Colorectal Cancer is currently the third most common cancer in the world and the rectum is affected in approximately one-third of cases [1
]. The mainstay of curative treatment is surgical resection with lymph node dissection and total or partial mesorectal excision (TME/PME) [2
]. Depending on tumor stage localization surgery can be accompanied by (neo)adjuvant (radio)chemotherapy [4
Surgery can be performed by either traditional open surgery or minimally invasive approaches. International multicenter studies such as the COLOR II trial [5
] or COREAN [6
] have proven the equivalence of laparoscopy and open surgery for oncological outcomes (e.g., 5-year disease-free survival), while ALaCaRT [8
] and ACOSOG [9
] showed equivocal results. Additionally, a current meta-analysis confirmed these findings [10
]. Furthermore, robotic-assisted surgery (RAS) has been recently implemented in the field of general surgery for performing complex procedures in rectal [11
], pancreatic [12
] or esophageal surgery [13
]. Advantages include 3D vision, a steady camera and the endowrist function with seven degrees of freedom. As for rectal cancer in particular, a recent meta-analysis found no differences between conventional laparoscopy and RAS in terms of oncological outcomes [14
]. Additionally, another meta-analysis compared open, RAS, laparoscopy, and transanal TME and found that laparoscopy and RAS enhance postoperative recovery, while open surgery and transanal TME may improve oncological resection [15
In combination with screening methods, the available treatment options have led to a lower mortality for rectal cancer in the western world over the last decades [16
], with an improvement of 5-year survival from 48.1% in the 1970s to now 67.7% [18
]. Due to these higher survival rates, there is now a considerably higher number of patients with long-term survival that have to deal with potential side effects of the treatment which can have a profound impact on patient’s quality of life (QoL). These influencing factors include but are not limited to bowel and urinary continence, sexual function as well as the immediate postoperative course.
Even in the established fields of RAS such as urology, there is no clear evidence indicating better oncologic outcomes for RAS when compared to laparoscopic surgery (or open surgery) [19
]. However, there is evidence that shows an improvement in functional outcomes for RAS [7
]. This might be true for rectal cancer as well. It is frequently hypothesized that the technological advancements of the robotic system might facilitate fine dissection in anatomical planes and could thus improve the intactness of nerves and other structures relevant for functional outcomes. Current studies focus mainly on oncologic outcomes or perioperative parameters [14
] or raise considerable concerns for methodologic quality, e.g., without critical appraisal of the quality of evidence or bias analysis [20
Therefore, the aim of this review is to compare functional outcomes for laparoscopic and RAS for rectal cancer.
Materials and methods
This systematic review was conducted in line with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [22
] and the AMSTAR 2 criteria [23
]. In addition, it was prospectively registered with PROSPERO (CRD42018104519).
2.1. Search strategy and information sources
A comprehensive database search including Medline (via PubMed), Web of Science and CENTRAL was performed as suggested by Goossen et al. [24
] until November 2018. However, more recent trials which were identified by manual search during the working process were also included if eligible. No restrictions in terms of language, year or study design were applied. Assistance of a librarian from Heidelberg University was sought to optimize the search strategy based on PICO criteria [25
P (patients)—Male or female patients over the age of 18 with disease of the rectum which requires elective rectal resection.
I (intervention)—Robotic-assisted rectal resection.
C (comparator)—Conventional laparoscopic rectal resection.
O (outcome)—At least one functional outcome.
S (Study design)—Comparative studies with all types of study designs.
RCTs and cohort studies were eligible since RCTs are often criticized to not necessarily reflect reality as there are always patient and surgeon specific factors which influence the choice of treatment. However, they were also analyzed separately in order to evaluate whether different study types lead to different outcomes. Furthermore, different types of resection such as AR, ISR, or APR were included. Depending on the study, high and/or low AR were reported together.
2.2. Screening process and data extraction
Two reviewers independently screened titles/abstracts and consecutively full-text of eligible studies. Disagreement was solved by consensus or a third reviewer. Data from included studies were extracted into a dedicated data sheet and pre-tested to prove its suitability. In addition, references and grey literature (e.g., abstracts) were searched. In case of similar studies from the same research groups or missing data, authors were contacted to evaluate whether studies report on the same patient clientele and to gather additional data. For example, in order to gain more information from the ROLARR trial [11
] for urinary and sexual dysfunction, an official data sharing agreement were signed. In case of multiple publications of one study, the main manuscript was included, but all of the manuscripts were read and included for data extraction in order to gather all available information.
2.3. Data items
Data were sought for (1) general information (e.g., year of publication, country), (2) study participant characteristics (e.g., age, disease, BMI) and inclusion/exclusion criteria, (3) type of intervention [laparoscopic vs. robotic-assisted, as well as type of resection (AR, APR, ISR)] and (4) functional outcomes [ileus, urinary retention, international index erectile function (IIEF), international prostate symptom score (IPSS) which includes typical lower urinary tract (frequency, urgency, straining, intermittency, weak stream, nocturia, incomplete emptying), Quality of life scores (QoL)]. Furthermore, analysis was stratified by study design (RCT vs. non-RCT). In case of different follow-up intervals or time periods, the value of the last available follow-up time point was used for analysis.
2.5. Exclusion criteria
Indication other than rectal cancer.
No stratified analysis by study arm.
Studies before 2001 (after the first reported robotic procedure for colorectal cancer) [26
Redundant patient population.
2.6. Risk of bias in individual studies
Risk of bias in the individual studies was assessed independently by two reviewers with the Newcastle–Ottawa scale (NOS) [27
]. Disagreements regarding the risk of bias assessment were resolved with a third party. Publication bias was assessed with the help of funnel plots.
2.7. Certainty in evidence
Certainty in evidence was assessed using the GRADE approach [28
]. This was done independently by two reviewers using the GRADE Pro Software (McMaster University and Evidence Prime Inc, Ontario, Canada). In case of disagreement a third reviewer was consulted.
2.8. Statistical analysis
Statistical analysis was performed by a biometrician (S.S.) at the Department of Medical Biometry and Informatics at Heidelberg University who was otherwise not involved in the study.
Endpoints were either binary or (quasi) continuous. In case of binary data, the odds ratio were used as effect measure and trials were pooled using the Mantel–Haenzel method to account for the sparse number of events [29
]. In case of 0 event either trial arm, 0.5 was added as continuity correction to both arms of the trial. The continuous endpoints were mainly estimated by the mean in each trial arm at pre-surgery and different follow-up times. We used the mean difference as effect measure for continuous endpoints, except for the IIEF total score, where trials reported different measures and the standardized mean difference estimated by Hedges’ g was used. If studies reported medians and interquartile ranges, the mean was estimated by the median the standard deviation by the width of the interquartile range divided by 1.35 [31
]. For all meta-analyses, random-effects models with the DerSimonian–Laird estimator for between-trial heterogeneity were estimated to account for unexplained variation. Continuous endpoints were pooled using the inverse variance method.
The individual trial results were additionally combined within type of trial as pre-specified subgroups to account for potential bias in the observational trials. Heterogeneity was explored and reported by the I2-statistic. Funnel plot asymmetry was assessed graphically for all endpoints, however, no test on funnel plot asymmetry was performed as the presence of subgroups on trial type may already have introduced asymmetry.
The analyses were carried out in the software R (version 3.5.1) [32
] and its extension meta (version 4.9.2) [33
2.9. Institutional review board
Approval by an internal review board was not applicable since only data which were already published was used for this study.
This systematic review and meta-analysis summarizes the existing evidence in terms of functional outcomes after robotic-assisted or laparoscopic rectal resection for cancer. Overall the results show advantages in favor of the robotic approach regarding ileus, urinary retention, urinary function, and quality of life, while there were no differences concerning sexual functioning.
When compared to open surgery, laparoscopy has shown to provide potential advantages during the perioperative period such as a faster return of bowel function [36
] and a shorter hospital stay after rectal cancer surgery [5
]. However, no consistent advantages regarding ileus were recorded in RCTs comparing laparoscopy to open surgery [36
]. Conversely, the present pooled analysis provides evidence that RAS may help reduce the rate of postoperative ileus compared to CL.
Urinary retention which requires prolonged catheterization was also significantly reduced for RAS compared to CL in the present analysis.
Mid- and long-term outcomes such as QoL and sexual, urinary and bowel functioning are very relevant to patients after the initial postoperative period. Despite the small number of included studies for this outcome, RAS was associated with improved QoL. The most prevalent assessment score used in the included studies was the QLQ C-30 which was developed by the European Organization for Research and Treatment of Cancer (EORTC) and is most commonly used to measure the QoL for cancer patients [37
]. Yet, systematic assessment of QoL has been underrepresented in surgical practice and research. In order to capture all facets of a disease and its impact on QoL, “patient reported outcomes” are gaining widespread acceptance. These are an effective approach to comprehensively collecting patients perception of the underlying disease [39
]. Therefore, these measurements should be incorporated into the clinical pathway and especially into surgical trials, in order to gain more insight into the impact of different treatment approaches on QoL and to confirm the hypotheses generated by the presented study.
In terms of urinary and erectile functioning it was noticeable that the outcome measures were mainly scores that were not primarily designed to evaluate patients after rectal surgery. The IPSS was first introduced into urology by the American Urological Association in order to assess patient perception of problems derived from benign prostatic hyperplasia [40
]. Therefore, it remains questionable whether this score is the optimal choice seeing as symptoms such as straining or weak stream could still be attributed to obstructive symptoms due to an enlarged prostate. Regarding these limitations, studies should check for potential baseline differences in different age groups and stratified randomization based on baseline characteristics should be considered. Additionally, comparability between sexes is not possible as women obviously require different scores.
Similar considerations can be made when assessing sexual function. For example, the ROLARR trial used the IIEF and FSFI to assess sexual function [11
]. Despite the high quality of the trial, only 57% of men and 36% of women completed the assessment scores which underlines the low emphasis and neglect of the relevance of functional outcomes in current trials. Additionally, less than 20% of the included studies reported erectile or sexual function at all. Moreover, while potential advantages regarding ileus, urinary retention rates, urinary symptoms as well as QoL for RAS over CL were found, no differences were found for sexual function. In consequence, in the present study it appears that RAS potentially provides better functional outcomes compared to CL.
Seen from a technical perspective, reasons for improved outcomes of RAS may include better visualization due to 3D vision, thus possibly enabling improved detection of risk structures and more accurate dissection due to the stable instrument platform and precise instrumentation. In addition, RAS seems to unfold its potential with more complex procedures. Hence for procedures with high complexity during the reconstruction phase such as pancreaticoduodenectomy or esophagectomy, the laparoscopic approach has failed to gain widespread acceptance due to its technical difficulty. Therefore, these procedures are mainly performed in specialized centers. For example, the recent Dutch Leopard-II study was terminated early due to increased mortality in the laparoscopic group [41
]. A subsequent meta-analysis comparing open and laparoscopic pancreaticoduodenectomy including only RCTs did not find any relevant advantage which would support the use of laparoscopy for this procedure [42
]. However, the robotic approach to pancreaticoduodenectomy seems to offer a promising alternative to the open procedure with comparable postoperative outcomes and low mortality and should thus be evaluated in RCTs [12
However, as controversially discussed, the surgeon herself or himself may be a main contributor to the surgical outcome. While there will always be interpersonal differences between surgeons which can hardly be eliminated, each individual has to go through a considerable learning curve. The presence of learning effects is well acknowledged in the surgical community and structured training programs have been introduced. However, it has been shown that previous experience in open or laparoscopic surgery does no translate into better performance with RAS [44
]. With this in mind, the learning curve of RAS for rectal cancer takes approximately 20 to 35 cases [45
]. In comparison, the learning curve of laparoscopic colorectal surgery has been shown to be between 80 and 150 cases depending on the outcome parameter in a systematic review and international multicenter analysis [48
]. Of the included studies in this manuscript, a considerable amount reported on their initial experience with RAS without adjusting for learning effects. An excellent example of accounting for surgeon experience is the ROLARR trial by Jayne et al. [11
]. Here it was clearly defined that participating surgeons had to have performed at least 30 minimally invasive rectal cancer resections with at least 10 robotic and 10 laparoscopic procedures. As a matter of fact, the ROLARR trial served as real-world example in a consecutive study by Corrigan et al. that accounted for experience of participating surgeons [49
]. While the original paper reported an OR of 0.61 (CI 0.31, 1.21; p
= 0.16) for the primary endpoint (conversion), the study by Corrigan et al. found that patients who were operated by surgeons with the mean experience of all ROLARR surgeons (153 laparoscopic cases; 68 robotic cases), had a lower OR of 0.40 favoring the robotic approach (CI 0.168, 0.953; p
= 0.039). Furthermore, for surgeons who had performed at least 100 robotic procedures, the odds of conversion were significantly lower with the robotic approach, independent from the number of laparoscopic procedures performed (i.e. 45, 91 or 180 cases). Overall there are learning effects for both approaches, RAS and laparoscopy, while RAS seems to be easier to adopt, thus less cases (and patients) are needed to optimize quality of surgical care.
Finally, since minimally invasive techniques are more often used, an emerging third option besides RAS and traditional laparoscopy is transanal TME (taTME). taTME offers some potential advantages but also seems to have some drawbacks especially in terms of implementation and learning curves. The available studies on this technique showed contradicting results, hence it remains to be further investigated according to IDEAL stages [50
]. In accordance with the previous paragraph stringent patient selection, dedicated training, and high case-volume in specialized centers should be integrated to avoid adverse outcomes [52
The major limitations of the present systematic review and meta-analysis are the heterogeneous reporting measures for the respective studies. For example, there is an original and a simplified version of the IIEF-5 which were both used in different studies. Further limitations include the use of only aggregated data as there was no access to primary data sources and no stratification by surgical approach (e.g., high and low anterior resection, Hartmann’s procedure, intersphincteric resection). For future trials it would thus be desirable to distinguish between the different procedures since this could influence functional outcomes. However, appropriate statistical measures were applied to account for these variations and all analyses were performed by a senior statistician. Additionally, the studies were not powered for functional outcomes as primary endpoints which might reduce the trust in the presented results.
Furthermore, there was no consistent definition of the operative approach in terms of clear differentiation. For example, between LAR and HAR. Both anatomical landmarks, such as above/below the peritoneal flap as well as distances from the anus in centimeter were reported. The limited quality of the included studies itself resulting from mostly non-randomized studies should not be considered as a limitation, since quality assessment is a crucial step of each systematic review which helps to understand drawbacks of recent studies.
Implications for future research
As discussed previously, the current evidence remains equivocal but this meta-analysis summarizes the existing data, identifies gaps of the current evidence and thus can serve as a basis for generating hypotheses and performing sample size calculations of further studies with functional outcome as primary outcome. In addition, the term functional outcomes must be further defined and clarified. For example, some scores were adopted from other specialties such as urology and might fail to assess further symptoms that are specific to rectal surgery. Along with this, patient characteristics and subgroups (body mass index, gender, previous surgery or neoadjuvant treatment) should be taken into consideration in order to provide patients with a more individualized treatment. Finally, full analysis for the learning curve and differences in tumor stages as well as the different procedures should be further investigated.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.