EEG-Derived Pain Threshold Index-Guided Remifentanil Titration During Laparoscopic Abdominal Surgery: A Randomized Trial

Abstract

Objective

To determine whether remifentanil titration guided by the pain threshold index (PTi), an electroencephalography-derived proxy for nociceptive tolerance rather than subjective pain, reduces intraoperative remifentanil infusion during laparoscopic abdominal surgery.

Methods

In this prospective, single-center, assessor-blinded randomized trial, adults undergoing elective laparoscopic abdominal surgery with propofol-remifentanil target-controlled infusion were assigned 1:1 to PTi-guided or conventional hemodynamic-guided remifentanil titration. The primary endpoint was intraoperative remifentanil infusion rate. The principal analysis used the full-analysis set and covariate-adjusted linear regression with heteroskedasticity-consistent robust standard errors.

Results

Among 200 randomized patients, 195 were analyzed (98 PTi-guided; 97 control). The primary endpoint was neutral: mean (SD) remifentanil infusion rate was 7.44 (1.43) versus 7.63 (1.41) µg kg⁻¹ h⁻¹; adjusted mean difference, −0.06 µg kg⁻¹ h⁻¹ (95% CI, −0.44 to 0.32; P=0.758). Per-protocol results were similar. Median (IQR) NRS at post-anesthesia care unit (PACU) admission was 2.0 (2.0–3.0) versus 3.0 (2.0–4.0), and Ramsay score was 2.0 (2.0–2.0) in both groups. In exploratory secondary analyses, rescue opioid was used in the PACU by 18/98 (18.4%) versus 31/97 (32.0%) patients and any rescue opioid through 48 h by 29/98 (29.6%) versus 44/97 (45.4%). Any postoperative nausea and vomiting (PONV) through 48 h occurred in 17/98 (17.3%) versus 29/97 (29.9%) patients. Total postoperative opioid exposure through 48 h was not reduced.

Conclusion

PTi-guided remifentanil titration did not reduce intraoperative remifentanil infusion. PACU pain and sedation scores were not materially different. Lower early rescue-opioid use and fewer PONV events through 48 h were exploratory secondary findings, were not accompanied by lower postoperative opioid dose, and require prospective confirmation.

Introduction

Adequate antinociception is a core component of general anesthesia, yet nociception cannot be reported directly by an unconscious patient. Opioid titration is therefore commonly guided by arterial pressure, heart rate, surgical timing, and clinician judgment, but these signals are nonspecific and are influenced by intravascular volume, vasoactive drugs, autonomic reserve, and hypnotic depth.^1–4

Objective nociception monitors were developed to reduce subjectivity in intraoperative opioid administration. Most systems use autonomic, photoplethysmographic, pupillary, or multiparameter signals, whereas the pain threshold index (PTi) is derived from frontal electroencephalography and is intended to estimate nociceptive tolerance during general anesthesia rather than measure subjective pain directly.^5–7 Trials of nociception-guided opioid titration have reported heterogeneous effects, ranging from lower remifentanil or fentanyl exposure to neutral perioperative pain outcomes.^8–13 Recent evidence synthesis suggests that effects differ by monitor technology, dosing algorithm, anesthetic regimen, and endpoint definition.^14–17

The premise of PTi is that EEG responses to noxious stimulation may provide an indirect estimate of nociceptive tolerance. Published PTi descriptions use two-channel frontal EEG wavelet analysis to combine multiscale waveform changes, delta, theta, alpha, and beta components, and spectral frequency features into a proprietary 0–100 score.¹⁸ WLi is also EEG-derived but is intended to reflect hypnotic depth. In a pediatric observational study, PTi predicted hemodynamic responses to intubation and incision better than WLi, suggesting potentially complementary information rather than established physiological independence.¹⁸

The PTi target of 40–60 follows the developer-recommended range reported in published work and has been used in a laparoscopic colectomy trial.^18,19 Observational cutoffs near 60–62 support further study of the upper trigger but do not validate both treatment thresholds in adults.¹⁸ A previous trial used 40–65.⁷ Accordingly, 40 and 60 were operational titration triggers in this trial, not diagnostic boundaries for pain or opioid excess; a value below 40 alone does not establish excessive antinociception.

Remifentanil’s rapid elimination provides a pharmacological basis for adjusting its infusion to changing surgical stimulation.²⁰ However, minimizing the dose is not the sole objective: adequate antinociception must be balanced against unnecessary opioid exposure. In a randomized abdominal-surgery trial, a higher remifentanil dose increased periincisional hyperalgesia and postoperative morphine requirements.²¹ In another randomized trial, Nociception Level-guided titration during propofol-remifentanil anesthesia reduced remifentanil administration and hypotensive events.⁹ These results support testing a feedback-based approach to dosing, but do not establish that the autonomic Nociception Level index and EEG-derived PTi are interchangeable.

There is also a rationale for studying brain-derived feedback specifically. A placebo-controlled trial using functional near-infrared spectroscopy found that remifentanil attenuated prefrontal responses to cardiac ablation in pediatric patients under general anesthesia; this supports the concept of assessing central antinociceptive effects, not the validity of PTi itself.²² More directly, a PTi-guided trial in gynecologic surgery reported less frequent use of PACU rescue analgesia despite similar intraoperative remifentanil consumption.⁷ Thus, PTi-guided remifentanil titration warrants evaluation for its potential to match drug delivery to changing antinociceptive requirements, with both intraoperative exposure and postoperative analgesic outcomes assessed. Whether this translates to improved recovery in a broader laparoscopic abdominal surgical population remains uncertain.

We therefore reanalyzed this assessor-blinded randomized trial using a CONSORT 2025-aligned full-analysis-set strategy (Supplementary file 1). The primary aim was to determine whether PTi-guided titration reduced intraoperative remifentanil infusion rate compared with conventional hemodynamic-guided titration. Secondary and post hoc analyses examined post-anesthesia care unit (PACU) pain and sedation scores, postoperative analgesia, time-specific Postoperative nausea and vomiting (PONV), hemodynamic and hypnotic trajectories, target attainment, treatment-effect heterogeneity, and potential mediation of rescue-opioid outcomes.

Materials and Methods

Trial Design and Oversight

This was a prospective, single-center, parallel-group, assessor-blinded randomized trial at the First Affiliated Hospital of Zhengzhou University, Zhengzhou, China. The study was performed in line with the principles of the Declaration of Helsinki and its later amendments. The Ethics Committee of Scientific Research and Clinical Trials of The First Affiliated Hospital of Zhengzhou University approved the protocol (reference S2023-K012-002; approval date: August 29, 2023), and all patients provided written informed consent. The trial was prospectively registered in the Chinese Clinical Trial Registry (ChiCTR2300078455; https://www.chictr.org.cn/showproj.html?proj=206473) on December 8, 2023. The registry listed December 12, 2023 as the anticipated recruitment start date required at registration; after registration and the study initiation meeting, actual participant recruitment began on January 1, 2024. No patient was recruited or enrolled between December 12 and December 31, 2023. Reporting follows CONSORT 2025 guidance and the ICH E9 (R1) estimand framework where applicable.^23–25 Patient recruitment ended on October 31, 2024. All participants received scheduled follow-up visits for efficacy evaluation and adverse event monitoring within 7 days after surgery. Patients or public contributors were not involved in trial design, conduct, analysis, or reporting.

The intraoperative anesthesiologist could not be blinded because the intervention required either use or concealment of PTi values. Patients, surgeons, PACU nurses, ward assessors, and the analyst generating the final tables were masked whenever feasible. The term sex refers to biological sex recorded in the medical record and was analyzed as male or female.

Participants, Randomization, and Interventions

Eligible patients were 18–65 years old, had American Society of Anesthesiologists physical status I–III, body-mass index 20–25 kg m⁻², and were scheduled for elective laparoscopic abdominal surgery expected to last 2–6 hours. Key exclusions were planned regional anesthesia, anticipated difficult airway, neurological disease, baseline hemodynamic instability, severe arrhythmia, severe hepatic or renal dysfunction, recent opioid use, substance misuse, uncontrolled psychiatric disease, participation in another trial within 90 days, and severe hematologic abnormality.

Patients were randomly assigned in a 1:1 ratio to receive either PTi-guided analgesia or conventional hemodynamic-guided analgesia. A computer-generated randomization sequence was created with SPSS Statistics version 26.0 (IBM, Armonk, NY, USA), secured by an independent researcher, and concealed from staff responsible for screening, enrollment, and allocation. Treatment assignments were placed in sequentially numbered, opaque, sealed envelopes; each envelope was opened in the operating room by the attending anesthesiologist only after eligibility had been confirmed. PTi readings were concealed for all patients in the control group.

All patients received electrocardiography (ECG), pulse oximetry, invasive arterial pressure, end-tidal carbon dioxide, temperature, and train-of-four monitoring. PTi and the wavelet index (WLi) were acquired using the HXD-I monitor (Beijing Easymonitor Technology Co., Ltd., Beijing, China). EEG electrodes were positioned 2 cm above the glabella, above the bilateral eyebrows, and at the bilateral mastoid processes. PTi was the value generated by the monitor. Published processing includes removal of the direct-current component, wavelet decomposition and reconstruction, spectral analysis, and weighted combination of EEG-derived features; the complete proprietary weighting coefficients are not disclosed.¹⁸ This describes the reported algorithm, not an independent reconstruction or validation in this trial. WLi is an electroencephalography-derived 0–100 index, analogous to the Bispectral Index, used to assess hypnotic depth during general anesthesia. The predefined intraoperative target range was 40–60 for both PTi and WLi. PTi values >60 and <40 were used as protocol triggers to increase and decrease remifentanil, respectively, rather than as diagnostic evidence of insufficient or excessive antinociception; WLi values >60 were interpreted as insufficient hypnotic depth and values <40 as excessive hypnotic depth.

Anesthesia was induced with propofol and remifentanil administered by effect-site target-controlled infusion using the Marsh and Minto models, respectively. Rocuronium was administered after loss of consciousness, and neuromuscular blockade was subsequently titrated using train-of-four monitoring.

During maintenance, propofol was titrated to WLi. The remifentanil effect-site concentration was adjusted in 0.5 ng mL⁻¹ increments or decrements at intervals of at least 5 min within a range of 1.0–4.0 ng mL⁻¹. In the PTi group, remifentanil was increased when PTi remained >60 for more than 60s and decreased when PTi remained <40 for more than 60s. Vasoactive drugs were administered for hemodynamic disturbances (hypertension: mean arterial pressure [MAP] >20% above baseline; hypotension: MAP >20% below baseline; tachycardia: heart rate >90 beats min⁻¹; bradycardia: heart rate <45 beats min⁻¹) when PTi was within the target range. If hemodynamic instability occurred with out-of-range PTi, remifentanil titration to restore PTi to 40–60 was prioritized; vasoactive drugs were administered only if hemodynamic targets remained unmet after two sequential remifentanil adjustments within 10 min, with hemodynamic safety overriding PTi targets. In the control group, remifentanil was titrated according to hemodynamic criteria, and vasoactive drugs were administered only if haemodynamics failed to recover after two remifentanil dose adjustments separated by 10 min.

Approximately 30 min before the anticipated end of surgery, intravenous sufentanil 0.2 µg kg⁻¹ and intravenous propacetamol 2 g were administered. Postoperative analgesia was provided with patient-controlled intravenous hydromorphone without a background infusion (10 mg hydromorphone in 100 mL of normal saline; 5 mL bolus; 15 min lockout). Following extubation, the patient was transferred to the PACU. Rescue analgesia was administered when the Numeric Rating Scale (NRS) pain score was ≥4 at rest. Intravenous sufentanil was used as rescue analgesia in the PACU, with repeat boluses permitted every 10 min until the NRS score was <4. On the ward, intravenous opioids or nonsteroidal anti-inflammatory drugs were used as rescue analgesia when the NRS pain score was ≥4. Rescue opioids were converted to morphine equivalent doses for analysis. A blinded anesthesiologist recorded resting NRS pain scores and Ramsay Sedation Scale scores at PACU admission, as well as cumulative rescue-sufentanil doses administered in the PACU. A separate blinded researcher collected ward outcomes at 24 and 48 h postoperatively, including total hydromorphone consumption and rescue analgesia. PONV scores were recorded at PACU admission, 24 h, 48 h, and postoperative day 7; PONV events were defined as any PONV score >0 at the relevant time point or time window. Intraoperative awareness was assessed at 24 and 48 h by a blinded researcher using the modified Brice questionnaire, a structured interview about memories immediately before and after anesthesia, any recall between these periods, dreaming, and unpleasant perioperative experiences.²⁶

Outcomes and Analysis Populations

The primary endpoint was intraoperative remifentanil infusion rate (µg kg⁻¹ h⁻¹). Secondary endpoints included propofol infusion rate, extubation time, PACU-admission NRS pain score, PACU-admission Ramsay Sedation Scale score, rescue-opioid use in the PACU and ward, hydromorphone consumption, total postoperative opioid exposure through 48 h, nonsteroidal anti-inflammatory drug rescue analgesia, intraoperative hemodynamic events, vasoactive drug use, PONV scores and event outcomes at PACU admission, 24 h, 48 h, and postoperative day 7, intraoperative awareness, intensive care unit (ICU) admission, and hospital length of stay.

MAP, heart rate, PTi, and WLi were recorded at eight protocol-defined time points: T0, 5 min after operating-room entry; T1, before induction; T2, loss of consciousness; T3, 3 min after tracheal intubation; T4, skin incision; T5, 5 min after pneumoperitoneum; T6, end of surgery; and T7, 5 min after extubation. Target attainment from T2 to T6 was defined as WLi 40–60, MAP within 20% of T1, and heart rate 45–90 beats min⁻¹. PTi target attainment was summarized descriptively in the PTi group only as the patient-level proportion of the five sampled T2-T6 values within 40–60, not as time spent in range.

The full-analysis set comprised randomized patients who received the allocated intraoperative intervention and had evaluable primary endpoint data; it was the principal efficacy population. The per-protocol set excluded major protocol deviations and was used for supportive analyses. Safety analyses included patients who received the allocated intervention and had safety data.

Statistical Analysis

All tests were two-sided. Baseline variables are described without significance testing using median (interquartile range [IQR]) or number (%). The primary endpoint was analyzed with linear regression adjusted for treatment, age, sex, body-mass index, American Society of Anesthesiologists physical status (I vs II–III), surgical category (gynecological, gastrointestinal, other), admission MAP, and admission heart rate, using heteroskedasticity-consistent (HC3) standard errors. Sensitivity analyses included unadjusted Welch comparison, Huber robust regression, additional adjustment for surgical duration, additional adjustment for baseline comorbidity and long-term treatment, permutation testing, and repetition in the per-protocol set.

Binary endpoints were analyzed using modified Poisson regression with robust variance to estimate risk ratios and marginal standardization to estimate risk differences.²⁷ PACU NRS and Ramsay scores and other skewed continuous endpoints used median regression; proportional-odds ordinal logistic models were used as sensitivity analyses for PACU score endpoints when model convergence permitted.²⁸ PONV scores were summarized as observed scores and analyzed primarily as events because the distributions were zero-inflated. The composite PONV endpoint through 48 h comprised any event at PACU admission, 24 h, or 48 h. A repeated-measures PONV analysis used generalized estimating equations with a Poisson log-link, exchangeable working correlation, categorical time, treatment, treatment-by-time interaction, and robust sandwich variance. Longitudinal MAP, heart-rate, and WLi trajectories were analyzed using generalized estimating equations with categorical time, treatment, treatment-by-time interaction, and robust sandwich variance.²⁹ Secondary outcomes were exploratory; false-discovery-rate q values are reported for key secondary endpoints.

Post hoc subgroup analyses of the primary endpoint used treatment-by-subgroup interaction terms and were interpreted for consistency rather than for confirmatory precision-medicine inference.^30–32 Exploratory mediation analyses for rescue-opioid outcomes used model-based g-computation and bootstrap intervals under the sequential ignorability assumption.^33–35 These post hoc analyses were hypothesis-generating and could neither establish nor exclude a treatment-effect mechanism. No imputation was performed because the primary and key secondary endpoints had no missing values in the full-analysis set. Detailed statistical analyses are provided in Supplementary method. Analyses were performed using Python 3.13.5 (Python Software Foundation, Wilmington, DE, USA) with statsmodels 0.14.6.

Sample Size

The protocol targeted 200 randomized patients, based on 90% power to detect a standardized difference of 0.5 in intraoperative remifentanil consumption with a two-sided alpha of 0.05 and allowance for attrition. The planning assumption was informed by a 30-patient pilot study of gynecological laparoscopic surgery, which reported mean (SD) remifentanil infusion of 7.63 (1.72) µg kg⁻¹ h⁻¹. The hypothesized 11.1% relative reduction corresponded to an absolute difference of approximately 0.85 µg kg⁻¹ h⁻¹; dividing this difference by 1.72 gives Cohen’s d of approximately 0.49, rounded to 0.5 for planning. The 11.1% reduction was an assumed target, not an observed treatment effect in the present trial.

Results

Participant Flow and Analysis Populations

Among 227 screened patients, 27 were excluded before randomization (15 did not meet inclusion criteria and 12 withdrew consent), and 200 underwent randomization. Surgery was canceled after randomization in two PTi-guided patients and three controls. The full-analysis set included 195 patients (98 PTi-guided; 97 control), the per-protocol set included 186 (94 and 92, respectively), and the safety set included 195 (Figure 1). No full-analysis-set patient had missing primary or key secondary endpoint data.

Figure 1.

CONSORT flow diagram. The diagram shows screening, randomization, allocation, follow-up, and inclusion in the full-analysis, per-protocol, and safety populations.

Baseline and Perioperative Characteristics

Baseline demographic, physiological, laboratory, and planned surgical characteristics are shown in Table 1. The groups were broadly similar on descriptive inspection, although pre-existing comorbidity, hypertension, long-term treatment before surgery, and admission MAP were more common or higher in the pain threshold index-guided group. Post hoc sensitivity analyses additionally adjusting for these imbalanced baseline covariates did not materially change the primary result (Supplementary Table S1).

Table 1.

Baseline and Perioperative Characteristics of Patients Included in the Full-Analysis Set

Characteristic	Overall (n=195)	PTi-Guided (n=98)	Control (n=97)
Baseline and other preoperative characteristics	Baseline and other preoperative characteristics	Baseline and other preoperative characteristics	Baseline and other preoperative characteristics
Age, yr	52.0 (44.0–58.0)	55.0 (44.0–59.0)	50.0 (44.0–56.0)
Sex
Male	58 (29.7)	33 (33.7)	25 (25.8)
Female	137 (70.3)	65 (66.3)	72 (74.2)
Height, cm	163.0 (160.0–169.0)	163.0 (160.0–169.0)	164.0 (159.0–168.0)
Body weight, kg	62.0 (59.0–67.0)	62.2 (59.0–67.5)	62.0 (58.0–66.0)
Body-mass index, kg m−2	23.4 (21.9–24.5)	23.3 (22.0–24.6)	23.4 (21.8–24.4)
ASA physical status
1	34 (17.4)	18 (18.4)	16 (16.5)
2	159 (81.5)	79 (80.6)	80 (82.5)
3	2 (1.0)	1 (1.0)	1 (1.0)
Baseline MAP, mm Hg	92.3 (87.3–100.0)	93.3 (88.2–101.2)	91.7 (87.0–98.0)
Baseline heart rate, beats min−1	77.0 (70.0–84.0)	76.0 (70.0–81.8)	78.0 (71.0–84.0)
Admission temperature, °C	36.3 (36.2–36.5)	36.3 (36.2–36.5)	36.3 (36.2–36.5)
Admission respiratory rate, breaths min−1	18.0 (14.0–20.0)	18.0 (14.5–20.0)	18.0 (14.0–20.0)
Preoperative history	Preoperative history	Preoperative history	Preoperative history
Pre-existing comorbidity	39 (20.0)	28 (28.6)	11 (11.3)
Long-term treatment before surgery	31 (15.9)	22 (22.4)	9 (9.3)
Previous surgery	80 (41.0)	37 (37.8)	43 (44.3)
Clinically significant ECG abnormality	4 (2.1)	4 (4.1)	0 (0.0)
Preoperative laboratory values	Preoperative laboratory values	Preoperative laboratory values	Preoperative laboratory values
Haemoglobin, g L−1	122 (109–131)	123 (110–131)	122 (108–131)
White blood cell count, 109 L−1	5.41 (4.47–6.78)	5.39 (4.44–6.88)	5.46 (4.50–6.61)
Platelet count, 109 L−1	227 (176–282)	222 (172–276)	235 (189–284)
Urea, mmol L−1	4.8 (4.0–5.9)	4.8 (3.9–5.9)	4.8 (4.0–6.0)
Creatinine, μmol L−1	59.0 (52.0–69.0)	59.0 (52.5–69.0)	60.0 (52.0–67.8)
Alanine aminotransferase, U L−1	13.0 (9.0–19.0)	13.0 (9.0–19.0)	13.0 (9.0–18.2)
Aspartate aminotransferase, U L−1	17.0 (14.0–22.0)	17.0 (14.0–23.0)	16.0 (14.0–21.0)
Albumin, g L−1	41.2 (39.4–43.0)	41.5 (39.3–43.1)	40.8 (39.5–42.8)
Surgical and intraoperative characteristics	Surgical and intraoperative characteristics	Surgical and intraoperative characteristics	Surgical and intraoperative characteristics
Type of surgery
Gynecological	101 (51.8)	49 (50.0)	52 (53.6)
Gastrointestinal	80 (41.0)	42 (42.9)	38 (39.2)
Hepatobiliary	6 (3.1)	2 (2.0)	4 (4.1)
Urological	8 (4.1)	5 (5.1)	3 (3.1)
Duration of surgery, min	171.0 (129.5–217.0)	178.0 (138.5–211.0)	163.0 (122.0–224.0)
Duration of anesthesia, min	192.0 (148.0–238.0)	197.5 (151.0–230.0)	185.0 (147.0–243.0)
Intraoperative fluid, mL	1250 (1000–1500)	1300 (1050–1500)	1250 (1000–1500)
Urine output, mL	200 (200–400)	200 (200–400)	200 (200–400)
Estimated blood loss, mL	50 (30–50)	50 (50–50)	50 (20–50)

Notes: Data are median (25th-75th percentile) or number (%), unless otherwise stated. The pain threshold index-guided group received PTi-guided remifentanil titration; the control group received conventional hemodynamic-guided remifentanil titration. Denominators are patients with available data; percentages may not total 100 because of rounding; no imputation was performed. Baseline pre-randomization characteristics are presented without statistical hypothesis testing.

Abbreviations: ASA, American Society of Anesthesiologists; ECG, electrocardiogram; MAP, mean arterial pressure; PTi, pain threshold index.

Primary Outcome

The primary endpoint was neutral. In the full-analysis set, mean (SD) intraoperative remifentanil infusion rate was 7.44 (1.43) µg kg⁻¹ h⁻¹ in the PTi-guided group and 7.63 (1.41) µg kg⁻¹ h⁻¹ in controls. The adjusted mean difference was −0.06 µg kg⁻¹ h⁻¹ (95% CI, −0.44 to 0.32; P=0.758). The per-protocol estimate was similar (−0.03 µg kg⁻¹ h⁻¹; 95% CI, −0.43 to 0.37; P=0.884). Unadjusted, robust-regression, additional-covariate, and permutation analyses all supported a neutral primary finding (Table 2 and Figure 2, and Supplementary Table S1).

Table 2.

Primary Endpoint - Intraoperative Remifentanil Infusion Rate (µg Kg⁻¹ H⁻¹)

Population	PTi-Guided Mean (SD), n	Control Mean (SD), n	Unadjusted Mean Difference (95% CI)	Adjusted Mean Difference (95% CI)	P value
Full-analysis set	7.44 (1.43), n=98	7.63 (1.41), n=97	−0.20 (−0.60 to 0.21)	−0.06 (−0.44 to 0.32)	0.758
Per-protocol set	7.46 (1.43), n=94	7.63 (1.44), n=92	−0.17 (−0.59 to 0.24)	−0.03 (−0.43 to 0.37)	0.884

Notes: Data are mean (SD). Effects compare PTi-guided titration with conventional hemodynamic-guided titration. Adjusted models included treatment, age, sex, body mass index, ASA physical status, surgical category, admission MAP, and admission heart rate with heteroskedasticity-consistent (HC3) robust standard errors.

Abbreviations: ASA, American Society of Anesthesiologists; CI, confidence interval; FAS, full-analysis set; MAP, mean arterial pressure; PPS, per-protocol set; PTi, pain threshold index.

Figure 2.

Primary endpoint and sensitivity analyses. (A) Distribution of intraoperative remifentanil infusion rate in the full-analysis set (pain threshold index-guided group, n=98; control group, n=97). Dots are individual patients, violins show the empirical distribution, boxes show median and interquartile range, whiskers show 1.5 times the interquartile range, and black points with error bars show mean and 95% confidence interval. (B) Mean differences compare pain threshold index-guided titration with conventional hemodynamic-guided titration; horizontal bars show 95% confidence intervals. The primary model was covariate-adjusted linear regression with heteroskedasticity-consistent (HC3) robust standard errors.

Abbreviations: PTi, pain threshold index.

Exploratory subgroup analyses did not show credible treatment-effect heterogeneity. Treatment-by-subgroup interaction P values were greater than 0.05 for sex, age, body-mass index, admission MAP, admission heart rate, long-term treatment, previous surgery, and surgical category; the ASA interaction was nominally lowest but imprecise and not adjusted for multiplicity (Supplementary Table S2 and Supplementary Figure S1).

Secondary and Exploratory Outcomes

PTi guidance was associated with fewer early rescue-opioid events but not with a clear shift in PACU pain or sedation scores. At PACU admission, median (IQR) NRS pain score was 2.0 (2.0–3.0) in the PTi-guided group and 3.0 (2.0–4.0) in controls; adjusted median difference was 0.00 (95% CI, −0.48 to 0.48; P=1.000), and an ordinal sensitivity model did not show a statistically supported difference in higher NRS scores. Ramsay scores were 2.0 (2.0–2.0) in both groups, with no adjusted difference. PACU rescue opioid was used in 18/98 (18.4%) PTi-guided patients and 31/97 (32.0%) controls (adjusted risk ratio 0.56; 95% CI, 0.35 to 0.91; adjusted risk difference −14.2 percentage points). Any opioid rescue through 48 h occurred in 29/98 (29.6%) and 44/97 (45.4%) patients, respectively (adjusted risk ratio 0.62; 95% CI, 0.45 to 0.85; adjusted risk difference −17.8 percentage points). Hydromorphone consumption and total postoperative opioid exposure through 48 h were not reduced, whereas nonsteroidal anti-inflammatory drug rescue analgesia was more frequent in the PTi-guided group. Any PONV through 48 h occurred in 17/98 (17.3%) and 29/97 (29.9%) patients, respectively (adjusted risk ratio 0.61; 95% CI, 0.38 to 0.97; adjusted risk difference −11.5 percentage points). These secondary findings are shown in Table 3 and Figure 3, and Supplementary Table S3. Supportive per-protocol analyses of PACU NRS, Ramsay sedation, and time-specific PONV endpoints were broadly consistent with the full-analysis-set findings (Supplementary Table S7).

Table 3.

Key Secondary Endpoints in the Full-Analysis Set

Endpoint	PTi-guided (n=98)	Control (n=97)	Effect (95% CI)	P value	FDR q
Propofol infusion rate, mg kg−1 h−1	8.45 (1.30)	8.48 (1.34)	Adjusted mean difference, −0.09 (−0.46 to 0.29)	0.649	0.893
Extubation time, min	10.0 (5.0–15.0)	7.0 (5.0–14.0)	Adjusted median difference 2.20 (−0.28 to 4.68)	0.082	0.180
PACU NRS pain score	2.0 (2.0–3.0)	3.0 (2.0–4.0)	Adjusted median difference 0.00 (−0.48 to 0.48)	1.000	1.000
PACU Ramsay Sedation Scale score	2.0 (2.0–2.0)	2.0 (2.0–2.0)	Adjusted median difference 0.00 (0.00 to 0.00)	0.982	1.000
PACU rescue opioid	18/98 (18.4)	31/97 (32.0)	RR 0.56 (0.35 to 0.91); RD −14.2 pp	0.018	0.100
Ward rescue opioid through postoperative day 2	18/98 (18.4)	21/97 (21.6)	RR 0.79 (0.50 to 1.25); RD −4.7 pp	0.320	0.586
Any rescue opioid through 48 h	29/98 (29.6)	44/97 (45.4)	RR 0.62 (0.45 to 0.85); RD −17.8 pp	0.003	0.037
PCIA hydromorphone through 48 h, mg	2.0 (0.5–5.2)	3.0 (1.0–5.0)	Adjusted median difference 0.27 (−0.67 to 1.20)	0.574	0.893
Total postoperative opioid exposure through 48 h, mg morphine equivalent	25.0 (6.2–67.2)	35.0 (10.0–60.0)	Adjusted median difference 1.56 (−9.40 to 12.51)	0.780	0.953
Any NSAID rescue analgesia	53/98 (54.1)	41/97 (42.3)	RR 1.36 (1.02 to 1.81); RD 14.6 pp	0.037	0.103
Any PONV through 48 h	17/98 (17.3)	29/97 (29.9)	RR 0.61 (0.38 to 0.97); RD −11.5 pp	0.035	0.103

Notes: Data are mean (SD), median (25th-75th percentile), or number (%), unless otherwise stated. Secondary endpoints are exploratory. Binary endpoints used modified Poisson regression with robust variance and marginal standardization; P values correspond to adjusted risk ratios, with adjusted risk differences reported as absolute effects. PACU NRS and Ramsay scores and skewed continuous endpoints used median regression. FDR q values use the Benjamini-Hochberg correction across the expanded key secondary endpoint set.

Abbreviations: CI, confidence interval; FDR, false-discovery rate; NRS, Numeric Rating Scale; NSAID, nonsteroidal anti-inflammatory drug; PACU, post-anesthesia care unit; PCIA, patient-controlled intravenous analgesia; PONV, postoperative nausea and vomiting; pp, percentage points; PTi, pain threshold index; RD, risk difference; RR, risk ratio.

Figure 3.

Postoperative rescue analgesia, postoperative nausea and vomiting, and opioid exposure. (A) Adjusted risk differences for rescue-analgesia and postoperative nausea and vomiting endpoints in the full-analysis set (pain threshold index-guided group, n=98; control group, n=97). Points show marginal standardized risk differences and horizontal bars show 95% confidence intervals; event counts are pain threshold index-guided versus control. (B) Empirical cumulative distribution of total postoperative opioid exposure through 48 hours expressed as morphine equivalents in the full-analysis set. Dotted vertical lines indicate group medians.

Abbreviations: NSAID, nonsteroidal anti-inflammatory drug; PACU, post-anesthesia care unit; PONV, postoperative nausea and vomiting; PTi, pain threshold index.

Repeated-measures trajectory models did not show significant treatment-by-time interactions for MAP, heart rate, or WLi (Supplementary Table S4 and Supplementary Figure S2). Averaged over all time points, MAP was modestly lower with PTi guidance (adjusted difference −2.1 mm Hg; 95% CI, −4.0 to −0.2), heart rate was numerically lower (−2.1 beats min⁻¹; 95% CI, −4.4 to 0.1), and WLi was similar (−0.6; 95% CI, −1.4 to 0.2). Target attainment from T2 to T6 was similar between groups for WLi, MAP, and heart rate.

In the PTi-guided group, the median patient-level proportion of sampled T2-T6 values within the PTi target was 0.60 (IQR, 0.40–0.80; Supplementary Table S4). This describes attainment at five scheduled observations, not the proportion of operating time in range. Sparse PTi recordings in controls and the lack of continuous monitor exports limited between-group assessment of PTi behavior.

The exploratory mediation models did not establish mediation of rescue-opioid outcomes through intraoperative remifentanil infusion rate, MAP target attainment, or heart-rate target attainment. For PACU rescue opioid, the estimated indirect effect through extubation time was −1.4 percentage points (95% CI, −4.8 to 0.2), and the interval included zero. These non-prespecified analyses do not demonstrate a causal pathway or rule out other mechanisms. Detailed exploratory mediation estimates are shown in Supplementary Table S6.

Hospital length of stay did not differ materially between groups (Supplementary Table S3).

Safety Outcomes

Other safety endpoints, including intraoperative tachycardia, bradycardia, hypertension, hypotension, hypoxemia, vasoactive drug use, intraoperative awareness, and ICU admission, did not differ materially between groups. Time-specific PONV events were 4/98 (4.1%) versus 6/97 (6.2%) at PACU admission, 13/98 (13.3%) versus 22/97 (22.7%) at 24 h, 2/98 (2.0%) versus 8/97 (8.2%) at 48 h, and 3/98 (3.1%) versus 3/97 (3.1%) on postoperative day 7. In the repeated-measures PONV model across all four time points, PTi guidance was associated with a lower overall PONV event risk (risk ratio 0.60; 95% CI, 0.37 to 0.96; P=0.033), with no evidence of treatment-by-time interaction (P=0.635). No patient was admitted to ICU and no intraoperative awareness was recorded. Detailed safety and time-specific PONV estimates are shown in Supplementary Table S5 and Supplementary Figure S3.

Discussion

In this assessor-blinded randomized trial, PTi-guided remifentanil titration did not reduce intraoperative remifentanil infusion rate, the primary pharmacologic endpoint. The conclusion was consistent in the per-protocol population and across sensitivity analyses, including additional adjustment for baseline imbalances. Subgroup analyses did not show credible heterogeneity; the exploratory trajectory, target-attainment, and mediation analyses cannot establish or exclude a treatment-effect mechanism. The newly checked PACU NRS and Ramsay scores also did not indicate a clinically distinct early pain or sedation profile between groups.

The neutral primary result should be interpreted in the context of a heterogeneous literature. Selected studies of surgical pleth index, nociception level, pupillary reflex dilation, and analgesia nociception index-guided dosing have reported reductions in intraoperative opioid or early pain outcomes.^8–13 However, other randomized trials have been neutral, and recent systematic reviews and a network meta-analysis show that effects vary by monitor type, dosing algorithm, anesthetic technique, and endpoint.^14–17 These data suggest that an electroencephalography-derived PTi algorithm should not be assumed to reproduce the pharmacologic effects observed with autonomic or multiparameter indices.

The reduction in early rescue-opioid events is clinically interesting but remains exploratory. Adjusted risk differences of −14.2, −17.8, and −11.5 percentage points correspond to approximately 14 fewer patients requiring PACU rescue opioid, 18 fewer requiring any rescue opioid through 48 h, and 12 fewer experiencing PONV through 48 h per 100 patients, respectively (Table 3). Such differences could be relevant to early recovery if confirmed. However, among these three outcomes, only any rescue opioid through 48 h had an FDR q value below 0.05; the q values were 0.100, 0.037, and 0.103, respectively. PACU NRS was numerically lower in the PTi-guided group, but adjusted median and ordinal models did not support a robust difference, and Ramsay sedation scores were essentially identical. These findings suggest that the lower PACU rescue-opioid use was not simply a consequence of measurable oversedation or a large fixed-time pain-score separation at PACU admission. PONV through 48 h was less frequent with PTi guidance, with the clearest absolute separation at 24 and 48 h, but this secondary signal was not accompanied by a between-group difference on postoperative day 7. These observations do not establish a sustained opioid- or nausea-sparing benefit. The lower early rescue-opioid and PONV signals were not accompanied by a robust reduction in hydromorphone consumption or total morphine equivalent opioid dose through 48 h, and nonsteroidal rescue analgesia was more frequent with PTi guidance. Modern postoperative pain and PONV guidance emphasizes patient-centered recovery, multimodal analgesia, and avoidance of simple opioid-dose endpoints as the sole measure of benefit.^36,37 A further concern in remifentanil-based anesthesia is opioid-induced hyperalgesia, but the present dataset did not include mechanistic measures that could distinguish hyperalgesia from differences in early rescue-analgesia practice.³⁸

The post hoc analyses refine interpretation. Subgroup analyses in randomized trials are frequently underpowered and prone to false-positive claims; our interaction tests were exploratory and focused on consistency.^30–32 Mediation analysis also rests on strong assumptions about unmeasured mediator-outcome confounding.^33–35 Randomization of treatment does not remove potential mediator-outcome confounding. The indirect-effect intervals for remifentanil infusion and extubation time included zero; these estimates neither demonstrate mediation nor identify why rescue-opioid use differed.

Strengths include randomized allocation, assessor masking, standardized target-controlled anesthesia, analysis of full-analysis and per-protocol populations, robust primary modeling, explicit baseline-balance assessment, interaction testing, repeated-measures modeling for hemodynamic and PONV trajectories, and transparent post hoc analyses. Limitations include the single-center design, inability to blind intraoperative anesthesiologists, restricted body-mass-index range, heterogeneous laparoscopic procedures, sparse PTi data in controls, absence of continuous high-frequency monitor exports, and limited power for subgroup, mediation, and uncommon safety outcomes. PACU NRS, Ramsay, and PONV scores were collected at clinically defined time points rather than continuously, and postoperative analgesia was clinically titrated; therefore, fixed-time ward pain scores were less informative than rescue-analgesia patterns and cumulative opioid exposure. The main postoperative pain and opioid outcomes covered 48 h, and scheduled adverse-event follow-up, including PONV assessment, extended through postoperative day 7. Persistent postsurgical pain, hyperalgesia, and later benefits or harms were not assessed. Sampled PTi target attainment does not establish continuous signal stability, discrimination of individual noxious events, or information independent of WLi.

Conclusions

In this single-center trial of laparoscopic abdominal surgery under propofol-remifentanil target-controlled anesthesia, PTi-guided remifentanil titration did not reduce intraoperative remifentanil infusion. PACU pain and sedation scores did not show a robust between-group difference. Early rescue-opioid events and PONV through 48 h were fewer with PTi guidance, but these secondary signals were exploratory, were not accompanied by a robust reduction in postoperative opioid dose, and do not establish a causal mechanism. The findings should not be extrapolated to other surgical populations, procedures, or anesthetic techniques. Prospective multicenter confirmation with continuous monitor exports, prespecified estimands, and patient-centered pain and recovery outcomes is required before a clinical benefit or opioid-sparing role can be claimed.

Funding Statement

This study was supported by Beijing Fresenius Kabi Pharmaceutical Co., Ltd., Beijing, China. The funder had no role in study design, data collection, analysis, interpretation, manuscript preparation, or the decision to submit the manuscript.

Clinical Trial Registration

Chinese Clinical Trial Registry, ChiCTR2300078455, https://www.chictr.org.cn/showproj.html?proj=206473; registered December 8, 2023.

Abbreviations

ASA, American Society of Anesthesiologists; CI, confidence interval; ECG, electrocardiogram; FDR, false-discovery rate; ICU, intensive care unit; IQR, interquartile range; MAP, mean arterial pressure; NRS, Numeric Rating Scale; NSAID, nonsteroidal anti-inflammatory drug; PACU, post-anesthesia care unit; PCIA, patient-controlled intravenous analgesia; PONV, postoperative nausea and vomiting; PTi, pain threshold index; RD, risk difference; RR, risk ratio; SD, standard deviation; WLi, wavelet index.

Data Sharing Statement

Deidentified individual participant data underlying the results reported in this article, together with the data dictionary, study protocol, statistical analysis plan, and analytic code, may be made available from the corresponding authors after publication to researchers with a methodologically sound proposal, subject to institutional and ethics approval and a data-use agreement. Requests may be submitted from the date of publication and will be considered for as long as the institution retains the study records under applicable policy. Public deposition is not planned because the dataset contains participant-level clinical information.

Ethics Approval and Informed Consent

This study was conducted in accordance with the Declaration of Helsinki and its later amendments. The Ethics Committee of Scientific Research and Clinical Trials of The First Affiliated Hospital of Zhengzhou University approved the protocol (reference S2023-K012-002; approval date: August 29, 2023). Written informed consent was obtained from all participants before enrollment.

Consent for Publication

The manuscript contains no identifiable individual data, images, videos, or recordings.

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 competing interest in this work.