BACKGROUND Lipodystrophy studies in HIV-infected patients have usually defined abnormalities in body fat by clinical evaluation and patient questionnaires. Despite the risk for bias with these subjective approaches, agreement analysis among the large number of variables employed was seldom performed. OBJECTIVE To analyze consistency between the usual approaches for definition of abnormalities in body fat distribution. DESIGN We evaluated agreement between the clinical and questionnaire findings for abnormalities in body fat in an HIV patient population under antiretroviral treatment followed in our institution, using different criteria for definitions of body fat abnormalities within the same data set. METHODS Kappa analysis for consistency and receiver-operator characteristic (ROC) curve analysis were performed. RESULTS Low levels of agreement between clinical and patient perspectives were observed. Only one combination of criteria showed adequate agreement results. The waist/hip ratio showed low levels of agreement with all other variables, and no clear discriminative point was observed by ROC curve analysis. The ratio between the trunk fat content and the leg fat content assessed by dual energy x-ray absorptiometry (DEXA) scan demonstrated better agreement and more clear discriminative values for both male and female patients. CONCLUSION Agreement analyses may help in the selection of the subjective variable methodology and in the inclusion of consistent and nonredundant objective measurements for diagnosis of abnormalities in body fat.