OBJECTIVES: To compare the incremental medical expenditure associated with alternative disease modifying anti-rheumatoid drug (DMARDs) choices in Rheumatoid Arthritis. METHODS: Retrospective cohorts were constructed from California Medicaid paid insurance claims between January 1, 1998 to December 31, 2005. Non-overlapping monthly panels were created from pharmacy claims for biologic (adalimumab and etanercept) and standard (methotrexate, lefluonomide, hydroxychloroquine and sulfasalazine) DMARDs. Final sample included 59,788 observations on 7,025 patients. Covariates included age, gender, race, location of beneficiary’s county in either Northern or Southern California, population density in beneficiaries county, exclusive fee-for-service reimbursement used in beneficiary’s county, Medicare and Medicaid dual eligibility, Charlson comorbidities index excluding Rheumatoid arthritis, and expenditures associated with pharmacy, out-patient, inpatient, inpatient-MD, LTC, and ER visits in the 3-months prior to treatment. We compared parameter estimates between naïve fixed effects (FE) and instrumental variables based fixed effects (IV-FE) panel data models. First lag of the observed treatment served as the instruments for the endogenous variables in IV-FE models to mitigate time-varying endogeneity. The primary dependant variable was total monthly expenditure. Secondary analysis included monthly expenditures associated with pharmacy, out-patient, inpatient, inpatient-MD, LTC, and ER visits.    RESULTS: Based on the FE model, as compared to methotrexate, incremental monthly total expenditure associated with adalimumab ($1623.4, p <0.001), etanercept ($1185.3, p <0.001) and leflunomide ($467.3, p <0.001) was significantly higher. Based on the IV-FE model, total expenditure associated with adalimumab ($2129.9, p <0.001), etanercept ($1604.1, p <0.001) and leflunomide ($686.8, p <0.001) exhibited significant increase in magnitude of the parameter estimates, again with baseline as methotrexate. Under identification test based on Anderson’s canonical correlation LM statistic, strongly rejected the null hypothesis in all the IV-FE models.      CONCLUSIONS: The incremental acquisition cost associated with adalimumab, etanercept and leflunomide may not be offset by commensurate reductions in routine and catastrophic resource utilization in the California Medicaid population.