#1700 Heterogeneity in characteristics of target population and definitions of renal outcome on evaluating the renoprotective effect of GLP-1RA in patients with diabetes: a systematic review and meta-analysis of observational studies versus randomized controlled trials

Abstract Background and Aims Growing research suggests that glucagon-like peptide-1 receptor agonist (GLP-1RA) might exert potential renoprotective effects in patients with diabetes. However, existing evidence from observational studies and randomized controlled trials (RCTs) was inconsistent, which may be attributed to differences in characteristics of population and definitions of renal outcomes. Therefore, this study systematically summarized the characteristics of target population and definitions of renal outcomes, and evaluated the heterogeneity between RCTs and observational studies in these aspects. Method A systematic search was conducted in PubMed and Embase databases, RCTs and observational studies investigating the renal effects of GLP-1RA in patients with diabetes were included. Characteristics of study design, baseline characteristics of target population, the definitions of renal outcomes, and the corresponding findings were extracted. Baseline characteristics of target population were compared between RCTs and observational studies using random effects meta-analyses with Knapp-Hartung adjustment. The effects of GLP-1RA on renal outcomes were estimated using random effects model based on the different definitions of renal outcomes. Data were pooled using hazard ratios (HRs) or odds ratios (ORs) with 95% confidence intervals (CIs). Meta-regression was performed to explore whether the differences in study designs, types of controls, and characteristics of target population could explain the heterogeneity of renal outcomes. Results Overall, 42 studies were included in analysis, comprising 10 RCTs and 32 observational studies. In the main analysis of observational studies, propensity score matching was used in 23 studies, average treatment effects for the treated and average treatment effects for the control were estimated in 17 and 6 studies, respectively. Inverse probability of treatment weighting was applied in 7 studies to estimate average treatment effects, and overlap weighting was employed in one study to estimate the average treatment effects for the overlap population. Additionally, propensity score regression adjustment was used in one study and the conditional average treatment effects were estimated. Compared to patients enrolled in observational studies, those in RCTs were on average 3.9 years older (P = 0.004), had a 3.5-years longer duration of diabetes (P = 0.03), and presented with worse renal function: the proportion of patients with albuminuria was 23.4% higher, and estimated glomerular filtration rate (eGFR) was 14.7 ml/min/1.73 m2 lower among RCT participants (albuminuria: P = 0.019; eGFR: P = 0.067). For renal outcomes, the meta-analysis suggested that GLP-1RA significantly reduced the risk of composite renal outcomes compared with placebo (HR 0.75 [95% CI 0.70–0.80], P < 0.0001), and dipeptidyl peptidase-4 inhibitor (DPP-4i) (HR 0.78 [95% CI 0.74, 0.82], P < 0.0001). While, GLP-1RA were associated with a significantly higher risk than sodium-glucose cotransporter 2 inhibitor (SGLT2i) (HR 1.29 [95% CI 1.08, 1.53], P = 0.005). Regarding other renal outcomes, including the development of macroalbuminuria, decline in eGFR, progression to end-stage renal disease, acute kidney injury, and renal related death, the effects of GLP-1RA on renal outcomes in patients with diabetes varied. Conclusion Substantial systematic differences were observed in the characteristics of target population and definitions of renal outcomes between RCTs and observational studies for patients with diabetes. Therefore, the efficacy seen in RCTs may not reflect real-world effectiveness.