Childhood cancer survival in Europe and the United States
Survival rates for most major adult cancers are higher in the United States compared with the survival rates in Europe. The objective of this study was to determine whether transatlantic differences in survival also are present in childhood cancers.
The authors analyzed 16,148 European patients and 3476 patients in the United States who were diagnosed with malignant disease at age < 15 years during 1985–1989. The patients were obtained from 34 EUROCARE cancer registries in 17 countries and from 9 SEER registries in the United States. The authors considered the major 14 diagnostic categories of the International Classification of Childhood Cancers. To increase the power of comparisons, they also considered all childhood cancers together. Observed survival was calculated by actuarial methods.
For all cancers combined, northern Europe had the highest 5-year survival rate at 75% (95% confidence interval [95%CI], 72–78%), and Eastern Europe had the lowest survival rate at 55% (95%CI, 52–58%). The survival rate in the United States was roughly comparable to the survival rates in Italy and other Western European countries at 70%. Northern Europe also had highest survival rate for patients with lymphoid leukemias (83%; 95%CI, 78–88%); whereas Germany, Italy, and the other Western European countries had survival rates similar to the average survival rate for patients in the United States (77%; 95%CI, 74–80%). The survival rate was 7–9% lower in Europe compared with the survival rate in United States for patients with neuroblastoma and Wilms tumors and 8% higher for patients with retinoblastoma (all significant). Small, nonsignificant differences were found for patients with osteosarcoma, ependymoma, and medulloblastoma (with a higher survival rate in the United States) and for patients with acute nonlymphocytic leukemia (with a higher survival rate in Europe). Very similar survival rates among the two populations were found for the other cancers.
Unlike the survival of adults with cancer, the survival of children with cancer in Europe (except Eastern Europe) is very similar to that in the United States. Childhood cancers are generally more responsive to therapy than adult cancers, but these results also may reflect wide accessibility of these treatments for most patients. These results are relevant to the interpretation of differences in adult cancer survival. Cancer 2002;95:1767–72. © 2002 American Cancer Society.
Survival rates for most major adult cancers are higher in the United States compared with the survival rates in Europe, especially among the oldest patients.1 It would be interesting to determine whether these trans-Atlantic survival differences also are present among children, not least because they may help us to interpret the differences in adult survival. Childhood malignancies are rare, accounting for < 1% of all cancers. Leukemia is the most common, representing about one-third of all diagnoses.2 Therapy has greatly improved over the last 30 years, especially for patients with hemopoietic malignancies and with solid tumors that are localized at presentation. This is principally due to more effective chemotherapy.3 Population-based survival studies and mortality statistics are essential for evaluating the effectiveness of healthcare provision and the availability of effective treatments for patients with malignant disease in different populations and countries.
The objective of this study was to compare the survival of populations in Europe and the United States for 14 major diagnostic groups of childhood cancer and for all childhood malignancies combined, and to interpret the differences. Cancer survival rates for children diagnosed in Europe during 1985–1989 were published recently by EUROCARE,4 and cancer survival data for children in the United States can be accessed from the publicly available SEER database.5 Both data sets are derived from population-based cancer registries.
MATERIALS AND METHODS
We analyzed survival in 16,148 European children and in 3476 children from the United States who were diagnosed with cancer at age < 15 years during the period 1985–1989. The data were contributed by the EUROCARE6 and SEER5 databases. We considered all children with a malignancy as defined by the International Classification of Diseases for Oncology7 behavior code 3 or higher. The European data were contributed by 34 cancer registries in 17 countries within EUROCARE:6 Denmark, Finland, Iceland, and Sweden (Northern Europe); Estonia, Poland, Slovakia, and Slovenia (Eastern Europe); England, Wales and Scotland (the United Kingdom); and Austria, France, Germany, Italy, The Netherlands, Spain, and Switzerland (Western Europe). Survival data were available by country, age class (< 1 year, 1–4 years, 5–9 years, and 10–14 years), and gender.
Data from the United States were contributed by nine SEER registries from five states (Connecticut, Utah, New Mexico, Iowa and Hawaii) and four metropolitan areas (Detroit, Atlanta, Seattle-Puget Sound, and San Francisco). The same age classes that were used in the EUROCARE data were examined. We grouped each cancer in the EUROCARE data set into 1 of 14 diagnostic categories according to the International Classification of Childhood Cancers (ICCC)8 (Table 1). The SEER data already were grouped into the ICCC categories. To increase the power of the comparisons, we also considered all childhood cancers combined. The entire population of eligible children was considered, with no selection for race. Both histologically verified and nonverified diagnoses were included, but patients known to registries by death certificate only (DCO) were excluded. Patients classified as DCO were very uncommon on both continents, and almost all diagnoses were verified microscopically in both the SEER (97%) and EUROCARE (95%) data sets. Slightly more patients were lost to follow-up in the EUROCARE data set (6% vs. 4%). More detailed information on the two databases is available.3, 4
Table 1. Five-Year Survival for European and U.S. Children Ages 0–14 Years Diagnosed with Cancer during 1985–1989
|Lymphoid leukemia||Ia||4663||878||75 (74–76)||77 (74–80)||76 (75–78)|
|Acute nonlymphocytic||Ib||915||146||40 (37–43)||34 (27–43)||42 (38–46)|
|Hodgkin disease||IIa||704||145||92 (90–94)||90 (86–95)||93 (91–95)|
|Non-Hodgkin lymphoma||IIb||860||141||75 (72–78)||73 (66–81)||77 (74–80)|
|Ependymoma||IIIa||300||79||48 (42–54)||55 (45–68)||49 (44–56)|
|Astrocytoma||IIIb||1,265||375||73 (70–75)||72 (67–77)||74 (71–76)|
|PNET/medulloblastoma||IIIc||759||153||48 (44–51)||52 (45–61)||49 (45–53)|
|Neuroblastoma||IVa||1,094||265||48 (45–51)||57 (51–63)c||48 (45–51)c|
|Retinoblastomab||V||470||92||95 (92–96)||87 (85–90)c||96 (93–97)c|
|Wilms tumor||VIa||951||221||83 (80–85)||90 (86–94)c||85 (82–87)c|
|Osteosarcomab||VIIIa||288||62||61 (55–66)||68 (56–80)||62 (57–68)|
|Ewing sarcoma||VIIIc||305||56||60 (55–66)||61 (49–76)||62 (57–68)|
|Rhabdomyosarcoma||IXa||624||124||62 (58–66)||60 (52–69)||64 (60–68)|
|Germ-cell: Testisb||Xc||106||20||95 (89–98)||100 (89–100)||98 (92–100)|
|Germ-cell: Ovaryb||Xc||78||21||87 (78–93)||90 (77–100)||92 (84–100)|
Observed survival was calculated by using the actuarial method. Relative survival was also calculated using the Hakulinen method9 for the European database and the Ederer method10 for the American database. Relative survival is not presented here, because it was identical to observed survival. Indeed, in young people, general mortality is negligible with respect to cancer death risks. Confidence intervals of crude and age specific proportions (with limits between 0 and 100) were estimated by logistic transformation of the survival probabilities(s) and the corresponding standard errors, based on the usual binomial formula [s (1 − s)/(number of patients)]1/2. When survival was based on a single patient or on very few patients, it was often 0% or 100%. In such instances, the binomial formula gives 0 values that do not reflect the true sampling variability. In such instances, we used the conservative expression [(0.5 * 0.5)/(number of patients)]1/2. This correction is important to avoid underestimating sampling variability of age-standardized proportions. P values also are given as estimates of the statistical significance of differences between the survival rates in two populations.
Because survival usually depends on patient age, and because the age distribution of children with cancer may differ between countries, survival rates for children in the United States were adjusted to the age distribution of European children with the same disease to ensure comparability. For age standardization, we used four age classes (ages < 1 year, 1–4 years, 5–9 years, and 10–14 years) for patients with neuroblastoma, two age classes (ages 5–9 years and 10–14 years) for patients with Ewing sarcoma, and three age classes (ages 0–4 years, 5–9 years, and 10–14 years) for most of the patients in the remaining diagnostic groups. For patients with four rare tumors, however, survival was estimated only within a single age class: age < 5 years for patients with retinoblastoma and testicular germ-cell tumors and age 10–14 years for patients with osteosarcoma and ovarian germ-cell tumors.
Survival estimates for each cancer were compared between Europe and the United States. Differences in survival for childhood cancers are difficult to evaluate, because these diseases fortunately are rare, and survival differences between Eastern and Western Europe for most childhood cancers were large;4 thus, we calculated survival rates for all childhood malignancies combined to provide an overall comparison between European areas or countries and populations in the United States. These survival estimates were adjusted according to age and ICCC cancer type by direct standardization to the distribution of patients in the entire EUROCARE childhood data base. Adjustment was made using three age strata (ages 0–4 years, 5–9 years, and 10–14 years) and 15 diagnostic categories: the 14 most common malignancies, accounting for 85% of all patients, and all remaining cancers combined.
Because the European data refer to 17 countries at differing levels of economic development and with differing social structures and health care systems, we defined 7 European areas as follows: There were three geographically related groups of registries with broadly similar survival rates: Finland, Iceland, and Southern Sweden (Northern Europe); England, Wales, and Scotland (the United Kingdom); and Estonia, Poland, Slovakia, and Slovenia (Eastern Europe). Denmark was kept separate, because its survival rates differed from those of the other northern Europe countries (like the rates for adult cancers11). The national registry of the former West Germany also was considered separately. For the remaining Central and Southern European registries, only Italy had a sufficient number of patients to warrant separate analysis. Data from the remaining countries (Austria, France, Spain, Switzerland, and The Netherlands) were combined.
The 5-year survival rates and 95% confidence intervals are shown in Table 1; P values < 0.05 are also indicated. The survival rates for patients with acute nonlymphocytic leukemia were 40% in Europe and 34% in the United States (not significant), although the survival rates for patients with other hemopoietic malignancies were similar for both continents.
The survival rate was 4–7% lower in Europe compared with the rate in the United States for patients with central nervous system tumors, ependymoma, and medulloblastoma (not significant), although the survival rates for patients with astrocytoma were similar. Among other solid tumors, the survival rate was 7–9% lower in Europe compared with the United States for patients with neuroblastoma, Wilms tumors (both significant), and osteosarcoma (not significant); and it was 8% higher in Europe for patients with retinoblastoma (significant).
Because survival was low in Eastern Europe, Table 1 also shows a European mean survival rate that excludes the Eastern European countries. These countries provided only 8% (1340 patients) of the European childhood cancer patients who were included in this study. The effect of this exclusion is a small increase in the average European survival rates, reducing the European survival deficit for patients with lymphoid leukemias and ovarian and testicular germ-cell tumors to 2% or less.
The 5-year survival rates for all malignancies combined and for lymphoid leukemia are shown in Table 2 according to the groups defined in above. Finland, Iceland, and Southern Sweden (Northern Europe) had the highest survival rate at 75%, and Eastern European countries had the lowest survival at 55% (for both, P < 0.05 vs. the United States). Among the other populations, survival rates ranged from 72% in former West Germany to 65% in Denmark, with intermediate survival rates of 67% for the other Western European countries and 66% in the United Kingdom. The survival rate in the United States was 70% and was roughly comparable with the survival rates in Italy and other Western European countries.
Table 2. Age and Site Standardized 5-Year Survival with 95% Confidence Intervals for All Cancers Combined and for Lymphoid Leukemia in Children Diagnosed during 1985–1989 in Europe and the United States
|Northern Europe||904||240||75 (72–78)b||83 (78–88)b|
|West Germany||5364||1603||72 (71–73)b||78 (76–80)|
|Italy||833||203||71 (68–74)||76 (70–82)|
|Other West European countries||1187||528||67 (64–70)||77 (73–81)|
|United Kingdom||5880||1613||66 (65–67)b||73 (71–75)|
|Denmark||640||171||65 (62–68)b||74 (67–81)|
|East European countries||1340||305||55 (52–58)b||62 (56–68)b|
|United States||3476||878||70 (68–72)||77 (74–80)|
We also calculated the 5-year survival rates for the same country groupings for lymphoid leukemias (ICCC Ia), which represented about 25% of all malignancies. Again, Northern Europe had the highest survival rate (83%; P < 0.05) compared with the United States; whereas West Germany, Italy, and the other Western European countries had survival rates similar to the average rate for the United States (77%). Eastern European countries had the lowest survival rate (62%; P < 0.05) compared with the United States.
Contrary to the situation with adult cancers, this study revealed that there are few systematic differences in childhood cancer survival between the United States and Europe. Before interpreting the findings, it is important to consider some issues of methodology. Differences in the age distribution of patients with cancer between the United States and Europe were accounted for by age standardization. Because of the small number of patients, we did not standardize patients by gender. For patients with lymphoid leukemia, neuroblastoma, and osteosarcoma, survival rates were higher in girls (data not shown); whereas, for patients with rhabdomyosarcoma, the outcome was better for boys. However, the gender distributions for these malignancies were very similar in the United States and Europe, and it is unlikely that gender confounded the comparison.
Other possible biases may arise from the quality and comparability of the cancer registry data. The major indicators of the data quality are the proportion of DCO patients, the proportion of microscopically verified diagnoses, and the proportion of patients lost to follow-up. However, DCO patients were very uncommon on both continents, and almost all diagnoses were confirmed microscopically in both the SEER data set and the EUROCARE data set. The confirmed diagnoses are particularly important, because childhood malignancies are classified mainly according to histologic type. Slightly larger proportions of patients were lost to follow-up in the EUROCARE data set (6% vs. 4%). This may bias comparisons in an unpredictable way, because patients can be lost to follow-up either because they are cured and move away from the territory of registration or because they die, and this information does not reach the registry. However, in several European registries, the proportion of patients lost to follow-up was very low (Northern Europe group, United Kingdom, Denmark, and Italy; data not shown). One indicator of the quality of diagnostic data is the proportion of patients classified in unspecified categories within each major diagnostic group: This included 4% of patients in both the EUROCARE data set and the SEER data set. These considerations of data quality suggest that cancer survival differences between American and European children are unlikely to be due to registration artifacts.
Lead-time bias also should be considered as a possible contributor to survival differences between the two continents. However, a previous SEER study12 in which solid childhood tumors were analyzed by disease stage at initial presentation provided no indication that earlier diagnosis accounted for increasing survival trends: The proportion of children diagnosed at an early stage of disease did not increase. Nevertheless, it is likely that earlier diagnosis may have been responsible for some lead-time bias for patients with neuroblastoma. The incidence of this disease was slightly greater in the United States compared with its incidence in Europe,2 probably due to the identification of patients who had previously undetected disease with minimal clinical symptomatology through widespread application of fetal ultrasound testing and noninvasive diagnostic tests in the United States.13 Thus, although there was no significant difference in survival with neuroblastoma for infants age < 1 year (SEER, 76%; EUROCARE, 79%), there were differences for children age 1–4 years (SEER, 50%; EUROCARE, 33%) and for children age 5–14 years (SEER, 34%; EUROCARE, 30%). The propensity of neuroblastoma to undergo spontaneous regression in infancy does not explain the better outcome of children with neuroblastoma in the United States. Survival differences for patients with Wilms tumors may be explained in part by differences in approach between the two largest European and American cooperative groups that set treatment standards: the International Society of Pediatric Oncology (SIOP) in Europe and the National Wilms Tumor Study (NWTS)14 in the United States. SIOP has focused on optimizing preoperative treatments, whereas NWTS has focused on postoperative treatments. The outcomes for patients who were treated according to these two approaches in the late 1980s were very similar, with 2-year survival rates in the range of 90%.14 These data reflect the results of clinical trials, although population-based, 2-year survival rates were very similar (SEER, 92%; EUROCARE, 87%), suggesting widespread availability of optimal treatment. The European countries with the highest age-adjusted, 2-year survival rates were Finland, with 90% (53 patients), and the French registries, with 100% (14 patients).4 The relatively low 5-year survival rate for children who were diagnosed with retinoblastoma in the United States during 1985–1989 (87%; based on only 92 patients), compared with 95% in Europe (470 patients), may have been a chance fluctuation. In a SEER series that covered a longer time span than the current analyses (1976–1994), the 5-year survival rate in children age < 5 years was 93% and did not change for children who were diagnosed during different periods (1976–1984 and 1985–1994).15
In conclusion, survival rates for patients with childhood cancer in Europe (except Eastern Europe) generally are high and are very similar to the survival rates in the United States. This probably is due in part to the fact that childhood cancers are generally more responsive to therapy compared with adult cancers, but it also indicates that patients have access to these treatments. Survival rates were very low in the countries of Eastern Europe; in Western Europe, survival rates from all childhood cancers combined ranged from 65% to 75%. Differences in the availability of effective treatments and in access to up-to-date therapeutic protocols are likely explanations for the international differences in survival rates for childhood cancers. The findings of this study are pertinent to the interpretation of differences in adult cancer survival: In most countries, the data for both adult cancers and childhood cancers are collected in the same way by the same cancer registries. The overall similarity of childhood cancer survival estimates reported here for Europe and the United States suggests that the transatlantic differences seen for almost all adult cancers are unlikely to be due mainly to bias.
The authors are grateful to the cancer registries for collecting data and completing the follow-up, to Donald Ward for help with the English, and to Emily Taussig for editorial assistance. They also thank Roberto Luksch for critically reading the article.