Deriving ethnic-specific BMI cutoff points for assessing diabetes risk

Abstract

Objective: The definition of obesity (BMI ≥ 30 kg/m(2)), a key risk factor of diabetes, is widely used in white populations; however, its appropriateness in nonwhite populations has been questioned. We compared the incidence rates of diabetes across white, South Asian, Chinese, and black populations and identified equivalent ethnic-specific BMI cutoff values for assessing diabetes risk.

Research design and methods: We conducted a multiethnic cohort study of 59,824 nondiabetic adults aged ≥ 30 years living in Ontario, Canada. Subjects were identified from Statistics Canada's population health surveys and followed for up to 12.8 years for diabetes incidence using record linkages to multiple health administrative databases.

Results: The median duration of follow-up was 6 years. After adjusting for age, sex, sociodemographic characteristics, and BMI, the risk of diabetes was significantly higher among South Asian (hazard ratio 3.40, P < 0.001), black (1.99, P < 0.001), and Chinese (1.87, P = 0.002) subjects than among white subjects. The median age at diagnosis was lowest among South Asian (aged 49 years) subjects, followed by Chinese (aged 55 years), black (aged 57 years), and white (aged 58 years) subjects. For the equivalent incidence rate of diabetes at a BMI of 30 kg/m(2) in white subjects, the BMI cutoff value was 24 kg/m(2) in South Asian, 25 kg/m(2) in Chinese, and 26 kg/m(2) in black subjects.

Conclusions: South Asian, Chinese, and black subjects developed diabetes at a higher rate, at an earlier age, and at lower ranges of BMI than their white counterparts. Our findings highlight the need for designing ethnically tailored prevention strategies and for lowering current targets for ideal body weight for nonwhite populations.

Figure 1

Association between the incidence rate of diabetes and BMI by ethnic group. The multivariate Poisson regression model included age, sex, BMI, BMI-ethnicity interaction, age-BMI interaction, income adequacy, survey year, and urban versus rural dwelling. Four knots were used to generate the restricted cubic splines. All estimates were weighted by the survey sample weight to allow for estimates to be generalizable to the overall Ontario population. Data were derived from the Ontario components of Statistics Canada’s NPHS and CCHS, 1996–2005.