Online support groups for women with breast cancer
Abstract
Background
Survival rates for women with a diagnosis of breast cancer continue to improve. However, some women may experience physical, psychological and emotional effects post diagnosis, throughout treatment and beyond. Support groups can provide opportunities for people to share their experiences and learn from others. As the number of online support groups increases, more and more women with breast cancer will likely access them.
Objectives
To assess effects of online support groups on the emotional distress, uncertainty, anxiety, depression and quality of life (QoL) of women with breast cancer.
Search methods
We searched for trials in the Cochrane Breast Cancer Specialised Register, the Cochrane Central Register of Controlled Trials (CENTRAL; 2016, Issue 4), MEDLINE, Embase and PsycINFO on 2 May 2016, and we handsearched journals and reference lists. We also searched the World Health Organization's International Clinical Trials Registry Platform (WHO ICTRP) search portal and clinicaltrials.gov on 2 May 2016.
Selection criteria
We included randomised controlled trials (RCTs) assessing effects of online support groups on women with a diagnosis of breast cancer and women who have completed breast cancer treatment. We included studies comparing online support groups with a usual care group, and studies comparing two or more types of online support groups (without a usual care group).
Data collection and analysis
Two review authors independently extracted data and assessed risk of bias. We presented outcome data using mean differences (MDs) and standardised mean differences (SMDs) along with 95% confidence intervals (CIs), and we used the fixed‐effect model when appropriate. We assessed the quality of the body of evidence using the GRADE approach.
Main results
We included six studies (492 women) that assessed online support groups for women with breast cancer. Online support groups in these six trials lasted from six to 30 weeks. Women participated in these groups between 1.5 and 2.5 hours per week, and investigators conducted all studies in the USA. Participants were predominantly white and well educated and were moderate to high earners. Four studies compared an online support group versus a control group, and the other two compared a 'moderated' versus a 'peer‐led' online support group, and a 'standard' versus an 'enhanced' online support group, respectively.
None of the included studies measured 'emotional distress' or uncertainty. One study (78 women) for which data for analysis were missing reported no positive effects of online support on 'distress' and 'cancer‐specific distress' versus support provided by a control group. Two studies measured anxiety: One study (72 women) found no difference in anxiety at the end of the intervention between the online support group and the control group (MD ‐0.40, 95% CI ‐6.42 to 5.62; low‐quality evidence), and the second study (184 women) reported a reduction in anxiety levels at the end of the intervention when comparing the 'standard' support group (run by participants without prompting from health professionals) versus an 'enhanced' online support group (in which participants were specifically asked by the researcher to respond to one another's need for support).
Five studies (414 women) measured depression. Three studies compared depression in the online support group with depression in the control group. Pooled data from two studies (120 women) showed a small to moderate reduction in depression in the online support group compared with control groups at the end of the intervention (SMD ‐0.37, 95% CI ‐0.75 to 0.00; very low‐quality evidence). The third study, a pilot study (30 women), provided no data for analysis but reported no difference in depression between participants in support and control groups at the end of the intervention. Of the remaining two studies that measured depression, one study (60 women) provided no extractable data for comparison but reported no difference in depressive symptoms between a 'moderated' and a 'peer‐led' support group; the other study (184 women) reported greater reduction in depression in the 'standard' support group than in the 'enhanced' online support group.
Three studies measured quality of life. One pilot study (30 women) provided limited data for analysis but reported no change in quality of life at the end of the intervention. Only two studies (140 women) provided data for pooling and showed no positive effects on quality of life at four months post intervention compared with controls (SMD ‐0.11, 95% CI ‐0.47 to 0.24; very low‐quality evidence). At 12 months post intervention, one study (78 women) reported that the intervention group did not attain better quality of life scores than the control group (MD ‐10.89, 95% CI ‐20.41 to ‐1.37; low‐quality evidence).
We found no data for subgroup analyses on stage of disease, treatment modality and types and doses of interventions. No studies measured adverse effects.
Authors' conclusions
This review did not find the evidence required to show whether participation in online support groups was beneficial for women with breast cancer, because identified trials were small and of low or very low quality. Large, rigorous trials with ethnically and economically diverse participants are needed to provide robust evidence regarding the psychosocial outcomes selected for this review.
PICOs
Plain language summary
Online support groups for women with breast cancer
Review question
We reviewed the evidence for effects of online support groups for women with breast cancer on emotional distress, uncertainty, anxiety, depression and quality of life.
Background
Women with a diagnosis of breast cancer can be affected physically, psychologically and emotionally. They are uncertain about the future and may need information and support to help them cope with their condition. Increasingly, people with cancer are accessing the Internet to seek the information and support that they need; many join online support groups. At this time, we know little about how participation in online support groups psychologically and emotionally affects women with breast cancer.
Study characteristics
We conducted a systematic search of the literature with no restrictions on language or country. We included in this review six studies, with a total population of 492 women with breast cancer. Five of the six studies had small samples. Study participants were predominantly 'white', well‐educated women with moderate to high income at any stage of breast cancer who were undergoing a range of treatments.
Online support groups in these six trials lasted six to 30 weeks and included eight to 15 members. Women participated in these groups between 1.5 and 2.5 hours per week. Investigators reported all trials in English and conducted their research in the USA.
Key results
None of the included trials measured emotional distress or uncertainty. Women who participated in online support groups showed no improvement in anxiety or quality of life when compared with those in control groups (which included women with similar characteristics who did not participate in online support groups). However, women who took part in online support groups showed a small to moderate reduction in depression when compared with those in control groups.
Results revealed no difference in depression between groups led by peers and those led by health professionals. However, women taking part in standard online groups (run by participants without prompting from health professionals) reported a greater reduction in depression and anxiety than those in other types of online groups (in which women were asked specifically by the health professional to respond to one another's need for support).
Quality of the evidence
Small studies of low or very low quality attributed mainly to poor study design and other shortcomings have provided evidence on the effectiveness of online support groups for women with breast cancer. Large, rigorous trials including ethnically and economically diverse participants are needed to provide robust evidence on the effectiveness of online support groups for women with breast cancer.
Authors' conclusions
Summary of findings
| Online support group for women with breast cancer | ||||
| Patient or population: women with breast cancer | ||||
| Outcomes | Relative effects* (95% CI) | Number of participants | Quality of the evidence | Comments |
| Online support group vs usual care | ||||
| Emotional distress | Not reported | ‐ | Emotional distress was not measured. However, 1 study reported that the intervention did not have a positive effect on "distress" and "cancer‐specific distress" | |
| Uncertainty | Not reported | ‐ | ‐ | None of the 6 included studies measured this outcome |
| Anxiety at end of intervention at 1 time point (end of a 12‐week intervention) | Mean anxiety in the intervention group was 0.4 lower at end of intervention (95% CI 6.42 lower to 5.62 higher) | 58 | ⊕⊕⊝⊝ | |
| Depression at end of intervention | Mean depression in the intervention group was 0.37 standard deviations undefined lower (95% CI 0.75 lower to 0) | 120 | ⊕⊝⊝⊝ | |
| Quality of life post intervention | Mean quality of life (at end of intervention) in the intervention group was 0.11 standard deviations undefined lower (95% CI 0.47 lower to 0.24 higher) | 140 | ⊕⊝⊝⊝ | |
| Quality of life 12 months post intervention | Mean quality of life in the intervention group (at 12 months post intervention) was 10.89 undefined lower (95% CI 20.41 lower to 1.37 lower) | 78 | ⊕⊕⊝⊝ | |
| *Risk in the intervention group (and its 95% confidence interval) is based on assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI) | ||||
| GRADE Working Group grades of evidence | ||||
| aLack of information on randomisation, concealment and blinding of participants and personnel | ||||
Background
Description of the condition
Breast cancer is the most common cancer among women, with an estimated 1.67 million new cancer cases diagnosed in 2012 worldwide (Ferlay 2012). Breast cancer incidence rates vary according to country, age, gender and socioeconomic group. In the USA alone, it was estimated that there will be more than 255,180 new cases of breast cancer among women in 2017 (American Cancer Society 2017). With early detection and advances in medical diagnostics and treatment, the number of people surviving cancer, in general, has increased significantly since the mid‐1990s. Five‐, 10‐ and 15‐year relative survival rates for women with a diagnosis of localised breast cancer in the USA were 89%, 83% and 78% in 2015, and death rates among women, especially younger women, have decreased steadily since 1989 (American Cancer Society 2017). Worldwide, there were 6.3 million women alive who had been diagnosed with breast cancer in the previous five years (International Agency for Research on Cancer 2013). The UK reported 1.6 million breast cancer survivors in 2010, and it is expected that this figure will grow at a rate of over 3% each year (Department of Health 2010).
Although survival rates have improved, women with breast cancer may experience many physical and psychosocial problems following diagnosis and during and after treatment, including fatigue, anxiety, depression, effects on body image, loss of employment, adverse effects of treatment and breakdown in relationships (Department of Health 2010). These events can impact a woman's quality of life (QoL). One review of 477 studies on the QoL of people with breast cancer reported that psychological factors predicted subsequent QoL or even overall survival (Montazeri 2008). In particular, breast cancer survivors 50 years of age or younger experience greater reduction in QoL than older survivors, along with distinct psychosocial and menopause‐related concerns, weight gain and physical inactivity (Howard‐Anderson 2012).
Description of the intervention
Support groups are a popular resource for people looking for information and support from peers to help them cope with their condition. In 2005, it was estimated that more than 400,000 Internet cancer support groups were available (Im 2005). The Internet will increasingly be the arena of choice for patients seeking psychosocial help in the future, not least because it can be accessed with relative ease by millions of people across wide geographical areas, from the comfort of their own homes. No consensus has been reached on the definition of support groups (Cancer Council Australia 2005), but they can be described as a form of peer support that consists of group members meeting face‐to‐face or communicating by telephone or via the Internet (including email and Facebook) for the purpose of sharing information and experiences and providing support on an issue or on topics of mutual interest.
Support groups vary in terms of membership, structure, leadership, delivery and setting. Some support groups are open to people with all types of cancer, and others are specific to one type of cancer. Some support groups are set up and led by health professionals as a psychosocial intervention (often based on cognitive‐behavioural theories) with clear outcomes (e.g. see Lepore 2011). This type of support group has a defined therapeutic intent. In contrast, support groups can be loosely structured, informal and flexible enough for group members to take part when they feel the need to do so and to discuss any relevant issue they want. These groups may be facilitated or 'serviced' by the people who set them up. Their main function is to provide a forum for group members to share information and experiences. Such groups have a 'supportive' intent. In reality, both types of groups may have therapeutic and supportive elements, and many groups may fall between these two types.
How the intervention might work
Support groups are based on the principle of self‐management, by which individuals take responsibility and become proactive in seeking ways to address their problems. For support groups, the underlying belief is that collectively, the group has a pool of knowledge and experiences that can benefit individuals who become members of these groups.
A cancer diagnosis and subsequent treatment can be perceived as overwhelming, especially if the person with cancer does not have the resources to cope. Support groups provide opportunities for people with cancer to compare experiences and learn about different ways that other people experience and cope with cancer. Social comparison is premised on the concept that humans have a need to look externally for images and information as they evaluate their own opinions and abilities (Festinger 1954).
Investigators have used the social cognitive theory (Bandura 1997) and the transactional theory of stress and coping (Lazarus 1984) as frameworks to explain the process of coping with a stressful event such as breast cancer. Information plays a crucial role in social cognitive theory, in that it enables individuals to cognitively frame and reframe their perceptions of the challenges they face during their cancer journey. The transactional theory of stress and coping posits that when people are faced with a threat, they appraise it in terms of how challenging or controllable it is. They also appraise their own and external resources at their disposal to help them face the threat; how they eventually cope depends on these appraisals (Parahoo 2013). The ability to appraise and cope with problems and difficulties engenders a sense of control and empowerment (van Uden‐Kraan 2008). Support groups can also reduce social isolation and loneliness (van Uden‐Kraan 2008).
Why it is important to do this review
There is little evidence about 'what works for whom' to inform the development of support groups for people with cancer, and for those in specific subgroups of cancer who may have different needs (Cancer Council Australia 2005). With increasing availability and access and improved skill in using online resources, it is likely that online support groups will increasingly play a key role in providing support for women with cancer. To date, no systematic review has examined the effectiveness of online support groups for women with breast cancer. Review authors must assess the state of research on this topic to make recommendations for future policy, practice and research.
Objectives
To assess effects of online support groups on the emotional distress, uncertainty, anxiety, depression and QoL of women with breast cancer.
Methods
Criteria for considering studies for this review
Types of studies
All randomised controlled trials (RCTs) assessing effects of online support groups on women with a diagnosis of breast cancer and those who have completed treatment for breast cancer. We included studies comparing online support groups with a usual care group, and studies comparing two or more types of online support groups (without a usual care group). We applied no language restrictions.
Types of participants
Included studies enrolled women with a diagnosis of breast cancer (any stage), disease free or not. Studies with mixed cancer populations and studies including partners were eligible for inclusion if they provided separate data for women with breast cancer. We included all types of treatment and applied no restrictions regarding age, ethnicity or setting.
Types of interventions
All types of support groups involving more than two participants, offered via the Internet in the form of messaging (on a dedicated website or through email) or chat rooms, were eligible. We included both professional and user‐led groups and combinations of these types of support. We excluded studies that evaluated a combination of face‐to‐face, telephone and online communication. We imposed no restrictions related to dose, frequency, intensity or duration of the intervention.
We compared online support groups against an inactive control intervention group (standard care or waitlist control) or against an active control intervention group (e.g. another form of psychological intervention).
Types of outcome measures
Primary outcomes
-
Emotional distress: assessed by validated instruments such as the Brief Symptom Inventory‐18 (BSI‐18) or the Profile of Mood States (POMS)
-
Uncertainty: assessed by a validated instrument such as the 28‐item Mishel Uncertainty in Illness Scale
-
Anxiety: assessed by validated instruments such as the State‐Trait Anxiety Inventory (STAI), the Beck Anxiety Inventory (BAI) and the Hospital Anxiety and Depression Scale‐Anxiety (HADS‐A)
-
Depression: assessed by validated instruments such as the Beck Depression Inventory (BDI, BDI‐II) and the Center for Epidemiologic Studies Depression Scale‐Revised (CESD‐R)
Secondary outcomes
-
QoL: assessed by validated instruments such as Medical Outcomes Study 36‐item Short Form (MOS SF‐36), the European Organisation for Research and Treatment of Cancer (EORTC) Quality of Life Questionnaire‐Core 36 (QLQ‐C30) and the Breast Cancer‐Specific Quality of Life Questionnaire (QLQ‐BR23)
Search methods for identification of studies
Electronic searches
We searched the following databases on 2 May 2016.
-
Cochrane Breast Cancer Specialised Register. Details of search strategies used by the Cochrane Breast Cancer Group (CBCG) for identification of studies and procedures used to code references are outlined in the CBCG module (onlinelibrary.wiley.com/o/cochrane/clabout/articles/BREASTCA/frame.html). We extracted trials with the key words "breast cancer", "support group", "online", "on‐line", "internet", "web‐based", "email", "chat room", "bulletin board", "computer" and "social network" and considered them for inclusion in the review.
-
Cochrane Central Register of Controlled Trials (CENTRAL; 2016, Issue 4). See Appendix 1.
-
MEDLINE (via OvidSP; from 2008 to 2 May 2016). See Appendix 2.
-
Embase via Embase.com (from 2008 to 2 May 2016) and Embase via OvidSP (from 2015 to present). See Appendix 3.
-
PsycINFO (via OvidSP; 2 May 2016). See Appendix 4.
-
World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) search portal for all prospectively registered and ongoing trials (apps.who.int/trialsearch/Default.aspx). See Appendix 5.
-
ClinicalTrials.gov (clinicaltrials.gov/). See Appendix 6
Searching other resources
Bibliographic searching
We searched the reference lists of identified relevant trials and reviews. We obtained a copy of the full‐text article for each reference reporting a potentially eligible study, but we found no new studies.
We searched the Internet for reports and other literature related to the objectives of this review, but we identified no additional studies, other than those obtained from the databases listed above.
Data collection and analysis
Selection of studies
In the first stage of selection, we assigned all studies (after removal of duplicates) an identification number. Two review authors (EM and KP) independently read all abstracts to decide whether we should include, exclude or wait for full versions of the papers. We contacted a third review author (LN) when we encountered discordance in the first pair’s decision, or when we needed further advice. At the second stage, we obtained full versions of all selected papers, and two review authors (EM and KP) independently read these papers. The third review author (LN) provided a third opinion when needed. We found no abstract or paper that required translation into English.
Selection criteria included randomised controlled trial, online support group and women with breast cancer (any stage). Additionally, we included studies that measured any of the primary outcomes (emotional distress, uncertainly, anxiety and depression) and the secondary outcome (quality of life).
We noted excluded studies in the Characteristics of excluded studies table, along with reasons for exclusion.
Data extraction and management
Two review authors (EM and KP) independently extracted data using a data extraction form developed for the purposes of this review. This form comprises items such as aims/objectives or hypotheses, study design (including randomisation method), sample (including age, ethnicity, setting and stage of cancer), follow‐up, type and nature of online support group, dose, frequency, intensity or duration, outcomes (e.g. QoL, uncertainty), outcome measures, statistical tests and findings. A third review author (LN) was available to resolve differences when necessary.
We extracted data from all publications pertaining to the same study. We considered the main paper presenting the study design and most outcomes as the primary reference. Figure 1 shows the flow chart detailing the selection of studies.
Study flow diagram.
Assessment of risk of bias in included studies
Two review authors (EM and KP) independently used the Cochrane 'Risk of bias' tool to assess risk of bias in the selected studies (Higgins 2011). A third review author (LN) resolved disagreements in ratings between these two review authors. We rated each of the following seven domains as having low, high or unclear risk of bias.
-
Sequence generation.
-
Allocation concealment.
-
Blinding of participants and personnel.
-
Blinding of outcome assessors.
-
Incomplete outcome data.
-
Selective outcome reporting.
-
Other sources of bias.
We contacted five of the six study authors (Changrani 2008; Klemm 2012; Salzer 2010; Vilhauer 2010; Winzelberg 2003) to request missing information and received a reply from only one of them (Vilhauer 2010). We used outcomes of the risk of bias assessment to assess the quality of evidence for each outcome according to GRADE recommendations (Guyatt 2011).
Measures of treatment effect
We considered all outcomes in the review (emotional distress, uncertainty, anxiety, depression and QoL) to be continuous outcomes. We used the mean difference (MD) or the standardised mean difference (SMD) together with 95% confidence intervals (CI), as appropriate. We obtained standard deviations (SDs) at baseline and at end of treatment from standard errors, CIs, t values or P values related to differences between means in the two groups (Higgins 2011). We compared means in intervention and control groups at follow‐up for each group, using available data.
Unit of analysis issues
We identified no cluster‐randomised or cross‐over trials. Three of the six included studies had more than one follow‐up time point, but we could extract limited data for analysis from published papers or from study authors.
Dealing with missing data
We contacted original study authors to request missing data and information on how they handled missing data when they conducted their data analyses, but we received only one response (as mentioned in Assessment of risk of bias in included studies section); no new information was available. When meta‐analysis was not possible owing to missing data, we provided a narrative commentary.
Assessment of heterogeneity
We noted the Chi2 test (Cochran 1954) and the I2 statistic (Higgins 2003) for two of the outcomes for which we found available data for pooling. We did not conduct a visual inspection of forest plots to assess statistical heterogeneity, as only two studies reported each of these two outcomes and they were similar in size. We narratively discussed different types of heterogeneity.
Assessment of reporting biases
We identified too few studies to carry out funnel plots and the Egger test to assess reporting bias (Egger 1997).
Data synthesis
We performed analysis using Review Manager 5 software (RevMan 2014) and used the inverse variance method to pool continuous measures (MDs) with a fixed‐effect model (DerSimonian 1986).
For analysis, we grouped studies as intervention versus usual care (standard care). When the same study provided one or more interventions (and no usual care group), we reported study results narratively.
We used the GRADE approach to rate the quality of evidence; we used GRADEPro software to generate summary of findings Table for the main comparison to report primary and secondary outcomes.
Subgroup analysis and investigation of heterogeneity
Data were insufficient for review authors to carry out separate analyses for 'user‐led' support groups and 'professional‐led' support groups. For the same reason, we did not conduct subgroup analyses regarding sources of heterogeneity, such as stage of disease and types and doses of interventions.
Sensitivity analysis
Studies and data were insufficient for review authors to perform a sensitivity analysis to assess robustness of results (e.g. excluding studies with high risk of bias).
Results
Description of studies
Results of the search
We conducted the search on 2 May 2016, and found 2453 records from the following databases: CBCG Specialised Register (194), CENTRAL (744), Embase (531), MEDLINE (273), PsycINFO (137) and the trial registries WHO ICTRP (32) and ClinicalTrials.gov (542). We found eight additional records from the reference lists of retrieved papers.
After removing duplicates, we screened the titles and abstracts of 1478 records. Fifty‐two papers were potentially eligible, and the first two review authors read the full texts of these. We excluded 46 studies (see Excluded studies) and selected six studies (Changrani 2008; Klemm 2012; Lepore 2014; Salzer 2010; Vilhauer 2010; Winzelberg 2003) for inclusion in this review. Figure 1 outlines the selection process.
Included studies
Participants
Investigators in the six included studies randomly allocated 492 women with breast cancer (mean 82; standard deviation (SD) 57.2; median 70; range 30 to 184) to online support or control groups. At final data collection time points, 416 participants (mean 69.3; SD 47; median 60; range 22 to 160) remained in the study, yielding an average attrition rate of 17.8% (range 13% to 27%). Slightly more participants dropped out of the intervention groups than out of the control groups, except in Vilhauer 2010, which reported that 35% dropped out of the waitlist group compared with 19% of the online support group.
Researchers allocated 277 participants to online support groups (mean 46; SD 27.5; median 42; range 16 to 96) compared with 215 to control groups (mean 35.8; SD 26.7; median 28.5; range 14 to 88). Two of the six studies (Changrani 2008; Salzer 2010) had approximately twice as many participants in the intervention group as in the control group.
Study authors carried out all six studies in the USA and recruited participants from different states. Investigators described three studies as 'pilot' (Salzer 2010), 'feasibility' (Changrani 2008) or 'feasibility pilot' (Vilhauer 2010).
Three studies (Klemm 2012; Lepore 2014; Salzer 2010) described more than 90% of participants as 'white'. Vilhauer 2010 had 100% 'white' participants, and Changrani 2008 described all participants as 'Hispanic'. Klemm 2012 reported 10% and Winzelberg 2003 4% 'Afro‐Americans'. Only one study (Winzelberg 2003) included 'Asians' (6%).
Three studies provided no information on participant income, and the other three studies reported participant annual earnings of between 40,000 and 50,000 $USD. The mean reported annual income in one study was over 83,000 $USD. Most participants in four of the six studies completed education at high school level or higher. Winzelberg 2003 described participants as 'highly educated'. The remaining two studies (Changrani 2008; Klemm 2012) provided no information on educational attainment.
Salzer 2010 did not provide details on the age profile of participants but mentioned that almost 40% were younger than 50 years of age. The other five studies reported participant mean age of 50.7 years (standard deviation (SD) 2.56 years).
Two studies (Changrani 2008; Klemm 2012) recruited women at all stages of breast cancer, and two other studies (Lepore 2014; Salzer 2010) recruited participants at stages I and II of breast cancer. Vilhauer 2010 reported that all participants had metastatic breast cancer, and Winzelberg 2003 only described participants as having had a primary breast carcinoma diagnosis. Lepore 2014 recruited only participants with distress or depression levels ≥ 8 on the Hospital Anxiety and Depression Scale (HADS). Winzelberg 2003 provided no information on stage of treatment of participants. The other five studies reported that all participants were receiving one or more forms of treatment (surgery, chemotherapy, hormonal therapy or radiotherapy).
Intervention
Of the six included studies, four (Changrani 2008; Salzer 2010; Vilhauer 2010; Winzelberg 2003) included one intervention group and one control/waitlist group. Klemm 2012 provided only two interventions (a moderated online support group and a peer‐led online support group). Lepore 2014 provided only two interventions (a standard Internet support group and an enhanced prosocial Internet support group). In all, the six included studies included eight intervention groups and four control/waitlist groups.
The aim of the four studies (Changrani 2008;Salzer 2010; Vilhauer 2010;Winzelberg 2003) that included one intervention and one control group was to find out whether Internet/online support groups improved psychosocial outcomes for women. These investigators also measured outcomes such as participation, satisfaction, personal growth and social support. The aim of the remaining two studies was to compare two different formats of Internet/online support groups in terms of the outcomes mentioned above. The aim of Klemm 2012 was to compare a moderated group versus a peer‐led group. Lepore 2014 sought to compare a standard Internet support group (S‐ISG) versus an enhanced prosocial Internet support group (P‐ISG) and hypothesised that the P‐ISG, which encouraged and facilitated participants to help others, would have a more positive impact than the S‐ISG in reducing depression and anxiety.
Only three of the eight interventions in this review were peer led (Klemm 2012; Salzer 2010; Vilhauer 2010).
Facilitators in the included studies were trained bilingual (English and Spanish) professionals (Changrani 2008); Master's degree prepared social workers experienced in providing online and telephone help for people with cancer and their carers (Klemm 2012); graduate level health professionals with more than 10 years' experience running Internet support groups (Lepore 2014) and mental health professionals (Winzelberg 2003). The main role of moderators or facilitators was to provide structure to support groups while encouraging participants to talk about and share their views on issues (preselected or not) of concern to participants. Among studies that provided moderated online support groups, Changrani 2008 had 'no set agenda', Klemm 2012 had 'preselected topics', Lepore 2014 had 'chat topics' and Winzelberg 2003 included conversations around 'weekly topics'.
Frequency and duration
Five studies (Changrani 2008; Klemm 2012; Lepore 2014; Vilhauer 2010; Winzelberg 2003) provided detailed information about frequency and duration of the interventions, which ranged from six to 30 weeks. Salzer 2010 did not describe frequency and duration of the intervention. Two studies (Changrani 2008; Lepore 2014) reported that participants logged in for 1.5 hours per week, and two others (Klemm 2012; Vilhauer 2010) reported that participants logged in for 2.5 hours per week. Winzelberg 2003 did not describe the exact duration of the intervention but reported that participants logged in on average 34 times (SD 29; range 3 to 22) and posted an average of 36 support group messages (SD 38; range 1 to 146) over the 12‐week duration of the intervention.
The size of support groups in each trial was as follows: Changrani 2008 ‐ eight to 10; Klemm 2012 ‐ up to 15; Lepore 2014 ‐ 14 to 17; Vilhauer 2010 ‐ 10 to 11; and Winzelberg 2003 ‐ 10 to 15. Salzer 2010 provided no information on group size.
Control group
Included studies provided sparse information on control groups. Changrani 2008 provided 'usual care' to participants; Salzer 2010 gave participants information on a cancer‐related website, and Vilhauer 2010 "sent a bi‐weekly breast cancer newsletter by email" to participants in both intervention and control groups. In Winzelberg 2003, participants in the waitlist group "were invited to participate in their own support group". Two studies (Klemm 2012; Lepore 2014) included no control group.
Only one study (Vilhauer 2010) set "not being users of online groups" as an inclusion criterion. The same study reported that one participant was a regular user of an online support group and another "occasionally accessed a large online bulletin board for all kinds of cancer patients". Among participants in Winzelberg 2003, 34% were already participating in another breast cancer support group or were receiving individual counselling at baseline. The other four studies (Changrani 2008; Klemm 2012; Lepore 2014; Salzer 2010) did not mention participation in other support groups in their selection criteria and did not report this information.
Four studies (Klemm 2012; Lepore 2014; Salzer 2010; Vilhauer 2010) specified access to computers and the Internet as an inclusion criterion. Changrani 2008 provided access to computers and the Internet as well as technical support for participants who did not have these facilities. Winzelberg 2003 reported that participants without access to a computer "were loaned, free of charge, a WebTV computer" and "were instructed on its use".
Finally, two studies (Lepore 2014; Salzer 2010) required that eligible participants had to be fluent in English, Klemm 2012 required that participants had to be "able to read and write English" and Winzelberg 2003 specified "being able to communicate in written English" as an inclusion criterion. In Changrani 2008, all participants were Spanish speaking. Vilhauer 2010 did not mention language as an inclusion criterion.
Excluded studies
We excluded 46 studies (see PRISMA flow chart in Figure 1) because they were not randomised controlled trials, did not meet the inclusion criterion of providing an online support group or included breast cancer as well as other cancers but did not provide separate data related to breast cancer. Eleven studies (see Characteristics of excluded studies) initially appeared relevant to this review, but after further assessment, we excluded these 11 studies after independent assessment by two review authors (EM and KP) and discussion with another review author (LN). We excluded four studies (Badger 2013; Børøsund 2014; Owen 2005; Schover 2013) because investigators tested online interventions other than online support groups. We excluded three studies (Gustafson 2001; Gustafson 2008; Ruland 2013) that tested other types of online interventions but had an online support group component, because they did not provide separate data for online support groups. We excluded three other studies (Hoybye 2010; Klemm 2002; Stephen 2013) because they included participants with various cancers (including breast) but did not provide separate data for breast cancer. We excluded one study (Heiney 2012) because it included elements of support groups but did not report outcomes of interest for this review.
Risk of bias in included studies
Refer to Figure 2 and Figure 3.
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Allocation
Three studies (Klemm 2012; Lepore 2014; Vilhauer 2010) used a 'coin flip' or a computer package to randomise participants to study groups and were therefore judged to be at low risk of bias. Changrani 2008 reported that "the randomisation protocol was compromised by selecting patients serially as they registered"; we judged this study to be at high risk of bias. The remaining two studies (Salzer 2010; Winzelberg 2003) reported that investigators randomly assigned participants to groups but did not explain how this happened; therefore, we judged these studies to be at unclear risk of bias.
One study (Lepore 2014) reported concealment of allocation by use of opaque envelopes and was assessed as being at low risk of bias. The other five studies provided no information on allocation concealment in published papers nor via personal communication; therefore, we classified these studies as having unclear risk of bias.
Blinding
Blinding of participants to group allocation is problematic in studies involving support groups or other psychosocial interventions, as participants knew what they were receiving (and no placebo was involved). Therefore, it is not surprising that the authors of five studies (Changrani 2008; Klemm 2012; Salzer 2010; Vilhauer 2010; Winzelberg 2003) did not mention whether participants were blind to group allocation, nor whether personnel, in particular, support group moderators, were aware of group allocation; we judged these studies to be at unclear risk of bias. The remaining study (Lepore 2014) described that "participants and interventionists" were not told of the study hypotheses; therefore, we considered this study to be at low risk of bias.
Four studies (Changrani 2008; Klemm 2012; Salzer 2010; Winzelberg 2003) provided no information on whether data assessors were blinded to group allocation. In one study (Vilhauer 2010), the author stated that she carried out all data entry and analysis but could not say whether data assessment was blind to group allocation (personal communication). Therefore, we judged these studies as having unclear risk of bias. In the remaining study (Lepore 2014), "trained researchers, who were blind to condition" collected the data; therefore, we assessed this study as having low risk of bias.
Incomplete outcome data
Researchers in five studies (Changrani 2008; Klemm 2012; Lepore 2014; Vilhauer 2010; Winzelberg 2003) provided information about participants lost during the study and reasons for attrition. These studies were assessed as having low risk of bias. One study (Salzer 2010) provided overall attrition rates at two time points, but did not provide data on how many participants in each group did not complete the study; it was assessed as having high risk of bias.
Selective reporting
Four studies (Changrani 2008; Klemm 2012; Lepore 2014; Winzelberg 2003) reported all outcomes. We judged these studies to be at low risk of bias. Salzer 2010 reported outcome data but information on precise attrition in groups was missing, and Vilhauer 2010 did not provide detailed results at two months post intervention. We judged these two studies (Salzer 2010; Vilhauer 2010) to be at unclear risk of bias.
Other potential sources of bias
Two studies (Lepore 2014; Salzer 2010) appeared to have no other potential sources of bias. We judged two studies (Changrani 2008; Klemm 2012) as having unclear risk of bias from other potential sources. Changrani 2008 was a small feasibility study that provided intentional, unequal allocation of participants to two study groups; and Klemm 2012 randomised women with more depressive symptoms to peer‐led groups. We determined that the remaining two studies (Vilhauer 2010; Winzelberg 2003) were at high risk of bias from other potential sources. Vilhauer 2010 was a small feasibility study in which 43% of participants were attending face‐to‐face support groups and 20% were receiving individual psychotherapy; and Winzelberg 2003 reported that 34% of women were participating in another breast cancer support group and in individual counselling at baseline.
Effects of interventions
See: Summary of findings for the main comparison Online support group for women with breast cancer
Refer to summary of findings Table for the main comparison. The primary outcomes in this review were emotional distress, uncertainty, anxiety and depression. Quality of life was a secondary outcome.
Emotional distress
None of the included studies measured 'emotional distress' as an outcome. Salzer 2010 used the Hopkins Symptoms Checklist (HSCL‐25) and the Profile of Mood States (POMS) to measure 'distress' and used the Impact of Events Scale (IES) to measure 'cancer‐specific distress'. They reported that participants (78 women) in the Internet peer‐to‐peer support group (intervention) "did worse" on these outcomes than those in the control group at four and 12 months post intervention.
Uncertainty
None of the six included studies measured uncertainty as an outcome.
Anxiety
Only two studies (Lepore 2014; Winzelberg 2003) measured anxiety as an outcome. Winzelberg 2003 used the 20‐item State‐Trait Anxiety Inventory (STAI) to measure anxiety as an outcome in a sample of 72 women with breast cancer when investigators compared a web‐based support group versus a waitlist control group. Researchers reported no statistically significant change in anxiety (MD ‐0.40, 95% CI ‐6.42 to 5.62; Analysis 1.1) between the two groups. We rated the quality of evidence as low owing to unclear risk of bias regarding randomisation, blinding of participants and personnel and imprecision (small sample).
Lepore 2014 compared a standard Internet support group (S‐ISG) with a prosocial Internet support group (P‐ISG) for 184 women with breast cancer (see Characteristics of included studies) and used the seven items that measured anxiety in the HADS. Health professionals facilitated sessions for both groups, and the study included only distressed participants (scoring above normal (≥ 8) for levels of depression and anxiety on the HADS). For S‐ISG, investigators "emphasized the exchange of information and emotional support between peers", and participants in enhanced P‐ISG "received written tips on how to recognise and respond to others’ need for support online" (p.4082). These researchers reported decreased symptoms of anxiety from baseline post intervention in both groups but noted that participants in the P‐ISG did not do as well as those in the S‐ISG, as they had hypothesised. Study authors commented that the lack of a usual control group prevented them "from estimating how much of symptom improvement was attributable to natural recovery" (p.4085).
Depression
Five studies (Changrani 2008; Klemm 2012; Lepore 2014; Vilhauer 2010; Winzelberg 2003) reported depression as a study outcome. Lepore 2014 (184 women) used the HADS to measure depression, and the other four studies used the 20‐item Center for Epidemiologic Studies Depression Scale (CESD). Three studies (Changrani 2008; Vilhauer 2010; Winzelberg 2003) included an online support group and a control group. Two studies (Klemm 2012; Lepore 2014) included no control group ‐ only comparison groups.
Vilhauer 2010 compared 'peer‐to‐peer' online support groups with a waitlist group and reported no difference in levels of depression post intervention. All participants were at the metastatic stage of breast cancer. This was a pilot study with a small sample size (30 women) and 27% attrition. No extractable data were available for analysis.
Changrani 2008 reported no statistically significant differences after comparing the effectiveness of an online support group versus a control group in reducing depression. Study authors described all participants (68 women) as "underserved immigrant Latinas". Winzelberg 2003 included 72 women with breast cancer (most were highly educated, white women) and reported that its web‐based support group intervention was more effective in reducing depression when compared with a "waitlist" control group. Data pooled from Changrani 2008 and Winzelberg 2003 showed a small to moderate decrease in depression among online support groups when compared with waitlist groups (SMD ‐0.37, 95% CI ‐0.75 to 0.00; two studies; 120 women; Analysis 1.2; Figure 4). We rated the quality of this evidence as very low owing to unclear risk of bias (compromised randomisation in one study (Changrani 2008) and lack of information on blinding of participants and personnel in both studies), imprecision (small sample) and inconsistency (population heterogeneity).
Forest plot of comparison: 1 Online support group versus usual care, outcome: 1.2 Depression at end of intervention.
Klemm 2012 included 60 women and used the CESD to compare "moderated" and "peer‐led" online support groups at three time points (six, 12 and 16 weeks). Investigators hypothesised that women with breast cancer taking part in moderated online support groups (facilitated by social workers, with preselected topics for discussion each week) would experience fewer depressive symptoms when compared with those participating in a peer‐led online support group. Results showed no significant differences (P > 0.05) between the two groups in depressive symptoms at the three time points. Study authors commented that the sample size may not have been large enough to permit detection of subtle changes in depressive symptoms.
Lepore 2014 compared an S‐ISG with a P‐ISG for 184 women with breast cancer (see section on anxiety, above). Investigators used the HADS to measure depression and reported that both groups experienced a reduction in depression, but those in the S‐ISG had significantly lower depression levels than those in the P‐ISG. The absence of a usual care control makes it difficult to know how much this reduction in depression was due to natural recovery.
Quality of life
Only three studies (Changrani 2008; Salzer 2010; Vilhauer 2010) measured quality of life; all used the Functional Assessment of Cancer Therapy – Breast Cancer (FACT‐B). Vilhauer 2010 (30 women) reported no statistically significant differences in reported quality of life when comparing an online support group with a "waitlisted control condition" group. This study provided no extractable data for analysis. Changrani 2008 (68 women) reported no statistically significant differences in quality of life resulting from an online cancer group intervention when compared with control. Salzer 2010 (78 women) compared an Internet peer‐to‐peer support intervention versus a control and noted that participants in the Internet group reported lower quality of life than those in the control group at four months post intervention. They concluded that these findings should be treated with caution, as the study was "underpowered to detect small–moderate effects" (p.445).
Pooled data from Changrani 2008 and Salzer 2010 showed no significant change in quality of life at four months post intervention (SMD ‐0.11, 95% CI ‐0.47 to 0.24; Analysis 1.3; Figure 5). Salzer 2010 provided no standard deviations and used pooled SDs for calculations. Although this study provided an overall attrition rate at two time points, investigators did not provide clear data on how many participants in each group did not complete the study and used baseline figures in study calculations. We rated the quality of evidence from both Salzer 2010 and Changrani 2008 as very low owing to high or unclear risk of bias (compromised randomisation in one study (Changrani 2008), lack of information regarding concealment and blinding of participants and personnel in both studies), imprecision (small samples) and high statistical heterogeneity (I² = 81%). Only Salzer 2010 reported data on quality of life at 12 months post intervention, revealing a decrease in quality of life in the intervention group compared with the control group (MD ‐10.89, 95% CI ‐20.41 to ‐1.37; Analysis 1.4). We rated the quality of this evidence as low owing to unclear risk of bias (lack of information regarding randomisation, concealment and blinding of participants and personnel) and imprecision (small sample).
Forest plot of comparison: 1 Online support group versus usual care, outcome: 1.3 Quality of life post intervention.
None of the included studies mentioned adverse effects.
Data were insufficient for subgroup analysis on stage of disease, treatment modality and type and dose of interventions. Overall risk of bias in the included studies was unclear or low.
Discussion
Summary of main results
None of the included studies measured emotional distress or uncertainty as an outcome of online support groups. Results showed no difference in anxiety levels between those who participated in online support groups and controls (based on low‐quality evidence from one study). Two studies showed a small to moderate reduction in depression levels among participants in online support groups compared with controls, but we rated the quality of this evidence as very low. Pooled data from two studies showed no difference in quality of life among women with breast cancer at four months post intervention between online support groups and control groups; we rated the quality of this evidence as very low. Similarly, another study showed no difference in quality of life at four months on the basis of low‐quality evidence.
Readers should treat the results of this review with caution as we obtained evidence regarding selected outcomes from a few small studies with several methodological weaknesses.
Overall completeness and applicability of evidence
We found few studies with large enough samples to adequately address the objectives of this review. Investigators described three of the six included studies as 'pilot' or 'feasibility', and two of six studies included no control group but instead tested two different formats of online support groups. Not all included studies measured all of the outcomes in this review, and even fewer studies provided data on each outcome. The overall implication of underpowered trials is a scarcity of data from which to draw conclusions regarding effects of online support groups. Lack of extractable data for pooling and selective reporting added to the difficulties of this review.
Applicability of evidence from included studies should be put in the context of types of participants and interventions. In four of the six studies, 90% to 100% of participants were ‘white’. One study described all participants as "underserved immigrant Latinas". The remaining study provided no information on ethnic background of participants. In the three studies that provided information on income, participants reported annual earnings of 40,000 to 50,000 $USD. The mean reported income in one study was over 83,000 $USD.
In four of the six studies that provided information about educational attainment, most participants had completed education at high school level or higher. The generalisability of review findings to an ethnically diverse, more deprived and less educated population (than participants in this review) is questionable. Generalisation to populations outside the USA (where all six studies were conducted) is also limited. Online support groups involve cultural norms and behaviours that are often specific to context. One of the authors of this review (Changrani 2008) explained that different cultures have different values. She compared traditional Western values, which stress the desirability of individualism, autonomy and competition, versus Hispanic cultural traditions, which emphasise the importance of collectivism, interdependence and cooperation.
Studies showed some degree of heterogeneity in terms of stage of disease. In one study, all participants were at the metastatic stage. Participants in all other included studies were at disease stages I and II. One study recruited only participants with distress and depression levels ≥ 8 on the Hospital Anxiety and Depression Scale (HADS). All participants received one or more forms of treatment.
Investigators also reported variation among interventions in terms of dose, frequency, content and format. Study duration ranged between six and 30 weeks. Participation in group sessions lasted between 90 and 142 minutes per week. Some sessions had 'no set agenda' and others focused on 'preselected' topics. Three of the eight interventions were peer led; social workers, health professionals or mental health workers moderated all others. In practice, variations among online support groups currently in operation are similar. Outcomes selected for this review (emotional distress, uncertainty, anxiety, depression and quality of life) seem to be more health professional centred than participant centred. Included studies also measured other outcomes such as participation, satisfaction, personal growth and social support.
Quality of the evidence
We have reported details about potential sources of bias in the Characteristics of included studies table and in summary of findings Table for the main comparison. Other weaknesses in the methods of most included studies may have compromised the integrity of these studies. With regards to anxiety, evidence from one large (184 participants) study (Lepore 2014) of high quality (see Risk of bias in included studies) shows that participants in the standard online support group reported less anxiety than those in the enhanced online support group of women with breast cancer. One study with 72 participants (Winzelberg 2003) provided evidence that online support groups did not reduce anxiety levels among women with cancer compared with control groups. Winzelberg 2003 reported that 34% of participants (9 control and 13 intervention) were participating in another breast cancer support group or were receiving individual counselling at baseline. However, "no data were collected on posttest participation in psychologic interventions" (p.1166). Participation in other groups could have confounded the results of this study. On the basis of these limitations, as well as the risk of bias assessment (see Risk of bias in included studies) and evaluation of criteria in terms of GRADE assessment, we rated the quality of evidence (regarding anxiety) derived from this study as low.
Similarly, we obtained evidence that participants in online support groups had a small to moderate reduction in depression (when compared with controls) in two studies (total population: 120): Winzelberg 2003 and Changrani 2008. Changrani 2008 is a feasibility study in which investigators allocated twice as many participants to the experimental group. Changrani 2008 researchers pointed out that the "randomisation protocol was compromised by selecting patients serially as they registered" (p.61). Limitations of these two studies (Changrani 2008; Winzelberg 2003) could have reduced the strength of evidence related to depression. On the basis of these limitations, as well as the risk of bias assessment (see Risk of bias in included studies) and evaluation of criteria in terms of GRADE assessment, we rated the quality of evidence (regarding depression) derived from these studies as very low.
Two studies included comparison groups but no usual care control groups. One large, rigorously conducted study (Lepore 2014) with 184 participants showed a greater reduction in depression in the standard online support group than in the enhanced online support group for women with breast cancer. Klemm 2012 was the only study (total population: 60) that provided evidence of no difference in depressive symptoms between participants in 'moderated' and 'peer‐led' groups. The report of study authors indicating that women with more depressive symptoms were randomised to the peer‐led group revealed that this evidence was compromised.
Three studies measured quality of life. Two studies (Changrani 2008; Salzer 2010) with a total population of 140 provided evidence that online support groups did not affect the quality of life of women with breast cancer. We have pointed out the limitations of Changrani 2008. The second was a small study (Salzer 2010) in which investigators intentionally allocated twice as many participants to the experimental group. As a result, "the study was underpowered to detect small‐moderate effects at P < 0.05 level" (p.445). Also, Salzer 2010 did not provide details of attrition in the two groups. On the basis of these limitations, along with the risk of bias assessment and evaluation of criteria in terms of GRADE assessment, we rated the quality of evidence (regarding quality of life at four months post intervention) as very low. Similarly, we rated evidence from this trial suggesting that quality of life did not improve at 12 months post intervention as low. The third study (Vilhauer 2010) was a ‘feasibility pilot’ study with a small sample (n = 30), in which investigators reported that 43% of participants were attending face‐to‐face support groups, and 20% individual psychotherapy. Results showed higher attrition at the end of the intervention in the control group (36%) than in the experimental group (19%). These factors could have introduced bias and confounders. This study did not provide extractable data for pooling, but study authors reported no differences between intervention and control groups.
Potential biases in the review process
A potential limitation of reviews in general involves missing key studies. Although we carried out an extensive search of the literature at the start of this review, the search for studies that could have been missed did not end until the review was completed. Most journals offer online facilities to help readers search for 'related articles' or similar articles from the same study authors. Use of these tools, as well as perusal of the reference lists of relevant studies and reviews, provided opportunities to validate results of the main search and to find new papers.
Another factor that may have affected the results of this review is the non‐availability of relevant studies published in the public domain (i.e. grey literature). Researchers conducted all included studies in the USA and published study findings in the English language. The possibility exists that relevant studies may have been carried out in other countries and published in languages other than English.
Although we contacted all but one of the included study authors, only one responded. Non‐availability of key data and information needed to clarify the trial process was beyond the control of the authors of this review (and of some authors of these studies, as some studies were conducted longer than a decade ago). If all data for pooling and other information had been available, results of this review may have been different.
Outcomes selected for reviews often reflect the interests of the review authors and, to some extent, what is available in the literature. Other outcomes such as 'patient satisfaction', 'connecting with others' and 'empowerment' may have presented online support groups in a different light.
With hindsight, we believe it would have been useful to include studies with other designs along with randomised controlled trials. Findings of these studies may have added strength to the review conclusions, although lack of controls would have added to the inconclusiveness of the evidence. We propose to be more inclusive in our selection criteria in future updates of this review.
Defining online support groups, as well as choosing to exclude studies of interventions that included some elements of support groups along with therapy, training or services, was a subjective exercise, albeit carried out collectively by the authors of this review. Others undertaking this review may have been more inclusive than we have been. We deliberated for a long time as to whether we should include 'high‐intensity' interventions that were in fact 'psychosocial interventions'. We decided to exclude them to protect the integrity and focus of our review (i.e. online support groups for women with breast cancer, whether led by participants or health professionals). Inclusion of 'high‐intensity interventions' in this review would have added heterogeneity and further diluted review findings. Scope and opportunity exist for a systematic review of online psychosocial interventions for women with breast cancer.
Agreements and disagreements with other studies or reviews
In this section, we compare studies of online support groups for women with breast cancer that used a design other than a randomised controlled trial. We also compare the findings of this review with those of relevant systematic reviews.
Lieberman 2003 was a pretest/post‐test study including 32 participants at all stages of breast cancer. They reported that participants’ depression (as measured by the Center for Epidemiologic Studies Depression Scale (CESD)) was "significantly reduced". This study lacked randomisation and control. Another pre‐test/post‐test study (Battenburg 2014) of 133 Dutch women with breast cancer reported that depression did not change significantly from baseline to six months after participation in an online peer‐led support group. This study had no control group.
A literature review on online cancer support groups (Klemm 2003) identified 10 studies, of which six focused on women with breast cancer. None of these studies included randomisation to groups, and none included control groups. The review revealed one study (Kraut 1998) in which investigators found that more time spent on the Internet led to higher levels of depression among participants. Internet use, in this case, referred to all types of social interactions conducted via the Internet, including participation in online groups. However, Klemm 2003 did not report specifically the effect of online support groups on depression.
Systematic reviews of support group studies of participants with cancer have reported that most of these studies involved women with breast cancer (Hoey 2008; Hong 2012). Hoey 2008 was a systematic review of peer support programmes for people with cancer, including face‐to‐face and online programmes. Studies included in this review were descriptive (n = 26), non‐randomised (n = 8) and randomised controlled trials (n = 8). Evidence of psychosocial benefit was mixed, and randomised controlled trials reported no significant effects on quality of life. In a systematic review of online support and resources for cancer survivors (Hong 2012), 14 of the 24 included studies focused on women with breast cancer, and only four of the 24 studies used a randomised controlled trial design. Although most of the included studies reported positive effects on psychosocial outcomes, none of the randomised controlled trials reported significant positive outcomes.
A systematic review (Griffiths 2009) of depression‐specific online support group studies (n = 28) included participants with cancer and those with other conditions such as mental disorder, diabetes and kidney disease requiring dialysis. However, Internet support groups focused more on patients with breast cancer than on patients with any other condition. Results showed that peer‐to‐peer Internet support groups had a positive effect on depressive symptoms; however, only two of the 17 studies reporting this effect had used a controlled trial design. Of the five studies that involved women with breast cancer, three reported significant moderate to large effects on depression, and only one (Winzelberg 2003) was a randomised controlled trial. Griffiths 2009 concluded that breast cancer online support groups were more likely to be associated with positive results with regards to depression than were online support groups for other patients.
All three systematic reviews (Griffiths 2009; Hoey 2008; Hong 2012) described methodological weaknesses in most of the studies that used a randomised controlled trial design. Griffiths 2009, in particular, concluded that the “most salient finding” of this review "was the paucity of high‐quality studies". This finding is significant because review authors reported "a trend toward an association between lower design quality and positive outcomes" (Griffiths 2009).
Overall, evidence from these studies and from systematic reviews indicates that online supports may have a positive effect on depression, although this finding is by no means conclusive. Results of the present review also showed a small to moderate reduction in depression, albeit from two studies with methodological weaknesses. Both this review and that of Hoey 2008 concluded that support groups did not have significant effects on quality of life. We found no systematic reviews against which to compare results of this review for the other outcomes (emotional distress, uncertainty and anxiety). These three systematic reviews (Griffiths 2009; Hoey 2008; Hong 2012) and this current review concur that large and robust studies on the effectiveness of online support groups for women with breast cancer are needed.
Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Forest plot of comparison: 1 Online support group versus usual care, outcome: 1.2 Depression at end of intervention.
Forest plot of comparison: 1 Online support group versus usual care, outcome: 1.3 Quality of life post intervention.
Comparison 1 Online support group versus usual care, Outcome 1 Anxiety at end of intervention.
Comparison 1 Online support group versus usual care, Outcome 2 Depression at end of intervention.
Comparison 1 Online support group versus usual care, Outcome 3 Quality of life post intervention.
Comparison 1 Online support group versus usual care, Outcome 4 Quality of life 12 months post intervention.
| Online support group for women with breast cancer | ||||
| Patient or population: women with breast cancer | ||||
| Outcomes | Relative effects* (95% CI) | Number of participants | Quality of the evidence | Comments |
| Online support group vs usual care | ||||
| Emotional distress | Not reported | ‐ | Emotional distress was not measured. However, 1 study reported that the intervention did not have a positive effect on "distress" and "cancer‐specific distress" | |
| Uncertainty | Not reported | ‐ | ‐ | None of the 6 included studies measured this outcome |
| Anxiety at end of intervention at 1 time point (end of a 12‐week intervention) | Mean anxiety in the intervention group was 0.4 lower at end of intervention (95% CI 6.42 lower to 5.62 higher) | 58 | ⊕⊕⊝⊝ | |
| Depression at end of intervention | Mean depression in the intervention group was 0.37 standard deviations undefined lower (95% CI 0.75 lower to 0) | 120 | ⊕⊝⊝⊝ | |
| Quality of life post intervention | Mean quality of life (at end of intervention) in the intervention group was 0.11 standard deviations undefined lower (95% CI 0.47 lower to 0.24 higher) | 140 | ⊕⊝⊝⊝ | |
| Quality of life 12 months post intervention | Mean quality of life in the intervention group (at 12 months post intervention) was 10.89 undefined lower (95% CI 20.41 lower to 1.37 lower) | 78 | ⊕⊕⊝⊝ | |
| *Risk in the intervention group (and its 95% confidence interval) is based on assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI) | ||||
| GRADE Working Group grades of evidence | ||||
| aLack of information on randomisation, concealment and blinding of participants and personnel | ||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Anxiety at end of intervention Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
| 2 Depression at end of intervention Show forest plot | 2 | 120 | Std. Mean Difference (IV, Fixed, 95% CI) | ‐0.37 [‐0.75, 0.00] |
| 3 Quality of life post intervention Show forest plot | 2 | 140 | Std. Mean Difference (IV, Fixed, 95% CI) | ‐0.11 [‐0.47, 0.24] |
| 4 Quality of life 12 months post intervention Show forest plot | 1 | Mean Difference (IV, Fixed, 95% CI) | Subtotals only | |
