Bispectral index during remimazolam versus propofol sedation at matched depths: a randomized, double-blind crossover trial in healthy volunteers

Abstract

Background

The bispectral index (BIS) was derived primarily from propofol data, and whether its thresholds apply to remimazolam is uncertain. We compared BIS between the two agents at matched sedation depths and examined early cognitive recovery.

Methods

In this prospective, randomized, double-blind, two-period crossover trial, 16 healthy volunteers (American Society of Anesthesiologists physical status I) received remimazolam and propofol in random order, ≥ 7 days apart. Each agent was titrated to moderate (Modified Observer’s Assessment of Alertness/Sedation [MOAA/S] 3) and deep (MOAA/S 1) sedation. The confirmatory outcome was mean BIS at deep sedation. All other outcomes were exploratory: BIS at moderate sedation and at loss of the eyelash reflex, emergence time, and cognitive recovery at 30 and 120 min.

Results

At deep sedation, mean BIS was higher with remimazolam (57.6 ± 7.5) than with propofol (46.5 ± 7.9; adjusted difference 11.1 units; 95% CI, 7.3 to 14.8; p < 0.001). In exploratory analyses, BIS was also higher at moderate sedation and at loss of the eyelash reflex (both p < 0.001), and emergence time did not differ (p = 0.606). At 30 min, processing speedand  executive function recovered faster after remimazolam (p = 0.012 and p < 0.001), whereas verbal memory did not differ between agents. All domains converged by 120 min. Adverse events were uncommon, and the trial was not powered to compare them.

Conclusion

At clinically matched sedation depths, remimazolam produced higher BIS values than propofol. Spectral data were not recorded, so the basis of this offset is undetermined, and the cognitive findings were exploratory. Whether this offset warrants agent-specific reference values requires validation in clinical populations.

Graphical Abstract

KEY MESSAGES

1. Remimazolam produced higher BIS values than propofol at all three clinically matched sedation depths in healthy young volunteers.

2. Because depth was matched by clinical endpoint, these higher readings do not indicate lighter sedation as conventionally defined.

3. Whether this offset warrants agent-specific reference values requires validation in clinical populations before implementation.

Introduction

Accurate monitoring of sedation depth is fundamental to safe anesthetic and procedural care. Inadequate depth may permit intraoperative awareness and excessive sympathetic activation, whereas excessive depth may cause cardiovascular and respiratory depression and has been linked to postoperative cognitive dysfunction, especially in older patients [1–3]. This balance is most critical in ambulatory settings, where rapid recovery and timely discharge are essential [4]. The bispectral index (BIS), the most widely used monitor derived from the electroencephalogram (EEG), condenses this signal into a single score (0, isoelectric; 100, fully awake) and allows real-time titration [5,6]. BIS-guided sedation lowers drug use, speeds recovery, and reduces intraoperative awareness [7,8]. These benefits notwithstanding, BIS is not a direct measure of the anesthetic state. Its behavior is agent-dependent, is influenced by age and frontal electromyographic activity, and it does not track hypnotic depth for ketamine, nitrous oxide or dexmedetomidine [9]. Trials of index-guided anesthesia have also given inconsistent results for postoperative delirium [10]. Contemporary practice therefore favors interpreting the raw electroencephalogram alongside clinical signs rather than relying on a single value. In particular, the algorithm was calibrated largely on propofol and volatile anesthetics, and whether its thresholds hold for agents with different pharmacodynamic profiles remains uncertain [11].

Remimazolam is one such agent, an ultrashort-acting benzodiazepine with rapid onset and offset, metabolism by tissue esterases independent of organ function, limited cardiovascular and respiratory depression, and reversibility with flumazenil. These properties suit ambulatory procedures and patients with organ dysfunction or hemodynamic instability [12,13]. It acts at γ-aminobutyric acid type A receptors through a mechanism different from propofol and produces a distinct EEG signature, most notably persistent beta-frequency (13–30 Hz) activity at deep sedation [14]. Because the algorithm was not derived from this pattern, whether BIS quantifies sedation depth accurately during remimazolam administration remains unresolved.

Several studies have reported that remimazolam yields higher BIS values than propofol at comparable sedation levels, and that this gap persists even when patients are clinically unresponsive [15–17]. Several limitations weigh on this evidence. BIS was usually a secondary measure, and findings were inconsistent. Critically, parallel-group designs cannot control for the wide between-person variability in EEG responses to sedatives, which can exceed 10 BIS units at the same sedation level [18,19], making drug effects difficult to separate from individual differences. The one prior crossover trial was conducted in surgical patients under general anesthesia [17]. That trial neither characterized the discrepancy across graded sedation levels nor related the elevated BIS to any concurrent measure of function. That the two agents yield different index values is therefore established. What remains undefined is the magnitude of the offset within the same person across graded depths, and whether it has any functional counterpart. If propofol-derived thresholds underestimate the depth of remimazolam sedation, the clinical stakes are direct. Clinicians may escalate doses needlessly, risking avoidable cardiorespiratory depression, or misjudge adequate sedation as insufficient.

We hypothesized that, at sedation depths matched by clinical endpoints, remimazolam yields systematically higher BIS values than propofol, and we anticipated that any such offset would reflect the cortical EEG signature rather than a true difference in sedation depth. To test this, we conducted a prospective, randomized, double-blind crossover trial in healthy volunteers, a model free of the surgical stimulation, opioid co-administration, and comorbidity that confound patient studies, in which each participant served as his or her own control. The single confirmatory aim was to compare BIS between remimazolam and propofol at deep sedation. Exploratory aims were to compare BIS at moderate sedation and at loss of the eyelash reflex, emergence time, and early cognitive recovery, assessed with the Digit Symbol Substitution Test (DSST), Number Connection Test (NCT), and Auditory Verbal Learning Test–Huashan (AVLT-H). These analyses were hypothesis-generating and were not designed to establish agent-specific reference values.

Materials and methods

Study design

This prospective, randomized, double-blind, two-period, two-treatment crossover trial was conducted at Sanming First Hospital Affiliated to Fujian Medical University (Sanming, China) between 6 December 2025 and 10 May 2026. The trial used a superiority framework for the single confirmatory endpoint, all other analyses were exploratory. The hospital’s institutional review board approved the protocol (approval number 2023-66), and the trial was registered with the Chinese Clinical Trial Registry (ChiCTR2500112548; https://www.chictr.org.cn/showproj.html?proj=290328) on 17 November 2025, before the first participant was enrolled. All participants provided written informed consent before any study procedure. The study followed Good Clinical Practice guidelines and the Declaration of Helsinki as revised in 2013. Reporting follows the CONSORT 2025 statement together with the CONSORT extension for randomized crossover trials [20,21]. Completed checklists for both are provided in the Supplementary Material. No patients or members of the public were involved in the design, conduct, reporting or dissemination of this trial.

Participant selection

We enrolled healthy volunteers of either sex, aged 18–45 years, with American Society of Anesthesiologists (ASA) physical status I who were recruited from the local university student population. Exclusion criteria were known hypersensitivity to remimazolam, propofol, soy, or egg products; a history of chronic alcohol use or substance use disorder (including benzodiazepines or opioids); use of any central nervous system–active medication within 72 h before each session; a history of neurological or psychiatric disorders; pregnancy or lactation; and any condition judged by the investigator to compromise protocol compliance or data integrity.

Randomization and blinding

An independent statistician generated a computer-based randomization sequence, stratified by sex, allocating participants 1:1 to sequence AB (propofol first) or BA (remimazolam first). Allocations were concealed in sequentially numbered, opaque, sealed envelopes that were opened only after enrollment. A minimum washout of 7 days separated the two sessions. An unblinded pharmacist prepared both drugs in visually identical 50-mL syringes with opaque labels, administered through identical syringe pumps (BeneFusion eSP; Mindray, Shenzhen, China). Participants, the BIS-recording investigator, the MOAA/S assessor, and the neurocognitive tester remained blinded throughout both sessions. The administering anesthesiologist was unblinded, as required for dosing safety, but took no part in outcome assessment. The statistician analyzed anonymized datasets with treatments coded as ‘Agent X’ and ‘Agent Y’, and unblinding occurred only after database lock. As specified in the protocol, participants were asked at the end of the second session to identify which agent they had received in each period.

Study procedures

On each study day, participants fasted according to ASA preoperative guidelines (≥6 h for solids, ≥2 h for clear liquids) [22]. A forearm intravenous cannula was inserted, and standard monitoring was established: continuous five-lead electrocardiography, pulse oximetry, oscillometric noninvasive blood pressure at 3-min intervals, and BIS monitoring (BIS Vista; Medtronic, Minneapolis, MN, USA) with a four-electrode frontal sensor (BIS Quatro) applied according to the manufacturer’s instructions. Supplemental oxygen (3 L/min via nasal cannula) was given throughout.

Sedation was titrated to two predefined clinical endpoints using a standardized starting regimen and adjustment rule for each agent, matched by clinical endpoint rather than by target-controlled infusion. Moderate sedation was defined as a score of 3 on the MOAA/S, at which the participant responds only to loud or repeated calling of his or her name, with speech slurred and the eyes closed but airway reflexes and spontaneous ventilation preserved. Deep sedation was defined as a score of 1, at which the participant does not respond to loud calling or to mild prodding or shaking and responds only to a painful trapezius squeeze, a state at which airway support may become necessary [23]. The complete scale is given in Table S1. Propofol (10 mg/mL) was started with a 0.75 mg/kg loading dose over 5 min and an initial maintenance infusion of 42 µg/kg/min [24,25], then titrated in 10 µg/kg/min steps every 5 min to reach and maintain MOAA/S 3, then increased to 83 µg/kg/min and titrated by the same rule to MOAA/S 1. Remimazolam (1 mg/mL) was started with a 0.05 mg/kg loading dose over 1 min and an initial maintenance infusion of 0.35 mg/kg/h [15,16], then adjusted in 0.05 mg/kg/h steps every 5 min to reach and maintain each target. For both agents, the target score was confirmed by the blinded assessor and held for 5 min, with BIS averaged over the final 2 min of that period. before BIS acquisition. During up-titration, the infusion was briefly held at loss of the eyelash reflex to record a stable 2-min epoch.

Outcome measures

The confirmatory outcome was the mean BIS value during deep sedation, defined as a sustained MOAA/S score of 1 (response only to a painful trapezius squeeze). BIS was recorded continuously with a 15-second smoothing window, and the mean over the final 2 min of the 5-min steady-state period was used for analysis. Epochs were analyzed only when the signal quality index was ≥50% and electromyographic (EMG) activity was <50 dB [26]. All BIS recordings were made by a single blinded investigator.

Exploratory outcomes were the mean BIS value at loss of the eyelash reflex, averaged over a 2-min stable epoch while titration was briefly held, and the mean BIS value during moderate sedation, averaged over the final 2 min of the steady-state period. Emergence time, defined as the interval from infusion cessation to return of verbal responsiveness (MOAA/S = 5), was also exploratory.

Cognitive function was assessed at baseline (before each session) and at 30 and 120 min after return to full alertness (MOAA/S 5). The battery sampled the domains in which the two agents were most likely to diverge. The DSST indexes processing speed and executive function and is a conventional measure of residual sedative effect [27]. The NCT indexes psychomotor speed and cognitive flexibility [28]. The AVLT-H indexes verbal memory, the domain most closely tied to the amnestic action of both agents, and is widely used in Chinese-speaking populations [29,30]. All three instruments are brief and have parallel forms, administered in a counterbalanced (Latin-square) order across the two sessions and three time points, each form appearing equally often at each assessment and balanced across treatment sequences.

Safety monitoring

Safety was monitored continuously throughout sedation and recovery. Predefined adverse events were respiratory depression (apnea >20 s or respiratory rate <8 breaths/min), hypoxemia (peripheral oxygen saturation <92%), hypotension (mean arterial pressure <65 mmHg or a > 30% decrease from baseline), and bradycardia (heart rate <50 beats/min). All safety data were recorded in real time and reviewed by an independent blinded investigator.

Sample size calculation

The sample size was calculated for a 2 × 2 crossover design using PASS 2025 (NCSS, Kaysville, UT, USA). In an internal crossover pilot in patients undergoing gastrointestinal endoscopy (n = 10), the within-subject standard deviation of the paired BIS difference at deep sedation was 9.8 units. A paired difference of 10 BIS units was taken as the minimal clinically important difference, as it approximates the BIS separation between adjacent sedation levels (e.g. MOAA/S 3 vs 2) in foundational propofol volunteer studies, including crossover designs [31,32]. With a paired t-test, a two-sided α of 0.05, and 90% power, 14 participants (7 per sequence) were required. Allowing for approximately 10% dropout and rounding to balanced sequences, 16 participants (8 per sequence) were enrolled.

Statistical analysis

Analyses were performed in R 4.3.2 (R Foundation for Statistical Computing, Vienna, Austria), using lme4 and lmerTest (with pbkrtest) for mixed-effects models with the Kenward–Roger approximation and emmeans for marginal means and contrasts. The statistical analysis plan was finalized before unblinding. All tests were two-sided at α = 0.05, and estimates are reported with 95% confidence intervals (CIs). Continuous variables are summarized as mean ± standard deviation or median (interquartile range), categorical variables as frequency (percentage). All treatment comparisons were within-participant. The mean BIS value during deep sedation was the single confirmatory endpoint. All other analyses were exploratory. Their p values are nominal and were not adjusted for multiplicity.

The confirmatory outcome was analyzed with a linear mixed-effects model (REML) with treatment (remimazolam vs propofol), period, and sequence as fixed effects and participant as a random intercept. Inference used the Kenward–Roger approximation for standard errors and denominator degrees of freedom. The sequence term, equivalent in this 2 × 2 design to the treatment-by-period (carryover) contrast, was retained in the model but not used to select or condition the analysis. The treatment coefficient is the period- and sequence-adjusted mean BIS difference (remimazolam–propofol) with its 95% CI. Model assumptions were assessed by residual and quantile–quantile plots, supplemented by the Shapiro–Wilk test (Figure S1). The confirmatory effect was re-estimated by paired t-test on within-participant differences (unadjusted for period) as a sensitivity analysis.

Exploratory BIS outcomes (moderate sedation and loss of the eyelash reflex) used the same model. Emergence time (right-skewed) was log-transformed and analyzed in the same framework, with the effect back-transformed to a ratio of geometric means with 95% CI. Neurocognitive outcomes were analyzed with linear mixed models of the 30- and 120-min post-emergence scores, with treatment, time, period, and treatment-by-time as fixed effects and the session-specific baseline as a covariate. Participant and session-within-participant were included as random intercepts. The treatment-by-time interaction tested whether recovery trajectories differed, and treatment contrasts at each time were estimated as Kenward–Roger marginal means.

When a prespecified transformation did not resolve non-normal residuals, the within-participant difference was assessed by Wilcoxon signed-rank test without adjustment for period. Adverse events were summarized descriptively and compared by exact McNemar test, with no p value reported where no discordant pairs occurred. The blinding assessment was analyzed by exact binomial test against a null proportion of 0.5, with Clopper-Pearson confidence intervals. The 7-day washout exceeded the context-sensitive decrement times of both agents, so carryover was considered implausible by design. All 16 participants completed both periods without protocol deviations or missing data, so the intention-to-treat and per-protocol populations were identical and no imputation was required.

Results

Between 6 December 2025 and 10 May 2026, 16 healthy volunteers were enrolled and randomized, 8 to each sequence (Figure 1). All completed both sessions without protocol deviations or missing data. Participants had a mean age of 23.1 years (SD, 3.6), 7 of 16 (44%) were women, and all were ASA physical status I. All were university students and were therefore of comparable educational background. Baseline characteristics are summarized in Table 1. Nine of 16 participants (56.3%; 95% CI, 29.9 to 80.2) identified the agents correctly, a proportion consistent with chance (exact binomial p = 0.804).

Figure 1.

CONSORT flow diagram of the two-period, two-sequence crossover trial. All 16 participants were enrolled, randomized to sequence AB (propofol → remimazolam) or BA (remimazolam → propofol), and completed both periods with no losses or exclusions.

Table 1.

Participant characteristics by randomized sequence.

Characteristic	Sequence AB (propofol first) (n = 8)	Sequence BA (remimazolam first) (n = 8)	All participants (n = 16)
Age, years	23.0 ± 4.1	23.1 ± 3.2	23.1 ± 3.6
Female sex, n (%)	3 (38)	4 (50)	7 (44)
Height, cm	165.4 ± 7.4	171.9 ± 6.8	168.6 ± 7.7
Weight, kg	58.1 ± 11.1	68.5 ± 13.8	63.3 ± 13.2
Body mass index, kg/m2	21.1 ± 2.2	23.0 ± 2.9	22.0 ± 2.7

Note: Values are presented as mean ± SD or n (%).

Confirmatory outcome

At deep sedation, remimazolam produced higher BIS values than propofol. Mean BIS was 57.6 ± 7.5 for remimazolam versus 46.5 ± 7.9 for propofol, and the adjusted mean difference was 11.1 units (95% CI, 7.3 to 14.8; p < 0.001), exceeding the prespecified 10-unit minimal clinically important difference (Table 2; Figures 2 and S2). The sequence term, equivalent to the carryover contrast, was not significant (p = 0.472). The unadjusted paired t-test gave a similar estimate (11.1 units; 95% CI, 7.5 to 14.7; p < 0.001).

Table 2.

Session-level outcomes.

Outcome	Propofol (n = 16)	Remimazolam (n = 16)	Treatment effect (95% CI)	p Value
BIS at deep sedation	46.5 ± 7.9	57.6 ± 7.5	11.1 (7.3–14.8)	<0.001
BIS at moderate sedation	70.9 ± 7.8	81.7 ± 8.6	10.8 (6.3–15.2)	<0.001
BIS at loss of eyelash reflex	61.9 ± 5.3	71.6 ± 7.2	9.8 (6.4–13.1)	<0.001
Emergence time, min	7.9 ± 1.7	7.6 ± 1.8	0.97 (0.84–1.11)	0.606

Note: Values are mean ± SD. Treatment effects for bispectral index (BIS) outcomes are adjusted mean differences in index units from mixed-effects models, expressed as remimazolam minus propofol. Emergence time was analyzed after log transformation, and the treatment effect is the geometric mean ratio for remimazolam versus propofol. BIS at deep sedation was the single confirmatory outcome. All other outcomes were exploratory and p values are nominal, with no adjustment for multiplicity.

Figure 2.

Bispectral index (BIS) at three matched sedation depths. Boxes show the median and interquartile range (IQR); whiskers, the most extreme values within 1.5 × IQR; and the diamond, the group mean. Faint lines are individual participants (n = 16), and the bold dashed line is each agent’s mean trajectory. Remimazolam (orange) exceeded propofol (blue) at every depth. Between-agent differences are reported in Table 2.

Exploratory outcomes

In exploratory analyses, the BIS difference between agents was consistent across all graded sedation levels (Table 2). During moderate sedation, mean BIS was higher with remimazolam than with propofol (81.7 ± 8.6 vs 70.9 ± 7.8; adjusted difference 10.8 units; 95% CI, 6.3 to 15.2; p < 0.001). The difference persisted at loss of the eyelash reflex (71.6 ± 7.2 vs 61.9 ± 5.3; adjusted difference 9.8 units; 95% CI, 6.4 to 13.1; p < 0.001), so remimazolam produced higher BIS than propofol at every depth (Figure 2). No difference in emergence time was detected (remimazolam, 7.6 ± 1.8 min; propofol, 7.9 ± 1.7 min; geometric mean ratio 0.97; 95% CI, 0.84 to 1.11; p = 0.606; Table 2).

Cognitive recovery differed by domain. Baseline DSST, NCT, and AVLT-H scores were comparable between sessions (Table S2). Treatment-by-time interactions were significant for the DSST (p = 0.014) and NCT (p = 0.023) but not the AVLT-H (p = 0.847), indicating faster early recovery of processing speed and executive function with remimazolam (Table 3; Figures 3 and S3). At 30 min, DSST scores were higher (adjusted difference 3.7; 95% CI, 0.9 to 6.6; p = 0.012) and NCT completion times were shorter (−7.9 s; 95% CI, −11.5 to −4.2; p < 0.001) after remimazolam than after propofol, whereas verbal memory (AVLT-H) did not differ (0.9; 95% CI, −1.9 to 3.7; p = 0.511). By 120 min, no domain differed between agents (all p ≥ 0.320; Table 3).

Table 3.

Neurocognitive recovery.

Outcome	Treatment-by-time p value	REM − PRO at 30 min (95% CI)	p Value	REM − PRO at 120 min (95% CI)	p Value
DSST score	0.014	3.7 (0.9 to 6.6)	0.012	−1.4 (−4.3 to 1.4)	0.323
NCT completion time, s	0.023	−7.9 (−11.5 to −4.2)	<0.001	−1.8 (−5.5 to 1.9)	0.328
AVLT-H total score	0.847	0.9 (−1.9 to 3.7)	0.511	1.3 (−1.5 to 4.0)	0.353

Note: Values are adjusted mean differences with 95% confidence intervals from baseline-adjusted mixed-effects models. Higher DSST and AVLT-H scores indicate better performance. For the NCT, lower completion time indicates better performance, so negative treatment contrasts favor remimazolam. All neurocognitive analyses were exploratory and p values are nominal, with no adjustment for multiplicity. AVLT-H, Auditory Verbal Learning Test–Huashan; DSST, Digit Symbol Substitution Test; NCT, Number Connection Test; PRO, propofol; REM, remimazolam.

Figure 3.

Neurocognitive recovery after remimazolam and propofol sedation. (A) Digit Symbol Substitution Test (DSST) score, (B) Number Connection Test (NCT) completion time in seconds, and (C) Auditory Verbal Learning Test–Huashan (AVLT-H) total score at baseline and at 30 and 120 min after emergence. Symbols show the group mean, error bars the standard error, the bold line the mean trajectory, and faint lines individual participants (n = 16). For the NCT, a lower value indicates faster completion and better performance. At 30 min, DSST and NCT recovery were faster after remimazolam, whereas AVLT-H did not differ; all domains converged by 120 min (Table 3).

Safety data

No serious adverse events occurred. Predefined adverse events (respiratory depression, hypoxemia, hypotension, and bradycardia) were uncommon. The trial had very limited power to detect differences in these outcomes, and no statistically significant difference was detected between agents (exact McNemar test; Table 4). Respiratory depression occurred in 1 of 16 participants with remimazolam and 2 of 16 with propofol (p > 0.99), and hypoxemia in 2 of 16 with propofol only (p = 0.500). Bradycardia occurred in 1 of 16 participants with each agent (p > 0.99), and no participant developed hypotension. These comparisons are descriptive and should not be read as evidence of equivalent safety.

Table 4.

Safety outcomes.

Adverse event	Remimazolam (n = 16)	Propofol (n = 16)	Exact McNemar p Value
Respiratory depression	1 (6)	2 (13)	>0.99
Hypoxemia	0 (0)	2 (13)	0.500
Hypotension	0 (0)	0 (0)	–
Bradycardia	1 (6)	1 (6)	>0.99

Note: Values are the number of participants (%). Exact McNemar tests were based on discordant within-participant pairs. Where no discordant pairs occurred, the test is undefined and no p value is reported (–). The trial was not powered to compare adverse event frequencies between agents; these comparisons are descriptive and should not be read as evidence of equivalent safety. No serious adverse events occurred.

Discussion

In this double-blind, crossover trial in healthy volunteers, remimazolam produced higher BIS values than propofol at all three clinically matched sedation depths. Because sedation depth was matched by clinical endpoint in every participant, the higher index values cannot be attributed to lighter sedation as conventionally defined. In exploratory analyses, early recovery of processing speed and executive function was faster after remimazolam, whereas verbal memory recovered similarly after both agents.

Our findings are consistent with previous reports of higher BIS values with remimazolam than with propofol [15–17]. The crossover design in healthy volunteers isolated the drug effect from the between-person EEG variability that limited earlier parallel-group studies, and measurement across graded depths showed the offset to be present at every level rather than confined to one. Higher BIS with remimazolam has been attributed to greater 95% spectral edge frequency (SEF95) and electromyographic activity [33]. We add a concurrent behavioral measure that earlier EEG-only studies did not obtain.

A plausible explanation for the offset lies in the electroencephalographic profile of remimazolam, which in previous work shows persistent beta-band (13–30 Hz) activity at loss of consciousness and a greater electromyographic contribution than propofol [14,33,34]. Because the BIS algorithm was calibrated mainly on propofol and volatile anesthetics, which suppress high-frequency cortical activity, retained fast-frequency activity of this kind may be read as lighter sedation [9]. Spectral data were not recorded during sedation, so this account rests on that literature rather than on the present trial. The comparable recovery of verbal memory is consistent with the similar amnestic effects of both agents, mediated through hippocampal pathways that frontal BIS does not capture [29].

The divergence between the speed and memory domains was an exploratory finding in 16 participants, and its magnitude is imprecise. It identifies a hypothesis for testing in an adequately powered clinical population rather than an established difference in functional recovery. A further caution applies to the matching itself. Matching by MOAA/S and by the eyelash reflex establishes equivalence of behavioral responsiveness, not of cortical state. Responsiveness depends on brainstem and thalamocortical function as well as on cortical dynamics, and two agents may reach the same score by different routes. The offset is therefore open to two readings. The index may misrepresent cortical states that are in fact similar, or it may report faithfully on states that genuinely differ at equal behavioral depth.

These data caution against assuming that a single BIS target applies across agents. The estimated offset exceeded the prespecified minimal clinically important difference, although the confidence interval included values below it and individual participants varied around the average. Even so, an offset of this magnitude is enough to move a reading across the boundary of the conventional target range, which was derived from propofol. Targeting a range of 40 to 60 during remimazolam sedation could therefore prompt unnecessary dose escalation in a patient who is already unresponsive, with attendant risk of cardiorespiratory depression or delayed emergence [35].

This study has limitations. The cohort comprised healthy young university students, so the offsets may not apply to older patients, to those with neurological comorbidity, or to settings involving opioids and surgical stimulation. Sedation was titrated to clinical endpoints rather than by target-controlled infusion, precluding concentration–effect modeling. Safety was not an adequately powered comparison. Three further limitations concern measurement. First, acute tolerance has been reported during target-controlled remimazolam infusion in healthy volunteers [36]. Because sedation here was titrated to a clinical endpoint rather than delivered at a fixed concentration, tolerance would be expected to increase dose requirement rather than displace the index at matched depth, and the offset did not increase progressively across the three sequentially recorded depths. Tolerance was not measured directly, however, and these findings should not be extrapolated to prolonged infusion. Second, we assessed blinding in participants only. The result was consistent with successful masking, although the assessment was imprecise at this sample size. Investigators were not asked to guess allocation, the injection discomfort associated with propofol may have permitted inference in individual sessions, and the clinical sedation ratings used for matching are assessor-dependent, so they cannot be regarded as fully protected from expectancy. Finally, these findings concern one processed index and should not be read as support for index-guided titration in isolation. Where agents with dissimilar electroencephalographic signatures are used, the raw waveform and spectrogram, interpreted with clinical signs, remain the more complete description of cortical state.

The main strengths are the double-blind crossover design and the multidimensional cognitive testing, which provided a concurrent behavioral measure alongside the index. Future work should establish and validate agent-specific BIS reference values in clinical populations using target-controlled infusion, and pair quantitative EEG spectral analysis with cognitive testing to clarify the basis of the dissociation between cognitive recovery and electrophysiological readings.

Conclusion

Remimazolam produced higher BIS values than propofol at clinically matched sedation depths in healthy young volunteers, consistently across all three depths. Whether this offset warrants agent-specific reference values requires validation in clinical populations before implementation. Opioids, surgical stimulation, age and comorbidity may alter both the index and its relationship to clinical sedation depth.

Funding Statement

This research was supported by the Fujian Strait Medical and Health Exchange Association Precision Anesthesia Research Project [No. 2024-HYHMZ-05], the Natural Science Foundation of Fujian Province [Nos. 2024J011489 and 2024J08258], and the Sanming City Joint-Funded Science and Technology Program Projects [No. 2023-S-104]. The funders had no role in study design, data collection, analysis, interpretation, manuscript preparation, or the decision to publish.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The study protocol and the statistical analysis plan are available from the corresponding author on reasonable request. De-identified individual participant data underlying the reported results are available on the same basis from researchers whose proposed use has been approved by an independent review committee, in accordance with research ethics guidelines and the confidentiality safeguards in the approved protocol.