Xuxi Zhang

Risk of bias All 20 studies included in our meta ‐ analysis were RCTs, but only one study 26 reported blinding of participants and personnel. Individual quality of most studies revealed low bias risk in most domains (Supplementary Table S2). Quality of the evidence for most results in our study was from moderate to high, except the results of PPI, CCOI and Follow ‐ up >6 months after the end of intervention due to limited studies (Supplementary Table S3). Overall results of peer support intervention effect All 20 studies provided related data of HbA 1c mean and HbA 1c SD as the outcome measure of peer support intervention effect. In 17 studies 13 ‐ 15, 17 ‐ 27, 29, 31, 32 , we used the HbA 1c results of 0 month follow ‐ up to represent the effect of intervention. However, 3 studies 16, 28, 30 only provided the results of 1.75 ‐ month follow ‐ up, 4.5 ‐ month follow ‐ up and 10.25 follow ‐ up respectively. We used these results of the 3 studies to represent the effect of intervention. Figure 2, Part A presented the combined results of the 20 studies in our meta ‐ analysis. The results showed heterogeneity between studies ( I 2 = 49.5%) was acceptable. The pooled results indicated statistically significant difference in HbA 1c outcomes between IG and CG of  ‐ 0.16% (95% CI  ‐ 0.25 to  ‐ 0.07) or  ‐ 1.7mmol/mol ( P < 0.001). The sensitivity of 20 studies was low and the publication bias was acceptable. The details of sensitivity analysis and publication bias assessment can be found in Supplementary Data. Results of subgroup analysis based on intervention duration We divided the results into three mutually exclusive groups ( ≤ 3 months, >3& ≤ 6 months,  ≥ 12 months) based on the intervention duration (no studies reported duration >6&<12 months) to perform the subgroup analysis (Fig. 2, Part B). In intervention duration ≤ 3 months group 16, 22, 27, 28, 30 , although there was no heterogeneity ( I 2 = 0%), the difference in HbA 1c outcomes between IG and CG of  ‐ 0.01% (95% CI  ‐ 0.16 to 0.15) or  ‐ 0.1mmol/mol was much smaller and no longer statistically significant ( P = 0.924). In the group of studies with duration >3& ≤ 6 months 13, 18, 21, 24 ‐ 26, 29, 31, 32 , the heterogeneity was reduced ( I 2 = 38.4%) and there were larger differences in HbA 1c outcomes with statistical significance  ‐ 0.28% (95% CI  ‐ 0.46 to  ‐ 0.09) or  ‐ 3.1mmol/mol ( P = 0.003). Similarly, among the group with duration ≥ 12 months group 14, 15, 17, 19, 20, 23 , difference in HbA 1c was greater,  ‐ 0.21% (95% CI  ‐ 0.34 to  ‐ 0.07) or  ‐ 2.3mmol/mol ( P = 0.002), but the heterogeneity increased substantially ( I 2 = 70.7%). Moreover, although there was no statistically significant difference among the three groups, the P value of heterogeneity among the three groups was 0.057 which was close to 0.05. 140 Chapter 6