Diagnostic algorithm in septic total knee arthroplasty failure – What is evidence-based?

Abstract

Prosthetic joint infection (PJI) is among the most common differential diagnoses of total knee arthroplasty failure. It is a challenging complication, not least because of the difficulty of establishing the correct diagnosis. The fact that no single diagnostic parameter or test has been identified that can accurately rule in or out PJI has led to an evolution of similar but competing definitions of PJI on the grounds of an array of criteria. This development has had very positive effects on the scientific evaluation of various methods of PJI diagnostics and treatment because of an increased comparability. However, it can be challenging to stay abreast of the evidence these definitions are based on. Also, the definitions alone do not necessarily entail an algorithm to aid in evaluating the right criteria in a sound order to be able to use the definitions as a sensible tool. The aim of this overview is to state the most recent evidence on the diagnostic parameters included in the most established PJI definitions and to exhibit and compare the few algorithmic approaches published.

Clinical symptoms of PJI are very rarely reported on in the literature, hence the evidence on their diagnostic value is poor. The only symptom that is part of the established PJI definitions is the presence of a fistula. Concerning serological markers, CRP and ESR are still the common denominator in the literature, most recently accompanied by D-Dimer as a potentially suitable marker that has been included in the most recent update of the International Consensus Meeting (ICM) criteria. Imaging plays a minor role in the diagnostic cascade because of inconsistent evidence, and no role whatsoever in the established definitions. The most important preoperative diagnostic measure is arthrocentesis and cultural and cytological analysis of the synovial fluid. The much acclaimed α-Defensin test has so far not been included in the established criteria due to inconsistent reports on its diagnostic accuracy, it is, however, in wide use and considered an optional diagnostic tool for inconclusive cases. The most diagnostic accuracy lies in the cultural and histological analysis of periprosthetic tissue biopsies, whether they are gathered in a small procedure or during arthroplasty revision.

Published algorithmic approaches to PJI diagnosis are much rarer than the well-established definitions by various associations. With their PJI definition, the American Academy of Orthopedic Surgeons (AAOS) published a consensus based flowchart for PJI diagnosis. Another algorithm was proposed as part of the endeavor of the MSIS and the first consensus meeting, also based on a consensus among experts. There have been two more recent publications of flowcharts based on the current evidence, one introduced at our institution in 2013, one established in 2020 by the German Society for Arthroplasty (AE).

1. Introduction

One of the core difficulties of the management of total knee arthroplasty (TKA) failure is to determine whether the cause lies in a prosthetic joint infection (PJI). As there is no single criterion with sufficient diagnostic accuracy to establish the diagnosis of PJI or to completely rule it out, orthopedic surgeons have to rely on a combination of clinical findings, technical examinations, and their own experience and judgment. In order to establish a reliable diagnostic standard, several institutions have postulated definitions of PJI along with distinct criteria and suggested algorithms. The first such endeavor was undertaken by the American Academy of Orthopedic Surgeons in 2010 with the publication of clinical guidelines for the diagnosis of PJI of the hip and knee.¹ In 2011, the Musculoskeletal Infection Society (MSIS) followed suit.² In 2013, the International Consensus Meeting on the Definition of Prosthetic Joint Infections (ICM) published a slightly amended definition of that by the MSIS and updated this definition along with the introduction of a scoring system that acknowledges “inconclusive” cases in 2018^3,4 (see Fig. 1). The Infectious Diseases Society of America (IDSA) published their definition in 2013, as well.⁵ In Europe, the so-called “Zimmerli criteria” of 2004 have been in wide use, and have been proposed in updated form as the European Bone and Joint Infection Society's (EBJIS) definition.^6,7 For a comparison of the stated PJI definitions, see Table 1.

Fig. 1.

Scoring sheet of the 2018 ICM Definition of PJI reproduced from .⁴ Original legend reads: “Proceed with caution in: adverse local tissue reaction, crystal deposition disease, slow growing organisms. CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; LE, leukocyte esterase; PMN, polymorphonuclear; WBC, white blood cell. ^aFor patients with inconclusive minor criteria, operative criteria can also be used to fulfill definition for PJI. ^bConsider further molecular diagnostics such as next-generation sequencing.”

Table 1.

Overview of the evidence-based definitions of PJI and the interpretation of their criteria, starting with the MSIS definition of 2011.

Definition	MSIS	IDSA	ICM	ICM	Proposed EBJIS
Year	2011	2013	2013	2018	2018
Criteria	Major:	Sinus tract communicating with the prosthesis	Major:	Major:	Sinus tract or purulence around the prosthesis
	Sinus tract communicating with the prosthesis	Purulence without other etiology surrounding the prosthesis	A sinus tract communicating with the joint	Sinus tract with evidence of communication to the joint or visualization of the prosthesis	Synovial fluid leukocyte count >2000 cells/ml or >70% granulocytes
	A pathogen is isolated by culture from at least two separate tissue or fluid samples obtained from the affected prosthetic joint	Acute inflammation seen on histopathological examination of the periprosthetic tissue	Two positive periprosthetic cultures with phenotypically identical organisms,	Two positive cultures of the same organism	≥23 granulocytes per 10 high-power fields in histopathology
		≥2 intraoperative cultures or combination of preoperative aspiration and intraoperative cultures yielding an indistinguishable organism (the growth of a virulent microorganism in a single specimen is also considered as indicative of a PJI)			Microbial growth in:
	Minor:		Minor:	Minor:	- Synovial fluid
	Elevated ESR (>30 mm/h) and CRP (>10 mg/L) concentration C7		Elevated synovial fluid WBC count (>10,000 cells/mL for acute infections; >3000 cells/mL for chronic infections) or ++ change on leukocyte esterase test strip	Elevated CRP (>100 mg/L for acute infections; >10 mg/L for chronic infections) or D-Dimer (unknown threshold for acute infection; >860 μg/L for chronic infection) (2 points)	- Periprosthetic tissue (2 samples out of at least 3 collected; for highly virulent organisms, e.g. S. aureus, streptococci, E. coli, or patients under antibiotics, one positive sample confirms infection)
	Elevated synovial leukocyte count		Elevated PMN% (>90% for acute infections; >80% for chronic infections)	Elevated ESR (no role for acute infections; >30 mm/h for chronic infections) (1 point)	- Sonication culture (>50 CFU/ml; under antibiotics, for S. aureus and anaerobes, < 50 CFU/ml can be significant)
	Elevated PMN%		Positive histological analysis of periprosthetic tissue (>5 neutrophils per high-power field in five high-power fields (d) observed on periprosthetic tissue at ×400 magnification)	Elevated synovial WBC count (>10,000 cells/mL for acute infections; >3000 cells/mL for chronic infections) or Leukocyte Esterase (++ for acute and chronic infections) OR Positive alpha-defensin (3 points)
	Purulence in the affected joint		A single positive culture	Elevated synovial PMN% (>90% for acute infections; >70% for chronic infections) (2 points)
	Isolation of a microorganism in one culture of periprosthetic tissue or fluid			Single positive culture (2 points)
	Greater than five neutrophils per high-power field in five high-power fields observed in histologic analysis of periprosthetic tissue at ×400 magnification			Positive histology (3 points)
Interpretation	1 of the 2 Major Criteria	≥1 Positive Criterion	1 of the 2 Major Criteria	1 of the 2 Major Criteria	≥1 Positive Criterion
	or		or	or
	≥4 of 6 Minor Criteria		≥3 of 5 Minor Criteria	≥6 points: Infected
				3–5 points: Possibly infected
				<3 points: Not infected

Even though these definitions differ in some aspects, they have been tantamount to raise awareness for the need to adhere to a set of established criteria in order to evaluate the diagnostic validity of old and new markers, and in order to compare the incidence, diagnostic pathways, treatment, and outcomes of PJI between centers across the globe. The following overview is intended to highlight what evidence the respective definitions are based on and to outline a recently established, evidence-based diagnostic algorithm defined by the work group on implant-associated infection of the German Society for Arthroplasty (AE).

2. Clinical symptoms

While an acute infection of TKA presents with the typical symptoms of pain, swelling, redness, warmth, and impaired range-of-motion, which are usually more prominent in the knee joint due to its thinner soft-tissue envelope, the clinical presentation of chronic, low-grade infections is equivocal and often unremarkable. Gomes et al. state that the most sensitive but least specific symptom for PJI is pain. It is also the most common reported symptom of both acute and low-grade PJI. As the second most frequent symptom is joint dysfunction, it becomes obvious that clinical symptoms alone cannot differentiate adequately between infection and the most common differentials, i.e., loosening, instability, and malalignment (Table 2). Furthermore, the consistency and quality of the reported symptoms in the literature is poor.⁸

Table 2.

Diagnostic accuracy and prevalence of various clinical findings in any PJI and in specific types of PJI. #CF: Observed clinical finding: (True +) + (False +). NR: Not Reported. ^aComposite for effusion/swelling, warmth, and/or erythema, ^bComposite for delayed healing, non-purulent wound drainage, and/or superficial dehiscence, ^cComposite for sinus tract, suppuration, purulent drainage, abscess, and/or extensive necrosis, ^dComposite for stiffness and/or reduced range of motion. Reproduced from .⁸.

Clinical Findings Composites	Clinical Findings Analysis for Non-specific Type of PJI				Incidence of Clinical Findings for Non-specific and Specific Types of PJI. Mean % (95% CI)
	#CF/Controls (# Infected THA/TKA/Other)	Sensitivity % (95% CI)	Specificity % (95% CI)	Accuracy % (95% CI)	Non-specific Type of PJI	Early Postoperative	Acute Hematogenous	Late Chronic
Pain	225/237 (77/18/0)	57.9 (47.3–67.9)	28.3 (22.6–34.5)	36.7 (31.5–42.2)	77.2 (74.5–79.9)	54.7 (51.3–58.0)	92.8 (89.1–96.6	82.9 (77.5–88.3)
Fever	17/205 (62/3/0)	13.8 (6.5–24.7)	96.1 (92.5–98.3)	76.3 (70.7–81.2)	31.2 (28.2–34.1)	32.5 (29.4–35.5)	75.5 (71.3–79.7)	14.0 (10.4–17.6)
Periarticular Inflammation	113/716 (316/220/0)	14.9 (11.9–18.4)	94.7 (92.8–96.2)	61.8 (59.0–64.6)	35.9 (33.2–38.6)	49.0 (44.8–53.3)	69.7 (60.1–79.2)	18.4 (14.9–22.0)
Superficial surgical site disturbances	487/1771 (538/392/0)	23.6 (21.3–26.1)	88.6 (87.0–90.0)	61.8 (59.9–63.5)	40.4 (37.5–43.2)	46.8 (40.6–53.0)	NR	24.1 (17.9–30.2)
Deep soft tissue involvement	24/62 (4/17/3)	43.7 (29.5–58.8	100.0 (97.0–100.0)	84.3 (78.0–89.4)	38.8 (35.4–42.1)	44.0 (40.7–47.4)	11.0 (8.1–14.0)	26.3 (21.8–30.9)
Joint dysfunction	NR	NR	NR	NR	74.4 (69.9–78.9)	NR	20.5 (1 reference)	41.7 (1 reference)
Total # PJI included (THA/TKA/Other)	976 (561/412/3)	-	-	-	2523 (1421/1071/31)	902 (491/408/3)	435 (152/274/9)	607 (385/181/41)

Interestingly, the presence of a sinus tract is ubiquitously declared to be sufficient to prove PJI. This is, however, based solely on the assumption that the implant is certainly colonized once it has contact with the exterior. There are no studies on this matter available.⁹

3. Laboratory parameters

The MSIS and ICM definitions are consistent in demanding C-reactive protein (CRP) and erythrocyte sedimentation rate (ESR) as standard diagnostic markers.^1,4,6,10 A threshold of 10 mg/l for CRP and 30 mm/h for ESR are stated. These thresholds were agreed upon to reach a sufficient sensitivity for low-grade PJI. In acute infections, both markers can be expected to be elevated well beyond these thresholds. The 2018 MSIS definition provides a threshold of 100 mg/l for CRP to account for acute infections. It is important to emphasize two critical downsides of serum inflammatory markers: First, they can be elevated due to numerous conditions unrelated to PJI. Second, the calculated diagnostic accuracy of CRP and ESR is derived from studies including both acute and low-grade PJI. Hence, it is likely to be overestimated and these markers are likely to be less sensitive in reality. Therefore, they should be used only as a screening test, triggering further investigations if one or both are elevated. Normal values do not exclude the presence of low-grade PJI.11, 12, 13, 14, 15 For an overview of the diagnostic value of CRP and ESR, see Table 3. The IDSA and proposed EBJIS definitions do not include CRP or ESR.^5,7

Table 3.

Overview of the diagnostic accuracy of CRP and ESR. LoE: Level of Evidence (Eastern Association for the Surgery of Trauma, EAST). CoR: Class of Recommendation (EAST).16, 17, 18, 19, 20, 21, 22 LR: Likelihood Ratio. PPV: Positive Predictive Value. NPV: Negative Predictive Value. Table translated from .²³

Year	Author	Examination	LoE	CoR	No. of patients	Sensitivity	Specificity	Positive LR	Negative LR	PPV	NPV
2007	Bottner et al.24		I	I	78
		CRP			0,95 (0,86–1,0)	0,91 (0,94–0,99)	10,86 (5,3–73,07)	0,05 (0,0–0,17)	0,8 (0,64–0,96)	0,98 (0,94–1,0)
		(1,5 mg/dl)
		ESR			0,81 (0,64–0,98)	0,89 (0,82–0,97)	7,69 (3,47–38,2)	0,21 (0,02–0,44)	0,74 (0,56–0,92)	0,93 (0,8–1,0)
		(32 mm/h)
2004	Savarino et al.25		I	I	26
		ESR				0,6 (0,3–0,9)	0,94 (0,82–1,0)	9,6 (1,64–16,1)	0,43 (0,09–0,86)	0,86 (0,71–1,0)	0,79 (0,61–0,97)
		(50 mm/h)
		CRP				0,38 (0,14–0,61)	0,7 (0,42–0,98)	1,25 (0,24–38,34)	0,89 (0,39–2,07)	0,67 (0,36–0,97)	0,42 (0,18–0,65)
		(2 mg/dl)
1980	Kamme et al.26		I	I	63
		ESR				0,89 (0,8–0,99)	0,72 (0,54–0,9)	3,2 (1,75–9,54)	0,15 (0,01–0,37)	0,83 (0,71–0,94)	0,82 (0,66–0,98)
		(30 mm/h)
2007	Greidanus et al.27		I	I	151
		ESR				0,82 (0,71–0,93)	0,88 (0,81–0,94)	6,7 (3,84–15,52)	0,2 (0,07–0,36)	0,74 (0,62–0,86)	0,92 (0,87–0,97)
		(30 mm/h)
		CRP				0,93 (0,86–1,0)	0,83 (0,76–0,9)	5,5 (3,57–10,23)	0,08 (0,01–0,18)	0,7 (0,58–0,82)	0,97 (0,93–1,0)
		(1,0 mg/dl)
2007	Della Valle		I	I	94
	et al.28
		ESR				0,9 (0,81–0,99)	0,66 (0,53–0,79)	2,66 (1,74–4,68)	0,15 (0,01–0,35)	0,67 (0,55–0,8)	0,9 (0,8–0,99)
		(30 mm/h)
		CRP				0,95 (0,89–1,00)	0,75 (0,64–0,87)	3,88 (2,45–7,86)	0,06 (0,02–0,18)	0,75 (0,63–0,87)	0,95 (0,89–1,0)
		(1 mg/dl)
2008	Schinsky et al.29		I	I	201
		ESR				0,96 (0,91–1,0)	0,39 (0,31–0,47)	1,58 (1,33–1,91)	0,09 (0,0–0,28)	0,37 (0,29–0,45)	0,97 (0,92–1,0)
		(30 mm/h)
		CRP				0,95 (0,89–1,0)	0,71 (0,94–1,0)	3,29 (2,45–4,69)	0,08 (0,0–0,18)	0,55 (0,45–0,65)	0,97 (0,94–1,0)
		(1 mg/dl)
2007	Fink et al.30		I	I	145
		CRP				0,73 (0,59–0,86)	0,81 (0,73–0,88)	3,81 (2,21–7,48)	0,34 (0,15–0,56)	0,59 (0,45–0,73)	0,89 (0,82–0,95)
		(1,35 mg/dl)

There has been increasing interest in serum D-dimer as a screening test for PJI, leading to its inclusion in the 2018 MSIS definition, where it can suffice as a minor criterion instead of CRP in combination with the ESR.^4,31 The proposed threshold lies at 860 ng/ml.⁴

Other serum markers such as interleukin 6 (IL-6), tumor necrosis factor α (TNF-α) and procalcitonin have not been included in the most common PJI definitions due to inconsistent reports on their diagnostic accuracies.^13,24, 32, 33

4. Imaging

Imaging findings are not part of the common PJI definitions. Plain radiographs are part of every basic workup of painful total knee arthroplasties. If loosening can be seen, radiographs cannot differentiate between septic and aseptic cases. The speed at which signs of loosening develop can merely suggest a potential septic background.³⁴ However, both acute and low-grade infections often appear unremarkable on plain radiographs. Even though there is evidence for a good diagnostic accuracy of magnetic resonance imaging (MRI) of the knee, the reports are not consistent enough for inclusion of the method in the PJI definitions or its recommendation in a diagnostic algorithm.³⁵ At our institution, we could show a high sensitivity and very high specificity of metal-artefact-reduced sequences (MARS) to differentiate aseptic causes of failure, septic and aseptic loosening, and PJI of the hip, suggesting a potential use for the method on failed TKA, as well.³⁶

The American Academy of Orthopaedic Surgeons (AAOS) diagnostic algorithm includes nuclear imaging, but only late in the cascade¹.^99mTc bone scintigraphy exhibits good sensitivity, albeit it is not specific for the detection of PJI.³⁷ Furthermore, signal alterations around the prosthesis are normal within 2 years after implantation.³⁸ Positron-emission tomography (PET) or labelled imaging such as anti-granulocyte scintigraphy exhibit higher diagnostic accuracy for this purpose (sensitivity 0.82 vs. 0.83 and specificity 0.87 vs. 0.79, respectively).^9,39,40

5. Synovial fluid analysis

Joint aspiration is a key diagnostic step in the work-up of suspected PJI. The analysis of the synovial white blood cell count (WBC) and differential is a main feature of the established PJI criteria and exhibits an excellent diagnostic accuracy.^9,41 The cell count and differential thresholds vary in the available literature. The ICM and MSIS criteria of 2013 and 2018 suggest a cut-off at 3000 cells/μl with 80% neutrophils for low-grade and 10,000 cells/μl with 90% neutrophils for acute infections. In the proposed EBJIS criteria, the threshold for low-grade infection is set lower, at 2000 cells/μl and 65% neutrophils. This is due to the use of more sensitive culture methods such as sonication fluid culture as gold standard^3,4,7 and again due to the weakness of some of the underlying studies including acute infections with considerably higher cell counts. To account for low-grade infections, a lower cut-off at 1.500 cells/μl and at 65% neutrophils was chosen by the AE. For an overview of the evidence for the aforementioned thresholds, see Table 4.

Table 4.

Overview of the diagnostic accuracy of synovial fluid analysis for detection of PJI in TKA. LoE: Level of Evidence (Eastern Association for the Surgery of Trauma, EAST). CoR: Class of Recommendation (EAST).16, 17, 18, 19, 20, 21, 22 LR: Likelihood Ratio. PPV: Positive Predictive Value. NPV: Negative Predictive Value. Table translated from .²³.

Year	Author	Examination	LoE	CoR	No. of patients	Sensitivity	Specificity	Positive LR	Negative LR	PPV	NPV
2007	Della Valle		I	I	94
	et al.28
		Culture				0,8	0,93	n.a.	n.a.	0,94	0,84
		Cell Count (3,0 × 103)				0,98 (0,93–1,9)	1	X	0,02	1	0,98 (0,94–1,0)
		Cell Differential				0,98 (0,93–1,0)	0,85 (0,75–0,95)	6,46 (3,75–18,75)	0,03 (0–0,1)	0,83 (0,73–0,94)	0,98 (0,94–1,0)
2007	Fink et al.30		I	I	145
		Culture				0,73 (0,59–0,86)	0,95 (0,91–0,99)	15,23 (6,64–125,4)	0,29 (0,14–0,45)	0,85 (0,73–0,97)	0,9 (0,85–0,96)
2008	Ghanem et al.42		I	I	429
		Cell Count				0,91 (0,86–0,95)	0,88 (0,84–0,92)	7,59 (5,45–11,81)	0,11 (0,05–0,16)	0,82 (0,76–0,88)	0,94 (0,91–0,97)
		Cell Differential				0,95 (0,92–0,98)	0,95 (0,92–0,97)	18,19 (11,62–38,43)	0,05 (0,02–0,09)	0,92 (0,87–0,96)	0,97 (0,95–0,99)
2007	Trampuz et al.43		I	I
		Cell Count (1,7 × 103)				0,94 (0,86–1,0)	0,88 (0,81–0,94)	7,76 (4,65–17,92)	0,07 (0,0–0,17)	0,73 (0,60–0,86)	0,98 (0,95–1,0)
		Cell Differential				0,97 (0,91–1,0)	0,98 (0,95–1,0)	48,04 (19,07–136,76)	0,03 (0,0–0,09)	0,94 (0,87–1,0)	0,99 (0,97–1,0)
2012	Zmistowski et al.44		I	I	150
		Cell Count (3,0 × 103)				0,93 (0,87–0,99)	0,94 (0,88–0,99)	14,35 (7,28–99)	0,07 (0,01–0,14)	0,93 (0,87–0,99)	0,94 (0,88–0,99)
		Cell Differential				0,93 (0,87–0,99)	0,83 (0,75–0,91)	5,52 (3,46–11,62)	0,08 (0,01–0,17)	0,84 (0,76–0,92)	0,93

Novel synovial markers have emerged more recently, some of which have been included in the updated ICM definition and the proposed EBJIS criteria. Leukocyte-esterase test strips are readily available and economical, showing a sufficient diagnostic accuracy to serve as an alternative to an elevated cell count in the 2018 ICM definition.^4,45 Results can be distorted by traumatic aspiration, therefore prior centrifugation is recommended to remove blood from the sample.⁴⁶

Synovial fluid α-Defensin has been propagated in recent years as an extremely accurate diagnostic tool.⁴⁵ It is available as a lateral flow test strip, enabling a bed-side result, and as a laboratory enzyme-linked immunosorbent assay (ELISA). However, more recent studies have shown a lower sensitivity of the lateral-flow test and, to a lesser extent, of the ELISA than previously reported when current PJI definitions are used as a gold standard. Renz et al. state that the α-Defensin test is suited as a confirmatory, not a screening test.^7,47,48 It is also applicable as a highly specific alternative to synovial fluid leukocyte count and differential in the early postoperative period, when the results of the latter are not reliable.^7,48

Cultivating synovial fluid to identify the infectious agent is part of every established diagnostic cascade. Apart from being a criterion to prove the infection, knowing the underlying microorganism and its resistogram is essential for treatment planning. It is universally accepted that the fluid should be cultivated for 14 days to account for fastidious and slow-growing organisms such as Cutibacterium (formerly Propionibacterium). For improved detection rates, it has been suggested to transfer the sample into pediatric aerobic and anaerobic blood culture bottles for transport.⁴⁹ Reported sensitivities of synovial fluid culture vary between 0.45 and 0.75, but the method is highly specific^50,51 (see Table 3). The results can be disturbed by prior antibiotic treatment, which is why a cessation of treatment of at least 2 weeks prior to aspiration is recommended.

In a recent study, we could show a higher accuracy for preoperative microorganism detection in synovial fluid samples using a multiplex PCR analysis with a sensitivity of 0.8 and a specificity of 1.0. While the method is labor-intensive and demanding on the organizational workflow and infrastructure, we see a benefit when using multiplex PCR in the diagnostic work-up of inconclusive cases.⁵² The 2018 ICM definition recommends consideration of molecular methods such as next generation sequencing for patients with inconclusive preoperative scores, as well.

6. Intraoperative tissue sample analysis and sonication

The ICM consensus is that at least 3 and no more than 6 biopsies should be taken for culture from the tissue surrounding the arthroplasty. Again, a prolonged cultivation of 14 days is to be ordered.^3,53 The diagnostic accuracy of tissue biopsy cultivation is slightly higher than that of synovial fluid culture, albeit with a broad range in the available literature.^11,53,54 It appears that, in contrast to synovial fluid culture, perioperative administration of antibiotics does not alter biopsy culture results.^55,56 Growth of the same organism on ≥ 2 tissue sample cultures is proof of infection by the MSIS, ICM and proposed EBJIS criteria, and additionally, offers useful information on the infectious agent and antibiotic susceptibility.

Sonication of the explanted components and sonication fluid culture has been shown to increase organism detection sensitivity (0.79) and specificity (0.99) when using a cut-off of 50 colony-forming units (CFU)/ml.⁴³

Histological work-up of the periprosthetic synovia-like interface membrane (SLIM) has long been included in the most widely established criteria. Some publications have advocated intraoperative frozen section, while others recommend postoperative microscopy of the garnered tissue, using the amount of neutrophils to classify the pathology.28, 29, 30, 57, 58, 59 While all reported methods have been shown to be highly specific, the classification system as proposed by Morawietz and Krenn has the advantage of including other possible pathologies than PJI, therefore offering valuable information concerning other differentials for arthroplasty failure.⁶⁰

7. Strategic approaches to PJI diagnosis

The aforementioned PJI definitions outline the goal by stating the criteria that can best prove or rule out PJI. Only some of them offer the way to achieve that goal. They rather sum up the established diagnostic methods, with a certain threshold to rule in infection depending on the proponents’ weighing of the criteria.

Over the years, it has become apparent that in addition to the set of criteria to be examined, there is a demand for a guideline on when to examine which criterion. The AAOS guideline suggested two separate diagnostic algorithms to be used depending on the patients' risk for PJI.¹ The suggestion was based on consensus. In short, the AAOS algorithm hinges on the patient's CRP and ESR: If they are negative, no further diagnostics are recommended and infection is declared to be “unlikely”. If positive, joint aspiration is recommended. If this does not yield unequivocal results, the algorithm calls for repeat aspiration. Should that be inconclusive, as well, frozen sections and intraoperative white blood cell count and differential are to be obtained in cases where surgery was planned. In others, nuclear imaging is recommended. In low-risk patients, a reevaluation after 3 months is preferred to nuclear imaging. Considering the more recent evidence stated above of PJI without elevated serological markers, this approach is likely to miss some septic arthroplasty failures.

A much simpler diagnostic algorithm was published by Parvizi et al. in 2011, based essentially on what soon after became the first MSIS and ICM definitions.¹⁰ It does, however, not consider when during the course of treatment the information evaluated becomes available. Like the MSIS and ICM definitions that followed, it basically defines overruling criteria that make infection likely and second tier criteria that do so when three of them are positive. Therefore, it is rather a visualization of the later defined major and minor criteria than an actual suggestion of a workflow. The ICM of 2013 included a diagnostic flow chart based on the one published by Parvizi et al. which included preoperative biopsies and nuclear imaging in inconclusive cases.⁶¹ This suggestion was also based on consensus.

The updated 2018 ICM definition does not include an algorithm. It does, however, take into account that the majority of the previously promoted criteria could only be examined by intraoperative tissue collection. Therefore, the new score is divided into three separate elements: The major criteria, proving the presence of PJI when fulfilled; a preoperative section including CRP or D-Dimer, ESR, and the synovial fluid workup (without culture); and an intraoperative section including culture, histology, and the presence of purulence. A score is assigned to each criterion, leading to a more differentiated diagnostic approach, where infection can be ruled out, possible, or proven depending on the score reached before and after surgery.

At our institution, a diagnostic algorithm based on the then available evidence was established and published in 2017.⁶² Its aim was to even out some of the pitfalls of the previous algorithms by relativizing pathological serology as the sole trigger for joint aspiration by adding patient risk factors, suggestive history, and symptoms within the short-to mid-term after surgery. Furthermore, the value of tissue biopsies for culture and histology was accommodated by suggesting arthroscopic or open biopsy before arthroplasty revision in cases where previous diagnostic measures are inconclusive. The algorithm was later reevaluated after clinical use in a high volume center.⁶³ Recently, the work group on implant-associated infection of the German Society for Arthroplasty (AE) assembled an evidence-based algorithm encompassing key features of our institution's 2017 publication²³ (see Fig. 2).

Fig. 2.

Algorithm for the diagnosis of PJI as published by the work group of implant associated infections of the German Society for Arthroplasty (AE). Reproduced from ²³ with permission of the AE and translated into English by the author.

a) Examples of symptoms of periprosthetic infection, algorithm cannot display all possible symptoms;

b) Leukocyte count and differential in synovia, histopathology, microbiology;

c) Abdomen/pelvis/spine imaging, transesophageal echocardiography, orthopantomogram, urine analysis, chest x-ray, (individually or depending on the pathogen and affected joint);

d) not to be used in case of rheumatic arthropathy, fracture and within 6 weeks postoperatively;

e) in highly virulent pathogens (S. aureus, E. coli), detection in a single sample is sufficient; in low-virulent pathogens (S. epidermidis, Cutibacteriumacnes), detection in ≥2 samples is required;

f) Histology: Periprosthetic membrane Type II/III;

g) Optional peri/intraoperative diagnostics (second line) if: I) No elective diagnostics have been carried out, e.g. aseptic exchange with suspected intraoperative infection or II) inconclusive result despite algorithm. Be aware of the susceptibility of the test to interference, but it can be used individually as indicative. Susceptible to interference by erythrocyte content in the sample; centrifugation if necessary to optimize the test;

h) Tests for antimicrobial peptides (e.g. alpha-defensin) are gaining increasing importance, but the level of evidence of the available studies is currently not sufficient to recommend their use.

8. Conclusion

Within only a decade, we have seen the development of several definitions of prosthetic joint infection and even the amendment of some of them. It is safe to assume that the advances in molecular diagnostics will bring about new markers that will have to be included in future new or updated definitions. Any of the established sets of criteria, be it that of the MSIS, ICM or EBJI, are very accurate tools for the diagnosis of PJI in experienced hands, and the decision on which of them to use is as much based on the clinician's experience and training as on personal taste.

Equally important and difficult as the decision which definition to apply in clinical practice is the decision which diagnostic measures to perform and in what order. It is certainly neither economical nor practical to use every tool at our disposal on every patient with a failed knee arthroplasty. Furthermore, a false-positive result can obscure the real cause of failure and harm the patient with unnecessary surgical procedures and antibiotic treatment. The algorithmic approaches presented here do not free clinicians from applying their experience and sound judgment. They can, however, keep them from missing an important piece of information or forgetting to look for the infectious focus of an acute, hematogenous infection. Putting our diagnostic and treatment decisions on the sound basis of the established PJI definitions, a standardized workflow based on evidence and practicability, and our judgment and experience as surgeons will serve to improve our chances in the struggle against prosthetic joint infections.

Declaration of competing interest

Rüdiger von Eisenhart-Rothe ist part of the executive biard of the German Society for Arthroplasty (AE).