Data on multimodal approach for early poor outcome (Cerebral Performance Categories 3-5) prediction after cardiac arrest

The data presented in this article are related to our research article entitled ‘Neurophysiological and neuroradiological multimodal approach for early poor outcome prediction after cardiac arrest’ (Scarpino et al., 2018) [1]. We reported two additional analyses, including results gathered from somatosensory evoked potentials(SEPs), brain computed tomography(CT) and electroencephalography(EEG) performed on 183 subjects within the first 24 h after cardiac arrest(CA). In the first analysis, we considered the Cerebral Performance Categories(CPC) 3, 4 and 5a,b (severe disability, unresponsive wakefulness state, neurological death and non-neurological death, respectively) as poor outcomes. In the second analysis, patients that died from non-neurological causes (CPC 5b) were excluded from the analysis. Concerning the first analysis, bilateral absent/absent-pathologic(AA/AP) cortical SEPs predicted poor outcome with a sensitivity of 49.3%. A Grey Matter/White Matter(GM/WM) ratio <1.21 predicted poor outcome with a sensitivity of 41.7%. Isoelectric/burst-suppression EEG patterns predicted poor outcome with a sensitivity of 33.5%. If at least one of these poor prognostic patterns was present, the sensitivity for an ominous outcome increased to 60.9%. Concerning the second analysis, AA/AP cortical SEPs predicted poor outcome with a sensitivity of 52.5%. GM/WM ratio <1.21 predicted poor outcome with a sensitivity of 50.4%. Isoelectric/burst-suppression EEG patterns predicted poor outcome with a sensitivity of 39.8%.


a b s t r a c t
The data presented in this article are related to our research article entitled 'Neurophysiological and neuroradiological multimodal approach for early poor outcome prediction after cardiac arrest' (Scarpino et al., 2018) [1]. We reported two additional analyses, including results gathered from somatosensory evoked potentials (SEPs), brain computed tomography(CT) and electroencephalography (EEG) performed on 183 subjects within the first 24 h after cardiac arrest(CA). In the first analysis, we considered the Cerebral Performance Categories(CPC) 3, 4 and 5a,b (severe disability, unresponsive wakefulness state, neurological death and non-neurological death, respectively) as poor outcomes. In the second analysis, patients that died from non-neurological causes (CPC 5b) were excluded from the analysis. Concerning the first analysis, bilateral absent/absent-pathologic(AA/AP) cortical SEPs predicted poor outcome with a sensitivity of 49.3%. A Grey Matter/White Matter(GM/WM) ratio o1.21 predicted poor outcome with a sensitivity of 41.7%. Isoelectric/burst-suppression EEG patterns predicted poor outcome with a sensitivity of 33.5%. If at least one of these poor prognostic patterns was present, the sensitivity for an ominous outcome increased to 60.9%. Concerning the second analysis, AA/AP cortical SEPs predicted poor outcome with a sensitivity of 52.5%. GM/WM ratio o1.21 predicted poor outcome with a sensitivity of 50.4%. Isoelectric/burst-suppression EEG patterns predicted poor outcome with a sensitivity of 39. 8%

Data
Prognosticating neurological outcome after CA is challenging and should require a multimodal approach that could be used to develop a prognostication algorithm.

Patient management
All CA comatose patients included in the analysis underwent to brain CT upon admission to the emergency room, and EEG and SEP evaluation within the first 24 after CA. The results of these instrumental tests did not affect ongoing patient management, in fact in our clinical practice the withdrawal of life support was not included and treatments were not suspended except where patients with confirmed brain death (BD) were concerned. Patient discharge to a long-term care unit or to a rehabilitation unit was decided, in agreement with the intensivists, according to these neurophysiological and neuroradiological findings associated with the clinical data. Neurological status was determined using CPC at two follow-up points: at hospital discharge, looking at the chart review, and, for patients surviving at hospital discharge, at least 6 months after CA, by telephone interview. SEP, EEG and brain CT evaluation SEP prognostic power was based on the evaluation of the presence/absence or of the amplitude value of the cortical responses (N20/P25 complex) on both hemispheres. Thus, we identified these six SEP patterns: NN, NP, PP, AN, AP and AA, in which N stands for normal (N20/P25 amplitude is normal), P stands for pathological (N20/P25 amplitude is o1.2 mV or the difference between the two sides is greater than 50%) and A stands for absent, if no reproducible cortical components could be identified in the presence of a cervical potential [1][2][3].
EEGs were classified according to the terminology for EEGs recorded in ICU [4]. The continuity and the voltage of the background activity were the main parameters taken in to account for EEG classification. Thus, the main patterns identified were: continuous; nearly continuous; discontinuous; burst-suppression; suppression; epileptiform discharges, low voltage (voltage o20 μV) and isoelectric. Isoelectric (voltage o2 μV) recordings were identified, although the original classification did not distinguish them from suppressed activity (voltage o 10 μV) [5].
Brain CT prognostic power is based on the GM/WM ratio as a measure of density. In particular, in our analysis, we performed density measurements limited to the basal ganglia level, according to a previously reported method [6], as the GM/WM ratio ¼(caudate nucleus þ putamen)/(corpus callosum þposterior limb of the internal capsule).
For further details regarding SEP and EEG recording and brain CT acquisition, refer to the supplementary data of the related research article [1].

Statistical analysis
We used the receiver operating characteristic (ROC) to determine the sensitivity at a specificity of 100% (false positive rate¼ 0%) for SEPs, EEG, and GM/WM ratio, in relation to poor outcome. We expressed the performance of each measure for predicting poor outcome as the area under the ROC curve (AUC). The dependent variable, outcome, was dichotomous (good/poor). Two set of analysis were performed. In the first one, we considered as a good outcome no or minor neurological deficits (CPC ¼1) and moderate disability (CPC ¼2), and as a poor outcome severe disability (CPC ¼3), Table 2 Single and multimodal approach-sensitivity and negative predictive values (at 100% specificity) for poor outcome prediction.

Parameter
CPC 3-4-5a-5b "poor" unresponsive wakefulness state (CPC ¼4) and death (CPC ¼5a-b). In the second analysis, having distinguished patients who died from neurological causes (CPC ¼5a) from those who died from nonneurological causes (CPC ¼5b), according to the suggestion of Sandroni et al. [7], we evaluated the predictive value of the three tests in all patients except in those that died from non-neurological causes. Then, in poor outcome, group we included subjects with CPC¼ 4 and CPC ¼ 5a. A p-value o0.05 was considered statistically significant. Statistical analysis was performed using the Stat-View Software package (SAS Institute).  Table 1 shows the demographic characteristics of the 183 subjects that were subjected to all three tests within the first 24 h after CA.

Multimodal approach at 100% specificity for poor outcome prediction (CPC 3-4-5a-5b)
After determining the optimal cut-off for a specificity of 100% for each parameter, we combined the results of the three tests in the hypothesis that the ominous prognostic findings of each test were not all present simultaneously in the same patient, thus evaluating whether the availability of more than one test in the same patient could increase the predictability of a poor outcome. Actually, 89 patients had at least one poor prognostic parameter (grade 2 SEPs, malignant EEG patterns, GM/WM ratio o 1.21). When two tests were considered, if at least one of the patterns predicting a poor outcome was present, the sensitivity increased by 49.3% (obtained with the best single performing test, SEPs) reaching a maximum of 57.7% (obtained by SEPs and brain CT combination). Finally, when all three tests were considered, if at least one of the patterns predicting a poor outcome was present, the sensitivity for a poor prognosis increased to a maximum of 60.9%.

Transparency document. Supporting information
Transparency data associated with this article can be found in the online version at https://doi.org/ 10.1016/j.dib.2018.05.118.