We thank Dr. Katz for pointing out the typographical error in our Table 5, in which the data reported were reversed. As stated in the text, more patients with Lyme disease than controls scored greater than 2 standard deviations lower than the mean on two or more memory tests [1]. Among 12 patients who had neurocognitive testing deficits, 4 had had primary neurologic manifestations that were probably related to previous neuroborreliosis (our Table 6). The others had memory deficits on psychometric testing but also had persistent arthralgias and sleep difficulties, which may have affected performance on these tests. Because the cases and controls were balanced with regard to age, sex, and education, differences in neurocognitive testing results were attributed to the presence or absence of previous Lyme disease.

Dr. Katz expressed concern that study patients were vulnerable to recall bias. Because they were "informed that they had received treatment for Lyme disease before optimal treatment protocols were established," they might have been likely to recall aches and pains that others might dismiss. At the time of recruitment, Lyme disease treatment protocols were still evolving, and we doubt whether such patients could classify their treatment as optimal or not. "Case status" might influence reporting of symptoms; however, this factor would not explain the higher prevalence of objective cognitive deficits and abnormal findings on physical examination among patients with Lyme disease compared with controls. Our study was done before fibromyalgia and a syndrome similar to the chronic fatigue syndrome after Lyme disease were reported [2, 3]. We acknowledge that selection mechanisms may have affected our outcomes. However, we believe that the consecutive nature of recruitment and the high rate of study participation (85%) mitigates this effect. We suspect that both cases and controls sought participation in our study because they were concerned about potential exposure to Lyme disease.

In response to Dr. McCaulley, we were conservative in our criteria for neurologic and rheumatologic sequelae of Lyme disease. The 13 patients judged to be persistently impaired were those who met the criteria (for persistent impairment) stated in our article (evidence of joint swelling or pain on range of motion, peripheral neuropathy, or objective memory loss not caused by other systemic illness). Thus, most study patients with previous Lyme disease did not have documented features of Lyme disease, such as frank arthritis or peripheral neuropathy, at the time of clinical assessment. Many patients with frequent arthralgias, fatigue, or difficulty sleeping did not seek further medical attention and managed their symptoms with anti-inflammatory agents or without treatment. A positive serologic test result and earlier manifestations were not part of the algorithm that determined whether patients had Lyme disease sequelae.

The antibiotic regimens that were used to achieve a clinical response in the five responders are noted in our Results section. In brief, two patients responded to 1 month of oral doxycycline, and three others responded to 2 to 6 weeks of parenteral ceftriaxone. With such small numbers of patients, we could find no differences between responders and nonresponders with regard to duration of infection, type of initial antibiotic therapy, or patient age. We agree that fixed injury rather than active infection may cause the symptoms in nonresponders (who are therefore unlikely to benefit from long courses of antibiotic therapy).

It is striking that the study by Dr. Asch and colleagues and ours show a relatively high prevalence of symptoms [1, 2], suggesting that persistent infection, residual injury, or other processes may be important in the pathogenesis of these symptoms. Our study, however, differs somewhat from those of Asch and colleagues [2] and others [3, 4] in that ours was a population-based investigation. Previous reports on the long-term outcomes of Lyme disease have been vulnerable to referral and selection bias that were either hospital- or clinic-based. In studying persons from a geographically defined area, the data more likely approximate the true prevalence of persistent symptoms among persons with previous Lyme disease who were treated in the mid-1980s.

We agree that Lyme disease is being recognized and treated earlier in endemic areas and that the incidence of incomplete recovery from Lyme disease is likely to decrease.