OBJECTIVE. Identify clinical predictors of Lyme arthritis among patients with acute monoarticular arthritis.
METHODS. A medical chart review was conducted of children ≤18 years of age with monoarticular arthritis who underwent arthrocentesis in a pediatric emergency department located in the northeast United States. Patients were classified into 3 categories of arthritis: septic, Lyme, or nonseptic non-Lyme arthritis. Historical, clinical, and laboratory data were compared to identify distinguishing features of Lyme arthritis.
RESULTS. One hundred seventy-nine patients were studied: 46 (26%) patients with septic arthritis, 55 (31%) patients with Lyme arthritis, and 78 (43%) patients with nonseptic non-Lyme arthritis. Compared with those with septic arthritis, patients with Lyme disease were more likely to have a tick-bite history, knee involvement, and less likely to have a history of fever or elevated temperature at triage. Erythrocyte sedimentation rate, C-reactive protein, joint white blood cell count, and joint neutrophil percentage were also statistically lower. In comparison to nonseptic non-Lyme arthritis, knee involvement and tick-bite history were predictors of Lyme. Erythrocyte sedimentation rate, joint white blood cell count, and joint neutrophil percentage were also statistically different. Multivariate analysis comparing Lyme to septic arthritis demonstrated fever history and elevated C-reactive protein level to be negative predictors of Lyme arthritis and knee involvement to be a positive predictor (model sensitivity: 88%; specificity: 82%).
CONCLUSIONS. Lyme arthritis shares features with both septic and nonseptic non-Lyme arthritis. This overlap prevents the creation of a clinically useful predictive model for Lyme arthritis. In endemic areas, Lyme testing should be performed on all patients presenting with acute monoarticular arthritis.
According to the Centers for Disease Control and Prevention, reports of Lyme disease caused by the spirochete Borrelia burgdorferi continue to rise with an estimated 64 000 cases occurring during the years 2003–2005.1 Approximately 50% to 60% of patients who are not treated in the early phase of Borrelia infection will develop arthritis.2 As arthritis is a late-stage finding of Lyme disease, serologic tests are reliably positive. Although these tests have been shown to have high sensitivity (97%) and specificity (99%) for Lyme arthritis, results can take hours to days to return.3
Lyme arthritis must be differentiated from other types of acute nontraumatic monoarticular arthritis such as septic arthritis and nonseptic non-Lyme arthritis. This differentiation is important for both treatment and prognosis. Septic arthritis can destroy cartilage with associated long-term morbidity if not treated promptly.4–6 Failure to diagnose and treat Lyme arthritis does not usually result in profound joint destruction but can result in persistent joint symptoms, as well as the risk of developing additional disease manifestations. In this investigation, we sought to identify clinical predictors of Lyme arthritis among children presenting with acute monoarticular arthritis.
Study Design, Setting, and Selection of Participants
We conducted a retrospective cross-sectional study of patients who presented to an academic pediatric emergency department (ED) with 54 000 patient visits per year. The pediatric center is located in a Lyme-endemic area of the northeastern United States. The study included patients ≤18 years of age who presented to the ED between December 2000 and September 2006 and in whom arthrocentesis was performed. Patients were excluded if they met any of the following criteria: preexisting septic arthritis or Lyme disease, preexisting rheumatologic disease, immunocompromise, polyarthritis, or a nonseptic arthritis diagnosis and use of antibiotics. The study was approved by the institutional review board, and data collection was compliant with the Health Insurance Portability and Accountability Act of 1996.
Data Collection and Processing
Patients were identified through an electronic hospital data system by querying the laboratory database for any ED patient who had joint fluid analysis ordered. The electronic medical charts were then reviewed by using a computerized data collection form with embedded data validation rules. The ED physician record was reviewed for historical and physical examination information. Hospital discharge summaries and follow-up visits to outpatient orthopedic, infectious disease, and rheumatology clinics were also reviewed. All records were primarily reviewed by a single author (Dr Thompson). Results of laboratory studies were downloaded from the hospital information systems: complete blood count, Lyme serology, joint fluid analysis, blood and joint fluid cultures, Gram stains of joint fluid, erythrocyte sedimentation rate (ESR), and C-reactive protein (CRP) level. ESR values were considered normal if they were ≤20 mm/hour. CRP values were considered elevated if they were ≥0.5 mg/dL. Lyme serology results were included if sent during the ED encounter or within 5 days of the original encounter. Missing historical data were coded as missing, whereas physician examination findings were marked as absent if not noted in the electronic physician record.
All patients were classified as having (1) septic arthritis, (2) Lyme arthritis, or (3) nonseptic non-Lyme arthritis. For the purposes of this study, septic arthritis was defined as a joint fluid culture yielding a pathogen or joint fluid with ≥40 000 white blood cells (WBC)/μL with a positive blood culture. Bacillus species and Staphylococcus nonaureus were regarded as contaminants. The diagnosis of Lyme arthritis was defined as a patient with an enzyme-linked immunoabsorbant assay Lyme index value of >1.2 with a positive Lyme IgG on Western blot. Western blot results were positive or negative based on the reference laboratory definitions. Serologic testing for B burgdorferi was performed by off-site commercial laboratories (ARUP [Salt Lake City, UT] and Immugen [Norwood, MA]). Nonseptic non-Lyme arthritis was defined as patients without septic arthritis who had a negative Lyme serology result. Patients who did not meet the definition of septic arthritis but who did not have a Lyme titer sent in the hospital system within 5 days of the index encounter were excluded (including those who had had Lyme serology sent from a referring physician's office or outside hospital but whose results were not available to reviewers). In addition, patients who did not meet the definition of septic arthritis or Lyme arthritis and who were on concurrent antibiotic therapy were excluded from analysis. All final classifications were reviewed by the primary and senior author for agreement.
Data were analyzed by using SPSS 13.0 (SPSS Inc, Chicago, IL) and Stata 9 (Stata Corp, College Station, TX). Distributions were described as mean values and SDs (normal distributions) and median and interquartile ranges for nonnormal distributions. χ2 analysis was used to compare categorical data. Analysis of variance and independent samples t tests were used to compare mean values of continuous variables. The Mann-Whitney U test was used to compare nonparametric continuous data. A P value of ≤.05 was considered statistically significant. Comparisons were made between Lyme and septic arthritis and Lyme and nonseptic non-Lyme arthritis. A priori, the authors determined that the most clinically important comparison was between Lyme and septic arthritis; for this comparison, logistic regression was used to create a multivariate model to predict Lyme arthritis. The number of variables included in the initial multivariate model was determined by the 10:1 rule based on the total number of cases being analyzed. The initial multivariate model was refined by using backward stepwise elimination. Variables in the initial model with P > .05 were removed sequentially, after which the remaining variables were inspected to ensure that confounding bias was not reintroduced. A variable was considered confounded if the β coefficient changed by >20%. Variables in the final model were inspected for colinearity defined as a >20% change in a variable's standard error.
A total of 435 children were identified as having had arthrocentesis performed in the ED during the study period. After application of exclusion criteria, 179 (41%) patients were studied. Exclusions included patients >18 years of age (n = 15), current treatment for preexisting septic or Lyme disease (n = 4), polyarthritis (n = 7), known rheumatologic or immunocompromised states (n = 12), and insufficient or clotted joint fluid (n = 23). One hundred ninety-five patients were also excluded if they did not meet the definition of septic arthritis and either lacked Lyme serology results (n = 171; 100 of whom presented with hip arthritis) or had received antibiotics before ED presentation (n = 24). All patients had peripheral WBC counts and absolute neutrophil count (ANC) tests performed. ESR and CRP levels were collected in 94% and 72% of patients, respectively.
We identified 46 (26%) patients with septic arthritis, 55 (31%) patients with Lyme arthritis, and 78 (43%) patients with nonseptic non-Lyme arthritis (Table 1). The joints involved included ankle (3%), elbow (6%), hip (36%), knee (52%), and shoulder (3%). Our study population was 60% male with a median age of 7.3 years (interquartile range [IQR]: 4.7–10.9 years). Patients had a median duration of symptoms of 3 days (IQR: 1–6 days). Sixty-eight percent of patients were admitted to the hospital from the ED and 46% underwent operative joint irrigation.
Clinical Parameters for Patient Subgroups
Of the 46 patients with septic arthritis, 45 had positive joint fluid culture results and 1 patient had bacteremia with an elevated joint WBC count of 43 120 cells per μL. Thirteen (28%) patients had associated bacteremia. Organisms isolated on culture included Staphylococcus aureus (31), group A Streptococcus (4), Streptococcus pneumoniae (2), Neisseria gonorrhea (1), Neisseria meningitidis (1), Enterococcus species (1), Salmonella species (1), group B Streptococcus (1), Haemophilus influenzae (1), Haemophilus parainfluenzae (1), and Actinomyces species (1). There were 3 patients who had positive joint fluid Gram stains but negative blood and joint cultures; 2 of these patients were classified as Lyme arthritis based on positive serologies and 1 patient was classified as nonseptic non-Lyme arthritis (Table 2).
Nonseptic Non-Lyme Arthritis
Of the 78 patients with nonseptic non-Lyme arthritis, final diagnoses included juvenile rheumatoid arthritis (2), toxic synovitis or post infectious arthritis (68), adjacent osteomyelitis (4), dermatomyositis (1), infectious bursitis (1), psoriatic arthritis (1), and Henoch-Schonlein purpura (1) (Table 2).
The majority of those with Lyme disease (84%) did not recall tick exposure. Likely secondary to its staging as a late manifestation of Lyme disease, there was not a difference in the seasonal presentation of Lyme arthritis with 51% of patients presenting outside the classic Lyme season of June through October (Table 2).
Comparisons of Arthritis Subgroups
The clinical characteristics of patients with Lyme arthritis compared with septic arthritis are shown in Table 3. When comparing patients with Lyme disease to those with septic arthritis, patients were more likely to have knee involvement (P < .001) and a history of tick bite (P = .02). Negative clinical predictors for Lyme included recent fever (P < .001) and temperature of ≥38.0°C at triage (P < .01). Patients with Lyme arthritis had lower ESR levels (P < .01), CRP levels (P < .001), joint WBC counts (P = .03), and percent joint neutrophils (P < .001) (Fig 1).
Table 4 compares the clinical characteristics of patients with Lyme arthritis to those with nonseptic non-Lyme arthritis. Knee involvement (P < .001) and tick-bite history (P = .03) were associated with Lyme arthritis. Lyme disease was associated with a higher ESR level (P < .001), joint WBC count (P < .001), and percent joint neutrophils (P < .001).
Multivariate analysis comparing Lyme arthritis to septic arthritis was performed on the basis of the methods described earlier. Five variables were determined to be the most important predictors of having Lyme arthritis: history of fever, CRP, knee involvement, tick-bite history, and joint WBC count. ESR and ANC were not included in the model given their colinearity with CRP and WBC count, respectively. The final model demonstrated history of fever (odds ratio [OR]: 0.22 [95% confidence interval (CI): 0.051–0.91]), knee involvement (OR: 12 [95% CI: 2.8–47]), and CRP (OR: 0.79 [95% CI: 0.68–0.93]) to be the most clinically significant factors, with elevated CRP and history of fever to be negative predictors of Lyme arthritis and knee involvement to be a positive predictor of Lyme arthritis. This model carried a Hosmer-Lemeshow value of 0.78, a sensitivity of 88%, and a specificity of 82% (Table 5).
Lyme disease has emerged as a common but clinically elusive infectious disease. Only first recognized in the 1970s, Lyme disease cases continue to increase each year. The Centers for Disease Control and Prevention reported over 60 000 new cases during the years 2003 through 20051; although the true incidence of Lyme is suspected to be much greater given underreporting. Important for those caring for children, 61% of reported cases involve patients aged 5 to14 years.1
The arthritis of late-stage Lyme disease usually presents as oligoarthritis of the large joints, classically the knee.2,7–13 Fever is variably present with arthritis,2,14 and the majority of patients do not recall a preceding tick bite.7,10,13,14 The infected joint is swollen, and patients report limited range of motion with varying degrees of associated pain.7,11,13 Complaints of joint erythema and warmth may be present but are less common.7 Laboratory evidence of inflammation is found in the majority of patients and may include elevated ESR levels, CRP levels, and peripheral WBC counts.2,7,11,13,14 The leukocyte count of the synovial fluid can have a wide range with counts >100 000 cells per μL and as low as 1000 cells per μL being reported.7,12,15 In addition, a polymorphonuclear leukocyte predominance to the synovial WBC count is generally found.2,7,11,13,14
Previous work has shown that patients with Lyme arthritis have high levels of anti-spirochetal protein antibodies.16 This allows the 2-tier systems of enzyme-linked immunoabsorbant assay and Western blot to reliably diagnose those with the disease.17B burgdorferi can also be detected by polymerase chain reaction from the synovial fluid.18 Unfortunately, serology and joint polymerase chain reaction results are not reported during the initial evaluation of the patient, therefore clinicians must use available clinical data to determine if arthrocentesis is needed to exclude septic arthritis and whether to initiate empiric antibiotic therapy pending results of any cultures or Lyme serology.
During the initial evaluation of the patient, it would be beneficial to differentiate the monoarthritis of Lyme disease from other forms of acute arthritis, most importantly septic arthritis. For this, the clinician must currently rely on history and physical examination findings, results of inflammatory markers, and results of joint fluid analysis when available. When septic arthritis is being considered, joint fluid is required for proper diagnosis. For septic arthritis, synovial fluid culture has a reported sensitivity of only 75% to 95%.19 In a recent meta-analysis, Margaretten et al20 reported on clinical and laboratory findings that help identify patients with septic arthritis. Studies that defined septic arthritis by one of the following were included: positive synovial fluid culture, positive Gram-stain, positive blood culture, response to antibiotics or aspiration of macroscopic pus from the joint. In their review, joint pain and swelling were found to be present in 85% and 78% of cases, whereas fever occurred in only 57% of patients. Specificity was not addressed in the majority of studies evaluating signs and symptoms of septic arthritis, and history and physical examination findings were not found to be useful in predicting septic arthritis. Likewise, the likelihood of septic arthritis only minimally increased when patients had an elevated peripheral WBC count or CRP and ESR levels because of the poor specificity of these markers. Synovial WBC count and the percentage of polymorphonuclear cells were the only factors found to significantly change the likelihood of septic arthritis. In another systematic review by Mathews et al,21 the management of septic arthritis was discussed. Without proper treatment, bacterial toxins can cause inflammatory articular changes that lead to subsequent joint destruction.6 Besides antibiotics, many orthopedists also feel that drainage of the purulent material from the joint space is essential for successful treatment as well.22
Of less importance, it would be helpful to distinguish Lyme from nonseptic non-Lyme arthritis to prompt physicians to confirm the diagnosis by serology and consider initiation of therapy. Based on previously published literature,23 patients with nonseptic non-Lyme arthritis also present with fever as well as monoarticular swelling and pain. For these patients, inflammatory markers and the synovial WBC count are variably abnormal. Patients with Lyme arthritis may also have false-positive antinuclear antibody and rheumatoid factor levels at the initial presentation.7,10
Our study is the largest study of acute monoarticular arthritis in children that attempts to identify predictors of Lyme disease. In the current study, patients with Lyme disease demonstrated similar clinical presentations to previous descriptions. Almost half of patients reported recent fever, whereas only 16% recalled a tick bite. Painful swelling of the knee was the presenting complaint in 85% of patients. The hip was the next commonly involved joint. Ten percent of the study patients with hip arthritis were diagnosed with Lyme disease. This presents an interesting perspective to the young patient presenting with suspected toxic synovitis. A large population of patients with Lyme disease was described as having joint warmth (60%) and non–weight-bearing status (38%). These percentages are higher than those reported by many other studies. Given that our study evaluated patients undergoing arthrocentesis, this feature may be overestimated in comparison to the general population of patients with Lyme arthritis. In comparison to septic arthritis, knee involvement, tick-bite history, and a lack of fever were found to be more commonly associated with Lyme arthritis. Knee involvement and tick-bite history were also found more commonly in Lyme arthritis than nonseptic non-Lyme arthritis. Unfortunately, poor specificity makes these features clinically insignificant.
Our patients with Lyme arthritis had elevated peripheral WBC counts, ESR values, and CRP values. These mild elevations are similar to those reported in previous studies. Unfortunately, although the median values of ESR and CRP were significantly different between Lyme and other forms of arthritis, the wide range of values prevents their use as discriminators. Joint WBC count was also found to be a poor discriminator of the etiology of the arthritis. The wide range of synovial white blood cell counts in Lyme disease is well documented, and this work further demonstrated the variability. The median synovial WBC count in Lyme arthritis was significantly greater than nonseptic non-Lyme arthritis and lower than septic arthritis. Again, however, the wide range of values precludes using joint WBC count to distinguish Lyme from septic or nonseptic non-Lyme arthritis.
For clinicians, the early diagnosis of septic arthritis is important for initiating proper treatment. In many settings, preferred treatment involves joint drainage and irrigation that is unnecessary for nonseptic arthritis. Although it was our intention to report discriminators of Lyme arthritis, the overlap for many clinical parameters with septic arthritis and nonseptic non-Lyme arthritis did not permit accurate differentiation. Clearly, clinicians must consider the diagnosis of Lyme disease in endemic areas, and the results of initial laboratory evaluation should not dissuade the clinician from considering Lyme disease. When logistic regression was used to create a model to predict patients with Lyme arthritis from those with septic arthritis, fever and CRP were negative predictors and knee involvement proved to be a positive predictor. Unfortunately, the actual predictive capability of the model was insufficient to be used clinically. With a sensitivity of 88% and a specificity of 82%, this model would properly identify only 41/46 patients with septic arthritis.
Our findings must be interpreted within the limitations of the study. First, this study was performed at a single institution in a Lyme endemic area. Our findings would be less applicable in geographic areas where B burgdorferi is not endemic. Second, our study investigated patients with monoarticular arthritis who underwent joint aspiration. Patients with monoarticular arthritis who were thought to be at low risk for septic arthritis and who therefore did not undergo arthrocentesis were not included in this study. We also excluded patients without Lyme serologies; there is likely a bias relating to testing for Lyme disease. Based on our results, some, if not all, of these patients should have been tested for Lyme disease at presentation including the large number of patients presumed to have toxic synovitis. As with all retrospective studies, many of the clinical features were dependent on interpretation of the medical documentation.
Lyme disease is a common cause of monoarticular arthritis in a Lyme-endemic area. Although testing is accurate and reliable, results are not available during the initial evaluation, and the clinician must make additional diagnostic and treatment decisions. Unfortunately, the clinical similarity between septic, Lyme, and nonseptic non-Lyme arthritis does not allow reliable distinction by clinical or laboratory features. This overlap in clinical presentation emphasizes the need to consider Lyme disease as a diagnosis in children with monoarticular arthritis.
- Accepted August 28, 2008.
- Address correspondence to Amy Thompson, MD, Children's Hospital Boston, Division of Emergency Medicine, 300 Longwood Ave, Boston, MA 02115. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
What's Known on This Subject
Lyme arthritis is a common cause of monoarticular arthritis in endemic areas. Physicians must differentiate Lyme arthritis from other types of acute monoarticular arthritis. This differentiation is important for both treatment and prognosis.
What This Study Adds
To our knowledge, our study is the largest study of acute monoarticular arthritis in children that attempts to identify predictors of Lyme disease.
- ↵Bacon RM, Biggerstaff BJ, Schriefer ME, et al. Serodiagnosis of Lyme disease by kinetic enzyme-linked immunosorbent assay using recombinant VlsE1 or peptide antigens of Borrelia burgdorferi compared with 2-tiered testing using whole-cell lysates. J Infect Dis.2003;187 (8):1187– 1199
- ↵Gerber MA, Zemel LS, Shapiro ED. Lyme arthritis in children: clinical epidemiology and long-term outcomes. Pediatrics.1998;102 (4 pt 1):905– 908
- ↵Saulsbury FT. Lyme arthritis in 20 children residing in a non-endemic area. Clin Pediatr (Phila).2005;44 (5):419– 421
- ↵Dressler F, Whalen JA, Reinhardt BN, Steere AC. Western blotting in the serodiagnosis of Lyme disease. J Infect Dis.1993;167 (2):392– 400
- ↵Aguero-Rosenfeld ME, Wang G, Schwartz I, Wormser GP. Diagnosis of lyme borreliosis. Clin Microbiol Rev.2005;18 (3):484– 509
- ↵Mathews CJ, Kingsley G, Field M, et al. Management of septic arthritis: a systematic review. Ann Rheum Dis.2007;66 (4):440– 445
- ↵Ruddy S, Harris ED, Sledge CB. Kelly's Textbook of Rheumatology. 6th ed. Philadelphia, PA: W. B. Saunders Company; 2001
- Copyright © 2009 by the American Academy of Pediatrics