Quality, comparability and methods of analysis of data on childhood cancer in Europe (1978–1997): Report from the Automated Childhood Cancer Information System project

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/.