European Journal of Cancer (2013) 49, 3262 3273

Available at www.sciencedirect.com

journal homepage: www.ejcancer.com

Worldwide trends in cervical cancer incidence: Impact

of screening against changes in disease risk factors
Salvatore Vaccarella , Joannie Lortet-Tieulent, Martyn Plummer, Silvia Franceschi,
Freddie Bray
International Agency for Research on Cancer, 150 Cours Albert Thomas, 69372 Lyon Cedex 08, France
Available online 8 June 2013

KEYWORDS
Cervical cancer
Incidence trends
Age-period-cohort
models
Impact of screening

Abstract Background: Cervical cancer trends in a given country mainly depend on the existence of effective screening programmes and time changes in disease risk factors, notably
exposure to human papillomavirus (HPV). Screening primarily inuences variations by period
of diagnosis, whereas changes in risk factors chiey manifest themselves as variations in risk
across successive birth cohorts of women.
Methods: We assessed trends in cervical cancer across 38 countries in ve continents, age
group 3074 years, using age-standardised incidence rates (ASRs) and age-period-cohort
(APC) models. Non-identiability in APC models was circumvented by making assumptions
based on a consistent relationship between age and cervical cancer incidence (i.e. approximately constant rates after age 45 years).
Findings: ASRs decreased in several countries, except in most of Eastern European populations, Thailand as well as Uganda, although the direction and magnitude of period and birth
cohort effects varied substantially. Strong downward trends in cervical cancer risk by period
were found in the highest-income countries, whereas no clear changes by period were found in
lower-resourced settings. Successive generations of women born after 1940 or 1950 exhibited
either an increase in risk of cervical cancer (in most European countries, Japan, China), no
substantial changes (North America and Australia) or a decrease (Ecuador and India).
Interpretation: In countries where effective screening has been in place for a long time the consequences of underlying increases in cohort-specic risk were largely avoided. In the absence
of screening, cohort-led increases or, stable, cervical cancer ASRs were observed. Our study
underscores the importance of strengthening screening efforts and augmenting existing cancer
control efforts with HPV vaccination, notably in those countries where unfavourable cohort
effects are continuing or emerging.
Funding: Bill and Melinda Gates Foundation (BMGF).
2013 Elsevier Ltd. All rights reserved.

Corresponding author: Tel.: +33 (0)4 72 73 80 97; fax: +33 (0)4 72 73 83 45.

E-mail address: vaccarellas@iarc.fr (S. Vaccarella).

0959-8049/$ - see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ejca.2013.04.024

S. Vaccarella et al. / European Journal of Cancer 49 (2013) 32623273

3263

1. Introduction

2.2. Statistical analysis

Invasive cervical cancer (ICC) is the third most common cancer in women worldwide, with an estimated
529,000 new cases in 2008. The burden of cervical cancer
varies considerably worldwide, with more than 85% of
the global burden occurring in low-to-medium-resource
countries, where it is still in many instances the most
common malignancy in women.1 Incidence and mortality rates of ICC have fallen over the past decades in a
number of countries, mainly in high-resource countries
following the introduction of screening programmes
for cervical cancer.25 However, stable or even rising
trends have been observed in countries where screening
activity is either lacking or suers from low-quality and
low-coverage.2,4,6
Persistent infection with oncogenic human papillomavirus (HPV) is considered a necessary cause of
ICC.7 Other cofactors, such as high number of sexual
partners, young age at rst sexual intercourse,8 multiparity,9 oral contraceptive use,10 smoking11 and HIV
infection,7 inuence either the risk of acquisition of
HPV infection or the progression to ICC. HPV infection
could not be accurately detected in large epidemiological
studies until the 1980s, and little is known on time trends
of HPV prevalence in dierent populations.12,13
The comparison of ICC trends in dierent countries
oers, therefore, an opportunity to assess the impact
of screening eorts set against background changes in
ICC risk factors.2 For this purpose, we performed ageperiod-cohort (APC) analyses using incidence data from
high-quality and longstanding population-based cancer
registries from 38 countries to examine ICC patterns
and trends across the major world regions.

Age-standardised incidence rates (ASRs) per 100,000

person-years were calculated by the direct standardisation method, using the World standard population as
a reference.15 ASRs were displayed on a semi-log plot
by 5-year time periods.
We obtained birth cohorts by subtracting age (midpoint of 5-year age band) from the central year of 5-year
calendar period of diagnosis, and plotted trends in incidence rates versus birth cohort and calendar period by
age on a logarithmic scale. APC models were used to
summarise time trends in terms of cohort and period
eects.1618 For each 5-year age-class a = 1, 2,. . ., A
and 5-year period of diagnosis p = 1, 2,. . ., P, the number of events and person-years (D, Y), corresponded to
5  5 year subsets of a Lexis diagram. Under the
assumption of a constant incidence rate k within the
5-year age classes and 5-year period the likelihood of
the contribution from each subset is proportional to
the likelihood for a Poisson observation D with mean
kY. The magnitude of the rates k(a, p) can be therefore
described as a function of age (a), period (p) and cohort
(c) using a log-linear model, with Poisson errors and a
logarithmic link function and with the log of the person-years at risk log(Y) as an oset. The nine distinct
values of a were 3034, 3539,. . ., 7074 years, corresponding to midpoints: 32.5, 37.5,. . ., 72.5 years,
whereas the number of distinct values of p and c varied
across registries (Table 1).
The APC model is a generalised linear model,18 and
the full model can be represented as:
log ka; p aa bp cc log ka; pY ap 
aa bp cc logY ap

2. Methods
2.1. Incidence data
New cases of ICC by age and calendar year of diagnosis were obtained from population-based cancer registries from the series Cancer Incidence in Five Continents
(CI5) Volumes I to IX.14 Population data were obtained
from the same sources. Registries were included in our
study if there was availability of at least 15 consecutive
years of data and they were included in the last volume
of CI5. The last year of diagnosis available in Volume
IX of CI5 was 2002, but more recent data accessible
online, up to 2010, were added, where available
(Table 1).
National data were available for 22 of the 38 eligible
countries. For the remaining populations, regional registries were aggregated to obtain a proxy of the national
incidence. The time span of observations at the country
level varied from 15 to 55 years and analyses were
restricted to ages 30 to 74 (Table 1).

The model suers from the problem of non-identiability on account of the inherent linear interdependency between the three variables. It is impossible to
determine the independent linear eects (slopes) of
the three APC variables. Only the curvature eects,
which represent departures from the linear trend, are
uniquely estimable for the three APC variables. The
age-adjusted sum of the period and cohort slopes, i.e.
the drift16 was used to estimate the annual percentage
change of the regular trend, a quantity that cannot
be attributed specically to period or cohort. We
computed the overall trend and the recent trend (the
relative change in the last two 5-year periods). A
two-sided 95% condence interval (95% CI) for each
estimate was also calculated.
The age distribution of ICC is inuenced by screening
practice. In countries with little or no cervical cancer
screening, ICC incidence rates rapidly increase until
the time premenopausal hormonal changes usually start,
at around the age of 45 years.8,19 Conversely, in screened
populations, incidence rates peak at approximately age

3264

S. Vaccarella et al. / European Journal of Cancer 49 (2013) 32623273

Table 1
Populations included in the analyses, observation period, number of incident cases and ASR for invasive cervical cancer, age 3074.
Country group

Registrya

North Europe

Denmark
19562010
Finland
19562010
Norway
19562010
Sweden
19612010
United Kingdom (UK), 19832007
England
The Netherlands
19892008

(55) 295
(55) 108
(55) 241
(50) 340
(25) 1709

1.6
1.5
1.3
2.5
1.4

21.2
8.2
21.0
14.0
13.7

2.8
3.9
1.4
2.7
4.1

(20)

576

4.7

12.3

1.6 (1.9 to 1.3)

Central/
South Europe

Austria
France*
Italy*
Spain*

19902009
19832002
19882002
19912005

(20)
(20)
(15)
(15)

330
181
199
193

2.4
1.1
1.5
1.5

17.8
15.8
12.8
12.6

3.7
3.5
2.0
0.6

(4.1
(4.1
(2.8
(1.5

3.4)
3.0)
1.2)
0.3)

4.6
2.5
3.1
1.3

(5.9
(4.2
(4.7
(3.1

3.4)
0.7)
1.5)
0.5)

APC
AC
AP
Ad

Baltic countries/
East Europe

Estonia
Latvia
Lithuania
Belarus
Poland*
Russian Fed.
Bulgaria
Croatia
Czech Republic
Slovakia
Slovenia

19682007
19832007
19822006
19782002
19882002
19942008
19942008
19882007
19842008
19782007
19632007

(40) 143
(25) 192
(25) 455
(25) 695
(15) 370
(15) 11356
(15) 1024
(20) 298
(25) 865
(30) 525
(45) 154

0.4
0.7
1.0
3.0
1.1
43.2
2.3
1.3
3.0
1.5
0.6

31.7
22.2
40.6
23.2
32.2
23.3
39.5
24.6
31.2
32.4
31.2

0.7
0.1
1.5
1.6
1.1
0.9
2.8
0.8
1.1
0.1
1.4

(0.9 to 0.5)
(0.3 to 0.5)
(1.2 to 1.8)
(1.8 to 1.4)
(1.7 to 0.4)
(0.8 to 1.1)
(2.4 to 3.2)
(1.2 to 0.3)
(1.2 to 0.9)
(0.1 to 0.3)
(1.5 to 1.2)

1.7
3.3
2.0
0.6
0.8
1.7
2.0
2.1
1.5
1.7
3.0

(0.4 to 3.9)
(1.4 to 5.3)
(0.8 to 3.2)
(1.5 to 0.4)
(2.1 to 0.5)
(1.4 to 1.9)
(1.2 to 2.8)
(3.5 to 0.7)
(2.3 to 0.6)
(0.6 to 2.9)
(4.9 to 1.2)

APC
APC
APC
AC
AC
AC
APC
APC
APC
APC
APC

Northern America/ United States of

Oceania
America (USA) Black*
USA White*
Canada*
Australia
New Zealand

Period

Incidence Person- ASR

Overall
(last
years
(19982002)c trend
5 years)b (last
(95% CI)d
5 years)b
(2.9
(4.0
(1.5
(2.8
(4.2

Recent trend
(95% CI)e

to
to
to
to
to

to
to
to
to

2.7)
3.8)
1.3)
2.6)
3.9)

1.3
1.5
1.0
0.2
2.3

(2.7
(3.8
(2.6
(1.6
(2.8

to
to
to
to
to

APC
modelf

0.1)
0.8)
0.6)
1.1)
1.7)

1.2 (0.1 to 2.3)

to
to
to
to

APC
APC
APC
APC
APC
APC

19752009 (35)

123

0.9

17.5

3.6 (3.9 to 3.4) 3.4 (5.5 to 1.2) APC

19752009
19832002
19842008
19842008

(35)
(20)
(25)
(25)

612
883
581
128

5.7
6.4
5.6
1.1

12.3
13.8
12.1
16.0

1.8
2.0
4.1
4.3

(1.9
(2.2
(4.3
(4.7

to
to
to
to

1.7)
1.7)
3.9)
3.9)

1.2
2.2
1.9
6.1

(2.2
(3.0
(2.9
(8.0

to
to
to
to

0.2)
1.4)
0.8)
4.1)

AC
AC
APC
Ad

to
to
to
to

to
to
to
to

2.6)
1.4)
0.3)
4.4)

Ad
Ad
AC
Ad

South/Central
America

Brazil*
Colombia*
Costa Rica
Ecuador*

19882002
19832002
19832002
19882002

(15)
(20)
(20)
(15)

144
214
274
108

0.2
0.4
0.8
0.3

68.3
55.7
38.3
43.0

4.1
2.1
1.9
5.2

(5.1
(2.6
(2.5
(6.3

3.1)
1.5)
1.4)
4.1)

4.6
0.4
1.8
6.6

(6.5
(2.1
(3.3
(8.7

Asia/Africa

Israel
Singapore
Philippines*
Thailand*
China*
India*
Japan*
Uganda*

19832007
19682007
19832002
19832002
19882002
19832002
19782002
19932007

(25)
(40)
(20)
(20)
(15)
(20)
(25)
(15)

155
175
363
246
483
482
397
148

1.5
1.0
1.0
0.4
3.8
0.9
3.3
0.2

11.6
23.0
43.4
65.2
12.8
63.4
12.0
104.3

0.6
2.2
1.5
0.6
2.7
3.4
5.4
1.9

(0.1 to 1.1)
(2.4 to 2.0)
(1.9 to 1.1)
(0.0 to 1.2)
(3.2 to 2.1)
(3.7 to 3.0)
(5.6 to 5.2)
(0.7 to 3.1)

2.2
3.9
3.0
2.6
1.4
1.6
4.8
0.6

(4.1 to 0.2)
(5.7 to 2.1)
(4.2 to 1.7)
(0.9 to 4.4)
(2.5 to 0.3)
(2.7 to 0.4)
(5.9 to 3.6)
(1.6 to 2.8)

AP
AP
APC
AP
AC
APC
APC
APC

Abbreviations: ASR, age-standardised incidence rate; CI, condence interval; APC, age-period-cohort.
a
List of regional registries, marked with *, which provided data and represent their country: Brazil (Goiania), Canada (British Columbia,
Manitoba, New Brunswick, Newfoundland, Nova Scotia, Ontario, Prince Edward Island, Saskatchewan), China (Hong Kong, Shanghai),
Colombia (Cali), Ecuador (Quito), France (Bas-Rhin, Calvados, Doubs, Ise`re, Somme, Tarn), India (Chennai), Italy (Florence, Lombardy Varese
Province, Modena, Parma Province, Ragusa Province, Romagna Province, Torino), Japan (Miyagi and Osaka Prefectures), Philippines (Manila),
Poland (Cracow City, Kielce, Warsaw City), Spain (Basque country, Granada, Murcia, Navarra and Tarragona), Thailand (Chiang Mai), Uganda
(Kampala), USA: SEER 9 registries (SEER: states of Connecticut, Hawaii, Iowa, New Mexico and Utah and metropolitan areas of San FranciscoOakland (California), Detroit (Michigan), Seattle-Puget Sound (Washington) and Atlanta (Georgia)).
b
Average annual number of cases or person-years (expressed in million person-years at risk) obtained for the most recent 5-year period.
c
Truncated age-standardised rates (world standard population) are computed for 19982002, except for Spain (19962000).
d
Estimated annual percentage change based on the trend variable from the net drift for the overall study period.
e
Estimated annual percentage change based on the trend variable from the net drift for the most recent two 5-year periods.
f
Refers to the most parsimonious model: Ad: age + drift; AC: age + cohort; AP: age + period; APC: age + period + cohort.

35 years, i.e. when the benecial eect of the removal of

precancerous lesions is rst observed. In all populations
however, ICC incidence is approximately constant after

age 45 unless age-specic ICC rates are further distorted

by dierential eectiveness of screening programmes
within dierent periods and cohorts (e.g. a lower uptake

S. Vaccarella et al. / European Journal of Cancer 49 (2013) 32623273

in older cohorts).8 We, therefore, constrained ICC incidence rates to be equal at ages 4549 and 6569, thus
enabling estimation of a unique set of parameters for
the period (with trends deecting downwards among
screened age groups from the time in which screening
was introduced) and cohort eects (sensitive to changes
in exposure to risk factors for ICC, e.g. oncogenic HPV
prevalence, in successive generations of women). The
choice of ages 4549 and 6569 years for setting ages
with equal ICC incidence, as opposed to more adjacent
age categories, allowed more exibility in describing different age curves.
While our assumption forces two age parameters to
assume the same value, no further constraints are necessary, and other parameters for age as well as for period
and cohort can be estimated by the APC model. The reference period was xed to be 19661970 (midpoint
1968.5) or the rst 5-year period for which incidence
data were available; the reference cohort was set to
approximately cover generations born in 19371941.
3. Results
3.1. ASRs of incident cervical cancer
Table 1 shows the incidence series available, by country, and the number of incident ICC cases, person-years
and truncated ASRs for the period 19982002 and ages
3074. Incidence rates varied approximately 10-fold
across study populations, with the lowest ASR in
19982002 observed in Finland (ASR = 8 per 100,000)
and the highest in Uganda (ASR = 104). Incidence rates
in North America, Australia and Europe ranged
between eight in Finland and 21 in Denmark and Norway. In Eastern Europe, ASRs were generally higher,
reaching 40 per 100,000 in Lithuania and Bulgaria.
ASRs in South American countries were also elevated
(between 38 in Costa Rica and 68 per 100,000 in Brazil),
while rates of ICC varied widely in Asian countries, with
high rates in Thailand (ASR = 65), India (ASR = 63),
but relatively low incidence rates in China (ASR = 13)
and Japan (ASR = 12).
3.2. Trends in the incidence of cervical cancer
Trends in ASRs in each country are plotted in Fig. 1.
The estimated annual percentage changes over the study
period and within the 10 most recent years are displayed
in Table 1. Age-standardised incidence rates signicantly declined in 29 of the 38 countries, although the
period of observation varied considerably across countries. Clear downward trends were seen in most European and South American countries, North America,
and Oceania, and in some countries in Asia (e.g. Japan,
China and India). In Northern Europe, however, the

3265

rate of change was lower in magnitude and non-signicant in the most recent period. Increasing incidence rates
were detected in several Eastern European countries in
the last 1020 years, notably in Estonia, Lithuania and
Bulgaria. Of note, ASRs in 1980s were still similarly
high in some countries from Northern Europe and Eastern Europe.
3.3. APC models: Calendar period and birth cohort eects
Fig. 2ad show estimates of age-specic incidence
rates of ICC and incidence rates ratios (IRRs) for calendar period and birth cohort based on APC model
parameters in 32 individual countries, grouped by geographical region. In the Supplementary Fig. S1, APC
results are shown for additional six countries. On
account of the constraint we imposed on the age eect,
countries could be approximately distinguished into
those in which age-specic incidence rates atten early
after ages 3034 (e.g. United Kingdom (UK), England
and United States (US) Whites) and those in which rates
continue to increase to age 4549 (e.g. China and India
but also many European countries). Finally, ASRs are
shown again above the period eect (Fig. 2ad).
3.3.1. Europe other than Eastern Europe
Period-specic declines in ICC were observed in all
countries. The decreases were recorded earlier and were
more pronounced in Northern Europe than elsewhere.
Increases in cohort-specic IRRs of ICC were seen in
cohorts born in the 1940s and thereafter. Cohort-specic
increases were preceded by decreasing risks in the
cohorts born in the rst decades of the 20th century
(Fig. 2a).
3.3.2. Eastern Europe
No substantial changes in period-specic IRRs were
detected in Eastern Europe countries. Similarly to other
European countries, increases in cohort-specic risks
were seen in cohorts born after 1940 or 1950. Rises in
cohort-specic IRRs in countries of the former Soviet
Union were preceded by decreasing risks in the cohorts
born in the rst decades of the 20th century (Fig. 2b).
3.3.3. Americas and Oceania
Declines in period-specic IRRs, comparable with
those observed in Europe (other than Eastern Europe),
were observed in the US Blacks and Whites and in Australia but not in the South American countries under
study, with the possible exception of Brazil. No clear
changes in cohort-specic IRRs were found in Canada
and Australia whereas in US cohorts born after the
1950 there was minor decline and increase in Blacks
and Whites respectively. Downward trends by birth
cohort were also seen in Latin American countries, notably in Ecuador (Fig. 2c).

3266

S. Vaccarella et al. / European Journal of Cancer 49 (2013) 32623273

ASR per 100000 (world)

Northern Europe

Central/Southern Europe

100

100

50

50

20

20
Denmark
Finland

10

Austria
France

Norway
Sweden

10

Italy
Spain

UK, England
The Netherlands

5
1960

1970

1980

1990

2000

2010

1960

ASR per 100000 (world)

Eastern Europe

1980

1990

2000

2010

North America/Oceania

100

100

50

50

20

20
Estonia
Latvia

USA, White
USA, Black

Lithuania

10

1970

Canada

10

Belarus
Russian Federation
Bulgaria

Australia

5
1960

1970

1980

1990

2000

2010

1960

South/Central America

1970

1980

1990

2000

2010

2000

2010

Asia/Africa

100

ASR per 100000 (world)

100
50

50

20

20
Brazil

10

Israel

Colombia
Costa Rica
Ecuador

10

5
1960

1970

1980

1990

2000

2010

Singapore
Philippines
Thailand
China
India
Japan
Uganda

1960

1970

1980

1990

Fig. 1. Age-standardised incidence trends of cervical cancer, by geographical region, age 3074 years. Abbreviation: ASR: age-standardised
incidence rate.

3.3.4. Asia and Africa

Decreases in period-specic IRRs were estimated in
Japan and India. Period eects were absent in other

Asian countries however, increasing cohort-specic

IRRs were estimated in women born after 1950s in
Japan and, possibly, in those born after the 1960s in

S. Vaccarella et al. / European Journal of Cancer 49 (2013) 32623273

Denmark

Finland
100

20

10

0.5

70 1920 1940

20

10

0.5

0.2

Age

Birth Cohort

50

70 1920 1940 1960 1980

1960 1980 2000

30

Norway
5

20

10

Rate Ratio

50

100

0.5

30

50

Birth Cohort

70 1920 1940

Period

Rates per 100000

Rates per 100000

10

10

50

20

10

0.2

0.5
Age

Birth Cohort

50

70 1920 1940 1960 1980

1960 1980 2000

30

UK, England

20

10

0.5

30

50

70 1920 1940

Period

Rate Ratio

Birth Cohort

10

50

20

10

0.5

0.2

Age

Birth Cohort

50

70 1920 1940 1960 1980

1960 1980 2000

30

France
5

20

10

Rate Ratio

50

0.5

30

50

Birth Cohort

70 1920 1940
Age

0.2

2000

100

Period

Rates per 100000

Rates per 100000

10

Period

Italy

100

Age

0.2

2000

100
Rates per 100000

Rates per 100000

10

50

Age

Period

The Netherlands

100

0.2

2000

Sweden

100

Age

Period

Rate Ratio

50

Period

Rate Ratio

30

Birth Cohort

50

10

50

20

10

0.5
Age

0.2

Cohort

Birth Cohort

1960 1980 2000

30
Period

Rate Ratio

Age

Rate Ratio

10
Rate Ratio

10

50

Rates per 100000

Rates per 100000

100

3267

50

Period

70 1920 1940 1960 1980

0.2

2000

ASR

The figure reports the estimates for age, cohort and period effects as estimated by the APC model. Age-standardised rates
(ASRs) are also plotted by period. Age effects and ASRs are shown on a rate per 100,000 scale; cohort and period effects
are on a relative risk scale.
Fig. 2a. Age-period-cohort analysis of trends in invasive cervical cancer incidence, in Europe other than Eastern Europe, age 3074 years.

China. In Singapore and India, IRRs decreased in

cohorts born after 1940. Relatively constant IRRs by
cohort were seen elsewhere in Asia. In Uganda, the sole
African country under study, a decrease was estimated

in successive cohorts born after 1955 alongside increasing period eects (Fig. 2d).
A summary assessment of the contribution of period
and cohort eects to the APC model is reported in Table 1.

S. Vaccarella et al. / European Journal of Cancer 49 (2013) 32623273

Estonia

Latvia
5

20

10

0.5
Birth Cohort

50

70 1920 1940

20

10

0.5

0.2

1960 1980 2000

30

Age

Birth Cohort

50

70 1920 1940 1960 1980

10

50

20

10

Rate Ratio

100

0.5
Age

Birth Cohort

2
30

50

70 1920 1940

Period

100

10

50

20

10

0.5

0.2

1960 1980 2000

30

Age

Birth Cohort

50

70 1920 1940 1960 1980

Poland
5

20

10

0.5
Birth Cohort

50

70 1920 1940

Period

Rate Ratio

50

30

10

50

20

10

0.5

0.2

1960 1980 2000

30

Age

Birth Cohort

50

70 1920 1940 1960 1980

10

50

20

10

Rate Ratio

100

0.5
Birth Cohort

2
30

50

70 1920 1940
Age

Period

0.2

2000

Croatia

Period

100
Rates per 100000

Rates per 100000

Bulgaria

Age

0.2

2000

100
Rates per 100000

Rates per 100000

10

Age

Period

Russian Federation

100

0.2

2000

Belarus
Rates per 100000

Rates per 100000

Lithuania

Period

Rate Ratio

30

Period

Rate Ratio

Age

Rate Ratio

50

10

50

Rate Ratio

10

Rates per 100000

Rates per 100000

100

100

10

50

20

10

0.5
Age

0.2

Cohort

Birth Cohort

1960 1980 2000

Rate Ratio

3268

30

50

Period

Period

70 1920 1940 1960 1980

0.2

2000

ASR

The figure reports the estimates for age, cohort and period effects as estimated by the APC model. Age-standardised rates
(ASRs) are also plotted by period. Age effects and ASRs are shown on a rate per 100,000 scale; cohort and period effects
are on a relative risk scale.
Fig. 2b. Age-period-cohort analysis of trends in invasive cervical cancer incidence, in Eastern Europe, age 3074 years.

The full APC model provided the most parsimonious of

the candidate models in 21 of 38 countries. In eight countries, including the Russian Federation, US Whites and
China, age plus cohort eects provided the best
description.

4. Discussion
This descriptive analysis suggests that, despite general
declines in ASRs globally, the risk of cervical cancer by
calendar period of diagnosis and birth cohort varied

S. Vaccarella et al. / European Journal of Cancer 49 (2013) 32623273

USA, Black

USA, White

50

20

10

0.5

20

10

0.5
Age

0.2

70 1920 1940 1960 1980 2000

30

50

Birth Cohort

Australia

50

20

10

0.5

30

50

Birth Cohort

Period

Rates per 100000

100

10
Rate Ratio

Rates per 100000

100

Age

10

50

20

10

0.5
Age

0.2

70 1920 1940 1960 1980 2000

30

50

Birth Cohort

20

10

0.5

50

Birth Cohort

Period

Rates per 100000

50

Rate Ratio

Rates per 100000

100

10

30

10

50

20

10

0.5
Age

0.2

70 1920 1940 1960 1980 2000

30

50

Birth Cohort

20

10

0.5

50

Birth Cohort

Period

Rates per 100000

10
Rate Ratio

Rates per 100000

100

50

30

Cohort

10

50

20

10

0.5
Age

0.2

70 1920 1940 1960 1980 2000

Age

0.2

Ecuador

100

Age

Period

70 1920 1940 1960 1980 2000

Costa Rica

0.2

Colombia

100

Period

70 1920 1940 1960 1980 2000

Brazil

Age

0.2

70 1920 1940 1960 1980 2000

Canada

Period

Rate Ratio

50

Rate Ratio

30

Period

50

30
Period

50

Birth Cohort

Period

Rate Ratio

Birth Cohort

10
Rate Ratio

10

Rates per 100000

100

Rate Ratio

Rates per 100000

100

Age

3269

0.2

70 1920 1940 1960 1980 2000

ASR

The figure reports the estimates for age, cohort and period effects as estimated by the APC model. Age-standardised rates
(ASRs) are also plotted by period. Age effects and ASRs are shown on a rate per 100,000 scale; cohort and period effects
are on a relative risk scale.
Fig. 2c. Age-period-cohort analysis of trends in invasive cervical cancer incidence, in the Americas and Australia, age 3074 years.

substantially by country and world region in the last

decades. Strong downward trends in ICC risk by period
were observed in the highest-income countries in contrast to the mainly stable period curves seen elsewhere.

Successive cohorts of women born in 194050 and thereafter exhibited either increases in ICC risk (e.g. in nearly
all European countries, Japan and China), no substantial changes (e.g. in North America and Australia) or

S. Vaccarella et al. / European Journal of Cancer 49 (2013) 32623273

Israel

Singapore
100

20

10

0.5

30

50

Period

20

10

0.5
Age

0.2

70 1920 1940 1960 1980 2000

30

50

Birth Cohort

Thailand
100

10
5

20

10

0.5
Birth Cohort

30

50

Period

Rate Ratio

50

Rates per 100000

Rates per 100000

100

Age

10

50

20

10

0.5
Age

0.2

70 1920 1940 1960 1980 2000

30

50

Birth Cohort

20

10

Rate Ratio

50

0.5
Birth Cohort

50

Period

Rates per 100000

Rates per 100000

100

10

30

10

50

20

10

0.5
Age

0.2

70 1920 1940 1960 1980 2000

30

50

Birth Cohort

20

10

Rate Ratio

50

0.5
Birth Cohort

50

Period

Rates per 100000

Rates per 100000

100

10

30

Cohort

10

50

20

10

0.5
Age

0.2

70 1920 1940 1960 1980 2000

Age

0.2

Uganda

100

Period

70 1920 1940 1960 1980 2000

Japan

Age

0.2

India

100

Period

70 1920 1940 1960 1980 2000

China

Age

0.2

70 1920 1940 1960 1980 2000

Philippines

Period

Rate Ratio

Birth Cohort

Rate Ratio

Age

10

50

Rate Ratio

50

Rate Ratio

10

Rates per 100000

Rates per 100000

100

30
Period

Rate Ratio

3270

50

Birth Cohort

Period

0.2

70 1920 1940 1960 1980 2000

ASR

The figure reports the estimates for age, cohort and period effects as estimated by the APC model. Age-standardised rates
(ASRs) are also plotted by period. Age effects and ASRs are shown on a rate per 100,000 scale; cohort and period effects
are on a relative risk scale.
Fig. 2d. Age-period-cohort analysis of trends in invasive cervical cancer incidence, in Asia and Uganda, age 3074 years.

a declining risk (e.g. Ecuador and India). The varying

period and cohort patterns in ICC trends across countries can be largely attributed to two independent factors: (1) the existence, duration, and quality of

screening programmes over calendar time; and (2)

changes in ICC risk factors, notably sexual behaviour
and, hence, the probability of HPV exposure, aecting
consecutive generations of women.

S. Vaccarella et al. / European Journal of Cancer 49 (2013) 32623273

In spite of generally favourable incidence trends, pronounced cohort-specic risk increases were seen in several European countries and Japan, often following
previous decreases among older generations. The
decreased risk of CC for cohorts of women born
between the two world wars is probably due to
improved hygienic conditions and increased access to
healthcare.20 In some other countries (e.g. US Whites,
Singapore and China) the inversion in cohort-specic
risk was only visible in women born after 1960 and
needs to be conrmed with a longer observation period.
Many aspects of sexual behaviour, including earlier age
at rst sexual intercourse and multiple lifetime partners,
have changed substantially starting from generations of
women born during or after the Second World War, as
shown by sexual behaviour surveys and rises in the prevalence of sexually transmitted infections.2124 Although
long-term data on the prevalence of HPV infections
are not available, indirect evidence documents substantial increases in HPV seropositivity12 and prevalence of
HPV-associated precancerous lesions in screened populations suggesting infection was becoming more widespread among young women.25
According to estimated period eects, the benecial
impact of screening in counteracting the underlying
cohort-specic increases in ICC risk was clearest in the
Nordic countries. In these countries, well-organised
screening programmes have been in place for an extended
period of time (e.g. regional screening available from 1967
in Denmark; national screening available from 1963 in
Finland),26 and declines of the formerly high ASRs were
driven by period-specic decreases. Such attributes in the
trends were also observed, although less pronounced, in
Southern European countries, where screening programmes started later and were less well-organised than
in Northern Europe.2729 Of note, period eects in the
US were similar in Whites and Blacks, despite opposing
directions in the cohort eects.
Screening activities have existed for a few decades in
Eastern Europe, Latin America30,31 and Israel32 but they
have not so far produced favourable period eects. In
South America, several countries have attempted to establish national screening programmes in the past two decades
without, however, being able to achieve high-quality and
coverage.33 Screening in Israel is opportunistic with less
than 40% of women being adequately screened.32 Recent
decreases in ASRs, accompanied by modest declines in period-specic risk, observed in Brazil, Singapore, Poland and
a few other countries in Eastern Europe (e.g. see Supplementary Fig. S1), may reect, however, recent improvements in screening programmes. Elsewhere, little or no
screening activities have been in existence, and there was
no evidence of decreases of either ASR or period-specic
risk, except for India and Ecuador.
A possible explanation for the decline in ASRs in
countries without adequate screening like India may

3271

be a decrease in ICC risk among younger, compared

to older generations. Improvements in education and
socioeconomic status may have resulted in delays in
age at marriage and childbearing34 and declines in multiparity. In conservative societies delays in marriage may
be correlated with delays in the age at rst intercourse
and HPV infection, which are associated with ICC risk.8
Multiparity is also a risk factor for ICC,9 and might
have played an important role in secular declines in
some South American and Asian countries for which
birth rates have diminished since 1965 (from an average
of six to less than three by 2000).35 The possibility exists,
however, that in these countries coverage and intensity
of screening have increased progressively in consecutive
generations of women. The impact of screening would
then have been visible as cohort, rather than period,
eect. In China, where fertility rates have also diminished,35 cohort-specic risks have been observed to be
increasing in cohorts born after 1960. If conrmed, a
rise in ICC in young women in any country calls for
urgent preventive measures to avoid a projected increase
in the future. Uganda is the only African country we
have been able to include in our age-period-cohort analysis and the only population to show increases in both
cervical cancer ASRs and period-specic risks. As previous studies suggested,36 these increases may be related to
the heavy burden of HIV and AIDS since the 1980s.
The major strengths of the present analysis are the
use of the most recent and highest quality incidence data
from ve continents and of an innovative APC
approach to disentangle dierent determinants of cervical cancer trends, building on previous work examining
ICC trends in Europe.2 Epidemiological and experimental evidence showed that, although ICC is caused by
HPV, it is also inuenced by sex hormones and its incidence is approximately constant after age 45 years.8,19
By appropriately constraining age-specic rates, we have
been able for the rst time to apply the age-periodcohort model systematically to both unscreened and
screened populations and to derive a unique and meaningful set of parameters for period and cohort eects.
The limitations of our study include the small number
of cancer registries for which more long-term observation is available and the underrepresentation of lowand medium-resource countries. In addition, the lack
of national incidence data in 16 of the 38 studied countries required the use of regional incidence data from
one or more cancer registries. The extent to which this
partial information can be considered representative of
the entire national population is uncertain, particularly
for very large and heterogeneous countries, such as Brazil, India and China.
In conclusion, our study showed in many, though not
all, of the 38 eligible countries steady increases in cervical cancer risk in successive generations of women born
after 1940 or 1950. On inspection of the ASRs, in

3272

S. Vaccarella et al. / European Journal of Cancer 49 (2013) 32623273

countries where eective screening programmes have

been in place for a long duration, the consequences of
such unfavourable cohort eects have been largely
avoided; declining overall incidence rates mimic those
of the period eects. Elsewhere, in countries with an
absence or only low-quality screening programmes, the
rising risks in consecutive birth cohorts have caused an
increase or a lack of decrease in overall ASRs. Our study
underscores the heterogeneity of cervical cancer internationally and the importance of strengthening screening
eorts, building on cancer control eorts with the implementation of HPV vaccination, notably in those countries where unfavourable cohort eects are continuing
or emerging.
Contributors
S.V. and F.B. conceived and designed the study.
J.L.T. provided cancer incidence estimates adapted from
the Cancer Incidence in Five Continents database. S.V.,
M.P. and J.L.T. contributed to data collection and data
analysis. S.V., F.B. and S.F. wrote the draft manuscript.
All authors contributed to the interpretation of the data
and approved the nal manuscript.
Role of the funding source
The funders had no role in the design of the study; the
collection, analysis and interpretation of the data; the
decision to submit for publication; or the writing of
the manuscript. The views expressed in this publication
are those of the authors and not necessarily those of the
funders.
Conict of interest statement
None declared.
Acknowledgements
This work was supported by a grant from the Bill and
Melinda Gates Foundation (BMGF), USA, (Grant
Number OPP1053353).
Appendix A. Supplementary data
Supplementary data associated with this article can
be found, in the online version, at http://dx.doi.org/
10.1016/j.ejca.2013.04.024.
References
1. Ferlay J, Shin HR, Bray F, Forman D, Mathers C, Parkin DM.
GLOBOCAN 2008, cancer incidence and mortality worldwide:
IARC CancerBase No. 10 [Internet]. Lyon, France: International
Agency for Research on Cancer; 2010. Available from: http://
globocan.iarc.fr.

2. Bray F, Loos AH, McCarron P, et al. Trends in cervical squamous

cell carcinoma incidence in 13 European countries: changing risk
and the eects of screening. Cancer Epidemiol Biomarkers Prev
2005;14:67786.
3. Reimers LL, Anderson WF, Rosenberg PS, Henson DE, Castle
PE. Etiologic heterogeneity for cervical carcinoma by histopathologic type, using comparative age-period-cohort models. Cancer
Epidemiol Biomarkers Prev 2009;18:792800.
4. Arbyn M, Raifu AO, Weiderpass E, Bray F, Anttila A. Trends of
cervical cancer mortality in the member states of the European
Union. Eur J Cancer 2009;45:26408.
5. Peto J, Gilham C, Fletcher O, Matthews FE. The cervical cancer
epidemic that screening has prevented in the UK. Lancet
2004;364:24956.
6. Arbyn M, Antoine J, Magi M, et al. Trends in cervical cancer
incidence and mortality in the Baltic countries, Bulgaria and
Romania. Int J Cancer 2011;128:1899907.
7. IARC. Monographs on the evaluation of carcinogenic risks to
humans volume 100B: a review of human carcinogens: biological
agents. Lyon: International Agency for Research on Cancer; 2012.
8. Plummer M, Peto J, Franceschi S. Time since rst sexual
intercourse and the risk of cervical cancer. Int J Cancer
2012;130:263844.
9. International Collaboration of Epidemiological Studies of Cervical Cancer. Cervical carcinoma and reproductive factors: collaborative reanalysis of individual data on 16,563 women with
cervical carcinoma and 33,542 women without cervical carcinoma
from 25 epidemiological studies. Int J Cancer 2006;119:110824.
10. International Collaboration of Epidemiological Studies of Cervical Cancer. Cervical cancer and hormonal contraceptives: collaborative reanalysis of individual data on 16,573 women with
cervical cancer and 35,509 women without cervical cancer from 24
epidemiological studies. Lancet 2007;370:160921.
11. International Collaboration of Epidemiological Studies of Cervical Cancer. Carcinoma of the cervix and tobacco smoking:
collaborative reanalysis of individual data on 13,541 women with
carcinoma of the cervix and 23,017 women without carcinoma of
the cervix from 23 epidemiological studies. Int J Cancer
2006;118:148195.
12. Laukkanen P, Koskela P, Pukkala E, et al. Time trends in
incidence and prevalence of human papillomavirus type 6, 11 and
16 infections in Finland. J Gen Virol 2003;84:21059.
13. Afgeijersstam V, Wang Z, Lewensohn-Fuchs I, et al. Trends in
seroprevalence of human papillomavirus type 16 among pregnant
women in Stockholm, Sweden, 19691989. Int J Cancer
1998;76:3414.
14. Curado MP, Edwards B, Shin HR, et al. Cancer incidence in ve
continents, vol. IX. IARC Scientic Publication No. 160. Lyon:
IARC; 2007.
15. Segi M, Kurihara M. Cancer mortality for selected sites in 24
countries (19501957). 2nd ed. Sendai: Tohoku University of
Medicine; 1960.
16. Clayton D, Schiers E. Models for temporal variation in cancer
rates. I: age-period and age-cohort models. Stat Med
1987;6:44967.
17. Clayton D, Schiers E. Models for temporal variation in cancer
rates. II: age-period-cohort models. Stat Med 1987;6:46981.
18. Holford TR. Age-period-cohort analysis. In: Armitage P, Colton
T, editors. Encyclopedia of biostatistics. 1st ed. Chichester,
UK: John Wiley & Sons; 1998. p. 8299.
19. Pike MC, Pearce CL, Wu AH. Prevention of cancers of the breast,
endometrium and ovary. Oncogene 2004;23:637991.
20. Parkin DM, Bray FI, Devesa SS. Cancer burden in the year 2000:
the global picture. Eur J Cancer 2001;37:S466.
21. Bajos N, Bozon M, Beltzer N, et al. Changes in sexual behaviours:
from secular trends to public health policies. AIDS
2010;24:118591.

S. Vaccarella et al. / European Journal of Cancer 49 (2013) 32623273

22. Samandari G, Speizer IS. Adolescent sexual behavior and reproductive outcomes in Central America: trends over the past two
decades. Int Perspect Sex Reprod Health 2010;36:2635.
23. Hogan DP, Sun R, Cornwell GT. Sexual and fertility behaviors of
American females aged 1519 years: 1985, 1990, and 1995. Am J
Public Health 2000;90:14215.
24. Haldre K, Part K, Ketting E. Youth sexual health improvement in
Estonia, 19902009: the role of sexuality education and youthfriendly services. Eur J Contracept Reprod Health Care
2012;17:35162.
25. Peto J, Gilham C, Deacon J, et al. Cervical HPV infection and
neoplasia in a large population-based prospective study: the
Manchester cohort. Br J Cancer 2004;91:94253.
26. Lynge E, Madsen M, Engholm G. Eect of organized screening on
incidence and mortality of cervical cancer in Denmark. Cancer Res
1989;49:215760.
27. Anttila A, von KL, Aasmaa A, et al. Cervical cancer screening
policies and coverage in Europe. Eur J Cancer 2009;45:264958.
28. Anttila A, Ronco G. Description of the national situation of
cervical cancer screening in the member states of the European
Union. Eur J Cancer 2009;45:2685708.
29. Ronco G, van Ballegooijen M, Becker N, et al. Process performance of cervical screening programmes in Europe. Eur J Cancer
2009;45:265970.

3273

30. Zeferino LC, Pinotti JA, Jorge JP, Westin MC, Tambascia JK,
Montemor EB. Organization of cervical cancer screening in
Campinas and surrounding region, Sao Paulo State, Brazil. Cad
Saude Publica 2006;22:190914.
31. de Quadros CA, Victora CG, da Costa JS. Coverage and focus of
a cervical cancer prevention program in southern Brazil. Rev
Panam Salud Publica 2004;16:22332.
32. Shavit O, Raz R, Stein M, et al. Evaluating the epidemiology and
morbidity burden associated with human papillomavirus in Israel:
accounting for CIN1 and genital warts in addition to CIN2/3 and
cervical cancer. Appl Health Econ Health Policy 2012;10:8797.
33. Gakidou E, Nordhagen S, Obermeyer Z. Coverage of cervical
cancer screening in 57 countries: low average levels and large
inequalities. PLoS Med 2008;5:e132.
34. Dhillon PK, Yeole BB, Dikshit R, Kurkure AP, Bray F. Trends in
breast, ovarian and cervical cancer incidence in Mumbai, India
over a 30-year period, 19762005: an age-period-cohort analysis.
Br J Cancer 2011;105:72330.
35. The World Bank Group. The World Bank; 2010. Available from:
http://web.worldbank.org/.
36. Wabinga HR, Parkin DM, Wabwire-Mangen F, Nambooze S.
Trends in cancer incidence in Kyadondo County, Uganda, 1960
1997. Br J Cancer 2000;82:158592.