Seasonal variations in the onset of childhood leukaemia and lymphoma.

Infection has long been suspected as a possible factor in the aetiology of leukaemia and lymphoma. If seasonal variation in the onset of disease could be shown in any of the diagnostic subgroups of leukaemia or lymphoma, this would provide supportive evidence of an aetiology linked to exposure to infection. All cases in the Manchester Children's Tumour Registry (aged 0-14 years at diagnosis) with acute lymphoblastic leukaemia (ALL), acute non-lymphocytic leukaemia (ANLL), Hodgkin's disease (HD) or non-Hodgkin lymphoma (NHL) between 1 January 1954 and 31 December 1996 were included in an analysis of seasonal variation in the month of first symptom and the month of diagnosis. Cases of common acute lymphoblastic leukaemia (c-ALL) diagnosed from 1979 onwards were also analysed separately. The groups considered for analysis were: all cases of ALL (n = 1070), ALL diagnosed between 18 and 95 months of age (n = 730), ALL diagnosed over 95 months of age (n = 266), c-ALL (n = 309), ANLL (n = 244), all infant acute leukaemias (ALL and ANLL under 18 months; n = 107), HD (n = 166) and NHL (n = 189). Using the Edwards method, both c-ALL and HD demonstrated significant seasonal variation (P = 0.037 and 0.001 respectively) in date of first symptom, with peaks occurring in November and December respectively. Using this method, no indication of seasonal variation was found in the other diagnostic groups for date of first symptom or in any of the diagnostic groups for date of diagnosis. For comparison with a previous study, a further analysis based on date of diagnosis for all ALL cases, using summer-winter ratios, showed a significant summer excess. These results provide supportive evidence for an infectious aetiology for c-ALL and HD, and possibly for all ALL, which warrants further investigation.