Human Papillomavirus Vaccination For Adults Aged 30 To 45 ... - PLOS

Discussion

This comparative analysis evaluated the cost-effectiveness of increasing the upper age limit for HPV vaccination utilizing 2 well-validated and comprehensive modeling platforms. We found that vaccination to older ages (30, 34, 40, 45 years) was inefficient or associated with unfavorable ICERs in the US context. Sensitivity analysis revealed that assumptions about cervical screening compliance, vaccine costs, and the natural history of noncervical HPV-related cancers could have major impacts on the estimated cost-effectiveness of the vaccination strategies. However, even with the most extreme assumptions, both models almost universally found that the cost-effectiveness of HPV vaccination for adults aged 30 to 45 years was greater than $200,000 per QALY. The only exception was in the scenario assuming imperfect screening compliance, where one model found that vaccinating up to ages 40 and possibly 45 years could be cost-effective if assuming an upper-bound willingness-to-pay threshold of $200,000 per QALY.

It should be noted that the strategies of vaccinating adults aged 30 to 45 years were assessed in the context of 2 prevention strategies already well underway in the US that serve to limit the incremental gains that can be achieved: HPV vaccination in younger cohorts of females and males, which provides some herd immunity benefits for the older unvaccinated cohorts, and cervical screening for women. Additionally, vaccination at older ages can only be effective in those who are susceptible to infection (i.e., those not currently infected but exposed to new infections as they age), and yet the probability of new exposures decreases with age [48]. Furthermore, due to the long median dwell time between infection and cancer development, the benefits of vaccination tend to be accrued several decades into the future and are thus discounted in cost-effectiveness analysis.

A strength of this analysis is that we have employed 2 independent, well-validated models that have been used for a number of evaluations in different settings. Where we vary in inputs, assumptions, or structure reflects the uncertainty in the mechanism of disease or data. This comparative analysis was also done in the context of CDC requesting several independent modeling groups to contribute to the ACIP deliberations. Despite variations across all models, we found that conclusions were qualitatively consistent among those models that were not industry-funded [25,49].

Additionally, this evaluation has been able to harness the extensive work done by the CISNET-Cervical consortium to obtain and standardize empirical data for the US, which were used to inform model inputs. The CISNET-Cervical consortium has focused on detailed modeling of cervical screening recommendations and outcomes; in our sensitivity analysis, both groups considered an imperfect cervical screening scenario, in which both models incorporated detailed empirical data for screening and referral to diagnostic evaluation, and simulated the full recommended range of downstream surveillance and treatment outcomes for screen-positive women. This detail is important because prior evaluations have shown that critical cost savings from vaccination are accrued through the avoidance of referrals for management of screen-detected abnormalities and the range of complex follow-up and surveillance sequalae that such referrals routinely generate [50,51]. On the other hand, cervical screening is a very effective mechanism for secondary prevention, with the absolute effectiveness increasing, particularly in women over the age of 25 to 30 years. In general terms, the impact of secondary prevention of cervical cancer is critical to incorporate, and the CISNET models were able to achieve this in great detail.

Our analysis has several limitations. Where possible, we erred in the direction of making assumptions that were favorable to increasing the age of vaccination, including assuming no delay between the reduction in HPV infections from vaccination and the reduction in genital warts. A similar “incidence-based” approach was used for noncervical cancers by both models, in which we assumed that there could be an impact of HPV vaccination on incidence of noncervical cancers as early as 5 years after vaccination. This minimum lag time was consistent with Chesson and colleagues [24,25] but may be longer in reality. Such assumptions about short dwell times serve to improve the estimated cost-effectiveness of adult vaccination since the health benefits are realized earlier (and therefore not heavily discounted). Unfortunately, evidence to support modeled assumptions about the time between infection and cancer development is very limited. In the case of cervical cancer, where we have leveraged relatively rich data to inform the natural history dwell time from causal HPV infection to invasive cervical cancer, both Harvard and Policy1-Cervix models have estimated the median dwell time to be roughly 26 years [26,27]. Noncervical cancers are typically diagnosed at an older age than cervical cancer (10 to 20 years later) [21]; although cervical screening allows for earlier diagnosis of invasive cancer, this might suggest that the period of time between a causal HPV infection and cancer diagnosis is longer for many noncervical cancers. In sensitivity analysis, both models examined the impact of extending the minimum dwell time to 40 years for noncervical cancers, which, as expected, greatly diminished the cost-effectiveness of vaccination up to age 45 years. Data on the natural history, namely the carcinogenic progression, of the noncervical cancers are uncertain. To assess this uncertainty in more detail, we conducted extensive sensitivity analysis on the prevaccination incidence of the noncervical diseases, the percent of each disease that is attributable to HPV, the survival rate of each cancer, and disutilities associated with each of the HPV-related diseases, and found that vaccination up to ages 30 or older was not cost-effective. However, when more reliable data become available, it will be important to revisit the analyses and update results.

We did not take into consideration potential changes in the future burden of the noncervical cancers (other than through vaccine impact) and instead assumed the current underlying age-specific incidence and mortality rates remained constant over time. To the extent that the incidence rates of these cancers are rising (e.g., oropharyngeal cancers in men), we may be underestimating the overall benefit of HPV vaccination, although it is not clear if the incremental costs and benefits between the different age thresholds for vaccination would be altered.

We also made the favorable assumption that no individuals received less than the recommended dosage. If incomplete dose-course provides reduced or no effectiveness, then the cost-effectiveness of vaccination would be reduced. Furthermore, we made the favorable assumption that HPV-9 was 95% effective at preventing new vaccine-targeted HPV infections in both females and males up to age 45 years and that this protection lasted over their lifetime. There is no trial evidence directly reporting HPV-9 effectiveness in mid-to-older adult females or males. To date, 2 large trials have shown that vaccination with HPV-2 or HPV-4 for females up to age 45 years is somewhat effective at preventing incidence of persistent HPV infection after 4 years of follow-up [8,9]. Additionally there are no trials indicating direct effectiveness of any HPV vaccine in males beyond age 26 years.

The ACIP decision to not recommend catch-up vaccination beyond age 26 years took into consideration the results from our models, which predicted additional health gains when extending HPV vaccination to older ages, but at a disproportionately higher cost. The results from our 2 independent models suggest that HPV vaccination for adult women and men aged 30 to 45 years is unlikely to represent good value for money in the US.