Introduction Predictive equations for DXA measures of total and abdominal adiposity from simple anthropometry for an Indian population will enable more accurate estimates of adiposity to be derived in other studies.

Methods Using 1230 men and 636 women from Hyderabad, India, multiple linear regression was performed to generate predictive equations for DXA measures of total body fat (grams) and total abdominal fat in the L1–L4 intervertebral region (grams). Equations were developed separately for each sex on a training set (60% of sample) and tested on the validation set (40% of sample).

Results For total body fat in males, a simple equation based on two variables (height and weight) gave an R² of 0.83 and SEE (SE of the estimate) (square root of the sum of observed- predicted values/number of observations) of 2200 g whereas a more complex equation (additionally including triceps skinfold, waist circumference and calf skinfold) gave an R² of 0.94 (SEE=1600 g). In females, hip circumference and calf skinfold alone explained 92% of the variance in total body fat (SEE 2300 g) increasing to 95% with waist circumference, subscapular skinfold, weight and calf circumference included (SEE 1800 g). Waist circumference was the best predictor of fat in the L1–L4 region for both sexes (R²=0.88 and 0.89 respectively); more complex equations achieved an R² of 93%. Predictive equations for both traits produced an SEE<0.5 SDs, indicating good accuracy.

Conclusion DXA measures of adiposity can be derived with a high degree of precision from simple anthropometric measures in an Indian population.