Psychological interventions for women with metastatic breast cancer

  • Review
  • Intervention

Authors


Abstract

Background

Psychological symptoms are associated with metastatic breast cancer. This is the basis for exploring the impact of psychological interventions on psychosocial and survival outcomes. One early study appeared to show significant survival and psychological benefits from psychological support while subsequent studies have revealed conflicting results. This review is an update of a Cochrane review first published in 2004 and previously updated in 2007.

Objectives

To assess the effects of psychological interventions on psychosocial and survival outcomes for women with metastatic breast cancer.

Search methods

We searched the Cochrane Breast Cancer Group Specialised Register, MEDLINE (OvidSP), EMBASE (OvidSP), PsycINFO (OvidSP), CINAHL (EBSCO), online trials and research registers in June/July 2011. Further potentially relevant studies were identified from handsearching references of previous trials, systematic reviews and meta-analyses. 

Selection criteria

Randomised controlled trials (RCTs) and cluster RCTs of psychological interventions, which recruited women with metastatic breast cancer. Outcomes selected for analyses were overall survival, psychological outcomes, pain, quality of life, condition-specific outcome measures, relationship and social support measures, and sleep quality. Studies were excluded if no discrete data were available on women with metastatic breast cancer. 

Data collection and analysis

Two review authors independently extracted the data and assessed the quality of the studies using the Cochrane Collaboration risk of bias tool. Where possible, authors were contacted for missing information. Data on the nature and setting of the intervention, relevant outcome data, and items relating to methodological quality were extracted. Meta-analyses was performed using a random-effects or fixed-effect Mantel-Haenszel model, depending on expected levels of heterogeneity.

Main results

Ten RCTs with 1378 women were identified. Of the seven RCTs on group psychological interventions, three were on cognitive behavioural therapy and four were on supportive-expressive group therapy. The remaining three studies were individual based and the types of psychological interventions were not common to either cognitive behavioural or supportive-expressive therapy. A clear pattern of psychological outcomes could not be discerned as a wide variety of outcome measures and durations of follow-up were used in the included studies. The overall effect of the psychological interventions across six studies, on one-year survival, favoured the psychological intervention group with an odds ratio (OR) of 1.46 (95% confidence interval (CI) 1.07 to 1.99). Pooled data from four studies did not show any survival benefit at five-years follow-up (OR 1.03, 95% CI 0.42 to 2.52). There was evidence of a short-term benefit for some psychological outcomes and improvement in pain scores.

Authors' conclusions

Psychological interventions appear to be effective in improving survival at 12 months but not at longer-term follow-up, and they are effective in reducing psychological symptoms only in some of the outcomes assessed in women with metastatic breast cancer. However, findings of the review should be interpreted with caution as there is a relative lack of data in this field, and the included trials had reporting or methodological weaknesses and were heterogeneous in terms of interventions and outcome measures.  

Résumé scientifique

Interventions psychologiques chez les femmes atteintes d'un cancer du sein métastatique

Contexte

Le cancer du sein métastatique est associé à des symptômes psychologiques.Cette affirmation constitue la base justifiant l'étude de l'impact des interventions psychologiques sur les résultats psychosociaux et la survie. Une étude ancienne avait semblé démontrer l'existence de bénéfices significatifs associés au soutien psychologique en termes de résultats psychologiques et de survie, tandis que les études subséquentes avaient produit des résultats contradictoires. Cette revue est une mise à jour d'une revue Cochrane publiée pour la première fois en 2004 et précédemment mise à jour en 2007.

Objectifs

Évaluer les effets des interventions psychologiques sur les résultats psychosociaux et la survie chez les femmes atteintes d'un cancer du sein métastatique.

Stratégie de recherche documentaire

Nous avons consulté le registre spécialisé du groupe Cochrane sur le cancer du sein, MEDLINE (OvidSP), EMBASE (OvidSP), PsycINFO (OvidSP), CINAHL (EBSCO), ainsi que les registres d'essais cliniques et de recherche en ligne en juin/juillet 2011.D'autres études potentiellement pertinentes ont été identifiées au moyen de recherches manuelles dans les références bibliographiques des précédents essais, revues systématiques et méta-analyses.

Critères de sélection

Les essais contrôlés randomisés (ECR) et les ECR en cluster portant sur des interventions psychologiques et recrutant des femmes atteintes d'un cancer du sein métastatique.Les critères de jugement choisis pour l'analyse étaient la survie globale, les résultats psychologiques, la douleur, la qualité de vie, les mesures de résultats spécifiques à la maladie, les mesures des relations et du soutien social et la qualité du sommeil. Les études ont été exclues lorsqu'aucune donnée spécifique n'était disponible concernant des femmes atteintes d'un cancer du sein métastatique.

Recueil et analyse des données

Deux auteurs de revue ont extrait les données de manière indépendante et évalué la qualité des études à l'aide de l'outil d'évaluation du risque de biais de la Collaboration Cochrane.Dans la mesure du possible, les auteurs ont été contactés afin d'obtenir des informations manquantes.Les données relatives à la nature et à l'environnement de l'intervention, les données de résultats pertinentes et les éléments en rapport avec la qualité méthodologique ont été extraits.Des méta-analyses ont été effectuées à l'aide d'un modèle à effets aléatoires ou d'un modèle de Mantel-Haenszel à effets fixes, selon les niveaux d'hétérogénéité attendus.

Résultats principaux

Dix ECR portant sur 1 378 femmes ont été identifiés. Sur les sept ECR portant sur des interventions psychologiques de groupe, trois examinaient une thérapie cognitivo-comportementale et quatre une thérapie de groupe de soutien par l'expression. Les trois études restantes examinaient des interventions psychologiques individuelles qui n'étaient ni des thérapies cognitivo-comportementales, ni des thérapies de soutien par l'expression. Aucune tendance claire n'a pu être dégagée concernant les résultats psychologiques en raison de la grande variabilité des mesures de résultats et des durées de suivi utilisées dans les études incluses. L'effet global des interventions psychologiques sur la survie à un an dans les six études était favorable au groupe de l'intervention psychologique, avec un rapport des cotes de 1,46 (intervalle de confiance (IC) à 95 %, entre 1,07 et 1,99). Les données combinées issues de quatre études ne révélaient aucun bénéfice en termes de survie lors d'un suivi à cinq ans (rapport des cotes de 1,03, IC à 95 %, entre 0,42 et 2,52).Il existait des preuves de bénéfice à court terme concernant certains critères de jugement psychologiques ainsi qu'une amélioration des scores de douleur.

Conclusions des auteurs

Les interventions psychologiques semblent efficaces pour améliorer la survie à 12 mois mais pas lors d'un suivi à plus long terme, et sont efficaces pour réduire les symptômes psychologiques, mais uniquement pour certains des critères de jugement évalués chez des femmes atteintes d'un cancer du sein métastatique. Néanmoins, les résultats de cette revue devraient être interprétés avec précaution en raison des données relativement limitées dans ce domaine et du fait que les essais inclus présentaient des faiblesses méthodologiques et de notification et utilisaient des interventions et des mesures de résultats hétérogènes.

Plain language summary

Psychological interventions for women with metastatic breast cancer

Cancer that has spread beyond the breast (metastatic breast cancer) is frightening and distressing and can lead women to experience psychological symptoms, such as depression. There is a belief that these psychological symptoms can make the cancer worse.

Treatment for psychological symptoms is sometimes offered to women with metastatic breast cancer, either individually or in a group. In 1989 one study examined the effect of this treatment, offered to a group of women, and it found that it led to them feeling psychologically better and living longer. Subsequent studies did not seem to replicate the findings from the 1989 study, causing uncertainty about the effects of psychological treatments for women with metastatic breast cancer. 

This review examined the studies to date to see what effect psychological treatments had on women with metastatic breast cancer. We found 10 studies with a total of 1378 women with metastatic breast cancer. Three of the studies used a psychological treatment known as cognitive behavioural therapy (CBT), four studies used supportive-expressive group therapy (SEGT), while the remaining three studies used treatments that were delivered on an individual basis and were neither CBT nor SEGT.

We performed statistical analysis and found that the odds ratio (a measure of association between an intervention and an outcome) for survival of women with metastatic breast cancer one year after receiving psychological treatment was 1.46, suggesting that there was an association between the psychological treatment and improved survival. This finding was not found when looking at the odds ratio of survival at five years. We also found some evidence that psychological treatments in the short term (for example, one year) may produce a small reduction in pain and improve some psychological symptoms. However, making comparisons across these studies was difficult as they differed in their conduct, treatments and measures used. Moreover, we cannot rule out that the psychological treatments could also cause psychological harm.

Résumé simplifié

Interventions psychologiques chez les femmes atteintes d'un cancer du sein métastatique

Un cancer qui se propage au-delà du sein (cancer du sein métastatique) est effrayant, stressant, et peut entraîner l'apparition de symptômes psychologiques tels qu'une dépression. On pense que ces symptômes psychologiques sont susceptibles d'aggraver le cancer.

Un traitement des symptômes psychologiques individuel ou en groupe est parfois offert aux femmes atteintes d'un cancer du sein métastatique. En 1989, une étude examinant les effets d'un tel traitement chez un groupe de femmes avait rapporté qu'elles se sentaient mieux psychologiquement et vivaient plus longtemps.Les études subséquentes n'avaient pas reproduit les résultats de cette étude de 1989, suscitant des incertitudes concernant les effets des traitements psychologiques chez les femmes atteintes d'un cancer du sein métastatique.

Cette revue a examiné les études disponibles à ce jour afin de déterminer les effets des traitements psychologiques chez les femmes atteintes d'un cancer du sein métastatique. Nous avons identifié 10 études portant sur un total de 1 378 femmes atteintes d'un cancer du sein métastatique. Trois de ces études examinaient un traitement psychologique connu sous le nom de thérapie cognitivo-comportementale (TCC), quatre études examinaient une thérapie de groupe de soutien par l'expression (TGSE), tandis que les trois autres études examinaient des traitements individuels qui n'étaient ni des TCC, ni des TGSE.

Nous avons effectué une analyse statistique et avons découvert que le rapport des cotes (une mesure de l'association entre l'intervention et le résultat) de la survie des femmes atteintes d'un cancer du sein métastatique un an après le traitement psychologique était de 1,46, ce qui suggérait une association entre le traitement psychologique et l'amélioration de la survie. Ce résultat n'était pas observé pour le rapport des cotes de la survie à cinq ans. Nous avons également identifié des preuves indiquant que les traitements psychologiques à court terme (un an, par exemple) pourraient produire une petite réduction de la douleur et améliorer certains symptômes psychologiques. Néanmoins, il était difficile d'effectuer des comparaisons sur l'ensemble des études car elles utilisaient différents plans, traitements et mesures. De plus, la possibilité que les traitements psychologiques puissent également provoquer des dommages psychologiques ne peut pas être exclue.

Notes de traduction

Traduit par: French Cochrane Centre 1st July, 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.

Laički sažetak

Psihološke intervencije za žene s metastatskim rakom dojke

Rak koji se proširio izvan dojke (metastatski rak dojke) izaziva strah i stres, i može kod žena dovesti do pojave psiholoških simptoma poput depresije. Vjeruje se da ti psihološki simptomi mogu pogoršati rak.

Za psihološke se simptome ženama s metastatskim rakom dojke ponekad ponudi liječenje, bilo pojedinačno ili u grupi. Jedna je studija 1989. g. ispitivala učinak terapije za psihološke simptome ponuđenog skupini žena, te je utvrdila da su se tada psihički osjećale bolje i živjele dulje. Čini se da naknadne studije nisu ponovile ove rezultate studije iz 1989. godine, što dovodi do nesigurnosti zaključaka o učincima psiholoških tretmana za žene s metastatskim rakom dojke.

Ovaj Cochrane sustavni pregled ispitivao je studije provedene do danas kako bi se vidjelo kakav su učinak psihološki tretmani imali na žene s metastatskim rakom dojke. Pronađeno je 10 studija s ukupno 1378 žena s metastatskim rakom dojke. U tri studije korištena je psihološka terapija poznata kao kognitivna bihevioralna terapija (KBT), četiri studije koristile su suportivno-ekspresivnu grupnu terapiju (SEGT), dok su preostale tri studije koristile tretmane koji su primijenjeni individualno, a nisu bili ni KBT niti SEGT.

Proveli smo statističku analizu i utvrdili da je omjer šansi (mjera povezanosti između intervencije i ishoda) za preživljenje žena s metastatskim karcinomom dojke jednu godinu nakon primitka psihološkog tretmana bio 1,46, što ukazuje na povezanost između psihološkog liječenja i poboljšanja preživljenja. Ovaj rezultat nije dobiven kada se gledao omjer šansi za preživljenje nakon pet godina. Također smo našli pojedine dokaze da psihološki tretmani mogu kratkoročno (na primjer jedna godina) dovesti do malog smanjenja boli i poboljšati neke psihološke simptome. Međutim, bilo je teško provesti usporedbe tih studija jer su se razlikovale u načinu provođenja, tretmana i mjera koje su se koristile. Nadalje, ne možemo isključiti da psihološki tretmani također mogu uzrokovati psihološku štetu.

Bilješke prijevoda

Hrvatski Cochrane
Prevela: Katarina Vučić
Ovaj sažetak preveden je u okviru volonterskog projekta prevođenja Cochrane sažetaka. Uključite se u projekt i pomozite nam u prevođenju brojnih preostalih Cochrane sažetaka koji su još uvijek dostupni samo na engleskom jeziku. Kontakt: cochrane_croatia@mefst.hr

Background

Description of the condition

Breast cancer is the most common cancer in women worldwide, comprising 13.7% of all female cancers and accounting for 485,503 deaths in 2008 (Ferlay 2010). The five-year survival from breast cancer presentation is approximately 70% to 90%, depending on country, and is improving (Kmietowicz 2000; McKenna 2002). An important prognostic indicator is the stage of breast cancer at the time of diagnosis. Although metastatic breast cancer is incurable it is treatable and, as survival can be years, it may be regarded as a chronic relapsing and remitting disease (Leonard 2000; Smith 2006).

Whether surviving a longer or shorter time, all women with advanced breast cancer, and their families, are not only living with uncertainty about the future, the burden of treatment, and the threat of dying, but also dealing with the existential, emotional, social and psychological difficulties their situation brings. Indeed, many believe psychological health can influence the outcome of cancer with 85% of cancer patients believing psychological functioning can predict disease progression (Lemon 2004). This belief has been the basis of research exploring the impact of interventions on survival and psychological variables such as depression in patients with cancer.

Prevalence of psychological morbidity in cancer patients

Up to a quarter (16% to 25%) of all newly diagnosed cancer patients experience depression or a clinically significant adjustment disorder with depressed mood using DSM-IV criteria (Sellick 1999). Deflorio 1995 reviewed 49 studies of the subsequent incidence of depression in cancer patients. The results showed a range in the incidence of depression from 1% to 53%, and this variability was considered to be due to methodological differences and variable diagnostic criteria. Different populations being studied may also account for some of the variation. The diagnosis of depression in cancer patients, particularly those with advanced cancer, is also made more difficult by the similarity between the somatic symptoms of depression and the symptoms of advanced cancer (Bottomley 1998; Lloyd-Williams 1999). 

There are plausible explanations that implicate hypothalamo-pituitary-adrenal (HPA) axis activation with cancer progression (Reiche 2004). The disruption of HPA axis activation may lead to cytokines and inflammatory molecules being involved in the functioning of the cellular immune system to enhance the malignant potential of tumours. Also, depression has been shown to increase pro-inflammatory cytokines that may influence cancer outcomes (Seruga 2008). This appears to resonate with findings of immune response changes following psychological interventions (Larson 2000; van der Pompe 1997).     

Prevalence of psychological morbidity in women with breast cancer

Many studies have examined the psychological well-being of women with early breast cancer. Most have focused on women at diagnosis, their reactions to mastectomy versus breast conserving surgery, or while undergoing adjuvant treatments. Up to 30% of women diagnosed with breast cancer will develop psychological morbidity, either an anxiety or depressive disorder, within one year of diagnosis (Bleiker 2000; Glanz 1992; Irvine 1991; Maguire 2000).

Less work has been done on the psychological effects of metastatic breast cancer (Edelman 1999; Fulton 1999). The reason for this may be the difficulties involved in research with such a vulnerable group. There are ethical problems in engaging in research with patients who have limited life expectancy.

Description of the intervention

Social support and psychological morbidity

According to Helgeson 1996 and Helgeson 1998, the advantageous effects of social support are thought to operate in two ways. First, network membership and social interaction directly increase positive thinking, emotions and behaviour (the 'main effect' hypothesis). Second, social networks help alleviate the damaging effects of stressful life events by providing potential coping resources such as emotional, informational and practical support (the 'stress-buffering' hypothesis).

These hypotheses contribute to the basis for many psychological interventions that have been proposed and sometimes evaluated in women with metastatic breast cancer. Marital or partner relationships may also be very important (Giese-Davis 2000). Pistrang 1995 found that for women with breast cancer (not metastatic) even a good helping relationship with another person did not compensate for a poor helping relationship with a partner, and that high levels of empathy and low levels of withdrawal were associated with good communication in the marital relationship. These factors may lead to a specific pattern of psychological issues in women with metastatic breast cancer, and interventions to improve such problems may need to be different in this patient group.

Psychosocial interventions for cancer patients

The recognition of psychosocial need has led to the development of psychosocial interventions for cancer patients. Fawzy 1995 reviewed the four main types of interventions: 1) education; 2) individual psychotherapy; 3) cognitive behavioural training; and 4) group interventions.

Educational interventions for individual patients with cancer aim to reduce feelings of inadequacy, confusion, helplessness and loss of control by providing information about the disease process, coping with the disease, and the resources available. Fawzy 1995 found that in three of the five studies of educational interventions, patients in the intervention groups had decreased anxiety. In one study, levels of depression improved in the intervention sample, and in another, compliance and appointment keeping improved.

Individual psychotherapy is a one-to-one interaction between therapist and patient. Its aims are to decrease distress and increase morale, self-esteem and coping by increasing the patient's sense of mastery over the situation and helping to overcome the practical problems encountered by the patient. Fawzy 1995 reviewed 13 studies where individual psychotherapy was at least part of the intervention. Ten of these suggested positive effects on coping, affect or quality of life.

Jacobsen 1998 defined the cognitive behavioural method as involving the identification and correction of thoughts, feelings and behaviours that contribute to the development and maintenance of symptoms. These behavioural and cognitive techniques are then employed to try to achieve psychological change. The review by Jacobsen 1998 showed that four out of the five studies showed significant improvements in coping and emotional well-being or functioning in intervention participants compared with controls. Other reviews by Fawzy 1995, Meyer 1995 and Trijsburg 1992 all showed that psychological interventions, including cognitive behavioural interventions, have positive effects on emotional and functional adjustment in cancer patients.

Group interventions centre on two types of support groups, those led by patients and those led by professionals. Leszcz 1998 proposed that group interventions offer an opportunity for the provision of social support, destigmatization, here-and-now practice of communication skills, instillation of hope, and added meaning in life through helping others. One of the principal manifestations of interventions on this basis is termed 'supportive-expressive group therapy' (Spiegel 1989), which includes building bonds, expressing emotions, detoxifying death and dying, redefining life priorities, increasing the support of friends and family, improving the doctor-patient relationship, and improving coping skills.

It is important to note, however, that only very few cancer patients attend available support programs when they are offered. Taylor 1986 found that less than 10% of patients attended support groups. Of these, the participants tended to be white and middle-class. Stalker 1990 found that of 200 patients nearly half said they would be interested in support groups, but only six had actually sought out group services. This raises significant questions about the generalisability of the evidence. It is equally important to note that these interventions have to date been almost exclusively theoretically or researcher driven and may not reflect approaches that women would find most important in addressing their psychological morbidity.

How the intervention might work

Psychological responses and survival

Psychological responses to diagnosis and psychosocial therapies have been reported to be associated with differences in survival among cancer patients. Greer 1979 reported on a prospective study of 69 patients with early breast cancer. The women's psychological response to diagnosis, assessed at three months post-operatively, appeared to be related to the five-year survival outcome. Watson 1999 repeated the study with 578 women, this time assessing women at four to 12 weeks and at 12 months post-diagnosis. The results indicated a significantly increased risk of all-cause mortality at five years in women with high scores on scales for depression or helplessness and hopelessness, although no significant differences were found for fighting spirit. However, Buddeberg 1991 observed, during their prospective three-year longitudinal study involving 107 women with breast cancer, that indicators of coping were less predictive in regard to the course of the disease as women change their coping strategies over time. They felt it was difficult to say whether measuring a woman's psychological response on two occasions early in the course of her disease can be considered to predict her responses over the next five years.

Petticrew 2002 reviewed 26 prospective studies to investigate the association between psychological coping styles and survival. Any positive findings were typically confined to small or methodologically flawed studies. In a more recent meta-analysis of 25 prospective studies, Satin 2009 found a 25% higher mortality rate for patients with cancer with depressive symptoms and a 39% higher mortality rate for those with major depression, after adjusting for prognostic factors. The evidence that was synthesised was however variable in terms of cancer types, and this was reflected in there being considerable heterogeneity in the findings.

Intervention studies have yielded conflicting results. Fawzy 1993 and Richardson 1990 showed survival benefits for patients in the intervention group, as did the Spiegel 1989 report of supportive-expressive group therapy for women with metastatic breast cancer. Other studies by Linn 1982 (on male patients only) and Ilnyckyj 1994 (with patients with a range of disease types) failed to show a survival benefit. Possible explanations include the use of different types of interventions as well as issues surrounding the selection of patients, further limiting their generalisability.

Why it is important to do this review

Despite conflicting results from studies to date, interest in this line of research has remained high. This may be partly because specific experiences in women with metastatic breast cancer (such as the effects on personal and body image and effects on relationships) are not replicated among the studies of patients with other types of cancer. Also, the effects of psychological interventions for women with metastatic breast cancer remain as yet uncertain. We therefore undertook a systematic literature review of the effects of psychological interventions for women with metastatic breast cancer.

The 2007 updated Cochrane review of psychological interventions for women with metastatic breast cancer, undertaken by Edwards 2008, showed from the meta-analysis of five primary studies (511 women) that there was very limited evidence of benefit arising from group psychological therapies. There was evidence of a short-term benefit for some psychological outcomes, but they were not sustained at follow-up. There was no evidence of longer survival times in women allocated to receive a psychological intervention. Edwards 2008 found the relative lack of available data in this field as the main limitation of the review. There was also heterogeneity in the outcome measures and follow-up durations, and a lack of data in suitable format for combination.

This 2012 review is an update of the Cochrane review first published in 2004 and previously updated in 2007. For this 2012 update, we used a modified search strategy that included more search terms for the different topics and a randomised controlled trial (RCT) search string. We also devised a new search strategy for CINAHL as the database changed from being SP to EBSCOhost.

Objectives

To assess the effects of psychological interventions on key psychosocial outcomes and survival for women with metastatic breast cancer.

Methods

Criteria for considering studies for this review

Types of studies

We included published studies using a randomised controlled trial (RCT) design along with cluster-randomised controlled trials with the unit of randomisation as an institute or definable organisation (for example general practice, hospital, college).

Studies were included even if they were not 'intention to treat', owing to the nature of the patient group under study and the likely high loss of follow-up data (McWhinney 2002).

We excluded results of studies with other designs for the formal assessment of intervention effects. However, such studies and consumer literature were considered in interpreting the results of the review, and they informed the review authors' conclusions.

Types of participants

Studies involving women with metastatic breast cancer (that is stages three or four). This included women with metastatic disease present at first diagnosis ('contemporaneous' metastatic disease) and those in whom metastatic disease was diagnosed after the initial diagnosis and treatment phases of disease ('delayed' metastatic disease).

We excluded those studies involving patients with non-metastatic breast cancer (that is stages one or two) or where there were no separate data on the stage of breast cancer. Similarly, studies involving interventions for people with a range of conditions (that is including people with other cancers) were excluded if specific data regarding women with metastatic breast cancer could not be extracted.

Types of interventions

The intervention had to be based on one or more psychological or (broader) psychosocial categories as described in the Background to the review (Fawzy 1995). These were:

  1. education;

  2. individual psychotherapy;

  3. cognitive behavioural training; and

  4. group interventions.

The characteristics of these interventions are as described in the Background to the review. They provide the basis and details for the inclusion and exclusion criteria.

Studies were excluded if data from multi-faceted interventions were reported and it was not possible to extract specific data about the psychological or psychosocial intervention effects themselves.

Types of outcome measures

Primary outcomes

The outcomes of interest were: (a) a change in participants in the intervention group according to psychological measures, and (b) survival. Survival was measured at one, five and 10 years after the intervention, if available.

Psychological outcome measures, duration and follow-up were according to what was reported as opposed to being set by a prior hypothesis. The range of psychological outcome measures used in this field of research is large (Deflorio 1995; Fawzy 1995; Petticrew 2002). Therefore, obtaining a uniform outcome measure was problematic. The review addressed the following categories of patient outcomes, as described below, and includes the types of measures most frequently reported in the included studies.

  1. Generic psychological outcome instruments: Profile of Mood States (POMS); Center for Epidemiologic Studies Depression Scale (CESD-20); Mental Adjustment to Cancer Scale; Rationality/Emotional Defensive Scale; Courtauld Emotional Control Scale; Impact of Event Scale; Coopersmith Self-Esteem Inventory; Weinberger Adjustment Inventory; Stanford Emotional Self Efficacy Scale Cancer.

  2. Quality of life: European Organisation for Research and Treatment of Cancer (EORTC) QLQ C30.

  3. Symptom scores: Pain Rating Scale.

  4. Condition-specific outcomes: Cancer Rehabilitation Evaluation System; Functional Living Index for Cancer.

  5. Relationship and social support measures: Yale Social Support Index; Duke UNC Functional Social Support Questionnaire; Marlowe Crowne Social Desirability Scale; Perceived Emotional Support; Supportive Care Needs Survey.

To minimise problems of some studies contributing unduly to the overall review findings, we specified the principal outcome measures of this review to be Profile of Mood States (POMS) and survival data. Studies were excluded if they included unvalidated or subjective outcome measures. 

We also sought to review any reported economic evaluation of the healthcare costs associated with the psychological or psychosocial intervention.

Search methods for identification of studies

Electronic searches

The search strategy for the review included the following, and involved the Trials Search Co-ordinator for the Cochrane Breast Cancer Group.

The search strategies from the original and updated (2007) review were modified and used to interrogate the following databases: 

  1. Cochrane Breast Cancer Group Specialised Register (June 2011);

  2. MEDLINE (OvidSP) (1947 to June 2011), see Appendix 1;

  3. EMBASE (OvidSP) (1947 to June 2011), see Appendix 2;

  4. PsycINFO (OvidSP) (1806 to June 2011), see Appendix 3;

  5. CINAHL (EBSCOhost) (June 2011), see Appendix 4.

There were no date or language limitations applied. These electronic databases were chosen as they were most likely to identify relevant studies.

The search strategy for the 2007 update was edited for several reasons: (A) each subject area of the core structure was expanded with more search terms; (B) search strings for RCTs were used for this additional search in accordance with SIGN 2011 (for EMBASE) and Eady 2008 (for PsycINFO); (C) the search strategy for CINAHL could no longer be conducted on Ovid and so a separate strategy was created for EBSCO. As indexing in this field of literature is poor (Cook 2001), the original review (Edwards 2004) and update (Edwards 2008) search strategy centred on topic areas without restrictions for study design. That is, RCT filters were not used as early trials may not have been reliably indexed in this field and conventional search filters could have been inaccurate.

Searching other resources

  1. Handsearching reference lists of previous trials, systematic reviews and meta-analyses. 

  2. Trials and research registers were also searched in July 2011. These included: (A) ClinicalTrials.gov (http://www.clinicaltrial.gov/) (see Appendix 5), Current Controlled Trials (http://www.controlled-trials.com/) (see Appendix 6) and the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) search portal (www.who.int/trialsearch) (see Appendix 7).

Data collection and analysis

Selection of studies

The initial selection of publications from search outputs was undertaken by two review authors independently (MM and ACS). Refer to Figure 1 for details of the selection process during the 2012 review update. Titles and abstracts of studies generated by the search strategies were checked for possible inclusion against the eligibility criteria (see data extraction sheet, Appendix 8). Disagreements over selection were resolved by discussion between MM and ACS, and AE if required. The full reports were retrieved for any study that appeared initially suitable for the review and further checked. If the full reports of the studies did not have the required level of data (for example separate data based on the type or stage of cancer) to decide on inclusion, or missing data could not be retrieved from author contact, the study was excluded on that basis. The reasons for non-inclusion of studies have been noted (see Characteristics of excluded studies).

Figure 1.

Study flow diagram for 2012 review update.

Data extraction and management

Using the data extraction sheet, all papers were appraised and analysed by two review authors (MM and ACS). The following information was extracted from each paper, where available.

  1. General information: title, authors, year, country of origin, ethical approval, funding, duration.

  2. Eligibility: RCT (type and method of randomisation), women with metastatic breast cancer, psychological intervention type.

  3. Participants: sample size calculation, method of recruitment, inclusion and exclusion criteria, demographic descriptors, similarity of groups at baseline, the numbers through the process (including numbers eligible, excluded, withdrawn, lost to follow-up).

  4. Interventions: description (including nature, duration, frequency), setting, type of provider, quality (including details of training for the providers and measures to gauge validity of intervention), details of the control group.

  5. Outcome measures: primary and secondary outcomes as reported in the study, methods and timing of assessing effects, data on the validity of outcome measurement tools used, follow-up details.

  6. Statistical methods used.

  7. Results.

  8. Risk of bias criteria (see below).

For included studies, if further information was required (for example stage of breast cancer, statistical outcome data), the designated author was contacted by e-mail for discrete data, where possible.

Assessment of risk of bias in included studies

The methodological quality of the studies was analysed according to the Cochrane Collaboration’s tool for assessing risk of bias (Higgins and Green 2011). It is a domain-based evaluation, as opposed to being a scale or a checklist. The following elements of each domain of bias were rated as ‘low risk’, ‘high risk’ or ‘unclear risk’.

  1. Random sequence generation (selection bias): assessment of the method used to generate the allocation sequence to produce comparable groups.

  2. Allocation concealment (selection bias): assessment of the method used to conceal the allocation sequence to determine whether intervention allocations could have been foreseen in advance of, or during, enrolment.

  3. Blinding of participants and personnel (performance bias): if applicable, assessment of measures used to blind study participants and personnel from knowledge of which intervention a participant received and whether it was effective.

  4. Blinding of outcome assessment (detection bias): assessment of measures used to blind outcome assessors from knowledge of which intervention a participant received and whether it was effective.

  5. Incomplete outcome data (attrition bias): assessment of the completeness of outcome data for each main outcome, including reporting of and reasons for attrition and exclusions from the analysis, and if there were any re-inclusions.

  6. Selective outcome reporting (reporting bias): assessment of  the possibility of selective outcome reporting by the authors.

  7. Other sources of bias.

The risk of bias of included studies was described, including statements to support the authors’ judgement, in the Characteristics of included studies table. Owing to the nature of the intervention studies, assessing the blinding of treatment assignment was not appropriate.

Measures of treatment effect

Studies in this review derive from different psychological or psychosocial interventions, in particular the split between cognitive behavioural therapies and the (group) supportive-expressive therapies. Nevertheless, the essential characteristic of the interventions (based on psychological or psychosocial elements) renders data combination appropriate.

Studies were first examined for descriptions of the intervention and qualitative synthesis. In view of the large number of different outcome instruments used in this field (see Types of outcome measures), data for the following outcomes were combined from three or more studies for meta-analyses:

  1. survival (one and five-year);

  2. Profile of Mood States (POMS) full version (10 to 12 months);

  3. pain (at one year).

These durations of follow-up were driven by what was available and reported, and not in relation to a priori hypotheses. We recognise that such analysis can be subject to data censoring, representing a potential weakness for conclusions to be drawn, but it reflects the nature and paucity of data available. Survival data were extracted at designated follow-up periods after randomisation, rather than from diagnosis of metastases, to evaluate effects subsequent to the intervention.

Dealing with missing data

Intention-to-treat analyses were conducted. Wherever possible, authors were contacted for data that could not be imputed from the published reports. Thereafter, if no further useable data were available, the study was not included in analysis.

Assessment of heterogeneity

Statistical heterogeneity was assessed using the I2 statistic [(Q-df)/Q] x 100% with Q representing the Chi2 statistic and df its degree of freedom (Higgins and Green 2011). This describes the percentage variability in effect estimates that is due to heterogeneity rather than chance (sampling error). 

An I2 value greater than 50% suggests substantial heterogeneity. As an additional measure of the strength of evidence for heterogeneity, the observed I2 value was interpreted along with the P value (assuming significance at less than 0.05) from the Chi2 test and the extent to which the confidence intervals (CIs) for individual studies overlapped when graphically depicted in forest plots. In cases of substantial heterogeneity, we attempted to explain the variation. Because of this, and in particular the disparate nature of the interventions (cognitive behavioural and supportive-expressive therapies), the random-effects model was employed. If the observed I2 value exceeded 80% then meta-analysis was not undertaken.

Assessment of reporting biases

An insufficient number of studies contributed to the outcomes so funnel plots were not constructed for inspection for signs of asymmetry, which is otherwise an indication of publication bias.   

Data synthesis

For meta-analyses, data were pooled from three or more RCTs if comparable measures were used. The effect size for studies was summarised as an odds ratio (OR). 

Review Manager 5.1 software (RevMan 2011) was used to generate descriptive and inferential statistics. The software was used to calculate single study and pooled ORs for the dichotomous outcomes (for example survival) with 95% confidence intervals (CIs). ORs were calculated using the Mantel-Haenszel method and a random-effect model in cases where the I2 measure was 50% or more, or in the case where there was clear clinical heterogeneity (for example differences in psychological or psychosocial therapies). A random-effects model meta-analysis, which is more conservative than the fixed-effect model, assumes that the effect sizes varies across the different studies but that they still follow the same normal distribution. 

An OR greater than one would suggest more participants benefited in the intervention group compared to the control group and was graphically represented on the right of the axis of the forest plot.

Subgroup analysis and investigation of heterogeneity

Potential clinical variations causing heterogeneity were explored by subgroup analysis stratified according to the type of psychological or psychosocial intervention (principally cognitive behavioural versus supportive expressive therapies). 

Sensitivity analysis

A sensitivity analysis was conducted to evaluate possible causes of methodological heterogeneity. This was based on repeated analysis excluding trials regarded at highest risk of bias. This analysis focused on two domains of the 'Risk of bias' tool in which weaknesses are especially likely to be significant in this field of palliative care research: (a) high post-randomisation attrition rate (more than 40%); and (b) unblinded outcome assessment. We also considered sensitivity analysis in the case of a high risk of selection bias consequent to poor randomisation.

Results

Description of studies

Details of the included individual studies are summarised in the Characteristics of included studies table. The excluded studies are listed with reasons for their exclusion in the Characteristics of excluded studies. Studies are identified by the first author and publication year. 

Results of the search

In the 2012 review update, the electronic search strategies identified 860 new titles (see Figure 1) (for previous versions of this review refer to Edwards 2008). Two further titles were identified from other references. Seven hundred and sixty-two titles were not relevant to this review and were excluded. Of the 100 remaining titles, 55 duplicates were excluded. Forty-five full text articles were further assessed for eligibility; 27 were excluded as they did not satisfy one or more of the inclusion criteria. Ten studies (18 publications) met the inclusion criteria. Seven studies were included in the meta-analyses. Outcomes that did not meet the criteria for meta-analysis were described from nine studies.

Included studies

Ten studies published between 1989 and 2010 were included in the review, involving 1378 women with metastatic breast cancer. All 10 RCTs used individuals as the unit of randomisation. Individual studies are described in the Characteristics of included studies, and a summary of the studies is presented in Table 1.

Table 1. Summary of characteristics of included studies
  1. a Only difference between the trial groups was the intervention itself.

    Interventions: CBT (Cognitive Behavioural Therapy): SEGT (Supportive Expressive Group Therapy).

    Outcomes: CARES (Cancer Rehabilitation Evaluation System); CECS (Courtauld Emotional Control Scale); DR (Defensive Repression); DUFSS (Duke UNC Functional Social Support); EORTC C-30 (quality of life); FLIC (Functional Living Index for Cancer); HIP (Hypnotic Induction Profile); IES (Impact of Events Scale); MCDS (Marlowe Crowne Social Desirability Scale); MAC (Mental Adjustment to Cancer); PES (Perceived Emotional Support); POMS (Profile of Mood States); QoL (quality of life measure); RED (Rationally Emotional Defensiveness); SESESC (Stanford Emotional Self Efficacy Scale Cancer); SCNS (Supportive Care Needs Survey); WAI (Weinberger Adjustment Inventory).

Author, yearRecruitmentInterventionSettingControlaProviderOutcome
Aranda 2006Hospital

Face-to-face

Telephone

OutpatientsStandard careNurse

EORTC QLQ-30

SCNS

Cunningham 1998HospitalCBTUnclear

Standard care

Workbooks

Audiotapes

Telephone calls

Psychologist

Survival

POMS

FLIC

MAC, RED, DR DUFSS, MCDS

Edelman 1999Hospital, CommunityCBTUnclearStandard carePsychologist

Survival

POMS

Anxiety

Self esteem

Goodwin 2001HospitalSEGTClinicEducational material

Psychologist/

Psychiatrist/

Social worker

Survival

EORTC QLQ-30

POMS

Anxiety

Pain

Gotay 2007UnclearPeer delivered telephone interventionHomeUnspecifiedBreast Cancer survivor

CES-D

CARES-SF

Kissane 2007HospitalSEGTOutpatientsRelaxation classes

Psychologist/

Psychiatrist/

Social worker

Survival

EORTC QLQ-30

IES

Mini-MAC

Classen 2001Hospital, CommunitySEGTUnclear

Educational material

1 year consumer health library membership

Psychologist/

Psychiatrist/

Social worker

Survival

POMS

Pain

CECS, WAI

SESESC

IES

HIP

Low 2010

Hospital,

Another study

Emotional expressive writingParticipant choiceUnspecifiedResearch Assistant

IES

CES-D

PSQI

PES

Scholten 2001UnclearCBTTherapists officeUnspecifiedPsychologist/ psychiatristsQoL
Spiegel 1989HospitalSEGTUnclearStandard care

Psychiatrist/

Social worker,

Breast Cancer survivor

Survival

POMS

Pain

Recruitment and setting

Four studies were conducted in the United States of America (Classen 2001; Gotay 2007; Low 2010; Spiegel 1989), three in Australia (Aranda 2006; Edelman 1999; Kissane 2007), two in Canada (Cunningham 1998; Goodwin 2001) and one in Austria (Scholten 2001). The recruitment for eight studies was from secondary healthcare settings (Aranda 2006; Classen 2001; Cunningham 1998; Edelman 1999; Goodwin 2001; Kissane 2007; Low 2010; Spiegel 1989). Two studies also used media publicity in the wider community (Classen 2001; Edelman 1999) and one trial recruited from a larger, descriptive research study in which one of the authors already had involvement (Low 2010). It was unclear from where Gotay 2007 or Scholten 2001 recruited patients.

Outpatient clinics were used to deliver the interventions in two studies (Aranda 2006; Kissane 2007). One study each used a healthcare setting (Goodwin 2001), participants’ home (Gotay 2007), therapist’s office (Scholten 2001) or a venue of the participants’ choice (Low 2010). The intervention settings were unclear in four studies (Classen 2001; Cunningham 1998; Edelman 1999; Spiegel 1989).

Intervention

The interventions from seven studies fell into two categories: cognitive behavioural (Cunningham 1998; Edelman 1999; Scholten 2001), and supportive-expressive group therapies (Classen 2001; Goodwin 2001; Kissane 2007; Spiegel 1989); with a third reserved for three studies that did not have a common intervention type (Aranda 2006; Gotay 2007; Low 2010).

Cognitive behavioural group interventions

Cunningham 1998 used a group-based intervention that comprised of three components. The first component was a course of 35 weekly group meetings (typically involving five to eight participants per group) lasting approximately two hours each, and covering issues including support, problem solving and emotional reactions to the disease. The second component was a 20-week course of cognitive behavioural assignments to be completed at home and then discussed in the group setting. These covered thought and behaviour monitoring and introduced substitutes for patterns identified as problematic or resistant to change. In the third component, participants were invited to an 'Intensive Weekend Coping Skills Training course', which addressed skills such as relaxation, stress control, positive affirmations and goal setting.

The trial by Edelman 1999 involved eight weekly sessions of Cognitive Behaviour Therapy (CBT) and a family night, followed by three further sessions one month apart. Participants were taught a range of cognitive and behavioural coping strategies, including thought monitoring, cognitive restructuring, use of coping statements, effective communication, goal setting and relaxation. Sessions comprised information sessions on one of these themes followed by group discussion. Homework was also set each week, to be discussed at subsequent meetings.

The intervention used by Scholten 2001 used cognitive and behavioural approaches and included such strategies as problem solving, regaining control, and setting new goals for the future. They focused on the patients’ coping strategies, self-esteem and femininity, overcoming feelings of helplessness, negative thoughts and depression, and promotion of a fighting spirit. There were no set number of sessions or limits on their duration.

Supportive-expressive group therapy (SEGT)

There were four separate studies in this category. For Spiegel 1989, patients in the intervention group were invited to weekly group support sessions, lasting 90 minutes each, for one year. The topics and nature of the support covered: building bonds, expressing emotions, detoxifying death and dying, redefining life priorities, increasing the support of friends and family, improving the doctor-patient relationship, and improving coping skills. There were three groups in the intervention arm; one group was also selected to receive specific input regarding self-hypnosis for pain control.

Classen 2001 was modelled on the first trial (Spiegel 1989) with the intervention groups divided into three, each taking part in weekly, 90-minute group therapy for a year. However, the groups continued for ‘several years’ and for as long as eight years. The supportive environment for participants included encouragement to express difficult emotions, confront problems, strengthen relationships with family and physicians, and find enhanced meaning in their lives. Self-hypnosis for pain was again included in this intervention. All participants also received educational materials after each assessment.

For Goodwin 2001, the intervention comprised of weekly, 90-minute meetings for groups of eight to 12 women who had been asked to attend for at least one year. The therapy fostered support among the group members and encouraged expression of emotions and sharing of experiences about the effects of the illness on self and relationships with family or others. The participants also discussed strategies for coping and communicating. A monthly 90-minute session was available for family members and friends. All participants also received educational materials after each assessment.

Kissane 2007 used an intervention that involved weekly, 90-minute group therapy for a year. The intervention aimed to improve relationships with family, friends and physicians, and to create a network of support and foster coping skills. The women were encouraged to have interactions outside of the group intervention setting. The women were also offered three further one-hour relaxation classes over consecutive weeks, annually.

Other interventions: individual-based

For Aranda 2006, the trial’s psycho-educational intervention comprised two components. The first was a one-hour face-to-face meeting with a breast cancer nurse who actively listened to the participant’s concerns, offered empathy and support, and elicited understanding of their situation (orientation). Information provided was tailored to the woman’s individual concerns (tailored response). Coaching and rehearsal of self-care, stress reduction and communication strategies to tackle the participant’s main problem were provided (coaching and practising self-care and communication strategies). Each woman was also given relevant information cards on self-care and communication strategies and a copy of her personal self-care plan. Women were provided with a relaxation CD. The second session was conducted over the telephone one week later to ask whether the suggested strategies had ameliorated the concerns; to elicit and respond to any concerns remaining; to reinforce or modify planned self-care strategies or introduce new ones; and to prompt for further questions or new concerns.

Gotay 2007 evaluated a weekly peer-delivered telephone intervention of four to eight sessions (one or two calls per week), each focusing on the quality of life and concerns as expressed by the participant. After the first session, participants were sent standardised information pamphlets.  

The Low 2010 trial employed an emotionally expressive writing intervention where the women, in a setting of their choice (for example home), would write about cancer-related emotions during four 20-minute sessions. The participants received information for the writing exercises beforehand. At the beginning of each session they were called by a research assistant, read the instructions, and then telephoned 20 minutes later to stop writing.

Control group

Although it was unclear in one study (Low 2010), there was diversity among the remaining included studies concerning the nature of the comparator to the psychological intervention. Four studies described the control group as having standard care (Aranda 2006; Cunningham 1998; Edelman 1999; Spiegel 1989). In addition, Cunningham 1998 offered a workbook as part of a home study cognitive behavioural package and two audiotapes for self-directed practice of relaxation and mental imaging; this constituted the material available to the intervention group as part of their ‘Intensive’ weekend programme. However, the control group were not permitted to attend the course itself.

Although the control conditions were not specified in three studies (Gotay 2007; Low 2010; Scholten 2001), it would be reasonable to assume standard care was provided.  

Educational material was provided in two studies (Classen 2001; Goodwin 2001). Classen 2001 also offered one-year membership to a local consumer health library. Relaxation classes were offered in one study (Kissane 2007).

The studies did not report any difference between the trial groups except for the interventions.

Provider

The intervention was delivered by a psychologist, psychiatrist or social worker in seven studies (Classen 2001; Cunningham 1998; Edelman 1999; Goodwin 2001; Kissane 2007; Scholten 2001; Spiegel 1989). For the telephone-based intervention study, a trained research assistant was used (Low 2010). Breast cancer nurses were the providers in Aranda 2006. Counsellors in one study were mostly survivors of breast cancer recurrence (Gotay 2007). A co-provider in one study also had a history of breast cancer (Spiegel 1989).

Training of the provider

Four studies referred to training the providers of the intervention, but they varied in their descriptions of duration, content, trainer, format and adherence to guidelines (Aranda 2006; Goodwin 2001; Gotay 2007; Kissane 2007). Breast care nurses attended two training days, delivered by a team including the authors, that included the medical and psychosocial management of women with advanced breast cancer and role playing scenarios (Aranda 2006). Providers received training according to a published therapy manual (Goodwin 2001; Kissane 2007) as well as observation of training videotapes and attendance at a workshop, both by David Spiegel (Kissane 2007).

Only two studies reported monitoring the provider (Goodwin 2001; Kissane 2007). Providers were given feedback to ‘promote consistent delivery of the intervention’ by reviewing videotaped sessions monthly and during two-day workshops every nine to 12 months that were attended by David Spiegel or his associate Catherine Classen (Goodwin 2001). Therapists and supervisors met fortnightly and also quarterly over six years to receive feedback about fidelity and to promote uniform application of the therapy (Kissane 2007).

Outcomes
Survival outcome

Six studies that published outcomes pertaining to survival presented the data at one and five-year follow-up (Classen 2001; Cunningham 1998; Edelman 1999; Goodwin 2001; Kissane 2007; Spiegel 1989), and four also at 10 years of follow-up (Classen 2001; Cunningham 1998; Edelman 1999; Spiegel 1989). Ten-year survival for two studies were from diagnosis of metastases rather than study entry or randomisation (Cunningham 1998; Edelman 1999).

Three studies described survival from randomisation (Cunningham 1998; Goodwin 2001; Kissane 2007) and three from study entry (Edelman 1999; Classen 2001; Spiegel 1989). 

Psychological outcomes

Five studies used POMS (Classen 2001; Cunningham 1998; Edelman 1999; Goodwin 2001; Spiegel 1989). However, different versions of the scale were used: short version (Cunningham 1998) and full version (Classen 2001; Edelman 1999; Goodwin 2001; Spiegel 1989). They also used a different duration of follow-up for evaluation: six months (Classen 2001; Edelman 1999), eight months (Cunningham 1998), 10 months (Spiegel 1989) and 12 months (Classen 2001; Goodwin 2001).

Anxiety data were extracted primarily from the Tension-Anxiety subscale of the POMS. Two studies provided data: at six months (Edelman 1999) and 12 months (Goodwin 2001).

Four studies used various other measures to evaluate generic psychological outcomes: Mental Adjustment to Cancer (MAC) scale, Rationally Emotional Defensiveness scale and Defensive Repression (Cunningham 1998); mini-MAC (Kissane 2007); Coopersmith Self Esteem Inventory adult form (Edelman 1999); Weinberger Adjustment Inventory, Stanford Emotional Self Efficacy Scale Cancer, and Courtauld Emotional Control Scale (Classen 2001).

Impact of Event Scale (IES), a measure of trauma symptoms, was used in three studies (Classen 2001; Kissane 2007; Low 2010); although, in contrast to the two evaluating ‘total’ IES scores, Low 2010 assessed the IES domain ‘intrusion’. Another patient-based outcome measure, Life Event Scale, was used by one study (Classen 2001).

Quality of life

Health-related quality of life was used in four studies: three studies used the EORTC C-30 measure (Aranda 2006; Goodwin 2001; Kissane 2007), and one study that measured quality of life using semi-structured interviews and the visual analogue scale (Scholten 2001). 

Pain

Pain was assessed in three studies (Classen 2001; Goodwin 2001; Spiegel 1989). Goodwin 2001 and Spiegel 1989 used a visual analogue scale ranging from zero (no pain) to 10 (worst imaginable pain). For Classen 2001, it relied on using the Likert-like scales for assessing pain at that moment, its constancy and frequency.  

Condition-specific outcome measures

Two studies used condition-specific outcome measures: Functional Living Index measure (Cunningham 1998) and Cancer Rehabilitation Evaluation System Short Form (Gotay 2007).

Relationship and social support measures

Four studies used various measures to assess relationships and social support: Supportive Care Needs Survey (Aranda 2006); Duke UNC Functional Social Support questionnaire, and Marlowe Crowne Desirability Scale (Cunningham 1998); Yale Social Support Index and Single Item Measure of Social Support (Classen 2001); and Perceived Emotional Support (Low 2010).

Economic evaluation

No studies described any cost-effectiveness analysis for their chosen psychological or psychosocial intervention.

Other

One study used a symptom score: the Pittsburgh Sleep Quality Index (Low 2010). One study used the Hypnotic Induction Profile (Classen 2001). Two studies used measures to rate the intervention itself (Classen 2001; Kissane 2007).

Risk of bias in included studies

The risk of bias in included studies has been summarised in Characteristics of included studies and Figure 2.

Figure 2.

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

Allocation

Three studies did not describe their methods of sequence generation and so were rated as unclear risk (Edelman 1999; Scholten 2001; Spiegel 1989). Four studies provided insufficient information on allocation concealment and were similarly rated as having an unclear risk for selection bias (Edelman 1999; Goodwin 2001; Scholten 2001; Spiegel 1989). 

Blinding

Blinding of participants and providers of the intervention was not feasible in such trials. However, in terms of blinding outcome assessors, insufficient information was available to make judgements in all but two studies (Goodwin 2001; Spiegel 1989) and so were described as unclear risk (Aranda 2006; Classen 2001; Cunningham 1998; Edelman 1999; Gotay 2007; Kissane 2007; Low 2010; Scholten 2001). 

Incomplete outcome data

The data were not always clearly presented to establish what the numbers and reasons were for attrition and exclusions. For two studies there was insufficient information in the studies and related published articles to permit judgement on attrition bias (Cunningham 1998; Spiegel 1989). The method of follow-up was unclear (Cunningham 1998) and there were no discussions of the number of patients who completed follow-up questionnaires (Spiegel 1989). As Scholten 2001 reported a dropout rate of 60% affecting the 20 patients randomised to the intervention, it was deemed that this could induce a clinically relevant bias in the intervention effect estimate and thus was judged as high risk of attrition bias.

Selective reporting

It was clear from all of the published reports that the studies’ pre-specified expected outcomes of interest were reported. As a result they were all judged to be at low risk of reporting bias.

Other potential sources of bias

Several studies were judged to be ‘unclear’ for bias that was not part of the criteria already described above, but could be grouped according to intervention fidelity, baseline differences and other.

Intervention fidelity

There was a paucity of information concerning the fidelity of the interventions, not only in terms of method specificity but also the training in and maintenance of the intervention for all but two studies (Goodwin 2001; Kissane 2007). Fidelity of the intervention was attempted by videotaping the sessions and reviewing them weekly (Goodwin 2001) and by supervision throughout the study period (Kissane 2007). However, no trial provided information on what methods (for example scale) were employed to gauge the provider’s in-session proficiency or processes to address measures below particular thresholds of proficiency.

Baseline differences between intervention and control groups

In Aranda 2006, at study entry a greater proportion of patients in the intervention arm were undergoing current radiotherapy (93% versus 73%) and more patients in the usual care arm of the study had children who were minors (38% versus 14%).

In Goodwin 2001, a potential risk of bias existed due to differences between the groups at the time of diagnosis. The intervention group were younger than those in the control group at the time of diagnosis (mean age 44.0 years versus 46.6 years) and were less likely to be postmenopausal (14.6% versus 33.8%). They were also more likely to have had involvement of axillary nodes (70.9% versus 50.7%). The tumours of the women in the intervention group were more likely to have had oestrogen receptors (69.0% versus 54.5%) and progesterone receptors (63.9% versus 42.9%), and the use of adjuvant chemotherapy was more common in the intervention group (75.9% versus 55.6%).

In the trial by Gotay 2007, the arms were balanced at the time of randomisation but with two exceptions: more patients in the intervention group received chemotherapy (71% versus 57%) and more control patients received hormone therapy for recurrence (36% versus 49%).  

For Spiegel 1989, there was a ‘nearly significant’ difference in staging at initial diagnosis, with initial staging favouring the intervention group as it took place on average 59.8 months before the study began.

Other

For Aranda 2006, it was unclear whether participants recruited from public hospitals would have different health seeking behaviours compared to those recruited from the private hospital. 

In Classen 2001, they found a difference in terms of ethnicity in the samples at follow-up. There were more minority women in the control group, scoring higher for Suppression, Restraint, Repressive–Defensiveness, and Emotional Self-Efficacy. It was unclear whether women of different ethnic minority origins would have different health behaviours.

Effects of interventions

Survival outcome: one year

The pooled effect across six studies (n = 856) favoured the psychological interventions with an OR of 1.46 (95% CI 1.07 to 1.99) (Classen 2001; Cunningham 1998; Edelman 1999; Goodwin 2001; Kissane 2007; Spiegel 1989). A random-effects model was used as data from different intervention types were being pooled. There was no evidence of statistical heterogeneity (I2 = 0%). See Figure 3.

Figure 3.

Forest plot of comparison: 1 Psychological intervention versus usual practice, outcome: 1.1 Survival at one year.

Survival outcome: five years

The meta-analyses of four studies (n = 666) did not indicate a survival benefit for the psychological intervention group compared to the control group at five years, OR 1.03 (95% CI 0.42 to 2.52) (Classen 2001; Goodwin 2001; Kissane 2007; Spiegel 1989). A random-effects model was used as data from different intervention types were being pooled. There was evidence of statistical heterogeneity, I2 = 70%. Two studies were not considered in the meta-analysis of the five-year survival data as their data were from point of metastasis, not from point of randomisation (Cunningham 1998; Edelman 1999). See Figure 4.

Figure 4.

Forest plot of comparison: 1 Psychological intervention versus usual practice, outcome: 1.2 Survival at five years.

Survival outcome: 10 years

The 10-year survival outcomes did not meet the criteria for meta-analysis as there were only two studies with survival data from study entry or randomisation (Cunningham 1998; Edelman 1999; Analysis 1.3) (see Measures of treatment effect). See below, 'Description of non meta-analysed outcomes'.

Psychological outcome: Profile of Mood States (POMS) full version 10 to 12 months

Pooling three studies, all of supportive/expressive group therapy (SEGT) (n = 257), there was no statistically significant evidence of short-term outcome benefit from the intervention on psychological measures using POMS (Goodwin 2001; Classen 2001; Spiegel 1989), mean difference -11.54 (95% CI -28.68 to 5.60). There was evidence of heterogeneity (I2 = 71%) and a random-effects model was used. See Figure 5.

Figure 5.

Forest plot of comparison: 1 Psychological intervention versus usual practice, outcome: 1.11 Profile of Mood States (full version) at 10 to 12 months.

Pain at 12 months                                      

Pooling three studies, all of the SEGT (n = 279), showed improved pain scores with a mean difference of -0.58 (95% CI -0.98 to -0.18) (Classen 2001; Goodwin 2001; Spiegel 1989) (Analysis 1.13). There was no evidence of heterogeneity (I2 = 0%) and a fixed-effect model was used. See Figure 6.

Figure 6.

Forest plot of comparison: 1 Psychological intervention versus usual practice, outcome: 1.13 Pain at one year.

Sensitivity analysis

Two of the studies with outcomes that featured in the meta-analyses (Edelman 1999; Spiegel 1989) were found to have an unclear risk of selection bias. A sensitivity analysis was performed on the survival meta-analyses, removing data from both studies for the one-year analysis and removing Spiegel’s data from the five-year analysis.

While removing both sets of data from the survival meta-analysis at one-year increased the overall treatment effect (OR 1.74; 95% CI 1.21 to 2.48), removing Spiegel 1989 at five years caused a decrease in treatment effect (OR 0.80; 95% CI 0.40 to 1.64).

Description of non meta-analysed outcomes

The outcomes that did not meet the criteria for meta-analyses are described below and summarised in Table 2.

Table 2. Summary of outcomes (not meta-analysed)
  1. X = no significant difference

    √ - significant difference

    CARES (Cancer Rehabilitation Evaluation System); CES-D (depression measure); CECS (Courtauld Emotional Control Scale); DUFSS (Duke UNC Functional Social Support); FLIC (Functional Living Index for Cancer); IES (Impact of Events Scale); MCDS (Marlowe Crowne Social Desirability Scale); MAC (Mental Adjustment to Cancer); PSQI (Pittsburgh Sleep Quality Index); POMS (Profile of Mood States); QoL (quality of life measures); RED (Rationally Emotional Defensiveness); SCNS (Supportive Care Needs Survey).

Author, yearSurvivalPsychological outcome measuresCondition specificRealtionship & social supportOther
AnxietyCES-DPOMSSelf-esteemREDMACMini MACCECSQoLIESFLICCARES SFSCNSDUFSSMCDSPSQI
Aranda 2006         X   X   
Cunningham 1998X  X XX    X  XX 
Edelman 1999XX XX            
Goodwin 2001X       X       
Gotay 2007  X         X    
Kissane 2007X       

X

Emotional, physical

Social functioning

X      
Classen 2001X  X           
Low 2010  X       

X

Intrusion domain

     X
Scholten 2001         X       
Spiegel 1989                
1. Survival - mean and median

Spiegel 1989 found a highly significant difference between groups in terms of survival (P < 0.0001), with mean survival times from study entry of 18.9 months (SD 10.8) for the control group and 36.6 months (SD 37.6) for the intervention group.

Cunningham 1998 found no significant difference between treatment groups in terms of survival (P = 0.35), with median survival times from randomisation of 28.2 months (interquartile range (IQR) 25.8 to 51.7) for the control group and 23.6 months (IQR 18.9 to 34.9) for the intervention group.

Classen 2001 found no significant difference between treatment groups in terms of survival (P = 0.731), with median survival times from study entry of 33.3 months in the control group and 30.7 months in the intervention group.

Edelman 1999 found no significant difference between treatment groups in terms of survival (P > 0.1), with median survival times from recruitment of 1.07 years (IQR 0.63 to 1.48) for the control group and 0.97 years (IQR 0.59 to 1.46) for the intervention group.

Goodwin 2001 found no significant difference between treatment groups in terms of survival (P = 0.72), with median survival times from randomisation of 17.6 months in the control group and 17.9 months in the intervention group. The hazard ratio for death during the study in the intervention group compared to the control was 1.06 (95% CI 0.78 to 1.45, P = 0.72).

Kissane 2007 found no significant difference between treatment groups in terms of survival (P = 0.60), with median survival times of 18.3 months in the control group and 24.0 months in the intervention group. The hazard ratio for death during the study in the intervention group compared to the control was 0.92 (95% CI 0.69 to 1.24, P = 0.60).

2. Psychological outcomes
Anxiety

Edelman 1999 found no significant difference between treatment groups in terms of anxiety change scores at six months (P = 0.674), with mean changes of -0.28 (SD 6.90) for the control group and 0.39 (SD 5.56) for the intervention group.

Goodwin 2001 found a significant difference between treatment groups in terms of anxiety scores at 12 months, after adjusting for baseline scores (P = 0.002). The 12-month change score was 1.9 (SD 5.7) for the control group and -1.5 (SD 6.9) for the intervention group (favours the intervention group).

The Centre for Epidemiologic Studies – Depression Scale (CES-D)

Gotay 2007 found no significant difference between treatment groups in terms of reported depression at three months, adjusted for baseline scores (P = 0.50), with depressive symptoms exhibited in 40% of control patients and 47% of intervention patients (OR adjusted for baseline depression 0.80; 95% CI 0.41 to 1.54).

Low 2010 found no significant difference between treatments groups in terms of reported depression at three months, with follow-up mean scores (adjusted for baseline scores) of 13.2 (SD 8.24) for the control group and 12.8 (SD 8.24) for the intervention group.

Profile of Mood States (POMS)

Cunningham 1998 found no significant difference between treatment groups in terms of POMS (short version) change scores at eight months (P > 0.05), with mean changes of 1.54 (SD 11.05) for the control group and 0.18 (SD 11.94) for the intervention group.

Classen 2001 found no significant difference between treatment groups in terms of POMS (full version) change over 12 months (P = 0.20), with mean slopes of -0.14 for the control group and -0.21 for the intervention group.

Edelman 1999 found no significant difference between treatment groups in terms of POMS (full version) change scores at six months (P = 0.818), with mean changes of -0.15 (SD 28.55) for the control group and -1.87 (SD 30.36) for the intervention group.

Self-esteem

Edelman 1999 found no significant difference between treatment groups in terms of self-esteem change scores at six months (P = 0.341), with mean changes of 2.97 (SD 12.86) for the control group and -0.42 (SD 14.87) for the intervention group.

Rationality/Emotional Defensive scale (RED)

Cunningham 1998 found no significant difference between treatment groups in terms of RED change scores at eight months (P > 0.05), with mean changes of 1.89 (SD 4.38) for the control group and 1.32 (SD 3.98) for the intervention group. The findings were similarly non-significant at four and 14 months.

Mental Adjustment to Cancer (MAC)

Cunningham 1998 found no significant difference between treatment groups in terms of MAC change scores for any of its domains (fighting spirit, helplessness, anxious pre-occupation, fatalism and anxiousness) at four, eight or 14 months.

Mini-Mental Adjustment to Cancer (Mini-MAC)

Kissane 2007 found there was reduced helplessness-hopelessness for SEGT at all five time points for the intervention group compared to the controls (F = 4.20, P = 0.004).

Courtauld Emotional Control Scale (CECS)

Classen 2001 found a significant difference between treatment groups in terms of CECS (P = 0.01), with mean slopes of 0.03 (SD 0.86) for the control group and -0.29 (SD 0.78) for the intervention group.

Impact of Event Scale (IES)

Classen 2001 found a significant difference between treatment groups in terms of IES total change scores over 12 months (P = 0.03), with mean slopes of -0.29 for the control group and -0.61 for the intervention group.

Kissane 2007 found no significant difference between treatment groups (for participants with data for at least two of the five time points – ranging from six to 24 months) in terms of IES total scores (F = 3.33, P = 0.07).

Low 2010 found no significant difference between treatments groups in terms of the intrusion domain of the IES at three months, with follow-up mean scores (adjusted for baseline scores) of 10.1 (SD 0.96) for the control group and 8.7 (SD 0.94) for the intervention group.

3. Quality of Life
EORTC QLQ-C30

Goodwin 2001 found no significant difference between treatment groups in terms of the global EORTC QLQ-C30 score at 12 months (P = 0.76), with mean scores of 58.8 (SD 23.5) for the control group and 59.7 (SD 20.2) for the intervention group. No significant differences between groups were observed for any of the domains.

Aranda 2006 found no significant difference between treatment groups in terms of overall EORTC QLQ-C30 change scores (P > 0.05), with mean changes of -26.2 (SD 42.7) in the control group and -28.1 (SD 36.1) in the intervention group at one month. Mean changes of -33.6 (SD 36.6) in the control group and -22.6 (SD 39.1) in the intervention group were observed at three months. No significant differences between groups were observed for any of the domains.

Kissane 2007 found a significant difference between treatment groups (for participants with data for at least two of the five time points, ranging from six to 24 months) in terms of the social functioning domain of the EORTC QLQ-C30 (F = 4.56, P = 0.03) but found no significant differences in terms of the emotional (F = 2.00, P = 0.16) and physical (F = 0.18, P = 0.68) functioning domains of the EORTC QLQ-C30.

Visual analogue scale

Scholten 2001 evaluated patients’ and controls’ judgement of quality of life by two visual analogue scales (health-related quality of life and non-medical aspects of quality of life) and did not find any significant differences at three or six months.

4. Condition-specific outcome measures
Functional Living Index for Cancer (FLIC)

Cunningham 1998 found no significant difference between treatment groups in terms of FLIC change scores at eight months (P > 0.05), with mean changes of 2.48 (SD 24.06) for the control group and 2.91 (SD 24.48) for the intervention group.

Cancer Rehabilitation Evaluation System Short Form Psychosocial Distress Score (CARES-SF)

Gotay 2007 reported that 77% of control and 66% of intervention patients exhibited psychosocial distress at three months and that “in the multivariate model, there was no evidence that the intervention differentially impacted the outcome (OR 0.80; 95% CI 0.40 to 1.54)”.

5. Relationship and social support measures
Supportive Care Needs Survey (SCNS)

Aranda 2006 found that after adjusting for baseline score, there were no significant differences between the intervention and standard care group for the change in SCNS domain score in any of the domains.

Duke UNC Functional Social Support (DUFSS)

Cunningham 1998 found no significant difference between treatment groups in terms of DUFSS change score at eight months (P > 0.05), with mean changes of -1.11 (SD 3.82) for the control group and -0.09 (SD 6.23) for the intervention group. The findings were similarly non-significant at four and 14 months.

Marlowe Crowne Social Desirability (MCSD)

Cunningham 1998 found no significant difference between treatment groups in terms of MCSD change score at eight months (P > 0.05), with mean changes of 1.41 (SD 3.83) for the control group and 1.36 (SD 4.32) for the intervention group. The findings were similarly non-significant at four and 14 months.

6. Other
Pittsburgh Sleep Quality Index (PSQI)

Low 2010 report adjusted mean scores for PSQI at three-months follow-up and revealed no effect for intervention versus control condition: mean 7.1 (standard error (SE) 0.51) versus 6.6 (SE 0.51) (P = 0.01).

Discussion

Summary of main results

Effects of psychological interventions

Survival

The overall effect after meta-analysis suggests that psychological therapies for women with metastatic breast cancer exert a statistically significant albeit modest survival benefit at one year (OR 1.46, 95% CI 1.07 to 1.99). The finding is based on six studies involving 856 women. Subgroup analysis revealed that supportive-expressive therapies were also associated with survival benefit at one year (OR 1.56; 95% CI 1.07 to 2.28). The OR for cognitive behavioural therapies was consistent with a small benefit at one year, but this was not statistically significant (OR 1.14; 95% 0.61 to 2.15).

Pooled outcomes from four studies for survival at five-years follow-up did not show a statistically significant benefit from the interventions (OR 1.03; 95% CI 0.42 to 2.52).

Pain

Meta-analyses of three studies for pain at 12 months showed improved pain scores with a mean difference of -0.58 (95% CI -0.98 to -0.18) favouring the psychological intervention (all were Supportive-Expressive Group Therapy (SEGT)). 

Psychological outcomes

Three studies of group psychological therapies for women with metastatic breast cancer showed, according to the Profile Of Mood States (POM) (full version), evidence of a small benefit arising from these interventions at 10 to 12 months, but this was not statistically significant (mean difference -11.54; 95% CI -28.68 to 5.60).

In two Supportive-Expressive Group Therapy (SEGT) studies, there was some evidence of short-term benefit favouring the intervention group for some psychological outcomes, but which did not meet the review’s criteria for meta-analyses. Kissane 2007 found that there was reduced helplessness-hopelessness (a domain of the mini-Mental Adjustment to Cancer (mini-MAC)). They also found a significant difference in terms of the social functioning domain of the EORTC QLQ-C30. Classen 2001 found significant differences in terms of the Impact of Event Scale (IES) total change scores over 12 months and the Courtauld Emotional Scale (CECS).

The majority of other psychological, quality of life and social functioning outcome differences were not statistically significant between the intervention and control groups (see Table 2).

Quality of the evidence

The trials were heterogeneous according to clinical and methodological characteristics and this could affect the direction and magnitude of effects. Therefore, further to the sample sizes, there are specific points to consider when comparing across trials in order to infer the effect attributable to psychological therapies with any degree of precision, including: (i) the intervention type and fidelity; (ii) participants’ co-morbid state; (iii) type of control conditions; and (iv) quality of studies. 

Psychological interventions, type and fidelity

Confidence that psychological interventions are linked to the desired outcome can only be justified once there is adherence or fidelity to the intervention. The included trials had different types of psychological interventions and, where reported, there was a range of psychological intervention training and intervention quality maintenance methods making direct comparisons across the studies problematic. Studies were reviewed for evidence of both skill acquisitions that occurred in the training of providers and detailed information about what actually happened in the course of intervention delivery, but little information about the content of training, standardised training procedures, or methods for how supervision was rated to check the integrity of the intervention was available.

Participant characteristics

Certain complications arise in the conduct of studies of this nature, which may make them particularly vulnerable to bias and wide variations in outcomes. One concerns the aggressiveness of the breast cancer. This would not be known when beginning a trial and cannot, therefore, be controlled for. Whilst the aim is that the complications should be evenly distributed by randomisation, any variations that do arise in disease aggression may particularly affect short-term outcomes between the intervention and control groups. Likewise, when numbers remaining in the study decline at long-term follow-up, the risk of underlying disease characteristics affecting the study outcome differentially between intervention and control groups is high, as found by Fox 1998 when reviewing Spiegel 1989.

There were other studies where such issues were also potentially important. For instance, Goodwin 2001 refers to differences between the groups at the time of the initial diagnosis in terms of progesterone-receptor status and, as this was associated with survival, this favoured the intervention group (63.9% progesterone receptors versus 42.9%). However, at the time of randomisation they reported no baseline differences that seemingly affect prognosis. 

Participants who are newly diagnosed with metastatic breast cancer are likely to be different to those who have been living with it for longer. Five studies have taken this into account, with descriptive statistics demonstrating no significant baseline difference in the trial groups in terms of duration of disease (Classen 2001; Edelman 1999; Goodwin 2001; Kissane 2007; Low 2010). Two studies specified particular durations since diagnosis of recurrence or progression of breast cancer as part of their inclusion criteria (Aranda 2006; Gotay 2007). However, in three studies there were no reported data to characterise how long the women had lived with metastatic breast cancer and how they were represented in the trial groups, if at all (Cunningham 1998; Scholten 2001; Spiegel 1989).

Another consideration is that the period of recruitment, reported in all but two trials (Aranda 2006; Scholten 2001), spans almost 30 years, from 1978 (Spiegel 1989) to 2007 (Low 2010). In that time, the management of patients with metastatic breast cancer has evolved substantially in terms of diagnosis and treatment, and this has to be considered when comparing trials conducted in different generations.

Control conditions

The control group provision varied across the trials. Authors reported that they differed only in terms of the intervention, but they were not matched to the experimental condition in terms of frequency of contact, and this may be significant as intervention effectiveness could increase with contact time. Comparator conditions should allow for outcomes to be tested against the intervention content rather than the frequency of contact.

Quality of studies

The total number of participants contributing to this review update is small (1378 participants). Most of these studies had methodological issues. This inevitably weakens the validity of their and this review’s findings. Although all of the studies were RCTs, three studies provided insufficient detail to judge whether they observed random sequence generation (Edelman 1999; Scholten 2001; Spiegel 1989), while four of the studies did not have enough information to decide on whether there was adequate allocation concealment, the other determinant of selection bias (Edelman 1999; Goodwin 2001; Scholten 2001; Spiegel 1989). As the risk of selection bias was unclear for Edelman 1999 and Spiegel 1989, a sensitivity analysis was undertaken by removing them from the meta-analyses. This increased the combined effect estimate for survival at one-year follow-up (OR 1.74; 95% CI 1.21 to 2.48), while removing Spiegel 1989 from the five-year survival data decreased the effect (OR 0.80; 95% CI 0.40 to 1.64).

Although it was not possible to blind the participants or the providers, a lack of outcome assessment blinding may lead to detection bias. For all but two studies (Goodwin 2001; Spiegel 1989) it was unclear whether there was any blinding of outcome assessment. Even though the response instruments are validated, the manner of follow-up (for example level of persistence) could be influenced by the knowledge of which treatment group the patient had been assigned to.

Potential biases in the review process

The review was conducted in keeping with the principles of the UK NHS Centre for Reviews and Dissemination (NHS CRD 1996) and the Cochrane Handbook for Systematic Reviews of Interventions (Higgins and Green 2011). It comprised electronic database searches and manual follow-ups for further references to maximise recall.

This 2012 updated review differs from the previous Cochrane review in two ways. We undertook a comprehensive update of the literature using a modified search strategy that included more search terms for the different topics and strings for trial design (that is RCT). We also had to devise a new search strategy for CINAHL as the database changed from SP to EBSCO. Nevertheless, the conclusions of this updated review are similar to those of the original Cochrane review (Edwards 2004), and also to the previously updated 2007 version of the review (Edwards 2008).

As with all systematic reviews and meta-analyses, subjective judgement is involved at various points along the process, from identifying studies to data extraction and analysis. As a result, although the search strategies, data extraction and analysis were thorough, there is a possibility of missing relevant studies and data, and alternative analyses could be undertaken. 

However, the main limitation of the review comes from the relative lack of data in this field. This is likely to give a lack of power to detect genuine effects that may arise from the interventions, and therefore the results of these meta-analyses must be interpreted with caution. However, a specific feature of the data available from the identified studies was the heterogeneity of the outcome measures, follow-up durations, and the lack of data in suitable format for combination. As described above, for example, some studies presented an analysis of longitudinal trends in psychological outcomes whereas this review uses essentially cross-sectional data (for example POMS score at one year). The paucity of data also affected how we could analyse the data. For instance, we could only use point-in-time meta-analyses to extract data from the reported survival curves. However, this method is problematic as it cannot take into account dropouts and censoring, both features of the included trials. For example, the dropouts are likely to account for the disparity between the median survival and proportion of participants surviving, as reported by Edelman 1999. They report median survival times from recruitment of 1.07 years (IQR 0.63 to 1.48) for the control group and 0.97 years (IQR 0.59 to 1.46) for the intervention group. Therefore, one could reasonably expect just over 50% of patients would survive approximately one year in the control group and just under 50% in the intervention arm. However, Edelman 1999 further reported the proportions surviving one year to be 42/62 (68%) and 41/62 (66%) in the intervention and control groups respectively. See Figure 3. Thus this review is not able to utilise all the data currently available in the field for meta-analyses, a particular difficulty when the data are relatively sparse to begin with.

The apparent beneficial effects of the psychological interventions may also be influenced by bias from poor methodological quality. For instance, sequence generation or allocation concealment, or both, were unclear in four studies. Blinding of outcome assessment was unclear for eight of the studies, and may pose a high risk of bias, with the potential to overestimate the effect of the psychological intervention.

We did not report data on the socio-economic status, age or ethnicity of study participants, but participants were primarily Caucasian women and most studies were conducted in developed nations (USA, Australia and Canada). As such, it is not clear whether our findings generalise to women from different ethnic backgrounds or those in less developed countries.

We had insufficient studies to perform a meta-regression analysis, which could have determined some predictors of successful interventions. 

Authors' conclusions

Implications for practice

Psychological interventions appear to be effective in improving survival at 12 months but not at longer-term follow-up, and in reducing psychological symptoms in women with metastatic breast cancer only in some of the outcomes assessed. Given the extensive commitment of any breast cancer service required to provide group psychological therapies (either cognitive behavioural or supportive-expressive), the relative lack of evidence in its favour warrants a debate among patients, providers and commissioners of care (pending further clinical and economic evaluations) about whether to routinely offer such services to women diagnosed with metastatic breast cancer in comparison to other interventions for which there is evidence of benefit to patients.

It is, however, likely that group psychological therapies will continue to be offered to some women following diagnosis of metastatic breast cancer. It remains a common diagnosis and one that is accompanied by a devastating psychological effect that prompts patients to seek help, and healthcare professionals to offer help to meet patient needs.

The possibility of the interventions causing harm is not ruled out by the available data. As the wider palliative care literature indicates (Whatley 1998), the confidence intervals for the outcome data identified include the possibility of adverse outcomes associated with the provision of the interventions.

Implications for research

Further research is needed in this field. Few studies were identified regarding educational, psychotherapy or cognitive behavioural therapies for individual women with metastatic breast cancer. Evaluating the effects of such therapies remains important as previous reviews have produced heterogeneous results (Devine 1995; Meyer 1995). As described in the Background section, the association between psychosocial factors and breast cancer may be more modest than conventional wisdom suggests (McKenna 1999), but there is some limited evidence of bio-medical changes arising from providing interventions to women with early stage cancer (Larson 2000; van der Pompe 1997) and evidence that the adjustment process in breast cancer may have specific differences from those in other cancers (due to issues arising in body image, sexuality, effects on relationships etc). As such, there appears to be value in further assessing the effects of educational or psychological interventions at the individual level. We also share the view of Goodwin 2003 that many of the psychological changes made by individuals in long-term interventions may elude conventional psychometric assessment. Further research, of a rigorous qualitative nature, is required to develop a clearer understanding of the experience of living and eventually dying of cancer within the context of a long-term intervention.

In terms of quantitative research, the general lack of evidence of the effects of the group psychological interventions may also be a Type 2 error, failure to identify a true effect. In this review this could stem from either the relative paucity of data or the difficulty of measurement. Further studies, with adequate power and follow-up to agreed stages with the outcome measures agreed as the most valid and relevant to assess effectiveness, would be helpful in identifying both the psychological and survival effects of the interventions. The failure to identify effects for most of the psychological outcomes measured may be particularly vulnerable to this criticism as different studies used a wide range of outcome instruments and data combination was difficult. The standardisation of outcomes and instruments to assess them is vital in further studies. Where possible, decisions about which measures and at which durations of follow-up to assess them should be made in the light of the methods of the existing trials, aiming to add to aspects on which some data have been accrued already (for example POMS or pain scales at one year). However, it is clear that the complexity of the emotional issues faced by women with metastatic breast cancer, including existential concerns, make it difficult for research studies to obtain valid measures of subtle aspects of emotional adjustment. The validity of instruments, including responsiveness to changes in individuals, requires attention in future research in this field. Any further trial should be complemented by a health economic (cost-effectiveness and cost-benefit) evaluation so that any effects can be interpreted in the light of the resources required to achieve them. This should also take account of carer outcomes, for which there are virtually no reliable data at present.

It is also important to note that the interventions have been usually theoretically or researcher driven to date. They may or may not reflect approaches that women would find most important in addressing their psychological morbidity. Inductive empirical research with women with metastatic breast cancer is needed to identify the required content for further interventions.

A further implication for research is the need, in all future studies, to take into account the biology of metastatic breast cancer, particularly as the evidence grows. There was evidence of inconsistencies in how these issues were addressed (or not) in some of the included studies. For instance, greater attention is needed to the biochemical factors that can influence the growth and proliferation of breast cancer (Sørlie 2001; Sørlie 2003). Oestrogen receptors and progesterone receptors are the most widely studied markers in breast tissue. When compared with hormone receptor–negative tumours, hormone receptor–positive breast cancers exhibit stronger clinical responses to hormonal treatment and confer better prognosis. Growth factors (epidermal growth factors, transforming growth factors α and β, and insulin-like growth factors I and II), and various cytokines and lymphokines are also known to influence the behaviour and phenotype expression of breast cells. A greater understanding of the molecular subtypes based on gene-expression patterns in breast tumour tissue will help to clarify whether breast cancers are aetiologically and clinically heterogeneous.

The questions of how psychological therapies work in women with metastatic breast cancer and which components are necessary remain unanswered. Future research should evaluate different components of the intervention to understand if there is an optimum treatment modality and also duration. In order to fully evaluate the intervention, reports of trials in this field should, therefore, pay greater attention to methodological quality for more comparable group characteristics including the biology of metastatic disease, explicitly describe the intervention delivered and its fidelity, and use sensitive and validated outcome measures. This would also help with implementing positive results, provided there is sufficient information to replicate the intervention. Longitudinal studies identifying psychological and physiological mediators of outcome may also have value, and could help shed light on the basic processes by which psychological treatments are effective for this patient group. Further research is required to explore whether the involvement of family members in treatment can predict outcomes. Similarly, the use of pharmacotherapy as a co-intervention to directly deal with psychological symptoms that meet diagnostic thresholds for depression and anxiety needs to be made explicit.

Several gaps in the available literature make our comment on the effectiveness or otherwise of interventions for certain scenarios impossible. These include psychological interventions for individuals, despite the three studies identified by this review (Aranda 2006; Gotay 2007; Low 2010), evaluation of online group support facilities that have flourished in recent years (for example Lieberman 2003) and complementary therapies (for example Laidlaw 2005).

Uncertainty remains regarding the subgroups of patients who would benefit most from treatment and the characteristics of successful interventions. Considerable work remains outstanding to clarify these relationships and whether certain subgroups of women with metastatic breast cancer can be identified who may be most likely to benefit. Examples of the latter might include people with specific personality characteristics, and help-seeking behaviour patterns that might make interventions potentially more effective.

Acknowledgements

The authors of this updated review gratefully acknowledge the contributions of Drs Melanie Maxwell and Stephen Hailey for the first published review (Edwards 2004) and Drs Nicholas Hulbert-Williams and Richard Neal for the updated review (Edwards 2008). The authors also thank the Breast Cancer Review Group editorial panel and Mr Fergus Tai, Trials Search Co-ordinator, for undertaking the search of the Specialised Register.

Data and analyses

Download statistical data

Comparison 1. Psychological intervention versus usual practice
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Survival at 1 year6856Odds Ratio (M-H, Random, 95% CI)1.46 [1.07, 1.99]
1.1 Cognitive behaviour therapy2190Odds Ratio (M-H, Random, 95% CI)1.14 [0.61, 2.15]
1.2 Supportive/expressive group therapy4666Odds Ratio (M-H, Random, 95% CI)1.56 [1.07, 2.28]
2 Survival at 5 years4666Odds Ratio (M-H, Random, 95% CI)1.03 [0.42, 2.52]
2.1 Supportive/expressive group therapy4666Odds Ratio (M-H, Random, 95% CI)1.03 [0.42, 2.52]
3 Survival at 10 years4394Odds Ratio (M-H, Fixed, 95% CI)1.08 [0.59, 1.96]
3.1 Cognitive behaviour therapy2190Odds Ratio (M-H, Fixed, 95% CI)0.62 [0.27, 1.39]
3.2 Supportive/expressive group therapy2204Odds Ratio (M-H, Fixed, 95% CI)2.32 [0.87, 6.21]
4 Anxiety at 6 months163Mean Difference (IV, Fixed, 95% CI)1.67 [-1.42, 4.76]
4.1 Cognitive behaviour therapy163Mean Difference (IV, Fixed, 95% CI)1.67 [-1.42, 4.76]
5 Profile of Mood States (full version) at 6 months163Mean Difference (IV, Fixed, 95% CI)10.28 [-4.28, 24.84]
5.1 Cognitive behaviour therapy163Mean Difference (IV, Fixed, 95% CI)10.28 [-4.28, 24.84]
6 Self-esteem at 6 months163Mean Difference (IV, Fixed, 95% CI)2.61 [-4.26, 9.48]
6.1 Cognitive behaviour therapy163Mean Difference (IV, Fixed, 95% CI)2.61 [-4.26, 9.48]
7 Profile of Mood States (short version) at 8 months166Mean Difference (IV, Fixed, 95% CI)-0.38 [-5.93, 5.17]
7.1 Cognitive behaviour therapy166Mean Difference (IV, Fixed, 95% CI)-0.38 [-5.93, 5.17]
8 Functional Living Index at 8 months166Mean Difference (IV, Fixed, 95% CI)-1.93 [-13.68, 9.82]
8.1 Cognitive behaviour therapy166Mean Difference (IV, Fixed, 95% CI)-1.93 [-13.68, 9.82]
9 Courtauld Emotional Control Scale at 8 months197Mean Difference (IV, Fixed, 95% CI)-4.20 [-4.53, -3.87]
9.1 Supportive/expressive group therapy197Mean Difference (IV, Fixed, 95% CI)-4.20 [-4.53, -3.87]
10 Anxiety at 1 year1147Mean Difference (IV, Fixed, 95% CI)-0.30 [-2.63, 2.03]
10.1 Cognitive behaviour therapy00Mean Difference (IV, Fixed, 95% CI)0.0 [0.0, 0.0]
10.2 Supportive/expressive group therapy1147Mean Difference (IV, Fixed, 95% CI)-0.30 [-2.63, 2.03]
11 Profile of Mood States (full version) at 10-12 months3257Mean Difference (IV, Random, 95% CI)-11.54 [-28.68, 5.60]
11.1 Supportive/expressive group therapy3257Mean Difference (IV, Random, 95% CI)-11.54 [-28.68, 5.60]
12 Quality of Life (EORTC QLQ-C30 score) at 1 year1215Mean Difference (IV, Fixed, 95% CI)0.90 [-5.51, 7.31]
12.1 Supportive/expressive group therapy1215Mean Difference (IV, Fixed, 95% CI)0.90 [-5.51, 7.31]
13 Pain at 1 year3279Mean Difference (IV, Fixed, 95% CI)-0.58 [-0.98, -0.18]
13.1 Supportive/expressive group therapy3279Mean Difference (IV, Fixed, 95% CI)-0.58 [-0.98, -0.18]
Analysis 1.1.

Comparison 1 Psychological intervention versus usual practice, Outcome 1 Survival at 1 year.

Analysis 1.2.

Comparison 1 Psychological intervention versus usual practice, Outcome 2 Survival at 5 years.

Analysis 1.3.

Comparison 1 Psychological intervention versus usual practice, Outcome 3 Survival at 10 years.

Analysis 1.4.

Comparison 1 Psychological intervention versus usual practice, Outcome 4 Anxiety at 6 months.

Analysis 1.5.

Comparison 1 Psychological intervention versus usual practice, Outcome 5 Profile of Mood States (full version) at 6 months.

Analysis 1.6.

Comparison 1 Psychological intervention versus usual practice, Outcome 6 Self-esteem at 6 months.

Analysis 1.7.

Comparison 1 Psychological intervention versus usual practice, Outcome 7 Profile of Mood States (short version) at 8 months.

Analysis 1.8.

Comparison 1 Psychological intervention versus usual practice, Outcome 8 Functional Living Index at 8 months.

Analysis 1.9.

Comparison 1 Psychological intervention versus usual practice, Outcome 9 Courtauld Emotional Control Scale at 8 months.

Analysis 1.10.

Comparison 1 Psychological intervention versus usual practice, Outcome 10 Anxiety at 1 year.

Analysis 1.11.

Comparison 1 Psychological intervention versus usual practice, Outcome 11 Profile of Mood States (full version) at 10-12 months.

Analysis 1.12.

Comparison 1 Psychological intervention versus usual practice, Outcome 12 Quality of Life (EORTC QLQ-C30 score) at 1 year.

Analysis 1.13.

Comparison 1 Psychological intervention versus usual practice, Outcome 13 Pain at 1 year.

Appendices

Appendix 1. Search strategy - MEDLINE (OvidSP) (1966 to June 2011)

  1. breast neoplasm.mp. or exp Breast Neoplasms/

  2. breast cancer.mp.

  3. (breast adj5 carcinoma*).ti,ab.

  4. (breast adj5 cancer*).ti,ab.

  5. (breast adj5 neoplasm*).ti,ab.

  6. (breast adj5 tumour*).ti,ab.

  7. (breast adj5 tumor*).ti,ab.

  8. (breast adj5 (cancer* or tumor* or tumour* or carcinoma* or neoplasm* or carcinogen* or malignan* or metasta* or recur* or relaps*)).tw,ot.

  9. (advanced adj3 (cancer* or tumor* or tumour* or carcinoma* or neoplasm* or carcinogen* or malignan* or metasta* or recur* or relaps*)).tw,ot.

  10. (advanced adj5 (cancer* or tumor* or tumour* or carcinoma* or neoplasm* or carcinogen* or malignan* or metasta* or recur* or relaps*)).tw,ot.

  11. psychotherapy.mp. or exp Psychotherapy/ or exp Psychotherapy, Group/

  12. supportive expressive group therapy.mp.

  13. exp Cognitive Therapy/ or cognitive behavioural therapy.mp.

  14. group support*.mp.

  15. group intervention*.mp.

  16. exp Self-Help Groups/ or self-help group*.mp.

  17. Counseling/ or counsel.mp.

  18. counsel*.ti,ab.

  19. expressive writing.mp.

  20. psychotherapeutic*.mp. or Psychotherapy/

  21. randomized controlled trial.pt.

  22. controlled clinical trial.pt.

  23. randomized.ab.

  24. placebo.ab.

  25. clinical trials as topic.sh.

  26. randomly.ab.

  27. trial.ti.

  28. 21 or 22 or 23 or 24 or 25 or 26 or 27

  29. exp animals/ not humans.sh.

  30. 28 not 29

  31. exp Survival/ or survival.tw.

  32. quality of life.tw. or exp "Quality of Life"/

  33. adaptation, psychological/

  34. 31 or 32 or 33

  35. 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10

  36. 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20

  37. 35 and 36 and 30

  38. 35 and 34 and 30

  39. 37 and 38

Appendix 2. Search strategy - EMBASE (OvidSP) (1980 to June 2011)

  1. breast neoplasm.mp. or exp Breast Neoplasms/

  2. breast cancer.mp.

  3. (breast adj5 carcinoma*).ti,ab.

  4. (breast adj5 cancer*).ti,ab.

  5. (breast adj5 neoplasm*).ti,ab.

  6. (breast adj5 tumour*).ti,ab.

  7. (breast adj5 tumor*).ti,ab.

  8. (breast adj5 (cancer* or tumor* or tumour* or carcinoma* or neoplasm* or carcinogen* or malignan* or metasta* or recur* or relaps*)).tw,ot.

  9. (advanced adj3 (cancer* or tumor* or tumour* or carcinoma* or neoplasm* or carcinogen* or malignan* or metasta* or recur* or relaps*)).tw,ot.

  10. (advanced adj5 (cancer* or tumor* or tumour* or carcinoma* or neoplasm* or carcinogen* or malignan* or metasta* or recur* or relaps*)).tw,ot.

  11. psychotherapy.mp. or exp Psychotherapy/ or exp Psychotherapy, Group/

  12. supportive expressive group therapy.mp.

  13. exp Cognitive Therapy/ or cognitive behavioural therapy.mp.

  14. group support*.mp.

  15. group intervention*.mp.

  16. exp Self-Help Groups/ or self-help group*.mp.

  17. Counseling/ or counsel.mp.

  18. counsel*.ti,ab.

  19. expressive writing.mp.

  20. psychotherapeutic*.mp. or Psychotherapy/

  21. exp Survival/ or survival.tw.

  22. quality of life.tw. or exp "Quality of Life"/

  23. adaptation, psychological/

  24. Clinical trial/

  25. Randomized controlled trial/

  26. Randomization/

  27. Single blind procedure/

  28. Double blind procedure/

  29. Crossover procedure/

  30. Placebo/

  31. Randomi?ed controlled trial$.tw.

  32. Rct.tw.

  33. Random allocation.tw.

  34. Randomly allocated.tw.

  35. Allocated randomly.tw.

  36. (allocated adj2 random).tw.

  37. Single blind$.tw.

  38. Double blind$.tw.

  39. ((treble or triple) adj blind$).tw.

  40. Placebo$.tw.

  41. Prospective study/

  42. or/24-41

  43. Case study/

  44. Case report.tw.

  45. Abstract report/ or letter/

  46. or/43-45

  47. 42 not 46

  48. 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10

  49. 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20

  50. 21 or 22 or 23

  51. 47 and 48 and 49

  52. 47 and 48 and 50

  53. 51 and 52

Appendix 3. Search strategy - PsycINFO (OvidSP) (1974 to June 2011)

  1. breast neoplasm.mp. or exp Breast Neoplasms/

  2. breast cancer.mp.

  3. (breast adj5 carcinoma*).ti,ab.

  4. (breast adj5 cancer*).ti,ab.

  5. (breast adj5 neoplasm*).ti,ab.

  6. (breast adj5 tumour*).ti,ab.

  7. (breast adj5 tumor*).ti,ab.

  8. (breast adj5 (cancer* or tumor* or tumour* or carcinoma* or neoplasm* or carcinogen* or malignan* or metasta* or recur* or relaps*)).tw,ot.

  9. (advanced adj3 (cancer* or tumor* or tumour* or carcinoma* or neoplasm* or carcinogen* or malignan* or metasta* or recur* or relaps*)).tw,ot.

  10. (advanced adj5 (cancer* or tumor* or tumour* or carcinoma* or neoplasm* or carcinogen* or malignan* or metasta* or recur* or relaps*)).tw,ot.

  11. psychotherapy.mp. or exp Psychotherapy/ or exp Psychotherapy, Group/

  12. supportive expressive group therapy.mp.

  13. exp Cognitive Therapy/ or cognitive behavioural therapy.mp.

  14. group support*.mp.

  15. group intervention*.mp.

  16. exp Self-Help Groups/ or self-help group*.mp.

  17. Counseling/ or counsel.mp.

  18. counsel*.ti,ab.

  19. expressive writing.mp.

  20. psychotherapeutic*.mp. or Psychotherapy/

  21. exp Survival/ or survival.tw.

  22. quality of life.tw. or exp "Quality of Life"/

  23. exp Adaptation/ or adaptation.mp.

  24. double-blind.tw.

  25. randomized.tw.

  26. randomly assigned.tw.

  27. 24 or 25 or 26

  28. 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10

  29. 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20

  30. 21 or 22 or 23

  31. 27 and 28 and 29

  32. 27 and 28 and 30

  33. 31 and 32

Appendix 4. Search strategy - CINAHL (EBSCO) (1982 to June 2011)

S1. (MH "Clinical Trials+")

S2. PT Clinical trial

S3. TX clinic* n1 trial*

S4. TX ( (singl* n1 blind*) or (singl* n1 mask*) ) or TX ( (doubl* n1 blind*) or (doubl* n1 mask*) ) or TX ( (tripl* n1 blind*) or (tripl* n1 mask*) ) or TX ( (trebl* n1 blind*) or (trebl* n1 mask*) )

S5. TX randomi* control* trial*

S6. (MH "Random Assignment")

S7. TX random* allocat*

S8. TX placebo*

S9. (MH "Placebos")

S10. (MH "Quantitative Studies")

S11. TX allocat* random*

S12. S1 or S2 or S3 or S4 or S5 or S6 or S7 or S8 or S9 or S10 or S11

S13. (MH "Breast Neoplasms+") OR "breast neoplasm"

S14. "breast cancer"

S15. "breast carcinoma"

S16. "breast tumour"

S17. "breast tumor"

S18. "metastatic breast cancer"

S19. "metastatic breast neoplasm"

S20. "metastatic breast carcinoma"

S21. "metastatic breast tumor"

S22. "metastatic breast tumour"

S23. "advanced cancer"

S24. "advanced breast cancer"

S25. "advanced breast carcinoma"

S26. "advanced breast tumour"

S27. "advanced breast tumor"

S28. "advanced neoplasm"

S29. "advanced carcinoma"

S30. "advanced tumour"

S31. "advanced tumor"

S32. "recurrent breast cancer"

S33. "recurrent breast neoplasm"

S34. "recurrent breast carcinoma"

S35. "recurrent breast tumour"

S36. "recurrent breast tumor"

S37. "recurrent cancer"

S38. "recurrent neoplasm"

S39. "recurrent carcinoma"

S40. "recurrent tumour"

S41. "recurrent tumor"

S42. (MH "Psychotherapy+") OR "psychotherapy" OR (MM "Psychotherapy, Brief") OR (MH "Psychotherapy, Group+")

S43. "supportive expressive group therapy"

S44. (MM "Cognitive Therapy") OR "cognitive therapy" OR (MH "Cognitive Therapy (Iowa NIC) (Non-Cinahl)+")

S45. (MH "Support Groups+") OR (MM "Support Group (Iowa NIC)") OR "group support"

S46. "group intervention"

S47. "self-help groups"

S48. (MH "Counseling+") OR "counseling"

S49. "counsel"

S50. "expressive writing"

S51. (MH "Psychotherapeutic Processes+") OR "psychotherapeutic"

S52. (MM "Survival") OR "survival"

S53. (MH "Quality of Life+") OR "quality of life"

S54. (MH "Adaptation, Psychological+")

S55. S13 or S14 or S15 or S16 or S17 or S18 or S19 or S20 or S21 or S22 or S23 or S24 or S25 or 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 or S39 or S40 or S41

S56. S42 or S43 or S44 or S45 or S46 or S47 or S48 or S49 or S50 or S51

S57. S52 or S53 or S54

S58. S12 and S55 and S56

S59. S12 and S55 and S57

S60. S58 and S59

Appendix 5. Search Strategy - ClinicalTrials.gov (July 2011)

Basic search

  1. (breast cancer AND metastatic) AND (psychological AND intervention)

  2. (breast cancer AND metastases) AND (psychological AND intervention)

  3. breast cancer AND advance) AND (psychological AND intervention)

  4. (breast cancer AND advanced) AND (psychological AND intervention)

  5. (breast cancer AND metastatic) AND (psychology AND intervention)

  6. (breast cancer AND metastases) AND (psychology AND intervention)

  7. (breast cancer AND advance) AND (psychology AND intervention)

  8. (breast cancer AND advanced) AND (psychology AND intervention)

  9. (breast cancer AND metastatic) AND (psychosocial AND intervention)

  10. (breast cancer AND metastases) AND (psychosocial AND intervention)

  11. breast cancer AND advance) AND (psychosocial AND intervention)

  12. (breast cancer AND advanced) AND (psychosocial AND intervention)

  13. (breast carcinoma AND metastatic) AND (psychological AND intervention)

  14. (breast carcinoma AND metastases) AND (psychological AND intervention)

  15. breast carcinoma AND advance) AND (psychological AND intervention)

  16. (breast carcinomaAND advanced) AND (psychological AND intervention)

  17. (breast carcinoma AND metastatic) AND (psychology AND intervention)

  18. (breast carcinoma AND metastases) AND (psychology AND intervention)

  19. (breast carcinoma AND advance) AND (psychology AND intervention)

  20. (breast carcinoma AND advanced) AND (psychology AND intervention)

  21. (breast carcinoma AND metastatic) AND (psychosocial AND intervention)

  22. (breast carcinoma AND metastases) AND (psychosocial AND intervention)

  23. breast carcinoma AND advance) AND (psychosocial AND intervention)

  24. (breast carcinoma AND advanced) AND (psychosocial AND intervention)

Each basic search combined with each of the following advanced searches

Advanced search 1

  • Recruitment: all studies

  • Study results: studies with results

  • Study type: interventional studies

  • Condition: breast cancer AND (advanced or metastatic)

  • Intervention: psychological AND (education or therapy or intervention)

  • Outcome: survival OR psychological

  • Gender: studies with female participants

Advanced search 2

  • Recruitment: all studies

  • Study results: studies with results

  • Study type: interventional studies

  • Condition: breast cancer AND (advanced or metastatic)

  • Intervention: psychosocial AND (education or therapy or intervention)

  • Outcome: survival OR psychological

  • Gender: studies with female participants

Advanced search 3

  • Recruitment: all studies

  • Study results: studies with results

  • Study type: interventional studies

  • Condition: breast cancer AND (advanced or metastatic)

  • Intervention: psychiatric AND (education or therapy or intervention)

  • Outcome: survival OR psychological

  • Gender: studies with female participants

Advanced search 4

  • Recruitment: all studies

  • Study results: studies with results

  • Study type: interventional studies

  • Condition: breast cancer AND (advanced or metastatic)

  • Intervention: psychotherapeutic OR group intervention OR group support OR cognitive behaviour OR supportive expressive OR individual psychotherapy

  • Outcome: survival OR psychological

  • Gender: studies with female participants

Appendix 6. Search strategy Controlled Trials (July 2011)

  1. (breast cancer AND metastatic) AND (psychological AND intervention)

  2. (breast cancer AND metastases) AND (psychological AND intervention)

  3. breast cancer AND advance) AND (psychological AND intervention)

  4. (breast cancer AND advanced) AND (psychological AND intervention)

  5. (breast cancer AND metastatic) AND (psychology AND intervention)

  6. (breast cancer AND metastases) AND (psychology AND intervention)

  7. (breast cancer AND advance) AND (psychology AND intervention)

  8. (breast cancer AND advanced) AND (psychology AND intervention)

  9. (breast cancer AND metastatic) AND (psychosocial AND intervention)

  10. (breast cancer AND metastases) AND (psychosocial AND intervention)

  11. breast cancer AND advance) AND (psychosocial AND intervention)

  12. (breast cancer AND advanced) AND (psychosocial AND intervention)

  13. (breast carcinoma AND metastatic) AND (psychological AND intervention)

  14. (breast carcinoma AND metastases) AND (psychological AND intervention)

  15. breast carcinoma AND advance) AND (psychological AND intervention)

  16. (breast carcinoma AND advanced) AND (psychological AND intervention)

  17. (breast carcinoma AND metastatic) AND (psychology AND intervention)

  18. (breast carcinoma AND metastases) AND (psychology AND intervention)

  19. (breast carcinoma AND advance) AND (psychology AND intervention)

  20. (breast carcinoma AND advanced) AND (psychology AND intervention)

  21. (breast carcinoma AND metastatic) AND (psychosocial AND intervention)

  22. (breast carcinoma AND metastases) AND (psychosocial AND intervention)

  23. breast carcinoma AND advance) AND (psychosocial AND intervention)

  24. (breast carcinoma AND advanced) AND (psychosocial AND intervention)

Appendix 7. Search strategy WHO ICTRP (July 2011)

Basic search

  1. (breast cancer AND advance*) AND (psycholog* AND intervention*)

  2. (breast cancer AND metasta*) AND (psycholog* AND intervention*)

  3. (breast cancer AND advance*) AND (psychosocial AND intervention*)

  4. (breast cancer AND metasta*) AND (psychosocial AND intervention*)

  5. (breast cancer AND advance*) AND psychotherap*

  6. (breast cancer AND metasta*) AND psychotherap*

Advanced search

Advanced search 1

  • Condition: breast cancer% AND (advance% OR metasta%)

  • Intervention: psycholog% AND (education% OR therap% OR intervention%)

  • Recruitment Status: ALL

Advanced search 2

  • Condition: breast cancer% AND (advance% OR metasta%)

  • Intervention: psychosocial AND (education% OR therap% OR intervention%)

  • Recruitment Status: ALL

Advanced search 3

  • Condition: breast cancer% AND (advance% OR metasta%)

  • Intervention: psychiatric% AND (education% OR therap% OR intervention%)

  • Recruitment Status: ALL

Advanced search 4

  • Condition: breast cancer% AND (advance% OR metasta%)

  • Intervention: psychotherap% OR group intervention% OR group support% OR support group% OR cognitive behavio% train% OR cognitive behavio% therap% OR individual psychotherapy%

  • Recruitment Status: ALL

Advanced search 5

  • Title: Psychological interventions for women with metastatic breast cancer

  • Recruitment Status: ALL

Appendix 8. Data Extraction Sheet

DATE FORM COMPLETED:

REVIEWER ID:

STUDY ID: (surname/year)

NOTES:

STUDY :

Title:

Aim:

Year:

Citation:

Country of origin:

Funding:

Ethics approved (yes/no/unclear):

Duration:

ELIGIBILITY :

 Y/NComment
RCT?  
Women with Metastatic Breast Cancer  
Psychological Intervention type:  
-  Educational  
-  Individual psychotherapy  
-  Cognitive behavioural therapy  
-  Group  
Other factors?  

 

PARTICIPANTS :

Patient group involved:                                                Age:                                       Ethnicity:

 

How were participants recruited?:

Inclusion criteria:

Exclusion criteria:

Other social/demographic details:

Numbers:           

Total eligible  
Pre-randomisedExcluded 
Refused 
Other 
RandomisedINT 
CTRL 
Post-randomisedWithdrew 
Died 
Lost to FU 
Other 
  
Included in analysis
Total 
Control 
Intervention 

 

Any differences in baseline risk?:

Other comments:

INTERVENTIONS :

Intervention type & detail:        

Education [ ] :

                Individual Psychotherapy [ ] :

                Cognitive Behavioural Training [ ] :

                Group Interventions [ ] :

                Other [ ] Comments:

Intervention delivery (e.g. frequency, duration):

Intervention setting:

Details of control group intervention:

Details of providers:                                     

Intervention quality:

Other:

OUTCOMES :

Authors outcome measures;

Primary:

Secondary:

Others:

Method of assessing outcome:

Timing of outcome assessment (e.g. frequency, length of FU):

Method of follow up for non-responders :

Adverse events:

Any data regarding the validity of outcome measurement tools?:

RESULTS :

Continuous outcomes
OutcomeTiming of Outcome AssessmentIntervention GroupControl GroupNotes
Mean/mean changeStandard deviationNMean/mean changeStandard deviationN
         
         
         
         
         
Dichotomous outcomes
OutcomeTiming of Outcome assessmentIntervention groupControl GroupNotes
Observed (n)Total (N)Observed (n)Total (N)
       
       
       
       
       
       

What's new

DateEventDescription
16 June 2015AmendedA typographical error was noted in the Excluded studies section

History

Protocol first published: Issue 2, 2003
Review first published: Issue 2, 2004

DateEventDescription
30 June 2011New search has been performedPerformed searches for new studies on 30 June 2011
30 June 2011New citation required but conclusions have not changedFive new studies included, adding 742 patients. New search strategies developed to include trial design
14 May 2008AmendedChange to author by-line
26 March 2008AmendedConverted to new review format
29 January 2008New search has been performedReview updated
24 February 2004New search has been performedFirst published version

Contributions of authors

Dr Mustafa and Dr Carson-Stevens were responsible for adapting, developing and implementing the search strategies. They selected the studies for inclusion, undertook quality assessment and extracted the data.

Analyses were performed by David Gillespie.

Dr Mustafa wrote the first draft of the updated review, with sections of the effects of intervention written by David Gillespie.

All co-reviewers were involved in providing critical comments about the manuscript.

Professor Edwards also provided supervision of the review process and is the guarantor of the review.

Declarations of interest

None declared

Sources of support

Internal sources

  • No sources of support supplied

External sources

  • Nil, Other.

Notes

The review was updated in 2012 based on further literature searches done in June-July 2011. Five new studies were identified as a result of the updated searches.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Aranda 2006

MethodsRCT
Participants105 women with advanced breast cancer, English speaking, Australian
InterventionsTwo components to the psycho-educational intervention; group (face-to-face) session followed by an individual (telephone) session a week later. The group session was designed around the FOCUS framework; family involvement, optimistic attitude, coping effectiveness, uncertainty reduction and symptom management. Telephone encounter involved asking whether the suggested strategies ameliorated the concerns, as well as addressing new concerns and self-help strategies
OutcomesEORTC QLQ-30 version 2, supportive care needs survey. Data collected at baseline and then one month and 3 months post-baseline. Telephone follow-up was used to enhance response rate
Notes

No discussion of participant ethnicity

Intervention delivered by breast care nurses previously trained by study team. Quality assurance checks not described beyond initial training and analyses not undertaken for each breast care nurse

Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskFor each study site, an even number of folded intervention (20) and control (20) cards were thoroughly shuffled then placed in consecutively numbered, opaque, sealed envelopes. After baseline data collection, an envelope was sequentially drawn and opened by the breast care nurse/researcher
Allocation concealment (selection bias)Low riskAs above - opaque, sealed envelopes
Blinding of participants and personnel (performance bias)
All outcomes
Low riskNo statement made by authors regarding information giving about the intervention or whether participants were informed which group they had been allocated. But review authors judge blinding as not applicable
Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskA research assistant, not involved in the intervention delivery, administered all follow-up data collection
Incomplete outcome data (attrition bias)
All outcomes
Low riskOverall, 69% (72 out of 105) of patients completed the 1 month follow-up, 78% (36 out of 46) in the usual care arm and 61% (36 out of 59) in the intervention arm. Four of the 105 patients (4%) died prior to 1-month follow-up. Follow-up at 3 months was 57% (60/105), 65% (30/46) of usual care patients and 51% (30/59) intervention patients. Four of the 105 patients (9%) died prior to 3-month follow-up. Excluding the deaths, this represents overall response rates of 71% and 63% for the 1 and 3 month follow-ups, respectively
Selective reporting (reporting bias)Low riskHypothesis generated and reported on (main paper). "Those patients randomly allocated to the intervention group will report a decrease in psychological and informational needs and an increase in quality of life from baseline to follow-up compared to women receiving usual care."
Other biasUnclear risk

The intervention and control groups were not comparable at baseline. A greater proportion of patients in the intervention arm were undergoing current radiotherapy (93% vs. 73%, P = 0.01). Proportionately more patients in the usual care arm of the study had minor children (38% versus 14%, P = 0.02)

Participants were recruited from 3 public and 1 private hospital - health seeking behaviours possibly different

Classen 2001

MethodsRCT (stratified)
Participants125 women with metastatic breast cancer; recruited 1991-1996; USA
InterventionsSupportive-expressive group therapy: weekly meeting for at least 1 year; support to express difficult emotions, confront problems, strengthen relationships (social and professional), and find meaning in life. Control: educational materials for support and information
OutcomesCourtaulds Emotional Control Scale; Weinberger Adjustment Inventory; Stanford Emotional self-efficacy scale; Profile of Mood States; Impact of events scale
Notes 
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low risk"Participants were randomized by the project director..who used a computer-assisted adaptive randomization biased coin-design method to ensure comparability of medical status in treatment and control conditions"
Allocation concealment (selection bias)Low risk"Randomization was conducted for overall assignment to treatment versus control condition"
Blinding of participants and personnel (performance bias)
All outcomes
Low riskReview authors judge blinding as not applicable
Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskInsufficient information to permit judgement
Incomplete outcome data (attrition bias)
All outcomes
Low risk"The sample for each outcome slope of change n = 97, CECS; n = 62, WAI Restraint; n = 64, WAI Repressive-Defensiveness; n = 65, SESES-C) included in this article is reduced from the 125 women randomized at baseline for various reasons (2 women filling out the WAI were missing too many items on the Restraint Scale to score it). Because of the nature of the illness, many women died before they could complete a follow-up assessment, and some women declined to complete follow-ups because of illness-related factors (ns = 17, 35, 35, and 17 for the four scales, respectively)"
Selective reporting (reporting bias)Low riskAll primary and secondary outcome measures reported
Other biasUnclear risk"We found no significant treatment versus control differences at baseline on any of the demographic, medical, or outcome variables except for ethnicity in the samples with follow-ups. There were more minority women in the control group for Suppression, X2 (1, n = 97) = 10.79, P < 0.10; Restraint, X2 (1, n = 62) = 6.33, P < 0.01; Repressive–Defensiveness, X2 (1, n = 64) = 6.66, P < 0.01; and Emotional Self-Efficacy, X2 (1, n = 65) = 12.46, P < 0.01."

Cunningham 1998

MethodsRCT
Participants66 women with metastatic breast cancer; recruited 1992-1994; Canadian
InterventionsSupport + cognitive behaviour therapy sessions and homework exercises; including coping skills (relaxation and stress management); control group had home based CBT only, and regular phone contact to ensure compliance with study outcome evaluations
OutcomesSurvival; Profile of Mood States; Functional Living Index; Mental Adjustment to Cancer Scale; rationality/emotional defensiveness; Marlowe Crowne social desirability; Defensive Repression
NotesSurvival data from time of diagnosis of metastases not time of intervention
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low risk

"Randomization was performed centrally, stratified for study center"

Comment: probably done especially in light of variation of randomization ratio needed by study site

Allocation concealment (selection bias)Low risk

"Using sealed envelopes containing allocations arising from computer-generated random numbers tables."

Comment: probably done (as above)

Blinding of participants and personnel (performance bias)
All outcomes
Low riskNo evidence of blinding to participants or study group leaders. But review authors judge blinding as not applicable
Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskInsufficient information to permit judgement
Incomplete outcome data (attrition bias)
All outcomes
Unclear riskMethod of follow-up for non-responders unclear
Selective reporting (reporting bias)Low riskSpecified expected outcomes. All outcomes reported (analysis of survival data and psychological outcome data presented separately in two different publications)
Other biasLow riskStudy appears free from other sources of bias

Edelman 1999

MethodsRCT (block randomisation)
Participants124 women with metastatic breast cancer; recruited 1994-1997; Australian (English speakers)
InterventionsCognitive behaviour therapy-based; manual for cognitive skills; behavioural techniques at group sessions; home-work exercises; themes addressed: depression, communication, self-esteem, anxiety, self-image, relationships, anger management; control content uncertain
OutcomesSurvival; Profile of Mood States (POMS); Coopersmith Self-Esteem Inventory
NotesAnxiety data from POMS component score
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear risk

"For every 20 patients that were recruited a block randomisation procedure took place, with ten being randomised to each treatment condition."

Comment: random sequence not described

Allocation concealment (selection bias)Unclear risk

"All patients were aware of the randomisation process and that therapy would be offered to only half of the participants. Whilst some did not mind which condition they would be randomised to, some had a clear preference to attend groups and were disappointed at having been randomised to the control group."

Comment: implies an open process of allocation, probably informed intervention group of condition too

Blinding of participants and personnel (performance bias)
All outcomes
Low riskReview authors judge blinding as not applicable
Blinding of outcome assessment (detection bias)
All outcomes
Unclear risk"Ninety-two participants who remained in the trial completed questionnaires in the pre- and post-therapy periods, with numbers diminishing in the subsequent 3 and 6 month assessments."
Incomplete outcome data (attrition bias)
All outcomes
Low risk

"Only 92 patients were included in the analysis of the psychological outcome of the therapy, as patients who failed to complete post-therapy questionnaires were classified as ‘dropouts’ in that analysis."

"The baseline psychological data that was available for 28 patients who dropped out of the study was compared with that of the 92 patients who were retained. Compared to patients who remained in the study, those who dropped out showed significantly higher levels of anxiety (MD 12 6s. 6.5, P = 0.008), confusion (MD 7.5 6s. 5, P = 0.014), TMD (MD 51.5 6s. 14, P = 0.010), and lower self-esteem (MD 66 6s. 80, P = 0.015) and vigour (MD 11 6s. 14.5, P = 0.026)."

"To examine the possibility that the failure to find a survival effect was related to the non-participation in therapy by patients in the intervention group, a further survival analysis (using time of recurrence as baseline) was conducted, with dropouts being excluded from the sample. Once again, only medical prognostic variables predicted survival outcome."

Comment: data presented and consideration of confounders discussed in papers

Selective reporting (reporting bias)Low riskAll outcomes reported (analysis of survival data and psychological outcome data presented separately in two different publications)
Other biasLow riskStudy appears free from other sources of bias

Goodwin 2001

MethodsRCT
Participants235 women with metastatic breast cancer; recruited 1993-1998; Canadian
InterventionsGroup support, encouraged expression of emotions about cancer, effects of illness, relationships, strategies for coping and communicating
Control: educational materials and routine outcome assessments as per study protocol
OutcomesSurvival; Pain, Profile of Mood States; Quality of Life
Notes 
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low risk

"Randomization, performed centrally with the use of sealed envelopes containing allocations from a computer-generated table of random numbers, was stratified according to the center (seven sites) and the presence or absence of visceral metastases."

Comment: probably done

Allocation concealment (selection bias)Unclear risk

"No blinding was used in randomization or data collection."

Comment: uncertain how beneficial or otherwise this process would be for this psychological intervention; risk is dependant on differences between control and interventional groups

Blinding of participants and personnel (performance bias)
All outcomes
Low risk

"Women in the control group did not receive any psychological therapy as part of the study. Women in either group could participate in peer support groups, or therapist-led support groups for patients with various types of cancer that did not involve supportive– expressive therapy, and they could receive any necessary medical, surgical, or psychosocial care."

Comment: no formal group environment available for control group participants; therefore participants very aware which groups they were participating in

Review authors judge blinding as not applicable

Blinding of outcome assessment (detection bias)
All outcomes
Low risk

"Psychosocial questionnaires were scored by a research assistant who had no knowledge of the subjects’ treatment assignment."

Comment: probably done

Incomplete outcome data (attrition bias)
All outcomes
Low risk

"The rate of completion of the psychosocial questionnaires did not differ significantly between the two groups (70.5% of the women in the intervention group and 65.3% of those in the control group completed at least one follow-up questionnaire, P = 0.43)."

"Thirty women (19.0%) dropped out of group therapy after a mean of 3.5 months, and eight women did not attend any sessions."

Selective reporting (reporting bias)Low risk

"Only progesterone-receptor status was associated with survival (P = 0.02); the imbalance in the distribution of this characteristic favoured longer survival in the intervention group (data not shown)."

Comment 1: important finding; unclear why authors did not include data in paper

Comment 2: otherwise, all outcome measures reported against

Other biasUnclear risk

"Participation after randomization in support groups outside of the study was reported by 8.2% of the women in the intervention group (13 of 158 women) and 10.4% of those in the control group (8 of 77 women, P = 0.59)."

"Women in the intervention group were younger than those in the control group at the time of diagnosis (mean age, 44.0 years vs. 46.6 years; P = 0.04) and were less likely to be postmenopausal (14.6% vs. 33.8%, P = 0.001). They were also more likely to have had involvement of axillary nodes (70.9% vs. 50.7%, P = 0.002). The tumors of the women in the intervention group were more likely to have had estrogen receptors (69.0 vs. 54.5, P = 0.06) and progesterone receptors (63.9% vs. 42.9%, P = 0.01), and the use of adjuvant chemotherapy was more common in the intervention group (75.9% vs. 55.6%, P = 0.01)."

Gotay 2007

MethodsRCT (stratified)
Participants305 women with metastatic breast carcinoma; recruited 1998-2002; USA
InterventionsPeer-delivered telephone intervention. The intervention content reflected the most common domains in multidimensional models of quality of life and the primary concerns expressed by breast cancer patients. Patients received four to eight counselling/ information sessions delivered by telephone at weekly intervals, with one to two calls per week. After the first session, patients were sent a standardized packet of information, primarily National Cancer Institute (NCI) pamphlets. All counsellors except one were breast cancer recurrence survivors at least 1 year post-recurrence, the exception being a bone marrow transplant recipient
Outcomes

Primary: emotional well-being: CARES-SF; depressive symptom: CES-D.

Secondary: secondary analyses examined associations between baseline predictors and 3-month scores: CARES-SF psychosocial summary, CES-D, CARES-SF physical functioning, and CARES-SF global, using logistic regression

NotesPhase III trial. All counsellors except one were breast cancer recurrence survivors at least 1 year post-recurrence, the exception being a bone marrow transplant recipient
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskRandom assignment was performed through the SWOG Web Registration Program, which automatically conveys treatment assignment, through a call to the SWOG Statistical Center. Assignment to the control group (CG) or telephone intervention group (TG) was dynamically balanced by the stratification factors (age, recurrence site, and time since diagnosis) according to Pocock and Simon’s method
Allocation concealment (selection bias)Low riskSee above
Blinding of participants and personnel (performance bias)
All outcomes
Low riskReview authors judge blinding as not applicable
Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskInsufficient information to permit judgement
Incomplete outcome data (attrition bias)
All outcomes
Low riskExtent and reasons given for missing data, balanced across trial conditions
Selective reporting (reporting bias)Low risk"The primary outcomes were specified a priori: patient reports of emotional well-being and depressive symptoms at 3 months post-study enrolment. Patients completed most assessments three times: baseline and 3 and 6 months. Three months was selected as the primary time point because most patients would be available for participation at that time, whereas considerable numbers would have progressed and become too ill to respond by 6 months. Therefore, this article reports outcomes data for baseline and 3 months only."
Other biasUnclear risk"More TG [intervention] patients received chemotherapy, whereas more CG [control] patients received hormone therapy for recurrence. Statistically significantly more patients in the CG progressed (progression also includes death) during the 6 months on study"

Kissane 2007

MethodsRCT
Participants227 women with metastatic breast cancer; recruited 1996-2002; Australia
InterventionsSupportive-expressive group therapy, weekly 90 minute sessions, encouraged to participate for a year
OutcomesSurvival and Psychosocial (Monash Interview for Liaison Psychiatry (MILP), EORTC Quality of Life C-30, Impact of Event Scale, Mini-Mental Adjustment to Cancer Scale, Recording of details of self-help or other psychosocial treatments). Women were asked to rate their experience of SEGT and the relaxation classes. Assessment took place at baseline prior to randomisation and four follow-up points at 6-monthly intervals
Notes 
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskEligible consenting patients were independently allocated in a ratio of 2:1 to treatment and control arms respectively, by a stratified randomisation process (utilising an ‘adaptive biased coin design’) at the Statistical Centre of the Peter MacCallum Cancer Institute. Stratification ensured arms were balanced on three prognostic factors: (i) visceral or non-visceral metastases; (ii) less than three, or three or more visceral sites; and (iii) positive or negative oestrogen receptor status
Allocation concealment (selection bias)Low riskSee above - central allocation
Blinding of participants and personnel (performance bias)
All outcomes
Low risk

"For practical reasons, neither researchers nor participants were blinded to the randomization outcome."

Review authors judge blinding as not applicable

Blinding of outcome assessment (detection bias)
All outcomes
Unclear riskInsufficient information to judge
Incomplete outcome data (attrition bias)
All outcomes
Low riskFlowchart of randomised controlled trial of supportive-expressive group therapy for 227 women with metastatic breast cancer. Deaths and study withdrawals are recorded cumulatively down the chart. Treatment data were missing for 13 of the 227 (5.7%): five received treatment interstate and eight had medical records missing
Selective reporting (reporting bias)Low risk"We report here on the impact of SEGT on survival, psycho-social outcomes and treatment adherence"
Other biasLow riskStudy appears free from other sources of bias

Low 2010

MethodsRCT (parallel groups)
Participants62 women with metastatic breast cancer; recruited 2006-2009; USA
InterventionsEmotionally expressive writing during four 20 minute sessions within 3 weeks at participants convenience. A trained research assistant telephoned women at the beginning of each session to read the instructions to the participant then called again 20 minutes later to ask women to stop writing
OutcomesCentre for Epidemiologic Studies Depression Scale (CES-D); Impact of Event Scale (IES); negative somatic symptoms; The Pittsburgh Sleep Quality Index (PSQI); Perceived emotional support
NotesThis trial has attracted a particular type of patient (white, educated) including from another observational study. Payment of $80/participant
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low risk"Randomization schedule was created by a biostatistician using a computerized random numbers generator"
Allocation concealment (selection bias)Low risk"Participants were randomized to either the emotional or control writing condition and mailed a packet of sealed envelopes containing writing instructions as well as return envelopes and materials for the writing exercise."
Blinding of participants and personnel (performance bias)
All outcomes
Low risk

"Condition assignment was revealed to the assistant reading instructions during the first writing session. After each writing session, women mailed their essays to the research office."

Comment: no clear statement made by authors regarding control group details; no information regarding blinding of either participants or personnel. However, review authors deem this kind of blinding inappropriate for this type of intervention

Blinding of outcome assessment (detection bias)
All outcomes
Unclear risk"An independent rater unaware of condition assignment read all transcribed essays in random order and recorded which condition instructions they most reflected. The rater correctly classified 94% of the essays, indicating excellent adherence to writing instructions."
Incomplete outcome data (attrition bias)
All outcomes
Low risk

"One woman was not able to complete her writing sessions within the 3-week interval and was dropped from the study and the remaining 75 completed the writing...Eleven women died during the follow-up period, and two women did not return their questionnaires and could not be reached. Women who did not complete the project did not significantly differ on any demographic or cancer-related variable or on baseline levels of outcome variables."

"Women who did not complete the project did not significantly differ on any demographic or cancer-related variable or on baseline levels of outcome variables"

Comment: no statistical significance between categories following dropouts and deaths

Selective reporting (reporting bias)Low riskAll measures described in method reported in results
Other biasLow riskStudy appears free from other sources of bias

Scholten 2001

Methods"Prospective" RCT
Participants43 women with metastatic breast cancer (group 2); Austria
InterventionsIndividual psychotherapy. Modeled upon crisis intervention with cognitive and behavioural approach. Self-hypnosis and progressive muscle relaxation therapies
OutcomesQuality of life assessed using semi-structured interview (study entry and 3 months) an visual analogue scale (4 time points: study entry, after the intervention, 3 months, 6 months)
NotesPatients were divided into two groups according to their stage of breast cancer. 41 patients in group one had breast cancer stage 1 or 2. 43 patients in group two had metastatic breast cancer. Both group one and two were each randomised to control and intervention arms. The intervention was the same for both groups
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear risk"The 43 patients were then randomized into one group of 20 patients who were to receive psychosocial support and another group of 23 patients who were to be the waiting list control group and who received no psychosocial support during the observation period."
Allocation concealment (selection bias)Unclear riskInsufficient information to permit judgement
Blinding of participants and personnel (performance bias)
All outcomes
Low riskReview authors deem this kind of blinding inappropriate for this type of intervention
Blinding of outcome assessment (detection bias)
All outcomes
Unclear risk"The (semi-structured) interviews were tape recorded, and for each patient a descriptive rating was performed independently by the interviewer and by another experienced psychologist."
Incomplete outcome data (attrition bias)
All outcomes
High risk

"Overall drop-out rate of 60%" from the intervention arm that originally comprised 20 women with metastatic breast cancer

"Three (15%) rejected the offer at study entry. After the first session, 5 (25%) further patients discontinued support. A further 6 (30%) discontinued therapy after three to five sessions."

Selective reporting (reporting bias)Low risk"The outcome measurements in our investigation was subjective self assessed quality of life"
Other biasLow riskStudy appears free from other sources of bias

Spiegel 1989

  1. a

    CARES-SF: Cancer Rehabilitation and Evaluation System Short Form 
    CBT: cognitive behavioural therapy
    CES-D: Center for Epidemiologic Studies Depression scale
    CG: control group
    HLC: health locus of control
    IES: impact of event scale
    md: median
    MILP: Monash Interview for Liaison Psychiatry
    ns: non-significant
    POMS: Profile of Mood States
    PSQI: Pittsburgh Sleep Quality Index
    QLQ: quality of life questionnaire
    RCT: randomised controlled trials
    SEGT: supportive-expressive group therapy
    SESES-C: the Stanford Emotional Self-Efficacy Scale - Cancer
    TG: telephone group
    TMD: temporomandibular disorder 
    WAI: Weinberger Adjustment Inventory 

MethodsRCT (3:2 towards intervention)
Participants86 women with metastatic breast cancer; recruitment started 1977; USA
InterventionsGroup support including terminal illness problems, improving relationships and communication, discussing concerns, fears and living life to full; control group - assumed to be usual care
OutcomesSurvival at 1-10 years; Profile of Mood States; Health Locus of Control (HLC); self-esteem (Janis-Field); maladaptive coping response; denial measures; phobias
Notes1 of 3 intervention groups had self-hypnosis; other psychological outcomes reported as non-significant differences in HLC, denial or self-esteem; reduction in fear and maladaptive coping reported (no data)
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear risk

"After written informed consent was obtained (protocol approved by Stanford Human Subjects Committee), a battery of psychological tests was administered. The subjects were then randomly assigned to either the intervention or control groups"

Comment: method of randomisation not described

Allocation concealment (selection bias)Unclear riskNot described
Blinding of participants and personnel (performance bias)
All outcomes
Low riskNo blinding occurred. Those allocated to the Group Intervention attended three 90-minute sessions of group therapy. Review authors deem this kind of blinding inappropriate for this type of intervention
Blinding of outcome assessment (detection bias)
All outcomes
Low risk"14 subjects assigned to group therapy were too weak or ill at initial interview to participate; 6 died after entry but before the groups began and 2 others moved away. 12 subjects were lost from the controls; 4 were too ill to participate, 2 died, 4 refused to participate, and we were unable to contact 2. 30 of the 34 women in the intervention group and all 24 control women survived long enough to respond to at least one follow-up questionnaire during the year of study."
Incomplete outcome data (attrition bias)
All outcomes
Unclear risk

"To examine between-group differences among those patients who where more actively involved, we did the same Cox regression analysis on the 54 patients who completed both a baseline and at least one of the three follow-up questionnaires during the year."

"We made phone contact with the 3 survivors. For other than the primary survival analysis, these 3 were treated as though their date of death was July 1, 1988, when all death records had been obtained. If there was any bias resulting from this decision, it would be in the direction of minimising the impact of intervention, since all 3 were in the treatment group."

Comment: no discussion of the number of patients who completed some or all follow-up questionnaires

Selective reporting (reporting bias)Low risk"Here we describe a 10 year follow-up of the effect of psychological intervention on disease progression and mortality"
Other biasUnclear riskNearly significant difference in staging at initial diagnosis. Staging information, based on medical records at study entry, available for 70 of 86. Initial staging favoured intervention group taking place on average 59.8 months before study beginning

Characteristics of excluded studies [ordered by study ID]

StudyReason for exclusion
Andersen 2007Of the 227 participants, authors report 10% had Stage III breast cancer. However, there was not the required level of data with respect to that stage of breast cancer
Andersen 2008Linked to Andersen 2007
Antoni 2009The study had 3 patients with Stage III breast cancer. In addition to the small number, there was not the required level of data with respect to that stage of breast cancer
Arathuzik 1994Does address patients with metastatic breast cancer but is not a randomised controlled trial
Badger 2005Not metastatic disease patients
Bahmani 2010Includes non-metastatic breast carcinoma
Berger 2009The study had 31 patients with Stage III breast cancer. However, there was not the required level of data with respect to that stage of breast cancer
Brietbart 2010Study included patients with advanced cancer. However it did not present the data according to cancer type
Brown 2006Patients with advanced cancer involving brain, head and lung, ovarian, GI, and 'other'. There was not the required level of data with respect to cancer type
Carlson 2003Mixed cancer types not analysed separately
Cerny 1999Non-randomised study
Chan 2006Includes metastatic cancer patients but not analysed separately
Chujo 2005Not metastatic disease patients
Dirksen 2008Included 9 patients with Stage III breast cancer, but did not the required level of data with respect to that stage of breast cancer
Doorenbos 2006Included breast as a solid cancer site but there was not the required level of data with respect to stage of breast cancer
Edelman 2005Not metastatic disease or randomised trial
Edgar 2001Not the required level of data with respect to stage of breast cancer
Fukui 2000RCT of group psychological intervention but predominantly for women with Stage 1 or 2 disease and other medical therapy (radiation, chemotherapy, hormonal); only 3 out of 50 recruited patients had Stage 3 disease
Gellert 1993Follow up of matched but non-randomised groups with or without 'ECaP’ (Exceptional Cancer Patients) group psychosocial support therapy
Graves 2003Not metastatic disease patients
Helgeson 1999Aimed to recruit only patients with Stage 1 or 2 disease. In fact 6% of sample (approximately 19 women) with Stage 3 disease were recruited but results not reported separately and randomisation not stratified for disease stage
Hong 2010Not the required level of data with respect to stage of breast cancer
Ilnyckyj 1994RCT of psychological intervention for cancer patients, but breast cancer patients (for whom separate data are presented) were not randomised and disease status/metastatic status is not reported for this subgroup
Kissane 2003RCT of psychological intervention for breast cancer patients but all patients had either Stage 1 or 2 disease
Laidlaw 2005Excluded on methodological grounds although it was a (pilot) trial of a psychological intervention for women
with metastatic breast cancer. Small study with high loss to follow-up (23/37 participants = 62%, leaving 6, 5
and 3 participants in the study groups at this stage and insufficient information about characteristics to assess
bias). Only one outcome with data from the randomised design was POMS, and this was at 3 months, thus
not consistent with other outcome reporting/extraction for this review’s methods
Lapid 2007Included patients with newly diagnosed advanced cancer but there was not the required level of data with respect to cancer type other than brain, head and neck, lung, ovarian, GI and 'other'
Larson 2000Staging data only on 68% of sample, and of these 28 patients, only 3 had metastatic disease
Lee 2006Included 7 patients with at least Stage III breast cancer but there was not the required level of data with respect to those stages of breast cancer
Leon-Pizzaro 2007Included 10 patients with Stage III and IV gynaecologic and breast cancer but there was not the required level of data with respect to type or stage of breast cancer
Lev 2000Included 18 patients with breast cancer bu there was not the required level of data with respect to stage of breast cancer
Levine 2005Includes patients with metastatic cancer, but not analysed separately according to cancer site
Lieberman 2003Not metastatic disease patients
Linn 1982Only male patients in study
Mantovani 1996Not the required level of data with respect to cancer type
Marchioro 1996Does not address patients with metastatic breast cancer
McArdle 1996Does not address patients with metastatic breast cancer
Michalec 2005Includes patients with metastatic cancer, but not analysed separately according to cancer site
Nidich 2009Included women with stages II to IV breast cancer, but there was not the required level of outcome data with respect to stage of breast cancer
Northouse 2005Included metastatic disease patients, but not analysed separately
Rosenbaum 2004Not randomised trial
Rummans 2006Mixed cancer types, breast cancer patients not analysed separately
Salzer 2010Includes non-metastatic breast cancer
Sandgren 2007Included 23 women with Stage III breast cancer, but there was not the required level of outcome data with respect to that stage of breast cancer
Sloman 2002Not the required level of data with respect to cancer type
Targ 2002Entry criterion was within 18 months of diagnosis, not related to extent of disease; 80% of participants had Stage 1 or Stage 2 disease (non-metastatic); data for women with Stages 3 or 4 (metastatic) are not reported separately
Traeger-Muney 2003Not randomised design
van der Pompe 1997Only 6 out of 23 study patients had metastatic disease and effects of intervention on patients with metastatic
disease not reported separately
Wang 2006Not the required level of data with respect to cancer type

Ancillary