Purpose To determine whether sleeve gastrectomy (SG) or Roux-en-Y gastric bypass (RYGB) should be the optimal choice in patients stratified by diabetes duration and body mass index (BMI) level. Methods Classification tree analysis was performed to identify the influential factors for surgical procedure selection in real setting. Meta-analyses stratified by influential factors were conducted to compare the complete diabetes remission rates between SG and RYGB. The cost-effectiveness analysis was performed when results from meta-analysis remain uncertain. Results Among 3198 bariatric procedures in China, 824 (73%) SGs and 191 (17%) RYGBs were performed in patients with T2DM. Diabetes duration with a cutoff value of 5 years and BMI level with 35.5 kg/m(2) were identified as the influential factors. For patients with diabetes duration > 5 years, RYGB showed a significant higher complete diabetes remission rate than SG at 1 year: 0.52 (95% confidence interval (CI): 0.46-0.58) versus 0.36 (95% CI: 0.30-0.42). For patients with diabetes duration <= 5 years and BMI >= 35.5 kg/m(2), there was no significant difference between 2 procedures: 0.57 (95% CI: 0.43-0.71) for SG versus 0.66 (95% CI: 0.62-0.70) for RYGB. The cost-effectiveness ratios of SG and RYGB were 244.58 and 276.97 dollars per QALY, respectively. Conclusions For patients with diabetes duration > 5 years, RYGB was the optimal choice with regard to achieving complete diabetes remission at 1 year after surgery. However, for patients with diabetes duration <= 5 years and BMI >= 35.5 kg/m(2), SG appeared to provide a cost-effective choice.