On the detection of seasonal variation in the onset of disease

Abstract

Many studies have investigated the seasonality of onset of diseases over the calendar year, with a view to finding indications of their aetiology. This thesis studies statistical methodology for seasonality which can be summarised by a single peak or multiple peaks determined by patient characteristics or external influences. The circular, rather than linear, nature of the date of onset means that Normal distribution theory is not appropriate. Instead the von Mises distribution on a circle and angular regression methods are required. A computer program is developed to implement the methodology. The thesis re-examines, using angular methodology, summary data from published studies of acute lymphoblastic leukaemia (ALL). In addition, variation using individual dates of onset of ALL (in children and adults) from Singapore, Sweden and USA are examined. No clear picture emerges with respect to the seasonality of ALL although methodological difficulties with the published studies may obscure the overall pattern.