OBJECTIVES: To measure the impact of obesity on annual medical and productivity costs among working U.S. adults with arthritis. METHODS: We conducted a cross sectional study using Medical Expenditure Panel Survey data from 2003-2007. Working adults with arthritis (18-64 years old) were selected if they did not have pregnancy, malignancy, kidney dialysis, immunodeficiency, low body mass index (BMI<18.5 kg/m2), or unemployed status. Patients with arthritis were identified by ICD-9 codes of 714-715 or via patient self-report. Obese and normal- weight were defined as BMI of ≥30 kg/m2 or BMI of 18.5-<25 kg/m2, respectively. Loss of productivity was estimated by loss of workdays due to illness or injury and standard hourly wage by occupation. Medical costs were estimated using a generalized linear model with a log link function and gamma distribution. Costs of productivity loss were calculated using a two-part model to adjust for patients with zero costs. Using Oaxaca decomposition, differences in treatment costs between obese and normal-weight patients were decomposed into two parts: a) differences in characteristics (endowments) across groups, and b) differences between obese and normal parameters (coefficients). Costs attributable to obesity were defined as the costs by coefficients component. All costs were converted to 2009 U.S. dollars using price indices.  RESULTS: Among the 7345 working adults with arthritis, prevalence of obesity and normal-weight was 24.8% vs. 40.7%, respectively. The difference in medical costs between the groups was US$2380 (95%CI:US$1934-2825) due to endowments and US$379 (95%CI:US$367-392) due to coefficient components. Productivity loss costs in the obese patients were higher, at US$46(95%CI:US$ 41-51) due to endowments and US$441(95%CI: US$435-447) due to coefficient components. CONCLUSIONS: Use of the Oaxaca decomposition method suggested that the economic burden, particularly productivity loss costs, of obesity in patients with arthritis was substantial, adjusting for characteristics across groups.