Acute lymphoblastic leukaemia (ALL) and type 1 diabetes have an environmental aetiology and common epidemiological features. Incidence rates and national characteristics of both conditions were investigated in 40 countries worldwide. There was a significant positive correlation between diseases. Markers of wealth and affluence were significantly associated with high incidence.
- type 1 diabetes
Statistics from Altmetric.com
Acute lymphoblastic leukaemia (ALL) and type 1 diabetes in children appear to be biologically unconnected conditions, but there are common threads in both their epidemiology and possible aetiology. Over the past 20 years, the incidence of both diseases has been slowly rising in western countries, a phenomenon that cannot be explained by underlying genetic changes at the population level. Furthermore, rates of ALL and type 1 diabetes display geographical variation both internationally and within countries; in Yorkshire, UK the spatial distribution of both diseases is closely linked.
A common and recurring theme in the aetiology of ALL and type 1 diabetes has been the possible involvement of infections.1 One explanation involves a paradox of development in westernised societies: that common infections in infancy may protect from later disease by appropriate modulation of the naïve immune system. Contrarywise, in the absence of such early exposures, later infection, for example with social mixing of children, may precipitate abnormal immune reactions and disease. Inherited genetic factors may also influence susceptibility. This scenario has been referred to as the “hygiene hypothesis” in the context of both type 1 diabetes and allergy/asthma2 and the “delayed infection” hypothesis for childhood ALL.1 Although the underlying immunological pathology of childhood allergies and type 1 diabetes are dissimilar (T-helper 2 versus T-helper 1 T cell overactivity), their shared environmental associations are reflected in positive correlations between their respective international incidence rates.3 We therefore anticipated that type 1 diabetes and childhood ALL might show a similar association.
Firstly we tested for correlations in international incidence of childhood ALL and type 1 diabetes (ages 0–14); secondly we explored factors that might help explain variation in rates. The correlation (Pearson’s product moment test) in rates across 40 countries was investigated by taking comparable published temporal and geographical data on standardised incidence rates. Where rates were only available for parts of countries, those from the same nation were pooled using weights based on case frequencies. For example, one sub-area with an overall rate based on 1000 individuals would carry twice the weight as a second sub-area containing 500 individuals.
The analysis showed a significant positive correlation between the incidence rates of ALL and type 1 diabetes (r = 0.53, 95% CI 0.36 to 0.72) (fig 1). This pattern was similar across Europe (n = 26) and the rest of the world (n = 14) and the exclusion of Finland as the only outlier made no difference to the results.
Higher rates of diabetes are associated with wealth and affluence in European countries,4 so we tested for associations with diabetes worldwide and for the first time explored the relation between levels of national prosperity and ALL incidence. The explanatory variables were gross domestic product (GDP), infant mortality, and life expectancy. Additional factors previously linked to disease incidence were examined including population density (linked to diabetes and ALL), coffee consumption (diabetes), and cows’ milk consumption (diabetes). A recently suggested potential protective effect from vitamin D supplementation in early life for type 1 diabetes was indirectly measured by examining latitude and average hours of sunshine.
The logarithmic or reciprocal transformation was used for factors which were highly skewed (GDP, infant mortality and population density; table 1). The remaining explanatory variables and incidence rates themselves were all reasonably normally distributed. Correlations and separate linear regression models were applied to each disease, with each factor included as an independent variable. All statistical analyses were performed using Stata.
The correlation and regression analyses showed positive associations for GDP, infant mortality, and life expectancy with both ALL and type 1 diabetes (table 1).
Population density was associated (inversely) only with diabetes, while coffee and milk consumption were significantly associated with both conditions. Latitude and sunshine were both linked to diabetes alone. The R2 statistic showed that GDP and infant mortality along with milk and coffee consumption were factors which best explained the variation in incidence for ALL and diabetes, although this goodness of fit test should be interpreted with caution.
Our analysis clearly shows for the first time that the international incidence of ALL and type 1 diabetes are positively associated. Countries with either high or low incidence of either disease are likely to have a corresponding rate of the other condition.
Markers of national prosperity appear to explain some of the worldwide variation between countries for both ALL and type 1 diabetes, which is consistent with observations in European populations.4,5 The same links with socioeconomic variables can also be seen within countries at a smaller geographical scale. Exactly what specific “affluent” lifestyle factors may be responsible for the presence of high rates of disease cannot be clarified by the ecological approach. However, affluence lowers the prevalence of infectious diseases and changes patterns and timing of exposures to microbial pathogens in children. The international correlations and their links to affluence are therefore consistent with an infectious aetiology.
Our results for population density support previous observations for diabetes but not for ALL.4,6 The latter finding requires confirmation in further studies and may have arisen through naturally large variations in population density within countries. For diabetes and latitude, sunshine, and coffee and milk consumption, we confirm other ecological studies, but the new finding for ALL and milk and coffee needs further evaluation, particularly as our measure of consumption related to the whole population rather than among children.
All ecological analyses need to be interpreted with caution as markers measured across an area may well not apply to individuals and we cannot rule out the presence of confounding factors which may have influenced our results. More affluent countries may have disease registries of higher quality than less affluent nations. Historically, cancer registries have been recognised as playing a vital role in public health monitoring and have been established for several decades with standardised methods throughout the world (fig 1: data source 1). Although childhood diabetes registers are generally less well established, formal estimates of completeness were in excess of 85% (fig 1: data source 2) with no apparent indication that levels of completeness are related to affluence. It is therefore unlikely that differential data quality would explain the associations we have observed with affluence.
Despite these concerns, investigating geographical variation can point us in the direction of more focused aetiological searches. Future studies describing the joint incidence patterns of disease would be useful, for example in countries with a range of ethnic and demographic characteristics and also in relation to temporal trends. The striking parallels in the descriptive epidemiology of type 1 diabetes and ALL in children suggest that an exploration of common causal pathways linked to the immune response in early life and underlying genetic susceptibility in individuals would be fruitful.