Goldkuhle, Marius ORCID: 0000-0001-8747-6770 (2024). On time-to-event outcomes in evidence syntheses of randomized controlled trials: methods, challenges and guidance. PhD thesis, Universität zu Köln.
PDF
Goldkuhle_IPHS_Dissertation_META-TTE.pdf Download (19MB) |
Abstract
Time-to-event analyses are methodologically challenging, subject to distinct assumptions and their associated effect estimators, in particular the hazard ratio (HR), are not always straightforward to interpret. Previous studies suggest considerable limitations in the reporting of time-to-event analyses in study publications. Authors of evidence syntheses therefore face certain challenges when they include time-to-event analyses from randomized trials (RCTs) in their meta-analyses. The aim of the projects composing this dissertation was to characterize these challenges for the first time by means of three meta-epidemiological studies and to develop targeted methodological guidance to support authors of evidence syntheses in dealing with them. A meta-epidemiological study of 217 meta-analyses in 100 systematic reviews revealed significant differences in the characteristics, methods and reporting of time-to-event outcomes between the assessed reviews. Major limitations were found, for example, in the reporting of outcome definitions as well as general and time-to-event specific review methods. Review authors rarely considered characteristics of their included RCTs that may affect the reliability of time-to-event analyses, for example, informative censoring and non-proportional hazards. A second meta-epidemiological study systematically assessed the characteristics, methods and reporting of 315 time-to-event analyses in 235 RCTs included in meta-analyses of time- to-event outcomes. It demonstrated considerable variability in the methodology of the RCTs, for example, in the available types of time-to-event data, such as HRs, plots and P-values. Major limitations were evident, for instance, in the reporting of outcome definitions, censoring and follow-up data. In the RCT publications, trial characteristics affecting the reliability of time- to-event analyses were rarely considered and discussed. A third meta-epidemiological study examined the calculation of absolute effects (e.g., natural frequencies, number-needed-to-treat) to present the results of meta-analyses of time-to-event outcomes from a total of 96 Cochrane reviews. The study revealed that the corresponding estimates were often mislabeled and frequently calculated in a way that the direction of the effect of the underlying pooled HR was reversed. To address the identified problems, two systematic guidelines were developed according to the standards of the Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group. One of these guidelines illustrates how to deal with informative censoring in evidence syntheses, presents methods for determining a corresponding risk of bias in trial publications and explains how judgements for individual trials can be translated to a body of evidence. The second GRADE guideline presents alternatives for calculating absolute effects for time-to-event outcomes, discusses the selection of appropriate variables and possible limitations of each calculation approach. Based on three meta-epidemiological studies, this dissertation revealed significant problems with time-to-event analyses in meta-analyses of current evidence syntheses, in the publications of their included RCTs and in the communication of their results. Targeted guidance should now counteract some of the identified shortcomings in order to improve the conduct and reporting of meta-analyses of time-to-event outcomes in the future and thus ultimately the decisions based on them.
Item Type: | Thesis (PhD thesis) | ||||||||||||
Translated abstract: |
|
||||||||||||
Creators: |
|
||||||||||||
URN: | urn:nbn:de:hbz:38-746517 | ||||||||||||
Date: | 2024 | ||||||||||||
Place of Publication: | Köln | ||||||||||||
Language: | English | ||||||||||||
Faculty: | Faculty of Medicine | ||||||||||||
Divisions: | Unspecified | ||||||||||||
Subjects: | General statistics Life sciences Medical sciences Medicine |
||||||||||||
Uncontrolled Keywords: |
|
||||||||||||
Date of oral exam: | 26 September 2024 | ||||||||||||
Referee: |
|
||||||||||||
Refereed: | Yes | ||||||||||||
URI: | http://kups.ub.uni-koeln.de/id/eprint/74651 |
Downloads
Downloads per month over past year
Export
Actions (login required)
View Item |