Impact of vaccination on 14 high-risk HPV type infections: a mathematical modelling approach

Abstract

The development of high-risk human papillomavirus (hrHPV) infection to cervical cancer is a complicated process. We considered solely hrHPV infections, thus avoiding the confounding effects of disease progression, screening, and treatments. To analyse hrHPV epidemiology and to estimate the overall impact of vaccination against infections with hrHPVs, we developed a dynamic compartmental transmission model for single and multiple infections with 14 hrHPV types. The infection-related parameters were estimated using population-based sexual behaviour and hrHPV prevalence data from Finland. The analysis disclosed the important role of persistent infections in hrHPV epidemiology, provided further evidence for a significant natural immunity, and demonstrated the dependence of transmission probability estimates on the model structure. The model predicted that vaccinating girls at 80% coverage will result in a 55% reduction in the overall hrHPV prevalence and a higher 65% reduction in the prevalence of persistent hrHPV infections in females. In males, the reduction will be 42% in the hrHPV prevalence solely by the herd effect from the 80% coverage in girls. If such high coverage among girls is not reached, it is still possible to reduce the female hrHPV prevalence indirectly by the herd effect if also boys are included in the vaccination program. On the other hand, any herd effects in older unvaccinated cohorts were minor. Limiting the epidemiological model to infection yielded improved understanding of the hrHPV epidemiology and of mechanisms with which vaccination impacts on hrHPV infections.

Figure 1. Data and Modelling Overview.

The parameters of the sexual contact structure were estimated from the School Health Promotion study , FINSEX 2007 study , and marriage statistics . The type-specific clearance of new infections was estimated from the control arm of PATRICIA phase III HPV vaccine study in Finland . The contact structure and the type-specific clearance rates were used as input for the single-type transmission models. The multiple-type transmission model ties together the single-type models and produces the hrHPV prevalence, which was fitted to the age-specific hrHPV prevalence data by calibrating three model parameters (transmission probability, natural immunity, and the clearance rate of persistent infections).

Figure 2. The Pattern of Sexual Contacts in Finland.

Upper panel: the age-specific annual mean numbers of new sexual partners by lifetime partner number with the observed numbers (asterisks). Lower panel: the age-specific stratification of the population by lifetime partner number and the corresponding data (asterisks).

Figure 3. Transmission Model Structure for a Single HPV Type.

The vertical flow corresponds to changes in the epidemiologic states susceptible (S), infectious (I), recovered (R), and vaccine-protected (V). The flow from left to right corresponds to an increasing lifetime partner number (n). The arrows describe possible transitions between different states: 1. Acquisition of a new partner without acquiring infection; 2. Acquisition of a new partner with acquiring infection (primary force of infection); 3. Acquisition of infection from the current partner (secondary force of infection, for n >0 only); 4. Clearance of infection; 5. Waning natural immunity; 6–7 and 10. Acquisition of a new partner for infected, recovered, and vaccine protected; 8. Take of vaccine protection; 9. Waning vaccine induced protection. The formulae for all transition rates are presented in File S1.

Figure 4. The Age-specific High-risk HPV (hrHPV) Prevalence in the Steady-state Before and After Vaccination.

Unless otherwise stated, the results pertain to females under the base-case scenario. (a) The model prediction on the current hrHPV prevalence (upper curve) with the observed data (asterisks). The lower curves show the prevalence for three different single hrHPV types with low, moderate and fast clearance of infection (see Materials and Methods); (b) the prevalence of hrHPV and persistent hrHPV before and after vaccination; (c) hrHPV prevalence at different times since the onset of the vaccination program; (d) hrHPV prevalence in females and males, before and after vaccination; (e) HPV16 prevalence under different vaccine scenarios, waning vaccine protection induces a second peak in the prevalence curve; (f) model fits to hrHPV data under different waning rates of natural immunity (base-case, SIR, 0.2 1/year waning rate) and the corresponding post-vaccination prevalences.