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

Date & Time
Wednesday, September 6, 2023, 12:30 PM - 2:00 PM
Location Name
Pickwick
Session Type
Poster
Category
Bias
Authors
Wu P1, Jia Y1, Robinson K2, Tang J1
1Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, China
2Department of Medicine, School of Medicine, Johns Hopkins University, USA
Description

Background: The effect estimates from randomized clinical trials (RCTs) may vary across countries owing to genetic or economic differences. However, previous studies suggested that RCTs conducted in low/middle-income countries might be systematically different from those conducted in high-income countries.
Objectives: To estimate the discrepancies in effect estimates of RCTs between low/middle-income and high-income countries. We hypothesize that RCTs conducted in low/middle-income countries (LM-RCTs) produced larger effect estimates than those conducted in high-income countries (H-RCTs).
Methods: We screened Cochrane reviews for eligible meta-analyses that assessed the efficacy of interventions and produced statistically significant results. The RCTs using superiority design in eligible meta-analyses were classified as LM-RCTs or H-RCTs based on the location of recruiting centers. In each meta-analysis, LM-RCTs were included in the exposed group, while H-RCTs were in the control group. The main outcome was the ratio of risk ratio (RRR), hazard ratio (RHR), or odds ratio (ROR, with OR transformed from standardized mean difference) between LM-RCTs and H-RCTs. RRRs, RHRs, and RORs were pooled to form a single estimate using random-effects meta-analyses (DerSimonian-Laird method). Potential confounders and effect modifiers were assessed in a linear mixed-effect regression model. We planned to include 100 eligible meta-analyses.
Results: This was an interim analysis. We identified 48 meta-analyses from 4,763 Cochrane reviews. A total of 516 RCTs were included, consisting of 341 H-RCTs and 175 LM-RCTs. Among meta-analyses using OR (transformed from SMD), RR, and HR as the effect measure, LM-RCTs produced an effect estimate 1.75 (1.29-2.29, I2=87.3%), 1.21 (1.13-1.34, I2=19.8%), and 1.14 (1.02-1.26, I2=6.8%) times greater than H-RCTs, respectively. On average, LM-RCTs produced an effect estimate 1.37 (95% CI: 1.25-1.54, I2=80.2%) times greater than H-RCTs.
Conclusions: LM-RCTs generally produced an effect size 37% higher than H-RCTs, implying the impact of potential biases. Such discrepancies should be considered when conducting and interpreting the results of systematic reviews.
Patient, public and/or healthcare consumer involvement: Patients, the public, and/or healthcare consumers were not involved in this study.