Characteristics, reporting, and methods of trials included in time-to-event meta-analyses of systematic reviews: A meta-epidemiological review

Session Type
Oral presentation
Statistical methods
Goldkuhle M1, Kreuzberger N1, Hirsch C1, Iannizzi C1, Bora AM1, Bender R2, Van Dalen EC3, Hemkens LG4, Skoetz N1
1Evidence-based Medicine, Department I of Internal Medicine, Faculty of Medicine and University Hospital Cologne, University of Cologne, Cologne, Germany, Germany
2Department of Medical Biometry, Institute for Quality and Efficiency in Health Care, Cologne, Germany, Germany
3Princess Máxima Center for Pediatric Oncology, Utrecht, The Netherlands, Netherlands
4University Hospital Basel and University of Basel, Basel, Switzerland; Meta-Research Innovation Center (METRICS), Stanford University, Stanford, USA; Meta-Research Innovation Center Berlin (METRIC-B), Berlin Institute of Health, Berlin, Germany, Germany

Background: Previous studies demonstrated limitations in the reporting of time-to-event (TTE) analyses in trial publications. The challenges that systematic review authors face in trials they include for TTE meta-analysis remain unexplored.
Objectives: To explore the characteristics, reporting and methods of trials included in aggregate data TTE meta-analyses of Cochrane and non-Cochrane reviews.
Methods: We extracted data from trials included in pairwise, hazard ratio (HR)-based meta-analyses of primary outcomes and overall-survival/all-cause mortality of 50 systematic reviews. Reviews were identified from the Cochrane Database and Core Clinical Journals through systematic searches (02/2020; MEDLINE). For reviews, trials and their outcomes, we extracted, in duplicate, data on review/trial characteristics, outcome definitions, general and TTE analyses and specific characteristics relevant to TTE analysis.
Results: Reviews included 235 trials in eligible meta-analyses, resulting in 315 individual trial analyses. Most prominently assessed in trials (91%; 214/235) and reviews (88%; 44/50) was all-cause mortality/overall survival. Outcome definitions (61%; 132/315), censoring reasons (41%; 130/ 315) and follow-up specifications (56%; 175/ 315) for trial outcomes were frequently missing. Available TTE outcome data per trial were most frequently a HR combined with a log-rank p-value, survival curves and either time-point specific (16%; 50/315) or median survival times (15%; 47/315). Most prominently used individual measures were survival curves (83%; 263/315). HRs were reported for 76% (240/315) of trial outcomes. Review authors most frequently used reported HRs or recalculated TTE data from p-values (each 5%; 15/ 315), although TTE data sources were specified sporadically for individual trial outcomes. Reviews included most frequently intention-to-treat (64%; 202/315) and unadjusted analyses (25%; 80/315), but often did not specify either. Except for missing outcome data, for which numerical data was available in 57% (134/235), TTE relevant study characteristics, e.g., informative censoring and proportional hazards, were sporadically addressed in trial publications. Direct comparison to the reporting of their including reviews indicated that limitations and variability of trial reporting translates to the review level as well.
Conclusions: Reporting of trials included in TTE meta-analyses of recent systematic reviews appears variable and limited, similar to their including reviews.
Patient, public and/or healthcare consumer involvement: Not applicable