Azithromycin Compared with Amoxicillin in the Treatment of Erythema Migrans: A Double-Blind, Randomized, Controlled Trial
Benjamin J. Luft, MD; Raymond J. Dattwyler, MD; Russell C. Johnson, PhD; Steven W. Luger, MD; Elizabeth M. Bosler, MPH; Daniel W. Rahn, MD; Edwin J. Masters, MD; Edgar Grunwaldt, MD; and Shrikant D. Gadgil, MD
1 May 1996 | Volume 124 Issue 9 | Pages 785-791
Objective: To determine whether azithromycin or amoxicillin is more efficacious for the treatment of erythema migrans skin lesions, which are characteristic of Lyme disease.
Design: Randomized, double-blind, double-dummy, multicenter study. Acute manifestations and sequelae were assessed using a standardized format. Baseline clinical characteristics and response were correlated with serologic results. Patients were followed for 180 days.
Setting: 12 outpatient centers in eight states.
Patients: 246 adult patients with erythema migrans lesions at least 5 cm in diameter were enrolled and were stratified by the presence of flu-like symptoms (such as fever, chills, headache, malaise, fatigue, arthralgias, and myalgias) before randomization.
Intervention: Oral treatment with either amoxicillin, 500 mg three times daily for 20 days, or azithromycin, 500 mg once daily for 7 days. Patients who received azithromycin also received a dummy placebo so that the dosing schedules were identical.
Results: Of 217 evaluable patients, those treated with amoxicillin were significantly more likely than those treated with azithromycin to achieve complete resolution of disease at day 20, the end of therapy (88% compared with 76%; P = 0.024). More azithromycin recipients (16%) than amoxicillin recipients (4%) had relapse (P = 0.005). A partial response at day 20 was highly predictive of relapse (27% of partial responders had relapse compared with 6% of complete responders; P < 0.001). For patients treated with azithromycin, development of an antibody response increased the possibility of achieving a complete response (81% of seropositive patients achieved a complete response compared with 60% of seronegative patients; P = 0.043). Patients with multiple erythema migrans lesions were more likely than patients with single erythema migrans lesions (P < 0.001) to have a positive antibody titer at baseline (63% compared with 17% for IgM; 39% compared with 16% for IgG). Fifty-seven percent of patients who had relapse were seronegative at the time of relapse.
Conclusions: A 20-day course of amoxicillin was found to be an effective therapeutic regimen for erythema migrans. Most patients were seronegative for Borrelia burgdorferi at the time of presentation with erythema migrans (65%) and at the time of relapse (57%).
Lyme borreliosis is the most common vectorborne disease in the United States and Europe . Infection begins as a local process after Borrelia burgdorferi is inoculated into the skin by a feeding tick. In most persons, the initial sign of infection is the development of erythema migrans, which is characterized by an annular erythematous skin lesion. Amoxicillin and doxycycline have been advocated as the treatments of choice, primarily on the basis of small, often unblinded, randomized trials and retrospective analyses [2-4]. Azithromycin, an azalide (a new subclass of macrolide antibiotics), has been shown to have excellent in vitro and in vivo activity against B. burgdorferi in the laboratory [5, 6]. In a small, randomized, open study on the treatment of erythema migrans, a 5-day course of azithromycin was reported to be as effective as a 10-to 20-day course of doxycycline or amoxicillin with probenecid . To more thoroughly assess the efficacy of azithromycin and further define the treatment of erythema migrans, we conducted this large, multicenter, double-blind, randomized trial. Amoxicillin rather than doxycycline was chosen as the comparative agent to circumvent the problems associated with sun-related hypersensitivity reactions. During the study, we clarified important clinical questions about the natural history of this disease, its manifestations, and the usefulness of enzyme-linked immunosorbent assay (ELISA) for serologic testing.
Adult patients who had had erythema migrans diagnosed by a physician were recruited between June 1990 and October 1991 from 12 centers in eight states: New York (83 patients), Connecticut (85 patients), Missouri (35 patients), Wisconsin (23 patients), New Jersey (10 patients), Minnesota (6 patients), California (2 patients), and Rhode Island (2 patients). To be eligible for the study, patients had to be at least 12 years of age and had to weigh at least 45 kg. Erythema migrans lesions at least 5 cm in diameter were photographed for documentation. Pregnant or nursing women were not enrolled. Exclusion criteria included frank arthritis or objective evidence of central nervous system or cardiac (second- or third-degree block) involvement at time of presentation; evidence of meningismus or Bell palsy with pleocytosis [more than 7 cells/mm3]; and history of 1) nervous system, cardiac, rheumatic, or collagen vascular disease, 2) an immediate hypersensitivity reaction to ß-lactam antibiotics or macrolides, 3) any antibiotic therapy within 72 hours before enrollment or use of any antibiotic during the study other than those supplied, or 4) antibiotic treatment for Lyme disease during the preceding 12 months.
The study protocol was approved by the Institutional Review Board of each study center, and all participants gave written informed consent.
Clinical Diagnostic Evaluations
Patients were evaluated by a physician at baseline and 8, 20, 30, 90, and 180 days after initiation of therapy. Subjective symptom scores for 11 key symptoms (fatigue, joint pain, headache, muscle pain, anorexia, stiff neck, fevers, paresthesias, dizziness, cough, and nausea and vomiting) were recorded on a visual analog scale at each evaluation. Blood samples were obtained for B. burgdorferi serologic testing (IgG and IgM), hepatitis B serologic tests, and liver function tests. Electrocardiography was done. All tests except for the hepatitis B serologic test were repeated on days 8, 20, and 30. In addition, blood samples for B. burgdorferi serologic testing and electrocardiograms were obtained at the 90- and 180-day evaluations. All ELISAs for Lyme disease were done at the University of Minnesota as previously described . Patients were seen at unscheduled visits if indicated by their clinical condition.
Treatment and Study Design
Patients were stratified by the presence of flu-like constitutional symptoms (such as fever, chills, headache, malaise, fatigue, arthralgias, and myalgias) and then randomly assigned to one of the two treatment arms. Each center was given a randomization schedule for two types of presenting symptoms: erythema migrans alone and erythema migrans with flu-like symptoms. These randomization schedules consisted of sequential numbers to which the following study drug regimens were allocated: 500 mg of azithromycin once daily and placebo doses twice daily (to match the three-times-daily dosing regimen of amoxicillin); or amoxicillin, 500 mg three times daily. The drugs were provided by Pfizer Central Research in a double-matching (dummy) form so that all pills for both groups of patients were identical. All patients received the oral (active or placebo) medication for 20 days, but those in the azithromycin group received active drug for only 7 days. Both the clinical investigator and the patient were blinded to treatment assignments.
Efficacy was evaluated in the patients who returned for an examination on day 20 and had taken at least one half of their medication. Patients who withdrew from the study because of adverse events and who took less than one half of their medication were considered nonevaluable for efficacy analysis. Response was assessed by clearance or persistence of erythema migrans and presenting objective signs and by relief of symptoms (assessed using the visual analog scale), and was then graded according to the following criteria.
1. Complete response: complete clearance of erythema migrans and all objective signs and greater than 75% relief of presenting symptoms.
2. Partial response: 1) complete clearance of erythema migrans with persistent signs and 50% to 75% relief of symptoms or 2) persistent erythema migrans with complete clearance of signs and greater than 75% relief of symptoms.
3. Treatment failure: 1) persistent erythema migrans, persistent signs, and less than 50% relief of symptoms or 2) development of new signs and symptoms of disease before the examination on day 20.
Symptom relief was calculated as a percent reduction from baseline in the sum of the symptom scores on the visual analog scale. On subsequent examinations up to 180 days, patients were evaluated for relapse, which was defined as any objective evidence of arthritis, evanescent skin lesions, facial palsies, atrioventricular heart block, or peripheral or central nervous system disease, including meningitis.
Toleration of treatment was determined from treatment-related adverse events and laboratory abnormalities.
To compare responses to therapy, we used chisquare analysis or the Fisher exact test (two-tailed) for ordinal data and the t-test (two-tailed) for interval data. Confidence intervals on percentages were calculated using a normal approximation with are sine transformation.
The baseline demographic and clinical characteristics of the patients are shown in Table 1. The mean diameter of the largest erythema migrans lesion, the distribution of single and multiple erythema migrans lesions, the presence of concomitant flu-like illness, and the seropositivity rate did not differ significantly between the treatment groups. The baseline characteristics were similar, except that the azithromycin group had more patients with multiple erythema migrans lesions.
On physical examination at study entry, the most common signs of disease included lymphadenopathy (18%; lymphadenopathy was regional in 40 patients and generalized in 3), pain on neck flexion (13%), muscle tenderness (12%), and joint tenderness (11%) (Table 2). Of the 11 self-reported symptoms, the most frequent were fatigue (52%), joint pain (34%), headache (29%), muscle pain (26%), anorexia (24%), fever (22%), and stiff neck (21%) (Table 2). Electrocardiographic abnormalities (first-degree block) attributed to Lyme disease were present in four patients (2%); none had cardiac abnormalities by day 20. Although one patient, who had first-degree heart block, had a relapse (joint pain) at 180 days, his electrocardiogram continued to be normal. Mild liver function abnormalities were noted in more than 20% of patients; increases in aminotransferase levels occurred most frequently.
Of the 246 patients enrolled, 217 (88%) were evaluable for efficacy at day 20. Seven patients (2 receiving azithromycin and 5 receiving amoxicillin) were excluded because they had received less than 50% of their study medication as a result of adverse events; 17 patients (8 receiving azithromycin and 9 receiving amoxicillin) did not return for the follow-up examination at day 20; 2 patients (both receiving amoxicillin) were noncompliant; and 3 patients (all receiving azithromycin) did not meet entry criteria. The baseline signs of disease in the nonevaluable patients were regional lymphadenopathy (9 patients [31%]), muscle tenderness (1 patient [3%]), neck pain (2 patients [7%]), pharyngitis (4 patients [14%]), evanescent erythema migrans (3 patients [10%]), right upper quadrant tenderness (1 patients [3%]), and first-degree atrioventricular block (1 patient [3%]). The baseline symptoms of these patients were fatigue (13 patients [45%]), joint pain (8 patients [28%]), headache (7 patients [24%]), muscle pain (7 patients [24%]), anorexia (5 patients [17%]), stiff neck (11 patients [38%]), fever (4 patients [14%]), paresthesia (4 patients [14%]), dizziness (3 patients [10%]), and cough (2 patients [7%]).
Response to Therapy
Twenty days after the initiation of therapy, 93 of 106 patients treated with amoxicillin (88% [95% CI, 80% to 93%]) had had a complete response to therapy compared with 84 of 111 patients treated with azithromycin (76% [CI, 67% to 83%]) (P = 0.024) (Table 3). Furthermore, 3 patients treated with azithromycin had not responded or had worsened within the first 20 days (for example, they had persistent erythema migrans, joint tenderness, or meningitis). All patients treated with amoxicillin responded within the first 20 days. In contrast, when complete and partial responders were pooled and compared with nonresponders, no significant difference was seen between those treated with amoxicillin and those treated with azithromycin. The symptoms of partial responders at day 20 were compared, and the only significant difference between the two groups was that patients treated with azithromycin were significantly more likely (P = 0.017) to have persistent and more severe headache (mean visual analog scale scores ± SE of 6.4 ± 1.7 and 1.9 ± 0.7, respectively).
The clinical response to therapy at day 20 (complete response, partial response, or failure to respond) was correlated with the serologic status of the patient, defined as at least one positive IgG or IgM anti-Borrelia antibody titer (IgG more than 0.873 or IgM more than 0.631 optical density ratio) at any time point. In the azithromycin group, seropositive patients were significantly more likely than seronegative patients (P = 0.043) to have had a complete response (Table 4). In the amoxicillin group, no significant difference in response was seen regardless of the serologic status of the patients (data not shown
Individual analyses were done for the six sites at which most of the patients were enrolled: Old Lyme, Connecticut; Yale University, New Haven, Connecticut; Westchester Medical Center, Valhalla, New York; Long Island, New York; Cape Girardeau, Missouri; and Marshfield, Wisconsin (data not shown). No significant differences were noted in response to therapy at each of these sites. The overall response rates for each treatment group were similar whether the patients from each of these major sites were or were not included in the analysis.
Resolution of the largest erythema migrans lesion was similar in both treatment groups. Multiple erythema migrans lesions or the presence of concomitant flu-like symptoms did not significantly affect the response to therapy measured at day 20.
Treatment Failures and Relapses
Three azithromycin recipients did not respond to therapy during the acute phase of illness (one each on days 14, 15, and 20) compared with no amoxicillin recipients. The signs of failure included neck pain on flexion, muscle tenderness, and joint swelling; the major symptoms reported were fatigue, joint pain, headache, muscle pain, and stiff neck (Table 2). One of the three patients who did not respond had neck pain on flexion at follow-up, subsequently received a diagnosis of lymphocytic meningitis, was treated with intravenous ceftriaxone, and achieved complete resolution of all signs and symptoms. Another patient who did not respond was treated with doxycycline but did not return for subsequent evaluations. The remaining patient did not receive any alternate medications but achieved complete resolution of disease symptoms and continued to be free of symptoms at the 180-day examination.
Relapse was significantly more frequent (P = 0.005) for patients treated with azithromycin. Seventeen patients treated with azithromycin (16% [CI, 10% to 24%]) and 4 patients treated with amoxicillin (4% [CI, 1% to 10%]) developed late complications of infection (Table 5). A partial response at day 20 was predictive of relapse. Ten of 37 patients (27%) who had a partial response compared with only 11 of 172 patients (6%) who had a complete response at day 20 ultimately had relapse within 180 days (P < 0.001) (Table 5).
At the time of relapse, 8 of 17 patients from the azithromycin group (47%) had a positive anti-Borrelia antibody titer (5 [29%] for IgG and 6 [35%] for IgM). Only 1 of 4 patients (25%) in the amoxicillin group who had relapse had a positive antibody titer (both IgG and IgM). In total, 12 patients (57%) who developed objective signs and symptoms of relapse were seronegative for B. burgdorferi at the time of relapse. No significant correlation was seen between single and multiple erythema migrans lesions or the presence of flu-like illness and subsequent relapse (Table 6). For patients in the azithromycin group, the mean time to relapse was twice as long (138 days compared with 59 days) if patients presented without nonspecific signs and symptoms of generalized infection (P < 0.01) (Table 6).
Forty-three patients receiving azithromycin (35%) and 29 patients receiving amoxicillin (24%) reported adverse events. Diarrhea was significantly more common (P = 0.02) among the azithromycin recipients. Therapy was discontinued in 8 patients because of adverse events (2 patients in the azithromycin group had diarrhea, and 6 patients in the amoxicillin group had skin rash). Of these 8 patients, 7 were excluded from analysis because they had received less than 50% of their study medication. The remaining patient, who was treated with amoxicillin, developed a rash on day 19 but achieved a complete resolution of baseline signs and symptoms and was considered evaluable on day 20. This patient continued to be evaluated up to 180 days with no evidence of relapse. Therapy was not discontinued in any patients because of laboratory abnormalities.
Serologic Response to Infection
At the time of presentation, 75 of 217 evaluable patients (35% [CI, 28% to 41%]) were seropositive for B. burgdorferi Figure 1: Fifty-five patients (25%) had a positive anti-Borrelia IgM antibody titer, and 44 (20%) had a positive anti-Borrelia IgG antibody titer. By day 30, 159 patients (73% [CI, 67% to 79%]) had achieved a positive IgG or IgM antibody titer at one or more time points. No patients seroconverted from a negative to a positive antibody titer after day 30. Graphing the cumulative development of a positive anti-Borrelia antibody titer across time shows that of the 73% of patients (n = 159) who were or who became seropositive during the course of the study, more than 90% (n = 154) did so within the first 20 days (Figure 1). However, by day 180, fewer than 25% of patients (45 of 186) remained seropositive.
Patients with single or multiple erythema migrans lesions developed positive anti-Borrelia antibody titers at similar rates, but patients with multiple erythema migrans lesions were significantly (P < 0.001) more likely to have a positive IgG or IgM antibody titer at the time of study entry. Of the 38 patients with multiple erythema migrans lesions, 24 (63%) had a positive IgM antibody titer, and 15 (39%) had a positive IgG antibody titer. In contrast, of the 179 patients with single erythema migrans lesions, 31 (17%) had a positive IgM antibody titer, and 29 (16%) had a positive IgG antibody titer at time of presentation. The baseline or overall seropositivity of patients with and patients without flu-like syndrome did not differ.
In this large, randomized, multicenter, double-blind study of the treatment of erythema migrans, we compared two antibiotic regimens that have been purported to be highly effective in previous prospective, open trials [7, 9, 10]. Our study was designed with stringent entry criteria and standardized assessments of key signs and symptoms throughout the 180-day study period. As a result, we obtained crucial information on the comparative efficacy of these two treatment regimens and on the clinical and laboratory presentation of early Lyme disease and relapse of Lyme disease. We found amoxicillin to be significantly more effective than azithromycin in completely resolving the acute manifestations of erythema migrans and in preventing relapse within 180 days after infection. Our large sample size and rigorous study design may account for the differences between our findings and those of other studies. Given the high rate of efficacy found in smaller, open studies of antimicrobial agents such as doxycycline and amoxicillin , a large cohort of patients was needed to achieve the power to reach a conclusion about similarity in efficacy.
The presence of erythema migrans was the disease-defining entry criterion. The lack of microbiological confirmation of the diagnosis is a major limitation of this and all other studies on the treatment of erythema migrans (for instance, in Missouri, spirochetes morphologically consistent with B. burgdorferi have been seen in ticks, but no isolations from clinical specimens have been reported . As a result, we could not correlate the in vitro sensitivity of the organism to a given antibiotic and to the clinical outcome. It is important to realize that the causative agent of Lyme disease is not homogeneous and may have varying levels of susceptibility to different antimicrobial agents [13-16]. The current microbiological diagnostic limitations of these treatment trials and the emerging spectrum of this disease underscores the necessity for large, well-defined, randomized, double-blind, multicenter studies that include geographically distinct areas.
Amoxicillin was significantly more effective than azithromycin in completely resolving initial signs and symptoms. The differences in overall efficacy may be due to the dissimilar durations of therapy—7 days for azithromycin (despite its long half-life) compared with 20 days for amoxicillin—or a variation in the ability of these drugs to reach certain extracellular compartments, particularly in the central nervous system. Azithromycin has been shown to be among the most active antimicrobial agents against B. burgdorferi, with mean minimum inhibitory concentrations ranging from 0.015 to 0.06 µg/mL [5, 6]. However, azithromycin is similar to other antimicrobial agents in that it requires 48 to 72 hours to effectively kill B. burgdorferi in vitro . Thus, the duration of therapy may be an important factor in ultimate success.
During the 180-day follow-up period, patients treated with azithromycin (n = 17) were significantly more likely than patients treated with amoxicillin (n = 4) to have objective evidence of clinical relapse of Lyme disease. The most common manifestation was joint tenderness and swelling, but 5 patients had pain on neck flexion. The patients reported various cognitive and constitutional symptoms, such as fatigue, headache, and paresthesia. Although they are not conclusive, these subjective symptoms may indicate a subacute central nervous system infection known to occur in patients with Lyme disease . Notably, patients who had a partial response at day 20 were more likely to have relapse. In view of this finding, it is apparent that further studies are needed to delineate the optimal approach to patients with continuing signs or symptoms. The roles of parenteral antibiotics and the duration of therapy in this subpopulation of patients still need to be defined.
Current recommendations for the treatment of early B. burgdorferi infection are predominately founded on small, unblinded studies that lack the size and rigorous study design of our trial. In our study, amoxicillin did extraordinarily well compared with previous reports from studies of this and other antibiotics. Of the 106 patients treated with this antibiotic, all responded during the acute phase of therapy, and only 4 had relapse within 180 days. This finding contrasts strikingly with the activity reported for various other agents that are currently being used as standard therapies for erythema migrans [4, 18, 19].
It is important to note that the chronic manifestations of Lyme disease may occur months to years after the acute infection and that our study was limited to a 180-day follow-up period. Future studies with longer follow-up are needed to establish the true failure rate and long-term sequelae associated with these therapeutic regimens. Our study emphasizes the benefits of a double-blind, prospective study design in assessing therapy for an infectious disease for which long-term follow-up is necessary and clinical criteria are the sole measure of efficacy.
Patients treated with azithromycin were significantly more likely to have a complete response to therapy if they had a significant humoral response to B. burgdorferi during the first 20 days of treatment. This may indicate that early immune response plays a role in limiting the early disease or that the humoral response acts synergistically with azithromycin in treating this acute infection. The reason this same phenomenon was not seen in the amoxicillin group was probably that the high level of activity of amoxicillin cannot be improved further. Of the 21 patients who had relapse, 15 (71%) had clearly demonstrable increases in their anti-Borrelia antibody titer at one or more time points between the onset of illness and the date of relapse. In contrast, on the day of relapse, only 9 of the 21 patients (43%) had an elevated B. burgdorferi antibody titer. This prospective study confirmed our previous observation that a subpopulation of patients treated promptly but ineffectively for erythema migrans may ultimately develop later manifestations of Lyme disease and be seronegative on ELISA tests for B. burgdorferi at the time of relapse . The finding that patients can have relapse and be seronegative indicates that serologic assays are not reliable in the assessment of patients who have been treated for early Lyme disease. Great care must be taken to document all objective abnormalities in these patients. With the advent of new technologies such as Western blotting analysis, polymerase chain reaction, and antigen capture, these assays will be increasingly important in defining this difficult patient population. Although Western blots were not included in our study design because of the lack of standardization and dearth of available monoclonal antibodies at the time of the study, a comparative analysis of these serologic assays is under way.