Hans-Georg Kuhn

In our project, “From Beyond the Bedside Back to the Bench", we applied causal inference methods to inform analytical approaches in preclinical stroke research. Animal attrition can significantly bias the estimation of causal treatment effects, as surviving animals may not represent the entire Study population, undermining the internal validity of studies despite initial randomization. Our Study aimed to illustrate this bias and demonstrate how Directed Acyclic Graphs (DAGs) can be used to visualize and mitigate it. Using DAGs, we depicted the assumptions about the causal structure underlying observed data, revealing less intuitive biases such as collider stratification bias. We developed an illustrative causal model specific to reclinical stroke research, showing how animal attrition induces bias through the interaction of animal welfare, initial disease severity, and adverse treatment side.

Our simulations covered various scenarios and demonstrated substantial bias in estimating treatment effects, even when the treatment had no causal effect. We demonstrated that researchers could mitigate this bias during the analysis phase, even with data solely from surviving animals, providing a comprehensive understanding of the underlying causal processes. Our findings emphasize that collider stratification bias should be a major concern in preclinical animal studies with severe side effects and high post-randomization attrition. Recognizing and addressing this bias is crucial for maintaining the validity of causal inferences.

This collaborative project between Prof. Kuhn and Prof. Kurth's research group was designed as a sub-project of a PhD thesis for which Prof. Kuhn co-supervised a PhD student. The subproject was successfully completed with the submission of a manuscript to an internationally renowned journal for stroke research. The manuscript entitled "Rethinking animal attrition in preclinical research: expressing causal mechanisms of selection bias using directed acyclic graphs" is currently under revision at the Journal of Cerebral Blood Flow and Metabolism. The collaboration with Prof. Kuhn underscores the importance of applying causal modeling techniques to ensure accurate and reliable estimation of treatment effects in preclinical research. By integrating population-based causal inference methods, we aim to improve basic science practice and emphasize the critical link between advanced analytical methods and robust experimental design in preclinical studies.