CLINICAL REVIEW David W.

Eisele, MD, Section Editor

PREVALENCE OF HUMAN PAPILLOMAVIRUS IN

OROPHARYNGEAL AND NONOROPHARYNGEAL HEAD AND
NECK CANCER—SYSTEMATIC REVIEW AND META-ANALYSIS
OF TRENDS BY TIME AND REGION
Hisham Mehanna, PhD,1 Tom Beech, MSc,1 Tom Nicholson, MBChB,1 Iman El-Hariry, MD, PhD,2
Christopher McConkey, MSc,3 Vinidh Paleri, MS FRCS(ORL-HNS),4 Sally Roberts, DPhil5
1
Institute of Head and Neck Studies and Education, Coventry, United Kingdom. E-mail: [email protected]
2
Synta Pharmaceuticals, Lexington, MA, USA
3
Warwick Clinical Trials Unit, Warwick Medical School, Coventry, United Kingdom
4
Consultant Head & Neck, Department of Otolaryngology–Head and Neck Surgery, Newcastle upon Tyne Hospitals,
National Health Service Foundation Trust, Newcastle upon Tyne, United Kingdom and Honorary Clinical Senior Lecturer,
Northern Institute for Cancer Research, Newcastle University, United Kingdom
5
School of Cancer Sciences, Birmingham Cancer Research UK Cancer Centre, College of Medical and Dental Sciences,
University of Birmingham, Vincent Drive, Birmingham, United Kingdom

Accepted 12 October 2011

Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/hed.22015

to a rise in human papillomavirus (HPV)-related oro-

Abstract: Background. Little information has been reported
on regional and time trends of human papillomavirus (HPV) prev- pharyngeal cancer (HPV-OPC). According to the
alence rates of oropharyngeal cancer (OPC) and non-OPC. International Agency for the Research on Cancer
Methods. The study consisted of a systematic review and monograph, HPV appears to fulfill the criteria for
meta-analysis using random effects logistic regression models. causality of this particular cancer,1 and the disease is
Results. Overall HPV prevalence in OPC (47.7%; 95% confi- thought to be sexually transmitted.2 A systematic
dence interval [CI], 42.9–52.5%) increased significantly over review in 20053 noted that the prevalence of HPV
time: from 40.5% (95% CI, 35.1–46.1) before 2000, to 64.3% infection in OPC was 35.9% (95% confidence interval
(95% CI, 56.7–71.3) between 2000 and 2004, and 72.2% (95% [CI], 32.6–38.%). It was also noted that the preva-
CI, 52.9–85.7) between 2005 and 2009 (p < .001). Prevalence lence of HPV in OPC in the United States (47%; 95%
increased significantly in North America and Europe, and the
CI, 41.1–53%) was significantly higher than that in
significant gap between them that existed before 2000 (50.7%
vs 35.3%, respectively, p ¼ .008) has now disappeared (69.7% Europe and in Asia (28%; 95% CI, 24.4–32.2%). Three
vs 73.1%, respectively, p ¼ .8). Prevalence in non-OPC (21.8%; years later, a second review4 reported little change,
95% CI, 18.9–25.1%) has not increased over time (p ¼ .97) with HPV prevalence in OPC reported as 38.1%.
Conclusions. The sharp increase in the proportion of HPV- Recent studies, however, appear to suggest a rise in
positive OPC over the last decade has occurred at a faster prevalence of HPV-OPC. A retrospective cohort study
rate in Europe compared with that in North America. In con- from Stockholm, Sweden, noted a progressive increase
trast, the relatively low prevalence of HPV in non-OPC remains in the proportion of OPC biopsies over the past 3 dec-
unchanged. V C 2012 Wiley Periodicals, Inc. Head Neck 00:
ades that tested positive for HPV, using the same
000–000, 2012 detection method (23.3% in 1970s, 29% in 1980s, 57%
Keywords: human papillomavirus; head and neck neoplasms; in 1990s, 68% between 2000 and 2002, 77% between
oropharyngeal neoplasms; prevalence; meta-analysis 2003 and 2005, and 93% between 2006 and 2007).5 A
randomized trial in the United States recently
In the last decade, the incidence of oropharyngeal reported HPV prevalence of 63.5% in 323 patients.6
A rapid rise in the prevalence of HPV-OPC would
cancer (OPC) has shown a sharp increase, in contrast
have significant implications for patients, health serv-
to other head and neck cancers. Many attribute this
ice providers, and commissioners.7 In contrast to alco-
Correspondence to: H. Mehanna
hol- and smoking-related head and neck cancers,
HPV-OPC affects younger patients, who are often
Additional Supporting Information may be found in the online version of employed.8 Importantly, HPV-OPC has a much better
this article. response rate to treatment.6 This means that many
H.M. is the director of a research institute that does contract work for more patients are likely to survive, and have to cope
Glaxo Smith Kline, and has served on advisory boards for GSK and with the considerable sequelae of treatment, for much
Sanofi Pasteur. I.E.H. works for a pharmaceutical company.
longer than was previously expected. Patients
V
C 2012 Wiley Periodicals, Inc. and caregivers may therefore require more support

Time and Regional Trends in HPV Prevalence in Head Neck Cancer HEAD & NECK—DOI 10.1002/hed Month 2012 1
from health and social care systems to enable lence of HPV infection in cancer samples (fresh-frozen
them to return to work and normal social activity. Im- or paraffin-embedded) from a population of patients
portantly, HPV-OPC is potentially preventable by with head and neck squamous cell carcinomas, using
vaccination against HPV infection.9,10 A rapid rise in polymerase chain reaction (PCR)-based methods to
prevalence may also necessitate a re-examination of amplify either DNA or cDNA, or in situ hybridization
the cost-effectiveness of HPV vaccination of boys.11 (ISH) or other HPV detection methods, or using p16
For these reasons, it is important to obtain an accu- immunostaining as a surrogate marker for HPV.
rate representation of the current global and regional Studies were excluded if abstracts were not written
trends in the prevalence of this disease to enable bet- in the English language; were animal studies; had a
ter health strategy planning and development. sample size <10; included tumors other than squa-
Although there has been much emphasis and discus- mous cell carcinoma; were not upper aerodigestive
sion on HPV and OPC, little is mentioned about the tract cancers; reported prevalence of both malignant
association of HPV and nonoropharyngeal head and
and benign lesions together; or undertook analysis
neck cancers (non-OPC), despite an abundance of avail-
only of serology or mucosal brushings.
able data in the literature. Prevalence rates of HPV in
non-OPC are reported to be lower than those for OPC.3,4
However, it is not known whether this is just a lag Study Designs Included. Cross-sectional studies,
phase and whether prevalence in non-OPC may be now cohorts, and patients from randomized trials were
increasing and emulating the rise of HPV in OPC. We included. Studies that were population based, hospital
have therefore undertaken a systematic review and based, from a cancer registry, or from a pathology
meta-analysis of all studies reporting on the prevalence archive were included. Both prospective and retro-
of HPV in patients with head and neck cancer. The spective studies were included.
objectives were to examine recent trends over time in
the prevalence of HPV infection in OPC and non-OPC
and to compare them by geographical region, to ascer- Data Abstraction and Study Outcomes. A standar-
tain whether the rise in prevalence observed in some dized, piloted data-extraction database was set up
recent studies can be confirmed, and whether differen- using a commercial software program (Excel; Micro-
ces in the regional prevalence of the disease still persist. soft, Redmond, WA). The 2 reviewers independently
abstracted data from the selected articles and were not
MATERIALS AND METHODS aware of the other reviewer’s results. Data were col-
lected on study and sample characteristics, country, and
Search Strategy. A systematic computer-based time period of collection of study sample, the primary
search was performed of the following databases: sites included, type of sample used (fresh-frozen or for-
Cochrane Oral Health and ENT Groups Trials Regis- malin-fixed paraffin-embedded), type of HPV assay
ter, MEDLINE, Zetoc, CINAHL (Cumulative Index to used, and the proportion that tested positive for HPV
Nursing and Allied Health Literature), and the and for HPV-16. The outcomes included were total num-
National Cancer Trials Database. All entries from ber of patients tested and the proportion that were HPV
1966 to 2010 were searched. We used the following positive in OPC (TNM classification anatomic codes
Medical Subject Headings (MeSH) terms (all in the C01, C05.1, C0.52, C09.0, C0.91, C0.99, C10.0, C10.2,
‘‘explode’’ function): ‘‘Head and neck neoplasms’’ or C10.3) samples and in all head and neck squamous cell
‘‘mouth neoplasms’’ or ‘‘oropharyngeal neoplasms’’ or cancer samples in total. Outcomes were collated by
‘‘pharyngeal neoplasms’’ or ‘‘hypopharyngeal neo- region and by time period of sample collection (before
plasms’’ or ‘‘laryngeal neoplasms’’ AND ‘‘Carcinoma, 2000, 2000–2004, and 2005 to the present).
squamous cell’’ AND ‘‘Papillomaviridae’’ or ‘‘papilloma-
virus infections’’ or ‘‘DNA, viral.’’ To avoid publication
bias, reference lists of the articles retrieved and previ- Quality and Risk of Bias Appraisal. The quality of
ous reviews were manually searched for suitable stud- and risk of bias within the papers was critically
ies, as were the gray literature and conference abstract appraised separately by the 2 reviewers using scoring
books of the meetings of the American Society of Clini- criteria based on those proposed by Hayden et al12 and
cal Oncologists, European Society of Medical Oncolo- the PRISMA (Preferred Reporting Items for System-
gists and American Head and Neck Society from 2006 atic Reviews and Meta-Analyses) guidelines.13 For ev-
to 2010. Where data were missing, authors of studies ery included study, the presence or absence of each of
and meeting abstracts were contacted to provide the the 8 criteria was assessed. The criteria assessed were:
data. A full review protocol is available on request. whether recruitment period was stated; was the sam-
ple size >50; was the study prospective; were the cases
Inclusion and Exclusion Criteria. After receiving recruited consecutively; did the study adequately
training from the first author, 2 clinicians (T.B., T.N.) describe the patient characteristics; were >80% of
independently reviewed the abstracts and selected cases analyzed; did the study specify the sample type;
them for detailed assessment of the full articles. and the detection technique? The types of assay used
Abstracts were selected if they reported on the preva- were assessed to ensure that they were fit for purpose.

2 Time and Regional Trends in HPV Prevalence in Head Neck Cancer HEAD & NECK—DOI 10.1002/hed Month 2012
 At all times, any differences in opinion were resolved In all, 102 studies reported on HPV prevalence in 5396
by discussion, then by adjudication to a third reviewer patients with OPC, and 236 studies reported on HPV
(H.M.). If multiple studies used the same data, the most prevalence in 13,972 patients with non-OPC; 206 studies
recent study was used. Where a meeting abstract and a used PCR to detect HPV, 48 used ISH, and 11 used both.
subsequent manuscript reported data on the same Three studies scored HPV positivity in the tumors by
cohort, the published manuscript data were used. Southern blotting of isolated DNA and 3 studies by
immunohistochemistry for HPV antigens, both methods
Statistical Analysis. Transformation rates were considered to be of low sensitivity.15 Eighteen studies
modeled by applying a random effects logistic regres- reported p16 immunohistochemistry results as a surro-
sion model that assumed a fixed effect of the underly- gate marker of infection with high-risk HPV, either
ing mean log-odds of transformation rate and a reported as the only result in 4 studies, or in combination
random study effect. This allowed estimation of the in 14 studies with the other HPV detection methods. A
underlying mean and testing of the effect covariates total of 162 studies used paraffin-embedded tissues, 72
such as year of study, region, and method of detection. used fresh-frozen tissues, and 15 using a combination of
Analysis by year of study was based on the mid- paraffin-embedded and fresh-frozen samples. Twenty
point between the first and last years of the cohort. studies did not specify details of sample type used.
This was used in models separately as a continuous
variable and grouped as: <2000, 2000–2004, 2005þ. A Quality and Heterogeneity of Included Studies. The
forest plot was generated showing individual point quality assessment of the included studies is detailed in
estimates in each study, by study size. Appendix 2 (online Supporting Information). Of the 269
Heterogeneity, calculated by Pearson’s chi-squared studies included, 37 studies were prospective, 117 stud-
test, was studied using the I2 index statistic, which ies were retrospective, and the remaining 15 did not
estimates the proportion of total variation due to het- specify this. In all, 223 studies reported results on
erogeneity.14 I2 scores of <25% suggest low heteroge- >80% of their patient sample. Two studies recruited
neity and those of
75% suggest a high degree of patients from a national cancer registry, 171 recruited
heterogeneity. Funnel plots, showing the individual patients at hospital outpatient departments, 80
study proportion testing HPV positive in relation to recruited cases from pathology archives, and 7 recruited
study size, were also used to assess bias. patients as part of a randomized trial. Six studies did
Sensitivity analyses of the likely sources of bias not detail the source of patient recruitment.
were performed by excluding low-quality studies scor- A high degree of heterogeneity was seen in both
ing 3 in the quality criteria. We also performed other the studies on OPC (I2 ¼ 0.89, p < .001), and in the
sensitivity analyses to examine the effect of detection studies on non-OPC (I2 ¼ 0.93, p < .001).
techniques and sample types on the HPV prevalence.
Statistical analyses were performed using com- Oropharyngeal Head Neck Cancer
mercial software (SAS v. 9; SAS Institute Inc., Cary, Overall HPV Oropharynx Prevalence and Trends over Time. The
NC). Values of p < .05 were considered statistically overall pooled HPV prevalence in patients with OPC
significant. We reported our results in accord with was 47.7% (95% CI, 42.9–52.5). Prevalence increased
the PRISMA guidelines. significantly over time: 40.5% (95% CI, 35.1–46.1) in
the 54 studies that recruited patients before 2000,
RESULTS 64.3% (95% CI, 56.7–71.3) in cohorts recruited
between 2000 and 2004, and 72.2% (95% CI, 52.9–
Search Results and Study Selection. In total, the 85.7) in cohorts recruited between 2005 onward (p <
search strategy described earlier yielded 867 .001) (Table 1, Figure 2). Prevalence over time
abstracts, of which 96 were duplicates. Therefore, 771 increased significantly in Europe (p < .004) and in
studies were identified, of which 501 were identified North America (p < .002). No recent data were avail-
from the database search and 270 from conference able for Asia, Africa, or South America, and so trend
abstract booklets and reference lists. A total of 345 over time analyses could not be performed for these
articles were selected for full review. Figure 1 details regions.
the reasons for exclusion. We contacted the authors of
53 papers and received replies from 12. HPV Oropharyngeal Cancer Prevalence by Geographic
Region. Overall HPV prevalence differed by region of
Characteristics of Selected Studies. In total, 269 recruitment (Table 1). The overall prevalence was
studies, reporting on 19,368 patients, were selected 39.7% (95% CI, 32.8–47%) in Europe, 59.9% (95% CI,
for inclusion in the meta-analysis. Their characteris- 54.7–64.9%) in North America, and 32.5% (95% CI,
tics are detailed in Appendix 1 (online Supporting In- 23.9–42.4%) in all other regions.
formation Supplementary Table 3). They reported on There was a significant difference in HPV preva-
a mean of 76 patients (range, 10–1088) from 43 coun- lence between North America, Europe, and other
tries in 6 continents, recruited between 1970 and 2008. countries before 2000 (50.7% vs 35.3% vs 32.2%,

Time and Regional Trends in HPV Prevalence in Head Neck Cancer HEAD & NECK—DOI 10.1002/hed Month 2012 3
FIGURE 1. PRISMA flow diagram showing identification, review, and selection of articles included in the meta-analysis. PRISMA, Pre-
ferred Reporting Items for Systematic Reviews and Meta-Analyses.

respectively; p ¼ .008). However, these regional differ- no statistically significant differences between the
ences in HPV prevalence have disappeared in the most regions (p ¼ .07).
recent studies, with pooled HPV prevalence rates in
North America now slightly lower than those in Europe HPV Prevalence by Genotype. For OPC, HPV-16
(69.7% vs 73.1%, respectively; p ¼ .8). There were no was positive in 1353 of the 1414 cases (95.7%) that
recent data from Asia, Africa, or South America. tested HPV-positive. For non-OPC, HPV-16 was found
to be positive in 1626 of 2199 HPV-positive cases
Nonoropharyngeal Head Neck Cancer. Overall (73.9%).
HPV prevalence in non-OPC was 21.8% (95% CI,
18.9–25.1%), and there appeared to be a statistically
Sensitivity Analyses
nonsignificant, declining trend over time (Table 1).
Pooled HPV prevalence was 22.2% (95% CI, 18.4– By quality of study.Sensitivity analysis performed by
26.4%) before 2000, compared with 17.2% (95% CI, repeating the meta-analyses after excluding low-quality
11.9–24.4%) between 2000 and 2004, and 6.1% (95% studies (scoring 3 points of 8 on the quality assess-
CI, 0.7–39%) from 2005 onward (p ¼ .97). There were ment) showed no significant changes in the results.

4 Time and Regional Trends in HPV Prevalence in Head Neck Cancer HEAD & NECK—DOI 10.1002/hed Month 2012
 Table 1. Synthesized human papillomavirus prevalence trends over time and by region.

Overall Midyear Midyear Midyear p value p value (year

prevalence <2000 2000–2004 2005þ (midyear group
Group (95% CI) (95% CI) (95% CI) (95% CI) Midyear NK trend) trend)

OPC
All regions 47.7 (42.9, 52.5) 40.5 (35.1, 46.1) 64.3 (56.7, 71.3) 72.2 (52.9, 85.7) 40.5 (31.1, 50.7) <.0001 <.0001
S ¼ 102 S ¼ 54 S ¼ 22 S¼4 S ¼ 22
n ¼ 5396 n ¼ 2690 n ¼ 2037 n ¼ 150 n ¼ 519
0.89 0.82 0.9 0.8
N. America 59.9 (54.7, 64.9) 50.7 (42.6, 58.7) 67.6 (61.7, 72.9) 69.7 (46.8, 85.7) 55.6 (41.4, 69.0) 0.0002 0.002
S ¼ 43 S ¼ 19 S ¼ 17 S¼2 S¼5
n ¼ 2550 n ¼ 696 n ¼ 1678 n ¼ 45 n ¼ 131
0.8 0.73 0.79 0.76
Europe 39.7 (32.8, 47.0) 35.3 (28.7, 42.5) 59.0 (30.2, 82.7) 73.1 (39.4, 91.9) 36.2 (24.3, 50.0) 0.07 0.004
S ¼ 46 S ¼ 27 S¼4 S¼2 S ¼ 13
n ¼ 2278 n ¼ 1704 n ¼ 164 n ¼ 105 n ¼ 305
0.87 0.82 0.87 0.87
Other, NK, and 32.5 (23.9, 42.4) 32.2 (21.0, 45.9) — — 35.4 (18.5, 56.9) 0.46 —
mixed regions
S ¼ 13 S¼8 S¼1 S¼0 S¼4
n ¼ 568 n ¼ 290 n ¼ 195 n¼0 n ¼ 83
0.73 0.78
Non-OPC
All regions 21.8 (18.9, 25.1) 22.2 (18.4, 26.4) 17.2 (11.9, 24.4) 6.1 (0.7, 39.0) 26.3 (19.3, 34.8) 0.97 0.07
S ¼ 236 S ¼ 140 S ¼ 37 S¼5 S ¼ 54
n ¼ 13,972 n ¼ 2260 n ¼ 186 n ¼ 2419
0.93 0.96 0.88 0.91
N. America 12.8 (9.7, 16.6) 14.1 (10.1, 19.5) 9.8 (5.2, 17.5) — 15.1 (7.3, 28.7) 0.03 0.08
S ¼ 62 S ¼ 38 S ¼ 13 S¼2 S¼9
n ¼ 3803 n ¼ 2212 n ¼ 1204 n ¼ 40 n ¼ 347
0.86 0.87 0.85
Europe 23.7 (19.4, 28.7) 23.6 (18.5, 29.5) 23.2 (12.7, 38.4) 11.7 (0.9, 67.0) 25.9 (16.3, 38.5) 0.27 0.66
S ¼ 90 S ¼ 53 S ¼ 14 S¼3 S ¼ 20
n ¼ 4625 n ¼ 29490 n ¼ 539 n ¼ 146 n ¼ 991
0.9 0.9 0.87 0.93
Other, NK, and 28.8 (22.5, 36.1) 28.6 (20.4, 38.5) 23.9 (14.5, 36.7) — 31.7 (19.8, 46.7) 0.53 0.55
mixed regions
S ¼ 84 S ¼ 49 S ¼ 10 S¼0 S ¼ 25
n ¼ 5364 n ¼ 3766 n ¼ 517 n¼0 n ¼ 1081
0.95 0.96 0.84
Abbreviations: HPV, human papillomavirus; CI, confidence interval; OPC, oropharyngeal cancer; NK, not known; S, number of studies; n ¼ number of patients included; I2, index
statistic measuring heterogeneity.

By detection method. For OPC samples, HPV preva- reported a prevalence rate of 37.1% (95% CI, 16.8–
lence was found to be 44% (95% CI, 38.649.6%) 63.3%), which was statistically not different (p ¼ .17)
when tested by PCR in 3347 cases tested in 78 stud- compared with the prevalence rate of 44.7% (95% CI,
ies. HPV prevalence was found to be 58.4% (95% CI, 25.8–65.2%) reported by the 21 studies published in
49.2–67.1%) in 1254 patients tested in 17 studies 2005 onward.
using ISH. After adjusting for the time period during For further sensitivity analysis, we graded the
which the test was performed, there were no signifi- above-cited studies in accord with the sensitivity of
cant differences between prevalence rates identified the PCR detection methods and primers used and
by ISH or PCR (p ¼ .4) (Table 2). Exclusion of the compared their reported HPV positivity rates. There
studies that detected HPV by Southern blotting of iso- was no statistical significance (p ¼ .28) between the
lated DNA and/or by immunohistochemistry for HPV pooled prevalence of HPV positivity reported by stud-
antigens (both considered to be methods of low sensi- ies that used techniques of low or medium sensitivity
tivity) had no effect on results. (52.7%; 95% CI, 14.1–88.3%) compared with the prev-
To detect any bias attributed to improved sensitiv- alence reported by studies using techniques of high or
ity of detection techniques with time, we examined very high sensitivity (39.4%; 95% CI, 33.3–45.8%).
the prevalence rates of cohorts recruited before 2000. For non-OPC cases, HPV prevalence by PCR was
We compared those studies published before 2005 25.1% (95% CI, 21.8–28.7%), and by ISH was 13.9%
(potentially using less sensitive primer sets) with (95% CI, 8.1–22.7%). This difference was statistically
those published after that (using more sensitive tech- significant, even after adjustment for time period of
niques). Thirty-three studies published before 2005 testing (p ¼ .001).

Time and Regional Trends in HPV Prevalence in Head Neck Cancer HEAD & NECK—DOI 10.1002/hed Month 2012 5
By sample type. The type of sample (fresh-frozen or DISCUSSION
paraffin-embedded) did not appear to make a differ- The proportion of HPV-related OPC in Europe has
ence in the reported rates of HPV infection for either increased sharply and at a faster rate over the past
OPC or non-OPC samples (Table 2). decade compared with that in North America, such

FIGURE 2. Forest plot of mean prevalence of HPV in oropharyngeal cancer (vertical dotted line denotes the random effects estimate
of mean prevalence for each time period). HPV, human papillomavirus (Continued).

6 Time and Regional Trends in HPV Prevalence in Head Neck Cancer HEAD & NECK—DOI 10.1002/hed Month 2012
 FIGURE 2. Continued.

that the significant difference in prevalence, previ- lacking in countries other than those located in North
ously detected between them, has now disappeared. America and Western Europe.
The sharp increase (of >70%, from 40.5% to 72.2%) in
the proportion of HPV-related OPC over the last dec-
ade appears to be a real one, and is not a result of Implications of Findings. The significant difference
improvements in the sensitivity and performance of in the prevalence of HPV between OPC and non-OPC
detection techniques, as some have claimed. In con- head and neck cancers, and the absence of an
trast, the prevalence rates of HPV in non-OPC appear increase in HPV prevalence in non-OPC with time,
to be stable, with no regional differences, and are at a compared with the concurrent rapid increase in prev-
much lower rate than that in OPC. In addition, alence in OPC, supports the premise that these are
recent data on HPV prevalence are conspicuously different disease entities. The findings appear to

Time and Regional Trends in HPV Prevalence in Head Neck Cancer HEAD & NECK—DOI 10.1002/hed Month 2012 7
 Table 2. Synthesized prevalence of human papillomavirus by detection technique and sample type used.

Pooled rate by type of sample (95% CI), N ¼ no. of studies

Group Detection method No. of studies No. of patients All samples Fresh- frozen FFPE Both/other/NK

OPC ISH 17 1,254 58.4 — 59.6 55.9

(49.2, 67.1) (51.0, 67.6) (30.5, 78.5)
N ¼ 17 N¼1 N ¼ 13 N¼3
PCR 78 3,347 44 41.7 45.1 44.7
(38.6, 49.6) (33.5, 50.4) (37.7, 52.6) (18.3, 74.5)
N ¼ 78 N ¼ 30 N ¼ 38 N ¼ 10
Non-OPC ISH 38 1,502 13.9 29.3 10.3 30.1
(8.1, 22.7) (15.2, 49.0) (5.2, 19.4) (5.4, 76.4)
N ¼ 38 N¼5 N¼0 N¼3
PCR 186 11,868 25.1 24.6 25.7 23.2
(21.8, 28.7) (19.6, 30.5) (21.2, 30.7) (13.3, 37.3)
N ¼ 186 N ¼ 65 N ¼ 99 N ¼ 22
Abbreviations: CI, confidence interval; OPC, oropharyngeal cancer; ISH, in situ hybridization; PCR, polymerase chain reaction; FFPE, formalin-fixed paraffin-embedded; NK, not
known.
Note: p value for OPC: ISH versus PCR is .03, but adjusted for year is .40; for non-OPC: ISH versus PCR is .02 and adjusted for year is .001; for OPC: FF versus FFPE is .20,
but adjusted for year is .11; for non-OPC: FF versus FFPE is .15, but adjusted for year is .92.

confirm that this is not likely to be a lag phase, after tions such as chlamydia and gonorrhea. With evidence
which HPV prevalence in non-OPC is likely to of increasing incidence of HPV-positive cancers, this
increase significantly. On the contrary, it supports the study highlights the need for research into modalities
conclusion that there appears to be a predilection of of HPV oral transmission and into sexual health educa-
HPV to the oropharynx and that the etiology of non- tion to change behaviors that increase the risk of HPV
OPC is mainly not HPV driven. infection and resultant cancers.19
The implications of the rapid rise in HPV preva- Healthcare professionals involved in caring for
lence in OPC are considerable. Both the United patients with head and neck cancer are usually well
States and many Western European countries have equipped to break the news of cancer. However, oral
reported considerable increases in the incidence of acquisition of HPV may be through a sexually trans-
OPC.16,17 Therefore, the burden of disease caused by mitted route,2 which may have additional social
HPV is increasing multiplicatively in these countries. stigma. Research into the effects of a diagnosis of
Patients with HPV-OPC are younger and have much HPV-OPC on the patient, their partner, and their
better survival rates than those with ‘‘traditional’’ information and counseling, is needed to develop an
non-HPV head and neck cancer. In view of this rapid information and support package for patients with
rise in the burden of disease, health service planners HPV-OPC, and to train their clinicians for this.
and commissioners should consider making provisions The different demographic and treatment response
for the prolonged support that may be required by of HPV-OPC would suggest that this may be a distinct
the increasing numbers of patients with HPV-OPC. It disease entity. Research is needed into alternative
would also be prudent to consider rationally directed, treatment modalities that result in less long-term mor-
cost-effective therapy, and even effective screening bidity, and should be considered by researchers and
programs for this subset of the population with head funding bodies.
and neck cancer to reduce mortality.
HPV-OPC may potentially be prevented by vaccina-
tion against HPV, especially HPV-16.9,10 However, there Limitations of the Study. This is a meta-analysis of
is currently no definitive evidence to support this pre- the published literature and therefore reflects the
mise. Previous analyses of cost-effectiveness of HPV strengths and weaknesses of the individual studies
vaccination programs in males concluded that it was included. There is a high degree of heterogeneity
not cost-effective.18 These studies used much lower between studies, which is likely to be explained by the
prevalence rates of HPV-OPC than that determined by variability of design and execution between them. Not
the present study for their modeling calculations. The only do the studies differ in the type of detection method,
large rise in disease burden may have changed the they also differ in the time periods that they recruited
cost-effectiveness of HPV vaccination of males and cases, the sample type, the setting of recruitment, and
should be imminently reassessed. In addition, data sup- possibly the case mix. There were also differences in the
porting the potential effectiveness of HPV vaccination quality of the studies. However, sensitivity analyses of
in preventing OPC should also be sought. the likely main sources of bias—detection methods, sam-
Currently, most sexual health education interven- ple types, and quality of studies—did not demonstrate
tions and campaigns that promote sexual risk reduction any significant changes in the overall findings.
behaviors primarily address the human immunodefi- It should be noted that the data for European
ciency virus (HIV) and other sexually transmitted infec- studies are heavily weighted toward Western Europe,

8 Time and Regional Trends in HPV Prevalence in Head Neck Cancer HEAD & NECK—DOI 10.1002/hed Month 2012
with only 4 of the 90 studies reporting data from increase in HPV prevalence over several decades,
Eastern European states, and none of these reported despite using the same ‘‘modern’’ detection technique
recent data after 2005. Therefore, the data presented for the samples from all 4 decades.5
should be interpreted as pertaining mainly to West- There do not appear to be any significant differ-
ern Europe. We restricted our search to studies pub- ences in the detection rates when using fresh-frozen
lished in English, and this may have introduced bias or fresh-frozen paraffin-embedded samples. This
into the study. However, there were very few studies would suggest that by using an appropriate tech-
that were excluded for this reason. Like Asia, Africa, nique, HPV detection can be reliably performed on
and South America, little can be determined about either sample type.
the burden of HPV-OPV in Eastern Europe. Further There also appear to be no significant differences
research on the prevalence of HPV in head and neck in the HPV prevalence rates detected by PCR or ISH
cancers in Eastern Europe, Africa, Asia, and South in OPC. However, PCR tests appeared to detect
America is needed. much higher rates of HPV in non-OPC than ISH
techniques. Termine and colleagues4 also reported
head and neck site-specific variation in HPV preva-
Excluding Ascertainment Bias. It would appear lence between PCR- and ISH-based detection meth-
that the prevalence of HPV in non-OPC has remained ods. The reasons for this may be that viral loads are
unchanged over time. This decreasing trend cannot be much lower in non-OPC, as shown within the oral
explained. It may be partly that researchers are now cavity, compared with OPC,20 and therefore are more
more aware of the significance of HPV in the orophar- difficult to detect by ISH techniques compared with
ynx. Therefore, greater attention paid to the accurate more sensitive PCR-based methods that are capable
classification of the tumors may lead to fewer OPCs of detecting only a few copies of the viral DNA.
(which are more likely to be HPV positive) being incor- Whether a reduced viral load in non-OPC can be
rectly assigned as oral cancers. The overall result of attributed to a higher rate of HPV DNA integration
this may be a decline in HPV non-OPC rates. A con- in these tumors compared with OPC is not clear,20,21
comitant decline in the incidence of oral cancers would and clarification of this controversy is needed
be expected in that case. However, the incidence of oral through a systematic analysis of the HPV physical
cancers has increased in some countries, such as the state within OPC.
United Kingdom, during the same period,17 suggesting It is evident from the number of different method-
that this theory does not hold true. ologies described and reported in the literature that
In addition, the study of OPC incidence in Stock- techniques for the detection of HPV in OPC require
holm over 3 decades has shown significant progressive urgent consensus and standardization on the method-
increases in the proportion of tonsillar carcinoma being ology for HPV detection, to allow more accurate com-
caused by HPV.5 This has been steadily rising since parisons between studies.22 It is imperative that the
1970s, well before the link with HPV was detected or implemented methodologies are able to distinguish
became publicized. Therefore, HPV prevalence was ris- between ‘‘active’’ infections, as defined by viral onco-
ing long before clinician awareness may have changed gene expression, and those tumors that exhibit HPV
their behavior when assigning tumor site. positivity through laboratory contamination or the
It could be argued that the detected rise in HPV presence of a latent infection. This may have even
prevalence in OPC may also be attributed to an more bearing when, in the future, treatment selection
improvement in the sensitivity of detection methods for individual patients in routine practice will be car-
over time. We performed sensitivity analyses to exam- ried out on the basis of HPV status.
ine this potential effect, by examining the cohorts
diagnosed before 2000. We compared studies that
were undertaken after 2005 and those using newer, Acknowledgments. Contributions: H.M. con-
more sensitive techniques, to those studies that were ceived the idea; T.B., T.N., and S.R. acquired the data;
performed before 2005 and to those using potentially H.M., C.C.M., I.H., S.R., and V.P. analyzed and inter-
less sensitive HPV detection techniques and primers. preted the data; H.M., T.B., and T.N. drafted the
However, there was no significant difference in the manuscript; and C.M., I.H., S.R., and V.P. critically
reported prevalence rates for studies published before reviewed it. H.M. provided funding and administra-
2005 (potentially using primers with lower sensitiv- tive support. Data integrity: T.B. and T.N. had full
ity) compared with those published after 2005 or access to all of the data in the study and take respon-
between the purported more and less sensitive techni- sibility for the integrity of the data, and C.M. takes
ques. Therefore, ascertainment bias arising from responsibility for the accuracy of the data, analysis.
improved sensitivity of the detection technique is
highly unlikely to be the cause of the documented
increase in prevalence rates. This conclusion is also REFERENCES
supported by the Swedish study described earlier. It 1. Wennerberg J, Schildt EB, Bladström A, et al. World Health Or-
demonstrated that there has been a considerable ganization. IARC monographs on the evaluation of carcinogenic

Time and Regional Trends in HPV Prevalence in Head Neck Cancer HEAD & NECK—DOI 10.1002/hed Month 2012 9
 risks to humans: volume 90-human papillomaviruses. Lyon: 11. Sturgis EM, Dahlstrom KR. Inaccurate assumptions about oro-
International Agency for Research on Cancer; 2007. pharyngeal cancer. BMJ. 2009;339:4525 (abstract).
2. D’Souza G, Kreimer AR, Viscidi R, et al. Case-control study of 12. Hayden JA, Pierre Côté P, Bombardier C. Evaluation of the
human papillomavirus and oropharyngeal cancer. N Engl J Med quality of prognosis studies in systematic reviews. Ann Int Med
2007;356:1944–1956. 2006;144:427–437.
3. Kreimer AR, Clifford GM, Boyle P, et al. Human papillomavirus 13. PRISMA statement on transparent reporting of systematic
types in head and neck squamous cell carcinomas worldwide: a reviews and meta-analyses. http://www.prisma-statement.org/,
systematic review. Cancer Epidemiol Biomarkers Prev 2005;14: accessed August 11, 2010.
467–475. 14. Higgins JPT, Thompson SG. Quantifying heterogeneity in a
4. Termine N, Panzarella V, Falaschini S, et al. HPV in oral squamous meta-analysis. Statist Med 2002;21:1539–1558.
cell carcinoma vs head and neck squamous cell carcinoma biopsies: 15. Robinson M, Sloan P, Shaw R. Refining the diagnosis of oropha-
a meta-analysis (1988–2007). Ann Oncol 2008;19:1681–1690. ryngeal squamous cell carcinoma using human papillomavirus
5. Nasman A, Attner P, Hammarstedt L, et al. Incidence of human testing. Oral Oncol 2010;46:492–496.
papillomavirus (HPV) positive tonsillar carcinoma in Stockholm, 16. SEER National Cancer Institute. Surveillance epidemiology and
Sweden: an epidemic of viral-induced carcinoma? Int J Cancer end results. http://seer.cancer.gov/faststats/selections.php#Output,
2009;125:362–366. accessed October 21, 2010.
6. Ang KK, Harris J, Wheeler R, et al. Human papillomavirus and 17. National Cancer Intelligence Network. Profile of head and neck
survival of patients with oropharyngeal cancer. N Engl J Med cancers. 2010 http://library.ncin.org.uk/docs/100504-OCIU
2010;363:24–35. Head_and_Neck_Profiles.pdf, accessed October 21, 2010.
7. Mehanna H, Jones TM, Gregoire V, et al. Oropharyngeal carci- 18. Marra F, Cloutier K, Oteng B, Marra C, Ogilvie G. Effectiveness
noma related to human papillomavirus. BMJ 2010;340:1439 and cost effectiveness of human papillomavirus vaccine: a sys-
(abstract). tematic review. Pharmacoeconomics 2009;27:127–147.
8. El-Mofty SK. Human papillomavirus (HPV) related carcinomas of 19. Shepherd J, Peersman G, Weston R, et al. Cervical cancer and
the upper aerodigestive tract. Head Neck Pathol 2007;1:181–185. sexual lifestyle: a systematic review of health education inter-
9. FUTURE I/II Study Group, Dillner J, Kjaer SK, Wheeler CM, ventions targeted at women. Health Educ Res 2000;15:681–694.
et al. Four year efficacy of prophylactic human papillomavirus 20. Syrjanen S. Human papillomavirus (HPV) in head and neck can-
quadrivalent vaccine against low grade cervical, vulvar, and vag- cer. J Clin Virol 2005;32S:S59–S66.
inal intraepithelial neoplasia and anogenital warts: randomised 21. Kim SH, Koo BS, Kang S, et al. HPV integration begins in the
controlled trial. BMJ 2010;341:3493 (abstract). tonsillar crypt and leads to the alteration of p16, EGFR and
10. Paavonen J, Naud P, Salmerón J, et al. Efficacy of human papil- c-myc during tumor formation. Int J Cancer 2007;120:1418–1425.
lomavirus (HPV)-16/18 AS04-adjuvanted vaccine against cervical 22. Braakhuis BJ, Brakenhoff RH, Meijer CJ, et al. Human papil-
infection and precancer caused by oncogenic HPV types (PATRI- loma virus in head and neck cancer: the need for a standardised
CIA): final analysis of a double-blind, randomised study in young assay to assess the full clinical importance. Eur J Cancer 2009;
women. Lancet 2009;374:301–314. 45:2935–2939.

10 Time and Regional Trends in HPV Prevalence in Head Neck Cancer HEAD & NECK—DOI 10.1002/hed Month 2012