Dr. Kaplowitz and his coauthors and I agree on many important points. We agree that establishing causality beyond a reasonable medical doubt in cases of possible adverse drug reactions is difficult, especially when cases are infrequent (perhaps not much more common than "background noise") and the drug in question is new. We agree on the importance of postmarketing surveillance to help identify new adverse drug reactions and on the advantages of using MedWatch reports and published case reports to benefit from the experience and wisdom of practicing physicians. Furthermore, we agree that it is difficult to know how often an adverse drug event triggered by one member of a drug class extends to some or all members of that class.

However, we appear to disagree somewhat concerning how to think about pioglitazone or rosiglitazone when we prescribe it for our diabetic patients. In my editorial on the case reported by May and colleagues (1), I concluded that "pioglitazone probably caused the symptomatic hepatitis" but that is was difficult to be more definitive "without additional prior case reports and without a positive drug rechallenge." I also concluded that it was a bit too early to be sure if there would be a class effect but that it might be wise for physicians to prepare for that possibility. I cited the official recommendations of pioglitazone's manufacturers (Eli Lilly and Co. and Takeda Pharmaceuticals North American), which are approved by the U.S. Food and Drug Administration, and offer practical, reasonable suggestions for performing liver function tests before prescribing the drug to a patient; avoiding the drug in patients with underlying significant liver inflammation; following results of liver function tests periodically; and discontinuing therapy with the drug if clinically significant, persistent hepatitis develops. GlaxoSmithKline, the manufacturer of rosiglitazone, includes very similar precautions and recommendations in its package insert (2).

Dr. Kaplowitz and his coauthors state that they believe that it is "very difficult to conclude that the few reported hepatitis cases associated with rosiglitazone or pioglitazone are truly drug induced" and that "there is no scientific basis to consider severe liver injury as a class effect of thiazolidenediones." They may well be correct on both counts. However, their approach may perhaps reflect the point of view of consultants to GlaxoSmithKline, that a drug is "innocent until proven guilty beyond a reasonable doubt." My approach, as a practicing internist and clinical pharmacologist, might be closer to the following: "These drugs could be innocent of inducing hepatitis in patients, but there is now growing anecdotal evidence that they might do this, so let's prescribe them using some reasonable guidelines, similar to those negotiated by the U.S. Food and Drug Association and the manufacturer." Perhaps Drs. Kaplowitz, Lewis, and Watkins and I are simply coming to slightly different conclusions about similar preliminary data and case reports, depending on whether we view these data through rosi-tinted glasses.