Childhood Cancer Survival Trends in Europe: A EUROCARE Working Group Study
http://www.100md.com
《临床肿瘤学》
the Epidemiology Unit, Istituto Nazionale per lo Studio e la Cura dei Tumori
Pediatric Unit, Istituto Nazionale per lo Studio e la Cura dei Tumori, Milan
Laboratory of Epidemiology, Istituto Superiore di Sanità, Rome, Italy
Childhood Cancer Research Group, University of Oxford, Oxford, United Kingdom
Institute for Medical Biostatistics, Epidemiology and Informatics, German Childhood Cancer Registry, University Mainz, Mainz, Germany
ABSTRACT
PURPOSE: EUROCARE collected data from population-based cancer registries in 20 European countries. We used this data to compare childhood cancer survival time trends in Europe.
PATIENTS AND METHODS: Survival in 44,129 children diagnosed under the age of 15 years during 1983 to 1994 was analyzed. Sex- and age-adjusted 5-year survival trends for 10 common cancers and for all cancers combined were estimated for five regions (West Germany, the United Kingdom, Eastern Europe, Nordic countries, and West and South Europe) and Europe as a whole. Europe-wide trends for 14 rare cancers were estimated.
RESULTS: For all cancers combined, 5-year survival increased from 65% for diagnoses in 1983 to 1985 to 75% in 1992 to 1994. Survival improved for all individual cancers except melanoma, osteosarcoma, and thyroid carcinoma; although for retinoblastoma, chondrosarcoma, and fibrosarcoma, improvements were not significant. The most marked improvements (50% to 66%) occurred in Eastern Europe. For common cancers, the greatest improvements were for leukemia and lymphomas, with risk of dying reducing significantly by 5% to 6% per year. Survival for CNS tumors improved significantly from 57% to 65%, with risk reducing by 3% per year. Risk reduced by 4% per year for neuroblastoma and 3% per year for Wilms’ tumor and rhabdomyosarcoma. The survival gap between regions reduced over the period, particularly for acute nonlymphocytic leukemia, CNS tumors, and rhabdomyosarcoma. For rare Burkitt’s lymphoma, hepatoblastoma, gonadal germ cell tumors, and nasopharyngeal carcinoma, risk reductions were at least 10% per year.
CONCLUSION: These gratifying improvements in survival can often be plausibly related to advances in treatment. The prevalence of European adults with a history of childhood cancer will inevitably increase.
INTRODUCTION
Survival for childhood cancers has improved dramatically over the last half century. Before 1950, almost all children who developed cancer died. At present, approximately three out of four children who are diagnosed with cancer can be cured.1 The most recently available (1990 to 1994) survival data on European cancer patients2 provide an opportunity to assess progress in childhood cancer survival over a 12-year period. Cancer survival in Europe data are available as a result of EUROCARE, a cooperative, cancer registry-based project.3,4 EUROCARE childhood survival data have been compared with data available for the United States and found to be fairly similar.5 This is not the case for adult cancer survival, which is generally better in the United States.6 The European publications documented large differences in childhood cancer survival between different European populations and also revealed improvements in survival between 1978 and 1989.7 However, survival trends were reported for Europe as a whole,7 and no studies comparing trends between European countries or regions have been performed.
This study analyzes 5-year survival in European children with cancers diagnosed between 1983 and 1994 and incorporates data from more cancer registries and countries than were included in earlier EUROCARE studies. The aims of this study were to compare survival trends for important cancer types between European countries and regions, to obtain an idea of the impact of new diagnostic and treatment protocols introduced in 1980 to 1995, and to investigate whether childhood survival differences were reduced. This study also takes advantage of the large EUROCARE incidence database to present prognostic trends in Europe for the rarest childhood cancers.
PATIENTS AND METHODS
We analyzed survival in 44,129 European children diagnosed with cancer under the age of 15 years during the period of 1983 to 1994. We excluded patients diagnosed from 1978 to 1982 because several registries did not cover these years. All children with a malignant neoplasm as defined by second edition International Classification of Diseases for Oncology behavior code 3 or higher were included.8 The patients were contributed by 27 population-based cancer registries in 17 countries participating in EUROCARE-3.9 All the registries included in this analysis provided data for the whole period (1983 to 1994). The database includes data from five specialized (childhood) cancer registries (England and Wales; Germany, only West Germany in this study; Piedmont in Italy; Lorraine in France; and Valencia in Spain). Most cancers, with a single exception, were classified according to the 12 diagnostic groups in the International Classification of Childhood Cancers (ICCC).10 The exception regarded groups of intracranial and intraspinal neoplasms (International Classification of Childhood Cancers III and Xa); for these cancers, only malignant tumors, as specified by the EUROCARE protocol, were included. Furthermore, because astrocytoma data from the Finnish registry were not coded using International Classification of Diseases for Oncology morphologic codes, these data were excluded.
Table 1 presents a general description of the data analyzed, showing sample size, demographic characteristics, and main data quality indicators by country and registry. The registries of Denmark, Finland, England and Wales, Estonia, Iceland, Norway, Scotland, Slovakia, Slovenia, Sweden, and West Germany cover the entire populations of their countries. Other countries are represented by one or more local or regional registries. Table 1 also lists the proportions of boys (56%), children aged less than 5 years (45%), microscopically verified cancers (95%; between-country range, 88% to 100%), and patients lost to follow-up (1.5%). All patients were observed until at least the end of 1998. However, a number of patients were censored as alive without having the 5-year follow-up specified by the protocol, having been diagnosed less than 5 years before the end of follow-up. For most registries, except those of France, West Germany, Cracow, Eindhoven, and Geneva, the proportions of patients with less than 4 years of follow-up were less than 4%.
Both histologically verified and nonverified patients were included, but patients known to registries by death certificate only (DCO) or autopsy report only were excluded (337 patients, 0.7%). The number of patients classified in unspecified categories within each major diagnostic group was also low (1,558 patients overall, 3.5%). More detailed information on the EUROCARE database is available elsewhere.9,11
Observed survival was calculated by the actuarial method. Relative survival was also calculated12 but is not presented here because it corresponds closely to observed survival in young people because deaths as a result of competing risks are rare.
The survival trend data are presented for all cancers combined and for 24 selected malignancies. Survival for all cancers combined was adjusted for patient mix (calculated by weighting individual cancer survival figures with the number of patients in each cancer category as a proportion of the whole data set). For this purpose, rare tumors were pooled into a category termed other, which contained approximately 20% of all cancers; astrocytoma and primitive neuroectodermal tumor (PNET) were included in the general CNS tumors category.
Survival estimates for the 10 major types of tumors and for all malignancies combined were compared between populations. We defined five European regions as follows: Denmark, Finland, Iceland, Norway, and Sweden (Nordic countries); England, Wales, and Scotland (the United Kingdom); Estonia, Poland, Slovakia, and Slovenia (Eastern Europe); West Germany, which was kept separate because of the large number of patients; and France, Italy, the Netherlands, Spain, and Switzerland (Western and Southern Europe).
Five-year survival was compared for four 3-year diagnosis periods (1983 to 1985, 1986 to 1988, 1989 to 1991, and 1992 to 1994). Because survival for many of the cancers depends on age and sex and the distribution of these variables varies across countries and over time, survival estimates were adjusted for these variables. To avoid problems caused by a few cases in individual categories when cross stratifying by sex, age, region, and period, a model approach was used instead of the standard direct adjustment method. A proportional hazards model was fitted to individual survival data for the 10 cancer categories (the common cancers plus all cancers combined). Age (three classes: 0 to 4, 5 to 9, and 10 to 14 years for all tumors except neuroblastoma, for which the age division was < 1, 1 to 4, and 5 to 14 years), sex, period (n = 4), region (n = 5), and a period x region interaction term were included as covariates. The model was then used to calculate 5-year survival for each region, for each time period, and for the average sex and age distribution of the whole sample. Linear trends in survival during the analysis period (1983 to 1994) were estimated by a similar proportional hazards model using year of diagnosis as a continuous covariate instead of categoric periods. The parameter relative risk (RR), which was estimated from the continuous proportional hazards model, is the RR of dying of children diagnosed in a given year compared with those diagnosed a year previously and averaged over the whole study period of 1983 to 1994. We also refer to the reduction in risk of dying; this is 1 RR and is expressed as a percentage.
For the European pool, the following two sets of 5-year survival figures were calculated: (1) crude and (2) region population weighted and sex and age adjusted. The latter were estimated, for each time period, as the weighted average of the corresponding region-specific model-based survival values using the childhood population counts in each region as weights. The weighted and adjusted European survival figures are considered in the Discussion but not presented in Table 2.
RESULTS
Table 2 lists 5-year survival trends for 24 tumor types, all CNS tumors, and all cancers combined, estimated from the entire European pool. Age- and sex-adjusted survival trends for all cancer combined and for the 10 most common cancers for the five European regions are shown in Figures 1 and 2. The corresponding linear trends of RR of dying are listed in Table 3. We present the results by cancer type.
For all cancers combined, 5-year survival increased smoothly from 65% in children diagnosed in 1983 to 1985 to 75% in 1992 to 1994, with the risk of dying reduced by an average of 5% per year (Table 2). The rate of increase in survival was similar in all regions (Fig 1). Survival reached approximately 75% in 1992 to 1994, except in Eastern Europe, where it was 66%.
With few exceptions, 5-year survival improved for all the types of cancer (Table 2); the improvements were statistically significant for all common tumors and several rare tumors. For thyroid carcinoma, melanoma and, osteosarcoma, survival was stable. The Eastern European countries experienced the greatest improvement in survival, although improvement was not always statistically significant (Fig 2 and Table 3).
Considering the major cancers, greatest overall improvements were seen for leukemias and lymphomas (Table 2). Five-year survival for lymphoid leukemia was less than 80% in 1983 to 1985 but had surpassed 80% in the Nordic countries, Western and Southern Europe, the United Kingdom, and West Germany in 1992 to 1994 (Fig 2); notwithstanding an impressive improvement in Eastern European countries, the gap between these countries and other regions of Europe did not diminish.
For acute nonlymphocytic leukemia (ANLL), significant survival increases occurred in the United Kingdom and Eastern Europe, where the risk of dying reduced by 8% to 9% per year across the study period (Table 3). For all regions, 5-year survival was less than 50% in 1983 to 1985 but had increased to greater than 50% in the United Kingdom, West Germany, and the Nordic countries by 1992 to 1994 (Fig 2). In Western and Southern Europe, however, there was no change in survival for ANLL over the study period.
Five-year survival for Hodgkin’s disease (HD) was greater than 90% over the entire period and remained stable in most regions (Fig 2); however, survival in the United Kingdom improved significantly, where the risk of dying decreased by approximately 10% every year (Table 3). For non-Hodgkin’s lymphoma, progress was marked in Europe, with the RR of death decreasing by 6% per year over the whole period (Table 2); in West Germany, the risk of dying decreased significantly by 10% per year (Table 3).
Progress for solid tumors was less marked than for hematologic malignancies. Survival for CNS tumors improved significantly, with the risk of dying reduced by 3% per year (Table 2). The improvement was significant for the United Kingdom, West Germany, the Nordic countries, and Eastern Europe (Table 3). In Eastern Europe, 5-year survival for CNS tumors improved significantly from 47% to 62% over the study period, with risk of dying reduced by 4% per year; the improvements were particularly marked in Estonia and Slovenia. Progress occurred first in the Nordic countries and West Germany, followed by Western and Southern Europe and, subsequently, the United Kingdom (Fig 2). The survival gap between regions reduced in 1992 to 1994 (Fig 2).
For PNET medulloblastoma and astrocytoma, the risk of dying was reduced significantly by 3% per year (Table 2), and for astrocytoma, the survival gap between regions had reduced by 1992 to 1994 (Fig 2). For PNET and medulloblastoma, West Germany and the Nordic countries showed similar survival trends, which were characterized by overall improvement but a dip in survival in 1989 to 1991; a similar dip was observed for Western and Southern Europe at the end of the period (Fig 2).
For neuroblastoma, the risk of dying was reduced by 4% per year (Table 2), with significant improvements for West Germany and the United Kingdom (Table 3) and a reduction in the between-region survival gap. During 1992-1994, West Germany and Western and Southern Europe had 5-year survival rates that approached 70%, whereas for all the other regions, survival was similar and approximately 60% (Fig 2).
For Wilms’ tumors, the risk of dying decreased by 3% per year overall. In the Nordic countries, 5-year survival improved significantly from 81% to 93% (Fig 2); marked improvements also occurred in West Germany and Eastern Europe. In Eastern Europe, survival increased from 59% in 1983 to 1985 to 75% in 1992 to 1994. Survival in the United Kingdom and Western and Southern Europe remained stable at approximately 80% over the period.
For rhabdomyosarcoma, the overall risk of dying decreased significantly by approximately 3% per year (Table 2). The decrease was greatest (but not significant) in the Nordic countries (5% per year), Eastern Europe (6% per year), and Western and Southern Europe (6% per year). The gap between regions was reduced markedly over the period.
For rare childhood cancers (less than 1,500 patients), 5-year survival improved generally (Table 2), except for melanoma, osteosarcoma, and thyroid carcinoma, for which survival remained stable or decreased slightly. Progress was remarkable for Burkitt’s lymphoma, ependymoma, hepatoblastoma, Ewing’s sarcoma, gonadal germ cell tumors, and nasopharyngeal carcinoma. Five-year survival for ependymoma, hepatoblastoma, Ewing’s sarcoma, and nasopharyngeal carcinoma increased from less than 50% in 1983 to 1985 to more than 50% in 1992 to 1994. Survival for retinoblastoma also improved (nonsignificantly), although survival was already high at the outset.
DISCUSSION
Our major findings are as follows. First, from 1983 to 1994, survival improved for all childhood cancers combined in Europe; the risk of dying was reduced by 5% per year. Second, the greatest survival improvements among the common cancers were for leukemia and lymphoma. Third, for all cancers combined and some major tumors, the most marked survival improvements occurred in Eastern Europe. Although Eastern European survival was always lower than in the rest of Europe, the between-region gap reduced over the study period, particularly for ANLL, CNS tumors, astrocytoma, and rhabdomyosarcoma.
Differences in data quality and comparability between registries have to be considered as possible sources of bias that may detract from the validity of these findings. The main indicators of data quality are the proportion of DCO patients, proportion of patients with microscopically verified cancer, and proportion of patients lost to follow-up. In this study, DCO patients were rare except in Slovakia, Latina, and Estonia (Table 1). Furthermore, 95% of patients were confirmed microscopically. Microscopic confirmation is particularly important for childhood tumors, which are primarily classified by histologic type. It is noteworthy that few patients (1.5% overall) were lost to follow-up. All registries except Eindhoven, West Germany, and the French registries observed their patients for at least 5 years; in France, legislation makes it difficult to collect life status data.11 It is also noteworthy that data quality indicators did not vary greatly over the study period. The only exception was that the proportion of patients alive with follow-up of less than 4 years increased from 1.7% in 1983 to 1985 to 7.8% in 1992 to 1994.
This concentration, in 1992 to 1994, of patients alive with follow-up of less than 4 years could bias trend estimates if these censored patients for each diagnostic group, sex, and age group had different prognoses from patients not censored. Therefore, we analyzed the distribution of these patients according to diagnostic group. To do this, we assigned a 5-year survival probability for each patient equal to the mean observed survival according to tumor category. We found that these assigned probabilities were 76.6% for patients censored at 0 months and 76.6% for patients lost between 1 and 47 months. These figures are just greater than the mean survival for all patients (74.9%), as one would expect because many censored patients were from West Germany. Thus, these censored patients did not contain an excess of poor prognosis tumors.
We also examined all censored patients. We found, among patients in 1992 to 1994, that 73% had at least 4 but less than 5 years of follow-up. These patients did not influence the survival trends with time because, for all cancers, most improvements in survival occurred in the first few years after diagnosis. In particular, the differences in survival after 4 and after 5 years in both the latest and the penultimate diagnosis periods were absent or minimal. This analysis indicates that no major bias is a result of high censoring of patients diagnosed in the latest period.
Although our findings are derived from a large sample from 17 countries, the sample is not representative of the totality of European pediatric patients. Western and Southern Europe, including Spain, France, Switzerland, Italy, and the Netherlands, provided less than 12% of the total patients in their respective countries and are underrepresented in the pooled European estimates. Population-weighted survival estimates9,11 for tumor type were calculated in this study but did not differ sufficiently from the crude pooled survival estimates, and this is explained by the fact that survival in the most populous European region (Western and Southern Europe) was always close to the average pooled survival levels.
We defined five broad European regions. This grouping was necessary because of the low incidence of childhood cancers. The grouping was partly geographic but also based on broad similarity of survival figures in the countries grouped together, both in the present analysis and in past study.2 We included Denmark with the Nordic countries even though its survival was somewhat low because increases in survival for all cancers combined and for the major childhood tumors were similar in all Nordic countries.
Improvements in survival over the study period can often be related to advances in treatments. For example, therapies for lymphoid leukemia were intensified during this time; multidrug regimens were introduced, and therapy intensities were stratified according to patient risk group.13 The greatest improvements in lymphoid leukemia survival occurred in the Nordic countries, and this is probably related to a major initiative by the Nordic Society of Pediatric Hematology and Oncology. In 1981, the Nordic Society of Pediatric Hematology and Oncology began registering all lymphoid leukemias in children under 15 years of age in the five Nordic countries to monitor the impact of high-dose methotrexate combined with reduction of CNS irradiation and to encourage the use uniform protocols.14 This population-based initiative found that 5-year event-free survival increased from 56.5% in the early 1980s to 77.6% in the 1990s. Lack of appropriate chemotherapy agents may be an important reason for low survival for lymphoid leukemia in Eastern Europe.
The marked increase in ANLL survival in the United Kingdom over the study period may reflect widespread participation in the Medical Research Council’s AML10 trial, which was carried out in 1988 to 1995 and which had survival results that were better than those of all previous large-scale trials in this disease.15 Significant progress for ANLL also occurred in Eastern Europe overall (and in all individual countries except Slovakia), with the risk of dying reduced by approximately 8% per year. Much more modest improvements were seen in West Germany and the Nordic countries, where survival was high at the outset. No improvement occurred in Western and Southern Europe overall.
Survival for HD improved by an impressive 10% in the United Kingdom over the study period, although an explanation similar to that for ANLL is not apparent. There have been attempts to minimize the long-term sequelae of HD treatment in children by reducing the dose of the most toxic component (radiotherapy) of multimodality treatments. This may be the reason for the survival decrease seen for this disease in 1992 to 1994 in the Nordic countries (Fig 2); note, however, that both the number of patients (n = 63) and the number of deaths (n = 6) were small. Survival figures over periods longer than 5 years may provide better indications of the utility of the low-dose approach.
Therapy for non-Hodgkin’s lymphoma, like that for leukemia, became increasingly based on multidrug regimens over the study period.16 However, the only significant trend to increased survival (from 75% to 89%) was seen in West Germany, plausibly because of the activity of the German-Austrian Cooperative Groups.17
For CNS tumors, computed tomography, which was introduced in the 1970s, and magnetic resonance imaging, which was disseminated widely in the 1980s, have become standard tools for disease imaging and evaluation. Neuroimaging is particularly important as a guide to the extent of resection.18 Our analysis found that survival for CNS tumors increased first in the Nordic countries and West Germany and then in Western and Southern Europe, the United Kingdom, and finally Eastern Europe (Fig 2). This pattern may reflect staggered implementation of magnetic resonance imaging to assess these tumors. Use of chemotherapy for these tumors also increased over the period, with the aim of avoiding, delaying, or reducing radiation and hence maintaining quality of life. For PNET and medulloblastoma, trials to reduce the radiotherapy dose19 could be responsible for the reduction of survival found in some regions (Fig 2). The updated protocols recommend that radiotherapy be combined with high-dose chemotherapy after surgery.20
We found significant improvements in survival for PNET and medulloblastoma (from 48% to 60%) and astrocytoma (from 72% to 80%) in West Germany over the study period. However, both incidence and follow-up in West Germany for childhood CNS cancers were less complete than for the other childhood cancers, and the favorable survival trend should be considered cautiously. In Germany, many children with CNS tumors do not receive chemotherapy and, thus, are not seen by pediatric oncologists who are the main providers of incidence notifications to cancer registries. Furthermore, the completeness of reported incidence and the quality of follow-up for childhood CNS tumors has varied over time in relation to the number of children included in clinical trials; the start of a large clinical trial on CNS tumors in the mid-1980s coincided with a considerable increase in the completeness of registration.21 Five clinical trials on primary CNS tumors were ongoing in 2001 in Germany, involving 82% of reported incident cases of CNS tumors. Yet, this proportion is low compared with the overall figure of 92% of incident cases of childhood malignancies recruited to trials.21
Prognosis for neuroblastoma depends on age and disease stage at diagnosis; the disseminated, rapidly expanding disease often found in older children is particularly difficult to treat. We found that overall 5-year survival for neuroblastoma improved significantly from 42% to 54% in older children (1 to 14 years of age) characterized by poor prognosis; whereas for infants with good prognosis, survival increased only modestly (from 81% to 84%). These findings suggest an effective improvement in the therapeutic modalities used to treat this tumor.
A major improvement in survival for Wilms’ tumor occurred before the present study period. We found that survival from 1983 to 1994 remained stable in the United Kingdom and Western and Southern Europe but improved in the Nordic countries, West Germany, and Eastern European countries. In the United Kingdom, national trials were ongoing throughout this period.22,23 Also, for most of this period, joint International Society of Pediatric Oncology-German Society of Pediatric Oncology studies24 were being conducted principally in France, Germany, Italy, Spain, Netherlands, Sweden, Norway, Denmark, and Slovenia. An aim of these studies was to reduce the chemotherapy, radiotherapy, or both administered to patients who had characteristics associated with favorable outcome to limit long-term sequelae.25 For these reasons, any progress in Wilms’ tumor for patients diagnosed in 1983 to 1994 may become evident later. However, a survival improvement occurred in the Nordic countries over the study period, which may have been a result of wider recruitment of children in the clinical trials conducted in these countries.
For rhabdomyosarcoma, survival differences between the European populations were reduced effectively to zero from 1983 to 1994. Survival in the Eastern European countries increased markedly from the third to the fourth period as a result of the 100% survival in the eight patients from Cracow and Slovenia.
Important improvements in survival occurred for the rare childhood cancers of hepatoblastoma (5-year survival, 41% to 74%), gonadal germ cell tumors (5-year survival, 88% to 99% in boys; 80% to 90% in girls), and nasopharyngeal carcinoma (5-year survival, 47% to 84%). It is likely that the establishment of the Liver Tumor Study Group of the International Society of Pediatric Oncology26 had a major influence on outcomes for hepatoblastoma. For germ cell tumors, the improvement occurred mainly from the first to second period, coincident with the widespread introduction of platinum-based treatment regimens.27 For nasopharyngeal carcinoma, the increase occurred most markedly from the third to the fourth period (Table 2). The magnitude of the latter increase was unexpected. It could be a chance fluctuation, given the small number of patients, or perhaps be a result of earlier diagnosis after widespread introduction of imaging modalities.
For all cancers combined, 5-year survival increased smoothly over the study period in all regions (Fig 1). The improvement was greatest in the Eastern European countries and, importantly, was a feature of each of these countries individually.
The overall improvement may in part be attributed to improved supportive care in the intensive-care setting for acute infections and toxicity related to intensive chemotherapy and also for metabolic complications, life-threatening hemorrhage, and other effects of the disease on organ function.28 It is important to note that 5-year survival is not an exhaustive indicator of improvement in cancer prognosis; in particular, it does not correspond to the proportion of children who are cured. In fact, approximately 10% of 5-year survivors die during subsequent follow-up.29 This excess mortality is mainly a result of recurrence or progression of the primary tumor, but in approximately 20% of patients, the excess mortality is caused by a second malignancy.29 For this reason, the effects of the low-toxicity treatment regimens that aim to reduce side effects and the incidence of subsequent tumors have to be evaluated in terms of survival well beyond 5 years. The updating of follow-up in our patient sample will hopefully provide indications of long-term survival, cure rates, and incidence rates of second tumors.
We acknowledge the 10-year plus delay in the publication of our data. A continent-wide study, such as this study, requires considerable time for data reception, standardization, quality checks, and analysis. Future survival analysis projects will seek to reduce this time by encouraging registries to speed up data processing and quality checking and by using new methods to estimate recent survival figures. The latest European project, which analyzes survival data up to 2002, expects to publish in 2006.
Finally, the gratifying progress in pediatric cancer management that this study has illustrated, in combination with the increasing incidence of malignancies in children,30 will inevitably increase the prevalence of adults with a past history of childhood cancer. The health, psychological, and social needs of these people will need to be documented because they are likely to differ from those of the wider community.
Appendix
EUROCARE Working Group: Denmark: H.H. Storm (Department of Cancer Prevention and Documentation, Danish Cancer Society); Estonia: T. Aareleid (Estonian Cancer Registry); Finland: T. Hakulinen (Finnish Cancer Registry); France: G. Hedelin (Bas-Rhin Cancer Registry), I. Tron, E. Le Gall (Bretagne Childhood Cancer Registry), G. Launoy (Calvados Digestive Cancer Registry), J. Mace-Lesech (Calvados General Cancer Registry), J. Faivre (Cte d’Or Digestive Cancer Registry), G. Chaplain (Cte d’Or Gynecologic Cancer Registry), P.-M. Carli (Cte d’Or Malignant Hemopathies Registry), B. Lacour (Lorraine Childhood Cancer Registry), C. Berger, F. Freycon (Rhne-Alpes Childhood Registry), J. Estève (Biostatistics, Hospices Civils de Lyon, Universite Claude Bernard); Germany: P. Kaatsch (German Childhood Cancer Registry); Iceland: L. Tryggvadottir (Icelandic Cancer Registry); Italy: F. Berrino (Project Leader), C. Allemani, P. Baili, L. Ciccolallo, G. Gatta, A. Micheli, M. Sant, E. Taussig (Epidemiology Unit, Istituto Nazionale per lo Studio e la Cura dei Tumori, Milan), R. Capocaccia, E. Carrani, R. De Angelis, P. Roazzi, M. Santaquilani, A. Tavilla, F. Valente, A. Verdecchia (Istituto Superiore di Sanità, Rome), S. Ferretti (Ferrara Cancer Registry), P. Crosignani, G. Tagliabue (Lombardy Cancer Registry, Istituto Nazionale per lo Studio e la Cura dei Tumori, Milan), V. Ramazzotti, M.C. Cercato (Latina Cancer Registry, Istituto Nazionale per lo Studio e la Cura dei Tumori "Regina Elena," Rome), M. Vercelli, A. Orengo (Liguria Region Cancer Registry, National Cancer Research Institute, University of Genova), V. De Lisi, L. Serventi (Parma Cancer Registry), C. Magnani, G. Pastore (Piedmont Childhood Cancer Registry), L. Gafà, R. Tumino (Ragusa Cancer Registry), E. Paci, E. Crocetti (Tuscany Cancer Registry); Norway: F. Langmark, A. Andersen (Cancer Registry of Norway, Institute of Population- Based Cancer Research); Poland: J. Rachtan (Cracow Cancer Registry); Slovakia: I. Pleko, A. Obsitníková (National Cancer Registry of Slovakia); Slovenia: V. Pompe-Kirn (Cancer Registry of Slovenia); Spain: E. Ardanaz, C. Moreno (Navarra Cancer Registry), J. Galceran (Tarragona Cancer Registry), A. Torrella (Childhood Tumour Registry of Valencia), R. Peris-Bonet (National Registry of Childhood Tumours and Instituto López Piero, Valencia); Sweden: L. Barlow, T. Mller (Cancer Registry of Sweden); Switzerland: G. Jundt (Basel Cancer Registry), J.-M. Lutz, M. Usel (Geneva Cancer Registry); the Netherlands: J.W.W. Coebergh (Eindhoven Cancer Registry); United Kingdom: England and Wales: C. Stiller (Childhood Cancer Research Group, Oxford), M.P. Coleman (London School of Hygiene and Tropical Medicine); United Kingdom: Wales: J.A. Steward (Welsh Cancer Intelligence and Surveillance Unit); United Kingdom: Scotland: R. Black, D. Brewster (Cancer Information Group).
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
Acknowledgment
We thank Donald Ward for help with the English and Samba Sowe for editorial support. We also thank Roberto Luksch, Maura Massimino, and Andrea Ferrari for their helpful critique of the results.
NOTES
Supported by the EUROCARE-3 BIOMED-2 Programme Contract No. BMH4-CT98–3390 and the Compagnia di San Paolo, Torino, Italy; the United Kingdom Childhood Cancer Research Group has been supported by the Department of Health and Scottish Ministers of the United Kingdom.
The views expressed in this article are those of the authors and not necessarily those of the Department of Health and Scottish Ministers.
Authors’ disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Landis SH, Murray T, Bolden S, et al: Cancer statistics. CA Cancer J Clin 49:8-31, 1999
Gatta G, Corazziari I, Magnani C, et al: Childhood cancer survival in Europe. Ann Oncol 14:V119-V127 2003 (suppl 5)
Berrino F, Sant M, Verdecchia A, et al (eds): Survival of cancer patients in Europe, the EUROCARE Study. IARC Scientific Publications No. 132:1-463, 1995
Capocaccia R, Gatta G, Magnani C, et al (eds): Childhood cancer survival in Europe 1978-92. Eur J Cancer 37:671-816, 2001
Gatta G, Capocaccia R, Coleman MP, et al: Toward a comparison of survival in American and European cancer patients. Cancer 89:893-900, 2000
Gatta G, Capocaccia R, Coleman MP, et al: Childhood cancer survival in Europe and the United States. Cancer 95:1767-1772, 2002
Terracini B, Coebergh JW, Gatta G, et al: Childhood cancer survival in Europe, an overview. Eur J Cancer 37:810-816, 2001
WHO: International Classification of Diseases for Oncology (ICD-O) (ed 2). Geneva, Switzerland, WHO, 1990
Capocaccia R, Gatta G, Roazzi P, et al: The EUROCARE-3 database, methodology of data collection, standardisation, quality control and statistical analysis. Ann Oncol 14:V14-V27, 2003 (suppl 5)
Kramarova E, Stiller CA: The international classification of childhood cancer. Int J Cancer 68:759-765, 1996
Magnani C, Gatta G, Corazziari I, et al: Childhood malignancies in the EUROCARE study: The database and the methods of survival analysis. Eur J Cancer 37:678-686, 2001
Hakulinen T, Abeywickrama KH: A computer program package for relative survival analysis. Comput Programs Biomed 19:197-207, 1985
Pui CH: Childhood leukemias. N Engl J Med 332:1618-1630, 1995
Gustafsson G, Schmiegelow K, Forestier E, et al: Improving outcome through two decades in childhood ALL in the Nordic countries: The impact of high-dose methotrexate in the reduction of CNS irradiation—Nordic Society of Pediatric Haematology and Oncology (NOPHO). Leukemia 14:2267-2275, 2000
Hann IM, Stevens RF, Goldstone AH, et al: Randomized comparison of DAT versus ADE as induction chemotherapy in children and younger adults with acute myeloid leukemia: Results of the Medical Research Council's 10th AML trial (MRC AML10), Adult and Childhood Leukaemia Working Parties of the Medical Research Council. Blood 89:2311-2318, 1997
Sandlund JT, Downing JR, Crist WM: Non-Hodgkin's lymphoma in childhood. N Engl J Med 334:1238-1248, 1996
Patte C: Non-Hodgkin's lymphoma. Eur J Cancer 34:359-362, 1998
Prados MD, Berger MS, Wilson CB: Primary central nervous system tumors, advances in knowledge and treatment. CA Cancer J Clin 48:331-360, 1998
Bailey CC, Gnekow A, Wellek S, et al: Prospective randomised trial of chemotherapy given before radiotherapy in childhood medulloblastoma: International Society of Paediatric Oncology (SIOP) and the (German) Society of Paediatric Oncology (GPO)—SIOP II. Med Pediatr Oncol 25:166-178, 1995
Taylor RE, Bailey CC, Robinson K, et al: Results of a randomized study of preradiation chemotherapy versus radiotherapy alone for nonmetastatic medulloblastoma: The International Society of Paediatric Oncology/United Kingdom Children's Cancer Study Group PNET-3 Study. J Clin Oncol 21:1581-1591, 2003
Kaatsch P, Rickert CH, Kuhl J, et al: Population-based epidemiologic data on brain tumors in German children. Cancer 92:3155-3164, 2001
Pritchard J, Imeson J, Barnes J, et al: Results of the United Kingdom Children's Cancer Study Group first Wilms' Tumor Study. J Clin Oncol 13:124-133, 1995
Mitchell C, Jones PM, Kelsey A, et al: The treatment of Wilms' tumour: Results of the United Kingdom Children's Cancer Study Group (UKCCSG) second Wilms' tumour study. Br J Cancer 83:602-608, 2000
Graf N, Tournade MF, de Kraker J: The role of preoperative chemotherapy in the management of Wilms' tumor: The SIOP studies—International Society of Pediatric Oncology. Urol Clin North Am 27:443-454, 2000
Zoller G, Lakomek M, Ringert RH: Wilm's tumor 1992: State of research and results of current therapeutic concepts. Urologe A 31:360-367, 1992
Brown J, Perilongo G, Shafford E, et al: Pretreatment prognostic factors for children with hepatoblastoma: Results from the International Society of Paediatric Oncology (SIOP) study SIOPEL 1. Eur J Cancer 36:1418-1425, 2000
Pinkerton CR, Pritchard J, Spitz L: High complete response rate in children with advanced germ cell tumors using cisplatin-containing combination chemotherapy. J Clin Oncol 4:194-199, 1986
Keengwe IN, Stansfield F, Eden OB, et al: Paediatric oncology and intensive care treatments: Changing trends. Arch Dis Child 80:553-555, 1999
Moller TR, Garwicz S, Barlow L, et al: Decreasing late mortality among five-year survivors of cancer in childhood and adolescence: A population-based study in the Nordic countries. J Clin Oncol 19:3173-3181, 2001
Steliarovà-Foucher E, Stiller C, Kaatsch P, et al: Geographical patterns and time trends of cancer incidence and survival among children and adolescents since 1970s: The ACCIS project—An epidemiological study. Lancet 364:2097-2105, 2004(Gemma Gatta, Riccardo Cap)
Pediatric Unit, Istituto Nazionale per lo Studio e la Cura dei Tumori, Milan
Laboratory of Epidemiology, Istituto Superiore di Sanità, Rome, Italy
Childhood Cancer Research Group, University of Oxford, Oxford, United Kingdom
Institute for Medical Biostatistics, Epidemiology and Informatics, German Childhood Cancer Registry, University Mainz, Mainz, Germany
ABSTRACT
PURPOSE: EUROCARE collected data from population-based cancer registries in 20 European countries. We used this data to compare childhood cancer survival time trends in Europe.
PATIENTS AND METHODS: Survival in 44,129 children diagnosed under the age of 15 years during 1983 to 1994 was analyzed. Sex- and age-adjusted 5-year survival trends for 10 common cancers and for all cancers combined were estimated for five regions (West Germany, the United Kingdom, Eastern Europe, Nordic countries, and West and South Europe) and Europe as a whole. Europe-wide trends for 14 rare cancers were estimated.
RESULTS: For all cancers combined, 5-year survival increased from 65% for diagnoses in 1983 to 1985 to 75% in 1992 to 1994. Survival improved for all individual cancers except melanoma, osteosarcoma, and thyroid carcinoma; although for retinoblastoma, chondrosarcoma, and fibrosarcoma, improvements were not significant. The most marked improvements (50% to 66%) occurred in Eastern Europe. For common cancers, the greatest improvements were for leukemia and lymphomas, with risk of dying reducing significantly by 5% to 6% per year. Survival for CNS tumors improved significantly from 57% to 65%, with risk reducing by 3% per year. Risk reduced by 4% per year for neuroblastoma and 3% per year for Wilms’ tumor and rhabdomyosarcoma. The survival gap between regions reduced over the period, particularly for acute nonlymphocytic leukemia, CNS tumors, and rhabdomyosarcoma. For rare Burkitt’s lymphoma, hepatoblastoma, gonadal germ cell tumors, and nasopharyngeal carcinoma, risk reductions were at least 10% per year.
CONCLUSION: These gratifying improvements in survival can often be plausibly related to advances in treatment. The prevalence of European adults with a history of childhood cancer will inevitably increase.
INTRODUCTION
Survival for childhood cancers has improved dramatically over the last half century. Before 1950, almost all children who developed cancer died. At present, approximately three out of four children who are diagnosed with cancer can be cured.1 The most recently available (1990 to 1994) survival data on European cancer patients2 provide an opportunity to assess progress in childhood cancer survival over a 12-year period. Cancer survival in Europe data are available as a result of EUROCARE, a cooperative, cancer registry-based project.3,4 EUROCARE childhood survival data have been compared with data available for the United States and found to be fairly similar.5 This is not the case for adult cancer survival, which is generally better in the United States.6 The European publications documented large differences in childhood cancer survival between different European populations and also revealed improvements in survival between 1978 and 1989.7 However, survival trends were reported for Europe as a whole,7 and no studies comparing trends between European countries or regions have been performed.
This study analyzes 5-year survival in European children with cancers diagnosed between 1983 and 1994 and incorporates data from more cancer registries and countries than were included in earlier EUROCARE studies. The aims of this study were to compare survival trends for important cancer types between European countries and regions, to obtain an idea of the impact of new diagnostic and treatment protocols introduced in 1980 to 1995, and to investigate whether childhood survival differences were reduced. This study also takes advantage of the large EUROCARE incidence database to present prognostic trends in Europe for the rarest childhood cancers.
PATIENTS AND METHODS
We analyzed survival in 44,129 European children diagnosed with cancer under the age of 15 years during the period of 1983 to 1994. We excluded patients diagnosed from 1978 to 1982 because several registries did not cover these years. All children with a malignant neoplasm as defined by second edition International Classification of Diseases for Oncology behavior code 3 or higher were included.8 The patients were contributed by 27 population-based cancer registries in 17 countries participating in EUROCARE-3.9 All the registries included in this analysis provided data for the whole period (1983 to 1994). The database includes data from five specialized (childhood) cancer registries (England and Wales; Germany, only West Germany in this study; Piedmont in Italy; Lorraine in France; and Valencia in Spain). Most cancers, with a single exception, were classified according to the 12 diagnostic groups in the International Classification of Childhood Cancers (ICCC).10 The exception regarded groups of intracranial and intraspinal neoplasms (International Classification of Childhood Cancers III and Xa); for these cancers, only malignant tumors, as specified by the EUROCARE protocol, were included. Furthermore, because astrocytoma data from the Finnish registry were not coded using International Classification of Diseases for Oncology morphologic codes, these data were excluded.
Table 1 presents a general description of the data analyzed, showing sample size, demographic characteristics, and main data quality indicators by country and registry. The registries of Denmark, Finland, England and Wales, Estonia, Iceland, Norway, Scotland, Slovakia, Slovenia, Sweden, and West Germany cover the entire populations of their countries. Other countries are represented by one or more local or regional registries. Table 1 also lists the proportions of boys (56%), children aged less than 5 years (45%), microscopically verified cancers (95%; between-country range, 88% to 100%), and patients lost to follow-up (1.5%). All patients were observed until at least the end of 1998. However, a number of patients were censored as alive without having the 5-year follow-up specified by the protocol, having been diagnosed less than 5 years before the end of follow-up. For most registries, except those of France, West Germany, Cracow, Eindhoven, and Geneva, the proportions of patients with less than 4 years of follow-up were less than 4%.
Both histologically verified and nonverified patients were included, but patients known to registries by death certificate only (DCO) or autopsy report only were excluded (337 patients, 0.7%). The number of patients classified in unspecified categories within each major diagnostic group was also low (1,558 patients overall, 3.5%). More detailed information on the EUROCARE database is available elsewhere.9,11
Observed survival was calculated by the actuarial method. Relative survival was also calculated12 but is not presented here because it corresponds closely to observed survival in young people because deaths as a result of competing risks are rare.
The survival trend data are presented for all cancers combined and for 24 selected malignancies. Survival for all cancers combined was adjusted for patient mix (calculated by weighting individual cancer survival figures with the number of patients in each cancer category as a proportion of the whole data set). For this purpose, rare tumors were pooled into a category termed other, which contained approximately 20% of all cancers; astrocytoma and primitive neuroectodermal tumor (PNET) were included in the general CNS tumors category.
Survival estimates for the 10 major types of tumors and for all malignancies combined were compared between populations. We defined five European regions as follows: Denmark, Finland, Iceland, Norway, and Sweden (Nordic countries); England, Wales, and Scotland (the United Kingdom); Estonia, Poland, Slovakia, and Slovenia (Eastern Europe); West Germany, which was kept separate because of the large number of patients; and France, Italy, the Netherlands, Spain, and Switzerland (Western and Southern Europe).
Five-year survival was compared for four 3-year diagnosis periods (1983 to 1985, 1986 to 1988, 1989 to 1991, and 1992 to 1994). Because survival for many of the cancers depends on age and sex and the distribution of these variables varies across countries and over time, survival estimates were adjusted for these variables. To avoid problems caused by a few cases in individual categories when cross stratifying by sex, age, region, and period, a model approach was used instead of the standard direct adjustment method. A proportional hazards model was fitted to individual survival data for the 10 cancer categories (the common cancers plus all cancers combined). Age (three classes: 0 to 4, 5 to 9, and 10 to 14 years for all tumors except neuroblastoma, for which the age division was < 1, 1 to 4, and 5 to 14 years), sex, period (n = 4), region (n = 5), and a period x region interaction term were included as covariates. The model was then used to calculate 5-year survival for each region, for each time period, and for the average sex and age distribution of the whole sample. Linear trends in survival during the analysis period (1983 to 1994) were estimated by a similar proportional hazards model using year of diagnosis as a continuous covariate instead of categoric periods. The parameter relative risk (RR), which was estimated from the continuous proportional hazards model, is the RR of dying of children diagnosed in a given year compared with those diagnosed a year previously and averaged over the whole study period of 1983 to 1994. We also refer to the reduction in risk of dying; this is 1 RR and is expressed as a percentage.
For the European pool, the following two sets of 5-year survival figures were calculated: (1) crude and (2) region population weighted and sex and age adjusted. The latter were estimated, for each time period, as the weighted average of the corresponding region-specific model-based survival values using the childhood population counts in each region as weights. The weighted and adjusted European survival figures are considered in the Discussion but not presented in Table 2.
RESULTS
Table 2 lists 5-year survival trends for 24 tumor types, all CNS tumors, and all cancers combined, estimated from the entire European pool. Age- and sex-adjusted survival trends for all cancer combined and for the 10 most common cancers for the five European regions are shown in Figures 1 and 2. The corresponding linear trends of RR of dying are listed in Table 3. We present the results by cancer type.
For all cancers combined, 5-year survival increased smoothly from 65% in children diagnosed in 1983 to 1985 to 75% in 1992 to 1994, with the risk of dying reduced by an average of 5% per year (Table 2). The rate of increase in survival was similar in all regions (Fig 1). Survival reached approximately 75% in 1992 to 1994, except in Eastern Europe, where it was 66%.
With few exceptions, 5-year survival improved for all the types of cancer (Table 2); the improvements were statistically significant for all common tumors and several rare tumors. For thyroid carcinoma, melanoma and, osteosarcoma, survival was stable. The Eastern European countries experienced the greatest improvement in survival, although improvement was not always statistically significant (Fig 2 and Table 3).
Considering the major cancers, greatest overall improvements were seen for leukemias and lymphomas (Table 2). Five-year survival for lymphoid leukemia was less than 80% in 1983 to 1985 but had surpassed 80% in the Nordic countries, Western and Southern Europe, the United Kingdom, and West Germany in 1992 to 1994 (Fig 2); notwithstanding an impressive improvement in Eastern European countries, the gap between these countries and other regions of Europe did not diminish.
For acute nonlymphocytic leukemia (ANLL), significant survival increases occurred in the United Kingdom and Eastern Europe, where the risk of dying reduced by 8% to 9% per year across the study period (Table 3). For all regions, 5-year survival was less than 50% in 1983 to 1985 but had increased to greater than 50% in the United Kingdom, West Germany, and the Nordic countries by 1992 to 1994 (Fig 2). In Western and Southern Europe, however, there was no change in survival for ANLL over the study period.
Five-year survival for Hodgkin’s disease (HD) was greater than 90% over the entire period and remained stable in most regions (Fig 2); however, survival in the United Kingdom improved significantly, where the risk of dying decreased by approximately 10% every year (Table 3). For non-Hodgkin’s lymphoma, progress was marked in Europe, with the RR of death decreasing by 6% per year over the whole period (Table 2); in West Germany, the risk of dying decreased significantly by 10% per year (Table 3).
Progress for solid tumors was less marked than for hematologic malignancies. Survival for CNS tumors improved significantly, with the risk of dying reduced by 3% per year (Table 2). The improvement was significant for the United Kingdom, West Germany, the Nordic countries, and Eastern Europe (Table 3). In Eastern Europe, 5-year survival for CNS tumors improved significantly from 47% to 62% over the study period, with risk of dying reduced by 4% per year; the improvements were particularly marked in Estonia and Slovenia. Progress occurred first in the Nordic countries and West Germany, followed by Western and Southern Europe and, subsequently, the United Kingdom (Fig 2). The survival gap between regions reduced in 1992 to 1994 (Fig 2).
For PNET medulloblastoma and astrocytoma, the risk of dying was reduced significantly by 3% per year (Table 2), and for astrocytoma, the survival gap between regions had reduced by 1992 to 1994 (Fig 2). For PNET and medulloblastoma, West Germany and the Nordic countries showed similar survival trends, which were characterized by overall improvement but a dip in survival in 1989 to 1991; a similar dip was observed for Western and Southern Europe at the end of the period (Fig 2).
For neuroblastoma, the risk of dying was reduced by 4% per year (Table 2), with significant improvements for West Germany and the United Kingdom (Table 3) and a reduction in the between-region survival gap. During 1992-1994, West Germany and Western and Southern Europe had 5-year survival rates that approached 70%, whereas for all the other regions, survival was similar and approximately 60% (Fig 2).
For Wilms’ tumors, the risk of dying decreased by 3% per year overall. In the Nordic countries, 5-year survival improved significantly from 81% to 93% (Fig 2); marked improvements also occurred in West Germany and Eastern Europe. In Eastern Europe, survival increased from 59% in 1983 to 1985 to 75% in 1992 to 1994. Survival in the United Kingdom and Western and Southern Europe remained stable at approximately 80% over the period.
For rhabdomyosarcoma, the overall risk of dying decreased significantly by approximately 3% per year (Table 2). The decrease was greatest (but not significant) in the Nordic countries (5% per year), Eastern Europe (6% per year), and Western and Southern Europe (6% per year). The gap between regions was reduced markedly over the period.
For rare childhood cancers (less than 1,500 patients), 5-year survival improved generally (Table 2), except for melanoma, osteosarcoma, and thyroid carcinoma, for which survival remained stable or decreased slightly. Progress was remarkable for Burkitt’s lymphoma, ependymoma, hepatoblastoma, Ewing’s sarcoma, gonadal germ cell tumors, and nasopharyngeal carcinoma. Five-year survival for ependymoma, hepatoblastoma, Ewing’s sarcoma, and nasopharyngeal carcinoma increased from less than 50% in 1983 to 1985 to more than 50% in 1992 to 1994. Survival for retinoblastoma also improved (nonsignificantly), although survival was already high at the outset.
DISCUSSION
Our major findings are as follows. First, from 1983 to 1994, survival improved for all childhood cancers combined in Europe; the risk of dying was reduced by 5% per year. Second, the greatest survival improvements among the common cancers were for leukemia and lymphoma. Third, for all cancers combined and some major tumors, the most marked survival improvements occurred in Eastern Europe. Although Eastern European survival was always lower than in the rest of Europe, the between-region gap reduced over the study period, particularly for ANLL, CNS tumors, astrocytoma, and rhabdomyosarcoma.
Differences in data quality and comparability between registries have to be considered as possible sources of bias that may detract from the validity of these findings. The main indicators of data quality are the proportion of DCO patients, proportion of patients with microscopically verified cancer, and proportion of patients lost to follow-up. In this study, DCO patients were rare except in Slovakia, Latina, and Estonia (Table 1). Furthermore, 95% of patients were confirmed microscopically. Microscopic confirmation is particularly important for childhood tumors, which are primarily classified by histologic type. It is noteworthy that few patients (1.5% overall) were lost to follow-up. All registries except Eindhoven, West Germany, and the French registries observed their patients for at least 5 years; in France, legislation makes it difficult to collect life status data.11 It is also noteworthy that data quality indicators did not vary greatly over the study period. The only exception was that the proportion of patients alive with follow-up of less than 4 years increased from 1.7% in 1983 to 1985 to 7.8% in 1992 to 1994.
This concentration, in 1992 to 1994, of patients alive with follow-up of less than 4 years could bias trend estimates if these censored patients for each diagnostic group, sex, and age group had different prognoses from patients not censored. Therefore, we analyzed the distribution of these patients according to diagnostic group. To do this, we assigned a 5-year survival probability for each patient equal to the mean observed survival according to tumor category. We found that these assigned probabilities were 76.6% for patients censored at 0 months and 76.6% for patients lost between 1 and 47 months. These figures are just greater than the mean survival for all patients (74.9%), as one would expect because many censored patients were from West Germany. Thus, these censored patients did not contain an excess of poor prognosis tumors.
We also examined all censored patients. We found, among patients in 1992 to 1994, that 73% had at least 4 but less than 5 years of follow-up. These patients did not influence the survival trends with time because, for all cancers, most improvements in survival occurred in the first few years after diagnosis. In particular, the differences in survival after 4 and after 5 years in both the latest and the penultimate diagnosis periods were absent or minimal. This analysis indicates that no major bias is a result of high censoring of patients diagnosed in the latest period.
Although our findings are derived from a large sample from 17 countries, the sample is not representative of the totality of European pediatric patients. Western and Southern Europe, including Spain, France, Switzerland, Italy, and the Netherlands, provided less than 12% of the total patients in their respective countries and are underrepresented in the pooled European estimates. Population-weighted survival estimates9,11 for tumor type were calculated in this study but did not differ sufficiently from the crude pooled survival estimates, and this is explained by the fact that survival in the most populous European region (Western and Southern Europe) was always close to the average pooled survival levels.
We defined five broad European regions. This grouping was necessary because of the low incidence of childhood cancers. The grouping was partly geographic but also based on broad similarity of survival figures in the countries grouped together, both in the present analysis and in past study.2 We included Denmark with the Nordic countries even though its survival was somewhat low because increases in survival for all cancers combined and for the major childhood tumors were similar in all Nordic countries.
Improvements in survival over the study period can often be related to advances in treatments. For example, therapies for lymphoid leukemia were intensified during this time; multidrug regimens were introduced, and therapy intensities were stratified according to patient risk group.13 The greatest improvements in lymphoid leukemia survival occurred in the Nordic countries, and this is probably related to a major initiative by the Nordic Society of Pediatric Hematology and Oncology. In 1981, the Nordic Society of Pediatric Hematology and Oncology began registering all lymphoid leukemias in children under 15 years of age in the five Nordic countries to monitor the impact of high-dose methotrexate combined with reduction of CNS irradiation and to encourage the use uniform protocols.14 This population-based initiative found that 5-year event-free survival increased from 56.5% in the early 1980s to 77.6% in the 1990s. Lack of appropriate chemotherapy agents may be an important reason for low survival for lymphoid leukemia in Eastern Europe.
The marked increase in ANLL survival in the United Kingdom over the study period may reflect widespread participation in the Medical Research Council’s AML10 trial, which was carried out in 1988 to 1995 and which had survival results that were better than those of all previous large-scale trials in this disease.15 Significant progress for ANLL also occurred in Eastern Europe overall (and in all individual countries except Slovakia), with the risk of dying reduced by approximately 8% per year. Much more modest improvements were seen in West Germany and the Nordic countries, where survival was high at the outset. No improvement occurred in Western and Southern Europe overall.
Survival for HD improved by an impressive 10% in the United Kingdom over the study period, although an explanation similar to that for ANLL is not apparent. There have been attempts to minimize the long-term sequelae of HD treatment in children by reducing the dose of the most toxic component (radiotherapy) of multimodality treatments. This may be the reason for the survival decrease seen for this disease in 1992 to 1994 in the Nordic countries (Fig 2); note, however, that both the number of patients (n = 63) and the number of deaths (n = 6) were small. Survival figures over periods longer than 5 years may provide better indications of the utility of the low-dose approach.
Therapy for non-Hodgkin’s lymphoma, like that for leukemia, became increasingly based on multidrug regimens over the study period.16 However, the only significant trend to increased survival (from 75% to 89%) was seen in West Germany, plausibly because of the activity of the German-Austrian Cooperative Groups.17
For CNS tumors, computed tomography, which was introduced in the 1970s, and magnetic resonance imaging, which was disseminated widely in the 1980s, have become standard tools for disease imaging and evaluation. Neuroimaging is particularly important as a guide to the extent of resection.18 Our analysis found that survival for CNS tumors increased first in the Nordic countries and West Germany and then in Western and Southern Europe, the United Kingdom, and finally Eastern Europe (Fig 2). This pattern may reflect staggered implementation of magnetic resonance imaging to assess these tumors. Use of chemotherapy for these tumors also increased over the period, with the aim of avoiding, delaying, or reducing radiation and hence maintaining quality of life. For PNET and medulloblastoma, trials to reduce the radiotherapy dose19 could be responsible for the reduction of survival found in some regions (Fig 2). The updated protocols recommend that radiotherapy be combined with high-dose chemotherapy after surgery.20
We found significant improvements in survival for PNET and medulloblastoma (from 48% to 60%) and astrocytoma (from 72% to 80%) in West Germany over the study period. However, both incidence and follow-up in West Germany for childhood CNS cancers were less complete than for the other childhood cancers, and the favorable survival trend should be considered cautiously. In Germany, many children with CNS tumors do not receive chemotherapy and, thus, are not seen by pediatric oncologists who are the main providers of incidence notifications to cancer registries. Furthermore, the completeness of reported incidence and the quality of follow-up for childhood CNS tumors has varied over time in relation to the number of children included in clinical trials; the start of a large clinical trial on CNS tumors in the mid-1980s coincided with a considerable increase in the completeness of registration.21 Five clinical trials on primary CNS tumors were ongoing in 2001 in Germany, involving 82% of reported incident cases of CNS tumors. Yet, this proportion is low compared with the overall figure of 92% of incident cases of childhood malignancies recruited to trials.21
Prognosis for neuroblastoma depends on age and disease stage at diagnosis; the disseminated, rapidly expanding disease often found in older children is particularly difficult to treat. We found that overall 5-year survival for neuroblastoma improved significantly from 42% to 54% in older children (1 to 14 years of age) characterized by poor prognosis; whereas for infants with good prognosis, survival increased only modestly (from 81% to 84%). These findings suggest an effective improvement in the therapeutic modalities used to treat this tumor.
A major improvement in survival for Wilms’ tumor occurred before the present study period. We found that survival from 1983 to 1994 remained stable in the United Kingdom and Western and Southern Europe but improved in the Nordic countries, West Germany, and Eastern European countries. In the United Kingdom, national trials were ongoing throughout this period.22,23 Also, for most of this period, joint International Society of Pediatric Oncology-German Society of Pediatric Oncology studies24 were being conducted principally in France, Germany, Italy, Spain, Netherlands, Sweden, Norway, Denmark, and Slovenia. An aim of these studies was to reduce the chemotherapy, radiotherapy, or both administered to patients who had characteristics associated with favorable outcome to limit long-term sequelae.25 For these reasons, any progress in Wilms’ tumor for patients diagnosed in 1983 to 1994 may become evident later. However, a survival improvement occurred in the Nordic countries over the study period, which may have been a result of wider recruitment of children in the clinical trials conducted in these countries.
For rhabdomyosarcoma, survival differences between the European populations were reduced effectively to zero from 1983 to 1994. Survival in the Eastern European countries increased markedly from the third to the fourth period as a result of the 100% survival in the eight patients from Cracow and Slovenia.
Important improvements in survival occurred for the rare childhood cancers of hepatoblastoma (5-year survival, 41% to 74%), gonadal germ cell tumors (5-year survival, 88% to 99% in boys; 80% to 90% in girls), and nasopharyngeal carcinoma (5-year survival, 47% to 84%). It is likely that the establishment of the Liver Tumor Study Group of the International Society of Pediatric Oncology26 had a major influence on outcomes for hepatoblastoma. For germ cell tumors, the improvement occurred mainly from the first to second period, coincident with the widespread introduction of platinum-based treatment regimens.27 For nasopharyngeal carcinoma, the increase occurred most markedly from the third to the fourth period (Table 2). The magnitude of the latter increase was unexpected. It could be a chance fluctuation, given the small number of patients, or perhaps be a result of earlier diagnosis after widespread introduction of imaging modalities.
For all cancers combined, 5-year survival increased smoothly over the study period in all regions (Fig 1). The improvement was greatest in the Eastern European countries and, importantly, was a feature of each of these countries individually.
The overall improvement may in part be attributed to improved supportive care in the intensive-care setting for acute infections and toxicity related to intensive chemotherapy and also for metabolic complications, life-threatening hemorrhage, and other effects of the disease on organ function.28 It is important to note that 5-year survival is not an exhaustive indicator of improvement in cancer prognosis; in particular, it does not correspond to the proportion of children who are cured. In fact, approximately 10% of 5-year survivors die during subsequent follow-up.29 This excess mortality is mainly a result of recurrence or progression of the primary tumor, but in approximately 20% of patients, the excess mortality is caused by a second malignancy.29 For this reason, the effects of the low-toxicity treatment regimens that aim to reduce side effects and the incidence of subsequent tumors have to be evaluated in terms of survival well beyond 5 years. The updating of follow-up in our patient sample will hopefully provide indications of long-term survival, cure rates, and incidence rates of second tumors.
We acknowledge the 10-year plus delay in the publication of our data. A continent-wide study, such as this study, requires considerable time for data reception, standardization, quality checks, and analysis. Future survival analysis projects will seek to reduce this time by encouraging registries to speed up data processing and quality checking and by using new methods to estimate recent survival figures. The latest European project, which analyzes survival data up to 2002, expects to publish in 2006.
Finally, the gratifying progress in pediatric cancer management that this study has illustrated, in combination with the increasing incidence of malignancies in children,30 will inevitably increase the prevalence of adults with a past history of childhood cancer. The health, psychological, and social needs of these people will need to be documented because they are likely to differ from those of the wider community.
Appendix
EUROCARE Working Group: Denmark: H.H. Storm (Department of Cancer Prevention and Documentation, Danish Cancer Society); Estonia: T. Aareleid (Estonian Cancer Registry); Finland: T. Hakulinen (Finnish Cancer Registry); France: G. Hedelin (Bas-Rhin Cancer Registry), I. Tron, E. Le Gall (Bretagne Childhood Cancer Registry), G. Launoy (Calvados Digestive Cancer Registry), J. Mace-Lesech (Calvados General Cancer Registry), J. Faivre (Cte d’Or Digestive Cancer Registry), G. Chaplain (Cte d’Or Gynecologic Cancer Registry), P.-M. Carli (Cte d’Or Malignant Hemopathies Registry), B. Lacour (Lorraine Childhood Cancer Registry), C. Berger, F. Freycon (Rhne-Alpes Childhood Registry), J. Estève (Biostatistics, Hospices Civils de Lyon, Universite Claude Bernard); Germany: P. Kaatsch (German Childhood Cancer Registry); Iceland: L. Tryggvadottir (Icelandic Cancer Registry); Italy: F. Berrino (Project Leader), C. Allemani, P. Baili, L. Ciccolallo, G. Gatta, A. Micheli, M. Sant, E. Taussig (Epidemiology Unit, Istituto Nazionale per lo Studio e la Cura dei Tumori, Milan), R. Capocaccia, E. Carrani, R. De Angelis, P. Roazzi, M. Santaquilani, A. Tavilla, F. Valente, A. Verdecchia (Istituto Superiore di Sanità, Rome), S. Ferretti (Ferrara Cancer Registry), P. Crosignani, G. Tagliabue (Lombardy Cancer Registry, Istituto Nazionale per lo Studio e la Cura dei Tumori, Milan), V. Ramazzotti, M.C. Cercato (Latina Cancer Registry, Istituto Nazionale per lo Studio e la Cura dei Tumori "Regina Elena," Rome), M. Vercelli, A. Orengo (Liguria Region Cancer Registry, National Cancer Research Institute, University of Genova), V. De Lisi, L. Serventi (Parma Cancer Registry), C. Magnani, G. Pastore (Piedmont Childhood Cancer Registry), L. Gafà, R. Tumino (Ragusa Cancer Registry), E. Paci, E. Crocetti (Tuscany Cancer Registry); Norway: F. Langmark, A. Andersen (Cancer Registry of Norway, Institute of Population- Based Cancer Research); Poland: J. Rachtan (Cracow Cancer Registry); Slovakia: I. Pleko, A. Obsitníková (National Cancer Registry of Slovakia); Slovenia: V. Pompe-Kirn (Cancer Registry of Slovenia); Spain: E. Ardanaz, C. Moreno (Navarra Cancer Registry), J. Galceran (Tarragona Cancer Registry), A. Torrella (Childhood Tumour Registry of Valencia), R. Peris-Bonet (National Registry of Childhood Tumours and Instituto López Piero, Valencia); Sweden: L. Barlow, T. Mller (Cancer Registry of Sweden); Switzerland: G. Jundt (Basel Cancer Registry), J.-M. Lutz, M. Usel (Geneva Cancer Registry); the Netherlands: J.W.W. Coebergh (Eindhoven Cancer Registry); United Kingdom: England and Wales: C. Stiller (Childhood Cancer Research Group, Oxford), M.P. Coleman (London School of Hygiene and Tropical Medicine); United Kingdom: Wales: J.A. Steward (Welsh Cancer Intelligence and Surveillance Unit); United Kingdom: Scotland: R. Black, D. Brewster (Cancer Information Group).
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
Acknowledgment
We thank Donald Ward for help with the English and Samba Sowe for editorial support. We also thank Roberto Luksch, Maura Massimino, and Andrea Ferrari for their helpful critique of the results.
NOTES
Supported by the EUROCARE-3 BIOMED-2 Programme Contract No. BMH4-CT98–3390 and the Compagnia di San Paolo, Torino, Italy; the United Kingdom Childhood Cancer Research Group has been supported by the Department of Health and Scottish Ministers of the United Kingdom.
The views expressed in this article are those of the authors and not necessarily those of the Department of Health and Scottish Ministers.
Authors’ disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Landis SH, Murray T, Bolden S, et al: Cancer statistics. CA Cancer J Clin 49:8-31, 1999
Gatta G, Corazziari I, Magnani C, et al: Childhood cancer survival in Europe. Ann Oncol 14:V119-V127 2003 (suppl 5)
Berrino F, Sant M, Verdecchia A, et al (eds): Survival of cancer patients in Europe, the EUROCARE Study. IARC Scientific Publications No. 132:1-463, 1995
Capocaccia R, Gatta G, Magnani C, et al (eds): Childhood cancer survival in Europe 1978-92. Eur J Cancer 37:671-816, 2001
Gatta G, Capocaccia R, Coleman MP, et al: Toward a comparison of survival in American and European cancer patients. Cancer 89:893-900, 2000
Gatta G, Capocaccia R, Coleman MP, et al: Childhood cancer survival in Europe and the United States. Cancer 95:1767-1772, 2002
Terracini B, Coebergh JW, Gatta G, et al: Childhood cancer survival in Europe, an overview. Eur J Cancer 37:810-816, 2001
WHO: International Classification of Diseases for Oncology (ICD-O) (ed 2). Geneva, Switzerland, WHO, 1990
Capocaccia R, Gatta G, Roazzi P, et al: The EUROCARE-3 database, methodology of data collection, standardisation, quality control and statistical analysis. Ann Oncol 14:V14-V27, 2003 (suppl 5)
Kramarova E, Stiller CA: The international classification of childhood cancer. Int J Cancer 68:759-765, 1996
Magnani C, Gatta G, Corazziari I, et al: Childhood malignancies in the EUROCARE study: The database and the methods of survival analysis. Eur J Cancer 37:678-686, 2001
Hakulinen T, Abeywickrama KH: A computer program package for relative survival analysis. Comput Programs Biomed 19:197-207, 1985
Pui CH: Childhood leukemias. N Engl J Med 332:1618-1630, 1995
Gustafsson G, Schmiegelow K, Forestier E, et al: Improving outcome through two decades in childhood ALL in the Nordic countries: The impact of high-dose methotrexate in the reduction of CNS irradiation—Nordic Society of Pediatric Haematology and Oncology (NOPHO). Leukemia 14:2267-2275, 2000
Hann IM, Stevens RF, Goldstone AH, et al: Randomized comparison of DAT versus ADE as induction chemotherapy in children and younger adults with acute myeloid leukemia: Results of the Medical Research Council's 10th AML trial (MRC AML10), Adult and Childhood Leukaemia Working Parties of the Medical Research Council. Blood 89:2311-2318, 1997
Sandlund JT, Downing JR, Crist WM: Non-Hodgkin's lymphoma in childhood. N Engl J Med 334:1238-1248, 1996
Patte C: Non-Hodgkin's lymphoma. Eur J Cancer 34:359-362, 1998
Prados MD, Berger MS, Wilson CB: Primary central nervous system tumors, advances in knowledge and treatment. CA Cancer J Clin 48:331-360, 1998
Bailey CC, Gnekow A, Wellek S, et al: Prospective randomised trial of chemotherapy given before radiotherapy in childhood medulloblastoma: International Society of Paediatric Oncology (SIOP) and the (German) Society of Paediatric Oncology (GPO)—SIOP II. Med Pediatr Oncol 25:166-178, 1995
Taylor RE, Bailey CC, Robinson K, et al: Results of a randomized study of preradiation chemotherapy versus radiotherapy alone for nonmetastatic medulloblastoma: The International Society of Paediatric Oncology/United Kingdom Children's Cancer Study Group PNET-3 Study. J Clin Oncol 21:1581-1591, 2003
Kaatsch P, Rickert CH, Kuhl J, et al: Population-based epidemiologic data on brain tumors in German children. Cancer 92:3155-3164, 2001
Pritchard J, Imeson J, Barnes J, et al: Results of the United Kingdom Children's Cancer Study Group first Wilms' Tumor Study. J Clin Oncol 13:124-133, 1995
Mitchell C, Jones PM, Kelsey A, et al: The treatment of Wilms' tumour: Results of the United Kingdom Children's Cancer Study Group (UKCCSG) second Wilms' tumour study. Br J Cancer 83:602-608, 2000
Graf N, Tournade MF, de Kraker J: The role of preoperative chemotherapy in the management of Wilms' tumor: The SIOP studies—International Society of Pediatric Oncology. Urol Clin North Am 27:443-454, 2000
Zoller G, Lakomek M, Ringert RH: Wilm's tumor 1992: State of research and results of current therapeutic concepts. Urologe A 31:360-367, 1992
Brown J, Perilongo G, Shafford E, et al: Pretreatment prognostic factors for children with hepatoblastoma: Results from the International Society of Paediatric Oncology (SIOP) study SIOPEL 1. Eur J Cancer 36:1418-1425, 2000
Pinkerton CR, Pritchard J, Spitz L: High complete response rate in children with advanced germ cell tumors using cisplatin-containing combination chemotherapy. J Clin Oncol 4:194-199, 1986
Keengwe IN, Stansfield F, Eden OB, et al: Paediatric oncology and intensive care treatments: Changing trends. Arch Dis Child 80:553-555, 1999
Moller TR, Garwicz S, Barlow L, et al: Decreasing late mortality among five-year survivors of cancer in childhood and adolescence: A population-based study in the Nordic countries. J Clin Oncol 19:3173-3181, 2001
Steliarovà-Foucher E, Stiller C, Kaatsch P, et al: Geographical patterns and time trends of cancer incidence and survival among children and adolescents since 1970s: The ACCIS project—An epidemiological study. Lancet 364:2097-2105, 2004(Gemma Gatta, Riccardo Cap)