OBJECTIVES The ONCOTYROL Prostate Cancer Outcome and Policy (PCOP) Model is a state-transition micro-simulation model designed to evaluate prostate cancer (PCa) screening. We used the model to investigate how the benefits-harm balance of PCa screening is affected by the size of the latent prevalence pool. For this purpose, we recalibrated the natural history and detection component of the original PCOP model adopted from an earlier version of the Erasmus MISCAN model to match the higher prevalence observed by autopsy studies. The benefits and harms of screening predicted by the recalibrated model were then compared with predictions from the original model. METHODS For recalibration, we reprogrammed the natural history and detection component of the PCOP model as a deterministic state-transition model with stage- and grade-specific cancer states in the statistical software package R. All parameters were implemented as functions or variables and calibrated simultaneously in a single run using the ‘nlminb’ optimization algorithm available in R to minimize the deviation of model predictions from observed data. Calibration targets were observed data from autopsy studies, cancer registries and the European trial (ERSPC). Both the recalibrated and original models were identical except for calibrated parameters. RESULTS In total, we calibrated 46 parameters. Observed data could not be sufficiently fitted using the original set of parameters. Additional parameters, allowing for an interruption of disease progression in the stage- and grade-specific health states, and an effect modifier allowing for lower screening sensitivities in older men had to be implemented. Recalibration to higher prevalence demonstrated a considerable increase of overdiagnosis and decline of screening sensitivity, which significantly worsened the benefit-harm balance of screening regarding QALYs. CONCLUSIONS Benefit-harm predictions of models, which use calibration to simulate PCa progression in the unobservable latent phase, can be significantly affected by the assumptions on latent cancer prevalence.