Angular regression and the detection of the seasonal onset of disease

Abstract

Background. In examining the seasonality of onset of a disease over the year, investigators attempt to identify the peak of onset, and its magnitude. A second objective is to see if the day in which the disease manifests itself is related to subject-specific characteristics or environmental factors. Method. This paper describes appropriate statistical methodology for the situation where seasonality can be summarised by either a single peak or several peaks, possibly determined by patient characteristics or external influences. The circular, rather than linear, nature of the day of onset of a disease (irrespective of year) requires angular regression techniques to assess these relations, and the von Mises distribution replaces the normal distribution in this context. Results. The methods outlined are illustrated by a national study of those experiencing an attack of acute primary angle-closure glaucoma in Singapore. Conclusions. We recommend re-analyses of already published work on seasonality of disease using this angular methodology. We anticipate that this may provide both useful further insight into aspects of aetiology and case studies for the methods themselves.