This prospective study evaluates the role of new laboratory markers in the diagnosis of deep implant infection in 78 patients (41 men and 37 women) with a revision total knee or hip replacement.

The mean age at the time of operation was 64.0 years (19 to 90). Intra-operative cultures showed that 21 patients had a septic and 57 an aseptic total joint replacement. The white blood cell count, the erythrocyte sedimentation rate and levels of C-reactive protein, interleukin-6, procalcitonin and tumour necrosis factor (TNF)-α were measured in blood samples before operation. The diagnostic cut-off values were determined by Received Operating Characteristic curve analysis.

C-reactive protein (> 3.2 md/dl) and interleukin-6 (> 12 pg/ml) have the highest sensitivity (0.95). Interleukin-6 is less specific than C-reactive protein (0.87 vs 0.96). Combining C-reactive protein and interleukin-6 identifies all patients with deep infection of the implant. Procalcitonin (> 0.3 ng/ml) and TNF-α (> 40 ng/ml) are very specific (0.98 vs 0.94) but have a low sensitivity (0.33 vs 0.43).

The combination of C-reactive protein and interleukin-6 measurement provide excellent screening tests for infection of a deep implant. A highly specific marker such as procalcitonin and pre-operative aspiration of the joint might be useful in identifying patients with true positive C-reactive protein and/or interleukin-6 levels.

The clinical picture of delayed deep infection of an implant (type 2 according to Coventry1) is often similar to that of aseptic failure.1,2 In every patient with late implant failure, deep infection needs to be excluded. The white blood cell count (WBC), erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) levels are not able to identify all patients with deep infection.3 Chronic encapsulated infections, low-grade infections or infections with a fistula may result in less intensive systemic reactions and are sometimes associated with normal laboratory markers.4 Elevated CRP levels have also been recorded in patients with rheumatoid arthritis, obesity and in those of lower socioeconomic status.5,6 In the recent literature it has been suggested that procalcitonin and interleukin-6 are markers for bacterial infections.710 Procalcitonin is a precursor of calcitonin and elevated levels have been noted in patients with systemic bacterial infections.710 Interleukin-6 and tumour necrosis factor (TNF)-α are released by monocytes in response to a local infection. Both markers trigger the release of CRP in the liver. The diagnostic value of interleukin-6 in children with bacterial infections has been confirmed.10 Although both interleukin-6 and TNF-α seem to have a similar role in indicating a bacterial infection, little is known about TNF-α as a diagnostic marker for this purpose.

This study evaluates the diagnostic value of the WBC count, ESR, CRP, interleukin-6, TNF-α and procalcitonin for the diagnosis of deep infection of implants in patients with failed total joint replacement.

Materials and Methods

Between July 2003 and March 2004, 78 patients having a revision of a total hip or knee replacement were included in this prospective study. They were grouped into septic and aseptic revisions based on the findings of intra-operative culture and histology. At operation, at least two tissue samples were taken from the synovial tissue and the interface membrane; these were sent for routine aerobic and anaerobic bacterial cultures. The tissue samples were cultured for 24 to 48 hours (standard culture) and ten days (long-term culture). In three patients Staphylococcus capitis grew in one of the two long-term cultures. None had histological evidence of infection or a positive pre-operative joint aspiration11 and therefore the samples were considered contaminated and the patients were included in the aseptic revision group. There were 57 patients in this group.

A total of 21 patients had a positive culture and histological evidence of deep infection and were therefore included in the septic group (Table I).11 Of these, 16 had pre-operative joint aspirations with more than 20 000 white blood cells per ml and the remaining five had a fistula. Among these 21 patients, the standard cultures for seven patients were negative and the causative organisms was discovered in the long-term cultures only. In one patient an additional organism was found in the long-term culture while in 13 patients the organisms identified in both the standard cultures and long-term cultures were identical. Overall, 21 patients (11 men and 10 women) had a revision for an infected total joint replacement (Table II). The mean age at the time of surgery was 63.5 years (19 to 90) and was not significantly different to that of patients undergoing revision of an aseptic arthroplasty (p = 0.68). A total of 16 patients (76.2%) had a septic hip replacement and five patients (23.8%) a septic total knee replacement. In the septic revision group, 12 patients (57%) had radiological evidence of loosening of the implant and a further six (28.5%) had scintigraphic evidence of loosening. Four patients (19%) had signs of polyethylene wear on the pre-operative radiographs and in three (14%) the radiological appearances were normal. Two required haemodialysis, and three had a history of a tumour. Five (23.8%) had systemic inflammatory disease, including two with ankylosing spondylitis and three with primary chronic polyarthritis, the latter requiring systemic immunosuppressive medication. Four patients (19%) had diabetes mellitus, of whom three were treated with insulin. Four (19%) smoked more than ten cigarettes a day. No patient drank alcohol on a regular basis.

View this table:
Table I.

The organisms identified in the 21 patients with septic revision total joint replacement

View this table:
Table II.

The number of previous operations including total joint replacement, the body mass index and the time between the first symptoms and the revision procedure

The mean age of the 57 patients undergoing aseptic revision was 64.8 years (32 to 87). There were 30 men and 27 women; 34 (60%) had revision of a total hip replacement and 23 (40%) of a total knee replacement. Radiological evidence of loosening was present in 42 patients (74%) and in seven (12%) there was scintigraphic evidence of this. The remaining eight patients (14%) had revision total joint replacement for pain (8), instability (5), implant malposition (5) or a combination of the three. Radiological evidence of polyethylene wear was seen in 16 patients (28%). One patient had a hepatitis C infection and another had elevated liver enzymes of unknown aetiology. Two (3.6%) had neoplastic disease, and nine (16%) had diabetes mellitus but only one required insulin. Five patients (8.7%) had a systemic inflammatory disease including two with gout, one with ankylosing spondylitis and two with primary chronic polyarthritis. Five patients (8.7%) smoked more than ten cigarettes a day. No patient drank alcohol on a regular basis.

On the day before surgery, the WBC count, ESR, CRP, interleukin-6, procalcitonin and TNF-α levels were measured within 30 minutes of taking the blood samples (Table III).

View this table:
Table III.

Sample material, analysing technique and equipment, transport time and time required for analysis and the cost of the analysing agents

The diagnostic cut off levels were determined using the Receiver Operating Characteristic (ROC) curve analysis of MedCalc Software (Medcalc, Mariakerke, Belgium). This provides the level with the highest sensitivity and specificity for each marker examined. Based on ROC analysis of the data, the cut-off levels were WBC > 6 200/μl, ESR > 32 mm/hour, CRP > 3.2 mg/dl, interleukin-6 > 12 pg/ml, procalcitonin > 0.3 ng/ml and TNF-α > 40 nl/ml. The sensitivity, specificity, positive predictive value, negative predictive value and accuracy were calculated using MedCalc Software.

Laboratory data in the septic and aseptic group were compared using a t-test. A p-value < 0.05 was considered significant. The correlation between the body mass index (BMI) and the laboratory marker was assessed using the Pearson correlation coefficient.


The ESR and CRP levels were significantly higher in patients with septic revision arthroplasty (p < 0.0001) (Table IV). There was no difference in the WBC count (0.086). Procalcitonin (p = 0.0033), interleukin-6 (p = 0.0001) and TNF-α (p = 0.0011) were significantly increased in patients with a septic joint (Table V; Fig. 1).

View this table:
Table IV.

Comparison of the erythrocyte sedimentation rate (ESR), the C-reactive protein level and the white blood cell count between patients with a septic and aseptic revision total joint replacement

View this table:
Table V.

Comparison of, interleukin-6, procalcitonin, and tumour necrosis factor (TNF)-α in patients with septic and aseptic revision total joint replacements

Fig. 1

Boxplot of white blood cell count (WBC) (1000/μl), erythrocyte sedimentation rate (ESR) (mm/hour), C-reactive protein (CRP) (mg/dl), inter-leukin-6 (IL6) (pg/ml), procalcitonin (PCT) (ng/ml) and tumour necrosis factor (TNF)-α (ng/ml) in patients with septic (grey box) and aseptic (white box) revision total joint replacement.

The CRP and interleukin-6 had the highest sensitivity (0.95). CRP had a higher specificity (0.96 vs 0.87) and accuracy (0.96 vs 0.89) than interleukin-6 (Table IV). The WBC count (0.70), procalcitonin (0.33), TNF-α (0.43), and the ESR (0.81) had a low sensitivity. Seven patients with an aseptic revision arthroplasty had elevated interleukin-6 levels. All but one patient with procalcitonin levels of greater than 0.3 ng/dl had a deep infection (Table IV). There was no correlation between the type of organism and any of the inflammatory markers evaluated in the study.

In the aseptic patients there was no difference in the ESR, CRP, TNF-α or procalcitonin levels between patients with and without radiological evidence of polyethylene wear and/or osteolysis (p > 0.10). However, there was a trend towards higher interleukin-6 levels in patients with polyethylene wear and osteolysis (7.2 vs 19.2 pg/ml; p = 0.09). Based on the means and standard deviations in the current study, an alpha of 0.05 and a power of 0.80 more than 139 patients are needed in each group to show a statistically significant difference. Five of 16 patients with radiological evidence of polyethylene wear/osteolysis in the aseptic group had an increase of interleukin-6 above 12 pg/ml without positive cultures. The presence of radiological evidence of loosening had no effect on any of the serum markers (p > 0.10).

There was no correlation between the BMI and the CRP (Pearson correlation coefficient (r) = 0.02), procalcitonin (r = 0.11), interleukin-6 (r = 0.13) and TNF-α (r = 0.15).


The sensitivity and specificity for the WBC and ESR are low.3 Although CRP is a more accurate marker it is not able to identify all patients with deep infection of an implant.3

Most patients with deep infections of implants have normal WBC counts.12,13 In this study, 14 of 21 patients (67%) with deep infection had normal WBC counts (≤ 6300 μl). The low diagnostic value of the WBC count is confirmed in the literature14 and is not helpful in the diagnosis of deep implant infection. Spangehl et al14 evaluated 178 patients with 202 hip revisions and suggested that a combination of ESR and CRP would have the highest diagnostic accuracy. This view was supported by Sanzen and Sundberg,15 who assessed 23 patients with low grade total hip infections. Of six patients with normal CRP levels, five had an elevated ESR.15 The study suggested that the ESR is an important marker for the diagnosis of low grade infection, but since it did not evaluate the ESR in patients without deep implant infection the overall importance of the ESR remains unclear. The current information suggests that the ESR has little value in the diagnosis of infection since its sensitivity and specificity are low (Table VI).

View this table:
Table VI.

Sensitivity, specificity, positive predictive value, negative predictive value and accuracy of erythrocyte sedimentation rate (ESR) within the first 60 minutes, C-reactive protein level (CRP), interleukin-6, procalcitonin and tumour necrosis factor (TNF)-α. Cut-off levels are based on receiver operating curve analysis. For combining CRP and interleukin-6 (*): patients were considered septic if either CRP is greater than 3.5 mg/dl or IL-6 levels are greater than 12 pg/dl

The CRP has been thought to be the most accurate laboratory marker.14,16 In Spangehl et al’s14 study, it had a sensitivity of 0.96 and a specificity of 0.92, which is comparable with our data using 1.5 mg/dl as a cut off for infection (Table VI). Using a cut off of 3.2 mg/dl increased the specificity to 0.96 (Table VI). However, the CRP is not considered a perfect screening marker for deep infection since its sensitivity is not 100% and therefore some patients with an infected implant might present with normal CRP values. Increasing the currently-used cut-off levels for CRP will improve the diagnostic accuracy.

Although the sensitivity (0.95) of interleukin-6 is similar to CRP, its specificity is lower (0.87). This can be explained by the observation that increased interleukin-6 levels are found in some patients with aseptic loosening of a total hip replacement.17,18 Monocytes respond to polyethylene particles by secreting interleukin-6, and high concentrations of interleukin-6 have been found in the interface membrane surrounding loosened implants.19,20 Although the concentration of interleukin-6 in the peripheral blood increases after total joint replacement, clinical studies did not show a correlation between failure of an aseptic implant and increased levels of interleukin-6.21,22 In our study seven of 57 patients (12%) with aseptic failure had an increase in interleukin-6, which is therefore not a marker for aseptic loosening. However, six of seven patients with increased levels of interleukin-6 had evidence of polyethylene wear and/or osteolysis. Although the current study is under-powered to make a definite statement, patients with polyethylene wear/osteolysis have a trend towards higher interleukin-6 levels. Of 21 patients with deep infection, 20 (95%) had increased levels of interleukin-6. Interleukin-6 levels above 12 pg/ml and/or CRP levels above 3.2 mg/dl will identify all patients with deep infection and the combination is an excellent screening test to identify all such patients. To identify patients with true positive CRP or interleukin-6 elevation we recommend markers with a high specificity. We found that procalcitonin has a specificity of 0.98 and therefore deep infection is likely in patients with increased procalcitonin levels. Pre-operative joint aspiration is also helpful in patients with increased interleukin-6 or CRP levels.

It is likely that interleukin-6 will respond faster to the eradication of deep infection than the CRP. Further studies are needed to determine the role of interleukin-6 and CRP in the evaluation of persistent infection in the interval between explantation and re-implantation during two-stage revisions.

Procalcitonin had a high specificity in our study. Only one patient with aseptic loosening, but seven of 21 patients with sepsis, had elevation of the levels of procalcitonin. A number of studies have shown that procalcitonin is a very specific marker for the diagnosis of infection,79 but they all focus on the use of procalcitonin in systemic infection in the newborn. Procalcitonin might also be an interesting marker for the diagnosis of infection in patients with rheumatoid arthritis or other inflammatory conditions in which the CRP is less reliable. Schwenger et al23 were able to show that patients with Wegner granulomatosis and a bacterial infection have higher levels of procalcitonin than those with active Wegner granulomatosis alone (1.3 ng/ml vs 0.19 ng/ml). Scott et al24 found levels of procalcitonin of 2 ng/ml or more in patients with sickle cell disease to be suggestive of serious bacterial infection. Procalcitonin may be useful in further assessment of patients with a positive CRP and elevated levels of interleukin-6.

There is no information in the current literature on the value of TNF-α as a diagnostic marker for infection. We found that TNF-α is not as sensitive as CRP and interleukin-6 and not as specific as procalcitonin. It is unstable and the sample needs to be processed within 60 minutes of being taken. It also requires more than two hours to process each sample (Table III). Thus, it is not as useful in assessing patients with a failed total joint replacement.

The CRP and interleukin-6 are excellent screening markers to rule out deep infection of an implant. For CRP, a cutoff value of more than 3.2 mg/dl is more specific for deep infection than a level of 1.5 mg/dl. If patients with either increased CRP levels or increased interleukin-6 levels are considered infected, all patients with a deep implant infection are identified (sensitivity: 1.00). To avoid false positive results, estimation of the level of procalcitonin and pre-operative joint aspiration are recommended in patients with elevated CRP or interleukin-6. The ESR and TNF-α lack both sensitivity and specificity and are not helpful.


  • No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article.

  • Received November 29, 2005.
  • Accepted August 23, 2006.


View Abstract