The prevalence of cardiac symptoms in patients with Lyme disease varies from 4% to 10% [8, 9] in the United States and from 0.3% to 4% in Europe [10]. Reversible atrioventricular block [9] is the most frequently reported finding. However, more severe manifestations, such as serious rhythm abnormalities [11-13], myocarditis [7, 14], pericarditis [15], and pancarditis [16], have been reported. Cardiac involvement responds well to antibiotics, and a permanent pacemaker is rarely required [8, 15]. Recent reports, however, suggest that Lyme disease might cause chronic cardiac involvement [17-19]. These reports are similar to other studies of the long-term outcomes of Lyme disease, which showed musculoskeletal and neurologic sequelae [6, 20].

We sought to ascertain whether persons with a previous episode of treated Lyme disease have a higher prevalence of cardiac abnormalities than persons without previous Lyme disease.

Methods

Top Methods Results Discussion Author & Article Info References

Participants

From January 1993 through January 1996, all adult permanent residents of Nantucket Island, Massachusetts, who were 17 years of age or older (n = 6046) were identified through the Island's census listing. Vacated addresses were updated by the Nantucket Post Office. Residents listed in the census file were those who owned homes on the Island; of these residents, 85% lived on the Island year round. Participants were surveyed according to the total design method [21]: They received the first written reminder 3 weeks after the survey was mailed; if they did not respond, the second reminder letter was mailed after an additional 4 weeks. The one-page survey was mailed to participants up to three times. The survey inquired about basic demographic characteristics; previous diagnosis of Lyme disease; treatment or positive result on a serologic test for Lyme disease; and symptoms commonly found in patients with acute Lyme disease, such as erythema migrans, arthritis, meningitis, and facial palsy.

Sampling Strategy for Clinical Examination

With a random sampling strategy designed to maximize case finding, we used the method of Cochran [22], a computerized protocol that selected 5 potential case-patients and 2 potential controls balanced by age tertile and sex in order to enroll 500 participants (350 potential case-patients and 150 potential controls). If someone refused participation in the study, the next randomly generated participant "replaced" the refusal according to the computerized algorithm. The goal was to examine approximately 350 persons (175 case-patients and 175 controls) to provide 80% power to detect differences of 10% between groups, assuming that the prevalence of a particular abnormality in controls would be 5%. Potential case-patients included persons who reported a previous diagnosis of Lyme disease or a history of a positive result on a serologic test for B. burgdorferi. The two potential controls sampled for study included persons who responded that they had no history of Lyme disease; these persons were sampled to balance the one survey respondent who reported symptoms suggestive of possible previous acute Lyme disease [a history of a swollen joint; Bell palsy; meningitis; large, red, ringlike expanding rash; and persistent cognitive difficulties]) and the one who had no symptoms suggestive of possible Lyme disease (Figure 1)

View larger version (18K): [in this window] [in a new window]   Figure 1. Sampling strategy and study design used to determine case status and measure clinical and electrocardiographic outcomes in persons with and without previous Lyme disease. *Adults 17 years of age and older. Thirteen persons were excluded because of difficulties with electrocardiography or because the persons declined to be evaluated. CDC – Centers for Disease Control and Prevention.

Assessment of Clinical Outcomes

A random sample of potential case-patients and controls was then invited for a clinical evaluation that included a blinded and standardized medical history, physical examination, Lyme serologic testing (enzyme-linked immunosorbent assay capture and Western blot), and 12-lead electrocardiography. Information on cardiac involvement was gathered by asking the participants questions about a history of heart disease; use of cardiac medications; implantation of a pacemaker; and history of chest discomfort, shortness of breath, or palpitations. Each participant underwent 12-lead electrocardiography. Each electrocardiogram was subsequently coded by one blinded reader according to the following features: heart rate; rhythm; cardiac arrhythmias; axis; PR, QRS, and QT intervals; QRS structure; ST segments and T waves; and other abnormalities. Sinus bradycardia was defined as a heart rate less than 50 beats/min; tachycardia was defined as a heart rate greater than 100 beats/min. Electrocardiograms were classified as normal or as having minor abnormalities or major abnormalities according to criteria previously applied to ambulatory persons with positive results on Lyme immunoserology [23]; these criteria were slightly modified by a panel of cardiologists from our hospital. An electrocardiogram was considered normal if it showed a normal sinus rhythm (50 to 100 beats/min) with no pathologic findings. An electrocardiogram was classified as showing minor abnormalities if it had one or two of the following: 1) sinus rhythm less than 50 or 100 to 120 beats/min, 2) ectopic or wandering atrial focus, 3) nonspecific ST-segment and T-wave changes, 4) first-degree atrioventricular block (PR interval > 200 ms), 5) abnormal QRS axis (not within –30 degrees to 90 degrees), 6) intraventricular conduction delay (defined as QRS width between 100 and 120 ms), 7) distinct sinus arrhythmias or premature atrial or ventricular contraction, 8) ventricular hypertrophy, or 9) abnormal voltage. An electrocardiogram was classified as showing a major abnormality if it met one of the following criteria: 1) three or more of the preceding minor abnormalities; 2) QRS width greater than 120 ms; 3) a prolonged rate-corrected QT interval (>440 ms); 4) myocardial infarction; or 5) cardiac arrhythmia consisting of atrial fibrillation or flutter, junctional rhythm, idioventricular rhythm, ventricular tachycardia or fibrillation, or paced rhythm. All electrocardiograms were read by a single cardiologist who was blinded to the participants' outcomes.

To assess intra- and interobserver reliability in the interpretation of the electrocardiograms, a sample of 11 electrocardiograms was distributed to three independent physician reviewers who were blinded to the participants' Lyme disease status. Intraclass correlation coefficients were calculated for all continuous variables [24]. For categorical variables (the presence or absence of a particular feature), we calculated the percentage agreement. For intraobserver reliability, each reviewer read 11 electrocardiograms twice in random order. Intraclass correlation coefficients and percentage agreement were calculated as described above.

Classification of Lyme Disease

Participants were classified according to the CDC case definition of Lyme disease [25]: erythema migrans (>5 cm) or laboratory confirmation of infection and at least one late manifestation. Late manifestations may involve the musculoskeletal system (recurrent attacks of joint swelling and chronic arthritis), nervous system (lymphocytic meningitis, cranial neuritis, facial palsy, radiculoneuropathy, and encephalomyelitis), or cardiovascular system (acute second- or third-degree atrioventricular conduction block and myocarditis). Laboratory confirmation of infection was determined by the presence of IgG, IgA, and IgM antibodies to B. burgdorferi by ELISA capture antibody and Western blot analysis, as described elsewhere [26, 27]. Ratios of ELISA capture (sample optical density/control density) of 1.0 or more were considered increased. Western blot reactivity to five or more B. burgdorferi-specific polypeptides indicated previous infection.

Statistical Analysis

The demographic variables and electrocardiographic findings of case-patients and controls were initially compared by using the Student t-test for continuous variables and chi-square tests for categorical variables. For 2 x 2 tables with expected cell counts less than five, we used a two-tailed Fisher exact test. Multivariate logistic regression models were fitted with the presence or absence of abnormal electrocardiographic findings as dependent variables [28]. Predictor variables were sex, age, CDC case-definition status, and history of heart disease. Additional models were fitted for case-patients who met CDC case-definition status, with abnormal electrocardiographic findings used as the outcome variable. The Pearson correlation coefficient on a random sample of electrocardiograms was calculated to determine the inter- and intrarater reliability for continuous electrocardiographic variables (rates, intervals, and axis) and percentage agreement for categorical variables. We used the SAS statistical package on a Unix mainframe computer to analyze the data [29]. A P value of 0.05 or less was considered statistically significant for all analyses.

Results

Top Methods Results Discussion Author & Article Info References

Among the 3703 respondents to the adult census (61.3%), 443 (12%) indicated a history of Lyme disease and 3260 (88%) did not (Figure 1). A total of 569 adult respondents were selected for clinical evaluation; 85 (15%) refused. One hundred thirty-five of the 569 respondents either could not be scheduled for the evaluation (for example, because of timing conflicts) or could not be reached by mail or telephone. Overall, 349 adults aged 17 years and older were randomly selected and were evaluated. Thirteen persons were excluded from this analysis because they did not have an electrocardiogram or had an electrocardiogram that was uninterpretable as a result of technical difficulties.

To determine whether respondent bias affected our survey, we analyzed census data from nonrespondents and found that nonrespondents were more likely to be young men scoring on the low end of an occupational prestige score (that is, they had blue-collar professions) than were respondents (National Opinion Research Center) [30]. We telephoned a sample of these nonrespondents to determine whether they were more likely to have had Lyme disease than the survey respondents. Survey responses on a history of Lyme disease in this group did not differ from those for the 3703 respondents who returned the survey by mail (data not shown) (Phillips CB, Wright EA, Fossel AH, Sangha O, Dalgas M, Liang MH, et al. A population survey of Lyme disease on Nantucket Island, Massachusetts. In progress),

To determine whether there was a selection bias among persons who agreed to be evaluated, in our clinical assessment we compared the participants' responses to the survey's questions on symptoms with the nonparticipants' responses. Persons with a history of swollen or painful joints (P = 0.009) and forgetfulness (P = 0.04), regardless of Lyme disease history, were more likely to agree to a clinical evaluation than not (Phillips CB, Wright EA, Fossel AH, Sangha O, Dalgas M, Liang MH, et al. A population survey of Lyme disease on Nantucket Island, Massachusetts. In progress).

The analysis is based on 336 adults for whom complete data on medical history and a 12-lead electrocardiogram were available. One hundred seventy-six participants met the CDC case definition for Lyme disease, and 160 did not. Case-patients and controls did not differ significantly according to age, sex, and highest grade completed in school (Table 1). There was a difference of borderline significance in years of residence on the island (24.8 years for case-patients and 24.4 years for controls; P = 0.04). A total of 9.4% of the case-patients and 11% of the controls reported a history of heart disease. Case-patients did not report a significantly higher frequency of palpitations or chest discomfort at the time of evaluation. The average time from Lyme disease to study evaluation for the case-patients was 5.2 years (median, 4 years).

The clinical presentation and treatment of case-patients are outlined in Table 2. Overall, 94% of this group received antibiotic therapy for Lyme disease. No case-patient reported a history of acute carditis or the implantation of a pacemaker during their illness.

We tested interobserver and intraobserver reliability on a random sample of 11 electrocardiograms by using three independent readers who were blinded to the participants' disease status. For all electrocardiographic features, interobserver reliability ranged from 0.93 (QRS axis) to 0.96 (heart rate) for continuous variables (intraclass correlation coefficient) and from 82% to 100% agreement for the presence or absence of a particular feature. For most findings, percentage agreement was 100%; only right anterior hemiblock (82%), nonspecific ST-segment abnormalities (82%), hypertrophy (91%), and ST-segment depression (91%) showed slightly lower agreement. Intraobserver reliability was also assessed by using intraclass correlation coefficients, which ranged from 0.95 (heart rate) to 0.97 (PR interval). There was 100% agreement for all dichotomous features (present compared with not present) with one exception: nonspecific ST-segment abnormalities (90% agreement).

The electrocardiographic findings for case-patients and controls are presented in Table 3. No significant differences in heart rate; PR, QRS, and QT intervals; frequency of sinus bradycardia; frequency of sinus tachycardia; or QRS axis outside the –90-degree to 30-degree range were seen between groups. Case-patients (9.9%) had a slightly higher incidence of first-degree atrioventricular block than controls (5.1%), but the difference was not statistically significant (P = 0.12). Other conduction disturbances, the frequency of QRS structure or ST-segment abnormalities, PR depression, low voltage, and a history of myocardial infarction did not differ between case-patients and controls.

When both groups were analyzed according to the presence of minor or major abnormalities, no significant differences were seen between cash-patients and controls. Minor abnormalities were seen in 30.1% (95% CI, 23.3% to 36.8%) of case-patients and 29.6% (CI, 22.5% to 36.7%) of controls (P > 0.2). Major abnormalities were seen in 11.4% (CI, 6.7% to 16.1%) of case-patients and 14.1% (CI, 8.7% to 19.5%) of controls (P > 0.2). Differences between case-patients and controls in minor electrocardiographic abnormalities (30.1% compared with 29.9%; P > 0.2) or major abnormalities (11.3% compared with 12.7%; P > 0.2) were not significant. When we compared case-patients and controls after excluding controls with a positive result on ELISA capture or Western blot (n = 23), there were also no differences in the presence of minor or major electrocardiographic abnormalities between groups.

We used multivariate logistic regression models that controlled for age, sex, and previous heart disease to examine whether minor or major electrocardiographic abnormalities could be explained by previous Lyme disease. In these analyses, previous Lyme disease did not significantly predict the presence of either minor abnormalities (odds ratio, 1.09; P > 0.2) or major abnormalities (odds ratio, 0.78; P > 0.2). Age was associated with the presence of major cardiac abnormalities (odds ratio, 1.05; P < 0.0001). Male sex was associated with the presence of minor cardiac abnormalities (odds ratio, 1.91; P = 0.01).

In further models, we explored whether the presence of minor or major electrocardiographic abnormalities in case-patients could be explained by age, sex, history of heart disease, presence of a positive result on ELISA capture or Western blot, or amount of time since the original episode of Lyme disease (Table 4). Only one of the variables was significant: A history of heart disease predicted the presence of major electrocardiographic abnormalities in patients with previous Lyme disease. When we added a categorical treatment variable to determine whether the untreated case-patients (n = 11) were more likely to have major or minor electrocardiographic abnormalities, this variable did not significantly affect our findings (data not shown).

We also evaluated how the presence of self-reported heart disease (which may have predated acute Lyme disease) affected our findings. We fitted regression models among case-patients, excluding persons who reported a history of heart disease (n = 16) (Table 5). Except for age among case-patients with major electrocardiographic abnormalities, no variable significantly predicted the presence of minor or major abnormalities.

Discussion

Top Methods Results Discussion Author & Article Info References

We found no differences in the prevalence of cardiac abnormalities between persons with and those without previously treated Lyme disease. In particular, we could not find significant differences in the prevalence of current heart symptoms and electrocardiographic findings, including heart rate; PR, QRS, and QT intervals; axis; presence of arrhythmias; heart block; ST-segment and T-wave abnormalities; or other nonspecific electrocardiographic changes. The history of Lyme disease was also not associated with either major or minor electrocardiographic findings. Among case-patients who had not had heart disease, only age (not sex, time since infection, or current seropositivity) predicted major cardiac abnormalities.

Previous reports of carditis during acute Lyme disease have shown a relatively high percentage of cardiac symptoms. In a review of 84 patients with acute Lyme carditis assembled by the Centers for Disease Control between 1983 and 1986, Ciesielski and colleagues [31] reported that 69% of persons had palpitations, 19% had conduction abnormalities, 10% had myocarditis, 5% had left ventricular failure, and 21% required hospitalization. In addition, McAlister and colleagues [8] described the presence of an atrioventricular block in 45 of 52 patients (87%) with Lyme disease. In another study, 38% of patients with Lyme disease required a temporary pacemaker [32]. Our results do not support these higher frequencies, a fact that might be explained by a cohort effect or treatment differences among the study participants.

Some evidence suggests that carditis in Lyme disease may not be transient but rather may lead to chronic abnormalities. Several case series and case reports suggest that myocarditis and cardiomyopathy might occur years after the primary infection [17, 33, 34]. Stanek and colleagues [35] described a B. burgdorferi-seropositive person with chronic congestive heart failure whose condition significantly improved after ceftriaxone treatment. Borrelia burgdorferi was grown from the myocardial biopsy specimens of this patient. The benefits of ceftriaxone treatment have also been reported by Peeters and colleagues [18], who demonstrated the reversibility of cardiomyopathy in 8 of 9 patients with dilated cardiomyopathy. In one study of 175 heart transplantation candidates, 14 patients (8%) were seropositive but only 5 patients could recall an episode consistent with previous Lyme disease [36]. Findings from such case series may not be generalizable to a population-based analysis because the series describe more severe cases.

According to histologic findings, isolation of B. burgdorferi from endomyocardial biopsy specimens suggests that Lyme carditis is due to direct invasion of the organism [16, 17, 35]. Evidence of cardiac inflammation can also be shown by gallium scanning [37-39]. Studies of His bundles have located the site of heart block from Lyme disease primarily in the atrioventricular node, although sites below the His bundle have also been reported [8, 40]. Bundle-branch blocks have been described at different locations of the His-Purkinje system [8, 10, 12, 40], suggesting that B. burgdorferi might cause damage at multiple foci. It is of interest that we found a slightly longer PR interval and a higher prevalence of first-degree atrioventricular block (although these findings were not significant) among patients with previous Lyme disease. These data suggest that atrioventricular nodal disease may have subclinical sequelae in patients with Lyme disease.

Some limitations of our study require comment. First, we conducted electrocardiography at one time point; by doing so, we may have missed transient or intermittent cardiac abnormalities, particularly arrhythmias. However, the case-patients had no higher prevalence of clinical symptoms (for example, syncope) that might go along with arrhythmias. Second, electrocardiography is insensitive to subtle cardiac involvement. Although gallium scans [37-39] and anti-myosin-indium scans [41] have shown cardiac abnormalities in persons with Lyme disease, their predictive value in population studies has not been evaluated. Third, the potential for bias in symptom reporting may have occurred; however, recall bias does not affect electrocardiograms.

Cardiac abnormalities are part of the CDC case definition of Lyme disease (clinical carditis with high-degree heart block) and were also the outcome of interest in our study. We sought to determine whether acute carditis from Lyme disease resolves completely after the illness is treated or whether sequelae develop. With its population-based design, our study provides useful information about the prevalence of a variety of cardiac abnormalities, including carditis that may have lingered in patients with Lyme infection compared with controls. In addition, ours was a retrospective study of patients who had had Lyme disease several years before evaluation; thus, recollection of Lyme disease symptoms may have been enhanced by recall bias. We used strict criteria (CDC criteria) to determine case status and used Western blot analysis in a single laboratory to assist with case determination. Nevertheless, we could not have the clinical certainty about case status that would be found in a prospective analysis that accrued patients with Lyme disease at the time of infection.

Furthermore, no standardized system is available to categorize electrocardiograms with "minor" or "major" abnormalities, although our categorization system was modified from criteria used by Vidaillet and colleagues [23] on ambulatory persons. It is unlikely that different categorization systems would reveal significant differences because none of the individual electrocardiographic features revealed substantially different distributions among case-patients and controls.

Several previous reports suggest that Lyme disease may be associated with long-term neurologic or musculoskeletal sequelae, particularly in patients who were untreated or experienced long delays before initiation of antibiotic therapy [42-45]. Because most of the study patients had been treated for their Lyme disease, we could not determine the long-term cardiac outcomes of untreated Lyme disease. In addition, the average duration between the episode of Lyme disease and our follow-up was 5.2 years; this period may have been too short to demonstrate some chronic abnormalities.

Although our study had adequate power to detect differences of greater than 10% between case-patients and controls, our sample size was too small to allow us to detect less common electrocardiographic abnormalities. For example, a sample more than twice as large would be needed to detect significant differences similar to those found for first-degree atrioventricular block or bradycardia.

In conclusion, persons with previous Lyme disease in our population-based study did not have a higher prevalence of electrocardiographic abnormalities and cardiac symptoms than did persons without previous Lyme disease. Our data suggest that the overall cardiac outcomes in treated Lyme disease are favorable.

Dr. Fleischmann: Division of Cardiovascular Medicine, Department of Medicine, Brigham and Women's Hospital, 75 Francis Street, Boston, MA 02115.

Drs. Sangha and Lew, Ms. Fossel, and Ms. Phillips: R.B. Brigham Multipurpose Arthritis Center, 221 Longwood Avenue, Boston, MA 02115.

Dr. Wang: Department of Medicine, Massachusetts General Hospital, Fruit Street, Boston, MA 02114.