Comparison of Prognostic Value of Early Phase 1H Magnetic Resonance Spectroscopy and Diffusion Tensor Imaging With Neuron-Specic Enolase at 72 Hours in Comatose Survivors of Out-of-hospital Cardiac Arrest – A Sub-Study of the Xe-Hypotheca Trial

Background Guidelines recommend brain imaging for neurological prognostication after out-of-hospital cardiac arrest (OHCA). We aimed to evaluate the predictive accuracy of early phase diffusion tensor imaging (DTI) and proton magnetic resonance spectroscopy (1H-MRS) combined with neuron-specic enolase (NSE) and selected clinical variables for poor neurological outcome after OHCA. an earlier phase to assess the neurological outcome in this patient group. We assessed the predictive values for 6-month neurological outcome, dichotomized as good (mRS 0-2) and poor (mRS 3-6), of fractional anisotropy from DTI, several brain metabolites from 1H-MRS each obtained by MRI (magnetic resonance imaging) and neuron-specic enolase (NSE) performed in comatose survivors within 72 hours after OHCA. Additional post-hoc analyses were performed to nd a model with the best predictive power for poor neurological outcome at six months after OHCA.

Based on the current guidelines neurological prognostication is recommended in patients who exhibit an extensor motor response to pain at 72 hours or later after cardiac arrest [4]. A multimodal approach with a combination of clinical assessment, serum biomarkers, electroencephalography, somatosensory evoked potentials and neuroimaging is recommended during the early phase at 3-5 days after cardiac arrest.
However, a poor outcome cannot be predicted with certainty and the assessment may be confounded by contradictory results from the different assessment modalities [4,5]. Therefore, there is an unmet need for new biomarkers to improve the accuracy of early-phase prognostication in order to identify patients with a potential poor neurological outcome.
Among conventional methods of neuroimaging the value of diffusion weighted imaging (DWI) in neurological prognostication is well-demonstrated; however, DWI has been reported to underestimate the extent of ischemic injury during the rst three days after OHCA [6]. While gray matter has classically been thought to be more sensitive to hypoxic-ischemic brain damage, white matter is also highly vulnerable even in the early stages of ischemia [7]. Diffusion tensor imaging (DTI) is an extension of DWI that allows evaluation of microstructural integrity of brain white matter using directional assessment of water diffusion, thus potentially being more sensitive than DWI to detect white matter damage in OHCA patients [8]. We have previously demonstrated that impaired white matter micro-integrity, re ected by lower fractional anisotropy in non-survivors than in the survivors, was caused by demyelination in the early phase after OHCA [9].
Proton magnetic resonance spectroscopy (1H-MRS) is another advanced magnetic resonance technique with some evidence of prognostic value in hypoxic-ischemic brain damage in adults after stroke and cardiac arrest as well as in asphyxiated neonates [10][11][12].
DTI performed between 7 and 28 days after cardiac arrest has shown great promise to accurately predict neurological outcome [12]. As de ned in the original study protocol, the purpose of this study was to test the hypothesis that DTI and/or 1H-MRS could also be used at an earlier phase to assess the neurological outcome in this patient group. We assessed the predictive values for 6-month neurological outcome, dichotomized as good (mRS 0-2) and poor (mRS 3-6), of fractional anisotropy from DTI, several brain metabolites from 1H-MRS each obtained by MRI (magnetic resonance imaging) and neuron-speci c enolase (NSE) performed in comatose survivors within 72 hours after OHCA. Additional post-hoc analyses were performed to nd a model with the best predictive power for poor neurological outcome at six months after OHCA.

Study Design
Xe-HYPOTHECA trial (ClinicalTrials.gov NCT00879892) was a randomized 2-group single-blinded phase 2 clinical drug trial at two multipurpose intensive care units (ICU) in Finland. The study was approved by the ethics committee of the Hospital District of Southwest Finland and the institutional review boards of the Helsinki University Hospital and the Finnish Medicines Agency. All patients' next of kin or legal representative gave written informed assent within 4 hours after hospital arrival. Consent was sought from patients when they regained consciousness. As described earlier, an independent data and safety monitoring committee reviewed data after enrolment of every 4 patients and after each 6-month interval [9].
The study was conducted according to good clinical practice and the latest revision of the Declaration of Helsinki. Study design and methodology was consistent with the STARD guidelines for reporting diagnostic accuracy studies [13].

Participants
Consecutive comatose survivors of witnessed OHCA from an initial shockable rhythm admitted to the Turku and Helsinki University hospitals between August 2009 and September 2014 were screened for eligibility.
Detailed inclusion and exclusion criteria are listed in eAppendix in the Supplement.
We have previously reported the primary and secondary clinical end points of the Xe-HYPOTHECA trial [9,14]. The protocol of the Xe-HYPOTHECA trial has also been published [9].

Randomization and Blinding
The patients were allocated in a 1:1 ratio with random block sizes of 4, 6, and 8 to receive either therapeutic hypothermia treatment alone for 24 hours or inhaled xenon (LENOXe, Air Liquide Medical GmbH, Düsseldorf, Germany) in combination with hypothermia for 24 hours as described earlier [9]. The neurological end-point evaluators as well as the patients were blinded to the treatment.

Procedures
MRI imaging was scheduled to be performed within 16 hours of rewarming i.e. 36-52 hours after OHCA.
Patients were kept intubated and sedated (with sedation interruptions after completion of rewarming) until brain imaging was performed, regardless of neurological status. A predetermined prognostication protocol (eAppendix in the Supplement) was used to preclude premature decisions to withdraw life-sustaining therapy. DTI and 1H-MRS results were not available at the time of prognostication or at any time during the intensive care. The clinical outcome was evaluated at six months after OHCA with modi ed Rankin Scale (mRS) by experienced neurologists.
After rewarming was completed sedation interruptions were initiated and performed every 6 to 12 hours throughout intensive care stay. Motor score of the Glasgow Coma Scale was assessed during each sedation interruption by either trained intensive care nurse or on-duty intensive care physicians. NSE serum concentration (Immuno-Electro-Chemi-Luminescent assay, Roche Diagnostics GmbH, Mannheim, Germany) was determined at hospital arrival, and at 24 hours, 48 hours and 72 hours after OHCA. An electroencephalogram was recorded only if it was clinically indicated.
Siemens Magnetom Verio 3T scanner (Siemens Medical Solutions, Erlangen, Germany) with 12-element Head Matrix coil was used in both MRI centers. DTI data were acquired using diffusion weighted spin-echo echo planar imaging (SE-EPI) sequence with 20 diffusion encoding directions (see eTable1 in the Supplement for details).
FSL software library (version 6.0, Analysis Group, FMRIB, Oxford, United Kingdom) was used for processing the DTI images, following the tract-based spatial statistics (TBSS) processing [15,16]. This observerindependent and hypothesis-free method has the ability to spatially locate group differences in the DTI data. Mean fractional anisotropy value of white matter was calculated as a mean value of all the voxels in the skeleton. (see eMethods in Supplement for details).
1H-MRS data were acquired from the region of basal ganglia by utilizing Chemical Shift Imaging (CSI) technique (see eTable 2 for details). Acquired data were analyzed using the LCModel software (version 6.3-0C) [17]. An average of all analyzed voxels, except the ones containing cerebrospinal uid (CSF) were selected for the nal analysis (see Figure 1). The metabolite concentration values were corrected for relaxation effects (eMethods) but absolute concentration values were not feasible to use. Therefore, the amount of tNAA (total N-acetyl aspartate) and total choline were expressed as ratios over total creatine, i.e. tNAA/tCr and tCho/tCr, as it is expected to remain stable. In addition, apart from the tNAA/tCho ratio, the amount of tNAA and tCho was assessed as these individual parameters are related to neuronal density, activity and integrity [18].

Statistical analysis
The sample size of 110 patients was based on a power analysis of the fractional anisotropy values from brain magnetic resonance imaging, i.e. the primary end-point of the Xe-HYPOTHECA trial [9]. The categorical demographic data and baseline clinical characteristics between groups of mRS 0-2 and mRS 3-6 were compared with chi-square or Fisher's exact test. Two-sample t-test or Mann-Whitney U-test was used to test the differences in continuous demographic data and baseline clinical characteristics between the groups mRS 0-2 and mRS 3-6. The normality of continuous variables was evaluated visually using histograms. The mean differences in mean fractional anisotropy, 1H-MRS data and NSE at 48 and 72 hours after OHCA between the groups were tested with two-sample t-test. Age-, sex-, treatment-, and site-adjusted mean differences between the groups were compared with analysis of covariance. NSE values were logtransformed for statistical analysis due to positively skewed distribution. The associations of the mean fractional anisotropy, 1H-MRS and NSE values with 6-month mortality were analyzed by using Cox regression analysis after adjustment for age, sex, treatment, and site. The follow-up time for survival analysis was calculated from the time of cardiac arrest until death or 6 months. The observation was censored in the survival analysis if the patient was withdrawn from the study or was still alive at the end of the 6-month follow-up. Permutation-based voxel-wise statistical analysis with tract-based spatial statistics in conjunction with family-wise error correction was used for multiple comparisons across space to obtain group differences in the white matter tracts [9,15,16].
The results are expressed using adjusted hazard ratios (HR) with 95% con dence intervals (CI). The prognostic values of fractional anisotropy, tNAA/tCr and NSE 72 hours after OHCA and logistic regression derived combined models including selected clinical variables were evaluated post-hoc by calculating the area under the curve (AUC) of receiver operating characteristic (ROC) curve using a nonparametric method. Sensitivity, speci city, positive predictive value (PPV) and negative predictive value (NPV) for each prognostic variable was calculated. Optimal cut-off values were chosen by using the Youden Index (sensitivity+speci city-1).
A 2-sided p-value less than 0.05 was considered statistically signi cant. Statistical analyses were performed with SAS System for Windows, version 9.4 (SAS Institute Inc., Cary, NC) and SPSS Statistics for Macintosh, version 24 (IBM Corp., Armonk, NY). The hazard ratio plot was created using DistillerSR Forest Plot Generator (Evidence Partners, Ottawa, Canada).

Patients
Of the 224 patients screened for eligibility for the Xe-HYPOTHECA trial, 110 were included. Of these, 97 underwent magnetic resonance imaging in a median (inter-quartile range) time of 53 hours (47-64) after OHCA and 93 had both 1H-MRS and DTI data available (Fig. 2).
One patient was withdrawn six days after the index event by the next of kin and therefore 92 out of 93 had applicable mRS data (Fig. 2, eTable 3). At six months after OHCA, 61 patients had good (mRS 0-2) and 31 patients had poor (mRS 3-6) neurological functional outcome. Patient demographics and clinical characteristics are presented in Table 1. NSE, epileptic seizures and motor score NSE at 48 and 72 hours after OHCA were signi cantly higher in patients with poor neurological outcome than in the patients with good neurological outcome at six months ( Table 2, eTable 4). In addition, the number of patients with epileptic seizures, and motor score ≤ 3 at 72 hours, and time to return of spontaneous circulation (ROSC) were signi cantly greater in patients with poor neurological outcome ( Table 2). Nine of the 61 patients with good neurological outcome and 27 out of 31 patients with poor neurological outcome had motor score ≤ 3 at 72 hours after OHCA (Table 2). Ten patients responded appropriately to commands within 48 hours after OHCA; a further 32 patients were responsive to commands between 48 to 72 hours and 25 patients achieved this state later than 72 hours. 25 patients never achieved a motor score of 6, all of whom died. A single measure motor score of ≤ 3 at 72 hours provided the best prognostic value for poor neurological outcome as compared with other methods ( Table  2).

DTI and 1H-MRS results
Mean fractional anisotropy values of the DTI, and tNAA/tCr and tNAA/tCho ratios of the 1H-MRS were signi cantly higher in patients with mRS 0-2 than in patients with mRS 3-6 ( Table 2). Results of the tractbased spatial statistics analysis are visualized with a statistical parametric map (Fig. 3). Results of the age-, sex-, treatment-, and site-adjusted survival analyses are presented in Figure 4.
Receiver operating characteristic analysis and diagnostic tests  Table 2).

Discussion
The main nding of this study was that early stage quantitative DTI or 1H-MRS did not perform better than NSE at 72h for prognosticating poor outcome at six months after OHCA. This was the case also in a subgroup of patients with motor score less than 6 (does not obey commands).
A single measure motor score of ≤ 3 at 72 hours provided the best prognostic value for poor neurological outcome as compared with other methods; NSE at 72 hours was the next best. In earlier studies, a cut-off value ≤ 2 for motor score has revealed low speci city and high sensitivity between 70-80% [20,21]. Here we demonstrate similar sensitivity values with higher speci city; improvement in the latter attribute can be explained by homogenous cardiac arrest population that only included patients with a shockable primary rhythm. Current AUC of 0.85 for NSE is consistent with the values of 0.86 and 0.90 recorded in earlier large studies in TTM-treated patients [22,23]. However, comparing NSE results among studies may be problematic because cut-off values vary and a consistent threshold limit for 0% false positive ratio has not been recommended [4]. Higher NSE cut-off values predict worse outcome with cut-off values of ≥23 as compared with current cut-off value of 21 [4,23]. One explanation may be that in this study the best value of the rst 72 hours was used in order to obtain an assessment for each patient. As a result, in some patients only assessments of the rst 48 hours were applicable which may have led to higher false positive ratio.
Fractional anisotropy is a DTI-derived scalar value that re ects white matter tissue characteristics such as ber density, organization coherence, myelination, and axon diameter [24]. Lower fractional anisotropy values in ischemic white matter probably represent a combination of myelin damage, axonal degeneration, and edema, which all contribute to loss of directional diffusion in white matter tracts [25].
A very recent study demonstrated a prognostic value of decreased mean global white matter fractional anisotropy levels imaged 7 to 28 days after cardiac arrest for long-term neurological outcome with an area under ROC of 0.95 in a subset of 150 patients with a persistent unresponsiveness at day 7 [12]. According to earlier evidence most survivors regain consciousness within a week and usually all of them within 10 days after cardiac arrest [4,5,26,27,28]. However, Velly and colleagues revealed that as many as 22% of patients who were without a response to simple commands a week after cardiac arrest may still have a favorable outcome at six months after cardiac arrest [12]. Current lower prognostic value for fractional anisotropy at early subacute phase can be partly due to evolving white matter injury over time. This suggests that predictive value of fractional anisotropy varies depending on the used imaging time window [29]. Conventional DWI MRI seems to perform better than DTI/MRS in the early phase [30,31].
NAA is ubiquitous in the central nervous system, and is a marker of the integrity of mature neurons; persistent reductions in NAA have been used as a marker of neuronal loss and extent of neurological damage and have prognostic value for outcome in stroke and cardiac arrest [10,12,32]. These earlier interpretations were supported by the present study revealing that ratios of tNAA/tCr and tNAA/tCho in the basal ganglia are independent predictors for poor neurological outcome at six months.
In this study, without DTI and 1H-MRS, the best prognostic value was provided with a combined model of NSE and motor score at 72h and the presence of epileptic seizures within 72h with AUC of 0.92 as early as 72 hours for all consecutively enrolled OHCA patients with ventricular brillation as a primary rhythm.
Addition of DTI and 1H-MRS parameters to this model with or without NSE further increased its diagnostic value (AUC 0.98, p=0.037).
Although statistically signi cant, the absolute difference in diagnostic value is still small considering the cost and challenges of MR imaging of critically-ill patients. Therefore early phase DTI/MRS cannot be recommended for clinical practice or at least it should be reserved for very speci c cases. However, for research purposes, early phase DTI/MRS can be used as viable surrogate marker for white matter injury.
These results warrant further con rmation in another population including also patients with asystole and PEA as primary rhythm.
There are some limitations in this study. First, our results represent a two-center (single country) cohort of patients in successfully resuscitated cardiac arrest victims in whom a shockable rhythm was the initial rhythm at time of resuscitation. Therefore, further validation of the results is required in patients with asystole and pulseless electrical activity. Second, results of pupillary light re exes at 72 hours were applicable only in 49 patients and were therefore excluded from the nal analyses. Third, metabolite concentration ratios instead of absolute values or inter-subject metabolite concentrations were used. The use of absolute concentration values would have required the use of reference solutions for calibrations and information about coil loading and T2 attenuation, or the use of a water-referencing method; none of these options were available for this study. In routine clinical practice, absolute metabolite values are seldom available and thus this approach was not deemed feasible. Fourth, due to a possible bias caused by the neuroprotective effect of the xenon on the prognostic value of the current metrics all statistical analyses were adjusted with the treatment group. Fifth, our model was limited by small number of patients with study end point and therefore all variables with predictive value for poor outcome could not be included.

Conclusions
In summary, early stage DTI or 1H-MRS imaging was not better than NSE at 72h in predicting poor outcome in this patient group. Although the addition of neuroimaging to the best model using clinical and laboratory parameters improved the diagnostic performance, the costs and challenges of MR imaging of critical care patients may outweigh this small bene t. Current results should be prospectively validated in another population of cardiac arrest patients.

Consent for publication Not applicable
Availability of data and materials The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
TL is a member of the Trial Executive Committee for XePOHCAS, a phase 3 RCT designed to test the e cacy and safety of inhaled xenon in patients with post-cardiac arrest syndrome. TL reports grant from Finnish Foundation for Cardiovascular Research. RL is a paid governmental consultant o cial for the National Supervisory Authority for Welfare and Health. AS reports grants from Academy of Finland and Finnish Foundation for Cardiovascular Research during the conduct of the study; consulting fees from GE healthcare, Novartis, Abbot, and Astra Zeneca. Dr. Maze is a founder, board director, and equity shareholder of NeuroproteXeon, a company that intends to commercialize the use of xenon for ongoing acute neurological injury, including its administration to successfully resuscitated patients after out-of-hospital cardiac arrest. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.

Funding
The study was funded by Academy of Finland and by State Research Funding.
Authors' contributions Dr Laitio T had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.  Due to missing data mRS was applicable for N=92.