The effects of removing propoxyphene napsylate products from the Wisconsin Medicaid drug program formulary were examined. Internal analgesic expenditures and usage data for 3-month periods before and after the removal were compared (April through June 1984 versus the same period in 1985). After adjusting for price and reimbursement changes between the two study periods, overall expenditures were slightly higher after removal of these products. Expenditures per recipient, prescription, and unit all increased, as did the number of prescriptions per recipient. Expenditures, prescriptions, and recipients increased more for propoxyphene hydrochloride products as substitutes for propoxyphene napsylate products than for products in any other category. Increases also occurred for nonsteroidal anti-inflammatory products, suggesting they may have been chosen as replacement therapy. The proportion of napsylate prescriptions converted to hydrochloride prescriptions was larger for institutional patients than for noninstitutional patients. Although program expenditures did not decrease, as intended by the formulary change, other qualitative outcomes also should be considered, such as any therapeutic advantages the replacement products may have had for the patients.