Cervical cancer prevention : New indicators for the impact of health controls policies and individual risk assessment

Cervical cancer remains heterogeneous within the European Union (EU) in terms of transmission patterns, and prevention through vaccination and screening. As the prevalence of cervical cancer caused by vaccine-targeted HPV types is now decreasing with the implementation of vaccination programs, there is a growing need to adapt screening policies accordingly. In addition, emerging health technologies, such as HPV self-sampling, consideration of vaccination status, and epigenetic markers, are challenging traditional approaches to cervical screening in each country. In this work, we developed new indicators specific to cervical cancer and HPV genital infection to better capture the evolving impact of prevention programs, whether at the population level or through individual risk assessment. First, using global cancer registries, we modelled cervical cancer incidence and mortality through flexible methods that penalize deviations from average incidence, providing estimates that account for disease dynamics across countries. These rates were also be used to create reference curves through flexible quantile regression, allowing each country to assess its profile relative to others. To further support HPV vaccination strategies and help stakeholders identify both global and country-specific cervical cancer burdens, we developed a novel indicator analogous to life expectancy at birth. This estimates the expected number of cervical cancer cases in a cohort of girls based on current age-specific incidence rates and assess the number of cancers that could potentially be avoided in these cohorts, given a vaccination campaign effectiveness. One of the key challenges in modelling HPV transmission is the limited availability of sexual behaviour data, which is rarely captured in population studies. In contrast, HPV age-prevalence trajectories are more commonly reported in population surveys and directly reflect sexual behaviour, since biological probabilities of HPV transmission are assumed to be consistent across populations. To address this gap, we developed an unsupervised clustering algorithm for longitudinal binomial counts that accounts for heterogeneity. We assessed its validity and interpretability compared to models that do not account for heterogeneity and applied both approaches to global HPV prevalence data to identify population clusters. Following the development of the trajectory clustering model, we conducted a systematic review of HPV age-prevalence in EU countries from 2000 to 2023 and applied the clustering algorithm without heterogeneity. We identified 71 studies covering 94 populations across 20 countries, accounting for data from 1,567,274 women. Five distinct age-prevalence clusters were identified and highlight the importance of considering population heterogeneity in HPV transmission when modelling screening policy effectiveness. Moreover, in the absence of local sexual behaviour data, HPV transmission patterns could potentially be extrapolated from populations within the same trajectory cluster. Finally, emergence of new technologies allows the identification of host-cell epigenetic biomarkers in cervical lesions. We developed an unsupervised model using hypermethylation data, independent of the pathologist’s diagnosis, to construct a continuous latent score ranging from 0 to 1. By ranking the most informative epigenetic sites and guiding biomarker selection, this score can help better characterize precancerous lesions, particularly in cases where diagnostic discrepancies among pathologists are common. In conclusion, this work provides novel tools to better understand cervical cancer dynamics and HPV transmission patterns, supports vaccination and screening strategies, and contributes to the development of future precision screening approaches that can adapt to population heterogeneity and emerging technologies.