Protein S100 as outcome predictor after out-of-hospital cardiac arrest and targeted temperature management at 33 °C and 36 °C

Mortality in comatose out-of-hospital cardiac arrest (OHCA) patients admitted to an intensive care unit (ICU) is around 50%. Whereas initial ICU mortality is caused by hemodynamic failure in the majority of cases, later morbidity and mortality are due mainly to hypoxic brain damage [1, 2]. Withdrawal of life-sustaining therapies (WLST) based on presumed poor neurological prognosis is the predominant cause of death [2, 3]. To better guide therapy and to support decisions on WLST, there is a need for early and accurate outcome prediction tools in this severely ill population.

The S100 protein, a 21 kDa intracellular calcium-binding dimer, is implicated in neuronal differentiation, proliferation, and apoptosis [4]. Many subtypes of the S100 protein are known, but the most studied in humans are the brain-specific homodimers A1B (αβ) and BB (ββ) [5, 6]. S100 is a biomarker candidate for outcome prediction after cardiac arrest (CA) [7, 8], but previous small studies yielded a wide range of cutoff values for a poor outcome, and current guidelines do not advocate its use [9]. S100 is present mainly in white matter, predominantly in astroglial cells, in contrast to neuron-specific enolase (NSE), which is found principally in neurons and neuroendocrine cells [10]. S100 is also commonly present in extracerebral tissues [11, 12]. The Target Temperature Management after Out-of-Hospital Cardiac Arrest (TTM) trial, a multicenter clinical trial that randomized 939 patients to targeted temperature management of 33 °C or 36 °C, provides an opportunity to investigate the role of S100 as a prognostic marker after OHCA [13].

In this substudy of a large international trial, the use of S100 for outcome prediction after OHCA was assessed. S100 values were higher in patients with poor outcomes at all time points, with the best capacity for S100 to predict outcome being at 24 h. In multivariable analysis, S100 measurements at 24, 48, and 72 h were not significant predictors of outcome. The joint effect of the three measurements, however, improved the AUC, NRI, and IDI of a predictive model that included established clinical characteristics associated with outcome.

In previous, smaller studies, researchers compared S100 in two target temperature groups and could not detect a significant influence of temperature on S100 levels [18, 19]. In this study, S100 values were higher at 24 and 72 h in the 33 °C group than in the 36 °C group, which was explained by higher S100 values among patients with good outcomes in the 33 °C group. Because the intervention groups and their outcomes were very similar in all aspects other than the intervention temperature, we speculate that this difference might be related to the targeted temperature. In addition, the observed values among patients with good outcomes were well below the suggested cutoff levels for S100. Although we do not have a clear explanation for this result, we consider the finding to be of negligible clinical relevance.

S100 could distinguish patients with good and poor outcomes after OHCA because median values were higher in the poor outcome group, and this has been described in previous reports [18, 20‐25]. It is noteworthy that S100 values declined over time in both temperature groups and for both outcome groups, indicating an early peak of this biomarker, which might explain why the first sample (at 24 h after ROSC) showed the best results [26]. We did not collect blood samples before 24 h after ROSC, and higher levels prior to 24 h cannot be ruled out. However, a clear peak earlier than 24 h could not be determined in a previous study in which researchers investigated the kinetic profile of S100 [23]. Other studies have also confirmed a similar decline over time after 24 h in patients with good and poor outcomes [22, 25]. The early release and subsequent decline may be explained by the short half-life of approximately 2 h in combination with a low molecular weight, allowing a rapid transition through the blood-brain barrier [27]. This differentiates S100 from other biomarkers (e.g., NSE), where the kinetics between 24 and 72 h after CA are indicative of outcome [17]. The earlier peak of S100 and its relative strength over NSE and other biomarkers for outcome prediction at 24 h could potentially be of clinical use under certain circumstances, such as when prolonged care after rewarming might be considered unethical and several prognostic indicators point to a poor outcome. Another argument in favor of using S100 as an adjunct in prognostication after CA might be its availability in many centers, owing to its common use in the assessment of traumatic brain injury [28].

The cutoff values for S100 in this study are comparable with those described previously [7, 8, 25, 29]. Any differences might be due to different assays that might yield different values [20, 22, 23], different outcome measures [29], and sample size [23]. As with other biomarkers, an absolute cutoff value with an FPR of 0 for poor outcome may be unrealistic and would limit its use. A more feasible approach might be to choose a higher FPR, which might be acceptable when used in combination with other prediction tools [9]. In this study, a cutoff with an FPR of 5% would correspond to an S100 serum level of 0.25 μg/L at 24 h after ROSC.

As with any other prognostication method, prediction should be based on a protocol including a holistic approach and with multiple tests and parameters [9, 30]. Clearly, NSE outperformed S100 for outcome prediction after CA in the same patient cohort [17]. Adding S100 to our model including clinical characteristics and NSE did not further improve the accuracy of the model. Similar results have also been described by others when S100 was added to NSE [25]. Using a multivariable model with fewer variables, researchers in another study suggested the usefulness of S100 over NSE on admission [22]. Although the use of a combination of biomarkers for outcome prediction is intriguing, we failed to demonstrate any added value of S100 in a clinical model including NSE.

There was no clinically important effect of two different target temperatures on levels of S100. High S100 values are predictive of poor outcome after OHCA but do not add any real value to present prognostication models with or without NSE. S100 measured at 24 h and afterward is of limited value in clinical outcome prediction after OHCA, especially in a setting where NSE is available.