This population-based study spanning nearly 50 years showed increasing incidence and survival of childhood cancer in Estonia. Survival improvements varied by cancer site and were most substantial before the 2000s.
Cancer registration began in the Soviet Union in 1953. At the very early stages of cancer registration no standard disease classification nor age groups were used, data was published annually, but never corrected. The ECR dates back to 1968. This nation-wide population-based registry holds data for adult and childhood cancer cases, including benign CNS tumours since 1998. Reporting to the registry is mandatory for all physicians who diagnose or treat reportable tumours. Case registration according to international rules was implemeted step-by-step starting with the introduction of ICD-O morphology section in 1978 and the topography section in 1983. The 3rd edition of TNM classification of Malignant Tumors was translated into Estonian in 1980. ECR data was first included in the Cancer Incidence in Five Continents volume 6, published in 1992. All cases have been retroactively coded to ICD-O-3. Currently, the ECR follows international definitions and rules, including those for multiple primaries, issued by the European Network of Cancer Registries and the International Association of Cancer Registries7, for reporting incidence and survival.
Good data quality has been shown for adult cancers.8 According to a previous study,9 the estimated completeness of reporting of childhood cancer cases was 89.5% (81.2% for hematologic malignancies and 95.3% for solid tumors). As a result of the data quality study, all missing cases starting from 2000 were added to the cancer registry and regular linkage with all treating hospitals was established. Thus, we believe that the disruption of registry practices in the 2000s10 has no further impact on case registration and the estimated completeness of childhood cancer is now close to 100%. The validity of childhood cancer data has improved markedly over time – there were only two autopsy cases and no DCO cases since 2003. The proportion of microscopically verified cases has increased throughout the study period, now reaching 98%. ECR data has been included in major childhood cancer studies in Europe,2,11 being a good indication of the data quality.
Methods for diagnosing improved during the study period. Ultrasound became available for prenatal screening since the early 1990s (first ever used in 1984), MRI in the 2000s. Most recent changes in histology were introduced in 2010 (molecular pathology).
The main strengths of the study were nation-wide population-based cancer registry coverage with complete follow-up and high data quality, enabling the evaluation of incidence and survival trends over a 47-year period, with the most recent data for patients diagnosed in 2016. The main limitations of this study were the scarcity of cases (causing a fluctuation in trends and instability of survival estimates) and the quality of morphological verification in earlier periods that might have caused some misclassification of cases.
The aetiology of childhood cancers is still in large part unknown. Many hypotheses have been made about different environmental, parental and gestational factors and some genetic causes have been established, but the scarcity of data and studies make it difficult to draw definite conclusions.12 Childhood cancer incidence in Estonia increased during 1970–2016 at a rate of 0.48% per year, comparable to rates in Europe and Eastern-Europe in 1991–2010 (0.54% and 0.50% respectively). Even though hypothesised that studies, where data was analysed starting from the 1970s, may see increase in rates in large part due to the rise in the earlier decades13 (caused by improved diagnostics, decreased overall child mortality), we saw the opposite. The overall age-standardised incidence of childhood cancers in Estonia for the first half of the study period was notably lower than the European average (122.2 vs 140.0 per million) but reached the European rate for 1991–2010 in the second half (138.1 vs 137.5 per million). Regarding the fact that our study dates back to 1970, we can assume that in some part the increase in incidence can be explained by improvements in diagnosing and registration. Diagnostic drift must also be taken into consideration, especially for cancer sites where overall incidence for the site group remains stable through time but changes occur among subsites. Other possible explanation for incidence increase, especially in an atypical age group may be a delay in diagnosis, highly probable for that time period. On the other hand, not all can be attributed to improved diagnostics. Rise in incidence for some sites (e.g neuroblastomas, soft tissue sarcomas etc) and all sites combined, indicates an actual rise probably due to the risk factors mentioned above. The same applies for competing risk theory, as infant and child mortality have decreased since the 1970s.
Leukemias, the most commonly diagnosed malignancies, showed a slight increase in incidence. ASIR in Estonia was lower in both periods than shown for Europe in ACCIS studies, for the periods 1970–1999 and 1991–2010 (39.5 vs 44.8 and 42.1 vs 46.9 per million). Rate for Europe has increased steadily by 0.6% per year in both periods,14,15 whereas Estonian average increase of 0.2% per year for the whole period is more characteristic to Eastern-European trend (increase of 0.3% per year for 1978–199716 and 1% in 1991–2010).14 Since there was a substantial decrease in unspecified leukemias a restricted analysis was done starting from 1985 onwards when the proportion of unspecified cases became negligible to control whether the increase was solely due to improved diagnostics. Even though non significant, incidence increased from 1985 onwards for ALL (APC 1.4%, 95% CI -0.6–3.5), AML (APC 2.4%, 95% CI -3.4–8.5) and leukemias combined (APC 1.4%, 95% CI -0.2–3.0), referring to other underlying causes (data not shown). Several environmental and parental lifestyle, residential and occupational exposures have been associated with pediatric leukemias17 but among known risk factors, high birth weight and increased maternal age18 could be possible reasons here.19 Some studies show that early day-care attendance is protective against leukemia20,21 due to exposure to infections. Starting from early 2000s, children in Estonia attend day-care less and start at a later age, as mothers now have the option for prolonged and paid leave.
Incidence of lymphomas is high in Estonia (16.9 per million for 1970–2016), as has been described also for other Eastern-European countries.14 Rate is increasing in Europe by 0.26% per year, but stable in Estonia and declining in Eastern-Europe (APC − 0.1% and − 1.32%, respectively).14 Even though the overall rate is relatively stable, an immense increase by 8.8% was seen for Burkitt lymphomas, which generally contribute proportionally lower to lymphoid malignancies.17 Similar trend, but at a much slower rate of 1.4% per year, was seen for Spain in 1983–2007,22 while a decreasing trend was shown in Canada (APC − 2.54% for 1992–2010).23 As lymphomas tend to be more frequent in less developed countries and of infectious origin, our finding is difficult to reason. Since the rise has been more recent, it could partly be explained by improved diagnostics and misclassification, as the incidence of all other subgroups have declined. Classification has changed for miscellaneous lymphoreticular neoplasms (IId). In previous ICD-O classifications Langerhans cell histiocytosis was not a reportable disease. ICD-O-3 includes codes for both malignant and uncertain disease. Thus, increase in incidence seen in our data could be caused by the changes in classification.
Likewise, the incidence of unspecified neoplasms of CNS decreased significantly, while the overall trend remained relatively stable. Therefore, the observed increase in the incidence of ependymomas (IIIa) and other gliomas (IIId) needs to be interpreted with caution. In the earlier periods, the incidence rates for ependymomas and gliomas were lower in Estonia compared to Eastern-Europe (1.9 per million in Estonia compared to 3.3 per million in Eastern-Europe for ependymomas and 0.7 vs 2.2 per million for gliomas, respectively).24 More recent incidence rates for ependymomas are now higher in Estonia than in the Nordic countries and France (6.2 vs 4.2 vs 3.8, per million, respectively), whereas incidence rate for other gliomas has remained lower – 2.0, 3.5 and 4.9 per million, respectively.25,26 The absence of overall trend in incidence is most probably due to the changes in tumor classification and the tendency in earlier years to record benign cases as malignant.
Neuroblastoma incidence in Estonia increased significantly over the study period at a rate of 2.3% per year due to a significant increase in the 0–4 age group (APC 3.1%) and subsite IVa (AAPC 2.5%). Incidence of subsite IVa increased more for age group < 1 (APC 4.0%, 95% CI 1.4–6.6) compared to ages 1–4 (APC 2.7%, 95% CI -0.3–6.1) (data not shown). A similar overall trend was seen for Eastern-Europe in 1978–1997 (AAPC 2.7%).16 For other countries, the incidence seems to increase at a slower pace, for example in Canada (APC 0.74% in 1992–2010).23 Since we do not see a distinct leap in the incidence in any of the time periods, we can not conclude that the increase is solely attributable to improvements in diagnostics, e.g wider use of ultrasound (1990s), MRI (2000s) or molecular pathology (2010), which has been shown to improve the differentiation of neuroblastomas from other small round cell tumors like rhabdomyosarcoma and Ewing´s sarcoma27, but the latter two also show an increasing trend in Estonia. Several risk factors, such as maternal oral contraceptives or sex hormones use during pregnancy, a shorter gestational duration, and maternal alcohol consumption during pregnancy28, have been debated as possible causes, but need to be analysed further.
Hepatic tumors show an increasing trend in Estonia and elsewhere.16,23 Even though non-significant, the rise was noticeable in the youngest age group, with almost half of the cases diagnosed in infants less than 1 year of age (data not shown). In these cases, perinatal exposures and decreased premature infant mortality could be postulated causes.29 Almost 90% of the cases diagnosed in age group 0–4 were hepatoblastomas, which is associated with low birth weight.30 The proportion of low birth weight babies has remained the same in Estonia, but their survival has increased over time.19
The incidence of both intracranial and extracranial germ cell tumors increased significantly. The incidence of intracranial germ cell tumors is in accordance with the rate in Europe (about 1.0 per million)31 and in Germany (1.2 per million, and also increasing), but the incidence of extracranial germ cell tumors is higher in Estonia and showing a reversed trend compared to Germany (decreasing at 0.1%).32 We see two peaks in incidence, first one falls to the mid 1980s, when certain biochemical methods and computer tomography were introduced,33 and another spike more recently, which could refer to changes in clinical and treatment practices for childhood ovarian tumors.32 A more detailed analysis of clinical data by subsite is warranted.
Estonia regained its independence in 1991, at midpoint of the study period. The latter resulted in major reforms, where solidarity-based health insurance was established and the health care was changed from the state-run Semashko34 system to a more Western one. Integrated and hierarchically organised system, which was communicable diseases and primary care oriented, was replaced by decentralized model funded through social insurance.
Childhood cancer treatment in Estonia in 1970s to 1990s was scattered between departments and institutions. At that time, hematologic malignancies were treated in the general pediatric departments of two major hospitals in Tallinn and Tartu. Solid tumors were operated in the surgical departments together with adults until 1979, when it was partly moved to the surgical department of Tallinn Children´s Hospital. Radiation was the main treatment and chemotherapy was used scarcely at that time. Even though the centralisation of childhood cancer care began in 1992 when the first pediatric oncology wards were opened, children were still treated together with adults for a long time. CNS treatment was gradually brought to pediatric oncology departments in 1995–2000, whereas low grade tumors are still partly managed at adult centres. The arrival of first European treatment protocols in 1991 for solid tumors (SIOP, Société internationale d'oncologie pédiatrique) and 1992 for hematologic malignancies (NOPHO, Nordic Society for Pediatric Hematology and Oncology) set the base for twinning with European oncology organisations, enabling Estonia to reach new information and treatment protocols faster (Fig. 3). NOPHO protocols for ALL and AML were at first hand used partially in 1992–2005. More advances followed with the introduction of high dose methodrexate treatment (HDMTX) and bone marrow transplant from sibling donors in 1995. In 2000, when CNS treatment was transferred to children´s oncology department, modern protocols and chemotherapy became available. Starting from 2008 for ALL and 2012 for AML Estonian patients are included in NOPHO clinical trials. Even though Estonia has never participated in clinical trials for solid tumors, standard treatment protocols are used according to guidelines. Over time, the main issues of managing childhood cancer, have been the availability of drugs (more pronounced during regaining of independence), reaching new protocols and twinning for diagnosing rarer types of cancer.
All these improvements in diagnosing and treatment and changes in the health care organisation are reflected in survival outcomes. Starting from 23.8% in the 1970s, overall 5-year survival reached 73.0% by the latest period, which is not far below the European average for 2000–2007 (77.9%, 95% CI 77.4–78.3).2 Survival increase has been relatively steady over time, but the biggest rise occurred in the 1990s.
The survival change in leukemias (from 7.5–83.8%) has been even more remarkable, being now equal to the rate observed in Finland for 2001–2010.35 Two leaps could be seen in leukemia survival. First from 1985–1989 (25.4%) to 1990–1994 (51.3%), coinciding with the introduction of first treatment protocols and the second from 1995–1999 (52.0%) to 2000–2004 (65.2%), when HDMTX treatment and ABMT became available. Last increase in survival rates can be associated with the access to registry donors for ABMT, which seemed to have a stronger impact primarily on AML survival. But looking at the two latest periods, 2005–2009 and 2010–2016, we can see that survival is plateauing (from 81.6–83.8%) (seen for other sites and overall survival as well, data not shown), indicating that the current treatment methods are depleting and new biology-driven approaches are needed.36 The gap in survival between ALL and AML has narrowed over the study period, a trend that has also been shown in Switzerland and Finland.35,37 Biggest increments in 5-year survival increase for lymphomas were seen from 1990–1994 (49.9%) to 1995–1999 (70.3%) (data not shown), most likely due to the addition of HDMTX which has been shown to decrease relapse rate, especially for advanced disease.38 Survival outcome dependency on methotrexate dose has been described for non-Hodgkin lymphomas,39,40 which showed the biggest improvements in Estonia.
The largest improvements for CNS malignant tumors occurred from the 1980s to the 1990s, coinciding with the newly available treatment options and continued into the 2000s, reflecting treatment centralization. However, there has been a recent decline in CNS tumour survival as the latest estimate of 49.5% is well below the European rate of 57.5% for 2000–20072 and Finnish rate of 59.5% in 2001–2010.35 As shown in previous analysis, the survival of all CNS tumour patients (including benign and borderline tumors) was 70.9% in Estonia in 2010–2014,9 compared to 79.1% in Finland in 2001–2010.35 The survival estimates for CNS tumours may be somewhat affected by the use of different diagnostic methods and the classification of tumour behaviour24. As the number of unspecified cases has decreased, it is possible that some benign cases were previously recorded as malignant, thus included in the survival analysis and causing a higher survival estimate. It has been shown that the rate of unspecified cases was high for Estonia in earlier years.41 The recent drop in survival could also be explained by changes in classification – pilocytic astrocytoma was downgraded to uncertain behaviour in ICD-O-3. As proposed by Stiller et al.42 some cases of pilocytic astrocytoma could have been recorded as astrocytoma, NOS in earlier years, thus included in the analysis and causing a higher survival estimate. Our data shows a higher number of annual cases of astrocytoma, NOS in the first half of the study period and a higher number of pilocytic astrocytoma for the second.
Opposite to that shown for Northern-England,43 no significant changes in 5-year survival trends for neuroblastomas were observed in Estonia. Increase from 12.5% in the first decade to 47.4% in 1980–1989 marked the only substantial change in 5-year survival, which has stayed stable since. Even though the same pattern was shown for Finland,35 rates in Estonia have remained lower than those in Finland and Europe (for 2000–2010 40.0% in Estonia and 68.2% in Finland; for 2000–2002 72.0% in Europe).35,44 Age > 18 months, male sex, metastasis at diagnosis and amplification of MYC-N are associated with poorer disease outcome.45 Almost 60% of the cases diagnosed in Estonia were older than 18 months and with a M:F ratio of 1.3. Unfortunately, stage and tumor biology info was not available in registry data. A recent pilot study for the application of Toronto Staging confirmed the disease being diagnosed rather late, but not advanced, as majority of the cases were classified as stage L1 (local), giving a 3-year survival of 75%.46 As for several other sites, low number of cases annually may also influence the result.
The huge gap in the survival of renal and malignant bone tumors between Estonian and Finnish patients in the 1980s has closed due to rapid improvements in Estonia throughout the 1990s and 2000s.35 Hepatic tumors, the site group with one of the worst prognosis, is more prevalent in Eastern than in Western countries and seems to present a less favourable survival outcome for our region.15 In Estonia, survival for hepatic neoplasms fluctuates greatly. Most of the cases were diagnosed in infants and specific symptoms generally develop late, but without data on stage, we can not conclude, whether this is the case. More detailed analysis could help to draw attention to possible prenatal parental exposures or shortcomings in treatment. Survival of 100% that could be seen from our data, for the period of 2010–2016, is most probably a random finding due to the small number of cases and does not reflect the current situation for hepatic malignancies. Germ cell tumors are a heterogenous group of tumors, presenting different rates of survival, depending on age and histology. Survival in Estonia reached 76.2% for 2000–2009, but then dropped to 37.0% in the latest period. The drop seen here is most probably random as low number of cases are more prone to cause a fluctuation in survival.