OBJECTIVE - To determine whether the associations of BMI and fat distr
ibution with diabetes risk are modified by race. RESEARCH DESIGN AND M
ETHODS - Data from the National Health and Nutrition Examination Surve
y, Epidemiologic Follow-up Study (1971-1992), were used to investigate
potential interactions of BMI and fat distribution with race. Inciden
t diabetes was defined by self-report of physician-diagnosed diabetes,
hospital and nursing home discharge records, and death certificates.
RESULTS - Among the 1,531 black and 9,852 white subjects who were nond
iabetic at baseline, 1,139 (10.0%) developed diabetes during 20 years
of follow-up. Although the cumulative risk of diabetes increased with
baseline BMI in all four race-sex groups, the sex-specific odds ratios
(ORs) for black:white subjects decreased with increasing BMI. In part
icular, for BMI of 22 kg/m(2), the OR of diabetes for black:white indi
viduals was 1.87 and 1.76 (P < 0.01) for men and women, respectively;
for BMI of 32 kg/m(2), the OR decreased to 0.99 and 1.20 (NS) for men
and women, respectively. Skinfold ratio was also associated with incre
ased diabetes risk in all race-sex groups, but did not modify the asso
ciation between race and diabetes. CONCLUSIONS - These findings sugges
t that the effect of BMI on diabetes risk is different for black and w
hite Americans, with a larger risk for blacks than whites at low BMI a
nd an equivalent risk for both groups at high BMI. A lower degree of v
isceral adiposity among blacks at higher BMI or a greater impact of vi
sceral adiposity among blacks at low BMI may help explain the interact
ion of race and BMI on diabetes risk.