Mortality prediction model from combined serial lactate, procalcitonin and calprotectin levels in critically ill patients with sepsis: A retrospective study according to Sepsis-3 definition

Abstract

Objective

1) To evaluate the ability of baseline and on 24 h serum calprotectin, in comparison to canonical biomarkers (lactate and procalcitonin), for prognosis of 28-day mortality in critically ill septic patients; and 2) To develop a predictive model combining the three biomarkers.

Design

A single-center, retrospective study.

Setting

Intensive Care Unit of a university hospital.

Patients or participants

One hundred and seventy three septic pacientes were included.

Interventions

Measurement of baseline lactate, procalcitonin and calprotectin level and procalcitonin and calprotectin levels on 24 h.

Main variables of interest

Demographics and comorbidities, SOFA score on ICU admission, baseline lactate, procalcitonin and calprotectin on admission and on 24 h and 28-day mortality.

Results

1) On ICU admission, lactate was the only biomarker achieving a significant accuracy (AUC: 0.698); 2) On 24 h, no differences were found on procalcitonin and calprotectin levels. Procalcitonin and calprotectin clearances were significantly lower in non-survivors and both achieved a moderate performance (AUCs: 0.668 and 0.664, respectively); 3) A biomarker based-model achieved a significant accuracy (AUC: 0.766), trending to increase (AUC: 0.829) to SOFA score alone; y 4) Baseline lactate levels and procalcitonin and calprotectin clearance were independent predictors for the outcome.

Conclusions

1) Baseline and on 24 h calprotectina and procalcitonin levels lacked ability in predicting 28-day mortality; 2) Accuracy of clearance of both biomarkers was moderate; and 3) Combination of SOFA score and the predictive biomarker based-model showed a high prognostic accuracy.