Identification of Studies
We identified 279 studies after initially searching PubMed, Embase, and the Cochrane Library until to July 2021. A total of 199 studies remained after removing 80 duplicate studies. Then, a total of 150 ineligible studies were excluded after carefully screening titles and abstracts. Finally, we included 17 eligible studies [3, 19, 26,27,28,29,30,31,32,33,34,35,36,37,38,39,40] into the final statistical analysis after excluding 33 ineligible studies according to the following reasons: ineligible topic (n = 11), ineligible follow-up duration (n = 13), abstract (n = 8), and letter (n = 1). The process of identification and selection of eligible studies was displayed in Fig. 1.
Characteristics of the Included Studies
Among the 17 included studies [3, 19, 26,27,28,29,30,31,32,33,34,35,36,37,38,39,40], most were performed in European countries. The sample size of receiving RYGB in individual study was between 19 and 4434 except for one study which did not report the number of patients underwent this procedure. The number of patients underwent initial remission in individual study was ranging from 9 to 2254. All studies reported the HbA1c threshold for confirming relapse of T2DM, and 10 studies introduced the definition of T2DM relapse. Details of characteristics of all included studies were summarized in Table 1.
We included 1 RCT , 6 prospective cohorts [26, 28, 30,31,32, 37], and 10 retrospective cohorts [19, 27, 29, 33,34,35,36, 38,39,40] in the final analysis. Finally, RCT was rated as low quality based on Cochrane risk of bias, and the remaining 14 studies were identified as moderate-to-high quality because the total quality score of individual study based on the methodological items for non-randomized studies was between 7 and 9. We summarized the results of quality assessment in Table S2.
Meta-analysis of Long-Term Relapse Rate After Initial Remission
Among the included 17 studies, 16 studies reported the long-term relapse rate of T2DM after RYGB; the long-term relapse rate eventually reported in individual study was varying from 0.15 to 0.56 during the follow-up. Meta-analysis generated a long-term relapse rate of 0.30 (95% CI, 0.26, 0.34; P < 0.001, I2 = 86.7% [Pheterogeneity < 0.001]) after RYGB during follow-up. The result of individual study and pooled result was displayed in Fig. 2A.
Subgroup analysis according to HbA1c thresholds for defining long-term relapse of T2DM suggested that the long-term relapse rate was comparable between thresholds of 6.5 and 6.0%, with a pooled rate of 0.29 (95% CI, 0.25, 0.33; P < 0.001, I2 = 87.6% [Pheterogeneity < 0.001]) versus 0.34 (95% CI, 0.17, 0.70; P = 0.004, I 2 = 85.5% [Pheterogeneity = 0.009]), respectively. However, the pooled rate was 0.43 (95% CI, 0.36, 0.51; P < 0.001) in the threshold of 7.0% group, which was higher than that in the threshold of 6.5% and 6.0% groups. Subgroup analysis based on HbA1c threshold was summarized in Table 2.
Subgroup analysis of long-term relapse according to study design suggested that the pooled rate based on prospective and retrospective studies was 0.37 (95% CI, 0.27, 0.49; P < 0.001, I2 = 75.7% [Pheterogeneity = 0.001]) and 0.26 (95% CI, 0.23, 0.31; P < 0.001, I2 = 89.8% [Pheterogeneity < 0.001]), which were all lower than that based on RCT, with a pooled rate of 0.53 (95% CI, 0.32, 0.88; P = 0.015). Subgroup analysis based on study design was summarized in Table 2.
Meta-analysis of Initial Remission Rate After RYGB Surgery
Among the 17 included studies, 15 studies reported the initial remission rate after receiving RYGB surgery. The initial remission rate of individual study was ranging from 0.38 to 0.89, and meta-analysis revealed a pooled remission rate of 0.63 (95%CI, 0.55, 0.72; P < 0.001, I2 = 97.9% [Pheterogeneity < 0.001]). The pooled remission rate and remission rate of individual study were all displayed in Fig. 2B.
Subgroup analysis according to study design suggested that the initial remission rate reported by RCT was 0.79 (95% CI, 0.63, 0.99; P = 0.046), and the initial remission rate based on prospective and retrospective studies was 0.64 (95% CI, 0.53, 0.78; P < 0.001, I2 = 92.6% [Pheterogeneity < 0.01]) and 0.61 (95% CI, 0.50, 0.73; P < 0.001, I2 = 98.6% [Pheterogeneity < 0.001]), respectively. Subgroup analysis based on study design was summarized in Table 2.
Meta-analysis of the Risk of Recurrence for Comparison of RYGB and SG
Among 17 eligible studies, 3 publications including 4 studies reported the hazard ratio of recurrence when RYGB surgery compared to SG surgery. Meta-analysis suggested a pooled hazard ratio of 0.73 (95% CI, 0.66, 0.81; P < 0.001, I2 = 0.0% [Pheterogeneity = 0.82]) for the comparison of RYGB and SG, indicating that RYGB was associated with a lower risk of recurrence of T2DM compared to SG. The pooled result was displayed in Fig. 2C.
In order to examine the robustness of pooled results in terms of long-term relapse rate and initial remission rate, we conducted sensitivity analysis with the sequential omission of each individual study method, and sensitivity analysis suggested a robust pooled long-term relapse rate (see Fig. 3A) and initial remission rate (see Fig. 3B).