INFORM-Stroke

Investigating concordance between preclinical multicenter confirmatory trials and systematic reviews in stroke

Background

This project aims to relate the strength of evidence between two important tools of biomedical translation - multicenter confirmatory trials and preclinical systematic reviews. A robust body of evidence is needed to determine whether a treatment has been sufficiently studied to be tested in early clinical trials with a good translational chance of success. Recent methods of generating this evidence include multicenter trials that test a treatment option against a verum while adequately controlling for factors of internal and external validity. Alternatively, systematic reviews and meta-analyses can pool evidence from preclinical experiments and assess the combined effect, taking into account study quality. In doing so, systematic reviews are subject to the possible error that the validity depends on the quality of the included studies. Using the example of data from a large multicenter trial in the field of stroke in which six treatment options were tested against a control intervention, we will investigate to what extent the results of systematic reviews differ from those of the trial. In addition, an important systematic error in reviews, the unclear influence of deceased animals on the effect size, will be quantified by calculating this from the multicenter trial data and comparing the number of deceased animals with the number of reported losses in the studies included in the reviews. Finally, we will investigate the extent to which methodological decisions influence the effect estimates in systematic reviews by calculating all possible options of predefined variables using the newly developed methodology of a multiverse analysis.

Objectives

Within this project, the following research questions (RQ) will be addressed:
RQ1-6: What is the effect of treatment of each of the six identified interventions of the Stroke Preclinical Assessment Network (SPAN) multicentre trial compared with the corresponding non-active controls on stroke outcomes in animal models of occlusion of the middle cerebral artery?

In addition, the following research questions were formulated in relation to the role of preclinical systematic reviews:
RQ7: How consistent are the results of systematic reviews and a multicentre stroke trial?
RQ8: How do the loss of experimental animals and different approaches to dealing with dropouts affect the results of animal studies and systematic reviews?
RQ9: How stable are the results of preclinical meta-analyses and does stability affect the consistency of results with multicentre studies?

Methodology

These research questions will be investigated in three separate work packages. In the first work package, 'Systematic Reviews', existing systematic reviews will be updated for the six interventions identified in SPAN, or new ones conducted if none exist. In the second work package, 'Analysis of SPAN data', we will use the SPAN dataset, which contains a full description of the outcome of each animal included in the study, to estimate i) the extent of non-reporting of animal loss in systematic reviews, ii) how much effect estimates vary based on different imputation methods, and iii) how accounting for loss through imputation affects the consistency of the results of multicentre studies and systematic reviews. The third work package 'Multiverse meta-analysis' aims to test the stability of the results from work package 1 by using the multiverse meta-analysis methodology.

Expected results / Implications / Perspectives

The evidence generated through this project will have a direct impact on translational processes in stroke research. The results can be used as a basis for decision-making to determine when a systematic review and when a multicenter trial is more appropriate for translation of a treatment. This will increase confidence in methodologies and encourage more targeted applications. Further, this may lead to faster translation of experimental results into clinical testing and, in the best case, clinical application. In addition, the results of the project can be used to create synergies with other projects that aim to support experimental researchers in how to conduct their research more robustly and support more efficient translation. The findings of this project can provide targeted recommendations on how best to deal with missing values due to animal losses, or decision support for robust meta-analyses. In this way, the project supports the evidence base of past and ongoing projects in the area of preclinical systematic reviews of stroke models, such as on the effect of age and comorbidities as well as biological sex on stroke endpoints, which are conducted in the CAMARADES Berlin working group.

Cooperation partners

Prof. Malcolm Macleod (University of Edinburgh)
Prof. Patrik Lyden (University of Southern California)