The natural history of human papillomavirus (HPV)-induced cervical cancer is largely unobservable, yet the length of time spent with preclinical disease impacts the effectiveness of alternative screening policies. Mathematical simulation models are increasingly being used to capture the complex natural history processes to project the health benefits and economic consequences of alternative cervical cancer prevention approaches.
The cervical cancer working group of the Cancer Intervention and Surveillance Modeling Network (CISNET) represents four independently developed microsimulation models of the natural history of cervical carcinogenesis. We aimed to perform a comparative modeling exercise to explore the differences in natural history of cervical cancer by characterizing the age of acquisition of the causal HPV infection and identifying the implied dwell times for distinct preclinical phases of cervical disease among women that developed cervical cancer.
We used the four CISNET-cervix microsimulation models (Cancer Council New South Wales, Erasmus Medical Center, Harvard, and University of Minnesota) to project outcomes for a hypothetical cohort of individuals. The natural history models were calibrated to match observed data on age-specific HPV prevalence and HPV type distribution in precancer and cancer, but varied in their underlying structure and assumptions of the carcinogenic process. For women with a cervical cancer diagnosis and in the absence of screening or vaccination, we calculated the age of acquisition of the causal HPV infection, and dwell times associated with three phases of cancer development: 1) “HPV dwell time”, defined as the time from the acquisition of an HPV infection to development of a high-grade precancer, 2) “precancer dwell time”, defined as the time from the development of a high-grade precancer to asymptomatic cancer development, and 3) “sojourn time”, defined as the time from asymptomatic cancer development to clinical detection. Conditioned on developing cancer, we enumerated these estimates for: 1) all high-risk HPV infections, 2) HPV-16, 3) HPV-18, and 4) other non-HPV-16/18 genotypes.
Our findings have important implications for prevention policies, for example catch-up vaccination, vaccination at older ages (HPV-FASTER) and screening interval. The comparative analyses address concerns about model transparency and highlight important structural differences. As the complexity of microsimulation models increases, understanding the impact of differences between model structures can elucidate important drivers of cervical cancer policy and provide guidance for areas of future research.