Objective: To investigate day-by-day changes in procalcitonin and maximum obtained levels as predictors of mortality in critically ill patients.
Design: Prospective observational cohort study.
Setting: : Multidisciplinary intensive care unit at Rigshospitalet, Copenhagen University Hospital, a tertiary reference hospital in Denmark.
Patients: Four hundred seventy-two patients with diverse comorbidity and age admitted to this intensive care unit.
Interventions: Equal in all patient groups: antimicrobial treatment adjusted according to the procalcitonin level.
Measurements and main results: Daily procalcitonin measurements were carried out during the study period as well as measurements of white blood cell count and C-reactive protein and registration of comorbidity. The primary end point was all-cause mortality in a 90-day follow-up period. Secondary end points were mortality during the stay in the intensive care unit and in a 30-day follow-up period. A total of 3,642 procalcitonin measurements were evaluated in 472 critically ill patients. We found that a high maximum procalcitonin level and a procalcitonin increase for 1 day were independent predictors of 90-day all-cause mortality in the multivariate Cox regression analysis model. C-reactive protein and leukocyte increases did not show these qualities. The adjusted hazard ratio for procalcitonin increase for 1 day was 1.8 (95% confidence interval 1.3-2.7). The relative risk for mortality in the intensive care unit for patients with an increasing procalcitonin was as follows: after 1 day increase, 1.8 (95% confidence interval 1.4-2.4); after 2 days increase, 2.2 (95% confidence interval 1.6-3.0); and after 3 days increase: 2.8 (95% confidence interval 2.0-3.8).
Conclusions: A high maximum procalcitonin level and a procalcitonin increase for 1 day are early independent predictors of all-cause mortality in a 90-day follow-up period after intensive care unit admission. Mortality risk increases for every day that procalcitonin increases. Levels or increases of C-reactive protein and white blood cell count do not seem to predict mortality.