Abstract Valve size selection for transcatheter aortic valve replacement (TAVR) is currently based on cardiac CT-scan. At variance with patient-specific computer simulation, this does not allow the assessment of the valve-host interaction. We aimed to compare clinical valve size selection and valve size selection by an independent expert for computer simulation. A multicenter retrospective analysis of valve size selection by the physician and the independent expert in 141 patients who underwent TAVR with the self-expanding CoreValve or Evolut R. Baseline CT-scan was used for clinical valve size selection and for patient-specific computer simulation. Simulation results were not available for clinical use. Overall true concordance between clinical and simulated valve size selection was observed in 47 patients (33%), true discordance in 15 (11%) and ambiguity in 79 (56%). In 62 (44%, cohort A) one valve size was simulated whereas two valve sizes were simulated in 79 (56%, cohort B). In cohort A, concordance was 76% and discordance was 24%; a smaller valve size was selected for simulation in 10 patients and a larger in 5. In cohort B, a different valve size was selected for simulation in all patients in addition to the valve size that was used for TAVR. The different valve size concerned a smaller valve in 45 patients (57%) and a larger in 34 (43%). Selection of the valve size differs between the physician and the independent computer simulation expert who used the same source of information. These findings indicate that valve sizing in TAVR is still more intricate than generally assumed.