Improved 5‐year survival with robot‐assisted resection for locally advanced rectal cancer compared to laparoscopic and open surgery: A real‐world cohort study

Abstract

Aim

Robot‐assisted rectal cancer surgery offers short‐term benefits; however, data on its long‐term outcomes are lacking. In this study, we aimed to compare the short‐ and long‐term outcomes of open, laparoscopic and robot‐assisted rectal resection (ORR, LRR and RARR) for locally advanced rectal cancer using a large real‐world database.

Methods

This multi‐centre retrospective cohort study analysed 37,191 patients with clinical T3/T4a rectal cancer from the Japanese Medical Data Vision database (2018–2024). After overlap weighting for baseline balance, the effective sample size was 14,627. The primary outcomes were 5‐year overall survival (OS) and relapse‐free survival (RFS). The secondary outcomes included peri‐operative outcomes.

Results

After overlap weighting, the total number of patients was 2247, 10,339 and 2041 in the RARR, LRR and ORR groups, respectively. The mean age was 70 years, and 66% were male. The RARR group had the lowest post‐operative complication rate (17%), shortest duration of hospital stay (16 days) and lowest 90‐day mortality rate and total hospitalisation cost; and the highest 5‐year OS (94%) and 5‐year RFS (93%) compared with the LRR (5‐year OS: 86%, 5‐year RFS: 83%; p < 0.001) and ORR (5‐year OS: 78%, 5‐year RFS: 74%; p < 0.001) groups. RARR was significantly associated with improved OS, whereas higher risks were observed for LRR (hazard ratio [HR]: 2.50) and ORR (HR: 4.69).

Conclusion

Robot‐assisted surgery demonstrated superior short‐ and long‐term outcomes and the lowest total hospitalisation cost compared with other approaches. These results demonstrate the potential of robot‐assisted surgery as a new standard of treatment for locally advanced rectal cancer.

What does this paper add to the literature?

Our findings provide large‐scale, database‐driven evidence of the significant short‐ and long‐term survival benefits of robot‐assisted surgery for locally advanced rectal cancer in Japan, underscoring the potential of robot‐assisted surgery as a new standard for locally advanced rectal cancer treatment and providing critical insights that may shape future surgical approaches.

INTRODUCTION

Conventional laparoscopic surgery is an effective treatment for colorectal cancer; however, the long‐term oncological outcomes for rectal cancer are similar to those of open surgery [1, 2, 3]. Moreover, laparoscopic surgery can lead to more incomplete mesorectal excisions [4]. Robot‐assisted surgery represents the latest advancement in minimally invasive techniques and offers several advantages over conventional laparoscopic surgery.

The utilisation of robot‐assisted surgery for rectal cancer has been increasing globally. However, although extensive research has demonstrated its superiority in short‐term outcomes compared to laparoscopic and open techniques, robust data on its long‐term oncological efficacy remain scarce. Among three randomised controlled trials (RCTs) [5, 6, 7] that evaluated the oncological outcomes of robot‐assisted and laparoscopic approaches, only the REAL trial [7] demonstrated the advantages of robot‐assisted surgery in achieving complete total mesorectal excision (TME) and reducing the rate of circumferential resection margin (CRM) positivity. Similarly, a retrospective study showed that robot‐assisted TME is beneficial for improving long‐term outcomes of patients with locally advanced rectal cancer (T3/T4) [8], which has a high risk of local recurrence. In contrast, the COLRAR [6] and ROLARR trials [5] reported comparable outcomes, making the oncological benefits a matter of ongoing debate.

Studies based on database analyses have been gaining increasing prominence in recent years. The Japanese Medical Data Vision (MDV; Medical Data Vision Co. Ltd., Japan) is a comprehensive, large‐scale, real‐world claims database containing anonymised demographic and diagnostic patient information. Previous studies utilising similar databases, such as the National Clinical Database or Diagnosis Procedure Combination (DPC), have been reported [9, 10]; however, these studies only focused on short‐term outcomes.

With the global surge in robot‐assisted rectal cancer surgery [11], robust evidence of its long‐term outcomes is imperative. Clinical T3 and T4a rectal cancer cases are frequently encountered in daily practice and present a higher risk of local recurrence. These cases demand particular attention, as surgical techniques can critically influence prognosis [7]. Therefore, we aimed to compare the short‐ and long‐term clinical and oncological outcomes of open, laparoscopic, and robot‐assisted approaches in patients with locally advanced rectal cancer within a large, diverse cohort.

PATIENTS AND METHODS

Data source

In this retrospective cohort analysis using data from the MDV database, we focused on patients diagnosed with clinical stages I–III rectal cancer, specifically those with cT3 or cT4a who underwent rectal resection between April 2018 and June 2024. The MDV database comprises de‐identified inpatient and outpatient administrative claims and Diagnosis Procedure Combination (DPC) data from acute care Japanese hospitals that treat severe illnesses for brief periods. The DPC is a case mix classification system linked to a flat‐fee payment system. As of March 2023, the MDV database included approximately 43.2 million patients from 475 acute‐phase DPC hospitals (approximately 27% of DPC hospitals in Japan). Among these 475 hospitals, 236 were designated as cancer care hospitals. These hospitals (classified by prefecture, community, or locality) are designated by the Japan Ministry of Health, Labour and Welfare to provide specialised treatments and care to patients with cancer, co‐ordinate and co‐operate with each other and act as information centres for patients. Robot‐assisted rectal resection (RARR) became eligible for reimbursement in April 2018, enabling its identification in the MDV dataset from that point onward.

Study design

This retrospective, multi‐centre observational study utilised data from a hospital‐based administrative claims database. The International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD‐10‐CM) code C20 was used to identify cases of malignant neoplasm of the rectum. [12] Tumour staging was based on the American Joint Committee on Cancer‐Union for International Cancer Control Tumour, Node, Metastasis (AJCC–UICC TNM) staging system (8th edition). [13]

In cases where multiple tumours were diagnosed simultaneously, the tumour with the highest TNM stage was selected for analysis. Patients with incomplete medical records or those who underwent concurrent surgeries were excluded. The rectal resections analysed in this study were defined as low anterior resection (LAR) and abdominoperineal resection (APR), according to Japanese insurance criteria [12]. RARR, laparoscopic rectal resection (LRR), and open rectal resection (ORR) were also identified using the 10th Revision (ICD‐10) code C20 [12]. If a laparoscopic or robot‐assisted procedure was converted to an open resection, the case was considered as having undergone open resection. The dataset did not include information on tumour location (measured in centimetres from the anal verge), extent of lymph node dissection or creation of diverting stomas. Only the first diagnosed tumour was included in the analysis of patients diagnosed with multiple primary colorectal tumours.

Data collected

The following data were collected and used as baseline characteristics for the analysis: sex, age, body mass index (BMI), Charlson Comorbidity Index (CCI), clinical TNM stage, neoadjuvant and adjuvant therapy status, type of surgery, hospital category, total hospitalisation cost and follow‐up duration. Post‐operative complications included anastomotic leakage, septic shock, peritonitis, intra‐abdominal bleeding, ileus, intestinal ischaemia, surgical site infection, dysuria, urinary tract infection, myocardial infarction, pulmonary embolism, pneumonia, venous thrombosis and cerebrovascular disease. These complications were identified using ICD‐10‐CM codes.

Study outcomes

The primary outcomes were 5‐year overall survival (OS) and 5‐year relapse‐free survival (RFS) among the three surgical approaches. The secondary outcomes were short‐term outcomes, including 30‐day complication rates, 30‐day re‐operation rates, 30‐day readmission rates, 30‐ and 90‐day mortality rates, total hospitalisation cost, peri‐operative outcomes and 5‐year local recurrence.

Statistical analysis

The 5‐year cumulative incidences of mortality and metachronous recurrence were also estimated. The time‐to‐event was measured from the date of surgery to the last follow‐up or death (OS) and recurrence (RFS). Baseline characteristics were compared between groups using pairwise comparisons. Continuous variables are summarized as means with standard deviations (SD) or medians with interquartile ranges (IQR), while categorical variables are presented as counts and percentages. Statistical tests, including Welch's t‐test or the Mann–Whitney U test, chi‐squared test or Fisher's exact test, were applied as appropriate, depending on the data distribution.

Covariates included patient background factors, clinical data (CCI, TNM stage and neoadjuvant therapy) and surgical factors (surgery type, hospital category and follow‐up duration). The CCI was calculated using the ICD‐10 codes to assign comorbidity scores to each patient.

To adjust for baseline differences across the three surgical modalities (robot‐assisted, laparoscopic and open), generalized propensity scores were estimated using multinomial logistic regression including the following pre‐treatment covariates: age, sex, BMI category, CCI, clinical TNM stage (T3 vs. T4a; N0/N1/N2), neoadjuvant therapy (yes/no), index surgery type (LAR vs. APR), hospital category (university vs. non‐university) and follow‐up duration.

Next, overlap weighting (OW) was applied to target the average treatment effect in the overlap population (ATO). OW down‐weights extreme propensities, achieves excellent mean balance on observed covariates and typically yields a larger effective sample size than inverse‐probability weighting or matching in multi‐arm settings [14, 15]. We report the effective sample size after weighting.

Covariate balance after OW was assessed using absolute standardised mean differences (ASMDs), with ASMD <0.10 indicating adequate balance; we also examined propensity score overlap across arms and observed sufficient overlap (no practical positivity violations).

To control the false discovery rate, the Benjamini–Hochberg procedure [16] was applied to adjust p‐values, maintaining a Type I error rate of 0.05. [17] As three independent hypothesis tests were conducted, each with a Type I error rate of 0.05, the cumulative error rate was approximately 0.143. The Benjamini–Hochberg method addresses this by ranking p‐values and comparing them against increasing thresholds, striking a balance between controlling false‐positives and retaining statistical power.

For survival outcomes (OS and RFS), Kaplan–Meier curves and log‐rank tests were computed with overlap weights. Cox proportional hazards models were also fit with overlap weights to identify predictors of survival. We also present a multivariable overlap‐weighted Cox model, with covariates entered if they met p < 0.05 in univariate overlap‐weighted Cox screening.

As a sensitivity analysis, we performed stabilized inverse‐probability of treatment weighting (IPTW) based on the same multinomial propensity score to target the average treatment effect, followed by IPTW‐weighted Kaplan–Meier curves, weighted log‐rank tests and IPTW‐weighted Cox models.

All statistical tests were two‐sided, with an alpha level of 0.05; p‐values < 0.050 were considered statistically significant. Statistical analyses were performed using R software (version 4.3.1; R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Baseline characteristics

The patient flowchart is outlined in Figure 1. Among the 64,292 patients in the MDV database who underwent rectal resection between April 2018 and June 2024, 37,191 patients with primary rectal cancer were screened. After excluding patients with Stage 0 (n = 347) and Stage IV disease (n = 2567), clinical T1 (n = 5095), clinical T2 (n = 5929), clinical T4b (n = 2066), clinical TX (n = 1771), unknown clinical stage (n = 1623), incomplete medical records (n = 388) and those who underwent concurrent surgery (n = 444), the final cohort included 16,961 patients with T3/T4a rectal cancer: 3045, 11,426 and 2490 underwent RARR, LRR and ORR, respectively.

FIGURE 1.

Patient recruitment flow chart. LRR, laparoscopic rectal resection; ORR, open rectal resection; RARR, robot‐assisted rectal resection.

Before OW, the three groups showed differences regarding age, BMI, clinical stage, type of surgery and hospital category (Table S1). After OW, the effective sample size was 14,627 patients, including 2247, 10,339 and 2041 in the RARR, LRR and ORR groups, respectively. Table 1 shows that after OW, all baseline characteristics were balanced across the three groups (ASMD < 0.1; p > 0.05) (Figure 2). In the weighted cohort, the mean age was 70.06 years, 66% of patients were male, and most patients (64.2%) had a normal BMI; the average CCI was 1.29. Neoadjuvant therapy was administered to 13.7% of patients. The most common surgical procedure was LAR, performed in 81.5% of cases. Additionally, 91.7% of surgeries were not performed in university hospitals. The median follow‐up duration was 20.68 months.

TABLE 1.

Clinical T3/T4a baseline characteristics after overlap weighting.

Characteristics	Summary statistics			p‐value	ASMD
	RARR a (n = 2247)	LRR a (n = 10,339)	ORR a (n = 2041)		RARR a (n = 2247)	LRR a (n = 10,339)	ORR a (n = 2041)
Sex
Female (%)	34.55%	33.80%	34.77%	0.607	0.016	0.005	0.020
Male (%)	65.45%	66.20%	65.23%		0.016	0.005	0.020
Age (mean SD)	70.16 (10.91)	70.03 (11.22)	70.15 (11.48)	0.826	0.012	<0.001	0.011
BMI classification b
Normal (%)	64.02%	64.23%	64.21%	0.926	0.004	0.004	0.001
Underweight (%)	14.12%	13.59%	14.14%		0.015	0.001	0.016
Obese (%)	21.86%	22.17%	21.65%		0.007	0.005	0.013
CCI (mean SD)	1.31 (2.03)	1.27 (1.88)	1.33 (2.00)	0.760	0.020	0.010	0.031
Clinical stage
II (%)	45.65%	45.66%	44.53%	0.637	<0.001	0.022	0.023
III (%)	54.35%	54.34%	55.46%		<0.001	0.022	0.023
Clinical T stage
T3 (%)	81.49%	82.14%	81.52%	0.666	0.017	0.001	0.016
T4a (%)	18.51%	17.86%	18.48%		0.017	0.001	0.016
Clinical N stage
N0 (%)	45.65%	45.66%	44.53%	0.723	<0.001	0.022	0.023
N1 (%)	36.03%	35.47%	35.59%		0.012	0.009	0.002
N2 (%)	18.32%	18.86%	19.88%		0.014	0.040	0.026
Neoadjuvant therapy (%)	14.12%	14.47%	14.70%	0.856	0.010	0.017	0.007
Type of surgery
LAR (%)	81.66%	81.43%	81.37%	0.961	0.006	0.008	0.002
APR (%)	18.34%	18.57%	18.63%		0.006	0.008	0.002
Hospital category
University (%)	8.08%	8.31%	8.39%	0.883	0.008	0.011	0.003
Non‐University (%)	91.92%	91.69%	91.61%		0.008	0.011	0.003
Follow‐up months (Median IQR)	21.20 (9.35–33.84)	20.79 (8.52–35.81)	19.58 (8.17–35.64)	0.400	0.060	0.037	0.021

Abbreviations: APR, abdominoperineal resection; ASMD, absolute standardised mean difference; BMI, body mass index; CCI, Charlson comorbidity index; IQR, interquartile range; LAR, low anterior resection; LRR, laparoscopic rectal resection; ORR, open rectal resection; OW, overlap weighting; RARR, robot‐assisted rectal resection; SD, standard deviation.

^a Approximate effective sample size.

^b The Japan Society for the Study of Obesity (JASSO) defines BMI classifications as follows: underweight <18.5 kg/m²; normal 18.5 to <25 kg/m²; obese ≥ 25 kg/m².

FIGURE 2.

Love plot of pre−/post‐ overlap weighting balance. HPCATEGORY, Hospital category; TRACKING_MONTHS, Follow‐up months; T_group_T4, Clinical T stage, T4a; PROCEDURENAME_LAR, Type of surgery, low anterior resection; BMI_group_Underweight; Obese; Normal, body mass index, underweight <18.5 kg/m²; obese ≥25 kg/m²; normal 18.5 to <25 kg/m²; N_status N2; N‐; N1, Clinical N stage;N2; N0; N1; SEX_F, SEX, Female; CCI, Charlson comorbidity index.

Short‐term outcomes

The short‐term outcomes after applying OW are summarised in Table 2. The RARR group demonstrated the lowest rate of overall post‐operative complications (RARR: 16.54%, LRR: 19.95%, ORR: 29.68%; p < 0.001) and intra‐operative blood transfusion (RARR: 6.04%, LRR: 9.85%, ORR: 27.63%; p < 0.001). The RARR group also had the shortest duration of hospital stay (RARR: 15.69 [SD: 10.71] days, LRR: 18.87 [SD: 15.98] days, ORR: 25.38 [SD: 34.02] days; p < 0.001). Regarding the total hospitalisation cost to undergo rectal resection from admission to discharge, the RARR group had the least costly average healthcare cost (RARR: ¥1,849,029 [SD: ¥434,609], LRR: ¥1,934,626 [SD: ¥631,546], ORR: ¥2,012,968 [SD: ¥958,923]). Additionally, the average 90‐day mortality rate (RARR: 0.23%, LRR: 0.70%, ORR: 1.21%; p < 0.001) was significantly lower in the RARR than the LRR and ORR groups. Moreover, the RARR group demonstrated a tendency toward lower incidence rates of peritonitis, ileus, surgical site infection, dysuria, urinary tract infection and 30‐day mortality compared with other approaches.

TABLE 2.

Clinical stage T3/T4a short‐ and long‐term outcomes after overlap weighting.

Clinical outcome	Summary statistics			p‐value	FDR corrected p‐value
	RARR a (n = 2247)	LRR a (n = 10,339)	ORR a (n = 2041)	Overall	RARR‐ LRR	RARR‐ ORR	LRR‐ORR
Post‐operative complications (%)	16.54%	19.95%	29.68%	<0.001	<0.001	<0.001	<0.001
Anastomotic leakage (%)	3.06%	3.66%	3.90%	0.281	0.280	0.280	0.564
Septic shock (%)	0.00%	0.00%	0.00%	1.000	1.000	1.000	1.000
Peritonitis (%)	2.27%	2.75%	7.24%	<0.001	0.219	<0.001	<0.001
Intra‐abdominal bleeding (%)	0.14%	0.14%	0.10%	1.000	1.000	1.000	1.000
Ileus (%)	4.95%	5.26%	10.56%	<0.001	0.564	<0.001	<0.001
Intestinal ischaemia (%)	0.34%	0.23%	0.35%	0.369	0.527	1.000	0.527
Surgical site infection (%)	2.62%	3.20%	5.60%	<0.001	0.159	<0.001	<0.001
Dysuria (%)	3.55%	5.84%	5.82%	<0.001	<0.001	0.001	1.000
Urinary tract infection (%)	2.04%	2.56%	3.97%	<0.001	0.177	0.001	0.001
Myocardial infarction (%)	0.11%	0.07%	0.05%	0.510	0.940	0.940	1.000
Pulmonary embolism (%)	0.13%	0.14%	0.28%	0.239	1.000	0.486	0.374
Pneumonia (%)	0.44%	0.94%	1.71%	<0.001	0.022	<0.001	0.005
Venous thrombosis (%)	0.64%	0.81%	1.05%	0.342	0.427	0.427	0.427
Cerebrovascular disease (%)	0.32%	0.35%	0.61%	0.154	1.000	0.265	0.237
Intra‐operative blood transfusion (%)	6.04%	9.85%	27.63%	<0.001	<0.001	<0.001	<0.001
Length of stay (mean SD)	15.69 (10.71)	18.87 (15.98)	25.38 (34.02)	<0.001	<0.001	<0.001	<0.001
Cost, JPY (mean SD)	1849029 (434609)	1934626 (631546)	2012968 (958923)	<0.001	<0.001	<0.001	<0.001
Unplanned readmission (%)	4.90%	3.94%	6.07%	<0.001	0.060	0.093	<0.001
Post‐operative adjuvant therapy (%)	41.42%	46.68%	49.04%	<0.001	<0.001	<0.001	0.052
30‐day re‐operation	1.58%	2.43%	1.58%	0.004	0.028	1.000	0.028
30‐day mortality (%)	0.11%	0.39%	0.56%	0.064	0.106	0.087	0.343
90‐day mortality (%)	0.23%	0.70%	1.21%	<0.001	0.010	<0.001	0.020
3‐year OS b (95% CI)	97% (96–98%)	92% (91–93%)	86% (85–88%)	<0.001	<0.001	<0.001	<0.001
5‐year OS b (95% CI)	94% (91–97%)	86% (85–88%)	78% (75–81%)	<0.001	<0.001	<0.001	<0.001
3‐year RFS b (95% CI)	94% (93–95%)	89% (88–90%)	83% (81–85%)	<0.001	<0.001	<0.001	<0.001
5‐year RFS b (95% CI)	93% (91–95%)	83% (81–84%)	74% (71–77%)	<0.001	<0.001	<0.001	<0.001
3‐year local recurrence (%)	2.26%	3.21%	4.16%	0.002	0.026	0.001	0.032
5‐year local recurrence (%)	2.26%	3.49%	4.63%	<0.001	0.004	<0.001	0.017

Abbreviations: 95% CI, 95% confidence interval; FDR, false discovery rate; FDR, false discovery rate; LRR, laparoscopic rectal resection; ORR, open rectal resection; OS, overall survival; OW, overlap weighting; RARR, robot‐assisted rectal resection; RFS, relapse free survival; SD, standard deviation.

^a Approximate effective sample size.

^b Log‐rank test.

Long‐term outcomes

The 5‐year OS rate was highest in the RARR (94% [95% CI: 91%–97%]) than in LRR (86% [95% CI: 85%–88%]) and ORR (78% [95% CI: 75%–83%]) groups (Table 2). The Kaplan–Meier curves demonstrated that the RARR group consistently had the highest survival rates throughout the follow‐up period, followed by the LRR and ORR groups (Figure 3). The 5‐year RFS was also highest in the RARR, followed by LRR and ORR groups (RARR: 93% [95% CI: 91–95], LRR: 83% [95% CI: 81–84], ORR: 74% [95% CI: 71–77]), demonstrating that RARR consistently yielded the best outcomes (Figure 4). Additionally, the 5‐year local recurrence rate was lowest in the RARR, followed by the LRR and ORR groups (RARR: 2.26%, LRR: 3.49%, ORR: 4.63%; p < 0.001; Table 2).

FIGURE 3.

Kaplan–Meier curve of overall survival for clinical T3/T4a following OW. LRR, laparoscopic rectal resection; ORR, open rectal resection; OW, overlap weighting; RARR, robot‐assisted rectal resection.

FIGURE 4.

Kaplan–Meier curve of relapse‐free survival for clinical T3/T4a following OW. LRR, laparoscopic rectal resection; ORR, open rectal resection; OW, overlap weighting; RARR, robot‐assisted rectal resection.

Risk predictors for OS

Table 3 shows the univariable and multivariable Cox regression analyses of factors associated with poorer OS. Multivariable analysis identified LRR (HR: 2.50; p < 0.001), ORR (HR: 4.69; p < 0.001), male sex (HR: 1.54; p < 0.001), an age ≥75 years (HR: 2.54; p < 0.001), underweight (HR: 2.24; p < 0.001), a CCI ≥2 (HR: 1.20; p = 0.011), cStage III (HR: 1.89; p < 0.001), APR (HR: 1.57; p < 0.001) and treatment at non‐university hospitals (HR: 3.53; p < 0.001) as significant factors of poor survival. Conversely, RARR was associated with significant protective effects on OS.

TABLE 3.

Clinical stage T3/T4a Cox regression analysis of overall survival.

Factor	Univariable			Multivariable
	HR	95% CI	p‐value	HR	95% CI	p‐value
Surgical approach
RARR	Reference	—	—	Reference	—	—
LRR	2.45	1.82–3.30	<0.001	2.50	1.85–3.37	<0.001
ORR	4.52	3.30–6.18	<0.001	4.69	3.42–6.43	<0.001
Sex
Female	Reference	—	—	Reference	—	—
Male	1.35	1.16–1.59	<0.001	1.54	1.31–1.81	<0.001
Age
<75	Reference	—	—	Reference	—	—
≥75	2.45	2.13–2.82	<0.001	2.54	2.21–2.93	<0.001
BMI classification a
Normal	Reference	—	—	Reference	—	—
Underweight	2.01	1.67–2.42	<0.001	2.24	1.86–2.70	<0.001
Obese	0.89	0.74–1.06	0.184	0.97	0.81–1.16	0.708
CCI
0~1	Reference	—	—	Reference	—	—
≥2	1.24	1.08–1.44	0.003	1.20	1.04–1.39	0.011
Clinical stage
II	Reference	—	—	Reference	—	—
III	1.72	1.49–2.00	<0.001	1.89	1.64–2.19	<0.001
Neoadjuvant therapy
No	Reference	—	—	—	—	—
Yes	1.14	0.93–1.39	0.201	—	—	—
Post‐operative adjuvant therapy
No	Reference	—	—	—	—	—
Yes	1.11	0.97–1.28	0.139	—	—	—
Type of surgery
LAR	Reference	—	—	Reference	—	—
APR	1.69	1.43–1.99	<0.001	1.57	1.33–1.86	<0.001
Hospital category
University	Reference	—	—	Reference	—	—
Non‐University	3.20	2.16–4.73	<0.001	3.53	2.37–5.24	<0.001

Note: Bold values indicate statistically significant differences.

Abbreviations: APR, abdominoperineal resection; BMI, body mass index; CCI, Charlson comorbidity index; CI, confidential interval; LAR, low anterior resection; LRR, laparoscopic rectal resection; ORR, open rectal resection; RARR, robot‐assisted rectal resection.

^a The Japan Society for the Study of Obesity (JASSO) defines BMI classifications as follows: underweight<18.5 kg/m²; normal 18.5 to <25 kg/m²; obese ≥25 kg/m².

Sensitivity analysis

Sensitivity analyses using stabilised IPTW yielded effect estimates that were directionally identical and similar in magnitude to the primary overlap‐weighted results. IPTW‐weighted Kaplan–Meier curves, log‐rank tests and Cox models yielded similar qualitative conclusions for OS and RFS, as well as the key short‐term outcomes; no material differences in inference were observed. Full IPTW outputs and balance diagnostics are provided in the Supplementary Materials (Tables S2–S5, Figures S1 and S2).

DISCUSSION

This study represents the first large‐scale, real‐world analysis demonstrating significant long‐term survival benefits of robot‐assisted surgery over laparoscopic and open approaches in the treatment of locally advanced rectal cancer. Utilizing a robust national database, we found that robot‐assisted surgery yielded the highest overall survival, followed by laparoscopic and then open surgery. In addition to improved long‐term outcomes, the robot‐assisted approach was associated with favourable short‐term outcomes, including lower intraoperative transfusion rates, fewer postoperative complications, shorter hospital stays, lower 90‐day mortality and reduced total hospitalisation costs—highlighting its effectiveness in optimizing perioperative safety and long‐term prognosis.

To date, RCTs have not yet reported long‐term outcomes comparing robot‐assisted surgery with other approaches. While previous retrospective studies and meta‐analyses [18] have shown no significant differences, the REAL trial [7] has reported advantages of robot‐assisted surgery in achieving a more complete TME and reducing the rate of CRM positivity compared with laparoscopic surgery. Given that CRM positivity increases the risk of local and distant recurrence up to fourfold [19, 20], improved CRM status with the robot‐assisted approach likely contributes to better long‐term outcomes. Most recently, the American Society of Clinical Oncology 2024 presentation of the REAL trial [21] demonstrated significantly higher 3‐year recurrence‐free survival with robot‐assisted surgery versus laparoscopic surgery (87.3% vs. 83.6%, log‐rank p = 0.035), further supporting the long‐term oncological benefits observed in our study.

Previous reports have emphasised the technical complexity of treating locally advanced rectal cancer and the potential advantages of the robot‐assisted approach. One study [8] attributed better long‐term outcomes in T3/T4 cases to the ability of robot‐assisted surgery to facilitate meticulous, gentle dissection—minimising both surgical difficulty and the potential for microscopic cancer cell dissemination while securing wider resection margins. Other studies have also suggested that surgical stress responses, which influence tumour cell progression, differ by surgical modality [22], with robot‐assisted surgery inducing lower systemic stress than open surgery [23]. These findings reinforce our results and support the oncologic rationale for the robot‐assisted approach in T3/T4a disease.

Postoperative complications have been identified as independent prognostic factors for poor disease‐free survival [24], suggesting that the reduced complication rate in the robot‐assisted group may have contributed to lower short‐term mortality and improved long‐term survival, including the lowest local recurrence rates.

No RCTs have directly compared long‐term outcomes between robot‐assisted and open approaches in rectal cancer surgery, nor have any studies demonstrated a clear advantage of robot‐assisted surgery [25, 26]. The potential benefits of this point can be attributed to precise manoeuvres that could help enhance oncological safety.

Several trials have evaluated laparoscopic versus open surgery in this setting. Notably, the ALaCaRT [27] and ACOSOG Z6051 [19] trials failed to establish the non‐inferiority of laparoscopic surgery regarding TME quality and CRM positivity, while the COLOR II trial confirmed the safety of laparoscopic surgery [28] and its non‐inferiority in 3‐year local RFS [1]. However, thus far, no RCT has shown a clear long‐term oncological advantage of laparoscopic over open surgery. To our knowledge, this study represents the largest cohort to date suggesting a potential long‐term survival benefit of laparoscopic surgery compared with open surgery in the treatment of locally advanced rectal cancer.

In alignment with previous RCT findings [7], our study also demonstrated short‐term advantages of robot‐assisted surgery over laparoscopy in terms of complication rates, faster recovery, and improved urinary function [29, 30, 31, 32, 33]. Consistent with the observed reduction in complications, shorter hospital stays and lower 90‐day mortality, the robot‐assisted group incurred the lowest inpatient costs. Although prior studies generally report higher costs for robot‐assisted rectal surgery [34, 35, 36, 37], a Japanese DPC analysis [10] has similarly found that robot‐assisted surgery was associated with lower overall costs than laparoscopic surgery for LAR and tended to be lower for APR. To our knowledge, this is the first study to demonstrate lower total inpatient costs for robot‐assisted surgery across all procedures for locally advanced rectal cancer.

Multivariate analysis identified several risk factors for poor OS, including male sex (HR: 1.54; p < 0.001), underweight BMI (HR: 2.24; p < 0.001), CCI ≥ 2 (HR: 1.20; p = 0.011), undergoing APR (HR: 1.57; p < 0.001) and treatment at non‐university hospitals (HR: 3.53; p < 0.001), besides surgical approach and oncological parameters.

Sex‐based anatomical differences are an important consideration in rectal cancer surgery. The association between male sex and worse survival aligns with additional analyses from JCOG0212 [38]. In this cohort, over 60% of the patients were male; especially since the laparoscopic subgroup represented the largest patient cohort, the association between male sex and poorer prognosis may partly result from the overrepresentation of male patients in the laparoscopic subgroup. The male pelvis, which is typically narrower and deeper, poses technical challenges in rectal surgery and has been linked to higher rates of incomplete TME and CRM positivity, both associated with worse oncological outcomes [39, 40]. Robotic assistance, with wristed articulation and stable three‐dimensional visualisation, may mitigate these constraints in confined pelvic spaces [39]. Consistent with this mechanism, previous studies report that for mid‐ to low‐rectal cancers—particularly in male patients with cT3/4 disease—robot‐assisted TME can be associated with superior long‐term outcomes versus laparoscopy [40], whereas in female patients, specimen quality and oncological outcomes are generally comparable between approaches [41, 42]. These data support our observation that, despite the overall poorer prognosis in male sex patients, the robot‐assisted approach is associated with comparatively better outcomes than the other approaches.

Similarly, the associations of low BMI [43], higher comorbidity burden [44], and APR [45] with worse outcomes are consistent with prior literature. Regarding treatment facility, earlier studies have reported that surgeries performed at high‐volume centres are associated with better surgical quality, often reflected in CRM status [46]. Our study did not directly assess CRM positivity by facility; however, the survival advantage at university hospitals may indirectly reflect superior surgical technique and infrastructure. This warrants further investigation.

A key strength of this study lies in its use of the OW method [14], which achieved excellent balance on measured covariates and targets the overlap (ATO) population—patients who could plausibly receive any modality [14, 15], thus reducing bias and enhancing the validity of comparative analyses with respect to observed variables. In multi‐arm settings, OW down‐weights extreme propensities to improve stability, maintains excellent mean balance on observed covariates, and typically yields a larger effective sample size than both IPTW and PSM, while avoiding the sample loss and implementation complexity of three‐way matching. These properties are supported by standard references, and effective sample sizes after weighting are reported in the manuscript. Furthermore, the inclusion of a large, real‐world, multi‐institutional cohort provides a broader, more generalisable perspective than traditional RCTs, offering clinically relevant insights applicable to everyday practice. The combination of nationwide data and advanced statistical methodology represents a novel contribution to the field.

However, this study has limitations. Its retrospective, claims‐based design carries risks of selection bias and residual/unmeasured confounding; OW approximates certain attributes of randomization for observed covariates but does not replicate randomization or account for unmeasured factors. The absence of detailed pathological and intraoperative data limits granularity in outcome analysis. Moreover, hospitals not participating in the MDV database were excluded, which may impact generalisability. Additionally, cT4b cases were excluded due to their complexity and variability in outcomes, and further studies are needed to evaluate robot‐assisted surgery in this subgroup. Prospective cohort studies and RCTs focusing on cT3‐4 and Stages I–III rectal cancer are necessary to validate and expand upon these findings.

In conclusion, our findings provide the first large‐scale, database‐driven evidence of the significant short‐ and long‐term survival benefits of robot‐assisted surgery for locally advanced rectal cancer in Japan. The findings support the robot‐assisted approach as a potentially new standard of care in this setting. While further randomised studies are warranted, our results provide critical insights that may guide future surgical strategies.

AUTHOR CONTRIBUTIONS

Marie Hanaoka: Conceptualization; data curation; validation; writing – original draft; methodology. Hiroyasu Kagawa: Writing – review and editing. Ataru Igarashi: Formal analysis. Hiroshi Yoshihara: Formal analysis. Shinichi Yamauchi: Writing – review and editing. Peng‐Lin Lin: Software; data curation. Minkyung Shin: Project administration. Yusuke Kinugasa: Conceptualization; methodology; writing – review and editing; supervision.

FUNDING INFORMATION

None.

CONFLICT OF INTEREST STATEMENT

Yusuke Kinugasa received speaker honoraria from Intuitive Surgical Inc., Johnson and Johnson KK and Medtronic, Japan. Access to the Medical Data Vision database was funded by Intuitive Surgical Inc. Ataru Igarashi and Hiroshi Yoshihara received grants from Intuitive Surgical Inc. Peng‐Lin Lin and Minkyung Shin are employees of Intuitive Surgical Inc. The other authors declare no conflicts of interest, funding, or other sources of support related to this submitted article.

ETHICS STATEMENT

Ethics committee review and approval prior to study initiation were not required because the medical data vision (MDV) database has undergone extensive ethical review and approval in relation to contracts with individual hospital data providers. Because data elements in the MDV database have been de‐identified, informed consent was not required. This report followed the Strengthening the Reporting of Observational Studies in Epidemiology guidelines, and the study was conducted in accordance with the principles of the Declaration of Helsinki.

Supporting information

Data S1

Hanaoka M, Kagawa H, Igarashi A, Yoshihara H, Yamauchi S, Lin P‐L, et al. Improved 5‐year survival with robot‐assisted resection for locally advanced rectal cancer compared to laparoscopic and open surgery: A real‐world cohort study. Colorectal Dis. 2025;27:e70278. 10.1111/codi.70278

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.