The + indicates low risk of bias; −, high risk of bias; and ?, unclear risk of bias.
A, Analysis includes complete and partial responders in 6 randomized clinical trials (RCTs) of 6 antibiotic agents and 908 patients. B, Analysis includes 14 RCTs, 9 antibiotic agents, and 1793 patients. C, Analysis includes 9 RCTs, 6 antibiotic agents, and 1211 patients. OR indicates odds ratio; numbers in network graphs, number of comparisons.
A, Analysis includes 8 randomized clinical trials (RCTs), 6 antibiotic agents, and 926 patients. B, Analysis includes 3 RCTs, 3 antibiotic agents, and 437 patients; C, Analysis includes 7 RCTs, 7 antibiotic agents, and 710 patients. OR indicates odds ratio; numbers in network graphs, number of comparisons.
eTable 1. Definition of Treatment Response and Assessment of Treatment Response in the Individual Studies
eTable 2. Study Characteristics
eTable 3. Summary of Findings Table: Different Treatment Regimens
eFigure 1. Outcome: Response to Treatment
eFigure 2. Outcome: Any Reported Adverse Events
eFigure 3. Outcome: Any Skin-Related Adverse Events
eFigure 4. Outcome: Any Gastrointestinal Adverse Events
eFigure 5. Outcome: Jarisch-Herxheimer–like Reactions
eAppendix 1. The PRISMA Extension Statement for Reporting of Systematic Reviews Incorporating Network Meta-analyses of Health Care Interventions
eAppendix 2. Differences Between Protocol and Review
eAppendix 3. Search Strategy Database(s) in MEDLINE(R) (via Ovid on 2015-11-24)
Customize your JAMA Network experience by selecting one or more topics from the list below.
Torbahn G, Hofmann H, Rücker G, et al. Efficacy and Safety of Antibiotic Therapy in Early Cutaneous Lyme Borreliosis: A Network Meta-analysis. JAMA Dermatol. 2018;154(11):1292–1303. doi:10.1001/jamadermatol.2018.3186
What are the most effective and safest antibiotic treatments of early cutaneous Lyme borreliosis?
This network meta-analysis that included 2532 patients did not detect any significant differences in treatment response by antibiotic agent, dose, or duration. There were also no differences in the effect sizes among antibiotic agents and treatment modalities in treatment-related adverse outcomes, which were generally mild to moderate, and treatment failures were rare.
Neither the antibiotic agent nor the treatment modality contributes to comparative effectiveness or drug-related adverse outcomes in early cutaneous Lyme borreliosis.
Controversies about the choice of antibiotic agent and treatment modality exist in the management of erythema migrans in early cutaneous Lyme borreliosis (LB).
To conduct a network meta-analysis (NMA) of all randomized clinical trials on various antibiotic agents and treatment modalities in early cutaneous LB.
Electronic searches in MEDLINE, Embase, and the Cochrane Central Register of Controlled Trials were conducted from inception until July 2017. The reference lists of the included studies were hand searched, authors were contacted, and ongoing trials were searched at ClinicalTrials.gov.
One reviewer screened the titles and abstracts of the 9975 reports identified by the electronic searches. Full-text copies of 161 potentially relevant articles were obtained, and 2 reviewers independently assessed those articles for inclusion. Adults with a physician-confirmed early localized skin infection who were treated with antibiotics of any dose or duration were included.
Data Extraction and Synthesis
Two reviewers independently extracted data on study, patient, and intervention characteristics. Network meta-analyses on treatment effects and adverse outcomes were calculated with a frequentist approach using the R package netmeta. The Grading of Recommendations Assessment, Development and Evaluation guidance for NMA was used to assess the certainty of evidence.
Main Outcomes and Measures
Treatment effects for response to treatment (resolution of symptoms) and treatment-related adverse events.
Overall, 19 studies (2532 patients) were included. The mean patient age ranged between 37 and 56 years, and the percentage of female patients ranged from 36% to 60%. The antibiotics investigated were doxycycline, cefuroxime axetil, ceftriaxone, amoxicillin, azithromycin, penicillin V, and minocycline. Pooled effect sizes from NMAs did not suggest any significant differences in treatment response by antibiotic agent (eg, amoxicillin vs doxycycline odds ratio, 1.26; 95% CI, 0.41-3.87), dose, or duration (eg, doxycycline, 200 mg/d for 3 weeks, vs doxycycline, 200 mg/d for 2 weeks, odds ratio, 1.28; 95% CI, 0.49-3.34). Treatment failures were rare at both 2 months (4%; 95% CI, 2%-5%) and 12 months (2%, 95% CI, 1%-3%) after treatment initiation. There were also no differences in the effect sizes among antibiotic agents and treatment modalities for treatment-related adverse outcomes, which were generally mild to moderate. Certainty of evidence was categorized as low and very low mostly because of imprecision, indirectness, and study limitations (high risk of bias) of the included studies.
Conclusions and Relevance
This NMA suggests that neither the antibiotic agent nor treatment modality contributed to comparative effectiveness or drug-related adverse outcomes. This finding is relevant for physicians treating patients with LB and for patient decision making.
Lyme borreliosis (LB) is a disease caused by the spirochete Borrelia burgdorferi and transmitted by tick bites in most northern hemisphere countries of Europe, North America, and Asia.1-5 Countries in Western Europe have a large variance in incidence rates. The highest rates of LB have been reported in southern Sweden, with 464 per 100 000 person-years, and the lowest rates have been reported in Italy, with 0.001 per 100 000 person-years.6 The population-weighted incidence rate in Western Europe has been estimated by 1 study6 to be 22.04 per 100 000 person-years.
Early localized manifestations of LB include erythema migrans (EM) and, more seldom, borrelial lymphocytoma. Erythema migrans is detected in approximately 70% to 90% of patients with LB7-10 and can have a range of appearances, including the classic bull’s-eye lesion, but atypical lesions are also common.11-15 Erythema migrans typically occurs at the site of the tick bite 3 to 32 days after the bite.12,14,16 Most individuals show a single lesion although approximately 10% to 20% of patients with LB show multiple EM lesions (MEM), which are caused by hematogenic dissemination of the bacteria via the bloodstream into the skin. Early disseminated manifestations of LB include neuroborreliosis (affecting 10%-20% of patients with LB)3,9 and heart problems (affecting 4%-10% of patients with LB).17 Late manifestations (affecting 2%-10% of patients with LB) may involve the joints18 but may also involve the nervous system or the skin. Lyme borreliosis has a substantial disease burden of 10.6 disability-adjusted life-years (DALYs) per 100 000 persons based on the incidence rate in northern Europe (Netherland) in 2010, with 9.1 DALYs for persisting symptoms, 0.9 DALYs for disseminated LB, and 0.6 DALYs for EM.19 Compared with other more common cutaneous diseases, EM has a low disease burden. For example, an update from the Global Burden of Disease Study 201320 reported 55 DALYs for viral skin diseases and 16 DALYs for cellulitis.
Patients with early cutaneous manifestations of LB are generally treated with antibiotics to eliminate the spirochete to cure immediate disease and to prevent late manifestations. The broad-spectrum antibiotic doxycycline is the mainstay of therapy for cutaneous manifestations of LB.21 Other first-line antibacterial agents are phenoxymethylpenicillin (also known as penicillin V), amoxicillin, or cefuroxime axetil.22-24 However, some treatment failures have been reported with all antibiotic agents in use.25-27 In particular, undertreated infections may resurface and result in progression and dissemination to other organs, especially the nervous system and joints, causing long-term morbidity and costs.28-30 Hence, the aim of the present study was to conduct a systematic review of all randomized clinical trials focused on treatment response and drug-related adverse outcomes of various antibiotic regimens in adults with early cutaneous LB to provide evidence for the development of clinical practice guidelines and individual patient decision making.
This research was conducted and reported in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) extension statement for systematic reviews incorporating network meta-analyses (NMA) of health care interventions (eAppendix 1 in the Supplement).31 The methods of this systematic review were prespecified in a protocol published in 2016.32 Modifications to that protocol are described in eAppendix 2 in the Supplement.
To identify eligible studies, we conducted electronic searches in Ovid MEDLINE, Ovid MEDLINE in Process & Other Non-Indexed Citations, Ovid MEDLINE daily update, Ovid Embase, and the Cochrane Central Register of Controlled Trials (via the Cochrane Library) from inception until November 2015. An update search was performed July 2017. The search strategy was based on combinations of medical subject headings and keywords and was not restricted to any specific language. The search strategy used in Ovid MEDLINE is given in eAppendix 3 in the Supplement. Search strategies for the other databases were modified to meet the requirements of each database. In addition, we hand-searched reference lists of included studies and contacted authors. Searches for ongoing trials or trials completed but not published were conducted on ClinicalTrials.gov and the World Health Organization’s International Clinical Trials Registry Platform.33 One reviewer (G.T.) screened the titles and abstracts of all references identified by electronic searches. We obtained full-text copies of all potentially relevant articles, and 2 reviewers (G.T., H.H., K.B., M.H.F., or C.S.) independently assessed those articles for inclusion. Conflicts were resolved by discussion. The randomized clinical trials considered for this review included those studying adults with a clinically (physician-confirmed) early localized skin infection who were treated with antibiotics of any dose for any duration.
We extracted data on patients (eg, number of patients included and randomized, and the proportion of patients receiving a diagnosis of EM or MEM), interventions (eg, antibiotic agent, dosage, length of treatment), and outcomes (response to treatment and any treatment-related adverse event). For the outcome “response to treatment,” we conducted 2 analyses. First, we combined complete responders and partial responders and compared them with treatment failures (persistent EM, disseminated disease, or both). We then combined partial responders and treatment failures and compared them with complete responders. The definition and assessment of response to treatment as described in the individual studies are given in eTable 1 in the Supplement. All reported adverse events were extracted according to the definitions of the primary studies, and individual participants served as the unit of analysis. Because repeated adverse events can occur in the same participant, we considered the number of individuals with at least 1 event of a type, rather than the total number of any adverse event. In addition, we also categorized all adverse events involving the skin or gastrointestinal tract or hematologic or Jarisch-Herxheimer–like reactions using the individual participant as the unit of analysis. Outcome data for which only the number of events, and not the number of participants, was reported were not included.
Risk of bias was assessed according to the method described in the Cochrane Handbook for Systematic Reviews of Interventions.34 The following domains were considered: randomization sequence generation; allocation concealment; masking (blinding) of participants, trial personal, and outcome assessors; incomplete outcome data; selective outcome reporting; and other sources of bias (ie, bias due to problems not covered elsewhere).
The certainty of evidence for the outcomes response to treatment (≥12 months following start of treatment) and any treatment-related adverse event was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach for NMA, which is associated with specific comparisons, including estimates from direct and indirect comparisons.35,36 In brief, our certainty assessment addressed study limitations (risk of bias), incoherence (differences between direct and indirect estimates of effect as defined by the P values of z tests and the size and overlap of 95% CIs of direct and indirect estimates), imprecision (95% CIs that were wide or included or were close to a null effect around the point estimate from the indirect comparison), inconsistency (differences in estimates of effect across studies that assessed the same comparison), and indirectness (in an NMA, indirectness expresses concerns from indirect comparisons, a so-called intransitivity measure).35,37 Indirect estimates were potentially further downgraded for intransitivity (ie, differences in patient characteristics, differing (co-)intervention, differing extent to which intervention of interest was optimally administered, differing comparator, and differences in measurement of outcome). In the event of coherent direct and indirect effect estimates, the higher rating was assigned to the NMA estimate. On the basis of these criteria, the certainty of the evidence for each comparison and outcome was categorized as high, moderate, low, or very low. The results of the NMA are presented as suggested by the GRADE Working Group,35,36 in the Table. Data extraction, risk of bias assessment, and the GRADE assessment were independently conducted by the aforementioned reviewers.
We used NMA for synthesizing information from numerous studies addressing the same outcomes but involving different interventions.38,39 For a given comparison (A vs B), direct evidence is usually provided by studies that compare these treatments directly. In addition, indirect evidence for the A vs B comparison can be provided by synthesizing studies that compare A vs C and B vs C. Hence, NMA combines direct and indirect evidence across a network of studies into a single effect size. For the present NMA, we used a frequentist approach based on the graph-theoretical method by Rücker et al.40,41 For the main analyses, doxycycline irrespective of dose or duration was chosen as the reference treatment. In addition, analyses to investigate the different treatment modalities were conducted using administration of doxycycline, 200 mg/d for 14 to 15 days, as the reference treatment.42 Estimations of treatment effects were calculated based on a random effects model43 using the R package netmeta.44,45 The netmeta function netgraph was used to generate network plots.46
The results for the response to treatment outcome are presented for both short-term (≤2 months following start of treatment) and long-term follow-up (≥12 months following start of treatment). Treatment effects are expressed as odds ratios (ORs) with 95% CIs. Overall, when significant heterogeneity was found among comparable studies (I2 for heterogeneity >50%), pooled estimates were not provided (which was, however, not the case in the present study). Data for the response to treatment outcome were analyzed, if possible, on an intention-to-treat basis.47 For adverse events, we conducted a complete case analysis (ie, including only participants assessed at each specific follow-up time).
After elimination of duplicates among databases, we screened 9975 references, of which the full text was evaluated for 161 references (Figure 1). We included 19 randomized clinical trials (2532 patients) that compared different antibiotic agents, treatment strategies, or both, of which 17 studies were included in the NMA.23,48-65
The main study characteristics are presented in eTable 2 in the Supplement. Across trials, the mean patient age ranged between 37 and 56 years, and the percentage of female patients ranged from 36% to 60%. Four studies enrolled solely patients with solitary EM,48-50,65 8 studies included mixed populations showing solitary EM and MEM (indirect population, patients with MEM<19%),23,52-54,56,58,59,63 and 7 studies did not provide detailed information on the included patient population.51,55,57,60-62,64 The included studies were conducted with outpatients in the United States (7 studies, 1049 patients),17,23,53-57 Slovenia (5 studies, 806 patients),48-50,58,63 Austria (2 studies, 162 patients),51,61 Germany (2 studies, 139 patients),59,64 Croatia (1 study, 88 patients),52 Sweden (1 study, 100 patients)62 and Norway (1 study, 188 patients).65 Median follow-up from treatment initiation was 12 months (range, 6 weeks to 30 months). All studies were active comparator studies, that is, none included a placebo control. Doxycycline (administered at various dosages and durations) was the most frequently investigated drug, with 909 patients assigned.23,48,49,52,54-56,58,60,63,65 Other antibiotics investigated were azithromycin,50,52,53,55,58,59,63 penicillin V,51,59,61-65 cefuroxime axetil,48,50,54,56 and amoxicillin,50,53,65 whereas ceftriaxone64 or ceftriaxone in combination with doxycycline60 or with minocycline61 were investigated in only 1 trial each. Two studies included patients receiving probenecid in addition to amoxicillin to increase the concentration of amoxicillin.23,55
The results of the risk of bias assessment are provided in Figure 2. Five studies were judged to have a high risk of selection bias because of major limitations in the randomization process.48-51,63 In another 4 studies, patients, clinicians, and outcome assessors were aware of the treatment group, leading to a high risk of performance and detection bias.52,59,61,64 Five studies were associated with a low risk of bias for random sequence generation and adequate blinding of outcome assessors.53,54,56,60,65 Selective reporting could not be sufficiently assessed because of missing information regarding important outcome measures (ie, no study protocol was prepublished, or outcome measures were neither predefined nor clearly described). Five studies were poorly reported overall, making an adequate risk of bias assessment impossible.23,55,57,58,62
When antibiotic agents were compared irrespective of dose and duration, pooled effect sizes did not suggest significant differences in treatment response among agents for up to 2 months after treatment initiation (10 studies, 8 agents [nodes], and 1435 patients)17,24,48,52-56,59,61 nor at 12 or more months after treatment initiation (6 studies, 6 agents [nodes], 908 patients, and low and very low certainty) (Figure 3A and Table).48,50,59,60,63,65 Overall, more than 80% of patients showed a treatment response, 84% (95% CI, 80%-88%) at 2 or fewer months after treatment initiation and 80% (95% CI, 74%-85%) at 12 or more months following treatment initiation. The reasons for the treatment failures were mainly disseminated manifestations, including neuroborreliosis or damage of the nervous system (≤2 months, 4% [95% CI, 2%-5%]; ≥12 months, 2% [95% CI, 1%-3%]). For 136 of 908 patients (approximately 15%), treatment response could not be categorized; therefore, these data are missing in the response definition. Additional analyses showed that dosage and duration also did not influence effectiveness expressed as response to treatment, neither up to 2 months after treatment initiation (6 studies, 9 modalities [nodes], and 934 patients)23,48,49,52,55,60 nor at 12 or more months after treatment initiation (8 studies, 10 modalities [nodes], and 1235 patients)48-51,59,60,65,66 (eFigure 1 in the Supplement; individual study data are given in eTable 3 in the Supplement). Overall, NMA findings for the response to treatment outcome were also similar regardless of the definition of treatment responders (ie, combining complete responders and partial responders vs treatment failures or combining partial responders and treatment failures vs complete responders) or the definition of EM (combining patients with localized EM and MEM vs patients with localized EM) although these results are not shown in a figure.
Adverse events occurred in 31% (95% CI, 25%-37%) of patients and were generally categorized as mild and occasionally categorized as moderate.23,48,49,52-56,58-61,63-65 Irrespective of dose and duration, penicillin V showed the lowest risk of any adverse event when compared with doxycycline (OR, 0.42; 95% CI, 0.16-1.10; 14 studies, 9 agents [nodes], 1793 patients, and very low certainty; Figure 3B and Table).23,48,52-56,58-61,63-65 An additional analysis examining dosages and durations of the various treatment modalities, including an extended treatment regimen of doxycycline, 200 mg/d for 20 to 21 days, also did not show significant differences in adverse outcomes when compared with doxycycline, 200 mg/d for 14 to 15 days, as the reference treatment (12 studies, 15 modalities [nodes], and 1624 patients; eFigure 2 in the Supplement; individual study data are given in eTable 3 in the Supplement).23,48,49,52,53,55,58-60,63-65
The percentage of patients with skin-related adverse events, such as photosensitivity, photodermatitis, or urticaria, was 3% (95% CI, 2%-6%) of patients. Irrespective of dose and duration, compared with doxycycline, cefuroxime axetil showed a significantly lower risk of skin reaction (OR, 0.18; 95% CI, 0.07-0.52; 9 studies, 6 agents [nodes], and 1211 patients; Figure 3C).23,48,52,54-56,58,63,65 When different antibiotic treatment modalities were compared irrespective of dose and duration with doxycycline, 200 mg/d for 14 to 15 days, all antibiotics in the network showed a nonsignificant decreased risk of skin-related adverse events (7 studies, 8 modalities [nodes], and 1006 patients; eFigure 3 in the Supplement).48,49,52,55,58,63,65 A pairwise meta-analysis that was not connected to the NMA showed that when compared with doxycycline, 300 mg/d, 20 or 21 days, the risk of skin-related adverse events was significantly decreased with cefuroxime axetil, 1000 mg/d for 20 days (OR, 0.17; 95% CI, 0.06-0.51; 2 studies and 355 patients).54,56
Gastrointestinal tract adverse events, namely vomiting, diarrhea, and abdominal or gastric pain occurred in 14% (95% CI, 10%-20%) of patients. No significant differences were detected in a comparison of antibiotics irrespective of dose and duration, (8 studies, 6 agents [nodes], and 926 patients; Figure 4A).23,52,54-56,58,63,65 In an analysis comparing doxycycline, 200 mg/d for 14 to 15 days, with various dosages and treatment durations of other antibiotics, azithromycin, 500 mg/d for 5 days, showed the lowest risk of gastrointestinal tract adverse events (OR, 0.45; 95% CI, 0.20-1.01; 4 studies, 5 modalities [nodes], and 439 patients; eFigure 4 in the Supplement).52,58,63,65 Further direct comparisons that were not connected to the NMA did not reveal any differences between the effect estimates for any gastrointestinal adverse event (comparison I: azithromycin, 250 mg/d for 5 days, vs doxycycline, 200 mg/d for 10 days [OR, 4.85; 95% CI, 0.45-51.66; 1 study and 57 patients]55; comparison II: amoxicillin plus probenecid, 1500 mg/d for 21 days, vs doxycycline, 200 mg/d for 20 or 21 days [OR, 0.32; 95% CI, 0.01-8.01; 1 study and 75 patients]23; comparison III: cefuroxime axetil, 1000 mg/d for 20 days, vs doxycycline, 300 mg/d for 20-21 days [OR, 1.44; 95% CI, 0.48-4.30; 2 studies and 355 patients]).54,56
The percentage of patients with hematologic adverse events, such as thrombocytopenia, was 1% (95% CI, 0%-3%). An NMA conducted irrespective of dosage and duration showed no difference among the antibiotic agents (3 studies, 3 agents [nodes], and 437 patients; Figure 4B). There was also no difference for hematologic adverse outcome when different treatment modalities were considered within pairwise meta-analyses (comparison I: azithromycin, 500 mg/d for 5 days, vs doxycycline, 200 mg/d for 14-15 days [OR, 2.29; 95% CI, 0.09-57.91; 1 study and 82 patients]52; comparison II: cefuroxime axetil, 1000 mg/d for 20 days vs doxycycline, 300 mg/d for 20-21 days [OR, 0.95; 95% CI, 0.10-9.23; 2 studies and 355 patients]).54,56
Jarisch-Herxheimer–like reaction occurred in 15% (95% CI, 11%-20%) of patients. No differences were identified in the comparison of agents irrespective of dose and duration (7 studies, 7 agents [nodes], and 710 patients; Figure 4C).23,54,56,58,61,63,64 Additional analyses examining various antibiotic treatment strategies using doxycycline, 200 mg/d for 14 to 15 days, as the reference also did not reveal any significant differences (3 studies, 5 modalities [nodes], and 244 patients; eFigure 5 in the Supplement).58,63,64 Moreover, different direct comparisons that could not be included in the network because of missing connectivity also did not reveal any differences in risk of Jarisch-Herxheimer–like reactions (comparison I: cefuroxime axetil, 1000 mg/d for 20 days, vs doxycycline, 300 mg/d, 20-21 days, [OR, 2.03; 95% CI, 0.49-8.45; 2 studies and 355 patients]54,56; comparison II: penicillin V, 3.0-4.5 Mio IU/d for 20-21 days, vs minocycline, 200 mg/d for 21 days [OR, 0.86; 95% CI, 0.02-45.54; 1 study and 39 patients]61; comparison III: amoxicillin [1500 mg/d] in combination with probenecid [1500 mg/d for 21 days] vs doxycycline, 200 mg/d for 20-21 days, [OR, 2.56; 95% CI, 0.61-10.77; 1 study and 75 patients]).23
This NMA addressed treatment response and risk of treatment-related adverse events following treatment of EM in early cutaneous LB with various antibiotic agents and treatment regimens. The studies were conducted in the United States or in various European countries. Despite differences in the spectrum of pathogens that may be present in different regions and given the extended study period of more than 25 years, the results were generally consistent across the trials, suggesting good applicability of the evidence in countries with high risk of borreliosis.
When antibiotic agents were compared irrespective of dose and duration, pooled effect sizes revealed no differences in response to treatment among the interventions up to 2 months following treatment initiation nor at 12 or more months following treatment initiation. In addition, different antibiotic treatment modalities (including 10, 14-15, and 21 days of doxycycline administration) did not differently contribute to EM resolution or the reduction of associated symptoms. Overall, treatment failures were rare (<5%), and most participants enrolled in the examined studies had good clinical outcomes regardless of the antibiotic agent administered or treatment modality used. Treatment-related adverse outcomes occurred in approximately 35% of patients, but the majority of those events were mild. Overall, penicillin V, regardless of dosage and treatment duration, was the antibiotic with the lowest risk of any adverse event (the reduction in effect size compared with that for doxycycline was 58%). Photosensitivity was the most commonly reported skin-related adverse event, and this occurred only in patients who used doxycycline. However, the low or very low certainty of evidence precludes any further speculation. Additional analyses considering different dosage and duration treatment modalities, including an extended treatment regimen of doxycycline for 3 weeks compared with that for 2 weeks, did not show significant differences in drug-related adverse outcomes (eFigure 2 in the Supplement).
The overall certainty of evidence was generally low, occasionally very low, meaning that further research is likely to change the estimate of effects. The major issues were wide confidence intervals (imprecision), indirectness and study limitations (high risk of bias) due to inadequate randomization and the lack of blinding of patients, health care professionals who administered treatment, and outcome assessors. Poor reporting of the included studies made the interpretation of the evidence additionally challenging.
Overall, NMA provides a useful tool for comparing several competing interventions to develop evidence-based clinical guidelines. However, some challenges exist in network connectivity, consistency, and similarity of studies for the study design, settings, patients and (co)-interventions. Although statistical heterogeneity was low, studies differed particularly in baseline characteristics, including the duration of EM, the inclusion patients of with symptoms additional to EM, study settings, and follow-up times; the contribution of these differences to treatment outcomes was unknown. Another issue may be the potentially unclear validity of the clinical diagnosis of EM. We included studies that used a clinical diagnosis of EM for patient enrollment, and this was often described only as physician-documented EM; thus, we had to rely on the authors’ reports. The same issue was true for concomitant symptoms of the disease. It may be difficult for patients and clinicians to differentiate drug-related adverse events from concomitant symptoms of the disease, such as joint pain, neck stiffness, headache, tiredness, or myalgia, resulting in treatment failures. In addition, patient data for treatment responders and failures were not always sufficiently reported and thus could not be categorized and are lacking in the present analysis. Therefore, the corresponding results on response to treatment could be overestimated or even underestimated. The lack of a placebo group prevented assessment of the efficacy of drug treatment when compared with the natural course of disease. Our review was therefore able to only summarize outcome data for comparative effectiveness of antibiotics (relative efficacy) when used for the treatment of EM.
Patients with neuroborreliosis and damage of the nervous system may need to be treated different from patients with early cutaneous LB. A former review from our group67 and a Cochrane review68 addressing these patient populations concluded that there is a lack of well conducted controlled published studies. Therefore, currently, it is not possible to draw firm conclusions on the relative efficacy of accepted antibiotic drug regimens for the treatment of neuroborreliosis or neurologic manifestations. However, the majority of patients included in the assessed trials showed good outcomes, with symptoms resolving by 12 months of treatment initiation regardless of the antibiotic used.
Numerous systematic reviews and guidelines addressing the management of EM exist.17,21,28,42,66,69,70 For example, a descriptive review from 1999 concluded that oral beta-lactam antibiotics and tetracyclines are most effective as first-line treatment of EM and should be administered for 10 to 21 days.21 By contrast, the International Lyme and Associated Diseases Society emphasized in 2014 that antibiotic treatment regimens of less than 20 days should not be offered to patients with EM because failure rates may be unacceptably high.28 Sanchez et al71 provided an update on the treatment of EM in 2016, reporting a sufficient treatment response with a 10-day course of doxycycline. A current guideline from the German Dermatological Society also recommends doxycycline, 200 mg/d for 10 days with a maximum of 14 days, for adolescents and adults, but stresses that the treatment response of the individual patient needs to be closely observed.69 The latest evidence review by the National Institute for Health and Care Excellence in 201870 reported a lack of head-to-head studies comparing doxycycline and amoxicillin. Overall, the guideline concluded that doxycycline should be first-line therapy in EM. The present systematic review is the first review to also consider the results of a recently published study by Eliassen et al65, which directly compares doxycycline and amoxicillin. Conducting an NMA based on the best currently available evidence allowed us to perform quantitative analyses of interventions compared directly in head-to-head studies and of interventions that have not been directly compared in studies. This methodological approach is pivotal to guideline development in the absence of adequate head-to-head comparator evidence.72 Nonetheless, an NMA is not a substitute for a well-conducted randomized clinical trial.
Treatment failures were rare regardless of antibiotic agent or treatment regimen used. There was no high-certainty evidence that extended treatment schemes with doxycycline enhanced therapeutic efficacy in patients with early cutaneous LB. Placebo-controlled studies are not in line with ethical standards; therefore, a comparison between antibiotic treatment and no antibiotic treatment was not feasible in this patient population. The reasons for the designations of low or very low certainty for the relative efficacy and safety in our NMA were imprecision, risk of bias, and indirectness. Rigorously conducted studies with adequate sample size are needed to identify possible smaller differences in comparative effectiveness and to rule out relevant drug-related safety issues.
Accepted for Publication: July 23, 2018.
Corresponding Author: Christine Schmucker, PhD, Institute for Evidence in Medicine (for Cochrane Germany Foundation), Medical Center, Faculty of Medicine, University of Freiburg, Freiburg, Germany (firstname.lastname@example.org).
Published Online: October 3, 2018. doi:10.1001/jamadermatol.2018.3186
Author Contributions: Dr Schmucker had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Torbahn, Hofmann, Freitag, Dersch, Fingerle, Meerpohl, Schmucker.
Acquisition, analysis, or interpretation of data: Torbahn, Hofmann, Rücker, Bischoff, Freitag, Dersch, Fingerle, Motschall, Schmucker.
Drafting of the manuscript: Torbahn, Schmucker.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Torbahn, Rücker, Schmucker.
Obtained funding: Torbahn, Hofmann.
Administrative, technical, or material support: Fingerle, Motschall.
Supervision: Hofmann, Rücker, Dersch, Fingerle, Meerpohl, Schmucker.
Conflict of Interest Disclosures: None reported.
Funding/Support: This study was supported in part by funding from the German Dermatological Society and the German Paul-Ehrlich-Society to Drs Torbahn and Hofmann and from the German Society for Pediatric Infectious Diseases to Dr Torbahn.
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contribution: Cornelius Lehane, MD, Department of Anesthesiology and Critical Care, University Heart Center Freiburg–Bad Krozingen, Medical Center–University of Freiburg, Freiburg, Germany, provided proofreading and editing of an earlier draft of the paper but received no financial compensation for this contribution.