This was a single-center, retrospective cohort study. This study was approved by the Institutional Review Board (IRB) of the Seoul National University Hospital (IRB No. H-2312-069-1491). The requirement for written informed consent from the patients was waived because no personalized data were included. This study was conducted in accordance with the Declaration of Helsinki and adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines [14].

We retrospectively reviewed data from adult patients who were diagnosed with lung cancer and underwent elective VATS between 2018 and 2022. The study included patients who were assessed for surgical site pain in the outpatient clinic between 3 and 6 months after the initial elective VATS and who did not undergo any other thoracic surgery within 6 months after the index surgery. The following patients were excluded: those who received neoadjuvant or adjuvant chemotherapy or radiation therapy between 1 year before and 6 months after surgery; those who were included in the authors’ previous randomized controlled trials (RCTs) on multimodal analgesia [15,16]; those whose anesthesia was maintained with an agent other than sevoflurane or propofol; those who underwent thoracic epidural analgesia; those who received combined operation such as chest wall resection or sternotomy; and those who required reoperation after the index surgery.

The data utilized in this study were obtained through a retrospective review of electronic medical records using the Seoul National University Hospital Patient Research Environment (SUPREME). Demographic data was collected, such as age, sex, weight, height, body mass index (BMI), history of smoking, and American Society of Anesthesiologists (ASA) physical status. The preoperative data included history of analgesic and neuropsychiatric medication use and history of prior thoracic surgery on the same side. In the assessment of the history of analgesic use, tramadol, classified as a weak opioid, was treated as an opioid. In the perioperative phase, the authors collected the name of the surgery, type of anesthetic used, duration of anesthesia (hour), regional analgesia, and intraoperative opioids used. Regional analgesia encompassed the paravertebral block, intercostal nerve block, or wound infiltration. During the study period, intravenous (IV) fentanyl and remifentanil were the only opioids used for intraoperative analgesia at the authors’ institution. Referring to a previous study that converted two opioids into equivalent doses, the total amount was calculated by adding the respective amounts of both opioids administered [17]. This was then divided by the patient's weight and anesthesia duration to get the amount (µg) per minute and body weight. In addition, d the maximum pain intensity assessed using an 11-point numeric rating scale recorded on the day of discharge after surgery, postoperative day of chest tube removal, length of hospital stay, prescription of strong opioids 3 months after discharge from surgery, and records of surgical site pain assessed at the outpatient clinic. No imputation for missing values were performed, nor did the authors rectify any outliers for the variables analyzed.

No pre-medications are routinely administered at the authors’ institution. The attending anesthesiologists determined that the type of general anesthesia administered during the study period was either inhalational anesthesia or propofol-based TIVA. In patients undergoing inhalation anesthesia, a bolus dose of 1.5 to 2 mg/kg of propofol, 2 to 4 µg/kg of fentanyl or target-controlled infusion (TCI) of remifentanil, and 0.6 to 1 mg/kg of rocuronium was administered for anesthetic induction. In patients undergoing propofol-based TIVA, both propofol and remifentanil were administered via TCI for anesthesia induction, followed by a bolus of rocuronium. Anesthesia was maintained with sevoflurane or propofol according to the anesthetic technique employed, under the guidance of an electroencephalogram-based anesthetic depth monitor. Intraoperative analgesic control was achieved by titration of the remifentanil concentration or an additional injection of fentanyl. Based on the judgment of the attending anesthesiologists, IV ketorolac, nefopam, or acetaminophen was administered for postoperative analgesia prior to completion of the surgery.

Fentanyl-based intravenous patient-controlled analgesia (IV-PCA) was employed for postoperative analgesia. Since mid-2019, the authors’ institution has discontinued the use of basal infusion for IV-PCA [18]. Other protocols related to postoperative pain management were consistently applied to all patients included in this study. Water intake and ward ambulation were permitted 6 hours postoperatively, and an oral extended-release 75 mg tramadol/650 mg acetaminophen combination tablet or a 400 mg ibuprofen tablet was routinely administered upon the resumption of water intake. Additionally, IV fentanyl and oral oxycodone were administered as rescue analgesics.

The primary outcome measure was the presence of CPSP, which was assessed by reviewing outpatient electronic medical records between 3 and 6 months after surgery. A patient was considered to have CPSP if they complained of surgical site pain with any level of pain intensity at least once during the observational period. Patients were considered not to have CPSP if all outpatient records during the observational period indicated the absence of surgical site pain. If the surgical site pain was not assessed at any time point, the patient was excluded from the study.

Secondary outcomes included postoperative length of hospital stay and post-discharge opioid prescriptions in outpatient settings. Post-discharge opioid prescriptions were defined as outpatient prescriptions for morphine, oxycodone, hydromorphone, buprenorphine, fentanyl, tapentadol, or tramadol between 3 and 6 months postoperatively.

Continuous variables are presented as means with standard deviations or medians with interquartile ranges, depending on normality, which was evaluated using the Shapiro–Wilk test. Categorical variables are presented as frequencies with percentages. The main statistical approaches for evaluating the association between the type of anesthesia and the occurrence of CPSP included the following analyses.

First, to investigate the association between the type of anesthesia (propofol-based TIVA vs. sevoflurane) and the primary and secondary outcomes, a logistic regression was conducted with adjustment for confounders using stabilized IPTW (sIPTW). The propensity score for each patient was calculated using a logistic regression model that predicted the likelihood of receiving propofol-based TIVA vs. sevoflurane anesthesia, based on the following confounders: sex, age, BMI, ASA physical status, history of analgesic use, history of neuropsychiatric medication use, history of prior thoracic surgery on the same side, duration of anesthesia (hour), extent of surgery (non-anatomical resection such as wedge resection and metastatectomy, segmentectomy, lobectomy, and more extensive procedures such as sleeve lobectomy, bilobectomy, and pneumonectomy), regional analgesia, prolonged chest tube placement, and maximum pain score on discharge day. Prolonged chest tube placement was defined as maintaining the tube for ≥ 4 days postoperatively, based on a previous study that identified it as a risk factor for CPSP after thoracic surgery [19]. Due to significant imbalances between the two anesthesia types, the intraoperative remifentanil equivalent dose could not be included as a confounder in the sIPTW. The weights were calculated as 1/(probability of receiving TIVA) for patients who underwent TIVA and 1/(1-probability of receiving TIVA) for those who did not [20]. These weights were stabilized by multiplying by the proportion of patients in each respective group. Patients with a probability of 0 or 1 for receiving TIVA were excluded based on the positivity assumption. Extreme weights below the 1st percentile or above the 99th percentile were adjusted to the 1st and 99th percentile values, respectively [20]. The balance of variables between the two groups before and after sIPTW was assessed using the standardized mean difference (SMD), with an SMD < 0.10 for all covariates indicating well-balanced groups. Subsequently, a binary logistic regression and t-tests were conducted to investigate the association between TIVA and both primary and secondary outcomes before and after applying sIPTW. Additionally, to consider the imbalance in the intraoperative remifentanil equivalent dose during sIPTW, multivariable logistic regression was conducted, including both the type of anesthesia and a high-dose intraoperative remifentanil equivalent dose (≥ 0.2 µg/kg/min) after applying sIPTW. The cutoff value of high-dose remifentanil (≥ 0.2 µg/kg/min) was determined according to previous studies that reported an association between opioid-induced hyperalgesia and CPSP [21,22].

Second, a sensitivity analysis was performed using a multivariable binary logistic regression analysis for the occurrence of CPSP after VATS. This analysis incorporated the variables used in the propensity score calculation along with the type of anesthesia and high-dose intraoperative remifentanil equivalent dose (≥ 0.2 µg/kg/min). Variables that showed a P value of less than 0.2 in the univariable logistic regression analyses were selected for inclusion in the multivariable analyses. A multivariable regression model was constructed using the selected variables and optimized using backward stepwise selection. The variance inflation factor was used to assess multicollinearity among the variables included in the logistic regression analyses. The Hosmer–Lemeshow test was employed to assess the model's goodness of fit.

Third, interaction analysis was conducted between propofol and female sex, high-dose remifentanil (≥ 0.2 µg/kg/min), maximal pain intensity on the day of discharge, and attending surgeon to investigate how these variables influenced the association between propofol and the occurrence of CPSP.

Finally, to assess the potential effect of unmeasured confounders on the observed association, E-values were estimated. The E-value quantifies the minimum association strength that an unmeasured confounder would need with both treatment and outcome to entirely negate a specific observed association between the treatment and outcome, assuming that all measured covariates are accounted for [23].

Additionally, the authors aimed to investigate the selection bias caused by the exclusion of patients due to the absence of pain records. During the revision period, they conducted a comparative analysis of the baseline characteristics between patients who were excluded from the analysis due to the absence of pain records in the EMRs and those who were included. They also considered the potential influence of attending surgeons on the results, expecting variations in pain documentation at postoperative outpatient clinics.

In the statistical analysis, all tests were carried out using a two-tailed approach, and P values less than 0.05 were deemed to be statistically significant. All statistical analyses were performed using the R language and environment for statistical computing (ver. 4.3.3; R Foundation for Statistical Computing). Additional packages including ‘tableone’ (ver. 0.13.2) and ‘survey’ (4.2.1) were used for conducting sIPTW analysis [24,25], and ‘ggplot2’ (ver. 3.5.0) for generating figures.