Seasonality in the diagnosis of childhood acute lymphoblastic leukaemia

in the period 1971-94. These data showed a 16% excess of cases diagnosed in the summer months (May-October) in children (95% confidence interval 1.02-1.30), and a 20% excess in adults (95% confidence interval 1.03-1.37). Using the same fairly crude technique as Badrinath et al (1997) on two datasets (which covered slightly different age and diagnostic groups and were collected over different time periods) has produced mixed results. We found little evidence of seasonality in a national dataset, but have found seasonality (albeit less marked than in East Anglian data) in a regional dataset. Badrinath et al (1997) noted that it may be difficult to demonstrate a seasonality effect in a heterogeneous national population, unless account is taken of geographical heterogeneity. To investigate this issue further, we suggest that future work on seasonality needs more sophisticated analyses, controlling for broad geographical heterogeneity. If data are to be examined over very long periods, the influence of long-term temporal trends should be removed, or false-positive patterns of seasonality may be produced. EA Gilman', T Sorahan2, RJ Lancashire', GM Lawrence3, KK Cheng'