Exploring the characteristics, reporting, and methods of systematic reviews including time-to-event meta-analyses and outcome analyses: A meta-epidemiological review
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, CA, USA; Meta-Research Innovation Center Berlin (METRIC-B), Berlin Institute of Health, Berlin, Germany, Germany
Time-to-event meta-analyses are complex and often suffer from often inadequate reporting in trial publications.
Objectives: We examined the characteristics, reporting and methods of Cochrane, and non-Cochrane systematic reviews, including time-to-event meta-analyses based on aggregate data.
Methods: Based on an a-priori published protocol, we included aggregate data systematic reviews with at least one pairwise meta-analysis based on the hazard ratio without further restriction. The 50 most recent Cochrane reviews corresponding to our eligibility criteria up to 08/2020 were identified in the Cochrane Database of Systematic Reviews. A systematic search then identified a corresponding random sample (n=50) of systematic reviews from Core Clinical Journals (Medline; 08/02/2021) from the same timeframe (02/2017 - 08/2020). For each review and each meta-analyzed time-to-event outcome, we extracted data on general characteristics, included outcome definitions, general and time-to-event specific methods, and handling of specific randomized trial characteristics with relevance to time-to-event analysis. Review selection and data-extraction took place in duplicate and was based on a priori developed sheets.
Results: The 100 included reviews analyzed 217 individual time-to-event outcomes (median: 2; interquartile range: 1-2). Most frequently assessed were overall survival/all-cause mortality (41%; 89/217). Less than half of all time-to-event outcomes were clearly defined (48%; 104/217). Few reviews specified general methods, e.g., the types of included analysis (intention-to-treat, per protocol) (35%; 35/100) or the adjustment of eligible effect estimates (12%; 12/100). Sources of time-to-event summary data differed substantially in number and complexity between reviews. The most prominent data sources were directly reported trial hazard ratios (64%; 64/100) and reference to established guidance without further specification (46%; 46/100). Study characteristics with particular relevance to time-to-event analysis, e.g., variable follow-up, informative censoring, proportional hazards, were seldomly included in additional analyses (e.g., sensitivity analyses) and seldomly discussed throughout the reviews. Reporting among Cochrane reviews appeared more comprehensive in most of our assessed items.
Conclusions: The reporting of current systematic reviews, including time to event meta-analyses, appears inconsistent and often inadequate. Reporting standards for such reviews are required. Patient, public, and/or healthcare consumer involvement: Not applicable.