Investigation of generalisibility of clinical trial results to routine care using generalised evidence-synthesis applied to two clinical scenarios with comorbidity
Final Report Abstract
Objective of the project was to investigate the generalisibility of clinical trial results to routine care using generalized evidence-synthesis. In a first step, systematic literature reviews in the areas of generalized evidence-synthesis and post-stroke depression were performed. In a second step, data sources for two application scenarios (post-stroke depression, diabetes and diabetic foot syndrome) covering aggregated data from randomized trials and individual data from cohorts and routine care were identified. With the help of these data sources, adding information from the systematic reviews on treatment and prevention of post-stroke depression, a careful analysis of data of clinical cohorts, and intensive discussion in the group (all partners), two models (post-stroke depression, diabetes mellitus and diabetic foot) were developed as input for the evidence synthesis. In order to provide a R-package for generalized evidence-synthesis, several preparatory steps were taken by Dr. P. Verde. An R-package for Bayesian meta analysis of diagnostic tests was developed and extended and applied to the problem of generalisibility of treatments investigated in this project (bamdit; version 1.1-1, 8 December 2011), a review of R-methods for statistical computing was performed (e.g. bootstrap computations with R) and the R-code for the application scenario of diabetes and diabetic foot was provided in a report. It is planned to finalise the development of a full R-package for generalised evidence synthesis in a follow-up project. In step 3, techniques of generalised evidence were applied to the acquired data sources and the models specified for the two application scenarios. With the data available, only one application scenario could be investigated in detail: diabetes and diabetic foot/amputation. The information sources used were aggregated data coming from 7 RCTs and patient individual data coming from a cohort study. With the techniques chosen, different study-types could be combined by sharing common outcomes, aggregated and individual patient data could be combined, treatment effects of aggregated data by external validity adjusted, individual baseline risk identified and group of patients with highest risk determined, pieces of evidence for potential quality issues adjusted and treatment effects for subgroups of patients predicted. Taking into consideration external validity bias and the baseline risk of the cohort, a clear positive treatment effect could be demonstrated. It is expected that patients similar to those participating in the cohort will have a positive treatment effect.
Publications
-
Comment on: efficacy and safety of tigecycline: a systematic review and meta-analysis. J Antimicrob Chemother 2011; 66:2893-2895
Curcio D, Verde PE
-
Imbalanced mortality evidence for tigecycline: 2011, the year of the meta-analysis (correspondence). Clinical Infectious Diseases 2012; 55: 471-472
Verde PE, Curcio D
-
Statistical inference with computer simulation: an introduction to boostrap analysis with R. Estadistica 2012; 64:182
Verde PE
-
Therapy of post-stroke depression – a systematic review. Die Psychiatrie 2013; 10: 108-129
Wannagat W, Zielasek J, Gaebel W
-
Combining randomized and non-randomized evidence in clinical research: a review of methods and applications. Research Synthesis Methods 6,1, March 2015, Pages 45-62
Verde PE, Ohmann C
