Assessment and diagnosis of the acute hot joint: a systematic review and meta-analysis

Abstract

Objectives

Prompt diagnosis of septic arthritis (SA) in acute native hot joints is essential for avoiding unnecessary antibiotics and hospital admissions. We evaluated the utility of synovial fluid (SF) and serum tests in differentiating causes of acute hot joints.

Methods

We performed a systematic literature review of diagnostic testing for acute hot joints. Articles were included if studying ≥1 serum or SF test(s) for an acute hot joint, compared with clinical assessment and SF microscopy and culture. English-language articles only were included, without date restriction. The following were recorded for each test, threshold and diagnosis: sensitivity, specificity, positive/negative predictive values and likelihood ratios. For directly comparable tests (i.e. identical fluid, test and threshold), bivariate random-effects meta-analysis was used to pool sensitivity, specificity, and areas under the curves.

Results

A total of 8443 articles were identified, and 49 were ultimately included. Information on 28 distinct markers in SF and serum, differentiating septic from non-septic joints, was extracted. Most had been tested at multiple diagnostic thresholds, yielding a total of 27 serum markers and 156 SF markers. Due to heterogeneity of study design, outcomes and thresholds, meta-analysis was possible for only eight SF tests, all differentiating septic from non-septic joints. Of these, leucocyte esterase had the highest pooled sensitivity [0.94 (0.70, 0.99)] with good pooled specificity [0.74 (0.67, 0.81)].

Conclusion

Our review demonstrates many single tests, individually with diagnostic utility but suboptimal accuracy for exclusion of native joint infection. A combination of several tests with or without a stratification score is required for optimizing rapid assessment of the hot joint.

Rheumatology key messages.

• Rapid exclusion of septic arthritis is required in order to improve patient outcomes and avoid unnecessary admissions and antibiotic use.

• Our review identified many biomarkers with individually good diagnostic utility but suboptimal accuracy for excluding septic arthritis.

• A panel of synovial fluid and/or serum tests may optimize rapid assessment of hot joints.

Introduction

The presentation of an acutely hot swollen native joint is common in clinical practice. It can be due to numerous conditions, but it is important to promptly exclude septic arthritis, as it can rapidly destroy cartilage. Acute hot joints commonly arise due to crystal-induced disease (i.e. gout or pseudogout), OA, trauma, and a variety of systemic diseases. The relative incidence of each condition varies between populations, but an audit of 137 patients at our centre found 38.7% due to crystal arthritis (with almost equal proportions of gout and pseudogout), 19.7% to OA, 19.7% to inflammatory arthritis (including RA and PsA), 8.0% to septic arthritis, and 13.9% to other diagnoses (including traumatic hemarthrosis and osteomyelitis) [1].

All can present with fever, joint swelling, pain and stiffness, mimicking septic arthritis. Crystal arthritis and septic arthritis are particularly difficult to distinguish and may also co-exist. The mortality for in-hospital septic arthritis is 7–15%, despite antibiotic use. The incidence of bacterial arthritis in England is 1 in 49 000/100 000 person-years [2]. Infected joints should be identified and treated in a timely manner.

Early joint aspiration and SF analysis is essential for diagnosis and management of acute hot joints [3]. SF Gram stain, white cell count (WCC), crystal examination, and culture should be performed. However, the British Society for Rheumatology (BSR) guidelines state ‘patients with a short history of a hot, swollen and tender joint (or joints) with restriction of movement should be regarded as having septic arthritis until proven otherwise’ [4]. Patients are frequently admitted to hospital with antibiotic treatment until results become available. Crystal microscopy, to identify uric acid and calcium pyrophosphate crystals, can be done relatively quickly to aid diagnosis of gout and pseudogout, respectively. Ultrasonography is also increasingly available to aid diagnosis of crystal arthritis. However, Gram stain and culture results may not be available for hours, leading to diagnostic delay. Crystal arthritis and septic arthritis can co-exist, especially as an underlying diagnosis of gout increases the risk of septic arthritis [5].

It is of clinical and financial benefit to seek efficient methods of differentiating septic from non-septic joints. Multiple studies suggest the utility of various biochemical markers in differentiating between a septic and non-septic joint. An increasing number of studies have explored the utility of biochemical markers for the rapid exclusion of prosthetic joint infection (PJI), including SF alpha-defensin and calprotectin, and some are now routinely used in clinical practice [6–9]. However, similar tests are lacking for native hot joints [6].

This systematic literature review (SLR) evaluates the use of SF and serum markers in diagnosing an acute native hot joint, compared with the internationally recognized gold standard of clinical assessment and SF analysis (including crystal microscopy and cultures) [3].

Methods

This SLR was conducted in accordance with the Cochrane Handbook and principles for reviews on diagnostic test accuracy, and reported as per Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [10–12].

The protocol was developed by M.D., S.D. and N.G., and registered in the online PROSPERO database of systematic reviews (CRD42018117065) [13]. The search question, framed and structured using the ‘Patients, Intervention, Comparator or Control, Outcome, Type of Study (PICOT)’ format [10], was: What is the utility of testing SF and serum markers in the presentation of an acute hot joint, compared with the current gold-standard practice of clinical assessment combined with SF aspiration, microscopy and cultures? Information on scoping searches is available in the supplementary material, available at Rheumatology online.

The overall aim of this SLR was to identify tests that are able to identify or exclude septic arthritis in native acute hot joints. The secondary aim was to identify tests for identifying or excluding other common causes of an acute hot joint, crystal arthritis, and other inflammatory arthritides.

Participants

A study was included in this review if the participants presented with an acutely swollen hot native joint (i.e. symptoms of under 6 weeks duration) and were undergoing diagnostic tests, in either the SF, the serum or both, to aid diagnosis and management of the acute hot joint.

Interventions

Included studies used one or more serum or SF marker(s) for the diagnosis of an acute hot joint.

Diagnostic tests under study were stratified by the condition intended to be either diagnosed or ruled out by the test. The categories of conditions were: septic arthritis, crystal arthritis, and other inflammatory arthritis. Tests in these categories were then further subclassified by whether the test was for serum or SF.

Comparator or control

Clinical assessment and SF aspiration, with microscopy and culture, was deemed the reference standard [2].

Outcome

The following outcomes from all studies were recorded: sensitivity and specificity for the test(s) under study; positive and negative predictive values (PPVs and NPVs); and likelihood ratios. Receiver operating characteristic (ROC) curves, e.g. the false-positive vs the true positive rate, were also recorded, where reported.

Type of study

Observational studies of patients with a native acute hot joint presentation, undergoing test(s) to establish the diagnosis, were included. In most cases, this involved testing serum and/or SF for the exclusion of joint infection, but studies were also included if the test under study was seeking to diagnose or exclude crystal arthritis or inflammatory arthritis, the two main differential diagnoses for septic arthritis.

Reviews, meta-analyses, comments and editorials were excluded. Case series or case studies comprising 10 or fewer cases were excluded.

The search strategy was developed by M.D., S.D. and N.G., and is available in the supplementary material along with database information and inclusion/exclusion criteria, available at Rheumatology online.

All identified full-length articles were uploaded into EndNote VX9 (Clarivate Analytics, PA, USA). Duplicates were subsequently removed. Titles and abstracts were screened by two reviewers, to assess eligibility. The full articles that met initial inclusion criteria were subsequently examined in detail by two reviewers. Any disagreements between reviewers were resolved through discussion or with involvement from a third reviewer. Full-text screening was not performed for three articles due to inability to access the article despite contacting a university library and the authors.

Assessment of risk of bias, data extraction, and synthesis

Risk of bias for each included study was assessed using the Quality Assessment of Diagnostic Accuracy Studies version-2 (QUADAS-2) tool [14]. Details of QUADAS-2, with the results for each article, are included in the supplementary material, available at Rheumatology online. Data extraction was performed by one reviewer, with 20% repeated by a second reviewer for validation. Disagreements were discussed until a consensus was agreed.

For each selected article, in addition to basic information, the following information was extracted: level of evidence; study design; sample size; inclusion and exclusion criteria; prevalence of septic arthritis, crystal arthritis, and other arthritis; diagnostic marker(s) under study; and gold-standard test (comparator). With regards to diagnostic values, the following were extracted from each article: sensitivity, specificity, PPV, NPV, positive likelihood ratio, negative likelihood ratio, true and false-positive rates, and true and false-negative rates.

Meta-analysis

Bivariate random-effects meta-analyses were used to pool sensitivity, specificity, and areas under the curves (AUCs) for biomarkers that were directly comparable, for tests discriminating septic from non-septic arthritis. A test was eligible for meta-analysis if >1 study used the same marker, threshold and fluid (i.e. serum or SF). Studies testing the same markers at different thresholds and/or in a different fluid were not eligible for meta-analysis if not replicated in a second study. Meta-analyses were conducted in R (version 3.5.3) by fitting a generalized linear mixed model and using the R package lme4.

Results

A total of 8443 articles were identified through our initial search across the three included databases (Fig. 1). Ultimately, 49 articles were included in the review. Five animal studies were also initially identified, but were ultimately excluded due to lack of comparability between the animal studies and the human studies. Complete concordance was achieved between reviewers at all stages. Fig. 1 summarizes the article numbers and the retrieval process. Table 1 displays basic information on each included study, including the diagnoses studied and the biomarkers investigated.

Figure 1.

PRISMA flowchart of included papers

Table 1.

Details of studies included in systematic review

Source	Study quality/Level of evidence	Country	Sample size	Median age (year)	Study design	Inclusion criteria	Exclusion criteria	Prevalence of SA (%)	Prevalence of crystal arthritis (%)	Prevalence of other diagnosis (%)	Diagnostic marker	Gold standard
Aliste-Fernández et al. (2020) [57]	2b	ESP	205	64 (mean)	Prospective cohort	Acute or chronic arthritis with joint effusion	Nil described	7.8	22	70.2	SF WBC	Diagnosis of gout/pseudogout: polarized light microscopy to examine all samples for microcrystals
Baillet et al. (2019) [35]	1b	FRA	74	SA: 66, PG: 81, RA: 63	Prospective cohort	Acute monoarthritis with inflammatory SF (i.e. with WBC count >2000/mm3 and >80% neutrophils) consistent with SA	Nil described	35.14	37.84	27.03	Calprotectin, alpha-defensin	Positive SF/blood culture
Baran et al.(2014) [21]	2b	USA	96	47	Retrospective case–control	Hip or knee symptoms suggestive of SA requiring arthrocentesis	Age <18 years, inmate in the county jail, at the time of aspiration, SF analysis done at outside facility, results listed as contaminants, prosthetic joints, repeat aspirate shows negative culture	45.83	N/A	N/A	Percentage PMNs in synovial WBCs	Positive synovial culture
Berthoud et al. (2020) [45]	2b	FRA	233	61.8 (mean)	Prospective cross-sectional	Age>18 years, acute joint effusion on a native joint, atraumatic, evolving for <30 days	Nil described	10.7	44.6	44.7	SF lactate and glucose	Newman’s criteria and SF analysis when one of the following present: pathogen was isolated from SF; pathogen isolated from blood culture with typical clinical presentation for SA; arthrocentesis revealed purulent SF with clinical presentation for SA, absence of crystals and absence of other suitable diagnoses
Bonilla et al. (2011) [26]	2b	USA	63	N/A	Retrospective cohort	Patients with clinically suspected infection, inflammatory arthritis, or normal joints under evaluation by the rheumatologist	Not described	25.4	0	74.6 (inflammatory and normal)	SF 16S rDNA PCR	SF culture
Borzio et al. (2016) [28]	2b	USA	458	51.7	Retrospective cohort	Patients undergoing arthrocentesis and SA of the knee and shoulder	Incomplete clinical or imaging data, atypical patients, periprosthetic infections, postoperative SA, and associated proximal femoral osteomyelitis	4.8	N/A	N/A	SF WBC	SF culture
Bram et al. (2018) [29]	2b	USA	302	6.0 (4.5)	Retrospective cohort	All patients with suspected SA who underwent arthrocentesis and subsequent surgical irrigation and debridement	Patients who underwent foreign body removal, or with contiguous osteomyelitis and/or pyomyositis	34	N/A	16% had positive Lyme titres	SF Gram stain	SF culture
Carpenter et al. (2020) [46]	2b	USA	71	58 (mean)	Prospective cohort	Acute monoarticular knee symptoms, possible SA, with at least 10 ml SF aspirated.	Failure to obtain consent or use of antibiotics within 72 h	7	N/A	93	SF lactate, SF PCR, SF WBC	SF bacterial growth
Cohen et al. (2019) [36]	2b	ISR	1024	63	Retrospective cohort	All SF specimens that were analysed in the microbiology laboratory between 2002 and 2016 of a single general medical centre	Nil described	62.3	N/A	N/A	Bottled culture broth (Bactec)	Positive culture on agar or a combination of clinical findings highly supporting the diagnosis with a negative culture
Coiffier et al. (2013) [60]	2b	FRA	98	N/A	Prospective cohort	Patients evaluated for joint effusion at a rheumatology department at a single centre over 12 months	Nil described	7.1	30.6	62.3	Leucocyte esterase	Leucocyte count per mm3, with microbiological cultures for 72 h; polarized light microscopy
Coiffier et al. (2019) [37]	1b	FRA	95	Mean 57.7	Retrospective cross-sectional	Adults (≥18 years old) referred for acute monoarthritis or oligoarthritis (progression <6 weeks) on native joint and who received a diagnostic joint fluid puncture	Nil described	35.9	N/A	N/A	16s rDNA PCR	Newman’s criteria, SF direct examination and culture, blood culture
Colvin et al. (2015) [24]	2b	USA	5	N/A	Retrospective cohort	Clinical suspicion of SA	Patients with insufficient fluid or blood-stained SF	20	N/A	80	Leucocyte esterase	Positive SF culture
Couderc et al. (2015) [25]	1b	FRA	105	N/A	Prospective cohort	Patients with suspected SA	Prosthetic joint; trauma; <18 years	36.2	N/A	63.8	Serum WBC > 10 000/mm3; ESR > 15 mm; ESR > 50 mm; ESR > 100 mm; CRP > 15 mg/l; CRP > 100 mg/l; uric acid > 420 mg/l; SF WBC >10 000/μl; SF WBC >10 000/μl; SF WBC > 50 000/μl; SF WBC > 100 000/μl; PMNs > 90%; presence of microcrystals; positive direct Gram stain	SF culture with clinician assessment and diagnosis
Couderc et al. (2019) [39]	1b	FRA	39	Mean 64.5	Prospective cohort	Patients with suspected SA	Nil described	53.85	28.21	12.82	Serum and SF MMP MMP-2, MMP-9, tissue inhibitor of MMP (TIMP-1), cartilage oligomeric matrix protein (COMP), C-terminal telopeptide of type II collagen (CTX-II), and calprotectin (CALP)	Microorganisms from SF or blood cultures
Cunningham et al. (2014) [19]	3b	DEU	273	N/A	Prospective cohort	Adult patients hospitalized for suspicion of SA (prosthetic joints included; therefore, values for native joints calculated by reviewer)	Nil described	73	N/A	27	Gram stain	Positive SF culture
Curtis et al. (1983) [30]	2b	UK	238	Not described	Prospective cohort	Specimens that arrived in the laboratory for routine culture; further 38 fluids from patients with non-septic conditions	Specimens without refrigeration for more than 6 h	7.98	4.20	Viral arthritis 2.10; acute osteomyelitis 1.68%; osteogenic sarcoma 1.26%; osteoarthrosis 26.05%; recent trauma 15.97%; coagulopathies 1.26%; RA 28.99%; Reiter’s syndrome 3.36%; CTDs 6.30%	SF lactate	SF culture
Ferreyra et al. (2016) [31]	2b	FRA	208	59.6 ± 18.5	Retrospective cohort	Patients older than 18 years enrolled in either of two cohorts: SPECTROSYNO cohort of patients with acute or chronic monoarthritis, oligoarthritis, or polyarthritis investigated by joint aspiration; and DNAr16S cohort of patients who underwent joint aspiration for onset of monoarthritis or oligoarthritis within the last 6 weeks	Incomplete joint fluid data, cytological results expressed semi-quantitatively, SA due to non-pyogenic organisms, SA without microbiological documentation	13.46	Chondrocalcinosis 19.71%; gout 13.46%	RA 15.87%; SpA 14.90%; OA 8.65%; undifferentiated arthritis 13.94%	Absolute leucocyte count, absolute neutrophil count, differential neutrophil count, monosodium urate, CPPD	SF/blood culture for SA, clinical findings, blood and joint fluid test results, and the radiographic appearance for the other diagnosis
Foocharoen et al. (2011) [32]	2b	THA	40	52.3 ± 17.4	Retrospective cohort	Patients over 15 years of age with clinically suspected tuberculosis arthritis or having an unknown aetiology of their arthritis who were candidates for arthrocentesis	Not described	24.2% (all SA); 16.7% (TB arthritis); 7.5% non-TB bacterial SA	0	RA 20%; miscellaneous 57.5%	SF adenosine deaminase	SF culture for Mycobacterium tuberculosis
Garg and Goyal (2018) [61]	4 (“Use of a non-independent reference standard (where the ‘test’ is included in the ‘reference’, or where the ‘testing’ affects the ‘reference’) implies a level 4 study.”)	IND	100	Not described	Prospective, observational and cross-sectional	All the patients of any age with one or more joint effusions were included in this study	Uncontrolled diabetes mellitus, cutaneous soft tissue infections mimicking acute arthritis	10% (6% tuberculous arthritis; 4% SA)	3	OA 22%, RA 16%, trauma 3%	WBC	Gross examination, WBC, viscosity, % PMNs, Gram stain, culture, red blood cells presence, crystals presence
Gautam et al. (2017) [33]	2b	IND	27	22.32 ± 19.20	Prospective cohort	Any age; acute monoarticular disease of major joint; clinical symptoms: acute onset, fever, limping while walking, unable to bear weight on the affected extremity, severe pain even on gentle passive movement, pseudoparalysis in children	Poor skin condition; presence of sinus; presence of blood in aspirate; known case of haemophilia or any other bleeding disorder; patients who give definite history of antibiotic intake for the same condition; proven case of any other joint pathology	77.78	N/A	N/A	SF leucocyte esterase; Gram stain; CRP; ESR	SF culture
Gbejuade et al. (2019) [47]	2b	GBR	830	N/A	Retrospective cohort	Suspected SA with SF samples taken	Results show possible contaminants; patient details duplicated; samples sent for possible tuberculosis; culture and Gram stain reports unavailable	12	N/A	88	SF Gram stain	SF culture
Gratacos et al. (1995) [34]	3b	ESP	17 patients (20 samples) with proven SA	N/A	Retrospective case–control	Unclear—collected over 3-year period, non-consecutive	Nil described	16.81	21.84	61.34	Synovial D-lactic acid, % PMNs, WBC, gram stains	Included patients already had diagnosis by synovial culture
Hassas et al. Yeganeh (2020) [48]	3b	IRN	68	N/A	Retrospective case–control	Suspected JIA or SA	Nil described	50	N/A	50	SF leucocyte esterase	JIA: ILAR criteria; SF culture and SF WBC >50 000/ml
Jeng et al. (1997) [18]	2b	TWN	75	N/A	Prospective cohort	Patients with suspected bacterial arthritis	Nil described	27	27	46	SF TNFα	Positive SF bacterial culture
Kim et al. (2010) [17]	3b	KOR	80	N/A	Prospective cohort	Patients with suspected SA	Nil described	34	N/A	66	Multiplex PCR	Positive SF bacterial culture and clinician assessment
Kinugasa et al. 2019 [40]	1b	JPN	30	Mean 4.5	Prospective cohort	Clinical features of SA; laboratory data suggesting inflammatory disease (≥38.0°C, WBC count > 12 000 cells/mm3, or a serum CRP level of 2.0 mg/dl); increased SF on US or MRI. Paediatric cohort	Nil described	56.67	N/A	43.33	SF glucose; serum CRP; SF WCC	SF culture positive
Kunnamo (1986) [54]	3b	FIN	129	N/A	Mixed retrospective and prospective cohort	Patients undergoing joint aspiration for suspected SA or for therapeutic aspiration for intra-articular steroid	Nil described	10	N/A	90	SF WBC	Positive SF bacterial culture
Lenski and Scherer (2014) (1) [52]	2b	DEU	82	N/A	Retrospective cohort	All patients with culture-verified SA and gout arthritis, presenting during study period	Nil described	53	29	18	Serum WBC; serum CRP; serum uric acid; SF lactate; SF glucose; SF uric acid; SF LDH; SF WBC; SF total protein; SF IL6	SF culture, SF crystal microscopy
Lenski and Scherer (2014) (2) [53]	4	DEU	119	N/A	Retrospective case–control	All patients requiring arthrocentesis for suspected native SA, based on 3/5 of the following: pain, redness, swelling, heat, impaired ROM	Peri-prosthetic infections	53	N/A	47	SF IL6; SF total protein; SF glucose; SF lactate; SF WBC	Positive bacterial culture
Li et al. (2007) [51]	2b	USA	156 (13% paediatric)	53	Retrospective cohort	Adults and children undergoing arthrocentesis during study period	Dry taps	10	38	52	Serum WBC; serum ESR; SF WBC; combination of WBC/ESR/SF WBC	SF culture positive; intra-operative findings consistent with SA
Logters et al. (2009) [50]	2b	DEU	42	N/A	Prospective cohort	Acutely inflamed joints	Dry taps	21.4	N/A	78.6	SF cf-DNA; SF IL6; SF TNFα; SF IL1 beta; SF MPO	SF culture positive; intra-operative findings consistent with SA
Logters et al. (2010) [49]	1b	DEU	41	N/A	Prospective cohort	Acutely inflamed joints	Dry taps	30	N/A	70	SF tryptophan; SF kynurenine; Kyn/trpt ratio	SF culture positive; intra-operative findings consistent with SA
Lu et al. (2019) [78]	1b	CHI	70		Prospective cohort	Gouty arthritis, RA and OA involving the knee, and knee swelling with effusions by examination	Patients developing more than one joint disease; use of warfarin or antiplatelet therapy; presence of infection	N/A	28.6	71.4	Serum/SF urate ratio	Gouty arthritis was defined as presence of MSU in SF determined by compensated polarized light microscopy previously or currently
Martinot et al. (2005) [38]	1b	FRA	42	66.6	Prospective cohort	Patients hospitalized for acute arthritis, with: bacterial arthritis, crystal arthritis, RA	Patients without one of the three described conditions	26.1	31	42.9	Serum PCT; SF PCT	SF/serum culture; SF microscopy
McGillicuddy et al. (2007) [27]	2b	USA	49	63	Retrospective cohort	>16 years; diagnosed as SA by SF culture	Nil described	100	N/A	N/A	SF WBC	SF culture
Mico et al. (2015) [16]	3b	ESP	7	N/A	Prospective cohort	Joint fluid samples with higher probability of positive outcome	Nil described	85.7	N/A	14.3	SF FilmArray blood culture identification (PCR)	SF culture
Morgenstern et al. (2018) [65]	2b	GER	57	62	Prospective cohort	>18 years, with acute inflammatory native hip or knee	SF volume <5 ml aspirated	38.6	N/A	61.4	PCR, microcalorimetry	SF/synovial tissue culture positive, or local inflammation, increased SF leukocytes, absence of non-infectious arthritis
Mortazavi et al. (2019) [42]	2b	IRN	25	2.8 (mean)	Prospective cohort	Suspected hip or knee SA undergoing arthrocentesis in children ≤18 years	Findings suggestive of any other diagnosis; insufficient SF; rheumatic diseases; immunodeficiencies; autoimmune disease; renal failure	76	N/A	24	Leucocyte esterase strip	1 or more of the following: (1) positive SF culture, (2) positive bacterial smear, (3) WBC count in the SF >50 × 103 plus positive blood culture, (4) purulent SF
Omar et al. (2014) [62]	2b	DEU	146	59	Prospective cohort	Atraumatic joint effusion of shoulder/elbow/hip/knee	Nil described	13	31.5	55.5	SF leucocyte esterase; SF leucocyte esterase + glucose	SF crystal analysis; Newman criteria; one of following: SF/serum culture positive; purulent SF, no crystals
Omar et al. (2017) [15]	2b	DEU	102	61	Prospective cohort	Atraumatic joint effusion of shoulder/elbow/wrist/hip/knee/ankle	Patients with joint arthroplasty	14.7	N/A	85.3	SF glucose via glucometer	One of the following: Newman criteria; SF/serum culture positive; purulent SF in absence of crystals; negative microbiology, but SF WBC >50000/mm3 and % PMN >75%, no crystals
Shmerling et al. (1990) [59]	2b	USA	100	N/A	Prospective cohort	All SF samples received at haematology or chemistry lab at Beth Israel Hospital November 1987 – October 1988	Samples sent only to microbiology; bursal fluid specimens; repeated aspiration from same patient	8	25	66% (26% no diagnosis, 40% other diagnosis)	Protein, glucose, LDH, lactate, WBC, % PMN	SA—synovial culture. Crystal arthritis—microscopy for intracellular crystals. RA—ACR criteria.
Shu et al. (2019) [43]	2b	USA	40	51 (mean)	Prospective cohort	Arthrocentesis for a swollen or painful joint, with at least 1 ml extra SF obtained	Nil described	27.50	22.50	50	SF lactate, cloudy SF, warmth/erythema, WBC (50 000 or 100 000), PMN, Gram stain, micromotion tenderness	(1) SF culture positive or (2) SA diagnosed by orthopaedists with surgical intervention and i.v. antibiotics given during the hospital stay, even if cultures negative
Sigmund et al. (2019) [44]	2b	AUT	72	64 (mean)	Prospective cohort	Suspected SA	Insufficient data or SF	58.30	N/A	41.70	Automated multiplex PCR; SF culture; combination	Pathogenic organism in affected joint; pathogenic organism from another source, e.g. blood in the context of a hot, red joint suspicious of sepsis; typical clinical features and turbid joint fluid in the presence of previous antibiotic treatment; post-mortem or pathological features of SA; and leucocyte count >50 000/µl or PMN > 90% in SF
Thornton et al. (2019) [55]	2b	GBR	190	63	Retrospective cohort	Patients requiring native joint aspiration	Recent surgery or joint injection; prosthetic joints; paediatric patients; overlying cellulitis	78.9	9.5	82.6	Serum CRP	Consultant clinician overall interpretation of investigations and clinical presentation. Investigations include SF WBC, crystals, extended culture, peripheral WBC, CRP
Vaidya et al. (2018) [56]	2b	NPL	181	51.5	Retrospective cross-sectional	Pain and swelling in one or multiple joints onset ≤1 day	Chronic joint pain (>14 days), known RA/SpA/gout on treatment DMARDs/urate-lowering therapy, those not willing to give consent	0	52.50	47.50	SF to serum uric acid ratio ≥1.01 (diagnosis of gout)	Gout—ACR/EULAR 2015 criteria. RA—ACR/EULAR 2010 criteria. SpA—ASAS criteria. AnkSpA—modified New York criteria. Pseudogout—chondrocalcinosis by Xray or SF aspirate
Wang et al. (2014) [20]	2b	China	95	N/A	Retrospective cohort	Patients from outpatient clinic, 12 January – 13 June	Arthritis other than SA, RA, OA or GA; artificial joints; patients with SA excluded if culture negative; RA, OA and gout patients excluded if culture positive	24.20	11.60	64.20	Serum procalcitonin, SF procalcitonin	ACR criteria and bacterial culture of SF
Wiener et al. (2008) [23]	2b	Switzerland	30	N/A; mean = 58	Retrospective cohort	Referrals by orthopaedic surgeons of suspected SA 1 July 2004 – June 2007	Aspirate <10 ml	33.33	N/A	N/A	Lactate concentration and T2 relaxation time in MR (SF)	Positive synovial culture
Yang et al. (2008) [22]	2b	USA	121	N/A	Retrospective cohort	Suspected acute SA in ED, orthopaedic clinic or rheumatology clinic July 2006 – July 2007	Nil described	17.40	N/A	N/A	PCR assay 16S rRNA gene	Synovial culture
Zamani et al. (2012) [58]	2b	IRN	75	N/A	Retrospective cohort	Knee monoarthritis; joint effusion requiring arthrocentesis	Nil described	4	17.3	78.7	SF adenosine deaminase; SF hs-CRP	SF/serum culture; SF crystal microscopy

cf-DNA: cell-free DNA; MR: magnetic resonance; N/A: not available; LDH: lactate dehydrogenase; PCT: procalcitonin; SA: septic arthritis; ROM: range of movement; ED: emergency department; PG: pseudogout; WBC: white blood cell count; WCC: white cell count.

Articles in the final SLR included the following study types: prospective cohort (n = 25); prospective cross-sectional (n = 2); prospective case–control (n = 1); retrospective cohort (n = 18); retrospective cross-sectional (n = 2); retrospective case–control (n = 5); and a mixed retrospective and prospective cohort (n = 1). Four studies comprised a partial or complete paediatric cohort.

Conference abstracts (n = 782 after deduplication) were screened separately, with 26 meeting eligibility criteria (Supplementary Fig. S1, available at Rheumatology online). Only one study included as a full-text article was present in the conference abstracts eligible for inclusion. Risk-of-bias assessments for included full-text articles are available in Table 2.

Table 2.

Risk of bias assessment using the Newcastle-Ottawa Scale

Study	Risk of bias				Applicability concerns
	Patient selection	Index test	Reference standard	Flow and timing	Patient selection	Index test	Reference standard
Aliste-Fernández et al. (2020)
Baillet et al. (2019)
Baran et al. (2014)
Berthoud et al. (2020)
Bonilla et al. (2011)
Borzio et al. (2016)
Bram et al. (2018)
Carpenter et al. (2020)
Cohen et al. (2019)
Coiffier et al. (2013)
Coiffier et al. (2019)
Colvin et al. (2015)
Couderc et al. (2015)
Couderc et al. (2019)
Cunningham et al. (2014)
Curtis et al. (1983)
Ferreyra et al. (2016)
Foocharoen et al. (2011)
Garg and Goyal (2018)
Gautam et al. (2017)
Gbejuade et al. (2019)
Gratacos et al. (1995)
Hassas Yeganeh et al. (2020)
Jeng et al. (1997)
Kim et al. (2010)
Kinugasa et al. (2019)
Kunnamo et al. (1986)
Lenski and Scherer (2014)
Lenski and Scherer (2014)
Li et al. (2007)
Logters et al. (2009)
Logters et al. (2010)
Lu et al. (2019)
Martinot et al. (2005)
McGillicuddy et al. (2007)
Mico et al. (2015)
Morgenstern et al. (2018)
Mortazavi et al. (2019)
Omar et al. (2014)
Omar et al. (2017)
Proot et al. (2015)
Robinson et al. (2017)
Rohde 2000
Scharf 2015
Shmerling et al. (1990)
Shu et al. (2019)
Sigmund et al. (2019)
Thornton et al. (2019)
Vaidya et al. (2018)
Wang et al. (2014)
Wauters 2013
Wiener et al. (2008)
Yang et al. (2008)
Zamani et al. (2012)

Low Risk    High Risk     Unclear risk.

Differentiation of septic from non-septic arthritis

Most included studies focused on the differentiation of septic arthritis from non-septic causes of a hot joint (usually from inflammatory arthritis or gout). SF markers were investigated in 40 studies [15–54]; serum markers were investigated in seven studies [20, 25, 38, 39, 51, 52, 55]. The markers and thresholds are summarized in Supplementary Table S1, available at Rheumatology online. Most markers had been tested at multiple potential diagnostic thresholds, yielding a total of 27 serum markers and 156 SF markers.

The following serum markers were studied: pro-calcitonin, ESR, CRP, white blood cell (WBC) count, uric acid, TIMP-1, CTX-II and calprotectin. Of these, pro-calcitonin, ESR, CRP and uric acid were tested at multiple thresholds in serum. While serum pro-calcitonin was noted to have excellent specificity (up to a value of 1), its sensitivity was noted to be poor (0.087–0.727), with an additional study finding a sensitivity of 0 for pro-calcitonin at a serum value of 10 μg/l. CRP and ESR were conversely noted to have good sensitivity at multiple thresholds, but poor specificity. Of the remaining markers, TIMP-1 and CTX-II had high specificity (0.94 and 0.89, respectively) but poor sensitivity (0.57 and 0.61, respectively), with calprotectin having moderately good values for both (sensitivity 0.65, specificity 0.77).

There were many more SF than serum markers tested for the differentiation of septic from non-septic arthritis. Due to heterogeneity of study designs, outcomes and diagnostic thresholds, meta-analysis was possible for only eight tests in SF, all differentiating septic from non-septic joints. The results (pooled sensitivity, pooled specificity and AUCs) are summarized in Table 3 and Fig. 2. All tests were conducted in SF, with the following markers being investigated: glucose, lactate, leucocyte esterase, PMNs, pro-calcitonin, TNFα and white cell count (WCC). Overall, of these tests, leucocyte esterase had the highest pooled sensitivity [0.94 (0.70, 0.99)] with good pooled specificity [0.74 (0.67, 0.81)].

Table 3.

Sensitivity, specificity and areas under the curves for SF tests included in the meta-analysis

SF test	Number of articles in meta-analysis	Sensitivity (95% CI)	Specificity (95% CI)	AUC
Glucose (40 mg/dl)	2	0.59 (0.48, 0.69)	0.86 (0.75, 0.92)	0.593
Lactate (≥5 mmol/l)	2	0.56 (0.32, 0.78)	0.77 (0.67, 0.84)	0.768
Lactate (≥10 mmol/l)	2	0.36 (0.22, 0.53)	0.99 (0.96, 1.00)	0.852
Leucocyte esterase (++ or +++)	4	0.94 (0.70, 0.99)	0.74 (0.67, 0.81)	0.784
PMNs (>90%)	2	0.69 (0.41, 0.88)	0.65 (0.53, 0.75)	0.665
Pro-calcitonin (0.5 μg/l)	2	0.67 (0.26, 0.92)	0.93 (0.84, 0.97)	0.931
TNFα (36 pg/ml)	2	0.86 (0.49, 0.97)	0.88 (0.54, 0.98)	0.931
WBC (50 000/mm3)	5	0.56 (0.42, 0.69)	0.90 (0.87, 0.92)	0.895

A test was eligible for meta-analysis if >1 study used the same marker, threshold and fluid. Studies testing markers at other thresholds and/or in serum were not eligible for meta-analysis if not replicated in a second study. AUC: area under the curve; WBC: white blood cell.

Figure 2.

Pooled sensitivity, specificity and areas under curves (AUCs) for eligible SF tests (for differentiating septic from non-septic arthritis) included in meta-analyses

Differentiation of crystal from non-crystal arthritis

Two studies investigated SF biomarkers for the exclusion of gout [31, 56]. One of these also investigated the exclusion of CPPD or pseudogout, using the presence or absence of CPPD crystals [31]. A third study looked at the exclusion of all crystal arthritides (using SF WCC < 1650/mm³) [57]. With regards studies investigating the differentiation of gout from non-gout causes of a hot joint, one studied the utility of a SF to serum uric acid ratio of ≥1.01 [56], yielding a sensitivity of 0.896 (0.81, 0.95) and a specificity of 0.663 (0.56, 0.75). The second study investigated the utility of SF monosodium urate crystals [31]. For the utility of monosodium urate crystals to distinguish gout from non-gout causes of a hot joint, the sensitivity was 0.89 (0.72, 0.98) and the specificity was 1.00 (0.97, 1.00), as would be expected based on current clinical practice. For CPPD to differentiate between CPPD disease from non-CPPD hot joints, the sensitivity was 0.93 (0.80, 0.98) and the specificity was 0.88 (0.78, 0.94). The results from all studies are summarized in Supplementary Table S1, available at Rheumatology online.

Differentiation of inflammatory from non-inflammatory arthritis

The definition of ‘inflammatory arthritis’ varied between studies, but mostly referred to diagnoses including autoimmune inflammatory arthritis, crystal arthritis or septic arthritis. This differentiation of inflammatory vs non-inflammatory arthritis has limited clinical utility, due to the widely varying treatment for each of these diagnoses. Four studies focused on the differentiation of inflammatory from non-inflammatory causes of a hot joint, investigating the utility of eight SF biomarkers [adenosine deaminase, high-sensitivity CRP (hs-CRP), WCC, PMNs, glucose, total protein, lactate dehydrogenase (LDH), leucocyte esterase], summarized in Supplementary Table S1, available at Rheumatology online [58–61]. Of the SF markers tested, adenosine deaminase, CRP, WBC, LDH and leucocyte esterase were noted to have good levels of sensitivity and specificity, but the results were limited by the fact that the studies aimed to differentiate multiple inflammatory arthritides (including septic, crystal and rheumatoid) from non-inflammatory diagnoses such as OA, limiting their utility in clinical practice. No studies looked at serum biomarkers in the differentiation of inflammatory from non-inflammatory arthritis in the acute hot joint setting.

Meta-analysis was not possible for studies investigating markers for differentiating crystal from non-crystal arthritis, or inflammatory from non-inflammatory arthritis, due to heterogeneity in study design, markers and thresholds.

Discussion

This systematic review identified a large number of studies of medium-to-high quality, highlighting several single tests that may have diagnostic utility for differentiating septic from non-septic arthritis. It is important to promptly exclude joint sepsis in acute settings. Our review demonstrates that, based on current evidence, joint aspiration and SF testing remain necessary to facilitate this. However, our review identifies additional tests to those used in current clinical practice that may facilitate more efficient exclusion of septic arthritis.

Our review demonstrates many single tests with some evidence for diagnostic utility. Individually, all have suboptimal accuracy and sensitivity, when compared with the gold standard, for exclusion of native joint infection. A far greater number of SF than serum tests were identified. However, not all are readily available or validated for exclusion or diagnosis of septic arthritis, e.g. pro-calcitonin (PCT) and TNF-α. The individual biomarker with greatest sensitivity and specificity was SF leucocyte esterase, a relatively cost-effective, quick and easy test that could be conducted in acute clinical settings to give early indication of SA. However, further testing would be required before this could be used in place of SF microscopy and culture in routine clinical care.

It is possible that a panel of individual serum and synovial biomarkers may yield better early diagnostic accuracy for (1) diagnosis of SA, or (2) exclusion of SA.

A far greater number of markers, tested at multiple thresholds for differentiation of septic from non-septic joints, were identified in SF [15–40, 42–54, 62–65]. This may reflect the fact that, early in SA, infection may be confined to the joint space, therefore SF sampling, including culture, is key to diagnosis. SA arises due to bacterial deposits in the synovial membrane, leading to acute inflammation. Bacteria can easily enter the joint space as synovial tissue has no basement membrane [2]. The severity of infection may therefore not be represented by levels of serum markers early in the disease process.

In clinical practice, it is crucial to be able to distinguish reliably between septic and crystal arthritis, which have similar presentations. SF microscopy and culture are already able to do this; however, culture results take several hours to days to be processed. In addition, rarer causes of joint infection, such as tuberculosis, Brucella and fungi, will not be identified by usual culture methods. Polarizing microscopy to identify CPPD or uric acid crystals can be done relatively quickly, sometimes within the rheumatology unit [66, 67]. Ultrasonography is also increasingly available to identify the characteristic double-contour sign of gout, as well as the presence of CPPD [68]. However, even in the presence of confirmed gout, SA may co-exist [5]. Furthermore, in the emergency setting, such as in emergency departments or out of hours, these resources may not always be available. It is necessary to be able to promptly predict a diagnosis of infection accurately, with tests with high sensitivity and good specificity.

Other important causes of an acute hot joint include a flare of inflammatory arthritis or OA. We identified seven studies that tested the utility of serum markers for distinguishing septic from non-septic joints [20, 25, 38, 39, 51, 52, 55]. Markers included CRP, pro-calcitonin and ESR at multiple thresholds, as well as uric acid and calprotectin. It is important to note that serum uric acid often paradoxically decreases during flares, due to increased renal clearance, and is therefore not an accurate marker of acute gout [69]. Both the specificity and the sensitivity of serum CRP and ESR, regardless of threshold, were suboptimal. Pro-calcitonin was found to have high specificity, above 0.9, regardless of threshold, for SA. However, the sensitivity was consistently low (0.35–0.55) [20, 38]. This is a limitation, as it is important to reliably exclude SA when using this test in the acute setting.

Meta-analysis was possible for eight tests (of the same marker, fluid and threshold), identifying leucocyte esterase in SF as the marker with the most optimal pooled sensitivity and specificity for SA [24, 42, 48, 62]. This is a simple point-of-care ‘dipstick’ test, usually undertaken on urine, making this a potentially cost-effective and quick screening test for joint infection. The evidence for its use in PJI is well established, where SF leucocyte esterase may be used as part of the diagnostic work-up [70, 71], and the sensitivity and specificity for infection is high. However, there are limitations to its precision and readability, and it is most likely to be useful as part of a panel of tests.

Another SF biomarker commonly used in the diagnosis of PJI is alpha-defensin. However, we identified no studies on the use of this marker in native SA. It may be interesting for future studies to evaluate its use in the native hot joint setting. It is, however, important to note several differences between prosthetic and native joint infections, which may account for the limited use of markers such as alpha-defensin and leucocyte esterase, and the overall lack of research in native joint infection markers [72]. An important difference is in the pathogenesis; crucially, the formation of a biofilm in PJI aids the survival and growth of bacteria and, if left to mature, can lead to difficult-to-treat chronic PJI. Another key difference is the feature of blood cultures in the definition of SA, and its absence from PJI criteria [73, 74]. This has led to the development and use of alternative diagnostic tests in PJI, while blood cultures remain key to the diagnosis of SA, along with physician or clinical assessment.

Tests such as glucose, pH and WCC levels are easy to conduct, quick and cost-effective. Meta-analysis of glucose and WCC in SF at the same thresholds found high pooled specificity but low sensitivity. However, one study testing the use of a glucometer at a threshold of 1.4 mmol/l demonstrated greater sensitivity for joint infection [15]. Similarly, sensitivities for SF WCC, at various thresholds, were wide-ranging, suggesting further work is required in this area to assess its utility for distinguishing joint infection. It is also important to note that smaller joints may have a higher WCC than larger ones in the presence of similar clinical inflammation, which can also be misleading [75]. Lactate is another marker that is easy and cost-effective to test, and at a threshold of >10 mmol/l in SF, had the best specificity. However, the overall results were variable between the included studies. A marker that is less widely available, but may have diagnostic utility and was investigated by two studies, is SF PCT [20, 38]. PCT in serum has been shown to be a useful early indicator of sepsis [20], and from these initial two studies, it appears to be able to distinguish between infected and non-infected joints at certain thresholds. This is a biomarker that certainly is worthy of further research in this area.

Of the included studies, seven aimed to either distinguish crystal from non-crystal, or inflammatory from non-inflammatory arthritis [31, 56–61]. It is ultimately most important to distinguish an infected from a non-infected joint, perhaps reflected by the small numbers of studies focusing on non-infectious causes of a hot joint. Nonetheless, this raises the question of whether a panel of rapid tests for SF, each with a high sensitivity and specificity for a given diagnosis, may be of greater utility in the acute setting than a single test. This would not only exclude infection as a cause of the hot joint, but also provide an indication as to the possible cause, while more specific tests are awaited. For example, one included study found a sensitivity and specificity of 0.9 or greater for SF calcium pyrophosphate crystals and monosodium urate crystals for pseudogout and gout, respectively [31].

Studies have shown that no single biomarker, in serum or SF, can diagnose or exclude SA alone [25]. It requires a combination of clinical findings, physical examination and multiple laboratory investigations, from some of which it may be feasible at the point-of-care to form a panel or score. Scores using multiple biomarkers are reliable in identifying SA in acute hot joints, comprising SF lactate, WCC, crystals and glucose tests [76]. Similar criteria or scores are used in other specialties to aid diagnosis, such as in pleural effusions, in which Light’s criteria (comprising lactate dehydrogenase and protein levels) are used to distinguish a transudative from an exudative effusion, and pH is used to distinguish a complicated from an uncomplicated parapneumonic effusion [77]. Give the variation in the utility of the tests identified in this review, it is likely that a panel of multiple tests, rather than a single test, would be of greatest benefit and accuracy in the acute setting when assessing hot joints. Based on our results, such a panel may comprise SF lactate, WCC and leucocyte esterase, along with point-of-care uric acid identification, which is already available. However, further work, including validation of such tests, is required.

Strengths and limitations

Our review has several strengths. To our knowledge, this is the first review of this scale undertaken to compare tests used in the setting of an acute hot joint, in a largely understudied area. We identified a large number of papers, enabling comparison of many SF and serum markers at multiple thresholds and for multiple diagnoses in the acute hot joint setting. We were able to extract a large volume of data to facilitate these comparisons and analyse this topic in depth. However, our study was limited by the fact that many markers were tested at multiple thresholds, and meta-analysis is only possible where markers are tested for a given diagnosis at a specified threshold. Therefore, meta-analysis was only possible on a small number of markers. Nonetheless, valuable comparisons were possible even in the absence of meta-analysis for those studies where this was not possible. Additionally, the nature of acute hot joints means that multiple potential diagnoses are possible, which can help as well as hinder comparisons across studies. We stratified studies by septic vs non-septic, crystal vs non-crystal and inflammatory vs non-inflammatory arthritis, to facilitate comparisons between markers testing for the same diagnoses.

Conclusion

Our review demonstrates the potential of multiple individual tests in SF and serum that may be able to facilitate prompt exclusion of joint sepsis in the acute setting. Having a reliable score or testing panel to distinguish septic from non-septic arthritis in acute hot joints would ensure antibiotic treatment is delivered promptly where needed, and avoided where there is very low likelihood of infection, avoiding complications such as unnecessary admissions and antimicrobial use. Further work developing testing panels including serum, urine and synovial panels of tests to facilitate prompt bedside diagnosis is essential. If effective this could be employed in emergency care and even primary care setting to reduce the need for hospital admission and unnecessary antibiotic use.

Supplementary material

Supplementary material is available at Rheumatology online.

Data availability statement

Data available upon request.

Funding

This work is generously funded by the British Society for Rheumatology and a Wellcome Trust Seedcorn Award.

Disclosure statement: The authors have declared no conflicts of interest.