Methods

End-point definitionsDefinitions for infection and disease end points can be found in the accompanying article by Wheeler et al. [14]. The 6-month infection end point was validated using clinical trial data of the quadrivalent vaccine (see appendix A, which appears only in the electronic edition of the Journalappendix A, which is not available in the print edition of the Journal)

Data sourcesThe analysis of disease end points used the combined database of 2 phase 3 efficacy trials: protocol 013 (NCT00092521) [8] and protocol 015 (NCT00092534) [9], termed FUTURE I and FUTURE II, respectively. Both were phase 3, randomized, double-blind, placebo-controlled clinical trials designed to investigate the prophylactic efficacy of the quadrivalent vaccine (Gardasil; Merck and Co.), as described elsewhere [8, 9]. Data for the analysis of infection end points was derived from protocol 012 (NCT00092482), a substudy of protocol 013 [15]. The studies were designed to be of 4 years’ duration. Because of the high efficacy seen in FUTURE I and II, the independent data and safety monitoring board recommended vaccination of women in the placebo arm earlier than planned. The end-of-study data reported here includes ∼3.6 years of post–dose 1 follow-up

Although baseline samples were collected at enrollment, the trials allowed the enrollment of subjects who had been previously infected with or were currently infected with ⩾1 vaccine HPV type or ⩾1 of the 10 nonvaccine types analyzed here

PopulationsBetween December 2001 and May 2003, 17,622 women aged 16–26 years (there were two 15-year-olds) were enrolled in FUTURE I (n=5455, including the protocol 012 substudy [n=3578]) and FUTURE II (n=12,167). The trials enrolled women who reported having had 0–4 sex partners during their lifetime, except in Finland, where there was no such restriction. Subjects with a history of an abnormal Pap test result or treatment for genital warts (FUTURE I) were not enrolled. On day 1, subjects underwent a detailed genital examination, Pap testing, cervicovaginal sampling to detect HPV DNA, and serology for anti–HPV-6/11/16/18 testing. Subjects were randomly assigned (1:1) to receive intramuscular injections of HPV-6/11/16/18 vaccine or visually indistinguishable adjuvant-containing placebo on day 1 and at 2 and 6 months. Each protocol was approved by the institutional review boards

Clinical follow-up and laboratory testingThinPrep (Cytyc) cytology samples for Pap testing were collected on day 1, at month 7, and at 6-month (FUTURE I) or 12-month (FUTURE II) intervals thereafter. Cytology samples were classified using the Bethesda System 2001 [16]. Procedures for algorithm-based cytology, colposcopy, and biopsy referral have been described elsewhere [8, 9]. Biopsy material was first read for clinical management by pathologists at a central laboratory (Diagnostic Cytology Laboratories) and then read for end-point determination by a panel of up to 4 blinded pathologists, as described elsewhere [9]

Cervical biopsy samples, endocervical curettage samples, and samples from loop electrosurgical excision procedures and conization procedures obtained at any time during the studies were tested for 14 types (6, 11, 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, and 59), using a polymerase chain reaction (PCR)–based assay [17–19]

The following genital swab samples were obtained at each scheduled visit: an endo/ectocervical swab (one specimen) and a combined labial/vulvar/perineal swab plus a perianal swab (pooled to become a second specimen). Ascertainment of HPV infection involved HPV PCR analysis performed on these genital swab samples. All subjects were tested for the above 14 HPV types on day 1. At months 3, 7, 12, 18, 24, 30, 36, and 48, swab samples from subjects enrolled in protocol 012 were tested for 9 HPV types (16, 18, 31, 33, 35, 45, 52, 58, and 59)

Case definition and study objectivesThe primary objective was to determine whether administration of HPV-6/11/16/18 vaccine reduces the combined incidence of infection or disease associated with HPV-31 and -45 (the 2 most common types found in cervical cancer after HPV-16 and -18) and with HPV-31, -33, -45, -52, and -58 (the 5 most common HPV types found in cervical cancer after 16 and 18) [20]. Other end points were the incidence of infection or disease associated with nonvaccine A9 species members (31, 33, 35, 52, and 58), nonvaccine A7 species members (39, 45, and 59), and the 10 nonvaccine HPV types for which testing was available (31, 33, 35, 39, 45, 51, 52, 56, 58, and 59). The type(s) of HPV associated with each of the combined end points was based on genotyping HPV DNA detected in lesional tissue

Statistical methodsPrespecified analyses were done in a population that approximates sexually naive females. This analysis was restricted to subjects who received ⩾1 vaccination and, at enrollment, were seronegative and DNA negative for each of the quadrivalent HPV vaccine types (6, 11, 16, and 18); were DNA negative for each of 10 nonvaccine types (31, 33, 35, 39, 45, 51, 52, 56, 58, and 59); and had a normal Pap test result. In this article, this population is referred to as being negative for 14 HPV types. Protocol violators were included. Case counting began after day 1

A point estimate of vaccine efficacy (VE) and the 95% confidence interval (CI) were calculated on the basis of the observed split between vaccine and placebo recipients, adjusted for the accrued person-time in each study arm. The statistical criterion for success (P<.05) was equivalent to requiring that the lower bound of the CI for VE exclude 0%. An exact conditional procedure was used to evaluate VE under the assumption that the numbers of cases in the vaccine and placebo arms were independent Poisson random variables

We evaluated VE for combined HPV types (i.e., HPV-31/45–related CIN1–3/AIS), and for individual HPV types (i.e., HPV-31–related CIN1–3/AIS). For each analysis, including those which combined multiple HPV types, a woman was counted as having an end point only once. For example, in the analysis of HPV-31/45–related disease, a woman detected with an HPV-31–related CIN2 lesion at month 12 and an HPV-45–related CIN1 lesion at month 18 would be counted once toward HPV-31/45–related CIN1–3/AIS. When VE for individual HPV types was calculated, she would count once toward HPV-31–related CIN2–3/AIS and once toward HPV-45–related CIN1–3/AIS

Time-to-event plots shown were based on the Kaplan-Meier estimator [21]. The plots were not part of a formal survival analysis; rather, they give a visual demonstration of the divergence of incidence rates between the 2 vaccination arms over time. No direct estimation of VE should be inferred by extrapolating data from these plots

Results

FUTURE I and II enrolled 17,622 women, 99.9% of whom received ⩾1 dose of vaccine or placebo (98.4% received 2 doses and 97.2% received 3 doses). Baseline demographics for the individual studies [8, 9] and the combined studies [22] have been described elsewhere. In previously reports of VE [8, 9], the primary analysis population was per protocol. For the analyses of cross-protection, the prespecified primary analysis was done in type-specific populations that required women to be negative for the type being analyzed on day 1 only (unrestricted susceptible) rather than from day 1 through 1 month after receipt of dose 3 (per protocol) [8, 9]. Similar to the per-protocol analysis, women in the unrestricted susceptible population could be positive for other HPV types. In this primary analysis population, efficacy for HPV-31/45–related CIN1–3/AIS and high-grade lesions (CIN2–3/AIS) was 37.3% (95% CI, 17.0% to 52.8%) and 43.2% (95% CI, 12.1% to 63.9%), respectively. Efficacy for HPV-31/33/45/52/58–related CIN1–3/AIS and CIN2–3/AIS was 26.4% (95% CI, 12.9% to 37.8%) and 25.8% (95% CI, 4.6% to 42.5%), respectively. Although the primary results were favorable, the magnitude of benefit was confounded by cases of CIN2–3 associated with mixed prevalent/incident coinfection; therefore, prespecified supportive analyses were done in a population that was negative for the 14 tested HPV types, thus approximating sexually naive females. Reasons for exclusion from the efficacy analyses and baseline characteristics of the efficacy population that was negative for 14 HPV types are listed in table 2 and in figure 1figure 1. Baseline characteristics were generally well balanced between the vaccine and placebo arms. The placebo arm had a slightly higher proportion of black women (2.6% in the vaccine arm and 3.5% in the placebo arm) and virgins (9.3% in the vaccine arm and 10.4% in the placebo arm)

Subjects were followed for a mean of 3.6 years after dose 1, with 2068 (57.8% of enrolled subjects in protocol 012) included in the analyses of infection. Vaccination reduced the incidence of HPV-31/45 infection by 40.3% (95% CI, 13.9% to 59.0%) and HPV-31/33/45/52/58 infection by 25.0% (95% CI, 5.0% to 40.9%) (table 3). For individual HPV types, a significant reduction in infection was observed for HPV-31. A positive percent reduction was observed for the other 6 types analyzed, although the data did not reach statistical significance. We performed a post-hoc analysis where HPV-31 was removed from the composite end points (table 4). Efficacy for HPV-33/45/52/58 infection was 14.9% (95% CI, −11.3 to 35.0)

A total of 9296 subjects (53% of enrolled subjects in FUTURE I and II) were included in the analyses for cervical disease (tables 5 and 6). Efficacy for CIN1–3/AIS and for high-grade lesions (CIN2–3/AIS) associated with the 10 tested nonvaccine HPV types was 23.4% (95% CI, 7.8% to 36.4%) and 32.5% (95% CI, 6.0% to 51.9%), respectively. Vaccination reduced the incidence of HPV-31/45–related CIN1–3/AIS by 43.6% (95% CI, 12.9% to 64.1%) and of HPV-31/33/45/52/58–related CIN1–3/AIS by 29.2% (95% CI, 8.3% to 45.5%). Efficacy was driven primarily by reductions in HPV-31, -33, -52, and -58. No efficacy for disease was observed with respect to HPV-45, although of the 10 nonvaccine HPV types examined, HPV-45 was the least likely to be detected in CIN lesions. A post-hoc analyses where HPV-31 was removed is shown in table 4. Significant reductions in CIN1–3/AIS were observed when only the remaining 9 nonvaccine HPV types for which testing was conducted were considered

To address any potential ascertainment bias resulting from the higher frequency of colposcopy, biopsy, and definitive therapy among placebo recipients or from the inclusion of cervical biopsy samples in the analysis of infection, we did supportive analyses whereby infection was restricted to detection of HPV DNA in cervicovaginal/anogenital swab samples only. Here, the observed efficacy for HPV-31/33/45/52/58 infection was 23.1% (95% CI, 2.1% to 39.7%)

Figure 2 shows the divergence of incidence rates over time for CIN1–3/AIS and CIN2–3/AIS for the 10 tested nonvaccine HPV types. Over the course of follow-up, reductions in the incidence of disease became increasingly apparent for the 10 tested nonvaccine HPV types. As shown in figure 2, the censoring of subjects who reach an end point markedly decreases the sample size; therefore, the 95% CIs for the incidence rates grow wider with longer follow-up. Direct estimation of VE cannot be inferred by extrapolating data from these plots. Efficacy (tables 3–⇑6) was calculated on the basis of a fixed-event design with the analysis performed at a single point in time at the end of follow-up, with all subjects in the analysis population providing data

The cross-protective efficacy was most apparent and consistent for members of the A9 species. The combined incidence of HPV-31/33/35/52/58–related CIN1–3/AIS was reduced by 31.9% (95% CI, 11.8% to 47.6%). For the A7 species, a positive percent reduction was observed with respect to HPV-39– and HPV-59–related end points, although the reductions were not statistically significant. There was no evidence for efficacy with respect to HPV-51–related CIN (A5 species). A positive percent reduction (not statistically significant) was observed for HPV-56–related CIN (A6 species)

Efficacy findings for CIN1–3/AIS were consistent between the individual protocols, with the exception of HPV-35–, HPV-51–, and HPV-59–related end points, for which nonsignificant reductions were observed in protocol 013 but not protocol 015 (data not shown)

Although subjects were required to be DNA negative for all 14 HPV types on day 1 in order to be included in the efficacy evaluations, the presence of ⩾1 type in incident CIN lesions was common, consistent with prior natural history studies. A total of 308 incident CIN2–3/AIS lesions were observed in the placebo arm during the follow-up period (note: a women may have developed more than one lesion during the course of the studies, but for each row in the analyses of VE, each woman was counted only once). Of the 308 incident CIN2–3/AIS lesions observed in the placebo arm, 138 (44.8%) were associated with HPV-16–related A9 species, with 99 (32.1%) being associated with HPV-31/33/35/52 and/or 58 with no coinfection with vaccine HPV types and with 39 (12.7%) being associated with HPV-31/33/35/52 and/or 58 with coinfection with vaccine HPV types. In the vaccine arm, lesions associated with a mix of vaccine and nonvaccine HPV types were rare, because the prophylactic efficacy against disease due to vaccine HPV types approaches 100% [8, 9]

Discussion

HPV types other than 6, 11, 16, and 18 that cause a substantial proportion of cervical HPV disease, including cancer, are structurally related to HPV-16 and -18. It is important to examine the prophylactic efficacy of the HPV-6/11/16/18 vaccine in preventing the acquisition of infection and related cervical disease associated with these nonvaccine types in populations approximating those targeted by most HPV vaccination programs. Thus, the present study evaluated the cross-protective efficacy of the vaccine in a subset of the clinical trial population that, before vaccination, was negative for the 14 HPV types tested and had a normal Pap test result, approximating sexually naive females. This allowed for an assessment of cross-protective efficacy in the absence of prevalent infections. We observed a 32.5% reduction in CIN2–3/AIS associated with 10 nonvaccine HPV types that collectively cause ∼20% of cervical cancers. These findings are the first demonstration of cross-protection for any HPV vaccine against CIN2–3/AIS, disease end points that are cervical cancer precursors and that formed the basis of vaccine licensure. Reductions of similar magnitudes were observed for the combined end points of CIN1–3/AIS. The cross-protective efficacy was driven by the A9 species members, particularly HPV-31, -33, -52, and -58

Although we observed significant reductions in cervical disease due to HPV types not included in the vaccine, the clinical benefit of cross-protection (in terms of numbers of lesions prevented) should not be expected to be fully additive to the efficacy already observed against HPV-6/11/16/18–related disease, because many women have >1 CIN lesion and each lesion may be associated with a different HPV type. For example, among placebo subjects, the cumulative incidence of CIN2–3/AIS for the A9 species (HPV-31, -33, -35, -52, and -58) was 14.7 events per 1000 subjects. Of the 69 cases in the placebo arm, 22 (31.9%) occurred in women who also had an HPV-16– or HPV-18–related CIN2–3/AIS lesion. Had these women been immunized before exposure, the HPV-16/18 disease would have been prevented. In the absence of cross-protection, the prevention of HPV-16/18 infection would not have ended their risk of developing CIN2–3 due to a nonvaccine type. However, with our observed cross-protection, the risk of CIN2–3 due to a nonvaccine HPV type was reduced. As shown in figure 3, the added benefit of cross-protection resulted in the prevention of an additional 3 cases of CIN2–3, an increment of 4.3%. Despite an observed benefit, as with any vaccine the long-term reductions in the overall burden of disease—including reductions in Pap testing and abnormalities, colposcopy, and definitive therapy—have yet to be determined

The WHO recommends that demonstration of cross-protection be established by observed reductions in CIN and/or viral persistence. There is currently no international consensus on a definition for HPV persistence based on type-specific detection of HPV DNA by PCR [12]. We considered 6-month infection for the same HPV genotype. Single-time detection of HPV was not an end point, because a single positive HPV DNA test result may be due to transient infection, contamination, or early persistent infection. Previous reports of a bivalent HPV vaccine showed some level of cross-protection against infection with HPV-45, -31, and -52; however, disease end points were not reported [23]. Efficacy analyses using disease end points rather than infection alone are necessary to measure the clinical benefit of cross-protection

In the present study, when excluding the vaccine types, the cumulative incidence of CIN1–3/AIS through 3.6 years of follow-up was highest for HPV-56, -51, -31, -52, and -58, in descending order. Lesions associated with HPV-45 were rare. We observed reductions in the incidence of CIN lesions associated with A9 species members, particularly HPV-31. It is likely that second-generation HPV vaccines targeting a broader spectrum of oncogenic HPV types may be available within the next decade and would include those HPV types that cause the highest proportion of cervical cancers. Given that some of these types, such as HPV-56, contribute to <1% of the worldwide cervical cancer burden, it is noteworthy to observe reductions in HPV types that may not be included in second-generation vaccines

The present study has some limitations. Our combined end points included lesions with strong malignant potential (CIN3/AIS). However, the study was not designed with sufficient power to detect reductions in CIN2–3/AIS due to nonvaccine types or to measure reductions in individual nonvaccine HPV types, explaining why a posteriori there is not sufficient power to measure these reductions. Our analyses also required that women be DNA negative for nonvaccine types on day 1 only; thus, some women may have become infected before receiving all 3 doses. In addition, subjects in the placebo arm were more likely to be referred for colposcopic examination, biopsy, and definitive therapy, because of the lack of protection from HPV-6/11/16/18 infection. This could lead to potential ascertainment bias. It was therefore important to corroborate the observed reductions in cervical disease associated with nonvaccine types with infection data from swab samples only, which would not be subject to this potential bias

Because women remain at risk for HPV infection as long as they remain sexually active, it will be important to determine the duration of protection against HPV-related disease for both vaccine and nonvaccine types. It has not been possible to establish a minimum protective antibody titer for vaccine types; however, the quadrivalent vaccine elicits immune memory, a hallmark of long-term protection [24]. Follow-up of large cohorts will be required to establish the duration of efficacy of the vaccine for both vaccine and nonvaccine types

The efficacy of the vaccine among 16–26-year-old women with a lifetime history of ⩾4 sex partners was not evaluated. Although the present analysis focused on cross-protection in a population that was representative of an HPV-naive population, an ITT analysis of all women entering the FUTURE I and II trials regardless of baseline HPV status also demonstrated statistically significant cross-protective reductions in both HPV infections and CIN1 or greater due to nonvaccine oncogenic HPV types [14]

In conclusion, the results of the HPV-6/11/16/18 vaccine program provide strong evidence that implementation of HPV vaccination campaigns in HPV-naive preadolescent girls and young adult women has the potential to reduce cervical cancer rates worldwide. The demonstrated cross-protection against additional oncogenic HPV types may provide an extra measure of protection for young women immunized with the quadrivalent HPV vaccine

Author affiliationsDepartment of Medicine, Indiana University School of Medicine, Indianapolis (D.R.B.); Department of Virus, Hormones, and Cancer, Institute of Cancer Epidemiology, Danish Cancer Society/Rigshospitalet, Copenhagen, Denmark (S.K.K.); National Cancer Detection Clinic, Reykjavik, Iceland (K.S.); Department of Clinical Medicine, University of Bergen, and Department of Obstetrics and Gynecology, Haukeland University Hospital, Bergen, Norway (O.-E.I.); Institute of Public Health, Cuernavaca, Morelos, Mexico (M.H.-A.); Department of Molecular Genetics and Microbiology and Department of Obstetrics and Gynecology, University of New Mexico, Albuquerque (C.M.W.); Universidad del Rosario, Bogotá, Colombia (G.P.); Department of Epidemiology, University of Washington, Seattle (L.A.K.); KK Women’s and Children’s Hospital, Singapore (E.H.T.); Epidemiology HIV and STD Unit, Universidad Peruana Cayetano Heredia, Lima, Peru (P.G.); Department of Gynecology and Obstetrics, Emory University School of Medicine Atlanta, Georgia (K.A.A.); Microbiology and Infectious Diseases Department, Royal Women’s Hospital, and Department of Obstetrics and Gynecology, University of Melbourne, Melbourne, Victoria, Australia (S.M.G.); Department of Gynecology and Obstetrics, Medical University of Vienna, Vienna, Austria (S.L. and E.A.J.); Karolinska Institute at Danderyd Hospital, Stockholm, Sweden (S.-E.O.); Department of Obstetrics and Gynecology, University of Hong Kong, Hong Kong Special Administrative Region, China (G.W.K.T.); Department of Family Medicine and Obstetrics and Gynecology, Medical College of Georgia, Augusta, Georgia (D.G.F.); Department of Obstetrics and Gynecology, University Central Hospital, Helsinki, Finland (J.P.); Direction Risques Biologiques, Environnementaux et Occupationnels, Institut National de Santé Publique du Québec, Montréal, Québec, Canada (M.S.); Institut Catala d’Oncologia, IDIBELL, Barcelona, Spain (F.X.B.); Department of Medical Microbiology, Lund University, Lund, Sweden (J.D.); Departments of Gynecology and Obstetrics, Pathology, and Oncology, Johns Hopkins University School of Medicine, Baltimore, Maryland (R.J.K.); Department of Dermatology and Venereology, Center of Diagnostics and Treatment of Sexually Transmitted Diseases, Warsaw Medical University, Warsaw, Poland (S.M.); National Institute of Cancer, Bogotá, Colombia (N.M.); Department of Obstetrics and Gynecology, Duke University Medical Center, Durham, North Carolina (E.R.M.); Department of Virology, Ludwig Institute for Cancer Research, Sao Paulo, Brazil (L.L.V.); Merck Research Laboratories, West Point, Pennsylvania (F.J.T., C.R., A.T., J.B., L.C.L., K.E.D.G., H.L.S., M.J., T.M.H., and E.B.)

Potential conflicts of interestN.M. has received lecture fees, advisory board fees, and consultancy fees from Merck and Sanofi Pasteur MSD. S.-E.O. has received lecture fees from Merck. M.H.-A. has received lecture fees and grant support from Merck. O.-E.I. has received lecture fees from Merck and GlaxoSmithKline (GSK). C.M.W. has received funding through her institution to conduct HPV vaccine studies for GSK. K.A.A. has received consultancy and advisory board fees from Merck and has received funding through his institution to conduct HPV vaccine studies for Merck and GSK and nonvaccine clinical trials for Gen-Probe. F.X.B. has received lecture fees from Merck and GSK and has received funding through his institution to conduct HPV vaccine studies for GSK. J.P. has received consultancy fees, advisory board fees, and lecture fees from Merck. J.D. has received consultancy fees, lecture fees, and research grants from Merck and Sanofi Pasteur MSD. S.L. has received lecture fees from Merck and Sanofi Pasteur MSD. E.A.J. has received lecture fees from Merck, Sanofi Pasteur MSD, and GSK. S.K.K. has received consultancy fees and funding through her institution to conduct HPV vaccine studies for Sanofi Pasteur MSD and Digene. S.M.G. has received advisory board fees and grant support from Commonwealth Serum Laboratories and GSK and has received lecture fees from Merck. D.G.F. has received consultancy fees and funding through his institution to conduct HPV vaccine studies for GSK and lecture fees and consultancy fees from Merck. K.S. has received consultancy fees from Merck. S.M. has received lecture fees and advisory board fees from Merck. G.P. has received lecture fees and consultancy fees from Merck and Sanofi Pasteur MSD. D.R.B. has received lecture fees, advisory board fees, and intellectual property fees from Merck. M.S. has received lecture fees and grant support from Merck. Additionally, S.-E.O., C.M.W., M.H.-A., L.L.V., O.-E.I., G.W.K.T., F.X.B., J.P., J.D., E.H.T., S.L., E.A.J., S.K.K., G.P., D.G.F., K.S., M.S., L.A.K., and D.R.B. have received funding through their institutions to conduct HPV vaccine studies for Merck. F.J.T., C.R., A.T., J.B., L.C.L., K.E.D.G., H.L.S., M.J., T.M.H., and E.B. are employees of Merck and potentially own stock and/or stock options in the company