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


Baseline Characteristics and Outcomes of the Patients

Of the 1526 consecutive patients who were screened for inclusion in the study, 713 (46.7%) were excluded, and 813 patients were enrolled in the analyses (Figure 1). The baseline characteristics and outcomes of the patients are summarized in Table 1. The serum concentrations of PCT and CRP were significantly higher in nonsurvivors than in survivors (PCT: 0.97 [0.23; 5.51] vs. 0.12 [0; 1.02] μg/L, P = 0.000; CRP 66.70 [12.80; 140.00] vs. 11.95 [2.25; 56.15] mg/L, P = 0.000). APACHE II score, SOFA score, age, source of admission, treatments, sepsis, preexisting clinical conditions including hypertension, COPD, coronary disease, diabetes mellitus and chronic heart failure also differed significantly between the two groups. The median days of LOS in the ICU were 6 and 3 in the survival and nonsurvival groups, respectively (Table 1).

Figure 1.

Flowchart showing inclusion and exclusion of patients for the study

Risk Factor Analyses

In the univariate analyses, elevations in PCT concentrations (odds ratio [OR]: 1.010 [1.003; 1.018], P = 0.004) and CRP concentrations (OR: 1.008 [1.006; 1.011], P = 0.000) were associated with an increased risk of ICU mortality. Additionally, APACHE II score, age, sepsis, hypertension, diabetes mellitus, COPD, coronary disease, mechanical ventilation, RRT, use of norepinephrine and chronic heart failure also had this association, whereas WBC was not determined to be a risk factor (Table 2).

ROC Analyses

ROC analysis revealed that, as continuous predictors, none of the inflammatory markers, including PCT and CRP concentrations and WBC count at ICU admission, were considered satisfactory in discriminating between survival and nonsurvival. The AUC-ROC value of PCT for discriminating ICU mortality was 0.696 (95% CI: 0.650–0.743), that of CRP was 0.684 (95% CI: 0.633–0.735), and that of WBC was 0.568 (95% CI: 0.509–0.628). The same results were observed in the sepsis and nonsepsis subgroups (Supplementary Table 1). The optimal cut-off value of CRP for ICU mortality was 62.8 mg/L, and the sensitivity and specificity were 52.1% (95% CI, 42.8–61.2) and 76.7% (95% CI, 73.4–79.8), respectively. The positive and negative predictive values were 28.1% (95% CI, 22.3–34.5) and 90.2% (95% CI, 87.5–94.2), respectively.

In the multivariable analysis, only APACHE II score (odds ratio [OR], 1.166; 95% CI, 1.129–1.203; P = 0.000) and CRP concentration > 62.8 mg/L (OR, 2.145; 95% CI, 1.343–3.427; P = 0.001) were significantly associated with an increased risk of ICU mortality (Table 2), and this association did not seem to be significantly different between the sepsis and nonsepsis groups (Figure 2). Moreover, adding the variable of CRP > 62.8 mg/L to the APACHE II score did not significantly increase the AUROC (Figure 3).

Figure 2.

A high CRP concentration at admission in relation to predicted risk of ICU mortality stratified by sepsis and non-sepsis patients

Figure 3.

OC analysis of CRP > 62.8 mg/L and APACHE II score biomarkers and their combinations for intensive care unit mortality prediction

Effect of CRP on the Risk Reclassification of ICU Mortality

Although CRP > 62.8 mg/L as a dichotomized variable did not significantly provide incremental value to the AUC of the APACHE II score, it markedly improved risk reclassification over the APACHE II score alone, displaying a category-free NRI of 0.556 (P < 0.0001) and an IDI of 0.013 (P = 0.0245) (Table 3). Our analysis suggested that the addition of CRP to the APACHE II score considerably improved the prediction accuracy of ICU mortality, mainly due to increasing the correct predicted probabilities for without events.

Sensitivity Analyses

In the sensitivity analysis using a restricted cubic spline function for PCT, there was no obvious nonlinearity between PCT and the risk of ICU mortality (Figure 4A). However, CRP concentration > 62.8 mg/L was confirmed to be the optimal cut-off value in differentiating between patients with survival and those without survival (Figure 4B). The sensitivity and specificity of using CRP > 62.8 mg/L as a cut-off point to predict ICU mortality were 0.521 (95% CI, 0.43–0.72) and 0.767 (95% CI, 0.55–0.87), respectively. The positive and negative predictive values were 0.281 (95% CI, 0.214–0.516) and 0.902 (95% CI, 0.85–0.98), respectively.

Figure 4.

Relationship between risk of ICU mortality and PCT (A) and CRP (B) concentrations at ICU admission, allowing PCT and CRP as a nonlinear continuous predictor using a restricted cubic spline 3-knot function while adjusting for the APACHEIIscore, WBC, and CRP concentration. Shadow area show 95% CI