The association between body mass index and the relative frequencies of diseases in a sample of hospitalized patients

Body mass index (BMI) is known to differ among the various diseases, potentially due to etiologic causes, which can lead to bias in estimating the effects of other risk factors. The relationship between BMI and disease must be identified to control for this potential bias in epidemiological investigations. We used the data from our large hospital-based case-control study to analyze the difference in BMI (computed as kg/m2) by diagnosis, separately in males (n = 20,011) and females (n = 9,083) admitted to the hospital between 1977 and 1992. The difference in BMI between diagnoses is most clearly represented through the quartile distributions, whereas the mean tends to camouflage it. Although some associations between BMI and disease differed between the sexes, in general, fractures and diseases of the respiratory tract were associated with the lowest BMI and arthritis, cataract/glaucoma, and endometrial cancer with the highest BMI. Potential disease risk factors, including alcohol use, smoking, and education, showed a strongly negative and age a strongly positive association with BMI in females, but little or no association was found between BMI and these factors in males. The data presented provide information on the correlation between BMI and several diseases and the potential distortion of etiologic factors that can occur as a result of these correlations. The reported associations with BMI can be useful in the evaluation of independent risk factors or when selecting appropriate controls in epidemiological investigations of diseases.