Psychological interventions for women with non‐metastatic breast cancer
Abstract
Background
Breast cancer is the most common cancer affecting women worldwide. It is a distressing diagnosis and, as a result, considerable research has examined the psychological sequelae of being diagnosed and treated for breast cancer. Breast cancer is associated with increased rates of depression and anxiety and reduced quality of life. As a consequence, multiple studies have explored the impact of psychological interventions on the psychological distress experienced after a diagnosis of breast cancer.
Objectives
To assess the effects of psychological interventions on psychological morbidities, quality of life and survival among women with non‐metastatic breast cancer.
Search methods
We searched the following databases up to 16 May 2013: the Cochrane Breast Cancer Group Specialised Register, CENTRAL, MEDLINE, EMBASE, CINAHL and PsycINFO; and reference lists of articles. We also searched the World Health Organization International Clinical Trials Registry Platform (WHO ICTRP) search portal and ClinicalTrials.gov for ongoing trials in addition to handsearching.
Selection criteria
Randomised controlled trials that assessed the effectiveness of psychological interventions for non‐metastatic breast cancer in women.
Data collection and analysis
Two review authors independently appraised and extracted data from eligible trials. Any disagreement was resolved by discussion. Extracted data included information about participants, methods, the intervention and outcome.
Main results
Twenty‐eight randomised controlled trials comprising 3940 participants were included. The most frequent reasons for exclusion were non‐randomised trials and the inclusion of women with metastatic disease. A wide range of interventions were evaluated, with 24 trials investigating a cognitive behavioural therapy and four trials investigating psychotherapy compared to control. Pooled standardised mean differences (SMD) from baseline indicated less depression (SMD ‐1.01, 95% confidence interval (CI) ‐1.83 to ‐0.18; P = 0.02; 7 studies, 637 participants, I2 = 95%, low quality evidence), anxiety (SMD ‐0.48, 95% CI ‐0.76 to ‐0.21; P = 0.0006; 8 studies, 776 participants, I2 = 64%, low quality evidence) and mood disturbance (SMD ‐0.28, 95% CI ‐0.43 to ‐0.13; P = 0.0003; 8 studies, 1536 participants, I2 = 47%, moderate quality evidence) for the cognitive behavioural therapy group than the control group. For quality of life, only an individually‐delivered cognitive behavioural intervention showed significantly better quality of life than the control with an SMD of 0.65 (95% CI 0.07 to 1.23; P = 0.03; 3 studies, 141 participants, I2 = 41%, very low quality evidence). Pooled data from two group‐delivered studies showed a non‐significant overall survival benefit favouring cognitive behavioural therapy compared to control (pooled hazard ratio (HR) 0.76, 95% CI 0.25 to 2.32; P = 0.63; 530 participants, I2 = 84%, low quality evidence). Four studies compared psychotherapy to control with one to two studies reporting on each outcome. The four studies were assessed as high risk of bias and provided limited evidence of the efficacy of psychotherapy. Adverse events were not reported in any of the included studies.
Authors' conclusions
A psychological intervention, namely cognitive behavioural therapy, produced favourable effects on some psychological outcomes, in particular anxiety, depression and mood disturbance. However, the evidence for survival improvement is still lacking. These findings are open to criticism because of the notable heterogeneity across the included studies and the shortcomings of the included studies.
PICOs
Plain language summary
Use of psychological interventions in women diagnosed and under treatment for non‐metastatic breast cancer
Review question
We reviewed the evidence for the effect of psychological interventions on the psychological impact, quality of life and survival among women with non‐metastatic breast cancer (that is cancer that has not spread beyond the breast).
Background
Breast cancer is the most common cancer affecting women worldwide. Being a distressing diagnosis, considerable research has examined the psychological consequences of being diagnosed and treated for breast cancer. Breast cancer diagnosis and treatment can cause depression and anxiety and reduce quality of life. As a result, various psychological interventions have been utilised to help address the psychological distress experienced after a diagnosis of breast cancer.
Study characteristics
The evidence was current to May 2013. An intervention could be delivered in a group setting (group intervention), as one to one contact between a therapist and a patient (individual intervention) or in the form of couple therapy where the patient and her spouse attends the therapy sessions (couple intervention). The control group could receive educational leaflets or have access to seminars or relaxation classes. A comprehensive search of the literature was conducted and 28 studies comprising 3940 participants were included. The majority (24 out of 28 studies) of interventions were based on cognitive behavioural therapy, which involves changing a person's thoughts and behaviour. Four studies used psychotherapy as the intervention. Generally, the methods for assessing outcomes (such as anxiety, depression, quality of life) after the intervention and the timing of these assessments were not uniform across studies.
Key results
Women who received cognitive behavioural therapy showed important reductions in anxiety, depression and mood disturbance, especially when it was delivered to groups of women. An improvement in quality of life was observed when women received individual cognitive behavioural therapy compared to the control group. The effects on survival were uncertain because the results were imprecise.
The four psychotherapy studies reported limited information for each outcome. Therefore no firm conclusion could be made about the efficacy of psychotherapy.
Adverse events were not reported in any of the included studies.
Further research should aim to provide evidence for people to make informed decisions about whether the effects of these treatments are sustainable after discontinuation of the therapy.
Quality of the evidence
The quality of evidence ranged from very low quality (for example for quality of life, individually delivered intervention) to moderate quality evidence (for mood disturbance). The interventions varied between studies as did the methods and timing of outcome measures and treatment received within the control groups.
Authors' conclusions
Summary of findings
| Cognitive behavioural therapy versus control for women with non‐metastatic breast cancer | ||||
| Patient or population: Women with non‐metastatic breast cancer | ||||
| Outcomes | Illustrative comparative risks* (95% CI) | No of Participants | Quality of the evidence | Comments |
| Corresponding risk | ||||
| Cognitive behavioural therapy versus control | ||||
| Depression | The mean depression in the intervention groups was | 637 | ⊕⊕⊝⊝ | SMD ‐1.01 (95% CI ‐1.83 to ‐0.18) |
| Anxiety | The mean change in anxiety in the intervention groups was | 776 | ⊕⊕⊝⊝ | SMD ‐0.48 (95% CI ‐0.76 to ‐0.21) |
| Mood disturbance | The mean mood disturbance in the intervention groups was | 1536 | ⊕⊕⊕⊝ | SMD ‐0.28 (95% CI ‐0.43 to ‐0.13) |
| Quality of life: Group‐delivered intervention | The mean quality of life: group‐delivered intervention in the intervention groups was | 578 | ⊕⊕⊝⊝ | SMD 0.21 (95% CI ‐0.03 to 0.46) |
| Quality of life: Individually‐delivered intervention | The mean quality of life: individually‐delivered intervention in the intervention groups was | 141 | ⊕⊝⊝⊝ | SMD 0.65 (95% CI 0.07 to 1.23) |
| *The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||
| GRADE Working Group grades of evidence | ||||
| 1 A higher score on either tool indicated higher depression | ||||
Background
Description of the condition
Breast cancer is the most common cancer in women of all ages. About 1.3 million women will be diagnosed with breast cancer every year worldwide and about 465,000 will die from the disease making it the third leading cause of death in women, after heart disease and lung cancer (ACS 2009). Improved prevention and detection methods, as well as advances in medical treatment, have resulted in a trend toward increasing numbers of cancer survivors (ACS 2009). Survival gains achieved in breast cancer have produced a growing acceptance of breast cancer as a long‐term illness and have led to a greater emphasis on rehabilitation and subsequently the quality of life (QoL) of these women (Reynolds 2000). Despite the growing recognition of the need to address the psychological requirements of patients with cancer, the current practice has lagged in targeting the psychological elements of living with cancer resulting in disparity in comprehensive cancer care (Greer 2013).
Breast cancer is still a distressing diagnosis and, as a result, considerable research has examined the psychological sequelae of being diagnosed and treated for breast cancer. In line with the increasing adoption of a bio‐psychosocial model of health care, one focus of interest has been to determine whether a diagnosis of breast cancer is associated with specific psychological disorders, and what course these take in patients (Fann 2008; Okamura 2005; Reich 2008). Psychological morbidities (the incidence or prevalence of psychological disorders) such as anxiety, depression, stress, distress, difficulty in adjustment and decreased social interactions (Vos 2004) are common responses to the diagnosis and treatment of breast cancer. Such responses may arise from pain (Reddick 2005), fear of recurrence, treatment side‐effects, life stresses (Low 2006) and lymphoedema (McWayne 2005). Many women consider chemotherapy as the most distressing aspect of treatment as it is usually associated with unpleasant symptoms such as nausea, emesis, fatigue and alopecia (Boehmke 2005; Partridge 2001). The debilitating effects are more profound in younger women who also experience the sudden onset of early menopause with the attendant symptoms of hot flashes, decreased sexual desire and vaginal dryness (Baucom 2005; Partridge 2004). Notably, younger women may have specific fertility needs and concerns (Peate 2009).
The considerable data available in relation to psychological conditions associated with breast cancer suggest that depression and anxiety are the most commonly studied mood disorders (Fann 2008). Rates of major depression or anxiety in breast cancer patients have been estimated to range from 20% to 30% in the initial six months following breast cancer diagnosis in women with early stage breast cancer (Akechi 2001;Fallowfield 1990). However, a recent study suggests that the number of patients approaching the threshold for depressive and anxiety disorders (including borderline cases) is close to 50% in the first year after diagnosis, dropping rapidly in the second year to 25% and following a further gradual sustained decrease over the five‐year study period to 15% in the fifth year (Burgess 2005). On the other hand, results from the UK Standardisation of Radiation Therapy Trials (START) showed that about one third (35%) of women reported anxiety or depression, or both, which did not significantly change over five years of follow up (Hopwood 2010).
A review of 37 studies on the epidemiology of major depression in women with breast cancer suggested a rate of 10% to 25% (Fann 2008). In a large cohort of 4496 patients with cancer, the prevalence of depression in those with breast cancer was estimated to be around 52% (Zabora 2001). This wide variation in rates of psychological disorders may be attributed to methodological differences across studies, different characteristics of the groups studied, heterogeneous tumour stage and the stage at which assessment took place in relation to diagnosis or treatment. Moreover, the use of different diagnostic criteria for depression and anxiety may have contributed to the different rates reported in the above studies (Chochinov 1994). Predictors of psychological morbidity following breast cancer diagnosis and treatment were primarily related to the patient (namely younger age, previous psychological problems and a lack of social support) rather than to the disease or treatment (Burgess 2005). Socio‐economically deprived patients were also at risk of depression at three to four months after surgery (Christensen 2009). Adjuvant chemotherapy was found to increase the risk of depression or anxiety, or both, during but not after treatment (Burgess 2005). These side‐effects varied depending on the specific agents used in the adjuvant regimen as well as the dose and duration of treatment (Boehmke 2005).
One quarter of the women studied maintained clinically significant levels of distress over a 12‐month period (Millar 2005). Distress emerged or intensified when women expected symptoms to disappear but they continued to persist (Rosedale 2010). However, the most intense or frequently occurring symptom is not necessarily the most distressing to patients (Bárez 2009; Henselmans 2009). Most research has focused on the identification of predictive variables related to higher levels of symptom distress, such as age (Baider 2003), coping style (Ben‐Zur 2001), baseline anxiety and depression, and fear of recurrence (Lebel 2009). When compared to a cohort of matched women free of disease, women with breast cancer exhibited significantly higher levels of distress and different coping styles from their counterparts in the control group (Amir 2002).
In addition to specific psychological disorders such as anxiety and depression, over the last two decades QoL outcomes have been increasingly used as an outcome variable in breast cancer research (Hewitt 2004). These studies have collectively examined QoL outcomes for women diagnosed at different ages, various stages of the disease, and at different time intervals between diagnosis and treatment (Ganz 2002). Most but not all data suggest that a younger age and shorter duration of time from diagnosis are associated with poorer QoL (Ganz 2002; Mols 2005). Social support from family members and friends helps to decrease the negative effects of symptoms on QoL (Ashing‐Giwa 2009; Kulik 2005; Manning‐Walsh 2005) and improve women's adjustments and ability to cope (Bloom 1982). Social deprivation was also related to poor breast cancer prognosis (Vona‐Davis 2009).
In addition to the impact on psychological disorders and QoL, it has been suggested that psychological distress following breast cancer diagnosis and treatment may adversely affect survival (Spiegel 1989), although this latter outcome is controversial (Smedslund 2004). Moreover, fighting spirit was linked to recurrence‐free survival in one of the pioneering studies in this field (Greer 1979).
As a consequence of the effects of breast cancer on depression, anxiety and QoL, various psychological interventions have been utilised to help address the psychological distress experienced after a diagnosis of breast cancer. This systematic review seeks to assess the effectiveness of those therapies that have been subjected to controlled trials thus far.
Description of the intervention
Psychological intervention includes a wide range of therapeutic techniques and is poorly defined in the literature.
Cognitive behavioural therapy (CBT) is a psychotherapeutic approach that aims to solve problems concerning dysfunctional cognition, emotions and behaviours through a goal‐oriented systematic procedure. CBT includes a variety of approaches and therapeutic systems; some of the most well known include cognitive therapy, rational emotive behaviour therapy and multimodal therapy. CBT focuses on changing specific thoughts or behaviours or on learning specific coping skills (Hopko 2008), such as progressive muscle relaxation training, meditation, hypnotherapy, systematic desensitisation, biofeedback, behaviour modification or reinforcement and cognitive therapy. In the past decade, research has supported mindfulness‐based therapies such as meditation for a number of medical and psychological conditions accompanying breast cancer diagnosis and treatment (Carlson 2003; Lengacher 2009). In recent years, CBT therapists have witnessed a shift towards focused therapies such as acceptance and commitment therapy (ACT) (Dahl 2004).
Psychotherapy, or personal counselling with a psychotherapist, is an intentional interpersonal relationship used by trained psychotherapists to aid a client or patient in problems of living. It includes non‐directive, psychodynamic, existential, supportive, general or crisis intervention; no specific behavioural or coping skills are taught (Barsevick 2002).
Group psychotherapy, or group therapy, is a form of psychotherapy in which one or more therapists treat a small group of clients (Montgomery 2002). The term refers to any form of psychotherapy when delivered to a group, including CBT, interpersonal therapy and psychodynamic group therapies; the group process explicitly using mechanisms of change by developing, exploring and examining interpersonal relationships within the group.
A psycho‐educational intervention is the education offered to people who live with a psychological disturbance. Frequently, psycho‐educational training involves patient training courses in the context of treating a physical illness (Bäuml 2008). Family members are also included in the education. Patients and their relatives are empowered to understand and accept the illness and cope with it in a successful manner (Bäuml 2008). The goal is for the patient to understand and deal with the presenting illness. Also, the patients' own strengths, resources and coping skills are reinforced in order to avoid relapse and for them to contribute to their own health and wellness on a long‐term basis. The theory considers that the better knowledge the patient has of her illness, the better the patient can live with her condition.
How the intervention might work
The purpose of psychological support programmes in breast cancer is to promote awareness and education, provide emotional support, and assist women with problem solving so that they can go through the processes and cope better with cancer (Sandgren 2000). It has been suggested that understanding the uncertainty that is experienced plays a key role in positively influencing future behaviours (Montgomery 2010). For example, CBT was the most frequently used approach in studying the effect of psychological intervention in cancer patients (Moyer 2009; Redd 2001) and has been shown to be a valuable tool in relieving distress in various cancer populations (Mundy 2003), particularly amongst breast cancer patients (Tatrow 2006).
Why it is important to do this review
There is an accumulating amount of research concerned with the effects of psychological interventions on QoL and psychological morbidity in women with non‐metastatic or early stage breast cancer. Yet the strength of this relationship is unknown because studies have not been combined into a systematic review. Several items pertaining to psychological interventions in early stage breast cancer remain unresolved; for instance, which interventions work the best, for which patients, and other items related to the therapy (for example duration, dose, type and optimal time to start therapy). Hence, the true effect of psychological interventions on the QoL of women with non‐metastatic breast cancer remains unclear. We conducted this systematic review to explore the uncertainty arising from conflicting results in a number of studies in this area.
Previous reviews of psychological intervention were conducted on women with metastatic breast cancer, which limited the applicability of the review to the larger group of women with non‐metastatic breast cancer (Edwards 2008; Mustafa 2013). Indeed, one might question whether women with non‐metastatic breast cancer have different outcomes to those women with metastatic breast cancer, especially as early detection and treatment continue to improve and patients continue to live longer.
Objectives
To assess the effects of psychological interventions on psychological morbidities, QoL and survival among women with non‐metastatic breast cancer.
Methods
Criteria for considering studies for this review
Types of studies
Randomised controlled trials which compare any form of psychological or behavioural intervention with a placebo, waiting list control or an alternative form of psychological intervention.
Types of participants
Women with a histologically confirmed diagnosis of breast carcinoma of an early non‐metastatic stage (Grade I‐III) as defined by the American Joint Committee on Cancer (AJCC) TNM (tumour, lymph nodes, metastasis) staging system (American Joint Committee on Cancer 2009).
The following studies were excluded:
-
studies including women with distant metastasis (Grade IV) unless there were subgroup analyses;
-
studies including patients with other types of cancer unless there were subgroup analyses of breast cancer groups;
-
studies about psychological intervention in caregivers of women with breast cancer as they represent a different entity.
Types of interventions
A range of psychological interventions to prevent or treat psychological distress were eligible for inclusion:
-
cognitive behavioural techniques;
-
psychotherapy or counselling; and
-
psycho‐educational interventions.
We compared these interventions to an inactive control intervention; that is placebo, standard care or waiting list control: 'a group that is assigned to a waiting list to receive an intervention after the active treatment group does', or with an active control intervention (for example another form of psychological intervention). Studies with multi‐interventions were excluded unless data were extractable.
Types of outcome measures
Primary outcomes
-
Depression: depression score measured using any validated disease specific tool at the end of the study
-
Anxiety: anxiety score measured using any validated disease specific tool at the end of the study
-
Stress: stress score measured using any validated disease specific tool at the end of the study
-
Mood disturbance: mood disturbance score measured using any validated disease specific tool at the end of the study
Secondary outcomes
-
Quality of life (QoL): QoL score measured using any validated questionnaire at the end of the study
-
Coping: coping score measured using any validated disease specific tool at the end of the study
-
Adjustment: adjustment score measured using any validated disease specific tool at the end of the study
-
Survival: all cause patient survival at the latest study time point
Search methods for identification of studies
See: Breast Cancer Group methods used in reviews
There were no language limits. Articles in all languages were searched and relevant abstracts were translated.
Electronic searches
We searched the following databases.
-
The Cochrane Breast Cancer Group's (CBCG) Specialised Register maintained by the Cochrane Breast Cancer Group was searched on the 16 May 2013. Details of the search strategies used by the CBCG for the identification of studies and the procedure used to code references are outlined in their module (http://onlinelibrary.wiley.com/o/cochrane/clabout/articles/BREASTCA/frame.html). Trials with the key words 'psychological intervention', 'early breast cancer', 'cognitive behavioural technique', 'cognitive behavioural therapy', 'psychotherapy', 'psycho‐educational therapy', 'psychotherapeutic', 'CBT' and 'acceptance and commitment therapy' were extracted for consideration.
-
The Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2013, Issue 3). See Appendix 1.
-
MEDLINE (via OvidSP) (July 2008 to 16 May 2013). See Appendix 2.
-
EMBASE (via Embase.com) (2008 to 16 May 2013). See Appendix 3.
-
PsycINFO (via OvidSP) (1960 to 16 May 2013). See Appendix 4.
-
CINAHL (via EBSCO) (1982 to 16 May 2013). See Appendix 5.
-
The World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) search portal (http://apps.who.int/trialsearch/Default.aspx) for all prospectively registered and ongoing trials (16 May 2013). See Appendix 6.
-
Clinicaltrials.gov (http://clinicaltrials.gov/ct2/search) (16 May 2013). See Appendix 7.
-
Other databases: CancerLit, PsyLit, Iranmedex, Indmed (July 2008 to May 2013). See Appendix 8; Appendix 9; Appendix 10; Appendix 11.
Searches were initially undertaken using key words then followed by searches of related articles until previously selected papers were obtained. Key words used were: breast cancer, breast neoplasm, quality of life, well‐being, depression, anxiety, stress, distress, adjustment, coping, mental health, health‐related quality of life, psychological intervention, psychological morbidity, psychiatric morbidity, cognitive behavioral therapy, group psychotherapy, relaxation, supportive therapy, visual imagery and psychosocial intervention.
Searching other resources
Searches also included the following.
(a) Bibliography searching. The bibliographies of all included studies and review papers were searched in order to identify other potentially suitable studies. Articles cited by relevant studies were reviewed. Language restrictions were not imposed. Full translations of all non‐English language papers were conducted using local resources.
(b) Unpublished literature. Experts in this field were contacted. Letters were sent to all authors of included studies requesting information on unpublished data or ongoing studies.
(c) Handsearching of journals. A list of journals currently being handsearched by The Cochrane Collaboration is available at the US Cochrane Center Handsearch master list page (http://apps1.jhsph.edu/cochrane/masterlist.asp).
Data collection and analysis
Selection of studies
Two authors (GJ and DW) independently assessed the titles and abstracts of each identified trial for inclusion into the review. After the initial assessment, we obtained full versions of all potentially relevant articles. A third author (AH) was approached to resolve any discrepancies regarding eligibility.
If the results of a randomised controlled trial (RCT) were unpublished but available, and three authors (GJ, DW and AH) were satisfied with the quality of the data, data were included (where possible) and disclosed in the discussion section. Trials were included if randomisation and patient preference allocation arms analysis were performed. If these analyses were not completed, the trials were dealt with separately because of the risk of allocation bias.
Additional data or information were sought from the principal investigator of the trial concerned, where necessary.
Data extraction and management
Data from all relevant studies were extracted and entered into the 'Characteristics of included studies' table in RevMan 5.2 (RevMan). All studies were appraised independently by two review authors (GJ, DW). Any disagreement was resolved by discussion. Extracted data included the following.
(a) Participants: country of origin, sample size, setting, diagnostic criteria, age, ethnicity, date of study and data on baseline psychological morbidity for assessment of effect modifiers.
(b) Methods: study design, methods of allocation, allocation sequence concealment, blinding, exclusion of participants after randomisation, proportion and reasons for loss at follow up.
(c) Interventions: type, dose, length and frequency of intervention (for each intervention and comparison group).
(d) Outcomes: primary and secondary outcomes using validated instruments were extracted.
If mentioned, sources of funding were recorded in the 'Characteristics of included studies'.
Assessment of risk of bias in included studies
Two review authors (GJ and DW) graded and assessed each selected trial using a simple contingency form, addressing the seven specific domains discussed in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). The evaluations given by all authors were compared and any inconsistencies and disagreements were resolved by discussion. Each domain was assigned a judgement related to the risk of bias in that domain. Judgements used were: 'low risk of bias', ‘high risk of bias’, and 'unclear', which indicated unclear or an unknown risk of bias. The domains were:
-
sequence generation;
-
allocation concealment;
-
blinding of participants and personnel;
-
blinding of outcome assessors;
-
incomplete outcome data;
-
selective outcome reporting; and
-
other sources of bias.
Assessment of these domains for each trial are reported in the Risk of bias in included studies. We also categorised and reported the overall risk of bias of each of the included studies according to the following:
-
low risk of bias (bias is unlikely to seriously alter the results) if all criteria were met;
-
unclear risk of bias (bias raises some doubt about the results) if one or more criteria were assessed as unclear;
-
high risk of bias (bias seriously weakens confidence in the results) if one or more criteria were not met.
Measures of treatment effect
The data could have been presented as continuous (for example changes in depression scales), dichotomous (for example either depressed or not depressed), ordinal (for example categories on a QoL scale such as mild, moderate and severe) or time‐to‐event data (for example survival data). Decisions regarding if and how to combine these outcomes were made depending on how the data were collected by each trial. These decisions were guided by section 9.2 'Types of data and effect measures' in the Cochrane Handbook for Systematic Reviews of Interventions 5.0.0 (Higgins 2011).
For continuous data, we presented the change from baseline as the standardised mean difference (SMD) due to the same outcome being measured in a variety of ways using different scales. For future updates, the mean and standard deviation will be reported if possible (that is when the outcome measurements in all studies are made on the same scale).
Dichotomous data were not used in this version of the review. In future updates of the review, if presented with dichotomous data and the authors have specified a cut‐off point for determining clinical effectiveness, we will use this where appropriate. Otherwise, cut‐off points on rating scales will be identified and participants will be divided on the basis of whether they were clinically improved or not. For dichotomous outcomes in future updates, a Mantel‐Haenszel odds ratio with its associated 95% confidence interval (CI) will be estimated.
In the case of time‐to‐event data (overall all‐cause survival and progression‐free survival), intervention effects were expressed as hazard ratios.
All measures of effect included 95% CIs, P values, and for pooled measures the I2statistic.
Unit of analysis issues
No studies were identified that experienced unit of analysis issues and it is unlikely that studies of this type will be identified in future updates. If these complications are found in updates, guidance will be taken from Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) and changes between the protocol and the updated review will be highlighted.
Dealing with missing data
If data were missing from trials less than 10 years old, we tried wherever possible to contact the investigators or sponsors of these studies. We re‐analysed data according to the intention‐to‐treat (ITT) principle whenever possible. If authors had conducted a per protocol analysis instead of an ITT analysis these results were included. Where baseline and follow‐up data only were summarised, the change from baseline scores have been estimated assuming a common correlation of structure of 0.8 (section 16.1.3.2 in the Cochrane Handbook for Systematic Reviews of Interventions) (Higgins 2011). In cases where variability data were summarised at baseline but not at follow up, the variance was assumed to have remained unchanged. In cases where medians were presented with ranges, the mean was estimated by the median and the variance by using the range and the number of observations (Hozo 2005).
Assessment of heterogeneity
To check for statistical heterogeneity between studies, both the I2 statistic and Chi2 test of heterogeneity as well as visual inspection of the forest plots were used. The graphical representation of the data was inspected; if CIs for the results of individual studies have poor overlap it generally indicates the presence of statistical heterogeneity. In addition, the Chi2 test was performed to check for differences between the results of each included trial. A P value of 0.10, rather than the conventional level of 0.05, was used to determine the statistical significance. A low P value provides evidence of heterogeneity of intervention effects Higgins 2011. The I2 statistic was used to quantify inconsistency across studies.
We assessed clinical heterogeneity by examining the characteristics of the studies, the similarity between the types of participants and the interventions. We reported heterogeneity as important if it was at least moderate to substantial with the I2 statistic greater than 60% (Higgins 2011). If this could be explained by clinical reasoning and a coherent argument could be made for combining the studies, these were entered into a meta‐analysis. In cases where the heterogeneity could not be adequately explained, the data were not pooled.
The clinical diversity between the studies included in this review as well as the limited number of studies that could be combined for each intervention allowed us to make assessments of heterogeneity between the studies for only one of the comparisons.
Assessment of reporting biases
We followed the recommendations on testing for funnel plot asymmetry as described in section 10.4.3.1 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). In future updates of this review, tests for funnel plot asymmetry will be used only when there are at least 10 studies included in the meta‐analysis. When there are fewer than 10 studies the power of the tests is too low to distinguish chance from real asymmetry. We will visually inspect the funnel plots and discuss reasons for funnel plot asymmetry in the discussion. We will follow the proposed tests for funnel plot asymmetry as outlined in table 10.4.b in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).
Data synthesis
One review author (BC) analysed the data in RevMan and reported them in accordance with the advice in Chapter 9 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Where data could be pooled, for continuous outcomes an inverse variance approach was taken to meta‐analysis fitting a random‐effects model. The degree of heterogeneity was summarised with the I2 statistic. The time‐to‐event data were pooled using the generic inverse variance method and the hazard ratio presented.
Subgroup analysis and investigation of heterogeneity
Subgroup analyses were conducted to test the interaction between the variables determined a priori and the overall treatment effect. In this review, two of the six pre‐specified subgroup analyses were possible using the available data: (1) different doses of psychological intervention (≤ 20 hours versus > 20 hours) and (2) group versus individually‐delivered intervention.
If possible, in future review updates we will consider conducting the following.
-
Age of participants (≤ 50 years versus > 50 years).
-
Dose of psychological intervention (≤ 20 hours versus > 20 hours).
-
Duration of psychological intervention (≤ 8 weeks versus > 8 weeks).
-
Type of psychological intervention (individual versus group).
-
Type of therapy received (total mastectomy versus conservative surgery, chemotherapy, radiotherapy and hormonal therapy).
-
Time point at which the outcome (ie with placebo too) was assessed (≤ 4 months after surgery versus > 4 months).
Sensitivity analysis
The impact of the methodological quality on overall effect size was determined by sensitivity analyses. Sensitivity analyses were conducted to assess the robustness of the review results by carrying out the following, where possible:
-
where heterogeneity existed, the removal of those studies that were considered to be estimating a different effect;
-
removal of studies with high risk of bias and including studies with low risk of bias only.
Results
Description of studies
See: Figure 1; Characteristics of included studies table and Characteristics of excluded studies table.
Figure 1. Study flow diagram.
Results of the search
We retrieved 2854 references to studies. After examination of the titles and abstracts of these references, we deleted duplicates and eliminated all of those which did not match our inclusion criteria and those which were clearly ineligible. We obtained full text copies of the remaining 150 potentially eligible studies for further evaluation. Seven of these needed translation (four in Spanish (Gabaldon 1993; Greer 1992; Narváez 2008; Paez 2007), two Chinese (Feng 1998; Yan 2001) and one French (Cayrou 2005)). Five studies were translated prior to assessment of eligibility, one was excluded based on the abstract alone (Gabaldon 1993) and one is awaiting classification (see Characteristics of studies awaiting classification table).
Twenty‐eight studies proved to be eligible for inclusion in this review. One hundred and fourteen studies were excluded and reasons for the exclusion of 49 key studies are provided in the Characteristics of excluded studies table. For further details see the ’Study flow diagram’ (Figure 1).
Included studies
We included 28 randomised controlled clinical trials comprising 3940 participants. The studies were very heterogenous in the interventions studied, the reporting of outcomes and the point at which the outcome was measured. In order not to replicate participants and findings, we combined six parent trials with their matching substudies because the same participants were used but the authors reported different outcomes in different published versions of the study. Andersen 2004 had an additional two studies (Andersen 2007; Andersen 2008); Antoni 2006 had two additional studies (Antoni 2006B; Antoni 2009); Graves 2003 had one additional Doctor of Philosophy (PhD) publication (Graves 2001); Kissane 2003 had one more publication (Kissane 2004); Mishel 2005 had one additional publication (Gil 2006); and Vos 2004 had one additional publication (Vos 2007).
Characteristics of the trial setting and methods
Only randomised controlled trials were included. Of the studies that were included, 15 were conducted in the US, 6 in Europe, 1 in Japan, 2 in Australia, 2 in Canada, 1 in Israel and 1 in Brazil. All but two trials (Marchioro 1996; Richardson 1997) were conducted in 2000 or beyond. The duration of the studies was between 3 weeks and 12 months. Most studies had two groups (control and intervention), whereas three studies compared three different arms (Cohen 2007; Richardson 1997; Vos 2004).
Characteristics of the participants
The number of participants included in the individual studies varied widely from 14 to 575 participants. The age of the participants ranged from 18 to 80 years, with the most common lower and upper limits being 18 and 65 years, respectively. All of the women had been diagnosed with early stage breast cancer with no metastases. Three studies (Badger 2007; Baucom 2009; Manne 2007) included women with their partners, whereas all other studies included the women only. The overall sample was generally skewed towards Caucasians. In five studies participants were undergoing adjuvant treatment (Badger 2007; Classen 2008; Cohen 2007; Nunes 2007; Yates 2005) and in five studies they were awaiting chemotherapy (Andersen 2004; Antoni 2001; Antoni 2006; Garssen 2013; Marchioro 1996). In the remaining studies participants had completed adjuvant therapy.
Characteristics of the interventions
A wide range of interventions were evaluated. The majority (24 out of the 28 trials) of interventions were based on cognitive behavioural theory. Most therapeutic interventions were delivered face‐to‐face and only three were via the telephone (Badger 2007; Marcus 2010; Mishel 2005). In some studies, mixed approaches were used (face‐to‐face and telephone calls or individual counselling) (Fillion 2008; Loprinzi 2011; Yates 2005). Three studies delivered the intervention to couples as opposed to women only (Badger 2007; Baucom 2009; Manne 2007). Twenty of the interventions were delivered in groups, including one couple intervention (Manne 2007); and seven were delivered individually, including two couple studies (Badger 2007; Baucom 2009). The mean ± SD duration of the intervention was 14 ± 9.65 hours with a median of 12 (maximum 39 hours and minimum 1 hour). Only one study did not specify the duration of the intervention (Marchioro 1996). Most intervention sessions were delivered on a weekly basis.
The psychological interventions in the studies can be categorised into the following.
-
Cognitive behavioural interventions: cognitive, cognitive behavioural, and behavioural methods focused on changing specific thoughts or behaviours or on learning specific coping skills. Procedures coded here included progressive muscle relaxation training, meditation, systematic desensitization, biofeedback, and behaviour modification or reinforcement (Andersen 2004; Antoni 2001; Antoni 2006; Baucom 2009; Boesen 2011; Cohen 2007; Dolbeault 2009; Fillion 2008; Ferguson 2012; Fukui 2000; Garssen 2013; Graves 2003; Henderson 2012; Kissane 2003; Loprinzi 2011; Manne 2007; Marchioro 1996; Mishel 2005; Narváez 2008; Nunes 2007; Richardson 1997; Simpson 2001; Taylor 2003; Yates 2005).
-
Psychotherapy counselling: non‐cognitive and non‐behavioural verbal psychotherapy and counselling including psychodynamic, existential, supportive or general counselling, and crisis intervention (Badger 2007; Classen 2008; Marcus 2010; Vos 2004).
-
Informational and psycho‐educational: most of the trials included an element of education in the protocol to reinforce the existing therapy. No studies used this type of intervention solely; it was an add on to the existing psychological treatment.
Characteristics of the outcome measures
In general, the methods of measurement and the timing of the assessments were not uniform across studies. Even when the same tool was used, in some studies only some subscales or domains were used. In other instances the short form of the original questionnaire was used, which further increased the heterogeneity in the data.
Depression was measured by several tools such as: the Beck Depression Inventory (BDI), the depression subscale of the Hospital Anxiety and Depression Scale (HADS), and the Center for Epidemiological Studies‐Depression Scale (CES‐D).
Anxiety was measured using the State Trait Anxiety Inventory (STAI), the Beck Anxiety Inventory (BAI), the anxiety subscale of the HADS, the Hamilton Rating Scale for Anxiety (HAM‐A) and Smith Anxiety Scale.
Mood disturbance was mostly measured using the Profile of Mood States (POMS) and one study used the Affect Balance Scale (ABS) (Kissane 2003), which is composed of four positive and four negative subscales. The POMS has six subscales (anxiety, depression, vigour, anger, fatigue and confusion) and a global score. The Total Mood Disturbance Score (TMDS) is the sum of five scales (anxiety, depression, anger, fatigue and confusion) minus the score of the vigour scale. In some studies not all subscales were used, for example in Fillion 2008 only anxiety and depression subscales were used; in Antoni 2001 anxiety, depression and anger were used; however in Vos 2004 all five subscales were used. In other studies the short form was used (Mishel 2005; Richardson 1997).
Stress and distress were used interchangeably and were measured using the Impact of Events Scale (IES), perceived stress scale, Brief Symptom Inventory (BSI), and the Mental Health Inventory (MHI).
Coping and adjustment were measured using the Mental Adjustment to Cancer (MAC) scale, the Cancer Behavior Inventory (CBI), the 16‐PF personality questionnaire Form A, the Index Introject Questionnaire, the Dealing with Illness Inventory, the Coping Strategies Questionnaire (CSQ), the Ways of Coping Questionnaire (WCQ) and the Utrecht Coping List (UCL).
Quality of life outcomes were reported in nine trials and were measured using the EORTC C‐30 and the QLQ‐BR23, the Functional Assessment of Cancer Therapy‐Breast (FACT‐B) measure, the Fu (LASA) and Quality of life cancer survivors (QoL CS), Functional Living Index Cancer (FLIC), the Quality of Life Index (QLI), the Cancer Rehabilitation and Evaluation System short form (CARES‐SF) and the Linear Analog Self Assessment Scale. The EORTC score has been reported as a raw score or as a transformation score. To increase comparability the raw score was converted to the transformation score using the formula in the EORTC scoring manual.
All‐cause survival was reported in three studies only (Andersen 2008; Boesen 2011; Kissane 2004) and was measured as time from randomisation to death at the end of the follow up period.
When two scales were used to measure the same outcome, the one which was used mostly by other studies was chosen for consistency of results and to increase comparability. Tools can consist of many subscales or domains. Only the total score was included in the meta‐analysis.
Clinical diversity of the interventions
This is a pragmatic review and we have aimed to combine similar interventions where possible. Whilst we acknowledge the differences between interventions across studies, we are interested in investigating the generic type of intervention, for example CBT, control or psychotherapy. The composition of the standard care intervention (listed here as control) is dependent on the location and context of the study. In some studies the control group received educational leaflets, whereas in others the control included access to seminars. In one study the control group had access to relaxation classes (Kissane 2003).
Excluded studies
We reported the reasons for exclusion for some of excluded studies in the 'Characteristics of excluded studies' table. The most frequent reasons for exclusion were non‐randomised trials and inclusion of women with metastatic disease.
Risk of bias in included studies
We assessed the risk of bias for each included study and reported the judgements for the individual risk of bias domains in the ’Risk of bias’ table. We have also presented these in the ’Risk of bias’ summary in Figure 2. In addition, an overall risk of bias was determined for each study and nine were categorised as high risk of bias because one or more risk domains received a judgement of high risk (Classen 2008; Cohen 2007; Fukui 2000; Graves 2003; Kissane 2003; Loprinzi 2011; Narváez 2008; Richardson 1997; Simpson 2001). The remaining 19 studies were rated as unclear risk of bias because one or more criteria were assessed as unclear.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Some studies in this review did not provide sufficient details to enable accurate judgements to be made. We contacted 25 authors and were able to amend the judgements for a number of the risk domains after receiving a reply from eight trial investigators. However, if the authors mentioned in the e‐mail communication that blinding or concealment was used without clarifying the method that was used for randomisation or allocation concealment the judgement remained as unclear.
Allocation
Random sequence generation
The description of the method used in random allocation was stated clearly in 12 studies (Andersen 2004; Badger 2007; Baucom 2009; Boesen 2011; Classen 2008; Ferguson 2012; Fillion 2008; Fukui 2000; Kissane 2003; Loprinzi 2011; Mishel 2005; Yates 2005). In Cohen 2007 a systematic order was used to allocate women, which was judged as inadequate. In the remaining studies the information obtained about random allocation was not sufficient to permit a clear judgement.
Allocation concealment
Four studies were open labelled (Classen 2008; Cohen 2007; Narváez 2008; Kissane 2003) and were judged as high risk of bias. Inadequate reporting did not permit a clear judgement to be made for this domain in 19 studies, while sufficient information was obtained regarding allocation concealment in the remaining 5 trials (Andersen 2004; Badger 2007; Boesen 2011; Dolbeault 2009; Fillion 2008).
Blinding
Blinding of participants and personnel
Blinding of participants and personnel was not stated in 26 included studies. Two studies (Fukui 2000; Simpson 2001) clearly mentioned that it was not possible to blind participants or investigators to the intervention considering its nature. This was judged as unclear risk of bias because the effect of the lack of blinding on the outcome for this type of intervention is unclear.
Blinding of outcome assessors
Blinding of outcome assessors was achieved in five studies (Antoni 2001; Antoni 2006; Badger 2007; Cohen 2007; Ferguson 2012). In three studies it was not done (Classen 2008; Narváez 2008; Richardson 1997) and in the remaining studies the available information was insufficient to allow a clear judgement of this domain.
Incomplete outcome data
In seven studies incomplete outcome data appeared to have been adequately addressed (Andersen 2004; Antoni 2006; Henderson 2012; Kissane 2003; Manne 2007; Richardson 1997; Yates 2005). The incomplete data were reasonably well‐balanced across intervention groups, with intention‐to‐treat analyses reported. However, the high dropout rate and subsequent per protocol analysis of the data in two studies resulted in a judgement of high risk of bias for this domain (Graves 2003; Loprinzi 2011). In the remaining 19 studies, the number of dropouts was balanced between the groups. However, the percentage of dropouts (less than 28%) and subsequent per protocol analysis posed an unclear risk of bias.
Selective reporting
In Classen 2008 some outcomes were not reported and in Taylor 2003 only 12‐month outcomes were reported because the intermediate assessment showed similar results. In Yates 2005 secondary outcomes were not reported due to non‐significant results. In Henderson 2012 only major significant study outcomes were reported (with no data for distress and mental adjustment to cancer). The impact of selective reporting in these four studies was therefore unclear and this domain was judged as unclear risk of bias.
Other potential sources of bias
In 19 trials the information provided about this domain was sufficient and as a result this domain was judged as low risk of bias. In the remaining nine trials there was insufficient information to permit a clear judgement in this domain. Because of the non‐therapeutic nature of the intervention, none of the studies reported any conflict of interest with pharmaceutical companies.
Effects of interventions
Cognitive behavioural therapy (CBT) versus control
There were 24 studies that compared these intervention groups (Andersen 2004; Antoni 2001; Antoni 2006; Baucom 2009; Boesen 2011; Cohen 2007; Dolbeault 2009; Ferguson 2012; Fillion 2008; Fukui 2000; Garssen 2013; Graves 2003; Henderson 2012; Kissane 2003; Loprinzi 2011; Manne 2007; Marchioro 1996; Mishel 2005; Narváez 2008; Nunes 2007; Richardson 1997; Simpson 2001; Taylor 2003; Yates 2005).
Primary outcomes
Mean difference in depression
We identified nine studies that included this outcome. Seven studies carried out a group‐delivered intervention and measured the mean difference (MD) in depression using any validated disease specific tool at the end of the study (Ferguson 2012; Fukui 2000; Garssen 2013; Henderson 2012; Kissane 2003; Narváez 2008; Nunes 2007); and two carried out an individually‐delivered intervention (Marchioro 1996; Yates 2005).
From the nine studies with this outcome, seven studies were included in our analysis, five compared a group‐delivered intervention to placebo and two an individually‐delivered intervention compared to placebo. The pooled change from baseline standardised mean difference (SMD) comparing the group and individually‐delivered interventions to standard care was ‐1.01 (95% CI ‐1.83 to ‐0.18; P = 0.02; I2 = 95%; Chi2 test P < 0.00001) (Analysis 1.1; Figure 3) but the estimate was associated with a high level of uncertainty due to severe heterogeneity, which was introduced by Garssen 2013. After the removal of Garssen 2013 in a sensitivity analysis, the SMD was ‐0.43 (95% CI ‐0.90 to 0.04; P = 0.07; I2 = 82%; Chi2 test P < 0.0001) (Analysis 1.2).
Forest plot of comparison: 1 CBT versus control, outcome: 1.1 Standardised mean difference for the change from baseline in depression.
Whilst there was non‐significant statistical evidence showing a within group reduction in depression from baseline following the group‐delivered intervention compared to control, the majority of the studies exhibited modest positive effects favouring CBT. We rated this result as low quality evidence due to inconsistency in the magnitude of effect across the studies, unexplained heterogeneity introduced by one study (Garssen 2013), absence of allocation concealment in two studies and unclear risk of bias in the remaining studies with respect to the same domain.
Mean difference in anxiety
Eight studies reported the change in anxiety with the intervention delivered as either a group intervention (Dolbeault 2009; Fukui 2000; Garssen 2013; Kissane 2003; Loprinzi 2011; Narváez 2008; Nunes 2007) or an individually‐delivered intervention (Yates 2005). The eight studies included 393 patients in the intervention arm and 383 in the control arm. The change from baseline pooled SMD between the intervention and control on the anxiety scale was ‐0.48 (95% CI ‐0.76 to ‐0.21; P = 0.0006; I2 = 64%; Chi2 test P = 0.006) (Analysis 1.3; Figure 4). This suggested reduced anxiety in the CBT group compared to control but the substantial heterogeneity left doubt over the comparability of these studies.
Forest plot of comparison: 1 CBT versus control, outcome: 1.3 Standardised mean difference for the change from baseline in anxiety.
We rated this result as low quality evidence because allocation concealment was not done in two studies whereas it was judged as unclear risk of bias in the remaining included studies.
Mean difference in stress
Two studies reported the mean difference in stress (Antoni 2006; Nunes 2007). However, no estimate of the variability was reported and we were unable to estimate it in one study (Antoni 2006). No meta‐analyses were carried out. In Antoni 2006, 92 patients were allocated to the intervention arm and 107 to the control arm. The intrusion subscale of the Impact of event scale was used to measure stress and showed a significant reduction in distress over time with CBT relative to the control condition (both arms had a reduction in stress from baseline but the reduction with CBT was greater). This effect persisted at nine months post‐intervention. In Nunes 2007, 20 patients were allocated to the intervention arm and 14 to the control arm. The experimental group showed significantly reduced levels of stress from baseline measured by Lipp's Inventory of Stress Symptoms in adults (P < 0.05). The scoring was performed by means of three different stages related to the duration (Q1 = last 24 hours, Q2 = last week and Q3 = last month). Q1, for example, showed a within group reduction from baseline in the score (mean ± SD) for stress in the intervention group from 4.11 ± 2 to 2.79 ± 1.65.
Mean difference in mood disturbance
Ten studies reported the change in mood with CBT delivered as a group intervention (Andersen 2004; Boesen 2011; Classen 2008; Dolbeault 2009; Fillion 2008; Fukui 2000; Graves 2003; Henderson 2012; Kissane 2003) or an individually‐delivered intervention (Mishel 2005). Sufficient data were available from eight studies across all subgroups with 755 participants allocated to the intervention and 781 to the control arm. The SMD between intervention and control was ‐0.28 (95% CI ‐0.43 to ‐0.13; P = 0.0003; I2 = 47%; P = 0.07) (Analysis 1.4; Figure 5). This suggested consistent evidence that CBT, delivered either in a group or on an individual basis, resulted in within group reduction from baseline in mood disturbance compared to control. There was no evidence to suggest any difference in interventional effect across the three subgroups (P = 0.43; I2 = 0%).
Forest plot of comparison: 1 CBT versus control, outcome: 1.4 Standardised mean difference for the change from baseline in mood disturbance.
We rated the quality of evidence as moderate because three or more domains of risk of bias including allocation concealment were judged as unclear risk of bias in the included studies. In two studies (Classen 2008; Kissane 2003) allocation concealment was not done.
Secondary outcomes
Mean difference in quality of life (QoL)
Nine studies reported the change in QoL in a group‐delivered (Boesen 2011; Dolbeault 2009; Ferguson 2012; Fillion 2008; Garssen 2013; Graves 2003) or individually‐delivered intervention setting (Baucom 2009; Loprinzi 2011; Yates 2005). In these studies, 351 women were allocated to the intervention arm and 368 to the control arm. There appeared to be differences between the subgroups that were indicative of an important difference between the subgroups for this outcome. However, this view was not confirmed by the P value of 0.17. The SMD for the group‐delivered intervention was 0.21 (95% CI ‐0.03 to 0.46; P = 0.08; I2 = 47%; Chi2 test P = 0.09) (Analysis 1.5; Figure 6) and for the individually‐delivered intervention the SMD was 0.65 (95% CI 0.07 to 1.23; P = 0.03; I2 = 41%; Chi2 test P = 0.18).
Forest plot of comparison: 1 CBT versus control, outcome: 1.5 Standardised mean difference for the change from baseline in quality of life.
We rated the quality of evidence as low in the group‐delivered intervention because in three studies one or more of the risk of bias domains, including allocation concealment, was judged as unclear risk of bias and four different tools were used to measure QoL. The individually‐delivered intervention was rated as very low evidence because in the three included studies at least four domains of risk of bias including allocation concealment were judged as unclear risk of bias and each study used a different tool to measure the outcome.
Mean difference in coping
Six studies reported the change in coping when CBT was delivered either as a group intervention (Dolbeault 2009; Fukui 2000; Graves 2003; Henderson 2012; Simpson 2001) or as an individually‐delivered intervention (Marchioro 1996). No data could be extracted to carry out a pooled meta‐analysis, but two studies (Graves 2003; Marchioro 1996) with data are shown in Analysis 1.6. The total number of participants in these two studies was 33 in the intervention arm and 35 in the control arm. The intervention showed significant improvement in the treatment arm in Fukui 2000, Graves 2003, Simpson 2001 and Marchioro 1996 and no significant difference in Dolbeault 2009 and Henderson 2012. Mental Adjustment to Cancer was the most frequently used tool to measure coping. However, several versions and subscales were used. For example, all five scales were used in Fukui 2000 and Simpson 2001 whereas two scales were used in Dolbeault 2009. Other studies utilised other tools such as the Cancer Behaviour Inventory (Graves 2003), adaptation to cancer (Marchioro 1996) and dealing with illness (Henderson 2012).
Mean difference in adjustment
One study reported the change in coping adjustment delivered as a group intervention (Boesen 2011). In this study 102 patients were allocated to the intervention arm and 103 to the control arm. The authors used the Mental Adjustment to Cancer instrument which has five domains (fighting spirit, helplessness and hopelessness, anxious preoccupation, cognitive avoidance and fatalism) but none of the five domains suggested any evidence of a difference between the intervention and control groups.
Overall survival
Three studies reported on overall survival (Andersen 2004; Boesen 2011; Kissane 2004). One study did not describe their data because only nine deaths were experienced (Boesen 2011).
From the remaining two studies included in the analysis 268 patients were allocated to the intervention group and 262 were allocated to control. In Andersen 2004 54 of 227 (24%) women had died, 24 in the intervention arm and 30 in the assessment only arm. Breast cancer was the primary cause of death for 44 of the 54 patients (19 patients from the intervention arm and 25 from the assessment only arm). In Kissane 2004 the Kaplan‐Meier analysis revealed a median survival of 81.9 months (95% CI 64.8 to 99.0) in the intervention group compared with 85.5 months (95% CI 67.5 to 103.6) in the control arm. Information regarding the number of deaths in each arm was not reported.
The pooled hazard ratio (HR) from two studies (Andersen 2004; Kissane 2004) was 0.76 (95% CI 0.25 to 2.32; P = 0.63; I2 = 84%; Chi2 test P = 0.01) but exhibited substantial heterogeneity, so the generalisability of this result should be considered with caution (Analysis 1.7). We rated the quality of evidence as low because allocation concealment was absent in one of the two included studies and at least two other risk of bias domains were judged as unclear risk of bias in addition to the substantial heterogeneity.
Adverse events
No studies reported on adverse events.
Subgroup analysis
Dose of psychological intervention (≤ 20 hours versus > 20 hours) was tested for the outcomes and the overall treatment effect. Only anxiety and mood disturbance outcomes showed significant results favouring the shorter intervention (≤ 20 hours versus > 20 hours). However, one should be very cautious interpreting this result as the long intervention was evaluated in only a handful of included studies (Kissane 2003; Loprinzi 2011).
Psychotherapy versus control
There were four studies that compared psychotherapy and control (Badger 2007; Classen 2008; Marcus 2010; Vos 2004).
Primary outcomes
Mean difference in depression
Three of the four studies reported depression outcomes. Badger 2007 reported the mean difference in depression using the Center for Epidemiological Studies‐Depression Scale (CES‐D). The participants in the intervention arm (n = 38) showed a decrease in the depression score (mean ± SD) from 16.44 ± 1.74 to 8.82 ± 1.81 at 10 weeks whereas the participants in the control arm (n = 33) showed an increase in the depression score from 9.88 ± 1.79 to 14.25 ± 1.76. No meta‐analyses were carried out.
Marcus 2010 reported depression using the CES‐D. The study included 152 participants in the intervention arm and 152 in the control arm. For depression, the mean scores for both groups showed a significant reduction over time, with no difference by experimental group in the change from baseline. In contrast, when these scores were dichotomised a dramatically different pattern emerged. The control group showed no significant change from baseline to 18 months (P = 0.41) whereas the intervention group showed significant improvement (P = 0.0007) reflecting about a 50% reduction in the percentage scoring at or above the cutpoint suggestive of the need for a clinical referral. Group differences in change from baseline to 18 months approached statistical significance (P = 0.06) with an effect size of 0.23.
In Classen 2008, depression was one of the secondary outcomes, which was measured using the Hospital Anxiety and Depression Scale (HADS). The study included 177 women in the intervention arm and 175 in the control arm. It showed beneficial effects of treatment on the depression scale using both the model with imputed data and the reduced model F (1,253) = 5.4, P = 0.021 and F (1,212) = 6.3, P = 0.013, respectively.
Mean difference in anxiety
Two studies reported on anxiety. Badger 2007 reported the mean difference in anxiety. In this study, 38 women were allocated to the intervention arm and 33 to the control arm. Participants in the intervention arm showed a reduction in the anxiety score (mean ± SD) from 3.05 ± 0.34 to 2.81 ± 0.29 at 10 weeks of follow‐up whereas the control arm reported a decrease from 4.39 ± 0.33 to 3.19 ± 0.28. No meta‐analyses were carried out.
In Classen 2008, anxiety was one of the secondary outcomes that was measured using the HADS. The study included 177 women in the intervention arm and 175 in the control arm. It showed beneficial effects of treatment on the anxiety scale using both the model with imputed data and the reduced model F (1,253) = 4.5, P = 0.034 and F (1,212) = 3.9, P = 0.049, respectively.
Mean difference in stress and distress
Only one study reported on distress. Marcus 2010 reported distress using the Impact of event scale (intrusive subscale only). Mean scores for both groups showed a significant reduction over time, with no difference by experimental group in the change from baseline. In contrast, when these scores were dichotomised a dramatically different pattern emerged. The control group showed no significant change from baseline to 18 months (P = 0.41) whereas the intervention group showed significant improvement (P = 0.0007), reflecting about a 50% reduction in the percentage scoring at or above the cutpoint suggestive of the need for a clinical referral. Group differences in change from baseline to 18 months approached statistical significance (P = 0.07) with an effect size of 0.24.
Mean difference in mood disturbance
Classen 2008 reported on mood disturbance using the POMS scale. The study included 177 women in the intervention arm and 175 in the control. It showed a decrease in the POMS score in the intervention arm from a mean of 27.59 (SD 32.11) at baseline to 19.54 (SD 30.65) at 6 months and a further decrease to 13.69 (SD 30.67) at 24 months. On the other hand, the participants in the control arm showed a decrease in the POMS score from a mean of 21.67 (SD 29.07) at baseline to 16.36 (SD 32.18) at 6 months and a further decrease to 9.05 (SD 26.19) at 24 months. However, no significant effect was seen in POMS score between the two arms when the outlier observation was removed from the analysis.
In Vos 2004, 69 women with early stage breast cancer were randomised in a psychosocial group intervention program starting within 4 months after surgery (n = 34) or in the waiting list control (n = 35). It was not possible to extract the data as they were presented in a regression model.
Secondary outcomes
Mean difference in quality of life (QoL)
No studies reported this outcome.
Mean difference in coping
Vos 2004 reported on coping using a shortened 19 item version of the Utrecht Coping List in a study that included 69 women with early stage breast cancer randomised to a psychosocial group intervention program starting within 4 months after surgery (n = 34) or to the waiting list control (n = 35). It was not possible to extract the data as they were presented in a regression model.
Mean difference in adjustment
In Classen 2008, adjustment was one of the secondary outcomes that was measured using the Mini‐Mental Adjustment to Cancer scale. This study showed beneficial effects of treatment on the adjustment scale using both the model with imputed data and the reduced model F (1,253) = 5.2, P = 0.024 and F (1,212) = 5.3, P = 0.022, respectively.
Overall survival and adverse events
No studies reported these outcomes.
Discussion
Summary of main results
Psychological outcomes
Meta‐analysis showed significantly improved scores for anxiety, mood disturbance and depression favouring the psychological intervention (cognitive behavioural therapy). On the other hand, the intervention showed significant improvement on quality of life only in an individually‐delivered format. Group‐delivered interventions appeared more beneficial for depression and mood disturbance outcomes. Both forms of intervention (group and individual) were associated with significantly reduced anxiety compared to control. Although the shorter intervention (≤ 20 hours) showed favourable results for the anxiety and mood disturbance outcomes, one should be very cautious interpreting this finding due to the limited number of studies evaluating high doses of the intervention (> 20 hours). Further studies comparing various doses of the therapy are needed before making any conclusive judgement.
Survival
Psychological interventions for women with non‐metastatic breast cancer were not associated with survival. This finding is based on two studies involving 530 women (pooled HR 0.76, 95% CI 0.25 to 2.32; P = 0.63; I2 = 84%).
Overall completeness and applicability of evidence
This review considers randomised controlled trials (RCTs) of a wide range of psychological therapies for breast cancer (see Types of interventions). Studies included participants from different countries and backgrounds who had been diagnosed with, and treated for, breast cancer (see Types of participants). Many factors, however, limit the generalisability of the results. Most studies were conducted in the USA and Europe, limiting generalisability of results to the rest of the world. Although we have included the full range of psychological therapies for breast cancer, insufficient data preclude meta‐analysis of a number of outcomes in some treatment groups. There are a larger number of studies of group cognitive behavioural intervention than for other therapies. All studies but one (Andersen 2004) did not include participants with comorbid psychiatric diagnoses such as depression and anxiety therefore excluding individuals who are arguably more difficult to treat. This may have resulted in the exclusion of individuals with high levels of stress who are more or less likely to benefit from the treatment.
Quality of the evidence
The trials were heterogeneous in many of the clinical and methodological characteristics which could potentially affect the direction and magnitude of the effect. The main areas of heterogeneity are type of intervention, control condition, inclusion criteria, outcome measurement, timing of follow up and quality of studies.
Type of intervention
A wide range of psychological interventions was used and in most cases a combination of two or more approaches and techniques have been employed making it by far the most challenging factor when comparing various studies. Additionally, details on the content and integrity of the intervention were not always available.
Control condition
Studies used standard care control conditions and did not control for the potential effect of non‐specific structured activities taking place at home or at a hospital. Variation in standard care in the various trials should also be taken in to consideration when comparing trials conducted in different settings.
Inclusion criteria
Although previous reports showed that women with higher level of psychological disturbances benefited the most from psychological interventions compared to standard care, all but one trial excluded women with psychological morbidities. Even when women with higher levels of distress were included in the trial, they were rarely analysed separately.
Outcome measurement
There is a huge variation in the instruments used to assess various outcomes in the trials included in this review. Most tools are validated but a few are not. One should be cautious interpreting the results originating from different tools although measuring the same outcome.
Timing of follow up
There is a good body of literature showing that the longer time that has elapsed after diagnosis the better the psychological status of women is. Therefore, timing of measuring the outcome can make a difference in the magnitude and direction of the effect of the intervention. The included trials showed a considerable amount of heterogeneity in the timing of outcome assessments hence this should be taken into consideration when comparing trials with early and late assessments.
Quality of studies
One of the major methodological issues with this kind of intervention is that blinding of interventionists and patients is not possible due to the nature of the intervention. The absence of blinding on the magnitude and direction of the treatment effect is unclear and could potentially have led to performance bias. Although only RCTs were included in this review, four studies provided insufficient details on allocation concealment and nine trials reported insufficient details on sequence generation. Considering the nature of the intervention, which is primarily a preventive psychological intervention as opposed to a drug intervention, pharmaceutical company funding did not constitute a threat.
Potential biases in the review process
This review was conducted in line with the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Although the methodology for conducting systematic reviews is well established, subjective judgement is inevitable throughout the process. The main limitation of this review is the lack of sufficient data in many trials to make a clear judgement in various bias domains. As a result, a false estimate of the underlying truth is inevitable. Another limitation is the heterogeneity in the intervention delivered, outcome measurement, timing of follow up and participants’ characteristics. The paucity of data, especially for the survival outcome, resulted in ambiguous study specific effect estimates. Additionally, because most of the included trials were primarily conducted among Caucasians and in developed countries, generalisability of the results to different ethnic groups and countries might not be feasible. Finally, the cut‐off of 20 hours used in the subanalysis is almost certainly arbitrary with no 'scientific' rationale as such. This cut‐off was used for comparison purposes with a previously conducted systematic review in the same field (Naaman 2009).
One of the important limitations is that adverse outcomes to the intervention were not reported in the included trials. The iatrogenic effect of psychological intervention has scarcely been studied (Barlow 2010; Roback 2000). For example, one study showed that peer discussion groups had a negative effect on women with breast cancer (Helgeson 1999). One should note that in this review we included only professional‐led well described interventions and not peer group discussions. Nevertheless, the possibility of the intervention causing harm cannot be ruled out by the reported data.
Agreements and disagreements with other studies or reviews
We identified few meta‐analyses investigating psychological intervention in women with breast cancer (Duijts 2011; Mustafa 2013; Matsuda 2014; Naaman 2009). However, the inclusion criteria were not uniform across the reviews. They varied drastically in terms of the stage of the disease, type of intervention and the outcome measured. For example, Naaman et al included women at early and late stage of the disease (Naaman 2009), Duijts et al investigated behavioural techniques and physical exercise (Duijts 2011), Matsuda confined the search to English‐language trials (Matsuda 2014) and Mustafa et al included only women with metastatic breast cancer (Mustafa 2013). There is considerable debate over the effect of psychological intervention on survival. Most meta‐analyses of data from cancer patients found no survival benefit of such intervention (Chow 2004; Cwikel 1998; Newell 2002; Smedslund 2004; Zabalegui 2005). In line with the vast literature, our analysis showed no significant survival benefit for psychological intervention in women with non‐metastatic disease. In contrast, Mustafa et al reported a favourable effect of the intervention on survival at one year but this was not sustained at five years (Mustafa 2013). There is general agreement amongst researchers in the field that consistent long‐term reporting of the effects of psychological interventions on survival is necessary (Mustafa 2013; Newell 2002; Smedslund 2004; Williams 2006).
Consistent with previous reviews on cancer patients (Jacobsen 2008; Meyer 1995; Osborn 2006; Raingruber 2011; Reshe 2003) and reviews on women with breast cancer (Naaman 2009), our results showed favourable effects of psychological interventions on some psychological outcomes, particularly anxiety. However, no significant effect was found on quality of life (QoL), which is in agreement with some but not all reviews for patients with cancer (Newell 2002) and breast cancer (Matsuda 2014). Nevertheless, a small improvement was achieved in QoL when the disease specific measurement tool was used (Galway 2012). The benefits for psychological outcomes seem to be less evident when the disease has already metastasised to other parts of the body (Mustafa 2013). It is noteworthy that in some reviews depression was used interchangeably with mood disturbance. As a result, data originating from tools used to measure mood disturbance were pooled under depression as an outcome. This might explain the discrepancy in the effect size for depression. This also applies to other outcomes such as stress, distress, coping and adjustment. The variation in the terminology used to describe various psychological outcomes and the interchangeably make it difficult to classify and subsequently pool the trials.
Similar to a previous review on patients with breast cancer (Naaman 2009), our result shows that group therapy was generally superior to individual therapy for the treatment of anxiety and depression. However, this was not uniform across other reviews in which all cancer patients were included (Osborn 2006). This controversy might be related to the differences in patients included and the scarcity of trials investigating individually‐delivered interventions.
CBT carried the most beneficial effect, however one should be cautious interpreting this result because of the insufficient number of trials investigating other forms of psychological intervention and the inconsistency in defining what constitutes psychotherapy and other forms of psychological intervention.
Although in a previous review (Naaman 2009) a treatment duration time of greater than 20 hours compared to less than 20 hours yielded significant results, it was not possible to examine this subgrouping in this review because only one trial fulfilled the long duration criteria (Kissane 2003).
There are indications that psychological interventions targeting patients at higher level of stress have greater clinical benefit than when targeting women with low or normal levels of stress (Sheard 1999). It was not possible to investigate this relationship in this review due to the paucity of studies that included patients with such a profile. Furthermore, even when patients with a higher level of distress were included, their results were not reported separately.
Figure 1. Study flow diagram.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Forest plot of comparison: 1 CBT versus control, outcome: 1.1 Standardised mean difference for the change from baseline in depression.
Forest plot of comparison: 1 CBT versus control, outcome: 1.3 Standardised mean difference for the change from baseline in anxiety.
Forest plot of comparison: 1 CBT versus control, outcome: 1.4 Standardised mean difference for the change from baseline in mood disturbance.
Forest plot of comparison: 1 CBT versus control, outcome: 1.5 Standardised mean difference for the change from baseline in quality of life.
Comparison 1 CBT versus control, Outcome 1 Standardised mean difference in the change from baseline in depression.
Comparison 1 CBT versus control, Outcome 2 Standardised mean difference in the change from baseline in depression (excluding Grassen 2013).
Comparison 1 CBT versus control, Outcome 3 Standardised mean difference in the change from baseline mean change in anxiety.
Comparison 1 CBT versus control, Outcome 4 Standardised mean difference in the change from baseline mood disturbance.
Comparison 1 CBT versus control, Outcome 5 Standardised mean difference in quality of life.
Comparison 1 CBT versus control, Outcome 6 Standardised mean difference in the change from baseline coping.
Comparison 1 CBT versus control, Outcome 7 Overall survival (group delivered intervention).
Comparison 1 CBT versus control, Outcome 8 Standardised mean difference in the change from baseline in depression group delivered (excluding Grassen 2013).
| Cognitive behavioural therapy versus control for women with non‐metastatic breast cancer | ||||
| Patient or population: Women with non‐metastatic breast cancer | ||||
| Outcomes | Illustrative comparative risks* (95% CI) | No of Participants | Quality of the evidence | Comments |
| Corresponding risk | ||||
| Cognitive behavioural therapy versus control | ||||
| Depression | The mean depression in the intervention groups was | 637 | ⊕⊕⊝⊝ | SMD ‐1.01 (95% CI ‐1.83 to ‐0.18) |
| Anxiety | The mean change in anxiety in the intervention groups was | 776 | ⊕⊕⊝⊝ | SMD ‐0.48 (95% CI ‐0.76 to ‐0.21) |
| Mood disturbance | The mean mood disturbance in the intervention groups was | 1536 | ⊕⊕⊕⊝ | SMD ‐0.28 (95% CI ‐0.43 to ‐0.13) |
| Quality of life: Group‐delivered intervention | The mean quality of life: group‐delivered intervention in the intervention groups was | 578 | ⊕⊕⊝⊝ | SMD 0.21 (95% CI ‐0.03 to 0.46) |
| Quality of life: Individually‐delivered intervention | The mean quality of life: individually‐delivered intervention in the intervention groups was | 141 | ⊕⊝⊝⊝ | SMD 0.65 (95% CI 0.07 to 1.23) |
| *The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||
| GRADE Working Group grades of evidence | ||||
| 1 A higher score on either tool indicated higher depression | ||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Standardised mean difference in the change from baseline in depression Show forest plot | 7 | 637 | Std. Mean Difference (IV, Random, 95% CI) | ‐1.01 [‐1.83, ‐0.18] |
| 1.1 Group delivered intervention (less than 20 hrs) | 3 | 120 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.33 [‐0.80, 0.14] |
| 1.2 Group delivered intervention (more than 20 hrs) | 1 | 303 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.10 [‐0.32, 0.13] |
| 1.3 Group delivered intervention (less than 20 hrs), study removed due to heterogeneity | 1 | 70 | Std. Mean Difference (IV, Random, 95% CI) | ‐4.43 [‐5.32, ‐3.54] |
| 1.4 Individually delivered intervention(less than 20 hrs) | 2 | 144 | Std. Mean Difference (IV, Random, 95% CI) | 1.00 [‐3.19, 1.20] |
| 2 Standardised mean difference in the change from baseline in depression (excluding Grassen 2013) Show forest plot | 6 | 567 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.43 [‐0.90, 0.04] |
| 2.1 Group delivered intervention (less than 20 hrs) | 3 | 120 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.33 [‐0.80, 0.14] |
| 2.2 Group delivered intervention (more than 20 hrs) | 1 | 303 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.10 [‐0.32, 0.13] |
| 2.3 Individually delivered intervention(less than 20 hrs) | 2 | 144 | Std. Mean Difference (IV, Random, 95% CI) | 1.00 [‐3.19, 1.20] |
| 3 Standardised mean difference in the change from baseline mean change in anxiety Show forest plot | 8 | 776 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.48 [‐0.76, ‐0.21] |
| 3.1 Group delivered intervention (less than 20 hrs) | 5 | 358 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.44 [‐0.85, ‐0.03] |
| 3.2 Group delivered intervention (more than 20 hrs) | 2 | 323 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.53 [‐1.32, 0.26] |
| 3.3 Individually delivered intervention (less than 20 hrs) | 1 | 95 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.78 [‐1.20, ‐0.36] |
| 4 Standardised mean difference in the change from baseline mood disturbance Show forest plot | 8 | 1536 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.28 [‐0.43, ‐0.13] |
| 4.1 Group delivered intervention (less than 20 hrs) | 6 | 768 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.34 [‐0.57, ‐0.11] |
| 4.2 Group delivered intervention (more than 20 hrs) | 1 | 259 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.22 [‐0.47, 0.02] |
| 4.3 Individually delivered intervention (less than 20 hrs) | 1 | 509 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.15 [‐0.32, 0.03] |
| 5 Standardised mean difference in quality of life Show forest plot | 9 | Std. Mean Difference (IV, Random, 95% CI) | Subtotals only | |
| 5.1 Group delivered intervention | 6 | 578 | Std. Mean Difference (IV, Random, 95% CI) | 0.21 [‐0.03, 0.46] |
| 5.2 Individually delivered intervention | 3 | 141 | Std. Mean Difference (IV, Random, 95% CI) | 0.65 [0.07, 1.23] |
| 6 Standardised mean difference in the change from baseline coping Show forest plot | 2 | Std. Mean Difference (IV, Random, 95% CI) | Subtotals only | |
| 6.1 Group delivered intervention | 1 | 32 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.09 [‐0.78, 0.61] |
| 6.2 Individually delivered intervention | 1 | 36 | Std. Mean Difference (IV, Random, 95% CI) | ‐1.28 [‐2.01, ‐0.56] |
| 7 Overall survival (group delivered intervention) Show forest plot | 2 | 530 | Hazard Ratio (Random, 95% CI) | 0.76 [0.25, 2.32] |
| 8 Standardised mean difference in the change from baseline in depression group delivered (excluding Grassen 2013) Show forest plot | 4 | 140 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.30 [‐0.67, 0.07] |
| 8.1 Group delivered intervention (less than 20 hrs) | 3 | 120 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.33 [‐0.80, 0.14] |
| 8.2 Group delivered intervention (more than 20 hrs) | 1 | 20 | Std. Mean Difference (IV, Random, 95% CI) | ‐0.09 [‐0.97, 0.79] |
