Results from this independent trial support the strong protective effect of Cervarix against 12-month HPV16/18 persistent infections in the ATP cohort (5
). Protection was close to 90% against these two types, which are responsible for approximately 70% of cervical cancers (17
). In addition, we observed nearly 50% cross-protection against HPV31/33/45, which are associated with approximately 10% of cancers. The VE against HPV16/18 was only 50% when we considered all vaccinated women (ITT) and just 12% when we considered persistent infections with any oncogenic HPV type, even in ATP cohorts.
For these analyses, we chose the surrogate outcome of persistent infection, which is highly reproducible (18
), unlike histopathologic end points emphasized in previous reports. In previous work we have reported from Guanacaste, we compared the relative reproducibility and validity of CIN2 and CIN3 diagnoses by comparing community pathologists’ diagnoses with two independent reviewers from the United States (total, N
= 357). Two review pathologists agreed with 84% and 81%, respectively, of initial diagnoses of CIN3 compared with 13% and 31% of CIN2. Although CIN3 is a substantially more reproducible diagnosis than CIN2, the latter constitutes an important fraction of lesions in reported clinical trials (13
). In addition, the virologic outcome provides direct assessment of causality in the presence of multiple infections and has a relatively high positive predictive value for subsequent development of lesions (19
). The ITT analyses incorporate the reality of incomplete vaccination in mostly sexually active adults and can be extrapolated to other populations of similar age, sexual behavior, and compliance. In contrast, most women in the ATP analyses are probably naïve to HPV infection, allowing extrapolation to women vaccinated before sexual debut and who comply with vaccination regimens.
We observed statistically significant cross-protection against HPV31/33/45 as a group. There was no apparent efficacy against the very common persistent infections with HPV types other than HPV16, 18, 31, 33, and 45, an association that attenuated the overall efficacy against persistent infections with all oncogenic types down to 12%. The nominally significant deleterious effect on HPV51 may be a chance finding among many comparisons made and was not observed in the other large trials of Cervarix (20
). The 4-year follow up of our study was too short to observe whether other HPV types replace vaccine types in vaccinated cohorts. Natural history data do not indicate that one HPV type modifies the epidemiology of the other (21
), but we did not investigate whether the presence of a nonvaccine type modifies the vaccine's protection against infection with HPV16 or HPV18.
Inclusion of women regardless of serostatus, which is imperfectly measured, allowed us to observe the full impact of the vaccine in a population, including presumably immune women. The ATP VE against HPV16 among women seronegative for HPV16 was 92.2%, approximately twice as high as that in seropositive women. The attack rate of persistent infection was lower in seropositive than seronegative women in the control arm, likely reflecting natural protection by serum antibodies and possibly other immune mechanisms (23
) or reduced exposure a few years after initiation of sexual activity. The greater attack rate of persistent infection among seropositive than seronegative women in the vaccine arm may reflect high proportions of missed infections (possibly as the result of inadequate sampling of the genital tract, missed test results, or latent infections) in women who do not benefit from vaccination because they were infected before baseline. The absence of reduction in efficacy against HPV16 persistent infection among HPV18 seropositive women suggests that immune protection, rather than other correlates of sexual activity associated with antibody levels, explains the effect.
Similar efficacy against persistent infection with HPV16/18 in ATP analysis by age indicates that the vaccine is effective at protecting against new infections in unexposed women independent of age. The strong decrease in efficacy from 68% at ages 18 to 19 to 21% at ages 24 to 25 in the ITT cohort probably reflects that in the latter, there is a significantly larger fraction of women who have initiated sexual activity before vaccination and have been exposed to HPV. Rate reductions also clearly decrease with age and years since first intercourse, with the exception of virgins, who do not contribute time at risk until they start sexual activity. It should be noted, however, that the reduction in VE and rate reductions is only present in the group of women 24 to 25 years old. Interestingly, rate reductions tend to be greater among women with more partners (because their attack rate is greater) despite lower VE.
These results indicate that both susceptibility and rates of transmission are important parameters when the potential impact of prophylactic vaccines is assessed and have implications for vaccination efforts and screening policy. In the absence of a test to determine expected benefit of an individual woman, age appears to be clearly a criterion to consider for definition of target groups for vaccination. The decreases in estimates of VE seen with increasing age and time since sexual debut suggest that many infections that could eventually progress to cancer occur early and can only be prevented with adolescent vaccination.
The observation that vaccination did not substantially reduce the incidence of oncogenic infections has implications for screening programs. The positive predictive value of the tests will likely be reduced because many of the persistent infections by nonvaccine HPV types are unlikely to progress to significant lesions. The lack of reduction in infections with the lesser oncogenic types can lead to more diagnostic and therapeutic procedures than necessary in vaccinated cohorts.
Interestingly, we noted that VE against HPV16 and HPV18 tended to increase with time since vaccination, to 100% in the ATP cohort and to almost 95% in the ITT cohort, with a similar effect for the combined outcome of HPV31, 33, and 45 (to a maximum close to 60%). One possible explanation of increasing efficacy against persistent HPV16/18 infections with time since vaccination in the ATP cohort is waning influence of false-negative baseline HPV DNA results, for which efficacy is zero or low. This interpretation is supported by the reduced VE against HPV16 observed among women who were seropositive for anti-HPV16 antibodies. Similarly, the likely explanation in the ITT cohort is waning influence of baseline prevalent infections. In ITT, increased VE with time since vaccination reflects protection against new infections, but the impact of this protection in the out years needs to be interpreted in the context of the fact that exposure tends to be greatest early on after initiation of sexual activity, with reduced exposure typically observed with increasing age/time.
Most of the findings, including the stratified analyses, were similar when a 6-month HPV persistence end point was used, with the advantage that the number of end points was larger, indicating that 6-month persistence could serve as an adequate surrogate end point in HPV vaccine trials, particularly for the evaluation of HPV types that occur with lower frequency or have lower vaccine efficacies that require larger sample sizes to achieve statistical significance. In this study, for example, VE against HPV45 was not significant when the 12-month end point was used but was clearly so when the 6-month end point was used.
This analysis has some limitations and strengths. One of the limitations is that we had a relatively small sample size to accurately assess the lower efficacy of individual nonvaccine HPV types, as has been the case with other clinical trials of Cervarix (5
). Those multicentric trials as well as those reported for Gardasil recruited smaller number of women in multiple research centers. In contrast, the Costa Rica HPV trial was conducted in a homogeneous population of young women at high risk of HPV infection. In this context, the results can be extrapolated to similar groups of women in areas of high HPV prevalence. It should be noted, however, that a high prevalence of HPV in young women is very common in most areas of the world, particularly those in which the vaccine is being considered to control the problem of cervical cancer. Differences in sexual practices, in particular the distribution of age at first intercourse in the population, should be taken into account when designing HPV vaccination programs.
One of the strengths of this study is that it is a large trial in a stable community, which will allow long-term follow-up up of these cohorts. Moreover, the fact that the results of this trial are very similar to those obtained in the multicentric trials points to the generalizability of VE results. The fact that participation rates at enrollment were limited could also affect the external validity of the results but not the internal validity of the randomized trial. We used virological outcomes, which have some advantages because they are highly reproducible and do not present problems for causality assessment in the presence of multiple infections. However, the clinical significance of virological outcomes, particularly for nonvaccine types, is still under active debate.
In conclusion, the clear benefit of Cervarix against persistent HPV16/18 infections observed among unexposed women decreases with age and sexual experience. These findings, together with extensive data indicating that HPV is acquired early on after sexual debut (24
) and the possibility of natural immunity (23
), suggest limited value, in general, for vaccination beyond a few years after adolescence in areas of high prevalence of HPV infection and high risk of cervical cancer. Efforts that focus vaccination on women before sexual debut may be most effective at reaching the most vulnerable groups.