The Challenge of Identifying a Septic Joint

Authors: Raymond Beyda, MD; Lee Grodin, MD; Jackie Shibata, MD; Ted Segarra MD
Editors: Kelly Maurelus, MD FAAEM and Michael Bond, MD FAAEM 

Originally Published: Common Sense May/June 2017

The ED evaluation of patients with acute monoarthritis is often challenging given the broad differential diagnosis and significant morbidity which can result from missed septic arthritis (SA). The following articles aim to simplify the work-up of the undifferentiated hot and swollen joint.

Couderc M, Pereira B, et al. Predictive Value of the Usual Clinical Signs and Laboratory Tests in the Diagnosis of Septic Arthritis. CJEM. 2015. 17(4):403-410.

This study was a prospective observational cohort study of 105 adult patients. Patients with acute monoarthritis were evaluated with white blood count (WBC), erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), uric acid level, joint radiography and some with an ultrasound. Arthrocentesis and analysis of synovial fluid was performed with appearance of the aspirate documented as either clear, turbid, purulent, or hemorrhagic. The history, physical examination, and data obtained from the above studies were synthesized to estimate the probability of SA. Positive cultures from blood or synovial fluid confirmed the diagnosis of SA.

Of 105 patients, 29 (27.6%) had SA and 9 (8.6%) were considered to possibly have SA. The most common diagnosis of remaining 67 patients was crystal-associated arthritis. Multiple rheumatologic joint diseases were also found. Patients with SA were slightly younger (58.6 vs 60.5 years old, p=0.03) and their symptoms lasted longer (2.6 vs 1.8 weeks, p=0.04). Conversely, the non-SA patients presented more acutely, i.e., <24 hours (72.3% vs 46%, p=0.008). Patients with SA reported more chills (39.5% vs 17.9%, p=0.01) and localized erythema (52.6% vs 28.4%, p=0.01) as well as report of potential port of entry through the skin (71.1% vs 46.3%, p=0.01). Finally, a history of crystal-associated arthritis was more common among the non-SA group (28.4% vs 5.3%, p=0.004). ESR and CRP elevations were present more commonly in the SA group (mean ESR 76.1 vs 45.7 [mm], p=0.002 and mean CRP 135 vs 95.1 [mg/L], p=0.015). Suggestive radiographic findings were more frequent in SA patients (29.7% vs 5.1%, p=0.001). Purulent synovial fluid was more common among SA patients (74% vs 26%, p=0.001) and the synovial fluid WBC was higher (54,900 vs 15,000 [cells/microL], p<0.001) in the SA group.

Importantly, no finding was sufficiently sensitive to exclude SA. Positive Gram stain had 100% specificity but only 40% sensitivity. This finding had the highest positive likelihood ratio (LR) approaching infinity, with a disappointing negative LR of only 0.6. Radiographic findings were similarly highly specific (95%) but not sensitive (30%) with a positive LR of 5.8. In this study, radiographic findings suggestive of SA were defined as decreased joint space, demineralization of bone, bony erosions, and joint destruction. Grossly purulent synovial aspirates were somewhat diagnostically useful compared to other collected data with a positive LR of 4.7. Likelihood of SA increased as the synovial fluid WBC count increased with a positive LR of 3.1 for a WBC >50,000 and LR of 3.9 for WBC >100,000 (cells/microL).

One model derived in this study from multivariate analysis used chills (OR=4.7, 95% CI 1.3-17.1), a history of crystal-induced arthritis (OR=0.09, 95% CI 0.01-0.9), purulent synovial fluid (OR=8.4, 95% CI 2.4-28.5), and radiographic findings suggestive of SA (OR=7.1, 95% CI 13-37.9), with a corresponding area under the curve (AUC) of 0.84. A second model used chills, a history of crystal-induced arthritis, synovial WBC greater than 50,000 cells/mm3 (OR=6.8, 95% CI 1.3-36), a port of entry for infection (OR=3, 95% CI 0.9-10.2), and SA risk factors (OR=3.1, 95% CI 0.87-11) including diabetes, cancer, immunosuppression, alcoholism, chronic kidney disease, and chronic steroid therapy, with a corresponding AUC of 0.87.

While no single finding can sufficiently exclude SA, features from the history, physical, radiography, and synovial fluid analysis should all be considered when considering the diagnosis of SA.

Borzio R, Muchandani N, Pivec, R, et al. Predictors of Septic Arthritis in the Adult Population. Orthopedics. 2016. 39(4): 657-663.

This study identified 458 patients with possible SA. Patients with prosthetic or post-operative infections, incomplete clinical or imaging data, atypical pathogens, and femoral osteomyelitis were excluded (n=188). No patients were given antibiotics prior to joint aspiration. Objective laboratory data including ESR, hematocrit, WBC with differentiation, synovial fluid analysis, synovial gram stain and culture was compared.

Using the receiver operating characteristic (ROC) curves, the authors determined the positive and negative likelihood ratios for reported fever, documented fever, weight-bearing status, synovial fluid WBC count, and segmental leukocyte differential. The gold standard for diagnosis of SA was positive synovial fluid culture.

In this study, 22 patients had SA. Of these, 15 (68%) had positive gram stains. Patients with SA were younger than those with other causes. The two groups showed no statistical difference in regards to reported or documented fever, weight-bearing status, ESR, serum WBC count, serum neutrophil percentage, or mean synovial fluid WBC count.

However, the most helpful finding was the mean synovial WBC count. In those with SA, it was 70,581 cells/microL and in those without SA, it was 26,758 cells/microL. The cell count differential of synovial fluid in the two groups had no statistically significant differences (neutrophil, lymphocyte, monocyte, RBC). A synovial WBC count of 64,000 cells/microL was examined using the ROC curve to reveal the AUC. This produced a specificity of 90% and a sensitivity of 40% for SA. The LR for synovial fluid WBC of 40,000 cells/microL was 1.88 (95% CI, 0.90-3.93), for 50,000 cells/microL was 2.02 (95% CI, 0.94-5.11), and for 65,000 cells/microL was 2.8 (95% I 1.2-6.7).

The authors of this study suggest that a value above this cutoff value [64,000 cells/microL] should be interpreted as true SA and irrigation and debridement should not be delayed for additional studies.

Margaretten ME, Kohlwes J, Moore D, Bent S. Does This Adult Patient Have Septic Arthritis? JAMA. 2007;297(13):1478-88.

Margaretten, et al., performed a systematic review of studies from 1966-2007 containing original data from patients presenting with acutely swollen or painful joints presenting to a rheumatology clinic, ED, or while hospitalized. Articles were excluded if they did not include a gold standard. The gold standard for diagnosis of SA has traditionally been positive cultures from synovial fluid; however, the authors realized this is a suboptimal gold standard citing that sensitivities of such cultures range from 75-95%. As such, the authors expanded their gold standard to include any of the following: positive Gram stain, positive blood culture, culture-negative aspirate that was described as “macroscopic pus,” or response to antibiotics.

Investigators used LRs to distinguish SA from other causes of acutely swollen or painful joints. Only two of the studies reviewed examined risk factors for developing SA. Age older than 80, diabetes mellitus, rheumatoid arthritis, existing prosthesis and skin infection all had positive likelihood ratios (+LR) 2.5-3.5. However recent joint surgery had a +LR of 6.9 and skin infection in combination with joint prosthesis had a +LR of 15. HIV infection was found to have a +LR of 1.7.

Pain in the joint had a sensitivity of 85%, a history of swelling had a sensitivity of 78%, while sweats and rigors were poorly sensitive (27% and 19% respectively). Fever was evaluated in seven studies and had an overall sensitivity of 57%. Interestingly, one study calculated the LR of fever and found a value of 0.67, signifying that SA was less likely when the patient was febrile.

The authors found that elevated serum WBC count (>10,000/microL), ESR (>30mm/h) and CRP (>100 mg/L) all had limited diagnostic power, citing +LRs of less than two for each abnormal value. A synovial fluid WBC count less than 25,000/microL had a +LR of 0.32; >25,000/microL had a LR of 2.9; >50,000/microL had a LR of 7.7; lastly a synovial WBC count >100,000/microL had an LR of 28. The authors also included four studies looking at the differential cell count of synovial WBCs and found that differentials which had >90% polymorphonuclear (PMN) cells had a +LR of 3.4. Of note, one study measured synovial lactate dehydrogenase (LDH) and found that while a level >250 U/L was only 50% specific, this cutoff was 100% sensitive for ruling out SA.

There were several limitations to this review. First, only two of the studies included were prospective in design allowing for many biases inherent with retrospective data collection. Secondly, some studies included very few patients. For example, LDH was found to be 100% sensitive, but only 8 patients in that study were diagnosed with SA. Also, there exists a large variation in the results of the studies examined. For instance, the authors conclude that synovial WBC counts >50,000/microL had a respectable +LR of 7.7; however the actual five LRs reported from each study range from as low as 2.2 up to 19.0. This wide range makes it difficult to rely on this cutoff for clinical answers. Lastly, studies conducted outside the ED were included so results may not be directly applicable to patients presenting only to the ED.

This study still demonstrated some important findings. Specifically, recently post-operative patients or those with a previous prosthetic joint with overlying cellulitis have high odds of SA. History, physical exam, and serum blood tests, while helpful, cannot alone rule in or rule out SA. Synovial fluid analysis is vitally important when evaluating for SA. It may be prudent to empirically treat patients with high synovial WBC counts or with a high percentage of PMN cells on the differential count prior to availability of culture results. Synovial LDH may be very helpful in single-handedly ruling out the disease, but like all the data presented here, stronger, prospective, ED studies need to be performed before it can be relied upon for this highly morbid diagnosis.

Carpenter CR, Schuur JD, Everett WW, and Pines JM. Evidence-based Diagnostics: Adult Septic Arthritis. Acad Emerg Med. 2011. 18(8):781-796.

In this systematic review of adult SA, Carpenter and colleagues reviewed data from 32 diagnostic trials in order to identify characteristics of history, physical exam, and bedside laboratory tests which may correctly identify non-gonococcal (NGC) SA in patients presenting with a complaint of monoarticular arthritis.

Trials including patients with the presenting complaint of monoarticular arthritis were included as long as they contained sufficient detail to reconstruct partial or complete 2×2 contingency tables for the determination of diagnostic test characteristics using an acceptable criterion standard for SA. Studies were excluded if they did not include ED patients or focused primarily on gonococcal arthritis, pediatric patients, or therapy. Case reports and narrative reviews were also excluded.

The authors determined the prevalence of NGC SA in patients with monoarticular complaints to be 27%. Analysis revealed that very few aspects of history, physical exam, and serum testing had any significant effect on post-test probability. However, most of the analysis was limited by incomplete data in the original studies. The historical components found to be significant predictors of SA were prior joint surgery (+LR 6.9) and presence of a skin infection overlying a prosthetic hip/knee (+LR 15.0). Although joint pain was fairly sensitive (85-100%), none of the studies included adequate data for calculating specificity, and so LRs could not be calculated. From physical exam, most studies only described joint tenderness (sensitivity 100% in one study) and fever, but none had complete information for calculation of LRs. In addition, the cutoffs for fever differed greatly among studies, and most studies did not describe timing of the fever.

Lack of data also prohibited many conclusions concerning serum markers. Only two studies calculated the sensitivity and specificity of leukocytosis, but no studies found an acceptable sensitivity of overall diagnostic accuracy for diagnosing SA. ESR and CRP were assessed but the sensitivities varied widely among all studies. Procalcitonin, tumor necrosis factor, and multiple cytokines were also assessed but they showed only modest specificity, and very poor sensitivity. Blood culture sensitivity ranged from 23-36%, but no specificity data was available.

For synovial fluid analysis, the authors determined that a synovial WBC (sWBC) count >50,000/microL had a summary +LR of 4.7 and a -LR of 0.52. Further, for a sWBC count >100,000/microL, the summary +LR is 13.2 and the -LR is 0.86. However, there was significant heterogeneity (I^2 >25%) between the trials when comparing the sensitivities and specificities for these ranges, and this heterogeneity could only be reduced by excluding certain data sets. LR intervals were also calculated, but the data was calculated using only the four trials that reported sWBC counts with an adequate degree of information to allow for the calculations to be completed. They identified the following sWBC interval LRs: 1) 0-25,000/microL with +LR 0.33; 2) 25,000-50,000/microL with +LR 1.06; 3) 50,000-100,000/microL with +LR 3.59; and 4) >100,000/microL with +LR 13.2. One trial determined that a synovial lactate dehydrogenase level less than 250 units/L may be adequate to exclude SA. In addition, four trials revealed high +LR for synovial lactate levels >10mmol/L (+LR 21, -LR 0.16). However, the remaining markers, including the percentage of synovial polymorphonuclear cells, synovial glucose, synovial protein, and presence of positive Gram stain did not significantly affect the likelihood of having SA.

There are several major limitations of this review. The majority of the studies were retrospective, so the patients that were studied and the tests they received were possibly affected by verification bias, thus falsely increasing the sensitivity and decreasing the specificity within the study. Similarly, these studies were limited by spectrum bias, as the majority of the patients screened were skewed toward higher-severity illness, leading to false elevation of sensitivity. The authors conclude that history, physical, and serum testing do not significantly alter post-test probability. However, the majority of their data points did not include adequate information to calculate complete sensitivity and specificity and the studies had high degrees of bias and heterogeneity, thus making it unclear whether the lack of significant findings reflected a type II error.

Despite these significant limitations, the authors note the importance of determining a clinical decision rule for SA that might be able to incorporate some of the aspects of history and synovial fluid analysis that they identified as having higher LRs. This study does a good job of demonstrating a statistically sound method for determining when to continue testing (test threshold) and when to begin treatment (treatment threshold). Nevertheless, its actual calculations are significantly limited by the quality of the studies that it incorporated, and further prospective studies will be needed in order to ascertain the true sensitivities and specificities of all these markers.

Acute monoarthritis is a common chief complaint encountered in EDs and SA is a do-not-miss diagnosis. However, its exclusion is very challenging, with few data points possessing sufficient sensitivity to rule it out. Arthrocentesis is essential when the diagnosis is suspected. While focusing on single variables from the history, physical examination, blood work, radiography, and synovial analysis is limited in terms of making the diagnosis, incorporating several of these features in the right clinical context can significantly increase the probability of disease.

In particular, a history of prosthetic joints, overlying skin infection, grossly purulent synovial fluid, elevated sWBC, positive Gram stain, and synovial lactate >10mmol/L can increase the probability of disease. A logistical factor in the inclusion of synovial lactate in the work up is that oftentimes, the synovial fluid can be too thick to allow for accurate measurement by the analyzer, so the lab can add hyaluronidase to the sample to digest the proteins to liquefy it, making for easier analysis, and without affecting lactate levels. Given the significant +LRs in these studies, eliciting these features in the history and examination, ensuring adequate examination of the synovial fluid by obtaining a gram stain, sWBC, and synovial lactate, should be essential components in the work up of a suspected septic joint.