Validation of Microsimulation Models against Alternative Model Predictions and Long-Term Colorectal Cancer Incidence and Mortality Outcomes of Randomized Controlled Trials

Med Decis Making. 2020 Aug;40(6):815-829. doi: 10.1177/0272989X20944869.


Background. This study aimed to assess the validity of 2 microsimulation models of colorectal cancer (CRC), Policy1-Bowel and ASCCA.Methods. The model-estimated CRC risk in population subgroups with different health statuses, "dwell time" (time from incident precancerous polyp to symptomatically detected CRC), and reduction in symptomatically detected CRC incidence after a one-time complete removal of polyps and/or undetected CRC were compared with published findings from 3 well-established models (MISCAN, CRC-SPIN, and SimCRC). Furthermore, 6 randomized controlled trials (RCTs) that provided screening using a guaiac fecal occult blood test (Funen trial, Burgundy trial, and Minnesota Colon Cancer Control Study [MCCCS]) or flexible sigmoidoscopy (NORCCAP, SCORE, and UKFSST) with long-term follow-up were simulated. Model-estimated long-term relative reductions of CRC incidence (RRinc) and mortality (RRmort) were compared with the RCTs' findings. Results. The Policy1-Bowel and ASCCA estimates showed more similarities to CRC-SPIN and SimCRC. For example, overall dwell times estimated by Policy1-Bowel (24.0 years) and ASCCA (25.3) were comparable to CRC-SPIN (25.8) and SimCRC (25.2) but higher than MISCAN (10.6). In addition, ∼86% of Policy1-Bowel's and ∼74% of ASCCA's estimated RRinc and RRmort were consistent with the RCTs' long-term follow-up findings. For example, at 17 to 18 years of follow-up, the MCCCS reported RRmort of 0.67 (95% confidence interval [CI], 0.51-0.83) and 0.79 (95% CI, 0.62-0.97) for the annual and biennial screening arm, respectively, and the UKFSST reported RRmort of 0.70 (95% CI, 0.62-0.79) for CRC at all sites and 0.54 (95% CI, 0.46-0.65) for distal CRC. The corresponding model estimates were 0.65, 0.74, 0.81, and 0.61, respectively, for Policy1-Bowel and 0.65, 0.70, 0.75, and 0.58, respectively, for ASCCA. Conclusion. Policy1-Bowel and ASCCA's estimates are largely consistent with the data included for comparisons, which indicates good model validity.

Keywords: ASCCA; Policy1-Bowel; colorectal cancer; microsimulation; population modelling; validation.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Colorectal Neoplasms / epidemiology
  • Colorectal Neoplasms / mortality*
  • Computer Simulation / standards*
  • Computer Simulation / statistics & numerical data
  • Early Detection of Cancer / methods
  • Humans
  • Incidence
  • Randomized Controlled Trials as Topic / statistics & numerical data
  • Reproducibility of Results
  • Sigmoidoscopy / methods
  • Treatment Outcome*