BMI is widely used to measure obesity in outcome studies, but other easily collected measures are available to characterise bodily shape. OBJECTIVE: The objective of this study was to determine whether BMI and waist-to-hip ratio (W:H-R), as standard measures of obesity, collectively better explained the likelihood of obesity-related complications. METHODS: This retrospective study used data from the England Health Survey, a national representative survey of the general population. Data describing BMI, obesity related diseases and other anthropometric measurements were abstracted from the survey 1991 to 1994, and 1997 to 2002. Using this data, a model was developed to account for various factors that could effect the development of obesity related diseases: diabetes, stroke, heart attack, hypertension and other vascular diseases. Using logistic regression, the association of between BMI categories further broken down by W:H-R categories was determined. The results were standardized for age, sex and ethnicity to generate standardized odds ratios comparing the accuracy of different anthropometric measurements and its combinations to predict obesity related diseases. RESULTS: BMI and W:H-R were independent predictors of disease in obesity. Using both these measures a more precise prediction of the likelihood of developing obesity related complications. Odds ratios in the group with the highest BMI of the sample varied between 1.6 and 2.8 conditional on their W:H-R. Patients with a medium BMI but high W:H-R had a higher relative risk (odds ratio 1.8) than patients with a high BMI but low W:H-R (odds ratio 1.6). These patterns were even more obvious in diabetic patients. Similar results could be shown by combining BMI and waist circumference only. CONCLUSIONS: BMI in addition to W:H-R or waist circumference is a better and more precise predictor of obesity-related complications, and should where possible be preferentially used. Longitudinal obesity-related outcome data are needed.