Dr. Bubb finds my reasons for limiting the use of misoprostol to patients at greatest risk to be unconvincing. He believes that "good reasons exist to limit the population in which misoprostol is used" but does not enumerate them. I suggest that because misoprostol's effectiveness in preventing serious gastrointestinal complications [1] is unrelated to any demographic factor, the only reason to limit its use in a given population is its cost-effectiveness based on the predicted prevalence of serious complications. This fundamental concept in medical decision making, although perhaps not widely used by some, has been well described by others [1]. Silverstein and colleagues [2] did not measure costs, but I believe that my editorial acknowledges the limitations that this factor placed on my estimates. Dr. Bubb, however, proposes that "the avoidance of a medical work-up for patients with gastrointestinal symptoms but no serious complications" will reduce costs; this statement suggests that he missed the point that such symptoms were more common in the misoprostol group and were an added cost that I elected to exclude from my analysis.

Ms. Moore notes a recent article in which nabumetone, when compared with ibuprofen alone, was associated with a small number of gastrointestinal ulcers, an incidence similar to that observed with ibuprofen plus misoprostol. I assume that this point is an indirect criticism of my inclusion of ibuprofen as a medication associated with decreased gastropathy. In fact, however, I was not suggesting that the two medications were equivalent in this regard but wanted to indicate that, in practice, many rheumatologists are uncertain about the anti-inflammatory efficacy of nabumetone at a dose of 1 g per day. Because I am unaware of any thorough prospective outcome studies comparing these two medications in terms of functional improvement, I attempted to provide some practical choices as counterpoints to the use of other NSAIDs (for example, piroxicam or naproxen) that are associated with a greater estimated risk for upper gastrointestinal complications [3].

After rereading Dr. Silverstein's letter, it seems that he and I agree that 1) NSAID-related complications represent an important problem; 2) his study [2] has shown efficacy for misoprostol; 3) it would not be appropriate to treat every patient receiving NSAIDs with misoprostol; and 4) the use of demographic factors to select a high-risk population aids the clinician in the appropriate use of misoprostol.