Remimazolam, a novel ultra-short-acting benzodiazepine invented out of ‘soft drug’ development, has recently (January 2021) been approved as a general anesthetic by the Ministry of Food and Drug Safety (MFDS) in South Korea. It has advantages over currently used hypnotic agents including rapid onset/offset, predictable duration of action, organ-independent metabolism, availability of a reversal drug, and maintenance of stable hemodynamics, and is increasingly being used as a hypnotic component of total intravenous anesthesia (TIVA) [1–5]. Pharmacokinetics and pharmacodynamics of remimazolam were characterized by relatively high clearance, small steady-state volume of distribution, short elimination half-life, short context-sensitive half-life, linear kinetics, and dose-related depth and duration of sedation [6–8]. Due to these pharmacologic properties, remimazolam concentration accurately represents its hypnotic effect; thus, knowing accurate remimazolam concentrations is beneficial in anesthesia practice. As most drugs act at sites remote from the bloodstream, hysteresis, a phenomenon denoting time lag between the measured plasma drug concentration and the observed drug effect may occur [9]. Therefore, the pharmacologic effect can be determined based on either the plasma concentration (Cp) or effect-site concentration (Ce); however, Ce rather than the Cp reflects the clinical effects of intravenous anesthetics better [10,11].

Target-concentration controlled infusion (TCI) systems, which allow drugs to be administered to achieve and maintain a desired target concentration [12], would be useful in administering remimazolam during anesthesia for optimal dosing adjustment. Currently, no commercial TCI device for remimazolam is available; thus, it has no choice but to be administered as a variable rate continuous infusion at the infusion rate approved by the MFDS (i.e., 6 or 12 mg/kg/h for induction, followed by 1 mg/kg/h [2 mg/kg/h of maximal infusion rate] for maintenance).

The Schüttler model, a three-compartment model that can be applied for administering remimazolam via effect-site target-controlled infusion method including body weight as a covariate, significantly affecting the volume of distribution in the central compartment [8]. As a result, the Ce of remimazolam varies from one person to another depending on body weight, even if administered with the same body weight-based dose. Consequently, considering optimal dosage titration, it would be helpful to simulate the Ce while administering remimazolam with the conventional weight-based dosing strategy during the induction and maintenance of anesthesia.

The objective of this retrospective study, therefore, was to investigate the simulated Ce of remimazolam at the loss of responsiveness (LOR) to a verbal command during anesthesia induction and at the recovery of responsiveness (ROR) during emergence using the data obtained in the process of clinical practice of patients who underwent general anesthesia with remimazolam. In addition, the relationship between the simulated Ce of remimazolam at LOR and ROR and the patient’s factors, such as age and sex, were evaluated.

This retrospective study was approved by the Institutional Review Board (approval number: 2021-1747). The medical records of patients who underwent elective surgery under general anesthesia with remimazolam at our institution between August 4 and October 12, 2021, were retrospectively reviewed. Inclusion criteria included patients administered with remimazolam as a hypnotic component of TIVA throughout induction and maintenance of anesthesia and patients who administered remimazolam while conducting concentration simulations and having the complete result files of simulation.

General anesthesia was identically conducted according to our institution’s standard anesthesia protocol for general surgery in all patients. The patients in this study fasted from midnight on the day of surgery and received no premedication. Upon arrival in the operating room, standard monitoring processes, including an electrocardiograph, pulse oximeter, noninvasive blood pressure monitor, train-of-four, end-tidal carbon dioxide partial pressure, and the bispectral index (BIS^TM, Covidien, USA) were applied in all cases. All data were recorded continuously until the end of anesthesia. After pre-oxygenation with 100% oxygen via facial mask, the anesthetic dosing strategies were employed in all patients as follows. Remimazolam (Byfavo™ Inj, Hana Pharm Co., Ltd, Korea) was administered as an initial induction dose of 6 or 12 mg/kg/h at the discretion of the attending anesthesiologist using a syringe pump (Perfusor^®, Space, B. Braun Melsungen AG, Germany). After starting remimazolam infusion, LOR, defined as the patient becoming unresponsive to loud verbal commands (open your eyes), was assessed frequently to determine the induction of anesthesia. After confirming LOR, the infusion rate of remimazolam was adjusted to 1 or 2 mg/kg/h and remifentanil was administered via target Ce controlled infusion using the Minto model [13]. Endotracheal intubation was facilitated by administration of rocuronium at 0.6 mg/kg. Patients were then ventilated with a mixture of air and oxygen (fraction of inspired oxygen, 0.5); mechanical ventilation was adjusted to maintain an end-tidal carbon dioxide partial pressure of 35–45 mmHg. Consequently, the remimazolam infusion rate was adjusted within 0.3–2.0 mg/kg/h to maintain BIS values at less than 60 during the maintenance of anesthesia. Target Ce values of remifentanil were titrated within a range of 3–20 ng/ml to maintain stable hemodynamics (blood pressure and heart rate within 20% above or below baseline values).

In our institutional anesthesia practice for general surgery, whenever remimazolam is administered during general anesthesia in general surgery patients, simulations of the manual infusion mode of remimazolam is simultaneously performed using the Asan Pump software (version 2.1.3, Bionet, Korea; http://www.fit4nm.org/download/, last accessed: August 27, 2012) to evaluate Ce of remimazolam. Accordingly, in all patients of the present study, simulations of the manual infusion mode of remimazolam were simultaneously performed using the Asan Pump software. Concurrently, remimazolam Ce was automatically simulated using pharmacokinetic parameters and blood-brain equilibration rate constant (k_e0) proposed by Schüttler et al. [8]. The pharmacokinetic parameters and k_e0 used for simulating Ce of remimazolam are presented in Table 1. Whenever the infusion rate of remimazolam of the syringe pump was manually changed, similar changes in values were made in the Asan Pump program simultaneously, and then the simulated Ce was confirmed. The Asan Pump infusion profiles including the infusion rate of remimazolam for all patients were automatically recorded in a .csv file format. In addition, during the simulation, the user can record clinical events (e.g., endotracheal intubation, loss of consciousness, extubation) in the pop-up window of the Asan pump program. Thus, in our routine anesthesia practice using remimazolam with concentration simulations, the anesthesiologists recorded LOR in event recording window of the Asan pump program; the time points when the patients reached LOR were automatically recorded in simulation result files of the Asan pump program.

On the closure of the surgical wound, opioids (e.g., fentanyl 50 μg or oxycodone 4 mg) and ramosetron 0.3 mg were administered for postoperative pain, nausea and vomiting control. At the end of the procedure, remimazolam and remifentanil were discontinued. The neuromuscular blockade was antagonized using sugammadex or a combination of neostigmine and glycopyrrolate according to the attending anesthesiologist’s discretion. During emergence, recovery of anesthesia was assessed by ROR, defined as a patient’s ability to open their eyes in response to verbal commands. The anesthesiologists recorded ROR in the event recording window of the Asan pump program; the time points when the patients reached ROR were also automatically recorded in simulation result files. When patients met the extubation criteria (BIS values ≥ 80 and train-of-four ratio > 0.9 for more than 1 min), they were extubated and transferred to a post-anesthesia care unit (PACU). Flumazenil, a specific antagonist of benzodiazepines, was administered based on the discretion of the attending anesthesiologists when the patients became drowsy again in the operating room or PACU after recovering consciousness.

A total of the 81 eligible patients (40 patients in the remimazolam induction dose 6 mg/kg/h group and 41 patients in the remimazolam induction dose 12 mg/kg/h group) were included in data analysis. The demographic and clinical characteristics of the patients are shown in Table 2. During induction, the median (1Q, 3Q) simulated Ce of remimazolam at LOR was 0.7 (0.5, 0.9) μg/ml. All patients lost response to verbal command in a median of 1.9 min after the start of remimazolam infusion with the cumulative dose of 0.3 mg/kg. In the subgroup analysis (between induction dose 6 mg/kg/h and 12 mg/kg/h groups), there were significant differences in the simulated Ce of remimazolam at LOR, the time to reach LOR, and the cumulative dose of remimazolam to induce LOR. The median (1Q, 3Q) simulated Ce of remimazolam at LOR was higher in 12 mg/kg/h group than 6 mg/kg/h group (0.8 [0.7, 0.9] vs. 0.5 [0.4, 0.7] μg/ml, P < 0.001) (Fig. 1). The patients who received remimazolam 12 mg/kg/h reached LOR faster compared with those who received remimazolam 6 mg/kg/h (1.9 [1.7, 1.9] vs. 2.2 [1.9, 2.5] min, P < 0.001). The median (1Q, 3Q) cumulative dose of remimazolam required to induce LOR was larger in 12 mg/kg/h group than 6 mg/kg/h group (0.36 [0.33, 0.40] vs. 0.22 [0.18, 0.25] mg/kg, P < 0.001). There was no significant relationship between age and simulated Ce at LOR (P = 0.199 for 6 mg/kg/h group and P = 0.0674 for 12 mg/kg/h group, respectively) (Fig. 2). The simulated Ce of remimazolam at LOR was not significantly different according to the patient’s sex (P = 0.299 for 6 mg/kg/h group and P = 0.342 for 12 mg/kg/h group).

During the emergence, when the median (1Q, 3Q) simulated Ce of remimazolam was 0.3 (0.2, 0.4), patients recovered response to verbal command (Fig. 3). Age was significantly correlated with the simulated Ce at ROR with a negative slope (Ce at ROR = –0.0043 × age + 0.57, r = 0.30, P = 0.014) (Fig. 2). The simulated Ce of remimazolam at ROR was not significantly different according to the patient’s sex (P = 0.818).

When the relation between the simulated Ce of remimazolam at LOR or ROR and the patient’s factor was assessed, the simulated Ce at ROR correlated only with the patient’s age. Therefore, we only evaluated the pharmacodynamic relationship between the remimazolam Ce at ROR and the probability of ROR using a population pharmacodynamic modeling incorporating age as a covariate. After exploring the base model, we included age as a covariate in Ce_50,ROR (effect-stie concentration of remimazolam, associated with 50% probability for recovery of responsiveness) to investigate whether age has a significant effect on Ce value at ROR. This pharmacodynamic modeling including age significantly improved the performance of the basic model based on the likelihood ratio test, with a decrease in the OFV (total 6.8, P < 0.01). Table 3 summarizes the estimated pharmacokinetic parameters of the basic and final model and the non-parametric bootstrap results. The medians of the pharmacokinetic parameters estimated by the bootstrap data set were generally similar to the final pharmacokinetic parameter estimate with the 95% CIs being narrow overall, indicating that the final parameters were adequately estimated. The pharmacodynamic relationship between the probability of ROR and remimazolam Ce according to age evaluated using the final population parameters is shown in Fig. 4. The age values used for the predictions correspond to the hypothetical patients aged 30-, 50-, and 80-year as distributed within the studied population. The Ce_50,ROR and Ce_95,ROR (effect-site concentration of remimazolam, associated with 95% probability for recovery of responsiveness) in 30-, 50-, and 80-year-old patients was predicted to be 0.52 and 0.48, 0.45 and 0.41, and 0.34 and 0.31 μg/ml, respectively.

This study is a clinical investigation quantifying the Ce of remimazolam to induce LOR to verbal commands during anesthesia induction and ROR during emergence, using the data obtained in the process of clinical practice of patients who underwent general anesthesia with remimazolam as a hypnotic component. The simulated Ce at LOR and ROR were 0.7 and 0.3 μg/ml, respectively. Moreover, we demonstrated that age was significantly related to the Ce of remimazolam at ROR. These findings suggest that Ce simulations would be beneficial while administering remimazolam during anesthesia for optimal dosage adjustment, and possibly providing clinically important information for the precise dose adjustment of remimazolam to achieve rapid and safe induction and early recovery. In particular, elderly patients may recover from anesthesia at lower Ce of remimazolam; thus, timely tapering of remimazolam considering with Ce of remimazolam upon surgery completion would facilitate fast emergence from anesthesia.

This is the first research that determined the Ce of remimazolam to induce LOR and ROR to verbal commands during TIVA with remimazolam and remifentanil. Variability in the onset and offset profile of an anesthetic drug is likely due to the factors influencing the pharmacokinetics and the relationship between drug concentrations and their effect [16]. Ce rather than Cp better reflects the clinical effects of intravenous hypnotics [10,11]; however, administering remimazolam via effect-site target-controlled infusion method is still unavailable. In terms of optimal dosage adjustment, therefore, it would be clinically advantageous if the individual remimazolam Ce could be simulated and applied in clinical practice rather than simply traditional observations of autonomic response along with monitoring of depth of anesthesia. The feasibility of predicting the potential remimazolam effects by estimating the Ce of anesthetics using pharmacokinetic/pharmacodynamic simulation has also been reported [17].

A significant relationship between the simulated remimazolam Ce at ROR and age was observed in this study. We conducted population pharmacodynamic modeling, incorporating age as a covariate in Ce₅₀ at ROR, and demonstrated that age plays an important role in determining inter-individual variability of remimazolam Ce₅₀ at ROR. Based on our prediction of the probability of ROR according to age, elderly patients may emerge at lower remimazolam Ce than younger patients. Therefore, age has a considerable influence on the patients’ sensitivity to remimazolam from a pharmacodynamic point of view. These results are consistent with the findings of the previous studies in which increased sensitivity to the sedative effects of other benzodiazepines was observed in elderly patients [18,19]. The currently reported case series suggested that adequate titration of infusion rate of remimazolam based on the patient's sedation level and general condition during maintenance of anesthesia achieved well-tolerated anesthesia with prompt emergence from anesthesia in two super-elderly patients [5]. Consequently, the ability to predict the individual remimazolam Ce for ROR by simulation especially in elderly patients would be clinically meaningful in terms of allowing the dose of remimazolam to be adjusted to achieve fast recovery from anesthesia.

During anesthesia induction in the present study, the patients who received 12 mg/kg/h of remimazolam were unresponsive to verbal commands approximately 18 s faster than those who received 6 mg/kg/h of remimazolam. These findings are consistent with the results of the previous study, which revealed that remimazolam's induction dose had a statistically significant effect on the time to lose responsiveness to the shaking of patient’s shoulder (a 13-s rapid onset with 12 mg/kg/h groups when compared with 6 mg/kg/h groups) [2]. This difference in time to LOR in our study may be explained by more cumulative doses to induce LOR and higher simulated Ce at LOR in 12 mg/kg/h group compared to that of 6 mg/kg/h group (approximately 0.14 mg/kg more for cumulative doses and 0.3 μg/ml higher for Ce, respectively). Thus, though both induction doses are suitable, 12 mg/kg/h of remimazolam is recommended in certain cases where rapid induction is needed.

In the current study, patients were unresponsive to verbal commands in median 132 and 114 s with cumulative doses of remimazolam 0.22 and 0.36 mg/kg after starting remimazolam at 6 and 12 mg/kg/h, respectively. However, a prior study reported that faster time and less cumulative doses until patients lost responsiveness to even stronger stimuli such as the shaking of their shoulder than those of the present study with equal infusion rates of remimazolam (102 s and 88.7 s for time to LOR, 0.17 and 0.29 mg/kg for cumulative doses to LOR, respectively) [2]. These different findings might result from the presence or absence of concomitantly used opioids. In a previous study, 0.25–0.5 μg/kg/min of remifentanil was infused just before remimazolam infusion to completion of tracheal intubation, whereas remifentanil infusion was commenced after confirming LOR in our study. The synergic relationship between opioids and benzodiazepine has been well-established [19]; it corresponds with the results of a recent study conducted to assess pharmacodynamic drug-drug interaction potential of remimazolam with remifentanil, which showed a high degree of synergism between two anesthetic agents [20]. In addition, the administration of remifentanil 10 min before remimazolam revealed a 20 s faster time to lose consciousness compared with when remifentanil was started 2 min before remimazolam [21].

Several limitations of this study should be considered. First, we have assumed the simulated Ce values of remimazolam derived from the Schüttler model to be applicable to the enrolled study population. Although the Schüttler model has drawbacks that these models have been constructed only in 20 healthy male volunteers without external validation, these can be considered reliable and unique models as they include pharmacokinetic parameters and the k_e0 of remimazolam obtained in a single population of adults. However, the true Ce values at LOR and ROR can be more variable than presented in this study. There has been a report that the k_e0 of remimazolam is different according to race [20]. In this study, remimazolam Ce was simulated using the k_e0 derived from the Schüttler model which has been developed in the European population. In general, if the pharmacokinetic parameters are the same, but the k_e0 is different, even though the same dose is administered, the time to peak effect is longer (slow onset) when the k_e0 is small, and the Ce at that time is also low [22,23]. Accordingly, based on the facts reported in the previous study that the k_e0 of remimazolam in the Asian race is approximately 50% low than that in the non-Asian race [20], there is possible that the Ce calculated in Koreans is lower. However, solid conclusions could be drawn by population pharmacokinetic/pharmacodynamic analysis in the Korean population. Thus, we are planning to develop a pharmacokinetic/pharmacodynamic model of remimazolam in the Korean population. Second, in this study, we investigated the simulated Ce anesthesia at the time when the patients were unresponsive to verbal commands during anesthesia induction; however, this referred to Ce only on unresponsiveness to stimuli rather than the absence of consciousness [24]. Therefore, the Ce of remimazolam to induce loss of consciousness can be different from the simulated Ce value at LOR presented in our study. Lastly, owing to the retrospective nature of the present study, LOR and ROR were not assessed in a standardized time period. Also, we did not evaluate various factors influencing LOR and ROR such as obesity, ethnicity, and co-administered drugs. Therefore, further well-designed prospective trials are warranted to quantify the exact Ce of remimazolam to induce LOR and ROR and establish thoroughly the clinical variables related to LOR and ROR during anesthesia with remimazolam.

In conclusion, the present study reports the simulated Ce of remimazolam for attaining loss and recovery of responsiveness to verbal commands in patients who underwent general anesthesia with remimazolam. Our study results suggest that it would be helpful to simulate the Ce while administering remimazolam with the conventional dosing strategy during anesthesia for optimal dosage adjustment. In particular, as there is a possibility that elderly patients may recover from anesthesia at lower Ce of remimazolam, it would be essential that remimazolam infusion should be timely ceased at the end of the surgery based on the simulated Ce of remimazolam for early recovery.