Prophylactic effect of intraoperative sodium oxybate on postoperative delirium in older patients undergoing major orthopedic surgery: a randomized clinical trial

Abstract

Background

Postoperative delirium (POD) represents a significant challenge in perioperative care, particularly among older surgical patients. This acute neuropsychiatric syndrome is associated with prolonged hospitalization, increased mortality, and long-term cognitive decline. Sleep disturbance has emerged as a significant modifiable risk factor for POD. Sodium oxybate (SO), a gamma-aminobutyric acid B (GABA_B) receptor agonist with established sleep-enhancing properties, presents a promising therapeutic approach for POD prevention. The objective of this trial was to investigate whether prophylactic intraoperative sodium oxybate reduces POD incidence in older patients (≥ 65 years) undergoing major orthopedic surgery.

Methods

This randomized, double-blind, placebo-controlled trial enrolled 332 older patients undergoing elective spine and joint replacement surgery. Participants received either sodium oxybate (30 mg kg⁻¹) or saline after anesthetic induction. Stratified randomization allocated equal numbers to morning and afternoon surgery groups. The primary outcome was POD incidence within seven postoperative days, assessed using the Confusion Assessment Method (CAM).

Results

POD incidence showed no significant difference between groups in unstratified population (10.3% vs. 13.5%, P = 0.372). However, subgroup analysis revealed protective effects in morning surgery patients (7.3% vs. 18.5%, relative risk (RR) = 0.395, 95% confidence intervals (CI) = 0.161–0.968, P = 0.033), while no effect was observed in the afternoon surgery group (13.3% vs. 8.5%, P = 0.318). Among patients with delirium, no significant differences were observed in delirium severity, onset timing, delirium duration, or subtype distribution after false discovery rate (FDR) correction. No significant differences were found in sleep quality, maximal pain score, or safety parameters between groups after FDR correction.

Conclusions

Intraoperative sodium oxybate demonstrates possible time-specific efficacy, significantly reducing POD incidence exclusively in older patients undergoing morning orthopedic surgery, while demonstrating an acceptable safety profile with no significant adverse effects on anesthesia recovery or hemodynamic parameters, suggesting a potential chronotherapeutic approach to POD prevention.

Trial registration

Chinese Clinical Trial Registry, ChiCTR2300078594. Registered on 2023–12-13.

Graphical Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12916-025-04590-1.

Key points

• Sodium oxybate reduced postoperative delirium risk when administered during morning orthopedic surgeries in older patients but showed no beneficial effect during afternoon procedures.

• Among patients who developed postoperative delirium, no significant differences were observed in delirium severity, onset timing, delirium duration, or subtype distribution after FDR correction

Supplementary Information

The online version contains supplementary material available at 10.1186/s12916-025-04590-1.

Background

Postoperative delirium (POD) represents a significant challenge in perioperative care, particularly among older surgical patients [1]. While current approaches emphasize multi-component interventions, their implementation often faces practical challenges in routine clinical care [2, 3]. Furthermore, the heterogeneous nature of POD manifestations complicates the standardization of preventive protocols across diverse surgical populations [3]. Understanding the underlying pathophysiological mechanisms of POD could inform more targeted and feasible preventive strategies.

The pathophysiology of POD involves multiple contributing factors, including inflammatory responses, neurotransmitter imbalances, and disrupted sleep–wake cycles [4, 5]. Among these, sleep disturbance has emerged as a particularly significant modifiable risk factor, especially in older populations. Sleep fragmentation and insomnia not only increase delirium risk but may also exacerbate cognitive vulnerability in the perioperative period [6, 7]. Despite this recognized association between sleep disruption and delirium, pharmacological interventions targeting sleep architecture for POD prevention remain understudied, representing a crucial gap in current preventive strategies.

Sodium oxybate (SO), a partial GABA_B receptor agonist with established hypnotic properties [8], presents a promising therapeutic approach for POD prevention. Approved by the US Food and Drug Administration since 2002 for treating a series of sleep symptoms [9], SO demonstrates unique effects on sleep architecture that closely mimic physiological sleep patterns [10]. By enhancing the effect of gamma-aminobutyric acid, the brain's major inhibitory neurotransmitter, it typically increases slow-wave sleep (SWS, characterized by delta waves) and promotes non-rapid eye movement (NREM) sleep. Slow-wave sleep was identified as a protective factor against postoperative delirium [11]. Therefore, we hypothesize that sodium oxybate may potentially reduce the incidence of postoperative delirium by augmenting perioperative slow-wave sleep. Besides, our preliminary research has shown that intraoperative administration of SO (30 mg kg⁻¹) improved postoperative sleep quality¹². These findings, combined with the established relationship between sleep disruption and delirium, suggest potential therapeutic benefits in POD prevention, particularly given the drug's favorable safety profile and well-characterized pharmacokinetics.

The present study investigates whether prophylactic intraoperative administration of subanesthetic-dose SO reduces POD incidence in older patients (≥ 65 years) undergoing major orthopedic surgery. This study aims to provide evidence for a novel preventive strategy that could be readily integrated into existing perioperative care protocols.

Methods

Study design and patients

This randomized, double-blind, placebo-controlled trial was conducted at Shengjing Hospital of China Medical University. The study protocol was approved by the Medical Ethics Committees of Shengjing Hospital and registered in the Chinese Clinical Trial Registry (ChiCTR2300078594). Written informed consent was obtained from all participants before enrollment. We followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline.

Patient selection

We enrolled patients aged ≥ 65 years undergoing elective spine and joint replacement surgery under general anesthesia with an American Society of Anesthesiologist (ASA) class ≤ III and expected surgery duration ≥ 2 h. Exclusion criteria includes severe visual or hearing impairment; inability to communicate; Mini-Mental State Examination (MMSE) score < 20; preoperative history of psychotic disorders; cardiovascular conditions including left ventricular ejection fraction less than 30%, sick sinus syndrome, severe sinus bradycardia (heart rates < 50 beats per minute), or atrioventricular block ≥ second degree; severe hepatic dysfunction (≥ Child–Pugh class C); severe renal dysfunction; anticipated postoperative intensive care unit admission; and allergic reactions or contraindications to SO.

Randomization and blinding

Participants were randomly allocated in a 1:1 ratio using stratified block randomization, with stratification by surgery timing (morning: before 12:00; afternoon: at or after 12:00). The randomization sequence was computer-generated by an independent statistician and implemented through sequentially numbered, opaque, sealed envelopes 30–60 min before surgery. Double-blinding was maintained throughout the study period, with all research personnel, surgical team members, outcome assessors, and data collectors blinded to treatment allocation until completion of statistical analysis.

Baseline data collection

Comprehensive baseline data including demographics, medical histories (cardiovascular diseases, respiratory disorders, central nervous system pathologies, diabetes), cognitive function (MMSE), psychological status (Hospital Anxiety and Depression Scale, HADS) with anxiety (HADS-A) and depression (HADS-D) subscales, and sleep quality (Athens Insomnia Scale, AIS) were collected preoperatively.

Intervention procedure

Based on preliminary dose-finding studies, the intervention group received sodium oxybate (30 mg kg⁻¹, Xi-an Hanfeng Pharmaceutical Co., Ltd.) diluted to 100 mg mL⁻¹ with saline solution, a dose that produced sleep and cognitive improvements without severe central nervous system depression, while the control group received an equal volume of saline, both administered at 60 mg kg⁻¹ h⁻¹ after anesthetic induction. Study medications were prepared in identical 20 mL syringes by an independent researcher.

General anesthesia was induced with propofol (2 mg kg⁻¹), sufentanil (0.3–0.5 μg kg⁻¹), and cis-atracurium (0.15–0.2 mg kg⁻¹). Anesthesia was maintained with 1% sevoflurane in a 50% nitrous oxide and 50% oxygen mixture (2 L min⁻¹) and remifentanil (0.02–0.2 μg kg⁻¹ min⁻¹). Cardiovascular parameters were maintained within 20% of baseline by adjusting sevoflurane concentration and remifentanil dosage while maintaining end-tidal minimal alveolar concentration (MAC) of 1.0–1.3. Postoperative analgesia was managed using patient-controlled intravenous analgesia (PCIA), administered using sufentanil 2 μg kg⁻¹ in 100 mL of 0.9% saline (background infusion 2 mL h⁻¹, bolus 0.5 mL, lockout time 15 min). Loxoprofen sodium (60 mg) was used as rescue analgesic for pain scores ≥ 4.

Outcome measures

The primary outcome was the occurrence of POD within 7 days post-operatively, assessed twice daily using the Confusion Assessment Method (CAM). A positive CAM assessment at any point during this period constituted an occurrence of POD. Secondary outcomes included delirium severity measured by the Delirium Rating Scale (DRS)−98, time to POD onset (days), duration of POD (days), maximum pain scores within seven postoperative days (numerical rating scale, pain-NRS, 0–10), maximum sleep disturbance scores (numerical rating scale, sleep disturbance-NRS, 0–10). Pain was assessed at 3 h, 6 h, 24 h, and day 2–7 once daily at 8:00 am post-surgery, while subjective sleep quality was evaluated at 8:00 am on postoperative days 1–7.

Safety indicators

Anesthesia recovery time was measured in minutes from discontinuation of anesthetic agents to achievement of a modified OAA/S (Observer’s Assessment of Alertness/Sedation Scale) score of 5. Post-Anesthesia Care Unit (PACU) length of stay was documented from admission to discharge. Hemodynamic monitoring included the assessment of bradycardic events, defined as heart rate < 50 beats per minute during both intraoperative and postoperative phases. Respiratory function was monitored through continuous pulse oximetry, with postoperative oxygen desaturation episodes recorded when SpO₂ fell below 90% for > 30 s. All safety parameters were documented by trained personnel blinded to group allocation according to institutional protocols.

Statistical analysis

Sample size calculation was based on an assumed baseline POD incidence of 15%, derived from institutional data demonstrating approximately 15.0% POD incidence among older patients (≥ 65 years) undergoing major orthopedic surgery at the study institution. The hypothesis was that SO would achieve an absolute risk reduction of 10 percentage points (from 15 to 5%). With α = 0.05 and 80% power, 302 patients (166 for each group) were required, increased to 332 to account for 10% attrition. Our prespecified primary analysis was based on the intention-to-treat principle, without accommodations for protocol deviations, treatment nonadherence, or incomplete intervention implementation.

Continuous variables were analyzed using Student’s t-test or Mann–Whitney U test, while categorical variables were compared using chi-square or Fisher’s exact tests where appropriate. Risk ratios with 95% confidence intervals were calculated using log-binomial models. Secondary outcome analyses were considered exploratory due to multiple comparisons. Multiple imputation method was used to handle missing data. A prespecified subgroup analysis was conducted based on stratification by procedural timing (morning versus afternoon subgroups). The primary outcome measure (incidence of POD) was evaluated using Pearson’s chi-square test. Secondary outcomes and safety indicators were similarly subjected to between-group comparative analyses. Statistical significance was defined at a threshold of P < 0.05 using two-sided tests. For exploratory subgroup and secondary analyses involving multiple comparisons, we performed false discovery rate (FDR) correction using the Benjamini–Hochberg procedure. The FDR level was set at 0.05, and both unadjusted and FDR-adjusted P-values are reported. Analyses were performed using SPSS version 27.0 (IBM, Armonk, Corp., NY, USA).

Results

Study population

From December 2023 to April 2025, 415 patients were screened, of whom 20 had MMSE scores < 20, 5 had contraindications for sodium oxybate (severe bradycardia), 27 declined to participate, and 31 had their operations canceled or transferred to regional anesthesia. A total of 332 participants (166 in the SO group and 166 in the control group) were enrolled in the study. One patient in the SO group and three patients in the control group were lost to follow-up; thus, 328 participants (165 patients in the SO group and 163 patients in the control group) were included in the final analysis (Fig. 1). The participants had a median (inter-quartile range (IQR) age of 68 (65–71) years, a median (IQR) body mass index(BMI) of 24.7 (22.6–26.8), a median (IQR) MMSE score of 26 (24–28), a median (IQR) HADS-A score of 4 (2–7), a median (IQR) HADS-D score of 3 (1–6), and a median (IQR) AIS score of 4(1–8). 134 (40.9%) were male and 194 (59.1%) were female. Two hundred nine (63.7%) of the patients received spine surgery, 93 (28.4%) received joint replacement surgery, while 26 (7.9%) received other types of orthopedic surgeries. One hundred sixty-seven (50.9%) of the participants had cardiovascular disease, 12 (3.7%) had respiratory system disease, 25 (7.6%) had central nervous system disease, and 46(14.0%) had diabetes. Baseline characteristics were well-balanced between the two study groups, including similar distributions of surgical procedure types (open vs. minimally invasive spine surgery, hip vs. knee replacement) and PCIA use (Table 1). Surgical duration was comparable between the SO and control groups overall (median (IQR): 110 (90–135) vs. 114 (94–140) min, P = 0.639) and within both morning (P = 0.146) and afternoon (P = 0.554) surgery subgroups, indicating similar surgical complexity across treatment arms and timing strata. Additionally, surgical duration was similar between morning and afternoon surgeries (P = 0.357), suggesting comparable case complexity across timing periods (Additional file 1: Supplementary Table 1).

Fig. 1.

CONSORT flow diagram representing patient recruitment and group allocation. Of 415 patients screened, 332 were randomized equally to sodium oxybate (n = 165) or control (n = 163) groups, with complete follow-up and analysis for all participants

Table 1.

Baseline demographic, clinical, and psychometric characteristics

	SO (N = 165)	Con (N = 163)	Overall (N = 328)
Age, median (IQR), year	68 (65–71)	68 (65–72)	68 (65–71)
BMI, median (IQR)	24.8 (22.7–26.6)	24.6 (22.3–27.1)	24.7 (22.6–26.8)
Gender, male, No. (%)	67 (40.6%)	67 (41.1%)	134 (40.9%)
MMSE, median (IQR)	26 (24–28)	26 (25–28)	26 (24–28)
HADS-A, median (IQR)	4 (1.5–7)	4 (2–7)	4 (2–7)
HADS-D, median (IQR)	3 (1–7)	3 (1–6)	3 (1–6)
AIS, median (IQR)	4 (0–8)	4 (1–8)	4 (1–8)
Surgery type
Spine surgery, No. (%)
Open spine surgery	71 (43.0%)	69 (42.3%)	140 (42.7%)
Minimally invasive endoscopic surgery	35 (21.2%)	34 (20.9%)	69 (21.0%)
Joint replacement, No. (%)
Hip replacement	17 (10.3%)	23 (14.1%)	40 (12.2%)
Knee replacement	27 (16.4%)	26 (16.0%)	53 (16.2%)
Others, No. (%)	15 (9.1%)	11 (6.7%)	26 (7.9%)
Co-morbidities
CVS disease, No. (%)	87 (52.7%)	80 (49.1%)	167 (50.9%)
Respiratory system disease, No. (%)	7 (4.2%)	5 (3.1%)	12 (3.7%)
CNS disease, No. (%)	11 (6.7%)	14 (8.6%)	25 (7.6%)
Diabetes, No. (%)	21 (12.7%)	25 (15.3%)	46 (14.0%)
Surgery duration, median (IQR), minute	110 (90–135)	114 (94–140)	111 (90–138)
PCIA use, No. (%)	104 (63.0%)	100 (61.3%)	204 (62.2%)

AIS Athens Insomnia Scale, BMI body mass index, CVD cardiovascular disease, CNS central nervous system, HADS-A Hospital Anxiety and Depression Scale—Anxiety subscale, HADS-D Hospital Anxiety and Depression Scale—Depression subscale, MMSE Mini-Mental State Examination, NRS numerical rating scale, PCIA patient-controlled intravenous analgesia. *P < 0.05

Primary outcome

There was no significant difference in POD incidence within postoperative 7 days between the SO and control groups (SO: 17/165 (10.3%) vs. Con: 22/163 (13.5%), risk ratio (RR) = 0.763, 95% CI = 0.421 to 1.384, P = 0.372). Subgroup analysis revealed a significant prophylactic effect of SO against POD in the morning surgery population, with an incidence of 7.3% (6/82) in the SO group compared to 18.5% (15/81) in the control group (RR = 0.395, 95% CI = 0.161 to 0.968, P = 0.033). However, this protective effect was not observed in the afternoon surgery subgroup, where differences failed to reach statistical significance (SO: 11/83 (10.3%) vs. Con: 7/82 (8.5%), risk ratio (RR) = 1.553, 95% CI = 0.633 to 3.808, P = 0.318) (Fig. 2).

Fig. 2.

Forest plot showing risk ratios (RR) for postoperative delirium. No significant difference was found in delirium incidence between the SO and control groups in unstratified population (SO: 17/165 (10.3%) vs. Con: 22/163 (13.5%), risk ratio (RR) = 0.763, 95% CI = 0.421 to 1.384, P = 0.372). SO reduced POD incidence in morning surgery subgroup (6/82 (7.3%) vs. 15/81 (18.5%), RR = 0.395, 95% CI = 0.161 to 0.968, P = 0.033). SO did not affect POD incidence in afternoon surgery subgroup (SO:11/83 (10.3%) vs. Con: 7/82 (8.5%), risk ratio (RR) = 1.553, 95% CI = 0.633 to 3.808, P = 0.318)

Secondary outcomes

Exploratory subgroup analyses by surgery timing showed that among patients who developed POD (N = 39), the maximal DRS-98 scores were comparable between the SO and control groups (unadjusted P = 0.243, FDR-adjusted P = 1.000). No significant between-group differences in maximal DRS-98 scores within either the morning surgery subgroup (unadjusted P = 0.604, FDR-adjusted P = 1.000) or the afternoon surgery subgroup were seen (unadjusted P = 0.876, FDR-adjusted P = 1.000) (Table 2).

Table 2.

secondary outcomes: maximal delirium severity rating scale scores, delirium onset day, duration of delirium, and delirium subtype

	SO	Control	P	FDR Corrected P	Hodges-Lehmann estimator	95% CI
DRSmax, median, IQR
Total (N = 17 vs. 22)	11 (6–17)	13 (9–18)	0.243	1.000	− 2	− 6, 3
Morning subgroup (N = 6 vs. 15)	13 (8–19)	13 (9–21)	0.604	1.000	− 2	− 16, 5
Afternoon subgroup (N = 11 vs. 7)	7 (6–17)	11 (7–13)	0.876	1.000	− 1	− 5, 8
Time to delirium onset (days), median, IQR
Total (N = 17 vs. 22)	2 (1–3)	2 (1–3)	0.693	1.000	0	− 1, 0
Morning subgroup (N = 6 vs. 15)	1 (1–3)	2 (1–3)	0.553	1.000	0	− 2, 0
Afternoon subgroup (N = 11 vs. 7)	2 (1–4)	2 (1–3)	1.000	1.000	0	− 1, 1
Delirium duration (days), median, IQR
Total (N = 17 vs. 22)	2 (1–4)	2 (2–3)	0.326	1.000	0	− 1, 0
Morning subgroup (N = 6 vs. 15)	1 (1–2)	3 (2–3)	0.025*	0.300	− 1	− 2, 0
Afternoon subgroup (N = 11 vs. 7)	2 (1–3)	2 (1–4)	0.860	1.000	0	− 2, 1
Delirium subtype (hypoactive), %
Total (N = 17 vs. 22)	5 (29.4%)	8 (36.4%)	0.648	1.000	N/A	N/A
Morning subgroup (N = 6 vs. 15)	2 (33.3%)	6 (40.0%)	1.000#	1.000	N/A	N/A
Afternoon subgroup (N = 11 vs. 7)	3 (27.3%)	2 (28.6%)	1.000#	1.000	N/A	N/A

^*P < 0.05; ^#Fisher’s exact test

Time to delirium onset showed no significant between-group differences (unadjusted P = 0.693, FDR-adjusted P = 1.000). No significant differences were seen in the morning subgroup (unadjusted P = 0.553, FDR-adjusted P = 1.000) and afternoon subgroup (unadjusted P = 1.000, FDR-adjusted P = 1.000) regarding time to delirium onset (Table 2).

Analysis of delirium duration revealed no statistically significant difference in the unstratified study population (unadjusted P = 0.326, FDR-adjusted P = 1.000) or in the afternoon surgery subgroup (unadjusted P = 0.860, FDR-adjusted P = 1.000). It is worth noting that, within the morning surgery subgroup, SO administration reduced delirium duration (median (IQR): SO: 1 (1–2) days vs. control: 3 (2–3) days; Hodges-Lehmann estimator = − 1, 95% CI = − 2 to 0, unadjusted P = 0.025). However, this finding did not survive correction for multiple comparisons (FDR-adjusted P = 0.300) (Table 2).

The distribution of POD subtypes demonstrated no statistically significant differences between SO and control groups in the total study population (unadjusted P = 0.648, FDR-adjusted P = 1.000). Similarly, subgroup analyses revealed no significant differences in POD subtype distribution in either morning (unadjusted P = 1.000, FDR-adjusted P = 1.000) or afternoon (unadjusted P = 1.000, FDR-adjusted P = 1.000) surgical population (Table 2).

In the aggregate sample (N = 328), maximal pain NRS scores showed comparable values between SO and control groups in the unstratified population (unadjusted P = 0.051, FDR-adjusted P = 0.183), morning surgery subgroup (unadjusted P = 0.385, FDR-adjusted P = 0.463), and afternoon surgery subgroup (P = 0.061, FDR-adjusted P = 0.183). Among patients who developed postoperative delirium (N = 39), SO-treated patients reported numerically higher pain intensity compared with controls (median (IQR): 7 (4–8) vs. 4 (3–6); unadjusted P = 0.010; FDR-adjusted P = 0.122). This difference was not significant after correction for multiple comparisons, nor was it observed in morning (unadjusted P = 0.207, FDR-adjusted P = 0.414) or afternoon (unadjusted P = 0.151, FDR-adjusted P = 0.362) surgery subgroups (Table 3).

Table 3.

Secondary outcomes: maximal pain numerical rating scale scores, worst sleep numerical rating scale scores

	SO	Control	P	FDR Corrected P	Hodges-Lehmann estimator	95% CI
Whole population
Pain NRSmax, median, IQR
Total (N = 165 vs. 163)	4 (2–7)	4 (2–6)	0.051	0.183	1	0, 1
Morning subgroup (N = 82 vs. 81)	4 (2–6)	4 (2–5)	0.385	0.463	0	0, 1
Afternoon subgroup (N = 83 vs. 82)	5 (2–7)	4 (1–6)	0.061	0.183	1	0, 2
Sleep NRSworst, median, IQR
Total (N = 165 vs. 163)	3 (2–5)	4 (1–6)	0.386	0.463	0	− 1, 0
Morning subgroup (N = 82 vs. 81)	4 (2–6)	4 (1–6)	0.661	0.661	0	− 1, 1
Afternoon subgroup (N = 83 vs. 82)	2 (1–5)	3 (1–5)	0.308	0.463	0	− 1, 0
Population with delirium
Pain NRSmax, median, IQR
Total (N = 17 vs. 22)	7 (4–8)	4 (3–6)	0.010*	0.122	3	0, 4
Morning subgroup (N = 6 vs. 15)	5 (3–8)	4 (2–4)	0.207	0.414	1	− 1, 4
Afternoon subgroup (N = 11 vs. 7)	7 (4–8)	5 (3–7)	0.151	0.362	2	− 1, 4
Sleep NRSworst, median, IQR
Total (N = 17 vs. 22)	5 (3–6)	4 (2–6)	0.506	0.552	− 1	− 2, 2
Morning subgroup (N = 6 vs. 15)	6 (4–7)	4 (2–5)	0.059	0.183	− 2	− 4, 0
Afternoon subgroup (N = 11 vs. 7)	5 (2–8)	3 (0–6)	0.357	0.463	2	− 2, 5

^*P < 0.05

Subjective sleep quality evaluation revealed no statistically significant differences in maximal sleep disturbance scores between SO and control groups across multiple analyses. In the aggregated patient population, no differences were observed in the unstratified analysis (unadjusted P = 0.386, FDR-adjusted P = 0.463), morning surgery subgroup (unadjusted P = 0.661, FDR-adjusted P = 0.661), or afternoon surgery subgroup (unadjusted P = 0.308, FDR-adjusted P = 0.463). Similarly, when the analysis was restricted to patients who developed POD, sleep disturbance NRS scores remained comparable between groups in the unstratified population (unadjusted P = 0.506, FDR-adjusted P = 0.552), morning subgroup (unadjusted P = 0.059, FDR-adjusted P = 0.183), and afternoon subgroup (unadjusted P = 0.357, FDR-adjusted P = 0.463) (Table 3).

Safety indicators

Safety profile analysis revealed no significant differences between SO and control groups across all safety parameters. Anesthesia recovery time demonstrated equivalence in the unstratified population (P = 0.935), morning subgroup (P = 0.822), and afternoon subgroup (P = 0.829). Similarly, PACU length of stay exhibited comparable durations in the unstratified population (P = 0.925), morning subgroup (P = 0.843), and afternoon subgroup (P = 0.960). Incidence of perioperative bradycardia showed no significant differences in the unstratified population (P = 0.285), morning subgroup (P = 0.360), and afternoon subgroup (P = 1.000). Postoperative oxygen desaturation events also showed comparable incidence in the unstratified population (P = 0.361), morning subgroup (P = 0.210), and afternoon subgroup (P = 0.983). (Additional file 2: Supplementary Table 2).

Discussion

This research is a double-blind randomized controlled study investigating the preventive effect of intraoperative intravenous infusion of SO on POD in older patients undergoing major orthopedic surgery. The study found that in patients undergoing morning surgery, infusion of 30 mg kg⁻¹ of SO after anesthesia induction significantly reduced the incidence of POD.

Based on the analysis of primary outcomes, we found that in the overall population without stratification by surgical timing, SO did not significantly reduce the risk of POD. However, according to our pre-specified stratified subgroup analysis, a significant reduction in POD risk was observed in the morning subgroup. In contrast, the afternoon subgroup showed no significant effect on POD risk. The reduction in POD incidence from 18.5% to 7.3% with SO administration corresponds to an absolute risk reduction of 11.2% and a relative risk reduction of approximately 61%. Current pharmacological interventions for POD prevention, including dexmedetomidine and melatonin, typically demonstrate risk reductions between 30 and 47% [13–18]. The magnitude of SO’s effect is particularly noteworthy given POD’s complex multifactorial etiology and the historical challenges in achieving substantial risk reduction through single-agent interventions.

Several potential mechanisms may underlie SO’s preventive effect. As a GABA_B receptor agonist, SO might reduce neuroinflammation [19] and oxidative stress [20], processes increasingly recognized as central to delirium pathogenesis [21, 22]. The drug’s known effects on slow-wave sleep architecture [23] could help maintain normal sleep–wake cycles during the critical immediate postoperative period. Additionally, SO’s modulation of neurotransmitter systems might prevent the acute cholinergic deficiency often implicated in POD development [24].

The results further indicate that SO’s preventive effect on POD demonstrates time-specific characteristics, with significant efficacy in morning surgery populations but potentially no effect in the afternoon surgery population. This time-dependent efficacy may derive from the pharmacological properties of the drug itself [25], or because the morning surgery population exhibits inherently higher POD incidence rates, allowing drug intervention to produce significant risk reduction effects. In this study, the POD incidence rate in the morning subgroup control was 18.5%, compared to 8.5% in the afternoon subgroup, suggesting that morning surgery patients may be more susceptible to POD, thereby amplifying SO’s intervention effects. Previous research has identified significant time-specific effects of SO on sleep improvement, with morning administration demonstrating greater enhancement of sleep quality compared to afternoon dosing [12]. Although there is currently no direct evidence confirming the precise mechanism by which SO prevents POD, its time-specific effects on sleep improvement suggest that SO’s pharmacological action may exhibit temporal specificity. The nervous system of morning surgery patients may possess a time-dependent sensitivity window to SO, potentially related to the secretion patterns of endogenous regulatory factors such as cortisol [26] and melatonin [27]. Diurnal changes in metabolic enzyme activity [28] could also modulate drug absorption and receptor binding dynamics. Additionally, environmental factors specific to afternoon surgeries, such as variations in operating room conditions and surgical staff fatigue levels [29], may influence the drug’s effectiveness. Additionally, it is noteworthy that the 95% CI for relative risk in the afternoon surgery subgroup is relatively wide (0.6330–3.8007). This may indicate considerable heterogeneity in SO’s preventive effect on POD within the afternoon population, suggesting that the current sample size may be insufficient to adequately ensure the statistical power of SO’s preventive effects when administered in the afternoon. In contrast, the morning group’s 95% CI is narrower (0.1614 to 0.9676), lending greater credibility to our conclusions regarding the efficacy of morning administration.

Further exploration into whether SO could reduce delirium severity (DRS-98) in patients who already developed POD revealed no significant improvement effects across the total population, morning subgroup, or afternoon subgroup. These results indicate that SO can significantly reduce POD incidence in morning surgery populations, but cannot mitigate delirium severity in patients who develop POD. Possible reasons for these results include: POD occurrence and severity may involve different pathological mechanisms; SO may primarily influence delirium-triggering mechanisms while having limited impact on severity. To explore whether SO affects POD disease progression, the study further compared time to delirium onset and delirium duration between SO and control groups. No significant difference was found in onset timing. In the morning surgery subgroup, SO-treated patients had shorter delirium duration, though this difference did not remain significant after FDR correction for multiple comparisons. Although this finding did not survive correction for multiple comparisons, it may provide a preliminary signal for the potential circadian-related effects of SO on POD that warrants validation in future studies. SO may prevent POD by regulating GABA_B/glutamate balance [30] and improving sleep quality in morning surgery patients, though its effect appears limited once delirium severity is established, possibly due to more complex neurotransmitter dysregulation that extends beyond GABA-mediated mechanisms [31].

Given that hypoactive delirium comprises a significant proportion of geriatric POD cases with substantial adverse outcomes [32], the study further investigated whether SO could reduce hypoactive POD incidence. Results revealed no significant differences in hypoactive delirium proportions between SO and control groups across the total population, morning subgroup, and afternoon subgroup. This may be attributed to older patients’ predisposition to hypoactive delirium potentially relating to age-associated frontal lobe dysfunction [33]. These structural and functional changes may determine delirium presentation patterns that are less susceptible to pharmacological intervention. Additionally, the preservation of natural delirium phenotypes in breakthrough cases suggests a preventive effect of SO against POD occurrence but exhibits no modulatory effect on phenotypic manifestations in populations with delirium. The mechanism may involve SO’s action on preventive neural pathways and inflammatory cascades during the pre-delirium phase, reducing initial triggers for delirium onset. However, once the threshold for delirium is breached, the phenotypic expression appears to follow predetermined pathophysiological patterns of POD. Similarly, current empirical evidence indicates that prophylactic interventions other than SO implemented prior to POD onset can reduce incidence by approximately 30–40%; however, therapeutic interventions initiated after POD manifestation demonstrate limited efficacy in attenuating its severity [34–36].

To investigate whether SO’s preventive effect on POD and its ability to shorten disease course originate from improvements in postoperative sleep and pain, the study compared maximum postoperative pain scores and worst subjective sleep quality scores between the SO group and control group. Results revealed that in the overall population, without distinguishing between patients with and without delirium, SO had no significant effect on maximum postoperative pain scores or worst sleep quality scores. However, in the patients with delirium, SO increased maximum postoperative pain scores, though this difference did not remain significant after FDR correction. In the patients with delirium, SO did not affect sleep quality scores. These results indicate that SO’s preventive effect on POD is not mediated by improving sleep quality or alleviating pain intensity, but may originate from other mechanisms such as regulating GABAergic system function, modulating neuroinflammation [26], or improving functional connectivity between specific brain regions [37]. SO’s effect on these systems may not be reflected in sleep scores. Surprisingly, in patients with delirium, SO was associated with potentially increased pain scores, which may indicate that SO might regulate certain neurotransmitter systems [38, 39], inversely affecting pain signal processing [40, 41], especially in patients already experiencing neural dysfunction; or this enhanced pain perception might be a side effect of preventive and reparative mechanisms of cognitive function [42]. However, given that this finding did not survive FDR correction for multiple comparisons and the relatively small sample size (N = 39 patients with delirium), this observation should be interpreted with considerable caution and requires validation in larger studies before definitive conclusions can be drawn.

Due to SO’s relatively long duration of action [43], there were concerns about possible delayed awakening and prolonged PACU stay. Although the drug itself does not affect spontaneous breathing, it may reduce respiratory rate [44], and considering potential synergistic effects with other general anesthetics, it could possibly lead to decreased postoperative oxygen saturation. In addition, SO also exhibits the adverse effect of inducing bradycardia [44].Therefore, the study further analyzed safety indicators, including anesthesia recovery time, PACU length of stay, postoperative oxygen desaturation incidence, and perioperative bradycardia incidence. Results showed no significant between-group differences in any of these parameters.

This study has certain limitations: First, its single-center nature may raise questions about the generalizability of the research conclusions. Second, this study did not address long-term cognitive dysfunction. Third, the secondary analyses of POD characteristics were limited by small sample sizes, particularly in subgroup analyses, resulting in insufficient statistical power to draw definitive conclusions. These exploratory findings should therefore be interpreted as hypothesis-generating and require validation in larger, adequately powered studies. Fourth, this study did not conduct pharmacokinetic or pharmacodynamic measurements, which limited the ability to provide more in-depth information for mechanistic exploration. Future research should consider conducting multicenter randomized controlled trials, and investigate effects across different surgical population types to clarify the efficacy of SO. Additionally, adequately powered studies specifically designed to examine delirium characteristics, including potential interactions with circadian timing, are needed to validate the exploratory findings from this study and elucidate the specific features of SO’s effects on POD. Pharmacokinetic and pharmacodynamic studies should be incorporated to further explore underlying mechanisms.

Conclusions

This study demonstrated that intraoperative intravenous infusion of sodium oxybate reduced the risk of POD in patients undergoing morning surgery. With an acceptable safety profile, sodium oxybate shows potential as an effective pharmacological preventive measure for POD. However, it is important to recognize the preliminary nature of these findings. Larger, adequately powered trials specifically designed to assess POD outcomes in morning surgery patients are required to confirm these results.

Abbreviations

POD

Postoperative delirium

GABA_B

Gamma-aminobutyric acid B

SO

Sodium oxybate

CAM

Confusion Assessment Method

RR

Relative risk

CI

Confidence interval

FDR

False discovery rate

SWS

Slow-wave sleep

NREM

Non-rapid eye movement

CONSORT

Consolidated Standards of Reporting Trials

ASA

American Society of Anesthesiologists

MMSE

Mini-Mental State Examination

HADS-A

Hospital Anxiety and Depression Scale-Anxiety

HADS-D

Hospital Anxiety and Depression Scale-Depression

AIS

Athens Insomnia Scale

MAC

Minimal alveolar concentration

PCIA

Patient-controlled intravenous analgesia

DRS-98

Delirium Rating Scale-98

NRS

Numerical rating scale

OAA/S

Observer’s Assessment of Alertness/Sedation

PACU

Post-anesthesia care unit

IQR

Interquartile range

BMI

Body mass index

Authors’ contributions

Study conception and design: MC, HX, PZ. Patient recruitment and data collection: FW, JZ, XY, SW. Study conduct and protocol implementation: MC, HX, TX, LZ. Data analysis and interpretation: MC, HX, FW, JZ. Statistical analysis: XY, SW. Manuscript drafting: MC, HX. Critical revision of manuscript: TX, LZ, PZ. Study supervision: PZ. MC (Meiying Cui) and HX (Hang Xue) contributed equally to this work as co-first authors. PZ (Ping Zhao) was responsible for the overall study conduct and final manuscript approval. All authors reviewed and approved the final manuscript version and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Funding

This work was supported by the Outstanding Scientific Fund of Shengjing Hospital No.202208 to Ping Zhao.

Data availability

Yes. Data types: Deidentified participant data. Additional Information: On request. How to access data: On request: zhaoping_sj@163.com. When available: With publication. Who can access the data: researchers whose proposed use of the data has been approved.

Declarations

Ethics approval and consent to participate

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The study was approved by the Medical Ethics Committees of Shengjing Hospital (No. 2023PS1183K).

Written informed consent was obtained from all participants before enrollment.

Consent for publication

Not applicable. This study reports aggregate data only. No individual participant data, images, videos, or other materials requiring consent for publication are included.

Competing interests

The authors declare no competing interests.