Estimated impact of human papillomavirus vaccines on infection burden: The effect of structural assumptions

Abstract

Mathematical models have been used to estimate the impact of human papillomavirus (HPV) vaccines on infection burden and cervical cancer. Models assume different mechanisms of naturally acquired immunity against re-infection, but processes of latency and reactivation of latent infection have not been explored. This study uses an individual-based dynamic model to simulate randomised controlled trials (RCTs) for vaccine efficacy, using different assumptions about naturally acquired immunity and viral latency after clearance of HPV infection. Model estimates of vaccine effectiveness are compared to those from published RCTs. We then estimate the impact of the bivalent vaccine on HPV-16 and -18 infection burden in South Africa under these different assumptions. When assuming no latency, simulated vaccine effectiveness overestimates results from RCTs and the model cannot match the observed difference in vaccine effectiveness between total vaccinated cohorts and more HPV-naïve cohorts. The reduction in HPV-16 and -18 burden by 2045, following roll-out of vaccination in 2014, does not depend on assumptions about natural immunity, but models that assume no latency predict ∼25% greater reduction in HPV-16 and -18 burden than models that include reactivation of latent infection for all men and women. Mathematical models that do not allow for reactivation of latent HPV infections may therefore overestimate the long-term impact of HPV vaccines.