OBJECTIVES: To calibrate a developed model simulating the clinical and health economic impact of different implementation scenarios of artificial intelligence in breast cancer surveillance strategies.

METHODS: A discrete event simulation (DES) model was developed, describing two major interacting components: the clinical pathway of breast cancer follow-up and a tumour growth model reflecting breast cancer locoregional recurrence (LRR) and distant metastasis (DM). Natural history of breast cancer progression was simulated through the 5-year risks of LRR and DM using the INFLUENCE 2.0 nomogram, and was impacted by observed individual patient, primary tumour and treatment characteristics.

Subsequently, tumour growth parameters, including volume doubling time (VDT) was calibrated to match the expected incidence numbers and mode of detection (symptomatic or at routine visit) of LRR and DM. Simulated patients with oligo-metastatic (<3 DM) disease had an increased probability of survival compared to simulated patients with metastatic disease (>3 DM). Goodness-of-fit tests were used to assess model fits.

RESULTS: Calibrated patient, primary tumour and treatment characteristics differed from real-world data by <2%. The simulated model outcomes were comparable with the expected conditional annual risks over the 5-year period and simulated an incidence of 3426 per 100,000 LRR (target: 3000/3%) and 5747 per 100,000 DM (target: 6300/6,3%). The model outcomes agree well with the expected symptomatic and routine detections of recurrence.

Kolmorogov-Smirnov tests of both survival models of simulated patients with metastatic and oligo-metastatic disease indicated no significant difference between the observed distributions and the simulated survival models.

CONCLUSIONS: A simulation model was calibrated to simulate a realistic breast cancer population during follow-up and the underlying disease model for tumour growth of LRR and DM. The calibrated model is ready to support healthcare providers and developers regarding decisions on AI implementation through quantifying health and monetary benefits under a variety of breast cancer surveillance scenarios.