The Long-Term Clinical Outcomes of Lyme Disease: A Population-Based Retrospective Cohort Study

Methods

Participants

In February l991, we did a follow-up analysis of residents of Argylla Road in Ipswich, Massachusetts, an endemic coastal area for Lyme disease. The incidence and clinical course of Lyme disease among residents of this area have been reported previously [13]. Participants were recruited by calling consecutive households located in the Argylla Road area, the epicenter of infection, to ask if they would be interested in enrolling in a study about Lyme disease in their area. Potential participants were told that the study involved a history, physical examination, and serologic analysis for Lyme disease. Information about whether a person ever had a previous diagnosis of Lyme disease was obtained and used to assign tentative status (with or without Lyme disease) for study participants.

We recruited participants until we had 50 tentative persons with Lyme disease and 50 tentative controls. Once the potential Lyme disease group was filled, calls were made consecutively to fill the potential control group. Residents 18 years of age or older were invited to participate in the study. This protocol was approved by the Brigham & Women's Hospital Committee for the Protection of Human Subjects.

Confirmation of Lyme Disease

For inclusion in the Lyme group, persons needed a previous diagnosis of Lyme disease by a physician and needed to fulfill the Centers for Disease Control and Prevention (CDC) criteria for Lyme disease (a history of physician-documented erythema migrans or a late manifestation of Lyme disease confirmed by a positive Lyme serologic test result, or both [14]). This information was obtained through patient interview and then medical record review to determine if patients fulfilled criteria for Lyme disease.

Previous study records, local physician reports, and previous serologic test results were available for confirmation of Lyme disease. Persons without a previous clinical history of Lyme disease were classified as controls. The status of the participants (with or without Lyme disease) was determined independent of the clinical assessment, using a protocol that did not include any outcome data.

Assessment of Clinical Outcomes

A blinded investigator determined outcomes in a standardized manner independent of Lyme disease status. All patients completed a standardized questionnaire, had electrocardiography, and had a neuropsychological battery of tests. The questionnaire included data on demographics, comorbidity, education, review of systems, medications, memory and cognitive function, and the Short Form-36 health status measure (a reliable, previously validated measure of physical, psychological, social, and role functions [15]).

A physical examination was done by one observer blinded to Lyme disease status. It included a joint examination (the American College of Rheumatology Glossary examination) that measured swelling and pain through passive range of motion [16] and a neurologic evaluation of strength and deep tendon reflexes, light touch, and vibration sensation with a 128-Hz tuning fork (at the elbow, wrist, fibula, and ankles). Pain and swelling indices from the joint examination (the American College of Rheumatology Glossary examination) were summed and recorded as a global score. A vibration test result of a distal gradient was considered present if the participant reported diminished vibratory sensation at a distal compared with proximal site. Each participant had an electrocardiographic study that was interpreted blindly by a cardiologist uninvolved with the clinical assessment. All outcomes were determined by one investigator who had no knowledge of whether participants were in the Lyme or control groups.

The neuropsychological battery of tests measured immediate and delayed verbal memory, attention, conceptualization, fine motor dexterity, and perceptual discrimination. Tests included the California Verbal Learning Test [17], Wechsler Memory Scale (visual reproduction and verbal paired associates subtests [18]), Shipley abstraction subtest [19], Stroop test [20], Trailmaking test [21], and Purdue Pegboard Test [22].

The California Verbal Learning Test measures verbal memory. Participants are asked to learn a list of 16 words during five trials; recall on the fifth trial is recorded (trial 5). This is then followed by a distracter list. The original list is recalled after the distracter list is learned (short recall) and then recalled again after a 20-minute delay (long recall). This is a challenging test of memory for patients with superior premorbid experience. Normative values are available for young and elderly adults [17]; the range of normal is between 11 and 15 words for trial 5 and is between 10 and 15 words for the long-recall subtest for persons between 45 and 54 years of age. A clinically significant change in the California Verbal Learning Test would be recalling 4 more words or 4 fewer words.

All tests were administered according to published procedures. Test scores were transformed into standard scores calculated from published, age-corrected normative data. Participants with a score of 2 or more SDs from age-adjusted means were considered impaired. All results were reviewed by a neuropsychologist (RK) who was not involved with the participant's evaluation, to determine those patients who were in need of further clinical evaluation.

Participants with swelling or pain (joint examination test result), evidence of a distal gradient (vibration test result) or persistent symptoms of paresthesias in an extremity, or impairment on two or more neurocognitive tests were sent for further clinical evaluation. Nine patients were evaluated at the Lyme disease clinic at Tufts-New England Medical Center and 4 were evaluated by other neurologist or rheumatologist consultants to determine if these abnormal screening test results were accompanied by objective findings. This evaluation included lumbar puncture, electrophysiologic studies, magnetic resonance imaging, detailed neuropsychological tests [8], joint radiographs, or arthrocentesis.

Serologic Evaluation

All patients had serologic testing after the history and examination. Serum samples were stored at −70°C and were tested for IgG antibodies to B. burgdorferi by indirect enzyme-linked immunosorbent assay (ELISA [23]); for IgM, IgG, and IgA antibodies to the spirochete by antibody-capture enzyme immunoassays [24]; and for the pattern of IgG reactivity to spirochetal polypeptides by Western blotting. In general, Western blot reactivity varied with the degree and duration of dissemination of Lyme disease. For example, patients with early localized infection or erythema migrans might react to only 2 to 8 B. burgdorferi polypeptides, those with meningitis might react to at least 8 to 14 polypeptides, and those with arthritis or late central nervous system disease might react to as many as 18 to 25 polypeptides (Berardi VP. Personal communication). The isolate used for antigen preparations was the B. burgdorferi G39/40 strain obtained through low passage [24]. Indirect ELISA titers greater than 400 and ELISA capture ratios (sample optical density/control optical density) of 1.0 or more were considered as increased test results. Western blot reactivity to five or more B. burgdorferi-specific polypeptides indicated previous infection [25].

Silver Stain Method

The Dieterle silver impregnation stain used was a modification made by one of us (PHD) in 1985 [26]. This standard approach has yielded a constant clean yellow background of cerebral cortex sections with no silver impregnation of anatomic neural processes and dendrites. Spirochetes are easily seen as black to blue-black cells against the yellow tissue. Specificity for nonstaining of normal tissue fibers (procollagen, elastin, basement membrane material, and neural dendrites and filaments) and documentation of the cytologic structure of Borrelia spirochete strains were further tested in a large extended study [27].

Controls routinely used in each stain assay consisted of NP40 strain that was injected into human normal breast tissue removed for cosmetic surgery and was paraffin-embedded in the usual manner (negative control), and rat gonad tissue infected with the Reiter strain of treponemal spirochetes (positive control). We tested an extensive panel of polyclonal and monoclonal antibodies with multiple Borrelia strains. Monoclonal antibodies were inefficient in delineating the spirochetes in formalin-fixed sections when compared with the less specific but more sensitive polyclonal antibodies [27].

Statistical Analysis

Comparisons of the prevalence of symptoms and signs between patients in the Lyme group and controls were made using chi-square tests of association and Fisher exact tests [28]. To test which symptoms were most predictive of Lyme disease status, a stepwise logistic regression analysis of variables with a 0.05 significance level was used [29]. Differences in continuous variables (the Short Form-36 score, the joint index, and the neurocognitive test scores) between patients with Lyme disease and controls were tested using linear regression analysis adjusting for age and sex [26].

Risk factors for Lyme disease sequelae (tertiles of duration from disease onset to treatment and the type of antibiotic therapy) and their serologic correlates were studied using chi-square analyses and the Student t-test, respectively. General linear modeling techniques were used to examine the relation between performance on the California Verbal Learning Test and the number of Western blot reactive bands adjusting for the potential confounders, age and sex. Data were analyzed with the SAS statistical package on an IBM PS2 computer [30]. All P values were two-tailed.

Results

Of 124 participants who were contacted, 106 (85%) agreed to participate. Previous sample size calculations and study resource considerations allowed for a minimum of 80 participants to be studied. Data on the first 81 participants are presented below. Thirty-eight participants met CDC criteria for previous Lyme disease (Lyme group) and 43 did not (controls). The groups did not differ in mean age, sex, educational attainment, race, or marital status. Most patients (90%) had 2 or more years of college; 50% had graduate degrees (Table 1). All but 2 persons who were categorized as tentatively having Lyme disease during the initial screening evaluation met CDC criteria for Lyme disease. No tentative controls met CDC criteria.

Eighteen patients declined participation; 4 of them reported a previous diagnosis of Lyme disease. The mean age and sex distribution among patients with Lyme disease and control participants compared with those who refused did not differ (49.0 years for participants with Lyme disease compared with 57.3 years for persons refusing to participate [P = 0.40]; 49.1 years for control participants compared with 53.8 years for persons refusing [P = 0.43]).

The mean duration from disease to study evaluation was 6.2 years (range, 1 to 11 years). The initial presentation in 38 patients with Lyme disease is summarized in Table 2: fifty-five percent had a history of erythema migrans, 24% had a history of neuritis (meningitis, seventh nerve palsy, radiculoneuritis), 13% had carditis in the form of high-degree atrioventricular block. Eleven persons who had erythema migrans subsequently developed a late manifestation of Lyme disease.

Initial antibiotic treatment consisted of the recommended regimens at that time. Twenty-eight patients with Lyme disease received oral tetracycline, penicillin, or erythromycin for 10 to 21 days. Five persons received intravenous penicillin for 10 days and two received intravenous ceftriaxone for 10 to 14 days as initial therapy. Only two patients with Lyme disease had received amoxicillin, and one had received doxycycline as initial oral treatment for Lyme disease. Ten of the 38 patients with Lyme disease reported relapses within 1 year of treatment (fatigue, persistent arthritis or arthralgias, headaches, or difficulty with memory and concentration) and had had repeated antibiotic treatment (5 patients with intravenous ceftriaxone). Eight patients reported reinfection (defined as a solitary erythema migrans lesion at a different site at least 1 year later) and were retreated; 1 participant had been reinfected 4 times, each time with the characteristic erythema migrans rash.

Table 3 outlines the prevalence of persistent symptoms at the time of evaluation in patients with Lyme disease and controls. Patients with Lyme disease had on average more symptoms than did controls (number of symptoms of Lyme disease, 2.2 ± 1.8 [mean ± SD] compared with control symptoms, 0.7 ± 1.1; P =0.0001). Compared with controls, patients with Lyme disease had more symptoms of diffuse arthralgias; numbness, tingling, or burning pain in an extremity; coordination difficulties; fatigue, concentration, and sleep difficulties; and emotional irritability. The percentage of patients reporting depressive symptoms was similar between groups. Further, the Lyme group had lower global health status scores compared with controls.

All symptoms (Table 3) were entered into a stepwise logistic regression model to determine the variables that were most predictive of previous Lyme disease. Only two symptoms met the 0.05 significance level for entry into the final model. These included arthralgias (odds ratio, 8.6; 95% CI, 2.9 to 25.1; P = 0.001) and numbness, tingling, or burning pain in an extremity (odds ratio, 10.1; CI, 1.0 to 98.8; P = 0.05). Inclusion of a variable incorporating arthralgias and sleep difficulties did not add significantly to the previous model (P = 0.09). The degree of fatigue was highly correlated with sleep difficulties (r = 0.54).

Results from the physical examination and the electrocardiogram are outlined in Table 4. No acrodermatitis chronica atrophicans was seen. More abnormal joint test results and musculoskeletal pain were detected in patients with Lyme disease than in controls. The most common site of musculoskeletal pain in patients with Lyme disease was large joints previously afflicted with Lyme arthritis. In patients with Lyme disease, pain on motion was most frequently found in the shoulder (5 patients), hip (5 patients), ankle (4 patients), knee (2 patients), and proximal interphalangeal joints (2 patients). One patient with Lyme disease had a shoulder effusion and another had knee and ankle swelling. The age- and sex-adjusted joint index was higher in the Lyme group compared with controls. Eleven patients with Lyme disease compared with 6 controls had diminished distal vibratory sense. No differences in muscle strength or reflexes were detected between groups. On the electrocardiogram, no differences were noted in the number of abnormal patterns between groups but one patient with Lyme disease had complete heart block (discussed below).

Table 5 shows age- and sex-adjusted means for the scores on the neurocognitive tests. Participants with previous Lyme disease had more neurocognitive deficits than did participants without previous Lyme disease; 12 patients with Lyme disease scored more than 2 standard deviations from the mean on two or more memory tests compared with only 5 controls (P = 0.01). Controls, on average, did better than patients with Lyme disease on all tests, although these mean differences were subtle because group scores were within the normal range. Patients with previous Lyme disease did worse on the 20-minute delayed recall task of the California Verbal Learning Test and had longer completion times on the dominant-hand subtest of the Purdue Pegboard test. Similar lower scores were noted among patients with Lyme disease on trial 5 and the short-recall task of the California Verbal Learning Test as well as the nondominant-hand subtest of the Purdue Pegboard test and the Visual Reproduction recall test when compared with controls, but these results were not statistically significant.

Serologic Findings

Although 25 of 38 patients with Lyme disease were initially seropositive by indirect ELISA immunoassay (13 had had previous physician-documented erythema migrans and had no initial Lyme serum testing), only 18 still had increased IgG responses to B. burgdorferi by indirect ELISA. At the time of follow-up clinical assessment, four additional patients with Lyme disease had antibody responses below the cutoff by ELISA but had reactivity to more than five B. burgdorferi polypeptides on Western blot. Four patients with Lyme disease had increased antibody responses by capture enzyme immunoassay, 3 had increased IgG antibody, and 1 had IgG antibody and increased IgA. Patients with Lyme disease who had a positive follow-up serologic test result were more likely to have had symptoms for more than 30 days before antibiotic treatment (P = 0.003) and to have had a history of disease of longer duration (P = 0.002). Only two controls had borderline ELISA titers and negative Western blot results. Three other controls had reactivity to five B. burgdorferi polypeptides; they had negative ELISA titers and no history of Lyme disease.

Lyme Disease Sequelae

Thirteen of 38 patients with previous Lyme disease (34%; CI, 19% to 49%) had evidence for musculoskeletal, neuropathic, or neurocognitive impairment likely caused by Lyme disease because all had earlier classic manifestations of Lyme disease, a positive serologic test result, and no other disease to explain their symptoms. Their results are outlined in Table 6.

Patients 1, 2, and 3 reported use-related arthralgias, mostly in joints that had been previously inflamed during acute Lyme disease. Patient 1 also had fatigue and concentration difficulties. However, on physical examination they had no detectable abnormal joints and were successfully treated with nonsteroidal anti-inflammatory agents. Patients 4 and 5 had fatigue, occasional concentration difficulties, and recurrent joint pain. They both had intermittent knee effusions that on arthrocenteses yielded cell counts less than 1000. Radiographic studies showed compartmental joint-space narrowing suggestive of degenerative joint disease. Patient 4, in addition, had had second degree atrioventricular block with acute Lyme disease that resolved with penicillin treatment. Her irregular rhythm recurred 2 years later, resolved temporarily with ceftriaxone retreatment, but progressed to complete heart block requiring a pacemaker. Patient 6 had a shoulder effusion that resolved after treatment with doxycycline.

Patient 7 had symptoms of distal paresthesias and clumsiness in the lower extremities with spasticity and impaired vibratory sensation on examination. He had normal cervical and head magnetic resonance imaging scans but had spinal fluid pleocytosis, increased levels of protein, and selective concentration of antibody in the spinal fluid and was diagnosed with spastic diplegia. His syndrome was most likely caused by myelopathy, and he improved after 6 weeks of intravenous ceftriaxone. Patients 8 and 9 both had paresthesias, sensory loss to pinprick, and vibration in the lower extremities with electrophysiologic evidence of axonal polyneuropathy from Lyme disease. Both patients improved objectively on examination after antibiotic therapy, one with doxycycline and the other with intravenous ceftriaxone.

Four patients reported persistent memory or concentration difficulties. Each patient had verbal memory deficits (neurocognitive test result). Each of these patients had had prominent headache or meningismus during acute Lyme disease; and two had lymphocytic meningitis. Patients 10 and 11 had previously diagnosed neuroborreliosis with meningeal symptoms and spinal fluid pleocytosis that was treated with antibiotics. Although their symptoms improved after therapy, they still had residual encephalopathy with verbal memory deficits.

Patient 12 had had high fever, meningeal symptoms, and subsequent arthritis in 1982. She was noted to have a positive serologic test result for Lyme disease 4 years later and was treated with 2 weeks of parenteral penicillin. She later developed a progressive speech disorder, bradykinesia, and abnormal ocular motor function. Magnetic resonance imaging of the brain showed scattered white matter lesions in the hemispheres and pons, and she was diagnosed with supranuclear palsy. Lumbar puncture showed no selective concentration of antibody in the spinal fluid. Nevertheless, she was re-treated with 2 weeks of parenteral ceftriaxone in 1989 that had no effect on her neurologic symptoms. During the time of observation, this patient died. At autopsy, lymphoid mononuclear cells were observed surrounding the intracerebral vessels in one section. Using Dieterle silver stain, a spirochete was present in the cortex and another was exterior to a leptomeningeal vessel.

Patient 13 had recently diagnosed late neuroborreliosis with intrathecal IgA production and had been treated with 6 weeks of intravenous ceftriaxone 1 year before our study. His neurocognitive test results showed improvement compared with his pretreatment scores, but he still had residual verbal memory deficits.

Risk Factors for Lyme Disease Sequelae

The thirteen patients who had residual symptoms from Lyme disease had a longer duration of disease before treatment (35.0 ± 29.5 months compared with 2.7 ± 6.1 months; P < 0.0001) than those without symptoms; higher specific IgG titers (3410 ± 6427 compared with 421 ± 497; P = 0.03); and more bands on Western blot, a surrogate for duration of infection (14.8 bands compared with 4.3 bands; P = 0.0007). Duration of disease was also correlated with the degree of neurocognitive impairment. In age- and sex-adjusted regression analyses, more Western blot bands were associated with poorer performance on the long-delay recall task of the California Verbal Learning Test (r = −0.05; CI, −0.089 to 0.011; P = 0.04).

Although all patients with previous Lyme disease had been previously treated with antibiotics, only three had received amoxicillin or doxycycline as initial oral therapy (Table 2). Suboptimal therapy (oral tetracycline, penicillin, or erythromycin [31]) was not more frequent in those with Lyme disease sequelae compared with those without sequelae (P = 0.38).

Discussion

This is the first population-based, controlled study of the long-term consequences of Lyme disease. Previous case series have shown that the long-term sequelae of Lyme disease in patients attending neurology or Lyme disease clinics include Lyme neuropathy or encephalopathy [8, 32], fibromyalgia or the chronic fatigue syndrome [9, 12, 33], and subtle joint pain syndromes [10]. The former two diagnoses typically respond to antibiotic treatment; the others do not. To ascertain the true prevalence of these sequelae, we systematically studied persons with a history of Lyme disease from a geographically defined endemic area and compared them with persons residing in the same area who had no evidence of previous Lyme disease. Our data show a higher prevalence of recurrent arthralgias, memory loss, and other neurologic symptoms in persons with a history of Lyme disease compared with persons of similar age and sex, and our data confirm previous reports [8, 10].

Previous studies of patients with late Lyme encephalopathy have shown deficits in verbal memory, learning, and reaction time [8, 34-36]. Spirochete-like structures have been seen in brain biopsy specimens from two patients with apparent Lyme encephalitis, but the pathologic basis of Lyme encephalopathy is currently unknown [37, 38]. Patients with Lyme disease scored as well as controls did on four of six tests of neurocognitive performance, but differences between patients with Lyme disease and controls were apparent on the California Verbal Learning Test (a test of verbal memory) and the Purdue Pegboard Test (a test of visuomotor ability and dexterity) that were consistent with previous data [34, 36, 39] from patients with Lyme encephalopathy presenting to neurology or Lyme disease clinics. In our study, these deficits were found in unselected persons; the most prominent deficits were found in patients with previous neuroborreliosis.

Some of our patients with Lyme disease reported having diffuse muscle and joint pain, fatigue, and concentration and sleep difficulties, which are reminiscent of fibromyalgia. Two patients with Lyme disease who had received further clinical evaluations for joint symptoms were diagnosed with fibromyalgia-like illness because of diffuse arthralgias, fatigue, and concentration difficulties. Borrelia burgdorferi infection has been reported [9, 12] to trigger fibromyalgia. This syndrome differs from Lyme arthritis, which typically causes joint pain or swelling in one or a few joints and responds to antibiotics [6]. Antibiotic therapy for post-Lyme disease fibromyalgia is usually not effective [12]. Although some of our patients with diffuse arthralgias also said they had memory difficulties, neurocognitive test results did not indicate verbal memory impairment. A recent study [34] showed a greater degree of memory impairment in patients with Lyme encephalopathy than in those with fibromyalgia.

Patients with Lyme disease had more abnormal musculoskeletal findings on physical examination than did participants without Lyme disease. They reported more recurrent joint pain than did controls, had poorer health status scores, and more impairment on the Purdue Pegboard test, all measures of the effect of late joint morbidity. In individual patients, persistent symptoms were centered in large joints previously afflicted with Lyme arthritis, although some of the patients had a history of wrist or proximal interphalangeal involvement. One participant had recurrent shoulder arthritis that resolved with doxycycline treatment. Osteoarthritis would rarely produce wrist, shoulder, and proximal interphalangeal arthritis, which contributed to the Purdue Pegboard scores. Osteoarthritis, therefore, would not likely be the cause of impairment on this test. Degenerative arthritis was noted in knee and ankle joints of some patients; these joints were also common sites for Lyme arthritis. These findings are consistent with previous studies [6, 10] on the natural history of Lyme arthritis showing that oligoarticular Lyme arthritis tends to improve over time but that late manifestations, such as subtle joint pain, may persist.

Not all symptoms described by patients with previous Lyme disease are attributable to previous B. burgdorferi infection, even though the Lyme group had significantly more symptoms than did controls. In multivariate analyses, the presence of arthralgias was by far the best predictor of previous Lyme disease. The combination of arthralgias and sleep difficulties, suggesting fibromyalgia-like illness, was also noted to be predictive of previous Lyme disease, but this was redundant with arthralgias alone and did not add any predictive power to the model. Clinicians who evaluate patients with Lyme disease who have persisting symptoms may wish to ask patients about joint symptoms that might suggest Lyme-related sequelae.

Most patients with Lyme disease in our study were without symptoms at the time of clinical assessment. Thirteen persons (34%; CI, 19% to 49%), however, had some persistent symptom(s) (either neuropathy, memory problems and/or subtle joint pain or arthritis) despite having previously received antibiotics; only five who were retreated improved objectively. Whether lack of response to antibiotic treatment in the remaining patients indicates fixed residual injury or other mechanisms was not determined but such a response is similar to the limited response to penicillin seen in some patients with advanced neurosyphilis [40]. A longer duration of disease before treatment was a clinically significant correlate for persistent symptoms, suggesting that early eradication of B. burgdorferi before its sequestration into the central nervous system or joint compartments is important in preventing sequelae.

As with other retrospective studies, some misclassification may have occurred between groups. The rate of misclassification was minimized by access to previous study records, local physician reports, and previous serologic test results. Patients with Lyme disease and controls had the same serologic analyses. Although the potential for misclassification exists, its effect is likely to be small because status (with or without Lyme disease) was determined independent of study outcomes.

Despite our high rate of participation (85%), some residents declined participation in our study; thus, selection mechanisms may have affected outcomes. The consecutive nature of recruitment, the low rate of study refusal, and the similar demographic profiles of participants and nonparticipants mitigate against any selection bias. Further, some controls sought participation in our study because of other illnesses. This would bias our results against finding differences between groups.

We recognize that some of our findings were of borderline statistical significance and that we did multiple comparisons increasing the likelihood of positive findings caused by chance. Although we had previous strong hypotheses that a higher prevalence of joint, neurocognitive, and neuropathic sequelae would occur among persons with previous Lyme disease, this was an exploratory analysis. Findings associated with marginal P values (P > 0.01) should be regarded as tentative, needing confirmation in another study.

Our patients contracted Lyme disease in the mid-1980s, when clinical syndromes of Lyme disease and optimal antibiotic therapies were still evolving. Although all patients with Lyme disease were treated, few received the more recently recommended antibiotic therapies [31] and some had months to years of disease before treatment. Thus, some of our findings may not be generalizable to present day patients with Lyme disease.

Our data show that persons with previous Lyme disease have more musculoskeletal impairment compared with those without previous Lyme disease who are of similar age and sex. Persons with a history of Lyme disease also have more verbal memory deficits than those without a history of Lyme disease. These sequelae correlate with longer duration of infection and suggest that prompt treatment with antibiotics may best prevent long-term morbidity.