Point pattern analysis of the spatial proximity of residences prior to diagnosis of persons with Hodgkin's disease.

There is some evidence to suggest that the etiology of Hodgkin's disease includes an infectious component. One approach to investigating this possibility is a formal assessment of the geographic and temporal variation in the incidence of the disease. The present study was designed to better evaluate space-time patterns of residence prior to diagnosis of persons with Hodgkin's disease. Lifetime residential histories were obtained from 279 incident cases of Hodgkin's disease diagnosed between 1974 and 1982 in a defined population in northwestern Washington State, and a similar number of population controls matched to the cases by age, sex, and socioeconomic status. A method of point pattern analysis was modified for use in this context to compare the spatial proximity of case residences to that expected under the null assumption of no difference between the spatial pattern of case and control residences. Analyses were applied within temporal groupings of residences specified a priori assuming 1) no specified latency, 2) four different 5-year latent intervals, and 3) three age at exposure intervals. Assuming no latency, there is no evidence that case residences are more or less clustered than would be expected by chance. Assuming latent intervals of up to 15 years prior to diagnosis, case residences are less clustered than expected, particularly among those who developed Hodgkin's disease after the age of 40 years. In contrast, there is suggestive evidence that persons diagnosed with Hodgkin's disease after age 40 years lived closer together than expected as young children and teenagers. These results illustrate the need to focus such analyses on specific time intervals defined a priori to most likely represent periods of greatest etiologic relevance. To the extent that these findings are not the result of some unknown artifact of the method itself, they may serve to focus additional attention on childhood environment as a particularly important period in the etiology of this disease. Furthermore, the analytic method employed may be useful in identifying space-time clustering using population-based data in other etiologic settings.