Article type
Year
Abstract
Background: Previous work demonstrated flaws in the analysis and interpretation of time-to-event outcomes by authors of recent Cochrane reviews. As part of a Cochrane Networks Innovation Fund project, we conducted a survey of Cochrane editors.
Objectives: To identify which training and informational resources around time-to-event analyses editors use and suggest to Cochrane review authors. Furthermore, to identify difficulties and questions editors frequently encounter from review authors or face themselves.
Methods: Based on previous methodological work we developed a survey that included 27 quantitative and narrative items. A general part of the survey addressed questions on review production, methodological expertise as well as relevant training and informational resources. The specialized part consisted of questions that addressed particular difficulties with time-to-event analyses that could arise along the review development process. The survey was distributed among the editorial staff of all Review Groups within the Cochrane Cancer Network. Furthermore, senior editors of other Networks were invited to forward it to editors if they found it relevant to their work. Its results were discussed and appraised by the authors of this abstract.
Results: Overall, we received 12 responses (all except one from within the Cancer Network). The majority of respondents were statistical editors. Most respondents rated their knowledge on time-to-event analyses three on a scale of five (“familiar with the basic methods”). Review Groups frequently recommend specific time-to-event training materials to authors, of which the most prominent one is the instruction paper authored by Tierney and colleagues in 2007 (1). Difficulties and questions were identified in all areas addressed by the survey. The most problematic according to quantitative items were the underlying assumptions of analytic methods (67%; 8/12), the reconstruction of data from primary reports (58%; 7/ 12) and the interpretation of effects (58%; 7/12). Besides the reconstruction of survival data, narrative responses revealed the reconstruction of survival data, the proportional hazards assumption, the timing of randomization, competing events, the concept of censoring and absolute effects as areas with increased uncertainty.
Conclusions: Our findings suggest that Cochrane editors frequently encounter issues with the crucial concepts underlying time-to-event analyses. Several particularly problematic areas are not yet covered by training and information resources. However, in the course of our Cochrane Networks Innovation Fund project targeted resources are currently under development.
1. Tierney JF, Stewart LA, Ghersi D, Burdett S, Sydes MR. Practical methods for incorporating summary time-to-event data into meta-analysis. Trials. 2007;8:16-.
Patient or healthcare consumer involvement: None
Objectives: To identify which training and informational resources around time-to-event analyses editors use and suggest to Cochrane review authors. Furthermore, to identify difficulties and questions editors frequently encounter from review authors or face themselves.
Methods: Based on previous methodological work we developed a survey that included 27 quantitative and narrative items. A general part of the survey addressed questions on review production, methodological expertise as well as relevant training and informational resources. The specialized part consisted of questions that addressed particular difficulties with time-to-event analyses that could arise along the review development process. The survey was distributed among the editorial staff of all Review Groups within the Cochrane Cancer Network. Furthermore, senior editors of other Networks were invited to forward it to editors if they found it relevant to their work. Its results were discussed and appraised by the authors of this abstract.
Results: Overall, we received 12 responses (all except one from within the Cancer Network). The majority of respondents were statistical editors. Most respondents rated their knowledge on time-to-event analyses three on a scale of five (“familiar with the basic methods”). Review Groups frequently recommend specific time-to-event training materials to authors, of which the most prominent one is the instruction paper authored by Tierney and colleagues in 2007 (1). Difficulties and questions were identified in all areas addressed by the survey. The most problematic according to quantitative items were the underlying assumptions of analytic methods (67%; 8/12), the reconstruction of data from primary reports (58%; 7/ 12) and the interpretation of effects (58%; 7/12). Besides the reconstruction of survival data, narrative responses revealed the reconstruction of survival data, the proportional hazards assumption, the timing of randomization, competing events, the concept of censoring and absolute effects as areas with increased uncertainty.
Conclusions: Our findings suggest that Cochrane editors frequently encounter issues with the crucial concepts underlying time-to-event analyses. Several particularly problematic areas are not yet covered by training and information resources. However, in the course of our Cochrane Networks Innovation Fund project targeted resources are currently under development.
1. Tierney JF, Stewart LA, Ghersi D, Burdett S, Sydes MR. Practical methods for incorporating summary time-to-event data into meta-analysis. Trials. 2007;8:16-.
Patient or healthcare consumer involvement: None