EUROPEAN JOURNAL OF CANCER
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
available at www.sciencedirect.com
journal homepage: www.ejconline.com
Quality, comparability and methods of analysis of data
on childhood cancer in Europe (1978–1997): Report from
the Automated Childhood Cancer Information System project
E. Steliarova-Fouchera,*, P. Kaatschb, B. Lacourc, V. Pompe-Kirnd, S. Esere,
A. Mirandaf, A. Danzong, A. Ratiuh, D.M. Parkini
a
Descriptive Epidemiology Group, International Agency for Research on Cancer, 150 Cours Albert Thomas, 69372 Lyon Cedex 08, France
German Childhood Cancer Registry (GCCR), University of Mainz, Germany
c
French National Registry of Childhood Solid Tumours, Faculty of Medicine, Vandœuvre-les-Nancy, France
d
Cancer Registry of Slovenia, Ljubljana, Slovenia
e
Cancer Registry of Izmir, Izmir, Turkey
f
Cancer Registry of Southern Portugal, Lisbon, Portugal
g
Cancer Registry of Doubs, France
h
Institute of Oncology, Bucharest, Romania
i
Clinical Trials Service Unit & Epidemiological Studies Unit, University of Oxford, Oxford, UK
b
A R T I C L E I N F O
A B S T R A C T
Keywords:
In collaboration with 62 population-based cancer registries contributing to the Automated
Childhood cancer
Childhood Cancer Information System (ACCIS), we built a database to study incidence and
Europe
survival of children and adolescents with cancer in Europe. We describe the methods and
Epidemiology
evaluate the quality and internal comparability of the database, by geographical region,
Incidence
period of registration, type of registry and other characteristics. Data on 88,465 childhood
Survival
and 15,369 adolescent tumours registered during 1978–1997 were available. Geographical
Registry
differences in incidence are caused partly by differences in definition of eligible cases.
Data quality
The observed increase in incidence rates cannot be explained by biases due to the selection
of datasets for analyses, and only partially by the registration of non-malignant or multiple
primary tumours. Part of the observed differences in survival between the regions may be
due to variable completeness of follow-up, but most is probably explained by resource
availability and organisation of care. Further standardisation of data and collection of additional variables are required so that this study may continue to yield valuable results with
reliable interpretation.
Ó 2006 Elsevier Ltd. All rights reserved.
1.
Introduction
In European populations, about 1% of all malignant neoplasms occur in patients aged less than 20 years.1 This low
frequency represents a major difficulty for studies of putative
risk factors and clinical management, and the problem is fur-
ther accentuated when the number of cases is split into a
wide variety of tumour types, most of which are uncommon
in adults. As a result, international data on childhood and
adolescent cancer are sparse.2–6
The Automated Childhood Cancer Information System
(ACCIS) is a European project aimed at collection, analysis,
* Corresponding author: Tel.: +33 4 72 73 84 66; fax: +33 4 72 73 86 50.
E-mail address: steliarova@iarc.fr (E. Steliarova-Foucher).
0959-8049/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ejca.2006.05.007
1916
EUROPEAN JOURNAL OF CANCER
interpretation and dissemination of data on cancer incidence
and survival of children and adolescents in Europe.7 To date,
the ACCIS database contains data from some 80 populationbased cancer registries, which cover about 50 % of the
population aged 0–14 years and about 25% of the population
aged 15–19 years living in the 35 participating countries. Over
160,000 childhood and adolescent cancers were diagnosed
during the period 1970–2001 in the 1300 million person-years
of observation. The wide coverage and large size of this study
permit identification of geographic and temporal differences
in incidence across Europe, and so provide useful information
for generating aetiological hypotheses.8 Simultaneously,
differences in population-based survival of various groups of
patients help to identify areas for improvement in the
management of childhood and adolescent cancer patients.5,6,9
However, valid conclusions about differences in cancer
incidence or survival between populations and over time
can be drawn only from comparable datasets of high quality
with complete registration and follow-up. This paper describes the methods of data collection and analysis and evaluates internal quality and comparability of the part of the
ACCIS database that was selected for detailed analyses of
incidence and survival of individual and combined tumour
types. It also discusses the extent to which the comparison
of incidence rates and survival may be influenced by the differences in methods of registration and follow-up and identifies priorities for further improvement of data comparability
across Europe.
2.
Material and methods
2.1.
Setting up the ACCIS database
Some 100 European population-based cancer registries were
invited to participate in the ACCIS study. Acceptable data
were received from 78. To participate, the registries submitted a file of records for all cancer cases incident over a specified registration period (from about 1970 onwards) in
children (aged 0–14 years at diagnosis) and adolescents (age
15–19 years) resident in the defined registration area at the
time of diagnosis. All malignant and non-malignant tumours
of the central nervous system (CNS) defined in the International Classification of Childhood Cancer10 (ICCC) were to
be included. Non-malignant tumours of the meninges and
the pituitary and pineal glands were also included (see
Appendix). Each record contained identification and demographic variables (registration number, sex, age, date of
birth), data concerning the incident cancer (date of incidence, topography, morphology, behaviour and grade) and
information on follow-up (date of last contact and vital status at this date). The participating registries also supplied
population data and a questionnaire describing registration
methods, coding systems and data sources for respective
data items, to aid understanding and interpretation of their
data.
Collection of data started in mid-2000. All information received by November 2002 was included in the ACCIS database, stored at the International Agency for Research on
Cancer (IARC) and consolidated in September 2003. Analyses
presented in this issue are based on this version and disre-
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
gard any later data submissions from cancer registries. The
early closing date of the study in some registries results in
part from early date of data submission. All versions of the
database contain anonymous records and any particular case
can be identified only in the originating registry. All contributing registries agreed to participate in this project, which
was approved by the ethical committee of the IARC. All records were verified and standardised centrally in collaboration with the registries concerned. The ACCIS Scientific
Committee evaluated overall quality and comparability of
all submitted datasets.
2.2.
Selection of datasets for analyses
Central verification of data aimed at detection of errors, both
random and systematic. Logical consistency between variables was verified using standard10 and new procedures.
Unspecified tumours were converted to specified whenever
possible, according to the Guidelines of European Network
of Cancer Registries (ENCR).11 Registrations of retinoblastoma
and Wilms’ tumours were reviewed to identify bilateral cases.
Coding of records was standardised. Most valid basis of
diagnosis was simplified to four values: microscopically verified (MV), clinically diagnosed, reported from death certificate
only (DCO), and unknown. In the registries that do not have
access to information permitting identification of new registrations from death certificates the proportion of DCO cases
could not be calculated (see Table 2). If necessary, topography
and morphology were converted to the second edition of the
International Classification of Diseases for Oncology (ICD-O2).12 The International Classification of Childhood Cancer14
[the conversion table is reprinted in the Annex] was then used
to present results. The multiple primary tumours were identified (and redefined, if necessary) in compliance with the
IARC/IACR recommended definition.12 Length of survival time
in days was calculated centrally as the difference between the
date of diagnosis and death (if deceased), or closing date (if
alive by that date). In the registries, where the follow-up interval exceeded 1 year, the survival time was extended by a half
of the follow-up interval for those subjects with latest date of
follow-up before the closing date of the study (‘lost to followup’), to account for the average time at risk of dying, according to the actuarial assumption underlying the survival analysis method.15
Population data originated from reports published officially by statistical offices of the respective countries (Table
1). The accuracy of the estimated population at risk is fundamental to the calculation of valid rates of incidence. For each
registration area, the population figures were provided for
every combination of covered calendar year, sex and single
year of age. When, exceptionally, population data for single
calendar years were not available, they were estimated using
linear interpolation. If population figures were unavailable for
single ages in some calendar years, the census data were used
to estimate the proportion of each year of age and these were
applied to the available age groups. Population data were
examined for consistency of temporal changes by sex and
age to discern any random errors due to data transfer. Most
of these procedures were conducted using standard software10,16 or tailored macros in Stata17 and Excel.18
Table 1 – Background information on the cancer registration procedures in the cancer registries selected as comparable and included in the analyses of incidence and
survival in children (age 0–14 years) and adolescents (age 15–19 years) (Source: ACCIS)
Country
coverage
(%)
Incidence
date
definition
Topography
codes
BELARUS, National
1999
100
ENCR
ICD-O-2
DENMARK, National
1943
100
other
ESTONIA, National
1968
100
FINLAND, National
1953
FRANCE, Brittany
Morphology
codes
Multiple
primaries
definition
Follow-up procedures
Source of population data
ICD-O-2
IARC/IACR
Ministry of Statistics and
Analysis of the Republic of
Belarus
ICD-O-1
ICD-O-1, ICD-O-2
IARC/IACR
ENCR
ICD-8, ICD-O-1,
ICD-O-2
ICD-O-1, ICD-O-2
IARC/IACR
100
ENCR
ICD-7, ICD-9
Motnac, 1951
IARC/IACR
1991
5
ENCR
ICD-O-2
ICD-O-2
–
FRANCE, Lorraine
1983
4
ENCR
ICD-O-2, ICD-O-3
ICD-O-2, ICD-O-3
IARC/IACR
FRANCE, PACA & Corsica
FRANCE, Rhone Alps
1905
1987
7
10
other
ENCR
ICD-O-1, ICD-O-2
ICD-O-2
ICD-O-1, ICD-O-2
ICD-O-2
other
other
Active follow-up using
medical records and death
certificates. Patient’s vital
status reviewed at least once
a year.
Patients data linked with
central population registry
and death registry at least
once a year.
Automatic annual link of the
registry’s database with
death certificate database.
Date of death or emigration
recorded. Complementary
manual linkage.
Annual computerised record
linkage with population
registry for date of death and
emigration and with the files
of Statistics Finland for
causes of death.
Changes in vital status
notified to the registry.
Complementary data from
physician and medical
records. Vital status verified
actively once in 3 years.
Medical records and
certificates of death used in
active and passive follow-up
every 3–4 years.
Passive follow-up.
Follow-up is based on
medical records (once a year)
and contact with treating
physician (every 5 years).
Follow-up incomplete for
some 10% of cases.
Denmark statistics
Statistical Office of Estonia,
Tallinn
Statistics Finland
INSEE, France
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
Year
start
EUROPEAN JOURNAL OF CANCER
Country and covered
region
INSEE, France
INSEE, France
INSEE, France
1917
(continued on next page)
1918
Table 1 – continued
Country
coverage
(%)
FRANCE, Doubs
1976
1
FRANCE, Herault
FRANCE, Isere
FRANCE, Manche
1985
1979
1994
FRANCE, Bas-Rhin
Incidence
date
definition
Topography
codes
Morphology
codes
Multiple
primaries
definition
Follow-up procedures
other
ICD-O-1, ICD-O-2
ICD-O-1, ICD-O-2
IARC/IACR
1
2
1
ENCR
ENCR
other
ICD-O-2
ICD-O-2
ICD-O-2
ICD-O-2
ICD-O-2
ICD-O-2
–
IARC/IACR
IARC/IACR
1975
2
ENCR
ICD-O-2
ICD-O-2
IARC/IACR
FRANCE, Haut-Rhin
1988
1
ENCR
ICD-O-2
ICD-O-2
IARC/IACR
FRANCE, Somme
1982
1
ENCR
ICD-O-2
ICD-O-2
IARC/IACR
FRANCE, Tarn
GERMANY, NCR (only former East)
1982
1976
1
100
ENCR
ENCR
ICD-O-2
ICD-O-2
ICD-O-2
ICD-O-2
IARC/IACR
IARC/IACR
Follow-up based on
systematic monthly review
of death certificates in the
registration area and in local
press. Active follow-up for
survival studies mainly from
population offices.
Non-systematic follow-up.
–
Follow-up for cases incident
in 1994 and 1995 only. Based
on monthly matching of
registry file with cancer
death certificates and annual
verification of regional
population office after a
formal request.
Follow-up based on
systematic monthly review
of death certificates (all
causes) in the registration
area. Cases without date of
death followed-up actively
every 5 years through
regional population office.
Follow-up for cases incident
in 1988–1991 only. Vital
status from population
offices, hospital records and
postal survey.
Active follow-up through
medical records and
population offices in the
département every 3 years.
–
Follow-up for cases incident
in 1970–1987. Passive followup conducted yearly. Losses
to follow-up varied 1–5%,
and some 30% in year 1987.
Source of population data
INSEE, France
INSEE, France
INSEE, France
INSEE, France
INSEE, France
INSEE, France
INSEE, France
INSEE, France
Staatliche Zentralverwaltung
für Statistik, Berlin, Germany
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
Year
start
EUROPEAN JOURNAL OF CANCER
Country and covered
region
100
other
ICD-O-2
ICD-O-2
other
HUNGARY, National
1971
100
ENCR
Tumour types,
ICD-O-2
Tumour types,
ICD-O-2
-
ICELAND, National
1955
100
other
ICD-O-2
ICD-O-1, ICD-O-2
other
IRELAND, National
1994
100
ENCR
ICD-O-2
ICD-O-2
IARC/IACR
ITALY, Piedmont paediatric
1965
6
ENCR
ICD-O-2
ICD-O-2
other
ITALY, Marche
1990
2
ENCR
ICD-O-1
ICD-O-1
IARC/IACR
ITALY, Ferrara
1991
0.5
ENCR
ICD-O-2
ICD-O-2
IARC/IACR
ITALY, Latina
1983
1
ENCR
ICD-9
ICD-O-2
IARC/IACR
ITALY, Liguria
1986
1
ENCR
ICD-O-1
ICD-O-2
other
ITALY, Lombardy
1976
1
ENCR
ICD-O-2
ICD-O-2
IARC/IACR
Active follow-up once a year
through hospitals,
physicians, family. Losses to
follow-up are traced via
population registry 3 years
after last contact at the latest.
Follow-up based on annual
reports from regional
childhood oncology centres.
Active seach for losses to
follow-up at least once a
year.
Vital status updated
automatically from National
Statistics Office once a
month.
Registry files matched
automatically with death
certificate file once a year
(automated cancer
registration).
Vital status checked every 3
years in municipal
population offices.
Vital status checked every 3
years in regional population
offices.
Vital status checked every 2
years in consultation with
regional population offices
and healthcare files.
Vital status verified in
regional population offices
every 2–3 years. Causes of
death from Nominative
Registry of all causes of
death.
Follow-up for date of death
or emigration within Italy
was conducted through the
Municipality Roster and
Ligurian Mortality Registry.
Central Statistical Office of
Hungary: Demographical
yearbook of Hungary,
Budapest
Icelandic National Roster
Central Statistics Office.
Census 91 and 96. Stationery
Office, Dublin
Official files from ISTAT
(census years) or intercensus
estimates
Sistema Statistica Nasionale,
Instituto Nasionale di
Statistica (ISTAT), Rome
Emilia Romagna Regional
Statistical and Informative
Unit (see:
www.regione.emiliaromagna.it)
Official files from ISTAT
(census years) or intercensus
estimates
Assessorato BilancioFinanze, Andamento della
popolazione, Servizio
Statistica, Comune di
Genova, Census and
intercensual estimates.
ISTAT
(continued on next page)
1919
Active follow-up through
regional population offices,
hospital records, other.
Statistical yearbooks. Federal
Statistical Office (ed.).
Metzler-Poeschel, Stuttgart,
Germany
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
1991
1980
EUROPEAN JOURNAL OF CANCER
GERMANY, GCCR (East and West)
GERMANY, GCCR (only former West)
1920
Table 1 – continued
Year
start
Country
coverage
(%)
ITALY, Macerata
1991
0.5
ITALY, Parma
1978
ITALY, Piedmont general
Incidence
date
definition
Morphology
codes
Multiple
primaries
definition
Follow-up procedures
ENCR
ICD-O-2
ICD-O-2
IARC/IACR
1
ENCR
ICD-O
ICD-O-2
IARC/IACR
1985
1
ENCR
ICD-O-2
ICD-O-2
other
ITALY, Ragusa
1981
1
ENCR
ICD-O-1
ICD-O-1
IARC/IACR
ITALY, Sassari
1992
1
ENCR
ICD-10
ICD-O-2
IARC/IACR
ITALY, Tuscany
1985
2
other
ICD-O-2
ICD-O-2
other
ITALY, Umbria
1994
1
ENCR
ICD-O-2
ICD-O-2
other
ITALY, Veneto
1987
3
other
ICD-O-1
ICD-O-1
IARC/IACR
Vital status checked every 3
years in the municipalities
offices of resident
population.
Population offices consulted
once a year to determine
vital status of the regional
residents.
Follow-up through annual
automatic linkage with
municipal population
registry and active follow-up
in population registries of
other Italian cities.
Vital status verified yearly in
regional population register.
Death certificates for all
causes of death received
quarterly from local health
authority.
Vital status verified in
regional population and all
causes mortality registries
annually.
Vital status through
automatic linkage with
regional mortality registry.
Bi-annual active search for
alive cases through regional
population office and
regional health authority
database.
Annual check of vital status
in regional population offices
and death certificates.
Registry data linked with all
causes mortality files.
Complementary search
through population file of
regional health units and
population offices. Average
interval of checks: 3 years.
Source of population data
National Statistic system
ISTAT, Rome
ISTAT 1981 and ISTAT 1991
Sistema Statistico Nationale.
Citta di Torino Ufficio di
Statistica. Annuario
Statistico 1997
Censuses 1981, 1991 and
intercensual estimates,
ISTAT, Rome
ISTAT
Censuses of the 1981 and
1991 and intercensual
estimates taking into
account births and deaths
and the migration rate for
Central Italy, provided by the
Tuscany Region
Umbria in figures; Sistema
statistico Nazionale, Perugia,
Italy
Municipal registries
estimates of population at
31st December. National
Institute of Statistics
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
Topography
codes
EUROPEAN JOURNAL OF CANCER
Country and covered
region
ENCR
ICD-O-2
ICD-O-2
other
Follow-up by linkage with
mortality database and review
of death certificates.
NETHERLANDS, National
1989
100
ENCR
ICD-O-1
ICD-O-2
IARC/IACR
NETHERLANDS, Eindhoven
1995
7
ENCR
ICD-O-1
ICD-O-1
IARC/IACR
NETHERLANDS, DCOG
1973
100
other
-
Four histology groups
other
NORWAY, National
1952
100
ENCR
ICD-7, ICD-O-2
MOTNAC, ICD-O-2
IARC/IACR
SLOVAKIA, National
1976
100
ENCR
ICD-O-1
ICD-O-1
other
SLOVENIA, National
1950
100
ENCR
ICD-10
WHO24.1, ICD-O-1,
ICD-O-2
IARC/IACR
SPAIN, National
1990
55
other
ICD-O-1
ICD-O-1
other
Follow-up for the patients
aged 0–14 at diagnosis
conducted in a framework of a
special ad hoc study.
Follow-up for vital status
through links with municipal
health administration, central
genealogy registry of all
deaths, other paediatric
oncology centres.
Follow-up for vital status
conducted actively (attending
physician) and passively
(national death registry) once
a year.
Follow-up conducted in a
framework of a special ad hoc
study. Survival time was
provided by the registry.
Passive followed-up for vital
status through compulsory
notifications from clinicians
and autopsy reports.
Complementary active followup by manual matching of all
death certificates for the entire
country.
Follow-up for vital status at
least once a year by automatic
linkage of the registry records
with population register of the
country. Childhood cancer
patients also followed-up
through special out-patient
department of the Institute of
Oncology in Ljubljana.
Active follow-up for vital
status once a year through
treating physicians. Date of
death obtained from hospitals.
Central Office of Statistics,
Malta, Annual Demographic
Reviews of the Maltese
Islands, census of population
and housing in Malta, 1995
(www.magnet.mt/home/cos/
index.html)
Statistics Netherlands
Statistics Netherlands
Statistics Netherlands
Statistics Norway
Statistical Office of Slovak
Republic, Bratislava
Statistical Office of Republic
of Slovenia, Ljubljana
National Institut of Statistics
of Spain
(continued on next page)
1921
100
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
1991
EUROPEAN JOURNAL OF CANCER
MALTA, National
1922
Table 1 – continued
Year
start
Country
coverage
(%)
SPAIN, Albacete
1991
1
SPAIN, Asturias
1982
SPAIN, Basque Country
Incidence
date
definition
Morphology
codes
Multiple
primaries
definition
Follow-up procedures
Source of population data
ENCR
ICD-O-2
ICD-O-2
IARC/IACR
Proyeccion de Poblacion del
Instituto Nacional de
Estatistica (www.ine.es)
2
ENCR
ICD-O-2
ICD-O-2
IARC/IACR
1986
15
other
ICD-O-2
ICD-O-1
other
Date of death collected
manually once a year in the
mortality registry of all
causes of death in the region.
Annual follow-up for date of
death by automatic linkage
(date of birth and sex) with
mortality database of cancer
deaths (1982–1994) and all
causes (1985–1994).
Vital status verified by
linkage of the registry file
with mortality dataset for
period 1986–2000.
SPAIN, Canary Islands
1993
4
ENCR
ICD-O-2
ICD-O-2
IARC/IACR
–
SPAIN, Girona
1994
1
ENCR
ICD-O-1
ICD-O-2
IARC/IACR
SPAIN, Granada
1985
2
other
ICD-O-2
ICD-O-2
other
SPAIN, Mallorca
1988
1
ENCR
ICD-O-1
ICD-O-1
IARC/IACR
SPAIN, Navarra
1973
1
ENCR
ICD-O-2
ICD-O-2
other
Vital status verified yearly by
automatic record linkage
with mortality database
covering the Catalonia
region.
Active follow-up for ad-hoc
survival studies through
death certificates, hospital
discharge records and other
clinical records. Linkage with
Spain, National.
A special follow-up survey
for ACCIS study. Vital status
verified from hospital or
physician records and death
certificates of cancer deaths
within the island.
Follow-up for vital status
within the region by monthly
manual checks of death
certificates, complemented
with annual automatic
linkage. Other sources:
clinical records, other health
registries, municipality
registry.
SADEI. Caracteristica de la
poblation de Asturias; Censo
de poblacion de Asturias
(1991); Padron municipal de
habitantes
EUSTAT (Basque Institute of
Statistics) censo de
poblacion y vivienda, 1991.
Gioberno Vasco Vitoria
Gasteiz
National Institut of Statistics
of Spain
Institut d’Estatistica de
Catalunya (www.idescat.es)
Instituto de estadistica de
Andalucia. Un siglo de
demografia en Andalucia: la
poblacion desde 1900.
Sevilla, IEA,1999
L’esperanca de vida a les
baleares. Institut balear
d’estadistica. Palma, Espana
Gobierno de Navarra.
Departamento de Economia
y Hacienda. Servicio de
Estadı´stica. Estadı´stica de
Población de Navarra.
España
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
Topography
codes
EUROPEAN JOURNAL OF CANCER
Country and covered
region
1
ENCR
ICD-O-1
ICD-O-2
other
SPAIN, Zaragoza
1960
2
ENCR
ICD-O-1, ICD-O-2
ICD-O-1, ICD-O-2
IARC/IACR
SWITZERLAND, Basel
1970
5
ENCR
ICD-O-1
ICD-O-1
other
SWITZERLAND, Geneva
1970
5
ENCR
ICD-O-1, ICD-O-2
MOTNAC, ICD-O-1
other
SWITZERLAND,
Graubunden & Glarus
1989
3
ENCR
ICD-O-2
ICD-O-2
IARC/IACR
SWITZERLAND,
St. Gallen Appenzell
1980
8
ENCR
ICD-O-1, ICD-O-2
ICD-O-1, ICD-O-2
other
Annual automatic linkage
with the mortality registry of
all causes of death in
Catalonia.
Registry records matched
yearly with hospital records
and cancer death
certificates.
Continuous automatic
reporting of deaths of the
registered patients from
national database of deaths.
Vital status verified ad-hoc
for survival studies by
contacting control bureau of
the community of
residence.
Continuous verification of
vital status through
anonymous cancer death
certificates and national
anonymous mortality data.
Active follow-up through
population file of Geneva at
the end of every calendar
year.
Date of death through
questionnaires sent on every
5th anniversary of incidence
date to community’s
population offices.
Additional source of
information: death
certificate and obituaries in
newspapers.
Vital status verified on the
5th and 10th anniversary of
the incidence date of cancer
from community registers of
residents. Data on death
from cancer death certificate
of the residents of the
registration area and
obituaries in
newspapers.
Catalonia Statistical Institut
Instituto Nacional de
Estatistica. Madrid Censo de
poblacion de Zaragoza
(1984)
Statistisches amt Kanton
Basel-Stadt, Statistisches
amt Kanton BaselLandschaft
Cantonal Office of Statistics
(Geneva)
Office Federal de la
Statistique, Neuchatel,
Suisse
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
1980
EUROPEAN JOURNAL OF CANCER
SPAIN, Tarragona
Bundessaunt fur Statistik,
Schwiz (Swiss Federal Office
for Statistics); Census data
1980 and 1990 and
intercensual estimates based
on registration of births,
deaths and population
movements
(continued on next page)
1923
1924
Table 1 – continued
Country
coverage
(%)
SWITZERLAND, Valais
1989
4
TURKEY, Izmir
1993
UNITED KINGDOM,
England & Wales
UNITED KINGDOM,
Northern Ireland
Incidence
date
definition
Topography
codes
Morphology
codes
Multiple
primaries
definition
Follow-up procedures
ENCR
ICD-O-1, ICD-O-2
yes, ICD-O-1 & ICD-O-2
IARC/IACR
4
ENCR
ICD-O-2
ICD-O-2
IARC/IACR
Vital status verified routinely
every 5 years after incidence
using questionnaires
completed in the population
offices of 160 municipalities.
Information on death from
death certificates of
residents. Tracing a case
nationally is possible.
Follow-up complete for cases
incident in 1989–1993.
–
1962
100
ENCR
ICD-O-2
ICD-O-2
–
1993
100
ENCR
ICD10
ICD-O-2
other
Passive follow-up for vital
status and emigration
facilitated by flagging cancer
cases in national population
register (NHSCR). Quarterly
updates. Sources: death
certificates (nationally, all
mentioning neoplasms),
hospitals, clinical trials.
Complementary active
requests once a year.
Patients followed-up for
multiple primary tumours,
but data not provided for
ACCIS.
Yearly electronic matching
of registry files with a file of
all deaths from all causes for
Northern Ireland. Extra
information from death
certificates, hospital records,
if necessary. Follow-up
incomplete for some 10% of
cases.
Source of population data
Recensement federal de la
population de 1990. La
population des communes.
Berne, Suisse
Primary heath centers,
responsible for registration
of all people living in the
covered areas, and for
forwarding the figures to the
Provincial Health Directorate
(unpublished data)
Office for National Statistics
(mid-year population
estimates)
General Registers Office,
Oxford House
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
Year
start
EUROPEAN JOURNAL OF CANCER
Country and covered
region
–, not available / not applicable
Topography/Morphology codes: the coding systems used to code tumour types during the study period. Using several systems (marked with ‘‘yes’’ in the relevant columns) usually implies that
the original codes were converted into the most recent coding system.
UNITED KINGDOM,
Scotland
1959
100
other
ICD-9, ICD-10, ICD-O-2
ICD-O
other
Follow-up for vital status by
linking registry files with
national population death
records. Emigrations of
registered cases notified by
the national population
register (NHSCR) which flags
all registered cancer
patients.
General Register Office for
Scotland (mid-year
population estimates)
EUROPEAN JOURNAL OF CANCER
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
1925
Finally, processed data from each registry were evaluated
at meetings of the ACCIS Scientific Committee, according to
accepted19 as well as newly developed criteria. Completeness
of registration was supported by a reasonable value for the
overall incidence rates, no inexplicably low rates for tumour
subgroups, consistent information on registration procedures, previous publications, etc. Incidence rates for particular tumour types prone to under-registration, especially in
childhood cancer registries, such as brain tumours, retinoblastoma, bone tumours, skin tumours and thyroid cancers,
were examined in particular. Quality of the tumour-related
information was assessed from percentage of microscopically
verified cases and those with diagnostic method unspecified.
A low proportion of tumours in the unspecified categories
suggested data with high precision of diagnosis and classification. It was not possible to compare the proportion of
DCO cases in all registries, but registration procedures were
analysed in the registries where DCOs were not registered.
Other indicators examined were the overall proportion of
the records qualified as ‘unlikely’, annual frequency distribution of cases, and sex- and age-specific distribution of selected tumour types.
Several indicators were developed to evaluate completeness of follow-up in the registries. The proportion of cases
lost to follow-up was based on the codes provided by the registries for the cases declared lost to follow-up before the closing date of the study. The percentage of cases followed-up
(follow up 1+ days) represents the proportion of incident
cases included in survival analyses. The proportion of cases
with follow-up of 5 years or more was calculated as a ratio
of the total number of cases with follow-up of at least 5 years
and without a date of death prior to closing date of the study
(numerator) to total number of cases with follow-up of more
than 0 days and without a date of death prior to closing date
of the study (denominator). The proportion of cases followed
up for at least 5 years indicates the reliability of the estimated
5-year cumulative survival probabilities. The mean follow-up
time was calculated for cases with non-zero follow up and not
deceased by the closing date of the study. The follow-up indicators relate to the total number of cases not deceased by the
closing date of the study (rather than to the total number included in survival analysis), to remove the influence of fatality within a cohort.
As a result of evaluation, each dataset was either assessed
as good quality (comparable) and included in the analyses or
excluded. The reasons for the decision were communicated
to the registries. Of the 78 cancer registries included in the
ACCIS database, 15 were considered to be non-comparable
and were therefore excluded from the analyses presented in
this and other papers of this volume. Most commonly, the excluded registries showed one or several signs of incompleteness of registration: very low incidence rates, either overall
or tumour-specific, and diagnosis based exclusively on microscopic verification. A few registries were excluded on the
grounds of unusually high incidence rates in settings where
it was not possible to be sure that non-residents had been removed from the dataset. Some datasets were excluded
mainly because of an unacceptably high proportion of
unspecified tumours, preventing valid international comparisons by tumour group. Finally, a dataset could also be
Region
Registry
Period
Timetrend
Person-years
Age
1978–
1983–
1988–
1993–
Non-
0–14
15–19
1982
1987
1992
1997
mal.
years
years
Years
British Isles
IRELAND, National
1994–1997
UNITED KINGDOM,
1978–1995
3417728
+
Systematic registration
Age
176922141
%
%
%
%
1378156
–
–
–
100
*
–
29
27
27
17
*
*
Laterality
Number of cases
Age
Age
DCO
Unknown
%
%
0–14
15–19
years
years
n
n
%
231
96
0
91
<1
434
Rb, Wt
Basis of diagnosis
MV
21112
–
NOS
C.a.
Follow-up
Closing date
1+ days
1926
Table 2 – Datasets contributed by the European cancer registries for the analyses of incidence and survival in children (age 0–14 years) and adolescents (age 15–19 years),
with indicators of coverage, data quality and follow-up (Source: ACCIS)
Notes
5+ years
Median
%
%
<1
6
<1
31.12.1998
99
0
3
3
3
0
31.1.2001
98
99
12.3
P Nb
England & Wales
UNITED KINGDOM,
1993–1996
1568121
500977
–
–
–
100
20279630
7793722
28
26
23
23
20710420
–
223
115
77
0
0
27
0
31.12.1999
100
12
1.2
2436
1281
94
<1
0
4
<1
31.12.1999
99
82
10.6
–
Northern Ireland
1978–1997
+
Scotland
East
BELARUS, National
1989–1997
ESTONIA, National
1978–1997
+
6551135
2110021
–
–
45
55
25
26
26
24
*
Wt
3200
Wt
810
330
96
0
0
7
0
1.9.2000
99
72
6.6
93
<1
0
15
<1
31.12.1998
96
66
7.3
HUNGARY, National
1978–1997
+
43147391
–
27
27
25
22
*
Rb, Wt
4875
–
96
–
0
2
<1
1.1.2000
99
72
9.5
SLOVAKIA, National
1978–1997
+
25936857
8433782
25
25
25
24
*
Rb, Wt
3289
1281
94
2
0
8
<1
31.12.1997
91
66
7.9
GERMANY, NCR
1978–1989
+
38959777
15342870
43
41
16
*
Rb, Wt
4831
2640
98
0
<1
5
<1
31.12.1987
80
64
6.3
9.3
–
P
P
S
(only former East)
North
1978–1997
+
19093480
7407903
28
25
24
23
*
Rb, Wt
2775
1400
93
<1
2
10
<1
31.12.1997
97
75
FINLAND, National
1978–1997
+
19265469
6789298
26
25
24
25
*
Wt
3012
1295
98
0
<1
9
2
31.12.1998
98
73
8.8
ICELAND, National
1978–1997
+
1275103
429753
25
25
25
25
*
Rb, Wt
174
82
98
0
0
5
0
31.12.2000
100
84
10.7
NORWAY, National
1978–1997
+
17002822
6188827
26
25
24
24
*
2360
1196
97
<1
<1
11
0
1.1.2000
100
80
10.7
ITALY, Piedmont
1978–1997
+
13015954
–
32
26
22
20
*
1970
–
94
<1
0
6
0
31.12.1999
100
87
11.5
P o2
–
–
P o3
Rb, Wt
paediatric
ITALY, Marche
1990–1997
1559718
–
–
39
61
243
88
–
9
10
0
30.9.2000
100
62
6.2
ITALY, Ferrara
1991–1995
174945
100278
–
–
43
57
28
26
80
4
0
17
0
31.12.1998
96
67
5.7
ITALY, Latina
1983–1997
1413695
576820
–
36
33
30
152
90
91
<1
3
17
<1
31.12.1998
99
78
7.6
ITALY, Liguria
1988–1995
565441
303677
–
–
66
34
90
71
81
<1
0
7
<1
15.4.2000
99
89
8.2
ITALY, Lombardy
1978–1997
+
2667855
1143476
30
27
23
20
405
238
94
<1
0
4
<1
23.9.1999
99
66
7.1
ITALY, Macerata
1991–1997
271141
115992
–
–
29
71
ITALY, Parma
1978–1995
+
962924
442951
33
29
25
13
ITALY, Piedmont
1988–1997
1097222
545600
–
–
56
44
+
Rb
*
Wt
*
38
27
89
–
8
6
0
30.9.2000
100
64
6
139
81
93
0
0
6
0
1.4.1999
100
90
12
186
109
97
<1
0
4
<1
31.5.2001
99
80
8.3
o3
o2
general
ITALY, Ragusa
1983–1997
883057
346031
–
35
33
31
ITALY, Sassari
1992–1995
+
304735
143235
–
–
26
74
ITALY, Tuscany
1988–1997
1376966
671048
–
–
54
46
ITALY, Umbria
1994–1996
315902
140191
–
–
–
100
ITALY, Veneto
1990–1996
1730175
816316
–
–
45
55
MALTA, National
1991–1997
570071
197534
–
–
29
71
*
SLOVENIA, National
1978–1997
8220264
2919236
26
26
25
23
+
SPAIN, National
1990–1995
10377774
SPAIN, Albacete
1991–1997
473835
–
199238
–
–
53
47
–
–
30
70
Rb, Wt
112
65
95
0
<1
11
0
30.3.2000
100
84
9.1
41
30
90
0
4
8
0
30.12.1999
100
77
5.4
223
169
63
<1
0
13
<1
31.12.1998
99
56
5.5
Wt
59
36
87
0
0
9
2
31.12.1999
100
31
4.5
Rb, Wt
288
199
95
<1
0
8
<1
31.12.1998
99
54
5.2
Rb
78
27
96
0
<1
3
0
31.12.1999
98
69
6
*
Rb, Wt
990
400
98
0
0
5
0
31.12.1999
99
76
9.9
*
Rb, Wt
1371
92
0
2
4
0
31.12.2000
98
91
6.1
40
92
1
0
8
0
15.9.2000
99
65
6.1
*
Wt
–
57
P o4 Z
SPAIN, Asturias
1983–1997
2768535
1235271
–
38
34
28
374
208
94
2
<1
12
<1
31.12.1997
96
63
6.8
SPAIN, Basque Country
1988–1994
2522128
1234417
–
–
74
26
359
210
95
1
0
10
<1
31.12.2000
97
100
9.5
o4
SPAIN, Canary Islands
1993–1996
1069190
479540
–
–
–
100
150
81
87
4
3
3
0
–
SPAIN, Girona
1994–1997
326767
155384
–
–
–
100
29
97
0
1
5
0
31.12.1997
99
0
2.4
o4
SPAIN, Granada
1988–1997
1695848
–
–
53
47
Rb, Wt
208
98
0
<1
2
0
31.12.1999
98
68
6.6
G
SPAIN, Mallorca
1988–1995
892125
–
–
64
36
Rb
132
98
0
0
5
0
31.12.1998
95
82
7.2
o4
+
–
369826
*
49
–
68
–
–
–
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
South
DENMARK, National
EUROPEAN JOURNAL OF CANCER
UNITED KINGDOM,
1988–1997
1978–1996
1988–1997
1979–1997
1994–1996
1978–1996
1988–1997
1983–1996
1983–1997
1991–1997
FRANCE, Rhone Alpes
FRANCE, Doubs
FRANCE, Herault
FRANCE, Isere
FRANCE, Manche
FRANCE, Bas-Rhin
FRANCE, Haut-Rhin
FRANCE, Somme
FRANCE, Tarn
GERMANY, GCCR
1989–1997
SWITZERLAND,
42
1473
1376
Rb, Wt
Rb, Wt
47
0
99
99
65
97
88
P
National Cancer Registry of the former German Democratic Republic. Data for 1978–87 contributed only to analyses of time trends for Europe as a whole. Data on children for 1988–89 were pooled with GCCR and included in West.
Cases with unspecified histology, including the ICCC categories Ie, IIe, IIIf, Vic, VIIc, VIIIe, IXe, XIIb, Xe (M-8000 to M-8004 only) and XIf (C76 to C80.9 only)
Overlapping registration areas: for the overlapping years, data from the registry with larger coverage are included in each analysis, according to availability (see Table 1 and text)
Paediatric cancer registry; age range for all registrations is 0–14
Provence, Alps, Côte d’Azur
Cancer registry with at least 90% of retinoblastoma cases with known laterality
Survival analyses were possible only for a restricted dataset (see Table 1)
Registrations with unknown basis of diagnosis
Cancer registry with at least 90% of Wilms’ tumour cases with known laterality
Covers only selected areas (see Table 1)
NOS
o1-o5
P
PACA
Rb
S
Unknown
Wt
Z
For explanation, see text and footnote 1).
Systematic screening for neuroblastoma within (part of) registration area (see Table 1)
S
P o5
o5
S o5
P
P Nb
S
S
Nb o1
P Nb o1
NCR
8.5
2.5
5.1
8.1
8.6
8.8
8.3
5.8
9.3
3.4
–
4.9
7.2
8.7
4.2
–
–
1.9
P
P
Number of cases
97
47
56
73
80
81
70
64
87
33
–
49
82
72
33
–
–
31
3.9
4.7
o4
o4
Nb
46
99
100
100
100
–
99
41
48
52
6
4.6
–
8.4
o4
Microscopically verified cases
1.12.1998
1.2.2001
25.5.2000
31.12.1999
30.6.2000
1.1.2000
59
91
100
–
–
92
93
61
44
–
69
7
9.8
n
0
0
0
0
<1
1.7.1999
31.12.1998
31.12.1998
31.12.1998
–
15.8.2000
31.12.1995
31.12.1997
31.5.2000
–
–
1.6.2001
99
100
–
93
66
69
National German Childhood Cancer Registry (until 1990 covering only West and since 1991 the reunified Germany)
3
5
5
2
5
0
5
3
0
0
4
<1
0
<1
0
<1
<1
<1
1.6.2000
31.3.1998
1.1.1999
1.1.2000
–
31.12.1996
97
97
Notes
MV
0
0
3
0
0
0
5
4
2
2
4
5
4
7
2
4
2
7
0
<1
0
0
0
0
31.12.1998
31.12.1997
Median
General cancer registry, which has only contributed data for age-range 0–14
0
<1
0
0
–
–
2
0
0
0
0
<1
<1
0
0
3
0
4
2
1
2
2
4
0
<1
5+ years
Follow-up
1+ days
GCCR
99
97
94
98
99
99
–
–
–
–
–
–
–
–
–
–
–
–
<1
0
<1
2
<1
7
9
%
Closing date
Dutch Childhood Oncology Group
31
101
30
105
84
94
97
100
99
98
97
95
97
91
96
94
45
–
–
–
–
–
1
<1
7
%
C.a.
G
–
251
1360
–
–
79
77
228
14
231
100
125
97
98
93
98
96
7
<1
0
%
NOS
Registrations from death certificate only
67
202
48
177
144
2199
478
2665
9124
12153
117
184
190
498
40
628
211
–
–
–
144
89
94
2
%
Unknown
DCOG
Wt
Rb, Wt
Wt
Wt
Wt
Rb, Wt
Rb
Rb, Wt
Rb
–
%
95
DCO
Basis of diagnosis
MV
Carcinoid of appendix (C18.1, M-8240)
*
971
Rb, Wt
262
527
Wt
332
Rb, Wt
161
403
Rb, Wt
124
144
n
years
15–19
Age
209
272
n
years
0–14
Age
Number of cases
DCO
56
34
56
26
33
25
22
*
*
*
*
*
*
*
*
Laterality
C.a.
–
71
31
27
50
20
100
26
51
19
50
32
32
71
100
18
31
mal.
Non-
Systematic registration
Cases followed-up for 1 or more days, as a percentage of all cases in the registries with follow-up
44
33
44
24
32
24
22
58
39
29
33
36
50
26
–
26
49
25
50
38
33
29
–
25
33
17
%
1997
1993–
Cases followed-up for 5 or more years, as a percentage of all those not deceased by the closing date
–
33
–
25
35
24
26
–
61
–
35
38
–
27
–
26
–
27
–
30
35
–
–
28
36
25
%
1992
1988–
5+ years
Included in time trend analyses
–
–
–
25
–
27
29
–
–
–
–
–
–
27
–
21
–
28
–
–
–
–
–
30
–
28
%
1987
1983–
1+ days
Systematic registration of non-malignant (Non-mal.) tumours
156055
517611
124578
465088
370892
–
1569267
7130225
–
–
342810
629396
468970
1436570
96325
1489370
581332
765760
–
–
–
–
1084075
1219766
642620
+
452140
1450053
367257
1192884
938187
57636767
3909153
19372876
74685000
93284500
915738
1678648
1403085
3725275
292671
4149282
1499508
2109443
11178532
10876853
7357727
3809039
2931115
2954825
1572447
30
%
years
790001
1982
15–19
0–14
years
1942415
1978–
Person-years
Age
Age
Not applicable
1989–1997
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
trend
Time-
*
SWITZERLAND, Valais
St. Gallen Appenzell
SWITZERLAND,
1983–1997
1978–1997
SWITZERLAND, Geneva
Graubunden & Glarus
1978–1997
1983–1997
SWITZERLAND, Basel
NETHERLANDS, Eindhoven
NETHERLANDS, DCOG
1989–1995
1978–1997
NETHERLANDS, National
(only former West)
GERMANY, GCCR
1983–1990
1984–1996
FRANCE, PACA & Corsica
(East and West)
1983–1997
1993–1996
TURKEY, Izmir
FRANCE, Lorraine
1978–1996
SPAIN, Zaragoza
1991–1997
1983–1997
SPAIN, Tarragona
FRANCE, Brittany
1978–1996
Years
Period
SPAIN, Navarra
Registry
–
West
Region
EUROPEAN JOURNAL OF CANCER
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
1927
1928
EUROPEAN JOURNAL OF CANCER
excluded because it was not possible to verify compatibility of
the local classification system of tumours with international
standards.12 Compared with the complete ACCIS database,
application of the data quality and comparability requirements resulted in a selection of 80% of the total person-years
available in children and 51% in adolescents, for the analyses
related to the period 1988–1997. For the analyses of time
trends (1978–1997), the selected person-years represented
89% for children and 71% for adolescents.
2.3.
Overview of the datasets selected for analyses
Sixty-two cancer registries were qualified for inclusion in
the analyses of incidence and 57 in the analyses of survival
(Table 2). Their geographical coverage is illustrated in Fig. 1,
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
separately for childhood and adolescent populations. Table
1 shows the basic characteristics of the registries. The year
of start of registration is an indicator of likely stability of
the incidence rates. Country coverage is the percentage of
the national population at risk served by the cancer
registry.
Deviation from the ICD-O classification of topography and
morphology potentially introduces artificial differences in
incidence and survival between registries for specific tumour
groups. The most striking example was seen for Finnish data
(Table 1), where dissimilar coding system has resulted in a
large proportion of unspecified lymphomas (ICCC subgroup
IIe) [Izarzugaza and colleagues, this issue] and no distinction
between astrocytoma (IIIb) and other gliomas (IIId).13 But even
the different editions of the ICD-O were not directly compara-
Fig. 1 – Europe, showing the countries represented in the ACCIS database and included in various analyses presented in this
volume. (a) For children aged 0–14 years and (b) for adolescents aged 15–19 years. The percentages give an approximate
coverage of the population at risk in the countries without national cancer registration in the 1990s. Source: ACCIS.
EUROPEAN JOURNAL OF CANCER
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
1929
Fig. 1 – continued
ble for some tumour types. For example, the ICD-O-1 codes
are insufficient to define optic nerve glioma (topography code
C72.3 in ICD-O-2), rhabdoid sarcoma (M-8963 in ICD-O-2) and
some other tumour types. Unless the original records contain
the necessary information, the direct conversion is impossible for these tumour types.
The nature of the follow-up procedures indicates the reliability of survival statistics for a given period and area. The
degree of completeness and accuracy of follow-up depend
to some extent on systems of data collection. Direct access
to a national source of identifiable death certificates would
probably yield the most complete follow-up.
Table 2 provides details of the individual datasets included in the analyses. The datasets differed by size (ranging from 300,000 to 177 million person-years) and type of
registry (paediatric or general), administrative area covered
(national or regional), calendar period and number of years
of registration (3–28 years), availability of follow-up data,
etc.
2.4.
Constitution of dataset for analysis of time trends
(1978–1997)
All registries with at least 15 years of registration during 1978–
1997 were included in the analyses of temporal trends for this
20-year period (Table 2). Two other registries were also included: Somme, and PACA (Provence, Alps, Côte d’Azur and
Corsica), because of their substantial contribution to the
French data, despite the slightly shorter period covered. All
three datasets from German cancer registries were also included in the analyses of trends in children, since between
them, they did cover the required 15-year period. To study
trends in adolescents, data from the cancer registry of the
former German Democratic Republic were included for the
1930
EUROPEAN JOURNAL OF CANCER
period 1978–1987 in the analyses of time trends for Europe as
a whole. Counting the German Childhood Cancer Registry
(GCCR) only once, 33 and 31 registries with sufficiently long
registration period thus contributed to the analyses of time
trends of incidence and survival in children, respectively (Table 2). Slightly fewer registries were included in time trends
analyses of incidence (26) and survival (24) for adolescents.
The study period was divided into four successive 5-year periods, 1978–1982, 1983–1987, 1988–1992 and 1993–1997. Registries included in the time trend analyses contributed at
least 3 years’ data to any of these time periods. If only 1 or 2
years were contributed to any of the 5-year period, those
years were excluded from analyses. The distribution of person-years over the four calendar periods is given in Table 2
for each registry.
2.5.
Definition of geographical regions
All 62 datasets were used to describe the geographical, age
and sex-specific patterns of incidence and survival in Europe
for the most recent 10-year period (1988–1997). Participating
countries were grouped into five regions, as shown in Table
2, and defined according to the United Nations (UN) definition
of the European regions,20 with four exceptions:
The United Kingdom (UK) and Ireland constituted a separate region labelled the British Isles, given the large number
of cases available for the UK. These two countries were withdrawn from Northern Europe in the UN classification.
Estonia, one of the Baltic republics, was grouped with Eastern Europe (instead of Northern Europe in the UN classification), because of similar socio-economic determinants with
those in the other countries grouped in East. This also helped
to increase the number of cases available for this region.
Turkey, represented by the western city of Izmir (Western
Asia by the UN categorisation), was grouped with other Mediterranean countries into Southern Europe. The Turkish data
contributed only to the analyses of incidence.
The dataset for the national registry of the former German
Democratic Republic (GDR) was split as follows: the years
1978–1987 were used in the analysis of trends of incidence
and survival for Europe as a whole in children and adolescents. The remaining 2 years covered by the registry of the
former GDR (1988–1989) were pooled with the childhood cancer datasets from the former Federal Republic of Germany
(FRG), and this pool contributed to the West, while the same
2 years were excluded from all analyses for adolescentsj.
Table 2 shows that some areas overlap partially, for example, when a particular geographical area was covered both by
a general and a paediatric cancer registry. To avoid including
such overlapping areas twice in the analysis, only one of the
j
This exceptional treatment of the registry of the former GDR
was motivated by three simultaneous facts. Given the geopolitical
situation of the former GDR considered in the above point (2), it
would belong to the East region. However, data for the former
GDR as a distinct registration area were only made available
within ACCIS until 1989. Since during the relevant period the
former GDR represented about 50% of the population at risk of the
East region, the discontinuation of the data series after 1989
would have had a disproportionate effect on the time trends of
incidence and survival of the East.
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
relevant registries was selected for each particular analysis.
As a rule, paediatric registries covered a larger population at
risk than the overlapping general registries, and were thus included preferentially in the analyses concerning the age
range 0–14 years. However, if a paediatric registry could not
contribute to the time trend analyses, the related general registry was included instead. Since paediatric cancer registries
do not provide data on adolescents, results for this age group
(15–19 years) are based exclusively on information from general cancer registries. A triple overlap can be observed in Table 2 for the Netherlands. The three registries are the
National Cancer Registry (since 1989), the regional cancer registry South based in Eindhoven (covering the entire study period 1978–1997), and the Dutch Childhood Oncology Group
(DCOG), the latter providing national data for childhood leukaemia for the entire study period. All three were used,
according to data availability: DCOG contributed to all analyses of childhood leukaemias, data from the National Cancer
Registry were used for geographical analyses (period 1988–
1997) of all tumours of childhood (except leukaemia-specific
analyses) and incidence analyses of all tumours of adolescents. Eindhoven data were used in time-trend analyses of
incidence and survival in childhood (except leukaemia-specific analyses) and adolescents, as well as in the analysis of
survival in adolescents for the period 1988–1997, because survival data from the National Cancer Registry were available
only for the childhood age-range (Table 1).
2.6.
Variations in registration practices
The registries systematically recording data on non-malignant CNS tumours are flagged in Table 2. This information
is based on the affirmation of this practice by the registry
and the presence of a reasonable proportion of non-malignant tumours in the contributed dataset. The majority of registries were able to provide data on laterality of
retinoblastoma and Wilms’ tumour. Datasets in which this
information was available for at least 90% of cases were included in the analyses by laterality (Table 2).
To calculate the proportion of unspecified tumours in each
contributory dataset, the following tumour types were included: the complete ICCC subgroups Ie, IIe, IIIf, VIc, VIIc,
VIIIe, IXe, XIIb, the unspecified histologies M-8000 to M-8004
in Xe, and the ill-defined and unspecified sites (C76 to C80.9
only) in the subgroup XIf (c.f. Annex).
According to ICD-O-2,12 carcinoid of appendix (C18.1)
should be coded to M-8240/1, i.e. ‘uncertain’ behaviour and
would be, therefore, excluded from the ACCIS database. However, in some registries the behaviour code /3 (malignant) was
reported for this tumour type. The percentage of the cases
diagnosed with a ‘malignant’ carcinoid of appendix (C18.1,
M-8240) is shown in Table 2.
Table 2 also permits evaluation of the length and completeness of follow-up in those registries following up patients for vital status. Some registries contributed a shorter
time period to analyses of survival than to analyses of incidence (Manche and Haut-Rhine (France), the registry of the
former German Democratic Republic, Valais (Switzerland)
and the national cancer registry of the Netherlands, Tables 1
and 2). Legislation in some countries (e.g. Germany, France,
EUROPEAN JOURNAL OF CANCER
the Netherlands, Italy and Spain) does not allow access to the
nation-wide mortality database of identifiable deceased individuals, which greatly reduces the opportunity for a cancer
death being registered in patients who had escaped registration during their life-time. This lack of information is partly
substituted in some registries of Italy and Spain through an
access to regional deaths records, which permit registration
of cancer deaths as DCO cases at least in a selected resident
population.
2.7.
Statistical analysis of incidence
Incidence rates were calculated as the average annual number of cases per million person-years.24 The age-standardised
incidence rate (ASR) for the age-range 0–14 years is the
weighted average of the age-specific incidence rates using
the weights of the World standard population25 for the age
groups 0, 1–4, 5–9 and 10–14k. 95% confidence intervals (95%
CI) were calculated using the Poisson approximation, or exactly, if less than 30 cases were observed.26
Pairs of age-specific and ASRs were compared using rate
ratio and standardised rate ratio (SRR), respectively and their
95% CIs.24 Differences between several incidence rates were
evaluated using Poisson regression models and expressed as
incidence rate ratio (IRR), adjusted for sex, age group and region, as applicable. The reported P-values test the null
hypothesis of no difference in the relevant populations, compared with the reference.
Change of incidence rates over time was evaluated using
Poisson regression models, adjusted for sex, age group and region and expressed as an average annual percent change
(AAPC). The related P-values report the probability of the
slope of the regression line of incidence rate over time being
consistent with zero, i.e. no change over the years.
2.9.
Statistical analysis of survival
The actuarial life-table method15,27 was used to analyse survival of groups of cases defined by period of diagnosis and
possibly other variables, such as age, sex, region of residence,
diagnostic group, etc. Only cases with non-zero survival time
were included in survival analyses. Five-year observed survival is the cumulative actuarial probability of surviving to
k
1931
the 5th anniversary of the incidence date. The 95 % CIs of
the cumulative survival were calculated according to Kalbfleisch and Prentice.28
Differences in survival of two or more groups of patients
were compared for the entire survivorship curves using the
log-rank v2 test.26 A log-rank test for trend was used to test
a gradual change in survival curves over the successive time
periods.29
3.
Results
3.1.
Study size
Screening practices
According to the information received from the cancer registries, systematic population-based screening for neuroblastoma in very young children was limited to a few
registration areas. In the West, children were screened in
the region Rhone-Alps in France, during 1990–199621 and in
Germany, screening was offered to about 50% of eligible children during 1995–2001.22 In the British Isles, neuroblastoma
screening was conducted among children aged less than 1
year in two small areas of the UK.23 The possible impact on
the incidence rates and survival and comparison between
the regions is discussed in full in a paper on neuroblastoma
[Spix and colleagues, this issue].
2.8.
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
Weights being 2.4, 9.6, 10 and 9, respectively.
Overall, 88,465 tumours in children and 15,369 in adolescents
were included in various analyses (excluding the overlapping
registrations). In addition, 2199 childhood leukaemia cases
from DCOG were included in leukaemia-specific analyses in
replacement of 863 leukaemia cases of the two other Dutch
registries [Coebergh and colleagues, leukaemias, this issue].
The precise sample size varied between the different
analyses.
The underlying person years at risk included in the analyses of time trends are illustrated in Fig. 2. The person-years of
observation for children were some 6–12-times more than for
adolescents. This was due to the wider age-range and more
complete geographical coverage of the childhood population,
especially via the large paediatric cancer registries. On the
other hand, distribution of person-years across regions and
over time was much more homogeneous in adolescents than
in children. The drop in person-years for Europe seen in
Fig. 2(b) at the late 1980s represents the end of the contribution of data for adolescents by the cancer registry of the former GDR. In Fig. 2(a), the withdrawal of this registry from
childhood person-years is evident as the dip at 1990 in the
curve for Europe. In fact, case ascertainment in the territory
of the former GDR was incomplete for the year 1990 and it
is thus missing from the ACCIS database. The same area
was covered since 1991 (for children only) by the national German Childhood Cancer Registry (GCCR) (Table 1). Inclusion of
the area of the former GDR registry explains the increase in
person-years for the West since 1991 seen in Fig. 2(a). Other
minor irregularities to the curves for the West and South represent entries to and exits from the time trends dataset of
some smaller registries. The European curve for children thus
reflects the late arrival of the West, the early departure of England and Wales and the temporary lack of data from GDR in
1990.
3.2.
Quality indicators
Table 3 documents overall quality of the dataset used for time
trends analyses of incidence and survival. The proportion of
microscopically verified cases varied between 90% and 99%
for all periods, regions and age-ranges. The lowest proportion
was observed in the British Isles and highest in the West
(since 1983). The proportion of DCO cases was usually lower
then 1% (Table 3), although only the registries where registration of cases from death certificates is possible contributed to
this percentage. The proportion of cases with unknown
method of diagnosis was also very low, with a maximum of
1932
EUROPEAN JOURNAL OF CANCER
a
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
b
Region:
North
EUROPE
North
British Isles
South
British Isles
South
East
West
East
West
40
5
30
Person years (millions)
Person years (millions)
Region:
EUROPE
20
10
0
4
3
2
1
0
1978
1983
1988
Calendar year
1993
1998
1978
1983
1988
Calendar year
1993
1998
Fig. 2 – Distribution of person-years at risk included in the analyses of incidence time trends presented in this volume over
calendar years. (a) Children (age 0–14 years), (b) adolescents (age 15–19 years). Europe includes the person-years contributed
by the former GDR for the period 1978–1989, which are not included in any of the shown European regions. Source: ACCIS.
5% in early 1990s in the British Isles. There were no substantial differences in these quality indicators between children
and adolescents, regions and 5-year periods.
Overall, 3.7% of cases included in the time-trends dataset
for children were unspecified tumour types. Their proportion
decreased from 6.1% in 1978–1982 to 2.9% in 1993–1997 (Table
3). In adolescents, the corresponding percentages were higher: 8.4% on average, decreasing from 8.9% to 7.1% between
the first and the last period. The reduction in the proportion
of unspecified tumours was fairly general (although the proportions were high and stable in the North and low and stable
in the British Isles, Table 3).
Table 4 describes the size and the quality of the dataset
used to examine geographical, sex- and age-specific patterns
of incidence and survival for the 10-year period 1988–1997.
The quality indicators and the follow-up data in this dataset
are consistent with those observed in the time-trend dataset:
the additional cancer registries (62 instead of 33) therefore did
not markedly modify the data quality.
3.3.
Impact of the selection of registries on incidence time
trends
Eleven of the 33 registries included in the analyses of incidence trends contributed only to the final three 5-year periods
(Table 2). The possible effect of these ‘late’ entries, and other
deficiencies in complete coverage within the time-trends data
set is examined in Table 5. Each pair of the adjacent datasets
shown in Table 5 compares the ‘long’ study period (four quinquennia) with ‘short’ one (three quinquennia). Analyses were
also carried out for the period ending in 1995 (datasets 5 to 8),
because the last 2 years of the study period were covered less
well (Fig. 2). The effect of the on-and-off contribution of the
German datasets was examined by comparing the datasets
3, 4, 7 and 8 (excluding German data) to the corresponding
ones (1, 2, 5 and 6, respectively). The analyses were carried
out for both children and adolescents, although the scope
for artefacts due to variable person-years contribution was
much smaller in the time-trends dataset for adolescents.
The increase in incidence rates was consistent in all eight
datasets, although formal statistical significance was not
attained in children within the datasets 4 and 8, covering
the period 1983–1997 and excluding the German datasets.
This lack of significance is the result of the drastic reduction
in the study power, since the overall incidence rates for the
common period in datasets 4 and 8 were not lower than
those in the other compared datasets. Formal comparison
of the ASR for datasets 6 and 8 in children yielded non-significant result: SRR = 1.04, 95% CI = (0.96, 1.13). Dropping
the last 2 years 1996 and 1997 (datasets 5 to 8) reduced the
level of incidence rates slightly in both children and adolescents, which is consistent with the overall tendency of secular increase. Table 5 thus documents that the incidence
trends are unlikely to be the result of the late addition of
newer registries to those with long registration periods. On
the contrary, the registries covering only the final three quinquennia may have slightly reduced, rather than increased,
the rate of change in incidence rates, as seen from the comparison of ASR for the common periods in adolescents and
in children with German data included.
3.4.
Effect of incomplete follow-up on survival results
Table 3 shows the indicators of the completeness of follow-up
for the subsets of data used for time trends analysis of
survival. Overall, 96% of children registered in the registries
Table 3 – Numbers of cases and indicators of data quality by region and age group used for time trend analyses of incidence and survival in children (age 0–14 years) and
adolescents (age 15–19 years) in Europe, 1978–1997 (Source: ACCIS)
Region
Period
Children (age 0–14 years)
Person-years Cases NOS Nonmal.
a
%
n
Adolescents (age 15–19 years)
Basis of diagnosis
Follow-up
MV DCO Unknown 1+ days 5+ years
%
%
%
%
%
%
Person-years Cases NOS Nonmal.
%
%
n
%
Basis of diagnosis
Follow-up
MV DCO Unknown 1+ days 5+ years
%
%
%
%
%
96
98
98
98
98
77
95
23
1978–82
1983–87
1988–92
1993–97
121284732
170572963
168426435
150824389
20
28
28
25
13907
20890
21906
20408
6
4
3
3
3
3
3
3
94
95
95
96
<1
<1
<1
<1
<1
<1
2
<1
96
97
97
93
98
87
88
36
20622735
21169170
14240018
13246564
30
31
21
19
3177
3530
2541
2641
9
9
8
7
1
2
2
3
95
96
96
96
<1
<1
<1
<1
<1
<1
<1
<1
British
Isles
1978–82
1983–87
1988–92
1993–97
57031152
52615567
52790201
34764851
29
27
27
18
6179
6189
6632
4548
3
3
3
4
3
3
3
4
93
92
90
92
<1
<1
<1
<1
1
2
5
3
98
98
98
99
99
99
99
90
2261447
2171285
1791372
1569618
29
28
23
20
315
344
320
302
8
8
5
5
0
0
0
0
92
94
94
95
<1
0
0
<1
0
0
0
0
99
99
100
99
100
100
100
45
East
1978–82
1983–87
1988–92
1993–97
19725395
20006154
18992858
16910976
26
26
25
22
2129
2327
2326
2192
9
6
5
4
2
3
3
3
93
94
95
96
<1
<1
<1
<1
0
0
0
0
92
95
96
98
93
84
89
31
2561929
2437063
2681816
2862995
24
23
25
27
340
336
441
494
15
9
7
5
2
1
2
2
89
94
97
95
6
<1
<1
2
0
0
0
0
90
94
97
95
100
99
100
4
North
1978–82
1983–87
1988–92
1993–97
15090315
14037183
13584110
13925266
27
25
24
25
1996
2042
2045
2238
11
9
8
11
4
3
4
5
95
96
96
94
<1
0
<1
<1
1
<1
<1
1
97
98
98
99
99
100
99
26
5535837
5470333
5063716
4745895
27
26
24
23
932
984
985
1072
10
12
10
8
3
4
3
4
97
96
97
97
<1
<1
<1
0
<1
<1
<1
<1
97
99
99
100
99
99
98
26
South
1978–82
1983–87
1988–92
1993–97
9046622
10520563
9090335
7744451
25
29
25
21
1168
1381
1281
1196
9
7
6
5
5
3
2
2
90
95
96
96
3
<1
<1
<1
<1
<1
<1
<1
97
99
99
99
98
99
96
34
1707033
2629396
2623252
2356491
18
28
28
25
220
404
444
443
8
10
7
7
1
2
2
2
92
95
97
95
3
<1
1
1
0
1
<1
<1
97
97
99
99
93
99
94
18
West
1978–82
1983–87
1988–92
1993–97
3941628
57355305
67496965
77478845
2
28
33
38
470
6879
8828
10234
22
1
<1
<1
2
3
3
3
84
98
99
99
0
<1
0
0
1
<1
<1
<1
99
98
95
87
92
89
77
16
1496339
2298998
2079862
1711565
20
30
27
23
239
331
351
330
8
12
9
9
<1
2
<1
<1
86
93
92
98
0
0
0
0
1
<1
<1
2
98
100
97
98
87
81
74
24
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
Europe
EUROPEAN JOURNAL OF CANCER
%
1+ days Cases followed-up for 1 or more days, as a percentage of all cases in the registries with follow-up.
5+ years Cases followed-up for 5 or more years, as a percentage of all those not deceased by the closing date.
DCO
Cases registered from death certificate only.
MV
Microscopically verified diagnosis.
n
Number of cases.
Non-mal. Includes non-malignant tumours located in CNS and classified in the ICCC group III and subgroup Xa.
NOS
Cases with unspecified histology, including ICCC subgroups Ie, IIe, IIIf, VIc, VIIc, VIIIe, IXe, XIIb, Xe (M-8000 to M-8004 only) and XIf (C76 to C80.9 only).
a Europe includes the data of the former German Democratic Republic, which are not included in any of the regions.
1933
1934
EUROPEAN JOURNAL OF CANCER
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
Table 4 – Numbers of cases and indicators of data quality by region and age group used for geographical analyses of
incidence and survival in children (age 0–14 years) and adolescents (age 15–19 years) in Europe, 1988–1997 (Source: ACCIS)
Region
Age group
Person-years
Cases
Non-mal.
NOS
C.a.
Basis of diagnosis
MV
DCO
Unknown
Follow-up
1+ days
5+ years
Median
%
n
%
%
%
%
%
%
%
%
Years
Europe
0–14
0
1–4
5–9
10–14
15–19
401843320
25662950
105450671
134925501
135804198
44479551
90
6
24
30
30
10
53717
5073
19379
14677
14588
8272
3
3.2
1.7
3.7
3.9
1.9
4
5.5
3.2
4.1
4.2
6.7
0.2
0
0
0.1
0.5
1.9
97
96.8
98.3
96.3
96.1
98.1
0.3
0.6
0.2
0.2
0.2
0.4
1.1
1.2
0.7
1.1
1.5
0.4
95
92
96
95
96
99
62
62
63
62
61
56
5.8
5.7
5.9
5.8
5.7
5.5
British Isles
0–14
0
1–4
5–9
10–14
15–19
92540901
6366126
25510910
30981391
29682474
5240123
95
7
26
32
30
5
11837
1079
4584
3136
3038
968
3.5
4.7
1.9
4.2
4.9
1.3
3.8
6.8
2.4
3.8
4.7
8.5
0
0
0
0
0
0.6
96.5
95.3
98.1
95.8
95.1
98.7
0.9
2.4
0.5
0.8
0.9
0.1
3.9
4.5
2.3
3.9
6.2
0
99
95
99
99
99
100
89
91
90
89
88
49
8.2
8.4
8.4
8.0
8.0
4.9
East
0–14
0
1–4
5–9
10–14
15–19
56614254
3220674
14033863
19130650
20229067
5544811
91
5
23
31
33
9
7718
570
2468
2216
2464
935
2.7
2.6
1.9
3.1
3.1
2.2
5.1
10.2
5.8
4.6
3.7
5.9
0
0
0
0
0
0.7
97.3
97.4
98.1
96.9
96.9
97.8
0.2
0.5
0.2
0.2
0.1
1.2
0
0
0
0
0
0
98
91
98
98
99
96
64
62
65
66
61
42
6.2
6.3
6.4
6.5
5.9
4.4
North
0–14
0
1–4
5–9
10–14
15–19
27509376
1919279
7492182
8959113
9138802
9809611
74
5
20
24
24
26
4283
436
1653
1069
1125
2057
4.4
4.8
2.3
6
5.9
3.7
9.3
12.6
7.1
11.5
9.2
8.9
0.7
0
0
0.8
2
2.2
95.6
95.2
97.7
94
94.1
96.3
0.3
0.5
0.4
0.2
0.1
0.1
0.9
1.1
0.7
1.2
0.6
0.5
98
95
98
99
100
99
59
58
59
59
60
58
5.8
5.9
5.8
5.9
5.9
5.7
South
0–14
0
1–4
5–9
10–14
15–19
39317250
2238515
9210623
12714912
15153200
11536094
77
4
18
25
30
23
5534
453
1824
1460
1797
2153
1.8
1.3
0.9
2.3
2.4
1.7
5.4
5.7
3.8
6.3
6.1
7.8
0.1
0
0
0
0.2
0.3
98.2
98.7
99.1
97.7
97.6
98.3
0.3
0.2
0.3
0.3
0.3
0.9
1.1
1.5
1
1
1.1
0.5
99
97
99
99
99
99
72
72
71
71
72
63
6.1
6.2
6.0
6.1
6.1
6.1
West
0–14
0
1–4
5–9
10–14
15–19
185861539
11918356
49203093
63139435
61600655
12348912
94
6
25
32
31
6
24345
2535
8850
6796
6164
2159
2.9
2.7
1.6
3.7
3.7
0.4
2.4
2.7
2.1
2.6
2.6
3.1
0.2
0
0
0.1
0.7
4.4
97.1
97.3
98.4
96.3
96.3
99.6
0
0
0
0
0
0
0.1
0
0.1
0.1
0.1
0.4
92
89
93
91
92
98
47
49
48
47
45
52
4.4
4.5
4.5
4.2
4.2
5.0
1+ days Cases followed-up for 1 or more days, as a percentage of all cases in the registries with follow-up.
5+ years Cases followed-up for 5 or more years, as a percentage of all those not deceased by the closing date.
C.a.
Carcinoid of the appendix (C18.1, M-8240).
DCO
Cases registered from death certificate only.
MV
Microscopically verified diagnosis.
n
Number of cases.
Non-mal. Includes non-malignant tumours located in CNS and classified in the ICCC group III and subgroup Xa.
NOS
Cases with unspecified histology, including ICCC subgroups Ie, IIe, IIIf, VIc, VIIc, VIIIe, IXe, XIIb, Xe (M-8000 to M-8004 only) and XIf
(C76 to C80.9 only).
conducting follow-up were included in survival analyses. In
adolescents, the start of follow-up was even more complete
(98%), although at the end of 5 years, fewer cases were still
tracked in adolescents (71%) than in children (74%). Followup was most complete in the British Isles and least complete
in the West. The proportion of followed-up patients in the
first three 5-year periods was largely complete and did not
vary much, while large differences were observed between
the regions in the last period. Besides the efficiency of the
tracing procedures, the completeness of the follow-up at 5
years obviously depended on the length of the registration
period (if it was shorter than 5 years), and the timing of the
closing date of the study (Table 2).
The proportion of patients followed-up for 5 years or more
was far from 100% for some cancer registries (Table 2), which
reduced comparability between the regions (Table 3). This was
mostly caused by the early closing date of the study, so that
many cases were withdrawn alive (censored) before the 5-
Table 5 – Simulation of incidence time trends with various subsets of data (Source: ACCIS)
Dataset
Period
Modifications
Children (age 0–14 years)
Adolescents (age15–19 years)
Registries Person-years Cases Common period
n
Millions
Total period
Registries Person-years Cases Common period
n
ASR
AAPC
P > jzj
n
Millions
Total period
n
Rate
AAPC
P > jzj
1978–1997
1983–1997
22
11
413
198
51,896
25,215
134.6
132.7
1.2
1.2
<0.0001
<0.0001
18
8
44
5
7914
798
179.9
171.2
2.0
2.6
<0.0001
0.002
3
4
1978–1997 Germany excluded
1983–1997 Germany excluded
21
10
374
30
47,065
3938
135.1
136.8
1.2
0.7
<0.0001
0.14
17
7
38
5
6783
798
179.3
171.2
2.0
2.6
<0.0001
0.002
5
6
1978–1995
1983–1995
22
11
396
169
49,348
21,210
133.6
130.6
1.2
1.0
<0.0001
<0.0001
18
8
39
4
6985
690
177.0
167.5
2.0
2.7
<0.0001
0.01
7
8
1978–1995 Germany excluded
1983–1995 Germany excluded
21
10
357
27
44,517
3521
134.0
136.2
1.2
0.6
<0.0001
0.095
17
7
33
4
5854
690
175.8
167.5
1.9
2.7
<0.0001
0.01
Common period includes the overlapping years of both compared datasets. Total period refers to all years shown in Period column. Average annual percentage change (AAPC) was calculated from
Poisson regression model, with year as explanatory variable, adjusted for sex, region and, for children only, age group. The registries included can be found in Table 2.
n, number; ASR, age-standardised incidence rate (World standard); Rate, age-specific incidence rate.
Cohort A: 10,000 cases
Cohort B: 10,000 cases
Alive at
Dead Withdrawn 5-year
the end during
during
OS
of 5 years 5 years
5 years
n
Follow-up < 5 years
No withdrawals traced
Follow-up P 5 years
All withdrawals
confirmed deceased
All withdrawals
confirmed alive
1/2 withdrawals
deceased, 1/2 alive
n
n
%
7.5% withdrawals
Cohort C: 10,000 cases
Alive at
the end
of 5 years
Dead
during
5 years
Withdrawn
during
5 years
5-year
OS
Alive at
the end
of 5 years
Dead
during
5 years
Withdrawn
during
5 years
5-year
OS
n
n
n
%
n
n
n
%
15% withdrawals. Number of deaths is the same as in cohort A.
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
Table 6 – Example of modification of 5-year survival estimate in relation to the proportion of withdrawals and their eventual trace-back in a fictitious study of 10,000
subjects (Source: ACCIS)
EUROPEAN JOURNAL OF CANCER
1
2
15% withdrawals. Number of deaths is lower than in
cohort A, proportionately to the number of withdrawals.
5492
3008
1500
68.9
6000
3250
750
66.9
5250
3250
1500
66.3
6000
4000
0
60.0
5250
4750
0
52.5
5250
4750
0
52.5
6750
3250
0
67.5
6750
3250
0
67.5
6750
3250
0
67.5
6375
3625
0
63.8
6000
4000
0
60.0
6000
4000
0
60.0
1935
OS, observed survival, calculated from life table. Number of losses in each follow-up year for cohorts B and C were double those in cohort A.
1936
EUROPEAN JOURNAL OF CANCER
year follow-up was completed. Early censoring may bias the
results of survival analyses, as demonstrated in Table 6. The
5-year survival estimate of 66.3% for group B is based on a cohort of 10,000 cases followed-up incompletely, with 15% of
cases censored at the closing date. Depending on the fate of
the patients, the final 5-year survival (based on completed follow-up) may be as low as 52.5% or as high as 67.5%. With the
same number of deaths, the corresponding range is narrower,
60.0% to 67.5% in the cohort A with only 7.5% of withdrawals.
Assuming that the proportion of deaths among withdrawals
is the same as in the cohort with completed follow-up, the
5-year actuarial probability will be overestimated at an early
closing date, as seen for the cohort C in Table 6.
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
whether all registries or only those with systematic registration of non-malignant tumours were included. Only a marginal difference (for Europe as a whole) was observed for
the same comparison in adolescents, with higher rate in the
dataset composed of selected registries. This was due to a larger weight of North, the region with highest incidence rates,
among the selected (50% of the cases) than among all (24%)
registries. Overall, the incidence rates of non-malignant tumours in children were higher by some 14% when only selected registries were included, compared with the
unrestricted dataset. In adolescents the selection of the registries did not influence the rates of non-malignant tumours
significantly (except in the East), although their rates also
seemed to be considerably higher in the restricted than in
the unrestricted dataset for British Isles, South and West (Table 8). Fig. 3 illustrates the increasing incidence of malignant
and non-malignant tumours over the study period 1978–1997
in Europe as a whole (Fig. 3(a)), and in the individual regions
(Fig. 3(b and c)). The differences between the regions were
somewhat wider in the dataset destined for analysis of the
most recent period 1988–1997 than in the dataset for time
trends (Table 7), reflecting the larger representation of registries registering only malignant tumours, especially in the
South and West, and for adolescents in the British Isles. Table
9 shows that the incidence of both malignant and non-malignant tumours was increasing, slightly faster for non-malignant tumours (non-significantly in adolescents). The rate of
increase was slightly higher when only the registries with
systematic collection of non-malignant tumours were
included.
Selection of the registries according to the registration of
non-malignant tumours did not seem to influence much the
overall survival, as Fig. 4 shows for children. Similar results
were seen in adolescents: a difference between the total
and the restricted dataset was only seen for the British Isles,
where five-year survival with malignant tumours was 72%,
3.5.
Consequence of variable registration of nonmalignant tumours
About 97% of all childhood cancer cases were malignant tumours, the remaining 3% were non-malignant tumours
occurring in the CNS. Due to a good representation of the registries with systematic registration of non-malignant tumours in the childhood datasets, the percentage of nonmalignant tumours was only slightly higher in the restricted
dataset (composed exclusively of the registries with systematic registration of non-malignant tumours, Table 7). In adolescents, the proportion of such registries was lower on
average (except in the North), which resulted in less comparable proportions of non-malignant tumours between the full
and the restricted datasets in Table 7. On the whole, there
seem to be slightly fewer non-malignant tumours among
adolescents than among children, which is consistent with
the lower incidence rates of CNS tumours in adolescents [Stiller, Desandes, Danon and colleagues, this issue] than in children [Peris-Bonnet and colleagues, this issue]. Table 8 shows
that during the most recent 10-year period there were no differences in incidence rates of malignant tumours in children,
Table 7 – Proportion of non-malignant tumours of the total registrations in the various datasets (Source: ACCIS)
Geographical analyses (period 1988–1997)
All registries
n
Time trends analyses (period 1978–1997)
Selected registries
All registries
Selected registries
% nonmalignant
% PY
n
% nonmalignant
n
% nonmalignant
% PY
n
% nonmalignant
3.0
3.5
2.7
4.4
1.8
2.9
86
90
94
100
63
84
46,033
10,595
7289
4283
3505
20,361
3.4
3.9
2.8
4.4
2.7
3.3
77,111
23,548
8974
8321
5026
26,411
3.3
3.5
2.7
3.9
3.0
3.1
92
91
89
100
77
93
70,855
21,112
8164
8321
3876
24,551
3.5
3.9
2.9
3.9
3.8
3.2
Adolescents (age 15–19 years)
Europe
8272
1.9
British Isles
968
1.3
East
935
2.2
North
2057
3.7
South
2153
1.7
West
2159
0.4
48
36
81
100
37
9
4097
346
751
2057
755
188
3.6
3.8
2.8
3.7
4.1
2.1
11,889
1281
1611
3973
1511
1251
2.6
0.0
2.0
3.7
1.7
1.0
74
0
76
100
64
26
8698
0
1281
3973
850
332
3.4
–
2.5
3.7
2.7
1.8
Children (age 0–14 years)
Europe
53,717
British Isles
11,837
East
7718
North
4283
South
5534
West
24,345
Selected registries are those with systematic registration of non-malignant tumours (Table 2). n, total number of registrations, %PY, percent
person-years covered by the selected registries.
EUROPEAN JOURNAL OF CANCER
1937
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
Table 8 – Number (n) and incidence rate of the malignant and non-malignant tumours in the European regions in children
and adolescents incident in the period 1988–1997 (Source: ACCIS)
Malignant tumours
All registries Selected registries
n
Rate
Children (age 0–14 years)
Europe
52,109 129.7
British Isles 11,419 123.4
East
7511 132.7
North
4094 148.8
South
5435 138.2
West
23,650 127.3
Adolescents (age 15–19 years)
Europe
8117 182.5
British Isles
955 182.3
East
914 164.8
North
1980 201.8
South
2117 183.5
West
2151 174.2
Non-malignant tumours
Comparison
n
Rate
Rate ratio
44,453
10,177
7084
4094
3411
19,687
128.7
122.7
132.7
148.8
138.7
125.4
0.99
0.99
1.00
1
1.00
0.99
3951
333
730
1980
724
184
183.9
177.2
163.3
201.8
171.7
166.8
1.01
0.97
0.99
1
0.94
0.96
All registries Selected registries
95% CI
n
Comparison
Rate
n
Rate
Rate ratio
0.98–1.00% 1,608
0.96–1.03%
418
0.95–1.05%
207
–
189
0.90–1.12%
99
0.97–1.00%
695
4.0
4.5
3.7
6.9
2.5
3.7
1580
418
205
189
94
674
4.6
5.0
3.8
6.9
3.8
4.3
1.14
1.12
1.05
1
1.52
1.15
1.13–1.15%
1.08–1.15%
1.00–1.11%
–
1.35–1.71%
1.13–1.17%
1.00–1.01%
0.95–1.00%
0.99–0.99%
–
0.76–1.15%
0.91–1.00%
3.5
2.5
3.8
7.9
3.1
0.7
146
13
21
77
31
4
6.8
6.9
4.7
7.9
7.4
3.6
1.95
2.79
1.24
1
2.36
5.60
0.52–7.36%
0.17–45.20%
1.09–1.41%
–
0.28–19.80%
0.08–417.27%
155
13
21
77
36
8
95% CI
Selected registries are those with systematic registration of non-malignant tumours (Table 2). Rate is age-standardised (World standard)
incidence rate for children and age-specific rate for adolescents. Rate ratio is standardised incidence ratio for children and simple rate ratio for
adolescents.
95% CI = (69, 75) in the complete data set and 63%, 95%
CI = (57, 69) in the restricted dataset, reflecting the absence
of Scotland in the latter. The next largest difference concerned the survival of children with non-malignant tumours
in the South, with the corresponding survival of 89%, 95%
CI = (68, 96) in the complete, compared with 92%, 95%
CI = (68, 98) in the restricted dataset.
3.6.
Influence of the registration of multiple primary
tumours
A half percentage of the tumours in children and about 1% in
adolescents were multiple primaries: second, third, fourth or
fifth malignancies in a single individual. This percentage varied slightly between the regions and with age at diagnosis (Table 10). Information on second and higher primary tumours
was not provided from the large paediatric registry of England
and Wales (Table 1), which represented a substantial component of the British Isles. Assuming the odds of multiple to first
primaries in each age group being the same as in the other
European regions, the ASR of the multiple primaries for British Isles in 1988–1997 would be 0.76 per million and the corresponding European ASR would be 0.81 per million: only
slightly higher than those shown in Table 10. In the 1988–
1997 dataset, multiple primaries accounted for less than 1
per million in children and about 3 per million in adolescents
(Table 10). The highest incidence rates of second and higher
primary tumours were observed in North. Incidence of multiple tumours increased between the first and the last 5-year
period almost five-fold both in children and adolescents (Table 10). However, the incidence rates of first primary tumours
were also increasing significantly (Table 10).
Multiple primary tumours were included in survival analyses, but their inclusion did not affect survival markedly, as
shown in Fig. 5, because their proportion was very small.
For example, 5-year survival of 49,651 European children inci-
dent in 1988–1997 was 71.89%, 95% CI = (71.56, 72.36), while 5year survival of 49,407 children with first primaries only was
72.12%, 95% CI = (71.70, 72.53). In the North and West, the regions with highest proportion of multiple primaries, the 5year survival differed by 0.24 and 0.27 percentage points,
respectively. Nevertheless, survival of children with multiple
primaries was considerably lower than that of children with
first primary tumours. This disparity was observed in each
period except the first (Fig. 5(a)) and in each region except
the British Isles (Fig. 5(b)), due to a deficiency of multiple
primaries. While survival of children with first primaries
has improved significantly over the 20 years (P < 0.0001),
survival of children with multiple primaries did not change
(P = 0.76).
The most common groups of tumours occurring as second
or higher primary tumours were the CNS tumours (26%), followed by leukaemias (20%), carcinomas (15%) and lymphomas (13%). At the level of ICCC subgroups, more than 10%
each were contributed by ANLL and astrocytoma. Since the
reporting of multiple primary tumours was incomplete and
non-comparable within ACCIS, further detail is not presented.
3.7.
Scale of registration of carcinoid of appendix as
malignant tumour
An additional potential source of artefact is the recording of
carcinoid of the appendix as a malignant tumour by some registries. There were 341 ‘malignant’ cases of carcinoid of the
appendix in the data used for analyses, none occurring before
the age of 5 years and the majority occurring in adolescents.
Since only 5 carcinoids of the appendix were reported from
two paediatric cancer registries, the results reported below
are based on the dataset composed of the general cancer registries only. Incidence rates of this tumour increased with age
and with period of diagnosis (Fig. 6(a)). Differences were observed between the regions, with especially high incidence
1938
EUROPEAN JOURNAL OF CANCER
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
Malignant
Non-malignant
1000
ASR per million
a
100
10
1
1978
1983
1988
1993
1998
Calendar year
British Isles
British Isles
Malignant:
East
West
North
c
180
18
160
16
140
14
ASR per million
ASR per million
South
East
West
North
b
Non-malignant:
South
120
100
80
60
12
10
8
6
40
4
20
2
0
0
1978
1983
1988
1993
1978
1998
1983
Calendar year
1988
1993
1998
Calendar year
Fig. 3 – Age-standardised incidence rates (ASR, world standard) of malignant and non-malignant tumours in European
children (age 0–14 years) over time. Based on all registries included in the time trends dataset (1978–1997), n = 77,111. Europe
includes data contributed by the former GDR, which are not included in any of the shown European regions. (a) Europe,
(b) malignant tumours, (c) non-malignant tumours. Source: ACCIS.
Table 9 – Number (n) and average annual percent change (AAPC) in incidence rate of the malignant and non-malignant
tumours in all cancer registries included in the time trends analyses in Europe, 1978–1997 (Source: ACCIS)
All registries
Selected registries
n
AAPC
P
n
AAPC
P
74,592
2519
1.10
1.66
<0.0001
<0.0001
68,363
2492
1.30
1.71
<0.0001
<0.0001
Adolescents (age 15–19 years)
Malignant
11,584
Non-malignant
305
2.00
2.02
<0.0001
0.087
8402
296
2.03
2.24
<0.0001
0.062
Children (age 0–14 years)
Malignant
Non-malignant
AAPC was calculated from Poisson regression model with year as explanatory variable, adjusted for sex, region and, for children only, age
group.
EUROPEAN JOURNAL OF CANCER
Five-year survival (%)
a
Malignant:
All registries
Selected registries
N=48,276
N=41,644
100
90
80
70
60
50
40
30
20
10
0
EUROPE British Isles
East
North
South
West
Region of residence at diagnosis
Five-year survival (%)
b
Non-malignant:
All registries
Selected registries
N=1,375
N=1,352
100
90
80
70
60
50
40
30
20
10
0
EUROPE
British Isles
East
North
South
West
Region of residence at diagnosis
Fig. 4 – Five-year survival of children diagnosed in Europe in
1988–1997 with (a) malignant tumours and (b) nonmalignant tumours. Data provided by all registries are
compared with those with systematic registrations of nonmailgnant tumours (Selected registries). N, number of cases
included in analysis. 95% confidence intervals are shown as
line sections. Source: ACCIS.
1939
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
rates in the North and West (Fig. 6(b)). The overall age-specific
rates of cancer in adolescents reported from these two regions
were therefore inflated by about 3 per million in the North and
8 per million in the West. The effect on the age-standardised
incidence rate in the age-range 0–14 years in the West is diluted by the contribution of the large German registry and
other paediatric cancer registries, with no or exceptional registrations of ‘malignant’ carcinoid of the appendix. In the
North, the inclusion of the carcinoid of the appendix diagnosed before the age of 15 years increased the rates by at most
one per million, that is from 162.8 to 163.9 in 1993–1997, the
period with highest incidence of these tumours.
Survival of patients with carcinoid of the appendix is
excellent: only 2 of all 243 patients followed up have died,
possibly of other causes. The variable proportion of these tumours in the region-specific datasets was therefore thought
to contribute to differences in survival by time period or region. However, in adolescents of the North the overall 5-year
observed survival (OS) was actually slightly higher for the
complete cohort (OS = 78.6%, 95% CI = (77.0, 80.0), n = 3085)
than for the cohort free from these cases (OS = 78.3%, 95%
CI = (76.7, 79.8), n = 3040). The corresponding figures in the
West for the total dataset of 660 cases were OS = 74.2%, 95%
CI = (70.3, 77.6) and for the dataset free of carcinoid of the
appendix (n = 647) it was OS = 73.6% 95% CI = (69.7, 77.1). This
is because in the smaller cohorts there were fewer patients at
risk of dying for the same number of deaths. Unequal presence of carcinoid of the appendix in different regions did
not enhance regional differences in survival.
3.8.
Systematic differences between paediatric and general
cancer registries
Paediatric cancer registries (Table 2) contributed a substantial
proportion of person-years during the most recent 10-year period 1988–1997, as seen in Fig. 7. However, their representation
Table 10 – Numbers (n) and incidence rates by tumour multiplicity in Europe (Source: ACCIS)
Children (age 0–14 years)
First primary
1988–1997
1978–1982
1983–1987
1988–1992
1993–1997
AAPC (1978–1997)
P
Europe
British Isles
East
North
South
West
Adolescents (age 15–19 years)
Multiple primaries
First primary
n
ASR
% of total
n
ASR
n
53,447
11,831
7704
4247
5512
24,153
137.8
131.1
140.7
158.9
148.0
134.9
0.5
0.1
0.2
0.8
0.4
0.8
270
6
14
36
22
192
0.65
0.06
0.22
1.27
0.46
1.01
8,192
961
929
2022
2145
2135
128.2
183.4
167.5
206.1
185.9
172.9
13,885
119.3
20,838
126.9
21,812
133.8
20,277
140.1
1.1061
<0.0001
0.2
0.2
0.4
0.6
22
52
94
131
4.2478
<0.0001
0.18
0.30
0.53
0.81
3167
3510
2508
2609
153.6
165.8
176.1
197.0
1.9694
<0.0001
Multiple primaries
Age-specific rate % of total
n
Age-specific rate
1.0
0.7
0.6
1.7
0.4
1.1
80
7
6
35
8
24
1.80
1.34
1.08
3.57
0.69
1.94
0.3
0.6
1.3
1.2
10
20
33
32
0.49
0.95
2.32
2.42
2.087
0.352
Multiple primaries are second, or higher order tumours occurring in a single individual. They are underrepresented in ACCIS for children of the
British Isles (see text). ASR, age-standardised rate (world standard); AAPC, average annual percent change in incidence rate, derived from a
Poisson regression model with year as explanatory variable, adjusted for sex, region and, for children only, age group.
1940
Total
First primaries
Multiple primaries
N=72,398
N=72,135
N=255
Five-year survival (%)
80
60
40
20
1978-82
1983-87
1988-92
1993-97
Period of diagnosis
b
7
6
5
4
3
2
1
Total
First primaries
Multiple primaries
N=49,651
N=49,407
N=238
0
1978-82
1983-87
1988-92
1993-97
Calendar period
80
Region:
60
b
40
20
British Isles
East
North
South
West
Region of residence at diagnosis
Fig. 5 – Five-year survival of children (age 0–14 years) in
Europe according to tumour multiplicity. (a) Europe, dataset
1978–1997; (b) Europe, dataset 1988–1997. Multiple
primaries are second, or subsequent tumours occurring in
a single individual. They are underrepresented in ACCIS for
children of British Isles (see text). N, number of cases
included in analyses. 95% confidence intervals are shown as
line sections. Source: ACCIS.
differed greatly by region, and in the North all the contributing cancer registries were general. Incidence rates recorded by
the two types of registries within the age-range 0–14 years are
compared in Fig. 8. While the age-specific incidence rates of
all specified tumours are very similar, at least up to the age
of 11 years, there is a constant excess of unspecified tumours
recorded by the general cancer registries. The overall agestandardised rate (ASR) was higher in general than in the paediatric registries; but this was largely due to the higher rates
of unspecified tumour types (Table 11). This pattern was also
observed when the North (with no paediatric cancer registry)
was excluded from the dataset of the general cancer registries
(Table 11). Overall survival was slightly higher in paediatric
than in general cancer registries. Survival of children with
unspecified tumour types was markedly inferior to survival
of children with specified tumour types and this was observed
to a larger extent in the paediatric than in the general cancer
registries. However, since the proportion of unspecified tumours in the paediatric datasets was about half of that in
the general ones, the overall survival was higher in paediatric
Incidence rate per million
Five-year survival (%)
5-9
10-14
8
100
0
15-19
Age group:
a
100
0
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
Incidence rate per million
a
EUROPEAN JOURNAL OF CANCER
British Isles
East
North
South
West
8
7
6
5
4
3
2
1
0
5-9
10-14
15-19
Age group (years)
Fig. 6 – Age-specific incidence rates of carcinoid of appendix
(M8240, C181) in children and adolescents in Europe.
(a) Dataset 1978–1997 (number of cases = 215), (b) dataset
1988–1997 (number of cases = 171). The total Europe
includes data contributed by the former GDR for the period
1978–1987, which are not included in any of the shown
European regions. Source: ACCIS.
than in the general cancer registries (Table 11). With the
North excluded, survival of patients with unspecified tumours was similar in general and paediatric cancer registries
(Table 11): which implies that patients with ‘unspecified’ tumours have better prognosis in the North than in the other
regions.
3.9.
Systematic differences between national and regional
cancer registries
Despite the large number of regional cancer registries contributing to the ACCIS database (44 of 62), cancer registries cover-
EUROPEAN JOURNAL OF CANCER
General
Registry type:
Paediatric
Person-years (millions)
200
150
100
50
0
British Isles
East
North
South
West
84%
72%
0%
38%
79%
Region
Fig. 7 – Distribution of person-years at risk for the period
1988–1997 in the age-range 0–14 years by the five European
regions, according to the type of reporting cancer registry.
Percentages show coverage by the paediatric cancer
registries in each region. Source: ACCIS.
General, specified
General, NOS
Registry type
and tumour type:
Paediatric, specified
Paediatric, NOS
Incidence rate per million
250
200
150
100
50
0
0
1
2
3
4
5
6
7
8
9 10 11 12 13 14
Age (years)
Fig. 8 – Age-specific incidence rates of childhood cancer in
Europe 1988–1997, as registered by the paediatric and
general cancer registries (number of cases = 55,227). The
following tumour groups were included among unspecified
(NOS): the ICCC categories Ie, IIe, IIIf, VIc, VIIc, VIIIe, IXe,
XIIb, the morphology codes M-8000 to M-8004 in Xe and
topography codes C76 to C80 in XIf. Source: ACCIS.
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
1941
ing a whole country contributed a large proportion of personyears (54%). Contrary to the information shown in Table 2, all
regional registries covering jointly the whole country were
considered as being national in this comparison. All regional
cancer registries are thus found in only two of the five regions, South and West (Fig. 9). As shown in Table 12, the incidence rates reported by the regional cancer registries were
significantly higher than those reported by the national cancer registries, irrespective of the tumours specificity (the difference was larger for unspecified tumours). The difference
persisted virtually unchanged in all ages, with additional excess in infants (Fig. 10). This was due mainly to more numerous lymphomas (O/E = 29/15, OR = 2.4, 95% CI = (1.9, 7.1)) and
neuroblastoma (O/E = 350/296, OR = 1.3, 95% CI = (1.2, 4.0))
among the specified tumour types. Part of this discrepancy
in age 0 may be due to earlier diagnosis (or reporting) in the
areas covered by regional registries, since the excess incidence is not observed at age 1 year (Fig. 10).
Survival of children with unspecified tumour types was
similar whether they were followed-up by a national or regional cancer registry, while the pattern was more complicated
for the specified tumours. Within Europe as a whole, slightly
higher survival was observed in the regional cancer registries.
However, this was lower than the survival observed in the national registries of the two combined regions, South and West
(including Germany, Malta, Netherlands and Slovenia), by
some three percentage points. Of note is the lower completeness of the follow-up in the dataset for the combined region
(Table 12).
4.
Discussion
Sixty-two population-based cancer registries contributed to
the analyses. They differed by size, geographical location,
administrative area, age-range and other criteria for registration, time period available, methods of registration and follow-up of the patients, data management systems,
availability of data sources, data protection legislation, etc.
These different settings contributed to the complexity of the
ACCIS database, its exploitation and interpretation of results.
We have chosen to include a maximum of cancer registries, to
reach a maximum sample size and preserve the heterogeneity within each analysis. The possibility that some of the differences in incidence and survival represent artefacts, caused
by the registration process, must therefore be considered.
Only the datasets judged to be of high quality and completeness were selected for the analyses, with a considerable
loss of the person-years available within ACCIS. A high percentage of microscopically verified cases indicated high precision of diagnoses. Cases detected through death certificate
only (DCO cases) were rare in the registries with access to
information that allows registration of DCO cases. While lack
of access to national mortality database may seriously alter
incidence rates in patients of all ages due to missed DCO
cases, it would cause much less of a problem in childhood
populations, for several reasons. First, the probability of cancer remaining undiagnosed in a child is much lower than in
an elderly person. Secondly, due to the rarity of childhood
cancers, diagnostic and treatment centres for children with
cancer are organised into networks and usually work with
1942
EUROPEAN JOURNAL OF CANCER
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
Table 11 – Numbers of cases, data quality indicators, incidence rates and population-based survival as recorded by general
and paediatric cancer registries in Europe for the period 1988–1997 (Source: ACCIS)
All tumours
Specified tumours
Unspecified tumours
General
G (excl.
North)
Paediatric
General
G (excl.
North)
Paediatric
General
G (excl.
North)
Paediatric
Incidence
n
MV (%)
DCO (%)
Unknown (%)
ASR
95% CI(ASR)
SRR
95% CI(SRR)
17,640
94
0.3
0.4
144.9
(142.5, 147.2)
1.06
(1.04, 1.08)
13,357
94
0.3
0.3
140.4
(137.4, 143.4)
1.03
(1.01, 1.05)
37,587
96
0.3
1.4
136.5
(135.5, 137.4)
16,441
97
0.2
0.2
135.0
(132.8, 137.2)
1.02
(0.999, 1.04)
12,558
97
0.2
0.2
132.0
(129.2, 134.8)
1.00
(0.97, 1.02)
36,566
97
0.2
1.0
132.8
(131.8, 133.8)
1,199
53
2.2
3.2
9.8
(9.7, 10)
2.68
(2.61, 2.75)
799
49
3.4
2.4
8.4
(8.2, 8.6)
2.29
(2.22, 2.36)
1,021
56
3.3
14.9
3.7
(3.6, 3.7)
Survival
n
Follow-up (%)
w1–5 (%)
w1–10 (%)
5+years (%)
Median (years)
5-year OS
95% CI (OS)
v2
15,003
85
31
66
57
5.6
71
(71, 72)
7.8
10,785
81
31
65
56
5.5
69
(68, 70)
52.4
36,833
98
26
63
65
5.9
72
(72, 73)
13,980
85
31
66
57
5.6
72
(71, 73)
3.3
10,146
81
31
66
56
5.5
70
(69, 71)
34.1
35,939
98
26
63
64
5.9
73
(72, 73)
1,023
85
28
56
54
5.3
60
(57, 63)
9.1
639
80
24
50
56
5.4
53
(48, 57)
0.15
894
88
15
45
73
6.2
53
(50, 57)
Unspecified tumours: includes ICCC categories Ie, IIe, IIIf, VIc, VIIc, VIIIe, IXe, XIIb, the morphology codes M-8000 to M-8004 in Xe and
topography codes C76 to C80 in XIf.
G (excl. North), general cancer registries outside Northern region; n, number of cases included in analyses; MV (%), percentage of cases with
microscopic verification of diagnosis; DCO (%), percentage of cases registered from death certificate only; Unknown (%); percentage of cases
with unknown basis of diagnosis; ASR, age standardised rate (World standard); SRR, standardised rate ratio comparing ASR in general with ASR
in paediatric cancer registries; 95% CI, 95% confidence interval; Follow-up (%), percentage of cases included in survival analyses of the total
registered; w1–5 (%), percentage of cases withdrawn from survival analysis during the first 5 years of follow-up; w1–10 (%), percentage of cases
withdrawn from survival analysis during the first 10 years of follow-up; 5+ years (%), percentage of cases followed-up for at least 5 years in the
registries with follow-up, among those who did not deceased by the closing date of the study; median (years), median follow-up time in years
of the cases included in survival analyses, among those who did not deceased by the closing date of the study; 5-year OS, 5-year observed
survival (cumulative actuarial survival probability at 5 years since diagnosis); v2, statistic of the log-rank test of equality of survival in general
and paediatric cancer registries; P, associated P-value.
Coverage:
Regional
National
Person-years (millions)
200
150
100
50
0
British Isles
100%
East
100%
North
100%
South
10%
West
78%
Region
Fig. 9 – Distribution of person-years at risk contributed for
the period 1988–1997 and the age-range 0–14 years by the
five European regions, according to the country coverage of
reporting cancer registry. Percentages show coverage by the
national cancer registries in each region. Source: ACCIS.
cancer registries. Thirdly, favourable prognosis reduces the
proportion of potential deaths in the childhood population.
On the other hand, late deaths do occur in childhood cancer
cases and access to the relevant information may increase
incidence rates slightly and improve the completeness of
the follow-up. Access to nominative death certificates, centralised nationally, would help to improve data quality and
completeness and their evaluation. Alternatively, anonymous
death records may be used to compare incidence, survival
and mortality from cancer in the relevant childhood population and possibly to estimate the extent of potential loss of
cases. In our database, the decreasing proportion of DCO
cases during the 1980s may be explained by a reduced proportion of registries having access to mortality databases, by improved ascertainment of cases from primary sources, or
possibly by more efficient retrospective finding of cases first
notified via a death certificate (DCN)l. It was not possible to
evaluate the proportion of DCN cases in the present data.
The proportion of cases with unknown source of diagnosis
l
These are the cases first notified from death certificate, which
are then traced back to find further information of the case in
medical records.
EUROPEAN JOURNAL OF CANCER
1943
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
Table 12 – Numbers of cases, data quality indicators, incidence rates and population-based survival as recorded by national
and regional cancer registries in Europe for the period 1988–1997, with the indicators of data quality (Source: ACCIS)
All tumours
Specified tumours
Unspecified tumours
National
National
(S&W)
Regional
National
National
(S&W)
Regional
National
National
(S&W)
Regional
Incidence
n
MV (%)
DCO (%)
Unknown (%)
ASR
95% CI(ASR)
SRR
95% CI(SRR)
43,690
96
0.3
1.1
137.3
(136.4, 138.2)
0.94
(0.92, 0.97)
19,852
99
0.0
0.0
135.4
(134.5, 136.2)
0.93
(0.9, 0.95)
11,537
94
0.3
0.8
145.9
(142.2, 149.7)
41,981
97
0.2
0.8
132.0
(131.1, 132.8)
0.94
(0.92, 0.97)
19,381
99
0.0
0.0
132.2
(129.3, 135.1)
0.95
(0.91, 0.98)
11,026
95
0.2
0.6
139.7
(136.1, 143.3)
1709
54
3.2
9.5
5.4
(5.3, 5.4)
0.86
(0.83, 0.89)
471
21
0.0
0.0
3.2
(3.2, 3.2)
0.51
(0.49, 0.53)
511
57
2.2
5.3
6.3
(6.1, 6.5)
Survival
n
Follow-up (%)
5+yrs (%)
w1–5 (%)
w1–10 (%)
Median (years)
5-year OS
95% CI (OS)
v2
P
40,852
94
63
27
63
5.9
72
(71, 72)
4.1
0.0429
17,422
88
48
41
75
4.5
76
(75, 76)
25.71
<0.0001
10,984
95
60
29
67
5.6
73
(72, 74)
39,374
94
63
27
63
5.9
72
(72, 73)
6.09
0.0136
17,052
88
47
41
75
4.5
76
(75, 77)
19.84
<0.0001
10,545
96
60
29
68
5.6
73
(72, 74)
1478
70
63
26
49
5.7
58
(55, 60)
1.01
0.3145
370
22
58
24
54
4.4
53
(47, 58)
0.02
0.8776
439
13
63
26
56
5.9
55
(50, 60)
Unspecified tumours, includes ICCC categories Ie, IIe, IIIf, VIc, VIIc, VIIIe, IXe, XIIb, the morphology codes M-8000 to M-8004 in Xe and
topography codes C76 to C80 in XIf. National (S& W), national cancer registries of the South and West regions only; n, number of cases included
in analyses; MV (%), percentage of cases with microscopic verification of diagnosis; DCO (%), percentage of cases registered from death
certificate only; Unknown (%), percentage of cases with unknown basis of diagnosis; ASR, age standardised rate (World standard); SRR,
standardised rate ratio comparing ASR in national with ASR in regional cancer registries; 95% CI, 95% confidence interval; Follow-up (%),
percentage of cases included in survival analyses of the total registered; w1–5 (%), percentage of cases withdrawn from survival analysis during
the first 5 years of follow-up; w1–10 (%), percentage of cases withdrawn from survival analysis during the first 10 years of follow-up; 5+ years
(%), percentage of cases followed-up for at least 5 years in the registries with follow-up, among those who did not deceased by the closing date
of the study; Median (years), median follow-up time in years of the cases included in survival analyses, among those who did not deceased by
the closing date of the study; 5-year OS, 5-year observed survival (cumulative actuarial survival probability); v2, statistic of the log-rank test of
equality of survivor curves in national versus regional cancer registries; P, associated P-value.
describes the level of difficulties in finding information about
eligible cases, but more detailed enquiry in each registry
would be necessary to understand why the source of information was not identified. In any case, the values of indicators
derived from the variable ‘basis of diagnosis’ do not suggest
that variations in quality and comparability of data between
the regions and over time contributed significantly to observed differences in incidence or survival.
In the future, registration completeness should be evaluated more formally, for example by reporting number of
sources per case,19 if this information is available in the collaborating registries and by reporting DCN cases. Registration
flow30 might be available from those registries, which distinguish date of incidence and date of registration and can determine for their area the probability that cancer is mentioned
on the death certificates of cancer patients who die. Other
methods of evaluation of completeness, such as capture-recapture31 may be applicable at registry level and evaluated
qualitatively within ACCIS. Mortality to incidence ratio (M/I
ratio),19 the usual measure of registration completeness is difficult to apply to childhood cancer data, because the cause of
death data may not be specific enough and mortality is postponed markedly to older ages.
Multiple primaries were incompletely represented within
ACCIS, but their variable registration did not materially influence the comparison between the regions and periods. More
emphasis will be devoted to this topic after the completion
of records.
Differences in coding of tumours resulting in artificial differences for some diagnostic subgroups were of special concern in the North [Stiller, Marcos-Gragera, Ardanaz and
colleagues; and Stiller, Desandes, Danon and colleagues, this
issue]. These differences persist despite some 15 years of coordination and common recommendations for data definition,
classification, coding and management, within Europe.11 This
study shows in particular the need for homogenous coding of
carcinoid of the appendix. Adherence to the latest edition of
ICD-O across the registries would facilitate international comparison, which in turn helps further data standardisation.
We have shown that the increase of incidence previously
reported from the ACCIS study7 is unlikely to be due to different registries contributing to different periods. Neither can
the inclusion of non-malignant tumours and multiple primaries explain the observed increase, since we have seen that
the incidence increased significantly irrespective of behaviour
or multiplicity of tumour. The incidence trends are further
1944
EUROPEAN JOURNAL OF CANCER
Regional, specified
Registration coverage
and tumour type:
Regional, NOS
National, specified
National, NOS
250
Incidence rate per million
200
150
100
50
0
0
1
2
3
4
5
6
7
8
9 10 11 12 13 14
Age (years)
Fig. 10 – Age-specific incidence rates of childhood cancer in
Europe 1988–1997, as registered by the national and
regional cancer registries (number of cases = 55,227). The
following tumour groups were included among unspecified
(NOS): the ICCC categories Ie, IIe, IIIf, VIc, VIIc, VIIIe, IXe,
XIIb, the morphology codes M-8000 to M-8004 in Xe and
topography codes C76 to C80 in XIf. Source: ACCIS.
discussed elsewhere7,32 [Kaatsch and colleagues; and Stiller,
Desandes, Danon and colleagues, this issue].
Some of the differences in observed survival between the
regions (and over time) might be explained by unequal completeness of follow-up in different data sets, since incomplete
follow-up may bias survival estimates.33 Therefore, proportion of cases lost to follow-up (or, more generally, withdrawals) should be kept to a minimum. Losses to follow-up are
especially of concern in the registries without an access to
official (national) mortality data. However, in our data the follow-up indicators did not differ systematically between national and regional cancer registries. Rather, the difference
depended on geographical region, with access restricted more
in South and West than in other regions. Paediatric cancer
registries, which have to invest extra efforts to ensure follow-up of their cases after they quit the childhood age-range,
had actually a slightly more complete follow-up than the general cancer registries (although this advantage is likely to
diminish with lengthening of follow-up).
We have identified systematic differences between the
data provided by general and paediatric cancer registries.
The observed incidence patterns suggest a more careful
recording of tumours in the paediatric registries, accompanied perhaps by a lack of access to some data generated outside the paediatric healthcare structures, especially for older
children. Simultaneously, the higher rates of unspecified
tumours observed in general cancer registries may reflect
either more complete registration or less efficient checking
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
procedures (due to large total number of cases) and higher
probability of random errors, which may go unnoticed in
childhood age-range. Overall, however, the observed differences do not preclude pooling the data, and are rather beneficial for overall comparison, at least until the potential
problems could be identified.
Our attempt to compare quality of data provided by national as opposed to regional cancer registries was only partially fulfilled, since the two categories also differed in other
aspects (geographic coverage, type of registry, etc.). In any
case, the observed patterns, notably higher incidence rates
in regional than national cancer registries document the need
for further standardisation of cancer registration techniques.
Geographical grouping of countries was justified by a combination of factors, such as previous observations of similar
incidence and survival patterns3,5 and balanced numerical
representation of each region in analyses. Although there
are differences in incidence and survival between countries
within each region [Stiller, Marcos-Gragera, Ardanaz and colleagues; Sankila and colleagues; Stiller, Desandes, Danon and
colleagues; and Pritchard-Jones and colleagues, this issue],
the purpose of this study was, by comparing large geographical areas, to identify consistent patterns that might not have
been obvious when comparing smaller geographical areas.
The inter-regional comparison is certainly influenced by the
definition of the geographical regions, which is somewhat
arbitrary and susceptible to future modification, depending
on data availability.
4.1.
The British Isles
Two countries contributed data for the British Isles (UK and
Ireland), but these provided a considerable proportion of person-years for the childhood age-range. The large specialised
registry of England and Wales covered the childhood population for the entire study period (except the last 2 years), and
so had a major influence on the results for the region as a
whole. The incidence rates were at the lower end of the scale
and would have been little higher (possibly up to 1 per million)
had multiple primary tumours been included for England and
Wales. The low incidence might be explained by efficient
exclusion of non-eligible cases, e.g. non-malignant, nonchildhood, unconfirmed, etc., considering the outstanding tradition of the UK’s Childhood Cancer Research Group (CCRG), to
which the cancer registry is affiliated.35 Notification of cases is
voluntary, so under-reporting cannot be excluded, although it
is unlikely. The childhood cancer registry of England and
Wales receives data from general registries, both regional
and national, including the general registry of Scotland.36 A
link with the databases of clinical trials is also maintained.36
This cancer registry was used as a gold standard in a study
of completeness of childhood cancer registration and recorded 109% of the registrations made by the general cancer
registries across the country during the period 1971–1984.37
About 6% of the British population was represented by the
ethnic minorities,36 the majority being of south Asian or African origin, who typically have low incidence rates of childhood cancer.38 However, it is unlikely that this small
proportion of population would influence the overall rates
markedly.
EUROPEAN JOURNAL OF CANCER
The high quality of follow-up data in the CCRG and Scotland39 and the information available in ACCIS may be partly
responsible for the average level of observed survival in the
British Isles in children [Sankila and colleagues, this issue].
The complete follow-up in the British Isles may theoretically
result in lower estimates of survival than in the other regions
with a less complete follow-up, as shown in the above example in Table 6. However, for the earlier 5-year diagnostic periods, follow-up was fairly complete in all regions and the rank
for the British Isles was the same [Magnani and colleagues,
this issue]. This would suggest that the censored observations
and the withdrawals in the other regions would mostly comprise patients with long-term survival, as seen recently for
the German childhood cancer registry.34
The dataset used to describe the adolescent population of
the British Isles was represented by Ireland, Northern Ireland
and Scotland and excluded England and Wales, for which data
on adolescents were not available. The incidence rates attained average levels [Stiller, Desandes, Danon and colleagues,
this issue] and would have been slightly higher, should all the
registries include non-malignant tumours. The follow-up of
the adolescents incident in 1988–1997 in the British Isles was
markedly lower than that for children, especially in the Irish
registries. The level of bias due to incomplete follow-up is
therefore comparable with other regions.
4.2.
The East
The cancer registries grouped in the East region were all national, relying either upon compulsory notification (in the
general registries of Estonia and Slovakia) or the largely active
ascertainment of cases (in the paediatric cancer registries of
Belarus and Hungary). Several other registries, operating in
central and Eastern Europe were excluded from analyses because of lack of data comparability. The cancer registry of
the former German Democratic Republic could have also been
included in the East, had the data from the same registration
area been available until the end of the study period. It will
hopefully be possible in the future to study the impact of
the socio-economic changes on cancer burden in the young
population of this part of Germany, although dilution of such
effects would be expected due to migration of the population
between the two parts of Germany since the 1990s.
The East seemed to be the most perilous region, in terms of
relatively high incidence rates [Stiller, Marcos-Gragera, Ardanaz
and colleagues, this issue] and low survival [Sankila and colleagues, this issue]. The high incidence rates can be explained
entirely by a rapid rise in the incidence of thyroid cancer in
Belarus due to the Chernobyl accident [Steliarova-Foucher,
Stiller, Pukkala and colleagues, this issue]. Exclusion of the
thyroid cancer cases of Belarus from the East would result in
an ASR of 132.4 (instead of 140.9), ranking the East second
lowest, just above the British Isles. Incidence rate of childhood
cancer in the East is low compared with the high incidence in
adults, especially in males.1 Under-registration is therefore
unlikely, unless limited specifically to childhood ages.
Low survival in the East [Sankila and colleagues, this issue]
is the likely consequence of specific socio-economic setting,
as discussed in detail elsewhere [Pritchard-Jones and colleagues, this issue], since completeness of follow-up is similar
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
1945
to that of other regions. The recent political changes in this
part of Europe will undoubtedly contribute to changes in the
patterns of incidence40 and survival, as can be judged from
the differences in survival already seen between the countries
grouped within this region [Pritchard-Jones and colleagues,
this issue]. The first prerequisite for future comparative studies is the access to data. This seems to be threatened due to
the new data protection constraints currently replacing the
previous legislative vacuum in the countries grouped within
the East.41 Such restrictions will lead at best to underestimation of incidence rates, at worst to no data at all.
4.3.
The North
The North was the most homogeneous region, with 100% national coverage by general cancer registries and coordinated
registration practices.42 As a result, the representation of the
Northern region was almost unchanged in the various analyses (geographical, time trends, children, adolescents, survival,
etc.). In addition, the cancer registries in Nordic countries have
an access to other national databases (population registers),
which is especially valuable for follow-up of cancer patients,
using linkage via the unique personal identification number.
The Northern region was also outstanding in having the
highest incidence rates [Stiller, Marcos-Gragera, Ardanaz
and colleagues; and Stiller, Desandes, Danon and colleagues,
this issue] and the most favourable survival [Sankila and colleagues; and Stiller, Desandes, Danon and colleagues, this issue]. These patterns may be explained partly by differences in
diagnostic and registration routines, documented by the highest percentage of non-malignant tumours and multiple
primaries, as well as the large number of the carcinoids of
appendix (up to 3 per million in adolescents). Misleading results may have been obtained for some tumour groups due
to the use of outdated classification systems, during the period under study. For example, the low rate of neuroblastoma
[Spix et al., this issue] and high rate of soft tissue sarcomas
[Pastore, Peris-Bonet, Carli and colleagues, this issue] are difficult to explain otherwise than by classification mismatch
[Stiller, Marcos-Gragera, Ardanaz and colleagues, this issue].
Finally, the high proportion of unspecified tumours can result
from high level of automation and possibly a lack of validation of individual records for children. The relatively high survival of patients with unspecified tumours in the North may
suggest that this group may include some tumour types that
would be classified among specified in other regions.
Case mix of the total incidence data may partly explain the
high survival, especially if a larger proportion of tumours with
a favourable prognosis were included in the analyses for the
North than for the other regions. The Nordic cancer registration system based on fine cross-linkage of various registries
may be especially prone to include all non-fatal tumours,
which are not notified (or not subject to notification) in other
regions. This system also ensures high completeness of
follow-up. Coordinated management of childhood cancer
patients through the Nordic Society of Paediatric Haematology and Oncology (NOPHO), as well as the high socio-economic level of the Scandinavian countries certainly also
contribute to the superior outcome in the North [Sankila
and colleagues, this issue].
1946
4.4.
EUROPEAN JOURNAL OF CANCER
The South
The South was the most heterogeneous of all regions,
including national, regional, general and paediatric cancer
registries in the Mediterranean area (Italy, Malta, Slovenia,
Spain and, for incidence only, Turkey). Relatively high incidence rates (second rank after North) were observed [Stiller,
Marcos-Gragera, Ardanaz and colleagues; and Stiller, Desandes, Danon and colleagues, this issue], despite the lowest
proportion of non-malignant tumours. This is partly due
to the high incidence of haematopoietic neoplasms, reported from Spain in the past and in this volume2,3 [Coebergh, Reedijk, de Vries and colleagues; Clavel and
colleagues; and Izarzugaza and colleagues, this issue], consistently with findings in Hispanic populations elsewhere
in the world.2,3,43 The South also has the largest proportion
of person-years coming from regional cancer registries
(which tend to record higher incidence rates than national
ones). Rather high incidence rates were accompanied by
average levels of survival [Sankila and colleagues; and Stiller, Desandes, Danon and colleagues, this issue]. The excess
incident cases therefore did not have a better prognosis.
Procedures for excluding cases occurring among non-residents should be examined in these registries, since the regional hospitals may attract non-residents with cancers
that are difficult to treat.
4.5.
The West
For children, the West region was dominated by the patterns
of German Childhood Cancer Registry, which provided 80% of
the data for trends analysis and 70% of that used for geographical comparison. It overshadowed the contributions
from France, Netherlands and Switzerland. The incidence
data for this region are therefore affected by underreporting
of brain tumours in this registry.44 Observed survival data
might be influenced by less complete follow-up (92% of cases
included in survival analyses compared with about 98% in
other regions) and this may contribute to the West (together
with the North) having relatively high survival. However, in
a recent German study with the closing date of December
200434 it became evident that the follow-up considered in this
study was more complete for deceased patients than for survivors, which largely invalidates the assumption of overestimation of survival in the West due to incomplete follow-up.
A specific issue of the West is also the temporary implementation of systematic screening programme for neuroblastoma
in some areas and years covered by this study, which might
have contributed to the increased level of incidence rate of
this tumour, although its effect is relatively moderate [Spix
and colleagues, this issue].
Data on adolescents in the West were provided by the registries of France, Netherland and Switzerland. This dataset
was marked by a low proportion of registries recording nonmalignant tumours and high proportion of carcinoid of
appendix (ASR of almost 8 per million).
The West was least well represented in the early years of
the study period (especially in children), which might have
introduced random fluctuations to the yearly rates and affect
time trends. These problems are compounded by the varying
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
constitution of the datasets for different analyses, notably
due to the contributions of German and Dutch registries.
However, these irregularities did not seem to much affect
the overall European time trends. A better equilibrium of
the West region (as defined for the purposes of this study) will
be achieved in the future, when the contribution of the Dutch
National Cancer registry has increased with the passing of
time45 and the two French national paediatric registries46,47
have joined the study. Newer cancer registries from other
countries in this region may also be able to contribute.
5.
Conclusion
Despite common recommendations and standardisation of
data, the subsets of the ACCIS database are not strictly comparable. However, we have identified a number of artefacts
and quantified their possible impact on the comparisons.
Geographical differences between the regions are relatively
small and removal of the detected artefacts would reduce
them to some extent. Further standardisation of data collection and processing is therefore necessary, especially in children and adolescents. Improvements have already been
recorded in some registries, for example in the German Childhood Cancer Registry,34 cancer registry of Doubs, France [A.
Danzon, personal communication] and some Italian cancer
registries [C. Magnani, personal communication], partly as a
result of this international collaboration within ACCIS. The
construction of the ACCIS database, data standardisation,
evaluation, analyses and interpretation was a long process,
and new data are now available within the collaborating registries, since the current version of the database was ‘closed’
at the end of 2002. Extension of the database and its further
exploration is imminent and will help to answer many of
the questions generated in this series of articles.
Conflict of interest statement
None declared.
Acknowledgement
The ACCIS project was funded by the European Commission
from the Europe Against Cancer programme (contracts
SI2.126875, SI2.321970 and SPC.2002303), jointly with the
International Agency for Research on Cancer. Data analyses
were partly funded by the French Ligue Nationale Contre le
Cancer, Comité du Rhône. ESF was partially supported by
the Communauté urbaine de Lyon and Federal Ministry of
Health of the Federal German Government.
The authors wish to thank Mr Nicolas Mitton (France) for
his input in the set-up and management and exploration of
the ACCIS database, the members of ACCIS Scientific Committee for steering the study and the Guest Editors for comments on earlier drafts. The ACCIS Scientific Committee
consists of Franco Berrino (Italy), Jan Willem Coebergh (The
Netherlands), Brigitte Lacour (France), Peter Kaatsch (Germany), Max Parkin and Charles Stiller (UK) and Eva Steliarova-Foucher (IARC).
EUROPEAN JOURNAL OF CANCER
The following collaborators from the cancer registries contributed actively to the ACCIS study: S.V. Petrovich, O. Budanov (Belarus); H. Storm, N. Christensen (Denmark); T.
Aareleid (Estonia); T. Hakulinen, R. Sankila, E. Pukkala (Finland); E. Le Gall, I. Tron (Brittany, France), B. Lacour, E. Desandes (Lorraine, France), J.L. Bernard, P. Pillon, J.C. Gentet (PACA
and Corsica, France), F. Freycon, C. Berger (Rhone Alps,
France), L. Remontet (Francim, France), A. Danzon, M. Mercier
(Doubs, France), J.P. Daurès, B. Tretarre (Herault, France), F.
Ménégoz (Isère, France), A.V. Guizard (Manche, France), M.
Velten (Bas-Rhin, France), A. Buemi (Haut-Rhin, France), N.
Raverdy (Somme, France), M. Sauvage, P. Grosclaude (Tarn,
France); P. Kaatsch, B. Eisinger, R. Stabenow (Germany); D.
Schuler, Z. Jakab, G. Borgulya (Hungary); L. Tryggvadottir, J.G.
Jonasson, K. Bjarnadottir (Iceland); H. Comber, F. Dwane (Ireland); C. Magnani, G. Pastore (Piedmont, Italy), F. Pannelli, C.
Pascucci (Marche, Macerata, Italy), S. Ferretti (Ferrara, Italy),
E. Conti, V. Ramazzotti, M.C. Cercato (Latina Province, Italy),
M. Vercelli, A. Puppo (Liguria, Italy), P. Crosignani, G. Tagliabue, A. Tittarelli (Lombardy, Italy), V. De Lisi, P. Sgargi (Parma,
Italy), R. Zanetti, S. Patriarca (Piedmont, Italy), R. Tumino (Ragusa, Italy), M. Budroni, D. Piras (Sassari, Italy), E. Paci, E. Crocetti (Tuscany, Italy), F. La Rosa, F. Stracci (Umbria, Italy), P.
1947
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
Zambon, S. Guzzinati (Veneto, Italy); M. Dalmas (Malta);
J.W.W. Coebergh, J. van Dijck, A. Wit, H. de Ridder-Sluiter
(Netherlands); F. Langmark, A. Johansen, A. Andersen (Norway); I. Plesko (Slovakia); M. Primic Žakelj, V. Pompe-Kirn
(Slovenia); R. Peris-Bonet, B. Giner (Spain), E. Almar Marques,
A. Mateos Ramos (Albacete, Spain), J. Ramon Quiros Garcia, A.
Cañada Martı́nez (Asturias, Spain), I. Izarzugaza (Basque,
Spain), A. Alemán Herrera (Canary Islands, Spain), P. Viladiu,
R. Marcos, A. Izquierdo (Girona, Spain), C. Martı́nez Garcia
(Granada, Spain), A. Obrador, I. Garau (Mallorca, Spain), E.
Ardanaz (Navarra, Spain), J. Borràs, J. Galceran (Tarragona,
Spain), J. de la Bárcena Guallar, M.C. Martos Jiménez (Zaragoza, Spain); G. Jundt (Basel, Switzerland), C. Bouchardy, M.
Usel (Geneva, Switzerland), J. Allemann, H. Frick (Graubünden
and Glarus, Switzerland), T. Fisch, S. Ess (St Gallen Appenzell,
Switzerland), F. Joris, D. de Weck (Valais, Switzerland); S. Yalcin Eser (Izmir, Turkey); C.A. Stiller, M.F.G. Murphy, G.J. Draper
(England and Wales, UK), A. Gavin, C. Fox, W. Hamill, R. Middleton (Northern Ireland, UK), D. Brewster, L. Bhatti, A.
McDonald (Scotland, UK). We also acknowledge the collaborators from the other 15 registries, whose data were not included in the analyses because of reduced comparability
and hopefully will be included in the future.
Appendix. International Classification of Childhood Cancer (ICCC)14 (Reprinted with the permission from
the International Journal of Cancer)
Diagnostic group
ICD-O-2 codes
Morphology
I LEUKAEMIA
(a) Lymphoid leukaemia
(b) Acute non-lymphocytic leukaemia
(c) Chronic myeloid leukaemia
(d) Other specified leukaemias
(e) Unspecified leukaemias
9820–9827, 9850
9840, 9841, 9861,
9867, 9891, 9894,
9863, 9868
9830, 9842, 9860,
9890, 9892, 9893,
9941
9800–9804
Topography
9864, 9866,
9910
9862, 9870–
9900, 9930–
II LYMPHOMAS AND RETICULOENDOTHELIAL NEOPLASMS
(a) Hodgkin’s disease
9650–9667
(b) Non-Hodgkin lymphoma
9591–9595, 9670–9686, 9690–
9714, 9723
(c) Burkitt’s lymphoma
9687
(d) Miscellaneous lymphoreticular neoplasms
9720, 9731–9764
(e) Unspecified lymphomas
9590
III CNS AND MISCELLANEOUS INTRACRANIAL AND INTRASPINAL NEOPLASMS
9383, 9390–9394
(a) Ependymomab
(b) Astrocytoma
9380
9381, 9400–9441
(c) Primitive neuroectodermal tumours
9470–9473
(d) Other gliomas
9380
9382a, 9384a
9442–9460, 9481
8270–8281, 8300, 9350–9362,
(e) Other specified intracranial and intraspinal
9480, 9505, 9530–9539
neoplasmsb
8000–8004
(f) Unspecified intracranial and intraspinal neoplasmsb
C72.3
C70.0–C72.0, C72.4–C72.9
C70.0–C72.9, C75.1–C75.3
(continued on next page)
1948
EUROPEAN JOURNAL OF CANCER
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
Appendix – continued
Diagnostic group
ICD-O-2 codes
Morphology
IV SYMPATHETIC NERVOUS SYSTEM TUMOURS
(a) Neuroblastoma and ganglioneuroblastoma
(b) Other sympathetic nervous system tumours
Topography
9490, 9500
8680, 8693–8710, 9501–9504,
9520–9523
V RETINOBLASTOMA
9510–9512
VI RENAL TUMOURS
(a) Wilms’ tumour, rhabdoid and clear cell sarcoma
(b) Renal carcinoma
(c) Unspecified malignant renal tumours
VII HEPATIC TUMOURS
(a) Hepatoblastoma
(b) Hepatic carcinoma
(c) Unspecified malignant hepatic tumours
VIII MALIGNANT BONE TUMOURS
(a) Osteosarcoma
(b) Chondrosarcoma
(c) Ewing’s sarcoma
(d) Other specified malignant bone tumours
(e) Unspecified malignant bone tumours
IX SOFT-TISSUE SARCOMAS
(a) Rhabdomyosarcoma and embryonal sarcoma
(b) Fibrosarcoma, neurofibrosarcoma and other
fibromatous neoplasms
(c) Kaposi’s sarcoma
(d) Other specified soft tissue sarcomas
(e) Unspecified soft tissue sarcomas
8960, 8964
8963
8010–8041, 8050–8075, 8082,
8120–8122, 8130–8141, 8143,
8155, 8190–8201, 8210, 8211,
8221–8231, 8240, 8241, 8244–
8246, 8260–8263, 8290, 8310,
8320, 8323, 8401, 8430, 8440,
8480–8490, 8504, 8510, 8550,
8560–8573
8312
8000–8004
8970
8010–8041, 8050–8075, 8082,
8120–8122, 8140, 8141, 8143,
8155, 8190–8201, 8210, 8211,
8230, 8231, 8240, 8241, 8244–
8246, 8260–8263, 8310, 8320,
8323, 8401, 8430, 8440, 8480–
8490, 8504, 8510, 8550, 8560–
8573
8160–8180
8000–8004
9180–9200
9220–9230
9231, 9240
9260
9363, 9364
8812, 9250, 9261–9330, 9370
8000–8004, 8800, 8801, 8803,
8804
8900–8920, 8991
8810, 8811, 8813–8833, 9540–
9561
9140
8840–8896, 8982, 8990, 9040–
9044, 9120–9134, 9150–9170,
9251, 9581
8963
9231, 9240, 9363, 9364
9260
8800–8804
C64.9, C80.9
C64.9
C64.9
C22.0, C22.1
C22.0, C22.1
C40.0–C41.9
C40.0–C41.9, C80.9
C40.0–C41.9
C40.0–C41.9
C00.0–C63.9,
C00.0–C39.9,
C00.0–C39.9,
C00.0–C39.9,
C65.9–C76.8
C47.0–C80.9
C47.0–C76.8
C44.0–C80.9
EUROPEAN JOURNAL OF CANCER
1949
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
Appendix – continued
Diagnostic group
ICD-O-2 codes
Morphology
X GERM-CELL, TROPHOBLASTIC AND OTHER GONADAL NEOPLASMS
9060–9102
(a) Intracranial and intraspinal germ cell tumoursb
(b) Other and unspecified non-gonadal germ cell
9060–9102
tumours
(c) Gonadal germ cell tumours
(d) Gonadal carcinomas
(e) Other and unspecified malignant gonadal tumours
9060–9102
8010–8041, 8050–8075, 8082,
8120–8122, 8130–8141, 8143,
8155, 8190–8201, 8210, 8211,
8221–8241, 8244–8246, 8260–
8263, 8290, 8310, 8320, 8323,
8430, 8440, 8480–8490, 8504,
8510, 8550, 8560–8573,
8380, 8381, 8441–8473
8590–8670, 9000
8000–8004
XI CARCINOMAS AND OTHER MALIGNANT EPITHELIAL NEOPLASMS
(a) Adrenocortical carcinoma
8370–8375
(b) Thyroid carcinoma
8010–8041, 8050–8075, 8082,
8120–8122, 8130–8141, 8155,
8190, 8200, 8201, 8211, 8230,
8231, 8244–8246, 8260–8263,
8290, 8310, 8320, 8323, 8430,
8440, 8480, 8481, 8500–8573
8330–8350
(c) Nasopharyngeal carcinoma
8010–8041, 8050–8075, 8082,
8120–8122, 8130–8141, 8155,
8190, 8200, 8201, 8211, 8230,
8231, 8244–8246, 8260–8263,
8290, 8310, 8320, 8323, 8430,
8440, 8480, 8481, 8504, 8510,
8550, 8560–8573
(d) Malignant melanoma
8720–8780
(e) Skin carcinoma
8010–8041, 8050–8075, 8082,
8090–8110, 8140, 8143, 8147,
8190, 8200, 8240, 8246, 8247,
8260, 8310, 8320, 8323, 8390–
8420, 8430, 8480, 8542, 8560,
8570–8573, 8940
(f) Other and unspecified carcinomas
8010–8082, 8120–8155, 8190–
8263, 8290, 8310, 8314–8323,
8430–8440, 8480–8580, 8940,
8941
XII OTHER AND UNSPECIFIED MALIGNANT NEOPLASMS
(a) Other specified malignant tumours
(b) Other unspecified malignant tumours
a Behaviour code /1 is included.
b Behaviour codes /0 and /1 are included.
8930, 8933, 8950, 8951, 8971–
8981, 9020, 9050–9053, 9110,
9580
8000–8004
Topography
C70.0–C72.9, C75.1–C75.3
C00.0–C55.9, C57.0–C61.9,
C63.0–C69.9, C73.9–C75.0,
C75.4–C80.9
C56.9, C62.0–C62.9
C56.9, C62.0–C62.9
C56.9, C62.0–C62.9
C73.9
C11.0–C11.9
C44.0–C44.9
C00.0–C10.9,
C23.9–C39.9,
C50.0–C55.9,
C63.0–C63.9,
C75.0–C80.9
C12.9–C21.8,
C48.0–C48.8,
C57.0–C61.9,
C65.9–C72.9,
C00.0–C21.8,
C42.0–C55.9,
C63.0–C63.9,
C73.9–C75.0,
C23.9–C39.9,
C57.0–C61.9,
C65.9–C69.9,
C75.4–C80.9
1950
EUROPEAN JOURNAL OF CANCER
R E F E R E N C E S
1. Ferlay J, Bray F, Pisani P, Parkin DM. GLOBOCAN 2000. Cancer
incidence, mortality and prevalence worldwide, version 1.0.
IARC CancerBase No. 5. Lyon: IARC Press; 2001.
2. Parkin DM, Stiller CA, Draper GJ, Bieber CA, Terracini B, Young
JL, editors. International incidence of childhood cancer. IARC
scientific publication no. 87. Lyon: International Agency for
Research on Cancer; 1988.
3. Parkin DM, Kramárová E, Draper GJ, et al., editors. IARC
scientific publications no. 144, International incidence of childhood
cancer vol. 2. Lyon: IARC; 1998.
4. Ries LAG, Smith MA, Gurney JG, et al., editors. Cancer
incidence and survival among children and adolescents:
United States. SEER Program 1975–1995. National Cancer
Institute, SEER Program. NIH Pub. No. 99-4649, Bethesda, MD;
1999.
5. Capocaccia R, Gatta G, Magnani C, Stiller C, Coebergh JW,
editors. Childhood cancer survival in Europe 1978–1992: the
EUROCARE study. Eur J Cancer 2001;37:671–816.
6. Gatta G, Capocaccia R, De Angelis R, Stiller C, Coebergh JW,
EUROCARE Working Group. Cancer survival in European
adolescents and young adults. Eur J Cancer 2003;39:
2600–10.
7. Steliarova-Foucher E, Stiller C, Kaatsch P, et al. Geographical
patterns and time trends of cancer incidence and survival
among children and adolescents in Europe since 1970s: the
ACCIS project. Lancet 2004;364:2097–105.
8. Little J. Epidemiology of childhood cancer. IARC scientific
publications no. 149. Lyon: International Agency for Research
on Cancer; 1999.
9. Gatta G, Capocaccia R, De Angelis R, Stiller C, Coebergh JW,
EUROCARE Working Group. Cancer survival in European
adolescents and young adults. Eur J Cancer 2003;39:2600–10.
10. Kramárová E, Stiller CA, Ferlay J, et al. International
classification of childhood cancer 1996. IARC technical report no.
29. Lyon: International Agency for Research of Cancer; 1996.
11. Parkin DM, Tyczynski JE, Démaret E, editors. Standards and
guidelines for cancer registration in Europe. IARC technical
publication no. 40. Lyon: International Agency for Research of
Cancer; 2003. p. 69–73.
12. Percy C, Van Holten V, Muir CS, editors. International
classification of diseases for oncology. second ed. Geneva: World
Health Organisation; 1992.
13. Teppo L, Hakama M, Sankila R. Finland, Finnish Cancer
Registry, 1980–1989. In: Parkin DM, Kramárová E, Draper GJ,
et al., editors. International incidence of childhood cancer. IARC
Scientific Publications No 144, vol. 2. Lyon: IARC; 1998. p.
261–3.
14. Kramárová E, Stiller CA. The International Classification of
Childhood Cancer. Int J Cancer 1996;68:759–65.
15. Berkson J, Gage RP. Calculations of survival rates for cancer.
Proc Staff Meet Mayo Clin 1950;5:270–86.
16. Ferlay J, Burkhard C, Whelan S, Parkin DM. Check and
conversion programs for cancer registries (IARC/IACR Tools
for Cancer Registries). IARC technical report no. 42, Lyon,
2005. 29 March 2006. Available from http://www.iacr.com.fr/
iacr_iarccrgtools.htm.
17. Intercooled Stata 8.2 for Windows.
18. MicrosoftÒ Excel 2000.
19. Parkin DM, Chen VW, Ferlay J, Galceran J, Storm HH, Whelan
S. Comparability and quality control in cancer registration. IARC
technical report no. 19. Lyon: International Agency for
Research on Cancer; 1994.
20. United Nations, 2005. Definition of major areas and regions.
11 October 2005. Available from http://esa.un.org/unpp/
index.asp?panel=5.
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
21. Chauvin F, Mathieu P, Frappaz D, et al. Screening for
neuroblastoma in France: methodological aspects
and preliminary observations. Med Pediatr Oncol
1997;28:81–91.
22. Schilling FH, Spix C, Berthold F, et al. Children may not
benefit from neuroblastoma screening at 1 year of age.
Updated results of the population based controlled trial in
Germany. Cancer Lett 2003;197(1–2):19–28.
23. Powell JE, Esteve J, Mann JR, et al. Neuroblastoma in Europe:
differences in the pattern of disease in the UK. SENSE. Study
group for the Evaluation of Neuroblastoma Screening in
Europe. Lancet 1998;352(9129):682–7.
24. Boyle P, Parkin DM. Statistical methods for registries. In:
Jensen OM, Parkin DM, MacLennan R, Muir CS, Skeet R,
editors. Cancer registration: principles and methods. IARC scientific
publications no. 95. Lyon: International Agency for Research
on Cancer; 1991. p. 126–58.
25. Doll R, Payne P, Waterhouse JAH. Cancer incidence in five
continents, vol. I. Geneva,Berlin: UICC,Springer; 1966.
26. Kirkwood B. Essentials of medical statistics. Oxford:
Blackwell Scientific Publications; 1988.
27. Parkin DM, Hakulinen T. Analysis of survival. In: Jensen OM,
Parkin DM, MacLennan R, Muir CS, Skeet R, editors. Cancer
registration: principles and methods. IARC Scientific Publications
No. 95. Lyon: International Agency for Research on Cancer;
1991. p. 159–76.
28. Kalbfleisch JD, Prentice RL. The statistical analysis of failure
time data. New York: John Wiley & Sons; 1980.
29. Peto R, Pike MC, Armitage P, et al. Design and analysis of
randomized clinical trials requiring prolonged observation of
each patient, II, analysis and examples. Br J Cancer
1977;35:1–39.
30. Bullard J, Coleman MP, Robinson D, Lutz J-M, Bell J, Peto J.
Completeness of cancer registration: a new method for
routine use. Br J Cancer 2000;82:1111–6.
31. Robles SC, Marrett LD, Clarke EA, Risch HA. An application of
capture-recapture methods to the estimation of
completeness of cancer registration. J Clin Epidemiol
1988;41(5):495–501.
32. Steliarova-Foucher E, Stiller C, Kaatsch P, Berrino F,
Coebergh JW, ACCIS Scientific Committee. Childhood
cancer incidence trends in Europe, 1970–1999. Lancet
2005;365:2088.
33. Enstrom JE, Austin DF. Interpreting cancer survival rates. The
available data on survival are not a sensitive measure of
progress in cancer control. Science 1977;195:851–74.
34. Kaatsch P, Blettner M, Spix C, Jürgens H. Follow-up of
long-term survivors after childhood cancer in Germany. Klin
Pädiatr 2005;217:169–75.
35. Draper GJ, Birch JM, Bithell JF, et al. Childhood cancer in Britain.
Incidence, survival and mortality. Studies on medical and population
subjects No. 37. Office of Population Censuses and
Surveys. London: Her Majesty Stationery Office; 1982.
36. Stiller CA, Allen M, Bayne A, et al. National Registry of
Childhood Tumours, England and Wales, 1981–1990. In: Parkin
E, Kramárová E, Draper GJ, et al., editors. International incidence
of childhood cancer. IARC scientific publications no. 144, vol.
2. Lyon: IARC; 1998. p. 365–7.
37. Hawkins MM, Swerdlow AJ. Completeness of cancer and
death follow-up obtained through the National Health Service
Central Register for England and Wales. Br J Cancer
1992;66(2):408–13.
38. Stiller CA, McKinney PA, Bunch KJ, Bailey CC, Lewis IJ.
Childhood cancer and ethnic group in Britain: a United
Kingdom children’s Cancer Study Group (UKCCSG) study. Br J
Cancer 1991;64(3):543–8.
39. Campbell J, Wallace WHB, Bhatti LA, Stockton DL, Rapson T,
Brewster DH. Childhood cancer in Scotland: trends in incidence,
EUROPEAN JOURNAL OF CANCER
40.
41.
42.
43.
mortality and survival 1975–1999. Edinburgh: Information and
Statistics Division; 2003.
Hrusak O, Trka J, Zuna J, Polouckova A, Kalina T, Stary J. Czech
Pediatric Hematology Working Group Acute lymphoblastic
leukaemia incidence during socio-economic transition:
selective increase in children from 1 to 4 years. Leukemia
2002;16:720–5.
Geryk E. Awareness medal for Czech registry. Eur J Cancer
2004;40:2010.
Association of Nordic Cancer Registries. 25 October 2005.
Available from http://ncu.cancer.dk/ancr/.
Stiller CA, Parkin DM. International variations in the
incidence of childhood lymphomas. Paediatr Perinat Epidemiol
1990;4(3):303–24.
4 2 ( 2 0 0 6 ) 1 9 1 5 –1 9 5 1
1951
44. Kaatsch P, Rickert CH, Kühl J, Schüz J, Michaelis J. Populationbased epidemiologic data on brain tumors in German
children. Cancer 2001;92:3155–64.
45. Coebergh JWW, Van Dijck JAMM, Janssen-Heijnen MLG, Visser
O, editors. The Netherlands Cancer Registry. Childhood Cancer in
the Netherlands 1989–1997. Utrecht: Association of
Comprehensive Cancer Centres; 2000.
46. Clavel J, Goubin A, Auclerc MF, et al. Incidence of
childhood leukemia and non-hodgkin’s lymphoma in
France – National Registry of Childhood Leukemia and
Lymphoma, 1990–1999. Eur J Cancer Prev 2004;13:
97–103.
47. Registre National des Tumeurs Solides de l’Enfant. 25 October
2005. Available from www.chu-nancy.fr/rntse/.