Article type
Year
Abstract
Background: Prognostic factors are associated with the risk of a subsequent outcome in people with a given disease or health condition. They have a broad array of uses; for example, they help define disease at diagnosis; they inform individual treatment strategies; and they inform the design and analysis of clinical trials. Studies to identify prognostic factors are abundant in the literature, and meta-analysis is needed to combine them.
Methods: Meta-analysis using individual participant data (IPD), where the raw data are synthesised from multiple studies, has been championed as the gold-standard for synthesising prognostic factor studies. In this talk I will examine the feasibility and conduct of this approach, using a systematic review of currently published IPD meta-analyses of prognostic factors studies.
Results: Forty-eight published IPD meta-analyses of prognostic factors were identified; only three were published before 2000 but thereafter a median of four articles exist per year. These show that having IPD is advantageous, allowing modelling assumptions to be checked; variables to be analysed on their continuous scale; non-linear relationships to be examined; and results to be adjusted for other variables. However, the IPD approach also raises many challenges, such as large cost and time required to obtain and clean IPD; unavailable IPD for some studies; different sets of prognostic factors in each study; and variability in study methods of measurement. Their conduct can also be improved; continuous variables are often categorised without reason and more sophisticated methods to meta-analyse them are required; and publication bias is rarely examined.
Conclusions: IPD meta-analysis of prognostic factor studies is a huge improvement over an aggregate data meta-analysis approach. However, many practical, methodological and clinical problems still remain. The gold-standard is a prospectively planned meta-analysis using IPD.
Methods: Meta-analysis using individual participant data (IPD), where the raw data are synthesised from multiple studies, has been championed as the gold-standard for synthesising prognostic factor studies. In this talk I will examine the feasibility and conduct of this approach, using a systematic review of currently published IPD meta-analyses of prognostic factors studies.
Results: Forty-eight published IPD meta-analyses of prognostic factors were identified; only three were published before 2000 but thereafter a median of four articles exist per year. These show that having IPD is advantageous, allowing modelling assumptions to be checked; variables to be analysed on their continuous scale; non-linear relationships to be examined; and results to be adjusted for other variables. However, the IPD approach also raises many challenges, such as large cost and time required to obtain and clean IPD; unavailable IPD for some studies; different sets of prognostic factors in each study; and variability in study methods of measurement. Their conduct can also be improved; continuous variables are often categorised without reason and more sophisticated methods to meta-analyse them are required; and publication bias is rarely examined.
Conclusions: IPD meta-analysis of prognostic factor studies is a huge improvement over an aggregate data meta-analysis approach. However, many practical, methodological and clinical problems still remain. The gold-standard is a prospectively planned meta-analysis using IPD.