A

oa©

1

1

cnca63oSS

afathsC

f

0d

Vaccine 24S3 (2006) S3/11–S3/25

Chapter 2: The burden of HPV-related cancers

D. Maxwell Parkin a,∗, Freddie Bray b,1

a Clinical Trial Service Unit & Epidemiological Studies Unit, Richard Doll Building, University of Oxford,Old Road Campus, OXFORD OX3 7LF, UK

b Cancer Registry of Norway, Institute of Population-Based Research, Montebello, Oslo, Norway

Received 3 March 2006; accepted 15 May 2006

bstract

On the basis of current evidence regarding human papillomavirus (HPV) and cancer, this chapter provides estimates of the global burdenf HPV-related cancers, and the proportion that are actually “caused” by infection with HPV types, and therefore potentially preventable. Welso present trends in incidence and mortality of these cancers in the past, and consider their likely future evolution.

SV(n

cgatipcrtvr[

W

2006 Elsevier Ltd. All rights reserved.

. Burden of HPV-related cancer

.1. Cancer of the cervix

Cancer of the cervix uteri is the second most commonancer among women worldwide, with an estimated 493,000ew cases and 274,000 deaths in 2002 [2]. Some 83% of theases occur in developing countries, where cervical cancerccounts for 15% of female cancers, with a risk before age5 of 1.5%, while in developed countries it accounts for only.6% of new cancers, with a cumulative risk (ages 0–64)f 0.8% [2]. The highest incidence rates are observed in sub-aharan Africa, Melanesia, Latin America and the Caribbean,outh-Central Asia, and South East Asia (Fig. 1).

Fig. 2 shows incidence rates recorded in cancer registriesround 1995 [3]. There is a range in incidence of at least 20-old. In general, the lowest rates (less than 15 per 100,000)re found in Europe (except some Eastern European coun-ries), North America and Japan. The incidence is generally

igher in the developing countries of Latin America (age-tandardized incidence rates (ASR) 33.5 per 100,000) and thearibbean (ASR 33.5), sub-Saharan Africa (ASR 31.0), and

∗ Corresponding author. Tel.: +44 1865 743743; fax: +44 1865 743985.E-mail addresses: max.parkin@ctsu.ox.ac.uk (D.Maxwell Parkin),

reddie.bray@kreftregisteret.no (F. Bray).1 Tel.: 47 2245 1334.

vlicfZ

i

264-410X/$ – see front matter © 2006 Elsevier Ltd. All rights reserved.oi:10.1016/j.vaccine.2006.05.111

outh-Central (ASR 26.5) and Southeast Asia (ASR 18.3).ery low rates are observed in China, and in Western Asia

Fig. 1); the lowest recorded rate is 0.4 per 100,000 in Ardabil,orthwest Iran [4].

The majority of cases of cervical cancer are squamous cellarcinomas (SCCs); adenocarcinomas are less common. Ineneral, the proportion of adenocarcinoma cases is higher inreas with a low incidence of cervical cancer, and this his-ology may account for up to 25% of cervical cancer casesn many Western countries [3] (Fig. 2). The relatively highroportion of adenocarcinomas is a consequence of cytologi-al screening, which historically, probably had little effect ineducing the risk of adenocarcinoma of the cervix, becausehese cancers, and their precursors, occur within the cer-ical canal (from the glandular epithelium), and were noteadily sampled by scraping the epithelium of the ectocervix5].

Mortality rates are substantially lower than incidence.orldwide, the ratio of mortality to incidence is 55%. Sur-

ival rates vary between regions with quite good prognosis inow-risk regions (73% at 5 years in US registries [6] and 63%n Europe [7]), but even in developing countries, where manyases present at relatively advanced stages, survival rates are

air (for example, 30.5% in the African population of Harare,imbabwe [8]).

Because cervical cancer affects relatively young women,t is an important cause of lost years of life in the developing

S3/12 D. Maxwell Parkin, F. Bray / Vaccine 24S3 (2006) S3/11–S3/25

cidenc

w2iiEttfa

1

1

aerIiSpo

p

bravledai4c

1

wTnomi

Fig. 1. Age-standardised (World) in

orld. Yang et al. [9] found that it was responsible for.7 million (age-weighted) years of life lost (YLL) worldwiden 2000 and it is the biggest single cause of YLL from cancern the developing world. In Latin America, the Caribbean andastern Europe, cervical cancer makes a greater contribution

o YLL than diseases such as tuberculosis, maternal condi-ions or AIDS. It also makes the largest contribution to YLLrom cancer in the populous regions of sub-Saharan Africand South-Central Asia.

.2. Cancers of the external genitalia

.2.1. Cancer of the penisOn a global basis, cancer of the penis is a rare cancer,

ccounting for less than 0.5% of cancers in men. In West-rn countries, the ASR is less than 1 per 100,000. Incidenceates greater than this are observed in cancer registries inndia, Southeast Asia (Thailand and Viet Nam), Latin Amer-ca (Paraguay, Puerto Rico, Peru, Brazil) and in Eastern andouthern Africa (Fig. 3). Incidence in Jewish populations is

articularly low (0.04 per 100,000 in the Jewish populationf Israel in 1993–1997) [3].

The importance of circumcision in determining the risk ofenile cancer has been evident for many years, and this has

ifco

e rates of cervical cancer 2002 [2].

een confirmed by case-control studies which suggest that theisk is reduced about threefold [10,11]. Circumcision protectsgainst sexually transmitted infections, like HIV [12], by pre-enting accumulation of infected vaginal secretions, or moreikely by simply reducing the surface area of non-keratinizedpithelium. The geographical correlation between the inci-ence of penile and cervical cancers (Fig. 4c), has suggestedcommon aetiology, as has concordance of these two cancers

n married couples [13]. HPV-DNA is detectable in about0–50% of all penile cancers, and serological studies haveonfirmed the role of HPV-16 and -18 [1].

.2.2. Cancers of the vulvaFig. 5 shows some incidence rates of cancers of the vulva

orldwide. ASRs mostly lie between 0.5 and 1.5 per 100,000.he geographical pattern is not immediately obvious, and isot similar to that of cervical cancer, since high rates arebserved in several European populations (Scotland, Den-ark, Spain, Italy), and low rates in several populations

n sub-Saharan Africa, Southeast Asia, and Latin Amer-

ca. Bosch and Cardis [14], using cancer incidence datarom 1978 to 1982, found that the incidence of cervicalancer was strongly correlated with penile cancer, and ofther cancers of the female genitalia; however, this is not

D. Maxwell Parkin, F. Bray / Vaccine 24S3 (2006) S3/11–S3/25 S3/13

Fig. 2. Age-standardised (World) incidence rates of cervical cancer by histological subtype in selected cancer registries circa 1993–1997 [3].

Fig. 3. Age-standardised (World) incidence rates of penile cancer in selected cancer registries circa 1993–1997 [3].

S3/14 D. Maxwell Parkin, F. Bray / Vaccin

Fig. 4. Correlation between incidence rates of cancer of the cervix, andc(ip

o(

ttct

1

cionod

ca

gcoacc

1

cfipbeScfamsba[

1

yssfogsiPLaodiiratatc

ancer of the vulva (a), vagina (b) and penis (c). Rates are age-standardisedWorld) based on data circa 1993–1997 from all cancer registries acceptedn volume VIII of Cancer Incidence in Five Continents. Weights for size ofoints are person-years of observation in each registry [3].

bvious when rates for the period 1993–1997 are comparedFig. 4a).

Although the majority of tumours are SCCs, distinct sub-ypes are recognised, particularly the warty and basaloidypes, which are associated with HPV infection and the pre-ursor lesions of vulvar intraepithelial neoplasia (VIN), andhe verrucous type, which is not [15].

.2.3. Cancers of the vaginaCancers of the vagina are less frequent than vulvar can-

ers: the ASR is 0.3–0.7 per 100,000 in most countries. Theres no clear association with incidence of cancers of the vulva

r cervix (Figs. 4b and 5). Many SCCs are preceded by vagi-al intraepithelial neoplasia (VAIN). Clear-cell carcinomaf the vagina is a well-known complication of exposure toiethylstilboestrol in utero [16]; simultaneous or prior gynae-

aat

e 24S3 (2006) S3/11–S3/25

ological malignancy leads to an increased risk, especially ifssociated with pelvic irradiation [15].

The number of new cases of cancer of the penis and femaleenitalia worldwide can be estimated from the number ofervical cancer cases, and the ratio of incidence rates at eachf these sites to those of cervical cancer, by world area andge group [3]. The estimated numbers in 2002 were 40,000ases of cancer of the female genitalia and 26,300 cases ofancer of the penis worldwide.

.3. Cancers of the anus

Cancers of the anus are those arising in the analanal—tumours of the external skin (anal margin) are classi-ed along with skin cancers. The canal is lined in its upperart by colorectal-type of mucosa, and in its lower thirdy squamous epithelium, with specialized transitional zonepithelium in between. Therefore, cancers are predominantlyCCs, adenocarcinomas, or basaloid and cloacogenic car-inomas. In most populations SCC is twice as common inemales as in males (Fig. 6). Incidence is particularly highmongst homosexual males, so the incidence rates amongales in some populations (e.g. San Francisco) are corre-

pondingly elevated (Fig. 6), and the risk is increased furthery infection with HIV. Risk is increased by cigarette smoking,nal intercourse, and the number of lifetime sexual partners17].

.4. Cancers of the mouth and oro-pharynx

Cancer of the mucosa of the oral cavity and of the phar-nx are frequently considered as a group, as there are manyimilarities in their epidemiology, treatment and progno-is. Cancers of the nasopharynx and salivary glands dif-er histologically and aetiologically. The incidence rates ofro-pharyngeal cancers vary widely and are highest in theeographical regions where tobacco use and alcohol con-umption are popular. A high incidence of these cancerss observed in the Indian subcontinent, Oceania (Australia,apua-New Guinea), Switzerland, The Netherlands, France,atin America (Brazil) and Southern Africa (Fig. 7). In 2002,n estimated 405,000 cases occurred worldwide, two-thirdsf them in developing countries, and there were 211,000eaths [2]. The highest incidences among males are reportedn Northern and Eastern France, while the highest incidencesn females are in India and Pakistan [3]. The male:femaleatio of occurrence varies from 2 to 15:1, depending on thenatomical sub-site, with the extreme ratios characteristic ofongue, mouth and pharyngeal cancers. Cancers of the mouthnd of the tongue generally predominate in developing coun-ries, whereas pharyngeal cancers are common in developedountries, including Central and Eastern Europe.

Although HPV is accepted as an aetiological factor for oralnd pharyngeal cancers, the major risk factors are tobaccond alcohol, with the effects of these two agents being mul-iplicative [18]. Tobacco exposure occurs through smoking

D. Maxwell Parkin, F. Bray / Vaccine 24S3 (2006) S3/11–S3/25 S3/15

nd vagi

iA4[s

S

Fig. 5. Age-standardised (World) incidence rates of vulvar a

n the high-risk populations of Europe, Australia and Latin

merica, and has been estimated to be responsible for about1% of these cancers in men, and 11% in women worldwide19]. Chewing tobacco is the major exposure in the Indianubcontinent, Papua New Guinea, and to a lesser extent, in

ftoa

Fig. 6. Age-standardised (World) incidence rates of anal cancer by histologic

nal cancer in selected cancer registries circa 1993–1997 [3].

outheast Asia [18]. Chewing has traditionally been in the

orm of ‘betel’—a large quid made up of varying propor-ions of tobacco, lime and condiments wrapped in the leaff the piper betel. Recently, sales of commercially preparednd marketed powdered tobacco have become increasingly

al subtype and sex in selected cancer registries circa 1993–1997 [3].

S3/16 D. Maxwell Parkin, F. Bray / Vaccine 24S3 (2006) S3/11–S3/25

ates of

phE

rt

aalt

2

cvc

cimota

Htaltcssoctc9dcoD

iifc

Fig. 7. Age-standardised (World) incidence r

opular, and there are indications that these may be even morearmful. Oral snuff (“dipping”) is used in some countries inurope (Sweden) and in the US amongst the young [18].

Alcohol increases the risk of these cancers, and this riskelates to the quantity of alcohol consumed, rather than theype of beverage [20].

Epidemiological studies worldwide have linked a gener-lly impoverished diet, particularly one lacking in vegetablesnd fruits, with an excess risk for these cancers. Cancer of theip is associated with sunlight exposure and outdoor occupa-ions in white populations.

. Quantification of the role of HPV in human cancer

Oncogenic HPV can be detected by PCR in virtually allases of cervical cancer, and it is generally accepted that theirus is necessary for the development of cancer, so that allases of cancer can be “attributed” to infection [1].

With respect to SCCs of the vulva and vagina, penile can-er, and anal cancer, the relative risk associated with HPVnfection is difficult to quantify, because of the small size of

ost studies, and the absence of comparable measurementsf prevalence of infection in normal subjects. The fraction ofhese cancers attributed to HPV infection is therefore gener-lly equated with the proportion of cancer cases infected with

or

d

cancer of the mouth and pharynx * 2002 [3].

PV in various series. In studies using PCR methodology,he prevalence of HPV in vaginal cancer is about 60–65%nd 20–50% in vulvar cancers [1] (75–100% in the basa-oid and warty type associated with VIN and only 2–23% ofhe keratinizing carcinomas [21]). HPV-related vulvar can-er occurs in younger women than the typical keratinizingquamous histology related to chronic inflammatory precur-ors. For anal cancer, Frisch et al. [22] observed in a seriesf 386 cases from Denmark and Sweden, 95% and 83% ofancers involving the anal canal in women and men, respec-ively, were positive for oncogenic HPV. Prevalence in a USase-control study (of about 300 subjects) was approximately0% of cancers, and HPV-16 was the most frequent HPV typeetected (73%), followed by HPV-18 (6.9%) [17]. For penileancer, Bezerra et al. [23] found HPV DNA in 30% of 71 casesf penile cancer from Brazil, and Rubin et al. [24] found HPVNA in 42% of 148 cases from the US and Paraguay.Several studies have investigated the prevalence of HPV

n cancers of the mouth and pharynx. In a review of 60 stud-es, Kreimer et al. [25] calculated the average to be 23.5%or cancers of the oral cavity and 35.6% for oro-pharyngealancers. HPV-16 was the predominant type (87% and 68%

f HPV-infected oro-pharyngeal cancers and oral cavity,espectively).

The population attributable fraction (PAF) of a disease isefined as the proportion of the new cases of disease observed

/ Vaccine 24S3 (2006) S3/11–S3/25 S3/17

itowchacpfiTbvotococwatcwpflDT

gtfmitlstmaHc5c

3

alpto

-attr

ibut

able

canc

erin

2002

:by

sex

and

HPV

type

Attr

ibut

able

toH

PV(%

)O

fw

hich

,HPV

16an

d/or

18(%

)[R

efs]

Mal

esFe

mal

esB

oth

sexe

s

Tota

lca

ncer

sA

ttrib

utab

leto

HPV

Attr

ibut

able

toH

PV16

/18

Tota

lca

ncer

sA

ttrib

utab

leto

HPV

Attr

ibut

able

toH

PV16

/18

Tota

lca

ncer

sA

ttrib

utab

leto

HPV

%al

lca

ncer

Attr

ibut

able

toH

PV16

/18

%al

lca

ncer

100

70[2

8]a

00

049

2,80

049

2,80

034

4,90

049

2,80

049

2,80

04.

5434

4,90

03.

1840

63[2

4]26

,300

10,5

0066

000

00

26,3

0010

,500

0.10

6600

0.06

4080

[29,

30,2

1]0

00

40,0

0016

,000

12,8

0040

,000

16,0

000.

1512

,800

0.12

9092

[17,

22]

14,5

0013

,000

12,0

0015

,900

14,3

0013

,100

30,4

0027

,300

0.25

25,1

000.

233

95[2

5]17

5,90

052

0050

0098

,400

2900

2800

274,

300

8200

0.08

7800

0.07

1289

[25]

42,5

0051

0045

0096

0011

0010

0052

,100

6200

0.06

5500

0.05

5,80

1,80

033

,900

28,1

005,

060,

700

527,

200

374,

700

10,8

62,5

0056

1,10

05.

1740

2,90

03.

71

sin

diff

eren

twor

ldar

eas,

repo

rted

byM

unoz

etal

.[28

],w

eigh

ted

byth

eco

rres

pond

ing

num

ber

ofca

ses

inea

ch[2

].

D. Maxwell Parkin, F. Bray

n a population that are attributable to exposure to a risk fac-or. The PAF therefore provides an upper limit to the numbersf cases that could be prevented, for example, if vaccinationas 100% effective. The estimated numbers of cases of can-

er attributable to infection with all HPV types worldwideave been published [26]. All cases of cervical cancer weressumed to be related to HPV, while for the ano-genital can-ers, the estimated attributable fraction was based upon theroportion of tumours in which the virus can be detected; theractions assumed were 40% of cancers of the external gen-talia (vulva, vagina, penis) and 90% of cancers of the anus.his method, however, overestimates attributable fractions,y including some cancer cases in which the presence of theirus was coincidental without, for example, expressing viralncoproteins. The estimate of HPV-attributable cancers ofhe oral cavity and pharynx, on the other hand, makes usef data from the International Agency for Research on Can-er (IARC) multi-national study [27] in which prevalencef HPV16 E6 or E7 antibody in controls (1.6%) as well asases (6.4% mouth cancers, 15.3% oro-pharyngeal cancers)as available. An allowance was also made for the fact that

ssessment of HPV-DNA presence by PCR assay may leado an overestimation of cases in which the virus is aetiologi-ally involved, as suggested by the lower proportion of casesith E6/E7 expression (4.6% of oral cancers and 12% of oro-haryngeal cancers). The resulting attributable fractions (3%or oral cancer, 12% for oro-pharynx) are, therefore ratherower than the percentage of tumours with detectable HPV-NA (4% and 18%, respectively). The results are shown inables 1 and 2.

Although the study of HPV prevalence in oral and pharyn-eal cancers by Herrero et al. [27] has the advantage of beinghe largest available (1670 cases), and allows attributableractions to be corrected for the presence of HPV in nor-al control subjects, the prevalence of HPV in tumour tissue

n this study (4% of oral cancers, 18% of oro-pharyngealumours) was considerably lower that the averages calcu-ated by Kreimer et al. [25], based on 3611 cases from 60tudies (32.5% and 35.6%, respectively), so it is possible thathe HPV-attributable cases at these sites in Tables 1 and 2

ay be considerably underestimated. Table 1 also providesn estimate of the numbers of cancer cases attributable toPV 16 and/or 18, based on the percentage of HPV-positive

ancers at each site that contained HPV-16 or -18. Of the61,000 cancers estimated to be caused by HPV, 71.8% wereaused by HPV-16 and -18.

. Time trends of cervical cancer

Time trend studies make use of published rates of mortalitynd incidence. Comparative studies of mortality, in particu-

ar, must take into account the deaths certified as “Uterus,art unspecified”. The proportion of uterine deaths ascribedo this category varies widely, both between countries, andver time, and it can be very high, for example, over 50% Ta

ble

1H

PVin

fect

ion

Site

Cer

vix

Peni

sV

ulva

,vag

ina

Anu

sM

outh

Oro

-pha

rynx

All

site

sa

Perc

enta

ge

S3/18 D. Maxwell Parkin, F. Bray / Vaccine 24S3 (2006) S3/11–S3/25

Table 2HPV infection-attributable cancer in 2002: developed & developing countries (source: [26])

Site Attributable toHPV (%)

Developed countries Developing countries

Total cancers Attributable to HPV % all cancer Total cancers Attributable to HPV % all cancer

Cervix 100 83,400 83,400 1.7 409,400 409,400 7.0Penis 40 5200 2100 0.0 21,100 8400 0.1Vulva, vagina 40 18,300 7300 0.1 21,700 8700 0.1Anus 90 14,500 13,100 0.3 15,900 14,300 0.2Mouth 3 91,200 2700 0.1 183,100 5500 0.1O

A

ifeLptr“tatpimpmaO

ruabetttads

twcts

3

3

es

oirtAic4muwimMhiotf1950s and 1960s. The trends in incidence and mortality inthe Nordic countries are the best known examples. Declinesin incidence parallel the extent and coverage of the organ-

ro-pharynx 12 24,400 2900

ll sites 5,016,100 111,500

n France and Italy for women aged over 30 in 1995. Someorm of “reallocation” of these deaths to more specific cat-gories is generally necessary before rates are calculated.oos et al. [31] have described appropriate methods, and haverovided “corrected” mortality figures for 35 European coun-ries. A further problem arises when incidence and mortalityates are calculated using the entire female population as thepopulation-at-risk”; women who have had a total hysterec-omy for reasons other than the presence of cervical neoplasiare not at risk of cervical cancer, and should be excluded fromhe population-at-risk. However, this is difficult since therevalence of hysterectomy is generally unknown, althought can be substantial in some age groups and countries, and

ay vary by time as well as place and age. Correction of theopulation-at-risk may have a substantial impact on the esti-ated rates of incidence and mortality, especially in older

ge groups, although the impact on the observed trends inntario and England and Wales was not significant [5].Cytological screening programmes have been shown to

educe incidence or mortality of cervical cancer in pop-lations where they have been intensive or systematicallypplied. Often, these reductions have taken place against aackground increase in incidence that affects successive gen-rations (birth cohorts). These changes in risk are most likelyo result from changing sexual behaviour with increasedransmission of oncogenic HPV types. The study of timerends of cervical cancer has therefore been important inssessing both changes in exposure among women born inifferent generations, as well as the effectiveness of organisedcreening programmes.

The following section reviews the historic and current pat-erns of cervical cancer worldwide, and provides scenariosith regards to future rates at the global level. Trends in cervi-

al cancer are also compared with other HPV-related cancerso assess the extent to which HPV-related cancers share theame high-risk (HR) HPV as risk factors.

.1. Trends before and after introduction of screening

.1.1. Developed countriesCurrent patterns of incidence worldwide reflect the net

ffect of two influences: the underlying risk of disease (pre-umably related to transmission of HPV) and prevention

Fc

0.1 27,700 3300 0.1

2.2 5,827,500 449,600 7.7

f its manifestation as invasive cancer by effective screen-ng. The geographic pattern described earlier is relativelyecent. Before the introduction of screening programmes inhe 1960’s and 1970’s, the incidence in most of Europe, Northmerica, and Australia/New Zealand was much as we see

n developing countries today [5] (Fig. 8). Rates of cervi-al cancer incidence and mortality have declined in the last0 years in many western countries (Fig. 9). The declines inortality predated the introduction of screening in some pop-

lations (e.g. the US), and was ascribed to factors correlatedith improving socioeconomic levels [5], as well as a steady

mprovement in stage at diagnosis, and more effective treat-ent and improved survival certainly played a major role [5].ore recently, the beneficial effects of screening programmes

ave been recognised. Since the efficacy of cervical screen-ng was never tested in randomised controlled trials, muchf the evidence of its effectiveness is based on the reduc-ion in incidence and mortality from invasive cervical cancerollowing the introduction of organised programmes in the

ig. 8. Time trends in age-standardised (World) incidence rates of cervicalancer incidence in four Nordic countries [48].

D. Maxwell Parkin, F. Bray / Vaccine 24S3 (2006) S3/11–S3/25 S3/19

Fig. 9. (a) Time trends in age-standardised (World) incidence rates of cervical cancer incidence based on data from selected cancer registries accepted insuccessive Volumes of Cancer Incidence in Five Continents [48]. (b) Time trends in age-truncated (World, ages 20–34) incidence rates of cervical cancerincidence based on data from selected cancer registries accepted in successive volumes of Cancer Incidence in Five Continents [48].

S / Vaccin

itm

ibciiwhitruct

3

ibdisicicuiacDCot

3h

chttwa

ctico

c

dfdhilp

aobFaopcnvat

3

ltrIAiwrdrpiilessnsca

2sneyli

3/20 D. Maxwell Parkin, F. Bray

sed programmes, and were most marked in the age groupsargeted by these programmes [5]. Similar observations were

ade in Canada and the US [5].In spite of the overall declines in crude or age-adjusted

ncidence and mortality in Western countries, increases haveeen reported among young women, presumably reflectinghanging sexual habits and increased transmission of HPVn younger generations. This phenomenon was first describedn England and Wales [32], where successive generations ofomen born since the mid-1930s have been at increasinglyigh-risk. Similar observations have been made elsewheren Europe [5]. In some countries of Eastern Europe, wherehere has been little or no screening, mortality rates have beenising rapidly (e.g. Bulgaria, Romania, and Russia), partic-larly amongst recently born generations. In more affluentountries, however, there is evidence of recent declines (e.g.he Czech Republic, Hungary and Poland).

.1.2. Developing countriesLess information is available regarding trends in develop-

ng countries. In general, incidence and mortality rates haveeen relatively stable in many countries, or shown modesteclines [5]. The absence of an overall decline – as observedn high resource populations – probably reflects the lack ofcreening implementation, or where programmes have beenntroduced, their low population coverage and poor qualityytology. In Fig. 9, declines in incidence are relatively slightn Bombay, India, in contrast to the dramatic declines inervical cancer in Shanghai. The declines in Chinese pop-lations have been attributed to screening, treatment andmproved female genital hygiene, although increasing ratesre seen among younger women and this has been linked tohanging economic circumstances and sexual behaviour [5].espite the steady decline in cervical cancer, incidence inali, Colombia (Fig. 9) – rates in the 1990s were one-thirdf those observed three decades earlier – the population con-inues to have one of the highest rates in the world [3].

.2. Trends in cervical cancer incidence by age andistology

Temporal studies have often failed to separate adeno-arcinomas from SCCs, even though there may be someeterogeneity in their aetiology, and in their susceptibilityo detection by cytology screening. Screening using the Papest has been shown to effectively detect SCC in early stages,hile it has been considered much less effective at detecting

denocarcinoma [5].In addition to the problems of “Uterus unspecified” can-

ers, and the changes in prevalence of hysterectomy, alludedo above, the study of trends by histological subtype must takento account changes over time in the proportions of cases of

ervical cancer coded as unspecified or of ill-defined histol-gy.

The impact of screening on the incidence of cervical can-er is difficult to separate from the effects of other factors in

3

N

e 24S3 (2006) S3/11–S3/25

etermining the rate of diagnosis. Screening has been in useor several decades in many parts of the world, so that it isifficult to estimate what the risk of cervical cancer wouldave been in the absence of screening. There is also lack ofnformation on the extent and quality of screening, particu-arly when a large proportion of tests are outside organisedrogrammes.

One informal way to evaluate changing risk patternslongside the impact of screening, is the joint assessmentf the effects of age at diagnosis, period of diagnosis, andirth cohort using age-period-cohort (APC) modelling [33].or cervical cancer, changing rates among successive gener-tions point to modifications in the population prevalencef persistent infection with oncogenic HPV types, whileeriod effects act as surrogate measures of events that quicklyhange incidence or mortality with the same order of mag-itude regardless of the age group under study. They can beiewed as representing the effects of cytological screenings the intervention should deflect trends downwards acrossargeted age groups over the same period of time.

.2.1. Trends in SCC of the cervixSince most cervical cancers are SCC, studies of all types

argely reflect trends in rates of this histology. Fig. 10 showshe cervical SCC trends in Sweden and Slovakia as part of aecent APC analysis of trends in 13 European countries [34].n this analysis, the non-identifiability problem inherent inPC models (the indexes of age, period and cohort are not

ndependent, but are exactly linearly dependent on each other)as addressed assuming a constant pattern of age-specific

isk over time. There were large declines in risk by calen-ar period in Northern Europe, but a pattern of escalatingisk in successive generations born after 1930 in many Euro-ean countries. The decreased incidence of SCC in Swedens accompanied by a decline in the period slope, implying thatt is the consequence of screening. A recent study reportedittle or no increase in risk in young women in Sweden (due toffective screening), and a levelling off of risk in cohorts bornince the mid-1940s [35]. A bleaker picture emerges fromeveral Southern and Eastern European countries (e.g. Slove-ia and Slovakia (Fig. 10) where trends by period remainedtable, but marked increases in risk among consecutive birthsohort have been observed, related to an absence of screeningmong a population where sexual behaviour has changed.

An earlier birth cohort analysis of cervical SCC trends in5 countries [36] reported declining rates of SCC in succes-ive generations in the US (except US Hispanic), Australia,on-Maori women of New Zealand, and a number of North-rn and Western European countries, in both younger (age<50ears) and older women (age 50–74 years). In contrast, in Fin-and, the UK, Slovakia, Slovenia and Israel, SCC rates havencreased in recent birth cohorts of young women.

.2.2. Trends in adenocarcinoma of the cervixStudies in the last twenty years in Europe and

orth America have reported increasing rates of cervical

D. Maxwell Parkin, F. Bray / Vaccine 24S3 (2006) S3/11–S3/25 S3/21

F f cervicS ed the u

ay[ii1itaosmfiIttirIrsHt

ebtiwir

4i

rHtii

ig. 10. Effects of age, birth cohort and calendar period in the time trends olovakia. Estimates were obtained from an age-period-cohort model that fix

denocarcinoma. Most have noted increasing rates amongounger women (age<40 years) [5]. The European study37] reported increases in age-adjusted adenocarcinoma ratesn the majority of 13 countries as a result of a generationalncrease in risk that first affected women born in the early-930s through to the mid-1940s. As noted earlier, differencesn geographic and temporal profiles of the two main sub-ypes of invasive cervical cancer is in keeping with the rel-tive inability of cytological screening to reduce the ratesf invasive adenocarcinoma. In Fig. 10, there are period-pecific increases in risk in both Sweden and Slovakia, whichay partly be the result of increasing accuracy in the speci-city of the diagnosis of adenocarcinoma with calendar time.nterestingly, the trends in cohort-specific risk in both coun-ries coincide more with the corresponding cervical SCCrends. Interpretation is, however, complex as cohort-specificncreases among recent generations of women may be theesult of an increasing prevalence of oncogenic HPV types.f cohort trends are stable however, it may be that increasing

isk has been diminished by screening programmes (e.g. forquamous cell but not adeno tumours in Northern Europe).owever, other reasons related to sexual behaviour and atti-

udes to screening cannot be discounted.

lrar

al squamous cell carcinoma and adenocarcinoma incidence in Sweden andnderlying age curve [34,37].

Recent work suggests that the efficacy of cytologicalxaminations has improved during the 1990s, and may haveeen responsible for some reductions in adenocarcinoma inhis period [38]. A European study [37] reported declinesn period-specific risk of adenocarcinoma among youngeromen in the U.K., Denmark and Sweden during the 1990s,

ndicating that the Pap test may have had a recent impact ineducing incidence in Europe.

. Quantifying the contributions of changes in HPVnfection and screening on incidence of cancer

Although observed trends are the net result of changes inisk, as determined by infection and persistence of oncogenicPV, and prevention of invasive disease via screening, quan-

ifying their relative contributions is difficult. There is littlenformation on changes in incidence or prevalence of HPVnfection. Increases in HPV-16 have been reported in Fin-

and in women aged in their twenties [39], while in Sweden,ises in HPV-16 during the 1970s and early-1980s in womenged under 35 have been reported [40]. It would be useful toelate the prevalence and distribution of HR-HPVs to birth

S / Vaccin

cfao

hmlwiomcto[

5

cicw

tsbvftwC[

esmtpsbcccg

Fo

3/22 D. Maxwell Parkin, F. Bray

ohorts to better understand how sexual behaviour of dif-erent populations subjects women to an increasingly higherverage risk of cervical cancer in certain countries relative tothers.

In a health services context, it is important to knowow many cases are being prevented by screening and howany cases may be prevented in the future. Studies in Eng-

and and Wales [41] have attempted such a quantification,here an improved programme has successfully countered

ncreases. In estimating the cases prevented in the past,ne must specify an appropriate model and provide infor-ation on the historical screening process. However, pre-

ise quantitative information on organised screening prac-ices is difficult to obtain, and there is little informationn the extent and distribution of opportunistic screening5].

. Trends in other HPV-related cancers

There are few systematic studies of time trends of can-

ers of the external genitalia. In Norway, marked increasesn the incidence of VIN were observed, but not of invasiveancer between 1973 and 1992 [42]; a similar observationas observed in the US [43]. Possibly early diagnosis and

uibl

ig. 11. Time trends in age-truncated (World, ages 30–64) incidence rates of five Hf Cancer Incidence in Five Continents [48].

e 24S3 (2006) S3/11–S3/25

reatment of in situ carcinoma reduces increases in inva-ive vulvar carcinoma incidence. For anal cancers, there haveeen marked incidence increases for over 50 years [44]. Con-ersely, there have been recent declines in mortality ratesrom oral and pharyngeal cancers in many European coun-ries, as well as in Australia and North America (coincidingith lung cancer); however, they are still rising in most ofentral and Eastern Europe, and until very recently, in Japan

45].In the context of examining the impact of HPV, it is inter-

sting to compare cervical cancer trends in the absence ofcreening, with those of the external genitalia and of theouth and of the pharynx. The observation of simple associa-

ions between these trends would be expected if a substantialroportion of cancers at these sites due to infection with theame HR-HPV types. In Fig. 11, there are no striking resem-lances in the trends in HPV-related cancers and cervicalancer in the four countries studied. However, there is someorrelation between the temporal patterns of cervical can-er and those for cancers of the penis and of other femaleenitalia, at least in the Colombian, Chinese and Indian pop-

lations. This may be anticipated given that the PAR estimatesndicate that about two-fifths of these cancers can probablye attributed to HPV (40%), whereas the proportion is muchower for mouth and pharyngeal cancer.

PV-related cancers in four cancer registries accepted in successive Volumes

D. Maxwell Parkin, F. Bray / Vaccine 24S3 (2006) S3/11–S3/25 S3/23

Table 3Predicted number of cervical cancer cases in 2010 and 2020 by world area and age. Projections assume rates estimated for 2002 hold into the future. Adaptedfrom [2]

2002 2002 (% burden ofcervix cancer 2002)

2010 (% changefrom 2002)

2020 (% changefrom 2002)

2020 (% burden ofcervix cancer 2020)

World 493,000 100 584,500 (19%) 702,500 (42%) 100Women aged <65 396,500 80 470,000 (19%) 549,000 (38%) 78Women aged ≥65 96,500 20 114,500 (18%) 153,500 (59%) 22

Less developed areas 409,000 83 505,500 (23%) 639,500 (56%) 83Women aged <65 336,000 82 413,500 (23%) 507,500 (51%) 79Women aged ≥65 73,000 18 92,000 (25%) 132,000 (80%) 21

More developed areas 83,000 17 89,000 (06%) 92,500 (11%) 17Women aged <65 60,000 72 63,500 (05%) 62,500 (03%) 67

2

6

rb(neiwpap[

oiie[cw

dlsca

6

tahastab

cd

va2pacag9tb

D

R

Women aged ≥65 23,000 28

. Projections of HPV-related cancers

Planning cancer services requires knowledge of the cur-ent and likely future patterns of occurrence. Predictions cane used to make decisions on future provision of servicesor post hoc, to investigate why their expected impact wasot achieved). Both screening and HPV infection will influ-nce future rates, and accurate predictions will depend uponnformation on both changing levels of persistent infectionith HR-HPVs and the relative impact of interventions in theopulation. The assessment of the impact (epidemiologicallynd economically) of HPV vaccine, requires fairly com-lex mathematical modelling with a scenario-based approach46].

Simpler approaches usually involve a projection basedn past trends that represent the composite effect of chang-ng risk and preventive interventions. Here the APC models commonly applied, wherein the period and birth cohortffects are proxies for events that cannot be measured directly47]. The cancer burden measured by the number of newases, is calculated by multiplying the predicted cancer ratesith population forecasts.In countries where the HPV vaccine is likely to be intro-

uced, predicted numbers based on such extrapolations areikely to become increasingly inaccurate; however, they doerve to quantify the effectiveness of a proposed vaccine byomparison of the numbers predicted in its absence with thosectually observed subsequently.

.1. Global projections of cervix cancer 2010 and 2020

We assume here that current incidence rates will apply inhe future, and calculate expected numbers using appropri-te population forecasts. Anything more complex would beard to justify, given past trends are quite different betweennd within world regions, and the temporal landscape will

urely be radically altered by the aforementioned interven-ions. Irrespective of changing risk, population growth andgeing are extremely important in determining likely futureurden, and demographic changes will continue to have major

5,222 (09%) 30,000 (31%) 33

onsequences over the next half-century, particularly in theeveloping world.

Table 3 shows the predicted number of new cases of cer-ical cancer based on the estimated incidence rates in 2002pplied to short-term population projections in 2010 and020. In the absence of changing risk or intervention, it isrojected that by 2020, 0.7 million cases will occur, about40% increase from 2002. Four-fifths of the cervical can-

er cases worldwide in 2002 were in less-developed regions,nd, as a result of the much more rapid ageing and populationrowth in these areas, this proportion could increase to over0% by 2020. The biggest relative increase will occur amonghe elderly—an 80% increase from 410,000 cases to 640,000y 2020.

isclosed potential conflicts of interest

Authors have disclosed no potential conflicts of interests.

eferences

[1] IARC Monographs on the Evaluation of Carcinogenic Risks toHumans, vol. 90. Human Papillomaviruses. Lyon: IARC; 2005.

[2] Ferlay J, Bray F, Pisani P, Parkin DM. GLOBOCAN 2002 cancerincidence. Mortality and prevalence worldwide. IARC CancerBaseNo. 5 version 2.0. Lyon: IARC Press; 2004.

[3] Parkin DM, Whelan SL, Ferlay J, Teppo L, Thomas DB, editors.Cancer incidence in five continents, vol. VIII. Lyon: IARC ScientificPublications No. 155 IARC; 2002.

[4] Sadjadi A, Malekzadeh R, Derakhshan MH, Sepehr A, NouraieM, Sotoudeh M, et al. Cancer occurrence in Ardabil. Resultsof a population-based cancer registry from Iran. Int J Cancer2003;107(1):113–8.

[5] International Agency for Research on Cancer. Handbooks of Can-cer Prevention, vol 9. Cervix Cancer Screening. Lyon: IARC Press;

2004.

[6] Ries LAG, Eisner MP, Kosary CL, Hankey BF, Miller BA, Clegg L,Mariotto A, Feuer EJ, Edwards BK, editors. SEER Cancer StatisticsReview, 1975–2002. Bethesda, MD: National Cancer Institute; 2005.http://seer.cancer.gov/csr/1975 2002/.

S / Vaccin

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

3/24 D. Maxwell Parkin, F. Bray

[7] Sant M, Aareleid T, Berrino F, Bielska LM, Carli PM, FaivreJ, et al. EUROCARE-3: survival of cancer patients diagnosed1990–94—results and commentary. Ann Oncol 2003;14(Suppl.5):v61–v118.

[8] Gondos A, Chokunonga E, Brenner H, Parkin DM, Sankila R, BorokMZ, et al. Cancer survival in a southern African urban population.Int J Cancer 2004;112(5):860–4.

[9] Yang BH, Bray FI, Parkin DM, Sellors JW, Zhang ZF. Cervi-cal cancer as a priority for prevention in different world regions:an evaluation using years of life lost. Int J Cancer 2004;109(3):418–24.

10] Maden C, Sherman KJ, Beckmann AM, Hislop TG, Teh CZ,Ashley RL, et al. History of circumcision, medical conditions,and sexual activity and risk of penile cancer. J Natl Cancer Inst1993;85(1):19–24.

11] Tseng HF, Morgenstern H, Mack T, Peters RK. Risk factors forpenile cancer: results of a population-based case-control studyin Los Angeles County (United States). Cancer Causes Control2001;12(3):267–77.

12] Auvert B, Taljaard D, Lagarde E, Sobngwi-Tambekou J, Sitta R,Puren A. Randomized, controlled intervention trial of male circum-cision for reduction of HIV infection risk: the ANRS 1265 trial.PLoS Med 2005;2(11):e298.

13] Smith PG, Kinlen LJ, White GC, Adelstein AM, Fox AJ. Mortal-ity of wives of men dying with cancer of the penis. Br J Cancer1980;41(3):422–8.

14] Bosch FX, Cardis E. Cancer incidence correlations: genital, uri-nary and some tobacco-related cancers. Int J Cancer 1990;46(2):178–84.

15] Tavassoli FA, Devilee P. WHO classification of tumours. pathology& genetics of tumours of the breast and female genital organs. Lyon:IARC Press; 2003.

16] Herbst AL, Ulfelder H, Poskanzer DC. Adenocarcinoma of thevagina. Association of maternal stilbestrol therapy with tumorappearance in young women. N Engl J Med 1971;284(15):878–81.

17] Daling JR, Madeleine MM, Johnson LG, Schwartz SM, Shera KA,Wurscher MA, et al. Human papillomavirus, smoking, and sex-ual practices in the etiology of anal cancer. Cancer 2004;101(2):270–80.

18] IARC Monographs on the Evaluation of Carcinogenic Risks toHumans, vol. 37. Tobacco Habits other than Smoking; Betel-Quidand Areca-Nut Chewing; and some Related Nitrosamines. Lyon,France: IARC; 1985.

19] Parkin DM, Pisani P, Lopez AD, Masuyer E. At least one in sevencases of cancer is caused by smoking. Global estimates for 1985.Int J Cancer 1994;59(4):494–504.

20] IARC Monographs on the Evaluation of Carcinogenic Risksto Humans, vol. 44, Alcohol Drinking. Lyon, France: IARC;1988.

21] Trimble CL, Hildesheim A, Brinton LA, Shah KV, Kurman RJ.Heterogeneous etiology of squamous carcinoma of the vulva. ObstetGynecol 1996;87(1):59–64.

22] Frisch M, Fenger C, van den Brule AJ, Sorensen P, Meijer CJ, Wal-boomers JM, et al. Variants of squamous cell carcinoma of the analcanal and perianal skin and their relation to human papillomaviruses.Cancer Res 1999;59(3):753–7.

23] Bezerra AL, Lopes A, Landman G, Alencar GN, Torloni H, Villa LL.Clinicopathologic features and human papillomavirus dna prevalenceof warty and squamous cell carcinoma of the penis. Am J Surg Pathol2001;25(5):673–8.

24] Rubin MA, Kleter B, Zhou M, Ayala G, Cubilla AL, Quint WG,

et al. Detection and typing of human papillomavirus DNA in penilecarcinoma: evidence for multiple independent pathways of penilecarcinogenesis. Am J Pathol 2001;159(4):1211–8.

25] Kreimer AR, Clifford GM, Boyle P, Franceschi S. Human papillo-mavirus types in head and neck squamous cell carcinomas world-

[

e 24S3 (2006) S3/11–S3/25

wide: a systematic review. Cancer Epidemiol Biomarkers Prev2005;14(2):467–75.

26] Parkin DM. The global health burden of infection-associated cancersin the year 2000. Int J Cancer 2006;118(12):3030–44.

27] Herrero R, Castellsague X, Pawlita M, Lissowska J, Kee F, BalaramP, et al. Human papillomavirus and oral cancer: the InternationalAgency for Research on Cancer multicenter study. J Natl CancerInst 2003;95(23):1772–83.

28] Munoz N, Bosch FX, Castellsague X, Diaz M, de Sanjose S, Ham-mouda D, et al. Against which human papillomavirus types shallwe vaccinate and screen? The international perspective. Int J Cancer2004;111:278–85.

29] Daling JR, Madeleine MM, Schwartz SM, Shera KA, CarterJJ, McKnight B, et al. A population-based study of squamouscell vaginal cancer: HPV and cofactors. Gynecol Oncol 2002;84:263–70.

30] Iwasawa A, Nieminen P, Lehtinen M, Paavonen J. Human papil-lomavirus in squamous cell carcinoma of the vulva by polymerasechain reaction. Obstet Gynecol 1997;89:81–4.

31] Loos AH, Bray F, McCarron P, Weiderpass E, Hakama M,Parkin DM. Sheep and goats: separating cervix and corpus uterifrom imprecisely coded uterine cancer deaths, for studies ofgeographical and temporal variations in mortality. Eur J Cancer2004;40(18):2794–803.

32] Hill GB, Adelstein AM. Cohort mortality from carcinoma of thecervix. Lancet 1967;2(7516):605–6.

33] Holford TR. The estimation of age, period and cohort effects forvital rates. Biometrics 1983;39(2):311–24.

34] Bray F, Loos AH, McCarron P, Weiderpass E, Arbyn M, MollerH, et al. Trends in cervical squamous cell carcinoma inci-dence in 13 European countries: changing risk and the effectsof screening. Cancer Epidemiol Biomarkers Prev 2005;14(3):677–86.

35] Bergstrom R, Sparen P, Adami HO. Trends in cancer of thecervix uteri in Sweden following cytological screening. Br J Cancer1999;81(1):159–66.

36] Vizcaino AP, Moreno V, Bosch FX, Munoz N, Barros-Dios XM,Parkin DM. International trends in the incidence of cervical cancer.I. Adenocarcinoma and adenosquamous cell carcinomas. Int J Cancer1998;75(4):536–45.

37] Bray F, Carstensen B, Moller H, Zappa M, Zakelj MP, LawrenceG, et al. Incidence trends of adenocarcinoma of the cervixin 13 European countries. Cancer Epidemiol Biomarkers Prev2005;14(9):2191–9.

38] Mitchell H, Hocking J, Saville M. Cervical cytology screening his-tory of women diagnosed with adenocarcinoma in situ of the cervix:a case-control study. Acta Cytol 2004;48(5):595–600.

39] Laukkanen P, Koskela P, Pukkala E, Dillner J, Laara E, Knekt P, et al.Time trends in incidence and prevalence of human papillomavirustype 6, 11 and 16 infections in Finland. J Gen Virol 2003;84(Pt8):2105–9.

40] af Geijersstam V, Wang Z, Lewensohn-Fuchs I, Eklund C, SchillerJT, Forsgren M, et al. Trends in seroprevalence of human papil-lomavirus type 16 among pregnant women in Stockholm, Sweden,during 1969–1989. Int J Cancer 1998;76(3):341–4.

41] Peto J, Gilham C, Fletcher O, Matthews FE. The cervical can-cer epidemic that screening has prevented in the UK. Lancet2004;364(9430):249–56.

42] Iversen T, Tretli S. Intraepithelial and invasive squamous cell neo-plasia of the vulva: trends in incidence, recurrence, and survival ratein Norway. Obstet Gynecol 1998;91(6):969–72.

43] Sturgeon SR, Brinton LA, Devesa SS, Kurman RJ. In situ and inva-

sive vulvar cancer incidence trends (1973 to 1987). Am J ObstetGynecol 1992;166(5):1482–5.

44] Melbye M, Rabkin C, Frisch M, Biggar RJ. Changing patterns ofanal cancer incidence in the United States, 1940–1989. Am J Epi-demiol 1994;139(8):772–80.

/ Vaccin

[

[

D. Maxwell Parkin, F. Bray

45] Blot WJ, Devesa SS, McLaughlin JK, Fraumeni Jr JF. Oral andpharyngeal cancers. Cancer Surv 1994;19-20:23–42.

46] Taira AV, Neukermans CP, Sanders GD. Evaluating human papil-lomavirus vaccination programs. Emerg Infect Dis 2004;10(11):1915–23.

[

[

e 24S3 (2006) S3/11–S3/25 S3/25

47] Bray F, Moller B. Predicting the future burden of cancer. Nat RevCancer 2006;6(1):63–74.

48] Parkin DM, Whelan S, Ferlay J, Storm H. Cancer incidencein five continents, vol. I–VIII. Lyon: IARC CancerBase No. 7;2005.