Spatial Patterning of Type 1 and Type 2 Diabetes Incidence in Youth
Abstract Number: 1003-P
Authors: ANGELA D. LIESE, ANDREW B. LAWSON, MICHELE NICHOLS, MAYA ZHELIAZKOVA, JAMES HIBBERT, DWAYNE PORTER, ELIZABETH J. MAYER-DAVIS, DANA DABELEA, DEBRA A. STANDIFORD, LENNA L. LIU, RICHARD F. HAMMAN, RALPH B. D'AGOSTINO, Columbia, SC, Denver, CO, Cincinnati, OH, Seattle, WA, Winston-Salem, NC
Institutions: Cincinnati, OH; Columbia, SC; Denver, CO; Seattle, WA; Winston-Salem, NC
Results: While the etiology of both Type 1 (T1) diabetes mellitus (DM) and Type 2 (T2) DM is not fully understood, both are complex diseases caused by gene-environment interactions. It has recently been suggested that common environmental determinants may operate in the etiology of both diseases. The geographic patterning of disease may reveal important environmental etiologic clues. No data on spatial patterning of T1 and T2 DM incidence exist for youth in the United States. Using data from two centers in the SEARCH for Diabetes in Youth Study, a multi-center population-based surveillance system of diagnosed DM, we aimed to evaluate whether T1 and T2 DM exhibit an overlay of spatial patterning.
Between 2002 and 2003, a total of 299 incident cases occurred in 46 counties in South Carolina (SC) and 333 cases in 64 counties in Colorado (CO) among youth aged 10 through 19 years of age. County of residence at diagnosis was available for 99.1% of youth (81% self-report, 18.1% medical record). DM type was based on the physician’s diagnosis. Because T1 and T2 incidence differs markedly between races, we conducted race-stratified analyses, focusing on African Americans (AA) in SC, Hispanics (HISP) in CO, and non-Hispanic whites (NHW) in SC and CO. To estimate the number of expected cases of T1 and T2 per county, published DM-type, age group, and race-specific incidence rates from SEARCH were applied to 2002 Census denominators. We then fitted multivariate conditional autoregressive models which generate smoothed, county-specific relative risks (RRs) of T1 and T2 and estimated the spatial correlation parameter between T1 and T2 RR within a Bayesian framework.
Applying spatial smoothing models, none of the 110 counties exhibited significantly elevated RR of T1 or T2 DM. The evaluation of spatial correlation between T1 and T2 revealed that there was no overlay of spatially structured risk between both diseases as the confidence intervals (CI) of the spatial correlation parameters consistently included the null value. This was true for both racial groups in SC (NHW: CorrRR=0.42, 95%CI -0.70, 0.96; AA: CorrRR=0.40, 95%CI -0.79, 0.95) and CO (NHW: CorrRR=0.25, 95%CI -0.85, 0.97; HISP CorrRR=-0.62, 95%CI -0.93, 0.75). In conclusion, at the level of geographic patterning, our study does not lend support to the hypothesis that T1 and T2 DM share a common environmental etiology.