Hughes and colleagues have previously estimated the per-partnership transmission probability of HPV to be 0.8 [
]. Our estimate of a similarly high transmission probability of HPV 16 per sexual partnership is based on the observed prevalence of antibody responses to this virus. The transmission probability contributes to the potential for spread of the infection and is one of the key variables in determining the basic reproductive number of the virus. A high transmission probability and consequent high basic viral reproductive number requires a high efficacy and high vaccine coverage to eliminate the infection [
]. How reliable is our high estimate of the transmission probability? While it is possible to construct narrow confidence intervals for our estimate, this can be done only by fixing the many other parameter values. However, we can be comfortable that the value lies above 0.4, because this estimated transmission probability emerges when we use the extreme values of lifetime reported partners as yearly new partners. If we have underestimated sensitivity or overestimated the loss of a specific antibody response, then the prevalence used in comparison to model estimates will be high, which would lower the estimated transmission probability.
Further, caution is required because our assumption of lifelong acquired protection is important. If protection is lost, infected individuals would move back into the susceptible class, and the epidemiology of HPV 16 would be very different, with a high observed prevalence and seroprevalence possible with a much lower basic reproductive number and transmission probability. This would greatly reduce the vaccine efficacy and coverage required for elimination with a vaccine that did generate immune protection. It is also important to note that our estimate is per sexual partnership, not per sex act. The transmission probability per partnership is obviously a function of the transmission probability per act and the number of sex acts, but to estimate the transmission probability per act requires either more detailed sexual behaviour or epidemiological data. While we did not explicitly explore the effect of condom use on our estimation for the transmission probability, a meta-analysis of 20 studies suggested that, while condoms may not prevent HPV infection, they may protect against genital warts, cervical intraepithelial neoplasia 2 or 3 and ICC [
]. Also, using data from the Finnish maternity cohort and the distribution of reported numbers of children per woman in the sexual behaviour component takes this limitation of condom use into account.
We explored whether the observed increase in cervical cancer incidence among Finnish women aged 35–39 y, from 3.7 cases per 100,000 women in 1985 to 7.7 per 100,000 in 1999 [
], could be explained by changes in sexual behaviour or increases in smoking. While some of the change could be explained through observed changes in sexual behaviour, these were not entirely sufficient. This suggests that other cofactors also contributed to the observed increase. For example, the decrease in age at sexual debut could expose those with immature cervical transformation zones (the region of the cervix where columnar epithelium replaces squamous epithelium with an existing predilection to HPV infection) to HPV infection [
], increasing susceptibility further and enhancing the effect of changes in sexual risk behaviour. Thus, decreased age at sexual debut could increase disease in young women in two ways: (1) through increasing time since infection and (2) through increased vulnerability of the immature transformation zone. The two would be difficult to differentiate epidemiologically, and we included only the former mechanism in our model. In the model, changes in rates of sexual partner change rather than the age at sexual debut had a greater effect on ICC incidence (unpublished data). This increased risk associated with sexual partners reflects risk of HPV infection rather than ICC risk once infected.
Also, it is notable in this context that a new HPV type (HPV 45) is present in 19% of the Finnish ICC cases appearing in fertile-aged women, i.e., especially those women who have experienced the increase of ICC incidence since early 1990′s (M. Lehtinen, unpublished data). It is possible that an increase in HPV 45 cancers accounted for some of the increase seen in the overall incidence for cervical cancer among women aged 35–39 y.
Including trends in tobacco smoking and an associated increase in progression to ICC improved the model prediction prior to 1979. However, since the percentage of women smoking between 1978 and 1995 was consistently approximately 20%, smoking trends do not explain changing incidence between 1985 and 1999. A more careful exploration of age-specific patterns of tobacco smoking, to look at cohort effects, may be warranted.
Oral contraceptive use and parity are aetiological cofactors that may also account for some of the cervical cancer incidence increase [
], whereas organised screening programmes for cervical cancer in Finland have been unchanged since 1990, and changes in registration or diagnostic practices are not thought important [
]. Changes in other cofactors, such as diet and other sexually transmitted infections, might also play a role in the increasing incidence of cervical cancer in Finland.
Screening was incorporated into the model by representing the same mass-screening coverage and efficacy rates reported in 1999 over time starting in 1950, because screening coverage has not changed in the last 40 y. In 1999 the Finnish national screening programme invited 80% of 30-y-old women to screening, screened 60% of those invited—that is, only 49% of those eligible [
]. Low participation in organised screening was found to correlate well with high or increased incidence of cervical carcinoma, so improved screening participation would reduce the incidence of cancer [
]. Among older women, aged 40–55 y, on average 72% of eligible women are screened. With the 75% increase seen in ICC among 35- to 39-y-olds, improved coverage to detect precancerous lesions in 30-y-old women is necessary. For simplicity, changing sexual behaviour practices and regular and/or sporadic screening done in the private sector were not modelled.
The impact of vaccination in our model is comparable with findings from previous modelling exercises [
], where high coverage of women alone, vaccination before sexual debut, and long-term protection (or boosters) providing three to four decades of protection are required to substantially reduce ICC incidence. A vaccine providing protection of less than 15 y may generate unexpected outcomes by shifting susceptibility in women to an older age group, where they could have increased risk of persistence. However, if age-specific patterns of persistence are derived from following cohorts of infected women, because older women infected at entry are more likely to have a persistent infection, it seems possible that an observed increase in persistent infection with increased age is an artefact. It is possible that type replacement with nonvaccine oncogenic HPV types could have a comparable effect, with a perverse increase in cervical cancer incidence. This could be avoided by including new oncogenic HPV types in the booster vaccines.
In developed countries with low cervical cancer incidence, vaccine coverage, in terms of both the target population and oncogenic HPV types, needs to be high in order to sustain the low ICC incidence associated with effective screening. Continued screening programmes have the potential to detect precancerous lesions in those not vaccinated and lesions associated with nonvaccine HPV types. The most effective strategy is vaccination combined with current screening protocols, compared to both screening alone and vaccination alone. However, cytological screening is costly, and in choosing a cervical cancer prevention strategy, health economic assessments of available options are warranted.