Human Papillomavirus Vaccine Effectiveness and Herd Protection in Young Women

The study population comprised young women recruited from a university-affiliated, hospital-based primary care clinic (Cincinnati Children’s Hospital Teen Health Center) and the Cincinnati Health Department (obstetrics and gynecology clinic and a sexually transmitted disease clinic). The study was approved by the hospital’s and health department’s institutional review boards, and participants provided written informed consent. Parental consent was waived for participants <18 years of age to protect patient confidentiality because history of sexual contact was an inclusion criterion. To participate, individuals had to be 13 to 26 years of age and sexually experienced, which was defined as having had sexual contact (oral-genital or genital-genital, with a male or female partner).²² Individuals were not eligible if they had participated in previous surveillance studies. Participants were enrolled by using a sequential recruitment strategy; 95% to 98% of those approached in each study agreed to participate. We collected the following 4 waves of data: wave 1 (2006–2007, N = 371), wave 2 (2009–2010, N = 409), wave 3 (2013–2014, N = 400), and wave 4 (2016–2017, N = 400). A survey instrument was used to collect data on sociodemographic and behavioral variables that may be risk factors for type-specific HPV infection; details about survey development and validity are described in previous articles.^15,22 Cervicovaginal swabs were collected by self-swab or clinician swab from each female participant, and samples were genotyped for HPV by using a Roche Linear Array test, a polymerase chain reaction amplification technique that uses an L1 consensus primer system and a reverse line blot detection strip to identify 36 HPV genotypes.²³ In previous studies, researchers have demonstrated that the results of self- and clinician testing for vaginal HPV in women are highly concordant.^24,–26 Vaccination status was defined as having received at least 1 HPV vaccine dose before enrollment, and it was assessed by using the Ohio statewide immunization registry and the electronic health record, both systems with high reliability.²⁷ Documentation in at least 1 of these systems was available for 98% of the participants who were vaccinated. Vaccination status was determined by a positive record in either or both databases and, in the small number of cases in which no information was available, by self-report.

We first examined participant characteristics and behaviors using descriptive statistical methods. We then determined if there were differences by study wave in participant characteristics using univariable methods (eg, χ² test, Fisher’s exact test, Kruskal-Wallis test, or analysis of variance). Because there were between-wave differences for some variables, we conducted a propensity score analysis adjusted by inverse probability of treatment weighting. This adjusts for selection bias, ensuring that any differences noted in HPV prevalence across the study waves were due to HPV vaccine introduction instead of differences in measured participant factors, which may have been confounders. These factors included sociodemographic characteristics, gynecologic history, sexual history, and enrollment site (see Table 1 for all factors included). We then determined HPV prevalence for all women in the study, women who were vaccinated, and women who were unvaccinated across the 4 study waves; these results were unadjusted and adjusted for the propensity score. Methodologic details are described in previous articles.^15,22 Total vaccine effectiveness (the relative infection risk in individuals who were vaccinated versus the infection risk in individuals who were unvaccinated before the introduction of a vaccine) was assessed by comparing vaccine-type HPV prevalence in women who were vaccinated in waves 2, 3, and 4 with women in wave 1, all of whom were unvaccinated.²⁸ 

Finally, we conducted logistic regression analyses to determine the odds of HPV prevalence across the study waves, unadjusted and adjusted for the propensity score. The independent variables were study waves (eg, wave 4 versus wave 1), and the dependent variables were prevalence of at least 1 of the HPV types included in the 9-valent vaccine (HPV-6, -11, -16, -18, -31, -33, -45, -52, and -58), at least 1 of the HPV types included in the 4-valent vaccine (HPV-6, -11, -16, and -18), or at least 1 of the 5 HPV types in the 9-valent but not the 4-valent vaccine (-31, -33, -45, -52, and -58). Separate models were estimated for all women in the study, those who were unvaccinated, and those who were vaccinated. Data were analyzed by using SAS software version 9.4 (SAS Institute, Inc, Cary, NC).

A total of 1580 participants were enrolled across the 4 studies. Participant characteristics are shown in Table 1. The range of mean ages across all waves was 18.7 to 19.1 years. Between 62.5% and 76.8% of participants were recruited from the Teen Health Center and 23.3% to 37.5% were recruited from the health department. The majority of participants (69.9%–77.3%) identified as African American or multiracial, 48.9% to 56.8% reported at least 2 male lifetime sexual partners, 26.3% to 38.5% used condoms during their last sexual intercourse, and 16.9% to 31.8% had smoked at least 100 cigarettes in their lifetime. The rate of HPV vaccination (defined as receipt of at least 1 dose) increased from 0% in wave 1 to 59.2% in wave 2, 71.5% in wave 3, and 84.3% in wave 4. Virtually all participants in waves 1 to 3 received the 4-valent vaccine; in wave 4, 88% received the 4-valent and 12% the 9-valent vaccine. Overall, 97% of participants received the 4-valent vaccine.

A propensity score analysis was performed to adjust for between-wave differences in participant characteristics (Table 1). The P values after propensity score adjustment were all >.05, and the standardized differences for most of the variables were <20%, indicating that variables were successfully balanced.

The proportions of women across the 4 study waves who were positive for 9-valent HPV types, 4-valent HPV types, and the 5 HPV types included in the 9-valent but not the 4-valent vaccine are shown adjusted for propensity score in Table 2 and in Figs 1–3. Among women who were vaccinated, the proportion infected with ≥1 9-valent vaccine–type HPV decreased from waves 1 to 4 from 46.6% to 13.5% (71% decline), the proportion infected with ≥1 4-valent vaccine–type HPV decreased from 35.0% to 6.7% (80.9% decline), and the proportion infected with ≥1 of the 5 types in the 9-valent but not the 4-valent vaccine decreased from 23.4% to 7.3% (68.8% decline); all were statistically significant. Among women who were unvaccinated, the proportion of women infected with ≥1 9-valent vaccine–type HPV decreased from 43.4% to 42.1% (3.0% decline; nonsignificant) from waves 1 to 4, the proportion infected with ≥1 4-valent vaccine–type HPV decreased from 32.4% to 19.4% (40.1% decline; significant), and the proportion infected with ≥1 of the 5 types in the 9-valent but not the 4-valent vaccine increased from 22.9% to 36.1% (57.6% increase; nonsignificant). Vaccine effectiveness was 71.7% in wave 2 versus wave 1, 90.6% in wave 3 versus wave 1, and 80.1% in wave 4 versus wave 1.

The logistic regression model results are shown in Table 3. In these models, we demonstrate the odds of infection (in waves 4 vs 1, 4 vs 2, and 4 vs 3) with 9-valent vaccine–type HPV, 4-valent vaccine–type HPV, and the 5 additional types in the 9-valent vaccine by vaccination status. Among women who were vaccinated, the odds of infection decreased significantly from wave 1 to wave 4 for 9-valent vaccine–type HPV (adjusted odds ratio [aOR] 0.18), 4-valent vaccine–type HPV (aOR 0.13), and the 5 additional HPV types included only in the 9-valent vaccine (aOR 0.26). Among women who were unvaccinated, the odds of infection did not change significantly from wave 1 to wave 4 for 9-valent vaccine–type HPV; it decreased significantly from wave 1 to wave 4 for 4-valent vaccine–type HPV (aOR 0.50) and increased significantly for the 5 additional HPV types only in the 9-valent vaccine (aOR 1.90).

To our knowledge, this is the first study in which trends in 4-valent and 9-valent vaccine–type HPV among women who are vaccinated and women who are unvaccinated in the United States >10 years after HPV vaccine introduction are examined, and it is the first in which trends after 9-valent HPV vaccine introduction are examined. Consistent with our hypotheses, we found that from 2006 to 2017, the prevalence of 4-valent vaccine–type HPV and 9-valent vaccine–type HPV decreased significantly among women who were vaccinated, and the prevalence of 4-valent vaccine–type HPV decreased significantly among women who were unvaccinated. In exploratory analyses, we demonstrated a significant decrease in the 5 additional types targeted by the 9-valent vaccine among women who were vaccinated and, unexpectedly, a significant increase in the 5 additional types targeted by the 9-valent vaccine among women who were unvaccinated.

The significant decline (81%) in 4-valent vaccine–type HPV in women who were vaccinated and the high degree of vaccine effectiveness when comparing women who were vaccinated with women who were unvaccinated (90.6% in wave 3 versus wave 1 and 80.1% in wave 4 versus wave 1) provide evidence of direct protection by the 4-valent vaccine among the women who were vaccinated in this community, and it suggests high vaccine effectiveness in a real-world setting. This degree of effectiveness is remarkable given the fact that vaccination was defined as having received ≥1 dose (ie, was not defined as having completed the vaccination series) and that women in this study were likely at a substantially higher risk for preexisting HPV infection than those in the HPV vaccine clinical trials because of their reported sexual behaviors. The high efficacy of the licensed HPV vaccines and high rates of vaccination in this study sample likely contributed to this substantial decrease in HPV infection, even in young women at a high risk for sexually transmitted infections (STIs). Our findings of a decrease in vaccine-type HPV among women who were vaccinated extend the findings of those studies conducted over different time frames and in different populations and settings,^{16,–19,21,29} and they support the real-world effectiveness of the 4-valent vaccine, especially in younger age groups and countries with high vaccination coverage.^16,21 This decline in HPV prevalence is expected to result in a significant decrease in cervical precancers and cancers in the future. Of note, the findings should not be interpreted as suggesting that current recommendations for a 2- or 3-dose series and vaccinating before sexual initiation are not necessary. In additional exploratory analyses, we have demonstrated that receipt of 1 vs 3 doses was associated with 3.2 times the adjusted odds of vaccine-type HPV infection (P = .04) and that young women who did versus did not have sex before vaccination were more likely to be positive for vaccine-type HPV (8.8% vs 3.8%, P = .0021).

We also demonstrated a significant decline among women who were vaccinated in the prevalence of 9-valent vaccine–type HPV (71%), as hypothesized, and in the prevalence of the 5 additional HPV types included in the 9-valent vaccine (68.8%). All 5 additional types in the 9-valent vaccine are genetically related to HPV-16 (HPV-31, -33, -52, and -58) and HPV-18 (HPV-45). In clinical trials, researchers have demonstrated evidence of crossprotection against HPV types genetically related to HPV-16 and -18,^14,30,–32 and in studies conducted in clinical and community settings, researchers have also demonstrated evidence of crossprotection against HPV types genetically related to the 2-valent or 4-valent HPV vaccines^33,–35; however, it is not well established whether crossprotection occurs among younger women recruited from broader settings or after introduction of the 9-valent HPV vaccine. In this study, only 12% of women in wave 4 received at least 1 dose of the 9-valent vaccine; therefore, it is unlikely that these findings were due only to direct protection against the types in the 9-valent vaccine among women who were vaccinated. Instead, the decreases noted in 9-valent vaccine–type HPV and the 5 additional types in the 9-valent vaccine may represent evidence of crossprotection against genetically related types (-31, -33, -45, -52, and -58). In a separate analysis, we are examining trends in nonvaccine-type HPV in more depth by exploring the prevalence of nonvaccine-type HPV genetically related to HPV-16 and HPV-18 among women who received the 4-valent vaccine to examine for crossprotection and by exploring non–vaccine-type HPV genetically unrelated to vaccine-type HPV to examine for type replacement (C.S., L.D., D.B., et al, unpublished observations).

There was a significant decline of 40.1% in the prevalence of 4-valent vaccine–type HPV among women who were unvaccinated, suggesting herd protection. Evidence of herd protection was found in our previous study during the first 3 waves of data collection.¹⁵ In addition, evidence of herd protection was found in 2 studies conducted in Scotland among adult women attending cervical cancer screening,^33,35 in studies in Australia among adult women attending cervical cancer screening¹⁹ and recruited through a social networking site,¹⁸ and in the United States before 9-valent vaccine introduction.^29,36 As noted in a recent review,³⁷ evidence about herd protection will be a key component of cost-effectiveness analysis evaluating cervical cancer screening strategies.

We unexpectedly found a significant increase from 2006 to 2017 in the prevalence of the 5 additional types in the 9-valent vaccine among women who were unvaccinated. A theoretical explanation could be type replacement, which would be an increase in non–4-valent vaccine–type HPV created by an ecological niche after the 4-valent vaccine was introduced. However, this phenomenon is thought to be unlikely given the genetic stability of HPV and that HPV types do not seem to compete for trophic epithelial niches.³⁸ A more likely explanation is differences between women who are unvaccinated and women who are vaccinated. For example, if women who are unvaccinated versus women who are vaccinated are more likely to practice riskier behaviors that would increase their risk of acquiring HPV, they would be more likely to acquire non–vaccine-type HPV. In a previous analysis of the first 3 waves of data from this study, we found that women who were unvaccinated did differ significantly from women who were vaccinated in ways that could increase their risk for HPV acquisition, supporting this explanation.³⁹ For example, women who were unvaccinated versus women who were vaccinated were more likely to lack health insurance and to have had at least 1 new sexual partner in the past 3 months. Continued community-level research is needed to determine if this trend reverses and if the prevalence of the 5 additional types included in the 9-valent vaccine begin to decline once a higher proportion of young women have received the 9-valent vaccine. In addition, the findings of differences between women who are unvaccinated and women who are vaccinated underscore the importance of understanding predictors of nonvaccination and designing clinical interventions to reach those youth who are unvaccinated and who may be at an elevated risk for HPV.

A limitation of this study was the small proportion of women who received the 9-valent vaccine in wave 4, which limited our ability to examine the trends in HPV prevalence due to the effects of the 9-valent vaccine. Also, the clinic-based recruitment strategy could limit generalizability to all young women in this age group in the United States. Instead, the sample could be viewed as generalizable to a group of adolescent and young adult women at a relatively high risk for HPV and other STIs and with a fairly high vaccination rate. Finally, risk behaviors were assessed by self-report, which may limit validity.

Eleven years after the introduction of the HPV vaccine, we noted significant decreases in 4-valent vaccine–type HPV, 9-valent vaccine–type HPV, and the 5 additional HPV types included only in 9-valent vaccine among women who were vaccinated, suggesting 4-valent vaccine effectiveness in a real-world setting and possible crossprotection against genetically related HPV types. The significant decrease in 4-valent HPV types among women who were unvaccinated suggests herd protection. Although these findings are important for clinical care and public health policy, continued surveillance will be important to assess for waning vaccine effectiveness, herd protection, and the impact of 9-valent vaccine introduction.

aOR

adjusted odds ratio

HPV

human papillomavirus

STI

sexually transmitted infection

We acknowledge the support provided by Charlene Morrow, RN, Susan Glynn, BA, Rachel Thomas, BS, and Lisa Higgins, RPh, as well as the support of the Cincinnati Children’s Hospital Teen Health Center and Cincinnati Health Department staff.