Utility of magnetic resonance spectroscopy in predicting favorable outcome in adult comatose patients following cardiac arrest

Abstract

Aim

To correlate brain metabolites with clinical outcomes using magnetic resonance spectroscopy (MRS) in cardiac arrest (CA) patients and assess their prognostic performance compared to quantitative apparent diffusion coefficient (ADC) maps.

Methods

Comatose CA patients who underwent MRI and concurrent MRS were included. Primary outcome: coma recovery at hospital discharge; secondary outcome: good neurological function at 6 months (Cerebral Performance Index 1–2, vs. 3–5). Six metabolites were measured in the posterior cingulate gyrus (PCG), parietal white matter, and brainstem. Mean ADC values, and percentage of voxels with ADC <450 and <650 × 10⁻⁶ × mm²/s were computed for whole brain and specific regions. Prognostic performance was compared using Receiver Operating Characteristic (ROC) curves.

Results

Of 94 patients, 25 (27 %) achieved coma recovery, and 22 (23 %) had a good outcome at 6 months. N-acetylaspartate/Creatine (NAA/Cr) in the PCG was most discriminative for coma recovery (median 1.29, IQR 0.21 vs. 0.86, 0.32; p-value <0.0001). NAA/Cr had the highest area under the curve for coma recovery (0.9, 95 % CI 0.84–0.96) and good outcome at 6 months (AUROC 0.88, 95 % CI 0.82–0.95), significantly outperforming all quantitative ADC measurements, except mean ADC of the PCG for the secondary outcome (adj. p-value = 0.064). Multivariable models incorporating NAA/Cr or ADC, alongside clinical and EEG variables, demonstrated improved performance compared to models with clinical and EEG variables alone, though the difference was not statistically significant. Adding MRS to established early predictors of favorable outcome increased the specificity from 67 % to 93 % at 100 % sensitivity.

Conclusion

MRS-derived NAA/Cr in the PCG is a valuable predictor of good outcome in comatose CA patients, outperforming quantitative ADC measurements for coma recovery. Further studies are needed to optimize MRS acquisition for multimodal neuroprognostication.