This study provides helpful inputs on French public health policy on HPV vaccination. However, we would like to highlight the fact that two critical assumptions were neither explored nor discussed in the paper, although they are likely to strongly impact the results.
First, there is high uncertainty in the natural immunity of HPV (i.e. whether patients who clear HPV after a natural infection remain at risk of HPV infection or not). The assumption made on natural immunity has a direct impact on the choice of model structure, which can range from Susceptible-Infected-Susceptible (SIS) models (assumption of no natural immunity) to Susceptible-Infected-Removed (SIR) models (assumption of full immunity following natural infection). With regard to HPV, natural immunity does not seem to play a role in controlling re-infections in women [1]. Furthermore, it has been recently shown that prevalent serum levels (following natural infection) are not indicative of reduced risk of subsequent HPV 16 infection in men [2]. Model calibration ensures that the model produces some reliable epidemiological estimates in terms of incidence, prevalence and/or mortality following infection; however, a SIR model will require higher transmission rates than a SIS model to fit the observed data, as a proportion of the population becomes naturally immune and is therefore not contributing to the transmission of the disease. Hence, a model assuming higher levels of natural immunity will mechanically predict a lower vaccine effect. Baussano et al. (2011) [3] explored the implications of using different structural assumptions when assessing the impact of HPV vaccination on lifetime incidence of cervical cancer. The SIS model predicted that the vaccine would lead to higher rates of reduction than the SIR model across the scenarios. When focusing on the vaccination of 12-year old girls in an unscreened environment, the SIS model predicted a reduction of the lifetime incidence of cervical cancer which was 16% and 21% higher than the SIR estimates in the unvaccinated and vaccinated cohorts respectively. Depending on the vaccination strategy and clinical assumptions (e.g. duration of vaccine protection), the differences between models could be as high as 30% of expected cases. Van de Velde et al. (2010) [4] recently reported similar findings. As the authors used an SIS approach, it would have been nice to discuss the corresponding implications.
Second, the study assumed an annual vaccine coverage rate of 30% in 14-19 year old girls and 10% in 20-24 year old women. Although these coverage rates may sound low, they are annual coverage rates should not to be confused with the cumulative vaccine coverage rate, which corresponds to the proportion of the target population that has ever received the vaccine at any point in time.
Overall, it is a very interesting piece of work assessing the effect of HPV vaccination in France. We just felt that the major assumptions in the study should be clearly stated as they are likely to result in an overestimation of the effect of vaccination or a misinterpretation of the results. It would have been very useful and informative to run some additional sensitivity analyses in order to better understand the impact of these assumptions.

Aline Gauthier and Yiling Jiang
Amaris, London, UK

[1] Trottier H, Ferreira S, Thomann P, Costa MC, Sobrinho JS, et al (2010) Human papillomavirus infection and reinfection in adult women: the role of sexual activity and natural immunity. Cancer Res 70: 8569-77.
[2] Lu B, Viscidi RP, Wu Y, Lee JH, Nyitray AG, et al (2012) Prevalent serum antibody is not a marker of immune protection against acquisition of oncogenic HPV16 in men. Cancer Res 72: 676-85.
[3] Baussano I, Garnett G, Segnan N, Ronco G, Vineis P (2011) Modelling patterns of clearance of HPV-16 infection and vaccination efficacy. Vaccine 29: 1270-7.
[4] Van de Velde N, Brisson M, Boily MC (2010) Understanding differences in predictions of HPV vaccine effectiveness: A comparative model-based analysis. Vaccine 28: 5473-84.