The Prognostic Role of Procalcitonin in Critically Ill Patients Admitted in a Medical Stepdown Unit: A Retrospective Cohort Study

Abstract

Procalcitonin (PCT) is a a marker of bacterial infection. Its prognostic role in the critically-ill patient, however, is still object of debate. Aim of this study was to evaluate the capacity of admission PCT (aPCT) in assessing the prognosis of the critically-ill patient regardless the presence of bacterial infection. A single-cohort, single-center retrospective study was performed evaluating critically-ill patients admitted to a stepdown care unit. Age, sex, Simplified Acute Physiology Score II (SAPS-II), shock, troponin-I, aPCT, serum creatinine, cultures and clinical endpoints (in-hospital mortality or Intensive Care Unit (ICU) transfer) were collected. Time free from adverse event (TF-AE) was defined as the time between hospitalization and occurrence of one of the clinical endpoints, and calculated with Kaplan-Meier curves. We engineered a new predictive model (POCS) adopting aPCT, age and shock.We enrolled 1063 subjects: 450 reached the composite outcome of death or ICU transfer. aPCT was significantly higher in this group, where it predicted TF-AE both in septic and non-septic patients. aPCT and POCS showed a good prognostic performance in the whole sample, both in septic and non-septic patients. aPCT showed a good prognostic accuracy, adding informations on the rapidity of clinical deterioration. POCS model reached a performance similar to SAPS-II.

Figure 1

Study flow showing the number of patients included and excluded from the study according to the inclusion and the exclusion criteria.

Figure 2

ROC Curve analysis for aPCT in the overall sample. ROC curve analysis showed a good predictive value of aPCT for the composite endpoint (death or transfer in ICU) in the overall sample (AUC:0.690; 95%CI:0.642–0.732; p < 0.05).

Figure 3

ROC Curve analysis for aPCT in the subgroups of “infective” and “non-infective” patients. ROC curve analysis showed a good predictive value of aPCT for the composite endpoint (death or transfer in ICU) both in the subpopulation of “infective” (AUC:0.660; 95%CI:0.61–0.70; p < 0.05) and “non-infective” patients (AUC:0.732; 95%CI:0.64–0.80; p < 0.05). When comparing the above-mentioned ROC curves, we observed a significantly better performance of the ROC curve in “non-infective” patients (p < 0.05).

Figure 4

ROC Curve analysis for POCS, SAPS-II, PCT and TnI in the overall sample. In the overall sample, the prognostic performance of aPCT was similar to TnI (AUC_aPCT vs AUC_TnI p = 0.271) but inferior to SAPS-II (AUC_SAPS-II vs AUC_aPCT p = 0.0244). We also compared the prognostic performance of POCS model with other indicators as SAPS-II, aPCT or TnI. We did not find any statistically significant difference between AUC_POCS and AUC_SAPS-II (p = 0.317), while AUC_aPCT and AUC_TnI were significantly lower than AUC_SAPS-II (AUC_aPCT vs AUC_SAPS-II p = 0.024; AUC_TnI vs AUC_SAPS-II p = 0.0009).

Figure 5

Kaplan-Meier survival analysis within WP group, according to aPCT values and adopting 0.5 ng/mL as cutoff. This curves confirmed that TF-AE can be significantly modified by aPCT levels at a cutoff of 0.50 ng/mL (p < 0.0001).