Incorporating single-arm studies in meta-analysis of randomised controlled trials: a simulation study

Area of Science:

• Biostatistics
• Evidence Synthesis
• Health Technology Assessment

Background:

• Real-world data (RWD) from non-randomized studies, such as single-arm studies, are increasingly used to complement randomized controlled trials (RCTs).
• Challenges exist in integrating RWD with RCT data due to potential biases and differing heterogeneity.
• Meta-analysis methods need to effectively handle mixed data sources for reliable evidence synthesis.

Purpose of the Study:

• To compare different meta-analysis methods for synthesizing aggregate data from both RCTs and single-arm studies.
• To evaluate the performance of contrast-based and arm-based methods under various simulation scenarios.
• To identify the most robust approach for combining diverse evidence in health technology assessments.

Main Methods:

• A simulation study was conducted to compare contrast-based (Begg & Pilote, 1991) and arm-based (Zhang et al, 2019) meta-analysis methods.
• Scenarios varied the proportion of RCTs vs. single-arm studies, magnitude of bias, and between-study heterogeneity.
• Methods were applied to a health technology assessment dataset including three RCTs and 11 single-arm studies.

Main Results:

• Zhang et al

Conclusions:

• The hierarchical power and commensurate prior methods (Zhang et al) provide a robust approach for synthesizing aggregate data from RCTs and single-arm studies.
• These methods effectively balance accounting for bias and heterogeneity while reducing uncertainty in estimates.
• The study focused specifically on pairwise meta-analysis using aggregate data.