Correspondence: Patrick S Moran M.Sc., B.Eng., Health Technology Assessment, Health Information and Quality Authority, George's Court, George's Lane, Dublin 7, Ireland. Email: email@example.com
Medline and Embase were searched for studies comparing robot-assisted radical prostatectomy with open prostatectomy and conventional laparoscopic prostatectomy. Random effects meta-analysis was used to calculate a pooled estimate of effect. The 95% prediction intervals are also reported. One randomized study and 50 observational studies were identified. The results show that compared with open surgery, robot-assisted surgery is associated with fewer positive surgical margins for pT2 tumors (relative risk 0.63, 95% confidence interval 0.49–0.81, P < 0.001) and improved outcomes for sexual function at 12 months (relative risk 1.60, 95% confidence interval 1.33–1.93, P = <0.001), and, to a lesser extent, urinary function at 12 months (relative risk 1.06, 95% confidence interval 1.02–1.11, P < 0.01). Compared with conventional laparoscopic prostatectomy, robot-assisted surgery is associated with a slight increase in urinary function at 12 months (relative risk 1.09, 95% confidence interval 1.02 to 1.17, P = 0.013). The overall methodological quality of the included studies was low, with high levels of heterogeneity. The use of prediction intervals as an aid to decision making in regard to the introduction of this technology is examined. Clinically significant improvements in positive surgical margins rates for pT2 tumors and sexual function at 12 months associated with robot-assisted surgery in comparison with open surgery should be interpreted with caution given the limitations of the evidence. Differences between robot-assisted and conventional laparoscopic surgery are minimal.
Since its introduction in 1999, robot-assisted radical prostatectomy has grown in popularity at the expense of open and conventional laparoscopic prostatectomy. This has occurred in the absence of high-quality data supporting an improvement in long-term outcomes. Previous systematic reviews have found evidence of better perioperative results, but concluded that there were insufficient data to support the superiority of robot-assisted surgery in relation to oncological and functional outcomes associated with the procedure, including cancer reoccurrence and return of sexual and urinary function.[2, 3] Extensive diffusion of a health technology in the absence of supporting evidence gives cause for concern, given the numerous examples that exist of widely-used treatments that were later shown to be ineffective or harmful. The rapid diffusion of robot-assisted prostatectomy is being matched by the rate of publication of peer-reviewed papers on the technique; a search for Medline articles with “robot” and “prostatectomy” in the title shows an average increase of approximately 25% every year between 2007 and 2011. This emphasises the need for systematic reviews of robot-assisted radical prostatectomy to keep pace with the literature, in the hope that additional evidence can provide greater clarity in regard to its effectiveness compared to alternative surgical approaches.
Medline, Embase, CINAHL, the Cochrane Library and the Journal of Robotic Surgery were searched for English or French language studies that compared robot-assisted radical prostatectomy with either open prostatectomy or conventional laparoscopic prostatectomy in men with prostate cancer. Our search terms and methodology filters are modelled on a previously published search strategy. The outcomes of interest were positive surgical margins (PSM) for pT2 and pT3 tumours, post-operative sexual function and urinary continence at 12 months, estimated blood loss (ml), transfusion rate, complication rate, operative time (minutes) and length of hospital stay (days). Post operative sexual function was defined as the ability to maintain an erection sufficient for intercourse with or without the help of phosphodiesterase type 5 inhibitors or through the use of validated sexual function questionnaires. Continence was defined as either no leakage, the use of 0 to 1 pads per day or through the use of validated continence questionnaires. The time period covered by the search was January 2000 to March 2011.
Randomized controlled trials, controlled clinical trials and observational studies comparing robot-assisted radical prostatectomy to open or conventional laparoscopic radical prostatectomy that included data on the specified outcomes of interest were eligible for inclusion. All studies were individually reviewed by two independent reviewers (PM, MON or CT), and any disagreements were resolved through discussion. Study quality was assessed using a previously published quality appraisal tool based on the design and performance of individual studies.
Data extraction was carried out by two reviewers independently (PM, MON), using data extraction forms prepared in advance. All extracted data was cross-checked and any disagreements were resolved after discussion between the two reviewers.
The quality of included studies was measured using an assessment form developed by Hailey and modified by Ho et al., which scores studies out of a maximum of 15 points based on study design (5 points) and performance (10 points). Five study performance criteria relating to issues such as patient selection, analytical methods and reporting of outcomes were scored individually to a maximum of two points depending on how these were reported in each study. Studies were then rated A to E based on the cumulative scores.
For continuous variables the sample size, mean and standard deviation were required. A number of studies reported median rather than mean, and range or interquartile range rather than standard deviation. Where a median or range was reported, the mean and standard deviation, respectively, were estimated. Studies that gave no information on the standard deviation or range were excluded from the meta-analysis.
Heterogeneity was assessed using the I2 statistic. Random effects meta-analysis was used to calculate a pooled estimate of effect. To aid interpretation of these results and to estimate the potential effect of robot-assisted surgery applied in a new setting, 95% prediction intervals were calculated. These provide an estimate of the interval within which outcomes from a new robotic prostatectomy program would fall, based on the data seen in the studies identified in the present review. In the presence of heterogeneity, this interval is wider than the 95% CI for the estimate of average effect and may be useful for service providers considering the range of likely results that can be expected on the introduction of robot-assisted surgery for radical prostatectomy. Meta-regression was used to determine if differences in outcomes could be partly explained by surgeon experience. A binary classification was used whereby a study was classified as experienced if the surgical team had completed 30 or more robot-assisted prostatectomies before commencement of the study.
For continuous variables, the WMD is reported, and binary outcomes are reported as a relative risk. Potential outliers were identified on the basis of standardized residuals that exceeded ±1.96, and publication bias was assessed using a regression test for funnel plot asymmetry. The alpha level was set at 0.05 for all statistical tests.
One randomized single-surgeon study and 50 observational studies were identified from the literature search (see Table 1 and Fig. 1). Of the observational studies included, 27 were retrospective comparisons or studies using historical comparison groups, and 23 were prospective observational studies.
A total of 37 studies compared robot-assisted and open surgery, nine compared robot-assisted with conventional laparoscopic surgery, and five provided comparative data for robot-assisted, open and conventional laparoscopic surgery in the same study. A total of 33 of the included studies originated in the USA, with the 18 remaining studies being carried out in France, Italy, South Korea, Sweden, Australia, Switzerland, Hong Kong and Taiwan. The average sample size in the included studies was 458, ranging from 40 to 1904, with patient numbers in the robot-assisted arms ranging from 20 to 1413. Mean age, BMI and preoperative PSA levels for patients in the included studies are shown in Table 2.
Table 2. Patient characteristics
Robot-assisted prostatectomy vs open prostatectomy
Robot-assisted prostatectomy vs laparoscopic prostatectomy
A total of 31 studies involved more than one surgeon in the intervention arm or failed to report surgeon numbers. A total of 62% (31/51) of studies provided information on surgeon experience; however, there were differences in how that experience was reported. The degree of surgeon experience in robot-assisted prostatectomy in the included studies ranged from surgeons carrying out their first series of cases,[10-17] to surgeons who had carried out over 300 procedures using the device. Therefore, the influence of the training curve for inexperienced surgeons is included in the overall estimate of effect calculated in the meta-analysis.
Differences between how various intraoperative and postoperative outcomes were reported in the included studies were noted. A total of 12 studies (24%) reported how operation time was defined. The most common operative time definition was skin-to-skin time (8/12),[10, 14, 15, 17, 19-22] with other studies using total operating time[23-25] or alternative definitions.[11, 12] Continence was defined in 15/51 studies (29%), with 12[10, 12, 17, 19-23, 26-29] using a definition of either no leakage or the use of 0 to one pads per day, whereas three studies[18, 30, 31] used continence questionnaires. A definition of what was considered the regaining of sexual function was defined in 12/51 studies.[10, 12, 17, 18, 20, 21, 23, 27, 29-32] This was most commonly defined as the ability to maintain an erection sufficient for intercourse with or without the help of phosphodiesterase type 5 inhibitors. Reporting of complications varied widely, with most of the studies that provided this data including a list of intraoperative and postoperative complications recorded as part of the individual study. Some studies used a standardized approach to complications reporting, whereas others categorized complications as major or minor, or provided a list of the complications recorded.
As most included studies were either prospective or retrospective observational studies, they generally scored ≤2 points on study design. Of the 51 studies assessed for study quality, three were rated as high quality, 14 as good quality, 25 as fair to good and nine as poor to fair quality (see Table 1).
Robot-assisted versus open radical prostatectomy
Pooled data from 15 studies including almost 3000 patients showed that robot-assisted surgery is associated with fewer PSM for pT2 tumors (RR 0.63, 95% CI 0.49–0.81, P < 0.001, I2 = 28%; Table 3 and Fig. 2). This result is not consistent with the results for pT3 tumors, which show no evidence of benefit (RR 1.06, 95% CI 0.85–1.34, P = 0.59, I2 = 58%), using data from 15 studies including just under 1200 patients.
Table 3. Meta-analysis results: robot-assisted surgery versus open surgery
Operative time (min)
Hospital stay (days)
Blood loss (mL)
Positive surgical margin
Pooled results from nine studies with almost 2000 patients show that men who undergo robot-assisted radical prostatectomy are more likely to regain sexual function within 1 year of surgery than those undergoing open radical prostatectomy (RR 1.60, 95% CI 1.33–1.93, P = <0.001; Fig. 3). However, the data shows a high level of heterogeneity (I2 = 70%). For urinary function at 12 months, the results of the meta-analysis of seven studies with over 1800 patients showed a slight, but statistically significant, increase in urinary function associated with robot-assisted prostatectomy (RR 1.06, 95% CI 1.02–1.11, P = 0.009).
In line with results reported elsewhere, robot-assisted radical prostatectomy was associated with a significant decrease in estimated blood loss and transfusion rate compared with open surgery. There was a high level of heterogeneity associated with the difference in blood loss between the two procedures (I2 = 98%), reflecting the different ways this was estimated in included studies. The results for transfusion rate showed greater consistency (RR 0.23, 95% CI 0.18–0.29, P < 0.001, I2 = 17%).
The risk of complications was lower for robot-assisted than for open radical prostatectomy (RR 0.74, 95% CI 0.56–1.00, P = 0.047, I2 = 72%). Operative times for robot-assisted surgery were found to be approximately 40 min longer than for open radical prostatectomy (WMD 37 min, 95% CI 17–58, P < 0.001, I2 = 98%). The level of inconsistency between the results of the 17 studies included in the present analysis was extremely high, with one study reporting statistically significant shorter operating times for robot-assisted radical prostatectomy.
Results for the difference in mean length of hospital stay for patients after radical prostatectomy showed overall stays were shorter for patients undergoing robot-assisted surgery (WMD −1.5 days, 95% CI −2.2 to −0.9, P < 0.001). Large differences were observed between the differences in length of stay observed between USA (WMD −0.7 days, 95% CI −1.2 to −0.2, P = 0.006) and European studies (WMD −2.1 days, 95% CI −3.1 to −1.1, P < 0.001).
Robot-assisted versus laparoscopic prostatectomy
No statistically significant difference between robot-assisted and conventional laparoscopic prostatectomy was observed for PSM rates (pT2 or pT3), return of sexual function, estimated blood loss or transfusion rate, complication rates, rate of conversion to open surgery or operative time (Table 4). The results from three studies reporting return of urinary function at 12 months showed a slight, yet statistically significant, increase in urinary function favoring robot-assisted surgery (RR 1.09, 95% CI 1.02 to 1.17, P = 0.013). Analysis of USA studies on their own showed a statistically significant decrease in length of stay; however, there was a high level of heterogeneity associated with this outcome (WMD −0.9 days, 95% CI −1.6 to −0.1, P = 0.022, I2 = 91%).
Table 4. Meta-analysis results: robot-assisted surgery versus laparoscopic surgery
Operative time (min)
Hospital stay (days)
Blood loss (mL)
Positive surgical margin
Sensitivity and subgroup analysis
Mean outcome values for robot-assisted surgery were similar across studies, comparing it with open and conventional laparoscopic surgery (Table 5). However, operative times were longer and estimated blood loss was lower in studies comparing robot-assisted surgery with open surgery.
Table 5. Mean outcome values for robot-assisted surgery
Mean outcome value
Compared with open surgery
Compared with laparoscopic surgery
Operative time (min)
Hospital stay (days)
Estimated blood loss (mL)
Conversion to open (%)
Positive surgical margin (%)
Regaining sexual function (%)
Regaining urinary function (%)
In the comparison with open surgery, a single statistical outlier was identified in the meta-analyses of operative time, length of hospital stay, PSM (pT3) and sexual function. Two outliers were found in the meta-analysis of transfusion rates. Only one study was identified as an outlier more than once. In the comparison with conventional laparoscopic surgery, the meta-analysis of operative time had one outlier. In all cases, the removal of outliers had a small effect on the estimate of the average effect size. In the comparison with open surgery, significant publication bias was found in three outcomes: hospital stay, transfusion rates and estimated blood loss. For the first two, removal of outliers eliminated the evidence of publication bias. In the case of blood loss, which did not have any statistical outliers, there was more systematic evidence of publication bias.
Subgroup analysis was carried out for length of stay using only European studies, on the assumption that differences in the configuration of health systems in different countries would be a confounder for this outcome. When only European data were used (USA and Asian studies excluded), the mean reduction in length of stay for robot-assisted versus open surgery was 2 days (95% CI 1.2–2.8).
Meta-regression using surgeon experience as a dichotomous covariate was carried out for seven outcomes in the comparison of robot-assisted to open surgery. For the length of stay subgroups and functional outcomes, there were insufficient studies available to justify meta-regression. Surgeon experience was not a significant predictor in any of the meta-regressions and had a minor impact on the levels of heterogeneity observed.
The 95% prediction intervals for each outcome are presented in Tables 3 and 4. The only outcomes for which robot-assisted surgery offers a highly probable benefit over open surgery are for estimated blood loss and requirement for transfusion. In comparison with conventional laparoscopic surgery, there is no evidence that a beneficial effect will be observed for robot-assisted surgery in a future study based on the 95% prediction intervals.
Despite the high rate of publication in this area, the overall methodological quality of the evidence identified through a systematic review of the literature remains low. This is a recurring theme in systematic reviews of robot-assisted prostatectomy,[62-64] and the results of the present review highlight that this remains a major concern. The vast majority of identified studies use observational study designs, with approximately half being retrospective studies or studies that use historical controls. The risk of bias inherent in these study designs, such as recruitment and allocation bias, might lead to misrepresentation of the true effect of robot-assisted surgery. One might speculate that the direction of other types of bias (e.g. confirmation and optimism bias) might favor the robot in case series studies that report clinical outcomes after the introduction of a robot-assisted surgery program within individual institutions. In such situations, the principles of equipoise and autonomy are bound to be in conflict, as presumably the setting up of a robot-assisted surgery program would require the support of surgeons who believe that such a service is in the best interests of their patients as opposed to its effectiveness being genuinely contested.
In the comparison with open surgery, robot-assisted surgery results in improved outcomes for urinary function at 12 months (RR 1.06, 95% CI 1.02–1.11, P = 0.009), sexual function at 12 months (RR 1.60, 95% CI 1.33–1.93, P = <0.001) and the rate of positive surgical margins for pT2 tumors (RR 0.63, 95% CI 0.49–00.81, P < 0.001). Although the difference observed in urinary function is of questionable clinical significance given the modest degree of benefit shown coupled with the low methodological quality of the evidence on which it is based, differences in sexual function and positive surgical margins might indicate that there are real advantages for patients whose surgery is carried out with the aid of a robot. The difficulty here is trying to ascertain the potential effect of bias on these results. In order to investigate the effect of study quality on PSM rates, the results from studies rated “high” or “good” quality were compared with those classified as “fair to good” and “poor to fair”. The results showed that using data from studies with lower methodological quality (8 studies with 1060 patients) increased the observed difference between robot-assisted surgery and open surgery (RR 0.58, 95% CI 0.41 to 0.83). When the analysis was restricted to studies of higher methodological quality (6 studies with 1130 patients), no significant difference was observed between both surgical approaches (RR 0.77, 95% CI 0.51 to 1.17). As the power to detect differences is reduced by excluding studies, this does not imply that no relationship exists; however, it might indicate that poorer study designs tend to overestimate rather than underestimate the effectiveness of robot-assisted surgery for this outcome. For the other outcome, where a clinically-important difference was observed (sexual function at 12 months), there was insufficient data to carry out a subgroup analysis based on study quality (5 studies with 1238 patients vs 2 studies with 312 patients).
When significant heterogeneity is observed in a meta-analysis, random effects approaches are commonly used. The computed confidence bound applies to the average effect, and often does not reflect what the treatment effect could be in a new study. In the face of such uncertainly, the use of prediction intervals is recommended rather than confidence intervals, particularly when making decisions about the introduction of the technology with a given healthcare institution. This approach has the benefit of providing a more realistic distribution of the effect sizes that could be expected after the introduction of robot-assisted surgery, given the heterogeneity observed in the available pool of evidence. The 95% prediction intervals can be used to determine the range of treatment effect that might be observed in a future study. In the comparison of robot-assisted and laparoscopic surgery, it is possible for all outcomes that no effect will be observed. Compared with open surgery, other than blood loss and requirement for transfusion, all outcomes might show no treatment benefit with robot-assisted surgery. These findings reflect the heterogeneity of evidence and uncertainty in the extent of treatment benefit.
Differences between robot-assisted and conventional laparoscopic radical prostatectomy were minimal. The quantity of pooled evidence used in this comparison was substantially less than that used for the comparison with open prostatectomy. There is also a much greater degree of similarity between the two approaches being compared, both being minimally-invasive surgical techniques. This results in a situation where not only are the actual differences likely to be lower than for the comparison with open surgery, but the available statistical power for the detection of these differences within this meta-analysis is also lower.
Given the major differences that exist between healthcare systems, it might be misleading to pool data from different countries. The high level of heterogeneity observed when USA and European data is combined might indicate that we not comparing like with like in this instance. Furthermore, earlier discharges might not always equate to better outcomes or a more efficient service.
The need for standardized guidelines for reporting complications in oncological urology has been highlighted previously,[66, 67] and their absence was apparent in the present review. Reporting of complications in included studies varied widely, with most of the studies providing a list of intraoperative and postoperative complications recorded as part of the individual study. Some studies used a standardized approach to complications reporting, and others categorized complications as major or minor, or provided a list of the complications recorded. Therefore, it was not surprising that a high degree of heterogeneity was observed. Inconsistent reporting of this outcome, combined with the poor evidential quality discussed earlier, makes it difficult to draw any conclusions from the results of the meta-analysis for the difference in complication rates between robot-assisted and open radical prostatectomy.
High levels of heterogeneity were seen for many of the outcomes reported. This is thought to be at least partly attributable to the poor level of consistency between the way some outcomes are defined, measured or reported combined with inherent differences that exist between studies that have been carried out in separate health systems by surgeons of differing degrees of training and experience.
Despite continued growth in the quantity of published literature relating to robot-assisted radical prostatectomy, the overall methodological quality of the available evidence remains low. In addition, data on long-term outcomes, such as cancer reoccurrence and mortality, is lacking.
The majority of the evidence base is comprised of prospective or retrospective case series. This casts doubt over the reliability of the results, as these study designs carry a high risk of bias. Furthermore, the present review shows that the direction of this bias would tend to overestimate the benefits of robot-assisted surgery. In the absence of better quality evidence, it might be advisable to use prediction intervals as an adjunct to confidence intervals to estimate the likely effects of introducing robot-assisted surgery, as an analysis that takes account of where future observations are likely to fall might provide useful information for healthcare decision makers considering the introduction of a robot-assisted surgery programme.
Based on the results of the present review, robot-assisted radical prostatectomy is associated with decreased rates of positive surgical margins for pT2 tumors, as well as improvements in return of sexual function within 12 months, when compared with open radical prostatectomy. Estimated blood loss and transfusion rates are also reduced, but operating times are longer. When compared with conventional laparoscopy, robot-assisted surgery was associated with shorter hospital stays for USA studies only, whereas the difference seen in urinary continence rates was marginal. All of these results must be considered in the context of the limitations of the evidence underpinning them.