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

Materials and Methods

Search Strategy and Data Sources

A systematic electronic search was conducted via MEDLINE (PubMed) and Scopus up to February 2019, with no date and study size restriction, to retrieve studies that assessed the diagnostic accuracy of both ESR and CRP as index tests in an adult or a pediatric population for the diagnosis of acute location-confined or systemic inflammation of mild or moderate degree. The search was conducted using controlled vocabulary (ie, the National Library of Medicine's Medical Subject Headings). The terms searched were "erythrocyte sedimentation rate AND C-reactive protein."

First, two reviewers (I.L. and A.P.) independently performed screening of titles and abstracts of all studies resulting from the search to identify potentially relevant articles. Afterward, the two authors reviewed the full texts of the remaining articles to assess if they were eligible for the study and finally selected articles qualified for inclusion. The selected articles were thoroughly reviewed by both authors and the final decision on article inclusion was reached mutually. Any disagreement concerning study eligibility or data interpretation was resolved through consensus or, if required, by consulting a senior author (M.P.).

Selection Criteria

Studies were included if they assessed the accuracy of ESR and CRP for the diagnosis of inflammation that was medically confirmed by either clinical criteria and/or diagnostic tests (imaging techniques or microbiologic findings). However, studies evaluating different clinical outcomes were considered. Articles were excluded if any of the following conditions was met: (1) the study did not address the research question, (2) it was not a diagnostic accuracy study, (3) it was not in English, (4) it was not an original article, (5) not all data required for the study were available, (6) no full text was available online from archives at our university library or other sources such as Researchgate, and (7) the study used biological matrices other than blood. To be eligible, studies had to have a well-defined reference standard for the confirmation of the respective inflammatory condition, as well as report diagnostic accuracy data (diagnostic sensitivity and specificity) or allow their calculation from the number of true-positive, false-positive, false-negative, and true-negative cases using a 2 × 2 contingency table. Data from all included studies were extracted into an electronic file and included authors, publication year, setting, target population and target condition, type of study, outcome of the study, index tests and reference standard, cutoff values for ESR and CRP, total number of participants included in the study as well as the number of those with negative or positive outcome, diagnostic sensitivity and specificity for ESR and CRP, and the combined use of ESR and CRP, where available. Studies included in this meta-analysis were both retrospective and prospective, with either case-control or cohort design, or cross-sectional. The included studies were further grouped into supercategories, based on the main outcome of the study (ie, orthopedic infections, rheumatic diseases, and others). The latter included all studies dealing with various outcomes that do not enter in either of the first two categories but still evaluate the diagnostic accuracy of the index tests in different inflammatory conditions. In the selected studies, CRP was determined by either immunoturbidimetric or immunonephelometric assays, while for ESR, the gold-standard Westergren method or various automated techniques were used.

Quality Assessment

Articles were screened for methodologic quality by two independent reviewers (I.L. and A.P.) according to the Quality Assessment of the Diagnostic Accuracy Studies–Revised criteria (QUADAS-2), a structured questionnaire used to evaluate the quality of primary diagnostic accuracy studies. It consists of four key domains that assess patient selection, index test, reference standard, flow and timing of the given study in terms of risk of bias, and concerns regarding applicability. These two judgments are estimated through multiple-choice signaling questions that can be scored as "yes," "no," or "unclear," and based on it, the final decision on risk of bias and concerns regarding applicability can be defined as "high," "low," or "unclear."[19] We used the Access database available online ( for data extraction and study evaluation. Any disagreements between the two reviewers were resolved by discussion or consultation with a third reviewer (M.P.). QUADAS-2 results were finally summarized and analyzed to evaluate overall quality of the included studies.

Statistical Analysis and Data Synthesis

The primary analysis was to estimate and compare ESR and CRP sensitivity and specificity values in diagnosing acute location-confined or systemic inflammation reported in included studies. Several studies, however, reported different accuracy results, each one calculated by using different cutoff values for both ESR and CRP. Accordingly, we considered two different options for analyzing these data and obtained meta-analysis results. In the first case, studies reporting multiple cutoff values were evaluated considering only sensitivity and specificity results for the cutoff closest to the median cutoff value of each supercategory. Bivariate mixed-effects regression was performed for making inferences about average sensitivity and specificity using the MIDAS packages of Stata v 13.1. Meta-regression was used to investigate heterogeneity, and a measure for inconsistency (I2) was calculated. Study type (retrospective, prospective, or cross-sectional), population type (adult or pediatric), total sample size, and publication date were used as categorical or continuous variables for evaluating the covariates effect in meta-regression. In the second case, all cutoff values reported in the included studies were considered. Summary receiver operating characteristic (SROC) curve, as well as pooled sensitivity and specificity at each specific threshold, was calculated using the R package diagmeta developed by Steinhauser et al[20] for meta-analysis of diagnostic accuracy studies with multiple cutpoints. Furthermore, an optimal threshold across studies was determined using maximization of the Youden index.[21]