relationship between BMI and outcome variables was examined using categorical variables. The skewed distribution of outcomes, even after transformation, and the presence of variables, such as the M-HAQ, with multiple scores of 0 were the reasons for transforming a continuous into a categorical variable. We recognized that power was reduced by dichotomizing variables. However, even with this conservative approach, the results are striking.Third, they raise the possibility that, despite the fact that we found no ethnic difference in CRP concentrations (median 3 mg/liter [interquartile range [3][4][5][6][7] in Caucasians compared with 4 mg/liter [interquartile range 3-7] in nonCaucasians; P ϭ 0.20), our finding of higher CRP concentrations in obese patients could merely reflect a greater proportion of non-Caucasians in the overweight and obese groups, because some studies have shown higher concentrations of CRP in African Americans. This is not the case; when race (categorized as Caucasian and non-Caucasian) was included in the model, the association between BMI and CRP remained significant (P Ͻ 0.001 for the unadjusted and all adjusted models). Furthermore, in a subgroup analysis that included only Caucasian patients, the relationship between BMI and CRP remained significant (P Ͻ 0.001 unadjusted, adjusted for age and sex, and adjusted for age, sex, SLEDAI, and SDI).By identifying obesity as a contributor to quality of life in patients with lupus, we have drawn attention to a modifiable factor that may not only increase cardiovascular risk, but also decrease functional capacity.