C-reactive Protein Concentration as a Risk Predictor of Mortality in Intensive Care Unit

A Multicenter, Prospective, Observational Study

Rong Qu; Linhui Hu; Yun Ling; Yating Hou; Heng Fang; Huidan Zhang; Silin Liang; Zhimei He; Miaoxian Fang; Jiaxin Li; Xu Li; Chunbo Chen


BMC Anesthesiol. 2020;20(292) 

In This Article


Study Design and Setting

The study was conducted from September 1, 2018, to August 1, 2019, in four general ICUs of tertiary care hospitals in the Guangdong Province, China, which were multidisciplinary ICUs admitting patients from all medical areas with a specialty in surgery, including cardiothoracic surgery. When admitted into the ICU, patients were assessed for inclusion in the study. The inclusion criteria were: 1) length of ICU stay more than 24 h; 2) age over 15 years old; 3) informed consent signed. The exclusion criteria included any of the following: length of stay (LOS) in the ICU < 24 h; patients with thyroid tumors (e.g., thyroid adenoma or thyroid carcinoma); and inability to provide informed consent or unavailability of a proxy for informed consent. The primary endpoint was all-cause ICU mortality. The protocol was approved by the Institutional Ethics Committee of each participating center and was performed according to the ethical standards of the Declaration of Helsinki. Written informed consent was obtained from each patient or their legal surrogates. The study was registered at http://www.chictr.org.cn/showproj.aspx?proj=29522 (ChiCTR1800017806).

Laboratory Measurements

All of the biomarkers were measured in a central laboratory, and all of the samples were labeled using study identification numbers without personal identifiers or clinical conditions. PCT was measured by Elecsys BRAHMS PCT (Roche Diagnostics GmbH, Germany; normal range, ≤0.05 μg/L). CRP was measured by an immunoenzyme analyzer (Hitachi 917, Tokyo, Japan; normal range, ≤5 mg/L). WBC counts were measured using an XE4000i automatic hemocyte analyzer. Blood samples for the purpose of study were collected only within 1 h after ICU admission, and clinicians decided the time and frequency of testing according to the actual clinical situation during the ICU stay.

Data Collection

In addition to PCT and CRP concentrations and WBC count, we also collected the demographic and clinical characteristics of each patient, including sex, age, treatment, preexisting chronic conditions, sepsis, Charlson score, source of admission, SOFA score, APACHE II score, and LOS in the ICU. ICU mortality data were collected by reviewing medical records in the in-hospital patient data management system. Sepsis was diagnosed according to the Surviving Sepsis Guidelines.[4]

Statistical Analysis

Continuous variables are expressed as the median (IQR) and were compared with the Mann-Whitney U test; categorical variables are expressed as numbers (%) and were compared by the χ 2 test or Fisher's exact test between the survival and nonsurvival groups. All analyses were 2-tailed and conducted by SPSS for Windows (version 26.0; IBM, Chicago, IL, USA) and R Statistical Software (version 5.3.0). A P value < 0.05 was considered statistically significant.

Discrimination was evaluated using the area under the curve (AUROC) derived from the conventional receiver operating characteristic (ROC) curve. AUROC of > 0.5, > 0.6, > 0.70 or > 0.80 were considered poor, fair, satisfactory or good, respectively.[12] AUROCs, as a measure of classification accuracy, were further compared with or without CRP added to the APACHE II score using the nonparametric approach of DeLong and Clarke-Pearson.[14] Univariate and multivariate logistic regression analyses were used to detect factors independently associated with nonsurvival. The optimal cut-off values for individual biomarkers were determined using Youden's index. To evaluate the utility of the biomarkers for risk classification, we determined the category-free net reclassification improvement (NRI) and the integrated discrimination improvement (IDI), as previously described.[15,16]

In consideration of the possibility that dichotomized cutoffs may not accurately capture the usefulness of PCT and CRP, sensitivity analysis was used to assess whether alternative cut points for these biomarkers were more appropriate. We used a restricted cubic spline function with 3 knots for PCT and CRP to allow nonlinearity as continuous predictors in a multivariable model.[12,17,18]