Clinical and Economic Outcomes of Liposomal Bupivacaine, Ropivacaine, and Patient-Controlled Intravenous Analgesia for Adult Video-Assisted Thoracoscopic Surgery: A Retraospective Cohort Study

Abstract

Objective

To evaluate the effectiveness, safety, and cost-effectiveness of three analgesic regimens—PCIA (patient-controlled intravenous analgesia), R (ropivacaine), and LB (liposomal bupivacaine) local infiltration—in adult patients undergoing VATS (video-assisted thoracoscopic surgery).

Methods

Adults undergoing unilateral VATS in 2024 were retrospectively assigned to PCIA, R, and LB groups. IPTW (inverse probability of treatment weighting) balanced baseline confounders. The primary effectiveness outcome was the 72-h cumulative pain burden (area under the curve, AUC). Secondary outcomes included NRS (numerical rating scale) scores and early recovery indicators; independent factors for postoperative pain were also analyzed. Safety was assessed by adverse event incidences. CEA (cost-effectiveness analysis) was conducted from the Chinese healthcare perspective, with robustness validated by one-way and probabilistic sensitivity analyses (OWSA, PSA).

Results

829 patients (PCIA: 274; R: 258; LB: 297) were included. After IPTW, baseline characteristics were balanced (SMD < 0.1). The LB group had the lowest 72-h AUC and overall NRS score compared to PCIA (β =−11.01, P = 0.017; 0.80, 95% CI: 0.70–0.89, P = 0.022) and R (β = −30.19, P < 0.001; 1.07, 95% CI: 0.96–1.18, P < 0.001). LB group had the shortest time to first ambulation, and both LB and PCIA groups had shorter hospital stays than the R group (all P < 0.05). No significant differences were found in rescue analgesia incidence or consumption (P > 0.05). Advanced age and tumor history protected against postoperative pain, whereas alcohol history was a risk factor. No significant differences were found in adverse event rates, including postoperative nausea and vomiting, among the groups (all P > 0.05). Economically, PCIA had the highest total cost and R the lowest. LB was dominant over PCIA; versus R, the ICER (incremental cost-effectiveness ratio) for LB was USD 123.29 per 1-point NRS reduction. Sensitivity analyses confirmed robustness.

Conclusion

For VATS postoperative analgesia, LB is associated with favorable analgesia and earlier recovery. It may be cost-effective, depending on acceptable WTP thresholds. PCIA and R remain safe and viable alternatives. These findings warrant further validation in large-scale prospective studies.

Introduction

Postoperative pain control in thoracic surgery is of critical importance.¹ In recent years, VATS (video-assisted thoracoscopic surgery) has become the mainstream surgical approach for lung cancer treatment, owing to its advantages of less trauma, less bleeding, and faster recovery.^2,3 However, “minimally invasive” does not mean painless; due to factors such as compression from the trocar ports and continuous irritation from drainage tubes, as many as 50% to 75% of patients still experience moderate-to-severe acute pain in the early postoperative period.^4–6 Therefore, adequate analgesia after VATS is of great significance for promoting early recovery, reducing the incidence of complications, and preventing chronic pain.⁷ In this context, exploring optimal postoperative analgesia regimens is particularly urgent.

In current clinical practice of postoperative analgesia, based on differences in administration routes and formulation technologies, commonly used analgesic regimens mainly include: PCIA (patient-controlled intravenous analgesia), LIA (local infiltration analgesia) with conventional amide local anesthetics represented by ropivacaine, and LIA with LB (liposomal bupivacaine), which relies on sustained-release technology. As a common approach for systemic postoperative analgesia, PCIA offers advantages such as rapid onset and individualized dosing. However, it is associated with a high incidence of opioid-related side effects, including nausea, vomiting, constipation, and pruritus. Moreover, consumables such as infusion pumps are mostly out-of-pocket expenses, imposing a certain financial burden on patients.^8–10 Ropivacaine, a commonly used amide local anesthetic in clinical practice, is characterized by sensory-motor dissociation and relatively low cardiovascular and central nervous system toxicity.^11,12 Nevertheless, its analgesic duration following a single local infiltration is limited and cannot effectively cover the peak period of acute postoperative pain lasting 2–3 days, thus failing to meet patients’ needs for medium-to-long-term analgesia.^13,14 In contrast to conventional regimens, LB, as an ultra-long-acting sustained-release formulation, achieves gradual drug release via multilamellar liposome delivery technology, providing up to 72 hours of continuous analgesia with a single administration.¹⁵ This formulation was approved by the US FDA in 2011 and was launched in China in 2022. Multiple meta-analyses have demonstrated that LB significantly reduces resting pain scores at 24–72 h postoperatively and decreases postoperative opioid consumption in joint replacement and spinal surgeries.^16,17 However, whether these findings apply to thoracic surgery remains unclear. VATS pain has distinct characteristics—intercostal nerve irritation, pleural inflammation, and dynamic pain aggravated by breathing and coughing—so findings from other surgeries cannot be directly extrapolated, and evidence specifically in this field remains limited.¹⁸ In addition, the unit cost of LB is substantially higher than that of conventional local anesthetics, and whether its potential advantages can translate into clinically meaningful benefits that justify the extra costs remains to be determined.

In summary, the optimal choice of analgesic regimens following VATS still requires further systematic evaluation and validation with more clinical evidence. However, to date, real-world studies that systematically compare the three regimens—PCIA, ropivacaine, and LB—in terms of efficacy, safety, and cost‑effectiveness remain limited, particularly regarding direct comparative evidence specifically in the VATS setting. Therefore, the present study, based on real–world clinical data, systematically evaluates the differences in efficacy, safety, and cost-effectiveness among the above three regimens for postoperative analgesia after VATS, aiming to provide evidence for informed clinical decision-making.

Materials and Methods

Data Source

This study adopted a retrospective cohort study design and included patients who underwent VATS at a tertiary hospital in Jiangsu Province from January to December 2024 and received postoperative analgesia with PCIA, ropivacaine, or LB. Patient data were collected from the hospital’s electronic medical record system, including: (1) baseline and clinical characteristics, such as age, sex, ASA (American Society of Anesthesiologists) physical status, and comorbidities; and (2) outcome measures, including postoperative NRS (numeric rating scale) scores, rescue analgesia, treatment costs, and adverse events. These data were used for a systematic evaluation of the efficacy, safety, and cost-effectiveness of each analgesic regimen.

Inclusion and Exclusion Criteria

Inclusion Criteria

(1) age ≥ 18 years; (2) ASA physical status II or III; (3) VATS with one of the following three postoperative analgesic regimens: LB for local infiltration, ropivacaine for local infiltration, or PCIA.

Exclusion Criteria

(1) presence of contraindications to local anesthetics, severe hepatic or renal insufficiency, coagulation disorders, infection at the surgical/puncture site, pregnancy, or lactation; (2) history of severe psychiatric disorders or cognitive impairment that could affect cooperation or postoperative evaluation; (3) preoperative chronic opioid use or opioid tolerance; (4) history of ipsilateral thoracic surgery; (5) intraoperative conversion to thoracotomy, concurrent surgery at other sites, reoperation within 72 h postoperatively due to complications, or switching of the analgesic regimen for any reason; (6) combined use of two or more local anesthetics, or intraoperative continuous infusion of other non-opioid analgesics; (7) duplicate enrollment of the same patient; (8) missing core clinical data (eg, NRS scores at key postoperative time points, rescue medication records) that precluded effective outcome assessment.

Perioperative Anesthesia and Analgesia Management

Intraoperative Anesthesia Management

All patients received a standardized general anesthesia induction regimen with intravenous sufentanil 0.5 μg·kg⁻¹, propofol 2 mg·kg⁻¹, and rocuronium 0.6 mg·kg⁻¹. After induction, a double-lumen endobronchial tube was inserted for OLV (one-lung ventilation). Anesthesia was maintained with inhaled sevoflurane and continuous intravenous infusion of propofol and remifentanil, with the infusion rates adjusted dynamically based on BIS (bispectral index) monitoring to maintain BIS values between 40 and 60. All anesthetic maintenance drugs were discontinued immediately at the end of surgery.

Sufentanil redistributes rapidly after intravenous administration, with a distribution half-life of approximately 1.4 min; by the end of surgery, the induction dose has been sufficiently redistributed and metabolized.¹⁹ Remifentanil used for intraoperative maintenance has a context-sensitive half-life of approximately 3 to 5 min, and its residual effects at the end of surgery are negligible.²⁰ Therefore, the consistent intraoperative anesthesia management across groups effectively controlled the potential confounding effect of intraoperative medications on postoperative pain outcomes.

Postoperative Analgesia Regimens

PCIA (patient-controlled intravenous analgesia) group: At the end of surgery, an electronic infusion pump was connected for continuous analgesia for 48 h. The pump regimen consisted of sufentanil 2.5 μg·kg⁻¹ diluted with normal saline to a total volume of 100 mL, with a background infusion rate of 2 mL·h⁻¹, a bolus dose of 2 mL, and a lockout interval of 15 min.

R (ropivacaine) group: Before wound closure, ropivacaine hydrochloride injection (10 mL:75 mg; Zhejiang Xianju Pharmaceutical Co., Ltd., China) was diluted with normal saline to a total volume of 20 mL (final concentration 0.375%) and infiltrated layer-by-layer into the tissues around the surgical incision and drainage tube sites for LIA.

LB (liposomal bupivacaine) group: Before wound closure, 20 mL of undiluted liposomal bupivacaine injection (Aihengping^®; 20 mL:266 mg; Jiangsu Hengrui Pharmaceutical Co., Ltd., China) was aspirated and administered using the same method as in the R group for LIA.

Rescue Analgesia Management

Postoperatively, when patients reported pain with a resting NRS score ≥ 4, or experienced intolerable breakthrough pain, rescue analgesia was administered according to the ward’s standardized protocol (eg, oral tramadol, or injection of pethidine or bucinnazine). The total consumption of rescue analgesics was uniformly converted to IV-MME (intravenous morphine milligram equivalents) for subsequent statistical analysis.

Effectiveness and Safety Evaluation

Primary Outcome

The cumulative pain burden within 72 h postoperatively (ie, area under the pain score-time curve, AUC). The AUC was calculated using the trapezoidal rule based on the resting NRS scores at 2, 24, 48, and 72 h postoperatively. Unlike a single time-point NRS measurement, the AUC integrates both pain intensity and duration over the 72-hour observation period, reflecting the overall pain burden.

Secondary Outcomes

Overall mean resting NRS score assessed using a GEE (generalized estimating equation) model (hereinafter referred to as the overall NRS score), NRS scores at specific time points (2, 24, 48, and 72 h postoperatively) and their longitudinal trends, as well as other clinical outcomes including rescue analgesia (incidence and MME consumption), LOS (length of hospital stay), time to first ambulation, anesthesia recovery time, temperature variation, and rate of return to normothermia.

Safety Outcomes

Adverse events during the treatment period were collected from the medical records, including PONV (postoperative nausea and vomiting), constipation, dizziness, fever, hypotension, arrhythmia, and hepatic/renal injury. Safety was evaluated based on the incidence of each specific adverse event.

Economic Evaluation

Perspective, Time Horizon, and Cost Identification

This study conducted a CEA (cost-effectiveness analysis) from the perspective of the Chinese healthcare system. Regarding cost identification, only direct medical costs were included; indirect costs (eg, productivity loss, caregiver time lost) and intangible costs were excluded due to substantial individual variation and the difficulty of accurately quantifying them from retrospective data. In addition, because the time horizon of this study (72 h postoperatively) was less than one year, no discounting of costs or effects was applied, in accordance with the China Guidelines for Pharmacoeconomic Evaluation.

All direct medical cost data were retrospectively extracted from the real-world electronic medical record system. Total direct medical costs consisted of the following five components: analgesic intervention cost, rescue analgesia cost, adverse event management cost, intraoperative anesthesia cost, and basic hospitalization cost. The analgesic intervention cost for the R group and LB group was calculated as the unit drug cost, which was USD (US dollars) 0.70 and USD 53.64, respectively; for the PCIA group, this cost comprised the fixed pump cost (USD 34.12 per set), daily equipment maintenance cost (USD 11.23 per day), and the cost of drugs in the pump. Basic hospitalization costs (including bed charges, routine nursing care, consumables, and surgical procedures) were adjusted by deducting the aforementioned four cost components to avoid double-counting. All cost data were uniformly converted to USD using the official exchange rate published by the National Bureau of Statistics of China for 2024 (1 USD = 7.1217 CNY).

Base–Case Analysis

The overall NRS score within 72 h postoperatively was used as the effectiveness outcome measure in this study.²¹ The NRS score is an inverse outcome measure, where a lower score indicates better analgesic efficacy. Cost input parameters were derived from weighted means and standard errors estimated by an inverse probability of treatment weighting (IPTW)- GLM (generalized linear model); effectiveness input parameters were extracted from overall marginal means calculated by the weighted GEE model. Details of the modeling procedures are described in Statistical Analysis.

The ICER (incremental cost-effectiveness ratio) was calculated using the following formula:

(1)

Where C represents the weighted mean total cost, E represents the overall NRS score; subscripts 2 and 1 denote the intervention group and the reference group, respectively. The numerator, C₂–C₁, represents the incremental cost of the intervention group relative to the reference group (a positive value indicates higher cost in the intervention group), and the denominator, E₁–E₂, represents the reduction in the overall NRS score attributable to the intervention (ie, health benefit; a positive value indicates better efficacy in the intervention group). The ICER represents the additional total direct medical cost required per 1-point reduction in the overall NRS score for the intervention group compared with the reference group. If a regimen had both lower total direct medical cost and lower overall NRS score than the reference group (ie, lower cost and better analgesic efficacy), it was defined as a dominant strategy.

Sensitivity Analysis and Willingness-to-Pay Threshold

Uncertainty in model parameters was assessed through OWSA (one-way sensitivity analysis) and PSA (probabilistic sensitivity analysis).

In the OWSA, the analgesic intervention costs (at fixed unit prices) in the R group and LB group were varied by ±30% from the baseline value; for the remaining cost parameters and NRS score parameters, the variation ranges were set according to the 95% CIs (confidence intervals) estimated from the corresponding models. The results were presented as tornado diagrams.

The PSA was performed using 1,000 Monte Carlo simulations, with parameter sampling based on the weighted means and standard errors of each group. Cost parameters were assumed to follow a gamma distribution, and effectiveness parameters were assumed to follow a normal distribution. Because no universally accepted WTP (willingness-to-pay) threshold has been established for a 1-point reduction in the overall NRS score, a specific WTP threshold was not set in this study. Instead, based on the simulation results, a CEAC (cost-effectiveness acceptability curve) was constructed to present the probability of each intervention being cost-effective across different WTP values.

Sample Size Calculation

The sample size was estimated based on the 72-h AUC in this study. Using preliminary data from 20 patients in the R group (AUC_72h = 112.90 ± 55.13), the minimal clinically important difference was set as a 15% reduction in overall pain burden,²² corresponding to an absolute difference of 16.94, with Cohen’s d = 0.31. Under a two-sided α = 0.05 and a power of 90%, the required sample size per group was calculated to be 224 patients.¹ Given that this study included three parallel groups, the total sample size was at least 672 patients.

(2)

Statistical Analysis

All data management, statistical analyses, and visualization were performed using R version 4.3.1. A two-sided P < 0.05 was considered statistically significant.

Normality Testing and Descriptive Statistics

The Shapiro–Wilk test was used to assess the normality of continuous variables. Variables following a normal distribution were presented as mean ± SD (standard deviation); non-normally distributed variables were presented as median and interquartile range (M[IQR]). Categorical variables were presented as frequencies and percentages.

Inverse Probability of Treatment Weighting

To control for baseline confounding, this study employed multinomial IPTW to generate a weighted pseudo-population, with the ATE (average treatment effect) as the target estimand. Propensity scores were calculated using a multinomial logistic regression model that included 13 baseline covariates: operative time, sex, age, BMI (body mass index), ASA physical status, cardiovascular comorbidities, diabetes, tumor, neurological disorders, type of resection, anesthesia, smoking, and alcohol consumption. To reduce the influence of extreme weights, the obtained weights were trimmed at the 1st and 99th percentiles.

The absolute SMD (standardized mean difference) was used to evaluate covariate balance, and Love plots were generated to visually present the balance of all baseline variables before and after weighting. In addition, the ESS (effective sample size) after weighting was calculated for each group to assess statistical power.

Effectiveness Outcomes and Analysis of Factors Influencing Pain

Group differences in the primary outcome, the 72-h AUC, were assessed using a weighted GLM, with adjustment for five covariates: sex, age, BMI, ASA physical status, and type of resection. For the secondary outcomes, longitudinal resting NRS scores were analyzed using a weighted GEE. The model included intervention group, time point, and their interaction term (Group × Time), specified a Gaussian distribution family, and used a first‑order autoregressive (AR(1)) structure as the working correlation matrix. After adjustment for the aforementioned five covariates, overall effects were evaluated using Type III Wald chi–square tests, and EMMs (estimated marginal means) with 95% CIs were extracted at each time point. P–values for post hoc multiple comparisons between groups were adjusted using the Holm method. Given that rescue analgesic consumption data were highly right-skewed with a substantial number of zero values, a two-part model was used for evaluation: a weighted logistic regression model was first used to analyze the rate of rescue analgesia use across groups, followed by a weighted GLM to assess the differences in consumption within the subgroup of patients who received rescue analgesia. Other continuous outcomes were assessed using a weighted GLM for group comparisons.

To further explore factors independently associated with postoperative pain, univariate and multivariate weighted GEE models were constructed with adjustment for group and time point; variables with P < 0.1 in the univariate analysis were entered into the multivariate model. Finally, two sensitivity analyses were performed to validate the robustness of the primary models: (1) replacing the working correlation matrix of the GEE with an exchangeable structure; and (2) constructing a CLMM (cumulative link mixed model) with patients as random intercepts, based on the ordinal nature of the NRS scores.

Safety Outcomes

Binary safety outcomes such as adverse events were analyzed using weighted logistic regression models. For specific adverse events (eg, pneumothorax, oliguria) in which the number of events within a group was zero, leading to model non‑convergence or fitting failure, group differences were assessed using the Rao-Scott chi–square test.

Economic Parameter Estimation

The means and 95% CIs of each direct medical cost component were estimated using a weighted GLM with a Gaussian distribution. To avoid variance inflation, the model included only the intervention group as the independent variable without additional adjustment for covariates. The extracted estimates were used directly as input parameters for the subsequent cost-effectiveness analysis. Finally, OWSA and PSA were performed to evaluate the uncertainty of the CEA.

Results

Study Population and Baseline Characteristics

A total of 829 eligible patients who underwent VATS lung resection were finally included in this study, including 274 in the PCIA group, 258 in the R group, and 297 in the LB group (Figure 1). Before weighting, the three groups differed in some baseline characteristics, including type of resection and anesthesia. After IPTW, the weighted pseudo‑sample sizes for the PCIA, R, and LB groups were approximately 274, 258, and 290, respectively, with corresponding ESS of approximately 259, 244, and 282, respectively. The ESS in each group was approximately 94% of the original sample size, indicating that statistical power was well preserved after IPTW. Detailed baseline characteristics of the three groups before and after weighting are presented in Table 1.

Figure 1.

Flow diagram of patient selection and group allocation.

Abbreviations: LB, liposomal bupivacaine; NRS, numerical rating scale; PCIA, patient-controlled intravenous analgesia; R, ropivacaine; VATS, video-assisted thoracoscopic surgery.

Table 1.

Baseline Characteristics of the Study Cohort Before and After Inverse Probability of Treatment Weighting (IPTW)

	Baseline Data of Patients Before IPTW				Baseline Data of Patients After IPTW
	PCIA (n = 274)	R (n = 258)	LB (n = 297)	SMD	PCIA (N = 274)	R (N = 258)	LB (N = 290)	SMD
Operative time	1.45 (1.00, 1.92)	1.48 (1.08, 1.92)	1.50 (1.03, 2.08)	0.126	1.50 (1.00, 2.00)	1.50 (1.08, 1.93)	1.50 (1.00, 2.08)	0.028
Age	63.00 (54.25, 70.00)	65.00 (56.00, 71.00)	63.00 (55.00, 69.00)	0.108	63.00 (54.00, 70.00)	64.00 (56.00, 70.00)	63.00 (54.34, 69.00)	0.031
BMI	23.44 (21.68, 25.31)	23.40 (21.44, 25.61)	23.56 (21.64, 25.51)	0.112	23.51 (21.78, 25.42)	23.27 (21.38, 25.39)	23.53 (21.62, 25.43)	0.021
Male	109 (39.8)	100 (38.8)	105 (35.4)	0.091	103 (37.7)	100 (38.6)	107 (37.0)	0.034
ASA
II	225 (82.1)	203 (78.7)	249 (83.8)	0.133	225 (82.1)	211 (81.9)	240 (82.5)	0.016
III	49 (17.9)	55 (21.3)	48 (16.2)		49 (17.9)	47 (18.1)	51 (17.5)
Smoking	57 (20.8)	55 (21.3)	41 (13.8)	0.194	50 (18.3)	49 (19.1)	50 (17.4)	0.046
Alcohol consumption	17 (6.2)	23 (8.9)	22 (7.4)	0.103	20 (7.3)	20 (7.6)	22 (7.5)	0.010
Comorbidities
Cardiovascular	106 (38.7)	104 (40.3)	111 (37.4)	0.060	102 (37.0)	99 (38.3)	110 (37.9)	0.026
Diabetes	31 (11.3)	28 (10.9)	29 (9.8)	0.050	26 (9.4)	26 (10.2)	30 (10.2)	0.025
Tumor	19 (6.9)	18 (7.0)	8 (2.7)	0.188	15 (5.3)	14 (5.2)	12 (4.2)	0.049
Neurological	8 (2.9)	13 (5.0)	5 (1.7)	0.191	8 (2.9)	9 (3.3)	8 (2.8)	0.027
Resection type
Lobe	201 (73.4)	202 (78.3)	236 (79.5)	0.145	211 (77.1)	198 (76.8)	225 (77.5)	0.015
Segmentectomy	50 (18.2)	33 (12.8)	25 (8.4)	0.293	36 (13.2)	35 (13.7)	35 (12.0)	0.050
Wedge resection	23 (8.4)	23 (8.9)	36 (12.1)	0.125	27 (9.8)	24 (9.5)	31 (10.5)	0.035
Anesthesia
Anesthesia_1	119 (43.4)	153 (59.3)	151 (50.8)	0.320	138 (50.2)	131 (51.0)	149 (51.3)	0.021
Anesthesia_2	155 (56.6)	105 (40.7)	146 (49.2)		136 (49.8)	126 (49.0)	142 (48.7)

Notes: Data are presented as median (interquartile range) or number (percentage). The sample size (n) before weighting represents the actual number of patients in each group. The sample sizes (N) after IPTW represent the weighted sample sizes for each group, calculated using stabilized weights trimmed at the 1st and 99th percentiles, and have been rounded to the nearest integer. For anesthesia methods, Anesthesia_1 indicates intravenous anesthesia, and Anesthesia_2 indicates intravenous inhalational anesthesia. SMD represents the maximum absolute standardized mean difference across all three possible pairwise group comparisons (PCIA vs R, PCIA vs LB, and R vs LB). A Max SMD of < 0.1 is generally considered to indicate an optimal balance of covariates between groups.

Abbreviations: ASA, American Society of Anesthesiologists; BMI, body mass index; IPTW, inverse probability of treatment weighting; LB, liposomal bupivacaine; PCIA, patient-controlled intravenous analgesia; R, ropivacaine; SMD, standardized mean difference.

After IPTW adjustment, all baseline covariates (including age, BMI, ASA physical status, various comorbidities, and type of resection) were balanced among the three groups, with the maximum SMD for all pairwise comparisons being less than 0.1 (Figure 2).

Figure 2.

Love plot of covariate balance before and after inverse probability of treatment weighting. The plot displays the absolute standardized mean differences (SMDs) for all baseline covariates across the three pairwise group comparisons (PCIA vs R, PCIA vs LB, and R vs LB). Red circles represent the unweighted cohort (Before Weighting), and blue triangles represent the weighted pseudo-cohort (After Weighting). The vertical dashed line indicates the reference threshold for covariate balance (absolute SMD = 0.1). Following IPTW adjustment, the absolute SMDs for all covariates converged to < 0.1.

Abbreviations: ASA, American Society of Anesthesiologists; BMI, body mass index; IPTW, inverse probability of treatment weighting; LB, liposomal bupivacaine; PCIA, patient-controlled intravenous analgesia; R, ropivacaine; SMD, standardized mean difference.

Effectiveness Outcomes

Cumulative Pain Burden and Dynamic Changes in NRS Scores

Regarding the primary outcome, the 72-h AUC was calculated in this study to quantify the overall cumulative pain burden experienced by patients. The results showed that the LB group had the lowest AUC, indicating the least cumulative pain burden during the 72-hour postoperative period. Specifically, this AUC was significantly lower than that of the PCIA group (β = −11.01, P = 0.017) and that of the R group (β = −30.19, P < 0.001); meanwhile, the AUC of the PCIA group was also significantly lower than that of the R group (β = −19.19, P < 0.001). The detailed distribution of cumulative pain burden across the three groups is presented in Figure 3.

Figure 3.

Adjusted longitudinal pain trajectories and integrated pain burden over 72 hours postoperatively. Solid lines and points represent the adjusted marginal mean numeric rating scale (NRS) scores at each time point (2h, 24h, 48h, and 72h), estimated by the generalized estimating equations (GEE) model incorporating inverse probability of treatment weighting (IPTW). The shaded areas under the curves (down to the x-axis) provide a visual representation of the cumulative pain burden (area under the curve, AUC) for the PCIA, R, and LB groups. Across the 72-hour observation window, the LB group exhibited the lowest cumulative pain burden compared with the PCIA and R groups.

Abbreviations: LB, liposomal bupivacaine; PCIA, patient-controlled intravenous analgesia; R, ropivacaine.

Based on the GEE model, marginal means at each specific time point were extracted for trend analysis (Figure 3). The results showed that NRS scores at each postoperative time point gradually decreased across all three groups. Comparisons of overall marginal means between groups indicated that the LB group had the lowest overall NRS score (0.66, 95% CI: 0.57–0.76), which was significantly lower than that of the R group (1.07, 95% CI: 0.96–1.18, P < 0.001) and that of the PCIA group (0.80, 95% CI: 0.70–0.89, P = 0.022); in addition, the overall NRS score of the PCIA group was also lower than that of the R group (P < 0.001). The results based on the time‑dimension analysis were consistent with those of the overall cumulative burden, further supporting the differences in overall analgesic efficacy among the groups.

The two sensitivity analyses showed that the LB group consistently had the lowest overall NRS score, lower than that of the PCIA group (P = 0.021 and P < 0.001, respectively, for the two analyses) and that of the R group (both P < 0.001), and the PCIA group also had a lower score than that of the R group (both P < 0.001). The sensitivity analysis results were consistent with the primary analysis, fully supporting the robustness of the main conclusions (Supplementary Table S1).

Pain Assessment at Each Time Point and Postoperative Recovery Outcomes

The group differences in NRS scores at each specific postoperative time point were further evaluated in this study (Table 2, Figure 4). The results showed that, compared with the R group, the LB group had significantly lower NRS scores at 2, 24, 48, and 72 h postoperatively (all P < 0.05). Additionally, the PCIA group also had significantly lower NRS scores than the R group at 24 and 72 h postoperatively (both P < 0.05). At each specific time point, pairwise comparisons between the LB group and the PCIA group showed no statistically significant differences (all P > 0.05). Regarding postoperative rescue analgesia, the overall differences among the three groups in the incidence of rescue analgesia (P = 0.484) and specific drug consumption (P = 0.135) were not statistically significant; however, numerically, the incidence of rescue analgesia in the LB group (11.4%) was lower than that in the R group (13.5%), and among patients who actually received rescue analgesia, the median consumption (15.00 MME) in the LB group was also the lowest among the three groups.

Table 2.

Comparison of Numeric Rating Scale (NRS) Scores at Specific Time Points and Postoperative Recovery Indices Among the three Groups

	PCIA (N = 274)	R (N = 258)	LB (N = 290)	P
NRS scores, Mean (95% CI)
2h	1.52 (1.35, 1.69)	1.78 (1.57, 2.00)	1.45 (1.27, 1.63)*	0.037
24h	0.79 (0.62, 0.95)*	1.14 (0.94, 1.34)	0.60 (0.43, 0.76)***	<0.001
48h	0.61 (0.46, 0.77)	0.78 (0.63, 0.94)	0.43 (0.30, 0.56)***	0.001
72h	0.27 (0.18, 0.36)***	0.56 (0.43, 0.70)	0.18 (0.09, 0.27)***	<0.001
Rescue analgesia (MME)
Incidence	28 (10.2)	35 (13.5)	33 (11.4)	0.484
Amount in users, Median (IQR)	99.00 (6.70, 99.00)	99.00 (15.00, 102.30)	15.00 (6.70, 99.00)	0.135
Amount in users, Mean ± SD	67.20 ± 45.90	82.65 ± 72.60	52.29 ± 52.52
LOS (days)	6.00 (5.00, 7.00)*	6.00 (5.00, 8.00)	6.00 (5.00, 7.00)**	0.006
Time to first ambulation (hours)	21.22 (18.67, 23.83)	21.03 (18.17, 23.75)	20.83 (18.20, 23.22)*#	0.010
Anesthesia recovery time (hours)	0.98 (0.77, 1.20)*	0.92 (0.70, 1.18)	0.87 (0.67, 1.22)	0.039

Note: Data are presented as adjusted marginal means (95% confidence intervals), mean ± standard deviation, median (interquartile range), or number (percentage). Adjusted marginal means of NRS scores at 2, 24, 48, and 72 h postoperatively and corresponding intergroup differences were estimated using a generalized estimating equation (GEE) model with inverse probability of treatment weighting (IPTW). N indicates the weighted sample size. * P < 0.05, ** P < 0.01, *** P < 0.001 versus the R group; ^# P < 0.05 versus the PCIA group.

Abbreviations: CI, confidence interval; IQR, interquartile range; LB, liposomal bupivacaine; LOS, length of stay; MME, morphine milligram equivalent; NRS, numeric rating scale; PCIA, patient-controlled intravenous analgesia; R, ropivacaine; SD, standard deviation.

Figure 4.

Adjusted numeric rating scale (NRS) scores at specific postoperative time points. Data are presented as adjusted marginal means with 95% confidence intervals (error bars), derived from the generalized estimating equations (GEE) model incorporating inverse probability of treatment weighting (IPTW). * P < 0.05, *** P < 0.001 versus the R group. No statistically significant differences were observed between the PCIA and LB groups at any evaluated individual time point.

Abbreviations: LB, liposomal bupivacaine; PCIA, patient-controlled intravenous analgesia; R, ropivacaine.

Regarding postoperative recovery outcomes (Table 2), the time to first ambulation in the LB group was significantly earlier than that in the R group and the PCIA group (both P < 0.05). The overall difference in anesthesia recovery time among the three groups was statistically significant (P = 0.039), with the LB group having the numerically shortest median time (0.87 h). In terms of LOS, both the LB group and the PCIA group had significantly shorter stays than the R group (P < 0.01 and P < 0.05, respectively).

Factors Influencing Postoperative Overall Pain Level

In the univariate analysis evaluating postoperative overall NRS scores, the baseline variables that met the inclusion criterion (P < 0.1) were age, BMI, a history of tumor, and a history of alcohol consumption. After incorporating these variables into the multivariate model, the results showed that, in addition to the analgesic intervention regimen itself, age (β = −0.005, 95% CI: −0.01 to 0.00, P = 0.048) and a history of tumor (β = −0.170, 95% CI: −0.31 to −0.03, P = 0.014) were independently associated with lower postoperative overall NRS scores. In contrast, a history of alcohol consumption (β = 0.096, 95% CI: 0.01 to 0.18, P = 0.030) was independently associated with higher postoperative overall NRS scores. In addition, BMI did not show independent statistical significance after multivariate adjustment (P = 0.074) (Figure 5).

Figure 5.

Forest plot of independent factors associated with overall postoperative pain. Data present the beta coefficients and 95% confidence intervals (CIs) for baseline variables. All results are derived from the weighted multivariate generalized estimating equations (GEE) model. A negative beta coefficient (left of the dashed line) represents a protective factor that lowers pain scores. A positive beta coefficient (right of the dashed line) represents a risk factor that increases pain scores.

Abbreviation: BMI, body mass index.

Safety Outcomes

The main postoperative adverse events in the three groups were PONV and constipation. The incidences of PONV in the PCIA, R, and LB groups were 9.3%, 6.8%, and 7.7%, respectively, and the incidences of constipation were 3.7%, 1.8%, and 1.6%, respectively, with no statistically significant differences among the groups (all P > 0.05). In addition, the incidences of occasional adverse events such as dizziness, fever, hypotension, arrhythmia, and hepatic/renal injury were all below 2.0% in the three groups. Overall, none of the specific adverse events showed statistically significant differences among the three groups (all P > 0.05) (Table 3).

Table 3.

Comparison of the Incidence of Postoperative Adverse Events Among the Three Groups

	PCIA (N = 274)	R (N = 258)	LB (N = 290)	P
Liver injury	1 (0.5)	1 (0.5)	1 (0.2)	0.874
Kidney injury	2 (0.8)	2 (0.7)	-	0.335
PONV	25 (9.3)	18 (6.8)	22 (7.7)	0.583
Fever	1 (0.2)	1 (0.4)	-	0.539
Dizziness	3 (1.0)	2 (0.7)	2 (0.8)	0.911
Constipation	10 (3.7)	5 (1.8)	5 (1.6)	0.252
Hypotension	5 (1.8)	2 (0.8)	1 (0.4)	0.333
Atrial fibrillation	1 (0.4)	1 (0.3)	-	0.578
Oliguria	1 (0.4)	-	-	0.370
Sinus tachycardia	-	-	2 (0.5)	0.337

Note: Data are presented as number (percentage). The P values were derived from the Chi-square test or Fisher’s exact test to compare the incidence rates across the three groups. N indicates the weighted sample size. “–” Indicates no recorded events.

Abbreviations: LB, liposomal bupivacaine; PCIA, patient-controlled intravenous analgesia; PONV, postoperative nausea and vomiting; R, ropivacaine.

Economic Outcomes

Cost Differences

The detailed weighted direct medical costs for each cost component across the three groups are presented in Table 4. In terms of total costs, the PCIA group had the highest mean total cost (USD 598.82), followed by the LB group (USD 562.74), and the R group had the lowest (USD 512.99), with a statistically significant difference among the groups (P < 0.001). However, for the sub-cost components such as basic hospitalization costs and intraoperative anesthesia costs, no statistically significant differences were observed among the three groups (all P > 0.05). The mean total costs of the three groups were used as expected costs for the subsequent CEA.

Table 4.

Comparison of Weighted Direct Medical Costs Among the Three Groups (USD)

	PCIA (N = 274)	R (N = 258)	LB (N = 290)	P
Analgesic intervention cost	91.56 (87.84, 95.29)	0.70	53.64	-
Rescue analgesia cost	0.05 (0.03, 0.06)	0.20 (0.03, 0.37)	0.05 (0.03, 0.07)	0.211
Adverse event management cost	0.06 (0.03, 0.09)	0.09 (0.00, 0.19)	0.04 (0.01, 0.07)	0.509
Intraoperative anesthesia cost	271.81 (267.25, 276.37)	270.98 (266.11, 275.84)	277.39 (273.39, 281.38)	0.078
Basic hospitalization cost	235.34 (225.69, 244.98)	241.02 (230.29, 251.75)	231.62 (221.63, 241.60)	0.451
Total cost	598.82 (586.35, 611.28)	512.99 (500.85, 525.13)	562.74 (551.16, 574.31)	<0.001

Note: Data are adjusted marginal mean costs (95% CIs) derived from generalized linear models (GLM) with inverse probability of treatment weighting (IPTW). For the R and LB groups, analgesic intervention costs are fixed drug unit prices without variance, hence the omission of CIs and P values. N indicates the weighted sample size. All costs are calculated in United States Dollars (USD). “-”, not applicable.

Abbreviations: CI, confidence interval; LB, liposomal bupivacaine; PCIA, patient-controlled intravenous analgesia; R, ropivacaine.

Base-Case Analysis Results

The base-case analysis showed that the LB group had lower total costs and better analgesic efficacy than the PCIA group. Therefore, compared with the PCIA group, the LB group was both less costly and more effective, constituting a dominant strategy.

Compared with the R group, the LB group had an incremental cost of USD 49.75 and an incremental effectiveness of 0.40, yielding an ICER of USD 123.29 per 1-point reduction in the overall NRS score. That is, compared with the R group, the LB group required an additional total cost of USD 123.29 for each additional 1-point reduction in the overall NRS score. In this context, when the WTP threshold was ≤ USD 123.29, the R group was considered the most cost-effective strategy; conversely, when the WTP threshold was > USD 123.29, the optimal strategy shifted to the LB group.

Similarly, compared with the R group, the PCIA group had an incremental cost of USD 85.83 and an incremental effectiveness of 0.27, resulting in an ICER of USD 318.79 per 1-point reduction. Therefore, when the WTP threshold was > USD 318.79, the PCIA group was more cost-effective (Table 5).

Table 5.

Base-Case Results of the Cost-Effectiveness Analysis Among the Three Groups

Group	Cost (USD)	Effect (Point)	Incremental Cost (USD)	Incremental Effect (Point)	ICER (USD/Point)
R	512.99	1.07	-	-	-
PCIA	598.82	0.80	85.83	0.27	318.79
LB	562.74	0.66	49.75	0.40	123.29

Note: Costs are expressed in United States Dollars (USD). Effectiveness is measured as the adjusted overall marginal mean numeric rating scale (NRS) score, where lower scores indicate better analgesia. Incremental costs, incremental effects, and incremental cost-effectiveness ratios (ICERs) for the PCIA and LB groups were calculated with the R group as the reference. ICERs were calculated using unrounded estimates; manual calculations from table values may slightly differ. “-”, not applicable.

Abbreviations: ICER, incremental cost-effectiveness ratio; LB, liposomal bupivacaine; PCIA, patient-controlled intravenous analgesia; R, ropivacaine; USD, United States Dollar.

One-Way Sensitivity Analysis

The OWSA results showed that the parameters with the greatest influence on the ICER were the overall NRS scores of each group, followed by basic hospitalization costs and analgesic intervention costs. In contrast, fluctuations in other parameters had a minor impact on the model. Within the specified parameter variation ranges, the ICER values remained generally stable and consistent with the base-case analysis results, indicating that the base-case conclusions were relatively robust (Figure 6).

Figure 6.

Tornado diagrams of one-way deterministic sensitivity analyses. (A) LB group versus R group. (B) PCIA group versus R group. The vertical line represents the base-case incremental cost-effectiveness ratio (ICER). The horizontal bars indicate the range of ICER values across the plausible variation of each parameter.

Abbreviations: LB, liposomal bupivacaine; NRS, numeric rating scale; PCIA, patient-controlled intravenous analgesia; R, ropivacaine.

Probabilistic Sensitivity Analysis

The PSA results showed that in the comparison between the LB group and the R group, when the WTP was ≤ USD 139.45, the R group had a more than 50% probability of being cost-effective; when the WTP > USD 139.45, the optimal strategy shifted to the LB group. Similarly, in the comparison between the PCIA group and the R group, when the WTP > USD 323.55, the PCIA group became more cost-effective. The PSA results were generally consistent with the base-case analysis results, further confirming the robustness of the base-case conclusions (Figure 7).

Figure 7.

Cost-effectiveness acceptability curves (CEACs) from the probabilistic sensitivity analysis. (A) LB versus R group; (B) PCIA versus R group. Curves show the probability that each analgesic strategy is cost-effective across a range of willingness-to-pay (WTP) thresholds. The intersection points mark the WTP thresholds at which the optimal strategy switches.

Abbreviations: LB, liposomal bupivacaine; PCIA, patient-controlled intravenous analgesia; R, ropivacaine.

Discussion

Adequate and effective analgesia is a critical component of perioperative management and ERAS (enhanced recovery after surgery) practice in thoracic surgery.²³ This study performed a multi-dimensional parallel comparison of three postoperative analgesic regimens—PCIA, LB, and R (ropivacaine) for local infiltration—within a unified clinical setting. The findings showed that the R group had the lowest total direct medical cost but the less favorable analgesic efficacy; the PCIA group offered better analgesic efficacy than the R group but had the highest total cost and had a relatively higher number of adverse events; and the LB group was associated with more favorable analgesic efficacy within 72 h postoperatively and shorter early recovery indicators, with a total cost slightly higher than that of the R group but showed a cost‑effectiveness advantage under certain WTP thresholds. These findings provide real-world evidence to support individualized selection of postoperative analgesic regimens in thoracic surgery.

In the effectiveness evaluation, the LB group had the lowest cumulative pain burden and overall NRS score within 72 h postoperatively, as well as the shortest anesthesia recovery time and time to first ambulation among the three groups. Although there were no statistically significant differences in NRS scores at individual time points between the LB group and the PCIA group, the AUC integrates both pain intensity and duration, thus reflecting the analgesic advantage of LB over the full 72-h observation window. The prolonged analgesic effect of LB is attributable to its multivesicular liposome sustained-release technology: the drug is encapsulated within multilayered lipid vesicles and is released steadily as the vesicles gradually degrade, providing analgesic coverage for approximately 72 h and effectively covering the peak period of acute postoperative pain.⁸ In contrast, the R group had the highest NRS scores at all time points among the three groups, likely because the analgesic duration of a single ropivacaine injection is relatively short and does not adequately cover the continuous analgesia requirements for 72 h postoperatively.¹⁸ It is worth noting that while the absolute NRS differences across groups were relatively small—likely reflecting a floor effect given the adequate baseline analgesia achieved in all groups—the LB group still showed a roughly 17% to 38% relative reduction in pain scores compared with the PCIA and R groups. Evidence from acute pain research suggests that the threshold for meaningful absolute pain reduction is lower when baseline pain is less severe. For instance, Cepeda et al reported that a 20% to 35% relative pain reduction corresponds to patient-rated improvements ranging from “minimal” to “much” better in acute postoperative pain.²⁴ Likewise, a systematic review reported relative MCID (minimum clinically important difference) values for acute pain ranging from 13% to 85%.²² Taken together, the analgesic advantage of LB, though modest in absolute terms, may still be clinically relevant.

In addition to the NRS score benefits, LB was associated with earlier postoperative ambulation and shorter hospital stay. LB may have reduced pain-related limitations on coughing, deep breathing, and positional changes,¹⁸ while avoiding the central sedative effects associated with continuous opioid infusion in PCIA, which may support early mobilization. This is in line with the findings of a randomized controlled trial by Li et al²⁵ in VATS patients, which found that the time to first ambulation in the LB paravertebral block group was significantly shorter than that in the bupivacaine plus dexamethasone group. Regarding the LOS, Banks et al²⁶ found that the LB group had an LOS nearly 22 hours shorter than that of the standard bupivacaine group (34.8 h vs 56.5 h, P = 0.01); similarly, Bauerle et al²⁷ also reported that incorporating LB into the ERAS protocol significantly reduced the LOS. In the present study, the LOS in the LB group was significantly shorter than that in the R group, consistent with the aforementioned findings. In addition, although the overall incidence of rescue analgesia did not differ significantly among the three groups, among patients who required rescue analgesia, the median analgesic consumption in the LB group was the lowest of the three groups, a trend that suggests a possible opioid-sparing effect of LB.

However, the clinical value of LB remains controversial. A meta-analysis by Gong et al¹⁸ indicated that LB did not demonstrate a clear overall clinical advantage over conventional local anesthetics in thoracic surgery; in contrast, a meta-analysis by Hu et al²⁸ focusing on VATS found that LB reduced resting and dynamic pain scores at 24–72 h postoperatively and decreased opioid consumption. The findings of the present study are consistent with clinical application of LB for postoperative analgesia in thoracic surgery. The discrepancies across studies may be attributable to the heterogeneity of the included studies in previous meta‑analyses—the studies varied considerably in terms of surgical procedures, block approaches, and dosing regimens, which may have diluted the potential benefit of LB. In contrast, the present study performed a direct comparison under a unified surgical setting and dosing regimen, adding clarity to the clinical benefit of LB in a specific clinical pathway.

Regarding safety, the predominant postoperative adverse events in all three groups were PONV and constipation, with no statistically significant differences observed in the incidence of any adverse event, suggesting generally comparable safety profiles among the groups. The incidences of PONV and constipation in the PCIA group were numerically higher, a trend consistent with the mechanisms of opioid-induced inhibition of gastrointestinal motility and stimulation of the chemoreceptor trigger zone.^29,30 The opioid-sparing effect of LB may reduce opioid-related adverse events. In addition, although LB and R, as local infiltration regimens, carry theoretical risks of puncture-related complications such as hematoma and pneumothorax, no such complications were observed in this study, suggesting that under standardized operating conditions, local infiltration appears safe. These findings indicate that LB provides better analgesic efficacy without an observed increase in safety risks, further supporting its clinical application in postoperative analgesia for VATS.

The CEA in this study offers economic evidence that may inform the clinical selection of analgesic regimens. Compared with PCIA, LB provided better analgesic efficacy with lower total direct medical costs, constituting a dominant strategy. Compared with R, the least costly regimen, LB had an ICER of USD 123.29 per 1-point reduction in the overall NRS score, suggesting that an additional USD 123.29 was required for each additional 1-point reduction. Lacking a standard WTP threshold for acute pain, these findings should be viewed as relative associations rather than absolute cost-effectiveness. Consequently, clinical decision-making should integrate patients’ analgesic requirement intensity and subjective WTP: for patients at higher risk of pain with more clearly anticipated benefits, the incremental cost of LB is acceptable; for patients with limited budgets or lower analgesic requirements, R remains a reasonable choice. The ICER of PCIA was higher than that of LB, and its burden of opioid-related adverse events further limited its economic value. The OWSA and PSA results further supported the robustness of these conclusions.

Based on the above findings, a stratified decision-making approach according to patient characteristics may be considered. For patients with high analgesic requirements and a greater anticipated risk of pain—such as those with preexisting moderate-to-severe pain or those requiring early functional rehabilitation—LB could be the preferred option, offering a favorable balance between efficacy and cost-effectiveness. PCIA may be an alternative when local infiltration is not feasible (eg, in patients with coagulopathy or infection at the puncture site), although its use requires enhanced monitoring and prevention of opioid-related adverse effects. For patients with higher cost sensitivity or lower analgesic demands, the R regimen provides basic analgesia at the lowest cost and could be a reasonable choice. In addition, the multivariate analysis revealed that a history of alcohol consumption was associated with higher postoperative NRS scores, which may serve as a reference indicator for identifying patients at high risk of pain preoperatively and help inform individualized analgesic decision-making.

Several limitations of this study should be acknowledged. First, as a subjective pain assessment tool, the NRS score may be influenced by patients’ cognitive and emotional status. Second, this study employed a single-center retrospective design. Although IPTW was applied to balance known confounders, residual confounding from unmeasured variables cannot be entirely ruled out, and the generalizability of our findings requires confirmation in future multicenter prospective studies. Third, although the between-group differences in NRS scores reached statistical significance, the absolute values were relatively small. Given that the MCID for acute pain is context-dependent and varies across studies, the clinical relevance of the absolute differences in our study should be interpreted with caution. Fourth, the follow-up period in this study only covered the acute phase within 72 h postoperatively, and long-term outcomes such as chronic post-surgical pain were not assessed; thus, the long-term effects of the analgesic interventions remain to be explored. Fifth, our economic findings should be interpreted as associations rather than definitive cost-effectiveness. Given the absence of a universally accepted willingness-to-pay threshold for a 1-point reduction in the NRS score, and potential regional variations in drug pricing, actual clinical decision-making should integrate local healthcare policies and patients’ financial considerations. Sixth, this study only included patients undergoing elective unilateral VATS; whether the conclusions are applicable to thoracotomy or other thoracic surgical procedures remains to be verified. Future multicenter prospective studies with extended follow-up periods are warranted to further confirm our findings, assess the long-term impact of analgesic interventions, and conduct economic evaluations from a broader societal perspective, which would provide more comprehensive evidence to support clinical decision-making.

Conclusions

In summary, for postoperative analgesia in VATS, despite modest absolute differences in pain scores, LB was associated with favorable analgesic efficacy and early recovery indicators, and may be cost-effective under certain WTP thresholds. PCIA and ropivacaine remain safe and effective alternatives in different clinical scenarios. Future multicenter prospective studies are warranted to further validate these conclusions.

Funding Statement

This work was supported by the National Institute of Hospital Administration [grant numbers NIHAZX202415]; the Jiangsu Provincial Hospital Association [grant number JSYGY-2-2024-YS48]; and Jiangsu Pharmaceutical Association [grant number A202531].

Abbreviations

VATS, video-assisted thoracoscopic surgery; PCIA, patient‑controlled intravenous analgesia; LIA, local infiltration analgesia; LB, liposomal bupivacaine; ASA, American Society of Anesthesiologists; NRS, numeric rating scale; OLV, one-lung ventilation; BIS, bispectral index; IV-MME, intravenous morphine milligram equivalents; AUC, area under the pain score-time curve; GEE, generalized estimating equation; LOS, length of hospital stay; PONV, postoperative nausea and vomiting; CEA, cost-effectiveness analysis; USD, US dollars; GLM, generalized linear model; ICER, incremental cost-effectiveness ratio; IPTW, inverse probability of treatment weighting; OWSA, one–way sensitivity analysis; PSA, probabilistic sensitivity analysis; CIs, confidence intervals; WTP, willingness-to-pay; CEAC, cost-effectiveness acceptability curve; ATE, average treatment effect; BMI, body mass index; SMD, standardized mean difference; ESS, effective sample size; EMMs, estimated marginal means; CLMM, cumulative link mixed model; ERAS, enhanced recovery after surgery.

Data Sharing Statement

Data are available upon reasonable request from the first author Xiangyan Chen, contact: 1638157347@qq.com or the corresponding author Zongling Xia, contact: lysaxia@163.com.

Ethics Approval and Consent to Participate

This study was approved by the Medical Ethics Committee of The First People’s Hospital of Changzhou, with approval number (2025) Section 170. The requirement for informed consent was waived due to the retrospective nature of the study. Accordingly, all clinical data were strictly de-identified and anonymized prior to analysis to ensure patient confidentiality, with no identifiable information disclosed. This study fully respects and protects the autonomy and privacy of the subjects, in line with the requirements of the Declaration of Helsinki.

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.