Sensitivity analysis for exploring the variability and parameter landscape in virtual patient cohorts of multi-vessel coronary artery disease

Virtual patient cohorts (VPC) are crucial in <i>in silico</i> clinical trials, offering a promising, cost-effective and ethically advantageous alternative to real clinical randomized controlled trials to evaluate the safety and efficacy of clinical decision support tools and medical devices. This article focuses on the role of sensitivity analysis (SA) in evaluating a VPC created through a virtual cohort generator, which includes a one-dimensional pulse wave propagation model of the coronary circulation. Given the inherent limitations of clinical data, a synthetic VPC was generated that captured the global population variability of the fractional flow reserve distribution observed in the FAME study, a real-world randomized clinical trial. The synthetic VPC was created using random parameter variation and filtering with acceptance criteria, possibly inducing correlations between inputs. An SA methodology was employed that is able to account for correlations caused by acceptance criteria to explore the input-output relationship of the VPC and to explain its variability. The severity of the stenosis was found to be a key driver of the variability of the VPC. In general, the proposed SA approach, capable of handling correlated inputs, demonstrates an effective method for evaluating VPCs, providing a robust framework for <i>in silico</i> clinical trial applications.This article is part of the theme issue 'Uncertainty quantification for healthcare and biological systems (Part 2)'.