OBJECTIVES RA clinical studies using the SF-36 Physical Function (PF) scale and Health Assessment Questionnaire (HAQ) have identified limitations in each instrument's sensitivity across the full range of disease severity. Item Response Theory (IRT) estimates were used to develop a composite of both instruments (PF–HAQ) to provide a more sensitive measure of physical health. METHODS Data for 799 patients from a 2-year randomized control study of adalimumab in early RA (<3 years) were employed. Patients received adalimumab plus methotrexate; adalimumab monotherapy; or methotrexate monotherapy. Composite PF–HAQ scores were compared individually with PF and HAQ using 1) comparison of floor and ceiling effects; 2) ANCOVA models with ACR criteria classification or treatment as factors (covariates: sex, age, BMI); and 3) receiver operating characteristics (ROC) analyses using ACR50 criteria as a gold standard. RESULTS At baseline, 6.2% of patients were at floor for the PF. At endpoint, 37.7% and 14.2% of HAQ and PF scores, respectively, were at ceiling. IRT scores, by definition, have no ceiling or floor. Significant differences across treatment groups were obtained (F [2526] = 12.21, 7.48, 3.02, p<0.001 for all comparisons). PF–HAQ had significantly more power than either individual scale to detect treatment differences (F-statistic ratios of PF–HAQ with PF and HAQ were 1.6 and 4.0, respectively). PF–HAQ was better than PF at detecting differences in endpoint ACR criteria (F-statistic ratio for PF–HAQ to PF was 1.28). ROC analyses indicated that PF–HAQ provided better measurement precision (p<0.001) vs. PF and equivalent/better precision vs. HAQ (p=0.14) (AUC: PF 0.80, 95% CI: 0.79–0.81; HAQ: 0.83 [0.82–0.85]; PF–HAQ: 0.84 [0.83–0.85]). CONCLUSIONS Combining PF and HAQ measures into a single measure eliminated floor/ceiling effects, and provided greater efficiency in discriminating treatment effects, as well as greater sensitivity vs. common diagnostic criteria.