Optimal Dosing Strategies for Remimazolam Across Diverse Patient Populations: A Clinical Narrative Review

Abstract

Remimazolam is an ultra-short-acting benzodiazepine that is hydrolyzed by carboxylesterase 1 (CES1) to the inactive metabolite CNS 7054. Since regulatory approval in 2020 it has been adopted globally for procedural sedation and general anesthesia, supported by an expanding evidence base of randomized trials, population pharmacokinetic/pharmacodynamic (PK/PD) modeling studies, dose-finding sequential-allocation analyses, and meta-analyses. Synthesizing 92 studies published between January 2020 and April 2026, this narrative review defines dose optimization across diverse populations. For adult procedural sedation, a fixed 5 mg initial bolus followed by 2.5 mg maintenance titrations achieves moderate sedation with a hemodynamic and respiratory profile that compares favorably with propofol, whereas general anesthesia requires an induction rate of 6 to 12 mg kg⁻¹ h⁻¹ followed by maintenance at 1.0–2.0 mg kg⁻¹ h⁻¹. Dosing demands a 25–30% reduction in the elderly (≥ 65 years), higher weight-adjusted requirements in pediatric patients (1–18 years), and lean or corrected body weight scalars in obesity, while severe hepatic impairment reduces clearance by approximately 38% and opioid co-administration yields a 15–63% synergistic reduction in effective concentrations. Several quantitative estimates, however, derive from single-center trials with fewer than 30 participants per arm that have not been externally validated. To help clinicians weigh these recommendations, each population-level dosing estimate is graded by the strength and external validity of its supporting evidence (A, B, or C); most special-population point estimates rest on single-center studies (grade C) and are best regarded as starting points rather than established standards. Crucially, the halted Japanese Phase II intensive care unit trial (ONO-2745-04) serves as a key safety counter-narrative: continuous infusions exceeding 24 hours in critically ill patients may be associated with drug accumulation and delayed emergence, suggesting that organ-independent clearance may not be guaranteed when CES1 activity is compromised. Major evidence gaps persist for infants under one year of age, long-duration anesthesia, CES1 polymorphisms, and non-East-Asian populations.

Plain Language Summary

Remimazolam is a fast-acting sedative that anesthesiologists now use worldwide for short procedures and for general anesthesia. The body breaks it down rapidly with an enzyme called carboxylesterase 1, which is found mainly in liver cells but works independently of the slower liver pathways and the kidneys that clear most other sedatives, so doctors expected the dose to be easy to predict. In practice, the recommended dose still varies widely between different types of patients, and it has been hard for clinicians to know which numbers to trust.

We searched the published research up to April 2026 and brought together 92 studies. We compared the doses used in adults, older people, children, people carrying extra weight, and people whose liver or kidneys do not work well. We also graded how strong the evidence behind each recommendation is.

We found that older patients need roughly a quarter to a third less drug, children need more per kilogram of body weight, and much less drug is needed when a strong painkiller is given at the same time. We also found that the drug can build up and delay waking when it is given for more than a day to patients who are seriously ill, so its usual advantage over other sedatives may not apply to them. Many of the published dose figures come from single hospitals and small groups of patients, so clinicians should treat them as sensible starting points and adjust the dose for each patient rather than follow them as fixed rules.

Introduction

Remimazolam: An Ultra-Short-Acting Benzodiazepine

Remimazolam is an ultra-short-acting drug that represents a pharmacokinetic refinement, rather than a new mechanism, within the benzodiazepine class. Like its predecessors, it acts as a positive allosteric modulator of the γ-aminobutyric acid type A (GABA_A) receptor, but its molecular design as a soft drug confers rapid hydrolysis by non-specific tissue esterases — predominantly CES1 — to the pharmacologically inactive metabolite CNS 7054. CES1, the principal human carboxylesterase, is expressed most abundantly in hepatocytes and, to a lesser degree, in monocytes/macrophages and other tissues; its catalytic activity can be modified by functional genetic polymorphisms (notably the loss-of-function c.428G>A [p.G143E] variant), by disease states that accompany critical illness (systemic inflammation, hepatic hypoperfusion and hypoalbuminemia) and by drug–drug interactions with other CES1 substrates or inhibitors — influences examined in detail in the sections on ICU sedation and on CES1 pharmacogenomic uncertainty below. This metabolic pathway is intended to reduce dependence on hepatic cytochrome P450 enzymes and on renal elimination, with the consequence that single-bolus pharmacokinetics are largely preserved in mild-to-moderate organ impairment and the context-sensitive half-time remains short after prolonged infusion in healthy volunteers.1,2

Since first regulatory approval in Japan in 2020, remimazolam has been licensed in the United States (2020), the European Union (2021), China (2021), South Korea (2022) and Taiwan (2022), and uptake for procedural sedation, monitored anesthesia care (MAC) and general anesthesia induction and maintenance has been rapid.3–5 A growing body of evidence from randomized controlled trials, population PK/PD modeling studies and meta-analyses has begun to define dose–response relationships across diverse clinical settings and populations. Recent systematic reviews, meta-analyses, and multicenter randomized controlled trials suggest non-inferior or improved efficacy compared with propofol and midazolam for procedural sedation and anesthesia induction, together with a lower incidence of clinically significant hypotension and respiratory depression.6–15 However, dosing recommendations vary substantially across populations, and several published estimates of ED₅₀, ED₉₀ and ED₉₅ (the doses producing the target clinical effect in 50%, 90% and 95% of patients, respectively) derive from single-center dose-finding studies with small sample sizes that have not been externally validated.1,16–18 Equally important, the discontinuation of the Japanese Phase II program for long-term ICU sedation (ONO-2745-04, halted in 2013, predating widespread regulatory approval; the background and rationale for its discontinuation were published only recently, which is why this safety signal is being widely re-examined only now) highlights that the drug’s organ-independent metabolism may not be guaranteed in critically ill patients with multi-organ dysfunction — a caveat that should temper extrapolation of the favorable short-procedure evidence to long-duration or critical-care contexts.19–21

That caveat sharpens, rather than diminishes, the case for precise dosing. Each clinically used alternative carries limitations that remimazolam is intended to mitigate: propofol provides rapid, titratable hypnosis but causes dose-dependent hypotension, respiratory depression and injection pain and has no antagonist; midazolam is reversible with flumazenil but undergoes hepatic CYP3A metabolism that yields variable and sometimes prolonged recovery; and dexmedetomidine offers opioid-sparing sedation with preserved respiration but is limited by bradycardia, hypotension and comparatively slow arousal. Both directions of dosing error carry clinical cost: underdosing risks inadequate sedation, patient distress, intraoperative awareness or recall and interrupted or failed procedures, whereas overdosing risks hypotension, respiratory depression or apnea, delayed emergence and prolonged recovery-room stay — consequences amplified in the elderly and the critically ill.

This review is distinguished from the existing literature in three respects. First, unlike previous remimazolam narrative and systematic reviews — including efficacy/safety meta-analyses22 and indication- or population-specific reviews — it consolidates population-specific dose estimates into a single, critically appraised framework in which every recommendation carries an explicit A/B/C evidence grade (Table 1 and Table 2). Second, it benchmarks the principal published population PK/PD models against one another (Table 3), reporting their external-validation status and whether they are implemented on commercial target-controlled infusion platforms. Third, it integrates the recently published account of the discontinued Japanese ICU program and emerging CES1 pharmacogenomic data into practical, bedside dosing guidance. The present review shares a deliberate methodological scaffold — structured search, population tables and evidence grading — with our group’s recent dosing-focused review of dexmedetomidine in this journal;23 the two address pharmacologically and metabolically distinct agents (a CES1-hydrolyzed “soft” benzodiazepine versus an α2-adrenoceptor agonist) with different population-specific problems, and the shared structure is intended to make the evidence reproducibly comparable rather than to duplicate content.

Table 1.

Recommended Remimazolam Dosing Across Patient Populations and Indications

Population	Indication	Loading/Induction	Maintenance/Titration	Refs
Adults (general)	Procedural sedation, MAC	5 mg i.v. bolus over 1 min (fixed-dose regimen)	2.5 mg top-ups titrated to MOAA/S=3	[3, 24]
Adults (general)	General anesthesia	Age-dependent, 0.14–0.33 mg kg−1 across age bands – not a single recommended range; see Table 2; or 6–12 mg kg−1 h−1 infusion until LOC	Nil or 1.0–2.0 mg kg−1 h−1	[6, 14, 24]
Elderly (≥ 65 yr)	Sedation (MAC/CSEA)	Reduce 25–30% from adult dose	0.21–0.29 mg kg−1 h−1 titrated to MOAA/S — sedation only; for GA maintenance in elderly patients use the adult rate in the row above, titrated to depth-of-anesthesia monitoring	[12, 25, 26]
Pediatric (1–18 yr)	GA induction, sedation	Higher per-kg dose than adults; ED95 0.43–0.57 mg kg−1 for LOC (derived in children aged 1–12 yr; see Table 2 for the divergent 3–15 yr estimate)	Titrate to clinical endpoint. The EC50 for a UMSS score of 3 or greater falls from 777 ng mL−1 without opioid to 287 ng mL−1 at fentanyl 4 ng mL−1; these are EC50 values, not target concentrations — the regimen tested left 9.2–22.0% of children below UMSS ≥ 3, and model-based optimization required higher per-kilogram doses	[16, 27–29]
Hepatic impairment	Sedation, endoscopy, GA	Single-bolus 0.1 mg kg−1 preserved (AUC↑, Cmax unchanged)	Clearance↓38%; allow prolonged recovery (~ 17 min)	[24, 30, 31]
Obese	Bariatric, GA	Use lean (LBW) or corrected (CBW) body weight; ED95 ~ 0.179 mg kg−1 with sufentanil	Titrate to clinical endpoint; optimal scalar unresolved	[32–34]
ESRD (not on dialysis)	Dialysis access, sedation	Standard adult dose	Standard adult dose; no clearance data in patients receiving renal replacement therapy	[30, 35]
Cardiac interventional procedures (pediatric and young-adult)	Sedation for diagnostic and interventional catheterization	0.03–0.11 mg kg−1 bolus	0.3–1.2 mg kg−1 h−1 infusion, titrated to effect; no clinically significant hemodynamic depression reported	[36]
Sleep disorder/OSA	GA induction, DISE	ED95 ~ 0.237 mg kg−1 (vs 0.209 in non-SD); ~ 13–18% higher than baseline	Maintenance unstudied; pre-op PSQI advised	[18, 37]
With opioids	All indications	Reduce remimazolam 15–63% (fentanyl-dose dependent)	PK/PD-modelled synergy (pediatric); dose-sparing confirmed empirically in elderly	[17, 27, 38]
Infants < 1 yr	— (no prospective data)	Insufficient evidence	Do not extrapolate from older pediatric data	[11, 24, 27]
Long-term ICU sedation (> 24 h)	Not recommended for routine use	Limit to short courses (< 24 h) where used	Parent-drug accumulation reported in critically ill (ONO-2745-04); see Unresolved Controversies	[19, 20, 39, 40]

Notes: Rows in bold text indicate populations for whom current evidence is insufficient and routine use is not recommended. Doses are expressed per actual body weight unless a specific weight descriptor is stated. Arrows denote the direction of change relative to healthy volunteers: ↑, increased; ↓, decreased.

Abbreviations: AUC, area under the concentration–time curve; CBW, corrected body weight; Cmax, maximum plasma concentration; DISE, drug-induced sleep endoscopy; ED₉₅, 95% effective dose; ESRD, end-stage renal disease; GA, general anesthesia; ICU, intensive care unit; LBW, lean body weight; LOC, loss of consciousness; MAC, monitored anesthesia care; MOAA/S, Modified Observer’s Assessment of Alertness/Sedation; ONO-2745-04, Japanese Phase II ICU sedation trial; OSA, obstructive sleep apnea; PK/PD, pharmacokinetic/pharmacodynamic; PSQI, Pittsburgh Sleep Quality Index; SD, sleep disorder.

Table 2.

Pharmacokinetic and Pharmacodynamic Evidence and Safety Profile by Population, with Evidence-Strength Tags

Population	Key PK/PD Data	Safety Profile	Evidence Tag	Refs
Adults (procedural sedation)	● Fixed 5 mg + 2.5 mg top-up regimen effective and safe versus placebo and midazolam, including in high-risk ASA III/IV colonoscopy (efficacy 87.1% vs 0% placebo)41	● Lower hypotension (RR 0.5742/OR 0.2643) and respiratory depression (RR 0.4642/OR 0.2243) vs comparators ● Hypoxia risk rises with concomitant sedatives and higher BMI47	A (fixed-dose efficacy and safety: Phase III RCT + meta-analyses)/B–C (real-world cohorts)	[41–47]
	● Adult gastroscopy: 0.2 mg kg−1 non-inferior to propofol, with less hypotension46
	● Real-world cohorts: median ~ 7.5 mg in the emergency department;45 median time to alert ~ 12 min47
	● Broad-population popPK/PD model targets MOAA/S 2–344
Elderly (≥ 65 yr)	Age-dependent requirement: ~25–30% lower than in younger adults; cited elderly cohorts spanned 60–79 yr, 60–80 yr, and ≥ 65 yr.12 Induction bolus: ED95 0.249 mg kg−1 for LOC (Oh)25 to 0.297 mg kg−1 for inhibition of the intubation response (Qu).48 Sedation during CSEA for hip replacement: ED50 0.212 mg kg−1 h−1, ED95 0.288 mg kg−1 h−1 (Song Y and Zou).26 Painless gastroscopy with fentanyl co-administration: ED50 0.258 mg kg−1 (fentanyl 0.5 μg kg−1) → 0.095 mg kg−1 (fentanyl 1.5 μg kg−1) (Guo)17	Preserved hemodynamic stability25,26,48	B (LOC ED95, Oh)/C (intubation-response ED95, CSEA sedation ED95, and fentanyl co-administration ED50 — all small single-center dose-finding studies)	[12, 17, 25, 26, 48]
Adults (GA)	Age-banded induction bolus based on ED95 for LOC: 0.25–0.33 (< 40 yr), 0.19–0.25 (40–60 yr), 0.14–0.19 mg kg−1 (60–80 yr); no estimate published for > 80 yr. Age bands as corrected in Br J Anaesth 2022;129:831; the originally published Abstract and Discussion assigned these three ranges to patients aged < 40, 60–80 and > 80 yr in error, and the uncorrected version has been reproduced in several subsequent reviews.Single most representative value 0.19 mg kg−1 for a typical 60-year-old. Respiratory-depression ED50/ED95 0.14/0.27 mg kg−1, a margin over the LOC ED50/ED95 of 0.11/0.19 mg kg−1 that narrows with increasing age	Non-inferior efficacy vs propofol; less hypotension at clinical dose	A (non-inferiority/hypotension)/C (ED95 and age-banded doses, Chae; n = 20 per dose group)	[6, 14]
Pediatric (1–18 yr)	1–12 yr: ED50 0.30–0.42 and ED95 0.43–0.57 mg kg−1 for LOC in16	No drug-related serious adverse events; transient hypotension common in the procedural-sedation cohort	C (ED50/ED95 estimates from small single-center dose-finding studies of heterogeneous design; pediatric popPK model-based dosing not prospectively validated)	[16, 27, 28]
	3–15 yr: ED50 0.19 mg kg−1 in28
	6–18 yr: EC50 for UMSS ≥3 (procedural sedation) 777 ng mL−1 without fentanyl → 287 ng mL−1 at fentanyl 4 ng mL−1 in27
Hepatic impairment — severe (Child–Pugh C) for clearance/recovery; cirrhosis, class unspecified, for the sedation ED	● CL 38% lower/AUC 38% higher in severe hepatic impairment (Child–Pugh ≥ 10)24	Prolonged recovery to full alertness in severe HI (16.7 vs 8.0 min after 0.1 mg kg−1); no serious or severe AEs, all mild (numerically more TEAEs in impaired groups)	B for “no adjustment in mild–moderate, cautious titration in severe HI” (dedicated PK + label); ED50/ED95 cirrhosis figures C (pilot, different salt and endpoint)	[24, 30, 31]
	● Cmax not meaningfully altered after a single bolus, so no adjustment for mild–moderate impairment30 (besylate)
	● ED50 0.097/ED95 0.107 mg kg−1 for successful endoscopy in cirrhosis with sufentanil 0.15 μg kg−131 (tosilate; small pilot, wide and unstable CIs — treat as indicative, and not applicable to Child–Pugh B or C)
Obese/overweight adults (bariatric only for the ED anchor)	ED50 0.115/ED95 0.179 mg kg−1 (LBW) with sufentanil 0.5 μg kg−1 (LBW), bariatric surgery, up-and-down in.34 Weight-scalar data: per-kg induction dose normalizes on LBW/CBW rather than TBW in32,33	Stable cardiovascular profile; no respiratory depression, intraoperative awareness, or delayed emergence; PONV ~39%	C — ED anchor from a single small observational trial with a wide ED95 CI; scalar evidence converging on LBW/CBW but still single-center and heterogeneous	[32–34]
ESRD — non-dialysis for PK; dialysis (RRT) data pending	● Parent-drug clearance comparable to matched healthy controls in non-dialysis ESRD; no adjustment for renal impairment not requiring dialysis30 ● CNS 7054 (inactive metabolite) accumulates as glomerular filtration declines30 ● Effect of dialysis on CNS 7054 disposition untested ● Dialysis-population RCT versus dexmedetomidine registered and protocol-published; results pending15	● No serious adverse events in the dedicated PK study30 ● In transplant recipients: no adverse effect on graft function or urine output, no delayed graft function, less induction hypotension and injection pain than propofol, but longer time to LOC and longer PACU stay35	C — single non-dialysis PK study + one transplant RCT; no clearance data in patients on RRT	[15, 30, 35]
Cardiac surgery — valve replacement (bolus); CABG (infusion)	Bolus 0.2 mg kg−1 gives hemodynamic stability comparable to etomidate; 0.3 mg kg−1 increases MAP fluctuation, hypotension, and norepinephrine use. Less injection pain and myoclonus than etomidate at both doses	At 6 mg kg−1 h−1 infusion, failed non-inferiority to etomidate for post-induction hypotension in CABG (50.7% vs 34.2%; rate difference 16.5%, 95% CI 3.0–32.6, margin 12%); vasopressor requirement similar	B — two single-center RCTs, favorable at low bolus dose but non-inferiority not met vs etomidate by infusion	[49, 50]
Sleep disorders/OSA	In PSQI-defined sleep-disorder patients aged 18–64 yr, induction ED50 0.226/ED95 0.237 mg kg−1, higher than in non-sleep-disorder patients (ED50 0.191/ED95 0.209). These are opioid-free estimates, since sufentanil was given only after a bispectral index of 60 or below had been reached. Small single-center up-and-down trial (n = 27 and 29) with confidence intervals implausibly tight for the sample size — indicative only. No OSA dosing data	In OSA patients undergoing DISE, lower hypoxemia than propofol (25.0% vs 62.5%; RR 0.40) and fewer adverse events, with comparable sedation success — not absence of hypoxemia	C — small single-center trials; ED in non-OSA sleep-quality patients, OSA evidence limited to one comparative-safety RCT	[18, 37]

Notes: Evidence tags: A, multicenter randomized controlled trials or quantitative meta-analyses; B, single-center randomized trials or prospective dose-finding studies with ≥ 30 participants per group; C, small (< 30 per group) single-center dose-finding studies, retrospective cohorts, or extrapolation from population PK/PD simulation without prospective external validation. The arrow (→) denotes a decrease to the estimate obtained at the higher opioid exposure stated after the arrow.

Abbreviations: AE, adverse event; AUC, area under the concentration–time curve; CABG, coronary artery bypass grafting; CBW, corrected body weight; CI, confidence interval; CL, clearance; Cmax, maximum plasma concentration; CNS 7054, inactive carboxylic-acid metabolite of remimazolam; CSEA, combined spinal–epidural anesthesia; DGF, delayed graft function; DISE, drug-induced sleep endoscopy; EC₅₀, 50% effective concentration; ED₅₀, 50% effective dose; ED₉₅, 95% effective dose; ESRD, end-stage renal disease; GA, general anesthesia; HI, hepatic impairment; LBW, lean body weight; LOC, loss of consciousness; MAP, mean arterial pressure; OR, odds ratio; OSA, obstructive sleep apnea; PACU, post-anesthesia care unit; PK/PD, pharmacokinetic/pharmacodynamic; PONV, postoperative nausea and vomiting; popPK, population pharmacokinetic; PSQI, Pittsburgh Sleep Quality Index; RCT, randomized controlled trial; RR, risk ratio; RRT, renal replacement therapy; TBW, total body weight; TEAE, treatment-emergent adverse event; UMSS, University of Michigan Sedation Scale.

Table 3.

Published Population PK and PK/PD Models for Remimazolam: Comparative Appraisal

Model (Lead Author, Year)	Population	n	Structural Model	Key Covariates Retained	External Validation?	Embedded in Commercial TCI?
Schüttler et al, 2020 (Anesthesiology 132:636)51	Healthy male volunteers	20	3-compartment + effect compartment (MOAA/S-anchored)	Body weight (on V1 only)	No	Research prototype only
Zhou et al (Paion) 2020 (J Clin Anesth 66:109899)52	Healthy + surgical + ICU adults (pooled, 11 trials)	689 (popPK)/604 (popPK-PD)	3-compartment PK/PD on BIS	Sex, ASA class, extracorporeal circulation, time-dependent CL; PD: remifentanil synergy, cumulative dose, BMI, Asian race (weight allometric)	Internal only	Not in commercial pumps
Masui et al, 2022 (J Anesth 36:493)53	Surgical patients + healthy volunteers	662	3-compartment + virtual venous compartment	Adjusted body weight, age, sex, ASA-PS	Internal (VPC)	Not in approved pumps; TCI-oriented, implemented in TCI simulation software (TivaTrainer) with companion ke0
Eleveld et al, 2025 (Br J Anaesth 135:206)44	Mixed, broad age range	933 (20 pooled studies)	Allometrically scaled across age range	Weight, age, opioid effect (also sex, hepatic/renal, ECMO, ICU)	No (external validation pending)	Not yet (planned)
Colin et al, 2025 (Anesthesiology 143:368)27	Children/adolescents 6–18 yr	31	popPK/PD (proportional-odds + simplified Minto)	Weight (a priori allometric) + fentanyl PD interaction; no demographic covariate retained	No	No
Stöhr et al, 2021 (Br J Anaesth 127:415)30	Child–Pugh B/C (moderate+severe) + ESRD + matched controls	43 recruited (hepatic 11, renal 11, controls 21); PK-modelled ≈38	3-compartment recirculatory + popPK update	Hepatic function class (Child–Pugh ≥10 on CL; ≥9 on volumes) (+ allometric weight)	No	Reference for dose-adjustment principle, not for TCI

Notes: The key qualitative findings — no external validation, no TCI deployment, and a narrow covariate space — hold across all rows.

Abbreviations: ASA-PS, American Society of Anesthesiologists physical status; BIS, bispectral index; BMI, body mass index; CL, clearance; ECMO, extracorporeal membrane oxygenation; ICU, intensive care unit; ke0, effect-site equilibration rate constant; MOAA/S, Modified Observer’s Assessment of Alertness/Sedation; PD, pharmacodynamic; PK, pharmacokinetic; popPK, population pharmacokinetic; TCI, target-controlled infusion; V1, volume of the central compartment; VPC, visual predictive check.

Pharmacokinetic and Pharmacodynamic Background

The pharmacokinetic profile of remimazolam is characterized by rapid onset, short duration of action, and predictable dose–response relationships.4,24 Following intravenous administration in healthy adults, remimazolam is rapidly distributed (0.5–2 min) and hydrolyzed by tissue esterases, with an elimination half-life of approximately 7–11 min.24 Reported terminal half-lives ^{differ by model and sampling scheme,} ranging from 37–53 min in pooled analyses24 to 70 ± 10 min in a dedicated Phase I study.51 The context-sensitive half-time remains short even after continuous infusion for up to 4 h (6.8 ± 2.4 minutes).51,54

Population PK/PD modeling has identified key covariates influencing remimazolam disposition. Body weight and age are the principal determinants of clearance and volume of distribution, with elderly patients exhibiting enhanced pharmacodynamic sensitivity.27,44,53 A three-compartment population model derived from 662 surgical patients and volunteers additionally identified adjusted body weight, age, sex and ASA physical status as significant covariates of remimazolam clearance.53 Importantly, the organ-independent esterase metabolism means that hepatic impairment primarily affects continuous infusion pharmacokinetics rather than single-bolus disposition, whereas renal impairment has minimal effect on drug clearance.24,30

Purpose and Scope of the Review

Aims and Objectives

This focused narrative review aims to synthesize the current dose-optimization evidence for remimazolam across diverse patient populations and clinical indications, with explicit attention to the strength and external validity of the underlying studies. The specific objectives are: (1) to evaluate remimazolam dosing regimens across varied patient populations (including elderly, pediatric, obese, hepatic/renal impairment, critically ill, and patients with sleep disorders) and different procedural indications (induction, maintenance, procedural sedation, MAC); (2) to benchmark existing population PK/PD models guiding individualized dosing; (3) to identify and synthesize dose–response data including ED₅₀, ED₉₀, and ED₉₅ values in special populations; (4) to compare safety and efficacy outcomes across different dosing protocols; and (5) to deconstruct factors influencing dosing optimization, including age, comorbidities, and concomitant medications.

Search Strategy

This focused narrative review was informed by a structured literature search of PubMed/MEDLINE, Embase and the Cochrane Central Register of Controlled Trials (CENTRAL). Searches covered 1 January 2020 to 30 April 2026; the January 2020 start was chosen to coincide with the first regulatory approval of remimazolam (Japan) and the subsequent surge of post-approval clinical studies. In each database, controlled vocabulary (MeSH/Emtree) and free-text terms for remimazolam were combined with either a block of dosing, pharmacokinetic and pharmacodynamic terms or a block of sedation and anesthesia terms, using the structure remimazolam AND (dosing/PK-PD OR sedation/anesthesia); the full strings, database-specific limits and record yields are reproduced in Supplementary Appendix 1. The strategies were re-executed on 9 September 2026 with the publication-date window held at 1 January 2020 to 30 April 2026, and the record counts in Figure 1 derive from the exported results of that run. The reference lists of the included studies were additionally screened by backward citation searching, which identified no eligible study that had not already been retrieved by the database searches.

Figure 1.

Flow of records through the structured literature search. Records were identified in PubMed/MEDLINE, Embase and CENTRAL (English language; published 1 January 2020 to 30 April 2026; strategies re-executed 9 September 2026), de-duplicated, and reduced by pre-specified record-level rules; the 92 studies synthesized were selected by the authors from the resulting pool, and backward citation searching identified no additional studies. Because per-record exclusion reasons at the title/abstract and full-text stages were not documented, the diagram reports only the stages reproducible from the search exports and is not a PRISMA flow chart.

Abbreviations: CENTRAL, Cochrane Central Register of Controlled Trials; ICTRP, International Clinical Trials Registry Platform.

Eligibility Criteria

Pre-specified inclusion criteria were: (i) peer-reviewed randomized controlled trials, dose-finding sequential-allocation studies, prospective cohort studies, population PK/PD modeling analyses, and quantitative systematic reviews or meta-analyses; (ii) human participants of any age; (iii) studies reporting interpretable dose, ED, EC, or PK/PD parameters for procedural sedation, MAC, or general anesthesia; (iv) English-language full text available. We excluded all studies that focused on nonhuman research, sex differences, case reports without quantitative dosing data, conference abstracts without a full-text counterpart, editorials, narrative reviews not contributing primary data, and bioequivalence studies of remimazolam salt forms (besylate vs tosilate) reporting no new clinical PK/PD data.

Screening and Selection

Records exported from the three databases were de-duplicated by PubMed identifier, digital object identifier and normalized title, after which pre-specified record-level rules were applied: trial-registry entries retrieved through CENTRAL, records published or first indexed after 30 April 2026, records without an English-language full text, conference abstracts, animal-only studies, and errata or retraction notices were removed. The three publication-oriented database streams yielded 3,488 records: 1,215 from PubMed/MEDLINE, 1,664 from Embase, and 609 publication records from CENTRAL. CENTRAL yielded 1,772 records in total, comprising 609 publication records and 1,163 trial-registry records, including 971 records from WHO ICTRP and 192 from ClinicalTrials.gov. Trial-registry records were not peer-reviewed publications and were not carried forward into the publication-record pool. After removal of 1,702 duplicates and 370 pre-specified rule-based exclusions from the 3,488 publication-oriented records, 1,416 records remained (Figure 1). Within this pool, the 92 studies synthesized here were selected by the authors for their relevance to population-specific dose optimization according to the eligibility criteria above, and all 92 were confirmed to lie within the de-duplicated record set. The final selection of 92 studies was reviewed independently by two authors, who agreed on its composition. Because study selection for this narrative review was not documented as a formal two-stage screen with per-record exclusion reasons, Figure 1 reports only the stages that can be reproduced from the search exports and does not partition the 1324 non-selected records into title/abstract and full-text exclusions. The reference list contains 97 entries: the 92 synthesized studies together with five sources cited for context rather than as sources of remimazolam dosing data — our related dexmedetomidine dosing review, two publications on electroencephalographic monitoring of anesthetic depth, and two earlier publications on sleep physiology cited for mechanistic background in the section on patients with sleep disorders.

Methodological Scope

As this is a focused narrative review rather than a systematic review with meta-analysis, a flow diagram of the search and record-handling process is provided for transparency (Figure 1), but no PRISMA-compliant systematic review protocol was registered and no formal risk-of-bias assessment was undertaken. To preserve transparency about evidence strength, the consolidated dosing recommendations are presented in Table 1, and each population-level entry in Table 2 is tagged qualitatively as A, B or C according to the criteria described below.

Assessment of Evidence Strength

To convey the heterogeneity of evidence supporting individual dosing recommendations without implying systematic-review methodology, each row of Table 2 is annotated with an evidence tag: A, supported by multicenter randomized trials or by quantitative meta-analyses; B, supported by single-center randomized trials or by prospective dose-finding studies with at least 30 participants per group; and C, derived from small (fewer than 30 per group) single-center dose-finding studies, retrospective cohorts, or extrapolation from population PK/PD simulation without prospective external validation. Most ED₅₀, ED₉₀ and ED₉₅ point estimates for special populations fall into category C, and clinicians should regard them as best-available estimates rather than as established dosing standards. Quantitative summaries in this review — for example, the approximate 25–30% dose reduction in older patients and the 15–63% reduction with opioid co-administration — are narrative syntheses of individual studies rather than pooled meta-analytic estimates, and are presented as approximate ranges. The principal published population pharmacokinetic and PK/PD models underpinning these recommendations are compared in Table 3 with respect to their study populations, structural form, retained covariates, external-validation status, and implementation in target-controlled infusion systems.

Overview of Remimazolam Dosing Recommendations

The evolution of remimazolam dosing recommendations has progressed from early Phase I volunteer studies through pivotal Phase III trials to contemporary population PK/PD model-informed dosing frameworks. Initial dose-finding studies in healthy volunteers established the fundamental dose–response relationship and identified the therapeutic window for sedation and anesthesia.51,55–57 Subsequent regulatory trials refined these recommendations for specific clinical indications, with notable regional differences in approved dosing regimens reflecting variations in trial populations and regulatory requirements.7,19,52,58

Across the procedural sedation literature, evidence converges on effective sedation for procedures of less than approximately one hour; consensus is weaker for long-duration anesthesia exceeding 4 h and is absent for prolonged ICU sedation, where the discontinued Japanese Phase II program provides a counter-narrative. Divergence persists in three areas, each addressed in a dedicated section below: the choice of weight scalar in obesity,32,33 the wide spread of pediatric effective-dose estimates, and the degree of dose sparing conferred by opioid co-administration. Figure 2 renders the resulting recommendations as a bedside decision sequence, from patient assessment through indication-based baseline dosing, population-specific adjustment, opioid co-administration and titration, to reversal, recovery and the critical-care caveat. These divergences illustrate how strongly endpoint definition shapes a published estimate. Pediatric studies have demonstrated higher sedative potency for remimazolam (ED₅₀ 0.19 mg kg⁻¹) than propofol (ED₅₀ 1.11 mg kg⁻¹) for loss of consciousness (LOC).28 The ED₅₀ and ED₉₅ for laryngeal mask airway insertion responses vary significantly by age subgroup,29 but the overall outcome suggests higher doses for LOC and airway intervention in pediatric patients.16,29,59,60 Effective-dose estimates for preventing movement during surgical abortion vary with the co-administered opioid: ED₉₅ 0.242 mg kg⁻¹ with sufentanil 0.1 μg kg⁻¹,61 versus ED₉₀ 0.615 mg kg⁻¹ with remifentanil62 (both small single-center trials, n = 25 and 57, respectively), suggesting that endpoint definition, patient age, and body mass index critically influence dosing estimates.6,32,61–64

Figure 2.

Bedside dosing algorithm for remimazolam across patient populations.

Abbreviations: ED₉₅, 95% effective dose; EEG, electroencephalography; ESRD, end-stage renal disease; ICU, intensive care unit; LOC, loss of consciousness; MAC, monitored anesthesia care; MOAA/S, Modified Observer’s Assessment of Alertness/Sedation; OSA, obstructive sleep apnea; PSQI, Pittsburgh Sleep Quality Index.

The role of processed electroencephalography (pEEG) monitoring in guiding remimazolam titration represents another area of active investigation, with emerging evidence supporting density spectral array and bispectral index (BIS) monitoring for real-time dose adjustment.65,66 The integration of pEEG-guided dosing into clinical practice may help reconcile some of the observed variability across studies.44,67,68 Subsequent sections delineate remimazolam dosing details across various clinical settings, including procedural sedation, anesthesia and special patient populations.

Remimazolam in Procedural Sedation

Outpatient Procedural Sedation

Remimazolam has demonstrated efficacy and safety for outpatient procedural sedation across diverse settings, including dental surgery,69 emergency department procedures,45 and ambulatory endoscopy.70 In outpatient dental surgery, remimazolam provided adequate sedation with rapid recovery, facilitating same-day discharge.69 A key advantage of remimazolam in outpatient settings is its predictable offset, with recovery times consistently shorter than those observed with midazolam and comparable to propofol, while offering the additional safety net of flumazenil reversibility.1,71

Gastrointestinal Endoscopy

Gastrointestinal endoscopy represents the most extensively studied indication for remimazolam procedural sedation. A landmark randomized noninferiority trial by Zhu and colleagues demonstrated that remimazolam 0.2 mg kg⁻¹ for gastroscopy was non-inferior to propofol, with significantly less hypotension.46,72 Age-stratified dose-finding studies in elderly patients undergoing gastroscopy have determined an ED₉₀ of 0.4 mg kg⁻¹ for inhibiting responses to scope insertion73,74 and an ED₉₅ of 0.204 mg kg⁻¹ (95% CI 0.175–0.390) for adequate sedation at scope insertion with remifentanil 0.2 μg kg⁻¹ co-administration.75 Because these two estimates derive from different procedural endpoints and patient populations, they are not directly comparable and do not represent a dose–response inversion. The synergistic interaction with fentanyl has been quantified: in elderly patients undergoing painless gastroscopy, the remimazolam ED₅₀ fell from 0.258 mg kg⁻¹ with fentanyl 0.5 μg kg⁻¹ to 0.162 mg kg⁻¹ at 1.0 μg kg⁻¹ and 0.095 mg kg⁻¹ at 1.5 μg kg⁻¹, with corresponding ED₉₅ values of 0.272, 0.175 and 0.109 mg kg⁻¹. All three arms received fentanyl, so the study provides no opioid-free reference dose.17 These data underscore the importance of accounting for opioid co-administration when selecting remimazolam doses.17,38

Safety and Efficacy in Procedural Sedation

Pooled analyses of short-procedure randomized trials report a lower incidence of clinically significant hypotension and respiratory depression with remimazolam than with propofol; the absolute magnitude of these differences varies by procedure, opioid co-administration regimen and patient population.41–43,72 Remimazolam preserves patient satisfaction and procedural conditions comparable to propofol, with the added advantage of flumazenil reversibility in cases of oversedation and reduced injection pain.76

Remimazolam in Monitored Anesthesia Care and Cardiac Procedures

Remimazolam has shown particular promise for MAC during cardiac procedures, where hemodynamic stability is paramount.36 Comparative studies with dexmedetomidine for transcatheter aortic valve replacement (TAVR) sedation have demonstrated shorter arousal times with remimazolam. Importantly, remimazolam provides effective sedation and maintains stable hemodynamic parameters during TAVR when administered as a weight-based continuous infusion rather than a single bolus.77,78 In a retrospective pediatric and young-adult congenital cardiac cohort (n = 26, median age 18 yr), a bolus of 0.03–0.11 mg kg⁻¹ followed by a maintenance infusion of 0.3–1.2 mg kg⁻¹ h⁻¹ titrated to effect has been reported without clinically significant hemodynamic depression.36 Beyond MAC, remimazolam has also been evaluated for induction in cardiac surgery. In cardiac surgery, low-dose remimazolam (0.2 mg kg⁻¹) provided stable induction hemodynamics with less injection pain and myoclonus than etomidate, whereas 0.3 mg kg⁻¹ increased hypotension; in coronary artery bypass grafting it did not meet non-inferiority to etomidate for post-induction hypotension.49,50

In urological procedures, remimazolam-based MAC offered advantages over inhalation-based general anesthesia by reducing induction and recovery times with lower incidence of hypotension and less anesthesia costs.79

Current Dosing Strategies for Remimazolam Across Patient Populations

General Adult Population

Current dosing guidelines for remimazolam are tailored to clinical indication, patient population, and co-medications. For procedural sedation in adults, a fixed-dose regimen is widely applied, consisting of an initial bolus of 5 mg administered over 1 minute, followed by maintenance top-up doses of 2.5 mg titrated to achieve a Modified Observer’s Assessment of Alertness/Sedation (MOAA/S) score of 3. In general anesthesia, weight-based infusion pump guidelines utilize an induction rate of 6 to 12 mg kg⁻¹ h⁻¹ until LOC, followed by maintenance infusions typically set at 1.0–2.0 mg kg⁻¹ h⁻¹ and adjusted dynamically based on patient age and depth-of-anesthesia monitoring metrics.24

Population-Specific Dosing Considerations

Elderly Patients

The elderly population consistently demonstrates increased sensitivity to remimazolam, requiring an approximate 25–30% dose reduction compared with younger adults (≥ 65 years is retained here as the nominal reference category, although the cited elderly cohorts spanned 60–79 years for Oh et al, 60–80 years for Song Y and Zou, and ≥ 65 years for Song JC et al).12 The ED₉₅ for LOC and intubation in elderly patients is 0.249–0.297 mg kg⁻¹, compared with approximately 0.37 mg kg⁻¹ in younger adults.25,48 This magnitude is supported by two independent datasets: the age-stratified ED₉₅ series of Chae et al implies a 24–32% reduction between a 40-year-old and a 60- to 70-year-old, and Oh et al reported a 32% lower ED₉₅ in their elderly group (0.249 mg kg⁻¹). These two series are not on a common scale: Chae et al defined loss of consciousness within 5 min and Oh et al within 3 min, so the age-banded 60–80 yr range of 0.14–0.19 mg kg⁻¹ and the elderly ED₉₅ of 0.249 mg kg⁻¹ describe different endpoints rather than a contradiction. A further estimate of 0.336 mg kg⁻¹ has been reported for a bispectral index below 60 with sufentanil co-administration, indicating that the endpoint and its permitted time window, rather than age alone, determine the reported effective dose. Age-related reductions in clearance and increased pharmacodynamic sensitivity both contribute to the enhanced drug effect in this population.24,44,48 For elderly patients undergoing MAC with spinal–epidural anesthesia for hip replacement, the ED₅₀ was 0.212 mg kg⁻¹ h⁻¹ and the ED₉₅ was 0.288 mg kg⁻¹ h⁻¹.26 Remimazolam-based anesthesia with flumazenil reversal allows faster emergence than propofol in older patients undergoing spinal surgery.80

Pediatric Patients

In contrast to elderly patients, pediatric patients (aged 1–18 yr) require higher weight-adjusted doses. Published pediatric estimates are not a single coherent value: in children aged 1–12 years, an ED₅₀ of 0.30–0.42 mg kg⁻¹ and an ED₉₅ of 0.43–0.57 mg kg⁻¹ for loss of consciousness have been reported, whereas a dose-equivalence study in children aged 3–15 years reported a lower ED₅₀ of 0.19 mg kg⁻¹; these differences reflect divergent age ranges, endpoints and study designs (all graded C) rather than a consensus dose, so clinicians should titrate to a clinical endpoint rather than adopt a single recommended milligram-per-kilogram dose. Higher doses produce faster onset without hemodynamic instability or the need for pharmacological intervention to resolve transient changes in mean arterial blood pressure.11,16,28,29,59,81,82 Recent pharmacokinetic modeling in children and adolescents aged 6–18 yr has characterized the EC₅₀ for targeting UMSS ≥3 at 777 ng mL⁻¹ without opioid. However, when co-administered with fentanyl, required remimazolam EC₅₀ dropped in a concentration-dependent manner with rising fentanyl plasma concentrations: decreasing to 655 ng mL⁻¹ (at 1 ng mL⁻¹ of fentanyl), 533 ng mL⁻¹ (at 2 ng mL⁻¹ of fentanyl), and ultimately down to 287 ng mL⁻¹ when paired with a fentanyl steady-state concentration of 4 ng mL⁻¹.27

Patients with Hepatic Impairment

The peak concentrations (C_max) are preserved after a single bolus in patients with severe hepatic impairment (Child–Pugh C). However, a profound 38.1% drop in systemic clearance occurs in advanced liver disease. The reduction in clearance prolongs context-sensitive half-time, and yields delayed recovery (16.7 vs 8.0 min).24,30 This might be cautiously compared with the findings of Cao et al, who reported an ED₅₀ of 0.097 mg kg⁻¹ and an ED₉₅ of 0.107 mg kg⁻¹ for procedural sedation with adjuvant sufentanil 0.15 μg/kg in cirrhotic patients undergoing esophagogastric varices screening endoscopy.31 Taken together, these studies suggest that while single-bolus requirements may appear preserved, continuous or repeated dosing should be cautiously down-scaled to prevent drug accumulation and delayed emergence.30,31,41,83

Patients with Renal Impairment

Remimazolam is hydrolyzed by tissue esterases, predominantly CES1, so clearance is largely independent of renal function. In a dedicated pharmacokinetic study, plasma clearance in subjects with severe renal impairment not requiring dialysis was comparable to that of matched healthy controls, and all adverse events were mild.30 These data support standard adult dosing in this group. Two qualifications apply. First, the inactive metabolite CNS 7054 is renally eliminated and accumulates as glomerular filtration declines; although it has low GABAA affinity and is not expected to prolong sedation, its disposition during dialysis has not been formally studied.19,30 Second, in a randomized trial of 86 uremic patients undergoing living-donor kidney transplantation, remimazolam 0.2 mg kg⁻¹ followed by 1 mg kg⁻¹ h⁻¹ was associated with a longer time to loss of consciousness and a longer post-anesthesia care unit stay than propofol, so recovery profiles in this population may differ from those in patients with normal renal function.35 A randomized trial comparing remimazolam with dexmedetomidine for procedural sedation in ESRD patients is currently underway; results have not yet been reported.15

Obese Patients

Dosing in obese patients remains an area of active investigation. Available evidence suggests that corrected body weight or lean body weight-based dosing is preferred over total body weight, as total body weight-based dosing results in relatively higher total doses and potential overdosing.32,33 The ED₅₀ and ED₉₅ for obese patients undergoing bariatric surgery with sufentanil co-administration are 0.115 and 0.179 mg kg⁻¹, respectively.34

Patients with Sleep Disorders

A single dose-finding study has reported higher induction-dose requirements in patients with sleep disorders. The calculated ED₅₀ and ED₉₅ for LOC were 0.226 and 0.237 mg kg⁻¹ in the sleep disorder cohort, compared to 0.191 and 0.209 mg kg⁻¹ in patients without sleep disorder, respectively. This represents an approximate 13% to 18% increase in induction-dose requirements.18 The finding is mechanistically plausible given the GABAergic alterations in chronic sleep disruption but it has not been prospectively replicated, and the magnitude of effect at the population level remains uncertain.18,84,85

From a clinical standpoint, these data indicate that preoperative sleep quality assessment — using validated instruments such as the PSQI — should be integrated into the pre-sedation evaluation to guide remimazolam dose selection, while anesthetic requirements should be adjusted for broader hypersensitivity phenotypes.18,37 The overall favorable hemodynamic and recovery profile of remimazolam, combined with its immediate reversibility via flumazenil, offers an appealing safety profile for diverse procedural sedations.47 Nonetheless, the existing evidence is limited to a single dose-finding study with modest sample size, and prospective multicenter trials are warranted to validate these findings and to establish specific dosing algorithms for patients with varying severities of sleep-disordered breathing.18,86,87

Clinical Practice Recommendations

The population-specific evidence reviewed above can be reduced to a small number of operational rules. Based on the synthesized evidence, the following clinical practice recommendations are proposed to guide remimazolam dosing across diverse patient populations and procedural settings (Table 1 and Table 2). Table 1 summarizes remimazolam dosing recommendations across all patient populations, including indication, loading/induction dose, maintenance/titration dose, and supporting references.

Preoperative Assessment and Dose Planning

Systematic preoperative assessment should identify patient-specific factors that influence remimazolam dosing requirements, including age,12,16,54 body weight,32,33 hepatic and renal function,30,31 sleep disorder status,18 and planned concomitant medications.17,38

Titration Strategy and Sedation Endpoints

Remimazolam should be titrated to validated clinical endpoints, primarily the MOAA/S scale. Integration of pEEG monitoring, especially density spectral array, may provide additional guidance for dose titration.65,68

Opioid Co-Administration

When opioids are co-administered, remimazolam requirement falls by approximately 15–63%, depending on the opioid dose and clinical context. This synergistic interaction has been confirmed by PK/PD modeling and should be accounted for in the initial dose selection.17,27,38

Flumazenil Reversal and Post-Procedural Monitoring

Flumazenil reversal is an important safety feature of remimazolam. Reversal doses of 0.5 mg produce more rapid awakening and extubation than propofol.80 However, clinicians should be aware of the risk of re-sedation,88 particularly in the context of prolonged infusions19 or concurrent opioid administration.

Special Population Considerations

For each special population, dosing should be individualized based on the available quantitative dose–response data summarized in Table 1 and Table 2. Model-informed precision dosing (MIPD) frameworks, incorporating patient covariates identified by population PK/PD models, represent the most promising approach for optimizing remimazolam dosing across diverse clinical scenarios (Figure 2).

Unresolved Controversies in Remimazolam Dosing

Lessons from the Discontinued Japanese ICU Sedation Program

The clearest unresolved controversy in remimazolam dose optimization is the appropriate use, if any, of the drug for long-duration ICU sedation. The Japanese Phase II trial ONO-2745-04, halted in 2013, retains direct relevance to current dosing decisions despite predating widespread regulatory approval.19 The trial enrolled postoperative patients expected to require mechanical ventilation for at least six hours. After cessation of pre-existing sedatives, remimazolam was initiated at 0.1, 0.25 or 0.5 mg kg⁻¹ h⁻¹ once the Richmond Agitation–Sedation Scale (RASS) score reached −1 to −2, and titrated to RASS −1 to −4 up to a ceiling of 2 mg kg⁻¹ h⁻¹ for up to seven days. The pre-specified efficacy endpoint was met at interim analysis: 35 of 36 patients (97%) maintained the target RASS range for at least 70% of the administration time, confirming that titratable sedation was achievable in critically ill patients. Among the seven patients infused for 24 h or longer, however, parent-drug plasma concentrations rose far above model-predicted levels (peak 2742.7 ± 2406.8 vs 352.0 ± 194.7 ng mL⁻¹ in the 42 patients infused for less than 24 h), and every patient in this subgroup required at least 1 h for complete awakening, with a mean time to awakening of 418.0 ± 396.0 min — an offset roughly an order of magnitude slower than the 7–11-min elimination half-life and short context-sensitive half-time reported in healthy volunteers, and consistent with impaired, time-dependent clearance rather than simple dose-dependent accumulation.19,24

A time-dependent reduction in apparent clearance of approximately 25% after 22 hours of infusion (≈30% at 24 h) has been reported, tracking duration of infusion rather than cumulative dose.24,52 A subsequent prospective ICU pilot study, however, found no evidence of time-dependent accumulation and a stable clearance over infusions extending beyond 24 hours,20 underscoring that the phenomenon remains unresolved in critically ill patients. Three concurrent mechanisms are considered plausible: (i) reduced CES1 activity in the setting of systemic inflammation, organ dysfunction or compromised hepatic perfusion; (ii) accumulation of the inactive metabolite CNS 7054 in patients with acute kidney injury — clinically inert but with implications for assay interpretation; and (iii) an increased free fraction of remimazolam in patients with severe hypoalbuminemia, which enhances clinical effect without altering total clearance and is particularly relevant in septic shock and in burn injury.4,19,89,90

In a single-center prospective dose–response study (n = 23 postoperative ICU patients), Chen and colleagues identified an effective regimen of a 0.02–0.05 mg kg⁻¹ loading dose followed by maintenance starting at 0.2–0.35 mg kg⁻¹ h⁻¹ titrated to Narcotrend Index 65–94, with preserved respiratory and hemodynamic stability.91 Tang and colleagues reported three studies in this population: an open-label dose-finding study and two randomized pilot trials against propofol.39,92,93 In the deep-sedation pilot (RASS −4 to −5; 30 patients per arm), target RASS was reached a median of 100% of the sedation time in both groups, with no differences in ventilator-free hours, successful extubation, ICU length of stay or 28-day mortality; hypotension occurred at comparable rates (remimazolam 53.3% vs propofol 60.0%, p > 0.05) and no patient developed bradycardia, leading the investigators to conclude that remimazolam was effective and safe for short-term deep sedation.93 That hypotension nonetheless affected roughly half of patients in both arms indicates that the hemodynamic advantage of remimazolam demonstrated in procedural and lighter-sedation settings should not be extrapolated uncritically to deep-sedation targets in critically ill patients. Nevertheless, a single-center observational study found that long-term remimazolam sedation in mechanically ventilated ICU patients caused no excess in major clinical outcomes, such as adverse events or ICU mortality, when compared with traditional sedatives,40 and a European Phase 2 open-label pilot confirmed feasibility of titration to RASS −2 to 0 in critically ill patients, reporting preserved remimazolam clearance and rapid awakening despite critical illness.94 Recently, an ICU pharmacokinetic study has begun to characterize inter-individual variability of clearance in mechanically ventilated patients receiving 24-hour infusion.83

When remimazolam is used in critically ill patients, infusions should be time-limited (typically < 24 h) and paired with explicit monitoring of sedation depth, hepatic and renal function, serum albumin and hemodynamics, since recovery may be prolonged where CES1 activity is compromised. The Japanese experience is a reminder that a favorable pharmacokinetic profile in healthy volunteers need not predict drug behavior in critically ill patients with multi-organ dysfunction.19

Carboxylesterase 1 Polymorphisms and Pharmacogenomic Uncertainty

Unlike traditional benzodiazepines such as midazolam, which rely heavily on hepatic cytochrome P450 pathways, remimazolam’s parent clearance is largely independent of these mechanisms. Indeed, clinical data in healthy volunteers has confirmed that genetic polymorphisms in Cytochrome P450 3A (CYP3A) and Cytochrome P450 Oxidoreductase (POR) were not found to exert a significant influence on remimazolam pharmacokinetics, whereas Vitamin D Receptor (VDR) polymorphisms (rs1544410 and rs4516035) have been reported to affect remimazolam exposure and metabolite disposition and therefore cannot be grouped with them.95 Instead, CES1 is encoded by a gene with documented functional polymorphisms, most notably the loss-of-function variant c.428G>A (p.G143E), with allele frequencies that differ between ancestries. Because remimazolam clearance depends almost entirely on CES1-mediated hydrolysis, even moderate reductions in enzymatic activity could plausibly translate into clinically meaningful prolongation of effect; however, direct evidence linking CES1 genotype to remimazolam clearance in patients is currently absent. Until prospective genotype-stratified pharmacokinetic studies are performed, the proportion of inter-individual variability attributable to CES1 polymorphism, as distinct from disease state, age and weight, remains unknown.11,96

The Optimal Weight Scalar for Obese Patients

Available evidence supports lean or corrected body weight rather than total body weight for remimazolam dosing in obese patients, on the rationale that the drug’s volume of distribution does not scale linearly with adipose tissue.32,33 The single dose-finding study in bariatric surgery using sufentanil co-administration is, however, not sufficient to determine whether lean body weight, corrected body weight, or an allometric scalar based on fat-free mass best predicts EC across the BMI range.34 Until dedicated prospective dose-finding studies spanning BMI 30 to ≥ 45 kg m⁻² are performed, the lean body weight/corrected body weight recommendation should be regarded as a pragmatic default rather than a validated rule.

Re-Sedation After Flumazenil Reversal

Flumazenil’s elimination half-life (~ 1 h) is comparable to that of remimazolam after a short infusion but considerably shorter than the apparent terminal half-life observed after prolonged infusion in critically ill patients.19,51,54,90 Moreover, re-sedation following remimazolam anesthesia has been reported sporadically, particularly in the context of opioid co-administration and prolonged remimazolam exposure.88 In addition, target-controlled infusion studies in healthy volunteers have demonstrated modest acute tolerance to remimazolam, suggesting that maintenance requirements may increase during prolonged administration.97 Taken together, pre-discharge observation periods after flumazenil reversal should be at least 60 minutes, and longer where opioid co-administration has occurred or where the remimazolam infusion has exceeded one hour at higher rates. Routine prophylactic flumazenil administration is not supported by current evidence and should not be used as a substitute for careful titration.

Limitations, Gaps, and Future Research Directions

Several key limitations and knowledge gaps characterize the current evidence base for establishing optimal remimazolam dosing strategies, highlighting critical directions for future research:

Sample Size and Power Constraints

Many dose-finding studies suffer from small sample sizes that restrict statistical power and generalizability, with some ED₉₅ estimates derived from cohorts with fewer than 30 patients per group.17,18 Future multi-center trials with larger sample sizes are required to validate these dosing standards globally.

Pediatric and Neonatal Deficit

Pediatric dose–response data stem primarily from single-center studies with limited age-range representation, while dosing in infants under one year and neonates remains entirely unstudied.11,16,28,29,59,81,82 Robust, prospective dose-finding studies with age-appropriate PK/PD sampling are urgently needed.

Ethnic Bias and Homogeneity

The majority of high-tier trials have been conducted in East Asian populations, leaving the impact of broader ethnic diversity on remimazolam pharmacokinetics incompletely characterized. Future clinical trials must actively enroll ethnically diverse global patient cohorts to ensure generalizability.43

Lack of Long-Term ICU Safety Data

Long-term safety data for continuous intensive care sedation are notably absent following the early termination of the Japanese Phase II ICU trial.19 Dedicated prospective trials tracking continuous infusions beyond 24 hours in critically ill patients are warranted to quantify long-term drug accumulation profiles.

Confounding Study Heterogeneity

Cross-study comparisons are hindered by heterogeneous dosing protocols, varying sedation endpoints, and concurrent opioid administration, which create profound central nervous system synergy that obscures the parent compound’s independent PK/PD parameters. Standardizing titration endpoints and conducting comparative effectiveness research against established sedatives will help isolate these drug interactions.

Pathophysiological Protein Shifts

Remimazolam is highly protein-bound (~91%), causing its active “free fraction” to rise sharply during severe hypoalbuminemia (eg, sepsis, major burns). Existing models lack the mathematical granularity to guide titration during these real-time protein shifts, requiring future PK/PD trials to explicitly incorporate comorbidity indices and organ dysfunction severity.19,90

Unexplored Pharmacogenomics

Systematically characterizing CES1 polymorphisms — specifically the common loss-of-function c.428G>A (p.G143E) variant — remains largely unexplored in vivo. Future research should prioritize genotype-guided dosing trials alongside the development of commercial, MIPD tools that integrate patient covariates in real time.11,96

Unstandardized Objective Monitoring

Integration of pEEG-guided dosing protocols65,66 into routine clinical practice and in future comparative effectiveness research against established sedatives in diverse procedural and surgical settings, will further refine the evidence base for remimazolam dose optimization.

Conclusion

Current evidence supports remimazolam as an effective sedative for procedural sedation and short-duration general anesthesia, with a hemodynamic and respiratory profile that compares favorably with propofol. Organ-independent clearance of remimazolam is conditional on intact CES1 activity, requiring caution in patients with impaired function of this enzyme. Age is the principal modifiable determinant of dosing. Elderly patients (≥ 65 yr) require approximately 25–30% lower doses, whereas pediatric patients (1–18 yr) require higher weight-adjusted doses. Obese patients are best dosed using lean or corrected body weight, though the optimal scalar in obesity remains unresolved.

Population PK/PD modeling has clarified the dose–exposure–response relationship in healthy volunteers and selected surgical populations, although clinical-grade MIPD is not yet available on commercial target-controlled infusion (TCI) platforms. Synergistic interactions with opioids permit 15–63% reductions in EC of remimazolam, and conditions such as sleep disorders modulate induction thresholds. Important evidence gaps persist for infants younger than one year, for long-duration anesthesia, and for the critically ill. Validated pediatric population PK/PD models, externally validated adult population PK/PD models embedded in clinical infusion platforms, and prospective characterization of CES1 polymorphism effects are needed. Large-scale, multicenter randomized trials, ideally enrolling broader (non-East-Asian) populations, will be required to refine dosing recommendations and to establish their generalizability. For the practicing anesthesiologist or intensivist, the immediate value of this review is the consolidated, evidence-graded starting doses in Table 1 (with the accompanying bedside algorithm in Figure 2): each recommendation carries an A, B or C tag so that estimates resting on multicenter trials can be distinguished at a glance from those extrapolated from small single-center studies — an explicit calibration of confidence that is the review’s principal methodological contribution. These recommendations should be applied with the limitations of the narrative (non-systematic) format in mind, as no formal risk-of-bias appraisal or meta-analysis was performed. The clearest practical message concerns the critically ill: remimazolam is not a routine long-term ICU sedative, and where it is used, continuous infusion should be time-limited (typically < 24 h) with monitoring of sedation depth, hepatic and renal function, serum albumin and hemodynamics, anticipating prolonged recovery when CES1 activity is compromised. Finally, dosing should remain patient-centered: covariate-informed starting doses are a beginning rather than an endpoint and should be titrated to the individual, with clear communication about expected recovery, particularly in older and organ-impaired patients.

Funding Statement

This work received no external funding. No sponsor had any role in the design, conduct, analysis, or reporting of this review or in the decision to submit it for publication.

Abbreviations

ASA, American Society of Anesthesiologists; BIS, bispectral index; BMI, body mass index; CES1, carboxylesterase 1; C_max, maximum plasma concentration; EC, effective concentration; ED, effective dose; ESRD, end-stage renal disease; GABA_A, γ-aminobutyric acid type A; ICU, intensive care unit; LOC, loss of consciousness; MAC, monitored anesthesia care; MIPD, model-informed precision dosing; MOAA/S, Modified Observer’s Assessment of Alertness/Sedation; pEEG, processed electroencephalography; PK/PD, pharmacokinetic/pharmacodynamic; PSQI, Pittsburgh Sleep Quality Index; RASS, Richmond Agitation–Sedation Scale; TAVR, transcatheter aortic valve replacement; TCI, target-controlled infusion.

Data Sharing Statement

No new data were generated or analyzed in this review article.

Ethics Approval and Informed Consent

This article is a narrative review of previously published studies and did not involve any new studies of human or animal subjects performed by any of the authors. Ethics approval and informed consent were therefore not required.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors report no conflicts of interest in this work.