Predicting outcome after cardiac arrest with serum S-100B protein and procalcitonin: A prospective observational study
Despite significant advances in diagnosis and treatment predicting the neurological sequelae of cardiac arrest is difficult.1 Several laboratory tests have been investigated separately for their prognostic significance but no ideal biomarker for neurological damage postcardiac arrest has been identified.2,3 Among biomarkers investigated separately are S100B protein and procalcitonin (PCT). We hypothesised that combined biomarker determination of S100B and PCT could improve outcome prediction in patients resuscitated from cardiac arrest.
The Ethics Committee of the Medical University of Silesia (Katowice, Poland) approved the study with waiver of consent because the project was noninterventional (No KNW/0022/KB1/31/0; 16 March 2010). We enrolled 70 patients prospectively following successful return of spontaneous circulation after cardiac arrest, including 37 after in-hospital and 33 after out-of-hospital cardiac arrest. All of them were unconscious on admission, with a Glasgow Coma Score below 8.
Peripheral blood was sampled once on arrival in the ICU. Serum PCT level was assessed shortly after blood collection using a chemiluminescent enzyme immunoassay (LIAISON BRAHMS PCT assay). Samples for S100B protein concentration measurement were centrifuged and frozen at −70°C. They were collectively thawed and analysed once per quarter with an ELISA, using a Cobas e411 analyser (Hitachi, Tokyo, Japan) and reagents made by Boehringer Mannheim (Mannheim, Germany).
Detailed information on the outcome at hospital discharge was given by a consulting neurologist, who was unaware of the S100B and PCT data, generally after 3 to 5 days of weaning from sedation and opioid analgesia. The outcome was assessed using the Glasgow Outcome Scale (GOS)4 and categorised into poor (GOS 1 to 3) or good (GOS 4 to 5).
Mortality in the cohort (67% men; median age 63 years, interquartile range ([IQR] [54 to 72]) was 30%. In total, 38 (54%) had a good neurological outcome and 32 (46%) had a poor outcome.
Median S100B level was 0.27 μg l−1 (IQR [0.16 to 0.85]) and median PCT concentration was 1.92 ng ml−1 (IQR [0.86 to 3.41]). There was no correlation between S100B and PCT (R = 0.12; P = 0.3). S100B levels were higher in study participants with GOS 1 to 3 (0.36 μg l−1, IQR [0.21 to 1.0]) compared with those with GOS 4 to 5 (0.18 μg l−1, IQR [0.12 to 0.4]; P = 0.004). PCT levels were also higher in individuals with GOS 1 to 3 (2.43 ng ml−1, IQR [1.33 to 9.1]) compared with those with GOS 4 to 5 (1.5 ng ml−1, IQR [0.8 to 2.46]; P = 0.03; Table 1). Both S100B and PCT predicted the outcome with moderate accuracy (S100B: area under the receiver operating characteristic curve [AUROC]) = 0.7, 95% confidence interval (CI) [0.58 to 0.8]; P = 0.002 /PCT: AUROC = 0.65; 95% CI [0.52 to 0.76]; P = 0.02]. Combination of both biomarkers did not improve diagnostic accuracy (AUROC = 0.62; 95% CI [0.49 to 0.73]; P = 0.06).
Cardiac arrest may cause cerebral damage during the circulatory standstill phase of the event and also from ischaemia/reperfusion injury following the return of adequate circulation.5 The S100B protein is abundant in astroglial cells and can cross the blood–brain barrier after hypoxic injury of the brain following cardiac arrest.6 PCT is a serum acute-phase protein and may serve as a sensitive inflammatory marker of metabolic imbalance in the period following cardiac arrest.7 On this basis, we hypothesised that S100B and PCT may provide independent information regarding the extent of damage following cardiac arrest.
Changes in S100B levels shortly after the return of spontaneous circulation have been associated with neurological outcome.3 However, there is no consensus regarding a threshold value for S100B after cardiac arrest that will permit identification of brain injury as there is marked heterogeneity between studies. PCT concentrations immediately after cardiac arrest are also very variable but combining several biomarkers may improve prediction accuracy.