Epidemiological studies including new methods for cluster analysis of acute childhood leukaemia and brain tumours in Sweden

Abstract

Background: The aetiology in childhood cancer is essentially unknown. Epidemiological investigations as to whether the incidence rates have changed for paediatric cancers or whether clustering of cases occur may give clues to possible causal factors. One of the main purposes of the present work was to develop improved methods for spatial epidemiological investigations, especially cluster analyses and apply the methods evolved to the two most common forms of childhood malignancies: leukaemia and brain tumors.

Material and methods: Population-based materials of all cases of acute childhood leukaemia during the period 1973 to 1994 and registered brain tumours during the period 1973 to 1992 among children under 16 years of age were analysed. A geographical information system (GIS) was utilised in the management of spatially referenced data on patients and population. Analyses of geographical clustering in space, space-time and of space-time interaction were conducted by essentially new statistical methods, evolved in the work, namely a spatial scan statistic and a modified Knox test. For the brain tumours, analyses of temporal trends were performed by a logistic regression procedure.

Results: No statistically significant: geographically localised clusters of childhood leukaemia in Sweden were detected in space or in space-time. Statistically significant space-time interaction was found for acute lymphoblastic leukaemia (ALL) in the analyses using the modified Knox test statistic (p=0.01). Incidence rates in population centres, constituting 1.3% of Sweden's land area and approximately 80% of the population, compared to the rest of Sweden showed a statistically significant excess of (ALL), (OR 1.68, 95% CI 1.44-1.95) but not of acute non- lymphoblastic leukaemia (ANLL), (OR 1.13, 95% Cl 0.98-1.32).

There was no statistically significant increase in the incidence of acute childhood leukaemia in areas contaminated due to the Chernobyl reactor accident. Statistically significant increasing time trends were observed for the group of childhood malignant brain tumours as a whole (p = 0.0001) largely caused by an increase for the astrocytoma subgroup (p = 0.0001). The increase of astrocytoma rates was significantly larger for girls than for boys (p = 0.021).

Conclusions: No geographically localised clusters were found for acute childhood leukaemia and childhood brain tumours. The space-time interaction found for ALL indicates that environmental factors may be of importance to the aetiology of childhood ALL. No increased risk for ALL or ANLL was found after the heavy fallout in parts of Sweden after the Chernobyl reactor accident. The statistically significant increase of brain tumours, notably astrocytomas in girls, indicates the possible importance of some environmental factors to the aetiology. Within the project important improvements to the methodology of spatial analyses have been developed, which may possibly set a new standard for investigations of disease clusters and clustering.