Search for a command to run...
Abstract This study replicates human cardiac safety clinical trials using computational models, which include diurnal hormonal placebo effects, aiming to create fast and efficient drug screening tools to assess QT interval prolongation preclinically. A virtual cardiac population that simulates sex-specific electrophysiological variability influenced by diurnal hormonal cycles was created to closely mirror human physiology. The goal is to assess the impact of these hormone dynamics on placebo-induced responses and evaluate the virtual population’s accuracy in reflecting clinical trial outcomes. The study showed that the virtual population model successfully replicated about 50% of the observed QT variability seen in placebo groups. The computational framework facilitated direct comparison with human clinical trials, achieving critical concentrations, i.e., concentrations that cause a mean QTc effect of 10 ms, within a 0.6-fold range of clinical results. Additionally, the exposure-response slopes exhibited relative errors averaging 43%, which translates into mean absolute errors of 6-7 ms in comparison to clinical data. These results highlight its potential to reproduce human physiology and provide results consistent with real-world human clinical trials. Additionally, a comparison with a previously published statistical placebo model further supports this approach. By approximating the diurnal hormonal variations, as a component of the mechanistic basis of placebo, this method offers a robust decision-making tool for early-stage drug safety assessment. Highlights Diurnal hormonal fluctuations significantly contribute to intersubject variability observed in placebo responses. Modeling conservative diurnal hormonal variations can reproduce approximately 50% of QT interval prolongation variability within placebo groups. Comprehensive human clinical trials for cardiac safety drug testing, including placebo effects and placebo-corrected QT interval prolongation, can be accurately replicated through virtual human computational models showing mean absolute errors of 6-7 ms. Graphical Abstract