Pharmacological interventions for self-injurious behaviour in adults with intellectual disabilities

  • Review
  • Intervention

Authors


Abstract

Background

Self-injurious behaviour among people with intellectual disability is relatively common and often persistent. Self-injurious behaviour continues to present a challenge to clinicians. It remains poorly understood and difficult to ameliorate despite advances in neurobiology and psychological therapies. There is a strong need for a better evidence base in prescribing and monitoring of drugs in this population, especially since none of the drugs are actually licensed for self-injurious behaviour.

Objectives

To determine clinical effectiveness of pharmacological interventions in management of self-injurious behaviour in adults with intellectual disability.

Search methods

We searched the following databases on 19 February 2012: CENTRAL, MEDLINE, EMBASE, PsycINFO, CINAHL, Science Citation Index, Social Science Citation Index, Conference Proceedings Citation Index - Science, Conference Proceedings Citation Index - Social Science and Humanities, ZETOC and WorldCat. We also searched ClinicalTrials.gov, ICTRP and the reference lists of included trials.

Selection criteria

We included randomised controlled trials that examined drug interventions versus placebo for self-injurious behaviour (SIB) in adults with intellectual disability.

Data collection and analysis

Two review authors independently extracted data and assessed risk of bias for each trial using a data extraction form. We present a narrative summary of the results is presented. We did not consider meta-analysis was appropriate due to differences in study designs, differences between interventions and heterogeneous outcome measures.

Main results

We found five double-blind placebo-controlled trials that met our inclusion criteria. These trials assessed effectiveness and safety of drugs in a total of 50 people with intellectual disability demonstrating SIB. Four trials compared the effects of naltrexone versus placebo and one trial compared clomipramine versus placebo.

One of the naltrexone versus placebo trials reported that naltrexone had clinically significant effects (≥ 33% reduction) on the daily rates of three of the four participants' most severe form of SIB and modest to substantial reductions in SIB for all participants; however, this study did not report on statistical significance. Another trial reported that naltrexone attenuated SIB in all four participants, with 25 mg and 50 mg doses producing a statistically significant decrease in SIB (P value < 0.05). Another trial (eight people) indicated that naltrexone administration was associated with significantly fewer days of high frequency self injury and significantly more days with low frequency self injury. Naltrexone had different effects depending on the form and location of self injury. Another trial with only 26 participants found that neither single-dose (100 mg) nor long-term (50 and 150 mg) naltrexone treatment had any therapeutic effect on SIB.

Comparison of clomipramine versus placebo found no statistically significant benefit for any outcome measure, which included SIB rate and intensity, stereotypy and adverse events. However, it showed clinically significant improvement in the rate and intensity of SIB and stereotypy.

There were very few noteworthy adverse events to report in any of the four trials in which these were reported.

All trials were at high risk of bias, apart from one trial (Lewis 1996), which was probably at low risk of bias. The short period of follow-up was a significant drawback in the design of all five trials, as it did not allow long-term assessment of behaviour over time.

We were unable to examine the efficacy of antidepressants other than clomipramine, antipsychotics, mood stabilisers or beta-blockers as we did not identify any relevant placebo-controlled trials.

Authors' conclusions

There was weak evidence in included trials that any active drug was more effective than placebo for people with intellectual disability demonstrating SIB. Due to sparse data, an absence of power and statistical significance, and high risk of bias for four of the included trials, we are unable to reach any definite conclusions about the relative benefits of naltrexone or clomipramine compared to placebo.

Résumé scientifique

Interventions pharmacologiques pour les comportements d'automutilation chez l'adulte présentant une déficience mentale

Contexte

Chez les personnes présentant une déficience mentale, les comportements d'automutilation sont relativement fréquents et souvent persistants. Le comportement d'automutilation continue de constituer un problème pour les cliniciens. Il reste mal compris et est difficile à améliorer, malgré les progrès de la neurobiologie et des traitements psychologiques. On a fortement besoin d'une base plus solide de données probantes sur la prescription et le suivi des médicaments dans cette population, d'autant plus qu'aucun des médicaments n'est actuellement homologué pour les comportements d'automutilation.

Objectifs

Déterminer l'efficacité clinique des interventions pharmacologiques dans la prise en charge des comportements d'automutilation chez les adultes présentant une déficience mentale.

Stratégie de recherche documentaire

Nous avons effectué une recherche dans les bases de données suivantes le 19 février 2012 : CENTRAL, MEDLINE, EMBASE, PsycINFO, CINAHL, Science Citation Index, Social Science Citation Index, Conference Proceedings Citation Index - Science, Conference Proceedings Citation Index - Social Science and Humanities, ZETOC et WorldCat. Nous avons également cherché dans ClinicalTrials.gov, ICTRP et les références bibliographiques des essais inclus.

Critères de sélection

Nous avons inclus des essais contrôlés randomisés ayant comparé des interventions médicamenteuses à un placebo pour le comportement d'automutilation (CAM) chez les adultes présentant une déficience mentale.

Recueil et analyse des données

Deux auteurs de la revue ont, indépendamment, extrait les données et évalué le risque de biais pour chaque essai au moyen d'un formulaire d'extraction de données. Nous présentons un résumé narratif des résultats. Nous avons estimé que la méta-analyse n'était pas de mise en raison de différences dans les plans d'étude, de différences entre les interventions et de l'hétérogénéité des mesures de résultat.

Résultats principaux

Nous avons identifié cinq essais en double aveugle contrôlés par placebo qui répondaient à nos critères d’inclusion. Ces essais avaient évalué l'efficacité et l'innocuité de médicaments chez un total de 50 personnes souffrant de déficience mentale et présentant des CAM. Quatre essais avaient comparé les effets de la naltrexone à ceux d'un placebo et un essai avait comparé la clomipramine à un placebo.

Un des essais ayant comparé la naltrexone à un placebo avait rapporté que la naltrexone avait eu des effets cliniquement significatifs (≥ 33 % de réduction) sur les taux quotidiens de la forme la plus grave de CAM chez trois des quatre participants et réduit de manière modeste à substantielle les CAM chez tous les participants ; cette étude n'avait toutefois pas rendu compte de la signifiance statistique. Un autre essai avait rapporté que la naltrexone avait atténué les CAM chez les quatre participants, les doses de 25 mg et 50 mg produisant une diminution statistiquement significative des CAM (P <0,05). Un autre essai (huit personnes) avait indiqué que l'administration de naltrexone était associée à nettement moins de jours à fréquence élevée d'auto-mutilation et à significativement plus de jours durant lesquels l'auto-mutilation était peu fréquente. La naltrexone avait eu des effets différents selon la forme et l'emplacement de l'automutilation. Un autre essai avec seulement 26 participants avait constaté que ni le traitement de naltrexone à dose unique (100 mg), ni celui à long terme (50 et 150 mg) n'avaient eu d'effet thérapeutique sur les CAM.

La comparaison de la clomipramine à un placebo n'avait mis en évidence de bénéfice statistiquement significatif pour aucune mesure de résultat, dont le taux et l'intensité des CAM, la stéréotypie et les événements indésirables. Elle avait toutefois montré une amélioration cliniquement significative pour le taux et l'intensité des CAM et la stéréotypie.

Il y avait eu très peu d'effets indésirables notables à signaler dans les quatre essais où ils étaient rapportés.

Tous les essais étaient à risque élevé de biais, à part un essai (Lewis 1996) qui était probablement à faible risque de biais. La courte période de suivi constituait une faiblesse importante dans la conception des cinq essais, car elle n'avait pas permis l'évaluation à long terme du comportement.

Nous n'avons pas été en mesure d'examiner l'efficacité d'antidépresseurs autres que la clomipramine, les antipsychotiques, les stabilisateurs de l'humeur ou les bêta-bloquants car nous n'avons pas identifié d'essais pertinents contrôlés par placebo.

Conclusions des auteurs

Les essais inclus fournissaient de faibles preuves que tout médicament actif avait été plus efficace que le placebo pour les personnes souffrant de déficience mentale et présentant des CAM. Vues la rareté des données, l'absence de puissance et de signifiance statistique, et le risque élevé de biais pour quatre des essais inclus, nous n'avons pas été en mesure de parvenir à des conclusions définitives sur les avantages relatifs de la naltrexone ou de la clomipramine par rapport au placebo.

Plain language summary

Drugs as treatment for self-injurious behaviour in adults with intellectual disabilities

Self-injurious behaviour (SIB) among people with intellectual disability is relatively common and often persistent. It is difficult to treat, and presents challenges to those with caring responsibilities and to clinicians. Several types of drugs have been used by clinicians to help with this problem but none is licensed for self injury.

In this review we aimed to assess if medications used in the management of SIB in adults with intellectual disability are safe and help reduce the behaviours. We looked for studies that compared antidepressants, antipsychotics or mood stabilisers with no active medication (placebo). We found only five studies: four examined the effects of naltrexone, an opioid antagonist (which works through modulating pain perception) and one examined the antidepressant clomipramine. There were only a total of 50 participants included in these studies, which was not enough to determine whether or not the intervention resulted in any improvement. Three of the naltrexone studies and the one clomipramine study indicated the drugs resulted in clinical improvement in SIB, but more evidence is needed.

The data is too limited for us to be able to draw any firm conclusions about the benefits or safety of any medications for SIB in this population. We need more studies where many more participants are involved to know whether medications help in reducing SIB and if they are safe.

Résumé simplifié

Les traitements médicamenteux pour les comportements d'automutilation chez l'adulte souffrant de déficience mentale

Les comportements d'automutilation (CAM) sont relativement fréquents et souvent persistants chez les personnes souffrant de déficience mentale. Ils sont difficiles à traiter et constituent un vrai problème pour ceux qui ont la charge de ces personnes et pour les cliniciens. Plusieurs types de médicaments sont utilisés par les cliniciens pour affronter ce problème, mais aucun n'est homologué pour l'automutilation.

Dans cette revue, nous avons cherché à évaluer si les médicaments utilisés dans la prise en charge des CAM chez les adultes souffrant de déficience mentale sont sûrs et permettent de réduire ces comportements. Nous avons cherché des études ayant comparé des antidépresseurs, des antipsychotiques ou des stabilisateurs de l'humeur à un médicament non-actif (placebo). Nous n'avons trouvé que cinq études : quatre avaient examiné les effets de la naltrexone, un antagoniste des opiacés (qui agit en modulant la perception de la douleur) et un avait examiné l'antidépresseur clomipramine. Il y avait seulement 50 participants inclus au total dans ces études, ce qui n'était pas suffisant pour déterminer si oui ou non l'intervention avait causé une amélioration. Trois des études sur la naltrexone et la seule étude sur la clomipramine indiquaient que les médicaments avaient entraîné une amélioration clinique des CAM, mais on a besoin de davantage de données.

Les données sont trop limitées pour que nous soyons en mesure de tirer des conclusions définitives quant aux bénéfices ou à la sécurité de médicaments pour les CAM dans cette population. Nous avons besoin d'études supplémentaires impliquant beaucoup plus de participants pour savoir si les médicaments sont sûrs et aident à réduire les CAM.

Notes de traduction

Traduit par: French Cochrane Centre 17th May, 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.

Background

Description of the condition

Self-injurious behaviour (SIB) among people with intellectual disability (ID) is relatively common and often persistent (Murphy 1999; Kahng 2002). It is resistant to treatment (Hillery 2007) and presents challenges to those with caring responsibilities. 

A number of terms are used for ID with varying levels of acceptability across disciplines and professions. These terms include 'learning disability', 'intellectual handicap', 'intellectual retardation', 'mental retardation' and 'mental handicap'. The World Health Organization (WHO) International Classification of Diseases (ICD-10) (WHO 1992) uses the term 'mental retardation'. This is defined as a condition of arrested or incomplete development of the mind, characterised by impairment of skills manifested during the developmental period that contribute to the overall level of intelligence, that is, cognitive, language, motor and social abilities. ICD-10 classifies ID on the basis of intelligence quotient (IQ) scores as follows: mild (50 to 70), moderate (35 to 49), severe (20 to 34) and profound (under 20).

Several definitions exist for SIB. The widely known definition by Murphy (Murphy 1985, p15) defines it as: "Any behaviour, initiated by the individual, which directly results in physical harm to that individual. Physical harm includes bruising, lacerations, bleeding, bone fractures and breakages, and other tissue damage." This definition is supported by the definition set out in Diagnostic Criteria for Psychiatric Disorders for Use with Adults with Learning Disabilities/Mental Retardation (DC-LD 2001). More recently it has been defined as "a class of behaviours, often highly repetitive and rhythmic, that result in physical harm to the individual displaying the behaviour" (Fee 1992) and "the deliberate alteration or destruction of body tissue without conscious suicidal intent" (Bhaumik 2005). SIB in those with ID commonly presents as head butting; face slapping and slapping other body parts; hitting body parts against hard surfaces; biting; picking skin/wounds; scratching; pulling hair/nails; poking/sticking things into anus, nose, ears or mouth; and gouging of eyes.

The different terms used for this kind of behaviour include 'self harm', 'self mutilation', 'intentional self harm', 'deliberate self harm' and 'parasuicide'. In this review, we did not include studies where these terms were used for a deliberate suicide attempt.

As there are different terms and definitions, prevalence rates from different studies vary greatly. For example, one study found rates of only 4.9% (Cooper 2009), while another study reported rates of 17.4% (Collacott 1998). Borthwick-Duffy 1994 suggests that 10% to 50% of people with ID might, at some point, display SIB. According to this study, the prevalence in this client group is dependent on where people live, with 8% to 15% of those in hospital exhibiting SIB compared with generally lower rates for those living in other environments, such as special schools (3% to 12%), segregated day centres (3% to 10%) or at home (1% to 4%).

Studies have generally found an inverse relationship between IQ and the likelihood of self injury (Rojahn 2002; Hillery 2007). An epidemiological study of SIB in adults with ID in the UK (Collacott 1998) found both the chronological age and developmental quotient of individuals displaying SIB were lower than those without SIB, and no gender difference was found between those with and without SIB (Rojahn 2002). Several studies reported higher prevalence of autism spectrum disorders in people with ID (Bhaumik 2010; Saemundsen 2010) and autistic symptoms were found to be more common among those with SIB (Collacott 1998). Personality disorders have also been highly correlated with SIB (Horrocks 2003); however, diagnosis of personality disorder in those with ID is complex and difficult (Alexander 2003).

There are a number of contradictory hypotheses on the aetiology of SIB in people with ID. The theories include SIB as a learned behaviour; a form of communication; the result of neurochemical imbalance; a symptom of organic illness; a symptom of intermittent physical discomfort; a symptom of psychiatric illness such as depression, mania, phobias, psychosis, obsessive compulsive disorder, post-traumatic stress disorder, etc.; symptom of grief or trauma or a response caused by over-sensitivity to sensory input (Hillery 2007). SIB may have a self stimulatory function for those who are cared for in environments that lack both appropriate attention-giving and appropriate stimuli (either under- or over-stimulation) (Emerson 1995; Gates 2007). People with ID may display SIB because it results in a 'reward' for the individual, reinforcing the behaviour and increasing the likelihood that it will be repeated (Gates 1996). SIB is more likely to occur in people with specific conditions or syndromes such as Prader-Willi syndrome, Fragile X syndrome, Tourette's syndrome, Smith-Magenis syndrome, Cornelia de Lange syndrome and Lesh-Nyhan syndrome (Murphy 1999).

Some of the other theories about emergence of SIB explain that insecure attachment causing deficient socioemotional adaptation and unstable self with disturbed self regulation of self esteem could be underlying factors and that maintenance could be through operant conditioning (Rojahn 2007).

Consequences of self-injurious behaviour

SIB has physical, psychological and social consequences. There is contradictory evidence on its chronicity in people with ID, with some studies finding high levels of chronicity and others finding low chronicity over follow-up years (Griffin 1986; Murphy 1993; Emerson 2001; Totsika 2008; Cooper 2009). One study showed that 84% of the sample continued to self injure nearly 20 years on, with no significant mean changes in number of topographies or severity of self injuries across the group (Taylor 2010).

Exhibiting SIB may bring short-term positive results for the individual, but in the long term the outcomes are usually negative (Hillery 2007). There can be physical morbidity, for example, sensory deficit, neurological impairment and even death (Mikkelson 1986). There can be auditory and visual sensory damage (Wieseler 1995). It can cause increased stress to carers (Quine 1985), increased risk of suffering physical abuse from carers (Maurice 1982), and increased risk of institutionalisation (Lakin 1983). It has also been suggested that 'vicarious traumatisations' can occur as a result of supporting people who exhibit SIB (Beail 2007). In addition, people who self injure are more likely to be given psychotropic drugs (Aman 1993; Johnny 2009), and more likely to be kept on medication (Chadsey-Rusch 1989).

Description of the intervention

SIB remains poorly understood and difficult to ameliorate despite advances in neurobiology and psychological therapies (Symons 2011). As there are several contradictory hypotheses about the aetiologies as mentioned above, several interventions have been tried over the years including pharmacological and psychological/behavioural interventions. One of the challenges in using behavioural interventions is the dearth of empirical evaluation of these approaches (Emerson 2001a). One systematic review looking at the efficacy of behavioural approaches found mainly single case studies that yielded mixed results and authors highlighted the need for more research on effectiveness of these approaches (Prangnell 2010). Many people with ID and SIB have been prescribed psychotropic medications including opioid antagonists such as naltrexone, antipsychotics, antidepressants, mood stabilisers and beta-blockers. Some argue that psychotropic medication is prescribed too often in this population and there are a range of adverse effects reported (Einfeld 2001).

The scientific literature on the benefits of pharmacological agents for intellectually disabled individuals is beset with a number of problems (Deb 2007). Although no form of treatment has been demonstrated conclusively to be of general benefit, the literature suggests that therapeutic trials with treatments having predominant opiate antagonism (for example, naltrexone, naloxone), dopamine antagonism (for example, antipsychotics), serotonin (5-HT) reuptake inhibition (for example, antidepressants) and beta-blockers may be of value.

As mentioned earlier many psychiatric illnesses (as defined by multiaxial classification in DC-LD 2001), for example, phobias, psychosis, affective disorders, obsessive compulsive disorder and post-traumatic stress disorder can present with SIB as a secondary manifestation (Hillery 2007). Most of the drugs used in SIB are mainstay treatments for these psychiatric illnesses. Appropriate management of primary psychiatric illness may lead to improvement in SIB. Hence, our review does not include SIB that is a direct consequence of any psychiatric illness.

How the intervention might work

Opioid antagonists

Beta-endorphins are endogenous opiate-like substances in the brain, and self injury may increase the production and/or the release of endorphins. As a result, the individual experiences an effect like anaesthesia and, ostensibly, s/he does not feel any pain while engaging in the behaviour (Sandman 1983). Support for this explanation comes from studies in which drugs (opioid antagonists) that block the binding at opiate receptor sites (for example, naltrexone and naloxone) can successfully reduce self injury (Herman 1989). The findings suggest that some clients with SIB may have disturbances of the endogenous opiate system. There is also evidence to support the stress axis playing a significant role in the maintenance of complex episodes of self injury (Kemp 2008). Changes in the hypothalamic-pituitary-adrenal axis after SIB may predict differences in individual response to opiate blockers. Participants with elevated beta-endorphins immediately following SIB were noted to respond better to opioid antagonists (Sandman 2008). Opiate theories of self injury and the possible inter-relationship of self injury with pituitary-adrenal arousal again support the role of opioid antagonists.

Antipsychotics

Both traditional (first-generation antipsychotics) and newer antipsychotics (second-generation antipsychotics) have a similar blocking effect on D2 Dopamine receptors. Impact on additional receptors such as D1 and serotonin receptors add to the distinct profile of the newer drugs. Effect on SIB is most likely due to action via D1 receptor and serotonin receptor.

Clozapine is a neuroleptic known to have an affinity for blocking D1 receptors. Although clozapine is known to affect other neurotransmitter systems such as serotonin, glutamate, GABA (gamma-amino butyric acid), the successful treatment is consistent with the D1 hypothesis of SIB and suggests that clozapine could be an effective intervention in some cases (Hammock 1995). The considerable optimism for clozapine in treating SIB and aggression is because of its dopamine and serotonin receptor binding properties (Breese 1995). It may be that its relative efficacy is related to modulation of serotonin-dopamine interactions primarily in the nigrostriatal and mesolimbic pathways in the basal ganglia (Hammock 2001).

Antidepressants

The serotonergic system is most directly implicated in the pathophysiology of self injury and related behaviours. Antidepressants that act via the serotonergic system appear to be the most promising (Lewis 1996a). The efficacy of serotonin reuptake inhibitor clomipramine in the treatment of SIB was tested in individuals with severe and profound ID. Selective serotonin reuptake inhibitors (SSRIs) are used extensively based on their action upon aggression, depression and impulsivity. These characteristics may be observed in many people suffering from SIB (Yaryura-Tobias 1999).

Mood stabilisers/antimanic drugs

Support is emerging for the use of mood stabilisers such as lithium and carbamazepine (Winchel 1991). An open trial of lamotrigine in the treatment of SIB in an adolescent with profound ID showed a 50% reduction in its frequency as measured by standardised scales. Animal models of SIB suggest that the glutamate neurotransmitter systems involved in the generation of epileptic seizures may also have a role in the pathophysiology of SIB (Davanzo 1996). Adults with ID and aggressive or SIB may respond to divalproex sodium, and this drug is well-tolerated in the majority of participants (Ruedrich 1999). Exact mechanism of action is still unclear for the various mood stabilisers.

Beta-blockers

Central and peripheral mechanisms of actions suggest beta-blockers can be useful agents for management of aggression and self injury. Central mechanisms of action include specific blockade of adrenergic receptors in the central nervous system, membrane stabilisation, non-specific effects on dopamine and serotonin systems (Yaryura-Tobias 1999).

Why it is important to do this review

Most evidence for psychotropic agents is based on a small number of either prospective or retrospective case studies that included small numbers of participants and often used non-validated outcome measures for a short follow-up period. The evidence from single studies seem to be conflicting, with some studies suggesting improvement with atypical antipsychotics (McDonough 2000; Read 2007) and others suggesting no improvement in adults with ID (Ruedrich 2008). There have been some other attempts to appraise and synthesise the evidence for pharmacological interventions for SIB (Matson 2000; Symons 2004) but some of them included children and they did not restrict their inclusion to randomised controlled trials (RCTs) and had a different systematic approach.

Further, there was concern that psychotropic medication was prescribed excessively and inappropriately in this population, and most studies highlighted concerns regarding adverse effects (Einfeld 2001).

There was a strong need for a better evidence base in the prescribing and monitoring of drugs in this population, especially since none of the drugs are actually licensed for SIB. Given the potential challenge that clinicians face and the needs of this group of people, a systematic evaluation of the high-quality evidence for the effectiveness of pharmacological interventions seemed an important and timely undertaking.

Objectives

To determine the clinical effectiveness of pharmacological interventions in the management of SIB in adults with ID.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials (RCTs) including individually randomised, cluster-randomised and cross-over trials.

Types of participants

Adults with ID (mild to profound), aged 18 years or over, presenting with SIB occurring at least during most weeks of the preceding six months (as per diagnostic criteria in DC-LD 2001), and without additional psychiatric illness.

We did not include trials where the term SIB was used for parasuicide.

Types of interventions

Pharmacological interventions including any antidepressants, antipsychotics, mood stabilisers, opiate antagonist (naltrexone), beta-blocker (propranolol) and hypnotic (melatonin), regardless of dosage, against placebo. We excluded trials where there were other treatment controls. We included trials where participants had psychological therapies in the past.

The definition of antipsychotic medication used was that given in the British National Formulary (BNF 2010) section 4.2.1 and 4.2.2, in addition to that described in the individual trials:

  • phenothiazine group - chlorpromazine, fluphenazine, methotrimeprazine, pericyazine, perphenazine, pipotiazine, prochlorperazine, promazine, thioridazine, trifluoperazine;

  • butyrophenone group - benperidol and haloperidol;

  • diphenylbutylpiperidine group - pimozide;

  • thioxanthene group - flupentixol and zuclopenthixol;

  • substituted benzamide group - sulpiride;

  • tricyclic dibenzoxazepine group - loxapine and oxypertine;

  • newer antipsychotic drugs (atypical antipsychotics) - amisulpiride, clozapine, olanzapine, quetiapine, risperidone, aripiprazole, sertindole and zotiepine.

The definition of antidepressant medication used was that given in the BNF 2010 section 4.3.1, 4.3.2, 4.3.3 and 4.3.4, in addition to that described in the individual trials:

  • tricyclic and related antidepressant drugs - amitriptyline hydrochloride, clomipramine hydrochloride, dothiepin hydrochloride, doxepin, imipramine, lofepramine, nortriptyline, trimipramine, mianserin hydrochloride, trazodone hydrochloride;

  • monoamine oxidase inhibitors - phenelzine, isocarboxazid, tranylcypromine, moclobemide;

  • SSRIs - citalopram, escitalopram, fluoxetine, fluvoxamine, paroxetine, sertraline;

  • other antidepressant drugs - agolomelatin, duloxetine, mirtazapine, venlafaxine.

The definition of antimanic drugs used was that given in the BNF 2010 section 4.2.3, including lithium, valproate and carbamazepine.

The definition of opioid antagonists used was that given in the BNF 2010 section 4.10, including naltrexone and naloxone.

Types of outcome measures

Primary outcomes
  • Frequency of SIB

  • Intensity of SIB

  • Duration of SIB

  • Adverse events (effects of medication such as sleepiness, movement disorders, seizures, weight gain, etc.)

 The primary outcomes were measured by client/primary carer reports and validated rating scales.

Secondary outcomes
  • Attendance at day centre/college

  • Engagement in paid/unpaid work/leisure activities

  • Carer stress

Secondary outcome measures were measured by client/primary carer reports and validated rating scales.

Post hoc outcomes

Stereotypy, self restraint and compulsive behaviour were not originally specified in the 'types of outcome measures' when we published the protocol, but they were commonly reported outcomes in some of the included trials and many other studies that we identified. Hence they were post hoc inclusions and are discussed in the Effects of interventions, Discussion, Differences between protocol and review and other relevant sections.

Time points for measuring the outcomes
  • Three to six weeks from the start of intervention

  • Seven weeks to six months from the start of intervention

  • Seven months to one year

  • One to two years

Search methods for identification of studies

We first ran the searches in May 2011, covering all available years for each database. We repeated the searches in February 2012, this time limiting by date to find new studies that had been added to the database since the previous search. No language limits were used.

Electronic searches

We searched the following electronic databases.

  • Cochrane Central Register of Controlled Trials (CENTRAL), part of The Cochrane Library, 2012 (Issue 2), last searched 19 February 2012

  • CENTRAL, searched via The Cochrane Library (DVD), January 2011, searched 13 April 2011

  • Ovid MEDLINE, 1948 to February week 4 2012, last searched 19 February 2012

  • EMBASE (OVID), 1980 to week 7 2012, last searched 19 February 2012

  • PsycINFO (OVID), 1887 to February week 2 2012, last searched 19 February 2012

  • CINAHL (EBSCOhost), 1937 to 19 February 2012

  • Science Citation Index (SCI), 1970 to 17 February 2012, last searched 19 February 2012

  • Social Science Citation Index (SSCI), 1970 to 17 February 2012, last searched 19 February 2012

  • Conference Proceedings Citation Index - Science, 1990 to 17 February 2012, last searched 19 February 2012

  • Conference Proceedings Citation Index - Social Science and Humanities, 1990 to 17 February 2012, last searched 19 February 2012

  • ZETOC (conference search only), last searched 19 February 2012

  • WorldCat (theses search only), last searched 19 February 2012

We used the Cochrane highly sensitive search strategy for identifying randomised trials in MEDLINE (Lefebvre 2008). Appropriate search terms to identify randomised trials were used in the other databases when necessary. The search platform for PsycINFO changed during the review period, and CENTRAL was searched using the DVD version of The Cochrane Library in 2011 because the online version was unavailable. Search histories for all versions of the search histories for each database are reported in Appendix 1.

Searching other resources

The reference lists of the trials included in this review were searched to identify additional trials.

Data collection and analysis

Selection of studies

The titles and abstracts yielded by the search were independently screened by two review authors (SV and AG) against the inclusion criteria. The full text of papers or reports for trials that appeared relevant, or for which more information was needed to determine relevance, were obtained and independently screened by two review authors (FR and AG) to determine whether they met the criteria for inclusion. Disagreements about eligibility were resolved through discussion and, when disagreements were not resolved, a third review author (SV) was used to reconcile any differences. Additional information from the authors of the trials was sought when necessary to resolve questions about the relevance or methodology of a trial. The reasons for excluding trials were recorded. None of the review authors were blinded to the study authors, institutions or the journals of publication of the articles.

Data extraction and management

Two review authors independently extracted data for each trial using a data extraction form.

The following information was extracted or intended to be extracted.

Study information

  • Source (year, volume, pages)

  • Authors

  • Language

  • Type of report

  • Inclusion and exclusion criteria

Characteristics of trial

  • Type of study

  • Ethics committee approval

Characteristics of participants

  • Age

  • Sex

  • Number of participants

  • Degree of ID

  • Stereotypy

  • Current medications

  • Additional mental disorder

  • Additional information

Intervention details

  • Type of intervention

  • Dose and duration

Any deviations from protocol were recorded

Risk of bias in trial (see below)

Duration of follow-up

Outcomes

  • For each outcome: outcome definition; unit of measurement (if relevant)

  • For adjusted estimates: variables adjusted for in analysis

  • Results: number of participants allocated to each intervention group

  • For each outcome of interest: sample size; missing participants

  • Follow-up duration of each outcome

  • Reason for excluding studies

Data on outcomes were extracted as below.

  • For dichotomous outcomes (for example, adverse events), we planned to extract the number of participants in each treatment arm who experienced the outcome of interest and the number of participants assessed at end point, in order to estimate a risk ratio (RR).

  • For continuous outcomes we planned to extract the final value and standard deviation (SD) of the outcome of interest and the number of participants assessed at end point in each treatment arm at the end of follow-up, in order to estimate the mean difference (if trials measured outcomes on the same scale) or standardised mean differences (if trials measured outcomes on different scales) between treatment arms and its standard error.

Where possible, all data extracted were those relevant to an intention-to-treat analysis, in which participants were analysed in groups to which they were assigned.

The time points at which outcomes were collected and reported were noted.

Assessment of risk of bias in included studies

We used The Cochrane Collaboration's tool for assessing risk of bias (Higgins 2008). Two review authors (FR and SV) independently assessed the risk of bias within each included trial based on six domains (see below), with review authors' judging each risk of bias component to be at 'low risk of bias', 'high risk of bias' or 'unclear risk of bias'. Each trial was independently assessed by two review authors. Any disagreements were resolved by discussion and, when necessary, disagreements were reconciled by the third review author (AG). The tool was used to assess the following domains: sequence generation, allocation concealment, blinding (of participants, personnel and outcome assessors), incomplete outcome data, selective reporting and other possible sources of bias (for example, stopping the trial early without reasonable justification and adherence to protocol, changing methods during trial). We recorded the proportion of participants whose outcomes were not reported at the end of the trial.

Measures of treatment effect

No meta-analysis was possible in this review due to the differences in study designs, heterogeneity of interventions and differences in outcome measures. However, if trials can be identified in the future that can be statistically combined, they will be handled in the way that has been described in Appendix 2 (' Methods from the protocol archived for future updates' ).

Unit of analysis issues

See Appendix 2.

Dealing with missing data

Details of dropouts have been included in the review in 'Risk of bias' table under Characteristics of included studies. Reasons and numbers of dropouts have been reported based on available information. As the overall numbers of dropouts were small and unlikely to make significant difference to the summary of the overall review, no efforts were made to contact the authors for information on this. Data about the beneficial effects of single-dose naltrexone was not given in the Willemsen-Swinkels 1995 study. Several attempts were made to contact the authors of this paper but we were not successful. See also Appendix 2.

Assessment of heterogeneity

See Appendix 2.

Assessment of reporting biases

See Appendix 2.

Data synthesis

A narrative summary was conducted (see Effects of interventions) as meta-analysis was not possible due to differences in study designs (duration, cross-over phases with in the studies), heterogeneity of interventions (doses of drugs) and differences in how outcome measures were reported. Therefore it was not relevant to assess heterogeneity between results of trials and we were unable to assess reporting biases using funnel plots or conduct any subgroup analyses or sensitivity analyses. Also see Appendix 2.

Subgroup analysis and investigation of heterogeneity

See Appendix 2.

Sensitivity analysis

See Appendix 2.

Results

Description of studies

See: Characteristics of included studies; Characteristics of excluded studies.

Results of the search

We initially ran the search in May 2011 and identified 4585 citations including trial registers of which only 435 seemed potentially relevant from the screening of titles/abstracts by SV. These 435 records were screened through titles and abstracts again by AG and FR. Articles that did not clearly meet inclusion criteria were excluded at this stage, which resulted in 27 potentially eligible studies. The full texts were retrieved and read by AG, FR and SV. Sixteen of them, including an unpublished thesis, were clearly irrelevant therefore we do not describe them here. Six of the remaining 11 were excluded for the reasons described in the Characteristics of excluded studies table. Five completed RCTs were identified that met our inclusion criteria, which are included in our review and are described in the Characteristics of included studies table.

A further search run in February 2012 produced 553 citations, none of which SV and AG considered relevant.

Figure 1 explains results of the search.

Figure 1.

Study flow diagram

Included studies

The five included trials (Sandman 1990; Thompson 1994; Willemsen-Swinkels 1995; Lewis 1996; Symons 2001) randomised 57 participants but we have included and reported on only 50 participants in our review. The Willemsen-Swinkels 1995 included seven participants with autism but no SIB (seven out of 33 or 21%). We have not included these seven participants in our review. We included the Willemsen-Swinkels 1995 trial as it would have potentially been a waste of useful information and data on pertinent SIB outcomes. This trial included more participants with IDs and SIB than any of the other trials in this review.

Four trials (Sandman 1990; Thompson 1994; Willemsen-Swinkels 1995; Symons 2001) examined the effect of naltrexone (opioid receptor antagonist) versus placebo and one trial (Lewis 1996) examined clomipramine (serotonin reuptake inhibitor), an antidepressant, versus placebo. All trials were conducted in the USA except the Willemsen-Swinkels 1995 trial, which was conducted in the Netherlands.

Design

All were prospective, randomised, double-blinded, placebo-controlled trials and had a cross-over design (see 'Interventions' section below for further details). The trial of Sandman 1990 was the shortest in duration lasting four weeks. Thompson 1994 and Symons 2001 had three and four phases, respectively, with each phase lasting for two weeks. Willemsen-Swinkels 1995 lasted 15 weeks and Lewis 1996 extended over a 19-week period.

Participant characteristics

There were total of 50 participants with SIB in the five included trials. All participants had IDs. We have presented the characteristics of participants in two groups according to the pharmacological interventions used.

a) The four naltrexone trials

There were 42 participants with SIB in the four trials.

Age

The mean age in the trial of Sandman 1990 (24.8 years, SD 1.3, range 23 to 26 years) was lower than in Willemsen-Swinkels 1995 (29 years, SD 6, range 18 to 46 years), Symons 2001 (39.3 years, SD 8.3, range 27 to 45 years) and Thompson 1994 (34.6 years, SD 4, range 29 to 41 years).

Sex

All four participants in the trials of Sandman 1990 and Symons 2001 were male. There were four male and four female participants in Thompson 1994 and there were 27 males and six females in Willemsen-Swinkels 1995, seven of whom did not have SIB.

Degree of intellectual disability

The degree of ID was classified as severe to profound in all trials except in Willemsen-Swinkels 1995 where it ranged from mild to profound.

Concurrent conditions and medication

In Sandman 1990 , three participants were not on any medication while one was on thioridazine (Melaril). In Thompson 1994, one participant was on dilantin and three on carbamazepine for seizure disorders. Two participants in the trial of Symons 2001 had seizure disorder and were maintained on antiepileptics, while two participants were on antipsychotics. In the trial of Willemsen-Swinkels 1995, out of 33 participants, 24 were autistic of whom 17 presented with SIB (included in our review) and seven did not present with SIB (excluded from our review). The remaining nine participants in this trial who did not have diagnosis of autistic disorder presented with SIB. Five had history of epilepsy and were on antiepileptics. Eleven were receiving neuroleptics. Dosages of co-medications remained fixed throughout the study period.

Type of self-injurious behaviour

Forms of SIB in the four trials included head banging, body hitting, head hitting (including eye, nose, throat, ear and face), hand hitting, self biting (including arm and finger biting), self hitting, hair pulling, face pinching and hitting, self rubbing or scratching and rocking. In the trial of Sandman 1990, SIB consisted of kicking and hitting self, hitting against hard surfaces and banging ears with fist. In the trial of Symons 2001, the most common SIBs were head banging, body hitting, head hitting, self hitting and self rubbing. All participants in the Thompson 1994 trial displayed at least three different forms of self injury and had been treated with behavioural and pharmacological treatments, none having produced a sustained reduction of frequency of self injury of 50% or greater. Forms of SIB included hand to head, head to object, poking (nose, throat, eyes), slapping face, biting hands or arms, pinching self banging hands, kicking against objects. In the Willemsen-Swinkels 1995 trial, SIB took the form of face hitting and self scratching.

b) The clomipramine trial

This included eight participants who presented with SIB.

Age

The age of participants ranged between 21 and 39 years.

Sex

Three participants were female and five were male.

Degree of intellectual disability

The degree of ID was classed as severe in two participants and profound in six participants.

Concurrent conditions and medication

None of the participants in the clomipramine trial had a diagnosis of autistic disorder. However, authors acknowledged difficulty in making this diagnosis due to level of mental retardation in the participants. Three participants had seizure disorder but were not receiving anticonvulsant medication and had been seizure-free for at least six months prior to entering the trial. Five participants were receiving antipsychotic medication. Three were maintained on thioridazine (75 to 200 mg), one was maintained on same dose of thiozanthene (19 mg/day) and one on same dose of chlorpromazine (125 mg/day) throughout the study period.

Type of self-injurious behaviour

SIB included hand to face/head hitting, biting hand/wrist, rubbing, hitting fist to chest, pulling hair and kicking legs.

Interventions

There were only two pharmacological interventions used in the five included studies, naltrexone or clomipramine. We have presented the information on interventions in two groups according to the pharmacological agents used.

Naltrexone

Of the four trials examining naltrexone, the Sandman 1990 trial used a double-blind Latin-square design where the four participants received naltrexone tablets at one of these doses- 0, 25, 50 and 100 mg at 8.00 am on Monday and Wednesday only. As fixed doses were administrated, milligrams per kilogram transformation was carried out due to participants differing in weight. Symons 2001 reported that naltrexone dose was based on body weight (1.5 mg/kg) and administered orally at 8.00 am. Each phase lasted for two weeks and each phase was separated by one week. In the trial of Thompson 1994, each participant received 50 mg or 100 mg per day of naltrexone or placebo for a two-week period, with periods arranged in a different order for each subject. Half of the participants were randomly assigned to receive clonidine concurrently with naltrexone and the other half received placebo concurrently with naltrexone. All medications were administered once daily at 8.00 am in the form of three tablets. For the four participants receiving clonidine, placebo was gradually substituted over a six-day period (clonidine gradually reduced to 0.2 mg for three days and 0.1 mg for another three days). In the Willemsen-Swinkels 1995 trial at the beginning of week three, a single dose of naltrexone 100 mg or placebo was administered (33 people). For remainder of that week all participants received a placebo capsule each day. Subsequently, a four-week treatment of naltrexone or placebo was given, followed by a four-week washout period and then a cross-over to the other treatment. In the first four weeks, 19 participants received naltrexone 50 mg/day. As naltrexone 50 mg had no effect, the long-term dose for the second cohort of 14 participants was 150 mg/day.

Clomipramine

The trial of Lewis 1996 had a single-blind placebo condition (weeks one to two), which was followed by a double-blind placebo or clomipramine condition (weeks three to nine). This involved a titration up phase (weeks three to five) where clomipramine was increased by 25 mg every other day to maximum of 3 mg/kg body weight, a maintenance phase (weeks six to nine) during which the dose was 3 mg/kg body weight and a titration down phase (week 10) where clomipramine was reduced by 50 mg every day. Cross-over manipulation involved a second titration up phase (weeks 10 to 12), a second maintenance phase (weeks 13 to 16), a second titration-down phase (week 17), followed by a final single-blind placebo condition (weeks 18 to 19).

Outcomes

All five trials measured and reported data on SIB rate (frequency) but there were some differences in other outcomes that studies focused on. We present outcomes according to the pharmacological agents used.

Naltrexone

In the trial of Sandman 1990, SIB and stereotypy were scored in terms of frequency of occurrence of each behaviour by reviewing videotapes. The participants were videotaped for 10 minutes in their normal surroundings in the mornings and afternoon. Level of activity was rated on a five-point scale ranging from 'no activity' to 'very high activity' to see if effect of naltrexone was purely due to the sedative effect of naltrexone. Differences between placebo and naltrexone were computed when the placebo rate was the highest, during low rate periods and combined morning and afternoon measures. As the participants differed in their base rates during placebo conditions, the data for each participant was normalised for parametric statistical analysis. Normalisation was calculated within participants and the z score transform used in computing t tests of paired observations between treatment conditions. The Connors' Teachers Rating Scale (Conners 1969) was adapted and used to record the activity, aggression and emotional lability. The Vineland Adaptive Behaviour Scale (Sparrow 1984) was used to assess the social, motor, maladaptive, communication and daily living domains and presented as raw scores.

Symons 2001 study measured SIB by directly observing and recording its frequency for one participant and frequency and duration for other three participants. SIB recorded continuously was referred to as a 'bout'. Direct observation was conducted in the participants' daytime setting within groups of both staff and peers during activities of daily living. The effects of naltrexone was deemed to be clinically significant if there was at least a 33% reduction in self injury. The sequential dependency between antecedent staff interaction and self injury was examined for each participant. Yule's Q (transformed odds ratio bound by -1.0 and 1.0 that controls for the total number of coded behaviours) was calculated as an index of sequential dependency for self injury occurring within five seconds of staff prompting for time-based analyses or for self injury occurring as the next event.

In the trial of Thompson 1994, the target classes of SIB were the three most frequent patterns shown by each individual as revealed by videotaped samples and interviewing staff. Frequency and intensity of these target behaviours were recorded in five-minute blocks randomly distributed throughout the day programme hours. An average of nine five-minute observations per subject per day were scheduled.

Willemsen-Swinkels 1995 measured SIB rate on a five-point rating scale that formed part of Aberrant Behaviour Checklist (ABC) (Aman 1985). A list of target behaviours compiled by authors was rated on a five-point rating scale. At the end of a treatment period staff rated the participants behavioural change on the Clinical Global Impression Scale (CGIS) (National Institute of Mental Health 1985), which uses a seven-point scale ranging from 'very much improved' to 'very much worse'. Participants were observed twice at baseline (at beginning and end of placebo period), six and 24 hours after the single dose administration and after two and four weeks of daily treatment. Each observation was made up of two separate blocks, each being 10 minutes. Treatment emergent adverse effects were assessed through active questioning of the staff and particpants during treatment and at the end point. Liver functions were screened as well. The single-dose study was analysed with repeated measures analysis of variance using SPSS-PC software with drug as within-subject factor and diagnosis as the between-subject factor. The data relating to the long-term treatment were analysed in a similar way but with additional between-subject factor of time and dose. Baseline scores were used as covariates. The Barlett Box-F was used to test for homogeneity of variance. The Box-M was used as a multivariate test for equality of the variance-covariance matrixes. All testing was two tailed, with alpha = 0.05.

Clomipramine

In Lewis 1996, frequency of SIB was measured by direct observation by using general definitions for each category of behaviour. Intensity of SIB was quantified using a standard five-point Likert scale. Measurements in the classroom setting by teachers were obtained at the end of each week of the study period using the ABC (Aman 1985). Daily care provider ratings of SIB were obtained to assess whether drug treatment affected staff requirements for management of problem behaviours. The criterion for significant improvement was at least 50% reduction in SIB during clomipramine maintenance phase from placebo maintenance phase. Other outcomes reported in the trial included stereotypy, compulsions and self restraint. The effects of comipramine on SIB frequency and intensity were assessed by comparing the placebo maintenance phase to the drug maintenance phase for all participants by paired t test with alpha = 0.05. Percentage change in outcome measures from placebo to drug and the number of participants exhibiting a clinically significant response to clomipramine were also reported. Individual subject percentage change in SIB from average SIB frequency was reported in the living unit and classroom settings.

Measuring adverse events

In the trial of Symons 2001, side effects were monitored daily (vital signs) and weekly (adverse effects). In the trial of Thompson 1994, a drug side effects checklist was constructed as well as general observation and monitoring of side effects, based on the most frequently reported side effects of naltrexone and clonidine. This checklist was administered weekly by a staff member and answered on a three-point scale, which included, 'not at all', 'occasionally' and 'frequently'. Willemsen-Swinkels 1995 reported treatment emergent adverse effects. These were assessed by active questioning of staff and participants during treatment and end point. To evaluate the safety of naltrexone, liver functions were screened (levels of serum aspartate aminotransferase, serum alanine aminotransferase, alkaline phosphatase and bilirubin). Adverse events in the trial of Sandman 1990 were not reported. In the Lewis 1996 trial, nurses recorded weekly medication-related side effects using a modified version of Treatment Emergent Side Effects Scale (Rapoport 1985).

Excluded studies

Six studies were excluded for the reasons below and as described in the Characteristics of excluded studies table.

Three studies (Szymanski 1987; Sandman 1990; Garcia 1999) were excluded as they included only two participants each.

Three double-blind cross-over studies (Kars 1990; Zingarelli 1992; Bodfish 1997), which included between six and nine participants, were excluded because they were found to be unrandomised on a closer look.

Risk of bias in included studies

We used The Cochrane Collaboration's tool for assessing risk of bias (Higgins 2008). The risk of bias within each included trial was assessed based on six domains (see 'Risk of bias' tables under Characteristics of included studies, Figure 2, and Figure 3), with review authors judging each risk of bias component to be at 'low risk of bias', 'high risk of bias' or 'unclear risk of bias'. The tool was used to assess the following domains: sequence generation, allocation concealment, blinding (of participants, personnel and outcome assessors), incomplete outcome data, selective reporting and other possible sources of bias (for example, stopping the trial early without reasonable justification and adherence to protocol, changing methods during trial).

Figure 2.

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

Figure 3.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study

We recorded the proportion of participants whose outcomes were not reported at the end of the trial. Quality assessment was made of all included studies, to consider the following questions.

Sequence generation

In Lewis 1996, although allocation was random, method of sequence generation was not mentioned. Sandman 1990 used a Latin square design. Order of administration was random but method of sequence generation was unclear in Symons 2001. Similarly, though it was reported that participants were randomly assigned to treatment, sequence generation methods were not reported in Thompson 1994; no details were reported on method of allocation in Willemsen-Swinkels 1995. As highlighted above, we could not confirm whether the method of generation of the sequence of random numbers used to allocate participants to treatment arms was adequate in any of the trials other than Sandman 1990 making the risk unclear in all studies except Sandman 1990 where risk of bias was deemed to be low.

Allocation concealment

Only Lewis 1996 reported that allocation was done by pharmacists who were not involved in making observation or data collection. The other four studies made no reference to whether any process of concealment of allocation were used. All studies except Lewis 1996 where judged to be at unclear risk of selection bias.

Blinding

All five studies provided details about whether blinding was used and if so whether it was single- or double-blind. The risk of performance and detection bias was low across the five studies. For Lewis 1996, placebo periods at the beginning and end of study involving all participants were single-blind and all other phases were double-blind. In the Sandman 1990 study, the authors reported a double-blind Latin square design, which is most likely referring to blinding of participants and personnel. The authors for Thompson 1994, Willemsen-Swinkels 1995 and Symons 2001 reported a double-blind placebo-controlled trial, which most likely referred to blinding of participants and personnel.

Incomplete outcome data

Overall risk of attrition was deemed to be low for all the included five studies. Only one subject (12.5%) had to discontinue in the last week of clomipramine maintenance phase in Lewis 1996 due to side effects that did not respond to lowering of dosage. Data were collected and reported for this participant during the period present. One participant dropped out from the Thompson 1994 study due to an illness not related to self injury or medication treatment. Outcome data for 97% of participants were available in Willemsen-Swinkels 1995 where "one patient dropped out due to acute and severe increase in SIB. She had to be isolated for several weeks and her condition improved only gradually after naltrexone was stopped and lithium introduced. Her data were excluded from the study''. No intention-to-treat analysis was performed.

Selective reporting

All pertinent outcomes appear to have been reported in all the studies except Sandman 1990 making the risk of selective reporting low. Sandman 1990 did not report adverse events so selective reporting may have been an issue in this trial. It was unclear whether any other bias may have been present in the trials. Another area of concern was the low numbers of participants in each trial, but this was more an issue of the trials being very much underpowered than any risk of bias concern.

Effects of interventions

We have presented effects of intervention of naltrexone and clomipramine separately. Outcomes commonly reported in some of the included trials (but not in our protocol), which we thought were clinically useful, have been also included here.

Naltrexone (opioid receptor antagonist) versus placebo

Four trials reported a comparison of naltrexone versus placebo (Sandman 1990; Thompson 1994; Willemsen-Swinkels 1995; Symons 2001). The four trials included 49 participants in total but only 42 participants with SIB. The seven participants without SIB were from the Willemsen-Swinkels 1995 study. Where possible we have excluded these seven participants with autism and no SIB in this trial as the authors attempted to distinguish between subgroups of participants for some outcomes and analyses.

Meta-analysis was not possible due to differences in study designs (duration, cross-over phases with in the studies), heterogeneity of interventions (doses of drugs) and differences in how outcome measures were reported. Sandman 1990 was a Latin square design lasting four weeks, with fixed doses of 0, 25, 50 and 100 mg given on two days per week. Participants received each dose on a different week as determined by Latin square design. In Symons 2001 naltrexone dose was based on body weight (1.5 mg/kg) and lasted 11 weeks with four phases of treatment, each phase lasting for two weeks separated by one week of placebo condition. The Thompson 1994 trial lasted eight weeks. There was a two to three week drug-free baseline period at the beginning. Each participant then received 50 mg and 100 mg per day of naltrexone or placebo for a two-week period, with periods arranged in a different order for each subject. Number of calendar days per phase was typically 14 but data were collected only on weekdays. Due to holidays or illnesses some conditions for some participants were comprised of data from less than 10 days. The Willemsen-Swinkels 1995 trial lasted 15 weeks. It started with a two-week single-blind placebo period. At the beginning of week three, a single dose of naltrexone 100 mg or placebo was administered. For remainder of that week each participant received a placebo capsule each day. Subsequently a four-week treatment of naltrexone or placebo was given, followed by a four-week washout period and then a cross over to other treatment. In the first four weeks, 19 participants received naltrexone 50 mg/day. As naltrexone 50 mg had no effect, the dose for the second cohort of 14 participants was 150 mg/day.

Self-injurious behaviour rate

Sandman 1990 reported that naltrexone attenuated SIB in all four participants. The effect of naltrexone was dose dependent for three participants and the 100 mg produced the largest clinically noticeable decrease in SIB, although one participant responded favourably to 25 mg and 50 mg but not to 100 mg. All ratings on level of activity remained between two and four, which indicated that self injury reduction was not due to sedation from naltrexone. The 25 mg and 50 mg doses produced a statistically significant decrease in rate of SIB (t = 3.38, P value < 0.05 and t = 7.49, P value < 0.01 for 25 and 50 mg doses, respectively). However, the 100 mg dose did not achieve statistical significance.

The Symons 2001 trial reported that naltrexone had clinically significant effects (≥ 33% reduction) on the daily rates of three of the four participants' most severe form of self injury and resulted in modest to substantial reductions in SIB for all participants. The criteria for clinical significance was determined a priori based on previous research findings. Statistical data on change in SIB was not given. This study also examined sequential dependency between antecedent staff interaction and self injury for each subject. For all four participants, the magnitude of the sequential dependency between staff behaviour and self injury was significantly greater during treatment with naltrexone than during treatment with a placebo as shown by Yule's Q with z and P value. For participant one, sequential dependency (Yule's Q = 0.51) between staff prompting and self injury was significant during naltrexone treatment (z = 2.81, P value < 0.001) but not with placebo (Yule's Q = 0.45). The sequential dependency (Yule's Q = 0.45) between staff prompting and self injury for second participant was significant during naltrexone treatment (z = 2.66, P value < 0.004) but not placebo (Yule's Q = 0.25). For participant three, sequential dependency (Yule's Q = -0.31 ) was significant (z = -3.92, P value < 0.001) with naltrexone but not placebo (Yule's Q = 0.02). For participant four, again the sequential dependency was significant with naltrexone (Yule's Q = 0.45) with (z = 1.74, P value < 0.04).

The Thompson 1994 trial reported that the percentage of days (that is, days with low frequency of SIB) in the first quarter of the overall distribution of frequency of SIB during the baseline condition (see description of included studies section for further details) was significantly lower than in the 50 mg drug treatment phase (P value < 0.05). The percentage of days in the fourth quarter of the overall distribution in the baseline condition was significantly higher than in the 50 mg drug treatment phase (P value 0.05). The same pattern was found when comparing the baseline condition and the 100 mg drug treatment condition for the first and fourth quarters. These analyses indicate that naltrexone administration was associated with statistically significantly fewer days of high frequency self injury and significantly more days with low-frequency self injury. No significant differences were found in the percentages of days in the first and fourth quarter of the distribution between 50 and 100 mg naltrexone dosages (P value < 0.05) for the first and fourth quarter. Naltrexone had different effects depending on the form/location of the self injury. Because there were only data for a maximum of 10 days per phase and the measures in question tended to vary considerably from day to day, it was difficult to evaluate treatment effects within individuals. However, the trial reported that hand-to-head and head-to-object self injury were more likely to be reduced by naltrexone treatment than were other forms of self injury.

In the trial of Willemsen-Swinkels 1995, neither the single dose nor long-term treatment with naltrexone revealed any significant effects on SIB as measured in the ABC questionnaire and the target checklist. In fact, according to the CGIS ratings given by staff, treatment with placebo seemed more beneficial than the four-week treatment with naltrexone 50 mg/day. In addition, naltrexone treatment tended to be associated with an increase in stereotypic behaviour as rated by the care staff (pre- and post-treatment values given as mean and SD). This increase was independent of diagnosis or dose. The authors presented results graphically but numerical data were not reported. No significant correlations were found between behavioural changes following a single dose of naltrexone hydrochloride, 100 mg, and following a four-week treatment with 50 mg or 150 mg.

Adverse events
a) Effects on liver function

Thompson 1994 reported physiological measures that included assessing liver enzyme (serum glutamic oxaloacetic transaminase (SGOT) and gamma-glutamyl transpeptidase (GGT)) test values prior to beginning naltrexone, after two weeks of 50 mg and two weeks of 100 mg of naltrexone per day, and during the week following termination of naltrexone. This was assessed using a two-point rate reaction and a manual method (see Included studies). There were no consistent trends in SGOT and GGT from baseline to either naltrexone dose for any subject, and all values fell within the normal reference range. There were no statistically significant differences in SGOT and GGT levels between baseline, the 50 mg dosage condition (SGOT and GGT; P value > 0.05) and the 100 mg dosage condition (SGOT and GGT; P value > 0.05).

Results of liver function profiles for all participants in the Willemsen-Swinkels 1995 trial were normal and did not deviate from the normal range following the administration of naltrexone.

b) Other adverse events

Other physiological measures in Thompson 1994 included mean systolic and diastolic blood pressure and mean pulse rate. There were no statistically significant differences in any of these outcomes. It was also interesting to note that there was no statistically significant difference in the duration of sleep for participants during the naltrexone and clonidine combination phase and their average level of sleep.

The trial of Willemsen-Swinkels 1995 reported that one participant complained of nausea and tiredness. No severe adverse events were noted in the trial of Symons 2001. Information on adverse events was not reported in the Sandman 1990 trial.

Stereotypy

The Sandman 1990 trial reported the average difference of normalised values that reflected changes in stereotypy after treatment with naltrexone. There was no effect of naltrexone, or even a small increase of stereotypy at 50 mg. Inspection of the individual participants suggested that stereotypy decreased in two participants, increased at all doses in one participant and increased precipitously at 50 mg in one participant. None of the effects of naltrexone on stereotypy were statistically significant. In the trial of Willemsen-Swinkels 1995, naltrexone treatment tended to be associated with an increase in stereotypic behaviour as rated by care staff. This increase was independent of diagnosis or dose.

Clomipramine (serotonin reuptake inhibitor) versus placebo

The trial of Lewis 1996 reported a comparison of clomipramine versus placebo in eight participants. The effects of clomipramine on SIB frequency and intensity were assessed by comparing placebo maintenance phase to drug maintenance phase for all participants by paired t test with alpha = 0.05. A 50% reduction in SIB during the double-blind clomipramine maintenance phase from the placebo maintenance phase was set as the criterion for establishing clinically significant improvement. The number of participants meeting this criterion was compared to the number of participants who showed a similar level of change from the first to the second single-blind placebo phase participants. No statistically significant group differences were found when comparing placebo to active drug for any outcome measure. This reflected the small sample size, which included two obvious non-responders. Percentage change in outcome measures from placebo to drug and the number of participants exhibiting a clinically significant response to clomipramine were also reported. Individual subject percentage change in SIB from average SIB frequency was reported in the living unit and classroom settings.

Self-injurious behaviour rate

Six of the participants exhibited a clinically significant improvement (≥ 50% reduction from placebo) in the frequency of SIB. Four participants exhibited this level of improvement in the living unit, whereas three met the 50% or greater improvement criterion in the classroom setting. None of the participants exhibited a 50% or greater variation from one single-blind condition to the other in the living unit, whereas two participants did so in the classroom.

Self-injurious behaviour intensity

Three of the eight participants showed a 50% or greater reduction in the ratings of SIB intensity in the living unit whereas none of the participants met this criterion in the classroom setting.

Care provider intervention ratings were recorded as the mean percentage of study days that staff intervention was required for behaviour management for participants in the placebo and clomipramine phases. Behavioural intervention by staff was required on a substantially lower percentage of days during the clomipramine maintenance phase compared to placebo maintenance phase. Sixty three per cent of participants exhibited a clinically significant clomipramine-related reduction from placebo on this measure.

Adverse events

A total of eight side effects were reported for all participants during the placebo phase compared to six for participants during the clomipramine phase. Nine adverse events were reported during the trial. Of these six side effects, four (constipation, increased appetite, heart pounding, increased frequency of urination) appeared at increased frequency and severity relative to placebo. None of above side effects were treatment limiting. The most prevalent clomipramine-related side effect was heart pounding reported in 50% of the participants. Two of the participants (25%) had treatment limiting side effects other than those listed on the rating scale. These included seizure for one subject and worsening of mood and tachycardia for one subject during clomipramine maintenance phase participants.

Stereotypy

Three of the eight participants showed a 50% or greater reduction in stereotypy frequency, two of these noted in the living unit and one in the classroom.

Self restraint

None of the participants exhibited a 50% or greater reduction in self restraint in the living unit, whereas in the classroom two participants met this criterion.

Compulsive behaviour

Two participants, in addition to self injury and stereotypies, exhibited compulsive behaviour. Clomipramine treatment resulted in a marked and clinically significant reduction in this behaviour in one participant.

Discussion

Summary of main results

We found five double-blind, placebo-controlled trials that met our inclusion criteria. These trials assessed the effectiveness and safety of drugs in participants with IDs demonstrating SIB. Four trials compared the effects of naltrexone versus placebo and one trial compared clomipramine treatment versus placebo.

Four trials reported on naltrexone versus placebo and the findings of these trials gave mixed reports about the clinical value of naltrexone on the rate of SIB. Three of the four studies reported some benefit with naltrexone (Sandman 1990; Thompson 1994; Symons 2001); however, the Willemsen-Swinkels 1995 study, which had more participants than the total number of participants in the other three studies, did not report any benefit. The trial of Sandman 1990 reported that all four participants had decreased rate of SIB when treated with naltrexone. Three of these participants decreased their SIB as dose of naltrexone increased. There were no consistent effects of naltrexone on stereotypy. Similarly, the trial of Symons 2001 reported clinically significant reductions in SIB rate for three of the four participants. For all participants, the magnitude of the sequential dependency between staff behaviour and self injury was significantly greater during treatment with naltrexone than during treatment with placebo. In Thompson 1994, during naltrexone administration, there were fewer days with frequent head-banging and self biting, whereas there were more days on which blows to the head or self biting were infrequent. However, in Willemsen-Swinkels 1995, naltrexone treatment had no therapeutic effects on SIB. There were suggestions that naltrexone may have actually increased the incidence of stereotypic behaviour.

One trial reported a comparison of clomipramine versus placebo on SIB rate and intensity, stereotypy and adverse events (Lewis 1996). It reported clinically significant improvement (≥ 50% reduction from placebo) in the frequency of SIB in six of the eight participants in the trial. Three participants showed reduction in SIB intensity, three showed a reduction in stereotypy and two exhibited reduction in self restraint. Two participants exhibited compulsive behaviour and clomipramine treatment resulted in reduction in this behaviour in one of them. However, there were no statistically significant group differences when comparing placebo to active drug for any outcome measures.

Two trials reported that there were no significant differences in the results of liver function for all participants following the administration of naltrexone compared to normal levels of liver function (Thompson 1994; Willemsen-Swinkels 1995). No other severe adverse events were noted in three of the trials reporting naltrexone (Thompson 1994; Willemsen-Swinkels 1995; Symons 2001), but information on adverse events was not reported in the Sandman 1990 trial. Nine adverse events were reported in the clomipramine trial (Lewis 1996), and these ranged from none to much in severity using a three-point scale. Of the more severe adverse events, one participant in each of the placebo and clomipramine phases reported stiffness and tremor.

The five RCTs had varied lengths of follow-up period and outcomes were reported in the short term. It would have been useful if participants had been followed up for longer (possibly one to two years) as we would have been able to assess the benefits of treatment in the long term. It is also questionable whether the washout period in some of the trials was sufficient to avoid carry-over effects into the next phase. In Lewis 1996 there was no washout period, in Thompson 1994 some participants had a placebo period and others did not have this between the dose change. However, in the trial of Willemsen-Swinkels 1995 , the washout period was four weeks and in the Symons 2001 trial there were two weeks between the cross-over periods.

All comparisons were restricted to single trial analyses and the number of participants in each trial was very small. This review was not adequately powered as it was based on only 50 participants. In addition, meta-analysis was not possible due to the reasons described earlier, which did not allow us to make any firm conclusions and inferences.

We were unable to examine the efficacy of the following pharmacological interventions as we did not identify any relevant placebo controlled trials: antidepressants other than clomipramine, antipsychotics, mood stabilisers or beta-blockers regardless of dosage, against placebo.

Overall completeness and applicability of evidence

From this review, there is conflicting evidence that naltrexone is efficacious in management of SIB for people with IDs. It is also difficult to draw any firm conclusions about efficacy of clomipramine as there was only one study in this review although it did report some clinical benefit.

The review was incomplete in terms of comparing all available modalities of treatment for SIB, for example, there were no trials comparing antidepressants other than clomipramine, antipsychotics, mood stabilisers or beta-blockers regardless of dosage, against placebo.

Four of the trials in the review appeared to address adverse effects in the active drug and placebo phases of the trial, but again the numbers were small or were specified in the methods but not noted in the results section (so probably were not observed but this is not explicitly stated) so it was not possible to make definite conclusions.

It is important to note that all of the studies were done in large, state residential units. In the recent years large institutions are being closed and care is being provided in smaller community settings and peoples' own homes. The participants in the trials were all having severe-profound ID except Willemsen-Swinkels 1995 study which had participants with mild to profound ID. None of the trials lasted more than 19 weeks. Therefore it is difficult to generalise the findings of this review to a community sample.

Quality of the evidence

This review is based on five trials that compared pharmacological interventions with placebo for participants with IDs demonstrating SIB. The trials were randomised, double-blinded, placebo-controlled trials without significant drop-out rates but with low numbers in each trial. The trials included a total of only 50 participants with SIB and seven participants with autism who did not demonstrate SIB. All trials were at high risk of bias, apart from the trial of Lewis 1996, which was probably at low risk of bias, although the method of sequence generation could not be confirmed. The short period of follow-up was a significant drawback in the design of all five trials, as it did not allow long-term assessment of behaviour over time. Overall, the evidence is weak and incomplete and does not allow for a robust conclusion based on our objectives.

Overall, the quality of the evidence was very low (GRADE Working Group) as there was insufficient evidence to support the use of drugs for participants with SIB, although some of the trials found clinically significant benefits of using naltrexone and in one trial, clomipramine, compared with placebo.

Potential biases in the review process

A comprehensive search was performed, including a thorough search of the grey literature, and all references were sifted and data extracted by two review authors independently. We restricted the included studies to RCTs as they provide the strongest level of evidence available. Hence we have attempted to reduce bias in the review process.

The greatest threat to the validity of the review is likely to be the possibility of publication bias, that is, studies that did not find the treatment to have been effective may not have been published. We were unable to assess this possibility as the analyses were restricted to the results of a single trial.

Agreements and disagreements with other studies or reviews

There are other reviews on pharmacological interventions in the management of SIB but their inclusion criteria were different from our review. Symons 2004 conducted a quantitative synthesis of peer-reviewed published literature on efficacy of naltrexone from 1983 to 2003. They found 27 research articles with 86 participants. They reported clinical improvement of SIB in 80% of participants, and in 47% of these, it was reduced by 50% or more. However, the age range of participants was seven to 67 years, and 9% of studies were open-label, 6% single-blind and 85% double-blind.

Sohanpal 2007 did a systematic review of effectiveness of antidepressants in management of behaviour problems including SIB in adults with IDs from 1990 to 2005. They found one trial on clomipramine and nine on various SSRIs. In their narrative summary they reported mixed results for behaviour problems: less than 50% of cases showed improvement (some clinical, some statistical) and more than 50% of cases showed no improvement or deterioration.

Another literature review, by Matson 2000, found 34 studies relevant to SIB but many included children and the inclusion criteria appeared to be unrestrictive and included a wide range of study types. Of the 34 studies, they found three on use of SSRIs, one on clonidine, one on imipramine and another study on depakote. Out of these six studies, five reported overall decrease in frequency and severity of SIB; however, all of these were uncontrolled open trials. They found 13 trials on naltrexone, which were more methodologically sound, and two of them, Sandman 1990 and Willemsen-Swinkels 1995, are included in our review. Most of these studies on naltrexone indicated significant decrease in SIB. However some other naltrexone studies reported no change or worsening. They also found 15 studies on antipsychotics, most of which were single-case designs. All showed suppression of targeted behaviours including SIB but at the cost of suppressing adaptive behaviours where it was measured.

A narrative review by Villalba 2000 found three open trials of fluoxetine with 22 participants and authors report that these trials showed significant improvement in SIB but they also found one open study on paroxetine with 15 participants where measures of SIB did not change. They also found 19 studies on neuroleptics, three being double-blind studies that showed neuroleptics did not have any specific effect on SIB. There were 24 studies on use of naltrexone. Of these, 13 were double-blind involving 94 participants. The results of studies on naltrexone in this review were mixed and not supported by robust statistical power. In this review, an open trial with 28 participants on divalproex sodium showed significant reduction of SIB in 88% of participants.

These mixed results in the above reviews are in line with the findings of our review. They also had small numbers but did not restrict their inclusion to RCTs.

Authors' conclusions

Implications for practice

This review does not provide evidence for the efficacy of pharmacological interventions in managing SIB in adults with IDs. It therefore has no clear implications for clinical decision-making, leaving clinicians in the unsatisfactory position of having to rely on clinical experience and judgement. This means clinicians should be cautious in the use of these medications which, in many countries, may be prescribed 'off label'. The paucity of evidence needs to be discussed carefully with patients and carers, and their views need to be taken into account. If prescribed, the effects of medication on individual patients needs to be carefully monitored and regularly reviewed.

Implications for research

Given their widespread use, there is a need for further research to establish the efficacy, safety and acceptability of pharmacological interventions for SIB in this population. This may necessitate not only efficacy trials of pharmacological interventions (compared with placebo and compared one to another), but trials comparing pharmacological with other, non-pharmacological ways of management. Observational studies may be needed to answer questions of acceptability and safety (or adverse consequences)

We found no RCTs that examined the efficacy of newer and more commonly used medications, for example, SSRIs, and newer antipsychotics. Again, given their widespread use, this is a significant evidence gap.

The development of common outcome measures in studies of treatment efficacy and safety would help to address the issues of heterogeneity of outcome reporting, and facilitate the development of a more robust evidence base.

In the event that future evidence identifies the efficacy of certain medications, then finding the right therapeutic dose, regimen and type to minimise the adverse effects is also essential. International collaboration is needed to evaluate the efficacy and safety of various treatments due to the fact that only very small trials, predominantly in the USA, have been conducted in this area.

Acknowledgements

We would like to thank Geraldine Macdonald for her useful feedback and Laura MacDonald for her support during the writing of the review. We would also like to thank Andrew Bryant for his help with statistics. Margaret Anderson and Mark Bryant for the development of the search strategy and getting the articles and Marialena Trivella for her help with data extraction method.

Data and analyses

Download statistical data

This review has no analyses.

Appendices

Appendix 1. Search strategies

CENTRAL part of The Cochrane Library, 2012 (Issue 2) Searched 19 February 2012 (note that when MeSH descriptors were updated in 2012, the MeSH term Mental retardation/ was replaced by Intellectual disability/)

#1 MeSH descriptor Self-Injurious Behavior, this term only
#2 MeSH descriptor Self Mutilation, this term only
#3 MeSH descriptor Dangerous Behavior, this term only
#4 MeSH descriptor Stereotyped Behavior, this term only
#5 (self near/3 damage* ) or (self near/3 destruct* ) or (self near/3 harm* ) or (self near/3 hurt* ) or (self near/3 injur* ) or (self near/3 mutilat* ) or (self near/3 violen* ) or (self near/3 wound* )
#6 (parasuicid* or para-suicid* )
#7 ( (behav* near/3 challenging ) or (behav* near/3 disrupt* ) or (behav* near/3 disturb* ) or (behav* near/3 problem* ) )
#8 (#1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7)
#9 MeSH descriptor Intellectual Disability explode tree 1
#10 MeSH descriptor Developmental Disabilities, this term only
#11 ( (mental* near/3 retard* ) or (mental* near/3 retard* ) or (mental* near/3 disab* ) or (mental* near/3 deficien* ) or (mental* near/3 handicap* ) or (mental* near/3 subnormal* ) or (mental*near/3 sub-normal* ) )
#12 ( (intellectual* near/3 impair* ) or (intellectual* near/3 disab* ) or (learning* near/3 impair* ) or (learning* near/3 disab* ) )
#13 (feeble-mind* or feeble next mind* )
#14 (moron* or imbecil* )
#15 (de near/3 lange* )
#16 Cri NEXT du NEXT chat
#17 (Wagr near/3 syndrome )
#18 (down* near/3 syndrome* )
#19 (Rubinstein near/3 Taybi )
#20 (prader near/3 willi* )
#21 (williams near/3 syndrome )
#22 fragile next X
#23 (Martin next Bell or FRAXE or FRAXA )
#24 ( (lesh near/3 nyhan* ) or (Lesch near/3 nyhan* ) )
#25 (#9 OR #10 OR #11 OR #12 OR #13 OR #14 OR #15 OR #16 OR #17 OR #18 OR #19 OR #20 OR #21 OR #22 OR #23 OR #24)
#26 (#8 AND #25)

CENTRAL, The Cochrane Library DVD Version searched April 2011

#1 MeSH descriptor Self-Injurious Behavior this term only
#2 MeSH descriptor self mutilation this term only
#3 MeSH descriptor Dangerous Behavior this term only
#4 MeSH descriptor stereotyped behavior this term only
#5 ( (self in All Text near/3 damage* in All Text) or (self in All Text near/3 destruct* in All Text) or (self in All Text near/3 harm* in All Text) or (self in All Text near/3 hurt* in All Text) or (self in All Text near/3 injur* in All Text) or (self in All Text near/3 mutilat* in All Text) or (self in All Text near/3 violen* in All Text) or (self in All Text near/3 wound* in All Text) )
#6 (parasuicid* in All Text or para-suicid* in All Text)
#7 ( (behav* in All Text near/3 challenging in All Text) or (behav* in All Text near/3 disrupt* in All Text) or (behav* in All Text near/3 disturb* in All Text) or (behav* in All Text near/3 problem* in AllText) )
#8 (#1 or #2 or #3 or #4 or #5 or #6 or #7)
#9 MeSH descriptor mental retardation explode tree 1
#10 MeSH descriptor Developmental Disabilities this term only
#11 ( (mental* in All Text near/3 retard* in All Text) or (mental* in All Text near/3 retard* in All Text) or (mental* in All Text near/3 disab* in All Text) or (mental* in All Text near/3 deficien* in All Text) or (mental* in All Text near/3 handicap* in All Text) or (mental* in All Text near/3 subnormal* in All Text) or (mental* in All Text near/3 sub-normal* in All Text) )
#12 ( (intellectual* in All Text near/3 impair* in All Text) or (intellectual* in All Text near/3 disab* in All Text) or (learning* in All Text near/3 impair* in All Text) or (learning* in All Text near/3 disab*in All Text) )
#13 (feeble-mind* in All Text or feeble next mind* in All Text)
#14 (moron* in All Text or imbecil* in All Text)
#15 (de in All Text and lange* in All Text)
#16 Cri-du-chat in All Text
#17 (Wagr in All Text near/3 syndrome in All Text)
#18 (down* in All Text near/3 syndrome* in All Text)
#19 (Rubinstein in All Text near/3 Taybi in All Text)
#20 (prader in All Text near/3 willi* in All Text)
#21 (williams in All Text near/3 syndrome in All Text)
#22 fragile next X in All Text
#23 (Martin next Bell in All Text or FRAXE in All Text or FRAXA in All Text)
#24 ( (lesh in All Text near/3 nyhan* in All Text) or (Lesch in All Text near/3 nyhan* in All Text) )
#25 (#9 or #10 or #11 or #12 or #13 or #14 or #15 or #16 or #17 or #18 or #19 or #20 or #21 or #22 or #23 or #24)
#26 (#8 and #25)

Ovid MEDLINE(R) <1946 to February Week 2 2012> searched 19 February 2012
(Note that when MeSH descriptors were updated in 2012, the term 'Mental retardation/' was replaced by 'Intellectual disability/')

1 Self-Injurious Behavior/
2 Self Mutilation/
3 Stereotyped Behavior/
4 Dangerous behavior/
5 (self adj3 (damag$ or destruct$ or harm$ or hurt$ or injur$ or mutilat$ or violen$ or wound$)).tw.
6 (parasuicid$ or para-suicid$).tw.
7 (behav$ adj3 (challenging or difficult or disrupt$ or disturb$ or problem$)).tw.
8 or/1-7
9 exp Intellectual Disability/
10 Developmental Disabilities/
11 (mental$ adj3 (retard$ or disab$ or deficien$ or handicap$ or subnormal$ or sub-normal$)).tw.
12 ((intellectual$ or learning$) adj3 (impair$ or disab$)).tw.
13 feeble-minded$.tw.
14 moron$.tw.
15 imbecil$.tw.
16 de lange$.tw.
17 Cri-du-chat.tw.
18 (Wagr adj3 syndrome).tw.
19 (down$ adj3 syndrome$).tw.
20 (Rubinstein adj3 Taybi).tw.
21 (prader adj3 willi).tw.
22 (williams adj3 syndrome).tw.
23 fragile X.tw.
24 ((Martin Bell or FRAXE or FRAXA) adj3 syndrome$).tw.
25 (les?h adj3 nyhan$).tw.
26 or/9-25
27 randomized controlled trial.pt.
28 controlled clinical trial.pt.
29 randomi#ed.ab.
30 placebo$.ab.
31 drug therapy.fs.
32 randomly.ab.
33 trial.ab.
34 groups.ab.
35 or/27-34
36 exp animals/ not humans.sh.
37 35 not 36
38 8 and 26 and 37

Ovid MEDLINE(R) <1948 to March Week 5 2011>

1 Self-Injurious Behavior/
2 Self Mutilation/
3 Stereotyped Behavior/
4 Dangerous behavior/
5 (self adj3 (damag$ or destruct$ or harm$ or hurt$ or injur$ or mutilat$ or violen$ or wound$)).tw.
6 (parasuicid$ or para-suicid$).tw.
7 (behav$ adj3 (challenging or difficult or disrupt$ or disturb$ or problem$)).tw.
8 or/1-7
9 exp Mental Retardation/
10 Developmental Disabilities/
11 (mental$ adj3 (retard$ or disab$ or deficien$ or handicap$ or subnormal$ or sub-normal$)).tw.
12 ((intellectual$ or learning$) adj3 (impair$ or disab$)).tw.
13 feeble-minded$.tw.
14 moron$.tw.
15 imbecil$.tw.
16 de lange$.tw.
17 Cri-du-chat.tw.
18 (Wagr adj3 syndrome).tw.
19 (down$ adj3 syndrome$).tw.
20 (Rubinstein adj3 Taybi).tw.
21 (prader adj3 willi).tw.
22 (williams adj3 syndrome).tw.
23 fragile X.tw.
24 ((Martin Bell or FRAXE or FRAXA) adj3 syndrome$).tw.
25 (les?h adj3 nyhan$).tw.
26 or/9-25
27 randomized controlled trial.pt.
28 controlled clinical trial.pt.
29 randomi#ed.ab.
30 placebo$.ab.
31 drug therapy.fs.
32 randomly.ab.
33 trial.ab.
34 groups.ab.
35 or/27-34
36 exp animals/ not humans.sh.
37 35 not 36
38 8 and 26 and 37

Embase <1980 to 2012 Week 07> searched 19 February 2012

1 automutilation/
2 stereotypy/
3 (self adj3 (damag$ or destruct$ or harm$ or hurt$ or injur$ or mutilat$ or violen$ or wound$)).tw.
4 (behav$ adj3 (challenging or difficult or disrupt$ or disturb$ or problem$)).tw.
5 (parasuicid$ or para-suicid$).tw.
6 suicide attempt/
7 or/1-6
8 exp mental deficiency/
9 intellectual impairment/
10 developmental disorder/
11 (mental$ adj3 (retard$ or disab$ or deficien$ or handicap$ or subnormal$ or sub-normal$)).tw.
12 ((intellectual$ or learning$) adj3 (impair$ or disab$)).tw.
13 feeble-minded$.tw.
14 moron$.tw.
15 imbecil$.tw.
16 de lange$.tw.
17 Cri-du-chat.tw.
18 (Wagr adj3 syndrome).tw.
19 (down$ adj3 syndrome$).tw.
20 (Rubinstein adj3 Taybi).tw.
21 (prader adj3 willi).tw.
22 (williams adj3 syndrome).tw.
23 fragile X.tw.
24 ((Martin Bell or FRAXE or FRAXA) adj3 syndrome$).tw.
25 (les?h adj3 nyhan$).tw.
26 or/8-25
27 exp Clinical trial/
28 Randomized controlled trial/
29 Randomization/
30 Single blind procedure/
31 Double blind procedure/
32 Crossover procedure/
33 Placebo/
34 Randomi#ed.tw.
35 RCT.tw.
36 (random$ adj3 (allocat$ or assign$)).tw.
37 randomly.ab.
38 groups.ab.
39 trial.ab.
40 ((singl$ or doubl$ or trebl$ or tripl$) adj3 (blind$ or mask$)).tw.
41 Placebo$.tw.
42 Prospective study/
43 (crossover or cross-over).tw.
44 prospective.tw.
45 or/27-44
46 7 and 26 and 45

PsycINFO (OVID) searched 19 February 2012

1 Self Inflicted Wounds/
2 Self Injurious Behavior/
3 Attempted Suicide/
4 (self adj3 (damag$ or destruct$ or harm$ or hurt$ or injur$ or mutilat$ or violen$ or wound$)).tw.
5 (parasuicid$ or para-suicid$).tw.
6 (behav$ adj3 (challenging or difficult or disrupt$ or disturb$ or problem$)).tw.
7 Stereotyped Behavior/
8 or/1-7
9 exp mental retardation/
10 (mental$ adj3 (retard$ or disab$ or deficien$ or handicap$ or subnormal$ or sub-normal$)).tw.
11 ((intellectual$ or learning$) adj3 (impair$ or disab$)).tw.
12 feeble-minded$.tw.
13 moron$.tw.
14 imbecil$.tw.
15 de lange$.tw.
16 Cri-du-chat.tw.
17 Cri-du-chat.tw.
18 (Wagr adj3 syndrome).tw.
19 (down$ adj3 syndrome$).tw.
20 (Rubinstein adj3 Taybi).tw.
21 (prader adj3 willi).tw.
22 (williams adj3 syndrome).tw.
23 fragile X.tw.
24 ((Martin Bell or FRAXE or FRAXA) adj3 syndrome$).tw.
25 (les?h adj3 nyhan$).tw.
26 Developmental Disabilities/
27 or/9-26
28 8 and 27
29 clinical trials/
30 (randomis* or randomiz*).tw.
31 (random$ adj3 (allocat$ or assign$)).tw.
32 ((clinic$ or control$) adj trial$).tw.
33 ((singl$ or doubl$ or trebl$ or tripl$) adj3 (blind$ or mask$)).tw.
34 (crossover$ or "cross over$").tw.
35 random sampling/
36 Experiment Controls/
37 Placebo/
38 placebo$.tw.
39 exp program evaluation/
40 treatment effectiveness evaluation/
41 ((effectiveness or evaluat$) adj3 (stud$ or research$)).tw.
42 or/29-41
43 28 and 42

PsycINFO (EBSCOhost) searched April 2011

S40 S25 and S39
S39 S26 or S27 or S28 or S29 or S30 or S31 or S32 or S33 or S34 or S35 or
S36 or S37 or S38
S38 (evaluation N3 stud* or evaluation N3 research*)
S37 (effectiveness N3 stud* or effectiveness N3 research*)
S36 DE "Placebo" or DE "Evaluation" or DE "Program Evaluation" OR DE
"Educational Program Evaluation" OR DE "Mental Health Program Evaluation"
S35 (DE "Random Sampling" or DE "Clinical Trials") or (DE "Experiment
Controls")
S34 "cross over*"
S33 crossover* ch
Database - PsycINFO 4094 Edit S33
S32 (tripl* N3 mask*) or (tripl* N3 blind*)
S31 (trebl* N3 mask*) or (trebl* N3 blind*)
Search Screen - Advanced Search
Database - PsycINFO 0 Edit S31
S30 (doubl* N3 mask*) or (doubl* N3 blind*)
S29 (singl* N3 mask*) or (singl* N3 blind*)
S28 (clinic* N3 trial*) or (control* N3 trial*)
S27 (random* N3 allocat* ) or (random* N3 assign*)
S26 randomis* or randomiz*
S25 S6 and S24
S24 S7 or S8 or S9 or S10 or S11 or S12 or S13 or S14 or S15 or S16 or S17
or S18 or S19 or S20 or S21 or S22 or S23
S23 les#h N3 nyhan*
S22 (Martin Bell N3 syndrome*) or (FRAXE N3 syndrome*) or (Fraxa N3
syndrome*)
S21 fragile X
S20 williams N3 syndrome*
S19 prader N3 willi*
S18 Rubinstein n3 Taybi
S17 down* N3 syndrome*
S16 Wagr n3 syndrome*
S15 Cri-du-chat*
S14 de lange*
S13 DE "Developmental Disabilities"
S12 (DE "Learning Disorders")
S11 feeble-minded* or moron* or imbecil*
S10 (learning N3 impair*) or (learning N3 disab*) or (intellectual* N3
impair*) or (intellectual* N3 disab*)
S9 (mental* n3 retard*) or (mental* n3 disab*) or (mental* n3 deficien*)
or (mental* n3 handicap*) or (mental* n3 subnormal*) or (mental* n3
sub-normal*)
S8 (mental* n3 retard*) or (mental* n3 disab*) or (mental* n3 deficien*)
S7 DE "Mental Retardation" OR DE "Anencephaly" OR DE "Borderline Mental
Retardation" OR DE "Crying Cat Syndrome" OR DE "Down's Syndrome" OR DE
"Home Reared Mentally Retarded" OR DE "Institutionalized Mentally
Retarded" OR DE "Mild Mental Retardation" OR DE "Moderate Mental
Retardation" OR DE "Profound Mental Retardation" OR DE "Psychosocial
Mental Retardation" OR DE "Severe Mental Retardation" OR DE "Tay Sachs
Disease"
S6 S1 or S2 or S3 or S4 or S5
S5 (behav* n3 challenging) or (behav* n3 difficult) or (behav* n3
disrupt*) or (behav* n3 disturb*) or (behav* n3 problem*)
S4 (parasuicid* or para-suicid*)
S3 (self n3 damag*) or (self n3 destruct*) or (self n3 harm*) or (self n3
hurt*) or (self n3 injur* ) or (self n3 mutilat*) or (self n3 violen*) or
(self n3 wound*)
S2 DE "Attempted Suicide"
S1 DE "Self Inflicted Wounds" OR DE "Self Injurious Behavior" OR DE "Head
Banging"

CINAHL (EBSCOhost) 1937 to current searched 19 February 2012

S42 S26 and S41
S41 S27 or S28 or S29 or S30 or S31 or S32 or S33 or S34 or S35 or S36 or S37 or S38 or S39 or S40
S40 TI (evaluat* study or evaluat* research) or AB (evaluat* study or evaluat* research) or TI (effectiv* study or effectiv* research) or AB (effectiv* study or effectiv* research) OR TI (prospectiv* study or prospectiv* research) or AB(prospectiv* study or prospectiv* research) or TI (follow-up study or follow-up research) or AB (follow-up study or follow-up research)
S39 "cross over*"
S38 crossover*
S37 (MH "Crossover Design")
S36 (tripl* N3 mask*) or (tripl* N3 blind*)
S35 (trebl* N3 mask*) or (trebl* N3 blind*)
S34 (doubl* N3 mask*) or (doubl* N3 blind*)
S33 (singl* N3 mask*) or (singl* N3 blind*)
S32 (clinic* N3 trial*) or (control* N3 trial*)
S31 (random* N3 allocat* ) or (random* N3 assign*)
S30 randomis* or randomiz*
S29 (MH "Meta Analysis")
S28 (MH "Clinical Trials+")
S27 MH random assignment
S26 S24 and S25
S25 S8 or S9 or S10 or S11 or S12 or S13 or S14 or S15 or S16 or S17 or
S18 or S19 or S20 or S21 or S22 or S23
S24 S1 or S2 or S3 or S4 or S5 or S6 or S7
S23 les#h N3 nyhan*
S22 (Martin Bell N3 syndrome*) or (FRAXE N3 syndrome*) or (Fraxa N3 syndrome*)
S21 fragile X
S20 williams N3 syndrome
S19 prader N3 willi*
S18 Rubinstein n3 Taybi
S17 down* N3 syndrome*
S16 Wagr n3 syndrome*
S15 Cri-du-chat*
S14 de lange*
S13 feeble-minded* or moron* or imbecil*
S12 (learning N3 impair*) or (learning N3 disab*) or (intellectual* N3 impair*) or (intellectual* N3 disab*)
S11 (mental* n3 retard*) or (mental* n3 disab*) or (mental* n3 deficien*)
or (mental* n3 handicap*) or (mental* n3 subnormal*) or (mental* n3
sub-normal*)
S10 (MH "Developmental Disabilities")
S9 (MH "Mentally Disabled Persons")
S8 (MH "Mental Retardation+")
S7 (MH "Disruptive Behavior")
S6 (behav* n3 challenging) or (behav* n3 difficult) or (behav* n3 disrupt*) or (behav* n3 disturb*) or (behav* n3 problem*)
S5 (parasuicid* or para-suicid*)
S4 (self n3 damag*) or (self n3 destruct*) or (self n3 harm*) or (self n3 hurt*) or (self n3 injur* ) or (self n3 mutilat*) or (self n3 violen*) or (self n3 wound*)
S3 (MH "Suicide, Attempted")
S2 (MH "Injuries, Self-Inflicted")
S1 (MH "Self-Injurious Behavior")

Science Citation Index and Social Science Citation Index searched 19 February 2012

Conference Proceedings Index - Science and Conference Proceedings Index - Social Science and Humanities searched 19 February 2012

4 #3 AND #2 AND #1
# 3 TS=(random* or control* or trial* or groups* or effectiv* or evaluat* or placebo*)
# 2 TS=((behav* SAME (challenging or diffficult* or disrupt* or disturb* or problem*))) OR TS=("attempt* suicid*" or "suicid* attempt*" or "parasuicid*" or "para-suicid*") OR TS=(self same (injur* or harm* or damage* or hurt* or mutilat* or volen* or wound*))
# 1 TS=(((learning same ( impair* or disab*) ) or (intellectual* same ( impair* or disab*) ) or (developmental* same disab*))) OR TS=((mental* SAME (retard*or disab* or deficien* or handicap* or subnormal*or sub-normal*))) or TS=("de lange" or "Cri-du-chat" or "Wagr syndrome" or "down* syndrome" or "Rubinstein Taybi " or "prader willi" or "williams syndrome" or "fragile X" or "fraxe" or "fraxa" or "martin bell syndrome" or "lesch nyhan") OR TS=(feeble-minded* or moron* or imbecil*)WorldCat

WorldCat searched 19 February 2012
ti:(self injury OR self harm) limited to theses

ZETOC (Conference search only) searched 19 February 2012

"self harm*" disabilit*

ICTRP searched 19 February 2012

Self injur* or self-injur* or self harm* or self-harm*

ClinicalTrials.gov searched 19 February 2012

self harm OR self-harm or self -injur* OR self injur*

Appendix 2. Methods from the protocol archived for future updates

Measures of treatment effect  

We would use the following measures of the effect of treatment.

Dichotomous data

Where dichotomous data are presented, a RR with a 95% confidence interval (CI) will be calculated for each outcome in each trial (Higgins 2008).

Continuous data

Continuous data are analysed when other statistics, such as mean difference, P value, t-statistics, etc. are presented in the trial papers, or made available by the authors of the trials, or are calculable from the available data. Where outcomes are measured using the same scale, a mean difference (MD) will be calculated to determine the differences in mean scores between groups. Where similar outcomes are measured using different scales, a standardised mean difference (SMD) will be calculated (Higgins 2008).

Unit of analysis issues  

The authors will determine whether individuals were randomised in groups (that is, cluster-randomised trials) and whether results were reported at multiple time points.

For trials that use clustered randomisation, results will be presented with proper controls for clustering (robust standard errors or hierarchical linear models). If appropriate controls are not used and it is impossible to obtain the full set of individual participant data, the data will be controlled for clustering using the procedures outlined in Higgins 2008. When outcome measures are dichotomous, the number of events and number of participants per trial arm will be divided by the design effect [1 + (1 - m) * r], where m is the average cluster size and r is the intra-cluster correlation coefficient (ICC). When outcome measures are continuous, the number of participants per trial arm will be divided by the design effect, while leaving the mean values unchanged. To determine the ICC, the reviewers will use estimates in the primary trials on a study-by-study basis. However, where these values are not reported, the reviewers will use external estimates of the ICC that are appropriate to each trial context and average cluster size (Higgins 2008).

If cross-over trials are found, we plan to analyse the data separately. In order to include data from cross-over trials, we require data for all treatment periods and treatment groups regarding the number of allocated participants, exclusions and losses to follow up (Elbourne 2002). If a carry-over effect as evaluated on a clinical basis was suspected and data reported by period, then the analysis will be restricted to data from the first treatment period.

Multiple time points

When the results are measured at multiple time points, each outcome at each point will be analysed in a separate meta-analysis with other comparable studies taking measures at a similar time point post-intervention.

Dealing with missing data  

Missing data and dropouts will be assessed in the included studies. Reasons, numbers and characteristics of dropouts will be investigated and reported. Efforts will be made to contact the authors when further information or data are necessary. Any meta-analyses will use data from all original participants when possible, and will report when that is not the case.

If the missing data appear to be missing at random, we will analyse the available data as presented. If this is not the case, we will attempt to impute the missing data by 'intention-to-treat analysis'. If studies do not report intention-to-treat analyses, attempts will be made to obtain missing data by contacting the study authors. If dichotomous outcome data are not available, values will be imputed by assuming that all participants for whom data are missing did not experience a favourable outcome. If missing continuous data are not forthcoming, an available case analysis will be performed.

Where studies do not report response rates, values will be imputed using the method described by Furukawa 2005. Outcome data will be assumed to be normally distributed. Number of responders (n) will be calculated using the formula n = NΦ(x - μ/σ), where N is the number of participants at endpoint, x is 75% of the baseline mean score, μ is the endpoint mean score, σ is the SD of the end point mean and Φ is the cumulative distribution function. The validity of this method will be checked by investigating whether response rate from imputed values is equal to the response rate from observed value.

We will address the potential impact of missing data on the findings of the review in the discussion section.

Assessment of heterogeneity  

We will assess heterogeneity using:

  • visual inspection of forest plots which displays effect estimates and CIs for both individual studies and meta-analyses (Lewis 2001)

  • Chi2 test to assess whether observed differences in results are compatible with chance alone. A low P value (or a large Chi2 statistic relative to its degree of freedom) provides evidence of heterogeneity of intervention effects (variation in effect estimates beyond chance) (Higgins 2008)

  • I2 statistic to determine percentage of variability that is due to heterogeneity rather than sampling error or chance (Higgins 2008).

Assessment of reporting biases  

Funnel plots as described in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2008) will be used to investigate the relationship between effect size and standard error when possible. When such a relationship is found, clinical diversity will be examined as a possible explanation. Asymmetry could be attributed to publication bias or related biases.

Data synthesis  

Meta-analysis will be used when event rates or means and SDs are available or can be calculated and studies include similar populations (for example, sex, level of ID, type of SIB), interventions, methodology (for example, randomisation, measurements, time points, etc.) and outcome measurements. If these are not given, we will impute SDs using relevant data (for example, SDs or correlation coefficients) from other similar studies (Follman 1992) but only if, after seeking statistical advice, to do so is deemed practical and appropriate. Given that trials in this area are often conducted with small samples, any imputations (and the assumptions behind them) are likely to have an important impact. We will, therefore, follow, where possible, the method suggested by Higgins 2008 for weighting studies with imputed data.

Where the interventions are similar as explained above, we aim to synthesise results in a meta-analysis. For this, we will use both a fixed-effect and a random-effects model and compare to assess the impact of statistical heterogeneity. Unless contraindicated (for example, if there is funnel plot asymmetry), we plan to present the results from the random-effects model. In the presence of severe funnel plot asymmetry, we will present both fixed-effect and random-effects analyses, under the assumption that asymmetry suggests that neither model is appropriate.

In all cases, the data will be entered into Review Manager 5 in such a way that the area to the left of the 'line of no effect' indicates a 'favourable' outcome. Data will be inspected within a funnel plot to identify a relationship between study size and size of treatment effect, such as might be caused by publication bias.

Subgroup analysis and investigation of heterogeneity  

If sufficient studies are available, we will conduct subgroup analyses to explore the possible impact of the following factors:

  1. gender;

  2. level of ID

  3. type of pharmacological agents - opioid antagonists, antipsychotics, antidepressants, mood stabilisers/antimanic drugs, beta-blockers;

  4. personality disorders;

  5. autistic spectrum disorders.

These analyses will be exploratory as they involve non-experimental (cross-study) comparisons and we will treat any conclusions with caution.

Sensitivity analysis  

We will conduct sensitivity analyses to determine whether findings are sensitive to restricting the analyses to studies judged to be at low risk of bias. In these analyses, we restrict the analysis to: (a) only studies with low risk of selection bias (associated with sequence generation or allocation concealment); (b) only studies with low risk of performance bias (associated with issues of blinding); (c) only studies with low risk of attrition bias (associated with completeness of data). In addition, we will assess the sensitivity of findings to any imputed data. If cluster randomised trials are included, we will consider sensitivity of a range of ICC values.

Contributions of authors

All authors contributed to the development of this review. AG devised the search strategy. SV screened results of the search for duplicates and all three authors screened the abstracts and titles and retrieved potentially eligible papers and adjudicated with others where there was disagreement. All three reviewed the papers and made decisions about eligibility in discussion with each other. AG and FR extracted data, which was double-checked by SV. 'Risk of bias' tables were completed by FR and SV with adjudication from AG where there was disagreement.

Declarations of interest

  • Aynur Gormez - none known

  • Fareez Rana - none known

  • Susan Varghese - none known

Sources of support

Internal sources

  • Ridgeway Partnership NHS Trust, UK.

External sources

  • No sources of support supplied

Differences between protocol and review

Stereotypy, self restraint and compulsive behaviour were not originally specified in the 'types of outcome measures' when we published the protocol, but they were commonly reported outcomes in some of the included trials and many other studies that we identified. Hence they were post hoc inclusions and are discussed in the relevant sections.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Lewis 1996

MethodsDouble-blind placebo-controlled cross-over design in a state residential unit in the USA
Participants
  • Number of participants: 8

  • Age: 21-39 years

  • Sex: 3 female, 5 male

  • Degree of ID: severe-profound

  • Concurrent conditions and medication: 0 had diagnosis of autistic disorder. 3 participants had history of seizure disorder but were not receiving anticonvulsant medication and had been seizure free for at least 6 months prior to entering the study. 5 participants were receiving antipsychotic medication. 3 were maintained on thioridazine (75-200 mg), 1 was maintained on same dose of thiozanthene (19 mg/day) and 1 on same dose of chlorpromazine (125 mg/day) throughout the study period

  • Type of SIB: hand to face/head hitting, biting hand/wrist, rubbing, fist to chest, pulling hair, kicking legs

  • All participants manifested body stereotypies (for example, body rocking), in addition 4 manifested object stereotypy (for example, toy shaking) and 2 repetitive compulsive behaviours (for example, checking, ordering, hoarding, arranging, touching)

Interventions
  • Trial lasted 19 weeks

  • A single-blind placebo condition (weeks 1-2) was followed by a double-blind placebo or clomipramine condition (weeks 3-9)

  • This involved a titration-up phase (weeks 3-5), a maintenance phase (weeks 6-9) and a titration-down phase (week 10). Cross-over manipulation involved a second titration-up phase (weeks 10-12), a second maintenance phase (weeks 13-16), a second titration-down phase (week 17) followed by a final single-blind placebo condition (weeks 18-19)

  • The final single-blind placebo condition provided an ABA reversal design. During titration-up clomipramine was increased by 25 mg every other day to a maximum of 3 mg/kg body weight. During titration-down it was reduced by 50 mg every day

  • Each subject was randomly assigned by the pharmacy to receive either clomipramine or placebo during first titration and maintenance phases. There were 4 participants in each group (2, 5, 7 and 8 in 1 group)

Outcomes

Primary outcomes

  • Frequency of SIB: measured by direct observation by using general definitions for behaviour

  • Intensity of SIB: quantified using a standard 5-point Likert scale

The criterion for significant improvement was at least 50% reduction in SIB during clomipramine maintenance phase from placebo maintenance phase

  • Adverse effects: nurses recorded weekly medication-related side effects using modified version of Treatment Emergent Side Effects Scale (Rapoport 1985) and recordings of vital signs. Data were presented as (a) the number of participants for whom the side effects were reported, (b) the average scale rating (1 = none, 2 = slight, 3 = much) across participants and (c) the average number of weeks the side effects were reported

Secondary outcomes

  • To assess whether drug treatment affected staff requirements

Daily care providers ratings of SIB were obtained. This was accomplished by having staff members who worked directly with the subject rate whether the subject required no intervention (0), could be ignored (1), needed to be redirected (2) or required interruption - time out or restraint (3).These data were used to tabulate the percentage of days in each drug condition that active staff intervention were required (ratings of 2 or 3)

Other outcomes measured: stereotypy, compulsions, self restraint

Notes
  • Members of research team as well as professional and direct care staff members remained unaware of each participants medication status except participants 3 and 7 where blinding had to be broken as 1 experienced a seizure and the other developed persistent tachycardia and agitated mood

  • All participants were observed in classroom and living unit. 2 x 10-minute partial interval (10 seconds) direct observation sessions (1 per setting) were conducted on each of 3 days weekly for each subject. Total number of observation sessions per subject ranged from 83 to 113. A partial interval direct observation session was used to measure frequency of occurrence of SIB. At the end of each session, the data collector rated symptom intensity using a Likert scale

  • To assess interobserver agreement, a second observer simultaneously but independently, collected data during average of 185 of all sessions for all participants

Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskRandomly allocated, however, method is unclear. "Each subject was randomly assigned by pharmacy to receive either clomipramine or placebo during first titration and maintenance phases"
Allocation concealment (selection bias)Low riskCarried out by pharmacy who were not involved in making observations or data collection
Blinding (performance bias and detection bias)
All outcomes
Low riskPlacebo periods at the beginning and end of study involving all participants were single-blind and all other phases were double-blind, "Members of research team as well as professional and direct care staff members remained unaware of each participants medication status except participants 3 & 7". For participants 3 and 7 blinding had to be broken due to emerging signifcant side effects. However open data collection continued
Incomplete outcome data (attrition bias)
All outcomes
Low riskOnly one 1 subject (12.5%) had to discontinue in the last week of clomipramine maintenance phase due to side effects that did not respond to lowering of dosage. Data was collected and reported for this participant during that period
Selective reporting (reporting bias)Low riskAll pertinent outcomes appear to have been reported, including adverse events
Other biasUnclear riskInsufficient information to permit judgement

Sandman 1990

MethodsDouble-blind Latin square design of participants in a single programme at Fairview State Developmental Research Institute, Costa Mesa, CA
Participants
  • Number of participants: 4

  • Age: 23-26 years

  • Sex: male

  • Level of ID: severe-profound

  • Concurrent conditions and medication: only 1 participant was on a psychotropic, thioridazine (Melaril). 2 participants were maintained on their current medication (ferrous sulphate, vitamin B complex , dimetap extentabs, didoxacillin) to minimise changes to normal daily functioning

  • Type of SIB: head banging, self hitting

Interventions
  • Trial lasted 4 weeks

  • The participants received 0, 25, 50, 100 mg of naltrexone (fixed-dose administration)

  • Participants received each dose on a different week as determined by Latin square design. Naltrexone or placebo was administered at 8 am on Monday and Wednesday each week

  • All participants were videotaped for 10 continuous minutes in their normal surrounding twice a day, morning and afternoon

  • SIB, stereotypy and level of activity were evaluated by reviewing of recorded tapes

  • Level of activity was rated on 5-point scale

  • The Conners' Teachers Rating Scale (Conners 1969) was adapted for use assessing 3 domains on a 4-point scale: activity, aggression and emotional lability

  • The Vineland Adaptive Behavior Scales (Sparrow 1984) was assessed once a week to score social, motor, maladaptive, communication and daily living domains

Outcomes

Primary outcomes

  • SIB: naltrexone attenuated SIB in all of the participants; 25 and 50 mg doses producing a significant decrease t = 3.38, P value < 0.05 and t = 7.49, P value < 0.01 (all degrees of freedom = 3)

  • 100 mg dose failed to achieve statistical significance

  • The effect of naltrexone was dose dependent for 3 participants, 100 mg producing the largest decrease in SIB

Other outcomes

  • There was no clinically or statistically significant effect of naltrexone on stereotypy

  • Scores on the Conners' Teacher Rating Scale showed no significant change in the activity levels with the different dosing levels

Notes
  • Longer sampling period (2-48 hours) may explain the monotonic improvements noted in 3 of the participants with increasing dose

  • No overdosing effects seen with higher dose

  • Parallel analysis of other behaviours indicate naltrexone being specific for SIB

  • Rating activity indicate that naltrexone did not sedate or arouse participants

  • Lack of adequate data to evaluate parametrically the effect of dose on a mg/kg bias

  • Most self-abusive participants were most sensitive to naltrexone implying the degrees of therapeutic response to opioid blockers to be directly related to the extent of opioid disregulation

Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskThe study employed a double-blind Latin square design with people receiving 0, 25, 50 or 100 mg of naltrexone (fixed-dose administration) or placebo at 8 am on Monday and Wednesday
Allocation concealment (selection bias)Unclear riskOnly 4 participants selected. Method of allocation concealment was not described
Blinding (performance bias and detection bias)
All outcomes
Low riskThe authors reported a double-blind Latin square design, which is most likely referring to blinding of participants and personnel: "In a double blind, Latin Square design, the participants received 0, 25, 50, or 100 mg of naltrexone at 8.00 am on Monday and Wednesday only"
Incomplete outcome data (attrition bias)
All outcomes
Low riskAll participants were assessed for each outcome
Selective reporting (reporting bias)High riskAdverse events were not reported
Other biasUnclear riskInsufficient information to permit judgement

Symons 2001

MethodsDouble-blind, randomised, placebo-controlled trial, single centre, at a state residential facility in North America
Participants
  • Number of participants: 4

  • Age: 27-45 years

  • Sex: males

  • Degree of ID: severe to profound

  • Concurrent condition and medications: 2 participants were on antipsychotics. 2 participants had seizure disorder and were maintained on antiepileptics

  • Type of SIB: head banging, self hitting, hand biting, striking head

Interventions
  • Trial lasted 11 weeks

  • All participants had a trial of naltrexone hydrochloride

  • Naltrexone dose was based on body weight (1.5 mg/kg) and administered orally either in ABAB or BABA placebo-controlled double-blind single-subject experimental design

  • Each phase lasted for 2 weeks. There were 4 phases with phases separated by 1 week

  • Order of administration was random

  • Baseline rate of self injury was determined by directly observing each subject for approximately 8 h during the course of typically occurring daytime

Outcomes

Primary outcome

  • Self injury was measured by directly observing and recording its frequency for 1 participant and duration for other 3 participants. Self injury recorded continuously was referred as a 'bout'

  • Side effects were monitored daily (vital signs) and weekly (adverse effects)

  • Antecedent staff interaction defined for verbal and physical promoting were recorded by direct observation in the individuals daytime setting

Notes
  • Trained observers collected direct observational data daily in real time by using hand- held microcomputers

  • Prior to data collection, observers trained first by watching videotape and familiarising themselves with each participant's SIB and characteristic staff interaction with the subject

Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskOrder of administration was random, ABAB or BABA, but method of sequence generation was not reported
Allocation concealment (selection bias)Unclear riskNot reported
Blinding (performance bias and detection bias)
All outcomes
Low riskThe authors reported a double-blind placebo-controlled trial, which is most likely to refer to blinding of participants and personnel: "For all participants, naltrexone was based on body weight and administered orally at 8.00 am either in an ABAB or BABA placebo-controlled double blind single subject experimental design"
Incomplete outcome data (attrition bias)
All outcomes
Low riskAll 4 participants completed the study
Selective reporting (reporting bias)Low riskAll pertinent outcomes appear to have been reported, including adverse events
Other biasUnclear riskInsufficient information to permit judgement

Thompson 1994

MethodsDouble-blind, randomised, within-subject dosage (including placebo) comparison design at a large public residential unit in North America
Participants
  • Number of participants: 8

  • Age range: 29-41 years

  • Sex: 4 men and 4 women

  • Level of ID: severe to profound

  • Concurrent conditions and medication: 1 participant was on dilantin and 3 on carbamazepine for seizure disorders

  • Types of self injury: forms of SIB included hand to head, head to object, poking (nose, throat, eyes), slapping face, biting hands or arms, pinching self banging hands, kicking against object

Interventions
  • Trial lasted 8 weeks

  • There was a 2-3 week drug-free baseline period at the beginning. Each participant then received 50 mg and 100 mg per day of naltrexone or placebo for a 2-week period, with periods arranged in a different order for each subject

  • Number of calender days per phase was typically 14 but data were collected only on weekdays. Due to holidays or illnesses some conditions for some participants were comprised of data from less than 10 days

  • Half of the participants also received 0.3 mg of clonidine during active drug phases to reduce any possible opiate withdrawal effects

  • All medications were administered once daily at 8 am

Outcomes

Primary outcomes:

  • Self-injurious behaviours: the target classes of SIB were the 3 most frequent patterns shown by each individual as revealed by videotaped samples and interviewing staff

  • Frequency and intensity of SIB was recorded in 5-minute blocks randomly distributed throughout the day programme hours. Observation schedules for each subject were repeated every 2 weeks, such that each experimental phase was based on the same number and sequence of observation periods

  • Self injury intensity was estimated using 4-point scale

  • Adverse effects

    • Liver function: effects of naltrexone on liver function was assessed by liver enzyme tests (SGOT and GGT) performed on blood samples drawn at pretreatment, then every 2 weeks and during the week following termination of naltrexone for all 8 participants

    • Vital measures: body weight, supine pulse, blood pressure and body temperature were recorded twice daily

Notes
  • Authors hoped that not only would naltrexone block the reinforcing effects of the endogenous ligand occupying the opiate receptor, but that clonidine would diminish possibility of self injury maintained by avoidance of, or escape from, opiate-like withdrawal symptoms

Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNot reported. "participants were randomly assigned to a treatment sequence of naltrexone followed by placebo or placebo followed by naltrexone"
Allocation concealment (selection bias)Unclear riskNot reported
Blinding (performance bias and detection bias)
All outcomes
Low riskThe authors reported a double-blind placebo-controlled trial, which is most likely to refer to blinding of participants and personnel
Incomplete outcome data (attrition bias)
All outcomes
Low risk1 participant (participant 3) dropped out from study due to an illness not related to self injury or medication treatment
Selective reporting (reporting bias)Low riskAll pertinent outcomes appear to have been reported, including adverse events
Other biasUnclear riskInsufficient information to permit judgement

Willemsen-Swinkels 1995

  1. a

    ABC: Aberrant Behaviour Checklist; GGT: gamma-glutamyl transpeptidase; ID: intellectual disability; SGOT: serum glutamic oxaloacetic transaminase; SIB: self-injurious behaviour

MethodsDouble-blind, randomised, placebo-controlled cross-over study, single-centre (public institution) in the Netherlands
Participants
  • Number of participants: 33 (7 of these were excluded from our review as they did not present with SIB but were part of the study as they had autistic disorder)

  • Age: 18-46 years

  • Sex: 27 males, 6 females

  • Level of ID: mild to profound

  • Concurrent conditions and medication: 24 were autistic of whom 17 presented with SIB, 9 did not have diagnosis of autistic disorder but presented with SIB. 5 had history of epilepsy and were on antiepileptics. 11 were receiving neuroleptics. Dosages of co-medications remained fixed throughout the study period

Interventions
  • Trial lasted 15 weeks

  • It started with a 2-week single-blind placebo period during which participants were trained to receive a capsule (placebo) each day. At beginning of week 3, a single dose of naltrexone 100 mg or placebo was administered (n = 33). For remainder of that week all received a placebo capsule each day

  • Subsequently a 4-week treatment of naltrexone or placebo was given, followed by a 4-week washout period and then a cross over to other treatment

  • In the first 4 weeks 19 participants (2 with autism, eleven with autism and SIB, 6 with SIB) received naltrexone 50 mg/day. As naltrexone 50 mg had no effect, the long-term dose for the second cohort of 14 participants was 150 mg/day (5 with autism, 6 with autism and SIB, 3 with SIB)

Outcomes

Primary outcomes:

  • Rate of SIB

ABC (Aman 1985) and a list of target behaviours were used to rate the participants' behaviour. The behaviours were rated on a 5-point scale

Participants were observed twice at baseline (at beginning and end of placebo period), 6 and 24 hours after the single-dose administration and after 2 and 4 weeks of daily treatment. Each observation was made up of 2 separate blocks, each being 10 minutes. Inter-rater reliability across 2 raters was 90% for SIB

  • Adverse effects: treatment-emergent adverse effects were assessed by active questioning of staff and participants during treatment and end point. To evaluate safety of naltrexone, liver functions were screened (levels of serum aspartate aminotransferase, serum alanine aminotransferase, alkaline phosphatase and bilirubin)

Other outcomes: stereotypy which was recorded using the same checklist as above

Notes 
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNot reported, "participants were randomly assigned to a treatment sequence of naltrexone followed by placebo or placebo followed by naltrexone"
Allocation concealment (selection bias)Unclear riskNot reported
Blinding (performance bias and detection bias)
All outcomes
Low riskThe authors reported a double-blind placebo-controlled trial which is most likely to refer to blinding of participants and personnel
Incomplete outcome data (attrition bias)
All outcomes
Low risk"One patient dropped out due to acute and severe increase in SIB. She had to be isolated for several weeks and her condition improved only gradually after naltrexone was stopped and lithium introduced. Her data were excluded from the study". % analysed: 32/33 (97%)
Selective reporting (reporting bias)Low riskAll pertinent outcomes appear to have been reported, including adverse events
Other biasUnclear riskInsufficient information to permit judgement

Characteristics of excluded studies [ordered by study ID]

StudyReason for exclusion
Bodfish 1997Double-blind single subject design to evaluate effect of long-term treatment with naltrexone. 9 participants. There was no random allocation
Garcia 1999Only 2 participants
Kars 1990Double-blind cross-over study on 6 participants, but not randomised
Sandman 1983Only 2 participants
Szymanski 1987Only 2 participants
Zingarelli 1992Double-blind cross-over study. No randomisation. 8 participants. Groups were matched for age and intelligence quotient

Ancillary