Propofol Requirement and EEG Alpha Band Power During General Anesthesia Provide Complementary Views on Preoperative Cognitive Decline

Abstract

Background: Although cognitive decline (CD) is associated with increased post-operative morbidity and mortality, routinely screening patients remains difficult. The main objective of this prospective study is to use the EEG response to a Propofol-based general anesthesia (GA) to reveal CD. Methods: 42 patients with collected EEG and Propofol target concentration infusion (TCI) during GA had a preoperative cognitive assessment using MoCA. We evaluated the performance of three variables to detect CD (MoCA < 25 points): age, Propofol requirement to induce unconsciousness (TCI at SEF95: 8-13 Hz) and the frontal alpha band power (AP at SEF95: 8-13 Hz). Results: The 17 patients (40%) with CD were significantly older (p < 0.001), had lower TCI (p < 0.001), and AP (p < 0.001). We found using logistic models that TCI and AP were the best set of variables associated with CD (AUC: 0.89) and performed better than age (p < 0.05). Propofol TCI had a greater impact on CD probability compared to AP, although both were complementary in detecting CD. Conclusion: TCI and AP contribute additively to reveal patient with preoperative cognitive decline. Further research on post-operative cognitive trajectory are necessary to confirm the interest of intra operative variables in addition or as a substitute to cognitive evaluation.

Keywords: EEG signal; alpha band power; brain age; cognitive decline and dementia; general anesthesia (GA).

Conflict of interest statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Touchard, Cartailler, Levé, Serrano, Sabbagh, Manquat, Joachim, Mateo, Gayat, Engemann and Vallée.

Figures

Figure 1

Predicting CD from TCI and AP. We assessed CD prediction based on 3 models including the age (AGE), TCI and AP (HELP1), and the three variables altogether (HELP2). (A) We reported logistic based classification results for the three models in a form of ROC curves with their associated AUC, which were 0.832, 0.880, and 0.899 for AGE, HELP1 and HELP2, respectively. Despite better AUC for HELP2, model comparison showed that (HELP2) does not bring significantly more information than HELP (p = 0.032 vs. 0.064, likelihood ratio test). We therefore selected HELP1 to model the risk of CD occurrence. (B) We investigated the relationship between AP (y-axis), Propofol TCI (a-axis), preoperative MoCA (circles) and the probability to present CD estimated by HELP1 (color gradient). CD+ and CD– patients are depicted by black circles and white circles, respectively. Cold, purple colors indicated high probability to present CD. One can readily see that low MOCA scores were associated with lower TCI and lower alpha power. Some cases are discussed in Figure 4 concerning patients with MoCA and HELP mismatch [white circles in a cold zone, TCI 2.5 μg/ml, AP at 6 db (Figure 4C) and black circles in a warmer zone, TCI 3 μg/ml and alpha band at 10 db (Figure 4D)].

Figure 2

Marginal effects of TCI and alpha power of CD probability. The present triptych shows effect of TCI and AP on the patient's probability to be CD, estimated from the HELP model. On the panel (A), marginal effect means (circle) and standard deviation bar errors are reported for TCI (red) and AP (blue). The two variables contribute to predict the onset of CD among patients, although TCI appears to more strongly affect CD probability compared to AP. Panels (B,C) show conditional marginal effects reflecting change in one variable effect relatively to the other variable values. The AP effects is the most important at a TCI of 3 μg/ml(B), while TCI strongly impact prediction for an AP of 5.5 dB. Consequently, effect of AP in predicting CD is not homogeneous as TCI changes, with a maximal effect near 3 μg/ml.

Figure 3

Generalization testing on burst suppression data. To probe the generalization of the proposed model beyond the given sample we considered the observational data from Touchard et al. (2019) in which burst suppression was studied as a proxy for cognitive decline. The dataset comprised 56 patients sedated with Propofol. None of these patients was included in the previous analyses. In a first step (A), we applied the models as defined in the previous analyses (Figures 1, 2) to predict the occurrence of IES during the induction period. HELP1 (TCI + ABP) predicted IES (AUC=0.94, orange) better than AGE (AUC=0.80, yellow). Prediction did not increase significantly for HELP2 (TCI + ABP + Age) as compared to HELP1. We then inspected the HELP1 (IES) model (B). As in previous analyses, TCI (blue) and ABP (red) showed complementary average marginal effects (mean±SE: −0.25±0.09, −0.22±0.06 for TCI and ABP, respectively). We then directly quantified generalization performance (C) of the HELP1 (CD) on the IES dataset (AUC=0.91). The results suggest that the proposed HELP model generalizes beyond the observed sample and captures cognitive and physiological factors related to postoperative outcome.

Figure 4

The potential contribution of intra operative HELP model in relation to preoperative MoCA in the detection of cognitive decline of anesthetized patients. For two subjects (A,B), there are classification concordance between MoCA and HELP model. These are deliberately caricatural situations in order to clearly differentiate the intraoperative characteristics that exist between a subject with a low MoCA (A) and young subject without cognitive impairment (B). In these situations, neither the HELP model, nor the pre-operative cognitive evaluation provides information in relation to the date of birth. An 80-year-old subject (A) is at risk of cognitive complications and a 32-year-old patient (B) is not. Things could be more interesting in MoCA and HELP mismatch situations for middle-aged patients (C,D). If subject C presents a normal pre- operative MoCA and is relatively young, it is striking to observe his weak peroperative variables. In fact, this 62-year-old patient suffered a ruptured aneurysm at age 42 with a major subarachnoid hemorrhage that warranted a lengthy intensive care unit hospitalization and described memory and attention complaints later in life. HELP model could appear more robust to detect past and so maybe future brain suffering. On the contrary, the patient D (68 years old woman) presents a robust HELP score derived from the model, discordant with a low preoperative MoCA. The clinical history of this patient with difficult socioeconomic conditions shows chronic anxiety disorders and a marked apprehension of the upcoming surgery. Considering these two situations, the intraoperative variable could be used to help stratify perioperative cognitive risk when MoCA is taken in default, particularly for middle-aged patients (50 to 70 years old).