Objectives To compare the EQ5D index from the UK, US, and Japan preference weights models in patients with rheumatoid arthritis (RA), and to examine the relationship of each index with clinical outcomes for developing mapping algorithm.
Methods GO-FURTHER was a multicenter, randomized, placebo-controlled study. Adult patients with active RA were randomized to PBO + MTX (PBO group) or GLM (2mg/kg) plus MTX at wk 0, 2, and every 8 wk thereafter (GLM group). Patients in PBO group with <10% improvement in tender and swollen joint count from baseline at wk 16 entered early escape (EE) and received a 2 mg/kg GLM infusion at wks 16 and 20. Clinical and patient-reported outcomes measures including EQ5D domains, EQ VAS scale, Health Assessment Questionnaire (HAQ) were collected. The EQ-5D index was calculated using preference weights from the US (D1) 1, UK (N3)2 and Japanese3 population weights. Pearson's correlation and linear regression model were used to test correlation and established mapping algorithms. Comparisons between groups in EQ5D index were performed using t-test.
Results The study population had moderate to severe RA with a HAQ score of 1.6 and DAS28 score of 5.9 at the baseline. The UK model predicted a lower EQ5D index (0.36±0.33), compared to the US model (0.54±0.21) and Japan model (0.52±0.13).The differences in EQ5D index between UK and US or Japan model were moderate in patients with mild to moderate disability (HAQ-DI<1) (0.64 vs. 0.72 and 0.64), but were more than twofold in patients with severe disability (HAQ-DI>1.5) (0.19 vs.0.42 and 0.45). The Pearson's correlations coefficient of EQ5D index from the US model with HAQ, DAS28 score, pain, SF-36 PCS and MCS was -0.70, -0.40, -0.51, -0.57, -0.55, respectively (all p value <0.001), and were highly consistent across 3 models and sensitive to change. Comparing mapping algorithms based on linear regression models, HAQ model was equally predictive to each EQ5D index. However, the UK N3 model was more responsive to active treatment compared to US D1 or Japan model. Compared to PBO+MTX group, the change in EQ5D index in GLM+MTX group at wk24 were 0.26±0.32 vs. 0.11±0.37 in UK model (p<0.001), 0.17±0.21 vs. 0.11±0.23 in US model (p<0.01), and 0.13±0.16 vs. 0.05±0.17 in Japan model (p<0.001).
Conclusions Although HAQ was highly predictive to EQ5D index derived from US, UK or Japan model, the EQ5D scores from each model were not equivalent, and were significantly influenced by disease severity, warranting caution in the cost-utility analysis across trials.
Dolan P: Med Care 1997, 35: 1095-1108
Shaw JW, et al: Med Care 2005, 43:203-220
Tsuchiya A, et al Health Econ 2002, 11:341-345
Disclosure of Interest C. Han Employee of: Johnson & Johnson Pharmaceutical Services, LLC, C. Bingham Grant/research support from: Janssen Research & Development, LLC., R. Westhovens Grant/research support from: Janssen Research & Development, LLC., M. Weinblatt Grant/research support from: Janssen Research & Development, LLC., L. Kim Employee of: Janssen Research & Development, LLC., D. Baker Employee of: Janssen Research & Development, LLC., S. Peterson Employee of: Johnson & Johnson Pharmaceutical Services, LLC, E. Hsia Employee of: Janssen Research & Development, LLC.