Simulating results from trials of sigmoidoscopy screening using the OncoSim microsimulation model

Abstract

IntroductionProjection of the effect of cancer screening interventions are frequently conducted using complex simulation models. It is important that such models demonstrate their ability to replicate observational results on the effect of screening. We present results using the OncoSim-CRC microsimulation model to replicate results from four randomized trials (RCTs) of sigmoidoscopy screening for colorectal cancer (CRC). MethodsThe published results of four RCTs of sigmoidoscopy were reviewed. Two key outcomes were identified: the intention-to-treat hazard ratios (HR) for CRC incidence and CRC mortality for the screening versus control arms. Each RCT study arm was simulated within OncoSim-CRC using the study specific entry criteria, follow-up and observed participation and compliance rates. The ratio of predicted cases (deaths) between intervention arm and control arm was used to estimate the HRs. ResultsThe RCTs differed in the implementation of sigmoidoscopy screening and only one (PLCO) used more than one cycle. All four RCTs found significant reductions, HR <1, in CRC incidence (range 0.77–0.82) and three for CRC mortality (range 0.69–0.78). The four study cohorts were successfully simulated to match the age and sex structure and length of follow-up of the study cohorts. Each OncoSim-CRC trial-specific predicted reduction fell within the confidence intervals for the observed HR for CRC incidence and CRC mortality for the corresponding trial. The predicted ranges of HRs for incidence was 0.74–0.82 and for mortality was 0.66–0.76 for the four trials. ConclusionsOncoSim-CRC predicted reductions in CRC incidence and mortality agreed well with observed in RCTs of sigmoidoscopy screening.