Social Network Interventions for Improving Dietary Adherence among Adults with Type 2 Diabetes: A Systematic Review

Abstract

Context.

Adhering to recommended diets is essential for managing glycemic concentrations in type 2 diabetes (T2D). However, there is limited evidence on the effectiveness of interventions that leverage social networks to improve dietary adherence.

Objective.

Given the influence of social networks on health behavior, this review aims to assess the effectiveness of social network interventions involving families, friends, and peers in enhancing dietary adherence among people with T2D.

Data sources.

We searched for social network interventions in randomized controlled trials (RCTs) and controlled before-and-after studies aimed at improving dietary adherence and glycemic control across seven databases and two trial registries up to October 2024.

Data extraction:

Data extraction was conducted independently by 2 reviewers. Study quality was assessed using the RoB 2 tool for randomized trials and the ROBINS-I tool for non-randomized studies.

Results.

Ten studies were conducted from 2014 to 2023, including nine RCTs and one quasi-RCT. Most involved family networks (n=7), while two centered on peer support and three on significant others. Half of the studies reported improved dietary adherence, and six showed reduced hemoglobin A1C concentrations. Increased physical activity ranged from 18.6% to 23.6%, with three studies noting weight or BMI reductions. Two studies observed systolic (3.89 to 12.4 mmHg) and diastolic pressure (3.12 to 4.1 mmHg) decreases. One study reported a 0.52-point decrease in diabetes-related stress, and another noted a 27.6% improvement in quality of life. However, six studies had a high risk of bias, and two had unclear risks, mainly due to detection and attrition bias.

Conclusions.

This review suggests that social network interventions can enhance and sustain dietary adherence. However, the evidence is limited due to a lack of high-quality studies and considerable variability in intervention components. More rigorous research using standardized metrics and cost-effectiveness data is necessary to support evidence-based health service recommendations.

Systematic review registration:

PROSPERO registration number ID: CRD42023441223.

INTRODUCTION

Maintaining a healthy diet is crucial for individuals with type 2 diabetes (T2D) to ensure their overall health and well-being.^1–3. A high-quality diet that includes more fruits, vegetables, oily fish, beans, legumes, and whole grains is recommended for optimal weight control, healthy body mass index (BMI), and well-managed hemoglobin A1c (HbA1c) concenrations.^4–6 However, adhering to these dietary guidelines poses a significant challenge, with only 25% of individuals with T2D worldwide following their recommended dietary plan.^{7, 8}

Dietary risk factors significantly affect global health, contributing to approximately 11 million deaths and 255 million disability-adjusted life years (DALYs). In particular, diet-related mortality from T2D accounts for over 330,000 deaths and 24 million DALYs.⁹ Although intensive coaching and support from health professionals can help individuals adhere to dietary guidelines, many return to unhealthy eating habits once this support ends.¹⁰ To address this issue, research suggests that involving non-professional peers from an individual’s social network may provide a sustainable approach to promoting and maintaining long-term dietary changes.^11,12

Social networks consist of relationships with family, friends, colleagues, and neighbors, offering important social, psychological, and behavioral benefits. Two key theories explain these benefits: the stress-buffering hypothesis and the social contagion hypothesis (Figure 1).^13,14 The stress-buffering hypothesis suggests that social networks can protect individuals from the adverse effects of stress related to T2D and support healthy behaviors, such as dietary adherence.^15,16 Managing a diabetes diagnosis can create psychological stress, which is linked to lower dietary adherence.¹⁷ Social networks can assist in decision-making and emotional coping, promoting healthy lifestyles and reducing the risk of returning to unhealthy eating habits.^18–22

Figure 1. :

Conceptual framework for the relationship of social networks to health (Cohen ( 2014)¹³ Berkman et al. (2014) 14

The social contagion hypothesis posits that individuals are influenced by the behaviors of their social networks, which can lead to the adoption of both healthy and unhealthy lifestyles. For people with T2D, interventions using social networks have proven effective in promoting positive self-management behaviors.^{21, 23–27} A meta-analysis by Spencer-Bonilla et al. found that these interventions improved social support, resulting in lower HbA1c concentrations after three months. Additionally, stronger social support boosts self-efficacy for lifestyle changes, such as healthy eating.^{28, 29} People with T2D with supportive families are also more likely to follow dietary recommendations, underscoring the significant impact of social networks on dietary quality.^{24, 25, 30}

Although previous systematic reviews provide some evidence that social network interventions improved short-term HbA1c outcomes, the few studies identified had a high risk of bias.^23,27 In addition, these reviews did not address dietary adherence, a crucial pathway to improving glycemic control. Furthermore, since social networks significantly affect dietary habits, it is important to explore the impact of different network intervention approaches, as previous reviews primarily focused on dyadic approaches involving couples or family members. ^31,32 Therefore, this review aims to evaluate the effectiveness of social network interventions with family, friends, and peers in promoting dietary adherence among people with T2D. We compared these interventions to usual care, no intervention, or those without a social network element. We also explored whether different approaches—such as individual segmentation, induction, or alteration—impact their effectiveness. This review will provide valuable insights for healthcare practitioners and policymakers in developing effective strategies for managing T2D through improved dietary adherence.

METHODS

The protocol for this review was registered with PROSPERO (ID: CRD42023441223). The review adhered to standard systematic review methods and followed the guidelines outlined in the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) [Supplement 1].^{33, 34} This study did not require approval from the Institutional Review Board (IRB) since it used data from published studies.

Search Strategy

We searched several databases from inception to October 31, 2024, including the Cochrane Library’s Central Register of Controlled Trials (CENTRAL), the Cochrane Database of Systematic Reviews, PUBMED, EMBASE, EPOC, LILACS, Open Grey, ProQuest Dissertations and Theses, and Google Scholar. We did not impose any restrictions based on language or publication date. A detailed description of the search strategy is available in the published protocol.³⁵

Eligibility Criteria

We included quasi-experimental studies, randomized controlled trials, controlled before-and-after studies, and cohort studies with a control group that assessed the impact of social network interventions involving families, friends, and peers on improving dietary adherence and glycemic control among individuals with T2D. The target population consisted of adults aged 18 years and older who had received a T2D diagnosis, as defined by the World Health Organization (WHO) ³⁶ or the American Diabetes Association (ADA).³⁷ The interventions highlighted in this review comprised a social network component, which involves providing advice, arranging support, or offering social assistance to participants through their existing social networks - such as family, friends, or partners, or creating new social connections with other participants or peer mentors. The intervention also included dietary components or a combination of dietary and physical activity elements. One of the primary outcomes measured was changes in diet or adherence to dietary guidelines. A previously published review protocol describes the interventions and conceptual framework, summarized in Figure 2. Comparators for this intervention include no intervention, usual care, or an intervention that lacks an explicit social network component. The primary outcomes of the study are focused on dietary changes, specifically adherence to dietary recommendations or prescribed diet plans. Secondary outcomes include glycemic control, social network measures, social support, physical health indicators, knowledge related to diet and diabetes, diabetic complications, psychological effects (such as quality of life and stress), and metabolic outcomes like blood pressure.

Figure 2.

Literature Search Process and Results

Study Selection

A study eligibility form was used to screen studies for inclusion.³⁵ Two authors independently reviewed the titles and abstracts of the retrieved citations to identify potentially relevant studies. If a decision was reached based on the titles and abstracts, the full articles were then evaluated. Any discrepancies between the reviewers were resolved through consensus, and consultation with a third reviewer. The independent screeners demonstrated high reliability and agreement in their screening processes, achieving a kappa statistic of 0.74 and an agreement rate of 92% for the titles and abstracts and a kappa of 0.76 with a 97% agreement rate for the full texts.³⁸

Data Extraction

Three independent reviewers used a modified data extraction form from the Cochrane Collaboration to gather information from eligible studies.³⁹ The extracted data included details about the studies and their participants, the interventions implemented, and the outcomes measured (Table 1). The interventions were categorized based on their approach to creating social networks, following the 2012 classification system for social network intervention approaches. This system includes: 1) Individual Network Interventions, which use network data to identify individuals who can promote behavior change based on specific properties of the network; 2) Segmentation Network Interventions, which target specific groups of individuals clustered within a network; 3) Induction Network Interventions: these use existing social ties to disseminate information or encourage healthy behaviors, and 4) Alteration Network Interventions: these aim to modify the network structure by adding new members or breaking ties with individuals who promote unhealthy or risky behaviors.⁴⁰ Additionally, we adopted the umbrella term “Social Network Functions” to describe the core components of the intervention approaches.⁴¹ This term encompasses the network intervention strategy employed, the theoretical mechanisms involved, the definition and boundaries of the social network, methods of network recruitment (where applicable), training methods, and any details regarding the structure and characteristics of the social network or changes within it, described using network parameters (Table 2). Where necessary, we contacted the authors of the included studies for additional information or missing data.

Table 1.

Characteristics of included studies

Authors, years (Country)	Study Objectives	Setting (Intervention duration)	Study population (sample size & age)	Intervention	Outcomes measured	Key results
					Diet	Glycemic control
Sorkin et al. 2014 (US) 44	To assess the impact of a dyad-based lifestyle intervention on weight loss and dietary intake.	Two federally qualified health centers (4 months)	89 mother–daughter dyads (intervention =53, controls= 36) aged ≥18 years: mothers with T2D and their overweight/obese (BMI≥25 kg/m2) daughters	The intervention included four group meetings, eight home visits with a lifestyle community coach, and booster telephone calls by a lifestyle community coach. The intervention was designed based on the Diabetes Prevention Program's (DPP) Lifestyle Change Program. On the other hand, the control condition received educational materials sent to the participants' homes.	Diet: glycemic load, fat intake, fruit and vegetable intake Weight	• After adjusting for T1 (baseline) data and other covariates, the intervention group had significantly lower T2 (16-week) glycemic load (p<0.001) and saturated fat intake (p=0.004) than those in the control group. • The intervention was associated with less fruit intake at T2 (p=0.09) while vegetable intake did not differ between both groups at follow-up. • The intervention group had significantly lower T2 weights than the control group (p<.003).
Gray et al. 2021 (US) 45	To examine the effects of an in-home community health worker (CHW)-led intervention for adults with diabetes and incomes <250% of the federal poverty line on self-management behaviors	Three health systems in Washington State (12 months)	287 patients with T2D and hemoglobin A1c HbA1c ≥ 8%, aged 30–70 years.	Participants were randomized to the CHW intervention or usual care control arms. The CHW intervention focused on promoting the adoption of self-management behaviors that improve HbA1c, such as blood glucose monitoring, healthy eating, physical activity, medication taking, and smoking cessation. Participants in the control group received usual care and were offered one optional diabetes self-management educational visit after completing the 12-month outcome assessment.	Healthy eating plan (days) Dietary behaviors Physical activity, HbA1c	• At 12 months, the intervention arm followed diabetes-specific diet more days, but the changes were not significant (p=0.14). However, they significantly followed general diet recommendations more days per week (4.1 vs. 5.4; p=0.03), • At 12 months, intervention arm participants were more likely to skip meals a few times a month or less (RR = 1.24, 95% CI 1.02, 1.50), and were more likely to prepare meals mostly at home (RR = 1.19, 95% CI 1.07, 1.34) compared to controls. • Intervention arm participants also had 141 min of additional weekly physical activity compared to controls at 12 months (95% CI 46.4, 236.4). • Improvements in HbA1c were restricted to participants with HbA1c ≥ 10% at baseline.
McEwen et al. 2017 (US)46	To investigate the effects of a family-based self-management support intervention for adults with T2D	Hispanic urban neighborhoods in the Arizona border region (6 months)	157 dyads (intervention =83 dyads, wait list controls= 74 dyads), participants with T2D and a family member aged 35 to 74 years.	The family-based self-management support program included 3 successive components: six 2-hour educational and social support group sessions conducted weekly for 6 weeks, three 2-hour home visits scheduled weekly for 3 weeks, and three 20-minute telephone calls scheduled weekly for 3 weeks. The waitlist control program provided two-hour educational sessions weekly for 3 weeks.	Healthy eating plan (score) HbA1c Physical activity Diabetes self-care activities. Diabetes self-efficacy	• A significant difference between the 2 groups at baseline, t(155) = 2.04, P = .043, with diet self-management activities greater for the intervention group. The intervention effect was sustained for 6 months (3.04 vs. 4.32; p=0.001). • HbA1c decrease from 9.9% to 8.9% in the intervention group with no significant changes over time between both groups. • An increase in exercise self-management activities from baseline to T2 (3 months) for both groups, with the increase greater for the intervention group. But the intervention effect was not sustained for 6 months. • For total diabetes self-management, the intervention group increased from baseline to T2 while there was little change for the control group. The intervention effect was sustained for 6 months. • Diabetes self-efficacy for health behaviors, F(1.8, 168.6) = 4.50, P = .015; diabetes self-efficacy for general health, F(2, 190 = 3.55), P = .031; total diabetes self-efficacy, F(1.8, 173.7) = 4.98, P = .010 increased in intervention group than control group. The intervention effect was sustained for 6 months post-intervention.
Sreedevi et al. 2017 (India) 47	To assess the feasibility and effects of two low-cost interventions, yoga, and peer support, on glycemic and other outcomes among women with T2D.	Rural health training center (3 months)	124 women with T2D aged 30 to 65 years (yoga: 41; Peer: 42; Control: 41)	All the study participants were given an education on diet and were advised to exercise for at least 10 min per day. In the yoga arm, sessions by an instructor, consisting of a group of postures coordinated with breathing, were conducted for an hour two days a week. In the peer support arm, each peer mentor, after training, visited 13–14 women with T2D every week, followed by a phone call. The meeting was about applying disease management or prevention plans in daily life. The control group was given the usual standard of care, including continuing oral hypoglycemic drugs, advice on diabetic diet, and exercise for at least 10 min a day to a level of 150 minutes/week.	Diet: total calorie consumed HbA1c Fasting plasma glucose (FPG) Blood pressure (BP) BMI Medication adherence	• Dietary adherence in the yoga group increased from 40% to 80%. Total cholesterol levels in the peer group decreased by 5 mg% (95% CI–15,5.1) compared to an increase in yoga and control groups by 6 mg% (95% CI- 7.3–19.5) and 16 mg% (95% CI-2–34.3) respectively which was again not significant. • The differential fall in HbA1c was 0.3%(CI −0.85 – 0.34) in yoga; an increase in peer and control by 0.5%(CI −0.32,1.4) and 0.19%(CI–0.19,0.86) respectively was observed. Thus only yoga group showed a decrease in HbA1c, though not significant. • There was a mean fall of FPG by 6.5 mg/dl in yoga group (CI 34.5 − −44) as compared to 5.9 mg/dl in peer (37.6 – −46.1) and 1.7 mg/dl in control (−31.6 – 28) though this was not significant. • The yoga group has brought a significant decrease of 6.1 mm systolic (p = .08) and 3.1 mm diastolic blood pressure (p = .03) as compared to a marginal decrease in peer and an increase in control group. • BMI showed an increase in all three groups with the yoga group having the smallest increase, though this was not significant. • Morisky’s adherence score showed an increase in all three arms and the highest increase was seen in the control group at 0.7 (CI.09,1.4) though it was not significant.
Vissenberg et al. 2017 (Netherlands) 48	To evaluate a culturally sensitive social network intervention for people with T2D in lower socioeconomic groups, focusing on improving self-management behaviors and outcomes related to diabetes.	Respondent’s home and community center (16 months)	131 socioeconomically deprived patients aged 40 to 75 years with suboptimal glycemic control, with 69 assigned to the intervention group and 62 to the control group.	The intervention consisted of 24 group meetings for participants, 6 group meetings for their significant others, and two social network therapy sessions where both the participant and a significant other participated. The control group received the standard diabetes education for patients with T2D, which consists of information and education about T2D and self-management.	Diabetes-specific diet (days) Fat, fruits and vegetable consumption (days) Medication adherence Physical activity	• At baseline, both groups adhered to their diabetes-specific diet on five days and improved their adherence at 10 and at 16 months. However, the improvement was not significant (Interval. 5.30 vs 6.06 P=0.11; Control 5.26 vs. 5.71; p=0.77). • At baseline, fat consumption was low in both groups, while fruit and vegetable consumption was high. At 16 months, fat consumption had improved in the control group only, and fruit and vegetable consumption had decreased in both groups. • At baseline, medication adherence was high in both groups. At 16 months, the intervention group continued their high adherence pattern, compared to a slight decrease in the control group. However, the change was not significant. • The analyses show a greater and more diverse increase in physical activity in the intervention group compared to control group (p = 0.03).
Trief et al. 2019 (US) 49	To assess the effects of a telephone couples’ behavioral intervention on partners of persons with Type 2 diabetes	Health clinics (12 months)	240 couples (80 per arm) aged ≥ 21 years old in which one partner had T2D in poor glycemic control (HbA1c ≥ 7.5%). Couples were randomized to: couples intervention calls (CC; person with diabetes together with his/her partner), individual intervention calls (IC; person with diabetes alone), and individual diabetes education calls (DE; person with diabetes alone).	All arms received two calls covering comprehensive diabetes education. DE participants had no further intervention. CC and IC arms had 10 additional calls on behaviors (knowledge, self-monitoring, goal-setting and behavioral contracting to promote lifestyle changes in activity, diet, blood glucose monitoring and medication adherence).	Diet: fat intake weight/BMI Blood pressure Physical activity Diabetes distress (DD) Depressive symptoms (DS)	• There were no between-arm differences in percentage of energies from fat at any of the assessments. • No statistically significant differences in weight/BMI among the three arms at any follow-up. • Mean diastolic BP in the CC arm was significantly lower than DE at 4, 8 and 12 months (all P-values < 0.005), and significantly lower than IC at 12 months (P = 0.007). The IC mean was statistically significantly lower than DE at 4 months (P = 0.022). • At 4 and 8 months, there were statistically significantly higher physical activity scores in the CC compared with the IC arm (t = 2.23, P = 0.026 compared with t = 2.46, P = 0.015); this was not maintained at 12 months (t = 0.28, P = 0.768). • At 4, 8 and 12 months, the CC arm showed a significantly lower DD score, and greater reduction in DD, than the IC and DE arms (4 and 8 months, P < 0.001; 12-month, P < 0.05) with no significant differences between the IC and DE arms. • At 12 months, the CC arm showed a significantly lower mean DS score (P = 0.013) than IC and DE, this appears to reflect increases in partner DS in IC and DE arms, which did not differ.
Castillo-Hernandez et al. 2021 (Mexico) 50	To evaluate the effect of peer support, when added to a diabetes education program, on glycemic control and diabetes-related quality of life.	Community health centers (8 months)	58 patients with T2D aged >18 years	Participants were randomly assigned to 1 of 2 groups: 1) a peer support and diabetes self-management education group (PSEG) or 2) a conventional diabetes self-management education-only group.	Healthy eating plan (days) HbA1c BMI Blood pressure Physical Activity Quality of life	• The number of days per week after a healthy eating plan (≥2 days per week in the PSEG increased vs no change in the EG [p=0.005], • A1C level change was greater in the PSEG but not statistically different from the EG at 4 and 8 months of follow-up (4 months: PSEG 0.83% vs EG 0.44% [p=0.2]; 8 months: PSEG 1.29% vs EG 0.98% [p=0.3]) • BMI decreased in both groups but failed to reach a statistically significant difference. • Systolic and diastolic blood pressure (SBP and DBP, respectively) levels decreased at 4 months in the PSEG but increased in the EG. At 8 months, blood pressure levels continued to decrease in the PSEG; however, the difference between groups was no longer statistically different. • The PSEG improved their physical activity beyond the changes observed in EG, particularly at 4 months, where this difference was statistically different. • A clear and statistically significant difference between groups was observed in 2 areas at 8 months, namely Energy and mobility and diabetes control, favoring the PSEG.
Rosland et al. 2022 (US) 51	To determine if the Caring Others Increasing Engagement in Patient Aligned Care Teams (CO-IMPACT) intervention improves patient activation, diabetes management, and outcomes compared with standard care.	2 Veterans Health Administration primary care sites (12 months)	239 patient-supporter dyads. (intervention =116 dyads, wait list controls= 113 dyads). All patient participants were T2D adults aged 30 to 70 years with HbA1c concentrations greater than 8%.	Patient-supporter pairs underwent a health coaching session that emphasized dyadic information sharing and positive support techniques. Following that, they received biweekly automated monitoring telephone calls for 12 months, which reminded them of the actions they needed to take to achieve their diabetes goals. They were also provided with coaching calls aimed at helping them prepare for primary care visits and after-visit summaries. On the other hand, standard-care pairs were only given general diabetes education materials.	Healthy eating plan (days) Patient Activation measure (PAM) Diabetes self-efficacy HbA1c	• Patients in the CO-IMPACT arm had greater 12-month improvements in healthy eating (intervention effect, 0.71 d/wk; 95%CI, 0.20–1.22 d/wk; P = .007), • CO-IMPACT patients had greater 12-month improvements in PAM-13 scores (intervention effect, 2.60 points; 95%CI, 0.02–5.18 points; P = .048). • Patients in the CO-IMPACT arm also had greater 12-month improvements in self-efficacy in diabetes (intervention effect, 0.40 points; 95%CI, 0.09–0.71 points; P = .01) • However, the 2 arms had similar improvements in HbA1c concentrations (p=0.33).
Sukchaisong et al. 2022 (Thailand) 52]	To investigate the effectiveness of the Mindfulness-Based Diabetes Self- and Family Management Support Program (MDSFM) in improving HbA1c	Two health care centers (4 months)	80 adults with uncontrolled T2D (experimental group=40; usual care=40).	Participants in the experimental group received instructional booklets with behavioral logbooks and participated in the six-session MDSFM program in a separate room at the clinics, in addition to receiving usual care, whereas the control group received only usual care at the clinics’ outpatient section.	Healthy eating plan (scores) HbA1c Physical activity Medication adherence Perceived self-efficacy	• The experimental group had a significant improvement of dietary consumption than those in the control group (p < .001) with no improvement in the control group. • 92.5% of participants in the experimental group had decreased HbA1c and 52.5% of those reduced at least .5%, whereas 77.5% of those in the control group increased HbA1c. • The experimental group had a significant improvement in physical activity than those in the control group (p < .001) with no improvement in the control group. • The experimental group had a significant improvement of medication adherence than those in the control group (p < .001) with no improvement in the control group. • The experimental group had a significant improvement in perceived self-efficacy than those in the control group (p < .001) with no improvement in the control group.
Feng et al. 2023 (China) 53	To assess the effectiveness of an eHealth family-based health education intervention in patients with uncontrolled type 2 Diabetes via WeChat	Community health service centers (12 months)	220 patients (110 per group) with T2D aged 30–79 years and had a family member who could use WeChat and lived with the patient or visited them at least once a week.	The intervention group received 38 intervention articles through WeChat on knowledge of diabetes, complications, risk, and self-care activities (eg, diet, exercise, medication, lifestyle, glucose testing, and skills). In-person intervention with people with T2D 1 every 3 months and online intervention with family members on reminding patients to improve their self-care activities. The control group received usual care.	Diet: general diet and specific diet scores HbA1c Exercise, foot care, and risk perception (risk knowledge, personal control, worry, optimism bias, and personal risk)	In the intervention group, the scores for general diet (5.36 vs. 6.39; P<.001) and specific diet (3.59 vs. 4.31; P<.001) improved. In the control group, besides general diet (P=.002), there were no significant differences between baseline and follow-up. The intervention group had significantly lower HbA1c values (β=–.69, 95% CI –.99 to –.39; P<.001). In the intervention group, exercise (P=.002), foot care (P<.001), risk knowledge (P<.001), personal control (P<.001), worry (P=.02), optimism bias (P=.03), and supportive behaviors (P<.001) improved. In the control group, there were no statistically significant differences between baseline and follow-up.

Table 2.

Social network functions in included studies.

Authors, years (Country)	Social network definition	Network intervention strategy	Recruitment strategies	Training of Peer Educators/Leaders (where applicable)	Social Network Measures and Relevant Characteristics (where applicable)
Individual network interventions
Sreedevi et al. 2017 47	Three peer mentors identified from the community and trained.	1. Individuals (Low-threshold nodes) The investigation team selected peer leaders who adhered to the treatment and behavior change regime (early adopters) and had good communication and leadership skills. 2.Induction (Network outreach) Trained mentors provided support to the study participants in a ratio of 1:14. They visited participants once a week. They offered guidance on disease management and prevention, including advice on diet, exercise, medication, emotional stress, symptoms, and foot care. They also followed up with phone calls for further support. Theoretical Framework: Social influence and empowerment framework	The criteria for eligibility were having T2D for at least one year with an RPG ≤250 mg/dl in the last reading, judged by the investigation team to be generally adherent to treatment and behavior change regime, capacity and commitment to undergo the training required, an understanding of patients’ confidentiality, undertaking to liaise with the concerned doctor if unanticipated problems arose during the course of their peer support activity.	Number of training sessions: two-day training Duration of training sessions: not detailed. Characteristics of the trainer(s): facilitated by a physician and a nutrition specialist. Training elements: The physician explained the etiology of diabetes, changes in the body due to diabetes, complications due to poor glycemic control, and an outline of the drugs used, their mechanism of action, and the synergies with physical activity. The nutrition specialist explained all the nutritional and dietary aspects of diabetes; the psychologist trained the peer mentors on communication skills, empathy, and confidentiality. A training manual was also prepared for peer mentors based on the peer progress handbook for future reference.	In the peer support group, for the first two months the three peer supporters made regular visits and had a one-to-one discussion with the patient. On the visits, peer mentors could not regularly contact about 15% (5/32) of women. Of these 20% (1/5) of the women could not be directly contacted over the phone too. In the third month one of the peer mentors dropped out citing personal circumstances. Half of the women with T2D in the peer group were not enthusiastic about the idea of peer support at the beginning, though towards the end of the program 94% reported that it was useful.
Castillo-Hernandez et al. 2021 50	Peer leaders (supporters) known to have diabetes and an A1C of <8% in their most recent evaluation.	1. Individuals (Low-threshold nodes) The investigation team selected peer leaders who adhered to the treatment and behavior change regime and had good communication and leadership skills and asked to score at a high level in the diabetes Knowledge. 2.Induction (Network outreach) Leaders were trained to start each meeting with an “icebreaker” introduction, followed by a discussion session and goal setting. For the first four months, the discussion segment was based on the DSME theme of the previous education session, and for the remaining sessions, each PL identified the theme according to the groups' needs and interests. Theoretical Framework: Behavioral modeling	The investigation team identified PLs and patients. According to the health-care provider at their clinic, PLs were required to have good communication skills. After training on basic aspects of diabetes, communication, and leadership skills, according to the Peer Leader Manual, those who scored at a high level in the Diabetes Knowledge Questionnaire were selected. Their interest, level of involvement, and performance in the training process were also considered.	Number of training sessions: two sessions every week for 3 months. Duration of training sessions: 1.5-hours each. Characteristics of the trainer(s): facilitated by a dietitian and diabetes educator. Training elements: basic aspects of diabetes, communication and leadership skills, according to the Peer Leader Manual; government-sponsored DSME program named “A 7 Pasos del Control” (i.e. “Seven Steps to Achieve Control”).	The attrition rate for the study visits at 8 months was 4% (1 participant) in the PSEG group and 14% (3 participants) in the EG group. From the 20 peer-support meetings, participants in the intervention group attended an average of 8 support meetings (median, 9; range, 0 to 15).
Segmentation network intervention
Sorkin et al. 2014 44	Overweight/obese daughters that reside within 25 miles of their mother’s residence (co-residence allowed).	1. Segmentation (cohesive groups): Mother-daughter dyads received a 16-week intervention, consisting of: (a) four group meetings, (b) eight home visits with a lifestyle community coach, and (c) four booster telephone calls by a lifestyle community coach between home visits. 2.Induction (recruitment and interaction): Each participant was encouraged to recruit their overweight/obese daughters and to develop shared strategies for improving adherence to their diet and exercise plans, overcoming barriers, and providing support to address potential relapses. Theoretical Framework: Diffusion of innovation theory	Mexican American women with T2D were recruited from two federally qualified health centers and asked to recruit their adult and overweight/obese daughters interested in losing weight.	Not applicable	The intervention group reported significantly higher health-related social support and persuasion, as well as a decrease in undermining at T2 (16-week) compared to the control group (p < 0.001).
Gray et al. 2021 45	Family members and other social network members	1.Segmentation (cohesive groups) CHWs met with participants and their networks in their home to promote the adoption of self-management behaviors that improve HbA1c, such as blood glucose monitoring, healthy eating, physical activity, medication taking, and smoking cessation. 2.Induction (interaction) CHWs encouraged family members and other social network members to help them by supporting lifestyle changes and medication taking, attending clinic visits, and providing emotional support. Theoretical Framework: Social Cognitive Theory and Self-Regulation Theory	Participants were from three health systems.	Not applicable	Diabetes-related social support from significant others, family, friends was measured. However, no evidence was found that changes in social support mediated intervention-control differences in eating behaviors and physical activity.
McEwen et al. 2017 46	Adult family members who either lived in the same household as the participant or visited them weekly to share meals or do grocery.	1. Segmentation (cohesive groups): Patients and family members received a 12-week intervention program that included: (1) six 2-hour educational and social support group sessions, (2) three 2-hour home visits, and (3) three 20-minute telephone calls. 2.Induction (recruitment and interaction): Each participant with T2D was asked to identify a family member and promotora conducted the social support sessions, home visits, and telephone calls. Theoretical Framework: Diffusion of innovation theory	Potential participants were recruited by bilingual/bicultural promotoras. In addition, the Family Action Board (FAB) helped to identify, reach out, and motivate potential participants to participate. Inclusion criteria was having 1 adult family member who either lived in the same house as participant with T2D or saw them weekly to share meals or visit or shopping and were willing to participate.	Not applicable	Physician distress, F(2, 190) = 3.42, P = .035; regimen distress, F(2, 190) = 9.75, P < .001; interpersonal distress, F(1.9, 177.0) = 4.12, P = .020; and total diabetes distress, F(1.8, 172.7) = 9.07, P < .001. decreased in intervention group than control group. The intervention effect was sustained for 6 months.
Vissenberg et al. 2017 48	Patients’ significant others (partners, children, and friends)	1. Segmentation (cohesive groups): Participants and significant others were randomized together and attended two social network sessions together guided by group leaders. 2.Induction (recruitment and interaction): Each participant was encouraged to recruit a significant other; learned and practiced all the new behaviors and coping strategies together to stimulate social support, diminished hindering social influences and increased the ability of participants to deal with peer pressure and temptations that hindered self-management behaviors. Theoretical Framework: Guided by behavioral economics and social learning theory within a social networks’ framework.	Patients were selected from the records of 39 general practices in socioeconomically deprived neighborhoods in four Dutch cities.	Not applicable	Actively involving (significant) others in self-management seems a successful intervention strategy. Making an action plan with the help of others and practicing strategies together probably helped participants to formulate realistic goals and strategies, to receive more social support for self-management and to feel more confident about implementing their action plan.
Trief et al. 2019 49	Partners of patients with T2D and poor glycemic control (HbA1c ≥ 7.5%).	1. Segmentation (cohesive groups): The person with diabetes and his/her partner participated together. Their received 12 calls: two covering comprehensive diabetes education, and 10 calls on behaviors (knowledge, self-monitoring, goal-setting and behavioral contracting to promote lifestyle changes in activity, diet, blood glucose monitoring and medication adherence). 2.Induction (interaction): Education component emphasized team building and information sharing within the couple. Theoretical Framework Interdependence theory; social learning theory	Couples were recruited in upstate New York and northern California, enrolled by staff through chart review/recruitment letters, advertisements and community talks.	Not applicable	The CC arm showed a significantly higher and greater increase in relationship satisfaction (RS) scores than the DE (t = 2.86, P = 0.005), and the IC (t = 1.84, P = 0.067) arms at 4 months. At 8 months, the CC arm showed a statistically significantly higher RS score than the IC (t = 2.03, P = 0.043) and DE arms (t = 2.64, P = 0.009). At 12 months, there were no differences among the arms. Almost 40% of IC and DE participants reported that their partners were involved ‘a lot’ or ‘all of the time’ (e.g. listened to calls, read the materials).
Rosland et al. 2022 51	Patient-supporters are willing to participate to the program.	1.Segmentation (cohesive groups) Patient-supporter pairs underwent a health coaching session that emphasized dyadic information sharing and positive support techniques. 2.Induction (recruitment and interaction): Supporters were nominated by the patient and were advised to talk to the patient about their health at least twice per month and help the patient regularly with one or more aspects of their health care (e.g., filling medications, conducting home testing), but they did not receive pay to care for the patient. Theoretical Framework not detailed.	Potentially eligible patients were identified via administrative databases.	Not applicable	Patients in the CO-IMPACT arm had greater 12-month improvements in satisfaction with health system support for the family supporter participants’ involvement (intervention effect, 0.28 points; 95%CI, 0.07–0.49 points; P = .009).
Sukchaisong et al. 2022 52	Family member who lived with adults with T2D were primarily responsible for taking them to the doctor.	1.Segmentation (cohesive groups) Patients and family members received mindfulness-based diabetes self- and family management support program. 2.Induction (recruitment and interaction) Family members were encouraged to remind participants daily to take their prescribed medication. During the first and fifth sessions, family members were invited to learn about the disease and desired health behaviors and discuss potential behavior changes with the participants. Theoretical Framework: Information-Motivation-Behavioral Skills Model and Mindfulness approach	Participants were recruited in two health care centers purposively selected from a pool of 68 health care centers in Bangkok, Thailand	Not applicable	Not detailed
Feng et al. 2023 53	Adult family member as a supporter	1.Segmentation (cohesive groups) Patients and their family members received eHealth family intervention based on community management via WeChat. 2.Induction (recruitment and interaction) Patients were required to choose 1 family member who could use WeChat and lived with the patient or visited them at least once a week. In addition to receive 38 comprehensive articles on topics including basic knowledge, skills, and risk knowledge, family members were encouraged to remind patients to improve their self-care activities. Theoretical Framework Health Belief Model	Registered patients were recruited from 2 community health service centers in the urban area in Jiading District.	Not applicable	In the intervention group, non-supportive behavior of family members (P=.03) decreased. In the control group, there were no significant differences between baseline and follow-up.

Quality Assessment

Two authors assessed the quality of the included studies using the revised tool ‘Risk of Bias in Randomized Trials’ (RoB 2)^42]and the risk of bias in non-randomized studies of interventions (ROBINS-I).⁴³ The RoB 2 assessment for individual RCTs consists of six domains: (1) Random sequence generation (selection bias), (2) Allocation concealment (selection bias), (3) Blinding of participants and researchers, (4) Blinding of outcome assessment (detection bias), (5) Incomplete outcome data (attrition bias), and (6) Selective reporting (reporting bias). For each domain, the risk of bias can be categorized as (1) Low risk of bias, (2) Unclear, or (3) High risk of bias. On the other hand, the ROBINS-I tool assesses the following biases: (1) Bias due to confounding; (2) Bias in the selection of participants for the study; (3) Bias in the classification of interventions; (4) Bias due to deviations from intended interventions; (5) Bias due to missing data; and (6) Bias in the selection of the reported results. For these judgments, the options are (1) Low risk, (2) Moderate risk, (3) Serious risk, and (4) Critical risk of bias. An overall bias risk judgment was also made according to the guidelines (Table 3). Any disagreement between the two reviewers was resolved through discussion and consensus; if necessary, a third reviewer was involved in the arbitration process.

Table 3.

Risk of bias assessment for RCTs^a and quasi-experiments^b

RCTs	Risk of bias (High, low, unclear)
References	Random sequence generation	Allocation concealment	blindness of participants and researchers	blindness of outcome assessment	Incomplete outcome data	Selective reporting	Overall Risk of Bias
Sorkin et al, 2017 44	Low risk	Low risk	High risk	unclear risk	high risk	Low risk	High risk of bias
Gray et al, 202145	Low risk	Low risk	High risk	Low risk	low risk	Low risk	High risk of bias
McEwen et al, 2017 46	Low risk	Unclear risk	High risk	unclear risk	high risk	Low risk	High risk of bias
Sreedevi et al, 2021 47	Low risk	Low risk	Unclear risk	unclear risk	high risk	Low risk	High risk of bias
Trief et al, 2020 49	Low risk	Low risk	lower risk	Low risk	low risk	Low risk	Low risk of bias
Castillo-Hernandez et al, 2021 50	Low risk	Low risk	low risk	Low risk	Unclear risk	Low risk	Unclear risk of bias
Rosland et al, 2022 51	Low risk	Low risk	lower risk	Low risk	low risk	Low risk	Low risk of bias
Sukchaisong et al, 2022 52	Low risk	Unclear risk	Unclear risk	unclear risk	Unclear risk	Low risk	High risk of bias
Feng et al, 2023 53	Low risk	Low risk	Unclear risk	unclear risk	low risk	Low risk	Unclear risk of bias
Non-RCT	Confounding
Vissenberg et al. 2017 48	Moderate risk	low risk	low risk	low risk	low risk	Serious risk	Serious risk

^aRisk of bias assessment using RoB2.

^bRisk of bias assessment of risk of bias using ROBINS-I

Data Analysis

The studies included in this analysis varied significantly with regard to study design and analytical methodology, making it inappropriate to conduct a meta-analysis. As a result, we conducted a narrative synthesis of the ten eligible studies by summarizing, comparing, and contrasting the extracted data. We also chose not to provide a GRADE evidence summary for the review because some studies contained multiple intervention components, which could lead to a misleading summary of the evidence. For instance, one study used a dyad-based lifestyle intervention to enhance weight loss and improve dietary intake.⁴⁴ In contrast, another study focused on community health worker-led diabetes self-management and social support.⁴⁵

RESULTS

The search yielded 2,488 titles. After screening the abstracts, we identified 44 potentially eligible studies. Following a detailed review of the full texts, we excluded 34 studies that did not meet the inclusion criteria. This left us with ten studies included in this review, as illustrated in Figure 2.^44–53 Table 1 summarizes the characteristics of these included studies, Table 2 presents the social network functions, and Table 3 shows the risk of bias assessment for all included studies.

Sample Characteristics

As presented in Table 1, the ten studies were all published in English between 2014 and 2023. Most (n=6) were conducted in high-income countries, primarily in the United States ^{44, 45, 46, 49, 51} and the Netherlands.⁴⁸ However, four studies took place in low- and middle-income countries (LMICs), specifically in India,⁴⁷, Mexico,⁵⁰ Thailand,⁵² and China.⁵³ Among the ten studies, three were conducted in community settings.^{46, 48, 53} The sample characteristics showed considerable clinical heterogeneity, with participant sizes ranging from 58 to 287, ages ranging from 18 to 75 years, and follow-up durations varying from 3 to 16 months.

Intervention Characteristics

A majority of the included studies examined family networks that comprised partners,⁴⁹ children,⁴⁴ or other family members.^{45, 46, 48, 52, 53} Some studies focused on peer support,^{47, 50} while others looked at significant others, such as social network members, friends, or supporters.^{45, 48, 51} Many of the studies also engaged minority populations, such as ethnic minorities (n=3), women (n=2), and individuals from socio-disadvantaged communities (n=8). Additionally, most studies concentrated on one network intervention approach: two employed an individual network intervention approach, while eight used segmentation approaches (Table 2). These interventions integrated behavior change theories with theories regarding network influence, such as social cognitive theory, the health belief model, social influence, social learning theory, interdependence theory, and self-regulation theory. Notably, innovation diffusion theory was the most frequently cited theoretical framework regarding network elements.

Effects of Intervention

Primary Outcome: Dietary adherence

Seven of the ten included studies provided data on a diabetic-specific diet or healthy eating plan.^{45, 46, 48, 50–53} Two studies focused on fat intake,^{44, 49} fruits and vegetables,^{44, 48} and dietary behaviors.^{45, 53} Additionally, one study reported glycemic load,⁴⁴ and another focused on energy intake.⁴⁷

Five of the seven studies focused on diabetic-specific diets or healthy eating plans and reported improvements in dietary adherence. However, the criteria for evaluating adherence varied among these studies, and adherence was measured differently. For instance, McEwen et al. (US),⁴⁶ Sukchaisong et al. (Thailand),⁵² and Feng et al. (China)⁵³ found that a family-based self-management support program significantly increased dietary adherence scores by 1.28, 8.7, and 0.72, respectively. Similarly, Castillo-Hernandez et al. (Mexico)⁵⁰ and Rosland et al. (US)⁵¹ demonstrated that a peer or supporter-based program increased adherence to a healthy eating plan by 1 to 2 days. In contrast, the studies by Gray et al. (US)⁴⁵ and Vissenberg et al. (Netherlands),⁴⁸ which also provided family-based support, showed that the intervention group followed a diabetes-specific diet for more days. However, these changes were not significant at the 12- and 16-month marks.

In three studies on dietary intake, two examined fat intake ^{44, 49} and one focused on energy intake.⁴⁷ Sorkin et al. ⁴⁴ found that mother-daughter pairs receiving in-person lifestyle interventions had a significantly lower saturated fat intake by week 16 compared to the control group (p-value=0.004). Conversely, Trief et al.⁴⁹ reported no significant change in fat intake from a telephone behavioral intervention for couples. Additionally, Sreedevi et al.⁴⁷ noted that while participants appreciated peer mentoring, total energy consumption remained unchanged during the intervention.

Regarding glycemic load, fruits, and vegetables, Sorkin et al.⁴⁴ found that a dyad-based lifestyle intervention significantly reduced the glycemic load at the 16-week mark (p-value <0.001) in the intervention group compared to the control group. However, this intervention was linked to a decrease in fruit intake at T2 (p-value=0.09), while there was no significant difference in vegetable intake between the two groups at follow-up. Similarly, in the study by Vissenberg et al.,⁴⁸ which implemented a culturally sensitive social network intervention, fruit and vegetable consumption declined in the intervention group.

Two studies have examined the impact of social network interventions on dietary behaviors. In the first study conducted by Gray et al.,⁴⁵ it was found that participants in the in-home community health workers (CHWs)-led intervention and social support group had more general diet recommendations each week than those in the control group. Specifically, the intervention group had an average of 5.4 diet recommendations per week, compared to 4.1 recommendations in the control group. They also skipped meals less frequently and primarily cooked at home. In a second study by Feng et al.,⁵³ participants in the intervention group followed general diet recommendations for an average of 6.39 days per week, while the control group averaged 5.36 days after a 12-month family-based health education.

Secondary outcomes

Glycemic control

Seven studies reported data on HbA1c concentrations.^{45–47, 50–53} Among these studies, one specifically focused on fasting glucose concentrations.⁴⁷ Two of the seven studies showed a significant improvement in HbA1c concentrations after six months.^{52, 53} Sukchaisong et al.⁵² observed a notable decrease in HbA1c concentrations, with a reduction of 0.6% (p-value<0.001) following a four-month mindfulness-based diabetes self-management and family support program. Similarly, Feng et al.⁵³ reported a comparable result after a 12-month eHealth family-based health education intervention in China. In contrast, the study by Sreedevi et al.⁴⁷ found that the peer mentor group experienced an increase in HbA1c concentrations by 0.5% (CI −0.32, 1.4). Additionally, studies conducted by McEwen et al.,⁴⁶ Castillo-Hernandez et al.,⁵⁰ and Rosland et al.⁵¹ reported increases in HbA1c concentrations ranging from 1% to 1.6% following family, peer, and supporter-based interventions, respectively. However, these changes were not statistically significant. Gray et al.⁴⁵ indicated that improvement in HbA1c concentrations was limited to participants with baseline HbA1c concentrations of 10% or higher, where a community health worker-led intervention was provided.

Weight/BMI

Four studies examined the effects of interventions on weight/BMI.^{44, 47, 49, 50} The study by Sorkin et al.⁴⁴ found that participants who received a mother-daughter dyad-based lifestyle intervention lost significantly more weight at four months (p-value=0.03) than those who did not. Similarly, Trief et al.⁴⁹ in the US and Castillo-Hernandez et al.⁵⁰ in Mexico reported decreased weight/BMI, but their results did not reach statistical significance. However, Sreedevi et al.⁴⁷ found an increase in BMI following a peer support intervention in India.

Blood pressure

Three studies included in this review focused on blood pressure.^{47, 49, 50} The results indicated that a telephone-based couples’ behavioral intervention in the United States ⁴⁹ and a program that combined peer leader support with diabetes education in Mexico ⁵⁰ positively impacted blood pressure. The average systolic blood pressure in the intervention groups decreased by 3.89 mmHg and 12.4 mmHg, respectively. Similarly, these groups’ average diastolic blood pressure decreased by 3.12 mmHg and 4.1 mmHg. However, Sreedevi et al. found no significant effect of peer mentoring on blood pressure.⁴⁷

Physical activity

Seven studies collected data on physical activity.^{45, 46, 48–50, 52, 53} Castillo-Hernandez et al. ^[50] reported a significant increase of 869 METS/min per week in physical activity after eight months. Gray et al.⁴⁵ found that a CHW-led intervention, which included social support, resulted in an additional 141 minutes of physical activity per week after 12 months. Sukchaisong et al.⁵² noted increased physical activity from 12.1% before the intervention to 18.6% after 16 weeks. Feng et al. ⁵³ observed a 23.6% increase in physical activity 12 months following the intervention. Furthermore, McEwen et al. ⁴⁶ found a significant improvement in exercise self-management after implementing family-based self-management support (F(2, 188) = 3.77, p-value = 0.025). However, Vissenberg et al. ⁴⁸ and Trief et al. ⁴⁹ indicated that although physical activity increased significantly, these changes were not sustained after 12 or 16 months.

Medication adherence

Two included studies reported medication adherence.^{48, 52} The study by Sukchaisong et al^{. 52} revealed that mindfulness-based family diabetes support significantly improved medication adherence (p<001) in the intervention health center compared to the control group, which did not show any improvement. On the other hand, Vissenberg et al. ⁴⁸ reported that the intervention group had a continued high adherence pattern at 16 months, compared to a slight decrease in the control group. However, the change was not significant after diabetes self-management education and social support, which focused on reducing social influences that hinder self-management.⁴⁸

Diabetes self-efficacy

Three studies ^{46, 51, 52} included in this review reported outcomes related to diabetes self-efficacy. In the US, a family-based self-management support intervention led to an increase in diabetes self-efficacy for health behaviors (F(1.8, 168.6) = 4.50, p-value =0.015), diabetes self-efficacy for general health (F(2, 190) = 3.55, p-value =0.031), and total diabetes self-efficacy (F(1.8, 173.7) = 4.98, p-value =0.01 in the intervention group when compared to the control group ^[46]. Similarly, Rosland et al. ⁵¹ found that patients in the patient-supporter dyads group showed significant improvements in diabetes self-efficacy over 12 months (intervention effect, 0.40 points; 95%CI, 0.09–0.71 points; p-value =0.01). In Thailand, Sukchaisong et al. ⁵² discovered that a mindfulness-based diabetes self- and family management support program had a positive effect on diabetes self-efficacy in the intervention group compared to the controls (p-value < 0.001).

Diabetes distress/Quality of life

Two studies reported diabetes distress and quality of life.^49,50 The first study by Trief et al. found that couples who participated in a telephone behavioral intervention experienced an average decrease in diabetes-related stress levels by 0.52 points.⁴⁹ The second study by Castillo-Hernandez et al. showed that a combination of peer leader support and a diabetes education program significantly improved diabetes-related quality of life, with a reduction of 27.6%.⁵⁰

Quality Assessment

Overall, among the 10 included studies, six had a high or serious risk of bias, two had an unclear risk of bias, and 2 had a low risk of bias, leading to an overall low risk of bias across the included studies. The most common problem areas were the blindness of participants and researchers, the blindness of outcome assessment, and missing outcome data. Details of the risk of bias assessment for all included studies are available in Table 3.

DISCUSSION

Our review indicates that social network interventions can significantly improve adherence to diabetes-specific diets and overall health outcomes. Of the ten studies examined, five showed better adherence to diets, and six reported reductions in HbA1c concentrations. Additionally, physical activity improved between 18.6% and 23.6% in six studies, while three studies noted decreases in weight or BMI. Two studies found reductions in blood pressure, and one study showed a decrease in diabetes-related stress, along with a 27.6% improvement in quality of life. Family-based interventions, particularly those using a segmentation approach, were notably effective in enhancing dietary adherence. These findings highlight the advantages of social network interventions over individual-level strategies.^{23, 54, 55} However, their effectiveness varies by type, quality, and scope, with a critical mass of network sessions necessary for maximum impact.

In particular, the reviewed studies explored various interventions to improve dietary habits in individuals with T2D, including lifestyle programs, peer and family support models, and digital tools. Lifestyle programs by Sorkin et al.⁴⁴ and McEwen et al.⁴⁶ significantly reduced glycemic load and enhanced dietary self-management through group sessions, personalized support, and home visits. Peer and family support initiatives, such as those by Sreedevi et al.⁴⁷ and Castillo-Hernandez et al.⁵⁰, increased dietary adherence and physical activity by promoting accountability and shared learning. Digital interventions, like Feng et al.’s eHealth model ⁵³, effectively improved diet scores and HbA1c concentrations through continuous accessible education. While many of these interventions yielded positive dietary changes, some reported mixed results, highlighting the need for ongoing engagement and tailored approaches to achieve optimal outcomes.

The reviewed studies also implemented social network interventions to improve diabetes management using individual and segmentation strategies. Individual-centered interventions, such as those by Sreedevi et al. ⁴⁷ and Castillo-Hernandez et al. ⁵⁰, used trained peer mentors to foster communication and provide tailored support for lifestyle changes. In contrast, segmentation strategies, seen in studies by Sorkin et al. ⁴⁴ and McEwen et al.,⁴⁶ focused on cohesive groups, often involving family members through group education, home visits, and collaborative goal-setting. These approaches were grounded in behavioral theories and the diffusion of innovation. Additionally, Gray et al. ⁴⁵ and Feng et al. ⁵³ incorporated frameworks like the Health Belief Model and Social Cognitive Theory to enhance family support in self-care initiatives. Across these studies, while social network engagement improved self-management and created supportive environments, challenges such as participant attrition and inconsistent engagement levels underscored the difficulties in maintaining consistent participation.

Encouragingly, the studied social network interventions have effectively engaged hard-to-reach populations, including at-risk individuals, low-income groups, and minorities. These interventions demonstrate high participation and retention rates, crucial for enhancing health through social support and learning.⁵⁶ Research supports the social contagion theory, suggesting that health behaviors and beliefs can spread through social networks.^{57, 58} By incorporating network theories and social-ecological models, we can better understand how social factors influence individual beliefs, decisions, behaviors, and overall health. Additionally, these interventions have successfully promoted healthy behaviors and improved diabetes self-efficacy throughout 12 to 16 months. While traditional health interventions often fail to create lasting changes, this review highlights the importance of social networks in sustaining those changes. This is supported by habit formation theories and behavior change maintenance.^59,60.

Limitations and Implications for practice and future research

The strength of the evidence in this review is limited due to the lack of high-quality studies. Of the ten studies analyzed, only two RCTs had a low risk of bias, while many others demonstrated a high risk of bias and encountered methodological issues, such as detection and attrition bias. Furthermore, several included trials had small sample sizes or were quasi-experimental. Implementing stronger exclusion criteria or conducting sensitivity analyses could strengthen the reliability of these studies.^60,61 Additionally, most interventions were conducted in urban areas of low- and middle-income countries, yet rural communities often struggle with lower dietary adherence and social network challenges.⁶² This observation highlights the need to involve rural populations in social network interventions to improve health outcomes. Lastly, the absence of studies conducted in Africa restricts the applicability of the findings, indicating that social network interventions lack generalizability and context relevance, which suggests a need for innovative evaluation approaches.⁶³

Many social network interventions have demonstrated potential in improving dietary and health-related behaviors; however, fruit and vegetable consumption and energy intake results remain inconsistent.^44,47–49 Different factors, such as family structure, socioeconomic status, and baseline measurements, may influence these discrepancies. For example, Vissenberg et al. observed high fruit and vegetable intake among patients from underprivileged backgrounds,⁴⁸ whereas earlier findings by Everson-Hock et al. ⁶⁴ indicated mixed outcomes for lower socioeconomic groups. Sorkin et al. pointed out that participants with initially high consumption levels decreased their intake after 16 months.⁴⁴ Moreover, Sreedevi et al. highlighted challenges in meeting energy targets due to participants preparing meals for their families.⁴⁷ These results underline the intricacies of enhancing dietary practices and stress the importance of a more thorough understanding of the factors involved.

This review emphasizes the importance of having standardized methods for measuring dietary adherence to enhance comparability across different studies.⁶⁵ Dietary adherence refers to following a diet plan through self-monitoring, maintenance, and preventing relapses. Key factors that support adherence include motivation, understanding dietary recommendations, developing appropriate health beliefs and self-efficacy, setting achievable goals, and receiving social support. Maintaining adequate dietary adherence can lead to improved clinical outcomes for T2D and a better quality of life. However, many people with T2D struggle with adherence due to challenges in understanding and applying these factors.³ Thus, healthcare providers must ensure that patients grasp the concept of dietary adherence and can implement it in their daily lives. Moreover, further research is needed to explore dietary adherence and its components to evolve more effective measures to be communicated to people with T2D.

Additionally, while the studies reviewed suggest that social network intervention can enhance behaviors, assessing their cost-effectiveness and sustainability is crucial before large-scale implementation. Unfortunately, this review did not examine the operational costs, and none of the studies reported them, which are significant in low-resource settings. Nevertheless, these interventions could be cost-effective by using “free” human capital and enhancing the reach and sustainability of health behavior programs.⁶⁶ Some experts propose that these interventions may be particularly valuable in resource-limited areas, as they can bolster community engagement.⁶⁷ However, it is vital to address implementation challenges to develop efficient programs that inform public health policy. Targeting influential individuals within the network can maximize impact and reduce costs while leveraging existing community resources can improve efficiency and sustainability. Therefore, future social network intervention reviews should address cost and feasibility.

Furthermore, although some positive behaviors and health outcomes have been observed, the diversity of social network strategies complicates the identification of the most effective methods. Most interventions were not explicitly designed to test the causal effects of social network mechanisms or to distinguish between various approaches. However, network science can help consolidate evidence for these interventions, as noted by Tanner-Smith and Grant.⁶⁸ One promising approach is a multiphase optimization strategy (MOST), which assesses the unique effects of various intervention components, such as social network strategies and educational materials. This method can help identify the most effective and cost-efficient ways to evaluate network interventions. However, it will be essential to consider patients’ preferences, lifestyles, values, and the communication skills of health professionals, as these factors can significantly impact the implementation of these interventions.⁷⁰

CONCLUSION

Our findings underscore the critical role social networks may play in promoting healthy behaviors such as dietary adherence, glycemic control, physical activity, medication adherence, diabetes self-efficacy, and stress management. However, the strength of this evidence is constrained by the limited number of high-quality studies available and considerable variability in intervention components. Furthermore, existing interventions often fail to fully leverage social networks to optimize the diffusion and adoption of health-related behaviors. Current metrics for dietary adherence also lack alignment with the principles of sustainable healthy diets, emphasizing the need for more comprehensive and inclusive measures.

Notably, the absence of interventions in Africa—a region anticipated to face the most significant rise in type 2 diabetes burden—represents a critical gap in research and practice. Despite this, our review demonstrates the potential of social network interventions in diverse populations, including low-income groups, with both short- and long-term benefits. To address these gaps, future research and health behavior interventions must actively incorporate the social networks in which individuals are embedded. This approach is essential for developing culturally relevant, scalable, and sustainable strategies to mitigate the global burden of type 2 diabetes.

Supplementary Material

Supplementary

Supplement 1. Prisma 2020 Checklist