Mr. Ferguson suggests that empirical antibiotic therapy may be more cost-effective for patients whose symptoms may or may not represent late-stage Lyme disease. Our results do not support this view. Although we were pleased to read that his symptoms improved after a brief course of tetracycline therapy, we do not know whether the institution of antibiotic therapy and resolution of symptoms are causally related.

Dr. Cherney raises several concerns about our study. Given our stated research objectives, she questions whether biased assessments of study patients could have occurred. Primary outcomes were abstracted from patients' medical records by investigators who had no knowledge of the patients' Yale Clinic diagnosis. Assigning a diagnosis of active, previous, or no evidence of Lyme disease, however, required that study physicians carefully review patients' records. We do not believe that this requirement led to misclassification of patients' diagnostic status or threatened the validity of our findings. The possibility of diagnostic misclassification was also raised for three patients who were given a diagnosis of previous but not active Lyme disease and who had peripheral neuropathies. We agree that the most difficult diagnostic challenges exist in such patients with new, persistent, or recurrent neurologic or arthritic symptoms.

Dr. Cherney also questions how so many patients with no evidence of Lyme disease could have tested positive in outside laboratories but not in our own. Our results are consistent with those of previous investigations [1, 2] that showed high rates of seropositivity among similarly classified patients. Lack of standardization of the testing process [3] in combination with multiple testing is one potential explanation. Reported specificities range from 68% to 100% for Western blot assays [4] and from 73% to 100% for enzyme-linked immunosorbent assays [3]. The "average" specificity for these tests across all laboratories is, of course, unknown. If one assumes an average specificity of 85%, then a person who undergoes five separate assays would have a 0.85⁵ = 0.44 chance of all test results returning normal. Conversely, this person's likelihood of receiving one abnormal result by chance alone is 56%.

Although reports have claimed benefit from long-term antibiotic therapy [5], our data suggest that patients with persistent symptoms after documented Lyme disease do not improve with this treatment approach. We agree with Dr. Cherney that more appropriate therapies are needed for these patients. We share her view that well-designed clinical trials that produce effective treatments will benefit both patients and physicians. Finally, effective therapies are also needed for patients who present with nonspecific symptoms and have no evidence of previous or current Lyme disease.