LYMENET RESPONSE TO
Although Lyme disease has been recognized as a distinct clinical entity for nearly 20 years (1), there are still gaping holes in our knowledge about the most fundamental questions of incidence, diagnosis and treatment. This is due primarily to the many difficulties posed by the nature of the illness (non-specific symptoms, protean manifestations, long latency period and/or apparent asymptomatic infection in some individuals, etc.) and our limited arsenal in the areas of both testing and treatment. The resulting knowledge gap leaves room for reasonable people to disagree on almost every facet of Lyme disease research.
A recent study by Steere et al. (2) published in the April 14, 1993 issue of the Journal of the American Medical Association (JAMA) suggests that Lyme disease is widely overdiagnosed. Because the study's conclusions are so striking and its primary author is considered one of the world's leading experts in Lyme disease research, the publication of this article has generated considerable medical and media attention and will likely have significant impact on how Lyme disease is perceived in both the medical and lay communities.
We believe that the study contains multiple serious flaws which leave its conclusions unsupported. We are concerned that this paper's publication may lead to ill-advised complacency about Lyme disease in this country, may compromise patients' access to treatment at all stages of the disease -- thus exposing more of them to the risk of potentially devastating long term sequelae -- and may even influence the amount and direction of future allocations of funds for Lyme disease research. We thus feel it imperative to address the methodological and philosophical flaws in this work.
Our specific concerns about the Steere et al. paper are as follows:
The authors report that 98% of the patients (176 out of 180) found to have active Lyme disease, but none of the patients (0 out of 452) who had never had Lyme disease but who were evaluated for suspected Lyme at their clinic, were seropositive by enzyme-linked immunosorbent assay (ELISA) and/or Western Blot. If this is correct, it means the authors have developed an extremely sensitive and specific group of tests that vastly outperform any of the existing antibody tests or testing protocols that have been reviewed in the Lyme literature. Indeed, the authors state that of the 452 patients in the study who were determined to have never had Lyme disease, 203 (45%) had obtained "false" positive results from another laboratory.
It is difficult to accept uncritically the authors' claim that their antibody testing protocols are superior to others currently in use. Independent reviews of a wide variety of today's antibody detection tests -- immunofluorescent assays, enzyme-linked immunosorbent assays and assorted immunoblots -- have indicated generally dismal performance, marked by significant interlaboratory and intralaboratory variability (3-5). The authors offer no independent evaluations or persuasive arguments to distinguish their tests from others in current use. Instead, it appears that Steere and his co-authors are relying inappropriately on imperfect testing techniques -- from a single laboratory -- to make the diagnosis of Lyme disease. The presumption seems to be that their tests are better than others because the correlation between seropositivity and actual Lyme disease is highest; on the other hand, the definition of "actual Lyme disease" in the study is derived almost exclusively from their test results. The reasoning is entirely circular.
Moreover, the nearly exclusive reliance on serologies for diagnosis leads to the systematic underestimation of the prevalence of seronegative Lyme disease. This point cannot be overemphasized. Although false negative serologies are widely recognized as common in early Lyme disease, it is often claimed that they are extremely rare phenomena later in the course of the illness (6-8). The many cases of seronegative, culture-positive "late" Lyme disease that have been identified and reported (9-12), however, make this claim even more untenable than it is unverifiable. In one study on transplacental transmission of Borrelia burgdorferi (10), over half of the mothers who had adverse pregnancy outcomes and whose fetuses or neonates demonstrated the presence of B. burgdorferi were themselves seronegative. Thus, when the study population is determined by factors other than serology -- when the inevitable bias toward seropositivity is removed -- the true incidence of seronegative Lyme disease emerges.
One example from this study merits special attention:
"A 24-year old white woman was admitted in February 1985 in labor at term of her pregnancy. Ultrasound examination showed that the fetus was dead when she arrived at the hospital. Following the delivery of her stillborn infant and completion of the fetal autopsy, a retrospective interview established that she had acquired Lyme borreliosis in the first trimester of her pregnancy outside of Salt Lake City, Utah. Postpartum serological studies yielded conflicting results because the Centers for Disease Control found strongly reactive results by IFA and ELISA, as did the New York State Department of Health; however, the Yale University laboratory of Dr. Allen Steere could detect no evidence of specific antibodies for B. burgdorferi. Fetal viscera showed B. burgdorferi in the liver, adrenal, brain, heart and placenta. . . "
The implications of this finding for the patients studied in "The Overdiagnosis of Lyme Disease" are obvious and profound, given that over a quarter of the study population was determined by the authors to have received "false positive" test results from other laboratories. False negatives in Dr. Steere's laboratory may very well account for a significant portion of the discrepancy among results.
Steere et al. further compromise their estimates of the numbers of patients with Lyme disease by creating inappropriate conditions for diagnosis which in some cases are impossible to meet. The authors state that "a history of exposure in an area where B. burgdorferi has been recovered from ticks" was required for a diagnosis of Lyme disease in their study. This approach systematically excludes all patients from areas that have not been investigated for B. burgdorferi infestation. In light of the fact that thousands of clear-cut cases of Lyme disease, complete with physician-verified erythema migrans, and/or clinical findings and positive serologies, have been reported from "non-endemic" and unstudied areas, such a restriction is inappropriate. The patient in the study quoted above apparently contracted Lyme disease in Utah and was also seronegative on Dr. Steere's ELISA. Either of these two factors alone would have been sufficient in Steere et al.'s JAMA study to exclude the diagnosis of Lyme disease, though of course the patient clearly did suffer from B.burgdorferi infection.
It is remarkable that the authors made no attempt to use any direct testing methods, such as culture, histological tissue examination, or antigen or nucleic acid detection systems, to verify the accuracy of their diagnoses. Although the latter techniques are relatively new, several tests of considerable utility have been reported (13-15). If one is carrying out a study that purports to be the final word on the correct diagnosis of a disease, every effort should be made to use the latest and most accurate diagnostic techniques before publishing the study. The diagnostic criteria and methods used by Steere et al. are both outdated and marred by an unacceptable level of subjectivity on the part of the researchers.
Response to Treatment:
Steere et al. divide their patient population into three categories: patients with active Lyme disease, patients with a history of Lyme disease and another current illness, and patients with another illness. Interestingly, of the patients thought to have active Lyme disease, at least 52 had already been antibiotically treated before evaluation by the authors. Thus, Steere et al. implicitly acknowledge that treatment failures are a common phenomenon in Lyme disease.
Yet under the study protocols, lack of responsiveness to antibiotic therapy is a primary criterion for the determination that active Lyme disease is not present: "We did not find age, sex, or duration of symptoms to be of help in diagnosing fibromyalgia, but the presence of tender points upon examination and lack of response to antibiotic therapy were important clues in diagnosing fibromyalgia." These criteria are clearly not consistent with the authors' own findings regarding the prevalence of treatment failures among patients whom they determined to harbor active disease; further, it is well known that every one of the primary symptoms associated with fibromyalgia or chronic fatigue syndrome (persistent headache, fatigue, myalgias, arthralgias, sleep disturbance, etc.) are common in active Lyme disease and cannot be used for differential diagnosis. Finally, in cases where the patient did improve with antibiotic therapy but relapsed afterwards, the authors conclude anecdotally that the positive response was probably due to the placebo effect.
We strongly take issue with the fact that the alternate interpretation for these treatment responses -- that borrelial infection persisted after antibiotic treatment -- is completely ignored. There are now culture-confirmed treatment failures in the medical literature for all stages of Lyme disease [9,12,16-19], sometimes even after long term, high-dosage antibiotic therapy [17-19]. Other studies employing the polymerase chain reaction have indicated the persistence of B.burgdorferi-specific DNA in the cerebrospinal fluid of many patients who remain symptomatic after antibiotic therapy . In light of these findings, the authors' exclusive interpretation that treatment failure was due to misdiagnosis seems rather reckless. Although persistent symptoms after treatment by no means implies the continued presence of B. burgdorferi, unresponsiveness to short term antibiotic therapy cannot be interpreted to exclude it.
In addition, the use of psychiatric symptoms to exclude the diagnosis of Lyme disease (". . . psychiatric disorders such as anxiety, depression or somatization clearly played a role in the illness of some of these patients . . . ") is entirely inappropriate. Controlled studies have indicated that a high percentage (66%) of seropositive Lyme disease patients report an episode of major depression during the course of their illness, most (90%) for the first time . A wide variety of minor and major psychiatric disorders have been reported in Lyme disease [21-23], similar to the findings in neurosyphilis . The authors may be unaware of these findings or may be ignoring them; in either case, the study's patient evaluation techniques are compromised. Also disturbing is the authors' classification of the neurological sequelae that developed in 15 of the 156 patients who were determined to have previous Lyme disease and another current illness. Although discussion of these patients' symptoms was vague, the syndromes listed in the accompanying table -- vertigo, peripheral neuropathies, radiculopathy and seizure disorder -- have all been described in patients with active Lyme disease [7,8,25]. Steere and his colleagues ruled out active disease in these patients because "neurological test results for Lyme disease were negative." Such a conclusion is insupportable, as there are many reports of culture-positive and PCR-positive patients with active disease and normal neurological findings [15,26,27]. Further, many clinical findings in Lyme disease, such as Bell palsy or Lyme meningitis, often resolve even without treatment , but this certainly cannot be equated with microbiological "cure." The presence of subjective symptoms in combination with the absence of clinical findings after treatment is frustrating for both physician and patient, but it is clearly inappropriate to simply declare the patient borrelia-free in these circumstances. Finally, even if the authors are correct and active disease was not present in any of these 15 patients, the incidence of serious neurological sequelae in the "post-Lyme" patient population implies that the "post-Lyme syndrome" can be quite severe. The shunting of these patients into the "non-Lyme" category gives the impression that Lyme disease could not have been responsible for any of their symptoms, an extremely questionable conclusion that suggests that Lyme disease and its sequelae are more benign than they really are. On a broader level, the authors are asking us to believe that one-fifth of the entire study population had a history of Lyme disease, were all cured, and then went on to develop a variety of other illnesses, virtually all of which have identical symptoms to active Lyme disease. This strains credulity.
Implications and Conclusions:
The methodology used by Steere et al. for categorizing diagnoses with adverse antibiotic treatment outcomes can thus be summarized as follows:
Rather than engaging in a serious attempt to explore the rate of treatment failure in their study by utilizing direct detection methods, the authors chose to rely entirely on anecdotal observations and then reported these as objective fact: These patients do not have Lyme disease. This retrospective determination was made despite ongoing unchanged symptomatology, despite extensive literature documenting culture-confirmed persistent infection following antibiotic treatment, and without investigation via direct methods for the presence of B. burgdorferi. That this is offered as "science" in a leading medical journal is appalling.
Interestingly, the authors never followed up on the patients who were determined to have active Lyme disease and who were therefore treated with "appropriate" antibiotic regimens. Did all of these patients recover completely? If so, it would imply that Lyme disease is always easily cured, an assumption that is belied by both the medical literature and the authors' own willingness to re-treat. If there were later relapses, then the authors' contention that treatment failures are due primarily to misdiagnosis is obviously false, unless they had been mistaken in their diagnosis of Lyme disease in these patients, in which case the diagnostic conclusions of the study are rendered useless. No matter how these questions are answered, the internal contradictions of the study are exposed.
It is no longer acceptable to continue to deny that treatment failures often occur in Lyme disease. The literature now provides substantial documentation on this point, and is beginning to elucidate the mechanisms responsible. Early dissemination of B. burgdorferi to the central nervous system [27,28] and intracellular localization and persistence [29,30] have been identified as contributors to treatment failure in Lyme disease. Instead of anecdotal and subjective "studies" like "The Overdiagnosis of Lyme Disease," we urgently need to know the true incidence of Lyme disease, the percentage of patients in whom the Lyme spirochete survives after therapy, and the efficacy of further antibiotic treatment in these patients. Some authors are already attempting to come to terms with these issues [31,32]. Unfortunately, Steere et al. clearly are not. Instead, the authors have employed a flawed methodology that "defines away" Lyme disease and does a disservice both to patients and to unwary physicians who as a result of this article may adopt an inappropriately skeptical approach to the diagnosis .