Multivariate meta‐analysis using individual participant data

Abstract

When combining results across related studies, a multivariate meta-analysis allows the joint synthesis of correlated effect estimates from multiple outcomes. Joint synthesis can improve efficiency over separate univariate syntheses, may reduce selective outcome reporting biases, and enables joint inferences across the outcomes. A common issue is that within-study correlations needed to fit the multivariate model are unknown from published reports. However, provision of individual participant data (IPD) allows them to be calculated directly. Here, we illustrate how to use IPD to estimate within-study correlations, using a joint linear regression for multiple continuous outcomes and bootstrapping methods for binary, survival and mixed outcomes. In a meta-analysis of 10 hypertension trials, we then show how these methods enable multivariate meta-analysis to address novel clinical questions about continuous, survival and binary outcomes; treatment–covariate interactions; adjusted risk/prognostic factor effects; longitudinal data; prognostic and multiparameter models; and multiple treatment comparisons. Both frequentist and Bayesian approaches are applied, with example software code provided to derive within-study correlations and to fit the models.

Meta-analysis using individual participant data: one-stage and two-stage approaches, and why they may differ

“…We illustrate this using an IPD meta‐analysis by Riley et al 55, which used the 10 hypertension trials introduced in Section 2.4. Here, instead of focussing on the treatment effect, the objective was to examine whether smoking is a prognostic factor for high systolic blood pressure at follow‐up, after adjusting for baseline blood pressure and treatment group.…”

“…However, a standard two‐stage approach (as outlined in Section 2) does not account for such correlations unless a multivariate meta‐analysis model is used to jointly synthesise all parameters, whilst accounting for their within‐study, and possibly between‐study, correlations 55, 83. This could lead to the standard two‐stage approach having reduced efficiency and even a change in the summary estimates.…”

“…Figure 1(a) gives a continuous outcome example, from a random effects IPD meta‐analysis of 10 randomised trials to evaluate the effect of anti‐hypertensive treatment on systolic blood pressure 55. Again, the summary results are almost identical when using either the one‐stage (equation (14)) or two‐stage (equation (1) followed by equation (9)) approaches using REML estimation.…”

Increased risk of aminoglycoside-induced hearing loss in MDR-TB patients with HIV coinfection

“…Finally, we used a meta-analysis, a rigorous statistical method, to consolidate research findings from studies addressing a similar topic but conducted in diverse settings. 47,48 This approach enables the analysis to draw more decisive conclusions on effect size for a relationship between AG-induced hearing loss and HIV coinfection because of its greater statistical power and external validity. 47 …”

“…47,48 This approach enables the analysis to draw more decisive conclusions on effect size for a relationship between AG-induced hearing loss and HIV coinfection because of its greater statistical power and external validity. 47 …”