Outcome prediction in comatose patients after cardiac arrest based on progression of auditory discrimination and neuronal background activity at rest

Details

Ressource 1 Under indefinite embargo.
UNIL restricted access
State: Public
Version: After imprimatur
License: Not specified
Serval ID
serval:BIB_54E1C3893430
Type
A Master's thesis.
Publication sub-type
Master (thesis) (master)
Collection
Publications
Institution
Title
Outcome prediction in comatose patients after cardiac arrest based on progression of auditory discrimination and neuronal background activity at rest
Author(s)
JERAD N.
Director(s)
DE LUCIA M.
Institution details
Université de Lausanne, Faculté de biologie et médecine
Publication state
Accepted
Issued date
2021
Language
english
Number of pages
31
Abstract
1.1 Introduction: Outcome prognostication of post-anoxic coma continues to be challenging. Most clinical tests are based on qualitative observations and are mostly indicators of poor prognosis. Two markers for favourable outcome prediction in comatose patients after cardiac arrest have been introduced: the first based on the progression of electroencephalogram (EEG)- based auditory discrimination, the second derived from the EEG brain activity at rest. Here, I assessed whether the combination of these two markers could improve the overall outcome prediction in comparison to each of the two tests separately and with respect to the currently used clinical markers.
1.2 Method: On the first day after coma onset, we performed an EEG to record the brain activity at rest and the brain responses to auditory sequences while delivering tones following a mismatch negativity paradigm. On the second day, we repeated the recording for measuring the neural responses to auditory sequences. For the comatose patient’s outcome prediction, we computed the EEG power spectra on the first day and the progression of the EEG-based auditory discrimination between the first and second recording. Patient outcome was defined as favourable or unfavourable according to the best state they were within three months.
1.3 Results: Among the 138 patients, 80 had a favourable outcome; 31/80 had a high-power spectra value, and 21/80 had an increase in auditory discrimination. The power spectra’s predictive performance had better results on all parameters (PPV=1, NPV=0,54, Accuracy =0,64, Sensitivity= 0,39, Specificity=1) except for the sensitivity than the progression of auditory discrimination (PPV= 0,66, NPV= 0,52, Accuracy= 0,58, Senstivity =0,49, Specificity= 0,69). When we combined them, we slightly improved the accuracy and the sensitivity with respect to the power spectra only (PPV=1, NPV=0,56, Accuracy= 0,67, Sensitivity=0,43, Specificity=1). We observed that in addition to the patients with a high-power spectra value, already predicted on the first day, the improvement in auditory discrimination could predict the favourable outcome in other three patients. Overall –based on the combination of the two markers- the number of correctly predicted patients with favourable outcomes was 34 out of 80 patients, and the number of correctly predicted patients with unfavourable outcomes was 24 out of 58 patients.
1.4 Conclusion: Results show that combining the outcome prediction results based on resting state EEG on the first day of coma and on the progression of auditory discrimination between the first two days improves the accuracy of the outcome prediction. These results can be interpreted in terms of different degrees of preservation of the neural substrates underlying a healthy resting state activity and intact auditory processing. The fact that these indicators rely on different neural networks is an advantage: depending on the damage caused by the cardiac arrest, practicians could lean on one or the other with a certain flexibility to predict the patient’s outcome. Furthermore, those markers are easy to implement, and clinicians could use them in cases of uncertain outcomes.
Keywords
Auditory discrimination, Cardiac arrest, Coma, EEG
Create date
07/09/2022 15:52
Last modification date
21/09/2023 5:58
Usage data