It is clear that vaccination will eventually have a substantial impact on the rates of invasive cervical cancer and its precursors. However the relative impact on abnormal cytology is much less than that observed for cervical cancer and CIN3, due to the larger impact of other HPV types on this end point. The first effects will be seen in younger women, but even then it will take 10 or more years to be appreciable, especially if screening starts at age 25. In England we will not observe the projected benefit of vaccination on precursor lesions in women aged 20–24, as the screening programme only invites women over age 25. In fact (regardless of vaccination) we are likely to observe an increase in the number of abnormal smears and CIN3 detected during the first round of screening (at age 25) compared to historical data, due to prevalent lesions that might have been detected and removed if screening started earlier. However the effect should disappear by the second round of screening.
One of the main uncertainties in modelling the impact of vaccination is coverage. We have assumed coverage of 80% as this seems likely for a school-based programme, but estimates based on 70% coverage are also given. We have assumed that vaccination offers lifetime protection, but the current analysis only requires protection for the next 18 years (i.e. up to 2025). Data on the duration of protection offered by HPV vaccines are only available up to 7.3 years after vaccination (De Carvalh and Roteli-Martin, 2009
). These suggest no loss of protection, although the extent of protection remains unknown for longer periods of time and for cross-protection from HPV types not included in the vaccine.
Also, it is not clear how much efficacy will be lost if only two doses of the vaccine are received and also how many women will fall into this category. Preliminary data on antibody titres suggest two doses may be as effective as three (Dobson et al, 2009
). However, from experience with similar vaccines, most experts believe that a single dose is unlikely to be effective; therefore women who only received one dose are probably best grouped with unvaccinated women.
Our primary model has assumed no cross-protection, but we have also evaluated a model reflecting findings on cross-protection in the PATRICIA trial. Both models take into account that previous infection with one HPV type does not weaken the protection afforded by the vaccine to other types. We used CIN3 as one of our main end points. Another relevant clinical end point would be the reduction in women treated for cervical disease, as provided for in the PATRICIA trial. However the KC65 returns used to produce screening statistics for colposcopy episodes provide only data on treatment at first appointment. Because we have been unable to estimate the baseline number of women who get treated annually in England by age, it was therefore impossible to make this calculation.
Another major unknown is whether vaccinated women will reduce their attendance for screening. This is an important issue as screening and vaccination have benefits of similar magnitude, so the benefit of vaccination on cervical cancer incidence could be fully negated if vaccinated women choose not to be screened. Furthermore, based on associations with deprivation, it is likely that girls who are not vaccinated will, as women, be less likely to be screened. Although it is possible that some cultural or religious subgroups whose lifestyle puts them at low risk will choose not to participate in either programme, it seems likely that overall women who miss out on both programmes will have a greater disease burden than would be predicted by assuming that non-compliance to one intervention is independent of non-compliance for the other. In general non-screened women have a much higher risk than screened women so reaching this group remains a priority. Screening by self-sampling based on an HPV test may help to minimise the disease burden in this group, but developing methods to achieve a high vaccine coverage in deprived groups is also important. These issues will also be important for evaluating the joint cost-effectiveness of screening and vaccination programmes (Goldhaber-Fiebert et al, 2008
On a more positive note, it is likely that less frequent screening using HPV testing will become the norm, and the substantially greater level and duration of protection for this form of testing may reduce the disease burden further (Bulkmans et al, 2007
; Cuzick et al, 2008
; Dillner et al, 2008
). In the long term the combination of HPV vaccination and screening for HPV DNA promises to make cervical cancer a very rare disease and eliminate the need for frequent screening and high rates of colposcopic referral.