In response to Dr. Jacobson's letter, although 27 participants did receive antibiotic prophylaxis, they constituted only 21% of the 127 participants who had dental treatment in the 3-month period. More important, among those with indications for antibiotics (that is, patients with previously known cardiac valve abnormalities undergoing dental work), only 17 of 29 (59%) case-patients with dental treatment received antibiotic prophylaxis, compared with 3 (50%) of 6 controls (P = 1.0). In fact, the unadjusted odds ratio for the association between dental therapy and infective endocarditis was 0.5 (95% CI, 0.01 to 9.6) among those with antibiotic prophylaxis, compared with 0.3 (CI, 0.01 to 4.2) among those without antibiotic prophylaxis. Thus, not only were our results not due to widespread use of prophylaxis, but they also did not suggest a protective effect.

Mr. Adams recalls his history of mitral valve prolapse and two episodes of endocarditis, each following dental treatment. His unfortunate experience is certainly consistent with our results, which show a 19-fold elevated risk for endocarditis associated with mitral valve prolapse and a 37-fold increased risk associated with prior endocarditis. That his episodes of endocarditis were preceded by dental treatment, however, does not necessarily imply causation. Indeed, as our data show, while a large proportion of case-patients had recent dental treatment, the proportion of the general population reporting recent dental treatment was identical. Anecdotes of this type illustrate the need for controlled studies such as ours. Decisions about therapeutics and public health recommendations should not be based solely on collections of uncontrolled case reports.

Finally, Dr. Collignon asks about invasive dental procedures, stating that 18 case-patients and 7 controls had any of three specified invasive procedures; comparable numbers were 6 and 0 when the procedures were restricted to extractions. In fact, 36 (13.2%) case-patients and 27 (9.9%) controls had invasive dental treatment, a difference compatible with random variation (P = 0.2). A single statistically significant increased risk was observed with extractions, but this finding was based on small numbers, could not be adjusted for confounding, and was not apparent in the subgroup of patients with cardiac valvular abnormalities (the target of prophylaxis) or in those infected with dental flora. In addition, this increased risk was not confirmed in the other study on this question, which observed an odds ratio of 0.7 (CI, 0.3 to 1.8) with extractions (1). Thus, although we cannot exclude extraction as a cause of endocarditis, we are concerned that this may simply be a false-positive finding due to the multiple analyses we performed.