Space-time clustering of Burkitt's lymphoma in East Africa: analysis of recent data and a new look at old data.

Forty cases of Burkitts' lymphoma (BL) in North Mara, Tanzania, with onset between 1971 and 1977, were analysed for evidence of space-time clustering. Previous analyses in East Africa had produced conflicting results. The Knox method used in those analyses dichotomizes the space and time scales and does not take into account the degree of closeness. The Mantel method, a generalization of Knox's, does permit closeness between pairs of cases to enter into computations. To see whether this method could clarify matters, previously reported data sets from West Nile, Uganda and North Mara were reanalysed. Unexplained differences were found between West Nile and North Mara with respect to age, sex and temporal distributions of BL, and between the eastern and western parts of North Mara with respect to incidence. In West Nile between 1961 and 1965, there was clustering. All of the Mantel analyses and a few of the Knox analyses were highly significant (p less than 0.0025). Since 1966, evidence of clustering is weak. In North Mara, there was no statistical evidence of space-time clustering between 1971 and 1977, as there was none between 1964 and 1970. The conflicting results in East Africa are compatible with a model involving several factors which "move about", or with a single factor which is sporadic in some areas but constant in others. Alternatively, artifactual biases or coincidence may have created or wiped out the appearance of clustering. Results of space-time clustering analyses permit little discrimination between infectious and non-infectious etiologic hypotheses. Either one type or both may be operating.