This multicenter cohort prospective study was conducted in the area of Barcelona (Spain). We included consecutive patients (≥ 16 years) admitted from May 2011 to January 2017 to the intensive cardiac care unit of 4 referral hospitals after an OHCA with shockable rhythm who completed the MTH protocol and who had at least 2 serum NSE measurements during hospitalization. The patients were treated according to the current practice of each hospital and NSE measurement timing was at the discretion of the treating physician. Data for 6 months’ follow-up after hospital discharge were obtained using electronic medical record systems.

The study was approved by the Vall d’Hebron Hospital Research Ethics Committee.

After assessment of persistent unconsciousness, the initial cooling phase was achieved by cold saline solution infusion. When patients were admitted to the cardiac care unit, an intravascular cooling system (CoolGard Thermal Regulation System, Alsius Corporation, Irvine, USA) or a cooling system with cutaneous pads (Arctic Sun Temperature Management System, Medivance, Louisville, USA) were used with a target temperature of 33°C. The protocol for all participating hospitals included hypothermia at 33°C for a 24-hour period, with controlled progressive rewarming at a rate of 0.25°C/h until ≥ 36.5°C.

Serum NSE measurements were performed on fresh blood samples using a Liaison XL chemiluminescence test (DiaSorin, Saluggia, Italy). The normality threshold was 18 ng/mL. The variable Δ-NSE was defined as the percentage change in NSE from the first to the second value and was calculated as 100 x (NSE2-NSE1) / NSE1 (%). Δ-NSE was also considered in the analysis as a binary variable: Δ-NSE ≤ 0 (invariable or decreasing NSE) and Δ-NSE >0 (increasing NSE).

The following variables were collected: demographics, cardiovascular risk factors, time from OHCA to initiation of life support (cardiopulmonary resuscitation), time from OHCA to return of spontaneous circulation, time from OHCA to initiation of MTH, arterial blood pH and lactate (mmol/L) on arrival, time from OHCA to the first NSE sampling, time from OHCA to the second NSE sampling, and time from the first to the second NSE measurements.

In-hospital mortality and its causes (neurologic, cardiac or multifactorial) were analyzed. Neurological function was assessed by the Cerebral Performance Category (CPC) scale,18 which classifies patients in 5 different categories, from CPC 1 (no or mild disability) to CPC 5 (brain death). In-hospital CPC scale evaluation was conducted by the consultant neurologist who was unaware of NSE values. The time of evaluation was at discharge or before death, and also after 6 months of follow-up using health electronic records.

We carried out an exploratory analysis of NSE sampling distribution over time. To assess the timing of NSE measurements that could more accurately predict outcomes, we selected different measurement periods according to the median of NSE1 and NSE2 sampling distributions. We established that every tested interval had to include a minimal number of 20 patients. For OHCA-NSE1 timing, we tested 6 hour-width intervals; and for OHCA-NSE2 timing, due to wider interquartile range (IQR), we tested 8 hour-width intervals.

Quantitative variables are expressed as median and IQR. For our analysis, CPC was considered a binary variable: patients with CPC 1 or 2 were considered as having a good neurological outcome and patients with CPC 3-5 as having a poor neurological outcome. We performed a binary logistic regression analysis with poor or good clinical outcome as the dependent variable and Δ-NSE as the independent variable; diabetes mellitus (dichotomous variable), age, pH and lactate were considered as potential variables to be included in a predictive model (backward stepwise selection, P<.05 for inclusion and P ≥ .10 for exclusion).

To assess the timing of serial NSE determinations that best predicted outcomes, we created receiver operating characteristic curves of Δ-NSE and compared the areas-under-curve (AUC) of different periods of OHCA to NSE1 and of OHCA to NSE2 with the AUC in the global cohort.

Data analysis was performed with Stata/IC 14.2 (StataCorp LP).

Baseline characteristics are shown in Table 1. The median age of the cohort was 57.0 (range, 16-88 years). A total of 83.6% of patients were male and 19.5% had diabetes mellitus. Median time from OHCA to initiation of life support was 4 minutes, and was 25 minutes from OHCA to return of spontaneous circulation. Metabolic acidosis was prominent on arrival, with a median pH of 7.23; 13 patients (7.9%) had blood pH<7.0. Initial arterial blood lactate was high (> 2.0 mmol/L) in 67.3% of the patients; 19 patients (11.5%) had a lactate level above 8.0 mmol/L.

Table 1.

Comparison between low and high CPC patients at discharge

CPC, Cerebral Performance Category scale; CPR, cardiopulmonary resuscitation; DM, diabetes mellitus; IQR, interquartile range; NSE, neuron-specific enolase; NSE1, first neuron-specific enolase measurement; NSE2, second neuron-specific enolase measurement; OHCA, out-of-hospital cardiopulmonary arrest; ROSC, return of spontaneous circulation.

This table shows baseline characteristics, NSE serum levels and measurement intervals of the study global cohort, and also of the subgroups of patients with good and poor neurological outcome on discharge.

Δ-NSE (relative change in NSE)=100 x (NSE2-NSE1) / NSE1.

Data are expressed as No. (%) or median [interquartile range].

Patients with poor neurological status on discharge were more frequently diabetic, had longer time from OHCA to return of spontaneous circulation, lower pH and a higher lactate level on arrival (Table 1). In addition, advanced age and longer time from OHCA to initiation of life support showed a tendency to worse prognosis, but was not statistically significant.

NSE1 was obtained at a median of 18.8 (IQR, 13.0-26.0) hours after OHCA and NSE2 was obtained at a median of 65.1 (IQR, 48.1-81.1) hours after OHCA. Median time between the 2 measurements was 47.6 (IQR, 24.0-49.5) hours. Box plots showing the distribution of OHCA-NSE1 and OHCA-NSE2 timings are depicted in Figure 1 of the supplementary data.

Median NSE1 was 33.9 ng/mL, while median NSE2 was 27.0 ng/mL. Δ-NSE had high variability, with a median value of−16.7% (IQR−38.8% to 21.3%). One-hundred-and-eight patients (65.5% of the cohort) had a negative Δ-NSE value (decreasing NSE levels), and 57 patients (34.6%) had a positive value (increasing levels). The median Δ-NSE in the group of Δ-NSE ≤ 0 was−34.7% (IQR,−49.6% to−17.6%). For Δ-NSE >0, the median value was 75.0% (IQR, 20.6%-204.1%).

In-hospital mortality was 31.9% (53 of 166 patients). The main cause of death was severe anoxic encephalopathy (44 cases; 83.0%); cardiac death occurred in only 2 patients (3.8%).

In the whole cohort, CPC distribution at discharge was: CPC 1 in 72 patients (43.6%), CPC 2 in 24 patients (14.6%), CPC 3 in 11 patients (6.7%), CPC 4 in 33 patients (20.0%), and CPC 5 in 25 patients (15.2%). More than half of the patients (96, 58.2%) had a good neurological recovery according to the CPC scale.

Of the overall cohort, 89 patients (59.3%) were alive and had good neurological status at 6 months (CPC 1 or 2). Among patients who survived after discharge, 88 out of 97 (90.7%) were neurologically recovered after 6 months.

Both high NSE1 and NSE2, in absolute values, were significantly associated with mortality, CPC at discharge, and CPC at 6 months. For CPC at discharge, median NSE1 was 26.1 ng/mL for patients with CPC 1-2 and 42.8 ng/mL for those with CPC 3-5 (P<.001); median NSE2 was 17.2 ng/mL for CPC 1-2 and 54.0 ng/mL for CPC 3-5 (P<.001).

For patients with CPC 1-2 median Δ-NSE was−33.3% (IQR−49.0% to−14.0%) and for those with CPC 3-5 median Δ-NSE was+21.3% (IQR,−12.9% to+144.7%). Median Δ-NSE for patients who survived to discharge was−28.3% (IQR,−44.0% to−10.2%), and for those who died during admission, median Δ-NSE was+28.1% (IQR,−6.6% to+160.9%).

On univariate analysis, Δ-NSE was significantly associated with in-hospital all-cause mortality (OR, 1.013, 95% CI, 1.007-1.018; P<.001, for 1% increase), CPC 3-5 at discharge (OR, 1.017, 95% CI, 1.010-1.025; P<.001) and CPC 3-5 at 6 months (OR, 1.015, 95% CI, 1.008-1.023; P<.001). Thus, an increase in Δ-NSE was associated with poor neurological outcome, while a decrease was associated with good neurological outcome. Odd ratios are easier to interpret when Δ-NSE is considered as a binary variable (Δ-NSE >0 [increasing NSE] vs Δ-NSE ≤ 0 [invariable or decreasing NSE]): OR 9.28 (95% CI, 4.40-19.57) for in-hospital mortality, OR 11.24 (95% CI, 5.24-24.11) for CPC 3-5 at discharge and OR 11.14 (95% CI, 5.05-24.55) for CPC 3-5 at 6 months (Figure 2).

Figure 2.

Δ-NSE and outcomes. Neurological prognosis on discharge (A), neurological prognosis at 6 months (B) and in-hospital mortality (C) classified by relative change in NSE serum levels [Δ-NSE=100 x (NSE2-NSE1) / NSE1]. Δ-NSE ≤ 0 refers to invariable or decreasing levels, while Δ-NSE >0 refers to increasing NSE levels. Neurological prognosis was assessed by the CPC scale; numbers in the graphs correspond to CPC scale 1-5. Good neurological outcome is depicted in light gray (CPC 1 and 2) and poor neurological outcome is depicted in dark gray (CPC 3-5). CPC, Cerebral Performance Category; NSE, neuron-specific enolase; NSE1, first neuron-specific enolase measurement; NSE2, second neuron-specific enolase measurement.

(0.3MB).

Positive Δ-NSE had a sensitivity of 63.8% for predicting CPC 3-5 and negative Δ-NSE had a specificity of 86.5% for CPC 1-2; positive predictive value was 77.2% and negative predictive value was 76.9%. AUC for Δ-NSE was 0.8096 (95% CI, 0.7414-0.8779) for the whole cohort.

The final model for multivariate logistic regression analysis included age and arterial blood pH on arrival (Table 1 of the supplementary data). Δ-NSE persisted as a significant predictor of neurological prognosis at discharge: OR 1.016 (95% CI, 1.008-1.024), P<.001, for 1% increase. This was also true for Δ-NSE as a binary variable, as previously described: OR 11.58 (95% CI, 4.89-27.41), P<.001.

When we analyzed only patients with a high first NSE >33 ng/mL, increasing NSE was invariably associated with poor neurological prognosis (32 of 33 patients with CPC 3-5). On the other hand, 60.0% of patients with NSE1 >33 ng/mL and decreasing NSE had good prognosis (CPC 1-2) (Figure 2 of the supplementary data). Two extreme examples were 2 patients with a first NSE value of 79.3 ng/mL and 109.5 ng/mL, respectively, who were discharged with CPC 1 after experiencing a clear decrease in NSE (to 11.4 ng/mL and 33.2 ng/mL, respectively).

Compared with NSE2 as an absolute value, Δ-NSE could predict CPC at discharge in those patients with moderately elevated NSE2 (31 patients with NSE2 30-50 ng/mL), P=.0185. In this group of patients, NSE2 had no statistically significant predictive capacity (P=.2124).

Finally, we constructed a logistic regression model including Δ-NSE, age, diabetes, first blood pressure, first lactate and time from OHCA to return of spontaneous circulation, and we compared it with the same model without Δ-NSE. The AUC for the complete model was 0.8078, and AUC for the model without Δ-NSE was 0.7372, which was not statistically significantly lower (P=.2887).

The second objective of our study was to determine the sampling time intervals that best predicted CPC 3-5. Two groups of time intervals were considered: the interval from OHCA to the first NSE sampling, and the interval from the first to the second measurement. Comparing AUC of receiver operating characteristic-analysis, the best OHCA-NSE1 time interval appeared to be 18 to 24hours (AUC 0.9389, 95% CI, 0.8692-1.000, including 38 patients). The best OHCA-NSE2 time interval appeared to be 69 to 77hours (AUC 0.9910, 95%, CI 0.9657-1.000 with 21 patients). There was, however, remarkable overlap of CI in all assessed time intervals. AUC in both selected intervals were higher than AUC for the whole cohort (AUC 0.8096, 95% CI, 0.7414-0.8779) (Figure 3), and were even better than the AUC for absolute values (0.7640 for NSE1, 0.8986 for NSE2), although due to the small sample size in all intervals, the differences were not statistically significant. Table 2 of the supplementary data shows the tested intervals.

Figure 3.

ROC curves. Relative change in NSE serum levels [Δ-NSE=100 x (NSE2-NSE1) / NSE1], as a predictor of neurological outcome assessed by the CPC scale, considering good neurological outcome as CPC 1 or 2 and poor neurological outcome as CPC 3-5. Comparison of ROC curves of global cohort and 2 subgroups: (A) Δ-NSE with first NSE analysis conducted 18 to 24hours after OHCA and (B) Δ-NSE with second NSE analysis conducted 69 to 77hours after OHCA. Global cohort means Δ-NSE in the whole study population. CPC, Cerebral Performance Category; NSE, neuron-specific enolase; NSE1, first neuron-specific enolase measurement; NSE2, second neuron-specific enolase measurement; OHCA, out-of-hospital cardiopulmonary arrest; ROC, receiver operating characteristic.

(0.34MB).