Erythrocyte Sedimentation Rate and C-Reactive Protein in Acute Inflammation

Meta-Analysis of Diagnostic Accuracy Studies

Ivana Lapić, MSc; Andrea Padoan, PhD; Dania Bozzato, MSc; Mario Plebani, MD


Am J Clin Pathol. 2019;153(1):14-29. 

In This Article


The findings of this meta-analysis demonstrate that ESR and CRP are widely used as inflammatory markers and that the diagnostic performance is widely variable among studies. We evidenced that both ESR and CRP are well researched in the diagnosis of orthopedic infections and that CRP does not have a superior diagnostic accuracy over ESR for this purpose. However, CRP was observed to have a higher diagnostic sensitivity and specificity than ESR in the diagnosis of various other inflammatory conditions, while the results of the few studies dealing with rheumatic conditions yielded different results, depending on the assessed outcome. The high heterogeneity of diagnostic accuracy for ESR is in part related to the analytical reliability and the lack of harmonization between methods used in clinical practice.

Specifically, the results of our meta-analysis show that pooled sensitivity and specificity for both ESR (0.78 and 0.68, respectively) and CRP (0.79 and 0.70, respectively), as well as SAUC (0.80 and 0.81, respectively), for the diagnosis of orthopedic infections are rather similar, indicating that neither of the two tests has superior diagnostic accuracy over the other for this purpose. Our results are in concordance with previously published meta-analyses[51–53] that dealt with diagnostic accuracy of ESR and CRP in the diagnosis of prosthetic joint infections. The reported pooled sensitivities and specificities for ESR ranged from 0.75 to 0.88 and 0.70 to 0.73, respectively. For CRP, pooled sensitivity and specificity values were slightly higher, namely, from 0.87 to 0.88 and 0.73 to 0.79, respectively.

However, the observed heterogeneity among the studies was high, which prompted us to evaluate more deeply the characteristics of these studies and reported diagnostic accuracy results. Among the 16 studies dealing with orthopedic infections, four[26,30,31,42] dealt with the use of ESR and CRP in the assessment of implant-associated infections. The reported diagnostic sensitivities and specificities are concordant among three studies[26,30,31] with sensitivities of ESR ranging from 0.822 to 0.945, specificities of ESR from 0.66 to 0.868, sensitivities of CRP from 0.913 to 0.951, and specificities of CRP from 0.755 to 0.83. Moreover, Piper et al[42] demonstrated that these values are variable and highly dependent of arthroplasty site (ie, obtaining the lowest sensitivities and specificities of both tests for shoulder in contrast to other arthroplasty sites). This was emphasized in the meta-analysis by van den Kieboom et al[54] that investigated the diagnostic accuracy of serum inflammatory markers in the diagnosis of late fracture-related infections. They reported a pooled sensitivity and specificity for ESR of 0.451 and 0.793, respectively, while for CRP, these parameters were 0.77 and 0.679, respectively. In addition, these authors also showed high heterogeneity among studies, attributing it to the complexity of the clinical presentation of fracture-related infections in terms of location, severity, and duration, as well as to the heterogeneous patient study population.

Osteomyelitis was used as a target condition in five studies[28,38–40,45] with highly variable sensitivity and specificity values for both index tests. This can be attributed to the complex pathophysiologic basis of different subtypes of osteomyelitis that inevitably affects not only the clinical presentation of the disease but also the response of acute inflammatory markers, making it inconsistent among different patients.[55] Similarly, the studies dealing with septic arthritis[25,27,36,44,48,49] yielded inconsistent and highly variable diagnostic accuracy data. This was also observed in the study by Carpenter et al,[56] which emphasized that neither of the two tests is useful in the diagnosis of acute septic arthritis. The latter study also emphasizes that the issue of spectrum bias is especially evident in diagnostic accuracy studies for septic arthritis due to disease complexity. It is further explained that the heterogeneity among the results obtained in different studies can be attributed to inclusion of patients with different levels of disease severity. While higher severity can cause falsely elevated sensitivity, a healthier population will falsely increase specificity.[56] All studies from the orthopedic group that reported diagnostic accuracy data when either of the two tests was elevated,[30,31,42,45] predominantly using the optimal cutoff value derived from receiver operating curve analysis, showed a higher sensitivity of either single test alone, thus suggesting that the parallel use of ESR and CRP can be an effective adjunct in the assessment of orthopedic infections. Moreover, a study of clinical effectiveness and cost-effectiveness of the use of ESR and CRP testing in combination or individually showed that the combined use of elevated ESR and CRP results in periprosthetic infections can efficiently decrease the number of false positives, compared with ESR and CRP alone, thereby avoiding misdiagnosis, unnecessary downstream diagnostic procedures, and associated costs. On the other hand, the approach when either of the two tests is positive can produce fewer false negatives than these tests alone, with significant cost savings per false-negative cases avoided. Interestingly, the latter approach can lead to more misdiagnosis compared with the single assessment of ESR and CRP.[57] Based on all these data, it is clear why the recently published consensus document about diagnosis of orthopedic infections[12] recommends the use of ESR and CRP as first-line screening tests in suspected orthopedic infections. Moreover, they state that the recommended cutoff values are 30 mm for ESR and 10 mg/L for CRP. As a result, we observed that not only are these the most commonly used discriminating values, but meta-analysis modeling with multiple cutoffs showed that the optimal discriminating values are very close to the recommended ones—namely, 31.2 mm and 12.9 mg/L—indicating that the use of cutoff values used in routine practice can provide the best diagnostic performance of the two tests.

The wide use of ESR and CRP in the assessment of inflammation is evident by a number of studies dealing with different target conditions, the group in our study called "others." Obviously, high heterogeneity among studies is expected due to different pathophysiologic conditions. Both pooled sensitivity and specificity were higher for CRP (0.86 and 0.67, respectively) than ESR (0.77 and 0.59, respectively). This also results in a higher SAUC for CRP (0.86) compared with ESR (0.75). Since this group encompasses a variety of clinical conditions, we could assume that an overall higher diagnostic accuracy of CRP could be a consequence of its faster response in acute inflammation, but more important, its shorter half-life is independent of serum concentration or pathophysiologic circumstances, thus resulting in a consistent CRP clearance.[15] Also, underlying inflammation in various pathophysiologic conditions might not be accurately reflected by ESR determination since ESR does not measure a well-defined analyte but rather a physicochemical phenomenon.[16] Furthermore, the variety of cutoffs applied throughout the studies indicates that the underlying inflammation largely depends on the primary pathophysiologic condition. However, cutoffs for ESR can also depend on the measurement principle used since different ESR methods reflect distinct phases of the RBC sedimentation process. The optimal cutoff for ESR was twofold higher than the median one (51.6 mm vs 25 mm), which might presume that low ESR levels cannot be accurately associated with inflammation in these conditions. On the other hand, the optimal cutoff for CRP was almost equal to the median of the used ones (20.2 mg/L vs 20 mg/L), showing that it is a better indicator of inflammation-associated conditions.

Since ESR and CRP are part of established diagnostic criteria for rheumatic diseases, not many studies assessed their diagnostic accuracy in this area. The three studies identified through our systematic search equally provide variable diagnostic accuracy data, making us unable to draw a conclusion on a more appropriate test to use. Interestingly, the study by Kermani et al[34] reported almost equal sensitivities (0.842 and 0.864, respectively) and specificities (0.292 and 0.305, respectively) for ESR and CRP, with slight but nonsignificant changes of these variables when both tests were above the cutoff value. Specifically, reported sensitivity was 0.808 while specificity was 0.412. This suggests that their combined determination might not be essential for the diagnosis of GCA. However, in terms of clinical effectiveness and cost-effectiveness, the combined use yielded eight fewer false positives per 100 patients tested compared with the single use of either ESR or CRP. Accordingly, the cost savings per false positives and misdiagnoses avoided are substantial.[57]

Our study has some limitations. First, our search was restricted to only two scientific databases (ie, PubMed and Scopus) and to English-language articles. This surely may result in publication searching bias. Second, we limited our analysis to acute inflammatory conditions, thereby excluding a large number of studies that evaluated the diagnostic accuracy of ESR and CRP in chronic inflammation, such as inflammatory bowel disease or Crohn disease. Third, the heterogeneity observed among studies regarding sample size, study design, patient population, inclusion criteria, and disease severity could have affected the generalizability of the obtained results. Fourth, the use of different analytical methods in the evaluated studies, encompassing the gold-standard Westergren method and various modified and alternate automated techniques for ESR measurement, could have, at least partly, affected the high heterogeneity of diagnostic accuracy for ESR. These methods differ by the type of sample used, methodologic principle, and phase of the erythrocyte sedimentation measured. Therefore, large differences about the clinical usefulness of the same method are reported in literature.[16] Despite significative efforts in the field, harmonization of ESR methods is still an unsolved issue, and the results obtained by different analytical techniques might not be completely comparable, thus implying the use of method-tailored reference ranges and clinical decision limits.[16,58] Also, included studies did not report the sampling time relative to the onset of inflammation, which could have surely affected the diagnostic accuracy data due to their different kinetics.

In conclusion, the results obtained through this meta-analysis point to the wide use and usefulness of both ESR and CRP in the evaluation of distinct acute inflammatory conditions. Although disconcordance may be observed in certain settings, our results prove that they can fairly equally contribute to assessment of inflammatory conditions, especially those associated with orthopedic conditions. The combined use of ESR and CRP enhances diagnostic accuracy, and thus ESR can serve as a valuable adjunct in the assessment of inflammatory conditions. Although ESR lately has been considered an obsolete laboratory test with questionable clinical utility due to its nonspecificity, its firm role in the diagnostic criteria of specific clinical conditions and its contribution to the diagnosis of inflammatory conditions, especially when coupled with CRP, prove that this is not a laboratory test of the past and can aid in clinical decision making when used appropriately and measured accurately. Finally, the harmonization of ESR measurement, thanks to analytical improvements and automation, may lead to improved diagnostic accuracy of an old test with a bright future.[58]