Poisson regression modeling of temporal variation in incidence of childhood insulin-dependent diabetes mellitus in Allegheny County, Pennsylvania, and Wielkopolska, Poland, 1970-1985.

Contradictory observations have accumulated regarding a secular trend and/or an epidemic pattern in the incidence of insulin-dependent diabetes mellitus. In this study, insulin-dependent diabetes mellitus incidence below age 15 years was examined in Allegheny County, Pennsylvania, and in Wielkopolska, Poland, two areas diverse in terms of their geography and average risk for this disease. Numerator data were extracted from individual patient records, and annual denominator data were available for the years 1970-1985. Poisson regression models were used to disentangle the contributions of country, race, sex, age, period, and cohort effects to the observed variation in incidence. Poles and Allegheny County nonwhites were at greatly and moderately reduced risk, respectively, relative to Allegheny County whites. An increase in risk with age was significant and proportional in all three groups. There was significant time variability in Wielkopolska, where an epidemic began in 1982 and continued through 1985. This was a period rather than a cohort phenomenon and was a result of a recent outbreak of the disease rather than a long-term trend. In Allegheny County, changes in risk over the 16-year period were insignificant, although incidence doubled among whites aged 0-9 years during 1982-1983. The Poisson regression modeling provided a quantification and formal comparison of determinants of the incidence of insulin-dependent diabetes mellitus.