Assessment of the current state of the consideration of specific features of time-to-event analyses in meta-analyses: re-analysis of systematic reviews and implementation of guidance
Public Health, Healthcare Research, Social and Occupational Medicine
Final Report Abstract
Meta-analyses of time-to-event outcomes are associated with specific challenges. The aim of our project was to assess how authors of systematic reviews deal with these challenges by means of a meta-epidemiological study and to develop targeted guidelines to enhance the quality of meta-analyses of time-to-event outcomes. The meta-epidemiologic study was submitted for publication in two separate articles. Our first article systematically assessed the characteristics and reporting of time-to-event meta-analyses in 100 reviews and their 217 time-to-event review outcomes. We showed that reporting appears inconsistent and sometimes insufficient. Outcome definitions, for example, were available for less than half of the assessed outcomes. General review methodology was reported in a limited fashion. Time-to-event data from trial publications were obtained with wide variability. Consideration of trial characteristics relevant to the reliability of time-to-event analyses, such as variable follow-up between trials, informative censoring, competing events, and proportional hazards, was poor. In the second article, we systematically assessed the characteristics and reporting of trials in meta-analyses of a random sample of the previously considered reviews. A total of 235 individual trials with 315 time-to-event analyses were included in primary meta-analyses and overall survival meta-analyses of the respective reviews. Our results suggest that timeto-event analyses in trial publications suffer from reporting limitations as well. Outcome definitions, censoring mechanisms, and data on endpoint follow-up, for example, were not always available. Time-to-event data available in trial publications for individual analyses varied, and review authors rarely reported this information, e.g. sources of trial time-to-event data, for the trial estimators included in their reviews. Trial characteristics relevant to reliability of time-to-event end points were addressed only sporadically in trial publications. To address the identified shortcomings and to improve the quality of meta-analyses of timeto-event outcomes in a targeted manner, we are developing a reporting guideline that complements available methodological guidance. The development process will be overseen by a committee of international experts. Following the results of the meta-epidemiological study, 21 potential reporting elements, including explanations and rationales, were identified, and consented to by the committee. A stakeholder survey will prioritize the current elements and adjust them in terms of content and form. After the results of the survey, the final reporting guideline elements will be determined. In addition to a separate publication, it is planned to include guidance in the Cochrane Handbook and implement it via webinars.
Publications
-
Time-to-event analyses in meta-analyses: meta-epidemiological assessment of systematic reviews and their included randomized controlled trials (META-TTE) [Internet]. OSF
Kreuzberger, N.; Skoetz, N.; Bender, R.; van Dalen, E.; Hemkens, L.G.; Monsef, I.; Trivella, M. & Goldkuhle, M.
-
Transparent reporting of meta-analyses of time-to-event outcomes based on aggregate data from randomized trials of interventions (META-TTE reporting guideline) [Internet]. OSF
Goldkuhle, M.; Kreuzberger, N.; Bender, R.; Bora, AM; Burdett, S.; Hirsch, C.; Iannizzi, C.; Fisher, D.; Tierney, J. & Skoetz, N.
