To determine and compare the abilities of various anthropometric measurements to predict the development of non-insulin-dependent diabetes mellitus (NIDDM) in Pima Indian men and women.

A total of 290 male and 443 female Pima Indians were followed for up to 6 years for the development of NIDDM. A proportional hazards analysis was used to assess the ability of anthropometric measurements evaluated at baseline to predict NIDDM. Receiver operating characteristic (ROC) curves were used to compare individual variables in predicting NIDDM.

In separate models controlled for age and sex, body mass index (BMI), waist circumference, thigh circumference, waist-to-thigh ratio (WTR), weight, and percentage body fat (PBF) estimated by bioelectric resistance each predicted NIDDM, which developed in 30 men and 52 women. The highest incidence rate ratios (IRRs; for 1 SD of a variable) were for WTR in men and for PBF in women, although the confidence interval (CI) for PBF was wide. In stepwise analyses, WTR was the most significant predictor in men (IRR for 1 SD = 1.58, 95% CI = 1.20–2.07), and BMI was the most significant predictor in women (IRR for 1 SD = 1.65, 95% CI = 1.29–2.11). However, by ROC analyses, thigh circumference was the only variable significantly worse than WTR in men or BMI in women at predicting NIDDM.

Measurements such as waist circumference, WTR, weight, and BMI may be useful as more complicated measurements, such as PBF by bioelectrical resistance, for identifying groups of individuals whose body habitus places them at high risk of developing NIDDM.