Comparative perioperative outcomes of robot-assisted versus video-assisted thoracoscopic mediastinal tumor resection: a propensity score-matched analysis Authors: Jiahao Yu, Kaifang Pan, Zhe Chen, Lingyan Jiang, Shiyu Shen, Haifeng Xia Published: 2026 ABSTRACT ObjectivesTo compare perioperative outcomes of robotic-assisted thoracic surgery (RATS) versus video-assisted thoracoscopic surgery (VATS) for mediastinal tumor resection in a propensity score-matched cohort.MethodsThis retrospective study analyzed patients undergoing minimally invasive mediastinal tumor resection between January 2022 and November 2025. Patients were categorized into RATS and VATS groups. Propensity score matching balanced baseline characteristics including age, gender, body mass index, ASA score, Charlson Comorbidity Index, tumor size, tumor location, pathology, and prior thoracic surgery history. The primary outcome was postoperative complications of Clavien-Dindo grade ≥ II. Secondary outcomes included operative time, blood loss, chest tube duration, hospital stay, costs, and pain scores.ResultsAfter matching, 102 patients (51 per group) were included. RATS was associated with significantly less intraoperative blood loss [median difference − 15 mL (95% CI: − 20 to − 10), P < 0.001], shorter chest tube duration [median difference − 1 day (95% CI: − 1.7 to − 0.3), P = 0.003], and shorter hospital stay [median difference − 1 day (95% CI: − 1.7 to − 0.1), P = 0.009]. Total costs were higher in the RATS group [median difference +$4,400 (95% CI: $3,800–$5,000), P < 0.001]. Operative time, postoperative pain scores, and complication rates (grade ≥ II: 9.8% vs. 13.7%, P = 0.55) were comparable between groups. Using an alternative definition (grade ≥ III), complications occurred in 3.9% (RATS) vs. 5.9% (VATS), P = 0.65.ConclusionsFor mediastinal tumor resection, RATS offers modest perioperative benefits including reduced blood loss and shorter recovery times compared to VATS, but at substantially higher cost and without a reduction in major complications. RATS is a feasible and safe minimally invasive alternative, but its advantages over VATS are small and must be weighed against economic considerations. Routine adoption is not supported by the current evidence; selection should be individualized.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s13019-026-04465-3. Introduction The mediastinum, as a complex anatomic compartment, harbors a diverse spectrum of tumors. Surgical resection remains the cornerstone of treatment for most symptomatic or potentially malignant lesions [1]. The evolution from open thoracotomy to minimally invasive techniques, primarily video-assisted thoracoscopic surgery (VATS), has significantly reduced surgical trauma, pain, and recovery time [2]. However, VATS for mediastinal pathology presents unique challenges, including limited instrument maneuverability in a confined space, two-dimensional visualization, and difficult angles for dissection near vital structures like the phrenic nerve, great vessels, and pericardium [3, 4]. The introduction of robotic surgical system (Intuitive Surgical, Inc.) may have alleviated some of these limitations. The features of this system, including high-definition three-dimensional magnification, a seven-degree-of-freedom wrist-like instrument, and an improved ergonomic design, may enhance the precision of fine anatomical operations [5, 6]. In this study, all robotic procedures were performed using a multiportal technique with the Da Vinci Xi system, which, when optimized, can mitigate the arm movement restrictions sometimes associated with single-port robotic configurations. Although robotic-assisted thoracic surgery (RATS) has gained recognition in lung resection surgeries, its role and relative value in mediastinal tumor resection surgeries are still being defined [7, 8]. Prior comparative analyses are frequently constrained by limited sample sizes, inherent selection biases, and heterogeneity in tumor characteristics [9, 10]. Propensity score matching (PSM) reduces confounding bias from measured covariates in observational studies [11]. This study aimed to perform a PSM-based analysis to compare perioperative outcomes of RATS versus VATS for mediastinal tumor resection. Materials and methods Study design and patient selection A single-institution, retrospective cohort study was conducted after obtaining the Ethics Committee of the Fourth Affiliated Hospital of Soochow University approval (No.251282). The requirement for informed consent was waived. Consecutive patients who underwent elective minimally invasive resection of a mediastinal tumor (anterior, middle, or posterior compartment) between January 2022 and November 2025 were screened. The choice between RATS and VATS was not randomized but determined by a combination of factors: surgeon and patient preference, tumor characteristics (size, location, suspected invasion), and operating room availability. While no strict criteria dictated the choice, larger or more complex anterior mediastinal tumors were more frequently selected for RATS, as reflected in the pre-match baseline differences. This clinical decision process introduces the possibility of confounding by indication, which we aimed to mitigate through PSM. Exclusion criteria included: (1) tumors with evidence of direct invasion into surrounding major organs requiring extended resection; (2) planned open surgery or high likelihood of conversion anticipated preoperatively; (3) combined major pulmonary resection; and (4) incomplete medical records. Patients requiring unplanned conversion to thoracotomy were included on an intention-to-treat basis. Surgical techniques All operations were performed by one of two senior thoracic surgeons proficient in both techniques under general anesthesia with double-lumen endotracheal intubation. Both surgeons had performed more than 50 robotic-assisted thoracic procedures and over 100 video-assisted thoracoscopic procedures prior to the start of this study in January 2022, thereby exceeding the initial learning curve for both techniques as defined in the literature. This ensures that the outcomes reflect established proficiency rather than the learning phase for either approach. VATS Group: A standard 1-port or 2-port approach was used. Dissection was performed using standard thoracoscopic instruments. RATS Group: The Da Vinci Xi Surgical System was used. A 3-arm or 4-arm configuration was adopted. Camera port and instrumental ports were placed similarly to VATS but optimized for robotic docking. The robotic arms were managed by the console surgeon. The principles of complete resection (R0) with an intact capsule were identical for both groups. For anterior mediastinal masses suspected to be thymoma, the standard procedure was a complete thymectomy with en-bloc tumor resection. For other anterior, middle, or posterior mediastinal tumors, resection was typically limited to the tumor itself, preserving surrounding structures. Careful preservation of the phrenic nerves and great vessels was performed in all cases. Data collection and outcome measures Preoperative variables included age, gender, body mass index (BMI), American Society of Anesthesiologists (ASA) classification, Charlson Comorbidity Index (CCI), comorbidities, tumor size (largest diameter on CT), tumor location, final pathological diagnosis, and prior thoracic surgery history. The primary outcome measure was the incidence of major postoperative complications (Clavien-Dindo grade ≥ II) within 30 days. Secondary perioperative outcome measures included: (1) operative time (skin incision to skin closure); (2) intraoperative estimated blood loss; (3) duration of chest tube drainage (in days), with removal criteria typically including no air leak, serous output < 100 mL per 24 h, and full lung expansion on chest radiograph; (4) postoperative hospital length of stay; (5) hospitalization expenses, defined as the total direct in-hospital costs from the hospital’s perspective. This included costs for the operating room (encompassing all disposable surgical instruments, including robotic-specific supplies such as the robotic arms, drapes, and other single-use components), anesthesia, ward accommodation, routine medications, and all in-hospital diagnostic tests. Importantly, this analysis does not include the initial capital acquisition cost of the Da Vinci robotic system, instrument reprocessing costs, or long-term maintenance fees, as these represent institutional infrastructure investments rather than direct patient-level costs. Costs were recorded in Chinese Yuan (CNY) and converted to US Dollars (USD) using the March 2026 exchange rate (1 CNY ≈ 0.14 USD). No inflation adjustment was applied as all procedures occurred within a three-year period; and (6) postoperative pain assessed using the Visual Analog Scale (VAS, 0–10) by ward nursing staff on postoperative days 1, 2, and 3 as part of standard clinical care. Propensity score matching and statistical analysis Propensity scores were calculated using logistic regression including: age, gender, BMI, ASA score, CCI, tumor size, tumor compartment, tumor pathology category, and prior thoracic surgery history. Patients were matched 1:1 using nearest-neighbor matching with a caliper of 0.02 of the standard deviation of the logit of the propensity score. Balance was assessed using standardized mean differences (SMD < 0.1 indicating good balance). Missing data were minimal (< 2%) and handled by complete-case analysis. Continuous variables were compared using paired t-test or Wilcoxon signed-rank test; categorical variables using McNemar’s test. Effect sizes with 95% confidence intervals (CIs) are reported alongside P values. Sensitivity analyses included alternative calipers (0.01, 0.05) and covariate-adjusted regression using generalized estimating equations (GEE) on the matched cohort. GEE models used an exchangeable correlation structure to account for clustering within matched pairs. Covariates included all balancing variables. An identity link was used for continuous outcomes and a logit link for binary outcomes, with robust standard errors [12]. Analyses were performed using SPSS Statistics (Version 27.0, IBM Corp.) and R softwareVersion (Version 4.5.1, R Core Team). A two-sided P < 0.05 was considered statistically significant. Results Patient characteristics before and after propensity score matching A total of 158 patients were screened for eligibility. After applying exclusion criteria, 135 patients were included in the final analysis, comprising 59 who underwent RATS and 76 who underwent VATS. A STROBE-compliant patient flowchart is presented in Fig. 1. Fig. 1STROBE Patient Flowchart. Flowchart detailing the selection, exclusion, and propensity score matching of patients undergoing minimally invasive mediastinal tumor resection STROBE Patient Flowchart. Flowchart detailing the selection, exclusion, and propensity score matching of patients undergoing minimally invasive mediastinal tumor resection Prior to matching, significant differences in baseline characteristics were observed. The RATS group had a larger mean tumor size [3.4 (2.5–4.3) cm vs. 2.7 (2.0-3.4) cm, P = 0.002] and a higher proportion of anterior mediastinal tumors (37% vs. 22%, P = 0.039) compared to the VATS group. After 1:1 propensity score matching, 51 well-balanced pairs (n = 102) were generated. Thirty-three patients (8 RATS, 25 VATS) were unmatched, primarily due to differences in tumor size distribution. These patients had smaller tumors on average and were excluded from the primary comparative analysis. All baseline characteristics, including age, gender, body mass index, ASA score, Charlson Comorbidity Index, tumor size, location, pathology, and comorbidities, were comparable between the matched groups, with all standardized mean differences below 0.1 and no statistically significant differences (all P > 0.05) (Table 1). The excellent covariate balance achieved after matching is illustrated in a Standardized Mean Difference (SMD) Balance plot (Fig. 2). The distribution of propensity scores before and after matching showed substantial overlap (Fig. 3). Table 1Patient baseline characteristics before and after propensity score matchingCharacteristicBefore PSMAfter PSMRATS (n = 59)VATS (n = 76)P-valueSMDRATS (n = 51)VATS (n = 51)P-valueSMDAge, years (mean (SD))58.4 (12.1)60.2 (11.5)0.380.1559.1 (11.8)59.7 (11.4)0.790.05Gender, male (%)54% (32)50% (38)0.640.0853% (27)55% (28)0.840.04BMI, kg/m² (mean (SD))26.2 (4.5)25.8 (4.8)0.620.0926.0 (4.6)25.9 (4.7)0.910.02ASA score (I/II) (%)68% (40)63% (48)0.560.1065% (33)67% (34)0.840.04CCI, median (IQR)2 (1–3)2 (1–3)0.710.062 (1–3)2 (1–3)0.890.03Tumor size, cm, median (IQR)3.4 (2.5–4.3)2.7 (2.0-3.4)0.0020.552.4 (1.8–3.1)2.3 (1.7–3.0)0.700.07Tumor locationAnterior mediastinal37% (22)22% (17)0.040.3333% (17)29% (15)0.680.09Middle mediastinal9% (5)8% (6)> 0.990.008% (4)10% (5)0.730.07Posterior mediastinal54% (32)70% (53)0.050.3359% (30)61% (31)0.840.04Tumor pathologyThymoma29% (17)18% (14)0.150.2525% (13)22% (11)0.640.08Thymic carcinoma5% (3)3% (2)0.660.124% (2)4% (2)> 0.990.00Neurogenic tumor32% (19)42% (32)0.290.2135% (18)37% (19)0.840.04Bronchogenic cyst12% (7)14% (11)0.670.0714% (7)12% (6)0.770.06Pericardial cyst7% (4)9% (7)0.760.098% (4)10% (5)0.730.07Other benign15% (9)14% (10)0.820.0514% (7)16% (8)0.800.05ComorbiditiesHypertension42% (25)38% (29)0.620.0841% (21)39% (20)0.850.04Diabetes19% (11)16% (12)0.640.0818% (9)16% (8)0.800.05COPD12% (7)14% (11)0.670.0712% (6)14% (7)0.770.06PSM: propensity score matching; RATS: robotic-assisted thoracic surgery; VATS: video-assisted thoracic surgery; BMI: body mass index; ASA: American Society of Anesthesiologists; CCI: Charlson Comorbidity Index; SMD: standardized mean difference; COPD: chronic obstructive pulmonary disease Table 1Patient baseline characteristics before and after propensity score matchingCharacteristicBefore PSMAfter PSMRATS (n = 59)VATS (n = 76)P-valueSMDRATS (n = 51)VATS (n = 51)P-valueSMDAge, years (mean (SD))58.4 (12.1)60.2 (11.5)0.380.1559.1 (11.8)59.7 (11.4)0.790.05Gender, male (%)54% (32)50% (38)0.640.0853% (27)55% (28)0.840.04BMI, kg/m² (mean (SD))26.2 (4.5)25.8 (4.8)0.620.0926.0 (4.6)25.9 (4.7)0.910.02ASA score (I/II) (%)68% (40)63% (48)0.560.1065% (33)67% (34)0.840.04CCI, median (IQR)2 (1–3)2 (1–3)0.710.062 (1–3)2 (1–3)0.890.03Tumor size, cm, median (IQR)3.4 (2.5–4.3)2.7 (2.0-3.4)0.0020.552.4 (1.8–3.1)2.3 (1.7–3.0)0.700.07Tumor locationAnterior mediastinal37% (22)22% (17)0.040.3333% (17)29% (15)0.680.09Middle mediastinal9% (5)8% (6)> 0.990.008% (4)10% (5)0.730.07Posterior mediastinal54% (32)70% (53)0.050.3359% (30)61% (31)0.840.04Tumor pathologyThymoma29% (17)18% (14)0.150.2525% (13)22% (11)0.640.08Thymic carcinoma5% (3)3% (2)0.660.124% (2)4% (2)> 0.990.00Neurogenic tumor32% (19)42% (32)0.290.2135% (18)37% (19)0.840.04Bronchogenic cyst12% (7)14% (11)0.670.0714% (7)12% (6)0.770.06Pericardial cyst7% (4)9% (7)0.760.098% (4)10% (5)0.730.07Other benign15% (9)14% (10)0.820.0514% (7)16% (8)0.800.05ComorbiditiesHypertension42% (25)38% (29)0.620.0841% (21)39% (20)0.850.04Diabetes19% (11)16% (12)0.640.0818% (9)16% (8)0.800.05COPD12% (7)14% (11)0.670.0712% (6)14% (7)0.770.06PSM: propensity score matching; RATS: robotic-assisted thoracic surgery; VATS: video-assisted thoracic surgery; BMI: body mass index; ASA: American Society of Anesthesiologists; CCI: Charlson Comorbidity Index; SMD: standardized mean difference; COPD: chronic obstructive pulmonary disease Patient baseline characteristics before and after propensity score matching PSM: propensity score matching; RATS: robotic-assisted thoracic surgery; VATS: video-assisted thoracic surgery; BMI: body mass index; ASA: American Society of Anesthesiologists; CCI: Charlson Comorbidity Index; SMD: standardized mean difference; COPD: chronic obstructive pulmonary disease Fig. 2Standardized mean difference (SMD) balance plot. standardized mean differences (SMD) for all covariates before and after propensity score matching. All covariates have an SMD < 0.1 after matching, indicating excellent balance Standardized mean difference (SMD) balance plot. standardized mean differences (SMD) for all covariates before and after propensity score matching. All covariates have an SMD < 0.1 after matching, indicating excellent balance Fig. 3Propensity score distribution. histograms showing the distribution of propensity scores for the RATS and VATS groups before and after matching, demonstrating excellent overlap Propensity score distribution. histograms showing the distribution of propensity scores for the RATS and VATS groups before and after matching, demonstrating excellent overlap Perioperative outcomes Comparative outcomes for the matched cohorts are summarized in Table 2. Meanwhile, Fig. 4 illustrated the key perioperative outcomes that demonstrated statistically significant differences between the RATS and VATS groups. For completeness, perioperative outcomes for the entire unmatched cohort are provided in Table 3; the direction of effects was similar to those observed in the matched analysis, although baseline imbalances preclude direct comparison. Table 2Perioperative outcomes of matched cohortsOutcomeRATS group (n = 51)VATS group (n = 51)p-valueOperative time (min), median (IQR)62 (52–72)60 (50–70)0.42Estimated blood loss (mL), median (IQR)10 (8–15)25 (18–35)< 0.001Chest tube duration (days), median (IQR)3 (2–4)4 (3–5)0.003Postoperative length of stay (days), median (IQR)4 (3–5)5 (4–6)0.009Total costs (USD), median (IQR)$8,200 (7,300-9,500)$3,800 (3,200-4,500)< 0.001Postoperative pain scores (VAS), median (IQR) Day 13 (2–4)3 (2–4)0.72 Day 22 (1–3)2 (2–3)0.41 Day 32 (1–2)2 (1–3)0.38PSM: propensity score matching; RATS: robot-assisted thoracic surgery; VATS: video-assisted thoracic surgery; VAS: visual analog scale Table 2Perioperative outcomes of matched cohortsOutcomeRATS group (n = 51)VATS group (n = 51)p-valueOperative time (min), median (IQR)62 (52–72)60 (50–70)0.42Estimated blood loss (mL), median (IQR)10 (8–15)25 (18–35)< 0.001Chest tube duration (days), median (IQR)3 (2–4)4 (3–5)0.003Postoperative length of stay (days), median (IQR)4 (3–5)5 (4–6)0.009Total costs (USD), median (IQR)$8,200 (7,300-9,500)$3,800 (3,200-4,500)< 0.001Postoperative pain scores (VAS), median (IQR) Day 13 (2–4)3 (2–4)0.72 Day 22 (1–3)2 (2–3)0.41 Day 32 (1–2)2 (1–3)0.38PSM: propensity score matching; RATS: robot-assisted thoracic surgery; VATS: video-assisted thoracic surgery; VAS: visual analog scale Perioperative outcomes of matched cohorts PSM: propensity score matching; RATS: robot-assisted thoracic surgery; VATS: video-assisted thoracic surgery; VAS: visual analog scale Fig. 4Box Plots of key perioperative indicators following PSM. Estimated blood loss (A), chest tube duration (B), postoperative length of stay (C), and total costs (D) are compared between RATS and VATS. Statistically significant differences were observed between the groups. Abbreviation: RATS: robot-assisted thoracic surgery; VATS: video-assisted thoracic surgery; PSM: propensity score matching Box Plots of key perioperative indicators following PSM. Estimated blood loss (A), chest tube duration (B), postoperative length of stay (C), and total costs (D) are compared between RATS and VATS. Statistically significant differences were observed between the groups. Abbreviation: RATS: robot-assisted thoracic surgery; VATS: video-assisted thoracic surgery; PSM: propensity score matching Table 3Perioperative outcomes for the unmatched cohort (n = 135)OutcomeRATS group (n = 59)VATS group (n = 76)p-valueOperative time (min), median (IQR)63 (53–74)59 (49–68)0.08Estimated blood loss (mL), median (IQR)12 (8–18)22 (15–30)< 0.001Chest tube duration (days), median (IQR)3 (2–4)4 (3–6)0.002Postoperative length of stay (days), median (IQR)4 (3–5)5 (4–7)0.004Total costs (USD), median (IQR)$8,450 (7,400-9,800)$3,600 (3,000–4,300)< 0.001Postoperative pain scores (VAS), median (IQR) Day 13 (2–4)3 (2–4)0.68 Day 22 (1–3)2 (2–3)0.35 Day 32 (1–2)2 (1–3)0.42Major complications (Clavien-Dindo ≥ II), n (%)6 (10.2%)11 (14.5%)0.45PSM: propensity score matching; RATS: robot-assisted thoracic surgery; VATS: video-assisted thoracic surgery; VAS: visual analog scale; IQR: interquartile range Table 3Perioperative outcomes for the unmatched cohort (n = 135)OutcomeRATS group (n = 59)VATS group (n = 76)p-valueOperative time (min), median (IQR)63 (53–74)59 (49–68)0.08Estimated blood loss (mL), median (IQR)12 (8–18)22 (15–30)< 0.001Chest tube duration (days), median (IQR)3 (2–4)4 (3–6)0.002Postoperative length of stay (days), median (IQR)4 (3–5)5 (4–7)0.004Total costs (USD), median (IQR)$8,450 (7,400-9,800)$3,600 (3,000–4,300)< 0.001Postoperative pain scores (VAS), median (IQR) Day 13 (2–4)3 (2–4)0.68 Day 22 (1–3)2 (2–3)0.35 Day 32 (1–2)2 (1–3)0.42Major complications (Clavien-Dindo ≥ II), n (%)6 (10.2%)11 (14.5%)0.45PSM: propensity score matching; RATS: robot-assisted thoracic surgery; VATS: video-assisted thoracic surgery; VAS: visual analog scale; IQR: interquartile range Perioperative outcomes for the unmatched cohort (n = 135) PSM: propensity score matching; RATS: robot-assisted thoracic surgery; VATS: video-assisted thoracic surgery; VAS: visual analog scale; IQR: interquartile range Operative time was similar between groups [median difference + 2 min (95% CI: − 3 to + 7), P = 0.42]. Estimated blood loss was significantly lower in RATS [median difference − 15 mL (95% CI: − 20 to − 10), P < 0.001]. Chest tube duration [median difference − 1 day (95% CI: − 1.7 to − 0.3), P = 0.003] and hospital stay [median difference − 1 day (95% CI: − 1.7 to − 0.1), P = 0.009] were shorter in the RATS group. Total costs were significantly higher for RATS [median difference +$4,400 (95% CI: $3,800–$5,000), P < 0.001]. Postoperative pain scores, assessed using the visual analog scale (VAS), were comparable between the two groups on postoperative days 1, 2, and 3 (all P > 0.05). No conversions to open thoracotomy occurred in either group. Detailed postoperative complications are presented in Table 4. The overall incidence of major complications (Clavien-Dindo grade ≥ II) was comparable between groups (9.8% in RATS vs. 13.7% in VATS; P = 0.55). Using grade ≥ III as an alternative definition, rates were 3.9% vs. 5.9%, P = 0.65. While numeric differences were observed in specific complication types, such as pneumonia (3.9% vs. 5.9%) and atrial arrhythmia (2.0% vs. 3.9%), the small event numbers preclude definitive conclusions about procedure-specific risk profiles. One patient in each group required reoperation for bleeding (both grade IIIb). No perioperative mortality occurred in either group. Table 4Detailed postoperative complications (clavien-dindo classification)ComplicationRATS group (n = 51)VATS group (n = 51)Grade IGrade IIGrade IIIaGrade IIIbPneumonia2 (3.9%)3 (5.9%)0500Atrial arrhythmia1 (2.0%)2 (3.9%)0300Chylothorax1 (2.0%)1 (2.0%)0110Reoperation for bleeding1 (2.0%)1 (2.0%)0002Atelectasis requiring bronchoscopy01 (2.0%)0010Wound infection1 (2.0%)00100Prolonged air leak (> 5 days)01 (2.0%)1000Total patients with complications5 (9.8%)7 (13.7%)Some patients experienced more than one complication. Grades are based on the Clavien-Dindo classification system Table 4Detailed postoperative complications (clavien-dindo classification)ComplicationRATS group (n = 51)VATS group (n = 51)Grade IGrade IIGrade IIIaGrade IIIbPneumonia2 (3.9%)3 (5.9%)0500Atrial arrhythmia1 (2.0%)2 (3.9%)0300Chylothorax1 (2.0%)1 (2.0%)0110Reoperation for bleeding1 (2.0%)1 (2.0%)0002Atelectasis requiring bronchoscopy01 (2.0%)0010Wound infection1 (2.0%)00100Prolonged air leak (> 5 days)01 (2.0%)1000Total patients with complications5 (9.8%)7 (13.7%)Some patients experienced more than one complication. Grades are based on the Clavien-Dindo classification system Detailed postoperative complications (clavien-dindo classification) Some patients experienced more than one complication. Grades are based on the Clavien-Dindo classification system Sensitivity analysis results All sensitivity analyses supported the primary findings. Covariate-adjusted regression using generalized estimating equations (GEE) confirmed RATS was independently associated with reduced blood loss (β = −14.8 mL, 95% CI: −19.2 to − 10.4, P < 0.001), shorter chest tube duration (β = −1.0 days, 95% CI: −1.7 to − 0.3, P = 0.004), shorter hospital stay (β = −0.9 days, 95% CI: −1.7 to − 0.1, P = 0.012), and higher costs (β = +$4,438, 95% CI: $3,836 to $5,040, P < 0.001). Matching with alternative calipers (0.01 and 0.05) produced similar results. Exploratory subgroup analyses by tumor location and pathology (Supplementary Tables S1, S2) generally mirrored the main findings but should be interpreted cautiously due to small numbers. Discussion This propensity score-matched comparative study contributes to the accumulating evidence regarding the perioperative outcomes of robotic-assisted mediastinal tumor resection [13, 14]. After balancing baseline characteristics, our analysis indicates that robotic assistance may facilitate more refined dissection, which appears to translate into certain favorable recovery parameters. However, the clinical significance of these statistical differences warrants careful interpretation. The significantly lower intraoperative blood loss observed in the robotic group aligns with the technical features of the system, such as enhanced dexterity, tremor filtration, and three-dimensional visualization, which may allow for more meticulous handling of vascular structures within the mediastinum [6, 15]. Nevertheless, the absolute difference of approximately 15 mL between groups is unlikely to be clinically meaningful for patients. Furthermore, the difference in estimated blood loss may be partly attributable to technical factors in measurement. In VATS, the surgeon often uses the suction device for retraction and dissection, which can aspirate blood and irrigation fluid continuously, leading to a higher volume in the suction canister. In RATS, suction is typically performed on demand by the bedside assistant, and a small amount of blood (10–15 mL) might remain undetected in the pleural cavity, contributing to the lower reported estimate. This potential measurement bias suggests that the observed statistical difference in blood loss may not represent a clinically meaningful advantage. Operative time was comparable between the two groups, suggesting that the additional time required for robotic docking and instrument preparation did not adversely affect early recovery measures, which is consistent with some prior reports [16, 17]. The accelerated recovery trajectory in the RATS group, evidenced by shorter chest tube duration and hospital LOS, may reflect reduced tissue trauma and more precise surgical dissection [18, 19]. However, the median reduction of one day in both parameters, while statistically significant, represents a modest clinical benefit. The absence of a significant difference in early postoperative pain scores aligns with several studies comparing minimally invasive techniques [20, 21], though the underlying mechanisms warrant further investigation. A key strength of this study is the inclusion of tumor pathology in the matching process and its balanced distribution post-match (Table 1). Additionally, the detailed complication reporting in Table 4 provides greater transparency regarding procedure-specific risks. This mitigates a major source of confounding present in many prior series, allowing for a more reliable comparison of surgical techniques independent of tumor biology. The equivalent rate of major complications substantiates the procedural safety of RATS for mediastinal pathology, echoing findings from other comparative series [22]. The principal deterrent for widespread RATS adoption remains its substantial cost increment, approximately double that of VATS in our cohort, a critical factor in resource-conscious healthcare environments [23]. It is important to note that our cost analysis captures direct perioperative patient-level expenses, including robotic-specific disposable supplies, but does not encompass the substantial institutional capital investment required for robotic system acquisition and maintenance. The modest clinical benefits observed must therefore be judiciously weighed against economic implications, necessitating comprehensive cost-effectiveness analyses that incorporate longer-term outcomes and quality of life measures [24]. Furthermore, as a retrospective study with short-term follow-up currently available, we acknowledge the lack of long-term oncological outcomes as a limitation. While the primary goal was perioperative, future studies with extended follow-up are necessary to confirm the long-term oncological efficacy of RATS, especially for malignant pathologies. Future investigations should also incorporate patient-reported outcome measures (PROMs) to capture the patient’s perspective on recovery and quality of life. Our cost estimates are derived from a single Chinese institution using 2022–2025 billing data and reflect local pricing. These figures may not be directly generalizable to other healthcare systems. Nevertheless, the relative difference in costs between RATS and VATS (approximately twofold) is likely to persist across most systems. Importantly, our cost analysis captures direct perioperative patient-level expenses but does not include institutional capital investment for robotic system acquisition, maintenance, or reprocessing, nor long-term societal costs. Therefore, this is not a full cost-effectiveness analysis. In our pre-matching cohort, RATS was preferentially selected for larger tumors and a higher proportion of anterior mediastinal tumors, reflecting a clinical bias toward robotics for anatomically complex cases where enhanced dexterity and 3D visualization were deemed advantageous. These included large thymomas, tumors close to the phrenic nerve or great vessels, and reoperative cases. After PSM, these differences were eliminated, and the observed modest benefits of RATS were evident even in the balanced cohort. This suggests that robotic advantages are not solely driven by case complexity. In our current practice, we consider RATS for any mediastinal tumor where fine dissection near critical structures is anticipated, while VATS remains an excellent choice for small, non-invasive posterior or middle mediastinal lesions. These criteria require validation in prospective studies. Several limitations should be acknowledged. First, the retrospective, single-center design may affect generalizability. our findings should be considered hypothesis-generating. Second, residual confounding from unmeasured variables—such as tumor adhesion, fibrotic changes, exact relation to phrenic nerve or great vessels, and intraoperative judgment of difficulty—cannot be excluded despite PSM. Although we achieved good balance, the possibility of residual confounding from unmeasured variables such as tumor consistency, specific surgeon experience level with each platform during the learning curve, or detailed anatomical factors remains [25, 26]. Third, the heterogeneity of tumor types may obscure technique-specific effects; our sample size did not allow for meaningful subgroup analyses. Fourth, while we excluded cases with missing data, the rate was very low (< 2%). For larger datasets with more substantial missingness, more sophisticated methods like multiple imputation should be considered. Fifth, while necessary for creating balanced cohorts, the exclusion of 33 patients (24% of the original sample) during the PSM process limits the generalizability of our findings. The results presented here are primarily applicable to the subset of patients who were successfully matched and may not represent the entire population of patients undergoing minimally invasive mediastinal tumor resection. Sixth, the unblinded nature of the study is a limitation. Patients and caregivers were aware of the surgical approach, which could have introduced bias into the subjective assessment of postoperative pain (VAS scores) and potentially influenced decisions regarding discharge, despite the use of objective criteria. Seventh, while propensity score matching was used, the relatively modest sample size of the overall cohort (n = 135) and the final matched cohort (n = 102) is a limitation. This sample size may limit the generalizability of our findings and reduces the statistical power to detect clinically meaningful differences in less common outcomes, such as specific complication types (e.g., chylothorax, reoperation rates). Larger multi-center studies are needed to confirm these findings and to adequately power analyses for rare but important adverse events. Finally, the absence of long-term oncological follow-up is a limitation. Future studies should include patient-reported outcomes, long-term survival data, and formal cost-effectiveness analyses with QALYs. Conclusion For mediastinal tumor resection, RATS is associated with modestly reduced blood loss, shorter chest tube duration, and shorter hospital stay compared to VATS, but without a difference in major complications and at significantly higher cost. RATS is a feasible and safe minimally invasive alternative, but its advantages over VATS are small and must be weighed against economic considerations. Routine adoption is not supported by the current evidence; selection should be individualized based on patient anatomy, surgeon expertise, and institutional resources. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1. Supplementary Material 1.