Discrepant effect estimates in randomized clinical trials between high-income and low/middle-income countries: A meta-research study

Résumé fourni par la source

BACKGROUND: Randomized clinical trials (RCTs) in low/middle-income countries (LMICs) frequently produce evidence inconsistent with that conducted in high-income countries (HICs). This study aimed to compare RCT effect estimates in LMICs and HICs. METHODS: We identified RCTs from meta-analyses in six leading general medical journals and the Cochrane Database of Systematic Reviews between 2018 and 2023. Only RCTs that recruited participants after 2006 and used a negative control were included. Within each meta-analysis, RCTs sponsored by and enrolling participants in LMICs were designated as the exposed group, while those sponsored by and enrolling participants in HICs were the control group. Effect estimates were transformed into odds ratios (ORs); within each meta-analysis, ORs from LMICs and HICs were combined separately and compared as a ratio of ORs (RORs). RORs were combined across meta-analyses and subgrouped by patient-reported, investigator-assessed, and hard outcomes (e.g., mortality). An ROR larger than 1 indicated a larger OR from LMICs than from HICs. FINDINGS: 1,005 RCTs (423 from LMICs and 582 from HICs) were identified from 140 meta-analyses. The overall ROR was 1.73 (95% confidence interval: 1.44-2.08). The RORs were 1.94 (1.45-2.61), 1.78 (1.35-2.35), and 1.04 (0.81-1.34) for patient-reported, investigator-assessed, and hard outcomes, respectively. The ROR decreased to 1.04 (0.87-1.24) when restricted to RCTs with a low risk of bias. CONCLUSIONS: Our study suggested that RCTs from LMICs may produce larger effect estimates than those from HICs, with the difference substantially smaller for RCTs with a low risk of bias, indicating that bias may contribute to the discrepancy. FUNDING: This work was supported by the Shenzhen Science and Technology Program, Shenzhen Institutes of Advanced Technology, and China Postdoctoral Science Foundation.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.