Little is known about the time span of progression from high-grade cervical intraepithelial neoplasia (CIN2/3) to invasive cervical cancer. Estimation of this duration from longitudinal studies is not permitted, as CIN2/3 should be treated when detected. Cross-sectional data on the age-specific incidence of detected CIN2/3 and cervical cancer cases are readily available in national registries, but these data are difficult to interpret because neither the moment of lesion development nor the onset of invasive cancer is observed. We developed a statistical model for estimating the duration of time between CIN2/3 and preclinical cancer using Dutch national registries for the years 2000-2005. Human papillomavirus (HPV) genotype data were used to separate CIN2/3 and cancer incidences to obtain estimates for HPV-16-positive and HPV-16-negative lesions. The median time from CIN2/3 to cancer was estimated to be 23.5 years (95% confidence interval: 20.8, 26.6), and 1.6% of the lesions progressed to cancer within 10 years. The median duration for HPV-16-positive lesions was similar, but 2.4% of the HPV-16-positive lesions progressed to cancer within 10 years, as compared with 0.6% for HPV-16-negative lesions. Estimated durations of time to cancer are essential for reassessment of the optimal screening interval in light of vaccination and novel screening tests.
Keywords: cervical cancer; cervical intraepithelial neoplasia; disease progression; doubly censored data; human papillomavirus; mover-stayer model; natural history; registries.