We undertook a realist review [33] of the evidence on cancer OSGs. Realist reviews aim to build and test theories that explain how complex interventions work by examining how underlying causal processes (eg, mechanisms) in specific circumstances (eg, contexts) interact to produce results (eg, outcomes) [38]. We followed the steps by Pawson et al [33] and the RAMESES (Realist and Meta-Narrative Evidence Syntheses: Evolving Standards) quality standards for realist reviews (Checklist 1) [38]. Our methods were described in detail in a published protocol [39] that was registered on PROSPERO (International Prospective Register of Systematic Review; #CRD42021250046).

An integrated knowledge translation approach [40] was followed. Five categories of knowledge users (n=15 in total) were engaged in the project: (1) program developers who have experience developing OSGs; (2) community strategists and moderators that manage and moderate OSGs; (3) clinicians who refer patients to or deliver OSGs; (4) patients, survivors, and family members who have used or created their own “grassroots” OSGs; and (5) health care administrators and policymakers responsible for implementing OSGs and driving best practices. Knowledge users were consulted when designing the study protocol, developing the initial program theory, refining the program theory, and generating best practice recommendations, and they participated as coauthors in the writing of this paper.

An initial program theory was developed based on Bender’s multitheory framework on the use and effects of breast cancer OSGs [28], and a 2-hour consultation workshop with project team members and knowledge users. This multitheory framework draws on the Transactional Theory of Stress and Coping [41], Social Comparison Theory [42], the Technology Acceptance Model (TAM) [43], and the Theory of Planned Behavior [44]. As shown in our protocol paper [39], it suggests that people with cancer use OSGs to obtain information and emotional support from similar others to lessen negative appraisals of events, enhance feelings of control, and develop coping strategies, which in turn reduce or buffer anxiety. Beliefs about the usefulness and trustworthiness of OSGs influence adoption and engagement. The availability, anonymity, and low commitment afforded by the online medium are believed to enhance feelings of control. However, the effects of making social comparisons in OSGs may be positive or negative depending on the circumstances and the person’s goals.

The search strategies were developed by an information specialist (ME) and peer-reviewed by an external medical information professional following the PRESS (Peer Review of Electronic Search Strategy) guidelines [45]. Given the range of terms for OSGs and cancer diagnoses, our search strategy was intentionally broad, opting for sensitivity over precision (Multimedia Appendix 1). The following databases were searched from inception to February 15, 2025: MEDLINE, Embase, the Cochrane Database of Systematic Reviews, the Cochrane Central Register of Controlled Trials, Ovid EmCare Nursing, APA PsycINFO (all via Ovid), CINAHL (Ebscohost), and Scopus (Elsevier). In addition, 2 clinical trial registries were searched (ClinicalTrials.gov and WHO ICTRP), along with dissertations (ProQuest Digital Dissertations International) and books or chapters (University of Toronto Summon OneSearch).

The inclusion and exclusion criteria were based on a foundational systematic review of OSGs conducted by Eysenbach et al [20], with feedback from team members and knowledge users (Table 1). The papers gathered from the search were uploaded to Rayyan systematic review management software for screening and selection [46]. To assess interrater agreement, 4 researchers (SB, LRS, AC, and CS) independently screened the same 100 titles and abstracts using a pilot-tested screening tool (Table 2). As 90% agreement was established, the remaining records were divided among the 4 researchers to screen independently, with meetings held every 2 weeks to track progress and discuss and resolve uncertainties. A 10% random sample was audited by the lead author, and disagreements were resolved through discussion. The intentionally broad search strategy yielded a significant number of irrelevant papers (eg, negative answer to at least 1 screening question), the majority of which were excluded at the title and abstract screening stage.

The relevance and rigor of papers were assessed in 2 phases. Relevance was determined by assessing the extent to which the data could contribute to developing and testing context-mechanism-outcome configurations (CMOC) [33]. This was conducted by 3 team members (SB, LRS, and SS). A total of 2 team members independently assessed the relevance of each paper that passed the title and abstract screening and met to resolve uncertainties. A 10% random sample was audited by the lead author, and disagreements were resolved through discussion. Rigor was determined postanalysis for CMOCs with fewer than 5 supporting papers. This approach is recommended for realist reviews when there is sufficient data to develop and test CMOCs, as was the case for most CMOCs in this study [47]. Rigor was determined by assessing whether the methods were credible and trustworthy using the Mixed-Method Appraisal Tool [48]. To establish interrater agreement, 3 team members (SS, LRS, and JLB) independently assessed the same 5 papers using the Mixed-Method Appraisal Tool. Each remaining paper was independently assessed by 2 team members, who met to resolve uncertainties. The results of the rigor assessment are shown in Tables 3 and 4.

Qualitative study criteria [48] asked the following question: (S1) Are there clear research questions? (S2) Do the collected data allow for addressing the research question? (1.1) Is the qualitative approach appropriate to answer the research question? (1.2) Are the qualitative data collection methods adequate to address the research question? (1.3) Are the findings adequately derived from the data? (1.4) Is the interpretation of results sufficiently substantiated by data? (1.5) Is there coherence between qualitative data sources, collection, analysis, and interpretation?

Quantitative descriptive study criteria [48] included the following questions: (S1) Are there clear research questions? (S2) Do the collected data allow for addressing the research question? (4.1) Is the sampling strategy relevant to address the research question? (4.2) Is the sample representative of the target population? (4.3) Are the measurements appropriate? (4.4) Is the risk of nonresponse bias low? (4.5) Is the statistical analysis appropriate to answer the research question?

Quantitative nonrandomized study criteria [48] included the following questions: (3.1) Are the participants representative of the target population? (3.2) Are measurements appropriate regarding both the outcome and the intervention (or exposure)? (3.3) Are there complete outcome data? (3.4) Are the confounders accounted for in the design and analysis? (3.5) During the study period, was the intervention administered (or exposure occurring) as intended?

Quantitative nonrandomized study criteria [48] included the following questions: (S1) Are there clear research questions? (S2) Do the collected data allow for addressing the research question? (3.1) Are the participants representative of the target population? (3.2) Are measurements appropriate regarding both the outcome and the intervention (or exposure)? (3.3) Are there complete outcome data? (3.4) Are the confounders accounted for in the design and analysis? (3.5) During the study period, was the intervention administered (or exposure occurred) as intended?

Two team members (SB and LRS) extracted the study characteristics from each paper using a pilot-tested study extraction tool. Patient demographics (Multimedia Appendix 2 [22-24,28,29,49-210]) and intervention characteristics (Multimedia Appendix 3 [22-24,28,29,49-210]) were extracted. In parallel, 1 team member (SB) uploaded the papers into NVivo (version 12; Lumivero) qualitative analysis software to code sections of text that were relevant to supporting, refining, or refuting aspects of the program theory. A preliminary coding framework was developed based on the concepts in the initial program theory. A total of 2 researchers (SB and LRS) used the coding framework to code a random sample of 5 papers to determine agreement and refine the coding framework, and then 2 team members (SB and SS) coded each of the remaining papers. Coding consisted of initially grouping the data into high-level conceptual “buckets” that were defined iteratively through discussions with the lead author (JLB).

In the final step, data sorted into conceptual buckets were analyzed to determine whether the data were functioning as context, mechanisms, or outcomes [33]. This analysis was undertaken by 2 team members (JLB and SS) trained in realist methodology guided by an expert in realist methodology (GW). Further, 1 team member (SS) identified prominent patterns of contexts and outcomes in the data and sought to explain them through mechanisms by which they occurred, generating CMOCs. The findings were iteratively confirmed, refuted, or refined, first by JLB and then by GW through discussion with SS. The findings were shared with this project’s team and knowledge users for their feedback in a final 2-hour consultation workshop.

A total of 168 papers were included in the analysis, which involved a range of 1 (case study) to 180,000 (content analysis) study participants (Figure 1). Studies were conducted within a span of 25 years (1999‐2025). See Table 5 and Multimedia Appendices 2 and 3 for further details.

Once an individual finds an OSG that is a good fit, whether they experience positive outcomes depends on how they participate, the people involved, and the discussion that occurs. This is influenced by communication technology factors, group composition and dynamics (including the presence and nature of moderation), and the nature and content of interactions. These factors trigger underlying mechanisms such as feeling understood, accepted, cared for, valued, reassured, informed, and confident, which result in positive outcomes.

If an OSG is a good fit for an individual’s needs, and encourages supportive, inclusive discussion, users can experience various forms of social support: informational support (advice or guidance), emotional support (expressions of comfort, caring, and encouragement), appraisal support (help to assess the appropriateness or normativeness of thoughts, feelings, and experiences), and altruism-related esteem support (helping others with no expectation of reward). These types of support lead to positive or negative appraisals and coping mechanisms, which impact psychosocial well-being by affecting distress, isolation, self-esteem, and empowerment. Given that social support is a multidimensional construct, the same exchange in an OSG can yield multiple forms of social support, providing multiple benefits simultaneously.

Based on the findings, we have generated evidence-based practice recommendations to optimize the use and effects of OSGs for patients with cancer and survivors of cancer (Textbox 1).

This is the first realist review of the evidence on OSGs. The findings show that the functional provisions of social support gained from cancer OSGs can lead to changes in psychosocial outcomes such as distress, isolation, empowerment, and self-esteem, through positive and negative appraisals and coping efforts. However, whether an individual chooses to use OSGs as a source of support depends on their attitudes about OSGs, how well the OSG fits them and their needs, and their perceptions of control. In addition, whether users experience positive or negative outcomes depends on specific mechanisms that occur within OSGs that influence their reactions and reasoning. These underlying mechanisms are influenced by contextual aspects of the communication technology, group composition and dynamics, and nature and content posted in OSGs. Overall, this study confirms the potential benefits and pitfalls of OSGs for patients with cancer and survivors of cancer. It also highlights the complexity of factors that influence whether an individual uses and benefits from an OSG. These findings validate and extend the multitheory framework proposed by Bender et al [28] on the use and effects of breast cancer OSGs, resulting in a refined program theory to explain how, why, for whom, and in what contexts OSGs work, visually depicted in Figure 2.

Patients with cancer and survivors of cancer are driven to OSGs to address a specific need. Drawing on TAM [43], our findings demonstrate that the decision to use a cancer OSG is influenced by its perceived usefulness and ease of use. The TAM-3 [213] posits that perceived usefulness is determined by social influence (eg, subjective norm and image) and job performance characteristics (eg, job relevance, output quality, and tangible results), while perceived ease of use is determined by control beliefs (eg, computer self-efficacy, perceptions of external control, computer anxiety, and computer playfulness) and usability characteristics (eg, objective usability and perceived enjoyment).

Perceived usefulness is the strongest determinant of OSG use, shaped by the OSG’s ability to meet a specific need, its perceived trustworthiness (eg, quality, currency, and empathic nature of user-generated posts and moderation), and its cultural acceptability and sensitivity. Cross-sectional surveys across diverse patients with cancer and survivors of cancer show that unmet supportive care needs [15] and distress [214] predict OSG use. Qualitative research across diverse health-related OSGs has shown that once needs are met or if needs are not met, patients and survivors withdraw from OSGs [28,51,80], and as time passes, online help-seeking behaviors decrease [144]. As Massimi et al [80] have suggested, departure from an OSG is not necessarily a failing of the site’s design but potentially a logical reaction to changing life circumstances. OSGs must demonstrate cultural sensitivity and competence to be appealing, relevant, and effective for diverse patients [84]. The benefits of OSGs are enhanced when users share a common language, culture, and life experience, and when moderators are trained in cultural sensitivity and competency [85].

Perceived ease-of-use influences OSG adoption, but to a lesser extent than perceived usefulness. Control beliefs (including eHealth literacy [212]) and objective usability are important determinants of ease-of-use. For example, patients with cancer who were confident using the internet were 4 times more likely to use the internet and social media as a resource [214]. The anonymity and privacy of OSGs are frequently reported advantages that afford users control over their personal information and make users feel more comfortable sharing personal information [23,56]. Whereas privacy concerns contributed to a lack of control over personal information and discouraged participation in OSGs [151]. Furthermore, studies have shown that engagement declines and users leave when a site is frustrating to use [90]. For example, Holtz et al [91] found that conversion of in-person support groups to virtual (video-based) support groups was hindered by technology factors, including poor Wi-Fi connection, background noise, and users speaking over one another. Designing OSGs to ensure ease-of-use and protect privacy will enhance engagement and participation.

Whether patients with cancer and survivors of cancer benefit from OSGs depends on the mechanisms that occur in OSGs that influence their reactions and reasoning. According to the Transactional Theory of Stress and Coping by Lazarus and Folkman [41], how people interpret and appraise events is critical in determining the events’ stressfulness. There are two types of threat appraisals: (1) primary, which involve judgments about whether the event is a threat; and (2) secondary, which involve evaluations of personal and social resources available to cope with the event. Support from peers who have experienced a similar stressor can lessen negative appraisals of events and promote active coping efforts, thereby reducing or buffering anxiety. Several studies in this review confirm that by participating in an OSG, people can learn that their feelings and experiences are normal and to be expected, thereby improving their appraisal of their situation. The validation provided by peers normalizes the experience and reduces the stress associated with the uncertainty of the condition and the stigma imposed by society. Further, people can obtain information on effective ways of coping, which enhances feelings of preparedness, control, and self-esteem.

However, the effects of social support in cancer OSGs will be greatest if the support provided meets an individual’s needs. Optimal Matching Theory [215] posits that the benefits and deleterious effects of social support are due to the matching (or mismatching) of support in different contexts. If the desired type of support is provided, it will enhance psychological adjustment and well-being. For example, in a study of survivors of breast cancer, both misalignment of support desired and received, and unwanted support were associated with poor adjustment [216]. In another study, OSG users who received unwanted support fared worse, which likely led to feelings of helplessness, negatively impacting self-esteem [217]. According to Lazarus and Folkman [41], problem-focused coping, such as seeking information, is most effective when the stressor is controllable, whereas emotion-focused coping is more useful when the situation is not. However, studies have shown that coping strategies that enhance perceptions of control contribute to psychological adjustment regardless of the controllability of the stressor [218]. Collectively, these findings indicate that patients with cancer may require different types of social support at different times and highlight the importance of matching the support provided in OSGs with individuals’ needs for maximum benefit.

Optimal Matching Theory [215] also suggests that aligning the profile of the support provider with that of the support seeker may enhance the effectiveness of the support provided. According to the Social Comparison Theory by Festinger [42], people seek similar others in times of stress and uncertainty to judge the appropriateness of their thoughts, feelings, and behaviors. Not finding similar others or not being accepted by the group can enhance feelings of deviance or uniqueness, worsening anxiety and feelings of isolation. These experiences can cause people to withdraw from OSGs [51,53]. Shared qualities and experiences facilitate mutual identification, disclosure, and belonging because peers understand one another in a way that other network members cannot. The challenge in creating OSGs based on the similarity principle is that users are likely to vary on a host of sociodemographic, clinical, and lifestyle dimensions. OSGs could be better designed to facilitate optimal matching by increasing the specificity of forums and connecting people with similar others with search function optimization. Moderators could also be trained to facilitate group cohesion (sense of belonging, influence, mutual support, and identification with other group members [101,219]), which has been associated with better outcomes in professionally led support groups [220].

The variable effects of OSGs are likely due in part to the negative effects of social comparisons. First, lack of positive upward comparisons (eg, learning what to expect, how to cope, and that it is possible to get better) can increase anxiety and decrease perceptions of control [221]. This is likely why participants in unmoderated breast cancer OSG in a study by Salzer [29] experienced an increase in distress: it was comprised of newly diagnosed patients who had no opportunity to make positive upward comparisons with survivors of cancer. Second, negative social comparisons (eg, learning that you are not doing as well as everyone else or that it is possible to get worse) can increase anxiety, decrease self-esteem, and decrease perceptions of control [221]. For example, being exposed to patients in a breast cancer OSG who were doing poorly was anxiety-provoking and caused users to withdraw [48]. Similarly, loss of members within a lung cancer OSG evoked grief, distress, and concern for the future [145]. At the same time, patients with metastatic breast cancer felt silenced, marginalized, and helpless in mixed-stage OSGs, whereas in stage-specific OSGs, they were able to talk openly and felt understood [53]. While group cohesion can develop across diagnostically dissimilar professionally moderated support groups [219], these findings suggest that stage-specific groups might be more effective for unmoderated OSGs.

Importantly, our findings indicate that the nature and content of posts in OSGs may have the greatest effect on outcomes. Several studies of cancer-specific and mixed-cancer OSGs concluded that the most beneficial posts were those that involved the sharing of feelings and experiences [49,50,54,56,57,60,66,99,101,106,107,119,121,122,126-128,150,153,154,156,167,172,178-181,184,189,191,222]. Emotionally expressive writing is associated with positive psychosocial adjustment and well-being [223], and such disclosures foster feelings of connection and group cohesion [224]. However, expressing emotions is not enough to benefit maximally from an OSG. Several studies of breast cancer OSGs found that posts that express emotions and cognitive processing confer the most benefit for the poster and reader [150,184,189,191]. For example, posts in breast cancer OSGs that included a higher percentage of insightful disclosure words (eg, aware, know, realize, think, and understand) led to greater positive effects that were maintained over time [184,191]. Additionally, insightful disclosure posts were more likely to generate empathetic responses in an OSG for young adults with cancer [182]. Hence, posts that include emotional disclosures and cognitive processing are most helpful for the poster and the reader, and most likely to garner a helpful empathetic response, leading to further benefits. These findings highlight the importance of crafting messages for maximum benefit. However, for the health benefits to manifest, OSGs must provide a safe environment for disclosing emotions and stigmatizing experiences [224]. Community rules can establish acceptable behavior and foster a safe online environment, while anonymity and strong privacy controls can facilitate disclosure.

The extent to which an individual may benefit from OSGs also depends on how they participate, which is influenced by technology factors, group process, and coping style. To be most effective, OSGs need to be active to ensure users can access support when needed, enhancing perceptions of control. Hence, larger groups may be more effective because there are more active members to post and respond. However, ideal group size depends on modality, as live chats can become overwhelming with too many participants and posts [100]. While anonymity contributes to activity in OSGs by facilitating disclosure, it can also contribute to inappropriate behaviors [225]. In this context, people feel less restrained and, without visual cues, do not see the discomfort they may cause others [225]. OSGs that are active may also be more trustworthy because members can participate in regulating the accuracy of the posts. For example, in a breast cancer OSG, only 10 of 4600 posts were found to be false or potentially misleading, and 7 of them were corrected by OSG members within 4 hours of posting [226].

Ongoing investment in moderator training and facilitation guidelines covering conflict resolution, information verification, and cultural sensitivity is recommended [88,144,152]. Some studies report that patients with cancer prefer professionally moderated OSGs for fear of misinformation [99,151]. However, studies comparing peer vs professionally led breast cancer OSGs have found no differences in outcomes [165,196]. Some studies also suggest that people who participate actively in OSGs by posting benefit more than those who participate less actively or passively by lurking [106,191]. However, other studies found that lurkers benefit to a similar degree [227]. Variations in coping styles may be responsible for the variable effects of OSG participation. For example, frequent users of a breast cancer OSG who engaged in emotionally expressive coping (eg, actively expressed and processed emotions) experienced better psychological well-being than frequent users who approached their emotions less actively [106]. Engaging the community in implementing community guidelines and offering users tips on how to participate for maximum benefit could optimize the effects of cancer OSGs.

Overall, the evidence suggests that the people most likely to benefit from cancer OSGs are those who have unmet supportive care needs, perceive OSGs to be useful and easy to use, can find an OSG to which they feel connected, and that meets their needs. This means that people who have greater illness uncertainty (eg, newly diagnosed, experiencing new symptoms, or advancing disease), have a complex medical condition, greater symptom distress or social stigma associated with their condition, and lack sufficient support from offline family and friends, health care professionals, and other patients with cancer, may benefit more from cancer OSGs. Individuals with altruistic support needs, who are doing and coping well and who want to give back by supporting other patients with cancer in OSGs, can also benefit from OSGs.

In addition, most studies to date have examined cancer OSGs for women with breast cancer (71/92, 77% of single cancer studies) or have involved women with breast cancer (46/50, 92% of mixed cancer studies). This may mean that there are more OSGs available for women with breast cancer than for other people with other cancer types. Hence, women affected by breast cancer may have an easier path to benefitting from OSGs at present because there are likely more options for them to choose from to meet their needs.

The findings of this review also suggest that an individual’s coping style may impact whether they use and benefit from OSGs. Seeking information and support is reflective of an active coping style. Women are more likely to engage in active coping strategies and seek support than men [228], which may in part explain prior work demonstrating that women are more likely to use social media for cancer-related support than men [214]. Correspondingly, a population-based study of men with prostate cancer reported that active information seekers were 4.5 times more likely to use the internet as a source of information and support than those who were not [229].

Individuals with emotion-focused coping styles may also benefit more from OSGs, given that the most beneficial posts are those that express emotions and cognitive processing. However, studies of breast and prostate cancer OSGs have documented more information than emotionally supportive talk occurring in cancer OSGs [49,230]. Studies of breast, prostate, and mixed cancer OSGs have also identified gender differences, with some reporting that men are more likely to share information while women are more likely to share emotional support [131,166,230]. In contrast, the discourse analysis of ovarian and prostate cancer OSGs by Sullivan [49] concluded that both genders provide information and emotional support in OSGs, albeit in different ways. Men tended to engage in “report-talk,” providing information in the form of personal case histories, and demonstrated empathy by ensuring the information shared was accurate. In contrast, women more often engaged in “rapport-talk,” providing information and emotional support through the sharing of personal experiences and feelings.

However, not all patients with cancer seek support online or from OSGs specifically. Most studies included in this review reported that participants were college- or university-educated, age was variable, and income was inconsistently reported. Age, education, income, and broadband access remain well-known predictors of internet use [231]. In prior work involving patients with breast, testicular, and head and neck cancer in Canada, 5 variables correlated with OSG use: age, first spoken language, education, treatments received, and unmet supportive care needs [232-234]. In another study, confidence using the internet was associated with use of social media as a resource for cancer information and support [214]. This review similarly found that lower socioeconomic status may limit some individuals’ access to OSGs due to a lack of stable housing, internet access, or eHealth literacy.

Lastly, it is unclear how and to what extent patients with cancer and survivors of cancer who identify with non-White racial and ethnic backgrounds benefit from OSGs. Most studies (101/168, 60.1%) did not specify the racial or ethnic background of study participants. Of those that did, most studies (51/67, 76%) reported that most study participants were White North American or European. These findings mirror those of a review of cancer OSGs conducted over 20 years ago [140], indicative of little progress toward understanding the use and effects of English-language cancer OSGs among racialized ethnic groups and different linguistic communities. There was also a lack of studies on the use and effects of OSGs among 2SLGBTQIA+ (2-spirit, lesbian, gay, bisexual, transgender, queer, intersex, or asexual) patients with cancer. Only 2 such studies were identified: 1 focused on the OSG experiences of sexual minority women with breast cancer [159], and the other focused more broadly on the OSG experiences of sexual minority women with cancer [235]. Both highlighted the importance of designing OSGs to promote inclusivity and acceptance of diverse sexual and gender identities.

This study confirms and extends the multitheory framework we initially proposed to explain the use and effects of breast cancer OSGs. In doing so, the findings demonstrate the value of social support and technology adoption theory in explaining how online support interventions influence health, and realist methodology in uncovering the mechanisms through which complex social interventions exert their effects. Further research, as detailed below, is needed to test the proposed program theory.

The review identified several gaps in the literature. Most of the literature to date has focused on the use and effects of cancer OSGs among White, cisgender, heterosexual women with breast cancer from North America or Europe. Further research is needed to better understand and explore the applicability of our findings among men, ethnically and linguistically diverse minorities and majorities, 2SLGBTQIA+ patients with cancer, and people diagnosed with other common cancers, rarer cancers, and niche cancer groups at different stages of the cancer continuum. In addition, as most studies (109/168, 64.9%) were performed in the United States, further research is needed to determine how OSGs work for people in non-English speaking countries.

There is also a need for more research on the nature of posts that confer the most benefit for diverse gender and ethnocultural groups, how OSGs can be configured and moderated to optimize supportive exchanges for maximal benefit, and interventions designed to maximize benefits. The quality of reporting on the moderation of OSGs is generally poor, leaving many unanswered questions with respect to the training and qualifications of moderators and the extent and style of moderation.

Lastly, further research is needed to explore the implications of new technologies such as virtual reality (VR) and artificial intelligence (AI) in OSGs. VR could remove physical barriers to support, enabling “face-to-face” interactions through avatars and offering an interactive experience, but despite its potential, VR adoption may be limited due to low familiarity and accessibility [144]. In a series of 3 studies, Leung et al [236-238] demonstrated the potential of an AI cofacilitator to recommend resources, detect distress, and facilitate group cohesion in synchronous cancer OSGs with promising findings. The use of AI to support online content moderation raises ethical concerns regarding privacy, free speech, and bias [239]. Guidelines are needed for the use of AI in OSGs to balance the potential benefits with ethical considerations to ensure safe and inclusive online environments.

We followed a rigorous and systematic procedure to synthesize and analyze a substantial body of research using best practices in realist reviews, with guidance from an expert in realist methodologies and 5 categories of knowledge users. The findings consolidate and confirm prior research and provide new insight into the mechanisms underlying the effectiveness of OSGs and practical implications for improvement. However, as with any review method, the resulting synthesis is only as good as the primary data on which it is built. We limited our search to studies published in English, which in most cases focused on English-language OSGs. Another limitation we encountered in our review was that most studies of OSGs focused on women with breast cancer and included a disproportionate number of White, well-educated, and above-average-income earners. The predominance of breast cancer-related data may limit the generalizability of findings; however, using a realist approach enabled us to draw on data across cancer types, which showed commonalities. Second, many studies did not describe the OSGs in sufficient detail, including the technology used, the characteristics of the participants, the training and qualifications of the moderator, and the moderation strategy, and did not follow the CONSORT-EHEALTH (Consolidated Standards of Reporting Trials of Electronic and Mobile Health Applications and Online Telehealth) reporting guidelines [240]. While some studies predate these reporting guidelines, many did not. To improve reporting of OSG interventions, we recommend updating the CONSORT-EHEALTH guidelines to capture the contextual aspects of social-media-based interventions such as OSGs that influence their effects.

OSGs can be an effective source of support for patients with cancer and survivors of cancer that can address many supportive care needs and improve psychosocial well-being. However, outcomes depend on specific contexts and mechanisms that impact how well OSGs meet the needs of users. To maximize the effectiveness of OSGs, we recommend optimizing the design and implementation of OSGs to (1) help assess fit and address specific needs; (2) demonstrate trustworthiness; (3) enhance anonymity and control, and protect privacy; (4) enhance ease-of-use; (5) support connection and belonging; (6) encourage activity; (7) enhance the nature of content shared to boost therapeutic effects; and (8) monitor and adjust design and management strategies to optimize effectiveness.