The natural history of cervical cancer is only partly known, as most precancerous lesions are either treated or regress naturally prior to becoming clinical. Also, data on test characteristics is lacking because randomized controlled trials are never performed. However, microsimulation models allow us to mimic the natural history of cervical cancer. Calibrating unknown parameters in these models, such as durations of disease states, test characteristics and demographic assumptions, to observed data provides us with more insight into unobservable processes. Recently, large cohort studies have been added to this observed data providing new possibilities for analyses. In this study, we explored what could be learned from a microsimulation model about test characteristics when screening for cervical cancer.
The established MISCAN-Cervix microsimulation model was calibrated to the Dutch setting based on the latest data from the Netherlands Cancer Registry (NCR), the nationwide network and registry of histo- and cytopathology in the Netherlands (PALGA) and recently published cohort studies on HPV detection rates. To identify any differences in HPV-test sensitivity by disease grade, sensitivity was calibrated for each disease grade separately (i.e. cervical intraepithelial neoplasia (CIN)1, CIN2, CIN3 and cervical cancer). To test if false-negative tests were more likely to be attributable to the same (hard to reach) lesions, we allowed for the chance of systematically missing individual lesions with cytology testing in the model.
We found that the model fitted the observed data best when the sensitivity of HPV tests increases by disease grade. Especially the sensitivity for CIN1 was considerably lower than for higher CIN grades. For cytology testing, the fit of the model was best if we allowed for systematically missing about 11% of individual lesions.
The model was much better able to fit the observed data when adjusting HPV test sensitivity by disease grade and systematically missing a percentage of lesions at cytology testing. A possible explanation for an increasing sensitivity of HPV-tests by disease grade is that the infection persists for a longer period of time, was able to spread more, and therefore more easily detected. Systematically missing lesions at cytology testing is probably caused by the fact that some lesions are harder to reach with the cervical smear brush or spatula, such as adenocarcinoma. These findings could have clinical implications for screening practices as the screening guidelines depend heavily on test characteristics.