Comparison of the prognostic value of early-phase proton magnetic resonance spectroscopy and diffusion tensor imaging with serum neuron-specific enolase at 72 h in comatose survivors of out-of-hospital cardiac arrest—a substudy of the XeHypotheca trial

Purpose We compared the predictive accuracy of early-phase brain diffusion tensor imaging (DTI), proton magnetic resonance spectroscopy (1H-MRS), and serum neuron-specific enolase (NSE) against the motor score and epileptic seizures (ES) for poor neurological outcome after out-of-hospital cardiac arrest (OHCA). Methods The predictive accuracy of DTI, 1H-MRS, and NSE along with motor score at 72 h and ES for the poor neurological outcome (modified Rankin Scale, mRS, 3 − 6) in 92 comatose OHCA patients at 6 months was assessed by area under the receiver operating characteristic curve (AUROC). Combined models of the variables were included as exploratory. Results The predictive accuracy of fractional anisotropy (FA) of DTI (AUROC 0.73, 95% CI 0.62–0.84), total N-acetyl aspartate/total creatine (tNAA/tCr) of 1H-MRS (0.78 (0.68 − 0.88)), or NSE at 72 h (0.85 (0.76 − 0.93)) was not significantly better than motor score at 72 h (0.88 (95% CI 0.80–0.96)). The addition of FA and tNAA/tCr to a combination of NSE, motor score, and ES provided a small but statistically significant improvement in predictive accuracy (AUROC 0.92 (0.85–0.98) vs 0.98 (0.96–1.00), p = 0.037). Conclusion None of the variables (FA, tNAA/tCr, ES, NSE at 72 h, and motor score at 72 h) differed significantly in predicting poor outcomes in this patient group. Early-phase quantitative neuroimaging provided a statistically significant improvement for the predictive value when combined with ES and motor score with or without NSE. However, in clinical practice, the additional value is small, and considering the costs and challenges of imaging in this patient group, early-phase DTI/MRS cannot be recommended for routine use. Trial registration ClinicalTrials.gov NCT00879892, April 13, 2009. Supplementary Information The online version contains supplementary material available at 10.1007/s00234-022-03063-z.


Supplementary Information -Online resource 1
. Neurological outcome at six months after out-of-hospital cardiac arrest in the study group using the modified Rankin Scale score (page 7) Supplementary Table S4. Results of NSE in patients with good and poor neurological outcome at 6-months (page 8)

References of the Supplementary Information (page 9)
This supplementary material has been provided by the authors to give readers additional information about their work.

Neurological prognostication after out-of-hospital cardiac arrest in hypothermia treated patients
Neurological prognostication is accomplished if the patient remains unconscious 12 h after rewarming. All sedative medication must have been discontinued 12 h previously, and in hypothermia treated patients prognostication is performed after 72 h post resuscitation. Neurological examination is performed by consultant neurologist consisting:

Conclusion
As signs for poor prognosis are: 1. CT/MRI: general cerebral edema with sulcal effacement and wide-spread ischaemia with loss of margins of brain white and grey matter 2. EEG: generalized suppression (<20 V) or burst suppression, generalized epileptic activity or periodic epileptiformic discharges (PED) with background activity suppression, lack of spontaneous variation and lack of reactivity to external stimuli 3. Continuous refractory to treatments myoclonic status epilepticus with permanent unconsciousness 4. Serum NSE values: ascending trend 24-48 h 5. Unresponsiveness to painful stimuli or extension as the best motor response at 72 h 6. Absent brain stem reflexes at 72 h 7. Bilateral absence of thalamocortical sensory evoked potentials (SEP) 8. Generalized diminished cortical diffusion on MRI

Inclusion and exclusion criteria Inclusion criteria
Witnessed cardiac arrest Ventricular fibrillation Non-perfusing ventricular tachycardia Presumed cardiac origin Age 18-80 years Start of resuscitation by emergency medical personnel within 15 minutes Return of spontaneous circulation within 45 minutes Decision for therapeutic hypothermia treatment by attending physician

Exclusion criteria
Hypothermia (core temperature < 30 C) Unconsciousness before collapse (cerebral trauma, intoxication etc.) Computer tomography scan indicating cerebral pathological reason for cardiac arrest Responding to verbal commands after return of spontaneous circulation (ROSC) Pregnancy Coagulopathy Systolic arterial pressure less than 80mmHg lasting >30 minutes after ROSC Mean arterial pressure less than 60mmHg lasting <30 minutes after ROSC Hypoxemia (arterial oxygen saturation <85%) lasting >15 minutes after ROSC Factors making participation in follow-up implausible Enrolment in another interventional trial

Supplementary Methods. Complete magnetic resonance imaging and spectroscopy protocol DTI image processing details
Preprocessing of the diffusion tensor imaging data was done using the DTIprep quality control software. 1 The following steps were performed: Diffusion information checks (ensuring correct diffusion gradient orientations, gradient b-values). Inter-slice brightness artifact detection via normalized correlation analysis between successive slices within a single DWI volume. Interlaced correlation analysis for detection and removal of "venetian blind" artifacts and motion within a single DWI volume. Co-registration to an iterative average over all the baseline images. Eddy-current and motion artifact correction, including appropriate gradient direction adjustments. Residual motion detection to ensure all DWI volumes are well registered.
The diffusion tensor image model was fitted using the dtifit tool (FSL 6.0, Analysis Group, FMRIB, Oxford, United Kingdom). FA images were aligned to a typical FA image in the dataset using the nonlinear registration tool FNIRT, which uses a b-spline representation of the registration warp field. Then aligned dataset was affinetransformed into 1x1x1mm3 standard space (MNI152). Following the default tract-based spatial statistics (TBSS) pipeline, all individual patients' spatially normalized fractional anisotropy images were projected onto a skeletonized mean fractional anisotropy map for statistical voxel-wise regression analysis between the two patient groups 2, 3 . The skeleton represented the centers of all white matter structures that were generally common to the subject involved in a study. Each subject's aligned, non-skeletonized FA data were then projected on the mean FA skeleton in such a way that each skeleton voxel takes the FA value from the local center of the nearest relevant tract. 3 These projected FA values were used for voxel wise statistical analysis. 2 Mean FA value of white matter was calculated as a mean value of all the voxels in the skeleton.
As described previously, the tract-wise distribution of percentages of voxels with significantly lower fractional anisotropy was also analysed. 4 Localization and labeling of the tracts were confirmed and identified with JHU white-matter tractography atlas. 5

Metabolite exclusion criteria
Metabolite data with poor quality was omitted by using spectral and fitting quality parameters as exclusion criteria. The criteria were determined as follows: SNR had to be higher than 2 and FWHM lower than 0.15 ppm. Standard deviation of the fitting of tCr signal had to be less than 20%. Similar or corresponding limits have been used previously [6][7][8] , however, in this study the fitting quality of tNAA was not included into the criteria to avoid affecting the correlations between the metabolite level and patient outcome, for example, level of tNAA might decrease close to zero in case of severe neuronal damage, thus making the fitting procedure of tNAA more prone to errors.

Corrections for T2 and T1 relaxation effects in 1H-MRS data
Metabolite concentration values (Smeasured) were corrected for T2 and T1 relaxation effects as in 9 using the relaxation rates determined by Zaaraoui et al 10 as follows: where Scorrected is the T2 and T1 corrected metabolite signal.