Ordinal logistic regression models were constructed to examine the effect of vitiligo extent on QOL, using the DLQI to categorize the effect as none or mild (0-5), moderate (6-10), and very or extremely large (>10), as previously reported.7 Individual aspects of the DLQI were divided into 4 responses (not relevant/not at all, a little, a lot, and very much) as the dependent variables. The independent variables were affected BSA of at least 25%, the number of body parts affected, laterality and distribution of vitiligo lesions, and duration of vitiligo (in tertiles). Retention of covariates in multivariate models was based on consideration of confounding and effect modification. Sex, current age (continuous), and comorbid autoimmune disease (binary) were evaluated individually and in combination. Any variable that had a P value of less than .05 or changed the parameter estimate for the independent variables of interest by more than 10% was retained as a covariate in that model. Adjusted odds ratios (AORs) were calculated by including them in the full models. Linear interaction terms were tested and were included in final models if significant (P < .05). Contrast statements were used to generate estimates for AORs, 95% confidence intervals, and P values for each combination of predictors.