Limited data are available regarding the relationship of Borrelia burden in skin of patients with erythema migrans (EM) and the disease course and post-treatment outcome.
We studied 121 adult patients with EM in whom skin biopsy specimens were cultured and analyzed by quantitative PCR for the presence of Borreliae. Evaluation of clinical and microbiological findings were conducted at the baseline visit, and 14 days, 2, 6, and 12 months after treatment with either amoxicillin or cefuroxime axetil.
In 94/121 (77.7%) patients Borrelia was detected in skin samples by PCR testing and 65/118 (55.1%) patients had positive skin culture result (96.8% B. afzelii, 3.2% B. garinii). Borrelia culture and PCR results correlated significantly with the presence of central clearing and EM size, while Borrelia burden correlated significantly with central clearing, EM size, and presence of newly developed or worsened symptoms since EM onset, with no other known medical explanation (new or increased symptoms, NOIS). In addition, the logistic regression model for repeated measurements adjusted for time from inclusion, indicated higher Borrelia burden was a risk factor for incomplete response (defined as NOIS and/or persistence of EM beyond 14 days and/or occurrence of new objective signs of Lyme borreliosis). The estimated association between PCR positivity and unfavorable outcome was large but not statistically significant, while no corresponding relationship was observed for culture positivity.
Citation: Stupica D, Lusa L, Maraspin V, Bogovič P, Vidmar D, O’Rourke M, et al. (2015) Correlation of Culture Positivity, PCR Positivity, and Burden of Borrelia burgdorferi Sensu Lato in Skin Samples of Erythema Migrans Patients with Clinical Findings. PLoS ONE 10(9): e0136600. https://doi.org/10.1371/journal.pone.0136600
Editor: Brian Stevenson, University of Kentucky College of Medicine, UNITED STATES
Received: June 5, 2015; Accepted: August 5, 2015; Published: September 9, 2015
Copyright: © 2015 Stupica et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Data Availability: All relevant data are within the paper.
Funding: The microbiological analyses were funded by Baxter. The clinical part of the study was supported by the Slovenian Research Agency (grant number P3-0296). Baxter provided support in the form of salaries for authors MOR, AT and IL, but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The specific roles of these authors are articulated in the 'author contributions' section.
Competing interests: The microbiological analyses were funded by Baxter, the former employer of I. Livey, M. O'Rourke and A. Traweger. M. O'Rourke and I. Livey hold Baxter stocks/shares and a patent for a vaccine against Lyme Borreliosis. F. Strle was a consultant to Baxter regarding Lyme vaccine (function ended in 2013) and a member of the steering committee of the ESCMID Study Group on Lyme Borreliosis/ESGBOR. Patent details are as follows: 1. I. Livey, M. A. O’Rourke, M. Schwendinger and B.A. Crowe (2011), chimeric OspA genes, proteins, and methods of use thereof 31315/44557A1. 2. I. Livey, M. A. O’Rourke, M. Schwendinger and B.A. Crowe (2011), chimeric OspA genes, proteins, and methods of use thereof 31315/44557A2. 3. P. N. Barrett, G. Aichinger, B.A. Crowe, I. Livey and N. Wressnig (2012), compositions comprising chimeric OspA molecules and methods of use thereof 31315/47123A. 4. 6745WO1 BX2011T00615 / 31315/44557A1 PCT PCT/US11/36525. 5. 6745U1US BX2011T00613 / 31315/44557A1 USSN 13/107,787. 6. 6745WO2 BX2011T00616 / 31315/44557A2 PCT PCT/US11/36533. 7. 6745U2US BX2011T00614 / 31315/44557A2 USSN 13/107,796. There are no further patents, products in development or marketed products to declare. This does not alter the authors' adherence to all the PLOS ONE policies on sharing data and materials, as detailed online in the guide for authors.
Erythema migrans (EM) is the clinical hallmark of early Lyme borreliosis (LB), the most common tick-borne human disease in the Northern Hemisphere, caused by Borrelia burgdorferi s.l. [1–3]. Although antibiotic therapy is highly effective in resolving this manifestation, a proportion of patients (from zero to 40.8% in the USA and from zero to 23.4% in Europe  complain of persistent subjective symptoms lasting for more than 6 months after treatment, termed post-LB symptoms. The potential trigger mechanisms of these symptoms remain unclear and may be multifactorial. In our previous study we suggested that Borrelia skin culture positivity may predict a less favorable treatment outcome in EM patients . Skin culture positivity was found to be significantly associated with larger numbers of spirochetes in EM skin lesion specimens [6–8], suggesting that borrelial burden in the skin influences the course and post-treatment outcome of the disease. To further investigate the pre-treatment course and the post-treatment outcome in EM patients, we compared demographic, epidemiological, clinical, and microbiological findings according to the quantity of B. burgdorferi sensu lato as detected by quantitative PCR of EM skin samples taken before antibiotic treatment. We hypothesized that a higher borrelial burden in the skin of EM patients might influence the pre-treatment course and post-treatment outcome of the disease.
Setting and Patients
Patients were evaluated from June through December 2010 at the LB Outpatient’s Clinic, University Medical Center Ljubljana, Slovenia. Patients were enrolled in the study if they were ≥15 years old, with typical solitary EM defined according to the criteria reported by Stanek et al . Patients were excluded if they did not consent to the skin biopsy; had EM localized on head, neck or breasts; were pregnant or lactating; had received an antibiotic with known anti-borrelial activity within 10 days; had multiple EM lesions; had an extracutaneous manifestation of LB; or had an intercurrent episode of LB during follow-up. Patients received cefuroxime axetil 500 mg twice daily, or amoxicillin 500 mg three times daily for 15 days. The antibiotic treatment option was varied weekly during study entry.
The study was approved by the Medical Ethics Committee of the Ministry of Health of the Republic of Slovenia, No 127/06/10. All patients gave written consent for participation. For the two patients aged < 18 years, we additionally obtained their parents’ written consent.
Results on quantitative detection of Borrelia burgdorferi sensu lato in EM skin lesions of patients included in the present study have been reported in detail previously, together with some of the pretreatment patients' characteristics .
Evaluation of Patients
At all study visits (baseline, 14 days, 2, 6, and 12 months post-enrollment), participants underwent a physical exam and were asked an open question about health-related symptoms. Systemic symptoms that newly developed or had worsened since the onset of EM, but had no other known medical explanation were regarded as new or increased symptoms (NOIS). At day 14, patients were asked about medication compliance and adverse events.
Complete response was defined as return to pre-LB health status, while incomplete response was delineated as the presence of NOIS (partial response) and/or occurrence of new objective signs of LB, and/or persistence of B. burgdorferi sensu lato in skin as detected by culture at the site of the previous EM, and/or persistence of EM at 2 months or beyond post-treatment (failure). Persistence of EM means that EM could still be noticed on day light and at room temperature.
At baseline serum IgM and IgG antibodies to C6 peptide derived from the VlsE protein of B. burgdorferi sensu lato were determined using the C6 Lyme ELISA kit (Immunetics).
At baseline, paired, 3 mm, punch biopsies were obtained from the expanding edge of EM. One of the skin specimens was placed directly in a tube containing 6 ml BSK-B medium for culture, the second was placed in 70% ethanol for quantitative PCR analysis. Skin biopsy was repeated 2–3 months after the first biopsy in patients in whom the first skin biopsy specimen was culture positive for B. burgdorferi sensu lato. Cultures were examined weekly for the presence of spirochetes by dark-field microscopy and were interpreted as negative if no growth was established up until 12 weeks . Blood cultures for the presence of Borrelia were performed as previously described .
DNA was extracted from skin biopsy specimens using a QIAamp DNA Mini kit (Qiagen). A quantitative real-time TaqMan PCR assay amplifying and targeting a conserved 139bp fragment of the gene encoding the Borrelia 16S rRNA was used to detect borrelial nucleic acids in total DNA extracted from skin specimens. Real-time PCR assays yielding a Cq value of less than 40 were scored as positive. The number of targets per biopsy reflects the total spirochetal load per biopsy sample, however to normalize the data accounting for differences in biopsy sizes and differences in human skin at the site biopsied, the spirochete load was also expressed per 10,000 human genome equivalents . Isolates were typed according to their OspA sequences as outlined previously .
Data were summarized as medians with interquartile range (IQR) for numerical variables and as frequencies and percentages for categorical variables. The association between the test results (PCR, culture or number of Borrelia 16S rRNA targets) with the other variables was assessed using the chi-squared test with Yates' continuity correction, Mann-Whitney test or Spearman’s correlation; to control for false positives, the P values were adjusted using a multivariate permutation procedure .
To assess the association of the test results with incomplete response to treatment, we estimated three separate logistic regression models with incomplete response as the dependent variable, and skin culture positivity/PCR positivity/borrelial burden as a covariate. To take into account the multiple measurements repeated in each patient, the analyses were adjusted for a subject variable as a random effect, and for follow-up time (used as a categorical variable).
The shape of the association between EM size and the probability of positive test results was estimated using a logistic regression model, modelling the EM size as a restricted cubic spline; a similar (linear regression) model was used to model the number of Borrelia 16S rRNA targets and results were displayed graphically.
All the statistical analyses were performed using the R statistical language .
One hundred twenty one patients with solitary EM, 58.7% of whom were male, with a median age 54 (IQR 43–61) years, were enrolled in the study. Sixty one patients received cefuroxime axetil and 60 were treated with amoxicillin.
Paired skin samples were obtained from all the patients. Three of 121 cultures became contaminated and could not be analyzed further; of the 118 analyzable biopsies 65 (55.1%) were culture-positive. Of 62 identified Borrelia species, 60 (96.8%) were B. afzelii and 2 (3.2%) were B. garinii. Of the 121 biopsies, 94 (77.7%) were Borrelia-positive in PCR assay. All culture-positive biopsies were also positive by PCR. In three patients (2.5%), Borrelia was cultured from blood; skin biopsies from these patients were also positive by culture and PCR.
As reported previously  the number of Borrelia 16S rRNA targets detected by quantitative PCR for PCR positive patients varied from 12 to 17,458 target copies (median 905.5, IQR 188 to 2,415) and from 0.2 to 321 target copies per 10,000 human genome equivalents (median 15, IQR 5 to 39.75). In the present study the bacterial load was expressed in relation to human genomic DNA to compensate for variations arising from the DNA extraction procedure and/or the size and quality of the biopsy sample.
The number of Borrelia 16S rRNA targets detected in culture positive specimens (median 22, IQR 6 to 53) was significantly higher than in culture negative specimens (median 6, IQR 2 to 31.5; P<0.001).
In 81 of 118 (68.6%) patients discernible levels of antibody to C6 (C6 ELISA) were detected at presentation Serum samples from 81 of 118 (68.6%) patients tested positive for Borrelia antibodies (C6 ELISA) at enrollment (Table 1). Patients with positive serological results more often had positive skin culture and positive PCR results and had a higher borrelial burden than serologically negative patients. The association between serology and PCR positivity (P = 0.01) and serology and borrelial burden (P = 0.01) were statistically significant.
Pre-treatment characteristics according to culture positivity, PCR positivity and Borrelia burden
Patients characteristics at enrollment with regard to skin culture positivity and PCR results are outlined in Table 1. The EM lesions of patients with positive culture or PCR results exhibited central clearing more often and were typically larger in size; these types of EM also exhibited larger Borrelia burden. The association between these EM characteristics and the test results was stronger when the PCR, rather than the culture result, was used (Table 1). Interestingly, the estimated shapes of associations between EM size and culture positivity, PCR positivity, or Borrelia burden were similar (Fig 1). The estimated probability of having a positive culture or PCR result increased sharply with the size of EM for patients with EM smaller than 10 cm and remained stable for EM skin lesions ≥ 10 cm (Fig 1). Similarly, the estimated average Borrelia burden increased with the size of EM for patients with EM skin lesions ≤ 15 cm (Fig 1). The proportion of patients reporting NOIS at enrollment was higher in culture or PCR-positive vs. culture or PCR-negative patients, although the differences were not statistically significant (Table 1), however patients with NOIS at enrollment had a statistically significantly higher Borrelia burden (Fig 2). Among NOIS, fatigue was the most commonly reported systemic symptom, followed by headache, myalgia, and arthralgia. (Table 1).
All patients were compliant with taking the study drug assigned. No adverse event resulted in discontinuation of treatment.
In 61 of the 65 culture-positive patients re-biopsy was performed. All of the re-biopsies were culture-negative. Sixty re-biopsies were also PCR-negative while one re-biopsy was PCR-positive. The patient with PCR-positive re-biopsy did not have systemic symptoms at presentation. His skin lesion resolved in three days after starting antibiotics and he remained asymptomatic up till the 12-month visit. The median duration of EM skin lesion after starting antibiotic treatment did not differ significantly with regard to Borrelia burden. Failure was documented in two patients, both of whom were PCR and culture-positive, and both received amoxicillin. One of these patients developed meningitis by the 14-day visit and was given ceftriaxone and recovered uneventfully. The other patient was retreated with azithromycin because of faint residua of EM skin lesion persisting at the 6-month follow-up visit, which could still be detected at the 12-month visit, but no biopsy was performed for histopathology to assess whether the lesion represented active inflammatory disease. It is questionable whether this case truly represented a treatment failure because pale residua at the location of previous EM skin lesion may have represented post-inflammatory hyperpigmentation.
The large majority of patients (71% of PCR-positive patients and 81.5% of PCR-negative patients) had complete response from 2 months onward returning to pre-LB health status.
Patients in whom EM biopsies tested positive by PCR reported more NOIS at baseline and incomplete responses at 2, 6 and 12 months after enrollment (Fig 3). The logistic regression model for repeated measurements of incomplete response, with PCR positivity as explanatory variable, adjusted for time from enrollment (Table 2, Model B), indicated that the probability of incomplete response significantly decreased with time from enrollment (P = 0.05), most markedly between baseline and the first follow-up time point (OR for incomplete response at 14-day visit versus baseline equal to 0.69, 95% CI from 0.35 to 1.35, P = 0.27) and between second and sixth month (OR = 0.59, 95% CI: 0.27 to 1.32, P = 0.20). PCR-positive patients had a higher probability of incomplete response (OR = 1.83, 95% 0.84 to 3.95, P = 0.13), but the association with incomplete response was not statistically significant.
A similar logistic regression model was used with Borrelia burden as an explanatory variable, which indicated that patients with higher Borrelia burden had a higher probability for incomplete response (OR = 1.12, 95% CI: 1.00 to 1.26, P = 0.05, Table 2 Model C). Culture positivity was not associated with incomplete response (OR = 1.08, 95% CI: 0.58 to 2.03, P = 0.81, Table 2 Model A).
The frequency of NOIS decreased with time after treatment, however patients treated with amoxicillin reported more NOIS at the 12-month visit (8/53, 15%, 95% CI: 6.7 to 27.6%) compared to patients treated with cefuroxime (3/52, 5.8%, 95% CI: 1.2 to 15.9%) (Fig 3); the difference was not statistically significant, but its clinical meaningfulness could not be excluded (estimated difference in percentage of patients with NOIS: 9.3%, 95% CI: -4 to 22.8%). Furthermore, when the two treatment failures which occurred in the group of patients treated with amoxicillin were taken into account, the difference in the probability of incomplete response between the two treatment regimens was even more pronounced.
Antibiotic treatment of EM is very successful in terms of faster resolution of the skin lesion and accompanying symptoms, and more importantly it is highly effective in preventing development of other objective manifestations of LB . However, a considerable proportion of patients (up to 40%) complain of subjective symptoms ≥6 months after EM treatment. It is not yet clear whether these subjective symptoms are causally associated with antecedent borrelial infection and if so, what the potential underlying mechanisms are. Several therapeutic trials from USA and Europe have independently shown that in patients who reported post-treatment symptoms a greater frequency of multiple EM or accompanying symptoms such as fatigue, headache, arthralgias or myalgias were reported before treatment [14–20]. However, in other studies the severity of the initial illness did not influence treatment outcome [4, 21–23]. Seropositivity to Borrelia, either at presentation or at follow up, has uniformly been found not to be a risk factor for unfavorable treatment outcome [4, 16, 17, 20–24]. Strle et al recently reported positive correlation between high TH17-associated immune responses in EM patients and the presence of post-Lyme symptoms, suggesting that patient-inherent parameters may influence treatment outcome . In our previous study we evaluated EM skin culture result as a potential influential factor for disease course and treatment outcome. No significant association was ascertained between Borrelia skin culture results and selected pre-treatment characteristics, however a higher proportion of patients with incomplete response to treatment in the culture-positive group compared to the culture-negative group indicated less favorable treatment outcome in culture-positive patients .
The present study revealed that Borrelia load in culture-positive specimens was significantly greater than that detected in culture-negative specimens. This is consistent with findings of Liveris et al, who reported that the average number of spirochetes in culture positive specimens was more than double than in culture negative specimens , and by Li et al who found that recovery of B. burgdorferi in culture correlated with larger numbers of spirochetes in EM lesions . Published reports also indicate that Borrelia burden may contribute to the variability of EM clinical presentation [6–8]. Liveris et al found that larger numbers of spirochetes in EM skin lesions were significantly associated with a shorter duration of EM prior to diagnosis, with smaller skin lesions, and infection with a specific genotype of B. burgdorferi but not with the number or severity of pretreatment constitutional symptoms . Similarly, Li et al reported a higher density of organisms in smaller EM lesions (5–11 cm) during the first days (1–4 days) of illness . These results are consistent with ours, except that in our study peak Borrelia density was observed somewhat later  and in larger skin lesions than in the aforementioned studies. This “lag” is probably due to the difference of EM evolution with regard to the infecting Borrelia species; B. afzelii being the predominating agent in our study vs. B. burgdorferi in the studies of Liveris and Li.
However, as opposed to Liveris , in this study we found statistically significant association between Borrelia burden and the presence of pre-treatment constitutional symptoms accompanying EM.
Not much is known about the potential influence of Borrelia skin burden on treatment outcome in EM patients and the present study is the first to directly test the association between the amount of Borrelia present in skin lesions of EM patients and the respective treatment outcome.
Out of 121 patients there were two treatment failures (1.7%). In the first case, amoxicillin failure was evidenced by objective findings i.e., progression to meningitis documented by cerebrospinal fluid pleocytosis. The second case with persisting EM, was considered a failure due to our stringent criteria. Both treatment failures occurred among PCR-positive patients and both had uneventful course after being re-treated with antibiotics. Since treatment failure was a rare event in this and in our previous studies [4, 23], patients reporting NOIS represented the great majority in the group with incomplete response. The probability of having NOIS was significantly associated with Borrelia burden at the baseline visit, as well overall during follow-up, suggesting that bacterial load in skin could be at least one of the influential parameters for treatment outcome in EM patients.
Patients did not know their PCR results, but they knew their skin culture results. This might have biased our results, which is a drawback of our study. However the steady decline in proportion of symptomatic patients in the PCR-positive as well as in the PCR-negative group contradicts this assumption (Fig 3).
In conclusion, the higher Borrelia burden in skin biopsy specimens was associated with a higher chance for constitutional symptoms accompanying EM, but also negatively influenced the post-treatment outcome. Patients with higher Borrelia burden were more likely to have incomplete response, represented predominantly by post-Lyme symptoms, implicating that Borrelia burden might be involved in the pathogenesis of post-Lyme symptoms. If valid, our assumptions can only be considered relevant for European LB, caused by B. afzelli.
Conceived and designed the experiments: FS DS MO AT IL LL. Performed the experiments: MO AT IL DS VM PB DV. Analyzed the data: MO AT IL LL DS FS. Contributed reagents/materials/analysis tools: MO AT IL FS DS. Wrote the paper: DS FS LL MO.
- 1. Stanek G, Strle F. LB. Lancet. 2003; 362: 1639–1647. pmid:14630446
- 2. Steere AC, Bartenhagen NH, Craft JE, Hutchinson GJ, Newman JH, Rahn DW, et al. The early clinical manifestations of Lyme disease. Ann Intern Med. 1983; 99: 76–82. pmid:6859726
- 3. Steere AC. Lyme disease. N Engl J Med. 2001; 345: 115–125. pmid:11450660
- 4. Cerar D, Cerar T, Ružić-Sabljić E, Wormser GP, Strle F. Controlled prospective study on the development of subjective symptoms after treatment of early Lyme disease in Europe. Am J Med. 2010; 123: 79–86.
- 5. Stupica D, Lusa L, Cerar T, Ružić-Sabljić E, Strle F. Comparison of post-Lyme Borreliosis symptoms in erythema migrans patients with positive and negative Borrelia burgdorferi sensu lato skin culture. Vector Borne Zoonotic Dis. 2011; 11: 883–889. pmid:21083376
- 6. Liveris D, Wang G, Girao G, Byrne DW, Nowakowski J, McKenna D, et al. Quantitative detection of Borrelia burgdorferi in 2-millimeter skin samples of EM lesions: Correlation of results with clinical and laboratory findings. J Clin Microbiol. 2002; 40: 1249–1253. pmid:11923340
- 7. Li X, McHugh AG, Damle N, Sikand KV, Glickstein L, Steere CA. Burden and Viability of Borrelia burgdorferi in Skin and Joints of Patients with EM or Lyme Arthritis. Arth and Rheum. 2011; 63: 2238–2247.
- 8. O'Rourke M, Traweger A, Lusa L, Stupica D, Maraspin V, Barrett PN, et al. Quantitative detection of Borrelia burgdorferi sensu lato in erythema migrans skin lesions using internally controlled duplex real time PCR. PLoS One. 2013; 8: e63968. pmid:23696863
- 9. Stanek G, Fingerle V, Hunfeld KP, Jaulhac B, Kaiser R, Krause A, et al. Lyme borreliosis clinical case definitions for diagnosis and management in Europe. Clin Microbiol Infect. 2011; 17: 69–79. pmid:20132258
- 10. Ruzić-Sabljić E, Arnez M, Lotric-Furlan S, Maraspin V, Cimperman J, Strle F. Genotypic and phenotypic characterisation of Borrelia burgdorferi sensu lato strains isolated from human blood. J Med Microbiol. 2001; 50: 896–901. pmid:11599739
- 11. Livey I, O’Rourke M, Traweger A, Savidis-Dacho H, Crowe BA, Barrett PN, et al. A new approach to a Lyme disease vaccine. Clin Infect Dis. 2011; 52 Suppl 3: 266–270.
- 12. Westfall PH, Young SS. Resampling-Based Multiple Testing, NewYork: Wiley. 1993.
- 13. R Development Core Team. R: A language and environment for statistical computing, Vienna: R Foundation for Statistical Computing. 2009. Available: http://www.R-project.org. Accessed 2015 March 2.
- 14. Dattwyler RJ, Volkman DJ, Conaty SM, Platkin SP, Luft BJ. Amoxicillin plus probenecid versus doxycycline for treatment of EM borreliosis. Lancet. 1990; 336: 1404–1406. pmid:1978873
- 15. Luger SW, Paparone P, Wormser GP, Nadelman RB, Grunwaldt E, Gomez G, et al. Comparison of cefuroxime axetil and doxycycline in treatment of patients with early Lyme disease associated with EM. Antimicrob Agents Chemother. 1995; 39: 661–667. pmid:7793869
- 16. Massarotti EM, Luger SW, Rahn DW, Messner RP, Wong JB, Johnson RC, et al. Treatment of early Lyme disease. Am J Med. 1992; 92: 396–403. pmid:1313637
- 17. Nowakowski J, Nadelman RB, Sell R, McKenna D, Cavaliere LF, Holmgren D, et al. Long-term follow-up of patients with culture-confirmed Lyme disease. Am J Med. 2003; 115: 91–96. pmid:12893393
- 18. Steere AC, Hutchinson GJ, Rahn DW, Sigal LH, Craft JE, DeSanna ET, et al. Treatment of the early manifestations of Lyme disease. Ann Intern Med. 1983; 99: 22–26. pmid:6407378
- 19. Weber K, Neubert U. Clinical features of early EM disease and related disorders. Zbl Bakt Hyg A. 1986; 263: 209–228.
- 20. Weber K, Preac-Mursic V, Wilske B, Thurmayr R, Scherwitz C. A randomized trial of ceftriaxone versus oral penicillin for the treatment of early European LB. Infection. 1990; 18: 91–96. pmid:2185158
- 21. Luft BJ, Dattwyler RJ, Johnson RC, Luger SW, Bosler EM, Rahn DW, et al. Azithromycin compared with amoxicillin in the treatment of EM. Ann Intern Med. 1996; 124: 785–791. pmid:8610947
- 22. Weber K, Wilske B, Preac-Mursic V, Thurmayr R. Azithromycin versus penicillin V for the treatment of early LB. Infection. 1993; 21: 367–372. pmid:8132365
- 23. Stupica D, Lusa L, Ruzić-Sabljić E, Cerar T, Strle F. Treatment of erythema migrans with doxycycline for 10 days versus 15 days. Clin Infect Dis. 2012; 55: 343–350. pmid:22523260
- 24. Wormser GP, Ramanathan R, Nowakowski J, McKenna D, Holmgren D, Visintainer P, et al. Duration of antibiotic therapy for early Lyme disease. Ann Intern Med. 2003; 138: 697–704. pmid:12729423
- 25. Strle K, Stupica D, Drouin EE, Steere AC, Strle F. Elevated levels of IL-23 in a subset of patients with post-lyme disease symptoms following erythema migrans. Clin Infect Dis. 2014; 58: 372–380. pmid:24218102