Population-Level Impact of the Bivalent, Quadrivalent, and Nonavalent Human Papillomavirus Vaccines: A Model–Based Analysis
please e-mail: .
Population-l evel impact of the Bivalent, Quadrivalent, and Nonavalent Human Papillomavirus Vaccines: A Model-Based Analysis
Nicolas Van de Velde 0 1
Marie-Claude Boily 0 1
Mélanie Drolet 0 1
Eduardo L. Franco 0 1
Marie-Hélène Mayrand 0 1
Erich V. Kliewer 0 1
François Coutlée 0 1
Jean-François Laprise 0 1
Talía Malagón 0 1
Marc Brisson ) 0 1
0 Sacrement , 1050 Chemin Sainte-Foy, Québec , Canada , G1S 4L8 (
1 Affiliations of authors: URESP, Centre de recherche FRSQ du CHA uni- versitaire de Québec , Québec , Canada (NVdV, M-CB, MD, J-FL, TM , MB); Département de médecine sociale et préventive, Université Laval , Québec , Canada (NVdV, MD, J-FL, TM , MB); Department of Infectious Disease Epidemiology, Imperial College , London , UK ( M-CB); Division of Cancer Epidemiology, McGill University , Montreal , Canada ( ELF); Départements d'obstétrique-gynécologie et de médecine sociale et préventive, Université de Montréal et CRCHUM , Montréal , Canada ( M-HM); Départements d'obstétrique-gynécologie et de médecine sociale et préventive, Université de Montréal et CRCHUM , Montréal , Canada ( EVK); Epidemiology and Cancer Registry, CancerCare Manitoba , Winnipeg , Canada ( EVK); Community Health Sciences, University of Manitoba , Winnipeg , Canada ( EVK); Cancer Control Research, British Columbia Cancer Agency , Vancouver , Canada ( FC); Laboratoire de Virologie Moléculaire, Centre de Recherche, Centre Hospitalier de l'Université de Montréal , Montréal , Canada ( FC); Département de Microbiologie-Immunologie, Université de Montréal , Montréal , Canada (FC)
Methods Bivalent and quadrivalent human papillomavirus (HPV) vaccines are now licensed in several countries. Furthermore, clinical trials examining the efficacy of a nonavalent vaccine are underway. We aimed to compare the potential population-level effectiveness of the bivalent, quadrivalent, and candidate nonavalent HPV vaccines. We developed an individual-based, transmission-dynamic model of HPV infection and disease in a population stratified by age, gender, sexual activity, and screening behavior. The model was calibrated to highly stratified sexual behavior, HPV epidemiology, and cervical screening data from Canada. Under base case assumptions, vaccinating 12-year-old girls (70% coverage) with the bivalent (quadrivalent) vaccine is predicted to reduce the cumulative incidence of anogenital warts (AGWs) by 0.0% (72.1%), diagnosed cervical intraepithelial neoplasia lesions 2 and 3 (CIN2 and -3) by 51.0% (46.1%), and cervical squamous cell carcinoma (SCC) by 31.9% (30.5%), over 70 years. Changing from a bivalent (quadrivalent) to a nonavalent vaccine is predicted to reduce the cumulative number of AGW episodes by an additional 66.7% (0.0%), CIN2 and -3 episodes by an additional 9.3% (12.5%), and SCC cases by an additional 4.8% (6.6%) over 70 years. Differences in predicted population-level effectiveness between the vaccines were most sensitive to duration of protection and the time horizon of analysis.The vaccines produced similar effectiveness at preventing noncervical HPV-related cancers. The bivalent vaccine is expected to be slightly more effective at preventing CIN2 and -3 and SCC in the longer term, whereas the quadrivalent vaccine is expected to substantially reduce AGW cases shortly after the start of vaccination programs. Switching to a nonavalent vaccine has the potential to further reduce precancerous lesions and cervical cancer.
Human papillomavirus (HPV), with more than 40 genotypes
infecting the anogenital tract (1), is among the most common
sexually transmitted infections worldwide (2). Low oncogenic risk types
HPV6 and -11 are responsible for anogenital warts (AGWs) (3)
and recurrent respiratory papillomatosis (4). Infection with high
oncogenic risk types is a necessary cause of cervical cancer (5), with
HPV16 and -18 accounting for 70% of these cancers (6). High
oncogenic risk HPVs, most notably HPV16, are also associated
with other anogenital cancers (vulvar, vaginal, anal, penile) (7,8)
and head and neck cancers (9).
Two prophylactic HPV vaccines are now available: the bivalent
and quadrivalent vaccines that protect against HPV16 and -18
and HPV16, -18, -6, and -11, respectively. Both vaccines have
been shown to be very safe and highly effective against
vaccinetype persistent HPV infection and lesions among women [vaccine
efficacy = 98%–100% (10,11)]. In addition, clinical trials have
shown vaccine cross-protection against nonvaccine HPV types
(10,12–14). Vaccine efficacy against 6-month persistent infection
with HPV31, -33, and -45 were, respectively, 77% (95% confidence
interval [CI] = 67% to 84%), 43% (95% CI = 19% to 60%), and
79% (95% CI = 61% to 89%) for the bivalent vaccine and 46%
(95% CI = 15% to 66%), 29% (95% CI = −45% to 66%), and
8% (95% CI = −67% to 49%) for the quadrivalent vaccine (12,14).
Clinical trials are currently examining the potential efficacy of a
nonavalent vaccine including HPV16, -18, -31, -33, -45, -52, -58,
-6, and -11, which cause approximately 90% of cervical cancers
Given the recent availability of the bivalent and quadrivalent
vaccines in several countries, policymakers and clinicians are
currently deciding which HPV vaccine should be used. This choice
is complicated by the fact that the HPV vaccines have different
properties and intended benefits. The quadrivalent vaccine can
prevent HPV6– and HPV11–associated cervical lesions, AGWs,
and, potentially, recurrent respiratory papillomatosis. On the other
hand, recent results suggest that the bivalent vaccine may confer
greater cross-protection against high oncological risk HPV31,
-33, -45, -52, and -58 (10,12,14) and, potentially, longer duration
of protection (16,17). The timing of health benefits will also be
different between the HPV vaccines: substantial reductions of
AGW cases have been currently observed shortly after the start
of quadrivalent vaccination programs (18), whereas the potential
additional gains in prevention of cervical lesions and cancer from
higher bivalent cross-protection may not be observed for decades.
Given their different characteristics, the decision to switch vaccines
within ongoing programs can also lead to changes in the dynamics
of HPV-related diseases over time. Decision makers thus require
evidence on how the added cross-protection measured for the
bivalent vaccine within clinical trials translates into incremental gains
in HPV-related cancer prevention over time at the
populationlevel and how this compares to the added benefits of the
quadrivalent vaccine at preventing AGWs.
If the nonavalent HPV vaccine proves safe and effective, policy
makers will also have to decide whether or not to introduce this
second-generation vaccine. Given that the current HPV vaccines
have demonstrated partial efficacy against the additional HPV
types included in the nonavalent vaccine (HPV31, -33, -45, -52,
and -58), evidence is required on how this may impact the
incremental gains of the candidate vaccine. Finally, there are concerns
that adding more types to the current HPV vaccines may produce
immune interference (eg, lower antibody titers to the individual
types) and thus result in lower type-specific efficacy for the
nonavalent vaccine. Thus, evidence is required about the level of
vaccine efficacy and duration of protection necessary in order for the
nonavalent to be noninferior to the current HPV vaccines in terms
of cancer prevention.
Decisions regarding infectious disease control are
increasingly made with substantial input from mathematical models
(19). These models provide a formal framework to synthesize
and project results from various sources (eg, clinical trials and
epidemiological studies) to examine questions that cannot be
answered in a clinical trial setting. Currently, there is little
available evidence about the potential comparative population-level
effectiveness of the bivalent and quadrivalent vaccines in terms
of magnitude and timing of incidence reduction of HPV-related
diseases. The few modeling studies that have examined this
question did not incorporate type-specific cross-protection (20–23)
and/or herd immunity (20,22,24,25) and did not examine the
impact of switching HPV vaccines within ongoing programs on
the dynamics of HPV-related diseases. Furthermore, studies have
yet to examine the potential additional benefits of a candidate
nonavalent vaccine, or the level of vaccine efficacy and duration
of protection required for a nonavalent vaccine to be noninferior
to the bivalent and quadrivalent vaccines in terms of
The goal of this paper is to use mathematical modeling to
1) compare the population-level effectiveness of current HPV
vaccine strategies at preventing HPV-related diseases over time,
including recent evidence on type-specific cross-protective efficacy
and herd-immunity, 2) assess the potential impact of switching
HPV vaccines within current vaccination programs, and
3) examine the potential incremental gains of using a nonavalent vaccine.
We developed Agent-based Dynamic model for VaccInation
and Screening Evaluation (HPV-ADVISE), the first calibrated,
individual-based, transmission-dynamic model of sequential
partnership formation and dissolution and natural history of multitype
HPV infection and disease (26). An individual-based modeling
approach was chosen to be able to represent the various levels of
heterogeneity across the HPV-related disease control spectrum: sexual
behavior, health-seeking behavior (vaccination and screening), and
type-specific HPV transmission, infection, progression toward
disease, and vaccine efficacy. The model contains five fully integrated
components: 1) sociodemographic characteristics, 2) sexual behavior
and HPV transmission, 3) natural history of HPV-related diseases,
4) vaccination, and 5) screening and treatment. An in-depth
description of the model structure, model parameterization, calibration
data, and parameter values and figures of model fit are available on
the author’s website (http://www.marc-brisson.net/HPVadvise.pdf).
Sociodemographic Characteristics. Individuals enter the
simulated Canadian population prior to sexual debut. The modeled
population is heterosexual, open, and stable (ie, age-specific death rates
balance the birth rate). In the model, individuals are attributed three
different risk factors for HPV infection and/or disease: gender, four
levels of sexual activity, and five screening behavior levels.
Sexual Behavior and HPV Transmission. The sexual behavior
and transmission component is described in detail in Van de Velde
et al (26). HPV transmission is assumed to depend on sexual behavior
(level of sexual activity and mixing patterns), per sex-act probability
of transmission, and natural history of infection (duration of
infectiousness and natural immunity). Partnership formation and
dissolution are based on gender-specific, age-specific, and level of sexual
activity–specific partner acquisition and separation rates and mixing
patterns. Eighteen HPV types are modeled individually: HPV16,
-18, -6, -11, -31, -33, -45, -52, -58, -35, -39, -51, -56, -59, -66, -68,
-73, and -82. These types are assumed to be independent (no synergy
or competition) with respect to transmission, infection, persistence,
and disease progression; thus, any combination of multiple
infections is possible. Following clearance, individuals may develop
sametype natural immunity (ie, same-type reinfection is possible).
HPV-Related Diseases. HPV-ADVISE was developed to capture
the potential impact of HPV vaccination on AGWs, cervical cancer
(squamous cell carcinoma [SCC] and adenocarcinoma), and other
HPV-related cancers (vulva, vagina, anus, and head and neck):
Anogenital warts. Once infected with HPV6 or -11, individuals
have a joint probability of developing and being diagnosed with
AGWs or clearing their infection. Individuals can experience
multiple episodes of AGWs through recurrence of a persistent
infection, reinfection with a previously cleared HPV type, or infection
with a new HPV type. We assume that HPV6 and -11 are
responsible for 85% of all AGW cases (3).
Squamous cell carcinoma. The natural history of SCC is
represented by nine mutually exclusive health states: three HPV
infection states (susceptible, infected, and immune), three grades of CIN
lesions (CIN1, -2, and -3), and three stages of cancer (localized [I],
regional [II], and distant stages [III]) (27). Transition rates between
these states are type-specific.
Other HPV-related cancers. Similar to Smith et al (28), we
estimate the long-term impact of HPV vaccination on HPV-related
cancers by applying model predictions of the relative reductions
in type-specific HPV prevalence at equilibrium to the HPV-type
distribution among cervical adenocarcinomas and cancers of the
vulva, vagina, anus, and head and neck in North America (7–9).
Vaccination. HPV-ADVISE assumes that HPV vaccines prevent
infection but do not alter the natural history of disease in
individuals already infected by a vaccine type. Different vaccine efficacy
can be applied to any of the 18 HPV types included in the model
to examine the potential impact of cross-protection or the
nonavalent vaccine. Type-specific cross-protective vaccine efficacy values
were based on a comprehensive review of published clinical
trials results (10,12–14,29), including recently published results from
Wheeler et al (30). In our base case, vaccine efficacy against HPV
vaccine types is 100% (10,11,31), vaccine efficacies against
nonvaccine HPV types are the published type-specific efficacy against
persistent infection among HPV-naive females (Table 1), and
vaccine protection (including cross-protection) is lifelong.
A sensitivity analysis was performed to explore the impact of vaccine efficacy
and duration on model predictions. In scenarios with limited
vaccine duration, each vaccinated individual is given a specific
duration of protection against the vaccine types sampled from a normal
distribution (µ = 20 years; σ = 6). Hence, under this scenario,
3%, 20%, and 50% of individuals will have lost their protection
9, 15, and 20 years after vaccination, respectively. These
assumptions reflect available clinical data showing no evidence of waning
after 9.5 years of follow-up (32) and evidence of immune memory
8.5 years following vaccination (33).
Screening and Treatment. HPV-ADVISE mimics various
screening algorithms at the individual level by tracking and
simulating each woman’s screening history. In this study, the model
represents cervical cancer screening in Canada, which is
cytology based. Screening rates are a function of a woman’s
screening behavior level, previous screening test results, and age. The
model incorporates five screening behavior levels, which
represents the average time between two routine Papanicolaou tests in
the absence of an abnormal result (from screening every 1.25 years
[level 0] to never [level 4]). Screening behavior parameters were
estimated using Canadian-specific, population-based data (34,35).
Algorithms for the management and treatment of women with
an abnormal cytology (eg, repeat cytology, colposcopy or biopsy,
treatment) are dependent on the test result (ASC-US/LSIL,
HSIL/ASC-H/SCC (Atypical Cells of Undetermined Significance
[ASC-US], Low-grade Squamous Intraepithelial Lesion [LSIL],
High-grade Squamous Intraepithelial Lesion [HSIL], Atypical
Squamous Cells, cannot rule out a High grade lesion [ASC-H],
and SCC) and were based on Canadian guidelines (36–38) and
validated using empirical data (35). The sensitivity and specificity
of cytology and colposcopy are lesion-specific and were estimated
from two systematic reviews on the performance of cervical cancer
screening cytology (39,40) and a review of papers assessing the
success of colposcopy at diagnosing cervical lesions (41–44). Finally,
women with SCC have a stage-specific probability of developing
symptoms and being diagnosed outside of routine screening.
Individual-based models require a considerable amount of calibration
data in order to provide robust and valid predictions (parameters
for which values were derived through calibration are presented in
Supplementary Table 1, available online). The calibration procedure
developed in prior publications (26,45) was used to identify multiple
VE persistent cervical
Quadrivalent 46.2 (12) 28.7 (12) 7.8 (12)
VE CIN2+* (including lesions
co-infected with HPV16
and/or -18), %
VE CIN2+ (excluding lesions
co-infected with HPV16
and/or -18), %
Quadrivalent 70.0 (12) 24.0 (12) 0‡ (12)
Quadrivalent 57.4 (29) 0.0 (29)‡ 0.0 (29)§
parameter sets that fit highly stratified Canadian sexual behavior,
natural history, and screening data from the literature,
populationbased datasets, and original studies (see Supplementary Table 2,
available online) (34,35,46–55) and to account for conjoint parameter
uncertainty (26,45). The procedure is as follows:
1. Plausible prior distributions are defined for each of the 87
calibrated model parameters (minimum and maximum values
for each parameter are derived from the literature).
2. Different combinations of parameter values are drawn from
the prior distributions using Latin hypercube sampling [an
efficient sampling method of parameter space (56–58)].
3. Parameter sets are qualified as producing a “good fit” and
included in the posterior parameter sets if the associated
model predictions fall simultaneously within all prespecified
targets (ranges) of the observed data.
Such a procedure is computer and data intensive. Of 285 000
different combinations of parameters sampled (1 850 000 runs and
2 × 1013 person-years simulated) from the prior parameter
distributions, 10 parameter sets produced model results within the 639
prespecified data targets.
Model Outcomes Population-level vaccine effectiveness predictions are presented for three primary outcomes: AGWs in males and females, diagnosed CIN2 and -3, and SCC. Outcomes were modeled over 70 years
post-vaccination because 1) it is the time horizon required to reach
a stable post-vaccination equilibrium, 2) it allows the illustration of
the pre-equilibrium incidence dynamics, and 3) it shows the
maximum differences in effectiveness between the different vaccines.
Variability of model predictions is presented as the median, 10th,
and 90th percentiles of results from the posterior parameter sets,
referred to as the 80% range (80% R). Univariate sensitivity analysis
was performed by varying vaccination coverage, vaccine-type
efficacy, cross-protective efficacy, and duration of vaccine protection.
Population-Level Effectiveness of the Bivalent and
Figure 1 compares the population-level impact of vaccinating
12-year-old girls with the bivalent and quadrivalent vaccines,
taking into account cross-protection and assuming 70% coverage.
The model predicts that quadrivalent vaccine programs will lead to
a rapid decrease in the incidence of AGWs (eg, median = 10 years
[80% R = 9–11] to reach a 50% reduction in incidence) and that at
equilibrium (after 70 years) AGW cases will be reduced by 84.4%
(80% R = 77.7–85.0) (Figure 1, A). Overall the quadrivalent vaccine
is estimated to reduce the cumulative incidence of AGWs by 72.1%
over 70 years. In countries where the bivalent vaccine is used, the
Figure 1. Estimated population-level impact of vaccinating 12-year-old
girls with the bivalent and quadrivalent vaccines. Estimated
percentage change following vaccination in the incidence of anogenital warts
(AGWs) in males and females (A), diagnosed cervical intraepithelial
neoplasia 2 and 3 (CIN2 and -3) (B), and squamous cell carcinoma
(SCC) (C). Reduction in incidence of CIN2 and -3 and SCC in females
over the first 70 years of a girls-only vaccination program (D). Base
case: vaccine coverage = 70%, vaccine duration = lifetime, vaccine-type
efficacy = 100%, cross-protective vaccine efficacy presented in Table 1.
Average pre-vaccination incidence rate of AGWs in women = 124 per
100 000 women-years, AGWs in men = 157 per 100 000 men-years,
diagnosed CIN2 and -3 = 91 per 100 000 women-years, and SCC = 6 per
100 000 women-years. Population of 25 × 170 000 individuals. Incidence
of SCC is presented using a 3-year moving average.
incidence of AGWs will remain unaltered by vaccination (ie, 0.0%
reduction in AGW cases).
The bivalent and quadrivalent vaccines are predicted to
produce very similar short-term declines in cervical lesions and
cancer incidence (Figure 1, B and C). However, the bivalent vaccine
is estimated to produce larger reductions in the long-term. Under
base case assumptions (70% coverage, girls-only vaccination), the
model predicts that it would take the bivalent and quadrivalent
vaccine programs 19 (80% R = 17–23) and 20 (80% R = 19–27)
years, respectively, to decrease the incidence of diagnosed CIN2
and -3 by 50% (Figure 1, B), and 40 (80% R = 37–46) and 42
(80% R = 36–49) years, respectively, to decrease the incidence of
SCC cases by 50% (Figure 1, C). Furthermore, the bivalent and
quadrivalent vaccine programs are estimated to reduce the
incidence of diagnosed CIN2 and -3 by 62.1% (80% R = 57.0–71.5)
and 58.6% (80% R = 52.3–67.5), respectively, at equilibrium
(Figure 1, B), and to reduce the incidence of SCC cases by 70.5%
(80% R = 59.9–75.0) and 64.8% (80% R = 53.6–73.2), respectively
(Figure 1, C). Overall, the bivalent and quadrivalent vaccines are
estimated to reduce the cumulative number of diagnosed CIN2
and -3 by 51.0% (80% R = 44.3–54.8) and 46.1% (80% R = 40.2–
51.2), respectively, over 70 years after vaccination and to reduce the
cumulative number of SCC cases by 31.9% (80% R = 26.0–38.3)
and 30.5% (80% R = 24.5–38.2), respectively (Figure 1, D). These
differences in population-level vaccine effectiveness are due to the
predicted greater cross-protective efficacy of the bivalent vaccine.
For example, cross-protection produces a cumulative reduction
of 7.8% and 4.8% in CIN2 and -3 and 4.8% and 3.0% in SCC
incidence over 70 years for the bivalent and quadrivalent vaccines,
respectively (Figure 1, D).
Switching Vaccine Within Ongoing Vaccination Programs
Figure 2 shows the population-level impact of switching HPV
vaccines 5 years into a girls-only vaccination program. Under
base case vaccine assumptions (age at vaccination = 12 years;
coverage = 70%), switching from a quadrivalent to a bivalent
vaccine is estimated to prevent an additional 3.2% diagnosed
CIN2 and -3 cases and 1.8% SCC cases over 70 years compared
with continuing with the quadrivalent vaccine, but switching
increases the cumulative number of AGW cases by 62.7%
(Table 2). The increase (rebound) in AGW cases would occur about
5 years after changing to a bivalent vaccine, whereas the gains
in CIN2 and -3 and SCC prevention would start accumulating
after 15 and 45 years, respectively (Figure 2). Conversely, under
the above vaccination assumptions, changing from a bivalent to
a quadrivalent vaccine is predicted to increase the number of
diagnosed CIN2 and -3 and SCC cases by an additional 3.0%
and 1.3% over 70 years, respectively, but to reduce AGW cases by
66.7% (Table 2). Finally, our model predicts that if the nonavalent
vaccine proves to be highly effective, switching from a bivalent or
quadrivalent vaccine program could yield substantial incremental
benefits (Figure 2 and Table 2). Indeed, under base assumptions,
changing from a bivalent (quadrivalent) to a nonavalent vaccine
is predicted to reduce the cumulative number of AGW cases,
diagnosed CIN2 and -3 cases, and SCC cases over 70 years by
an additional 66.7% (0.0%), 9.3% (12.5%), and 4.8% (6.6%),
respectively (Table 2).
Bivalent to nonavalent after 5 years
Bivalent to quadrivalent after 5 years
Quadrivalent to nonavalent after 5 years
Quadrivalent to bivalent after 5 years
Figure 2. Estimated population-level impact of switching vaccines within
ongoing girls-only vaccination programs. Estimated percentage change
in the incidence of anogenital warts (AGWs) in males and females (A),
diagnosed cervical intraepithelial neoplasia 2 and 3 (CIN2 and -3) (B),
and squamous cell carcinoma (SCC) (C).The solid red (blue) lines
represent scenarios in which the bivalent (quadrivalent) is used continuously
throughout the 70-year time horizon. The dotted lines represent
scenarios in which the vaccine is switched. Change in the vaccine used is
modeled to occur 5 years after the start of vaccination programs. Base case:
age at vaccination = 12 years, vaccine coverage = 70%, vaccine
duration = lifetime, vaccine-type efficacy = 100%, cross-protective vaccine
efficacy presented in Table 1. Average pre-vaccination incidence rate of
AGWs in women = 124 per 100 000 women-years, AGWs in men = 157
per 100 000 men-years, diagnosed CIN2 and -3 = 91 per 100 000
womenyears, and SCC = 6 per 100 000 women-years. Population of 170 000
individuals. Each parameter set was run 25 times. Incidence of SCC is
presented using a 3-year moving average.
Differences in the population-level effectiveness of the bivalent,
quadrivalent, and nonavalent vaccines are dependent on
vaccination coverage, vaccine efficacy, duration of protection, and the
HPV outcome examined (Figure 3).
03 lau ed ra .05 to .80 to .11 to .60 to
– i − .5 − .0 − .5 − .9
0 m c %
2 2 4 4
–070 laum icedn ran% .667− .t5o− .667− .t6o− .627 .t3o6 .00¶
u in 08 8 7 9
C ( 6 6 (5
r ) )
e e 2
vo lw y itv ce eg .9
cen itan –05 lau iend ran .995 to− .995 to− .45 t5o .0¶
e e 0 m c % − .6 − .5
–04 –40 rase rase tob
s s y y d
5 9 0 ra ra in in e
. e e m
1 0 2 y y e e u
(6− (6− (5 ine ine iccn iccn ssa
in in va va is
).704 ).704 ).13 il.asapo litvaccavne litvaccavne litravaendu litravaendu -a11dn6V
e b b q q P
5 4 4 .78 0 illan the the the the tsnH
(4− (4− (3 iteh isgn isgn isgn isgn iaag
p u u u u )
ae d d d d E
ltan†e lt‡en t t§ ||lten ilitracn itaccen itaccen itaccen itaccen i(ayccV
litavne irtavuqdo taavonno ilrvaadenu iltvabneo taavonno I rcve=CN lirrasaveG lirrsaaveG lirrasaveG lirrsaaveG iffccaneeV
Figure 3. Sensitivity analysis. Vaccination coverage: Reduction in the
incidence of diagnosed cervical intraepithelial neoplasia 2 and 3 (CIN2
and -3) (A) and squamous cell carcinoma (SCC) (B) after 70 years of a
girls-only vaccination program for different vaccine coverage. Vaccine
characteristics of cross-protection: Reduction in the incidence of
diagnosed CIN2 and -3 (C) and SCC (D) after 70 years of a girls-only
vaccination program for different vaccine characteristics of cross-protection.
Nonavalent vaccine characteristics: Reduction in the incidence of
diagnosed CIN2 and -3 (E) and SCC (F) after 70 years of a girls-only
vaccination program for different nonavalent vaccine characteristics. Base
case: vaccine coverage = 70%, vaccine duration = lifetime, vaccine-type
efficacy = 100%, cross-protective vaccine efficacy presented in Table 1.
VDvac = vaccine duration for vaccine types; VDx = vaccine duration for
cross-protective types; VE CIN2+ (excluding) = type-specific vaccine
efficacy against CIN2+ excluding lesions co-infected with human
papillomavirus (HPV) 16 and/or 18; VE CIN2+ (including) = type-specific vaccine
efficacy against CIN2+ including lesions co-infected with HPV16 and/
or -18; VEvac = vaccine efficacy for vaccine types; VEX = vaccine efficacy
for cross-protective types. Average pre-vaccination incidence rate of
diagnosed CIN2 and -3 = 91 per 100 000 women-years and SCC = 6 per
100 000 women-years. Population of 170M 000 individuals. Each
parameter set was run 25 times.
Increasing coverage not only improves the predicted
population-level effectiveness of girls-only vaccination but also
amplifies the absolute and relative differences between the bivalent,
quadrivalent, and nonavalent vaccines (Figure 3, A and B). Using
published type-specific efficacy against CIN2+, which has greater
cross-protection than persistent infection, also increases
differences between the bivalent and quadrivalent vaccines in terms
of CIN2 and -3, but not SCC, prevention (Figure 3, C and D).
This is because a higher proportion of CIN2 and -3 cases than
SCC cases are caused by HPV31, -33, -45, -52, and -58
(crossprotective types). On the other hand, our model predicts that the
bivalent and quadrivalent vaccines produce similar
populationlevel effectiveness against CIN2 and -3 and SCC when duration of
cross-protection is 10 years (Figure 3, C and D). Finally, reducing
vaccine-type efficacy from 100% to 95% has little impact on
predictions (Figure 3, C and D).
Assumptions regarding duration of protection against vaccine
types (HPV16 and -18) have a greater impact on differences in
predicted population-level impact between the current vaccines than
does cross-protection. For example, if the bivalent vaccine confers
lifelong protection and the quadrivalent vaccine efficacy lasts for
only 20 years (10 years for cross-protective types), our model
predicts that the bivalent would prevent an additional 27% diagnosed
CIN2 and -3 cases and 18% SCC cases at equilibrium (assuming
70% coverage; Figure 3, C and D).
Because there is no published information about the nonavalent
vaccine’s efficacy, we examined which characteristics are required
in order for the candidate vaccine to be more effective at the
population-level than the existing vaccines. Our model suggests that
the vaccine-type efficacy of the nonavalent vaccine must be greater
than 80%–85% to produce higher population-level effectiveness
against SCC than the bivalent and quadrivalent vaccines
(assuming 100% vaccine-type efficacy, cross-protection, and lifelong
protection for the first-generation vaccines and lifelong protection
for the nonavalent [Figure 3, E and F]). Our model also predicts
that a nonavalent vaccine with an average 30 years of vaccine-type
protection would produce greater population-level effectiveness
against CIN2 and -3 and SCC than bivalent and quadrivalent
vaccines with lifelong vaccine protection (Figure 3, E and F).
Finally, the bivalent, quadrivalent, and nonavalent vaccines
are predicted to substantially reduce adenocarcinoma and other
HPV-related cancers (Figure 4). On the other hand, given that
HPV16 and -18 (primarily HPV16) are present in more than 90%
of vaginal, anal, and head and neck cancers attributed to HPV
(7–9), there is little difference between HPV vaccines in terms of
population-level effectiveness against these cancers (ie, differences
in vaccine efficacy against HPV31, -33, -45, -52, and -58 have little
impact on these cancers).
Our modeling analysis indicates that differences between the
bivalent and quadrivalent vaccines in preventing cervical and other
HPV-related cancers will be relatively small and would only be
observable decades after the start of vaccination programs. On the
other hand, the quadrivalent vaccine is expected to substantially
reduce AGW cases within the first decade of a girls-only
vaccination program. Furthermore, even though the bivalent and
quadrivalent vaccines are expected to substantially reduce HPV-related
diseases, switching to a nonavalent vaccine (if proven efficacious)
could produce substantial incremental gains in reduction of
diagnosed lesions and cervical cancer but only marginal benefits in the
prevention of other HPV-related diseases.
Our results have important implications for clinicians and
policymakers. They provide evidence of the comparative
populationlevel benefits of the bivalent and quadrivalent vaccines in the
short- to long-term. To put our results into perspective, in Canada
(30 million individuals; approximately one-tenth of the US
population), a quadrivalent vaccination program of 12-year-old girls with
70% coverage is predicted to prevent 1.9 million diagnosed AGW
cases in men and women, 560 100 diagnosed CIN2 and -3 cases,
Figure 4. Other human papillomavirus (HPV)–related cancers. Reduction
in other HPV-related cancers at equilibrium for the bivalent,
quadrivalent and nonavalent vaccines assuming 50% (A) and 70% (B)
vaccination coverage. Base case: vaccine duration = lifetime, vaccine-type
efficacy = 100%, cross-protective vaccine efficacy presented in Table 1.
ADC = adenocarcinoma. Population of 170 000 individuals. Each
parameter set was run 25 times.
and 20 800 SCC cases over 70 years. Switching from the
quadrivalent to the bivalent vaccine would produce a rebound in
diagnosed AGW cases (increase cases by 1.8 million) but would prevent
an additional 42 600 diagnosed CIN2 and -3 cases and 1400 SCC
cases over the same period. Given the long time-lag between the
age at vaccination and disease, the full benefit in prevention of
CIN2 and -3 and cervical cancer by switching from a quadrivalent
to a bivalent vaccine is not expected for 20–40 years after the start
of the vaccination programs. It should be pointed out that these
predictions assumed lifelong vaccine protection (vaccine and
crossprotective types). If the duration of cross-protection is shorter (eg,
10 years) for both vaccines, then the bivalent and quadrivalent
vaccines are predicted to produce very similar vaccine
effectiveness against HPV-related cancers. On the other hand, important
incremental benefits are predicted if the bivalent vaccine confers
substantially greater duration of protection than the quadrivalent
vaccine. Clearly, an important priority for future research is to
better understand whether the higher immunogenicity measured for
the bivalent vs the quadrivalent vaccine has implications on
longterm vaccine protection (16,17).
The relatively small incremental benefit of the bivalent
vaccine over the quadrivalent vaccine, when duration of protection
is similar, can be explained by the fact that more than 70%–80%
of cervical cancers (6) and 90% of other HPV-related cancers are
due to HPV16 and -18 (7–9) (HPV16 mostly for noncervical sites),
against which both vaccines are highly efficacious. Furthermore,
the additional type-specific vaccine efficacy provided by the
bivalent vs the quadrivalent vaccine is mainly limited to HPV31, -33,
and -45 (Table 1).
If the nonavalent HPV vaccine proves to be safe and effective,
policy makers will have to decide whether or not to introduce the
second-generation vaccine. In Canada, changing from a
quadrivalent to a nonavalent vaccine is predicted to prevent an additional
171 200 diagnosed CIN2 and -3 cases and 4700 SCC cases over
70 years assuming 100% vaccine efficacy, 70% coverage, and
vaccination of 12-year-old girls. The success of this
second-generation vaccine will depend on whether high HPV16 and -18 vaccine
efficacy will be achieved and maintained, and the level and
duration of cross-protection produced by the current HPV vaccines.
Our model predicts that vaccine efficacy against the vaccine types
would have to be greater than 80%–85% and duration of
protection longer than 30 years for the nonavalent vaccine to produce
incremental gains compared with a perfect bivalent or quadrivalent
vaccine (100% vaccine efficacy and lifelong protection). Given the
importance of protecting against HPV16 and -18 infection and
disease, it is unlikely that the nonavalent would be used if its efficacy
against these types is shown to be lower than the current HPV
vaccines. However, a point of uncertainty that will likely remain
when decisions are made regarding the nonavalent vaccine will be
its relative duration compared with the current vaccines.
The study has several methodological strengths. We developed
HPV-ADVISE, the first HPV microsimulation model that fully
integrates the various levels of heterogeneity across the
HPVrelated disease control spectrum—sexual behavior, health-seeking
behavior (vaccination and screening), type-specific natural history
of infection and disease, and vaccine efficacy. Second, the model
was calibrated to highly stratified data on sexual behavior, natural
history, and cervical cancer screening to identify multiple
goodfitting parameter sets. Using multiple parameter sets is important
to 1) capture uncertainty in key parameters because type-specific
and age-specific empirical data on the natural history of HPV
infection and disease are incomplete (26,45,59,60) and 2) increase
the robustness of predictions to changes in our understanding of
the biology and epidemiology of HPV (eg, natural immunity) (61).
Finally, the impact of vaccination was estimated by modeling HPV
types individually using the most recent type-specific estimates of
vaccine cross-protection (10,12–14,29). Previous modeling studies
that have examined the impact of cross-protection have grouped
multiple non-HPV16 and -18 types and used combined estimates
of vaccine efficacy from clinical trials (eg, efficacy against HPV31,
-33, -45, -52, and -58 infection or CIN2+ lesions). It has been shown
that grouping multiple types should be avoided when modeling
HPV vaccine effectiveness because it produces a fictional “super
bug,” which can lead to biased predictions (26). Using vaccine
efficacy against grouped HPV types can also lead to biased results
(29). This is because grouped vaccine efficacy is dependent on the
efficacy against each individual type and the relative distribution
of these types in the clinical trial population may differ from the
distribution in the population that is modeled.
Our study has limitations. First, estimates of type-specific
efficacies against nonvaccine HPV types for both vaccines
were derived from clinical trials with different populations and
designs (10,12,14), which could partly explain the differences in
cross-protection estimates between the vaccines. However, we
performed a systematic literature review to select similar vaccine
efficacy outcomes extracted from the most comparable populations
(ie, HPV-naive females) (29). Second, similarly to other studies
(23,28), the natural history of noncervical HPV-related cancers was
not modeled explicitly. Instead, model predictions of the reduction
in type-specific HPV prevalence at equilibrium were imputed to
the distribution of these types within the different HPV-related
cancers. Because HPV16 and -18 are the cause of most noncervical
HPV-related cancers, including more complex natural histories is
unlikely to influence our conclusion that the effectiveness against
other HPV-related cancers will be very similar between the HPV
vaccines (assuming equal vaccine-type efficacy and duration).
Third, similar to the majority of HPV vaccination modeling studies
(21,62–64), only heterosexual transmission was included. Given
that the probability of transmission and prevalence of HPV is very
high and that the population of men who have sex with men is
estimated to be small [3%–5% (34)], men who have sex with men are
unlikely to influence overall HPV transmission at the
populationlevel. Therefore, including men who have sex with men in our
model is expected to have little impact on our predictions of the
relative impact of girls-only bivalent, quadrivalent, and candidate
nonavalent vaccination programs. Fourth, similarly to all other
dynamic models of HPV vaccination in industrialized countries, we
did not model HIV as a potential modifying factor of HPV natural
history or vaccine efficacy because HIV prevalence is too low (65)
to influence the population-level effectiveness of girls-only HPV
vaccination. Finally, similar to the other HPV dynamic modeling
studies (21,62–64), we assume that sexual behavior remains constant
over time. Population-level data from North America suggest that
age at sexual debut, having multiple partners, and condom use have
remained relatively stable in the past 15–20 years (73–75). However,
even if sexual behavior changes in the next decades, this is unlikely
to impact our model conclusions given that our main outcomes are
comparisons in relative population-level effectiveness across HPV
vaccines (rather than absolute differences). To considerably modify
the relative effectiveness between the HPV vaccines, changes in
sexual activity would have to differentially impact acquisition of the
different HPV types (eg, HPV16 and -18 versus HPV31, -33, -45,
-52, and-58) and thus substantially influence the distribution of
types among cervical lesions, which is unlikely. However, if sexual
activity changes substantially, absolute differences in
populationlevel effectiveness for vaccination versus no vaccination would
increase or decrease in parallel and could also affect the relative
impact of the different HPV vaccines.
Although HPV-ADVISE was calibrated to Canadian data, our
main conclusions are relevant to other developed countries given
similarities in sexual behavior (34,66), HPV type distribution
(15,67,68), age profile of HPV prevalence (69), and cervical
cancer incidence and mortality rates between industrialized countries
(70). However, our results must be extrapolated to resource-poor
settings with caution due to differences in sexual behavior (34,66),
HPV epidemiology (69,70) and potential cofactors of HPV
infection and disease, such as high HIV prevalence (71).
This is the first modeling paper to provide evidence on 1) the
potential comparative population-level effectiveness of the
bivalent and quadrivalent HPV vaccines, taking into account the most
recent type-specific cross-protective efficacy estimates and
herdimmunity, 2) the impact of switching HPV vaccines within ongoing
programs, and 3) the potential additional benefits of a candidate
nonavalent vaccine. Population-level effectiveness is only one of
several important criteria in decisions regarding vaccination (72).
Although outside the scope of this study, future research should aim
at examining the cost-effectiveness of the bivalent and quadrivalent
vaccines and the price differential in order for both vaccines to be
deemed equally cost effective. Although previous models suggest
that the bivalent vaccine needs to be cheaper than the quadrivalent
vaccine to produce equal cost-effectiveness results, these studies
did not incorporate either type-specific cross-protection or herd
immunity (20–23,25). Future research should also focus on
identifying which outcome (persistent infection, CIN2+ including or
excluding coinfected lesions with HPV16 and/or -18) is the most
valid estimate of cross-protective vaccine efficacy and, most
important, estimating the duration of vaccine protection.
In summary, our model shows that HPV vaccines can
substantially reduce HPV-related diseases. Based on the higher
cross-protective efficacy reported in the current clinical trials, the bivalent
vaccine is expected to be slightly more effective at preventing
diagnosed cervical precancerous lesions and cervical cancer in the
longer term, whereas the quadrivalent vaccine will substantially
reduce AGW cases. Finally, the candidate nonavalent vaccine has
the potential to produce substantial incremental benefits.
Canada Research Chairs program (to MB); a team grant (no. 83320) from
the Canadian Institutes of Health Research; unrestricted research grant from
Merck Frosst Canada Ltd (to the Centre de recherche FRSQ du CHA
universitaire de Québec; co-PIs MB and M-CB); Frederick Banting and Charles
Best Canada Graduate Scholarships from the Canadian Institutes of Health
Research (to TM).
N. Van de Velde, M. Brisson, and M.-C. Boily designed HPV-ADVISE and had
full access to all of the data in the study and take responsibility for the integrity
of the data and the accuracy of the data analysis. N. Van de Velde programmed
the model. N. Van de Velde, M. Brisson, M.-C. Boily, and J.-F. Laprise tested
the model. E. Franco, E. Kliewer, F. Coutlée, and M.-H. Mayrand provided the
data necessary for model calibration and validation and provided comments on
the model structure. N. Van de Velde, M. Brisson, M. Drolet, T. Malagón, and
J.-F. Laprise performed the analysis. N. Van de Velde, M. Brisson, and M. Drolet
drafted the manuscript. All authors contributed to the interpretation of results,
critically revised the manuscript for important intellectual content, and approved
the final version submitted for publication. The funders had no role in design
and conduct of the study; collection, management, analysis, and interpretation
of the data; preparation, review, or approval of the manuscript; or the decision to
submit the manuscript for publication.
N. Van de Velde has consulted for Sanofi Aventis MSD and Merck.
M. Brisson has consulted and received reimbursement for travel expenses from
Merck Frosst and GlaxoSmithKline. E. Franco has served as occasional
consultant or advisory board member for Merck and GlaxoSmithKline. M.-H. Mayrand
is an FRSQ clinical research scholar. She received honoraria for lectures from
Merck and GlaxoSmithKline. Her institution has received unrestricted research
grants from Merck and Qiagen for her research projects. She is involved in
the nonavalent vaccine trial. E. Kliewer has consulted for Merck Frosst and
GlaxoSmithKline and has received reimbursement for travel expenses from
Merck Frosst. F. Coutlée participated in opinion expert meetings organized by
Qiagen in 2009 and GlaxoSmithKline in 2010, received a research grant from
Roche diagnostics in 2005–2007 and from Merck in 2006–2008, and received
honoraria for lectures from Roche Diagnostics (2009) and Merck (2011). All
other authors declare no conflicts of interest.
We thank Dr Jacques Brisson, PhD, of Laval University, for comments on
the manuscript and are indebted to the Imperial College High Performance
Computing Service for providing us with the computing power necessary to run
1. de Villiers EM , Fauquet C , Broker TR , et al. Classification of papillomaviruses. Virology . 2004 ; 324 ( 1 ): 17 - 27 .
2. Health Canada . Human papillomavirus (HPV). http://www.hc-sc.gc . ca/ hl-vs/alt_formats/pdf/iyh-vsv/diseases-maladies/hpv-vph-eng .pdf. Accessed February 1 , 2011 .
3. Garland SM , Steben M , Sings HL , et al. Natural history of genital warts: analysis of the placebo arm of 2 randomized phase III trials of a quadrivalent human papillomavirus (types 6, 11, 16, and 18) vaccine . J Infect Dis . 2009 ; 199 ( 6 ): 805 - 814 .
4. Pou AM , Rimell FL , Jordan JA , et al. Adult respiratory papillomatosis: human papillomavirus type and viral coinfections as predictors of prognosis . Ann Otol Rhinol Laryngol . 1995 ; 104 ( 10 )(pt 1): 758 - 762 .
5. Bouvard V , Baan R , Straif K , et al. A review of human carcinogens-part B: biological agents . Lancet Oncol . 2009 ; 10 ( 4 ): 321 - 322 .
6. Munoz N , Bosch FX , de Sanjose S , et al. Epidemiologic classification of human papillomavirus types associated with cervical cancer . N Engl J Med . 2003 ; 348 ( 6 ): 518 - 527 .
7. Backes DM , Kurman RJ , Pimenta JM , et al. Systematic review of human papillomavirus prevalence in invasive penile cancer . Cancer Causes Control . 2009 ; 20 ( 4 ): 449 - 457 .
8. De Vuyst H , Clifford GM , Nascimento MC , et al. Prevalence and type distribution of human papillomavirus in carcinoma and intraepithelial neoplasia of the vulva, vagina and anus: a meta-analysis . Int J Cancer . 2009 ; 124 ( 7 ): 1626 - 1636 .
9. Kreimer AR , Clifford GM , Boyle P , et al. Human papillomavirus types in head and neck squamous cell carcinomas worldwide: a systematic review . Cancer Epidemiol Biomarkers Prev . 2005 ; 14 ( 2 ): 467 - 475 .
10. Paavonen J , Naud P , Salmeron J , et al. Efficacy of human papillomavirus (HPV)-16/18 AS04-adjuvanted vaccine against cervical infection and precancer caused by oncogenic HPV types (PATRICIA): final analysis of a doubleblind, randomised study in young women . Lancet . 2009 ; 374 ( 9686 ): 301 - 314 .
11. Garland SM , Hernandez-Avila M , Wheeler CM , et al. Quadrivalent vaccine against human papillomavirus to prevent anogenital diseases . N Engl J Med . 2007 ; 356 ( 19 ): 1928 - 1943 .
12. Brown DR , Kjaer SK , Sigurdsson K , et al. The impact of quadrivalent human papillomavirus ( HPV; types 6 , 11 , 16, and 18 ) L1 virus-like particle vaccine on infection and disease due to oncogenic nonvaccine HPV types in generally HPV-naive women aged 16-26 years . J Infect Dis . 2009 ; 199 ( 7 ): 926 - 935 .
13. Wheeler CM , Kjaer SK , Sigurdsson K , et al. The impact of quadrivalent human papillomavirus ( HPV; types 6 , 11 , 16, and 18 ) L1 virus-like particle vaccine on infection and disease due to oncogenic nonvaccine HPV types in sexually active women aged 16-26 years . J Infect Dis . 2009 ; 199 ( 7 ): 936 - 944 .
14. Romanowski B. Efficacy of the HPV-16/18 AS04-adjuvanted vaccine against non-vaccine oncogenic HPV types: end-of-study results . Paper presented at 26th International Papillomavirus Conference; July 2010 ; Montreal, Canada.
15. de Sanjose S , Quint WG , Alemany L , et al. Human papillomavirus genotype attribution in invasive cervical cancer: a retrospective cross-sectional worldwide study . Lancet Oncol . 2010 ; 11 ( 11 ): 1048 - 1056 .
16. Olsson SE , Villa LL , Costa RL , et al. Induction of immune memory following administration of a prophylactic quadrivalent human papillomavirus (HPV) types 6/11/16/18 L1 virus-like particle (VLP) vaccine . Vaccine . 2007 ; 25 ( 26 ): 4931 - 4939 .
17. David MP , Van Herck K , Hardt K , et al. Long-term persistence of antiHPV-16 and -18 antibodies induced by vaccination with the AS04- adjuvanted cervical cancer vaccine: modeling of sustained antibody responses . Gynecol Oncol . 2009 ; 115 ( 3 ) (suppl):S1-6.
18. Read TR , Hocking JS , Chen MY , et al. The near disappearance of genital warts in young women 4 years after commencing a national human papillomavirus (HPV) vaccination programme . Sex Transm Infect . 2011 ; 87 ( 7 ): 544 - 547 .
19. Jit M , Brisson M. Modelling the epidemiology of infectious diseases for decision analysis: a primer . Pharmacoeconomics . 2011 ; 29 ( 5 ): 371 - 386 .
20. Brisson M , Van de Velde N , De Wals P , et al. The potential cost-effectiveness of prophylactic human papillomavirus vaccines in Canada . Vaccine. 2007 ; 25 ( 29 ): 5399 - 5408 .
21. Jit M , Choi YH , Edmunds WJ . Economic evaluation of human papillomavirus vaccination in the United Kingdom . BMJ. 2008 ; 337 :a769.
22. Dee A , Howell F. A cost-utility analysis of adding a bivalent or quadrivalent HPV vaccine to the Irish cervical screening programme . Eur J Public Health . 2010 ; 20 ( 2 ): 213 - 219 .
23. Jit M , Chapman R , Hughes O , et al. Comparing bivalent and quadrivalent human papillomavirus vaccines: economic evaluation based on transmission model . BMJ . 2011 ; 343 :d5775.
24. Capri S , Gasparini R , Panatto D , et al. Cost-consequences evaluation between bivalent and quadrivalent HPV vaccines in Italy: the potential impact of different cross-protection profiles . Gynecol Oncol . 2011 ; 121 ( 3 ): 514 - 521 .
25. Lee VJ , Tay SK , Teoh YL , et al. Cost-effectiveness of different human papillomavirus vaccines in Singapore . BMC Public Health . 2011 ; 11 (March 31): 203 .
26. Van de Velde N , Brisson M , Boily MC . Understanding differences in predictions of HPV vaccine effectiveness: A comparative model-based analysis . Vaccine . 2010 ; 28 ( 33 ): 5473 - 5484 .
27. Young J , Roffers S , Ries L , et al. SEER Summary Staging Manual - 2000 : Codes and Coding Instructions . Bethesda, MD: National Cancer Institute, National Institutes of Health ; 2001 . NIH publication 01-4969.
28. Smith MA , Lew JB , Walker RJ , et al. The predicted impact of HPV vaccination on male infections and male HPV-related cancers in Australia . Vaccine. 2011 ; 29 ( 48 ): 9112 - 9122 .
29. Malagon T , Drolet M , Boily M , et al.. Issues in comparing and interpreting clinical trial measures of cross-protection . Paper presented at 27th International Papillomavirus conference; September 2011 ; Berlin, Germany.
30. Wheeler CM , Castellsague X , Garland SM , et al. Cross-protective efficacy of HPV-16/18 AS04-adjuvanted vaccine against cervical infection and precancer caused by non-vaccine oncogenic HPV types: 4-year end-of-study analysis of the randomised, double-blind PATRICIA trial . Lancet Oncol . 2012 ; 13 ( 1 ): 100 - 110 .
31. Quadrivalent vaccine against human papillomavirus to prevent high-grade cervical lesions . N Engl J Med . 2007 ; 356 ( 19 ): 1915 - 1927 .
32. Naud P. Sustained immunogenicity and efficacy of the HPV-16/18 vaccine in women aged 15-25 years: follow-up to 9.4 years . Paper presented at 27th International Papillomavirus Conference and Clinical Workshop; September 2011 ; Berlin, Germany.
33. Roteli-Martins C , Naud P , De Borba P , et al. Sustained immunogenicity and efficacy of the HPV-16/18 AS04-adjuvanted vaccine: up to 8.4 years of follow-up . Hum Vaccin Immunother . 2012 ; 8 ( 3 ): 390 - 397 .
34. Statistics Canada. Canadian community health survey (CCHS- Cycle 3 . 1 - 2005 ). http://www23.statcan.gc.ca/imdb/p2SV.pl? Function=getSurvey& SurvId=3226&SurvVer=0&InstaId=15282&InstaVer=3&SDDS=3226& lang=en&db=imdb&adm=8&dis=2. Accessed January 1 , 2011 .
35. Demers A , Kliewer EV , Musto G , et al. Epidemiology of Cervical Abnormalities and Utilization of Related Health Care Resources . Winnipeg, Canada: CancerCare Manitoba ; 2009 .
36. BC Cancer Agency. Screening for the Cancer of the Cervix: An Office Manual for Health Professionals . 9th ed . Vancouver, Canada: BC Cervical Cancer Screening Program ; 2010 .
37. McLachlin CM , Mai V , Murphy J , et al. Ontario cervical cancer screening clinical practice guidelines . J Obstet Gynaecol Can . 2007 ; 29 ( 4 ): 344 - 353 .
38. CancerCare Manitoba. Manitoba Cervical Cancer Screening Program screening guidelines . http://www.cancercare.mb.ca/home/prevention_ and_screening/professional_screening_programs/cervical_cancer_screening/downloadorder_publications/. Accessed January 1 , 2011 .
39. Nanda K , McCrory DC , Myers ER , et al. Accuracy of the Papanicolaou test in screening for and follow-up of cervical cytologic abnormalities: a systematic review . Ann Intern Med . 2000 ; 132 ( 10 ): 810 - 819 .
40. Arbyn M , Bergeron C , Klinkhamer P , et al. Liquid compared with conventional cervical cytology: a systematic review and meta-analysis . Obstet Gynecol . 2008 ; 111 ( 1 ): 167 - 177 .
41. Gage JC , Hanson VW , Abbey K , et al. Number of cervical biopsies and sensitivity of colposcopy . Obstet Gynecol . 2006 ; 108 ( 2 ): 264 - 272 .
42. Chase DM , Kalouyan M , DiSaia PJ . Colposcopy to evaluate abnormal cervical cytology in 2008 . Am J Obstet Gynecol . 2009 ; 200 ( 5 ): 472 - 480 .
43. Da Forno PD , Holbrook MR , Nunns D , et al. Long-term follow-up of patients following negative colposcopy: a new gold standard and its implications for cervical screening . Cytopathology . 2003 ; 14 ( 5 ): 281 - 286 .
44. Cai B , Ronnett BM , Stoler M , et al. Longitudinal evaluation of interobserver and intraobserver agreement of cervical intraepithelial neoplasia diagnosis among an experienced panel of gynecologic pathologists . Am J Surg Pathol . 2007 ; 31 ( 12 ): 1854 - 1860 .
45. Van de Velde N , Brisson M , Boily MC . Modeling human papillomavirus vaccine effectiveness: quantifying the impact of parameter uncertainty . Am J Epidemiol . 2007 ; 165 ( 7 ): 762 - 775 .
46. Drolet M , Brisson M , Maunsell E , et al. The psychosocial impact of an abnormal cervical smear result [published online ahead of print June 21, 2011 ]. Psychooncology . 2011 ; doi:10.1002/pon. 2003 .
47. Drolet M , Brisson M , Maunsell E , et al. The impact of anogenital warts on health-related quality of life: a 6-month prospective study . Sex Trans Dis . 2011 ; 38 ( 10 ): 949 - 956 .
48. Ades S , Koushik A , Duarte-Franco E , et al. Selected class I and class II HLA alleles and haplotypes and risk of high-grade cervical intraepithelial neoplasia . Int J Cancer . 2008 ; 122 ( 12 ): 2820 - 2826 .
49. Richardson H , Kelsall G , Tellier P , et al. The natural history of type-specific human papillomavirus infections in female university students . Cancer Epidemiol Biomarkers Prev . 2003 ; 12 ( 6 ): 485 - 490 .
50. Mayrand MH , Duarte-Franco E , Coutlee F , et al. Randomized controlled trial of human papillomavirus testing versus Pap cytology in the primary screening for cervical cancer precursors: design, methods and preliminary accrual results of the Canadian cervical cancer screening trial (CCCaST) . Int J Cancer . 2006 ; 119 ( 3 ): 615 - 623 .
51. Coutlee F , Ratnam S , Ramanakumar AV , et al. Distribution of human papillomavirus genotypes in cervical intraepithelial neoplasia and invasive cervical cancer in Canada . J Med Virol . 2011 ; 83 ( 6 ): 1034 - 1041 .
52. Li N , Franceschi S , Howell-Jones R , et al. Human papillomavirus type distribution in 30,848 invasive cervical cancers worldwide: variation by geographical region, histological type and year of publication . Int J Cancer . 2011 ; 128 ( 4 ): 927 - 935 .
53. Liu S , Semenciw R , Probert A , et al. Cervical cancer in Canada: changing patterns in incidence and mortality . Int J Gynecol Cancer . 2001 ; 11 ( 1 ): 24 - 31 .
54. Statistics Canada. Health in Canada-CANSIM tables . http://www5. statcan.gc.ca/cansim/a33?RT= TABLE&themeID=1887&spMode=tables &lang=eng . Accessed September 10 , 2012 .
55. BC Cancer Agency. Cervical cancer screening program. 2009 annual report . http://www.bccancer.bc.ca/PPI/Screening/Cervical/about.htm. Accessed January 1 , 2011 .
56. Iman R , Conover W. Small sample sensitivity analysis techniques for computer models, with an application to risk assessment . Commun Stat Theor Meth . 1980 ;A9: 1749 - 1874 .
57. McKay M , Conover W , Beckman R. A comparison of three methods for selecting values of input variables in the analysis of output from a computer code . Technometrics . 1979 ; 21 ( 2 ): 239 - 246 .
58. Blower SM , McLean AR , Porco TC , et al. The intrinsic transmission dynamics of tuberculosis epidemics . Nat Med . 1995 ; 1 ( 8 ): 815 - 821 .
59. Kim JJ , Kuntz KM , Stout NK , et al. Multiparameter calibration of a natural history model of cervical cancer . Am J Epidemiol . 2007 ; 166 ( 2 ): 137 - 150 .
60. Jit M , Gay N , Soldan K , et al. Estimating progression rates for human papillomavirus infection from epidemiological data . Med Decis Making . 2010 ; 30 ( 1 ): 84 - 98 .
61. Gravitt PE . The known unknowns of HPV natural history . J Clin Invest . 2011 ; 121 ( 12 ): 4593 - 4599 .
62. Kim JJ , Brisson M , Edmunds WJ , et al. Modeling cervical cancer prevention in developed countries . Vaccine . 2008 ; 26 (suppl 10): K76 - 86 .
63. Brisson M , Van de Velde N , Boily MC . Economic evaluation of human papillomavirus vaccination in developed countries . Public Health Genomics . 2009 ; 12 ( 5-6 ): 343 - 351 .
64. Kim JJ , Goldie SJ . Health and economic implications of HPV vaccination in the United States . N Engl J Med . 2008 ; 359 ( 8 ): 821 - 832 .
65. Public Health Agency of Canada. National HIV prevalence and incidence estimates in Canada for 2008 . http://www.phac-aspc.gc.ca/aids-sida/publication/epi/2010/1-eng.php. Accessed May 1 , 2012 .
66. Wellings K , Collumbien M , Slaymaker E , et al. Sexual behaviour in context: a global perspective . Lancet . 2006 ; 368 ( 9548 ): 1706 - 1728 .
67. Guan P , Howell-Jones R , Li N , et al. Human papillomavirus types in 115,789 HPV-positive women: a meta-analysis from cervical infection to cancer [published online ahead of print February 9 , 2012 ]. Int J Cancer . 2012 ; doi:10.1002/ijc.27485.
68. Clifford GM , Smith JS , Aguado T , et al. Comparison of HPV type distribution in high-grade cervical lesions and cervical cancer: a meta-analysis . Br J Cancer . 2003 ; 89 ( 1 ): 101 - 105 .
69. Bruni L , Diaz M , Castellsague X , et al. Cervical human papillomavirus prevalence in 5 continents: meta-analysis of 1 million women with normal cytological findings . J Infect Dis . 2010 ; 202 ( 12 ): 1789 - 1799 .
70. International Agency for Research in Cancer. Globocan 2008 . Cervical cancer incidence and mortality worlwide 2008 . Available at: http://globocan.iarc.fr /factsheets/cancers/cervix.asp. Accessed May 1 , 2012 .
71. Adler DH , Kakinami L , Modisenyane T , et al. Increased regression and decreased incidence of HPV-related cervical lesions among HIV-infected women on HAART . AIDS. 2012 ; 26 ( 13 ): 1645 - 1652 .
72. Erickson LJ , De Wals P , Farand L. An analytical framework for immunization programs in Canada . Vaccine. 2005 ; 23 ( 19 ): 2470 - 2476 .
73. GlaxoSmithKline . Results summary for 580299/008 . http://www.gskclinicalstudyregister. com/result_detail.jsp?protocolId=580299%2f008& studyId=063E1C2C-1A99-427D-83F6-525AB3791945&compound= Human+Papillomavirus+Types+16+and+18+Vaccine. Accessed January 1 , 2011 .