The American Academy of Orthopaedic Surgeons (AAOS) has issued a strong recommendation for ESR and CRP measurement in assessing prosthetic joint infection [9]. This is based on the evaluation of several level I studies including data for periprosthetic infection in TKR [2, 5–7, 10].

Della Valle et al. [5] assessed 105 patients with painful TKR. Mean ESR was reported at 79.6 mm/h in the group subsequently diagnosed with infection. This was significantly greater (p < 0.001) than the noninfected group (mean 26.3). However, the ESR did range from 4 to 140 in patients with infection. Similarly, the CRP was measured at a mean of 123 in the infected group, again significantly greater (p < 0.001) than the noninfected group (mean 7.4). The range of CRP in infected cases was from 2 to 464. The combination of CRP and ESR was highly sensitive and specific (Table 26.1) with only one case of infection when both were normal.

Greidanus et al. [6] reviewed 151 knees in 145 patients presenting for revision TKR. A diagnosis of infection was established for 45 knees. Receiver operating characteristic (ROC) curve analysis was performed to determine the positivity criteria for the tests. By this method, thresholds of 22.5 mm/h for the ESR and 13.5 mg/L for the CRP were established. None of the patients included had been on antibiotic therapy during serum testing. With the use of these criteria, high levels of sensitivity and specificity were established (Table 26.1). Further, if the ESR/CRP tests were administered simultaneously and a positive result of either test was considered indicative of infection, the sensitivity was increased to 95 % although specificity decreased to 77 %. Conversely, if infection was only considered when both tests were positive, the sensitivity was lower at 88 % but specificity improved to 93 %.

It is important to understand the implications of negative and positive predictive values when reviewing the ESR and CRP in a particular case. In clinical terms, a more sensitive test may be used to rule out pathology while a more specific test may be used to increase the chance that pain is related to infection. From the study of Greidanus et al. [6], the negative predictive value of either test is 96 %, meaning there is only a 4 % chance that a symptomatic TKR is infected if both the ESR and CRP are <22.5 mm/h and 13.5 mg/L, respectively. If infection is only considered to be present if both tests are positive, the specificity is 93 % and the positive predictive value, indicating probability of infection, is 84 %.

The above cutoff points for the ESR assume a manual measurement of the erythrocyte sedimentation rate using the method of Westergren. More recently, automated methods of measuring the ESR have been introduced. Using these automated methods, the reference ranges depend on the anticoagulants used in the automated test, and occasionally, values under 11 using the Westergren method may be highly abnormal with this technique. In one study [11], the correlation was good with R² = 0.72; however, 12.5 % of patients (25/200) would have been treated differently if the automated method was relied upon. The surgeon should enquire what technique of measurement was used, and caution should be exercised if it was the automated method, particularly when the results are abnormal.