![[ Visit Client Website ]](images/banner.gif)
Monday, October 20, 2008: 2:45 PM
Grand Ballroom D (Hyatt Regency Penns Landing)
Purpose: To assess the performance of a microsimulation model of human papillomavirus (HPV) and cervical cancer by comparing model-projected outcomes to data from newly available studies.
Method(s): An empirically calibrated microsimulation model of HPV and cervical cancer has been used to evaluate health and economic outcomes of screening strategies, including HPV DNA testing. After the analysis was completed, data from three studies (two randomized trials of HPV DNA testing and one novel study of the long-term progression rate to invasive cancer) became available. We constructed large simulated cohorts, matched to study enrollment characteristics by age, background mortality, and screening history. Using multiple good-fitting parameter sets, we simulated each study to produce comparable model outputs and corresponding uncertainty ranges. Outcomes included the performance of the HPV DNA test in detecting high-grade cervical intraepithelial neoplasia (CIN) based upon concurrent infection with high-risk HPV, the reduction in risk of subsequent CIN after screening and treatment using HPV DNA testing, and the 5 to 30 year cumulative risk of invasive cancer among women with untreated high-grade CIN. Agreement was assessed based on overlap of model uncertainty ranges and study confidence intervals.
Results: The modeled HPV DNA test’s sensitivity and specificity for detecting high-grade CIN were 84.4% [range: 73.0 – 94.9] and 92.8% [range: 91.1 – 94.8], compared to study values of 94.6% [95% CI: 84.2 – 100.0] and 94.1% [95% CI: 93.4 – 94.8]. Adding HPV DNA testing to cytology detected more prevalent high-grade CIN in the model (relative rate 1.21 [range: 1.19 – 1.24]) and study (1.51 [95% CI: 1.13-2.02]) and reduced more incident cervical disease in the model (31% [range: 30 – 33]) and study (47% [95% CI: 2 – 71]). Modeled cumulative risk of developing invasive cancer for women with untreated high-grade CIN at 5, 10, 20, and 30 years was: 15.5%, 27.8%, 37.4%, 39.6% [ranges: 12.2 – 19.4; 22.2 – 34.5; 29.7 – 46.8; 30.9 – 49.7], compared to the study: 17.4%, 26.2%, 34.0%, 37.5% [95% CIs: 11.1 – 26.9; 18.4 – 36.5; 25.2 – 44.7; 28.4 – 48.3].
Conclusions: Concordance between the model and studies not used in model construction provides support for reasonable external consistency and projective validity. The performance of simulation models used for cost-effectiveness analyses should be iteratively re-evaluated as new data become available.
Method(s): An empirically calibrated microsimulation model of HPV and cervical cancer has been used to evaluate health and economic outcomes of screening strategies, including HPV DNA testing. After the analysis was completed, data from three studies (two randomized trials of HPV DNA testing and one novel study of the long-term progression rate to invasive cancer) became available. We constructed large simulated cohorts, matched to study enrollment characteristics by age, background mortality, and screening history. Using multiple good-fitting parameter sets, we simulated each study to produce comparable model outputs and corresponding uncertainty ranges. Outcomes included the performance of the HPV DNA test in detecting high-grade cervical intraepithelial neoplasia (CIN) based upon concurrent infection with high-risk HPV, the reduction in risk of subsequent CIN after screening and treatment using HPV DNA testing, and the 5 to 30 year cumulative risk of invasive cancer among women with untreated high-grade CIN. Agreement was assessed based on overlap of model uncertainty ranges and study confidence intervals.
Results: The modeled HPV DNA test’s sensitivity and specificity for detecting high-grade CIN were 84.4% [range: 73.0 – 94.9] and 92.8% [range: 91.1 – 94.8], compared to study values of 94.6% [95% CI: 84.2 – 100.0] and 94.1% [95% CI: 93.4 – 94.8]. Adding HPV DNA testing to cytology detected more prevalent high-grade CIN in the model (relative rate 1.21 [range: 1.19 – 1.24]) and study (1.51 [95% CI: 1.13-2.02]) and reduced more incident cervical disease in the model (31% [range: 30 – 33]) and study (47% [95% CI: 2 – 71]). Modeled cumulative risk of developing invasive cancer for women with untreated high-grade CIN at 5, 10, 20, and 30 years was: 15.5%, 27.8%, 37.4%, 39.6% [ranges: 12.2 – 19.4; 22.2 – 34.5; 29.7 – 46.8; 30.9 – 49.7], compared to the study: 17.4%, 26.2%, 34.0%, 37.5% [95% CIs: 11.1 – 26.9; 18.4 – 36.5; 25.2 – 44.7; 28.4 – 48.3].
Conclusions: Concordance between the model and studies not used in model construction provides support for reasonable external consistency and projective validity. The performance of simulation models used for cost-effectiveness analyses should be iteratively re-evaluated as new data become available.