Diagnostic Accuracy of Procalcitonin and C-reactive Protein Is Insufficient to Predict Proven Infection: A Retrospective Cohort Study in Critically Ill Patients Fulfilling the Sepsis-3 Criteria

Abstract

Background: New Sepsis-3 definitions facilitate early recognition of patients with sepsis. In this study we investigated whether a single initial determination of procalcitonin (PCT) or C-reactive protein (CRP) in plasma can predict proven sepsis in Sepsis-3 criteria-positive critically ill patients. We also investigated whether a decline in serial PCT or CRP can predict outcome in 28-day mortality. Methods: Patients, 18 years of age, at the intensive care unit with a suspected infection, a Sequential Organ Failure Assessment (SOFA) score of 2 points, and an index test PCT and CRP at admission were selected from a prospectively collected cohort. PCT and CRP were studied retrospectively with the Mann Whitney U-test and ROC analysis. Results: In total, 157 patients were selected; 63 of the 157 had proven sepsis, and sepsis could not be detected in 94 of the 157. Neither a single PCT nor CRP at admission was able to discriminate proven sepsis from nonproven sepsis (PCT, 1.8 ug/L and 1.5 ug/L, respectively, P = 0.25; CRP, 198 mg/L and 186 mg/L, respectively, P = 0.53). Area under the curve for both PCT and CRP for detecting proven sepsis was low (0.55 and 0.53). Furthermore, neither a decline from baseline to day 5 PCT nor CRP could predict 28-day mortality (PCT, 50% vs 46%, P = 0.83; CRP, 30% vs 40%, P = 0.51). Conclusion: PCT and CRP at admission were not able to discern patients with proven sepsis in Sepsis-3 criteria-positive critically ill patients. A decline of PCT and CRP in 5 days was not able to predict 28-day mortality.