As noted by the authors of the PJI diagnostic criteria , clinical assessment and judgement remain of critical importance for the diagnosis of these infections. Two main classifications systems have been developed to assist in the assessment of patients. Zimmerli et al. classified PJI as “early” developing in the first 3 months after surgery, “delayed” occurring 3 to 24 months after surgery and “late” occurring greater than 24 months after joint replacement surgery [7]. A similar classification system was developed by Tsukayama et al., with PJI classified as “early postoperative” developing within 4 weeks of the index procedure, “late chronic” developing after 4 weeks of the index procedure with an insidious clinical presentation and “acute haematogenous” developing after 4 weeks of the index procedure with an acute onset of symptoms [13]. These classification systems loosely relate to symptomatology and likely pathogens associated with PJI and inform management strategies, particularly with the classification system developed by Tsukayama and colleagues (Table 3.3) [3, 13].

In early infections , patients typically present with surgical wound complications, such as purulent discharge, erythema and swelling of the affected joint [7, 14–16]. Early infections are typically associated with organisms such as S. aureus [7, 14–16]. In acute haematogenous infections, patients often report sudden onset of pain at the site of a prosthetic joint that has been previously asymptomatic, with or without swelling or fever [7, 14–16]. Organisms such as Streptococcus species and S. aureus are classic organisms associated with acute haematogenous PJIs [7, 14–16]. In late chronic infections, many of the typical symptoms of infection are absent. Pain is the predominant feature, with patients reporting a history of slowly increasing pain involving the prosthetic joint [7, 14–16]. A discharging sinus, when present, is typically associated with chronic, indolent presentations [2, 7, 13, 15]. These infections are classically associated with coagulase-negative staphylococci [16]. As previously discussed, C. acnes has a predilection for infecting shoulder joints and tends to cause indolent infections [17–20]. Classic features of infection are frequently absent, with pain and stiffness of the shoulder joint being the predominant symptoms. Bruising along the surgical wound has been described as a pathognomonic sign of C. acnes shoulder arthroplasty infection [17–20].

Regardless of which classification system is applied, classification of PJI influences management decisions; debridement and retention of the prosthesis is a potential treatment option for early and acute haematogenous infections but not for delayed or late chronic infections [1, 4, 7].

The timing of prosthesis failure also aids clinical acumen. In a prospective cohort study by Portillo and colleagues, the underlying indication for prosthesis revision was examined, comparing the median time for revision due to PJI, mechanical failure (such as dislocation) and aseptic failure. Overall, the need for revision within 2 years of the index procedure was highly predictive of PJI with a calculated risk ratio of 2.9 (95% confidence interval 1.8–4.8). The authors also noted that infections occurring after 2 years were more likely to be acute haematogenous infections and were easily clinically distinguishable from aseptic failure [21].

A number of studies have assessed the utility of blood markers in the diagnosis of PJI. Elevated blood white cell count is a poor predictor of infection; in most series, fewer than 10% of patients with PJI have an elevated peripheral blood white cell count [7, 8, 22]. In a systematic review and meta-analysis performed by Berbari and colleagues, the pooled sensitivity of an elevated peripheral blood white cell count was 45% (95% confidence interval 41–49%) [22]. Two studies published subsequent to the meta-analysis by Berbari et al. have demonstrated diverse results (Table 3.4) [1, 2, 12, 22–30]. Glehr and colleagues examined blood markers in patients undergoing hip or knee revision surgery [23]. The study included 124 revision arthroplasties, of which 78 met the original MSIS diagnostic criteria for PJI. Applying a cut-off of 7355 white cells/μL, the sensitivity was 73%, higher than the sensitivity reported in the meta-analysis. In contrast, in a second study by Randau and colleagues, applying a threshold of 10,300 cell/μL, the sensitivity of peripheral blood white cell count was only 21.3% (95% confidence interval 10.7–35.7%) [24]. Despite the low sensitivity, measurement of blood white cell count is commonly performed as part of standard preoperative workup.

Other biochemical tests, such as erythrocyte sedimentation rate (ESR) and serum C-reactive protein (CRP) , are more useful diagnostic investigations for PJI [1–4, 22]. These tests are inexpensive, widely available and commonly performed in this clinical setting; furthermore, clinicians are familiar with their interpretation [31]. In the study by Berbari et al., the pooled sensitivity of ESR was 75% (95% confidence interval 72–77%), and serum CRP was 88% (95% confidence interval 86–90%) for the detection of lower limb PJI [22]. In addition, studies have examined the sensitivity of combining ESR and serum CRP. In a study by Schinsky et al. examining 235 revision total hip arthroplasties, the combination of elevated ESR and serum CRP had a sensitivity of 90% and specificity of 91% [30]. Similarly, Austin et al. examined blood markers in 296 patients undergoing revision total knee arthroplasty; combining ESR and serum CRP had a sensitivity of 96%; however, the specificity was much lower at 56% [29]. As noted by Berbari et al., the cut-off values for these inflammatory markers differed in studies [22]. In addition, the thresholds for these markers differ between early and late chronic infections and by prosthesis location [26].

In attempt to provide clarity, the recent International Consensus Meeting specified cut-offs, distinguishing between acute PJI (defined as less than 90 days) and chronic PJI (defined as greater than 90 days). For CRP, the threshold for acute infections was 100 mg/L and 10 mg/L for chronic infections. For ESR, the threshold for chronic infections was 30 mm/h: no threshold was stipulated for acute infections, due to poor specificity of ESR in the early postoperative period [3, 4]. There is a paucity of data to support the thresholds suggested, and the data that does exist is conflicting. In the study by Alijanipour et al. examining 1962 patients undergoing total hip or knee revision arthroplasties, 273 of whom underwent revision for PJI, the inflammatory markers differed between early and late chronic infections with higher values noted in early infections [26]. In addition, the median CRP was higher in knee PJI (133, interquartile range [IQR] 40–207) compared to hip PJI (84, IQR 55–101; P < 0.0001). For late chronic knee infections, the optimal cut-off values based on receiver operating characteristic (ROC) analysis for CRP was 23.5 mg/L with a sensitivity of 92% and specificity of 94% and for ESR 46.5 mm/h (sensitivity 87% and specificity 87%) [26]. In late chronic hip infections, Alijanipour and colleagues suggested a cut-off of 13.5 mg/L for CRP (sensitivity 90% and specificity 88%); however, the threshold suggested for ESR (48.5 mm/h) was associated with reduced sensitivity—78% [26]. For early infections, Alijanipour et al. suggested a threshold of 54.5 mm/h for ESR (sensitivity 80% and specificity of 93%) and 23.5 mg/L for CRP (sensitivity 87% and specificity 94%) for both hips and knee joints [26].

In contrast, studies by Bedair et al. suggested an optimal CRP threshold of 166 mg/L for early postoperative knee infections (sensitivity of 53%, 95% confidence interval 43–62%, and specificity 86%, 95% confidence interval 79–92%). The ESR threshold of 120 mm/h was associated with a sensitivity of 16% (95% confidence interval 9–23%) and specificity of 94%, (95% confidence interval 90–99%) [27].

In patients undergoing hip revision surgery for early postoperative infections, Yi et al. suggested optimal cut-offs for CRP of 93 mg/L for hips (sensitivity 88%, 95% confidence interval 77–98%) [28]. Similar to the observations of Bedair et al., the optimal cut-off for ESR was 44 mm/h with excellent sensitivity (92%, 95% confidence interval 83–100%) but low specificity (53%, 95% confidence interval 38–69%) [28]. The observed poor specificity with ESR in acute infections may relate to the fact that ESR is elevated in the postoperative period in uncomplicated joint surgery. Indeed, in an early study by Shih et al., ESR remained elevated out to a year in a small subset of patients [32]. In contrast, CRP peaks on day 2 and returns to normal within 3 weeks [31–34]. In addition, there may be potential overlap with other conditions that elevate inflammatory markers, such as rheumatoid arthritis [22]. Yi and colleagues also observed that combining ESR and CRP did not improve test performance in the acute setting [28]. Of further interest, in a study by Bracken et al., which compared CRP and ESR in 44 patients with fungal PJI compared to 59 patients with bacterial PJI, it was noted that these inflammatory markers did not differ between fungal and bacterial PJI [35].

These measures are of limited clinical benefit in predicting shoulder arthroplasty infections ; in a study by Piper et al., applying a cut-off of 26 mm/h for ESR and 7 mg/L for CRP, these markers had a sensitivity of 32% for ESR and 63% for CRP and a specificity of 93% for ESR and 73% for CRP [12]. In 23% of shoulder infections, the ESR and CRP were not elevated [12]. Similarly, in a study by Topolski et al., ESR (applying a cut-off of 22 mm/h) and CRP (applying a cut-off of 1 dL/L) were only elevated in 14% and 25% of patients, respectively [36].

The spectrum of reported test performance of these serum biomarkers is outlined in Table 3.4. In addition, Table 3.4 also highlights the range of definitions applied for PJI, thereby hampering the direct comparison of results between studies. Notwithstanding, ESR and CRP are relatively useful tests, given their broad availability and reasonable sensitivity; however, the optimal cut-offs according to chronicity of infection and location of the index prosthetic joint still require further study. In addition, the specificity of ESR is poor, particularly in early postoperative infections. As previously discussed, elevated ESR and CRP are minor criteria for the MSIS and revised MSIS diagnostic criteria [2, 3]. Of interest, in the diagnostic algorithm proposed by the American Academy of Orthopaedic Surgeons (Fig. 3.1), CRP and ESR are suggested as initial tests to exclude the diagnosis of PJI; however, given the variation in reported sensitivity, specificity and predictive value of these tests, the validity of this is unclear [37].

Other studied serum biochemical markers of infection include interleukin-6 , tumour necrosis factor-α (TNF-α) and procalcitonin C (Table 3.5) [1, 2, 22–25, 38, 39]. Interleukin-6 and TNF-α are cytokines released by monocytes and macrophages in the setting of infection [38]. Procalcitonin C is a precursor of calcitonin that has been shown to be a specific marker for a number of bacterial infections, including pneumonia and sepsis [40]. Bottner et al. compared the diagnostic utility of interleukin-6, TNF-α and procalcitonin C in 78 patients undergoing revision arthroplasties [38]. According to this study, patients were diagnosed with infection based on their having positive cultures or histological evidence of infection; however, details as to the exact aspects of these criteria were not further specified [38]. The authors noted that these serum biomarkers were all significantly elevated in patients meeting the above definition of infection, compared to patients with “aseptic loosening”. The observed sensitivity and specificity of interleukin-6 were 95% and 87%, respectively [38]. Bottner et al. also noted that in six of the seven patients with false-positive interleukin-6 results, there was associated polyethylene wear and osteolysis, which may limit its utility [38].

Interleukin-6 is elevated in the postoperative period for primary arthroplasty ; however, it returns to normal levels within 2 days of the operation [38]. Therefore, there is potential diagnostic utility of interleukin-6 over CRP and ESR in the early postoperative period if infection is suspected [41, 42]. In the meta-analysis by Berbari and colleagues, the pooled sensitivity and specificity for interleukin-6 was 97% (95% confidence interval 93–99%) with a pooled specificity of 91% (95% confidence interval 87–94%) [22]. In addition to the study by Bottner et al., the meta-analysis included two other small studies [22]. In a more recent study by Glehr et al. which was not included in the meta-analysis, preoperative performance of interleukin-6 did not predict infection [23]. Applying an optimal cut-off of 4.7 pg/mL, interleukin-6 had a sensitivity of 81% and specificity of 68%. Similar sensitivities have been reported in other studies [24, 25]. Interleukin-6 is not routinely available in most commercial laboratories, and the applicability and value of this marker remain unclear [31].

Five conflicting studies have examined procalcitonin in the setting of PJI; in the aforementioned study by Bottner et al., the sensitivity was 33% and specificity was 98% [38]. In contrast, in a study by Yuan et al. recruiting 71 patients undergoing revision hip arthroplasty, applying a cut-off of 0.5 ng/mL, the sensitivity of procalcitonin C was 80.0% and specificity was 73.9% [39]. As in the study by Bottner et al., the study by Glehr et al. demonstrated poor sensitivity of procalcitonin (48%); however, the specificity was 100% when a cut-off of 0.75 ng/mL was applied [23]. In the study by Randau and colleagues, procalcitonin performance was very poor (12.9%; 95% confidence interval 3.6–29.8%) when a higher threshold of 46 ng/mL was applied [40]. All studies noted no superiority of procalcitonin over CRP: Bottner et al. also noted that the cost of procalcitonin was 24 times higher than CRP [23, 38, 39]. Therefore, the value of procalcitonin in this clinical setting has not been established [24]. Finally Bottner et al. also examined the test characteristics of tumour necrosis factor-α, reporting a sensitivity of 43% and specificity of 94% applying a cut-off of 40.0 ng/mL [38]. These markers, interleukin-6, tumour necrosis factor-α and procalcitonin C, have not been incorporated into any diagnostic algorithm.