Meta-analysis of a binary outcome using individual participant data and aggregate data.

Area of Science:

• Biostatistics
• Epidemiology
• Health Services Research

Background:

• Synthesizing binary outcomes from health studies requires accounting for participant-level factors.
• Existing meta-analysis methods may not adequately address within-study covariate associations and ecological bias.

Purpose of the Study:

• To develop and validate novel meta-analysis models for binary outcomes, incorporating participant-level covariates.
• To demonstrate the utility of individual participant data (IPD) in accurately modeling within-study associations and mitigating bias.
• To produce both population-level and individual-level risk predictions.

Main Methods:

• Development of advanced meta-analysis models capable of handling individual participant data (IPD) or a mix of IPD and aggregate data.
• Modeling the within-study association between participant-level covariates and event probability.
• Application to traumatic brain injury (TBI) studies to synthesize six-month mortality risk in relation to age.

Main Results:

• The models successfully synthesized six-month mortality risk across 14 TBI studies, utilizing IPD from four studies.
• Increased individual age was significantly associated with higher six-month mortality risk.
• Clear evidence of ecological bias was detected, indicating that study-level mean age influenced individual mortality probability.

Conclusions:

• The developed meta-analysis models effectively synthesize binary outcomes while accounting for participant-level covariates and within-study associations.
• Individual participant data (IPD) is crucial for accurately modeling complex associations and reducing ecological bias.
• The findings highlight the importance of considering both individual and study-level factors in health outcome meta-analyses, as demonstrated in TBI mortality risk.