636 subjects who underwent knee (n
297), hip (n
221) or shoulder (n
64) arthroplasty, or spine implant (n
54) removal were analyzed. 487 subjects with knee (n
215), hip (n
187), or shoulder (n
45) arthroplasties and spine implants (n
40) removed met the preoperative definition of aseptic failure. 149 subjects who underwent removal of knee (n
82), hip (n
34), or shoulder (n
19) arthroplasties or spine implants (n
14) met the preoperative definition of orthopedic implant-associated infection.
CRP was significantly different in subjects with aseptic failure and infection of knee (median 4 and 51 mg/l, respectively, p<0.0001), hip (median 3 and 18 mg/l, respectively, p<0.0001), and shoulder (median 3 and 10 mg/l, respectively, p
0.01) arthroplasties, and spine implants (median 3 and 20 mg/l, respectively, p
0.0011). ESR was significantly different in subjects with aseptic failure and infection of knee (median 11 and 53.5 mm/h, respectively, p<0.0001) and hip (median 11 and 30 mm/h, respectively, p<0.0001) arthroplasties, and spine implants (median 10 and 48.5 mm/h, respectively, p
0.0033), but not shoulder arthroplasties (median 10 and 9 mm/h, respectively, p
Descriptive summary and comparison of aseptic failure versus orthopedic implant-associated infection subjects. Median (range) values are shown.
The sensitivities, specificities, positive and negative predictive values, and the p-value from logistic regression of CRP >10 mg/l, ESR >30 mm/h, and CRP >10 mg/l or ESR >30 mm/h to detect infection of knee, hip, and shoulder arthroplasties, and spinal implants are shown in . The combination of normal ESR (≤30 mm/h) and CRP (≤10 mg/l) predicted the absence of infection in 94, 94 and 91% of subjects undergoing knee or hip arthroplasty or spine implant removal, respectively, but only 77% of those subjects undergoing shoulder arthroplasty removal.
Sensitivity and specificity of CRP (>10 mg/l) and/or ESR (>30 mm/h) for the detection of infected knee, hip and shoulder arthroplasty and spinal instrumentation.
Optimized ESR cutoffs for knee, hip and shoulder arthroplasties and spine implants were 19, 13, 26, and 45 mm/h, respectively. Using these cutoffs, sensitivity and specificity to detect infection were 89 and 74% for knee, 82 and 60% for hip, and 32 and 93% for shoulder arthroplasties, and 57 and 90% for spine implants ().
Sensitivity and specificity of optimized CRP and ESR for the detection of infected knee, hip and shoulder arthroplasty and spinal instrumentation.
Optimized CRP cutoffs for knee, hip, and shoulder arthroplasties, and spine implants were 14.5, 10.3, 7, and 4.6 mg/l, respectively. Using these cutoffs, sensitivity and specificity to detect infection were for 79 and 88% knee, 74 and 79% for hip, and 63 and 73% for shoulder arthroplasties, and 79 and 68% for spine implants ().
Our study is a comprehensive analysis of CRP and ESR in subjects undergoing orthopedic implant removal with and without infection. In patients satisfying the definition for orthopedic implant-associated infection, CRP and ESR values were higher in knee arthroplasty and spine implant patients than in hip arthroplasty patients. Previous investigations have reported higherCRP and ESR values in knee than hip arthroplasty patients with infection 
We used receiver operating curve analysis to optimize CRP and ESR cutoffs. The optimized CRP cutoff value for hip arthroplasty infection was similar to the standard cutoff of CRP >10 mg/l often used in clinical practice. Optimized CRP and ESR cutoff values for knee arthroplasty in our study were >14.5 mg/l and >19 mm/h, similar to the ≥13.5 mg/l and ≥22.5 mm/h values derived by Greidenaus et al. using a similar approach 
Overall, CRP and ESR showed the lowest sensitivity for diagnosis of shoulder arthroplasty infection, even applying cutoffs optimized using receiving operating curve analysis. This may relate to the predominance of P. acnes
in shoulder arthroplasty infection 
Several investigators have examined the natural history of post-operative ESR and CRP after uncomplicated arthroplasty. CRP levels change more rapidly than ESR levels, and return to normal more rapidly following primary total knee or hip arthroplasty 
. CRP levels usually peak on the second or third day following total hip or knee arthroplasty 
, thereafter dropping to preoperative levels by the third week in total hip arthroplasty patients and by the end of the second month in total knee arthroplasty patients 
. CRP levels rise to a higher level postoperatively in total knee than hip arthroplasty patients 
. A similar rise and fall of CRP is noted postoperatively in total hip or knee arthroplasty patients with underlying rheumatoid arthritis 
. One study demonstrated that, after uncomplicated arthroplasty, ESR peaks on the fifth postoperative day 
, dropping close to preoperative levels at the end of the third month in total hip arthroplasty patients, and at the end of the ninth month in total knee arthroplasty patients 
. Other studies suggest, however, that the ESR, although usually normal by six months postoperatively, may be elevated for as long as one year after uncomplicated total hip arthroplasty 
For patients undergoing uncomplicated spinal surgery, investigators have described ESR peaking earlier than described above, on the fourth day, and normalizing within a two week period in the majority of patients 
. Additionally, levels in patients undergoing fusion surgery were higher than in those patients undergoing herniated disc removal 
. In patients with known vertebral osteomyelitis the ESR may be elevated for a prolonged period of time, even in the face of appropriate non-operative treatment 
. In the referenced series, many patients went on to a successful clinical outcome in spite of the elevated ESR, illustrating the poor specificity of the test when used alone to predict treatment failure 
. The current study differs in that the ESR was used in a predictive fashion preoperatively, in conjunction with the CRP, and a consistent definition of infection was applied.
In a series of 202 revision total hip arthroplasties, all subjects with infection had an elevated ESR (>30 mm/h) or CRP (>10 mg/l) 
. In our study, this was not the case.
There are several limitations of our study. ESR and CRP are nonspecific markers of inflammation and may be elevated by chronic inflammatory conditions (e.g., rheumatoid arthritis), surgical intervention, or systemic illness. Patients with underlying inflammatory arthritides were excluded from our study, but we did not assess for recent surgeries or systemic illnesses not involving the joint. We assessed ESR and CRP within the month prior to surgery, at a time when the study subjects had symptoms related to implant failure. Ideally, the timing of ESR and CRP measurement should be standardized (e.g., 24 hours prior to surgery). A final limitation is the lack of a Gold standard definition for prosthetic shoulder and spine implant infection.
In conclusion, CRP and ESR values are higher in knee arthroplasty and spine implant patients than in hip arthroplasty patients with infection, and show the lowest sensitivity for diagnosis of shoulder arthroplasty infection, even applying cutoffs optimized using receiving operating curve analysis.
Presented in part at the Musculoskeletal Infection Society Annual Meeting and 19th Open Scientific Meeting, Aug 7–8, 2009, San Diego, California and at the 29th Annual Interscience Conference on Antimicrobial Agents and Chemotherapy/Infectious Diseases Society of America 46th Annual Meeting, October 25–28, 2008.