Predictive Factors for Differentiating Between Septic Arthritis and Lyme Disease of the Knee in Children

Predictive Factors for Differentiating Between Septic Arthritis and Lyme Disease of the Knee in Children
                                       C OPYRIGHT  2016      BY   T HE J OURNAL   OF   B ONE   AND J OINT   S URGERY, I NCORPORATED

           A commentary by Elan J. Golan, MD, and
           Jeffrey D. Thomson, MD, is linked to the
           online version of this article at

           Predictive Factors for Differentiating Between
              Septic Arthritis and Lyme Disease of the
                         Knee in Children
Keith D. Baldwin, MD, MSPT, MPH, Christopher M. Brusalis, BA, Afamefuna M. Nduaguba, MD, and Wudbhav N. Sankar, MD

                              Investigation performed at The Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania

     Background: Differentiating between septic arthritis and Lyme disease of the knee in endemic areas can be challenging
     and has major implications for patient management. The purpose of this study was to identify a prediction rule to dif-
     ferentiate septic arthritis from Lyme disease in children presenting with knee pain and effusion.
     Methods: We retrospectively reviewed the records of patients younger than 18 years of age with knee effusions who underwent
     arthrocentesis at our institution from 2005 to 2013. Patients with either septic arthritis (positive joint fluid culture or synovial white
     blood-cell count of >60,000 white blood cells/mm3 with negative Lyme titer) or Lyme disease (positive Lyme immunoglobulin G on
     Western blot analysis) were included. To avoid misclassification bias, undiagnosed knee effusions and joints with both a positive
     culture and positive Lyme titers were excluded. Historical, clinical, and laboratory data were compared between groups to identify
     variables for comparison. Binary logistic regression analysis was used to identify independent predictive variables.
     Results: One hundred and eighty-nine patients were studied: 23 with culture-positive septic arthritis, 26 with culture-
     negative septic arthritis, and 140 with Lyme disease. Multivariate binary logistic regression identified pain with short arc
     motion, history of fever reported by the patient or a family member, C-reactive protein of >4 mg/L, and age younger than
     2 years as independent predictive factors for septic arthritis. A simpler model was developed that showed that the risk of
     septic arthritis with none of these factors was 2%, with 1 of these factors was 18%, with 2 of these factors was 45%, with
     3 of these factors was 84%, or with all 4 of these factors was 100%.
     Conclusions: Although septic arthritis of the knee and Lyme monoarthritis share common features that can make them
     difficult to distinguish clinically, the presence of pain with short arc motion, C-reactive protein of >4.0 mg/L, patient-
     reported history of fever, and age younger than 2 years were independent predictive factors of septic arthritis in pediatric
     patients. The more factors that are present, the higher the risk of having septic arthritis.
     Level of Evidence: Diagnostic Level III. See Instructions for Authors for a complete description of levels of evidence.

Peer Review: This article was reviewed by the Editor-in-Chief and one Deputy Editor, and it underwent blinded review by two or more outside experts. The Deputy Editor
reviewed each revision of the article, and it underwent a final review by the Editor-in-Chief prior to publication. Final corrections and clarifications occurred during one or
more exchanges between the author(s) and copyeditors.

        cute bacterial septic arthritis is most often a surgical                           present with fever, joint effusion, irritability, and inability to
        emergency. However, differentiating septic arthritis of the                        bear weight on the affected extremity. In addition, laboratory
        knee from Lyme disease in children with symptoms of                                measurements, including erythrocyte sedimentation rate (ESR),
knee pain and swelling can be very challenging. Clinical pre-                              C-reactive protein (CRP), and peripheral and synovial white
sentations of both conditions can be quite similar. Often patients                         blood-cell (WBC) counts, may be elevated in both conditions1-4.

Disclosure: One author of this study (K.D.B.) received funds from JBJS for manuscript preparation. On the Disclosure of Potential Conflicts of Interest
forms, which are provided with the online version of the article, one or more of the authors checked “yes” to indicate that the author had a relevant financial
relationship in the biomedical arena outside the submitted work.

J Bone Joint Surg Am. 2016;98:721-8     d
                TH E JO U R NA L O F B O N E & JO I N T SU RG E RY J B J S . O RG
                                                                                        P R E D I C T I O N A L G O R I T H M F O R S E P T I C A R T H R I T I S C O M PA R E D   WITH
                             V O LU M E 98 -A N U M B E R 9 M AY 4, 2 016
                                                d              d
                                                                                        LY M E D I S E A S E O F T H E K N E E I N C H I L D R E N

   TABLE I Patient Data

                                        Lyme Disease                     Culture-Negative Septic Arthritis                      Culture-Positive Septic Arthritis

           No. of cases*                      140                                        26                                                         23
           Mean age† (yr)                 8.03 ± 4.0                                 2.1 ± 1.7                                                 5.6 ± 1.6
             Male                               93                                       13                                                         12
             Female                             47                                       13                                                         11

   *The values are given as the number of patients. †The values are given as the mean and the standard deviation.

       In spite of the overlap in both clinical presentation and                        take several days to be finalized. As a result, frontline clinicians
laboratory testing, it is imperative that the treating physician                        are often faced with a challenging scenario: a child with knee
makes an accurate diagnosis because each condition has very                             pain and effusion and indeterminate serum and synovial lab-
different implications for treatment. Because pyogenic septic                           oratory values. In these circumstances, one must weigh the risk
arthritis can result in rapid articular cartilage destruction,                          of delayed or untreated septic arthritis against the potential for
emergent irrigation and drainage in the operating room are                              an unnecessary surgical procedure or serial aspirations, which
warranted. In contrast, Borrelia burgdorferi, the causative organ-                      are less than ideal in a children’s hospital where anxiety at
ism in Lyme disease, stimulates an immune complex-mediated                              multiple needle sticks is undesirable.
inflammatory reaction that does not endanger the articular car-                                 In the absence of a quick and accurate test to defini-
tilage to the same degree, but does cause systemic immune-                              tively diagnose Lyme disease, prediction algorithms that
driven medical complications5. Standard treatment generally                             combine multiple clinical and laboratory factors have been
involves an appropriate course of antibiotics and observation                           developed to identify patients at high risk for pyogenic ar-
for other sequelae of systemic Lyme disease.                                            thritis and to help to guide treatment decisions 6-9. Although
       In most pediatric medical centers, confirmatory testing                           these models have proven helpful in differentiating septic
for Lyme disease involves serum Western blot analysis for Lyme                          arthritis from transient synovitis of the hip, their utility in
immunoglobulin G (IgG), but the results of this testing may                             differentiating septic arthritis from Lyme disease of the knee

   TABLE II Univariate Analysis of Culture-Positive Septic Arthritis Compared with Lyme Arthritis for Clinical and Laboratory Findings

                                                         Culture-Positive Septic Arthritis* (N = 23)                    Lyme Disease* (N = 140)                       P Value

       Duration of symptoms (d)                                              3.2 ± 2.1                                             6.1 ± 9.3                           0.268
       History of antibiotic use                                               21.7%                                                   6.4%                            0.019
       History of fever                                                        73.9%                                                 40.7%                             0.001
       Radiographic effusion                                                   78.3%                                                 95%                               0.999
       Micromotion tenderness                                                  86.4%                                                   6.3%
                TH E JO U R NA L O F B O N E & JO I N T SU RG E RY J B J S . O RG
                                                                                        P R E D I C T I O N A L G O R I T H M F O R S E P T I C A R T H R I T I S C O M PA R E D   WITH
                             V O LU M E 98 -A N U M B E R 9 M AY 4, 2 016
                                                d              d
                                                                                        LY M E D I S E A S E O F T H E K N E E I N C H I L D R E N

   TABLE III Univariate Analysis of Culture-Negative Septic Arthritis Compared with Lyme Arthritis for Clinical and Laboratory Findings

                                                         Culture-Negative Septic Arthritis* (N = 26)                     Lyme Disease* (N = 140)                      P Value

      Duration of symptoms (d)                                               4.5 ± 3.2                                              6.1 ± 9.3                           0.625
      History of antibiotic use                                               23.1%                                                     6.4%                            0.030
      History of fever                                                        76.9%                                                   40.7%                             0.001
      Radiographic effusion                                                   96.2%                                                   95%                               0.999
      Micromotion tenderness                                                  58.8%                                                     6.3%
                   TH E JO U R NA L O F B O N E & JO I N T SU RG E RY J B J S . O RG
                                                                                       P R E D I C T I O N A L G O R I T H M F O R S E P T I C A R T H R I T I S C O M PA R E D   WITH
                                V O LU M E 98 -A N U M B E R 9 M AY 4, 2 016
                                                   d              d
                                                                                       LY M E D I S E A S E O F T H E K N E E I N C H I L D R E N

Materials and Methods                                                                  logical testing, blood and joint fluid cultures, Gram stains of the joint fluid,
                                                                                       ESR, and CRP.
W      e conducted a retrospective study of individuals younger than 18 years of
       age who presented to the emergency department of a large tertiary care
children’s hospital between 2005 and 2013 and in whom an arthrocentesis was            Statistical Analysis
performed for knee joint effusion. The study was approved by our institutional         Data were analyzed using SPSS version 22.0 (IBM). Descriptive elements of the
board review prior to data collection. Study patients were identified through a         data are presented as means (and ranges) and categorical variables are presented
query of the laboratory database of our electronic medical record system for all       as counts (and percentages). Cutoff values for continuous variables, CRP of
patients in the emergency department for whom joint or synovial fluid analysis          >4 mg/L and age younger than 2 years, were determined with receiver operating
was ordered. Identified patients were then cross-referenced with the labo-              characteristic curve analysis. Mann-Whitney U tests were used to evaluate
ratory results of Gram-staining, microbial culture, and synovial WBC count to          differences between groups in continuous or ordinal variables in cases where
identify patients with septic arthritis and were cross-referenced with the             the data were not normally distributed and to identify variables for multi-
results of the Lyme Western blot analysis to identify patients diagnosed with          variate analysis. A chi-square test was used to determine significance in the
Lyme disease.                                                                          case of binary or categorical data. Level of significance was established at a two-
        The emergency department records of each identified patient were                sided alpha level of p < 0.05. Variables that significantly influenced the diag-
reviewed to confirm the diagnosis of either pyogenic septic arthritis or Lyme           nosis of septic arthritis were included in a binary logistic regression with a
arthritis and to extract clinical and laboratory data. To create the cleanest          stepwise backward elimination method (with a cutoff for elimination of 0.1) to
cohort to minimize misclassification bias, we defined septic arthritis as either         determine factors predictive of septic arthritis. Simpler dichotomized models
the presence of a positive joint fluid culture irrespective of synovial cell            were generated using the risk factors and were confirmed using 22 log like-
count (culture-positive septic arthritis) or synovial WBC count of >60,000             lihood methodology. Models were generated for Lyme disease compared with
white blood cells/mm3 in a joint with a negative Lyme titer (culture-negative          culture-positive septic arthritis and for Lyme disease compared with all cases
septic arthritis). Lyme disease was defined as the presence of a positive serum         of septic arthritis.
Lyme IgG on Western blot based on our reference laboratory definitions.
Patients who did not meet the criteria for a diagnosis of septic arthritis and         Patient Demographic Characteristics
who did not have a Lyme titer sent within the same emergency department                During the study period, a total of 841 children were identified as having had an
evaluation were excluded from the study. To avoid misclassification bias,               arthrocentesis performed in the emergency department, of which 498 were
undiagnosed knee effusions and joints with both a positive culture and positive        knee aspirations. After cross-referencing identified patients with the results
Lyme titers were excluded. In patients with multiple peripheral blood                  of the microbial culture and Lyme disease immunoblot assays and applying
analyses and joint aspirations during the same hospitalization, we limited our         further exclusion criteria (Fig. 1), we identified 189 patients as our study
data analysis to the first peripheral laboratory values and to the initial aspirate     cohort (49 with septic arthritis and 140 with Lyme disease) (Table I).
sample.                                                                                        Of the 23 patients with culture-positive septic arthritis, organisms in-
        For each patient, we recorded data on demographic characteristics,             cluded methicillin-susceptible Staphylococcus aureus (MSSA) (11 patients),
medical comorbidities, historical elements, physical examination findings, and          methicillin-resistant Staphylococcus aureus (MRSA) (4 patients), Propionibacterium
results of laboratory testing, including complete blood-cell count, Lyme sero-         acnes (2 patients), Clostridium bifermentans (1 patient), Streptococcus pneumoniae

Fig. 1
Consolidated Standards of Reporting Trials (CONSORT) diagram of the study population. SA = septic arthritis, CNSA = culture-negative septic arthritis,
Dx = diagnosis, WB1 = positive white blood-cell count, and Hx = medical history.
                   TH E JO U R NA L O F B O N E & JO I N T SU RG E RY J B J S . O RG
                                                                                          P R E D I C T I O N A L G O R I T H M F O R S E P T I C A R T H R I T I S C O M PA R E D   WITH
                                V O LU M E 98 -A N U M B E R 9 M AY 4, 2 016
                                                    d             d
                                                                                          LY M E D I S E A S E O F T H E K N E E I N C H I L D R E N

    TABLE V Univariate Analysis of Operatively Treated Lyme Arthritis Compared with Nonoperatively Treated Lyme Arthritis

                                                                  Operative Group* (N = 46)                   Nonoperative Group* (N = 94)                           P-Value

           Temperature (F)                                                  99.3 ± 1.3                                    99.1 ± 1.2                                  0.08
           Serum WBC count       (·109   cells/L)                            11.2 ± 3.6                                     9.6 ± 2.8                                  0.010
           Absolute neutrophil count (cells/mm3)                            5,727 ± 2,122                                5,726 ± 2,121                                 0.061
           ESR (mm/hr)                                                        52 ± 21                                        37 ± 21
                TH E JO U R NA L O F B O N E & JO I N T SU RG E RY J B J S . O RG
                                                                                    P R E D I C T I O N A L G O R I T H M F O R S E P T I C A R T H R I T I S C O M PA R E D   WITH
                             V O LU M E 98 -A N U M B E R 9 M AY 4, 2 016
                                                d              d
                                                                                    LY M E D I S E A S E O F T H E K N E E I N C H I L D R E N

count alone may lead to a high rate of unnecessary surgical                                 Three previous studies have specifically sought to identify
procedures. Furthermore, the results of Lyme-specific serologi-                      predictive factors for differentiating septic arthritis from Lyme
cal testing can take several days and therefore are not helpful for                 disease. Thompson et al.11 evaluated 179 patients with mono-
making an urgent surgical decision. As a result, Lyme disease is                    arthritis of different joints (not just the knee) and identified
often mistaken for culture-negative septic arthritis and is                         history of fever and elevated CRP to be negative predictors of
treated as such with incision and drainage, unnecessarily ex-                       Lyme arthritis and knee involvement to be a positive predictor
posing these patients to the risks of a surgical procedure4.                        (model sensitivity of 88% and specificity of 82%). In spite of
       In the absence of an easily administered test with high                      this, the authors concluded that there was too much overlap in
sensitivity and specificity for definitively diagnosing septic ar-                    the data to create a clinically useful predictive model. Milewski
thritis, previous studies have used multiple physical examina-                      et al.15 compared 123 patients with Lyme disease with 51 pa-
tion and laboratory factors to create prediction models that                        tients with culture-positive septic arthritis, although several
help to identify patients at high risk for pyogenic arthritis.                      different joints were included in their analysis. Based on mul-
Kocher et al.8 developed a widely used set of criteria to differ-                   tivariate analysis, the authors found that refusal to bear weight
entiate septic arthritis from transient synovitis of the hip using                  was the most predictive factor, with weaker predictors being
4 independent clinical variables: history of fever, non-weight-                     presence of a fever, WBC count of >12 · 1,000/mL, and nu-
bearing status, ESR of ‡40 mm/hr, and serum WBC count of                            cleated cell count of >100,000 cells/mm3. Interestingly, CRP
‡12 · 109 cells/L. This model has since been validated internally                   and ESR were not found to be useful predictors, although the
and externally, and updated to reflect the additional diagnostic                     study was limited by a very low CRP collection rate during the
benefit of CRP6. Although the Kocher model is often applied to                       study period. A recent study by Deanehan et al. is the only
multiple joints even in the setting of Lyme disease, this pre-                      previous one, to our knowledge, that compared predictive
dictor was not originally intended to distinguish between                           factors for Lyme disease with those for septic arthritis specifi-
septic arthritis and Lyme disease and was not validated for the                     cally of the knee10. Unlike the aforementioned studies (and
knee. The septic arthritis cohort studied by Kocher et al. con-                     ours), the authors found ESR to be an independent predictor of
sisted of all patients with synovial leukocyte counts in excess of                  septic arthritis in Lyme disease-endemic areas; the other factor
50,000 cells/mm3 regardless of bacterial culture results and                        that they found to be predictive was an absolute neutrophil
therefore may have inadvertently included patients with Lyme                        count of ‡10,000 cells/mm3, which we also did not detect in our
disease and other misclassification biases. Indeed, a recent study                   study. However, their data were limited by a low CRP collection
by Deanehan et al. demonstrated synovial WBC counts in excess                       rate (77%) and a small number of patients with septic arthritis
of 50,000 cells/mm3 to be quite common in Lyme arthritis16,                         (19). In addition, the Lyme disease cohort included a significant
and these findings were reiterated by this current study.                            proportion of patients (58%) who did not undergo arthrocentesis.

Fig. 2
Map of reported cases of Lyme disease in the United States, 2014. (Reproduced from: CDC, National Center for Health Statistics. Lyme Disease. 2015
Oct 2. Accessed 2016 Jan 26.)
                  TH E JO U R NA L O F B O N E & JO I N T SU RG E RY J B J S . O RG
                                                                                      P R E D I C T I O N A L G O R I T H M F O R S E P T I C A R T H R I T I S C O M PA R E D   WITH
                               V O LU M E 98 -A N U M B E R 9 M AY 4, 2 016
                                                  d              d
                                                                                      LY M E D I S E A S E O F T H E K N E E I N C H I L D R E N

These latter types of patients can typically be triaged to ob-                               To provide a clinically useful prediction model, we de-
servation by the treating physician and are rarely, by clinical                       fined septic arthritis as either the presence of a positive joint
examination, suspected of having septic arthritis. These patients                     culture or as a synovial WBC count of >60,000 white blood
may therefore not represent the types of clinical presentations                       cells/mm3 with negative Lyme serological results. Although there
that are easily confused with those having true pyogenic arthritis.                   is a small risk of misclassification bias using this second def-
        In this present study aimed at differentiating septic ar-                     inition, ignoring the large number of patients who present to
thritis of the knee from Lyme monoarthritis, we found several                         emergency departments with culture-negative septic arthritis
independent predictive factors for pyogenic arthritis: patient or                     would limit the clinical applicability of any prediction rule.
family-reported history of fever, pain with short arc motion,                                This study had several limitations. First, as a chart review,
CRP of ‡4 mg/L, and age younger than 2 years. The probability                         it had the potential loss of data fidelity associated with retro-
of septic arthritis with any one factor present was 18% com-                          spective analysis. Second, as a single-center study in a Lyme
pared with 100% with all 4 factors present. Logistic regression                       disease-endemic area, the results of this study may not be ap-
of predictor variables showed that 87% of cases were correctly                        plicable to all geographical regions, particularly those with a low
predicted with our model. In our series, a patient with no risk                       prevalence of Lyme disease. The Centers for Disease Control and
factors had a 2% chance of being diagnosed with septic ar-                            Prevention (CDC) has reported that, in 2014, 96% of confirmed
thritis, and a patient with 4 factors had a 100% chance of septic                     cases of Lyme disease occurred in 14 U.S. states, concentrated in
arthritis. Notably, of 46 patients with Lyme disease who un-                          the Northeast and upper Midwest (Fig. 2)17. Therefore, the find-
derwent a surgical procedure, 11 patients had no clinical pre-                        ings of this study may be most applicable to these geographic
dictors. Thus, applying this clinical prediction rule may have                        regions. The strengths of the study include rigorous definitions
avoided a surgical procedure in 23.9% of patients in this                             of the cohorts and data acquisition and statistical analysis methods
cohort. Of the remaining 35 patients with Lyme disease who                            that limit misclassification and observer bias. Additionally, to our
underwent a surgical procedure, 19 had one clinical predictor,                        knowledge, our study is the first to identify predictive risk factors
14 had 2 predictors, 2 had 3 predictors, and no patient had all                       that, in our cohort, ensured the diagnosis of septic arthritis when
4 predictors. It is important to note that the results of this study                  all were present (i.e., the diagnosis was septic arthritis in 100%
may not be applicable in a non-endemic area with a lower                              of cases) and resulted in only one case of septic arthritis when
prevalence of Lyme disease.                                                           all 4 risk factors were absent (2%).
        Interestingly, we found pain with short arc motion to be                             In conclusion, our study offers a useful prediction algo-
the most clinically useful test for septic arthritis, with a sensi-                   rithm for septic arthritis compared with Lyme disease of the
tivity of 0.78 and a specificity of 0.92. Multivariate analysis                        knee. We believe that a patient older than 2 years of age without
showed that the adjusted odds of having septic arthritis with                         history of fever, no short arc pain, and CRP of
                    TH E JO U R NA L O F B O N E & JO I N T SU RG E RY J B J S . O RG
                                                                                           P R E D I C T I O N A L G O R I T H M F O R S E P T I C A R T H R I T I S C O M PA R E D   WITH
                                 V O LU M E 98 -A N U M B E R 9 M AY 4, 2 016
                                                      d               d
                                                                                           LY M E D I S E A S E O F T H E K N E E I N C H I L D R E N

7. Jung ST, Rowe SM, Moon ES, Song EK, Yoon TR, Seo HY. Significance of labo-               12. Bennett OM, Namnyak SS. Acute septic arthritis of the hip joint in infancy and
ratory and radiologic findings for differentiating between septic arthritis and transient   childhood. Clin Orthop Relat Res. 1992 Aug;(281):123-32.
synovitis of the hip. J Pediatr Orthop. 2003 May-Jun;23(3):368-72.                         13. Forlin E, Milani C. Sequelae of septic arthritis of the hip in children: a new
8. Kocher MS, Zurakowski D, Kasser JR. Differentiating between septic arthritis and        classification and a review of 41 hips. J Pediatr Orthop. 2008 Jul-Aug;28(5):
transient synovitis of the hip in children: an evidence-based clinical prediction al-      524-8.
gorithm. J Bone Joint Surg Am. 1999 Dec;81(12):1662-70.                                    14. Gillespie R. Septic arthritis of childhood. Clin Orthop Relat Res. 1973 Oct;
9. Molteni RA. The differential diagnosis of benign and septic joint disease in chil-      (96):152-9.
dren. Clinical, radiologic, laboratory, and joint fluid analysis, based on 37 children      15. Milewski MD, Cruz AI Jr, Miller CP, Peterson AT, Smith BG. Lyme arthritis in
with septic arthritis and 97 with benign aseptic arthritis. Clin Pediatr (Phila). 1978     children presenting with joint effusions. J Bone Joint Surg Am. 2011 Feb 2;
Jan;17(1):19-23.                                                                           93(3):252-60.
10. Deanehan JK, Kimia AA, Tan Tanny SP, Milewski MD, Talusan PG, Smith BG,                16. Deanehan JK, Nigrovic PA, Milewski MD, Tan Tanny SP, Kimia AA, Smith BG,
Nigrovic LE. Distinguishing Lyme from septic knee monoarthritis in Lyme disease-           Nigrovic LE. Synovial fluid findings in children with knee monoarthritis in Lyme dis-
endemic areas. Pediatrics. 2013 Mar;131(3):e695-701. Epub 2013 Feb 18.                     ease endemic areas. Pediatr Emerg Care. 2014 Jan;30(1):16-9.
11. Thompson A, Mannix R, Bachur R. Acute pediatric monoarticular arthritis: dis-          17. CDC, National Center for Health Statistics. Lyme Disease. 2015 Oct 2. http://
tinguishing Lyme arthritis from other etiologies. Pediatrics. 2009 Mar;123(3):959-65. Accessed 2016 Jan 26.
You can also read
Next slide ... Cancel