Objective: To assess the ability of the body mass index (BMI) to detect obesity-associated morbidity in subjects with a normal or short stature.
Methods: Information was obtained on 119 975 subjects from a cardiovascular risk factors detection program. Standardized questionnaires were used. Capillary glucose and cholesterol concentrations were measured. Diabetes, arterial hypertension and hypercholesterolemia were selected as end points. Sensitivity, specificity and the likelihood ratio for several BMI thresholds were calculated. ROC curves were constructed to identify the BMI cutoff points with best diagnostic performance. The area under the curve (AUC) was used to assess the proficiency of BMI.
Results: Short stature (height </=150 cm for women or </=160 cm for men) was found in 24 854 subjects (20.7%). These cases had a higher prevalence of type II diabetes and arterial hypertension even after adjusting for confounding variables. In addition, the frequency of the abnormalities was higher even at the lowest BMI values; the prevalence increased in direct proportion with the BMI, but at a lower rate compared to cases with normal stature. The AUC for every co-morbidity was smaller in short stature subjects. The likelihood ratio for detecting co-morbidities increased at the same BMI value in subjects with or without short stature.
Conclusions: The prevalence of obesity-associated co-morbidities is higher in subjects with short stature compared to those without it. The proficiency of BMI as a diagnostic tool is poor in short stature subjects. This problem is not resolved by decreasing BMI thresholds used to define overweight.