Aims To analyse quantitatively the association between the durability of glycaemic control and body weight changes during treatment. Materials and methods This study adhered to an appropriate methodology according to Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines. Studies with follow-ups >12 months, and final and intermediate assessments of haemoglobin A1c (HbA1c) and body weight were included. Four outcomes assessing therapeutic durability were extracted and synthesized using Stata statistical software, including changes in HbA1c, goal-achievement rate, failure rate and coefficient of failure (CoF). Results After 8.9 months of treatment, HbA1c levels declined from 8.03% [95% confidence interval (CI), 7.91-8.15; I-2 = 99.2%] to 7.15% (95% CI, 7.02-7.27; I-2 = 99.4%) and then gradually increased up to 7.72% (95% CI, 7.50-7.94; I-2 = 99.0%) 5 years later. The goal-achievement rate decreased from 54.8% (after 1 year of treatment) to 19.4% 5 years later. The CoF was 0.123 +/- 0.022%/year (P < .001). After stratification, the CoFs were 0.224 +/- 0.025%/year (P < .001) for weight gain, 0.137 +/- 0.034%/year (P < .001) for neutral weight and -0.024 +/- 0.032%/year (P = .450) for weight loss. After stratification by treatment approaches, the CoFs were 0.45%/year for insulin, 0.43%/year for sulphonylurea, 0.34%/year for thiazolidinediones, 0.29%/year for metformin, 0.16% for glucagon-like polypeptide-1 receptor agonists, 0.12% for surgery, -0.03% for sodium-glucose cotransporter-2 inhibitors and -0.21% for dipeptidyl peptidase-IV inhibitors. Conclusion Modest weight loss with a goal of 2-3% of body weight should be recommended to improve therapeutic durability and prevent beta-cell deterioration.