Comparison of bispectral index and patient state index during general anesthesia with remimazolam

Processed electroencephalographic indices, such as the Bispectral Index (BIS) and Patient State Index (PSI), are widely used to assess anesthetic depth; however, their comparative behavior under remimazolam remains unclear. In practice, either index may deviate from its target range despite stable anesthesia, and the response of the companion monitor is uncertain. This study examined whether the two indices moved in parallel or diverged when one departed from its nominal range during remimazolam anesthesia.

In this prospective, single-center study, BIS and PSI were simultaneously recorded at 21 predefined time points from induction to emergence in 60 adults who received remimazolam-based general anesthesia. Remimazolam was induced at 12 mg/kg/h and maintained at 1 mg/kg/h; all patients received neuromuscular blockade (train-of-four ≤ 1). Two trained observers captured synchronized monitor screenshots and documented the Signal Quality Index, electromyography (EMG), and suppression ratio. Analyses used 994 paired observations: pooled Pearson and repeated-measures correlations, simple linear regression (PSI vs. BIS), repeated-measures Bland-Altman analysis, and a linear mixed-effects model for BIS-PSI differences, including anesthetic phase and EMG as fixed effects and patient as a random intercept.

BIS and PSI showed a strong correlation (r = 0.89) and linear relationship (PSI = 0.936 × BIS - 7.26; R² = 0.79). The repeated-measures Bland-Altman bias was + 10.9 (95% CI 10.3-11.4) with 95% limit of agreement (LoA) - 7.8 to 29.5. Phase-wise modeling showed minimal discrepancies while awake and larger differences during intubation and emergence (Tukey-adjusted p < 0.001), even after EMG adjustment. Both indices tracked the anesthetic depth in parallel but were not numerically interchangeable.

During remimazolam anesthesia, BIS and PSI exhibited parallel temporal trends, indicating shared cortical dynamics despite distinct algorithms. When either index deviated, either elevated or suppressed, the other generally moved in the same direction, reflecting consistent responsiveness rather than device artifacts. Phase-dependent discrepancies and wide agreement limits suggest that each index should be interpreted within its reference framework. Understanding this "parallel yet nonidentical" behavior may support context-aware, device-specific depth monitoring.

University Hospital Medical Information Network Clinical Trials Registry (UMIN-CTR, UMIN000058212; registered on June 18, 2025).