sample size of 100+ participants (Caffey 1964; Carpenter 1990*; Herz 1991*; Pietzcker 1993*; Prien 1973; Schooler 1997; Wunderink 2007).

5. Interventions
5.1 Any Intermittent drug technique versus maintenance therapy

Out of the included studies, each made use of different intermittent drug techniques and various antipsychotics, however, most dosages were classified as either low/moderate (apart from McCreadie 1980; McCreadie 1982, in which both trials employed high doses of pimozide). All trials included a maintenance group consisting of participants taking their usual dosage of medication as prescribed in the form of maintenance, continuous or long-term therapy (these are discussed in Characteristics of included studies).

The majority of studies used a drug-holiday approach versus maintenance treatment (Blackburn 1961; Caffey 1964; McCreadie 1980; McCreadie 1982; Olson 1962*; Prien 1973; Remington 2011; Shenoy 1981), followed by a prodrome-based (early) intervention approach versus maintenance treatment (Carpenter 1987; Carpenter 1990*; Herz 1991*; Jolley 1989/1990; Schooler 1997) and gradually-increased drug-free periods versus maintenance treatment (Gaebel 2011; Wiedemann 2001; Wunderink 2007). One trial employed three treatment arms and compared both prodrome-based (early) versus crisis-based intermittent intervention versus maintenance treatment (Pietzcker 1993*); the results for this study were presented in the main results as well as in a subgroup analysis, to compare the effects of each intermittent intervention with each other. Types of antipsychotics and dosages varied between studies, and where some trials did not describe the specific drugs used nor the dosage, mean dosages of chlorpromazine or haloperidol equivalents were frequently employed; these details have been addressed in order to account for differences between studies.

5.1.1. Prodrome-based/early interventions (defined as treatment given on the early signs of relapse).

Five trials were included comparing an intermittent prodrome-based/early intervention to maintenance therapy. Two trials used the same approach of administering antipsychotic mediation on an 'as-needed' basis to participants who were otherwise drug-free, with moderate to high doses of antipsychotics given when prodromal symptoms occurred and discontinued when a stable state was achieved (mean daily dose 196 mg ± 163 mg chlorpromazine equivalents, Carpenter 1987; mean daily dose 4.4 mg ± 1.1 mg haloperidol/173 mg ± 69 mg chlorpromazine, Carpenter 1990*). Participants were treated within the context of a psychosocial intervention, which complemented the intermittent treatment strategy in Carpenter 1987 (involving assignment to a primary therapist) and complemented both treatment strategies in Carpenter 1990* (weekly individual therapy as well as an educational approach through involvement with family/significant others).

In Herz 1991*, participants received either placebo medication or their usual dose of maintenance antipsychotic medication, as well as weekly supportive group therapy sessions. When prodromal symptoms appeared, participants in the intermittent treatment group were given the active form of antipsychotic medication openly; dosage was at the discretion of the psychiatrist, depending on the severity of symptoms, but was usually twice the maintenance or baseline level (average cumulative antipsychotic drug dosage (mg chlorpromazine equivalents) over two years = 487.19 mg ± 370.68 mg). Once participants were clinically stabilised and considered in remission for two weeks, active medication was decreased again while placebo was simultaneously increased.

Participants in Jolley 1989/1990 either received fluphenazine in pre-trial doses or equivalent doses of placebo injections. Upon the emergence of prodromal symptoms or relapse, participants were given 5 mg to10 mg haloperidol for up to four weeks, and continued for a further four weeks after remission of symptoms (average cumulative antipsychotic drug dosage (mg haloperidol equivalents) over two years = 298 mg ± 249 mg - participants were withdrawn if relapse exceeded eight weeks).

An intermittent, 'targeted, early intervention' was employed in Schooler 1997, in which stabilised participants were given an injection of sesame oil (or "miglioil vehicle", placebo) every two weeks, and when participants showed prodromal signs of relapse, open-label rescue medication was added. In addition, participants received either one of two psychosocial interventions; psychoeducation-supportive family management (SFM) or applied family management (AFM), which included psychoeducational workshops for participants and their families over a period of 24 months.

Participants in Pietzcker 1993* were assigned to one of three treatment arms, including a prophylactic early intervention technique, in which antipsychotic therapy resumed as soon as prodromal symptoms appeared and were discontinued once stabilised (average cumulative antipsychotic drug dosage (mg chlorpromazine equivalents) over two years = 90 mg); the other treatment arms included maintenance treatment and a crisis-based intervention, addressed below.

5.1.2. Crisis intervention (defined as treatment given only in case of full relapse and discontinued again after re-stabilisation).

There was only one trial that employed a crisis-based intermittent technique (Pietzcker 1993*), classified as a 'antipsychotic crisis intervention', in which antipsychotic treatment was employed only when relapse occurred and was discontinued once stabilisation was achieved (average cumulative antipsychotic drug dosage (mg chlorpromazine equivalents) over two years = 110 mg).

5.1.3. Gradually increased drug-free periods (defined as increasing the cessation period of the treatment constantly).

The intermittent technique adopted in Gaebel 2011 involved stepwise removal of previously maintained antipsychotic treatment over a period of three months at the most, and was restarted at the occurrence of prodromal symptoms - mean daily dose of 1 mg/day haloperidol-equivalent (the mean dose covers an initial maintenance phase of about three to four weeks (mean daily dose 3 mg/day), a phase of about 10 weeks in which antipsychotics were tapered off (mean daily dose 1.7 mg/day), a phase of six months where antipsychotics were withdrawn completely (0 mg/day) and a two-week phase in which drug treatment was restarted).

Similarly, Wiedemann 2001 employed a 'targeted' medication approach, involving gradual decrease of antipsychotics after three months using a step-by-step discontinuation technique of 50% of antipsychotics every two weeks (mean daily dosage 201 mg ± 134 mg chlorpromazine (CPZ) equivalent). Where prodromal signs occurred, antipsychotic treatment was reintroduced and when re-stabilisation was attained, pharmacotherapy was tapered off once more. Participants in Wunderink 2007 received a 'guided discontinuation' treatment, in which dosage was gradually tapered and discontinued 'if feasible'. Tapering was guided by symptom severity levels and the preferences of the participant; if early warning signs of relapse emerged or positive symptoms recurred, clinicians were to restart or increase the dosage of anti-psychotics (average cumulative antipsychotic drug dosage (mg haloperidol equivalents) over two years = 4.36 mg).

5.1.4. Drug holiday (defined as stopping medication for fixed periods, and then reintroducing it (repeating this more than once)).

Blackburn 1961 subdivided their intermittent experimental group into two further groups of n = 15, with participants receiving either placebo for the whole length of the study, or receiving placebo for the first eight weeks of the study then reinstating antipsychotic medication for the final eight weeks of the study (either prochlorperazine, perphenazine, chlorpromazine, promazine, trifluoperazine - dosages not stated). Participants in Caffey 1964 received a reduced total dosage of either chlorpromazine or thioridazine (administered in standard 100 mg tablets) on an intermittent-schedule; they received their usual daily dosage on Monday, Wednesday and Friday, resulting in a reduction of dosage to 3/7 of what it had been previously. In McCreadie 1980, participants received intermittent pimozide (mean dose 8 mg/oral, maximum 32 mg every four days/week) and in McCreadie 1982, participants also received intermittent pimozide once a week (mean dose 10 mg to 60 mg/oral, mean 40 mg/intramuscular (IM) weekly). Olson 1962* 'alternated' either phenothiazine or chlorpromazine and nothing between participants in the intermittent strategy group (dosages not specified).

Participants were assigned to one of four intermittent interventions in Prien 1973: (1) intermittent five-day schedule A, where participants received their pre-study dosage Monday through Friday and placebo on Saturday and Sunday; (2) intermittent five-day schedule B, where participants received their pre-study dosage Monday, Wednesday, Thursday, Friday and Sunday, and placebo on Tuesday and Saturday; (3) intermittent four-day schedule A: participants received their pre-study dosage Monday through Thursday and placebo on Friday through Sunday; or (4) intermittent four-day schedule B: participants received their pre-study dosage Monday, Wednesday, Friday and Sunday, and placebo on Tuesday, Thursday and Saturday. However, these strategies are analysed collectively as one 'intermittent technique', as individually study data for each group were not reported. All participants had received stable doses of antipsychotic medication during the six months preceding the study; 48% of participants were receiving chlorpromazine (mean daily dose 462 mg), 46% were receiving thioridazine (mean daily dose 362 mg), 27% were receiving trifluoperazine (mean daily dose 15 mg), and 4% were receiving perphenazine (mean daily dose 28 mg), 25% of participants were receiving more than one drug. The intermittent technique employed in Remington 2011 required participants to receive the same daily dose administered every other day (either risperidone or olanzapine; dosages not specified). Participants in Shenoy 1981 received a six-week drug discontinuation technique, in which they were given a placebo injection and were returned to their routine active medication at the end of the study (dosages not specified).

5.2 Maintenance therapy

In each of the included studies, various antipsychotics were administered throughout maintenance therapy in comparison with the above mentioned intermittent strategies - all dosages were classified as either low/moderate. The majority of studies measured antipsychotic usage through chlorpromazine equivalents. Carpenter 1987 involved minimum daily chlorpromazine equivalent doses of 300 mg, combined with brief visits with a pharmacotherapist on alternate weeks; the mean daily dose totaling 720 mg ± 732 mg chlorpromazine equivalents; Carpenter 1990* employed the same continuous dose technique, but with a mean daily dose of 433 mg ± 46 mg chlorpromazine equivalents; participants in Herz 1991* received their usual dose of antipsychotic medication with an average cumulative antipsychotic drug dosage over two years of 424.84 mg ± 333.05 mg chlorpromazine equivalents; in Pietzcker 1993*, the type and application of antipsychotic drugs used throughout the study was not restricted and participants received continuous administration of medication, with doses individually adjusted in accordance with the patient's clinical demands at a given time, minimal dosage of 100 mg chlorpromazine equivalents, with a mean over two years of 210 mg; and in Wiedemann 2001, the same antipsychotic dose level was maintained throughout the study period, with a mean daily dosage of 314 mg ± 150 mg chlorpromazine equivalents) and haloperidol equivalents. Carpenter 1990* also presented a mean daily dose of 11.8 mg ± 4.4 mg haloperidol equivalent; participants in Gaebel 2011 were maintained on a specified drug regimen with a mean daily dose of 3 mg/day haloperidol equivalent; and participants in Wunderink 2007 were continuously treated with low-dose second-generation antipsychotics, measured as a mean daily dose of 2.94 mg haloperidol equivalent.

Five trials employed the use of continuous fluphenazine decanoate (participants in Jolley 1989/1990 continued to receive fluphenazine decanoate in clinically optimal (pre-trial) doses, but the average cumulative antipsychotic drug dosage was measured in haloperidol equivalents, which over two years equalled 1616 mg ± 598 mg; continued fluphenazine decanoate was employed in McCreadie 1980 , with a mean dose of 12.5 mg/IM (maximum 50 mg weekly); similarly, McCreadie 1982 continued participants on IM fluphenazine decanoate, with a mean dose of 2 mg to 25 mg; participants in Schooler 1997 were maintained on either a standard dose of fluphenazine decanoate of 12.5 mg to 50 mg every two weeks or a low-dose of 2.5 mg to 10 mg every two weeks; and participants in Shenoy 1981 were 'continued on their regular medication' and received a mean dose of 39.3 mg fluphenazine decanoate).

Participants in Blackburn 1961 had received initial daily doses of either prochlorperazine (15 mg to 150 mg), perphenazine (12 mg to 24 mg), chlorpromazine (50 mg to 800 mg), promazine (200 mg to 44 mg), trifluoperazine (6 mg) - when assigned to maintenance therapy, participants continued on their pre-study regimen. The same procedure was followed in Prien 1973, where participants had received stable doses of antipsychotic medication during the six months preceding the study; 48% of participants were receiving chlorpromazine (mean daily dose 462 mg), 46% were receiving thioridazine (mean daily dose 362 mg), 27% were receiving trifluoperazine (mean daily dose 15 mg), and 4% were receiving perphenazine (mean daily dose 28 mg); 25% of participants were receiving more than one drug.

Participants in Caffey 1964 continued to receive either chlorpromazine or thioridazine daily, administered in standard 100 mg tablets - participants on chlorpromazine had been receiving it for over two years at an average daily dose of 400 mg and participants on thioridazine had been receiving it for one year at an average daily dose of 350 mg.

Some trials did not describe the dosage or even the type of antipsychotic administered throughout maintenance treatment (participants were described in Olson 1962* as having received medication 'in standard form'; and in Remington 2011, the control group received "treatment as usual", receiving either risperidone, olanzapine or loxapine, without specifying the dosages).

5.3 Placebo

Three trials included a placebo group in addition to an intermittent and maintenance group. Particpants assigned to intermittent treatment in Blackburn 1961 were further subdivided into two groups to receive either a placebo drug or placebo-placebo regimen (thiamine chloride). Those who received placebo-placebo were kept on this regimen for the full length of the study. Results for the individual subdivided groups are presented only by medium term; at short term, the results are combined. In Caffey 1964, there were two corresponding placebo groups for both the intermittent and maintenance; in the former, participants received the same number of tablets they had prior to the study in placebo form on a daily basis, and participants in the latter received the same number of tablets prior to the study in placebo form on an intermittent basis (Monday, Wednesday and Friday only). The placebo group in Olson 1962* alternated monthly between the active drug (one half of the group received phenothiazine and one half received chlorpromazine) and placebo.

6. Outcomes: Scales providing useable data

Rating scales that provided useful data are described below.

6.1. Global state

6.1.1 Clinical Global Impression Scale (CGI, Guy 1970)
This scale is a three-item observer rated scale that measures severity of illness (CGI-S); global improvement or change (CGI-C); and therapeutic response. The CGI-S and CGI-C are rated on a seven-point scale from 1 = normal to 7 = extremely unwell; lower scores indicative of decreased severity and/or greater recovery, and the treatment response ratings take into account both therapeutic efficacy and treatment-related adverse effects, and range from 0 = marker improvement and no side effects, to 4 = unchanged or worse with side effects outweighing therapeutic effects. Each component of the CGI is rated separately, and the instrument does not provide a global score. Gaebel 2011 and Remington 2011 both reported data using the CGI-S component; a factor contributing towards the definition of relapse in Gaebel 2011 was a CGI-S change score of six or more, as well as PANSS positive score of more than 10, and a decrease in GAF score of more than 20. Pietzcker 1993* also noted deterioration on the CGI-S indicated by a score of > 7 as a determinant contributing towards their definition of relapse (see below).

6.1.2 Global Assessment Scale (GAS, Endicott 1976)
This single-item rating scale assesses one's overall functioning whilst diagnosed with a mental disorder during a specific time period. The scale scores from 1 = extremely unwell to 100 = extremely well and is divided into 10 equal intervals and the lower the mean value the more constant the symptomology. Carpenter 1987; Herz 1991*; Shenoy 1981; Wiedemann 2001 each reported data using the GAS; Herz 1991* defined 'relapse' as a decrease of ≤ 30 using GAS; and the definition of 'relapse' in Pietzcker 1993* included three criteria: i) an increase of > 10 in the psychosis factor (HOST, THOT, ACTV) of the BPRS; ii) a decrease of > 20 in the GAS; iii) and deterioration on the CGI scale indicated by a score of > 7. Relapse predicted by prodromal symptoms.

6.2. Mental state

6.2.1 Brief Psychaitric Rating Scale (BPRS, Overall 1962)
This scale measures positive symptoms and quantifies factors such as thought disorder, activation, hostility. somatic, hallucinatory, and depressive states. The original scale had 16 items, but a revised 18-item scale is more commonly used, with scores ranging from 0 - 126. Each item is defined on a seven-point scale from 0 = not present to 7 = extremely severe. Higher scores equate to severity of illness. In their study Remington 2011 defined relapse as a 20% increase in overall symptoms using the BPRS; however, Carpenter 1987 and Wiedemann 2001 were the only studies to report usable outcome data using this scale.

6.2.2 Positive and Negative Symptom Scale (PANSS, Kay 1987)
PANSS was developed from the BPRS and the Psychopathology Rating Scale. It is used to evaluate positive, negative and other symptom dimensions in schizophrenia. The scale has 30 items, each measured on a seven-point scoring system varying from 1 = absent to 7 = extreme. Gaebel 2011 and Wunderink 2007 were the only studies that reported data measuring both positive and negative symptoms.

6.2.3 Scale for the Assessment of Negative Symptoms (SANS, Andreasen 1982)
The SANS measures the incidence and severity of negative symptoms using a 25-item scale, using a six-point scoring system, where 0 = better to 5 = worse, where a higher score equals a more severe experience of negative symptoms. Gaebel 2011 was the only study that reported data using this scale.

6.3. General functioning

6.3.1 Global Assessement of Functioning (GAF, APA 1994)
The GAF subjectively measures overall psychiatric disturbance over a specified time period on a continuum from psychological or psychiatric sickness to health. Psychological, social and occupational functioning is assessed using a rating scale of 0 - 90, where 0 = severe symptoms and 90 = no present symptoms. Gaebel 2011 was the only study to report data from this scale.

6.3.2 Groningen Social Disabilities Schedule (GSDS, Wiersma 1990)
This scale measures social disabilities through a semi-structured interview with observer ratings of functioning in eight social role domains: vocational functioning, community integration, peer relationships, relationship with family members, parental functioning, partner relationship, house-keeping and self-care. Disabilities are rated on a four-point scale from no disability to serious disability. Total disability scores range from 0 - 21 and are calculated combining seven domains, excluding parental functioning because of limited applicability. Wunderink 2007 was the only study to report data from this scale, however data were skewed.

6.3.3 Level of Functioning Scale (LOFS, Hawk 1975)
The LOFS was designed to assess work functioning (occupational and overall level of function), social functioning (quality and quantity of social relationships, fullness of life, ability to meet basic needs) and hospitalisation. Scores from each area of functioning range from 0 = worst outcome, to 36 = best outcome, which is calculated by adding the nine outcome item scores.

6.4 Problem Appraisal Scale (PAS, Spitzer 1971)
The PAS in a one-page form, designed for use in emergency room or outpatient settings where extensive and detailed mental status examinations are generally not done. This scale provides scaled judgements of 38 areas of psychiatric disturbance (broken down into physical function; intellectual development; social relations; social performance; and other signs and symptoms). These dimensions are rated on a five-point scale, with 1 = none and 5 = severe over a time period of the last two weeks. Herz 1991* was the only study to report data using this scale.

6.3.5 Social Adjustment Scale (SAS, Weissman 1976)
The SAS contains 42 questions that measure either instrumental or expressive role performance in six major areas of functioning: work as a worker, housewife or student; social and leisure activities; relationships with extended families; marital role as a spouse, a parent and a member of the family unit. Each question is rated on a five-point scale, with a higher score indicating impairment. Wiedemann 2001 was the only study to report data using this scale.

6.4. Adverse effects

6.4.1 Abnormal Involuntary Movement Scale (AIMS, Guy 1976)
This scale measures the examination of involuntary movements (tardive dyskinesia) consisting of 12 items scored from 0 = none to 4 = severe, quantifying the severity of tardive dyskinesia. This trial used in short-term trials may also help to assess parkinsonian symptoms such as tremor. Only Gaebel 2011 and Remington 2011 reported usable data with this scale - however the data from each of these studies are skewed.

6.4.2 Extrapyramidal Side Effects Scale (EPS, Simpson 1970)
This is an observer-rated scale designed for assessing parkinsonian and related extrapyramidal side effects. The scale has 10 items - including gait, arm dropping, shoulder shaking, elbow rigidity, wrist rigidity, leg pendulousness, head dropping, glabella tap, tremor and salivation - each measured on a five-point scoring system from 0 = absent to 4 = severe). Gaebel 2011 and Jolley 1989/1990 both reported data from this scale.

6.4.3 Hillside Akathisia Scale (HAS, Fleischhacker 1989)
The HAS was used to measure akathisia; the subjective subscale has two subjective and three objective items for which anchored rating points are provided. The subjective items take into account a patient's sensation of restlessness and urge to move, and the objective items assess physical signs of akathisia present in the head, trunk, hands, arms, feet and legs. There are a total of five items, which are measured on a five-point scoring system from 0 = absent to 4 = present and not controllable. Gaebel 2011 was the only trial to report data using this scale.

6.4.4 Liverpool University Neuroleptic Side Effect Rating Scale (LUNSERS, Day 1995; Lambert 2003)
LUNSERS is a self-rating scale to measure antipsychotic side effects with 41 items. These side effects are rated on a five-point scale, from 0 = absent, to 4 = very much. Wunderink 2007 was the only trial to report data using this scale.

6.4.5 Udvalg for Kliniske Undersogelser Side Effects Rating Scale (UKU, UKU 1987)
The UKU is a clinician-rated scale developed to provide comprehensive side-effect ratings of psychopharmacological medications. There are 48 items to the scale, and each are defined by means of a four-point scale, starting from 0 = not or doubtfully present, ranging to 3 = severe. Gaebel 2011 was the only trial to report data using this scale, however, these data are skewed.

6.5. Quality of life

6.5.1 Lancashire Quality of Life Profile (LQLP, Oliver 1991)
This 27-item scale presents both objective quality of life indicators and subjective quality of life estimates; the subjective elements focus on five specific domains, including work and education, leisure and participation, religion, finances, living situation, legal and safety, family relations, social relations, and health. Participants rate their satisfaction with these factors on a seven-point scale (1 = 'can't be worse', to 7 = 'can't be better'), the higher the score indicating a better quality of life. This scale has been subject to thorough examination to assess its validity, but is nevertheless widely used in Europe as a measurement to assess quality of life (van Nieuwenhuizen 2001). Gaebel 2011 was the only trial to report data using this scale.

6.5.2 Quality of Life Scale (QLS, Heinrichs 1984)
The QLS is a 21-item clinician-administered scale rated from a semi-structured interview to assess symptoms and functioning. Each item is rated on a seven-point scale and requires judgement by a clinician/interviewer on all but two cases. Each item is composed of three parts; (1) brief descriptive statement to focus the interviewer on the judgement to be made, (2) a set of suggestive probes, (3) the seven-point scale with descriptive anchors for every other point - with a low score 0 = severe impairment, and 6 = normal/unimpaired functioning.

6.5.3 World Health Organization Quality of Life Scale (WHOQoL-Bref, O'Carroll 2000)
The WHOQoL-Bref is a 26-item self-report comprising satisfaction with health, psychological functioning, social relationships and environmental opportunities. Each item is scored on a five-point scale from 1 = poor to 5 = worse. Wunderink 2007 was the only trial that reported data using this scale.

Excluded studies

Eight studies were excluded altogether; Caffey 1975; Hymowitz 1980 and Uchida 2008 were each excluded because they were not randomised. Engelhart 2002; Levine 1980; Newton 1989 and Strauss 1990 were excluded because the trials delivered no usable data; and Docherty 2003 was excluded as there was no intermittent treatment group by which to draw comparison with continuous antipsychotic maintenance treatment.

Ongoing studies

There is one known ongoing study (UMIN Clinical Trials Registry (UMIN-CTR) unique trial number: UMIN000006011, Uchida 2013); as of January 2012, participants were in the process of recruitment, with a target date for completion of late 2013.

Awaiting assessment

There are no studies awaiting assessment at the time of writing this review.

Risk of bias in included studies

For a graphical overview please see Figure 2 and Figure 3.

Figure 2.

'Risk of bias' graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Figure 3.

Methodological quality summary: review authors' judgements about each methodological quality item for each included study.

Allocation

All 17 included trials were described as randomised; the methods used to describe allocation and concealment, however, were poorly described. Only four trials provided an adequate description of 'randomisation' (Gaebel 2011; Herz 1991*; Schooler 1997; Wunderink 2007), with six studies poorly describing the allocation methods used (Carpenter 1987; Carpenter 1990*; Olson 1962*; Pietzcker 1993*; Remington 2011; Wiedemann 2001) and the final seven provided no further description beyond stating that 'participants were randomised' or 'randomly assigned' (Blackburn 1961; Caffey 1964; Jolley 1989/1990; McCreadie 1980; McCreadie 1982; Prien 1973; Shenoy 1981). One study was rated as having a high risk of bias (Jolley 1989/1990) because it was stated that the participating clinicians were requested to refer patients whom they thought might benefit from the brief intermittent treatment approach.

Blinding

Overall, the description of blinding was poor throughout the records of the included studies; only five trials gave adequate description of methods used for blinding (Caffey 1964; McCreadie 1980; McCreadie 1982; Olson 1962*; Remington 2011). Three trials were rated as having a high risk of bias because, even though a double-blind method was expressed, treatment was open to some participants depending on their point of entry into the study ('lateral entry' participants in Gaebel 2011) and when participants required active antipsychotic medication as an intervention (Herz 1991*; Schooler 1997). Four studies did not employ blinding methods (Carpenter 1987; Pietzcker 1993*; Wiedemann 2001; Wunderink 2007) and one study was described as single (assessor) blind (Carpenter 1990*). The remaining studies were rated as an unclear risk of bias, as there was little/no detail as to blinding (Carpenter 1990*; Jolley 1989/1990; Prien 1973; Shenoy 1981).

Incomplete outcome data

For this review, high loss to follow-up was defined in trials where the number of participants lost exceeded 50% overall. Only one included study satisfied this criteria; Pietzcker 1993* had a high loss to follow-up, with only 44% completing the study. Therefore, only the outcome of 'leaving the study early' was included in the data and analysis, as per our protocol, which stated that should more than 50% of data go unaccounted for, we would not reproduce these data or use them within analyses.
Several studies had a moderate level of participants lost to follow-up (ranging between 34% to 48%), including McCreadie 1982; Olson 1962* and Schooler 1997. As did Carpenter 1987; Carpenter 1990*; Herz 1991* and Wiedemann 2001 studies, which presented completer-only data in their results. In Gaebel 2011, attrition was moderate, yet contentious, as the included participants were 're-randomised' following a previous trial by the same authors (one year of maintenance therapy comparing risperidone versus haloperidol); however, out of those randomised, n = 44 were treated and included in an intention-to-treat analysis.

Studies with low levels of attrition included Blackburn 1961; Jolley 1989/1990; McCreadie 1980; Remington 2011; Shenoy 1981; Wunderink 2007; each of these trials provided specific reasons why participants may have left the studies early. Caffey 1964 had a low attrition rate, and completer only data were used in their analysis, which included the participants in each group of the study that did not relapse, however, not all of the 'completer only' data were represented, making the results more susceptible to bias and therefore we rated this study as a 'high risk'. Follow-up data were not made clear in Prien 1973.

Overall, poor reporting at each stage of the included studies made it difficult to ascertain the true numbers of participants who left the study early - use of a flow chart would have been an appropriate measure to accurately reflect these numbers. Two studies did not make clear the number of participants lost to follow-up (Caffey 1964; Prien 1973), and most studies either (i) did not indicate whether they reported completer analyses; (ii) did not specify whether completer or intention-to-treat analyses were employed; or (iii) presented highly skewed data.

Selective reporting

The majority of studies were rated as having a high risk of bias, as not all outcomes were reported, in particular, data from rating scales went unreported from the time intervals specified in the study protocols. Only one study reported all specified outcomes and was therefore rated as a low risk of bias (Wunderink 2007). Two studies were rated as an 'unclear' risk; Carpenter 1987 reported scale data but only by two years, where it was specified that criterion instruments would be completed at six-month intervals; and Gaebel 2011, as mentioned above, was ambivalent; it was stated that participants from 13 German psychiatric hospitals were included, however, participants from only eight hospitals were mentioned, with no reasons for this reported.

Other potential sources of bias

1. Funding:

The majority of studies reported sources of funding, and included support grants from NIMH (Carpenter 1987; Carpenter 1990*; Herz 1991*); Department of Health and Social Services (Jolley 1989/1990); Netherlands Organisation for Health Research and Development (Wunderink 2007); the German Ministry of Research and Technology (Wiedemann 2001); Dumfries and Galloway Health Board (McCreadie 1982); the German Federal Ministry for Education and Research BMBF (Gaebel 2011); and a National Alliance for Research on Schizophrenia and Depression (NARSAD) Independent Award (Remington 2011). No studies described direct funding from pharmaceutical companies, but several trials acknowledged study drugs donated by Sandoz and Smith, Kline & French (Caffey 1964); Janssen Cilag and Wyeth-Pharma (Gaebel 2011); Janssen Pharamceutical and ER Squibb Limited (Jolley 1989/1990; McCreadie 1980; McCreadie 1982); and Eli Lilly Nederland B.V. (Wunderink 2007).

2. Rating scales:

The majority of studies did not describe whether raters of measurement scales were independent (Herz 1991*; Olson 1962*; Pietzcker 1993*; Prien 1973; Remington 2011; Schooler 1997; Wiedemann 2001; Wunderink 2007), and in two studies raters were not independent (Gaebel 2011; Shenoy 1981). Only seven studies gave details on how ratings scales were administered and the extent to which raters were independent (Blackburn 1961; Caffey 1964; Carpenter 1987; Carpenter 1990*; Jolley 1989/1990; McCreadie 1980; McCreadie 1982).

Effects of interventions

See: Summary of findings for the main comparison ANY INTERMITTENT DRUG TECHNIQUE compared with MAINTENANCE THERAPY for schizophrenia; Summary of findings 2 INTERMITTENT (EARLY-BASED) compared with MAINTENANCE THERAPY for schizophrenia; Summary of findings 3 INTERMITTENT (CRISIS INTERVENTION) compared with MAINTENANCE THERAPY for schizophrenia; Summary of findings 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) compared with MAINTENANCE THERAPY for schizophrenia; Summary of findings 5 INTERMITTENT (DRUG HOLIDAY) compared with MAINTENANCE THERAPY for schizophrenia; Summary of findings 6 ANY INTERMITTENT DRUG TECHNIQUE compared with PLACEBO for schizophrenia

Comparison 1. ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY

All 17 included studies were included in this comparison, which included any type of intermittent drug therapy, with a total n = 2252. Twelve of the included studies reported data for the primary outcome of relapse (n = 1327) and six reported data for the second primary outcome of hospitalisation (n = 661). All the remaining studies reported for outcomes of either death, global state, general functioning, service outcomes, adverse effects, quality of life and leaving the study early, with the primary objective of evaluating the effects of intermittent drug techniques on these outcomes - participants had to be clinically stabilised to some extent to participate in the study.

1.1 Relapse

Data demonstrate that there was a significantly greater risk of relapse for participants receiving intermittent drug therapy; results were homogenous, clinically and statistically significant in favour of maintenance therapy at each short term (n = 396, 4 RCTs, risk ratio (RR) 1.68, 95% confidence interval (CI) 1.00 to 2.81), medium term (n = 774, 5 RCTs, RR 2.41, 95% CI 1.50 to 3.86) and long term (n = 436, 7 RCTs, RR 2.46, 95% CI 1.70 to 3.54, Analysis 1.1).

1.2 Hospitalisation

At medium term, there was little difference between groups receiving intermittent therapy or maintenance therapy (n = 136, 2 RCTs, RR 1.03, 95% CI 0.07 to 14.98), however, results displayed significant heterogeneity (Chi² = 3.19, P = 0.07, I² = 69%) and were displayed using a random-effects model. At long term, results were homogenous and statistically significant in favour of maintenance therapy, with a greater risk of hospitalisation for people receiving intermittent therapy (n = 626, 5 RCTs, RR 1.65, 95% CI 1.33 to 2.06, Analysis 1.2).

1.3 Death

People on maintenance treatment were less likely to die in the long term than those on intermittent treatment (1/77 versus 4/78), however the data were from only two trials, and did not rule out the effects of chance (n = 155, 2 RCTs, RR 0.34, 95% CI 0.05 to 2.08, Analysis 1.3).

1.4 Global state
1.4.1. Average score (CGI-S, high = worse)

Symptom severity was measured using the CGI-S scale and did not significantly differ between intermittent treatment and maintenance treatment at medium term (n = 35, 1 RCT) or long term (n = 27, 1 RCT, Analysis 1.4); substantial heterogeneity was present (Chi² = 2.38, P = 0.12, I² = 58%) and therefore displayed using a random-effects model.

1.4.2. Average score (GAS, low = worse)

Symptomology was assessed using the GAS, and significantly favoured maintenance therapy in the short term (n = 31, 1 RCT, mean difference (MD) -3.61, 95% CI -5.67 to -1.55), but displayed little difference at both medium (n = 126, 2 RCTs, MD -0.45, 95% CI -4.55 to 3.66) and long term (n = 133, 3 RCTs, MD 1.32, 95% CI -2.75 to 5.39, Analysis 1.5); again, substantial heterogeneity was present (Chi² = 5.37, P = 0.07, I² = 62.7%).

1.4.3. Prodromal episodes

By long term, people receiving intermittent treatment were at a significantly greater risk of experiencing prodromal episodes, with significant favour for maintenance therapy (n = 155, 2 RCTs, RR 3.19, 95% CI 1.87 to 5.44, Analysis 1.6).

1.4.4. Significant improvement

One study also measured 'significant improvement', which was significantly demonstrated in people receiving maintenance therapy at both short term (n = 60, 1 RCT, RR 0.40, 95% CI 0.18 to 0.89) and medium term (n = 60, 1 RCT, RR 0.32, 95% CI 0.15 to 0.68, Analysis 1.7).

1.5 Mental state
1.5.1. Average score (BPRS, high = worse)

Mental state did not significantly differ between Intermittent and maintenance treatment over the medium term (n = 51, 1 RCT) or long term (n = 77, 2 RCTs) in trials that measured this outcome with the BPRS (Analysis 1.8).

1.5.2. Negative symptom score (PANSS negative symptom subscale, high = worse)

Negative symptoms were assessed in one RCT using the PANSS negative symptom subscale; there was no significant difference between intermittent and maintenance treatment groups by medium term (n = 128, 1 RCT) or long term (n = 128, 1 RCT, Analysis 1.9).

1.5.3. Negative symptom score (PANSS negative symptom subscale, high = worse, skew)

There were also skewed data reported using the PANSS negative symptom subscale, and these are reported separately (Analysis 1.10).

1.5.4. Negative symptom score (SANS negative symptom subscale, high = worse, skew)

This was also the case for assessing negative symptoms using SANS; these data are skewed and should be interpreted with caution (Analysis 1.11).

1.5.5. Positive symptom score (PANSS positive symptom subscale, high = worse)

Positive symptoms were measured using the PANSS positive symptom subscale, the results of which significantly favoured intermittent treatment by medium term (n = 128, 1 RCT, MD -0.80, 95% CI -1.58 to -0.02) but demonstrated no significant difference at long term (n = 155, 2 RCTs, MD 0.40, 95% CI -0.79 to 1.59, Analysis 1.12).

1.6 General functioning
1.6.1. Average endpoint (LOFS, low = worse)

Overall functioning was measured using the LOFS by one RCT, however there was no significant difference between intermittent or maintenance therapy (Analysis 1.13).

1.6.2. Social functioning score (GAF, low = worse)

Psychiatric disturbance was measured by one study using the GAF, which demonstrated significant favour of maintenance therapy by long term (n = 27, 1 RCT, MD -9.00, 95% CI -15.92 to -2.08, Analysis 1.14).

1.6.3. Social functioning score (GSDS, high = worse, skew)

The GSDS measured levels of social disabilities - note that these data are skewed and are presented in an additional table (Analysis 1.15).

1.6.4. Social functioning score (PAS, high = worse)

Psychiatric disturbance was measured using the PAS, but results were no different at both medium term (n = 75, 1 RCT) and long term (n = 56, 1 RCT, Analysis 1.16).

1.6.5. Social functioning score (SAS, high = worse)

Finally, social adjustment was measured in one RCT using the SAS, which again demonstrated no significant difference between interventions at both medium term (n = 51, 1 RCT) and long term (n = 51, 1 RCT, Analysis 1.17).

1.7 Adverse effects
1.7.1. Akathisia (HAS, high = worse, skew)

The HAS was employed to measure akathisia. These results should be interpreted carefully, however, as only one study reported data using this scale, and the data displayed are skewed, with high standard deviations (Analysis 1.18).

1.7.2. Extrapyrimidal side effects

Only one trial reported data for specific extrapyramidal side effects, which demonstrated a slightly higher (non-significant) instance of hypomimia (lack of facial expression) for maintenance therapy at medium term (n = 43, 1 RCT, RR 0.26, 95% CI 0.06 to 1.09) and long term (n = 43, 1 RCT, RR 0.17, 95% CI 0.02 to 1.33); no difference was found for rigidity between groups at medium term (n = 43, 1 RCT) and long term (n = 43, 1 RCT); no significant difference between groups for instance of tremor at medium term (n = 43, 1 RCT) and long term (n = 43, 1 RCT); no significant difference between groups for instance of akathisia at medium term (n = 43, 1 RCT), but with a significant difference favouring intermittent therapy at long term (n = 43, 1 RCT, RR 0.19, 95% CI 0.05 to 0.76); no significant difference between groups for instance of gait abnormality for maintenance therapy at medium term (n = 43, 1 RCT) and long term (n = 43, 1 RCT); no significant difference between groups for instance of parkinsonism at medium term (n = 43, 1 RCT), but with a significant difference favouring intermittent therapy at long term (n = 43, 1 RCT, RR 0.13, 95% CI 0.02 to 0.96); and a significantly higher risk of global non-liveliness with maintenance therapy at both medium term (n = 43, 1 RCT, RR 0.24, 95% CI 0.08 to 0.73) and long term (n = 43, 1 RCT, RR 0.09, 95% CI 0.01 to 0.61, Analysis 1.19).

1.7.3. Extrapyrimidal symptoms (EPS, high = worse, skew)

Parkinsonian and related extrapyramidal side effects were also measured using the EPS scale, which presented skewed data and are best viewed by inspection of the 'other data' table (Analysis 1.20).

1.7.4. Need for additional medication

There was no significant difference in the number of people requiring additional medication (n = 346, 2 RCTs, Analysis 1.21).

1.7.5. Side effects (LUNSERS, high = worse, skew)

Side effects measured using LUNSERS presented skewed data and should be interpreted carefully (Analysis 1.22).

1.7.6. Side effects (UKU, high = worse, skew)

Side effects measured using UKU presented skewed data and should be interpreted carefully (Analysis 1.23) .

1.7.7. Tardive dyskinesia

By medium term, there was no significant difference when measuring prevalence of tardive dyskinesia at medium term (n = 43, 1 RCT) or by long term (n = 165, 4 RCTs, Analysis 1.24).

1.7.8. Tardive dyskinesia (AIMS, high = worse, skew)

Intermittent treatment groups generally scored lower on the AIMS at both medium and long term - these results, however, are skewed and should be interpreted with caution (Analysis 1.25).

1.8 Quality of life
1.8.1. Average score (LQLP, low = worse)

There was no significant difference between groups when assessing quality of life scores at medium term, using LQLP (n = 27, 1 RCT) with lower scores in the intermittent treatment group in the one RCT that reported data (Analysis 1.26).

1.8.2. Average score (GLS, low = worse)

Data reported from one RCT using the QLS demonstrated no significant difference between groups by long term (n = 26, 1 RCT, Analysis 1.27).

1.8.3. Average score (WHOQoL-Bref, low = worse)

Using the WHOQoL-Bref, there was no difference between groups at both medium term (n = 128, 1 RCT) and long term (n = 128, 1 RCT, Analysis 1.28).

1.9 Leaving the study early/loss to follow-up

There was no significant difference between groups when assessing people leaving the study early for any reason by short term (n = 31, 1 RCT), with equivocal results at medium term (n = 155, 3 RCTs, RR 1.40, 95% CI 0.25 to 7.82); there was a trend favouring maintenance therapy over intermittent therapy by long term (n = 996, 10 RCTs, RR 1.63, 95% CI 1.23 to 2.15, Analysis 1.29). There was moderate heterogeneity between these results, however, and they are presented using a random-effects model (Chi² = 31.25, P = 0.003, I² = 58%).

Comparison 2: INTERMITTENT (EARLY-BASED) versus MAINTENANCE THERAPY

Five studies compared intermittent (early-based) therapy with maintenance therapy (Carpenter 1987; Carpenter 1990*; Herz 1991*; Jolley 1989/1990; Schooler 1997, n = 626).

2.1 Relapse

Data from two RCTs showed that people receiving intermittent (early-based) treatment were more likely to relapse, with significant favour of maintenance therapy at long term (n = 155, 2 RCTs, RR 2.33, 95% CI 1.32 to 4.12, Analysis 2.1).

2.2 Hospitalisation

At medium term, no significant difference between groups for rates of hospitalisation (n = 101, 1 RCT) but at long term, results displayed statistical significance in favour of maintenance treatment (n = 625, 5 RCTs, RR 1.66, 95% CI 1.33 to 2.08, Analysis 2.2).

2.3 Death

Only two studies reported the outcome of death, with people receiving maintenance therapy at greater risk in the long term (n = 155, 2 RCTs, RR 0.34, 95% CI 0.05 to 2.08, Analysis 2.3).

2.4 Global state
2.4.1. Average score (GAS, low = worse)

Medium-term data for global state using the GAS showed an equivocal effect between intermittent (early) treatment and maintenance therapy at medium term (n = 75, 1 RCT, MD -0.15, 95% CI -4.78 to 4.48) and long term (n = 82, 2 RCTs, MD 0.99, 95% CI -4.24 to 6.22, Analysis 2.4).

2.4.2. Prodromal episodes

Instances of prodromal episodes were also significantly more prevalent in intermittent (early) treatment, with statistically significant favour of maintenance treatment by long term (n = 155, 2 RCTs, RR 3.19, 95% CI 1.87 to 5.44, Analysis 2.5).

2.5 Mental state
2.5.1 Average score (BPRS, high = worse)

One study reported data using the BPRS, which demonstrated an equivocal effect by long term (n = 26, 1 RCT, MD 0.10, 95% CI -0.33 to 0.53, Analysis 2.6).

2.6 General functioning
2.6.1. Average endpoint (LOFS, low = worse)

In the one RCT that assessed level of functioning using the LOFS, results were equivocal by long term (n = 26, 1 RCT, MD 2.60, 95% CI -2.63 to 7.83, Analysis 2.7).

2.6.2. Social (PAS, high = worse)

There was no difference between groups in the one study that assessed functioning using the PAS, at medium (n = 75, 1 RCT) and long term (n = 56, 1 RCT, Analysis 2.8).

2.7 Adverse effects
2.7.1. Extrapyramidal side effects

Only one trial reported data for specific extrapyramidal side effects, which demonstrated no significant difference between groups for instance of hypomimia for maintenance therapy at medium term (n = 43, 1 RCT) and long term (n = 43, 1 RCT); no significant difference was found between groups instance of rigidity at medium term (n = 43, 1 RCT) and long term (n = 43, 1 RCT); no significant difference was found between groups for instance of tremor for maintenance therapy at medium term (n = 43, 1 RCT) and long term (n = 43, 1 RCT); no significant difference was found between groups for instance of akathisia at medium term (n = 43, 1 RCT), but with a significant outcome favouring intermittent (early) therapy at long term (n = 43, 1 RCT, RR 0.19, 95% CI 0.05 to 0.76); no significant difference was found between groups for instance of gait abnormality at medium term (n = 43, 1 RCT) and long term (n = 43, 1 RCT); no significant difference was found for instance of parkinsonism at medium term (n = 43, 1 RCT), but with a significant outcome favouring intermittent (early) therapy at long term (n = 43, 1 RCT, RR 0.13, 95% CI 0.02 to 0.96); and a significantly higher risk of global non-liveliness with maintenance therapy at both medium term (n = 43, 1 RCT, RR 0.24, 95% CI 0.08 to 0.73) and long term (n = 43, 1 RCT, RR 0.09, 95% CI 0.01 to 0.61, Analysis 2.9).

2.7.2. Extrapyramidal symptoms (EPS, high = worse, skew)

Parkinsonian and related extrapyramidal side effects were also measured using the EPS scale, which presented skewed data and are these are best viewed by inspection of the 'other' data table (Analysis 2.10).

2.7.3. Need for additional medication

In the one RCT that compared intermittent (early) treatment with maintenance therapy, there was no difference in the amount of people who required additional medication (n = 313, 1 RCT, Analysis 2.11).

2.7.4. Tardive dyskinesia

By medium term, there was no significant difference between groups when measuring prevalence of tardive dyskinesia at medium term (n = 43, 1 RCT) or by long term (n = 30, 1 RCT, Analysis 2.12).

2.8 Quality of life
2.8.1. Average score (QLS, low = worse)

Data reported from one RCT using the QLS demonstrated no significant difference by long term (n = 26, 1 RCT, Analysis 2.13).

2.9 Leaving the study early/loss to follow-up

There was a trend favouring maintenance therapy over intermittent therapy by long term (n = 562, 5 RCTs, RR 1.67, 95% CI 1.17 to 2.37, Analysis 2.14). There was moderate heterogeneity between these results, however, and they are presented using a random-effects model (Chi² = 9.45, P = 0.05, I² = 58%).

Comparison 3: INTERMITTENT (CRISIS INTERVENTION) versus MAINTENANCE THERAPY

Only one study reported data for this comparison.Pietzcker 1993* had three treatment arms, and so this comparison assesses the outcomes of participants in that study receiving intermittent (crisis) and maintenance therapy (n = 237). As loss to follow-up was at 56%, only data for the outcome of leaving the study early are presented.

3.1 Leaving the study early/loss to follow-up

By long term, there was a significantly greater number of participants receiving intermittent (crisis) therapy that left the study early or were lost to follow-up (n = 237, 1 RCT, RR 1.57, 95% CI 1.23 to 2.00, Analysis 3.1).

Comparison 4: INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY

Three studies provided data for this comparison (Gaebel 2011; Wiedemann 2001; Wunderink 2007, n = 260).

4.1 Relapse

Data from three RCTs showed that people receiving intermittent (gradually increased drug-free) treatment were more likely to relapse, with no significant difference at short (n = 128, 1 RCT) and medium term (n = 128, 1 RCT), but with significant favour of maintenance therapy at long term (n = 219, 3 RCTs, RR 2.76, 95% CI 1.63 to 4.67, Analysis 4.1). Heterogenity was present with long-term data, but this was only slight (Chi² = 2.89, P = 0.24, I² = 31%), and results remain in data and analysis using a fixed-effect model.

4.2 Global state
4.2.1. Average score (CGI-S, high = worse)

Symptom severity was measured by one RCT using the CGI-S scale, which demonstrated no significant difference between groups by long term (n = 27, 1 RCT, Analysis 4.2).

4.2.2. Average score (GAS, low = worse)

Results were equivocal when using the GAS at medium term (n = 51, 1 RCT, MD -1.54, 95% CI -10.42 to 7.34) and long term (n = 51, 1 RCT, MD 1.83, 95% CI -4.66 to 8.32, Analysis 4.3).

4.3 Mental state
4.3.1. Average score (BPRS, high = worse)

Using the BPRS, there was little difference in levels of psychiatric symptoms between people receiving intermittent or maintenance therapy, with no significant difference between groups at medium term (n = 51, 1 RCT) or long term (n = 51, RCT, MD 0.20, Analysis 4.4).

4.3.2. Negative symptoms score (PANSS negative symptom subscale, high = worse)

Negative symptom scores using PANSS tended to be slightly higher with maintenance therapy, but with no significant difference between groups at both medium (n = 128, 1 RCT) and long term (n = 128, 1 RCT, Analysis 4.5)

4.3.3. Negative symptoms score (PANSS negative symptom subscale, high = worse, skew)

Skewed data using PANSS in one RCT are presented in a separate table (Analysis 4.6).

4.3.4. Negative symptom score (SANS, high = worse, skew)

Data assessing negative symptoms using SANS were also skewed (Analysis 4.7).

4.3.5. Positive symptoms score (PANSS positive symptoms subscale, high = worse)

Positive symptoms were assessed using the PANSS positive symptom subscale; the one RCT that presented data at medium term demonstrated significant favour for intermittent treatment (n = 128, 1 RCT, MD -0.80, 95% CI -1.58 to -0.02), but an equivocal effect by long term (n = 155, 2 RCTs, MD 0.40, 95% CI -0.79 to 1.59, Analysis 4.8).

4.4 General functioning
4.4.1. Social functioning score (GAF, low = worse)

Social functioning scores using the GAF significantly favoured maintenance treatment at long term in the one RCT that reported data (n = 27, 1 RCT, MD -9.00, 95% CI -15.92 to -2.08, Analysis 4.9).

4.4.2. Social functioning score (GSDS, high = worse, skew)

Data presented were skewed with the GSDS (Analysis 4.10) and need interpreting with caution.

4.4.3. Social functioning score (SAS, high = worse)

Results were equivocal at both medium term (n = 51, 1 RCT, MD 0.06, 95% CI -0.10 to 0.22) and long term (n = 51, 1 RCT, MD 0.05, 95% CI -0.08 to 0.18, Analysis 4.11) when using the SAS.

4.5 Adverse effects
4.5.1. Akathisia (HAS, high = worse, skew)

All data from scales used to assess adverse effects/events are skewed and are presented in separate tables (Analysis 4.12).

4.5.2. Extrapyramidal symptoms (EPS, high = worse, skew)

There was little difference in EPS scores to measure extrapyramidal symptoms (Analysis 4.13), with skewed data presented separately.

4.5.3. Side effects (LUNSERS, high = worse, skew)

Side effects scores were skewed when using LUNSERS in the one RCT that used the scale and are reported separately (Analysis 4.14).

4.5.4. Side effects (UKU, high = worse, skew)

Results for side effects using the UKU by long term were also skewed (Analysis 4.15).

4.5.5. Tardive dyskinesia

The only binary outcome represents zero instances of tardive dyskinesia in either intermittent or maintenance groups throughout the period of the one study that reported such data (Analysis 4.16).

4.5.6. Tardive dyskinesia (AIMS, high = worse, skew)

AIMS scores to measure tardive dyskinesia were skewed and are presented separately (Analysis 4.17).

4.6 Quality of life
4.6.1. Average score (LQLP, low = worse)

Quality of life scores were equivocal when using the LQLP by long term (n = 27, 1 RCT, MD -0.50, 95% CI -1.38 to 0.38, Analysis 4.18).

4.6.2. Average score (WHOQoL-Bref, low = worse)

Again, results were equivocal using the WHOQoL-Bref at both medium term (n = 128, 1 RCT, MD -0.70, 95% CI -4.27 to 2.87) and long term (n = 128, 1 RCT, MD -0.90, 95% CI -5.39 to 3.59, Analysis 4.19).

4.7 Leaving the study early/loss to follow-up

There was no difference between groups for numbers leaving the study early for any reason by long term (n = 257, 3 RCTs, Analysis 4.20); these results displayed substantial heterogeneity (Chi² = 6.33, P = 0.04, I² = 68%) and are presented using a random-effects model of analysis.

Comparison 5: INTERMITTENT (DRUG HOLIDAY) versus MAINTENANCE THERAPY

Seven studies provided data for this comparison (Blackburn 1961; Caffey 1964; McCreadie 1980; McCreadie 1982; Olson 1962*; Remington 2011; Shenoy 1981, n = 627).

5.1 Relapse

Data demonstrate that there was a significantly greater risk of relapse for participants receiving intermittent drug therapy; results were homogenous, statistically significant in favour of maintenance therapy at medium term (n = 272, 3 RCTs, RR 2.15, 95% CI 1.25 to 3.68), however, results were equivocal by short term (n = 268, 3 RCTs, RR 1.59, 95% CI 0.94 to 2.70) and long term (n = 62, 2 RCTs, RR 1.70, 95% CI 0.54 to 5.38, Analysis 5.1).

5.2 Hospitalisation

There was no significant difference in numbers of hospitalisation between groups (n = 35, 1 RCT, Analysis 5.2).

5.3 Global state
5.3.1. Average score (CGI-S, high = worse)

At medium term, CGI scores were not significantly different between groups (n = 35, 1 RCT, Analysis 5.3).

5.3.2. Average score (GAS, low = worse)

In one RCT, results from the GAS significantly favoured maintenance therapy at short term (n = 31, 1 RCT, MD -3.61, 95% CI -5.67 to -1.55, Analysis 5.4).

5.3.3. Significant improvement

In the one trial that measured 'significant improvement', there was a significantly higher instance of improvement with maintenance therapy at short term (n = 60, 1 RCT, RR 0.40, 95% CI 0.18 to 0.89) and medium term (n = 60, 1 RCT, RR 0.32, 95% CI 0.15 to 0.68, Analysis 5.5).

5.4 Adverse effects
5.4.1. Need for additional medication

There was no significant difference between groups for instances of people requiring additional medication (n = 33, 1 RCT, Analysis 5.6).

5.4.2. Tardive dyskinesia

There was no significant difference between groups when assessing numbers of people who experienced tardive dyskinesia by long term (n = 50, 2 RCTs, Analysis 5.7).

5.4.3. Tardive dyskinesia (AIMS, high = worse, skew)

Tardive dyskinesia scores using the AIMS scale were skewed and are presented separately (Analysis 5.8).

5.5 Leaving the study early/loss to follow-up

Overall, there was no significant difference in lost to follow-up numbers between groups (n = 248, 6 RCTs, Analysis 5.9), however results demonstrated considerable heterogeneity (Chi² = 16.09, P = 0.007, I² = 69%) and are presented using a random-effects model.

Comparison 6: ANY INTERMITTENT DRUG TECHNIQUE versus PLACEBO

Three studies reported data for this comparison (Blackburn 1961; Caffey 1964; Olson 1962*, n = 498).

6.1 Relapse

Compared with placebo, there was significant favour for intermittent treatment at short term (n = 260, 1 RCT, RR 0.22, 95% CI 0.10 to 0.45) and medium term (n = 290, 2 RCTs, RR 0.37, 95% CI 0.24 to 0.58, Analysis 6.1), however, heterogeneity is present for the medium term results (Chi² = 2.59, P = 0.11, I² = 61%), and so caution should be employed in interpreting this data.

6.2 Global state
6.2.1. Significant improvement

There was no significant difference in the amount of people considered 'significantly improved' by medium term (n = 30, 1 RCT, Analysis 6.2).

6.3 Leaving the study early/loss to follow-up

People allocated to placebo were significantly more likely to remain in the study compared with the intermittent drug techniques used in two RCTS (n = 90, 2 RCTs, RR 1.97, 95% CI 1.28 to 3.01, Analysis 6.3).

Comparison 7: ANY INTERMITTENT DRUG TECHNIQUE (SPECIFIC DRUG) versus MAINTENANCE THERAPY (SPECIFIC DRUG)

We compared the differences between the various antipsychotics used in intermittent treatment compared with maintenance therapy at short term, medium and long term for our two primary outcomes of relapse and hospitalisation. Twelve studies provided data for the outcome of relapse (Blackburn 1961; Caffey 1964; Gaebel 2011; Herz 1991*; Jolley 1989/1990; McCreadie 1980; McCreadie 1982; Prien 1973; Remington 2011; Shenoy 1981; Wiedemann 2001; Wunderink 2007, n = 1327); however, some studies did not provide adequate information regarding the types of antipsychotics used, instead converting dosages into milligrams of chlorpromazine (Carpenter 1987; Carpenter 1990*; Herz 1991*; Wiedemann 2001) or haloperidol (Gaebel 2011; Jolley 1989/1990; Wunderink 2007) equivalents. Dosages were unclear in Remington 2011.

7.1 Relapse

Results demonstrated homogeneity, with favour of maintenance therapy with each specific drug comparison; some results were statistically significant, however, and demonstrate a higher risk of relapse for people receiving intermittent drugs in the following comparisons: various intermittent typical antipsychotics (moderate dose) versus maintained typical antipsychotics (moderate dose) at medium term (n = 551, 2 RCTs, RR 3.75, 95% CI 1.42 to 9.94); intermittent chlorpromazine equivalents (low dose) versus maintained chlorpromazine equivalents (low dose) by long term (n = 148, 2 RCTs, RR 2.62, 95% CI 1.30 to 5.28) and intermittent haloperidol equivalents (low dose) versus maintained haloperidol equivalents (low dose) by long term (n = 226, 3 RCTs, RR 2.53, 95% CI 1.60 to 4.01, Analysis 7.1).

7.2 Hospitalisation

Data demonstrated statistical significance in favour of maintenance therapy when comparing intermittent fluphenazine decanoate (low dose) versus maintained fluphenazine decanoate (low + moderate dose) by long term (n = 313, 1 RCTs, RR 1.81, 95% CI 1.33 to 2.48), with overall statistically significant favour for maintenance therapy between all groups (n = 661, 6 RCTs, RR 1.58, 95% CI 1.28 to 1.97, Analysis 7.2).

SENSITIVITY ANALYSIS: ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY

When included studies that implied randomisation or provided no further details regarding randomisation techniques were removed from the meta-analysis, this did not substantially alter the direction of effect or the precision of the effect estimates, with a significantly greater risk of relapse still demonstrated for people receiving any intermittent therapy by long term (n = 273, 3 RCTs, RR 2.19, 95% CI 1.41 to 3.42), and greater risk of hospitalisation for the same group by long term (n = 414, 2 RCTs, RR 1.76, 95% CI 1.31 to 2.36); therefore, where risk of bias for randomisation was 'unclear', all data have been employed from all studies. All of the included studies were judged to be at a high risk of bias across one or more of the domains of randomisation, allocation concealment, blinding and outcome reporting, which makes it difficult to draw meaningful conclusions from a sensitivity analyses, as we would need substantial power to show real differences with confidence.

When testing how prone the primary outcomes were to change when data only from people who complete the study to that point were compared to the intention to treat analysis, there was little difference in results - often showing results with greater significance. When comparing any intermittent drug technique to maintenance therapy, data still showed significantly greater relapses in the intermittent group at short term (n = 358, 3 RCTs, RR 2.12, 95% CI 1.12 to 3.98), medium term (n = 767, 5 RCTs, RR 2.94, 95% CI 1.71 to 5.05) and long term (n = 430, 7 RCTs, RR 2.51, 95% CI 1.72 to 3.67). Neither was there any significant difference in numbers of people hospitalised at long term (n = 620, 5 RCTs), with data still showing significant favour for maintenance therapy.

Discussion

Summary of main results

1. Specific

The review includes studies that span over five decades. Over this period, the landscape of psychiatry has changed, as have the types of participants involved in the studies; the methods by which trials are carried out; treatment administered and even the level of quality by which results are reported (Thornley 1998). All studies were included in the meta-analysis, however, and have been detailed in Characteristics of included studies for closer inspection. There were no differences in primary outcomes when a random-effects or fixed-effect model was used; results that displayed moderate heterogeneity were analysed using a random-effects model.

Comparison 1. ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY

This comparison encapsulates the various forms of intermittent antipsychotic treatment (which, for the purposes of this review, include intermittent treatment based on: prodrome/early-based intervention; crisis-based intervention; drug holidays; and gradually increased drug-free periods) and antipsychotic treatment (only a select few of included studies have provided adequate detail regarding types of antipsychotics and dosages administered). Therefore, one must remain mindful that these intermittent treatments, although uniformly meta-analysed within data and results, have employed differing: methods of administration; types of antipsychotics used; and methods by which trial authors determine when people receiving intermittent therapy (if receiving a form of early-based intermittent treatment that relies on recognition of prodromal symptoms, for example) are to re-start their treatment, if at all. With this in mind, the significance of the results for relapse at short, medium and long term indicates that any intermittent treatment (regardless of its administration), falls short to the stability of continuous maintenance therapy.

Only two studies reported death, with a higher (non-significant) rate in people receiving maintenance therapy. In Herz 1991*, the deaths recorded were due to 'natural causes', with no further information given (n = 2 receiving maintenance therapy), and a 'possible suicide' (n = 1 receiving intermittent therapy) who drowned after ingesting large amounts of alcohol. It is unclear to what extent, if any, these deaths could be said to be attributable to either intermittent or maintenance treatments. This is also the case with the deaths recorded by Jolley 1989/1990, in which n = 2 participants receiving maintenance treatment died; one death was a suggested suicide, and the other reported as 'attributable to an acute physical illness'.

Many scales were employed to measure adverse effects; however, all presented skewed data and were not included within a meta-analysis due to high standard deviations. Specific adverse effects were measured in only one randomised controlled trial (RCT) (Jolley 1989/1990), and demonstrated significant favour for intermittent treatment when assessing extra-pyramidal symptoms (EPS) side effects including: akathisia; parkinsonism; and non-liveliness.

Most continuous outcomes demonstrated no significant difference between intervention groups; were considerably skewed; or used different scales to measure the same outcome, making a meaningful interpretation of these results difficult.

Comparison 2: INTERMITTENT (EARLY-BASED) versus MAINTENANCE THERAPY

More people receiving intermittent treatment, given on the early/prodromal signs of relapse, were significantly more likely to experience full relapse compared to people receiving maintenance therapy. There was also a higher instance of hospitalisation with people receiving the early intermittent treatment; results were homogenous and demonstrated statistical significance by long term. These results from five RCTs, with a relatively large sample size (n = 620), demonstrate that maintenance therapy is more effective than early intermittent techniques at lowering the number of people hospitalised.

Results were equivocal when assessing general functioning; no two studies used the same scale to assess general functioning, and any real-life significance of these scale measurements have not been considered by trial authors.

We found that less specific adverse effects were experienced in people receiving intermittent treatment when given on the early/prodromal signs of relapse, demonstrating significant favour for intermittent treatment when assessing EPS symptoms including: akathisia; parkinsonism; and non-liveliness. Only one RCT with a small sample size (n = 43) reported data for this outcome. Data for extrapyramidal symptom scores using the EPS were highly skewed, making it difficult to draw any meaningful conclusions.

More people tended to leave the studies early when they were receiving intermittent treatment, with a significant favour of maintenance therapy; however, considerable heterogeneity (I² = 58%) was evident in the analysis. Upon visual inspection, removal of Carpenter 1987 restored homogeneity; this RCT documented people who dropped out as: n = 9 in the maintenance group (one refused assignment to regimen; seven discontinued treatment; one was non-compliant); and n = 7 in the intermittent group (two were hospitalised; two were dropped while hospitalised; three dropped out during clinical stability while medication-free). Taking this into account, it is difficult to ascertain whether those who 'discontinued' the intermittent treatment did so for treatment-related reasons or due to other circumstances.

Comparison 3: INTERMITTENT (CRISIS INTERVENTION) versus MAINTENANCE THERAPY

Only one study provided data for this comparison, and unfortunately, since attrition was over 50%, only data for the outcome of leaving the study early were included in the data and analysis. Maintenance therapy was significantly favoured, with a higher instance of people leaving the study early when given intermittent treatment. Numbers leaving the study early were adequately documented; a high number of participants in the intermittent treatment groups dropped-out due to difficulties of withdrawing antipsychotics, indicating that people who left the study early had negative outcomes. This finding would need to be confirmed through further research, using high-quality, randomised methods.

Comparison 4: INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY

Corresponding with the findings from Comparison 1 and 2, intermittent treatment, when employed via gradually increasing drug-free periods, had a higher instance of relapse, with a significant outcome by long term.

Due to the small amount of studies (three RCTs) and relatively small sample sizes (maximum of n = 128) providing data for this mental state outcomes, results are difficult to interpret meaningfully. Data for most continuous outcomes were either considerably skewed, or demonstrated no significant difference between the groups.

General functioning scores tend to favour maintenance treatment, using the SAS, and significantly when using the GAF - this judgement is based on a single study with only 27 participants. Furthermore, scores measuring social disabilities using the GSDS were considerably skewed.This makes the present results difficult to interpret - dichotomous data are a preferred method of reporting outcomes, particularly for adverse effects, in order to reflect the real-life significance of the outcome.

More people left the study early when receiving intermittent treatment via gradually increasing drug-free periods. Considerable heterogeneity associated with leaving the study early could be attributable to the different methods by which drug-free periods were gradually increased. Each of the three RCTs providing outcome data gradually decreased antipsychotic treatment in the intermittent group, however one trial sought complete removal over a period of three months at the most (Gaebel 2011), the second sought gradual decrease of the active drug after three months (Wiedemann 2001), and the third trial was open-label, dosages were gradually tapered-off guided by symptom severity levels and the preference of the participant, discontinued only 'if feasible' (Wunderink 2007). These factors could well influence whether or not a person left the study early due to inaccurate identification of times when medication should be administered or a lack of effective treatment structure, for example.

Comparison 5: INTERMITTENT (DRUG HOLIDAY) versus MAINTENANCE THERAPY

People who received intermittent drug treatment via drug holiday were significantly more likely to experience relapse in the short and medium term than people who received maintenance therapy, again confirming the results for the above comparisons that intermittent treatment is not as effective as continuous.

Only one RCT reported data for hospitalisation, in which intermittent therapy is favoured over maintenance; this contradicts results in the other comparisons, which was acknowledged by the trial authors, who stated their findings to be 'at odds' with the existing number of previous studies evaluating intermittent therapy. Further research is needed to clarify this issue.

There was no significant difference found when comparing instances of adverse effects between the two comparison groups, and numbers of people leaving the studies early were equivocal; reasons included: administrative (Blackburn 1961); refusal of medication (McCreadie 1980); exacerbation of positive symptoms, suicidal ideas and attempts, family requests to withdraw, tardive dyskinesia, complaints of tiredness (McCreadie 1982); 'behavioural' or 'inadvertent' attrition (Olson 1962*); withdrawal of consent (Remington 2011); and failure to attend appointments (Shenoy 1981).

Comparison 6: ANY INTERMITTENT DRUG TECHNIQUE versus PLACEBO

It is understandable that placebo-controlled trials in this area are uncommon and it could be argued that in certain circumstances are unethical. We identified three trials undertaken in the 1960s that made this comparison; each trial was poorly reported, and failed to describe randomisation methods.

The two trials reporting data for relapse (Blackburn 1961; Caffey 1964) both employed a drug holiday intermittent technique, in which antipsychotics were withdrawn and reintroduced on a regular or longer-term basis. Intermittent treatment was significantly favoured, with more people receiving placebo classified as relapsed. This finding goes some way to demonstrate that intermittent techniques are at least more effective than no treatment/placebo, the extent to which this result is applicable, however, could well be dependable on the type of intermittent therapy implemented.

Significantly more people receiving intermittent therapy left the study early when compared with placebo. Further, reasons for leaving were not made clear, described only as 'behavioural' or 'inadvertent' attrition, or 'lost for administrative reasons'. Whether this makes people leaving intermittent treatment likely to experience a negative outcome is difficult to establish. Differences between the study methods, however, may well have influenced favour for placebo - the two comparison groups in Blackburn 1961 examined the effects of placebo substituted for the first half of the study (eight weeks) then active medication reintroduced for the final half of the trial, versus placebo substituted for the full trial length; whereas Olson 1962* compared active drugs alternated monthly (intermittent) versus active drugs alternated monthly with placebo. Placebo effect may well have been evident in Blackburn 1961, which made use of placebo in each treatment arm; this may have resulted in the much higher rate of attrition in the study that employed placebo in only one group.

Comparison 7: ANY INTERMITTENT DRUG TECHNIQUE (SPECIFIC DRUG) versus MAINTENANCE THERAPY (SPECIFIC DRUG)

Not all studies described the type of antipsychotic medication administered or the dosages used, therefore, this comparison measured outcome data only from those trials where adequate details regarding the type of medication were given. There was significant favour for maintenance therapy when comparing:

  1. intermittent low/moderate dose typical antipsychotics versus maintained low/moderate dose typical antipsychotics at medium term;

  2. intermittent low dose chlorpromazine/haloperidol equivalents versus maintained low dose chlorpromazine/haloperidol equivalents at long term.

Only chlorpromazine or haloperidol equivalents were reported in some trials, which make it difficult to ascertain efficacy of specific drugs, if at all, when given intermittently.

Meta-analysis overall demonstrated significant favour of maintenance therapy for preventing hospitalisation, with significance found particularly when comparing intermittent low dose fluphenazine decanoate with maintained low/moderate dose fluphenazine decanoate.

2. General

Futhermore, because 'relapse' was defined differently in each individual study, there was not a single, fixed assessment of measurement. Out of the 17 studies included in this update, 15 reported our two primary outcomes of interest. The majority of the included studies had a low level of participants; only seven studies had a sample size of 100+ participants.

2.1 Heterogeneity

Some results are difficult to interpret due to high levels of heterogeneity; sources of heterogeneity have been explored and addressed in the above discussion, however, where results for the I2 are statistically significant (when testing for homogeneity), caution should be employed when adding trial data together.

2.2 Funnel plots

It is clear from Figure 4 that five small studies are scattered on the side of the plot, which suggests the possibility of publication bias. This is due to inequality of the scattered studies on both sides of the plot, as the left-hand side represents studies that are having or approaching either no effect or a negative effect. Therefore, it can be said that there is evidence of asymmetry.      

Figure 4.

Funnel plot of comparison: 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, outcome: 1.1 Relapse.

In Figure 5, three studies (two on the right and one on the left) have no effect since their standard errors (SE) range between 1 and 1.5. The funnel plot shows that the majority of large studies are scattered around the top of the plot along a lower SE of less than one. It can be concluded that the funnel plot is, to a large extent, symmetrical, raising the possibility of some potentially moderate bias. In other words, there may be some negative results that have not been published.

Figure 5.

Funnel plot of comparison: 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, outcome: 1.29 Leaving the study early/loss to follow-up.

Overall completeness and applicability of evidence

1. Completeness

Twelve of the 17 included studies reported our primary outcome interest of relapse; however, only six trials reported our second primary outcome of interest of hospitalisation. No study reported economic outcomes, which would be of interest to managers and policy makers in measuring potential cost-effectiveness and efficiency of the intervention, nor was satisfaction with treatment addressed. Furthermore, it was disappointing that only two studies reported death and only four reported tardive dyskinesia as outcome measures; two outcomes that could offer greater insight into an intervention that aims to reduce the risk of typical adverse effects of continuous antipsychotic treatment. Patient-oriented outcomes were addressed in few small-scale included studies; with only three studies reporting data for quality of life; four studies reporting mental state scores; and five studies reporting scores for general and social functioning. Numerous scales were employed (in some instances by single studies), which made it difficult to make any meaningful interpretation of results, as trial authors did not address the real-life significance of scale measurements, and combining all similar scale data in meta-analysis could overestimate authority of potentially weak data.

The available evidence is relevant to the review question, however, intermittent antipsychotic techniques were not always referred to explicitly, as some studies did not specify their treatment as intermittent, but 'targeted intervention', 'guided discontinuation' or 'dose reduction and family treatment'. This made the trial search problematic and the intervention question more indirect, as many other studies were identified via handsearching. Consequently, we worked mainly with published reports, which may have the result of missing other potentially relevant RCTs from the current review. When presented with problematic or incomplete data, we made efforts to contact relevant trial authors to clarify these issues.
Poor reporting exacerbated the problem of scarce data; three trials were excluded due to unusable data, and the majority of included studies did not report all outcomes as per protocol. This was particularly the case with scale data, where scale measurements were taken at designated time intervals, but no data were ultimately reported.

2. Applicability

Settings were not uniform between studies, with six large multi-centre trials carried out in Germany, the Netherlands and the US; two studies carried out hostel wards in Scotland (UK); four trials undertaken in psychiatric hospitals/institutes in the US, one in Germany and one in London (UK); with a further trial conducted in a Centre for Addiction and Mental Health, Canada, setting and location were not made clear in two other included studies. As studies included both inpatients and outpatients from various countries, the international applicability of the present available evidence needs considering in relation to the methods of psychiatric treatment employed and the types of antipsychotic medication available or recommended in individual countries. Participants in seven of the total included studies were inpatients, and in the remaining 10 studies participants were outpatients. Of the inpatient studies, all were undertaken between 1961 and 1982 with the exception of a study undertaken in 2001. With this is mind, one must take into account the applicability of the evidence when considering results from studies that included inpatients, particularly because more psychiatric treatments are increasingly obliged to take place in outpatient or community settings. There was a disproportionate amount of male participants compared to female participants in the included studies (n = 1555 males and n = 540 females), and two included studies did not specify gender in their reports. Ages of participants were described in each study, with a range from all studies of 18-60; means were provided in most of the reports. Furthermore, ethnicity/racial origin was not described in the majority of studies, with only six trials providing appropriate details.

Quality of the evidence

Overall, most quality criteria were poorly reported in the trials included in this review; large, well-designed and clearly reported trials are possible in this area (as has been demonstrated in the recent TREC trials - TREC-Rio-I; TREC-Vellore-I). Better reporting of the methods used to ensure trials of high quality as outlined in the CONSORT statement (Moher 2001) could have resulted in this review being more transparent. Future trialists in this area need to ensure that attention is given to the adequate reporting of quality criteria. Some data reported by the included studies were skewed, and without access to individual patient data, it is difficult to present these results in a meaningful way, or likewise to comment upon them with a degree of certainty.

Potential biases in the review process

The review protocol and process of study selection were strictly adhered to throughout the entire review process, and the process for searching for studies was thorough and data were extracted independently; however, this was made difficult due to numerical inconsistencies in parts of the reported data. Review authors adhered to the guidance from the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) and the Methodological Expectations for Cochrane Intervention Reviews (MECIR).

The first study search was narrow, and failed to identify several studies that suited inclusion criteria. With the intention of correctly identifying all relevant published studies, another trial search was conducted (see Appendix 1 for previous search term methods). Authors of included studies were contacted to obtain details of ongoing or unpublished studies, but there remains a possibility that other unpublished trials of the intervention exist which the review authors do not currently have access to. As a consequence, publication bias may have been perpetuated as we worked predominantly with published reports. Furthermore, it became clear that not all of the studies employing an intermittent technique of administering antipsychotics labelled it as such, referring instead to 'targeted intervention' or 'guided discontinuation'; it may therefore be that trials exist relying on essential elements of the intervention, but are not packaged and delivered in identical models.

Agreements and disagreements with other studies or reviews

The findings from this review supports the existing evidence that intermittent antipsychotic treatment is not as effective as continuous, maintained antipsychotic therapy in preventing relapse in people with schizophrenia. The findings of this review largely agree with current clinical guidelines (NICE 2010). These guidelines state that low-dose prescribing and use of intermittent dosing strategies may well minimise side effects in the long-term use of antipsychotic medication; however, the risk of symptom-worsening and relapse is stated to outweigh any benefits (NICE 2010). Instead, it is suggested that patients who refuse maintenance therapy or where another contraindication to such therapy exists (e.g. side-effect sensitivity), then targeted, intermittent dosage maintenance strategies may be considered, baring in mind the 'reasonable targets' for intervention, including presence of persistent symptoms, poor adherence to treatment regimen, lack of insight and substance use (Marder 2003). WHO guidance makes no mention of intermittent or targeted approaches (WHO 2009).

Authors' conclusions

Implications for practice

1. For people with schizophrenia

The results suggested that intermittent drug techniques do not play a preventative role for relapse or (re)hospitalisation when compared with maintenance therapy, and the assumption that intermittent antipsychotic treatment may minimise potential adverse effects associated with continuous, long-term antipsychotic treatment is far from being concluded. The results of this review included data from small, poorly-reported trials, where randomisation methods were not specified, therefore making any significant findings susceptible to a high risk of bias, with the quality of the majority of the evidence ultimately rated as 'low' or 'very low'. The report of higher relapses in the intermittent techniques may suggest an intolerance to drug withdrawal by not controlling positive symptoms. Those who are on drugs intermittently are unlikely to permanently discontinue antipsychotic medication; loss of substance, motivation and sense of purposiveness to medication is lost by treating individuals intermittently, yet the stigma of assignment to long-term treatment can make compliance more problematic.

2. For clinicians

The evidence that any intermittent drug technique is superior to maintenance therapy is weak, with a significantly higher instance of relapse at each time interval evident among people receiving drugs on an intermittent schedule, and a significantly higher rate of hospitalisation in the long term. Until further evidence is available concerning the potential range and extent of both benefits and harms of intermittent antipsychotics, clinicians should employ this method of treatment only under heavily supervised settings. Effectiveness of any psychiatric treatment relies on good communication between healthcare professionals and service users at every stage of treatment and care, with the treatment provided and information supplied tailored to individual persons needs and culturally appropriate (NICE); all relevant and suitable treatment options should be determined on an individual basis, through full-disclosure and discussion with the recipient of care. If intermittent drug techniques are to be undertaken there must be a thorough assessment taken prior to this concerning patient characteristics and history of treatment compliance. Throughout the therapy, clinicians must carefully control and monitor the withdrawal of antipsychotics and the process should be one made gradually for patient safety. In such circumstances, the evidence favours early intervention in comparison to crisis intervention, as attrition rates are marginally lower; by intervening early, symptoms may be more effectively controlled and treatment adherence improved.

3. For managers or policy-makers

The majority of the included studies provided insufficient information regarding the specific antipsychotic treatment employed and dosages administered in both intermittent and maintenance therapy groups. This makes it difficult to interpret any cost-effectiveness of intermittent treatment, on the assumption that similar antipsychotics (i.e. typical or atypical antipsychotics) and dosages would be administered on an intermittent as opposed to a continuous/maintained basis. There was no discussion of economic outcomes, specifically relating to potential cost-effectiveness of employing an intermittent antipsychotic regimen. Until further evidence is available concerning the potential range and extent of both benefits and harms of intermittent antipsychotic treatment, managers and policymakers should consider it an experimental therapy (see Table 1: suggested design of future studies). Well-reported economic data need to be well balanced with patient-oriented outcomes to help equip policy makers with much needed information regarding any cost-benefits of intermittent treatment.

Table 1. Suggested design for future study
  1. BPRS - Brief Psychiatric Rating Scale
    CGI - Clinical Global Impression.
    DSM - Diagnostic and Statistical Manual of Mental Disorders
    ICD - The International Statistical Classification of Diseases and Related Health Problems
    PANSS - Positive and Negative Syndrome Scale

Methods

Allocation: randomised, fully described in terms of methods of randomisation and allocation concealment.

Blinding: double/single blind, with methods of maintenance of blinding fully described.
Duration: two years.
Setting: outpatients.
Design: parallel.

ParticipantsDiagnosis: schizophrenia (DSM-IV/ICD-10/ RDC).
n => 300.
Sex: male and female.
Age: adults, with age specified in trial.
Interventions

1. Any intermittent drug technique (n = 150):

a. Prodrome-based/early intervention (defined as treatment given on the early signs of relapse).

b. Crisis intervention (defined as treatment given only in case of full relapse and discontinued again after re-stabilisation).

c. Gradually increased drug-free period (defined as increasing the cessation period of the treatment constantly).

d. Drug holiday (defined as stopping medication for fixed periods, and then reintroducing it - repeating this more than once).

Secifiying types of typical or atypical antipsychotics:

i. High dose (as defined by each study).

ii. Low or moderate dose (as defined by each study).

versus

2. Maintenance therapy, as defined in each study (n = 150):

Secifiying types of typical or atypical antipsychotics:

i. High dose (as defined by each study).

ii. Low or moderate dose (as defined by each study).

Outcomes

Relapse (operationally defined using either PANSS, CGI or BPRS).

Hospitaliation (as defined in each study).

Global state: CGI (Clinical Global Impression) or GAS (Global Assessment Scale) - preferably dichotomous outcomes/dichotomised scale data).

Mental state: BPRS or PANSS (preferably dichotomous outcomes/dichotomised scale data).

General functioning: GAF (Global Assessment of Functioning Scale - preferably dichotomous outcomes/dichotomised scale data).

Quality of life: QLS (Quality of Life Scale - preferably dichotomous outcomes/dichotomised scale data).

Adverse effects: including specific adverse effects, extrapyramidal symptoms, tardive dyskinesia.

Economic outcomes: including any cost effectiveness of treatment.

Death.

Leaving the study early: specific reasons including - any reason; loss to follow-up; treatment-related; non-treatment related; death; adverse effects.

NotesAny outcomes measured using scale-derived data should be interpreted in such a way as to make clear the real-life relevance of changes in scale score.

Implications for research

1. General

Adherence to the CONSORT statement (Moher 2001) would probably have resulted in this review being more conclusive. Clear descriptions of randomisation would have reassured users of these trials that selection bias had been minimised and well-described and tested blinding could have encouraged greater confidence in the control of performance and detection bias. The use of binary outcomes should take preference over continuous results because they are easier to interpret and the use of validated rating scales would have provided more usable data. The reporting of outcomes with their means and standard deviations again would have provided more usable data and facilitated synthesis of findings. When presenting data in a graph, the exact numbers and standard deviations should also be reported.

2. Specific

2.1. Reviews

The current review has sought to include all relevant randomised controlled trials comparing any form of intermittent drug treatment with the standard continuous/maintained antipsychotic therapy. However, the trial search failed to identify a number of relevant trials, owing in part to poor standards of reporting on behalf of trial authors. Future reviews would need to address the varying ways in which intermittent treatments are described and packaged to ensure completeness. Furthermore, the protocol to this review did not specify a comparison with specific named drugs (atypical or typical) nor dosages, which is certainly a comparison of interest for people receiving treatment as well as clinicians administering medication and was subsequently addressed in the current review (see Table 2 for differences between protocol and review).

Table 2. Differences between protocol and review
MajorMinor
1. There have been additions to Types of interventions: the protocol to the current review did not specify a comparison with specific named drugs (atypical or typical) nor dosages, which is certainly a comparison of interest for people receiving treatment as well as clinicians administering medication and was subsequently addressed in the final text.

1. The data collection and analysis section has been updated to reflect changes in methodology that have occurred whilst writing this review, namely inclusion of 'Risk of bias' tables and 'Summary of findings' tables.

2. Types of outcome measures have been modified/clarified from the original protocol - this decision was not influenced by the results. Economic outcomes of direct and indirect costs have been qualified as 'defined by each study', to take into account that what may constitute a direct and indirect cost can differ between trial authors. A further sub-category of 'cost-effectiveness' has been added, again a factor that would be 'defined by each study'.

3. Further, the protocol specified a global state outcome of 'no clinically important change in global state (as defined by individual studies)'; this has been changed to 'any clinically important change in global state (as defined by individual studies)'. The direction of the graphs have been modified to represent an unfavourable outcome for intermittent treatment when results are presented to the left of the line of no effect.

2.2 Research

The results demonstrate the efficacy of continuous antipsychotic use for managing schizophrenia; having said this, there are some individuals who demonstrate reduced treatment adherence or a higher sensitivity to side effects; it is for this population that the need for further research surrounding intermittent techniques is certainly needed. Furthermore, evidence that has suggested that first-episode patients seem to be more suited to intermittent treatment than multiple-episode patients needs further research (Gaebel 2002). Future research using large, methodologically sound and clearly reported trials should consider comparing typical and atypical antipsychotics for differences in tolerability among men and women equally, with varying degrees of illness specified, to assess why some individuals are particularly treatment-resistant to some forms of drugs than others. Methods of the intervention should be clearly defined, with specific drugs identified and dosages reported clearly, in order to facilitate higher standards of care provided by professionals, and a greater level of understanding for people with schizophrenia and their family members. Health economic outcomes should be addressed, in order to ascertain any potential cost benefits associated with intermittently administered antipsychotics, as efficiency of treatment must be demonstrated before becoming a realistic option for managers and policy makers. Any outcomes of further studies in this area should include validated scales that are acceptable to clinicians working in the field, recipients of care, researchers and those working with regulatory authorities, taking into account, however, the real-life significance and practicality of use of standardised outcome measures in practice (Gilbody 2002). The landscape of psychiatric care has transformed dramatically over the past 50 years, with emphasis placed on a person-centred approach, where people with schizophrenia are increasingly treated in outpatient or community settings. People with schizophrenia who are prescribed pharmacological treatment are most likely to also receive psychological therapy, and any further research and subsequent reviews into intermittent antipsychotic therapy could shed light onto the effects of the intervention combined with some form of psychological/supportive therapy.

Acknowledgements

We thank Dr. Adib Essali and Professor Clive Adams for encouragement, continuous support and great help. We thank Gill Rizzello and Tessa Grant for their support and Judith Wright for assistance in the literature searches. We would also like to thank William Carpenter, Wolfgang Gaebel, Marvin Herz, Mathias Riesbeck and Hiroyuki Uchida for their assistance in clarifying characteristics of their included studies.

The Cochrane Schizophrenia Group maintains a standard template for its methods section of reviews and protocols.We have used this and adapted it.

We would like to thank and acknowledge the contributions made by Kajal Joshi (who helped with trial selection, data extraction and writing the review), Anas Alomar, Khaled Boobes and Mohamad Awf Mouchli (who helped write the protocol). We would also like to thank Beccy Dickenson and Muhammad Shahzad Ashraf for peer reviewing the review.

Data and analyses

Download statistical data

Comparison 1. ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Relapse12 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
1.1 short term (≤12 weeks)4396Risk Ratio (M-H, Fixed, 95% CI)1.68 [1.00, 2.81]
1.2 medium term (13-25 weeks)5774Risk Ratio (M-H, Fixed, 95% CI)2.41 [1.50, 3.86]
1.3 long term (≥26 weeks)7436Risk Ratio (M-H, Fixed, 95% CI)2.46 [1.70, 3.54]
2 Hospitalisation6 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
2.1 medium term (13-25 weeks)2136Risk Ratio (M-H, Fixed, 95% CI)1.04 [0.31, 3.48]
2.2 long term (≥26 weeks)5626Risk Ratio (M-H, Fixed, 95% CI)1.65 [1.33, 2.06]
3 Death2 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
3.1 by long term (≥26 weeks)2155Risk Ratio (M-H, Fixed, 95% CI)0.34 [0.05, 2.08]
4 Global state: 1. Average score (CGI-S, high = worse)262Mean Difference (IV, Random, 95% CI)-0.05 [-0.76, 0.65]
4.1 medium term (13-25 weeks)135Mean Difference (IV, Random, 95% CI)-0.42 [-1.08, 0.24]
4.2 by long term (≥26 weeks)127Mean Difference (IV, Random, 95% CI)0.30 [-0.33, 0.93]
5 Global state: 2. Average score (GAS, low = worse)4 Mean Difference (IV, Fixed, 95% CI)Subtotals only
5.1 short term (≤12 weeks)131Mean Difference (IV, Fixed, 95% CI)-3.61 [-5.67, -1.55]
5.2 medium term (13-25 weeks)2126Mean Difference (IV, Fixed, 95% CI)-0.45 [-4.55, 3.66]
5.3 long term (≥26 weeks)3133Mean Difference (IV, Fixed, 95% CI)1.32 [-2.75, 5.39]
6 Global state: 3. Prodromal episodes2 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
6.1 by long term (≥26 weeks)2155Risk Ratio (M-H, Fixed, 95% CI)3.19 [1.87, 5.44]
7 Global state: 4. Significant improvement1 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
7.1 short term (≤12 weeks)160Risk Ratio (M-H, Fixed, 95% CI)0.4 [0.18, 0.89]
7.2 medium term (13-25 weeks)160Risk Ratio (M-H, Fixed, 95% CI)0.32 [0.15, 0.68]
8 Mental state: 1. Average score (BPRS, high = worse)2 Mean Difference (IV, Fixed, 95% CI)Subtotals only
8.1 by medium term (13-25 weeks)151Mean Difference (IV, Fixed, 95% CI)1.85 [-3.03, 6.73]
8.2 long term (≥26 weeks)277Mean Difference (IV, Fixed, 95% CI)0.10 [-0.32, 0.53]
9 Mental state: 2. Negative symptom score (PANSS negative symptom subscale, high = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
9.1 medium term (13-25 weeks)1128Mean Difference (IV, Fixed, 95% CI)-0.80 [-2.43, 0.83]
9.2 by long term (≥26 weeks)1128Mean Difference (IV, Fixed, 95% CI)-1.20 [-3.19, 0.79]
10 Mental state: 3. Negative symptom score (PANSS negative symptom subscale, high = worse, skew)  Other dataNo numeric data
10.1 by long term (≥26 weeks)  Other dataNo numeric data
11 Mental state: 4. Negative symptom score (SANS, high = worse, skew)  Other dataNo numeric data
11.1 by long term (≥26 weeks)  Other dataNo numeric data
12 Mental state: 5. Positive symptom score (PANSS positive symptom subscale, high = worse)2 Mean Difference (IV, Fixed, 95% CI)Subtotals only
12.1 medium term (13-25 weeks)1128Mean Difference (IV, Fixed, 95% CI)-0.80 [-1.58, -0.02]
12.2 by long term (≥26 weeks)2155Mean Difference (IV, Fixed, 95% CI)0.40 [-0.79, 1.59]
13 General functioning: 1. Average endpoint (LOFS, low = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
13.1 by long term (≥26 weeks)126Mean Difference (IV, Fixed, 95% CI)2.60 [-2.63, 7.83]
14 General functioning: 2. Social functioning score (GAF, low = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
14.1 by long term (≥26 weeks)127Mean Difference (IV, Fixed, 95% CI)-9.0 [-15.92, -2.08]
15 General functioning: 3. Social functioning score (GSDS, high = worse, skew)  Other dataNo numeric data
15.1 by long term (≥26 weeks)  Other dataNo numeric data
16 General functioning: 4. Social functioning score (PAS, high = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
16.1 by medium term (13-25 weeks)175Mean Difference (IV, Fixed, 95% CI)0.45 [-0.81, 1.71]
16.2 by long term (≥26 weeks)156Mean Difference (IV, Fixed, 95% CI)0.44 [-1.08, 1.96]
17 General functioning: 5. Social functioning score (SAS, high = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
17.1 by medium term (13-25 weeks)151Mean Difference (IV, Fixed, 95% CI)0.06 [-0.10, 0.22]
17.2 by long term (≥26 weeks)151Mean Difference (IV, Fixed, 95% CI)0.05 [-0.08, 0.18]
18 Adverse effects: 1. Akathisia (HAS, high = worse, skew)  Other dataNo numeric data
18.1 by long term (≥26 weeks)  Other dataNo numeric data
19 Adverse effects: 2. Extrapyramidal side effects1 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
19.1 hypomimia: by medium term (13-25 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.26 [0.06, 1.09]
19.2 hypomimia: by long term (≥26 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.17 [0.02, 1.33]
19.3 rigidity: by medium term (13-25 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.35 [0.01, 8.11]
19.4 rigidity: by long term (≥26 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.42 [0.09, 1.93]
19.5 tremor: by medium term (13-25 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.26 [0.03, 2.16]
19.6 tremor: by long term (≥26 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.21 [0.03, 1.65]
19.7 akathisia: by medium term (13-25 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.08 [0.00, 1.34]
19.8 akathisia: by long term (≥26 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.19 [0.05, 0.76]
19.9 gait abnormality: by medium term (13-25 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.15 [0.02, 1.11]
19.10 gait abnormality: by long term (≥26 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.08 [0.00, 1.34]
19.11 global parkinsonism: by medium term (13-25 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.35 [0.08, 1.54]
19.12 global parkinsonism: by long term (≥26 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.13 [0.02, 0.96]
19.13 global non-liveliness: by medium term (13-25 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.24 [0.08, 0.73]
19.14 global non-liveliness: by long term (≥26 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.09 [0.01, 0.61]
20 Adverse effects: 3. Extrapyramidal symptoms (EPS, high = worse, skew)  Other dataNo numeric data
20.1 by medium term (13-25 weeks)  Other dataNo numeric data
20.2 by long term (≥26 weeks)  Other dataNo numeric data
21 Adverse effects: 4. Need for additional medication2 Risk Ratio (M-H, Random, 95% CI)Subtotals only
21.1 by long term (≥26 weeks)2346Risk Ratio (M-H, Random, 95% CI)0.70 [0.20, 2.48]
22 Adverse effects: 5. Side effects (LUNSERS, high = worse, skew)  Other dataNo numeric data
22.1 by medium term (13-25 weeks)  Other dataNo numeric data
22.2 by long term (≥26 weeks)  Other dataNo numeric data
23 Adverse effects: 6. Side effects (UKU, high = worse, skew)  Other dataNo numeric data
23.1 by long term (≥26 weeks)  Other dataNo numeric data
24 Adverse effects: 7. Tardive dyskinesia4 Risk Ratio (M-H, Random, 95% CI)Subtotals only
24.1 by medium term (13-25 weeks)143Risk Ratio (M-H, Random, 95% CI)0.81 [0.37, 1.79]
24.2 by long term (≥26 weeks)4165Risk Ratio (M-H, Random, 95% CI)1.15 [0.58, 2.30]
25 Adverse effects: 8. Tardive dyskinesia (AIMS, high = worse, skew)  Other dataNo numeric data
25.1 by medium term (13-25 weeks)  Other dataNo numeric data
25.2 by long term (≥26 weeks)  Other dataNo numeric data
26 Quality of life: 1. Average score (LQLP, low = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
26.1 by long term (≥26 weeks)127Mean Difference (IV, Fixed, 95% CI)-0.5 [-1.38, 0.38]
27 Quality of life: 2. Average score (QLS, low = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
27.1 by long term (≥26 weeks)126Mean Difference (IV, Fixed, 95% CI)0.5 [-0.35, 1.35]
28 Quality of life: 3. Average score (WHOQoL-Bref, low = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
28.1 by medium term (13-25 weeks)1128Mean Difference (IV, Fixed, 95% CI)-0.70 [-4.27, 2.87]
28.2 by long term (≥26 weeks)1128Mean Difference (IV, Fixed, 95% CI)-0.90 [-5.39, 3.59]
29 Leaving the study early/ loss to follow-up141182Risk Ratio (M-H, Random, 95% CI)1.63 [1.18, 2.24]
29.1 short term (≤12 weeks)131Risk Ratio (M-H, Random, 95% CI)0.41 [0.04, 4.08]
29.2 by medium term (13-25 weeks)3155Risk Ratio (M-H, Random, 95% CI)1.40 [0.25, 7.82]
29.3 long term (≥26 weeks)10996Risk Ratio (M-H, Random, 95% CI)1.63 [1.23, 2.15]
Analysis 1.1.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 1 Relapse.

Analysis 1.2.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 2 Hospitalisation.

Analysis 1.3.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 3 Death.

Analysis 1.4.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 4 Global state: 1. Average score (CGI-S, high = worse).

Analysis 1.5.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 5 Global state: 2. Average score (GAS, low = worse).

Analysis 1.6.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 6 Global state: 3. Prodromal episodes.

Analysis 1.7.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 7 Global state: 4. Significant improvement.

Analysis 1.8.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 8 Mental state: 1. Average score (BPRS, high = worse).

Analysis 1.9.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 9 Mental state: 2. Negative symptom score (PANSS negative symptom subscale, high = worse).

Analysis 1.10.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 10 Mental state: 3. Negative symptom score (PANSS negative symptom subscale, high = worse, skew).

Mental state: 3. Negative symptom score (PANSS negative symptom subscale, high = worse, skew)
StudyInterventionMeanSDN
by long term (≥26 weeks)
Gaebel 2011Intermittent13.49.18
Gaebel 2011Maintenance9.12.819

Analysis 1.11.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 11 Mental state: 4. Negative symptom score (SANS, high = worse, skew).

Mental state: 4. Negative symptom score (SANS, high = worse, skew)
StudyInterventionMeanSDN
by long term (≥26 weeks)
Gaebel 2011Intermittent17.618.08
Gaebel 2011Maintenance5.79.719
Analysis 1.12.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 12 Mental state: 5. Positive symptom score (PANSS positive symptom subscale, high = worse).

Analysis 1.13.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 13 General functioning: 1. Average endpoint (LOFS, low = worse).

Analysis 1.14.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 14 General functioning: 2. Social functioning score (GAF, low = worse).

Analysis 1.15.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 15 General functioning: 3. Social functioning score (GSDS, high = worse, skew).

General functioning: 3. Social functioning score (GSDS, high = worse, skew)
StudyInterventionMeanSDN
by long term (≥26 weeks)
Wunderink 2007Intermittent5.84.565
Wunderink 2007Maintenance6.44.263
Analysis 1.16.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 16 General functioning: 4. Social functioning score (PAS, high = worse).

Analysis 1.17.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 17 General functioning: 5. Social functioning score (SAS, high = worse).

Analysis 1.18.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 18 Adverse effects: 1. Akathisia (HAS, high = worse, skew).

Adverse effects: 1. Akathisia (HAS, high = worse, skew)
StudyInterventionMeanSDN
by long term (≥26 weeks)
Gaebel 2011Intermittent0.00.08
Gaebel 2011Maintenance0.10.519
Analysis 1.19.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 19 Adverse effects: 2. Extrapyramidal side effects.

Analysis 1.20.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 20 Adverse effects: 3. Extrapyramidal symptoms (EPS, high = worse, skew).

Adverse effects: 3. Extrapyramidal symptoms (EPS, high = worse, skew)
StudyInterventionMeanSDN
by medium term (13-25 weeks)
Jolley 1989/1990Intermittent0.51.421
Jolley 1989/1990Maintenance2.82.722
by long term (≥26 weeks)
Gaebel 2011Intermittent0.00.08
Gaebel 2011Maintenance0.20.719
Jolley 1989/1990Intermittent0.41.221
Jolley 1989/1990Maintenance3.23.022
Analysis 1.21.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 21 Adverse effects: 4. Need for additional medication.

Analysis 1.22.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 22 Adverse effects: 5. Side effects (LUNSERS, high = worse, skew).

Adverse effects: 5. Side effects (LUNSERS, high = worse, skew)
StudyInterventionMeanSDN
by medium term (13-25 weeks)
Wunderink 2007Intermittent18.715.365
Wunderink 2007Maintenance20.313.863
by long term (≥26 weeks)
Wunderink 2007Intermittent24.536.665
Wunderink 2007Maintenance22.219.063

Analysis 1.23.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 23 Adverse effects: 6. Side effects (UKU, high = worse, skew).

Adverse effects: 6. Side effects (UKU, high = worse, skew)
StudyInterventionMeanSDN
by long term (≥26 weeks)
Gaebel 2011Intermittent0.41.18
Gaebel 2011Maintenance0.20.419
Analysis 1.24.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 24 Adverse effects: 7. Tardive dyskinesia.

Analysis 1.25.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 25 Adverse effects: 8. Tardive dyskinesia (AIMS, high = worse, skew).

Adverse effects: 8. Tardive dyskinesia (AIMS, high = worse, skew)
StudyInterventionMeanSDN
by medium term (13-25 weeks)
Remington 2011Intermittent0.00.017
Remington 2011Maintenance0.330.7818
by long term (≥26 weeks)
Gaebel 2011Intermittent0.00.08
Gaebel 2011Maintenance0.10.519
Analysis 1.26.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 26 Quality of life: 1. Average score (LQLP, low = worse).

Analysis 1.27.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 27 Quality of life: 2. Average score (QLS, low = worse).

Analysis 1.28.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 28 Quality of life: 3. Average score (WHOQoL-Bref, low = worse).

Analysis 1.29.

Comparison 1 ANY INTERMITTENT DRUG TECHNIQUE versus MAINTENANCE THERAPY, Outcome 29 Leaving the study early/ loss to follow-up.

Comparison 2. INTERMITTENT (EARLY-BASED) versus MAINTENANCE THERAPY
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Relapse2 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
1.1 long term (≥26 weeks)2155Risk Ratio (M-H, Fixed, 95% CI)2.33 [1.32, 4.12]
2 Hospitalisation5 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
2.1 by medium term (13-25 weeks)1101Risk Ratio (M-H, Fixed, 95% CI)4.08 [0.47, 35.25]
2.2 long term (≥26 weeks)5625Risk Ratio (M-H, Fixed, 95% CI)1.66 [1.33, 2.08]
3 Death2 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
3.1 by long term (≥26 weeks)2155Risk Ratio (M-H, Fixed, 95% CI)0.34 [0.05, 2.08]
4 Global state: 1. Average score (GAS, low = worse)2 Mean Difference (IV, Fixed, 95% CI)Subtotals only
4.1 medium term (13-25 weeks)175Mean Difference (IV, Fixed, 95% CI)-0.15 [-4.78, 4.48]
4.2 long term (≥26 weeks)282Mean Difference (IV, Fixed, 95% CI)0.99 [-4.24, 6.22]
5 Global state: 2. Prodromal episodes2 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
5.1 by long term (≥26 weeks)2155Risk Ratio (M-H, Fixed, 95% CI)3.19 [1.87, 5.44]
6 Mental state: 1. Average score (BPRS, high = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
6.1 long term (≥26 weeks)126Mean Difference (IV, Fixed, 95% CI)0.10 [-0.33, 0.53]
7 General functioning: 1. Average endpoint (LOFS, low = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
7.1 by long term (≥26 weeks)126Mean Difference (IV, Fixed, 95% CI)2.60 [-2.63, 7.83]
8 General functioning: 2. Social (PAS, high = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
8.1 by medium term (13-25 weeks)175Mean Difference (IV, Fixed, 95% CI)0.45 [-0.81, 1.71]
8.2 by long term (≥26 weeks)156Mean Difference (IV, Fixed, 95% CI)0.44 [-1.08, 1.96]
9 Adverse effects: 1. Extrapyramidal side effects1 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
9.1 hypomimia: by medium term (13-25 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.26 [0.06, 1.09]
9.2 hypomimia: by long term (≥26 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.17 [0.02, 1.33]
9.3 rigidity: by medium term (13-25 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.35 [0.01, 8.11]
9.4 rigidity: by long term (≥26 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.42 [0.09, 1.93]
9.5 tremor: by medium term (13-25 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.26 [0.03, 2.16]
9.6 tremor: by long term (≥26 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.21 [0.03, 1.65]
9.7 akathisia: by medium term (13-25 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.08 [0.00, 1.34]
9.8 akathisia: by long term (≥26 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.19 [0.05, 0.76]
9.9 gait abnormality: by medium term (13-25 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.15 [0.02, 1.11]
9.10 gait abnormality: by long term (≥26 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.08 [0.00, 1.34]
9.11 global parkinsonism: by medium term (13-25 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.35 [0.08, 1.54]
9.12 global parkinsonism: by long term (≥26 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.13 [0.02, 0.96]
9.13 global non-liveliness: by medium term (13-25 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.24 [0.08, 0.73]
9.14 global non-liveliness: by long term (≥26 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.09 [0.01, 0.61]
10 Adverse effects: 2. Extrapyramidal symptoms (EPS, high = worse, skew)  Other dataNo numeric data
10.1 by medium term (13-25 weeks)  Other dataNo numeric data
10.2 by long term (≥26 weeks)  Other dataNo numeric data
11 Adverse effects: 3. Need for additional medication1 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
11.1 by long term (≥26 weeks)1313Risk Ratio (M-H, Fixed, 95% CI)1.01 [0.87, 1.17]
12 Adverse effects: 4. Tardive dyskinesia1 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
12.1 by medium term (13-25 weeks)143Risk Ratio (M-H, Fixed, 95% CI)0.81 [0.37, 1.79]
12.2 by long term (≥26 weeks)130Risk Ratio (M-H, Fixed, 95% CI)0.56 [0.19, 1.70]
13 Quality of life: 1. Average score (QLS, low = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
13.1 by long term (≥26 weeks)126Mean Difference (IV, Fixed, 95% CI)0.5 [-0.35, 1.35]
14 Leaving the study early/ loss to follow-up5 Risk Ratio (M-H, Random, 95% CI)Subtotals only
14.1 long term (≥26 weeks)5562Risk Ratio (M-H, Random, 95% CI)1.67 [1.17, 2.37]
Analysis 2.1.

Comparison 2 INTERMITTENT (EARLY-BASED) versus MAINTENANCE THERAPY, Outcome 1 Relapse.

Analysis 2.2.

Comparison 2 INTERMITTENT (EARLY-BASED) versus MAINTENANCE THERAPY, Outcome 2 Hospitalisation.

Analysis 2.3.

Comparison 2 INTERMITTENT (EARLY-BASED) versus MAINTENANCE THERAPY, Outcome 3 Death.

Analysis 2.4.

Comparison 2 INTERMITTENT (EARLY-BASED) versus MAINTENANCE THERAPY, Outcome 4 Global state: 1. Average score (GAS, low = worse).

Analysis 2.5.

Comparison 2 INTERMITTENT (EARLY-BASED) versus MAINTENANCE THERAPY, Outcome 5 Global state: 2. Prodromal episodes.

Analysis 2.6.

Comparison 2 INTERMITTENT (EARLY-BASED) versus MAINTENANCE THERAPY, Outcome 6 Mental state: 1. Average score (BPRS, high = worse).

Analysis 2.7.

Comparison 2 INTERMITTENT (EARLY-BASED) versus MAINTENANCE THERAPY, Outcome 7 General functioning: 1. Average endpoint (LOFS, low = worse).

Analysis 2.8.

Comparison 2 INTERMITTENT (EARLY-BASED) versus MAINTENANCE THERAPY, Outcome 8 General functioning: 2. Social (PAS, high = worse).

Analysis 2.9.

Comparison 2 INTERMITTENT (EARLY-BASED) versus MAINTENANCE THERAPY, Outcome 9 Adverse effects: 1. Extrapyramidal side effects.

Analysis 2.10.

Comparison 2 INTERMITTENT (EARLY-BASED) versus MAINTENANCE THERAPY, Outcome 10 Adverse effects: 2. Extrapyramidal symptoms (EPS, high = worse, skew).

Adverse effects: 2. Extrapyramidal symptoms (EPS, high = worse, skew)
StudyInterventionMeanSDN
by medium term (13-25 weeks)
Jolley 1989/1990Intermittent0.51.421
Jolley 1989/1990Maintenance2.82.722
by long term (≥26 weeks)
Jolley 1989/1990Intermittent0.41.221
Jolley 1989/1990Maintenance3.23.022
Analysis 2.11.

Comparison 2 INTERMITTENT (EARLY-BASED) versus MAINTENANCE THERAPY, Outcome 11 Adverse effects: 3. Need for additional medication.

Analysis 2.12.

Comparison 2 INTERMITTENT (EARLY-BASED) versus MAINTENANCE THERAPY, Outcome 12 Adverse effects: 4. Tardive dyskinesia.

Analysis 2.13.

Comparison 2 INTERMITTENT (EARLY-BASED) versus MAINTENANCE THERAPY, Outcome 13 Quality of life: 1. Average score (QLS, low = worse).

Analysis 2.14.

Comparison 2 INTERMITTENT (EARLY-BASED) versus MAINTENANCE THERAPY, Outcome 14 Leaving the study early/ loss to follow-up.

Comparison 3. INTERMITTENT (CRISIS INTERVENTION) versus MAINTENANCE THERAPY
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Leaving the study early/ loss to follow-up1 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
1.1 long term (≥26 weeks)1237Risk Ratio (M-H, Fixed, 95% CI)1.57 [1.23, 2.00]
Analysis 3.1.

Comparison 3 INTERMITTENT (CRISIS INTERVENTION) versus MAINTENANCE THERAPY, Outcome 1 Leaving the study early/ loss to follow-up.

Comparison 4. INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Relapse3 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
1.1 short term (≤12 weeks)1128Risk Ratio (M-H, Fixed, 95% CI)2.91 [0.31, 27.22]
1.2 medium term (13-25 weeks)1128Risk Ratio (M-H, Fixed, 95% CI)2.33 [0.87, 6.22]
1.3 long term (≥26 weeks)3219Risk Ratio (M-H, Fixed, 95% CI)2.76 [1.63, 4.67]
2 Global state: 1. Average score (CGI-S, high = worse)1 Mean Difference (IV, Random, 95% CI)Subtotals only
2.1 by long term (≥26 weeks)127Mean Difference (IV, Random, 95% CI)0.30 [-0.33, 0.93]
3 Global state: 2. Average score (GAS, low = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
3.1 medium term (13-25 weeks)151Mean Difference (IV, Fixed, 95% CI)-1.54 [-10.42, 7.34]
3.2 long term (≥26 weeks)151Mean Difference (IV, Fixed, 95% CI)1.83 [-4.66, 8.32]
4 Mental state: 1. Average score (BPRS, high = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
4.1 by medium term (13-25 weeks)151Mean Difference (IV, Fixed, 95% CI)1.85 [-3.03, 6.73]
4.2 long term (≥26 weeks)151Mean Difference (IV, Fixed, 95% CI)0.20 [-2.77, 3.17]
5 Mental state: 2. Negative symptom score (PANSS negative symptom subscale, high = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
5.1 medium term (13-25 weeks)1128Mean Difference (IV, Fixed, 95% CI)-0.80 [-2.43, 0.83]
5.2 by long term (≥26 weeks)1128Mean Difference (IV, Fixed, 95% CI)-1.20 [-3.19, 0.79]
6 Mental state: 3. Negative symptom score (PANSS negative symptom subscale, high = worse, skew)  Other dataNo numeric data
6.1 by long term (≥26 weeks)  Other dataNo numeric data
7 Mental state: 4. Negative symptom score (SANS, high = worse, skew)  Other dataNo numeric data
7.1 by long term (≥26 weeks)  Other dataNo numeric data
8 Mental state: 5. Positive symptom score (PANSS positive symptom subscale, high = worse)2 Mean Difference (IV, Fixed, 95% CI)Subtotals only
8.1 medium term (13-25 weeks)1128Mean Difference (IV, Fixed, 95% CI)-0.80 [-1.58, -0.02]
8.2 by long term (≥26 weeks)2155Mean Difference (IV, Fixed, 95% CI)0.40 [-0.79, 1.59]
9 General functioning: 1. Social functioning score (GAF, low = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
9.1 by long term (≥26 weeks)127Mean Difference (IV, Fixed, 95% CI)-9.0 [-15.92, -2.08]
10 General functioning: 2. Social functioning score (GSDS, high = worse, skew)  Other dataNo numeric data
10.1 by long term (≥26 weeks)  Other dataNo numeric data
11 General functioning: 3. Social functioning score (SAS, high = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
11.1 by medium term (13-25 weeks)151Mean Difference (IV, Fixed, 95% CI)0.06 [-0.10, 0.22]
11.2 by long term (≥26 weeks)151Mean Difference (IV, Fixed, 95% CI)0.05 [-0.08, 0.18]
12 Adverse effects: 1. Akathisia (HAS, high = worse, skew)  Other dataNo numeric data
12.1 by long term (≥26 weeks)  Other dataNo numeric data
13 Adverse effects: 2. Extrapyramidal symptoms (EPS, high = worse, skew)  Other dataNo numeric data
13.2 by long term (≥26 weeks)  Other dataNo numeric data
14 Adverse effects: 3. Side effects (LUNSERS, high = worse, skew)  Other dataNo numeric data
14.1 by medium term (13-25 weeks)  Other dataNo numeric data
14.2 by long term (≥26 weeks)  Other dataNo numeric data
15 Adverse effects: 4. Side effects (UKU, high = worse, skew)  Other dataNo numeric data
15.1 by long term (≥26 weeks)  Other dataNo numeric data
16 Adverse effects: 5. Tardive dyskinesia1 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
16.1 by long term (≥26 weeks)185Risk Ratio (M-H, Fixed, 95% CI)0.0 [0.0, 0.0]
17 Adverse effects: 6. Tardive dyskinesia (AIMS, high = worse, skew)  Other dataNo numeric data
17.2 by long term (≥26 weeks)  Other dataNo numeric data
18 Quality of life: 1. Average score (LQLP, low = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
18.1 by long term (≥26 weeks)127Mean Difference (IV, Fixed, 95% CI)-0.5 [-1.38, 0.38]
19 Quality of life: 2. Average score (WHOQoL-Bref, low = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
19.1 by medium term (13-25 weeks)1128Mean Difference (IV, Fixed, 95% CI)-0.70 [-4.27, 2.87]
19.2 by long term (≥26 weeks)1128Mean Difference (IV, Fixed, 95% CI)-0.90 [-5.39, 3.59]
20 Leaving the study early/ loss to follow-up3 Risk Ratio (M-H, Random, 95% CI)Subtotals only
20.1 long term (≥26 weeks)3257Risk Ratio (M-H, Random, 95% CI)2.12 [0.70, 6.37]
Analysis 4.1.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 1 Relapse.

Analysis 4.2.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 2 Global state: 1. Average score (CGI-S, high = worse).

Analysis 4.3.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 3 Global state: 2. Average score (GAS, low = worse).

Analysis 4.4.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 4 Mental state: 1. Average score (BPRS, high = worse).

Analysis 4.5.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 5 Mental state: 2. Negative symptom score (PANSS negative symptom subscale, high = worse).

Analysis 4.6.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 6 Mental state: 3. Negative symptom score (PANSS negative symptom subscale, high = worse, skew).

Mental state: 3. Negative symptom score (PANSS negative symptom subscale, high = worse, skew)
StudyInterventionMeanSDN
by long term (≥26 weeks)
Gaebel 2011Intermittent13.49.18
Gaebel 2011Maintenance9.12.819

Analysis 4.7.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 7 Mental state: 4. Negative symptom score (SANS, high = worse, skew).

Mental state: 4. Negative symptom score (SANS, high = worse, skew)
StudyInterventionMeanSDN
by long term (≥26 weeks)
Gaebel 2011Intermittent17.618.08
Gaebel 2011Maintenance5.79.719
Analysis 4.8.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 8 Mental state: 5. Positive symptom score (PANSS positive symptom subscale, high = worse).

Analysis 4.9.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 9 General functioning: 1. Social functioning score (GAF, low = worse).

Analysis 4.10.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 10 General functioning: 2. Social functioning score (GSDS, high = worse, skew).

General functioning: 2. Social functioning score (GSDS, high = worse, skew)
StudyInterventionMeanSDN
by long term (≥26 weeks)
Wunderink 2007Intermittent5.84.565
Wunderink 2007Maintenance6.44.263
Analysis 4.11.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 11 General functioning: 3. Social functioning score (SAS, high = worse).

Analysis 4.12.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 12 Adverse effects: 1. Akathisia (HAS, high = worse, skew).

Adverse effects: 1. Akathisia (HAS, high = worse, skew)
StudyInterventionMeanSDN
by long term (≥26 weeks)
Gaebel 2011Intermittent0.00.08
Gaebel 2011Maintenance0.10.519

Analysis 4.13.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 13 Adverse effects: 2. Extrapyramidal symptoms (EPS, high = worse, skew).

Adverse effects: 2. Extrapyramidal symptoms (EPS, high = worse, skew)
StudyInterventionMeanSDN
by long term (≥26 weeks)
Gaebel 2011Intermittent0.00.08
Gaebel 2011Maintenance0.20.719

Analysis 4.14.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 14 Adverse effects: 3. Side effects (LUNSERS, high = worse, skew).

Adverse effects: 3. Side effects (LUNSERS, high = worse, skew)
StudyInterventionMeanSDN
by medium term (13-25 weeks)
Wunderink 2007Intermittent18.715.365
Wunderink 2007Maintenance20.313.863
by long term (≥26 weeks)
Wunderink 2007Intermittent24.536.665
Wunderink 2007Maintenance22.219.063

Analysis 4.15.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 15 Adverse effects: 4. Side effects (UKU, high = worse, skew).

Adverse effects: 4. Side effects (UKU, high = worse, skew)
StudyInterventionMeanSDN
by long term (≥26 weeks)
Gaebel 2011Intermittent0.41.18
Gaebel 2011Maintenance0.20.419
Analysis 4.16.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 16 Adverse effects: 5. Tardive dyskinesia.

Analysis 4.17.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 17 Adverse effects: 6. Tardive dyskinesia (AIMS, high = worse, skew).

Adverse effects: 6. Tardive dyskinesia (AIMS, high = worse, skew)
StudyInterventionMeanSDN
by long term (≥26 weeks)
Gaebel 2011Intermittent0.00.08
Gaebel 2011Maintenance0.10.519
Analysis 4.18.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 18 Quality of life: 1. Average score (LQLP, low = worse).

Analysis 4.19.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 19 Quality of life: 2. Average score (WHOQoL-Bref, low = worse).

Analysis 4.20.

Comparison 4 INTERMITTENT (GRADUALLY INCREASED DRUG-FREE PERIODS) versus MAINTENANCE THERAPY, Outcome 20 Leaving the study early/ loss to follow-up.

Comparison 5. INTERMITTENT (DRUG HOLIDAY) versus MAINTENANCE THERAPY
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Relapse6 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
1.1 short term (≤12 weeks)3268Risk Ratio (M-H, Fixed, 95% CI)1.59 [0.94, 2.70]
1.2 medium term (13-25 weeks)3272Risk Ratio (M-H, Fixed, 95% CI)2.15 [1.25, 3.68]
1.3 long term (≥26 weeks)262Risk Ratio (M-H, Fixed, 95% CI)1.70 [0.54, 5.38]
2 Hospitalisation1 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
2.1 medium term (13-25 weeks)135Risk Ratio (M-H, Fixed, 95% CI)0.26 [0.03, 2.14]
3 Global state: 1. Average score (CGI-S, high = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
3.1 medium term (13-25 weeks)135Mean Difference (IV, Fixed, 95% CI)-0.42 [-1.08, 0.24]
4 Global state: 2. Average score (GAS, low = worse)1 Mean Difference (IV, Fixed, 95% CI)Subtotals only
4.1 short term (≤12 weeks)131Mean Difference (IV, Fixed, 95% CI)-3.61 [-5.67, -1.55]
5 Global state: 3. Significant improvement1 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
5.1 short term (≤12 weeks)160Risk Ratio (M-H, Fixed, 95% CI)0.4 [0.18, 0.89]
5.2 medium term (13-25 weeks)160Risk Ratio (M-H, Fixed, 95% CI)0.32 [0.15, 0.68]
6 Adverse effects: 1. Need for additional medication1 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
6.1 by long term (≥26 weeks)133Risk Ratio (M-H, Fixed, 95% CI)0.15 [0.02, 1.07]
7 Adverse effects: 2. Tardive dyskinesia2 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
7.1 by long term (≥26 weeks)250Risk Ratio (M-H, Fixed, 95% CI)1.64 [0.82, 3.30]
8 Adverse effects: 3. Tardive dyskinesia (AIMS, high = worse, skew)  Other dataNo numeric data
8.1 by medium term (13-25 weeks)  Other dataNo numeric data
9 Leaving the study early/ loss to follow-up6248Risk Ratio (M-H, Random, 95% CI)1.15 [0.45, 2.90]
9.1 short term (≤12 weeks)131Risk Ratio (M-H, Random, 95% CI)0.41 [0.04, 4.08]
9.2 by medium term (13-25 weeks)3155Risk Ratio (M-H, Random, 95% CI)1.40 [0.25, 7.82]
9.3 long term (≥26 weeks)262Risk Ratio (M-H, Random, 95% CI)1.01 [0.46, 2.20]
Analysis 5.1.

Comparison 5 INTERMITTENT (DRUG HOLIDAY) versus MAINTENANCE THERAPY, Outcome 1 Relapse.

Analysis 5.2.

Comparison 5 INTERMITTENT (DRUG HOLIDAY) versus MAINTENANCE THERAPY, Outcome 2 Hospitalisation.

Analysis 5.3.

Comparison 5 INTERMITTENT (DRUG HOLIDAY) versus MAINTENANCE THERAPY, Outcome 3 Global state: 1. Average score (CGI-S, high = worse).

Analysis 5.4.

Comparison 5 INTERMITTENT (DRUG HOLIDAY) versus MAINTENANCE THERAPY, Outcome 4 Global state: 2. Average score (GAS, low = worse).

Analysis 5.5.

Comparison 5 INTERMITTENT (DRUG HOLIDAY) versus MAINTENANCE THERAPY, Outcome 5 Global state: 3. Significant improvement.

Analysis 5.6.

Comparison 5 INTERMITTENT (DRUG HOLIDAY) versus MAINTENANCE THERAPY, Outcome 6 Adverse effects: 1. Need for additional medication.

Analysis 5.7.

Comparison 5 INTERMITTENT (DRUG HOLIDAY) versus MAINTENANCE THERAPY, Outcome 7 Adverse effects: 2. Tardive dyskinesia.

Analysis 5.8.

Comparison 5 INTERMITTENT (DRUG HOLIDAY) versus MAINTENANCE THERAPY, Outcome 8 Adverse effects: 3. Tardive dyskinesia (AIMS, high = worse, skew).

Adverse effects: 3. Tardive dyskinesia (AIMS, high = worse, skew)
StudyInterventionMeanSDN
by medium term (13-25 weeks)
Remington 2011Intermittent0.000.0017
Remington 2011Maintenance0.330.7818
Analysis 5.9.

Comparison 5 INTERMITTENT (DRUG HOLIDAY) versus MAINTENANCE THERAPY, Outcome 9 Leaving the study early/ loss to follow-up.

Comparison 6. ANY INTERMITTENT DRUG TECHNIQUE versus PLACEBO
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Relapse2 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
1.1 short term (≤12 weeks)1260Risk Ratio (M-H, Fixed, 95% CI)0.22 [0.10, 0.45]
1.2 medium term (13-25 weeks)2290Risk Ratio (M-H, Fixed, 95% CI)0.37 [0.24, 0.58]
2 Global state: 1. significant improvement1 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
2.1 medium term (13-25 weeks)130Risk Ratio (M-H, Fixed, 95% CI)1.0 [0.24, 4.18]
3 Leaving the study early/ loss to follow-up2 Risk Ratio (M-H, Random, 95% CI)Subtotals only
3.1 by medium term (13-25 weeks)290Risk Ratio (M-H, Random, 95% CI)1.97 [1.28, 3.01]
Analysis 6.1.

Comparison 6 ANY INTERMITTENT DRUG TECHNIQUE versus PLACEBO, Outcome 1 Relapse.

Analysis 6.2.

Comparison 6 ANY INTERMITTENT DRUG TECHNIQUE versus PLACEBO, Outcome 2 Global state: 1. significant improvement.

Analysis 6.3.

Comparison 6 ANY INTERMITTENT DRUG TECHNIQUE versus PLACEBO, Outcome 3 Leaving the study early/ loss to follow-up.

Comparison 7. ANY INTERMITTENT DRUG TECHNIQUE (SPECIFIC DRUG) versus MAINTENANCE THERAPY (SPECIFIC DRUG)
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Relapse12 Risk Ratio (M-H, Fixed, 95% CI)Subtotals only
1.1 intermittent fluphenazine decanoate (low dose) vs maintained fluphenazine decanoate (moderate dose) - short term (≤12 weeks)131Risk Ratio (M-H, Fixed, 95% CI)0.0 [0.0, 0.0]
1.2 intermittent haloperidol equivalents (low dose) vs maintained haloperidol equivalents (low dose) - short term (≤12 weeks)1128Risk Ratio (M-H, Fixed, 95% CI)2.91 [0.31, 27.22]
1.3 various intermittent typical antipsychotics (moderate dose) vs maintained typical antipsychotics (moderate dose) - short term (≤12 weeks)1177Risk Ratio (M-H, Fixed, 95% CI)1.73 [0.53, 5.70]
1.4 intermittent haloperidol equivalents (low dose) vs maintained haloperidol equivalents (low dose) - medium term (13-25 weeks)1128Risk Ratio (M-H, Fixed, 95% CI)2.33 [0.87, 6.22]
1.5 various intermittent atypical antipsychotics (dosage unclear) vs maintained atypical/typical antipsychotics (dosage unclear) - medium term (13-25 weeks)135Risk Ratio (M-H, Fixed, 95% CI)1.41 [0.37, 5.40]
1.6 various intermittent typical antipsychotics (low dose) vs maintained typical antipsychotics (low dose) - medium term (13-25 weeks)160Risk Ratio (M-H, Fixed, 95% CI)1.88 [0.94, 3.75]
1.7 various intermittent typical antipsychotics (moderate dose) vs maintained typical antipsychotics (moderate dose) - medium term (13-25 weeks)2551Risk Ratio (M-H, Fixed, 95% CI)3.75 [1.42, 9.94]
1.8 intermittent chlorpromazine equivalents (low dose) vs maintained chlorpromazine equivalents (low dose) - long term (≥26 weeks)2148Risk Ratio (M-H, Fixed, 95% CI)2.62 [1.30, 5.28]
1.9 intermittent clozapine (low dose) vs maintained clozapine (low dose) long term (≥26 weeks)125Risk Ratio (M-H, Fixed, 95% CI)15.30 [0.92, 255.51]
1.10 intermittent haloperidol equivalents (low dose) vs maintained haloperidol equivalents (low dose) - long term (≥26 weeks)3226Risk Ratio (M-H, Fixed, 95% CI)2.53 [1.60, 4.01]
1.11 intermittent pimozide (high dose) vs fluphenazine decanoate (low dose) - long term (≥26 weeks)262Risk Ratio (M-H, Fixed, 95% CI)1.70 [0.54, 5.38]
1.12 various intermittent typical antipsychotics (low dose) vs maintained typical antipsychotics - long term (≥26 weeks)126Risk Ratio (M-H, Fixed, 95% CI)2.93 [0.38, 22.75]
2 Hospitalisation6661Risk Ratio (M-H, Fixed, 95% CI)1.58 [1.28, 1.97]
2.1 various intermittent atypical antipsychotics (dosage unclear) vs maintained atypical/typical antipsychotics (dosage unclear) - medium term (13-25 weeks)135Risk Ratio (M-H, Fixed, 95% CI)0.26 [0.03, 2.14]
2.2 intermittent chlorpromazine equivalents (low dose) vs maintained chlorpromazine equivalents (low dose) - long term (≥26 weeks)1101Risk Ratio (M-H, Fixed, 95% CI)1.53 [0.68, 3.42]
2.3 intermittent chlorpromazine equivalents (low dose) vs maintained chlorpromazine equivalents (moderate dose) - long term (≥26 weeks)1116Risk Ratio (M-H, Fixed, 95% CI)1.48 [0.97, 2.26]
2.4 intermittent chlorpromazine equivalents (low dose) vs maintained chlorpromazine equivalents (high dose) - long term (≥26 weeks)142Risk Ratio (M-H, Fixed, 95% CI)1.1 [0.60, 2.02]
2.5 intermittent fluphenazine decanoate (low dose) v maintained fluphenazine decanoate (low+moderate dose) - long term (≥26 weeks)1313Risk Ratio (M-H, Fixed, 95% CI)1.81 [1.33, 2.48]
2.6 intermittent haloperidol equivalents (low dose) vs maintained haloperidol equivalents (low dose) - long term (≥26 weeks)154Risk Ratio (M-H, Fixed, 95% CI)2.75 [1.00, 7.57]
Analysis 7.1.

Comparison 7 ANY INTERMITTENT DRUG TECHNIQUE (SPECIFIC DRUG) versus MAINTENANCE THERAPY (SPECIFIC DRUG), Outcome 1 Relapse.

Analysis 7.2.

Comparison 7 ANY INTERMITTENT DRUG TECHNIQUE (SPECIFIC DRUG) versus MAINTENANCE THERAPY (SPECIFIC DRUG), Outcome 2 Hospitalisation.

Appendices

Appendix 1. Previous search term

We searched The Cochrane Schizophrenia Group Trials Register (March 2006) using the phrase:
[(intermit* or drug?holiday* or drug?free* or internal?med*) in title, abstract and index fields in REFERENCE) OR (intermittent medication or drug-free period in interventions field in STUDY)]

Appendix 2. Previous data collection and analysis

1. Selection of trials
Material downloaded from electronic sources will include details of author, institution or journal of publication.

We (MM and AA) will inspect each report in order to ensure reliable selection. We will resolve any disagreement by discussion, and where there is still doubt, we will acquire the full article for further inspection. Once the full articles are obtained, we (MM and AA) will independently decide whether the studies meet the review criteria. If disagreement cannot be resolved by discussion, we will seek further information and add these trials to the list of those awaiting assessment.

2. Assessment of methodological quality
We will assess the methodological quality of each of included trials in this review using the criteria described in the Cochrane Handbook (Higgins 2005) and the Jadad Scale (Jadad 1996). The former is based on the evidence of a strong relationship between allocation concealment and direction of effect (Schulz 1995). The categories are defined below:

A. Low risk of bias (adequate allocation concealment)
B. Moderate risk of bias (some doubt about the results) 
C. High risk of bias (inadequate allocation concealment). For the purpose of the analysis in this review, trials will be included if they met the Cochrane Handbook criteria A or B.

The Jadad Scale measures a wider range of factors that impact on the quality of a trial. The scale includes three items:
1. Was the study described as randomised?
2. Was the study described as double-blind? 
3. Was there a description of withdrawals and drop outs?

Each item receives one point if the answer is positive. In addition, a point can be deducted if either the randomisation or the blinding/masking procedures described are inadequate. For this review we will use a cut-off of two points on the Jadad scale to check the assessment made by the Handbook criteria. However, we will not use the Jadad Scale to exclude trials.

3. Data collection 
We (MM and AA) will independently extract data from selected trials, while KB and MAM will separately re-extract information from the same trials. When disputes arise we will attempt to resolve these by discussion. When this is not possible and further information is necessary to resolve the dilemma, we will not enter data but we will add the trial to the list of those awaiting assessment.

4. Data synthesis 
4.1 Data types
We will assess outcomes using continuous (for example changes on a behaviour scale), categorical (for example, one of three categories on a behaviour scale, such as 'little change', 'moderate change' or 'much change') or dichotomous (for example, either 'no important changes' or 'important changes' in a person's behaviour) measures. Currently RevMan does not support categorical data so we will be unable to analyse these.

4.2 Incomplete data 
We will not include trial outcomes if more than 50% of people are not reported in the final analysis.

4.3 Dichotomous - yes/no - data 
We will carry out an intention-to-treat analysis. On the condition that more than 50% of people complete the study, we will count everyone allocated to the intervention, whether they completed the follow-up or not. We will assume that those who dropped out had the negative outcome, with the exception of death. Where possible, we will make efforts to convert outcome measures to dichotomous data. This can be done by identifying cut-off points on rating scales and dividing participants accordingly into 'clinically improved' or 'not clinically improved'. If the authors of a study have used a predefined cut-off point for determining clinical effectiveness, we will use this where appropriate. Otherwise, we will generally assume that if there has been a 50% reduction in a scale-derived score, this could be considered as a clinically significant response. Similarly, we will consider a rating of 'at least much improved' according to the Clinical Global Impression Scale (Guy 1976) as a clinically significant response.

We will calculate the relative risk (RR) and its 95% confidence interval (CI) based on the fixed-effect model. We will calculate the relative risk of statistically significantly heterogeneous outcomes using a random-effects model. When the overall results are significant, we will calculate the number needed to treat (NNT) and the number needed to harm (NNH) as the inverse of the risk difference.

4.4 Continuous data 
4.4.1 Normally distributed data: continuous data on clinical and social outcomes are often not normally distributed. To avoid the pitfall of applying parametric tests to non-parametric data, we will apply the following standards to all data before inclusion: (a) standard deviations (SDs) and means are reported in the paper or are obtainable from the authors; (b) when a scale starts from the finite number zero, the SD, when multiplied by two, is less than the mean (as otherwise the mean is unlikely to be an appropriate measure of the centre of the distribution, (Altman 1996); (c) if a scale started from a positive value (such as PANSS which can have values from 30 to 210), the calculation described above will be modified to take the scale starting point into account. In these cases skew is present if 2 SD > (S-Smin), where S is the mean score and Smin is the minimum score. Endpoint scores on scales often have a finite start and end point and these rules can be applied to them. When continuous data are presented on a scale which includes a possibility of negative values (such as change on a scale), it is difficult to tell whether data are non-normally distributed (skewed) or not. We will enter skewed data from studies of less than 200 participants in additional tables rather than into an analysis. Skewed data poses less of a problem when looking at means if the sample size is large and we will enter these into a synthesis.

For change data (endpoint minus baseline), the situation is even more problematic. In the absence of individual patient data it is impossible to know if data are skewed, though this is likely. After consulting the ALLSTAT electronic statistics mailing list, we have decided to present change data in MetaView in order to summarise available information. In doing this, we will assume either that data are not skewed or that the analyses could cope with the unknown degree of skew. Without individual patient data it is impossible to test this assumption. Where both change and endpoint data are available for the same outcome category, we will only present endpoint data. We acknowledge that by doing this we will exclude much of the published change data, but argue that endpoint data is more clinically relevant and that if we present change data along with endpoint data, it would be given undeserved equal prominence. We will contact authors of studies reporting only change data for endpoint figures. We will report non-normally distributed data in the 'other data types' tables.

4.4.2 Rating scales: A wide range of instruments are available to measure mental health outcomes. These instruments vary in quality and many are not valid, or even ad hoc. For outcome instruments some minimum standards have to be set. It has been shown that the use of rating scales which have not been described in a peer-reviewed journal (Marshall 2000) are associated with bias, therefore we will exclude the results of such scales. Furthermore, we stipulate that the instrument should either be a self-report or be completed by an independent rater or relative (not the therapist), and that the instrument could be considered a global assessment of an area of functioning. However, as it is expected that therapists would frequently also be the rater, we will include such data but comment on the data as 'prone to bias'.

Whenever possible we will take the opportunity to make direct comparisons between trials that use the same measurement instrument to quantify specific outcomes. Where continuous data are presented from different scales rating the same effect, we will present both sets of data and inspect the general direction of effect.

4.4.3 Summary statistic
For continuous outcomes, we will estimate a weighted mean difference (WMD) between groups, again based on the fixed-effect model. Afterwards, we will carry out a sensitivity analysis to find heterogeneous data. We will re-assess heterogeneous data using a random-effects model.

4.5 Cluster trials
Studies increasingly employ 'cluster randomisation' (such as randomisation by clinician or practice) but analysis and pooling of clustered data poses problems. Firstly, authors often fail to account for intraclass correlation in clustered studies, leading to a 'unit of analysis' error (Divine 1992) whereby P values are spuriously low, confidence intervals unduly narrow and statistical significance overestimated. This causes type I errors (Bland 1997Gulliford 1999).

Where clustering is not accounted for in primary studies, we will present the data in a table, with a (*) symbol to indicate the presence of a probable unit of analysis error. In subsequent versions of this review we will seek to contact first authors of studies to obtain intra-class correlation co-efficients (ICCs) of their clustered data and to adjust for this using accepted methods (Gulliford 1999). Where clustering has been incorporated into the analysis of primary studies, we will also present these data as if from a non-cluster randomised study, but adjust for the clustering effect.

We have sought statistical advice and have been advised that the binary data as presented in a report should be divided by a 'design effect'. This is calculated using the mean number of participants per cluster (m) and the ICC [Design effect = 1+(m-1)*ICC] (Donner 2002). If the ICC is not reported it will be assumed to be 0.1 (Ukoumunne 1999).

If cluster studies are appropriately analysed taking into account ICCs and relevant data documented in the report, synthesis with other studies will be possible using the generic inverse variance technique.

5. Investigation for heterogeneity
Firstly, we will consider all the included studies within any comparison to judge clinical heterogeneity. Then we will visually inspect graphs to investigate the possibility of statistical heterogeneity. This will be supplemented using, primarily, the I-squared statistic. This provides an estimate of the percentage of variability due to heterogeneity rather than chance alone. Where the I-squared estimate is greater than or equal to 75%, we will interpret this as indicating the presence of high levels of heterogeneity (Higgins 2003). If inconsistency is high, we will not summate data, but present this separately and investigate the reasons for heterogeneity.

6. Addressing publication bias 
We will enter data from all identified and selected trials into a funnel graph (trial effect versus trial size) in an attempt to investigate the likelihood of overt publication bias.

7. Subgroup analyses
We will carry out a subgroup analysis to compare results between different interventions, as defined in 'Types of interventions'.

8. General 
Where possible, we will enter data in such a way that the area to the left of the line of no effect indicates a favourable outcome for intermittent treatment.

Contributions of authors

Stephanie Sampson - completed data extraction, helped with completion of risk of bias tables and writing the review.

Clive Elliot Adams - helped with data extraction and writing the review.

Mouhamad Mansour - helped write the protocol.

Karla Soares-Weiser - completion of risk of bias tables.

Nicola Maayan - completion of risk of bias tables.

Declarations of interest

Stephanie Sampson - none known.

Clive Elliot Adams - none known.

Mouhamad Mansour - none known.

Karla Soares-Weiser - I currently work for Enhance Reviews Ltd, a company that carries out systematic reviews mostly for the public sector, we currently do not provide services for the pharmaceutical industry.

Nicola Maayan - I currently work for Enhance Reviews Ltd, a company that carries out systematic reviews mostly for the public sector, we currently do not provide services for the pharmaceutical industry.

Sources of support

Internal sources

  • Rida Saeed Centre, Damascus University, Syrian Arab Republic.

External sources

  • National Institute for Health Research (NIHR), UK.

    Cochrane Collaboration Programme Grant 2011, Reference number: 10/4001/15

Differences between protocol and review

See Table 2 to review differences between protocol and review that have been classified as either 'minor' or 'major'.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Blackburn 1961

MethodsAllocation: randomised.
Blinding: not stated.
Duration: 16 weeks.
Setting: inpatients, 'fully-privileged, open-ward', hospital.
Design: parallel.
Participants

Diagnosis: chronic schizophrenia.
N = 60.

Age: 20-40 years.
Sex: 60 M.

Racial origin: not stated.
Consent: not stated.

History: length of current hospitalisation ranged form three months to 129 months, with a mean of 45 months.

Inclusion: judged by psychiatric staff to be clinically tranquillised and to be making a satisfactory hospital adjustment on phenothiazine medication.

Exclusion: not stated.

Interventions

Initial daily drug dosages ranged from 15 mg to 150 mg prochlorperazine, 12 mg to 24 mg of perphenazine, 50 mg to 800 mg chlorpromazine, 200 mg to -400 mg promazine and 6 mg of trifluoperazine.

1. Intermittent: participants were further randomly subdivided into two groups of n = 15:

a. PD (placebo-drug): placebo (thiamine chloride) was substituted for phenothiazine for the first eight weeks, then drugs were re-introduced for the final eight weeks, n = 15.

b. PP (placebo-placebo): placebo was substituted for the whole 16 weeks of the study, n = 15.

2. Maintenance: DD (drug-drug) group remained on initial type and dosage of tranquillising medication throughout the period of the study, n = 30.

Outcomes

Global state: relapse: this included participants who were judged to have 'significantly deteriorated' indicated by a decrease in his adjustment score by at least one SE of measurement since initial evaluation, or if their behaviour had necessitated emergency resumption of drugs and/or transfer to a closed ward before the next scheduled rating period - by 8 and 16 weeks.

Global state: significant improvement: indicated by an increase in his adjustment score by at least one SE of measurement since initial evaluation, or if he was released from hospital in that period - by 8 and 16 weeks.

Leaving the study early.

Unable to use - Behaviour: Patient Adjustment Report; Taylor Manifest Anxiety Scale (no scale data reported).

NotesAvailable data from eight weeks combines both PD and PP groups, where data from the full 16 weeks presents individual group results. Meta-analysed data will combine both intermittent groups, but will present data comparing ANY INTERMITTENT DRUG TECHNIQUE vs PLACEBO separately at the end of the 16 weeks study period (medium term).
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskRandomised - no further description.
Allocation concealment (selection bias)Unclear riskNot described.
Blinding (performance bias and detection bias)
All outcomes
High riskNot described - 'the placebo was physically different from the phenothiazine' and thus identifiable to participants and staff as a change in medication. The nursing staff identified that this was indeed a placebo.
Incomplete outcome data (attrition bias)
All outcomes
Low riskFollow-up: 88% - 'seven participants lost for administrative reasons unrelated to clinical condition' (n = 25 DD; n = 14 PD; n = 14 PP). ITT used.
Selective reporting (reporting bias)High riskNot all outcomes reported.
Other biasUnclear risk

Funding source: not stated.

Rating scales: ratings completed independently by both the nurse and the aide on each ward at each of the three rating periods.

Caffey 1964

MethodsAllocation: randomised.
Blinding: double.
Duration: 16 weeks.
Setting: inpatients, 18-hospital collaborative study (hospitalised for two or more years).
Design: parallel.
Participants

Diagnosis: chronic schizophrenia (1/3 classified as paranoid sub-type).
N = 348.

Age: average 40 years.
Sex: 177 M.

Racial origin: not stated.
Consent: not stated.

History: average patient hospitalised for nearly 10 years; participants on chlorpromazine on average had been receiving a daily dosage of 400 mg over two years; participants on thioridazine on average had been receiving 350 mg daily for one year. All treated with fairly stable doses of at least 100 mg and not more than 800 mg daily of either chlorpromazine or thioridazine for at least three months immediately prior to the beginning of the study.

Inclusion: schizophrenia diagnosis.

Exclusion: CNS disease; history of seizures; prefrontal lobotomy.

Interventions

1. Intermittent: reduced total dosage on an intermittent-schedule (received usual daily dosage on Monday, Wednesday and Friday; 3/7 of their usual dosage, medication administered in standard 100 mg tablets), n = 89.
2. Placebo: corresponding placebo group 1 - patients received the same number of tablets prior to the study (daily); corresponding placebo group 2 - patients received the same number of tablets prior to the study (Monday, Wednesday or Friday), n = 171.

3. Maintenance: continue to receive either chlorpromazine or thioridazine daily (participants on chlorpromazine had been receiving it for over two years at an average daily dose of 400mg; participants on thioridazine had been receiving it for one year at an average daily dose of 350 mg (medication administered in standard 100 mg tablets), n = 88.

Outcomes

Global state: relapse: participants were judged clinically to have relapsed by the Principal Investigator and the rest of the treatment staff jointly - reported from 1-16 weeks.

Unable to use -

Mental state: PRP (no SD).

Adverse effects: IMPS (no SD).

Notes 
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskRandomised - 'randomly assigned to one of four groups' - no further description.
Allocation concealment (selection bias)Unclear riskNot described.
Blinding (performance bias and detection bias)
All outcomes
Low riskDouble blind - 'double blind control between corresponding active and placebo groups was maintained'...'the placebo tablets were identical in appearance to the commercial product'.
Incomplete outcome data (attrition bias)
All outcomes
High riskFollow-up: 73% (n = 254). Not all participants completed rating scales. ITT used.
Selective reporting (reporting bias)High riskNot all expected outcomes reported - only data presented for either relapse or non-relapse are available without SD and overall scores for all participants are not available; 'relapse' was not defined in this study.
Other biasUnclear risk

Funding source: drugs used in the study were donated by Sandoz (thioridazine-Mellaril) and Smith, Kline & French (chlorpromazine-Thorazine).

Rating scales: IMPS was completed independently by two psychiatrists or psychologists. The PRP was completed by two nurses or nursing assistants.

Carpenter 1987

MethodsAllocation: randomised.
Blinding: open label.
Duration: 2 years.
Setting: outpatients, Walnut Street Clinic, Maryland Psychiatric Research Center, Maryland, US.
Design: parallel.
Participants

Diagnosis: schizophrenia (76% were considered chronically ill, and approximately 50% were diagnosed with paranoid schizophrenia, with the remaining diagnosis fairly evenly distributed between RDC undifferentiated schizophrenia and schizoaffective disorder, mainly schizophrenia).
N = 42.

Age: mean 31 years.
Sex: 20 M; 22 F.

Racial origin: n = 34 Black; n = 8 White.
Consent: written consent form required, 'meeting both NIMH and local institutional board regulations'.

History: all participants had recently experienced a psychiatric episode and were in some intermediate stage of recovery.

Inclusion: 18-50 years old; clinical diagnosis of schizophrenia.

Exclusion: organic brain disorder; recent evidence of alcoholism or clinically significant drug abuse; poor physical health or a major medical illness requiring treatment.

Interventions

Upon referral and admission to the research clinic, all participants entered a 4-8 week stabilisation and evaluation period. When a participant was clinically stable, medication was withdrawn for a 4-week period during which baseline assessments were completed. Antipsychotic treatment was administered during this period only when there was an immediate need for symptom control. At the completion of the drug-free period, participants were then randomly assigned to a two-year treatment course, receiving one of two treatments:

1. Intermittent: targeted medication: medication administered on an 'as-needed basis' to participants who were otherwise drug-free. Antipsychotic treatment was initiated at moderate to high doses when prodromal experiences occurred, and discontinued when the participant returned to a stable clinical state. Participants were treated within context of psychosocial intervention and were assigned to a primary therapist (e.g. psychiatric social worker or masters-level psychologist) for weekly sessions of approximately 45 minutes - mean daily dose 196 mg ± 163 mg (chlorpromazine equivalents), n = 21.

2. Maintenance: continuous medication: minimum daily chlorpromazine-equivalent doses of 300mg* were administered combined with brief visits with a pharmacotherapist on alternate weeks - mean daily dose 720 mg ± 732 mg (chlorpromazine equivalents), n = 21.

Outcomes

Hospitalisation - by two years.

Leaving the study early - by two years.

Global state: BPRS; GAS - by two years.

General functioning: LOFS - by two years.

Quality of life: QLS - by two years.

Unable to use -

Global state: BPRS; GAS - taken at intervals of six months (no data reported for intervals).

General functioning: LFS - taken at intervals of six months (no data reported).

Quality of life: QLS - taken at intervals of six months (no data reported for intervals).

Notes

*Adjustments in dose above 300 mg were made on the basis of psychotic symptom manifestation and side effect information - medication was increased if prodromal symptoms of relapse were detected.

Participants who failed to complete the drug-free period after two attempts were admitted to the experimental phase while taking medication. During the study, participants who decompensated and required inpatient treatment were admitted to their local hospital until they were again candidates for outpatient treatment. At discharge, they continued the study provided they had been hospitalised for fewer than six months.

Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskRandomised - 'at the completion of the drug-free trial, the patients were randomly assigned using a stratification process involving age, sex and prognostic status'.
Allocation concealment (selection bias)Unclear risk'Randomly assigned' - no further details.
Blinding (performance bias and detection bias)
All outcomes
High riskNo blinding - 'open comparative trial'.
Incomplete outcome data (attrition bias)
All outcomes
Low riskFollow-up: 62% - drop outs were n = 9 continuous medication group (one refused assignment to regimen; seven discontinued treatment; one was non-compliant), n = 7 targeted medication group (two were hospitalised; two were dropped while hospitalised; three dropped out during clinical stability while medication-free) over the two-year period.
Selective reporting (reporting bias)Unclear riskCriterion instruments were completed at six-month intervals - however, only results by the full two years are available.
Other biasUnclear risk

Funding source: supported in part by NIMH grant MH-35996.

Rating scales: 'criterion instruments were completed at six-month intervals by an evaluator who was independent of the treatment team and whose sole function during the experimental treatment phase was to make independent assessments of outcome'. Trial author developed two of the psychosocial measure scales, including the Level of Functioning Scale and the QLS.

Carpenter 1990*

MethodsAllocation: randomised.
Blinding: single.
Duration: 2 years.
Setting: outpatients, Walnut Street Clinic, Maryland Psychiatric Research Center, Maryland, US.
Design: parallel.
Participants

Diagnosis: schizophrenia (RDC).
N = 116.

Age: mean 28.1 years.
Sex: 71 M; 45 F.

Racial origin: n = 73 Black; n = 43 White.
Consent: written consent form required.

History: all participants had recently experienced a psychiatric episode and were in some intermediate stage of recovery.

Inclusion: not stated.

Exclusion: organic brain disorder; recent evidence of alcoholism or clinically significant drug abuse; poor physical health; or a major medical illness requiring treatment.

Interventions

Upon referral and admission to the research clinic, all participants entered a 4 to 8 week stabilisation and evaluation period. When a participant was clinically stable, medication was withdrawn for a 4-week period during which baseline assessments were completed.

1. Intermittent: targeted administration of medication: medication administered on an 'as-needed basis' to participants who were otherwise drug-free. Participants remained drug-free until symptoms appeared that were suggestive of a prodromal phase of a psychotic episode, mean daily dose 173 mg ± 69 mg chlorpromazine/4.4 mg ± 1.1 mg haloperidol equivalents, n = 57.

2. Maintenance: continuous administration of medication: continued receiving medication - dose levels were raised when prodromal symptoms appeared and lowered when re-stabilisation occurred, mean daily dose of 433 mg ± 46 mg chlorpromazine/11.8 mg ± 4.4 mg haloperidol equivalents, n = 59.

Each group were further assigned to receive psychosocial treatment, involving individual therapy as well as involvement with the family or significant others (this was the same approach as in Carpenter 1987).

Outcomes

Hospitalisation - over two years.

Leaving the study early - by two years.

Unable to use -

Global state: BPRS, GAS - taken at intervals of six months (no data reported).

Quality of life: QLS - taken at intervals of six months (no data reported).

General functioning: LFS (no SD reported).

Notes

A variety of antipsychotics were administered: mean daily dose of haloperidol for targeted-medication group = 4.4 mg ± 1.1 mg; mean daily dose of haloperidol for continuous-medication group = 11.8 mg ± 4.4 mg.

Mean daily dose of chlorpromazine for targeted-medication group = 173 mg ± 69 mg; mean daily dose of chlorpromazine for continuous-medication group = 433 mg ± 46 mg.

Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskRandomised - 'at the completion of the drug-free trial, the patients were randomly assigned within age, sex and prognostic categories'.
Allocation concealment (selection bias)Unclear risk'Randomly assigned' - no further details.
Blinding (performance bias and detection bias)
All outcomes
Unclear riskSingle (assessor) blind.
Incomplete outcome data (attrition bias)
All outcomes
High riskFollow-up: 66% - n = 29 in targeted treatment group (19 were treatment-related drop-outs including refusal to continue treatment; n = 10 were non-treatment-related drop-outs); n = 11 in continuous treatment group (8 were treatment-related; 3 were non treatment-related). ITT used.
Selective reporting (reporting bias)High riskNot all outcomes reported - including data from BPRS, GAS, QLS.
Other biasLow risk

Funding source: supported in part by NIMH grant MH-35996 and MH-40279.

Rating scales: independent rater - 'to minimize bias in the evaluation of outcome, assessments were completed by an evaluator blind to the treatment administered in any given case'.

Gaebel 2011

MethodsAllocation: randomised.
Blinding: open.
Duration: 1 year.
Setting: outpatients, 8 German psychiatric university hospitals.
Design: 2x2.
Participants

Diagnosis: schizophrenia (ICD-10).
N = 44.

Age: 18-56, mean 33.1 years.
Sex: 25 M; 19 F.

Racial origin: n = 44 White.
Consent: written informed consent required.

History: first episode schizophrenia - defined in the study as 'the first inpatient treatment of the respective symptoms and no prior treatment with antipsychotics'.

Inclusion: had just completed one year of antipsychotic maintenance treatment after the first episode (defined as the first inpatient treatment of the respective symptoms and no prior treatment with antipsychotics) of schizophrenia (diagnosed according to ICD-10); were sufficiently stable (i.e. had no relapse in the first postacute year and did not fulfil any criteria of the decision algorithm for early intervention at the beginning of the second year; additional clinical assessment was performed by the treating psychiatrist; had been sufficiently compliant in keeping the assigned biweekly appointments; were aged between 18 to 56 years; sufficiently proficient in German; gave written informed consent after receiving detailed information about the study.

Exclusion: pregnancy; contraindication for antipsychotic treatment; mental retardation; organic brain disease; substance dependence; suicidal behaviour in previous history; serious physical disease; and participation in other incompatible traits.

Interventions

After an eight-week acute study and one year of double-blind antipsychotic maintenance treatment with low-dose haloperidol or risperidone (1 mg to 10 mg daily) plus either 8 weeks of psychoeducation vs 1 year CBT, stable first-episode participants were randomly assigned to 12 months of either:

1. Intermittent: respective antipsychotic treatment completely removed in a stepwise fashion over a period of three months (at the most) and restarted at occurrence of prodromal symptoms - mean daily dose of 1 mg/day haloperidol equivalent (the mean dose covers an initial maintenance phase of about 3-4 weeks (mean daily dose 2.9 mg ± 1.6 mg/day), a phase of about 10 weeks in which antipsychotics were tapered off (mean daily dose 1.7 mg ± 1.2 mg/day), a phase of six months where antipsychotics were withdrawn completely (0 mg ± 0 mg/d) and a 2-week phase in which drug treatment was restarted (4.5 mg ± 3.7 mg/day)), n = 21.

2. Maintenance: maintaining drug regimen from the end of the first year of treatment for the whole second year - mean daily dose of 3.1 mg ± 1.7 mg/day haloperidol equivalent, n = 23.

Participants received low dose atypical or typical antipsychotics throughout the course of the study - by the end of the trial, people in the maintenance group were receiving: risperidone (mean dose 2.2 mg ± 1.0 mg; n = 9); haloperidol (mean dose 2.7 mg ± 1.0 mg; n = 8); olanzapine (mean dose 11.8 mg ± 7.6 mg; n = 3); quetiapine (mean dose 495.1 mg ± 82.4 mg; n = 3). People in the intermittent treatment, when tapering off, received: risperidone (mean dose 1.5 mg ± 0.8 mg; n = 7); haloperidol (mean dose 1.3 mg ± 0.4 mg; n = 7); aripiprazole (mean dose 30.0 mg); n = 1); clozapine (mean dose 84.3 mg; n = 1); olanzapine (mean dose 4.5 mg ± 1.5 mg; n = 3); risperidone (open) (mean dose 1.0 mg; n = 1); ziprasidone (mean dose 38.7 mg ± 28.2 mg; n = 2).

In case of prodromal symptoms of an impending relapse, participants in both intervention groups received early drug intervention (MT = increased; IT = restarted), guided by a 'decision algorithm'.

Outcomes

Global state: relapse: an increase in PANSS positive score > 10, CGI change score ≥ 6 and a decrease in GAF score > 20 between two visits - by one year.

Global state: PANSS; SANS; CGI-S - by one year.

Leaving the study early - by one year.

Level of functioning: GAF - by one year.

Adverse effects: akathisia (HAS); extrapyramidal symptoms (EPS); side effects (UKU); tardive dyskinesia (AIMS) - by one year.

Quality of life: LQLP - by one year.

Unable to use-

Data from first year of study: concerned maintenance treatment - risperidone vs low-dose haloperidol (no intermittent treatment examined).

Attitude towards drugs: Drug Attitude Inventory (DAI) (not considered in this review).

Compliance: Compliance Rating Scale (CRS) (not considered in this review).

Depression ratings: Calgary Depression Rating Scale (CDSS) and Hamilton Depression Rating Scale (HDRS) (not considered in this review).

Stressful life events: Munich Event List (MEL) (not considered in this review).

'Marked clinical deterioration': defined as fulfilment of one of the three single relapse criteria or increase in PANSS positive score ≥ 7 with decrease in GAF score >15 between two visits (added post-hoc as a further outcome measure due to low prevalence of relapse).

Deterioration: according to Csernansky 2002 - defined as an increase (from start of the second year of treatment) in the sum of the PANSS positive and negative scores ≥ 25% or ≥ 10 points or a CGI-change score ≥ 6 (added post-hoc as a further outcome measure due to low prevalence of relapse).

Notes

*This study also had a 'secondary objective' of early drug intervention with antipsychotics vs benzodiazepines; the MT and IT groups were further split into subgroups (half of each group receiving either antipsychotics or benzodiazepines) but data were not reported individually, making a valuable and meaningful assessment of intermittent antipsychotic treatment difficult. The two results that have been presented (relapse and leaving the study early) represent the total group numbers.

**This study was subject to a sensitivity analysis: there was no substantive difference in the results when these data were included and they remain within the analysis.

Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskRandomised - 'openly randomly assigned' - randomisation for the 8-week initial treatment phase; randomisation for the first-year treatment phase; randomisation for the second year treatment phase; randomisation for 'lateral entry' participants who had originally dropped-out of the first year of the study and subsequently re-recruited to complete the second-year treatment phase. The randomisation for the second year (maintenance treatment vs intermittent treatment; for each: early intervention with the pre-given antipsychotic vs a benzodiazepine --> 4 conditions with a ratio of  1:1:1:1) was conducted by an independent institution in a different city (Coordinating Center for Clinical Trials of the Mainz University) by an approved biostatistician/ mathematician. Block randomisation was conducted [regarding study site and pre-treatment (risperidone/haloperidol)]'.
Allocation concealment (selection bias)Unclear riskUnclear - it is stated that participants in the first year of treatment were randomly assigned under double-blind conditions, but participants in the second year of treatment were 'openly randomly assigned to either continued double-blind' maintenance or intermittent treatment.
Blinding (performance bias and detection bias)
All outcomes
High riskUnclear - it is stated that treatment procedures for the primary hypothesis were conducted in an open manner for all participants; however, a double blind condition was seemingly employed for participants who completed the first year of the study and for 'lateral entry' participants treatment was open (randomisation and blinding procedures for the second year of study were provided by the Johann-Gutenberg University, Mainz, Germany).
Incomplete outcome data (attrition bias)
All outcomes
Unclear risk

First year of maintenance study: n = 159 - of these, n = 96 passed the first year and were assessed for eligibility; n = 37 (40%) were not eligible for inclusion in the second year of study (n =14 insufficient clinical stability, n = 8 unreliable attendance, n = 6 no consent, n = 6 wanted to remain on antipsychotic treatment, n = 1 wanted to discontinue antipsychotics treatment, n = 1 changed residence, n = 1 'lateral entry' not provided by study site).

Second year of study - n = 59 were included and further randomly allocated: however, n = 15 dropped out immediately after random assignment (in maintenance treatment, n = 6 withdrew consent: in intermittent treatment, n = 9 were not treated according to random allocation). Total n = 44 treated according to randomisation and included in ITT analysis (LOCF).

Follow-up: (for 1 year of IT vs MT) 61%.

Selective reporting (reporting bias)Unclear riskIn the first year of study, participants from 13 German psychiatric university hospitals were included; in the second year, however, participants from only 8 German psychiatric university hospitals were mentioned - 'there have been no formal or administrative reasons why 5 sites did not include patients in the second year'.
Other biasHigh risk

The original design was amended by addition of a 'lateral entry procedure' to allow first-episode participants to enter the second year after the initial first year of maintenance treatment'. The majority of the 'lateral entry' participants were patients who had participated in the first year of study but dropped-out for various reasons, then switched to another atypical antipsychotic at the discretion of the doctor and continued to attend the biweekly appointments.

Funding source: conducted within the framework of the German Research Network on Schizophrenia, funded by the German Federal Ministry for Education and Research BMBF (grants 01 GI 9932 and 01 GI 0232) risperidone and haloperidol were provided by Janssen Cilag, Germany. Lorazepam was provided by Wyeth-Pharma.

Rating scales: conducted by a study physician at every visit.

Herz 1991*

MethodsAllocation: randomised.
Blinding: double-blind.
Duration: 2 years.
Setting: stable schizophrenic outpatients, Buffalo, New York and New Haven, Conneticut (US).
Design: parallel.
Participants

Diagnosis: schizophrenia (RDC; DSM-III).
N = 101.

Age: 19 to 60 (mean 36 years old).
Sex: 54 M; 47 F.

Racial origin: n = 45 Black; n = 56 White.
Consent: informed consent required.

History: Research and Diagnostic Criteria of schizophrenia or schizoaffective depression; n = 62 with less than three (< 3) previous hospitalisations, n = 39 with three or more (> 3) previous hospitalisations.

Inclusion: Research and Diagnostic Criteria of schizophrenia or schizoaffective depression based on the Schedule for Affective Disorders and Schizphrenia Lifetime Version interview; at least two years admission in a psychiatric hospital or DSM-III schizophrenia diagnosis; currently receiving chlorpromazine, fluphenazine, haloperidol, trifluoperazine as maintenance therapy; in a stable state for at least 3 months; a significant other willing to cooperate.

Exclusion: organic mental disorder; uncooperativeness with treatment in the past; hallucinations and delusions that interfere with functioning; Research and Diagnostic Criteria of alcohol and drug dependence in the past two years; serious suicidal/ assaultive behaviour in past two years.

Interventions

Stage one: eight-week 'washout period'; last two weeks are drug-free; drug withdrawal of oral medication was 25% dose reduction every two weeks; depot fluphenazine was 33% dose reduction every two weeks. Those who showed prodromal symptoms were excluded, only those who passed stage one entered stage two and were randomly assigned to one of two double-blind treatment groups.

1. Intermittent: received placebo injection, mean daily dose 149.7 mg ± 179.3 mg chlorpromazine equivalents (average cumulative antipsychotic drug dosage during prodromal episodes over two years = 487.19 mg ± 370.68 mg), n = 50.
2. Maintenance: usual dose of antipsychotic medication mean daily dose 290.0 mg ± 146.7 mg chlorpromazine equivalents (average cumulative antipsychotic drug dosage during prodromal episodes over two years = 424.84 mg ± 333.05 mg), n = 51.

Participants in stage two were expected to attend weekly supportive group therapy sessions, to allow for weekly monitoring. Open family group meetings were held monthly for family members who could attend on a regular or irregular basis according to their own needs.

When a participant was considered to be experiencing a prodromal* episode, administration of study medication was stopped and the active form of the participant's study antipsychotic medication was given openly. The dosage was given at the discretion of the psychiatrist, depending on severity of symptoms, but it was frequently twice the maintenance or baseline level. After an episode was declared, participants were seen every other day for continuing evaluation and treatment. Once clinically stabilised and considered in remission for two weeks, active medication was decreased while double-blind study medication (placebo or active) was simultaneously increased over a two-week period.

Outcomes

Global state: relapse: defined as an increase in any problem appraisal scale PAS; ≤ 30 GAS; more than two days' duration of episode or consensus judgement of the clinical team where the above criteria is not met - by 24 months.

Hospitalisation - by 24 months.

Death - by 24 months.

Global state: GAS - by 6 months and 24 months .

Global state: prodromal episodes (worsening of symptoms of at least a days' duration; increase in any PAS; non psychotic and dysphoric symptoms + mild exacerbation of schizophrenia symptoms) - by 24 months .

Service outcomes: time to hospitalisation - by 6 months and 24 months.

General functioning: PAS social component average score - by 6 months and 24 months.

Leaving the study early - by 24 months.

Unable to use -

Global state: GAS (attrition rates were more than 50%).

Adverse effects: AIMS (no data reported).

Notes*Symptoms were considered prodromal if there was an increase in any Problem Assessment Scale (PAS) role-functioning item or nonpsychotic symptom from a baseline of none, slight or mild to moderate or marked, or from a baseline of moderate to marked.
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskRandomised - 'when a patient was deemed acceptable for randomisation, the coordinator called the statistician who consulted a table for randomisation. The table was in blocks in an attempt to have equal numbers of patients in each treatment group.There was no stratification used in the randomisation.'
Allocation concealment (selection bias)Unclear riskNot described.
Blinding (performance bias and detection bias)
All outcomes
High riskDouble blind - 'clinicians and research evaluators were blind regarding patients' treatment status'. However, when prodromal signs identified, antipsychotic medication was given openly.
Incomplete outcome data (attrition bias)
All outcomes
High riskFollow-up: 55% - protocol terminations: n = 7 participants from the maintenance group were dropped (n = 5 because their episodes lasted more than nine weeks, and n = 2 participants because they experienced three episodes in one year). n = 23 participants in the intermittent group were dropped (n = 10 participants because episodes lasted more than nine weeks and n = 13 participants because they had three episodes in the course of one year). A further n = 7 participants in the maintenance group an n = 8 from intermittent were dropped because of non-compliance or withdrawal from the study (of these n = 2 died of natural causes in the maintenance group, and n = 1 died in the intermittent group, classified as a 'possible suicide', who drowned after ingesting large amounts of alcohol). ITT used.
Selective reporting (reporting bias)High riskNot all expected outcomes reported - adverse effects using AIMS was mentioned but no data were reported.
Other biasUnclear risk

Funding: supported by grant MH37343 from the National Institute of Mental Health, Bethesda, Maryland.

Rating scales: unclear who administered the rating scales.

Jolley 1989/1990

Methods

Allocation: randomised.

Blinding: double.

Duration: 2 years.

Setting: psychiatric outpatients, Charing Cross Hospital, London, UK.

Design: parallel.

Participants

Diagnosis: schizophrenia (DSM-III) - in remission.
N = 54.

Age: mean 41 years.
Sex: 23 M; 31 F.

Racial origin: not stated.
Consent: not stated.

History: all participants had been free of florid symptoms for at least six months and had been stabilised for at least two months on a fixed dose of depot fluphenazine decanoate.

Inclusion: schizophrenia (DSM-III), all participants were required to be free of florid symptoms for at least six months and be stabilised for at least two months on a fixed dose of depot fluphenazine decanoate.

Exclusion: participants in whom relapse had entailed a definite risk to self or others in the past were excluded.

Interventions

1. Intermittent: equivalent doses of placebo injections, with brief intermittent course of antipsychotics begun at the earliest sign of relapse (prodromal symptoms - appearance of non-psychotic symptoms), average cumulative antipsychotic drug dosage (haloperidol equivalents) over two years = 298 mg ± 249 mg, n = 27.
2. Maintenance: continued to receive fluphenazine decanoate in clinically optimal (pre-trial) doses, average cumulative antipsychotic drug dosage (mg haloperidol equivalents) over two years = 1616 mg ± 598 mg, n = 27.

Haloperidol (5 mg to -10 mg) was given to patients in both the intermittent and control groups who developed prodromal symptoms or relapse. Treatment of prodromal symptoms continued for up to two weeks unless relapse occurred and treatment for relapse continued until four weeks after.

Outcomes

Global state: relapse: defined as the re-emergence of florid symptoms (delusions, hallucinations, bizarre behaviour, or thought disorder) or a deterioration of symptoms sufficient to warrant hospitalisation - by one and two years.

Hospitalisation - by one year.

Death - by two years.
Global state: prodromal symptoms: defined on a clinical basis - emergence of neurotic or dysphoric symptoms lasting for two days or more and causing distress to the patient - by one and two years.

Adverse effects: EPS, specific (including hypomimia; tremor; rigidity; gait abnormality; akathisia; global non-liveliness, global parkinsonism) - by six months, one and two years.
Leaving the study early.

Unable to use -

Global state: Early Signs Questionnaire (no data reported); GAS (no data reported); Manchester scale (no data reported); Symptom Checklist-90 (no SD reported).

Notes 
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear risk'Randomised' - no further description.
Allocation concealment (selection bias)High risk'Participating clinicians were requested to refer patients whom they thought might benefit from the brief intermittent treatment approach.'
Blinding (performance bias and detection bias)
All outcomes
Unclear riskDouble blind - 'placebo injections were substituted under double-blind conditions'.
Incomplete outcome data (attrition bias)
All outcomes
Unclear riskFollow-up: 91% - 'of the 49 patients for whom two year follow-up data were available, 19 had been withdrawn from double blind treatment before the end of the two year study period. Thirteen of these were re-established on depot antipsychotic drugs, and two patients in the intermittent treatment group and four controls received no further antipsychotics'. Of the five participants not followed up, n = 2 died in the control group (one suggested suicide, and one attributable to an acute physical illness, and the participant was withdrawn from the study at the time of diagnosis), n = 2 disappeared in the intermittent group, and n = 1 refused all attempts at follow-up.
Selective reporting (reporting bias)High riskAIMS; Early Signs Questionnaire; GAS; 'Manchester Rating Scale' and 'Symptom Checklist-90' were both mentioned as instruments used to measure psychotic symptoms at one year, but no data reported at either one or two years. Data reported at one year differs to data reported at two years.
Other biasUnclear risk

Funding: funded by grants from the Department of Health and Social Services, North West Thames Regional Research Fund, and the Priory Hospital, Roehampton. Depot fluphenazine decanoate and placebo were supplied by ER Squibb and Sons.

Rating scales: made during interviews with the participants, conducted by the community psychiatric nurse during home visits.

McCreadie 1980

Methods

Allocation: randomised.

Blinding: double.

Duration: 9 months.

Setting: inpatients, hostel ward of the rehabilitation unit of Gartnavel Royal Hospital, Glasgow (UK) and day patients attending the hospital's industrial and occupational therapy departments.

Design: parallel.

Participants

Diagnosis: 'definite schizophrenia' (Feighner 1972 criteria).
N = 34.

Age: mean 51 years.
Sex: 34 M.

Racial origin: not stated.
Consent: consent required from participant or next of kin.

History: n = 9 inpatients (average length of illness as estimated from first hospital admission 26 years (mean)) and n = 26 day patients (average length of illness 18 years (mean)).

Inclusion: physically fit; fulfilled Feighner criteria for 'definite schizophrenia'; in the opinion of both medical and nursing staff was well-controlled on and benefiting from long-acting intramuscular antipsychotic medication; consent from participant/ next of kin.

Exclusion: not stated.

Interventions

All participants were switched to fluphenazine decanoate (if not already receiving) at least three months before the trial began.

1. Intermittent: intermittent pimozide: mean dose 8 mg/oral, maximum 32 mg every 4 days/week, n = 16.

2. Maintenance: continued fluphenazine decanoate: mean dose 12.5 mg/IM, maximum 50 mg weekly, n = 18.

To ensure double-blind conditions, participants received active fluphenazine injections and placebo pimozide tablets, or placebo fluphenazine injections and active pimozide tablets.

Outcomes

Global state: relapse: defined as exacerbation of positive symptoms - by 9 months.

Adverse effects: dyskinesia, need for additional antiparkinsonian medication - by 9 months.

Leaving the study early - by 9 months.

Unable to use -

Mental state: Hamilton-Lorr Scale, Kraweicka Scale (depression and anxiety) (outcome reported but no usable data presented).

Behaviour: Wing Ward Behaviour Scale (independent behaviour) (outcome reported but no usable data presented).

NotesNumber of participants differ in the abstract and the full paper.
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskRandomised - no further description.
Allocation concealment (selection bias)Unclear risk'Blindly allocated' - no further description.
Blinding (performance bias and detection bias)
All outcomes
Low riskDouble blind - 'to ensure double-blind conditions, participants received active fluphenazine injections and placebo pimozide tablets, or placebo fluphenazine injections and active pimozide tablets.'
Incomplete outcome data (attrition bias)
All outcomes
Unclear riskFollow-up: 97% - one participant on pimozide refused all medication from the sixth month and was therefore withdrawn from the trial; he was counted neither as a relapse nor as a non-relapse.
Selective reporting (reporting bias)High riskNot all data reported for listed outcomes.
Other biasUnclear risk

Funding source: not stated - Janssen Pharmaceutical and ER Squibb acknowledged for supplying the medication and providing 'advice'.

Rating scales: participants' mental state was assessed independently by two psychiatrists.

McCreadie 1982

MethodsAllocation: randomised.
Blinding: double.
Duration: 10 months.
Setting: inpatients, hostel wards of Crichton Royal Hospital, Dumfries (UK).
Design: parallel.
Participants

Diagnosis: 'definite schizophrenia' (Feighner 1972 criteria).
N = 28.

Age: mean 55 years.
Sex: 28 M.

Racial origin: not stated.
Consent: consent required from participant or traceable next of kin.

History: average length of illness as estimated from first hospital admission 27 years (mean).

Inclusion: physically fit; fulfilled Feighner criteria for 'definite schizophrenia'; in the opinion of both medical and nursing staff was well-controlled on and benefiting from antipsychotic medication; consent from participant/ next of kin.

Exclusion: not stated.

Interventions

1. Intermittent: intermittent pimozide: once weekly, range 10 mg to 60 mg/oral, mean 40 mg/IM weekly, n = 13.

2. Maintenance: continued fluphenazine decanoate: range 2 mg to 25 mg/IM, mean 14 mg/IM biweekly, n = 15.

Pimozide was administered daily for the first week, four consecutive days weekly for the second, twice weekly for the third and once weekly thereafter on Monday mornings. The once weekly dose was four times the initial daily dose, subject to a maximum of 60 mg. The maximum of weekly dose of fluphenazine (which in most cases was given two-weekly, also on Monday mornings) was 50 mg.

Outcomes

Global state: relapse: defined as exacerbation of positive symptoms - by 10 months.

Adverse effects: tardive dyskinesia - by 10 months.

Leaving the study early - by 10 months.

Unable to use -

Mental state: Hamilton-Lorr Scale, Kraweicka Scale (outcome reported but no usable data presented).

Behaviour: Wing Ward Behaviour Scale (outcome reported but no usable data presented).

Notes

One participant was receiving fluphenazine decanoate and five participants were receiving daily pimozide before the study began; a further five participants were receiving intramuscular flupenthixol decanoate.

If participants were on anti-parkinsonian medication at the start of the trial, the dose of this was steadily reduced and the drug eventually discontinued if sign of parkinsonism were no more than 'mild'. Anti-parkinsonian medication was prescribed if signs were 'moderate' or 'severe'.

Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskRandomised - no further description.
Allocation concealment (selection bias)Unclear risk'Blindly allocated' - no further description.
Blinding (performance bias and detection bias)
All outcomes
Low riskDouble blind - 'to ensure double-blind conditions, participants received active fluphenazine injections and placebo pimozide tablets, or placebo fluphenazine injections and active pimozide tablets'.
Incomplete outcome data (attrition bias)
All outcomes
Low riskFollow-up: 58% - n = 5/13 on pimozide and n = 6/15 on fluphenazine were withdrawn (reasons included: exacerbation of positive symptoms; depression; suicidal ideas and attempts; family request to withdraw; tardive dyskinesia; complaints of tiredness).
Selective reporting (reporting bias)High riskNot all data reported for listed outcomes.
Other biasUnclear risk

Funding source: supported by a research grant from Dumfries and Galloway Health Board. Janssen Pharamceutical and ER Squibb provided 'advice', medication and 'other materials'.

Rating scales: participants' mental state was assessed independently by two psychiatrists.

Number of included participants differ in the abstract (N = 29).

Olson 1962*

MethodsAllocation: randomised.
Blinding: double.
Duration: 6 months.
Setting: hospital inpatients.
Design: longitudinal evaluation.
Participants

Diagnosis: schizophrenia.
N = 90.

Age: mean 51 years.
Sex: not stated.

Racial origin: not stated.
Consent: not stated.

History: not stated.

Inclusion: successfully tranquillized for at least the last 60 days; no significant physical illness exhibited; chronically mentally ill in the same hospital for the last 18 months or more.

Exclusion: not stated.

Interventions

1. Intermittent: drug-nothing (DN): alternated monthly between active drug (one half received phenothiazine and one half received chlorpromazine) and nothing, n = 30.

2. Placebo: drug-placebo (DP): alternated monthly between active drug (one half received phenothiazine and one half received chlorpromazine) and placebo, n = 30.

3. Maintenance: control (C): routine treatment by the ward physician and received medication in standard form (continuous medication in different form), n = 30*.

'The study contained three phases of medication, each followed by a phase of no medication - that is, three cycles, each containing two 30-day phases.'

Outcomes

Leaving the study early - by six months.

Unable to use -

Behaviour: ward behaviour ratings, clinical behaviour ratings (no SD reported).

Notes*Dosages were not specified.
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskRandomised - 'subjects were randomly selected from a roster of those on the continued-treatment service receiving ataractic medication, and were consecutively assigned to one of three groups'.
Allocation concealment (selection bias)Unclear riskNot described.
Blinding (performance bias and detection bias)
All outcomes
Low riskDouble - 'the procedure was blind to nurses, clinical interview rater and patients. The physician was informed of the date but not the direction of change between active drug and placebo to preclude the possibility of continuing high dosages which may have developed during the placebo period.'
Incomplete outcome data (attrition bias)
All outcomes
High riskFollow-up: 52% - attrition described as either 'behavioural' or 'inadvertent': n = 26 in DN group; n = 13 in DP group; n = 4 in the control group.
Selective reporting (reporting bias)High riskNot all reported outcomes presented data.
Other biasUnclear risk

Funding source: not stated.

Rating scales: clinical interview rater used - not stated whether an independent rater.

Pietzcker 1993*

MethodsAllocation: randomised.
Blinding: none - open treatment.
Duration: 2 years.
Setting: outpatients, Germany (multi-centre study): outpatient setting of the five psychiatric centres involved - university centres in Berlin, Düsseldorf, Göttingen and München (treatment observation lasted until September 1989).
Design: parallel.
Participants

Diagnosis: schizophrenia (ICD-9).
N = 364* (recruited between 1983-1987).

Age: 18-55, mean 34.7 years.
Sex: 169 M; 195 F.

Racial origin: not stated.
Consent: informed consent required.

History: average duration of illness: 7.2 years - 3 month stabilisation period.

Inclusion: 18-55 years of age; stable clinical condition for at least 3 months since the last acute episode or inpatient treatment; ICD-9 diagnosis pf schizophrenia

Exclusion: organic brain disease; drug and alcohol abuse; mental retardation; pregnancy; history of suicide attempts and/or serious legal violations during antipsychotic-free intervals; tutelage or treatment guardianship, or recent compulsory hospitalisation.

Interventions

1. Intermittent: antipsychotic crisis intervention (CI): antipsychotic treatment renewed after stabilisation only when relapse occurs, discontinued once stabilised (average cumulative antipsychotic drug dosage (mg chlorpromazine equivalents) over two years = 110 mg), n = 115.
2. Intermittent: prophylactic early intervention (EI): antipsychotic therapy resumed as soon as prodromal symptoms appear, discontinued once stabilised (average cumulative antipsychotic drug dosage (mg chlorpromazine equivalents) over two years = 90 mg), n = 127.

3. Maintenance: prophylactic maintenance medication (type and application of antipsychotic drugs used throughout the study was not restricted) (MT): continuous administration of medication, with doses individually adjusted in accordance with the patient's clinical demands at a given time; minimal dosage of 100mg chlorpromazine equivalents (average cumulative antipsychotic drug dosage over two years = 210 mg), n = 122.

Outcomes

Global state: relapse (definition included three criteria: i) an increase of > 10 in the psychosis factor (HOST, THOT, ACTV) of the BPRS; ii) a decrease of > 20 in the GAS; iii) and deterioration on the CGI scale indicated by a score of > 7. Relapse predicted by prodromal symptoms) - by 2 years.

Hospitalisation - by 2 years.

Leaving the study early - by 2 years.

Unable to use -

AIMS, BPRS, CGI, ESS, GAS: (no data reported for full 2 years).

Negative attitudes toward treatment (not specified in protocol).

Relapse: > 1 year - full participant numbers not available as study was still recruiting (incomplete data).

Rehospitalisation: > 1 year - full participant numbers not available as study was still recruiting (incomplete data).

Notes

*n = 3,910 schizophrenia suspect cases screened, of which n = 3,481 (89%) were excluded during initial recruitment phase due to either administrative reasons, patient refusal, not fulfilling inclusion criteria of age, or fulfilling exclusion criteria. A further n = 65 dropped out because of 'implicit/explicit treatment refusal'.

Criteria for withdrawal from study: patient refusal or unsatisfactory treatment cooperation; occurrence of side effects; intercurrent somatic illnesses; occurrence of a fourth relapse under the assigned treatment strategy; therapeutic objections of the attending physician to treatment continuation, as well as impossibility to withdraw the antipsychotics within six months (according to the schedule) or to maintain a patient without drugs for a minimum of four weeks in the intermittent strategies.

Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskRandomised - 'centrally randomised... using an adaptive randomisation strategy'.
Allocation concealment (selection bias)Unclear riskNot described.
Blinding (performance bias and detection bias)
All outcomes
High riskOpen label - 'a double-blind, placebo-controlled strategy...was not chosen, since early intervention treatment requires the patient's full knowledge of and responsibility for drug treatment and is thus best accomplished under naturalistic, open conditions.' Further, it was stated that 'non-blindness towards treatment strategy may have biased treating psychiatrists to overestimate the base rate of impending relapses, particularly in early intervention treatment, in order to legitimise initiation of treatment'.
Incomplete outcome data (attrition bias)
All outcomes
High risk

Follow-up: 44% (n = 159) at 2 years - n = 205 dropped-out.

Reasons for drop-outs fully addressed, and included: patients not showing up regularly (n = 107); impossible to withdraw antipsychotics within six months (n = 56); occurrence of exclusion criteria (n = 44); therapist refusal to continue treatment (n = 12); impossible to maintain patient withdrawn for >4 weeks (n = 11); patient refusal to continue (n = 10) (Multiple occurrence)

Overall drop-out rate: 42.6% in MT group (n = 52) 59.8% in EI group (n = 76) 67% in CI group (n = 77).

Selective reporting (reporting bias)High riskNot all reported outcomes presented data. AIMS, BPRS, CGI, ESS, GAS all mentioned as instruments of measurement to be taken throughout the course of the two-year study, but only data from the preliminary report (up to six months) were reported.
Other biasUnclear risk

Funding: not stated.

Rating scales: unclear who administered the rating scales.

Prien 1973

MethodsAllocation: randomised.
Blinding: double.
Duration: 16 weeks
Setting: inpatients, 18 Veterans Administration hospitals.
Design: parallel.
Participants

Diagnosis: chronic schizophrenia.
N = 375.

Age: mean 50.9 years.
Sex: 375 M.

Racial origin: not stated.
Consent: not stated.

History: mean duration of current hospitalisation was 10.3 years.

Inclusion: not stated.

Exclusion: organic brain disease; bedfast; had an incapacitating or terminal physical illness.

Interventions

All participants had received stable doses of antipsychotic medication during the six months preceding the study; 48% of participants were receiving chlorpromazine (mean daily dose 462 mg), 46% were receiving thioridazine (mean daily dose 362 mg), 27% were receiving trifluoperazine (mean daily dose 15 mg), and 4% were receiving perphenazine (mean daily dose 28 mg) - 25% of participants were receiving more than one drug.

Participants were assigned to one of five interventions:

1. Intermittent: five-day schedule A: participants received their pre-study dosage Monday through Friday and placebo on Saturday and Sunday.

2. Intermittent: five-day schedule B: participants received their pre-study dosage Monday, Wednesday, Thursday, Friday and Sunday, and placebo on Tuesday and Saturday.

3. Intermittent: four-day schedule A: participants received their pre-study dosage Monday through Thursday and placebo on Friday through Sunday.

4. Intermittent: four-day schedule B: participants received their pre-study dosage Monday, Wednesday, Friday and Sunday, and placebo on Tuesday, Thursday and Saturday.

5. Maintenance: daily schedule: participants continued on their pre-study regimen, n = 73.

Total participants for intermittent schedules: n = 301 (individual group data not reported).

Outcomes

Global state: relapse: defined as 'regression' that required the participant to be returned to hs is pre-study dosage before the end of 16 weeks - 'that decision was usually made jointly by the principal investigator at each hospital and the rest of the study staff' - by 16 weeks.

Unable to use -

Global state: Global Change Scale (GCS); Global Severity of Illness (GSI); BPRS; Nurses Observation Scale for Inpatient Evaluation (NOISE) (no data reported)

Notes

In the four intermittent schedules, there was no increase in daily dosage to compensate for the omission of two or three days of medication - The participant's weekly dosage was therefore reduced by two-sevenths (29%) or three-sevenths (43%) from the pre-study level.

*This study was subject to a sensitivity analysis and was considered only in the comparison ANY INTERMITTENT DRUG TECHNIQUE vs MAINTENANCE THERAPY: there was no substantive difference in the results when these data were included and they remain within the analysis.

Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskRandomised - no further description.
Allocation concealment (selection bias)Unclear riskNot described.
Blinding (performance bias and detection bias)
All outcomes
Unclear riskDouble - 'on drug-free days, placebo tablets identical in appearance were administered in place of the active medication'.
Incomplete outcome data (attrition bias)
All outcomes
High riskFollow-up: unclear - 'approximately 5% of the patients in each treatment group were dropped from the study for reasons other than relapse, such as intercurrent illness, AWOL status or discharge. None of the early terminations were considered to be related to study treatment.'
Selective reporting (reporting bias)High riskNo data reported for rating scales used.
Other biasUnclear risk

Funding: not stated.

Rating scales: administered by physician and nurse.

Remington 2011

MethodsAllocation: randomised.
Blinding: double.
Duration: 6 months.
Setting: outpatients, from the 'Schizophrenia Program', which offers both acute inpatient and ongoing outpatient care to over 4,000 individuals - Centre for Addiction and Mental Health, Toronto, Ontario, Canada.
Design: parallel.
Participants

Diagnosis: schizophrenia (DSM-IV).
N = 35.

Age: mean 39 years.
Sex: 20 M; 15 F.

Racial origin: not stated.
Consent: written informed consent required.

History: stabilised outpatients with a single, oral antipsychotic (excluding clozapine and quetiapine as the authors state 'there is theoretical and anecdotal evidence that abrupt discontinuation of these drugs may result in an increased risk of symptom exacerbation').

Inclusion: DSM-IV diagnosis of schizophrenia based on clinical interview, collaborative history, and chart review; capacity to provide written, informed consent; stabilised as outpatients with a single, oral antipsychotic (with the exception of clozapine and quetiapine) ≥ 3 months; evidence of adherence with current anti-psychotic treatment.

Exclusion: exposure to a depot antipsychotic ≥ 1 year; current diagnosis of substance abuse according to DSM-IV criteria; participants receiving clozapine and quetiapine.

Interventions

1. Intermittent: "same daily dose administered every other day" (n = 6 risperidone, n = 11 olanzapine), n = 17.

2. Maintenance: "treatment as usual" (n = 8 risperidone, n = 8 olanzapine, n = 2 loxapine), n = 18.

Dosages not specified.

Outcomes

Global state: relapse (20% increase in overall symptoms using BPRS) - by six months.

CGI-S - by six months.

Re-hospitalisation - by six months.

Side effects: (AIMS) tardive dyskinesia - by six months.

Leaving the study early - by six months.

Unable to use -

BPRS - by six months (no usable data were reported).

Simpson-Angus Rating Scale (SAS) - by six months (no usable data were reported).

Calgary depression scale - by six months (no usable data were reported).

Barnes Akathisia Scale - by six months (no usable data were reported).

Notes 
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskRandomised - 'participants were randomly assigned (1:1 ratio)', no further description.
Allocation concealment (selection bias)Unclear riskNot described.
Blinding (performance bias and detection bias)
All outcomes
Low riskDouble blind - pharmacy.ca (a company specialising in the preparation of drugs for experimental trials - Toronto, Ontario, Canada) was employed to provide, on an individualized basis, all antipsychotics at the appropriate dose and placebo when necessary in matching gelatin capsules.
Incomplete outcome data (attrition bias)
All outcomes
Low riskFollow-up: 74% - a total of n = 38 consented to the study, but three withdrew consent before the first follow-up visit. n = 26 completed the study, with n = 9 'partial completers' (n = 4 withdrew consent; n = 5 were withdrawn by the investigator - reasons included failure to attend scheduled visits and readmission to hospital). ITT used.
Selective reporting (reporting bias)High riskNot all reported outcomes presented data. No data for individual groups were reported for; Symptoms (Simpson-Angus Rating Scale), Calgary depression scale or the Barnes Akathisia Scale.
Other biasHigh risk

Funding source: supported through a National Alliance for Research on Schizophrenia and Depression (NARSAD) Independent Award to Dr. Remington.

Rating scales: not stated who administered rating scales.

Other potential conflicts of interest: Dr Remington has received grant/research support from Novartis and Merck (Germany). Dr Shammi has participated in speaker/ advisory boards for AstraZeneca, Pfizer, Novartis and Bristol-Myers Squibb. In the last three years, DR Kapur has received a grant/research support from AstraZeneca, Bioline, Bristol-Myers Squibb, Eli Lilly, Janssen (Johnson and Johnson), Lundbeck, Otsuka, Organon, Pfizer, Servier, and Solvay Wyeth.

Schooler 1997

MethodsAllocation: randomised.
Blinding: double blind.
Duration: 2 years.
Setting: outpatients, recruited form five centres; from Grady Memorial Hospital, Atlanta (Georgia) and San Francisco (California) (general public hospitals with psychiatric units), Hillside Hospital, New Hyde Park, New York (private psychiatric hospital in a medical centre), Whitney Clinic, New York (University hospital), Eastern Pennsylvania Psychiatric Institute (Philadelphia) (public psychiatric hospital in a medical centre).
Design: 3 X 2 factorial design.
Participants

Diagnosis: DSM-III-R diagnosis of schizophrenia (n = 249) (any subtype), schizoaffective disorder (n = 41) or schizophreniform disorder (n = 23) (79% met this criteria at study entry).
N = 313.

Age: 18-55 years (mean 36.5 years).
Sex: 207 M; 106 F.

Racial origin: n = 124 White.
Consent: informed consent from the participant and at least one family member.

History: 93% of participants were recruited during hospitalisation; the remainder were outpatients experiencing acute exacerbations.

Inclusion: DSM-III-R diagnosis of schizophrenia or schizoaffective disorder; age between 18-55 years; living with, or having more than superficial contact with, family of origin defined as a minimum of 4 hours of regular face-to-face contact per week; living close enough to the clinic to permit home visits; informed consent.

Exclusion: unequivocal liver damage; acute or chronic organic brain syndrome; DSM-III-R diagnosis of psychoactive substance dependence or substance abuse; pregnancy.

Interventions

Following consent, a 16-24-week stabilisation phase was initiated, defined by hospital admission date or by study entry for outpatients. Participants were randomly assigned to 'applied family management' (AFM1) or 'supportive family management' (SFM2) and then discharged, with the goal of stabilization.* Successfully stabilised participants entered the maintenance phase at which time they were randomised to one of three medication conditions:

1. Intermittent: fluphenazine decanoate: targeted, early intervention - an injection of sesame oil or miglioil "vehicle" (placebo) every 2 weeks (if subjects showed prodromal signs of relapse, open-label rescue medication was added; either oral fluphenazine or fluphenazine decanoate), n = 51 (AFM) n = 49 (SFM), total n = 100.

Treatment with combination of medication and family management continued for 24 months.

2. Maintenance: fluphenazine decanoate: continuous standard dose - 12.5-50mg every 2 weeks, n = 52 (AFM) n = 55 (SFM), total n = 107.

3. Maintenance: fluphenazine decanoate: continuous low dose - 2.5-10mg every 2 weeks, n = 54 (AFM) n = 52 (SFM), total n = 106.

Outcomes

Hospitalisation - by two years.

Adverse effects: need for additional medication (fluphenazine decanoate) - by two years.

Unable to use -

Relapse: mean time to psychotic relapse (no SD).

Relapse: mean time to first rescue medication (no SD).

Time to re-hospitalisation (no SD).

Family management parameters (exceeds the scope of the review).

Notes

Benztropine mesylate was the preferred antiparkinsonian mediation and temazepam or lorazepam was used for sleep. Fluphenazine decanoate was used as rescue medication is compliance was a concern.

1AFM: involved individual meetings (first weekly, then bi-weekly, then monthly) in home setting to assess functional status and knowledge of schizophrenia; maximum of 32 sessions; detailed treatment manual and educational material and training in communication skills and problem-solving.

2SFM: involved monthly family group meetings throughout stabilisation and maintenance phases; lasted roughly 1.5 hours; Q&A sessions provided by a clinician.

*Stabilistation criteria included: stable dosage of 12.5 mg to 50 mg fluphenazine decanoate every two weeks for four weeks without use of other antipsychotic or psychotropic medication; stable psychotic symptoms as assessed by BPRS for four weeks; no psychotic symptom (conceptual disorganisation, grandiosity, hallucinatory behaviour and unusual thought content) greater than moderate.

Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskRandomised - an adaptive randomisation algorithm was use in an attempt to maximise equality of randomisation using the following parameters: family management assignment; gender; AIMS score at the end of stabilisation (mild or greater in any body area vs less than mild); diagnosis (DSM-III-R schizophrenia vs schizoaffective disorder vs schizophreniform disorder); length of stabilisation phase (16 weeks vs > 16 weeks).
Allocation concealment (selection bias)Unclear riskNot described.
Blinding (performance bias and detection bias)
All outcomes
High riskDouble blind - blinding was maintained by administering the same volume of medication to all groups - if participants showed prodromal signs of relapse, open-label rescue medication was added. Participants who required rescue medication for a cumulative total of more than 140 days (20 weeks) were discontinued from receiving double blind medication and were treated openly. They continued with their assigned family management condition and received study assessments for the full two year maintenance phase.
Incomplete outcome data (attrition bias)
All outcomes
Low riskFollow-up: 59% - out of n = 528 acutely ill participants recruited at the beginning of the study, n = 215 did not enter the treatment phase. These numbers are accounted for; '105 could not be stabilised according to the study criteria, 21 were stable but only with the treatment above the study dose range (50mg of fluphenazine decanoate every 2 weeks), 80 withdrew consent and 9 were withdrawn for administrative reasons.'
Selective reporting (reporting bias)High riskHillside anchored version of the BPRS, modification of SANS, CGI severity and improvement scale, EPS symptoms - stated that scales will be used but no data were reported.
Other biasUnclear risk

Funding: not stated - all double-blind medication supplies as well as open label fluphenazine decanoate and fluphenazine hydrochloride were provided courtesy of Bristol-Myers Quibb Company, Princeton, New Jersey (US).

Rating scales: unclear who administered rating scales.

Shenoy 1981

MethodsAllocation: randomised.
Blinding: double.
Duration: 6 weeks*.
Setting: outpatients, McGuire VA Medical Center, Richmond, Virginia (US).
Design: parallel.
Participants

Diagnosis: schizophrenia (DSM-III) - average duration of illness mean 13 years.
N = 31.

Age: mean 37 years.
Sex: not stated.

Racial origin: n = 16 Black; n = 12 White.
Consent: written informed consent required.

History: for two years prior to the study, participants had been treated with fluphenazine decanoate, reporting to the clinic every three weeks for an injection. Prior to the two years on fluphenazine decanoate, participants had been treated with a variety of other antipsychotics for varying periods of time.

Inclusion: not stated.

Exclusion: not stated.

Interventions1. Intermittent: six-week drug discontinuation (drug-holiday): received a placebo injection (at the end of the study, participants were returned to their routine active medication), n = 17.
2. Maintenance: continued on regular medication, fluphenazine decanoate mean dose 39.3 mg, n = 14.
Outcomes

Global state: relapse (no definition).
Global state: GAS.

Leaving the study early.

Unable to use -

Adverse effects: AIMS (no SDs/full data reported).

Notes*Six weeks chosen as it is the 'best estimate of that period of discontinuation in which no relapse would occur'.
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskRandomised - no further description.
Allocation concealment (selection bias)Unclear riskNot described.
Blinding (performance bias and detection bias)
All outcomes
Unclear riskDouble blind 'fashion at 3 and 6 weeks' - placebo injection from 'a nurse not involved in the assessment'.
Incomplete outcome data (attrition bias)
All outcomes
Low riskFollow-up: 90% - 'three patients failed to complete the study for failure to appear for appointments'.
Selective reporting (reporting bias)High riskWeissman Social Adjustment Scale mentioned but no data were reported.
Other biasHigh risk

Funding: not stated.

Rating scales (AIMS and 'side effects scale) were completed by the same rater, 'raising the possibility that the correlations are partly due to a halo effect'.

Wiedemann 2001

MethodsAllocation: randomised.
Blinding: open.
Duration: 18 months.
Setting: inpatients, Max Planck Institute of Psychiatry, Munich, Germany.
Design: parallel.
Participants

Diagnosis: ICD-9 diagnosis and RDC criteria for schizophrenia and schizoaffective disorder.
N = 85.

Age: 17-55 years, mean 30 years.
Sex: 31 M; 21 F.

Racial origin: not stated.
Consent: informed consent required.

History: not stated.

Inclusion: ICD-9 diagnosis and RDC criteria for schizophrenia and schizoaffective disorder; age range 17-55 years old; living with partner or in close contact with relative (at least 10 hours a week) for at least three months before admission and likely to return to that household after discharge; living close enough to the clinic to permit at least one home visit and; informed consent.

Exclusion: evidence of organic central nervous system disorder; unequivocal liver damage; mental retardation; ICD-9 diagnosis of psychoactive substance abuse/ dependence; history of more than three relapses per year after the withdrawal of maintenance antipsychotic medication; pregnancy.

Interventions

After hospital discharge, a 'recommended' dose was maintained for three months - participants were then assigned to one of two groups (type of antipsychotic drug was not restricted: dosages converted into milligrams of chlorpromazine equivalents (CPZ)):

1.Intermittent: targeted medication: gradual decrease after three months using a step-by-step discontinuation technique of 50% of antipsychotics every two weeks; where prodromal signs occurred, antipsychotic treatment was reintroduced; when re-stabilisation was attained, pharmacotherapy was again tapered off (mean daily dosage during first year after discharge 161 mg ± 144 mg chlorpromazine equivalent), n = 24.

2. Maintenance: continuous standard dose: the same antipsychotic dose level was maintained throughout the 18-month study period (mean daily dosage during first year after discharge 312 mg ± 152 mg chlorpromazine equivalent), n = 27.

Of the 51 participants completing the study, n = 25 received clozapine (with n = 16 receiving clozapine versus n = 11 receiving typical antipsychotics in the continuous standard dose group; and n = 9 receiving clozapine versus n = 15 on typical antipsychotics in the targeted medication group).

Outcomes

Global state: relapse: defined as reoccurrence of psychotic symptoms with or without subsequent hospitalisation and operationalised following recommendations of Neuchterlein 1986: a rating of 'moderately severe' (= 5) for significant exacerbation or a rating of 'severe' (= 6) or greater fro relapse, representing an increase of at least 2 scale points in any one of the psychosis items of the BPRS - by 18 months.

Global state: BPRS, GAS - by 18 months.

General functioning: SAS - by 18 months.

Adverse effects: tardive dyskinesia - by 18 months.

Leaving the study early - by 18 months.

Unable to use -

Family burden global rating (not considered in this review).

IMPS (no data reported).

EPS; AIMS (no data reported).

Notes

Both interventions were combined with behavioural family treatment (not considered in this review).

Relapsed participants were kept in the study.

Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskRandomised - stratified block randomisation was applied with 'expressed emotion' and sex as strata.
Allocation concealment (selection bias)Unclear riskNot described.
Blinding (performance bias and detection bias)
All outcomes
High riskNon-blind: open label.
Incomplete outcome data (attrition bias)
All outcomes
High riskFollow-up: 55% - from n = 85 originally randomised, n = 25 dropped out before receiving any treatment; n = 9 received a maximum of six treatment sessions (four sessions of communication skills training and two psycho-educational sessions). After treatment, n = 4 dropped-out, the reasons for which are unclear. Not all data accounted for. Completer-only data used (no details as to treatment allocation).
Selective reporting (reporting bias)High riskAIMS, EPS and IMPS scale used but no data reported in the results.
Other biasUnclear risk

Funding: supported by a grant from the German Ministry of Research and Technology.

Rating scales: unclear who administered rating scales.

Wunderink 2007

  1. a

    General:

    CBT - cognitive behavioural therapy.
    EPS - extrapyramidal symptoms.
    IM - intramuscular.
    IV - intravenous.
    ITT - intention-to-treat.
    LOCF - last observation carried forward.
    SD - standard deviation.
    SE - standard error.
    vs - versus.

    Diagnostic tools:

    DSM - Diagnostic and Statistical Manual of Mental Disorders.
    ICD - The International Statistical Classification of Diseases and Related Health Problems.
    RDC - Research Diagnostic Criteria.

    Relevant rating scales:

    Global state
    BPRS - Brief Psychiatric Rating Scale.
    CGI - Clinical Global Impression.
    CGI-S - Clinical Global Impression Severity Scale.
    GAS - Global Assessment Scale.

    Mental state
    BPRS - Brief Psychiatric Rating Scale.
    IMPS - Inpatient Multidimensional Psychiatric Scale.
    NOSIE - Nurses' Observation Scale for Inpatient Evaluation.
    PANSS - Positive and Negative syndrome Scale.
    PRP - Psychotic Reaction Profile.
    SANS - Scale for the Assessment of Negative Symptoms.

    General functioning
    GAF - Gloabal Assessment of Functioning.
    GSDS - Groningen Social Disabilities Schedule.
    LOFS - Level of Functioning Scale.
    PAS - Problem Appraisal Scale.
    SAS - Social Adjustment Scale.

    Adverse Effects
    AIMS - Abnormal Involuntary Movement Scale.
    EPS - Extrapyramidal Side Effects scale.
    HAS - Hillside Akathisia Scale.
    LUNSERS - Liverpool University Neuroleptic Side Effect Rating Scale.
    SAS - Simpson-Angus Scale.
    UKU - Udvalg for Kliniske Undersogelser Side Effects scale.

    Quality of Life
    LQLP - Lancashire Quality of Life Profile.
    QLS - Quality of Life Scale.
    WHOQoL-Bref - World Health Organisation Quality of Life Scale.

MethodsAllocation: randomised.
Blinding: open.
Duration: 2 years.
Setting: outpatients, seven district mental health care centres and the Department of Psychiatry of the University Medical Center Groningen, Netherlands
Design: parallel.
Participants

Diagnosis: first episode schizophrenia or related psychotic disorder (DSM-IV diagnosis of schizophrenia, schizophreniform disorder, brief psychotic disorder, schizoaffective disorder, delusional disorder, or psychotic disorder not otherwise specified).
N = 131.

Age: 18-45 years.
Sex: 89 M; 39 F.

Racial origin: not stated.
Consent: written consent required.

History: participants received no prior antipsychotic medication for more than three months; spoke Dutch language; had estimated IQ score above 70.

Inclusion: participants had to show response of positive symptoms within six months of antipsychotic treatment and sustained remission during six months.

Exclusion: not stated.

Interventions

1. Intermittent: guided discontinuation: dosage was gradually tapered and discontinued if feasible (tapering guided by symptom severity levels and preference of the participant - if early warning signs of relapse emerged or positive symptoms recurred, clinicians were to restart/ increase the dosage of anti-psychotics - mean daily dose 4.36 mg haloperidol equivalent), n = 65.

2. Maintenance: all participants were continuously treated with low-dose atypical/second-generation antipsychotics (mean daily dose 2.94 mg haloperidol equivalent), n = 63.

By the end of the trial, people in the drug discontinuation group were receiving: risperidone (mean dose 2.4 mg ± 1.5 mg; n = 18); olanzapine (mean dose 2.1 mg ± 1.0 mg; n = 16); quetiapine (mean dose 6.5 mg ± 4.1 mg; n = 4); clozapine (mean dose 6.3 mg ± 2.4 mg; n = 2); zuclopenthixol (mean dose 4.5 mg ± 6.5 mg; n = 3). People in the maintenance treatment group were receiving: risperidone (mean dose 2.9 mg ± 1.7 mg; n = 21); olanzapine (mean dose 2.0 mg ± 1.2 mg; n = 22); quetiapine (mean dose 4.8 mg ± 5.9 mg; n = 7); clozapine (mean dose 3.7 mg ± 1.9 mg; n = 5); zuclopenthixol (mean dose 1.3 mg ± 0.8 mg; n = 3).

Outcomes

Global state: relapse (defined as clinical deterioration during at least one week, having consequences (augmentation of antipsychotic dosage, hospital admission or more frequent consultations) reported by the clinician and subsequently confirmed by PANSS positive subscale item scores assessed by a research team member, of at least one score of five (moderately severe)); PANNS (positive and negative symptoms) - by two years.

Social functioning: Groningen Social Disabilities Schedule (GSDS) - by two years.

Adverse effects: Liverpool University Neurolpetic Side Effect Rating Scale (LUNSERS) - by two years.

Quality of life: World Health Organisation Quality of Life Scale (WHOQoL-Bref) - by two years.

Notes

n = 45 were categorised as alcohol/cannabis dependent/ abusers.

The four most commonly used antipsychotics were all second-generation: risperidone, olanzapine, quetiapine, clozapine.

Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskRandomised - 'carried out by an independent agent, in separate blocks for the seven sites to prevent effects of site. Within these blocks a minimization procedure was applied for gender and age (under or over 25) - raters were blinded for the allotted strategy throughout.'
Allocation concealment (selection bias)Unclear risk'Randomly assigned' - no further description.
Blinding (performance bias and detection bias)
All outcomes
High riskNon-blind - open label.
Incomplete outcome data (attrition bias)
All outcomes
Low riskFollow-up: 98% - of the 131 participants included in the trial, n = 3 from the discontinuation group withdrew consent and were not included in the analyses.
Selective reporting (reporting bias)Low riskAll outcomes reported.
Other biasUnclear risk

Funding: direct funding from the Netherlands Organisation for Health Research and Development (The Hague) (DO945-01-001), Foundation for the Support of the Society for Christian Care of the Nervously and Mentally Ill (Bennekom), Foundation 'De Open Ankh' (Soesterberg) and Eli Lilly Nederland B.V. (Houten).

Rating scales: unclear who administered rating scales.

Characteristics of excluded studies [ordered by study ID]

StudyReason for exclusion
Caffey 1975Allocation: controlled design.
Docherty 2003Allocation: randomised.
Participants: chronic schizophrenia.
Interventions: continuous therapy.
Engelhart 2002Allocation: randomised.
Participants: schizophrenia.
Interventions: continuous therapy vs intermittent therapy vs low exposure therapy.
Outcomes: quality of life, symptoms, hospitalisation (no usable data reported).
Hymowitz 1980Allocation: controlled design.
Levine 1980

Allocation: randomised.

Participants: schizophrenia.

Interventions: abrupt discontinuation of antipsychotics vs continuous fluphenazine decanoate/ hydrochloride (not an intermittent technique).

Newton 1989Allocation: randomised, cross-over design.
Participants: schizophrenia.
Interventions: two-day drug holiday vs continuous medication.
Outcomes: serum haloperidol levels; tardive dyskinesia; mental state (no usable data).
Strauss 1990Allocation: randomised.
Participants: schizophrenia.
Interventions: intermitted from drug therapy ≥ 2 months vs long-term therapy.
Outcomes: cognitive function (no usable data).
Uchida 2008Allocation: controlled design.

Characteristics of ongoing studies [ordered by study ID]

Uchida 2013

Trial name or titleUchida 2013
MethodsAllocation: randomised.
Blinding: single (participant).
Duration: 1 year.
Setting: Japan
Design: parallel.
Participants

Diagnosis: schizophrenia.
Target sample size: n = 60.

Age: (lower age limit) 20 years-old <=.
Sex: male and female.

Inclusion: (1) DSM-IV diagnosis of schizophrenia or schizoaffective disorder, (2) having received a stable dose of risperidone or olanzapine for previous 3 months. 

Exclusion: (1) receiving another antipsychotic drug, (2) history of treatment with long-acting risperidone within 6 months, (3) past history of vagus nerve reflection with a blood test. 

Interventions

1. Intermittent risperidone/ olanzapine: D2 blockade of > 65%. Participants will be treated with an individually prepared regimen and observed for one year.

2. Continuous risperidone/ olanzapine: D2 blockade of > 65%. Participants will be treated with an individually prepared regimen and observed for one year.

Outcomes

Adverse effects: Abnormal Involuntary Movement Scale (AIMS), Simpson-Angus Scale, Barnes Akathisia Rating Scale (BARS).

Global state: Positive and Negative Symptom Scale (PANSS), Global Assessment Functioning Scale (GAF), Japanese version of the Calgary Depression Scale for Schizophrenics (JCDSS), Subjective Well-Being under antipsychotic drug treatment short form Japanese version (SWNS-J), Clinical Global Impression (CGI).

Starting dateAugust 2011.
Contact information

Hiroyuki Uchida

Keio University, School of Medicine

Department of Neuropsychiatry

Shinanomachi 35, Shinjuku-ku, Tokyo, Japan

Email: mcn41320@biglobe.ne.jp

NotesAs of January 2012, participants were in the process of recruitment - with a target completion date of late 2013.

Ancillary