Background Krakauer and Krakauer [2012] have developed an obesity measure, the ABSI (A Body Shape Index), based on waist circumference, height and weight, which has particular sensitivity to abdominal fatness. It is a strong predictor of all cause mortality in the USA population sample of 14,105 adults in the National Health and Nutrition Examination Survey (NHANES) 1999–2004. The ABSI is based on a novel logic. It is in the form of the ratio (waist measured) to (waist predicted from height and weight). It has little correlation with BMI and is designed to identify excess waist size irrespective of BMI.

Methods ABSI-UK was constructed in the same way as the Krakauer and Krakauer original version but from parameters of height, weight and waist from the Health Survey for England. The usefulness of the ABSI to predict CHD and all cause mortality in the 23 year follow-up of the 9003 men and women participants in the UK Health and Lifestyle Survey was compared with the usefulness of: ABSI-UK, BMI, waist, linear combination of BMI and waist, waist to hip ratio, and waist to height ratio, using Cox survival analysis adjusted for age and sex.

Results The ABSI indices were found to be the strongest predictors of all cause and CHD mortality in this sample. Differences in the relative distribution of height between the UK and the USA mean that ABSI-UK is a better predictor in the UK sample, but the improved performance may not be sufficient to justify a different measure. For all-cause mortality the rank ordering by effectiveness from best to worst by Wald statistic was ABSI-UK (60.5), ABSI (47.7), linear combination of waist and BMI (24.3), waist to height ratio (21.4), waist to hip ratio (12.8), waist (11.0), and BMI (0.117). BMI did not reach statistical significance whereas ABSI-UK had a HR of 1.50 (95% CI 1.35, 1.66).

Conclusion Where waist measurement is available in addition to height and weight, ABSI appears to be the preferred choice of obesity measure. Further tests of the ABSI obesity measure should be carried out in different samples. The role of ABSI as a confounder in analyses of the effects of other behaviours such as dietary measures should be investigated.