Ubiquitin C-terminal hydrolase L1 after out-of-hospital cardiac arrest

Background: We studied the prognostic ability of serum ubiquitin C-terminal hydrolase L1 (UCH-L1) after out-of-hospital cardiac arrest (OHCA), compared to that of neuron-specific enolase (NSE). Methods: In this post-hoc analysis of the FINNRESUSCI study, we measured serum concentrations of UCH-L1 in 249 OHCA patients treated in 21 Finnish intensive care units in 2010 – 2011. We evaluated the ability of UCH-L1 to predict unfavourable outcome at 12 months (defined as cerebral performance category 3 – 5) by assessing the area under the receiver operating characteristic curve (AUROC), in comparison with NSE. Results: The concentrations of UCH-L1 were higher in patients with unfavourable outcome than for those with favourable outcome: median concentration 10.8 ng/mL (interquartile range, 7.5 – 18.5 ng/mL) versus 7.8 ng/mL (5.9 – 11.8 ng/mL) at 24 h ( p < .001), and 16.2 ng/mL (12.2 – 27.7 ng/mL) versus 11.5 ng/mL (9.0 – 17.2 ng/mL) ( p < .001) at 48 h after OHCA. For UCH-L1 as a 12-month outcome predictor, the AUROC was 0.66 (95% confidence interval, 0.60 – 0.73) at 24 h and 0.66 (0.59 – 0.74) at 48 h. For NSE, the AUROC was 0.66 (0.59 – 0.73) at 24 h and 0.72 (0.65 – 0.80) at 48 h. The prognostic ability of UCH-L1 was not different from that of NSE at 24 h ( p = .82) and at 48 h ( p = .23). Conclusion: Concentrations of UCH-L1 in serum were higher in patients with unfavourable outcome than in those with favourable outcome. However, the ability of UCH-L1 to predict unfavourable outcome after OHCA was only moderate and not superior to that of NSE.


Editorial Comment
In this post-hoc analysis of the FINNRESUSCI study, serum ubiquitin C-terminal hydrolase L1 (UCH-L1), an enzyme mostly located in neurons of the cerebral cortex, was measured at 24 and 48 h following out-of-hospital cardiac arrest of any cause in 249 patients. The levels of UCH-L1 showed moderate predictive performance for unfavourable neurological outcome (cerebral performance category 3-5) at 12 months follow up and was not superior to the more commonly used neuron-specific enolase assay.

| INTRODUCTION
Hypoxic-ischemic brain injury (HIBI) is the most common reason for severe disability and death after cardiac arrest (CA). 1,2 Identifying imminent severe HIBI that results in poor prognosis is an essential part of patient management. 3 Biomarkers are one part of multimodal prognostication along with imaging and clinical and neurophysiological examinations. 4 Biomarkers have some special benefits: they are non-invasive and inexpensive, and the results are not confounded by sedative medications. 3 Neuron-specific enolase (NSE) is the biomarker recommended in the latest ERC-ESICM guideline. 4 However, when high NSE cut-off values are used to minimize the risk of falsely pessimistic prognosis, the sensitivity of NSE remains rather low, and many individuals with poor prognosis are not identified correctly. 5,6 In addition, haemolysis, 7,8 extracerebral sources, 9,10 stroke and traumatic brain injury (TBI) [11][12][13] may increase NSE concentrations. Moreover, we have previously shown that the prognostic value of NSE is dependent on the patient's age and the duration of CA. 14 Ubiquitin carboxyl-terminal hydrolase L1 (UCH-L1) is an enzyme mostly located in neurons of the cerebral cortex. 15,16 After TBI, UCH-L1 concentrations are associated with severity of injury. [17][18][19] Results from a previous study suggested that UCH-L1 may be a rather good predictor of unfavourable outcome in OHCA patients with a presumably cardiac cause of arrest. 20 To assess the prognostic ability of UCH-L1 in an unselected OHCA population, we analysed UCH-L1 concentrations in serum of patients included in the FINNRESUSCI study. 21 2 | METHODS

| Patient selection and data collection
Patients for this post-hoc study were included in the prospective FINNRESUSCI study 21

| Outcome definitions
We defined favourable outcome as cerebral performance category (CPC) 1-2 and unfavourable outcome as CPC 3-5 at 12 months after CA. CPC 1-2 means sufficient neurological function for at least independently managing basic activities of daily living, whereas CPC 3-5 describes severe disability, persistent vegetative state or death. 22 As short-term outcome, we used death during index hospital treatment.

| Laboratory analyses
Blood samples were drawn from FINNRESUSCI study patients, for whom written consent was provided by a next of kin. The samples were allowed to clot at room temperature for 60 min, after which they were centrifuged, and serum was stored at À70 C. We analysed serum concentrations of UCH-L1 in March 2015 by a commercial enzyme-linked immunosorbent assay kit (USCN, Wuhan, China). We analysed all samples in duplicate. The intra-and inter-assay coefficients of variation (CV) were <7.5% and <11.5%, respectively, for UCH-L1. Intra-assay CVs were determined in 16 aliquots of two serum pools analysed in the same run and inter-assay CVs were determined in 10 aliquots of two serum pools analysed in the consecutive runs. The calibrators covered the range 0.16-10 ng/mL for UCH-L1.
Serum samples were diluted 4-fold or up to 20-fold when needed prior to assay of UCH-L1.
We have previously published the concentrations and prognostic ability of NSE in this same study population. 14,23 In the current study, we measured the UCH-L1 concentrations and additionally compared the prognostic ability of UCH-L1 to that of NSE.

| Statistical analysis
We present categorical data as absolute numbers with percentages (95% confidence intervals [CIs]). For continuous data, like biomarker concentrations, we present medians with interquartile ranges. We tested normality of distribution with the Kolmogorov-Smirnov test.
For categorical data, we used a chi-square test or Fisher's exact test, as appropriate. For continuous variables, we used the independent samples t-test for data with normal distribution and the Mann-Whitney U test or Kruskal-Wallis test for data that were not normally distributed.
To assess prognostic ability, we determined the area under the receiver operating characteristic curve (AUROC) 24 with 95% CI for UCH-L1 and compared it with the previously determined AUROC of NSE. We used the bootstrap method to compare AUROCs.
We determined cut-off values for UCH-L1 at 24 and 48 h after CA to predict unfavourable outcome. We used the Youden method 25 to determine the optimal cut-off. To minimize the number of patients with false positive results, we determined cut-off values with high specificities (95% and 99%). We also calculated positive predictive values (PPVs), negative predictive values (NPVs) and positive likelihood ratios (LR+) with 95% CIs for those cut-off values.
We also evaluated the ability of UCH-L1 to identify patients with a high probability of favourable outcome. For this, we determined cut-off values with high (90%-99%) sensitivities. We calculated PPVs, NPVs and negative likelihood ratios (LRÀ) with 95% CIs for these cut-offs.
We used multivariable logistic regression analysis to analyse the independent association of UCH-L1 with unfavourable 12-month outcome and with risk of in-hospital death. We selected significant clinical variables to the models. To predict unfavourable outcome at 12 months, the variables included were patient's age, time from collapse to return of spontaneous circulation (ROSC), witnessed collapse and initial rhythm (shockable or non-shockable).
Data on whether the arrest was witnessed were non-significant in the model predicting death in hospital and thus were not included.

| Comparison with NSE
The AUROCs for NSE as a predictor of unfavourable outcome at      Table 3.
To evaluate the ability of UCH-L1 to identify patients with a high probability of favourable outcome at 12 months, we determined cutoff values to predict unfavourable outcome with sensitivities of 90%-99%. These high sensitivities resulted in low specificities (Table 4).

| DISCUSSION
In this post-hoc study on ICU-treated OHCA patients included in the observational FINNRESUSCI study, we found that UCH-L1 concentrations were elevated for patients with unfavourable outcomes. However, the power of UCH-L1 to discriminate between patients with poor prognosis and those with good prognosis was only moderate, as reflected by AUROCs below 0.7. The prognostic ability of UCH-L1 was not better than that of NSE. Moreover, the sensitivities of UCH-L1 corresponding to high (95%-99%) specificities were lower (9%-18%) compared to those we found for NSE (37%-46%) in our previous study on the same study population. 23 These results do not support the use of UCH-L1 to aid in prognostication and clinical decision-making in patients resuscitated from OHCA.
In some previous studies, the prognostic abilities of UCH-L1 neurofilament light in FINNRESUSCI study patients 23 was, albeit very good, somewhat poorer than that found in some other studies. 28,29 The most likely explanation for these differences is the unselected cohort of patients included in the observational FINNRESUSCI study compared to studies focusing on cardiogenic OHCA with most patients having a shockable initial rhythm. Neuronal biomarkers are quite good in predicting unfavourable outcome caused by HIBI, which is the most common reason of death after CA. 30 However, some deaths are the result of extracerebral causes, and neuronal biomarkers may not be able to predict those. We have previously shown that the prognostic ability of NSE is poor in old patients and in those with a short time from collapse to ROSC, 14 and a plausible explanation is that in these patient groups the cause of death may often be another cause than HIBI.
In our study outcome was assessed at 12 months after OHCA, whereas some other studies have assessed outcomes at 6 months. 20,27 Some of our patients may have initially recovered but later died of causes unrelated to the CA. It is also possible that there may be differences in laboratory methods. 31 In present-day prognostication after CA, not only identifying imminent unfavourable outcome but also identifying a high likelihood of favourable outcome is an important topic. [32][33][34] There is a need for reliable prognostic tools to detect individuals with good prognosis. In our study, cut-off values of UCH-L1 yielding high sensitivities resulted in low specificities. Thus, UCH-L1 demonstrated only limited value in identifying patients with good prognosis, in accordance with the recent study by Moseby-Knappe et al. 35 The timing of blood samples is a relevant issue. In our study, blood samples were drawn at 24 and 48 h after ROSC. In one study, the concentrations of UCH-L1 in CA patients with unfavourable outcome were high already at the time of ICU admission. 36 In a study on TBI patients, the mean half-life of UCH-L1 in serum was 13 h and median half-life 9 h. 19 After hypothermic circulatory arrest in canines, serum UCH-L1 at 8 h was able to predict brain damage. 37 In a study on hypoxic-ischemic encephalopathy in newborns, the highest UCH-L1 concentrations were detected at 0-6 h after delivery, and concentrations decreased during first 24 h. 38 As samples for biomarker measurements can be easily taken in the early phase of care, a sensitive biomarker with concentrations increasing rapidly after CA might be useful. The usefulness of UCH-L1 measured from very early samples may be worth investigating.

| Strengths and limitations
This study has several strengths. It is a multi-centre nationwide study with all five Finnish university hospitals and 14 of the 15 nonuniversity central hospitals participating. The study population was unselected, including patients with shockable and non-shockable initial rhythms and all types of CA. The outcome assessor was blinded for clinical data and biomarker results. However, there are also limitations. First, the blood samples were from patients included in the original study more than 10 years ago, and it is likely that there have been changes in the treatment of resuscitated patients. Second, we have no data on methods used for prognostication when considering withdrawal of life-sustaining treatments.

| CONCLUSION
In this post-hoc laboratory study on OHCA patients, UCH-L1 measured at 24 or 48 h after CA had limited ability to predict outcome.
UCH-L1 was not superior to NSE.