A comparison of operative outcomes between standard and robotic laparoscopic surgery for endometrial cancer: A systematic review and meta‐analysis

Abstract Background Evidence has been systematically assessed comparing robotic with standard laparoscopy for treatment of endometrial cancer. Methods A search of Medline, Embase and Cochrane databases was performed until 30th October 2016. Results Thirty‐six papers including 33 retrospective studies, two matched case–control studies and one randomized controlled study were used in a meta‐analysis. Information from a further seven registry/database studies were assessed descriptively. There were no differences in the duration of surgery but days stay in hospital were shorter in the robotic arm (0.46 days, 95%CI 0.26 to 0.66). A robotic approach had less blood loss (57.74 mL, 95%CI 38.29 to 77.20), less conversions to laparotomy (RR = 0.41, 95%CI 0.29 to 0.59), and less overall complications (RR = 0.82, 95%CI 0.72 to 0.93). A robotic approach had higher costs ($1746.20, 95%CI $63.37 to $3429.03). Conclusion A robotic approach has favourable clinical outcomes but is more expensive.

manuscript was used to avoid duplication. The exception was when different outcomes were reported in separate papers. It was not possible to include papers that looked at outcomes from large registries as many patients from the other studies would have been included in national and regional databases resulting in duplication. However, registry papers were retrieved from the search and assessed descriptively in the discussion of this paper.
Data were taken from the text and tables of the published papers.
The presentation of data depended on that reported in individual papers. For example, if a study reported both the pelvic and para-aortic lymph node yields, it was only possible to include this data in total lymph node counts if that data was reported. A similar situation was applied to the reporting of operative complications. To avoid a complication being counted twice and potentially prejudicing one arm, a conversion to a laparotomy in it's own right was not reported in the complication fields but treated separately. The same applied to blood transfusions. Where possible, complications were reported as 'total' but divided into 'major' and 'minor' in nature if reported as well as 'intra-operative' and 'post-operative' if separated in a paper's text. If the Clavien-Dindo classification was used in a paper, post-operative complications classed as III or above were defined as 'major' . Additional information clarifying data was sought from three authors and in one case this was provided. 7 Costs and charges were presented in United States Dollars. If this was reported in another currency then this was converted to Dollars using the exchange rate published for the middle year of the recruitment period from the Bank of England website (www.bankofengland. co.uk). The data were recorded using Review Manager. 8 Dichotomous data were presented as Risk Ratios using the Mantel-Haenszel method with random effects. 9 Continuous data were presented as means with standard deviations and analysed using the Inverse Variance method using random effects. 10 When continuous data were presented as medians with ranges, the data were converted for inclusion into the meta-analysis using the method described by Hozo et al.. 11 When only interquartile ranges were reported, the data could not be included into the meta-analysis.

| RESULTS
A flowchart of how papers were selected is given in Figure 1. This revealed 35 papers that were included in the study. 5, A further hand-search of review article references included one additional paper. 45 Therefore, a total of 36 papers were included in the analysis and these involved 8075 patients (3830 robotic and 4245 laparoscopic).
A list of papers included in the meta-analysis and the outcomes included are detailed in Table 1. This included 35 retrospective cohort studies of which two contained matched case-controls. 19,31 In addition, there was one randomised controlled study 32 (Table 1). Furthermore, seven papers reporting data from registries were carefully read and used for comparative discussion in the relevant section of this paper. [46][47][48][49][50][51] A summary of the outcomes is shown in Table 2. Across all studies, there was no statistically significant difference in the duration of surgery or operating room times (Table 2). However, the one randomized   but no difference in the total operating theatre time (Table 2). One study reported a longer time from arrival in theatre to the surgical incision for robotic surgery (Table 2). 39 The number of days stay in hospital was shorter in the robotic arm compared with standard laparoscopy (Figure 3).
There was no difference in the total number of lymph nodes removed in the two arms (Table 2). Furthermore, there were no differences between the pelvic and para-aortic lymph node yields when analysed separately ( Table 2).
The estimated blood loss was on average 57.  (Table 2).
For adverse outcomes, significant differences were not found for re-interventions, re-admissions, major complications, intra-operative complications, major intra-operative complications, post-operative complications or major post-operative complications ( Table 2) No differences could be demonstrated between the two groups for pain scores or post-operative analgesia usage (Table 2). However, data from two studies showed significantly less intra-operative narcotic analgesia usage in the robotic group (−40 mg morphine equivalents, 95%CI = −52.11 to −27.85 mg) although this was heavily weighted by one study. 22 No differences were demonstrated in the risk of recurrence (Table 2).
Six studies reported the total costs of surgery and could be used in a meta-analysis. All but one showed an increased cost with the robotic arm with a mean additional cost of $1869.42 (95%CI = $267.89 to $3470.94).

| DISCUSSION
These data are favourable towards the robotic arm for hospital stay, return to normal activity, return to a normal diet, conversion to laparotomy, operative complications and blood loss. The total cost is in favour of standard laparoscopy. All but three studies assessed are retrospective cohort reviews, two are matched retrospective reviews and one a randomized controlled study (Table 2). Therefore, the quality of the evidence is low although it is bolstered by large numbers of papers and patients. One criticism is that in many of the papers, the robotic arm consists of an early series for the surgical teams. Outcomes with robotic surgery improve with numbers performed 29 Other recent reviews and meta-analyses of the subject exist. 54,55 They do not include all the citations that are in this study nor the randomized controlled study. Some of these meta-analyses include registry studies even though some analyse the same databases and include patients reported in the institutional cohorts. However, the findings of less operative conversions, lower blood loss, and a shorter hospital stay are consistent findings within meta-analyses but this study also demonstrates less overall complications in the robotic arm as well as higher costs. 54,55 This study found significantly longer operating times for robotic surgery in the retrospective cohort studies. However, the one randomized controlled study showed shorter operating times for robotic surgery. 32 This may be due to the 'early series' effect described when a teams first few operations took longer than the later procedures in their series but in one study where the surgeon and team was already experienced in robotic surgery, longer operating times were still demonstrated. 5 It is possible that this is a power effect and a larger study

Pain and analgesia
Postoperative visual analogue pain score (0-10) This study demonstrates a shorter hospital stay for robotic cases.
This is supported by one registry study that showed a significantly lower proportion of women staying three nights or more in hospital. 51 One other registry study reports a non-significant shorter stay in the robotic group. 46 Return to normal activity is shorter for robotics in the one study that reports this outcome in the meta-analysis. 12 One registry study reports on this. 46 That study 46 reports on a 6.7 days quicker return to normal activity for the robotic arm but reports this as being non-significant. However, using the Inverse Variance method this would have 95% confidence intervals of 2.05 to 11.35 days shorter return to normal activity which supports the data we report.
The reduction in conversion to laparotomies and less complications might explain these findings as one would expect a patient who had a laparotomy or one who suffered complications to spend longer in hospital and take longer to return to normal activity.
In this analysis we demonstrated less blood loss in the robotic arm.
However, this could be perceived as a surrogate outcome as 50 mL This is likely to be related to the increased ergonomics of robotic surgery over standard laparoscopy. 57 However, the outcome is not supported in a registry study. 51 Re-operation and re-admission rates are also reported in registry studies without any demonstrable significant difference.
The findings of less overall complications may also be related to ergonomic reasons although it will be interesting to see with time how further studies not influenced by the 'early series' effect will alter the analysis of intra-operative, post-operative, and major complications. The registry studies have conflicting results for this outcome.
Total complication rates are very heterogeneous as they are dependent on the definition of a complication and the systematic way in FIGURE 4 Mean estimated blood loss (mL) following surgery for endometrial cancer morbidity' yet the analysis in a table showed significantly less medical complications, significantly less bladder injuries, and significantly less re-operations for robotic surgery compared with standard laparoscopy. 50 Another study by the same group showed a 4% increase in all complications and medical complications in the robotic arm. 49 FIGURE 6 All complications related to surgery for endometrial cancer FIGURE 7 Conversions to laparotomy following surgery for endometrial cancer The cost analysis is in favour of the standard laparoscopy arm of the study being $1869.42 less expensive. This is consistent with outcomes from a large registry study where standard laparoscopy was $1291.00 cheaper than a robotic approach to endometrial cancer. 50 This figure reduces to $688.00 for individual surgeons who perform more than 50 cases a year 48  One problem with analysing cost data in such a way is that different countries have variable healthcare reimbursement systems and wage costs. For example in some countries where there is social healthcare, surgeons are salaried by institutions and in other countries they charge separately. Therefore, a cost-benefit may exist in one healthcare system and not in another and it is difficult to interpret how this data would apply to a single institution although it is clearly of interest.
One matter to consider when assessing these outcomes is the innovation in new platforms over time. In early series, the Da-Vinci Standard® system will have been used, whereas in latter series the fourth generation of platform (DaVinci Xi®) may have been available.
To date there is no published data on the value of the updated systems on outcomes and it would be interesting to analyse this. Furthermore, different institutions have different protocols for para-aortic and pelvic lymph node dissections resulting in a heterogeneity of operations performed across institutions. If a consensus ever occurs on the role of lymphadenectomy in endometrial cancer then it would be wise to assess separate subgroups but this is not possible currently.
In summary, this study demonstrates that the current evidence is in favour of robotic assisted laparoscopy for endometrial cancer over standard laparoscopy for clinic outcomes but costs are probably greater. To date there are only 99 patients recruited to randomized controlled trials 32 and an increase in this number will undoubtedly provide stronger evidence.

CONFLICT OF INTERESTS
Marielle Nobbenhuis and Thomas Ind have proctored for Intuitive Surgical.

ETHICS
As this is a review no ethics was required.