Abstract
Meta‐analysis of individual participant data (IPD) is increasingly utilised to improve the estimation of treatment effects, particularly among different participant subgroups. An important concern in IPD meta‐analysis relates to partially or completely missing outcomes for some studies, a problem exacerbated when interest is on multiple discrete and continuous outcomes. When leveraging information from incomplete correlated outcomes across studies, the fully observed outcomes may provide important information about the incompleteness of the other outcomes. In this paper, we compare two models for handling incomplete continuous and binary outcomes in IPD meta‐analysis: a joint hierarchical model and a sequence of full conditional mixed models. We illustrate how these approaches incorporate the correlation across the multiple outcomes and the between‐study heterogeneity when addressing the missing data. Simulations characterise the performance of the methods across a range of scenarios which differ according to the proportion and type of missingness, strength of correlation between outcomes and the number of studies. The joint model provided confidence interval coverage consistently closer to nominal levels and lower mean squared error compared with the fully conditional approach across the scenarios considered. Methods are illustrated in a meta‐analysis of randomised controlled trials comparing the effectiveness of implantable cardioverter‐defibrillator devices alone to implantable cardioverter‐defibrillator combined with cardiac resynchronisation therapy for treating patients with chronic heart failure. © 2016 The Authors. Statistics in Medicine Published by John Wiley & Sons Ltd.
Highlights
Patient-centered outcomes research seeks, among other things, to understand which patients are most likely to benefit from available treatments, yet randomised controlled trials of new therapies are designed to answer primary efficacy questions, and often lack power to detect differential treatment effects across patient subgroups
Data in each individual study and report complete case analysis (CCA) [4]. This ‘unprincipled’ approach assumes that the data are missing completely at random (MCAR); that is, missing values are independent of any observed or unobserved factors or may depend on variables included in the analysis model but are independent of other observed values, a mechanism known as covariate-dependent missing completely at random (CD-MCAR) [3]
The study shows that the two approaches perform, the joint model provided slightly better coverage in settings with smaller samples (N = 100)
Summary
Patient-centered outcomes research seeks, among other things, to understand which patients are most likely to benefit from available treatments, yet randomised controlled trials of new therapies are designed to answer primary efficacy questions, and often lack power to detect differential treatment effects across patient subgroups. Meta-analysis of individual participant data (IPD) can aid the estimation of subgroup treatment effects by pooling information from multiple studies and using patient-level data to improve precision. An important concern when pooling data across studies relates to partially or completely missing data for some studies. The main problem with missing data is that individuals with incomplete information tend to be systematically different from those with complete data, and inferences based on the complete cases may be misleading. Recent studies considered methods for handling predictors or confounders that are missing for some studies [1, 2]. We consider the dual problem of outcomes that are partially or completely missing for some studies, and multiple outcomes measured per study participant
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call