Development and application of innovative evidence synthesis methods

Abstract

Systematic reviews and meta-analyses synthesise evidence from studies on a particular topic, and frequently inform clinical guidelines, policy, and practice. Yet, suboptimal methodological conduct may introduce biases or compromise the quality of these reviews. For instance, reviews that fail to identify and incorporate unpublished evidence may be affected by publication bias and selective outcome reporting bias; inadequate data quality and integrity checks threaten the validity and trustworthiness of results; and failure to plan for data sharing and re-use limits the utility of data, resulting in research waste.

This thesis explores some of the methodological shortcomings of current evidence synthesis approaches, proposes guidance and solutions to address these, and provides examples of how these solutions may be applied in practice. While some of the methods developed in this thesis are applicable to evidence synthesis in general, the focus is on an approach known as individual participant data meta-analysis (IPD-MA). Often regarded as the gold standard in meta-analyses, IPD-MA involve the central collection and synthesis of raw line-by-line data for each individual participant within each eligibility study. This approach offers many advantages, including

improved data quality and availability, greater capacity for harmonisation across studies, and the ability to conduct more complex analyses (such as subgroup analyses).

This thesis is presented in a series of eight manuscripts across nine chapters.