Age-dependent changes in the power spectrum conflate composite scores to assess brain frailty

Abstract

Objective: Evaluating age-related dependencies in the electroencephalogram (EEG) during induction of general anesthesia and their impact on composite scores used to assess frailty.

Methods: A composite score was derived from spectral edge frequency, total power, alpha power, and the effect-site concentration (Ce) of propofol. All these parameters are influenced by age, brain health, and dosage and speed of drug administration. Correlation coefficients and variance inflation factors were used to determine multicollinearity. Differences in the spectral EEG features of patients with "high" and "low" composite scores were assessed by the area under the receiver operator characteristic curve (AUC) as the statistical test.

Results: The EEG features, total power and alpha power, were strongly correlated (ρ = 0.82). But alpha power (ρ = 0.17) and total power (ρ = 0.2) were only weakly correlated with propofol, indicating a weak model. Additionally, the composite score showed a moderate negative correlation with age (ρ = -0.44). We also observed significant and strong (AUC < 0.3) differences in total power and the power of all EEG bands except gamma between patients with a "high" and a "low" score before loss of responsiveness (LOR).

Conclusion: Patient age significantly influences EEG-based parameters within the score. Importantly, significant differences in spectral EEG features between the groups were already observable before LOR. These differences could allow for early assessment of a patient's brain state and to titrate anesthetic dose before LOR. The study also shows that age should be considered as it can drive models for "frailty".

Significance: Age moderately influenced all subcomponents and should consequently be factored into score interpretation.

Keywords: Age; Anesthesia; EEG; Spectral power; TCI.

Fig. 1

Relationship between our composite score and its components and age. (A) The composite score’s correlation coefficient was ρ = -0.44 [-0.61;-0.22]. (B) The propofol concentrations correlation coefficient was ρ = -0.45 [-0.61; −0.25]. (C) The alpha-band power’s correlation coefficient was ρ = -0.32 [-0.52; −0.10. (D) The total power’s correlation coefficient was ρ = -0.36 [-0.55; −0.14].

Fig. 2

Boxplots and ROC curves when separating ”high-score” and ”low-score” patients at the time of BPTIVA calculation. AUC values as the effect size are symmetrical around 0.50, therefore both AUC < 0.30 or > 0.70 are considered acceptable.

Fig. 3

Differences in absolute EEG band power between ”high-score” and ”low-score” patients during anesthesia induction. (A) ”High-score” patients had significantly higher delta-band power throughout induction except during the first minutes. (B) ”High-score” patients had significantly higher theta-band power throughout induction. (C) ”High-score” patients had significantly higher alpha-band power throughout induction. (D) “High-score” patients had significantly higher beta-band power throughout induction. (E) ”High-score” patients had significantly higher gamma-band power starting around the time of LOR. (F) AUC values as the effect size are symmetrical around 0.50, therefore both AUC < 0.30 or > 0.70 are considered acceptable.

Fig. 4

Differences in relative EEG band power between ”high-score” and ”low-score” patients during anesthesia induction. (A) ”High-score” patients had significantly lower delta-band power starting around 3 to 5 min after LOR. (B) ”High-score” patients did not have significantly higher theta-band power apart from a short episode around 10 min after LOR. (C) ”High-score” patients had significantly higher alpha-band power throughout induction. (D) ”High-score” patients had significantly higher beta-band power starting around the time of LOR. (E) ”High-score” patients had significantly lower gamma-band power starting around 5 min after LOR. (F) AUC values as the effect size are symmetrical around 0.50, therefore both AUC < 0.30 or > 0.70 are considered acceptable.

Fig. 5

Differences in absolute (A-E) and relative (F-J) EEG band power across different levels of composite score values. (A) Patients with higher composite scores had significantly higher delta-band power throughout induction. (B) Patients with higher composite scores had significantly higher theta-band power throughout induction. (C) Patients with higher composite scores had significantly higher alpha-band power throughout induction. (D) Patients with higher composite scores had significantly higher beta-band power throughout induction. (E) Patients with higher composite scores had significantly higher gamma-band power starting around the time of LOR. (F) Patients with higher composite scores had significantly lower delta-band power starting around 3 to 5 min after LOR. (G) Patients with higher composite scores did not have significantly higher theta-band power apart from a short episode around 8 to 12 min after LOR. (H) Patients with higher composite scores had significantly higher alpha-band power after LOR and partly at induction. (I) Patients with higher composite scores had significantly higher beta-band power starting around the time of LOR. (J) Patients with higher composite scores had significantly lower gamma-band power starting around 7 min after LOR. AUC values as the effect size are symmetrical around 0.50, therefore both AUC < 0.30 or > 0.70 are considered acceptable.