The present study emulated a randomized clinical trial comparing laparoscopic and robotic-assisted proctectomy for rectal cancer, using the target trial methodology used in recent coloproctology studies [13, 14]. Our study found both approaches associated with similar pathologic and clinical outcomes. However, the robotic approach was associated with lower conversion rates to open surgery, shorter hospital stays, and increased 5-year OS, mainly in stage III disease.

Consistent with the existing literature [15, 16], the current study cohort consisted mainly of men aged ≥ 60 years, most of whom presented with stage III disease and underwent sphincter-saving surgery after neoadjuvant therapy. This observation suggests that the study population of a target trial is similar to that enrolled in prospective clinical trials. Interestingly, the robotic-assisted approach accounted for approximately 60% of patients. This finding is concordant with the observed time trend of increased use of robotic-assisted surgery for rectal cancer [15]. The present study entailed a contemporary cohort as it spanned a recent period (2015–2021) in which the preference for robotic-assisted surgery dramatically increased.

Because this study was not designed as a prospective randomized trial, several imbalances in patient, disease, and treatment characteristics between the two groups were noted. Importantly, robotic-assisted proctectomy was more often performed for men with advanced disease and elevated pretreatment CEA levels. This trend is expected as robotic-assisted surgery is presumed to have a benefit in challenging rectal cancer cases, such as men with narrow pelvises and locally advanced disease. A contemporary NSQIP analysis [17] also showed that robotic-assisted proctectomy was more likely to be performed in men with increased BMI and advanced T-stage rectal cancers, compared to laparoscopic proctectomy. Moreover, imbalances in surgical expertise among surgeons are unknown.

Patients who had robotic-assisted proctectomy in our study more often received neoadjuvant radiation. This observation may be explained by the more advanced disease or better compliance of patients in the robotic group with neoadjuvant treatments. Since patients who had robotic-assisted proctectomy received neoadjuvant therapy more often, this may explain the longer waiting time before surgery. Furthermore, as a previous study concluded [18], performing major procedures via the robotic platform may prolong wait times pre-surgery intervals presumably due to the limited number of available robotic systems.

After accounting for all observed imbalances between the two groups, using the propensity-score matching method, we were able to obtain two groups, balanced for the main patient and treatment confounders. More than 5500 patients were included in each group, yielding a larger sample size than the published randomized clinical trials [6, 7, 9,10,11] on the subject. The primary outcome analysis showed similar prevalence of positive CRM and surgical margins and a comparable number of harvested lymph nodes between the two approaches. The margin of difference in the primary outcome was ≤ 1%, and thus the lack of statistical significance is aligned with a lack of clinically meaningful differences.

The similar rates of positive CRM in our analysis were supported by the one randomized trial [7] that showed a similar incidence of involved CRM between the two groups (6.1% vs. 5.5%) and an interim analysis of the COLRAR trial, although a subgroup analysis of patients who had preoperative concurrent chemoradiotherapy showed a lower positive CRM rate in the robotic group [6]. Conversely, the REAL trial [11] showed that robotic surgery had a lower positive CRM rate than laparoscopic surgery (4% vs. 7.2%, p = 0.023). Furthermore, a recent meta-analysis of RCTs [19] showed a lower risk of positive CRM with robotic-assisted proctectomy (RR = 0.67, 95%CI 0.49–0.91). The comparable numbers of harvested lymph nodes in the two groups in our analysis were supported by the findings of two randomized trials [6, 7] and a meta-analysis [19]. Another pathologic parameter indicating the quality of surgery is the TME grade, which was not assessed in our study but was found to be comparable among laparoscopic and robotic-assisted proctectomy in two clinical trials [6, 7].

The main benefit of robotic-assisted surgery in our analysis was a 60% lower likelihood of conversion to open surgery. Although this benefit was refuted by the ROLARR trial that showed an unadjusted risk difference of 4.1% [9]; several meta-analyses [19,20,21] reported 47–68% lower odds of conversion to open surgery with robotic-assisted proctectomy compared to laparoscopic proctectomy. It should be noted that the definition and threshold for conversion from minimally invasive to open surgery may vary among surgeons and hospitals [22]. The lack of standardized criteria for what is considered conversion to open surgery may confound the interpretation of the lower likelihood of conversion with robotic surgery.

Overall, the present target trial substantiated the findings of previous small-scale randomized trials on similar clinical and pathologic outcomes of laparoscopic and robotic-assisted proctectomy. Given the known limitations in patient recruitment and follow-up, the large number of patients in each group may be challenging to recruit in a prospective clinical trial, even with a multicenter setting. However, there remains controversy on the effect of robotic-assisted surgery on CRM positivity and conversion to laparotomy. These controversies call for better standardization of the pathologic assessment of rectal cancer specimens and the definition of conversion to laparotomy.

The target trial methodology has several limitations related to the retrospective nature of the data used, missing data, the possibility of misclassifications, and failure to account for other variables that were not available, such as BMI, hospital volume, and the surgeon’s experience. Target trials attempt to emulate RCTs using large observational datasets, which are not feasible when single institutions’ data are used. Despite the lack of granular data on the availability of laparoscopy or robotics, perioperative management, learning curve, and case volume, a target trial adjusts for the main confounders such as demographics, disease stage and characteristics, and insurance type, while providing a large number of patients in each group, sufficient for well-powered and nuanced analyses of secondary outcomes. Although unobserved confounders may have impacted the study results, this impact is not expected to be major because of the narrow margin of difference in the main outcomes. Furthermore, the large number of patients included in the analysis should have helped reduce the risks of an erroneous conclusion. Although we matched the two groups for the facility type, and all hospitals contributing data to the NCDB are CoC accredited, assuming surgeons have completed an adequate learning curve for each procedure, we cannot verify if both laparoscopy and robotic platforms were available at each hospital. Both groups exhibited several significant baseline differences. However, after matching, the two groups were well balanced and matched for demographics, tumor characteristics, and treatment details. Although the effect of residual unobserved confounders should be considered, the study design emulated the outcomes of a randomized trial with selection bias minimized, albeit not eliminated. Finally, the definition of conversion to open surgery may not have been consistent among the hospitals that provide data to the NCDB.