Xuxi Zhang

According to the subgroup analysis of intervention duration, interventions ≤ 3 months has no significant positive effect on glycemic control. Both interventions >3& ≤ 6 months and interventions  ≥ 12 months have significant positive effect, and the effect of interventions >3& ≤ 6 months is greater than that of interventions  ≥ 12 months. The better effect of interventions lasting >3 months confirms the finding of a previous review 36 of self ‐ management interventions in diabetes that identified duration of intervention as major determinant of impact on HbA 1c . Therefore, the intervention duration with the best effect is >3& ≤ 6 months. Among all the studies included, however, we found no study with intervention duration >6&<12 months. The high heterogeneity among the group of interventions with duration ≥ 12 months and the current finding of somewhat greater benefits of intervention lasting >3& ≤ 6 months relative to those lasting  ≥ 12 months indicate that we still need more studies on peer support duration for further research. In addition, the effect of peer support on glycemic control weakens over time. According to the results measured during/immediately after the intervention, the effect is significantly positive and the efficacy during the intervention is better than that immediately after the intervention. The effect of peer support measured at  ≤ 6 months follow ‐ up interval is still positive but nonsignificant, while the efficacy measured at >6 months follow ‐ up interval is negative without significance, indicating more attention should be paid to maintain the effect of peer support after the end of intervention and we need to provide ongoing support. However, more studies should examine ongoing support, including specific time points and reinforcement methods. 37 There are still some limitations in our meta ‐ analysis. Firstly, although the 20 studies included are RCTs, of which 19 studies are not double blinded. Secondly, the control groups of some studies not only adopted usual care but also some additional interventions like education, appointments or follow ‐ up visits which may have compromised the sensitivity of studies to detect true effects of peer support. Thirdly, some subgroups in our subgroup analyses have small number of studies, some have high heterogeneity, and some have both. Fourthly, we did not conduct subgroup analysis controlling for other influencing factors of peer support like intervention intensity because of the limited studies. Conclusions In summary, peer support in these 20 studies achieves modest but statistically significant benefits on glycemic control for patients with T2DM. Analyses suggest possible differences among providers and type of peer support. Duration >3& ≤ 6 months is more likely effective and the effect of peer support on glycemic control weakens over time especially after the end of intervention. Both of them point to the importance of ongoing support and the importance of research investigating it such as through reinforcement methods within 6 months after the end of intervention. 6 145 Peer support among adults with type 2 diabetes