Lyme arthritis was first described in 1975 [2], and as the recognition of the clinical spectrum of infection with Borrelia burgdorferi increased, so did the diagnostic tests for what is now called Lyme disease [3]. The desire for a widely available, quick, inexpensive, sensitive, and reliable test has not yet been realized. Even so, the diagnosis of Lyme disease is becoming more test-oriented, and a positive test result, even in the absence of clinical findings suggesting Lyme disease, is mistakenly thought to be sufficient to make the diagnosis [4]. Serologic testing, most commonly by enzyme-linked immunosorbent assay (ELISA) and immunoblot (Western blot), is widely available, but the antigenic preparations, techniques, and interpretations have not been standardized, although criteria for immunoblot analysis have been proposed [5].

Indeed, the widely used shorthand terms "Lyme disease ELISA" and "Lyme disease test" have given the mistaken impression that these tests are specific for Lyme disease and that a positive test result equates with a diagnosis of Lyme disease. For this reason, I use the more precise term "B. burgdorferi seroreactivity" rather than "positive Lyme disease test." The mistaken assumption that seroreactivity is equivalent to the diagnosis of Lyme disease has led to many misdiagnoses [6-8] and is just another example of imbuing technology with false "power." "Enslavement" may not be too strong a term to describe this dependence on technology, new or old, and the resulting abdication of our roles as decision makers in favor of being collectors of test results.

As many as 3% to 5% of normal people may be seropositive by some assays [4]. False-positive ELISA test results for B. burgdorferi can occur in various inflammatory conditions, including rheumatoid arthritis, systemic lupus erythematosus [9], syphilis, and nonspirochetal infections such as malaria, Epstein-Barr virus infection, and subacute bacterial endocarditis [10]. Kaell and colleagues [10] have clearly shown that patients with nonspirochetal subacute bacterial endocarditis can be seroreactive for B. burgdorferi by ELISA but not have Lyme disease [10]. Finally, B. burgdorferi seroreactivity can persist long after Lyme disease has been treated and cured [11]. Thus, this test, like all others, demands proper interpretation by a careful clinician.

Because of variability in the serologic assays, some clinicians are eschewing serologic tests, stating rather boldly that the tests are unreliable and therefore that seronegativity is not a compelling argument against the diagnosis of Lyme disease. Although seronegativity in the first weeks of Lyme disease is common [3], most patients with later manifestations of Lyme disease will have serum (and often inflammatory fluid) antibodies to B. burgdorferi; thus, "seronegative Lyme disease" is in reality a rare entity. However, seroreactivity to B. burgdorferi by ELISA is not proof that the patient has Lyme disease [6-8]. Over-diagnosis in both seropositive and seronegative patients has contributed to an epidemic of "Lyme anxiety"

A recent addition to the diagnostic armamentarium in Lyme disease is the polymerase chain reaction (PCR) assay. This test, like culture of the organism, directly identifies the pathogen rather than measuring the host's immune response to it. Polymerase chain reaction can detect the DNA from as few as one to five organisms, even those that are nonviable. Different specific probes have been produced, and PCR has been used to detect B. burgdorferi DNA in many body fluids [12]. The appeal of PCR lies in its rapid turnaround time (2 days compared with 6 to 8 weeks for culture) and the avoidance of the difficulties of identifying the organism by culture or immunohistochemistry.

However, the very sensitivity of PCR can be a limitation if samples are not collected and analyzed meticulously. Contamination by even a single organism at any stage can cause a false-positive result. A true-positive PCR test result shows that B. burgdorferi DNA is present in that sample, but PCR cannot differentiate between live and dead organisms. Thus, a positive test result can suggest previous B. burgdorferi infection and the diagnosis of Lyme disease but is not proof of persistent infection. Borrelia burgdorferi produces large numbers of blebs, which are small membrane-bound bodies derived from outpouchings of the organism. Many of these contain B. burgdorferi DNA [13] and may persist in the synovium long after the organism has been killed and eliminated. Thus, blebs might be capable of producing a positive PCR result after the infection has been eradicated. Finally, false-negative results may occur because of the presence of polymerase "inhibitors" in some body fluids (for example, hemoglobin) or because of the paucity of organisms. Despite a very high specificity, the sensitivity may be as low as 70% [14] in the diagnosis of Lyme disease. Thus, some of the original enthusiasm for the test has paled. However, the precise sensitivity and specificity of the many PCRs for B. burgdorferi are not yet known; indeed, the central question in any study of the sensitivity and specificity of a test for Lyme disease is what, aside from well-defined erythema migrans, constitutes the disease; that is, how good is the independent a priori criterion or "gold standard" for defining the full clinical spectrum of Lyme disease?

We are probably beginning to emerge from the phase of disillusionment with PCR to a more balanced, realistic phase. The article by Bradley and colleagues [15] in this issue of Annals is a good example of a reasoned use of PCR. It is clear that infection with B. burgdorferi can cause Lyme arthritis, even in patients who have received previous adequate antibiotic therapy. However, the immunopathogenesis of synovitis is unclear [16]. Amplifiable B. burgdorferi DNA in the synovial fluid of patients with Lyme arthritis suggests that a local infection was present at some time and that it may have played a role in the initiation of synovitis.

However, the positive PCR results do not prove that the continuing presence of a viable organism is necessary for, or even occurs in, chronic synovitis; recall that Bradley and colleagues [15] could not grow the organism. Thus, a positive PCR result obtained from the synovial fluid of a patient who has already received adequate antibiotic therapy is not proof that further antibiotic therapy is indicated. On the other hand, treatment is probably indicated if a positive PCR result is obtained from the synovial fluid (processed carefully in a reputable laboratory) of a patient who has never been treated for Lyme disease. This would be also be true when tests of cerebrospinal fluid or other clinical samples were positive by PCR. If a fluid previously shown to test positive by PCR reverted to negative after the patient received appropriate antibiotic treatment, the therapy might be considered successful [17].

What then is the role of the PCR test? It may be useful in diagnosing early Lyme disease when the patient is still seronegative. Schwartz and colleagues [18] report that skin biopsy specimens from patients with erythema migrans, the pathognomonic skin lesion of Lyme disease, were often positive by PCR when the same sample was negative by culture and the patient was seronegative [18]. Similarly, cerebrospinal fluid may be positive by PCR when no measurable antibody to B. burgdorferi in serum or fluid is present and the cerebrospinal fluid is negative by culture [17, 19].

How much PCR adds to other immunologic assays of inflammatory fluid awaits comparison studies of PCR of synovial fluid with cellular [20] or antibody [21] immune reactivity with B. burgdorferi. Polymerase chain reaction may be more sensitive than antibody detection techniques in human Lyme neuroborreliosis [17, 19] and the murine experimental model [22] and clearly is more sensitive than current culture techniques [15, 18]. Our experience suggests that a few patients may be positive by PCR despite negative immunologic assay results in inflammatory fluid and blood (Sigal LH and Liebling M. Unpublished observation).

Thus, PCR represents a new, powerful tool for detecting "footprints" of B. burgdorferi, if not the presence of a live organism. When interpreted with knowledge of the clinical context and the limitation of the technique, PCR results can aid in making sound clinical judgments. When PCR and other tests are done without attention to these details, the resulting decisions may lead to added expense and suffering [5-7]. In this new era of PCR and other elegant diagnostic tools, it is important to avoid being enslaved by them and to remember that their value is only as great as the skill and judgment of the clinician using them.