OBJECTIVE: To determine possible differences in continuation among women initiating treatment with the selective estrogen receptor modulator raloxifene, versus those initiating treatment with estrogen-containing regimens. DESIGN: A pharmacy prescription database search for refill patterns. The study subjects were members of Kaiser Foundation Health Plan, a large health maintenance organization; 1,394 women age >or=60 years who filled index prescriptions for either raloxifene (n = 331) or systemic estrogens (n = 1,063) between April 1998 and March 1999. The main outcome measure was discontinuation based on prescription refill patterns through December 2000. RESULTS: At 24 months, the probabilities of discontinuing were 56% for women starting raloxifene compared to 72% for women starting estrogens. The likelihood of discontinuation was significantly less among women starting raloxifene than among those starting estrogen (hazard ratio = 0.75; 95% confidence interval = 0.64-0.88). Adjustments for age and prescriber specialty did not affect the risk. CONCLUSIONS: We conclude that discontinuation of estrogen by women well beyond the age of menopause is high; more than two-thirds discontinue within 2 years of starting. Women starting therapy with raloxifene are 25% percent less likely to discontinue their medication than those starting estrogen, providing some promise that long-term benefits of raloxifene may be more easily achieved than those of estrogen.