Expert Review of Anti-infective Therapy

The panel placed a high value on preventing disease, thereby avoiding both the unnecessary progression from a potentially preventable infection to one that is chronic and associated with significant morbidity and costs. The panel placed a high value on not causing the abrogation of the immune response. The panel also placed a high value on the ability of the clinician to exercise clinical judgment. In the view of the panel, guidelines should not constrain the treating clinician from exercising clinical judgment in the absence of strong and compelling evidence to the contrary.

The panel placed a high value on avoiding both the unnecessary progression from a potentially curable infection to one that is chronic and the morbidity and costs associated with chronic disease. The panel also placed a high value on the ability of the clinician to exercise clinical judgment. In the view of the panel, guidelines should not constrain the treating clinician from exercising clinical judgment in the absence of strong and compelling evidence to the contrary.

The panel placed a high value on reducing the morbidity associated with chronic Lyme disease and improving the patient’s QoL as well as on the need for individualized risk/benefit assessment and informed shared decision-making. The panel also placed a high value on the ability of the clinician to exercise clinical judgment. In the view of the panel, guidelines should not constrain the treating clinician from exercising clinical judgment in the absence of strong compelling evidence to the contrary.

The panel conducted a Medline search on 5 March 2013 for RCTs and meta-analyses, which investigated using a single dose of doxycycline for antibiotic prophylaxis of Ixodes scapularis bites. The search used this strategy: Ixodes scapularis bites OR erythema migrans/prevention OR erythema chronicum migrans/prevention OR Lyme disease/prevention and these filters: comparative study, clinical trial, meta-analysis, humans. The search identified 99 papers. Four trials of antibiotic prophylaxis following an I. scapularis bite that were conducted in the USA and two meta-analyses involving some or all of those trials were identified and reviewed [56–61]. Three trials were excluded because they investigated the efficacy of various 10-day antibiotic regimens rather than the efficacy of a single 200 mg dose of doxycycline [56–58]. Given that the two meta-analyses drew substantially from these trials, both were excluded. The fourth trial evaluated the effectiveness of a single 200 mg dose of doxycycline following a tick bite for the prevention of an EM rash at the bite site [59].

Overall, the quality of the evidence supporting the use of a single 200 mg dose doxycycline following a tick bite is very low (Table 2), implying that the true effectiveness of a single 200 mg dose of doxycycline is likely to be substantially different from the trial’s reported effectiveness rate [6].

Table 2. Quality of the evidence, in aggregate, supporting single-dose doxycycline for Lyme disease prophylaxis.

The single 200 mg dose doxycycline trial design employed an unvalidated and inappropriate surrogate and the duration of the observation period was inadequate. The reported 87% efficacy of single-dose doxycycline therapy was with regard to the observed reduction in the incidence of an EM rash at the bite site in the doxycycline subjects compared with the placebo subjects (Table 3) [59], but the reliability of this finding is diminished by its imprecision and its clinical significance is questionable (see quality of evidence discussion above). Therefore, the trial’s significant design deficiencies prohibit conclusions regarding the efficacy and, thus, the benefits of single-dose doxycycline therapy for the prevention of Lyme disease.

Table 3. Summary of findings regarding the effectiveness of single-dose doxycycline for prevention of erythema migrans rashes.

Treatment failure may result in seronegative Lyme disease. Although the single-dose doxycycline trial was not designed to determine whether this regimen could result in seronegative Lyme disease, the subject in the doxycycline arm who failed treatment remained negative on follow-up serologic testing, suggesting that this occurred [62,73]. Clinical trials, case reports and studies in non-human primates have also documented instances of seronegative disease [33,74–76]. While the mechanisms allowing for seronegative disease have yet to be fully investigated, antibiotic treatment has been shown to abrogate the immune response in Coccidioides spp. [77], primary syphilis [78], rheumatic fever [79] as well as Lyme disease [80,81]. It is postulated that antibiotic therapy reduces the antigenemia needed for the immune system to establish an immunologic response [77]. Inducing a seronegative disease state may lead to diagnostic and treatment delays, which are associated with poorer outcomes, and the development of a chronic form of the illness [16,27,32,82,83].

The potential harms of the single-dose oral doxycycline prophylactic regimen and the magnitude of those harms significantly outweigh its benefits. In assessing the risk–benefit profile, the panel considered the unknown efficacy of single dose prophylaxis in preventing the development of Lyme disease and the magnitude of the potential harm created by inducing a seronegative state, including its concomitant diagnostic and treatment delays and the resultant increased risk of developing a chronic form of the disease, which is more difficult to treat successfully. The panel also considered findings from a murine model, which demonstrated that the effectiveness of single-dose doxycycline is further reduced in dual infections involving B. burgdorferi and A. phagocytophilum, an important consideration in many regions of the USA. Additionally, the panel noted that the effects of this regimen on the clinical presentation, detection and prevention of other common Ixodes-borne co-infections are unknown.

Clinicians should not use a single 200 mg dose of doxycycline for Lyme disease prophylaxis. (Recommendation, very low-quality evidence)

Clinicians should promptly offer antibiotic prophylaxis for known Ixodes tick bites, in which there is evidence of tick feeding, regardless of the degree of tick engorgement or the infection rate in the local tick population. The preferred regimen is 100–200 mg of doxycycline, twice daily for 20 days. Other treatment options may be appropriate on an individualized basis (see remarks below). (Recommendation, very low-quality evidence).

During the initial visit, clinicians should educate patients regarding the prevention of future tick bites, the potential manifestations of both early and late Lyme disease and the manifestations of the other tick-borne diseases that may have been contracted as a result of the recent bite. Patients receiving antibiotic prophylaxis should also be given information describing the symptoms and signs of a C. difficile infection and the preventative effect of probiotics. Patients should be encouraged to immediately report the occurrence of any and all tick-borne disease manifestations and manifestations suggestive of a C. difficile infection (Recommendation, very low-quality evidence).

Lyme disease often results from unrecognized tick bites [32,84], which do not provide an opportunity for administering antibiotic prophylaxis. When antibiotic prophylaxis is employed for known bites, it is imperative that treatment begin without delay. A recent murine study demonstrated that prophylaxis was most effective when given immediately after a bite and that effectiveness diminished with treatment delays [85]. Although no studies to date have specifically investigated the efficacy of antibiotic prophylaxis for bites from other Ixodes species, it is reasonable to provide prophylaxis for such bites pending future research.

The evidence supporting use of 20 days of antibiotics is limited to the previously mentioned murine trials [69,70]. In the first trial, investigators demonstrated that a long-acting form of doxycycline, with measurable levels for 19 days, was 100% effective for preventing Lyme disease [69]. In the dual-exposure model, the long-acting form of doxycycline was 100% effective for preventing B. burgdorferi and A. phagocytophilum infections [70]. No long-acting, injectable doxycycline preparation is available for use in humans [62], which is why the panel recommends using 100–200 mg of doxycycline twice daily for a minimum of 20 days. One advantage to this regimen is that it would also address situations where patients are exposed to both B. burgdorferi and A. phagocytophilum.

Analysis of the single-dose doxycycline trial highlights the problems inherent in formulating treatment recommendations on the basis of a single study and demonstrates that a randomized, placebo-controlled study design, in and of itself is not a guarantee that the study will produce high-quality evidence. The panel recognizes that recommendations based solely on animal models are also problematic. Therefore, the panel encourages the NIH to fund appropriately designed trials in order to investigate the optimum duration of treatment for a known Ixodes bite.

Given that doxycycline dosages of 100 mg twice daily may not provide adequate levels in all tissues or in all patients [86], some clinicians may prefer to prescribe higher daily doses [52,86–89]. Regardless of the selected dose, clinicians should advise patients to take probiotics daily while on antibiotic therapy. Probiotics reduce the risk of C. difficile colitis and antibiotic-associated diarrhea [44,45].

‘Watchful waiting’ does not satisfy a strict definition of prophylaxis. Rather than acting to prevent disease, this option seeks the early identification and treatment of Lyme disease infections resulting from a known bite. The hallmark of early disease is the EM rash; and as previously noted, almost a third of reported surveillance cases of Lyme disease lack this finding [16,18,63]. Given the possible absence of an EM rash in a patient with a known bite, this option not only withholds primary preventative therapy, it potentially loses an opportunity to provide secondary prevention as well, should the early, non-EM manifestations of the infection be missed. However, patients wishing to avoid antibiotics may prefer this option, in which case clinicians should emphasize that patients must immediately report the occurrence of Lyme-related symptoms so that appropriate antibiotic therapy can be instituted.

In cases where doxycycline is contraindicated, clinicians may consider using other antibiotics known to be effective in Lyme disease, such as amoxicillin, cefuroxime or azithromycin, although there is no evidence to guide decisions with regard to the dose and duration of use for these agents. The excluded trials of antibiotic prophylaxis investigated the therapeutic efficacy of 10 days of amoxicillin, three-times daily [58]; penicillin, four-times daily [56,57] and tetracycline, four-times daily [57]. None of the trials was able to demonstrate efficacy, primarily due to the low incidence of disease in the placebo groups [56,57].

Some guidelines recommend that clinicians learn to estimate attachment times for recovered ticks based on their scutal index, but expertise is required to do this and it is unrealistic to assume that all clinicians can or will acquire such skills. In the single-dose doxycycline study, 9.9% of the bites from nymphal ticks that exhibited any degree of engorgement resulted in the development of an EM at the bite site [59]. Therefore, the panel determined that it was prudent to routinely offer prophylaxis under such circumstances and that withholding therapy from patients who failed to meet an arbitrary minimum tick attachment time was inappropriate. Similarly, the panel recognizes that clinicians frequently lack information regarding current infection rates for a given tick population (often because the research to establish local infectivity rates has not been done) and that tick infection rates in the same locale vary significantly on an annual basis [90]. Therefore, the panel concluded that meeting a specific tick infection rate should not be a prerequisite for antibiotic prophylaxis.

The panel conducted a Medline search on 5 March 2013 for prospective randomized clinical trials investigating the effectiveness of 5–20 days of oral azithromycin, cefuroxime, doxycycline, phenoxymethylpenicillin or amoxicillin for the treatment of EM. The search used the following strategy: (erythema migrans OR erythema chronicum migrans OR lyme OR lyme borreliosis) AND (amoxicillin/therapeutic use OR azithromycin/therapeutic use OR penicillin/therapeutic use OR cefuroxime/therapeutic use OR doxycycline/therapeutic use) AND (Clinical trial OR comparative study OR meta-analysis). The search identified 76 papers; 51 reported trial outcomes.

A preliminary assessment found that 27 papers described studies that either investigated antibiotic treatment of non-EM presentations (23); were primarily interested in disseminated disease (3) or did not involve any of the antibiotics of interest (1). These were not considered further. An additional 15 trials were excluded because additional review demonstrated that they were either retrospective studies (2); incompletely randomized (1); used a symptom list during post-treatment assessments that did not include commonly reported symptoms of the disease (7) or had a non-completion rate of 20% or higher (5). Thus, nine trials met the requirements for this GRADE analysis and were evaluated in detail (Tables 4 & 5) [46–49,53,74,88,91,92].

Table 4. Quality of the evidence, in aggregate, that supports restricting the antibiotic treatment of erythema migrans to 20 or fewer days.

Table 5. Summary of findings regarding the effectiveness of treating an erythema migrans rash with 20 or fewer days of antibiotics based on a re-analysis of the original trial data to reflect patient-centered outcomes.

The quality of the evidence addressing the effectiveness of 5–20 days of antibiotics for the treatment of EM is very low, implying that the true effectiveness of a 5–20 day course of antibiotics for the treatment of an EM rash is likely to be substantially different from the trials’ reported effectiveness rate.

The limitations of the evidence from the original trials reduce the reliability of their findings. Given that no trial directly compared all classes of agents (azithromycin, cefuroxime, doxycycline and phenoxymethylpenicillin/amoxicillin) and direct comparisons between individual trials are hampered by differences in outcome definitions, length of the observation periods and longitudinal data methodologies, the ability to draw valid conclusions regarding the relative effectiveness of commonly prescribed antibiotic regimens is impaired.

To provide comparative information on patient-centered outcomes by agent – information of clinical import to clinicians and patients – the original trial data were reanalyzed. To minimize biases due to variations in trial design, standardized, patient-centered definitions of treatment success and failure and uniform statistical methodology, utilizing the conservative approach of Barsic et al. [49], were applied to the original trial data. To avoid overstating the effectiveness of the investigated antibiotics, the panel specifically chose to assume that those who failed to complete the trial were treatment failures.

Success was defined as the complete resolution of EM and all associated symptoms and findings, without evidence of disease relapse or the development of new manifestations consistent with Lyme disease during the observation period. The panel viewed this outcome definition as the outcome that would matter most to patients and thought it was consistent with the expectation that the appropriate treatment of an EM rash should restore the patient to their pre-morbid baseline.

Failure included any outcome short of that. Subjects described by the investigators as failures and those who were retreated (regardless of the post-retreatment outcome) were considered failures for the purpose of this outcome analysis. Subjects who had ongoing symptoms at the final end point, including those described as ‘partial responders’, were also considered failures. In some instances, this resulted in subjects being re-categorized as failures. Subjects who were ‘unevaluable’, wrongly enrolled, non-compliant, withdrawn prematurely due to adverse reactions to their assigned antibiotic or lost to follow-up were also considered failures for the purpose of this analysis.

Success rates across the nine trials differed significantly. The lowest, 52.2% (CI: 30.6, 73.3), was in the phenoxymethylpenicillin arm of the 1992 trial by Strle et al. and the highest, 93.3% (CI: 68.1, 99.8), was in the high-dose cefuroxime arm in the trial by Eppes and Childs (see Supplementary Appendix III). The two arms with the highest success rates had exceptionally small sample sizes; one arm had 13 subjects, the other had 15 [91]. The two arms with the lowest success rates also had small samples sizes, 23 subjects in one and 26 in the other [46,53].

Success rates were subsequently regrouped by agent and treatment duration and weighted average success rates for the various regimens were then calculated. The outcome results from arms which had non-completion rates equal to or exceeding 20% were excluded from the calculations. As shown in Table 5, success rates for a given treatment duration vary by antibiotic class. Twenty days of phenoxymethyl-penicillin/amoxicillin had the highest overall success rate of all of the regimens, 84.4%, while 11–19 days of these same agents had the lowest success rate, 61.5%.

Serious adverse events, defined as allergic reactions, C. difficile infections, any adverse event resulting in withdrawal from study or change in therapeutic agent and any adverse event labeled by the investigators as ‘serious’ occurred in 20 of 1068 subjects (1.9%) (Table 5). None of the adverse events was specifically categorized as allergic reactions. The majority of serious adverse events involved the skin (11), including non-specific skin rash (6) [74], drug eruptions (4) [53] and serious photosensitivity reaction (1) [46]. Gastrointestinal adverse events were also common, including poor medication palatability in pediatric subjects (2) [91], nausea and vomiting (1) [48] and diarrhea (5) [49,74,88]. A single subject was treated for C. difficile infection shortly after completing treatment [91]. No deaths were reported.

Although the panel did not consider a Jarisch–Herxheimer reaction an adverse event, four EM trials reported a Jarisch–Herxheimer reaction in 60 of 351 subjects (17.1%) (range 12.1–18.7%) [47,53,88,91].

The harms associated with restricting treatment of an EM rash to 20 or fewer days of oral azithromycin, cefuroxime, doxycycline and phenoxymethylpenicillin/amoxicillin outweigh the benefits. In assessing the risk–benefit profile, the panel determined that the failure rates for antibiotic treatment of 20 or fewer days were unacceptably high and that for those who failed treatment, the magnitude of the potential harm created by delaying definitive treatment, which includes the increased risk of developing a chronic and more difficult to treat form of the disease, was too great.

Although it is generally assumed that antibiotic regimens of shorter duration will be associated with a lower rate of significant adverse events, adverse event rates for oral antibiotics are generally quite low regardless of the duration of use [105–107]. The panel concluded that while antibiotic treatment regimens of 20 or fewer days may result in slightly fewer significant adverse events compared with regimens of longer duration, that benefit does not offset the potential harms associated with the unacceptably high failure rates resulting from this treatment approach. Furthermore, as previously noted, the concomitant use of probiotics should reduce the risk of C. difficile colitis and antibiotic-associated diarrhea [44,45].

Treatment regimens of 20 or fewer days of phenoxymethyl-penicillin, amoxicillin, cefuroxime or doxycycline and 10 or fewer days of azithromycin are not recommended for patients with EM rashes because failure rates in the clinical trials were unacceptably high. Failure to fully eradicate the infection may result in the development of a chronic form of Lyme disease, exposing patients to its attendant morbidity and costs, which can be quite significant. (Recommendation, very low-quality evidence).

Clinicians should prescribe amoxicillin, cefuroxime or doxycycline as first-line agents for the treatment of EM. Azithromycin is also an acceptable agent, particularly in Europe, where trials demonstrated it either outperformed or was as effective as the other first-line agents [46–49]. Initial antibiotic therapy should employ 4–6 weeks of amoxicillin 1500–2000 mg daily in divided doses, cefuroxime 500 mg twice daily or doxycycline 100 mg twice daily or a minimum of 21 days of azithromycin 250–500 mg daily. Pediatric dosing for the individual agents is as follows: amoxicillin 50 mg/kg/day in three divided doses, with a maximal daily dose of 1500 mg; cefuroxime 20–30 mg/kg/day in two divided doses, with a maximal daily dose of 1000 mg and azithromycin 10 mg/kg on day 1 then 5–10 mg/kg daily, with a maximal daily dose of 500 mg. For children 8 years and older, doxycycline is an additional option. Doxycycline is dosed at 4 mg/kg/day in two divided doses, with a maximal daily dose of 200 mg. Higher daily doses of the individual agents may be appropriate in adolescents.

Selection of the antibiotic agent and dose for an individual patient should take several factors into account. In the absence of contraindications, doxycycline is preferred when concomitant Anaplasma or Ehrlichia infections are possibilities. Other considerations include the duration and severity of symptoms, medication tolerability, patient age, pregnancy status, co-morbidities, recent or current corticosteroid use [54,55], cost, the need for lifestyle adjustments to accommodate certain antibiotics and patient preferences. Variations in patient-specific details and the limitations of the evidence imply that clinicians may, in a variety of circumstances, need to select therapeutic regimens utilizing higher doses, longer durations or combinations of first-line agents. (Recommendation, very low-quality evidence)

Clinicians should provide ongoing assessments to detect evidence of disease persistence, progression or relapse or the presence of other tick-borne diseases. Lacking a test of cure, ongoing assessments are crucial for determining if treatment has been clinically effective (see remarks following Recommendation 2f). The first assessment should immediately follow the completion of therapy and subsequent evaluations should occur on an as-needed basis. (Recommendation, very low-quality evidence)

Clinicians should continue antibiotic therapy for patients who have not fully recovered by the completion of active therapy. Ongoing symptoms at the completion of active therapy were associated with an increased risk of long-term failure in some trials and therefore clinicians should not assume that time alone will resolve symptoms (see remarks following Recommendation 2f). There is a wide range of options and choices must be individualized, based on the strength of the patient’s initial response. Dosage ranges for oral agents are as noted in Recommendation 2b.

Strong-to-moderate responses favor extending the duration of therapy of the initial agent at the same dosage. Modest responses may prompt an increase in the dosage of the initial antibiotic or a switch to a different first-line agent. Tetracycline, with a total daily dose of 1000–1500 mg in three or four divided doses, is an additional option [50,109]. Due to its favorable pharmacokinetics, tetracycline may be more effective than doxycycline when initial therapy is non-curative [109].

Minimal or absent responses suggest a need for a combination of first-line agents, which includes at least one antibiotic that is able to effectively reach intracellular compartments [109,110]. Injectable penicillin G benzathine (Bicillin LA), totaling 1.2–3.6 million units weekly, or iv. agents such as ceftriaxone are other options. Intramuscular (IM) benzathine penicillin avoids the risks associated with gaining iv. access and it was effective in seemingly recalcitrant Lyme arthritis [111]. Ceftriaxone, 2 g iv. per day is known to be effective [16,17,32,33,54,112] and iv. cefotaxime [113], another cephalosporin, has also been recommended. iv. penicillin is less effective and requires more frequent dosing [114]. Additional iv. cell wall agents from the carbapenem and monobactam classes were effective in vitro, but have not been studied clinically [115].

Disease progression or recurrence suggests that the iv. agents or injectable penicillin G benzathine, as discussed above, may be required. For patients requiring antibiotic therapy beyond the initial treatment period, subsequent decisions regarding the modification or discontinuation of treatment should be based on the therapeutic response and treatment goals. Additionally, minimal or absent responses and disease progression require a re-evaluation of the original diagnosis (see remarks following Recommendation 2f). (Recommendation, very low-quality evidence).

Clinicians should retreat patients who were successfully treated initially, but subsequently relapse or have evidence of disease progression. Support for retreatment is drawn from the EM trials themselves. In seven of the nine trials reviewed in this analysis [46,48,53,74,88,91,92], subjects who had evidence of treatment failure during the observation period were offered retreatment. Regimens used either oral [46,48,53,74,88,91,92] or iv. antibiotics [48,53,74,88,92], with the choice of agent and route apparently reflecting the investigators’ clinical assessments and treatment preferences.

Therapeutic options include repeating the initial agent, changing to another oral agent or instituting injectable penicillin G benzathine or iv. ceftriaxone therapy. The previously listed dosage ranges for the individual agents would be appropriate. Choices must be individualized and based on several factors, including: the initial response to treatment; the time to relapse or progression; the current disease severity and the level of QoL impairments.

Prior to instituting additional antibiotic therapy, the original diagnosis should be reassessed and clinicians should evaluate patients for other potential causes that would result in the apparent relapse or progression of symptoms and/or findings (see remarks following Recommendation 2f).

The presence of other tick-borne diseases, in particular, should be investigated if that had not already been done. I. scapularis transmits several pathogens and the resulting infections may produce symptoms similar to those of Lyme disease. Thus, apparent relapse or disease progression following antibiotic therapy for Lyme disease may be indicative of a concurrent co-infection and not the failure to eradicate B. burgdorferi. The presence of other Ixodes-borne infections may increase the severity and duration of Lyme disease symptoms [116,117]. Treatment of dually infected patients has not been studied, therefore, the optimal antibiotic regimen for the Lyme disease component is unknown. The possibility of co-infections should not be casually dismissed. Two published surveys of Lyme disease patients found that many respondents were infected with more than one tick-borne pathogen [118,119]. A survey of 3090 patients diagnosed with Lyme disease found that laboratory confirmed cases of babesiosis and anaplasmosis were reported by 32.3 and 4.8% of respondents, respectively [119].

Following a long period of disease latency, minimal manifestations causing little deterioration in the patient’s QoL favor continued observation or repeating therapy with the initial agent; mild manifestations or QoL impairments may prompt a switch to a different first-line agent, tetracycline [50,109], or a combination of first-line agents (which includes at least one antibiotic that is able to effectively reach intracellular compartments) [109,110,120]. Intravenous or IM antibiotics such as injectable penicillin G benzathine or iv. ceftriaxone are other options.

Disease relapse or progression with mild manifestations or QoL impairments occurring within a few months of treatment suggests a need for longer regimens using either a combination of oral first-line agents, injectable penicillin G benzathine or iv. ceftriaxone. Regardless of the duration of disease latency, when disease manifestations or QoL impairments are significant or rapidly progressive, injectable penicillin G benzathine or iv. ceftriaxone may be required. Subsequent decisions regarding the modification or discontinuation of a patient’s treatment should be based on the individual’s therapeutic response and preferences (Recommendation, very low-quality evidence).

Clinicians should educate patients regarding the potential manifestations of Lyme disease, carefully explaining that disease latency can be prolonged. Education should also include information on preventing future bites, the manifestations of the other tick-borne diseases that they may have contracted as well as the symptoms and signs of a C. difficile infection and the preventative effect of probiotics. Patients should be encouraged to immediately report the occurrence of any recurrent or newly developing manifestation of Lyme disease as well as those suggestive of other tick-borne diseases or a C. difficile infection. Clinicians should emphasize that the need to report manifestations of tick-borne diseases never expires. (Recommendation, very low-quality evidence)

The panel conducted a Medline search on 5 March 2013 for RCTs investigating the effectiveness of antibiotic retreatment in patients with persistent manifestations of Lyme disease following treatment considered by some to be standard and appropriate antibiotic therapy for their stage of illness. The search used this strategy: chronic Lyme disease OR Lyme encephalopathy OR persistent Lyme disease AND antibacterial Agents/administration & dosage and this filter: clinical trial.

Five RCTs conducted in the USA were identified. Four met the inclusion criteria for this analysis [16–18]. A fifth trial had a non-completion rate in excess of 20% [87] and was excluded from this analysis on that basis. A Swedish trial was also excluded due to excessive incomplete data [125].

The four trials had unique designs. In one trial, Klempner et al. exclusively enrolled seropositive subjects and treatment consisted of 30 days of iv. ceftriaxone followed by 60 days of oral doxycycline or an identical placebo regimen [18]. A second trial by that same group used an identical design except enrolled subjects were exclusively seronegative [18]. Krupp et al. enrolled seropositive subjects with severe fatigue; participants received either 30 days of iv. ceftriaxone or an identical placebo [17]. Fallon et al. enrolled seropositive subjects with Lyme encephalopathy; treatment consisted of either 10 weeks of iv. ceftriaxone or an identical placebo [16].

The quality of the evidence regarding the effectiveness of antibiotic retreatment in patients with persistent symptoms following standard and appropriate antibiotic therapy for Lyme disease is very low (Table 6), implying that the true effectiveness of retreatment is likely to be substantially different from the effectiveness rates seen in the four NIH-sponsored retreatment trials.

Table 6. Quality of the evidence, in aggregate, that supports antibiotic retreatment in patients with persistent symptoms of Lyme disease.

Retreatment with ceftriaxone was effective in two of the four trials (Table 7). Krupp et al. found that 28 days of ceftriaxone was more effective than placebo (64 vs 18.5%; p < 0.001) for producing a clinically significant reduction in severe fatigue, a primary outcome [17]. The effect size was moderate to large [127]. Fallon et al. found that subjects treated with 70 days of iv. ceftriaxone achieved a moderate improvement (effect size = 0.81) in generalized cognitive function at 2 weeks post-therapy compared with those in the placebo arm (effect size = 0.30) (p = 0.053), although the preferential effect of drug versus placebo was not sustained at 14 weeks post-therapy [16]. The mechanisms leading to the subsequent loss of the cognitive gains are unknown; however, this long-term outcome may indicate that the offered therapy was incomplete. A planned secondary analysis demonstrated an interaction effect between baseline impairments and treatment, such that the ceftriaxone effect increased with higher baseline severity; this was demonstrated for the measures of pain and physical dysfunction at week 12 and sustained to week 24 [16]. On post hoc analysis, Fallon et al. also demonstrated a positive treatment effect on severe fatigue. Of the subjects in the trial by Fallon et al., who met the fatigue entrance criteria of the trial by Krupp et al., those who received ceftriaxone experienced significant reductions in the level of their fatigue compared with those who received placebo (66.0 vs 25.0%; p < 0.05).

Table 7. Summary of findings regarding the effectiveness of antibiotic retreatment in patients with persistent manifestations of Lyme disease.

The NIH-sponsored retreatment trials described 15 serious adverse events among the 221 subjects (6.8%) [16–18]. Each event was associated with ceftriaxone itself or the need for venous access; 60 days of oral doxycycline therapy was not associated with any significant adverse event. Six individuals experienced allergic reactions [16–18], including one case of anaphylaxis [17]. Seven events were related to the iv. line [16–18], four cases involved line-related infections (all on placebo) [16,17], two cases involved thrombi [16] and one subject developed a pulmonary embolus [18]. Additionally, there was one case of cholecystitis [16] and one case of gastrointestinal bleeding with fever and anemia [18].

The clinical population of patients with persistent manifestations of Lyme disease is heterogeneous; therefore, the risk–benefit assessment needs to be done on an individualized basis, taking into account the severity of an individual’s persistent disease, their responsiveness to treatment, their ability to tolerate side effects associated with additional and potentially long-term treatment as well as their willingness to accept the risk associated with antibiotic treatment or, conversely, the level of their desire to avoid treatment-associated risk.

The scientific evidence regarding potential etiologic mechanisms for the development of persistent manifestations of Lyme disease continues to evolve. Proposed mechanisms include immune dysregulation of various types, tissue injury, infection-induced secondary conditions, unrecognized or undertreated co-infections and persistent infection [129,130]. Of these, we think the weight of the evidence supports persistent infection, although other mechanisms may co-exist and the exact etiology for persistent manifestations may vary from patient to patient. Given this uncertainty, the panel concluded that the evidence at hand regarding persistent infection and the potential benefits of retreatment are adequate to support those who wish to treat but is not overwhelming enough to mandate treatment.

The panel also determined that there is no compelling evidence to support routinely withholding antibiotic retreatment from ill patients. While antibiotics are not always effective, the importance of providing patients with the opportunity to receive an adequate trial of antibiotic therapy is heightened by the lack of other effective treatment approaches. Palliative care may be helpful in addressing some symptoms in some cases, but it is important to bear in mind that palliative interventions also carry risks. Additionally, clinicians must not assume that palliative interventions would provide adequate treatment in the face of an underlying persistent infection. Therefore, in the panel’s judgment, antibiotic retreatment will prove to be appropriate for the majority of patients who remain ill and thus it is inappropriate to constrain clinicians from exercising their clinical judgment.

In making these determinations, the panel considered the strength of the evidence addressing the effectiveness of antibiotic retreatment, the burden of disease and the risks associated with various antibiotic options. The panel weighed each in light of the marked heterogeneity within this patient population.

Potential benefits include the restoration of health, improved QoL and prevention of further decline in health status. While complete restoration of health was not identified in any of the four retreatment trials, the moderate-to-large treatment effect on severe fatigue demonstrated in the trial by Krupp et al. and the sustained interaction effects between baseline severity and improvements in pain and physical functioning seen in the trial by Fallon et al. suggested to the panel that retreatment may improve the QoL of some patients.

Others have reached a similar conclusion. In a recent review of the four retreatment trials, Fallon et al. make the point that guidelines restricting the use of antibiotics in patients with persistent manifestation of Lyme disease are based on the erroneous dismissal of the treatment efficacy demonstrated in two of the trials [127]. The authors state that such guidelines ‘are not helpful to clinicians and patients’ [127].

In addition to the NIH-sponsored retreatment trials, retreatment was also shown to be beneficial in clinical trials of EM treatment and in a case series involving the treatment of late neurologic disease. Investigators in seven of the nine EM trials discussed above retreated subjects who failed initial therapy [47,48,53,74,88,91,92]. Decisions to retreat were often based on symptoms alone and investigators frequently reported on the success of retreatment. In three trials, biopsy specimens from the EM site were culture-positive for B. burgdorferi 1–3 months post-treatment [47,48,92]. In two of these, subjects were retreated with oral antibiotics and follow-up cultures 3 [47] or 4 months later [92] were negative. Thus, these trials simultaneously demonstrated persistent infection following standard therapy and the value of retreatment.

In a study by Logigian et al., one subject relapsed at 8 months post-treatment, was retreated, became well once again and remained so for the remainder of the study [33]. Several observational studies also demonstrated benefits from antibiotic retreatment [87,109,110,131].

The panel also considered the risk of withholding antibiotics from patients with a potentially treatable B. burgdorferi infection. Currently available laboratory tests are unable to confirm or deny persistent infection on a routine basis yet persisting infection has been demonstrated in patients with Lyme disease by PCR and culture [47,113,132–136]. A recently published xenodiagnostic study in humans demonstrated positive results in one of eight subjects with post-treatment manifestations of Lyme disease; a subsequent xenodiagnostic specimen obtained from the same subject 8 months later was also positive [137]. Animal studies have corroborated the human findings, documenting bacterial persistence by culture, PCR and histopathologic testing of post-treatment necropsy specimens and by xenodiagnosis [76,138,139]. Given these realities, withholding antibiotic retreatment risks allow an infection to continue unchecked.

The panel weighed the burden of chronic illness that Lyme disease imposes on patients. In the four retreatment trials analyzed here, the subjects’ QoL was consistently worse than that of control populations and reductions in employment or educational activities were common [16–18]. A community-based trial of antibiotic retreatment found the QoL of its subjects was the same or worse as that of individuals with depression, diabetes, heart disease, osteoarthritis and rheumatoid arthritis [87]. Surveys of Lyme disease patients further document the negative impact of persistent manifestations. One survey of openly recruited Lyme disease patients identified 2424 patients whose initial clinical diagnosis of Lyme disease was confirmed with positive serology and who had persistent manifestations of Lyme disease despite antibiotic treatment [140]. Of this cohort, 25% had received public support or disability benefits and the majority of respondents in this subset received these payments for 2 or more years. A second online survey identified 1087 respondants diagnosed with Lyme disease (based on the presence of an EM rash or positive two-tier testing that used the CDC interpretive criteria) who had ongoing manifestations of Lyme disease for 6 or more months [119]. Using a CDC metric of health-related QoL, the survey found that this group averaged 19.6 and 15.5 days/month of poor physical and mental health days, respectively. Not surprisingly, 71.6% rated their health as fair or poor. This rate is higher than that seen in other chronic diseases including congestive heart failure, fibromyalgia, post- stroke and post-myocardial infarction status, diabetes and multiple sclerosis and the survey findings corroborate those of the community-based retreatment trial mentioned above. By comparision, in a general population with an average age of 46, only 16% rated their health as fair or poor [119]. The respondants also reported significant economic impacts – 39.4% stopped working and an additional 28.3% reduced their work hours or role; 37.3% spent at least US$5000 in out-of-pocket Lyme-related expenses.

Given the severity of the QoL impairments, the panel viewed the need for clinical intervention as high.

Additionally, the panel considered that antibiotic risk varies by agent and route of administration. Although all of the regimens in the NIH-sponsored retreatment trials incorporated iv. ceftriaxone, the use of iv. antibiotics is discretionary and should be based on an individualized risk–benefit assessment. The risks associated with iv. antibiotics have two main origins. The first is the medication itself and includes allergic reactions and other adverse events, such as cholecystitis from ceftriaxone. The second source of risk is the iv. access device.

The risks associated with iv. access are well known. A meta-analysis of the risks associated with iv. access, in general, found that risks varied by access type; peripheral iv. catheters caused 0.5 bloodstream infections per 1000 intravascular device days, while surgically implanted long-term central venous devices – cuffed and tunneled catheters – caused 1.6 infections per 1000 intravascular device days [141].

Combined, there were seven device-related adverse events among the four retreatment trials and approximately 8110 days of device use, yielding 0.86 device-related adverse events per 1000 intravascular device days, which is lower than the rate found in the meta-analysis. Although the risk associated with iv. antibiotics is significant, in situations where the QoL impairments are substantial, retreatment with iv. antibiotics may be wholly appropriate.

There is substantial evidence on the clinical safety of amoxicillin, cefuroxime axetil, doxycycline and azithromycin, which are commonly used to treat Lyme disease [105,106]. In a community-based trial, none of the subjects randomized to amoxicillin experienced a serious adverse event [87]. Similarly, the trials by Klempner et al. confirmed the safety of oral doxycycline for longer-term use [18]. Regardless of treatment agent and route of administration, it is expected that the concomitant use of probiotics would reduce the risk of C. difficile colitis and antibiotic-associated diarrhea [44,45].

Values: The panel placed a high value on reducing the morbidity associated with chronic Lyme disease and improving the patient’s QoL as well as on the need for individualized risk/benefit assessment and informed shared decision-making. The panel also placed a high value on the ability of the clinician to exercise clinical judgment. In the view of the panel, guidelines should not constrain the treating clinician from exercising clinical judgment in the absence of strong compelling evidence to the contrary.

Clinicians should discuss antibiotic retreatment with all patients who have persistent manifestations of Lyme disease. These discussions should provide patient-specific risk–benefit assessments for each treatment option and include information regarding C. difficile infections and the preventative effect of probiotics (although none of the subjects in the retreatment trials developed a C. difficile infection). (Strong recommendation, very low-quality evidence. Note: In GRADE, a strong recommendation may be made in the face of very low-quality evidence when the risk–benefit analysis favors a particular intervention such that most patients would make the same choice.)

While continued observation alone is an option for patients with few manifestations, minimal QoL impairments and no evidence of disease progression, in the panel’s judgment, antibiotic retreatment will prove to be appropriate for the majority of patients who remain ill. Prior to instituting antibiotic retreatment, the original Lyme disease diagnosis should be reassessed and clinicians should evaluate the patient for other potential causes of persistent disease manifestations. The presence of other tick-borne illnesses should be investigated if that had not already been done. Additionally, clinicians and their patients should jointly define what constitutes an adequate therapeutic trial for this particular set of circumstances.

When antibiotic retreatment is undertaken, clinicians should initiate treatment with 4–6 weeks of the selected antibiotic; this time span is well within the treatment duration parameters of the retreatment trials. Variations in patient-specific details and the limitations of the evidence imply that the proposed duration is a starting point and clinicians may, in a variety of circumstances, need to select therapeutic regimens of longer duration.

Treatment options are extensive and choices must be individualized. Each of these options would benefit from further study followed by a GRADE assessment of the evidence and consideration of associated risks and benefits, but until this information is available, clinicians may act on the currently available evidence.

In choosing between regimens, clinicians should consider the patient’s responsiveness to previous treatment for Lyme disease, whether the illness is progressing and the rate of this progression; whether the patient has impaired immune system functioning or has received immunosuppressant corticosteroids [54,114] and whether other co-morbidities or conditions would impact antibiotic selection or efficacy. The possibility of co-infections should be investigated (see Recommendation 2e for discussion regarding co-infections complicating the diagnosis and treatment of Lyme disease).

Clinicians should also weigh the extent to which the illness interferes with the patient’s QoL, including their ability to fully participate in work, school, social and family-related activities and the strength of their initial response against the risks associated with the various therapeutic options. Antibiotic selection should also consider medication tolerability, cost, the need for lifestyle adjustments to accommodate the medication and patient preferences.

For patients with mild impairments who had a strong-to-moderate response to the initial antibiotic, repeat use of that agent is favored. Patients with moderate impairments or only a modest response to the initial antibiotic may benefit from switching to a different agent or combination of agents; the latter to include at least one agent that is able to effectively reach intracellular compartments [109,110]. Injectable penicillin G benzathine or iv. agents such as ceftriaxone are other options.

For patients with significant impairments and/or a minimal or absent therapeutic response, a combination of oral antibiotics or injectable penicillin G benzathine or iv. ceftriaxone alone, or in combination with other agents, is preferred. For patients who experienced disease progression despite earlier therapy, treatment with injectable penicillin G benzathine or an iv. agent, such as ceftriaxone, alone or in combination with other antibiotics, is advisable. Additionally, minimal or absent responses and disease progression require a re-evaluation of the original diagnosis. (Recommendation, very low-quality evidence)

Clinicians should re-assess patients immediately following the completion of the initial course of retreatment to evaluate the effectiveness of retreatment and the need for therapeutic adjustments. Reassessment may need to be done much earlier and with greater scrutiny in patients with severe disease or when the therapeutic intervention carries substantial risk.

For patients who improve yet continue to have persistent manifestations and continuing QoL impairments following 4–6 weeks of antibiotic retreatment, decisions regarding the continuation, modification or discontinuation of treatment should be based on several factors. In addition to the factors listed in Recommendation 3b, the decision to continue treatment may depend on the length of time between the initial and subsequent retreatment, the strength of the patient’s response to retreatment, the severity of the patient’s current impairments, whether diagnostic tests, symptoms or treatment response suggest ongoing infection and whether the patient relapses when treatment is withdrawn.

In cases where the patient does not improve after 4–6 weeks of antibiotic retreatment, clinicians should reassess the clinical diagnosis as well as the anticipated benefit. They should also confirm that other potential causes of persistent manifestations have been adequately investigated prior to continuing antibiotic retreatment. Decisions regarding the continuation, modification or discontinuation of treatment should consider the factors noted above as well as the definition of an adequate therapeutic trial.

Whenever retreatment is continued, the timing of subsequent follow-up visits should be based on the level of the therapeutic response, the severity of ongoing disease, the duration of current therapy and the need to monitor for adverse events (see remarks section below). (Recommendation, very low-quality evidence).