|Home | About | Journals | Submit | Contact Us | Français|
Despite a shift in practice patterns from open radical prostatectomy (ORP) to robot-assisted laparoscopic prostatectomy (RALP)  it has not been clearly established that RALP is as effective in treating prostate cancer as ORP. Early oncological outcomes demonstrating the effective treatment for prostate cancer are most commonly measured as a negative surgical margin status and the absence of biochemical recurrence (BCR).
At Memorial Sloan-Kettering Cancer Center (MSKCC), a high volume tertiary care cancer centre, both ORP and RALP are performed routinely by high volume experienced oncological surgeons. We sought to compare the positive surgical margin rates and the BCR-free survival (BCRFS) rates of patients undergoing radical prostatectomy (RP) stratified by surgical approach in a contemporary cohort of patients.
Following Institutional Review Board approval, we retrospectively identified 3005 men in the MSKCC prostatectomy database who underwent RP from January 2007 through December 2010. We only included patients whose prostatectomy was performed by surgeons performing the highest volume of ORP (JAE, PTS) or RALP (JAC, VPL) to reduce effects of the surgical learning curve. We excluded patients who underwent laparoscopic prostatectomy without robotic assistance or who underwent ORP or RALP by another surgeon at MSKCC (n = 1330) and patients who underwent salvage RP (n = 91). To compare only the predominant surgical approach performed by these high volume surgeons, we excluded patients who received the non-predominant surgery type (n = 125) by the four surgeons included in this study. Patients without BCR who received adjuvant therapy (n = 1) or who had radiographic or biopsy-proven metastatic disease preoperatively (n = 4) were also excluded. This left 1454 patients available for analysis. BCR was defined as PSA ≥ 0.1 ng/mL or PSA ≥ 0.05 ng/mL followed by further therapy for cancer.
ORP and RALP were performed with modifications to previously described surgical techniques [2,3]. Pelvic lymph node dissection (PLND) was performed in all patients with a 2% or greater risk of lymph node involvement using an established nomogram . In patients whose risk was below this value, the decision to perform a PLND was at the discretion of the surgeon. The majority of patients (n = 1363, 94%) underwent PLND and the boundaries of this dissection uniformly included external, obturator and hypogastric packets . Attempts were made to preserve the neurovascular bundles in all cases; complete or partial bilateral nerve sparing was accomplished in 91% of the cohort.
All patients who received prostate biopsies at outside institutions routinely have specimens secondarily reviewed at MSKCC by a dedicated team of pathologists. Following RP, prostate and seminal vesicles were step-sectioned transversely and the most apical tissue was sectioned in the sagittal plane. Lymph nodes were individually embedded in paraffin and examined microscopically.
Preoperative and postoperative oncological data and demographics on all patients undergoing RP at MSKCC are prospectively collected in the prostatectomy database. Patients treated at MSKCC are encouraged to have their postoperative care at MSKCC; those who receive care elsewhere routinely have their biochemical data sent to MSKCC to be entered into the database. PSA values are obtained at 6, 12 and 24 weeks following surgery, then every 6 months for the first 2 years, and then yearly thereafter.
BCRFS probabilities were estimated using Kaplan–Meier methods, censoring patients without BCR or with death of other causes at their last BCR-free follow-up. A multivariable Cox regression model was used to evaluate the association between surgical procedure and BCR using predicted risk from a predictive model (nomogram) for BCR, based on preoperative variables, as the covariate. Preoperative nomogram risk was determined using a previously externally validated risk calculator which incorporates the following clinical variables: PSA, both primary and secondary biopsy Gleason grade, clinical stage, number of positive biopsy cores and number of negative biopsy cores . In patients who underwent biopsies at outside institutions, nomogram-predicted values were based on a review of outside biopsy data by MSKCC pathologists. To determine whether differences between procedure types depended on baseline risk, e.g. if open surgery was superior only for high risk patients, an interaction term between nomogram risk and procedure type was included in a second model. To illustrate how recurrence risk for each procedure type varied with baseline risk, we created separate models for ORP and RALP; the difference between these two models can be interpreted as individualized difference in outcome for a patient with a given set of tumour characteristics. To explore whether our results were sensitive to the choice of covariates used to control for case mix, analyses were repeated using National Comprehensive Cancer Network (NCCN) risk groups in place of nomogram probability .
Finally, to explore whether differences between surgeons were greater than differences between surgical approaches, these analyses were repeated using individual surgeon in place of surgical approach, again adjusting for nomogram risk. Statistical analyses were performed using Stata 11.0 (StataCorp, College Station, TX, USA).
Preoperative patient characteristics by procedure type are summarized in Table 1. There was an approximately 2:1 split between ORP (n = 961, 66%) and RALP (n = 493, 34%). There were no important differences in preoperative characteristics by group. The entire cohort skewed toward higher risk patients. Overall, 63% had clinical Gleason ≥ 7 disease and 68% were intermediate to high risk using NCCN criteria. Not surprisingly, the groups were also similar in their pathological characteristics (Table 2): more than 80% had pathological Gleason scores ≥7, 37% had disease that extended beyond the prostate and 9% had node positive disease. Positive surgical margins were 15% in both groups; when stratified by NCCN criteria (Table 3) margin positivity was lower in the ORP group for low risk patients (8% vs 13%) but higher for high risk patients (31% vs 17%). Margin positivity rates showed little differences when patients were stratified by pathological T stage.
Median follow-up for patients without BCR was 1 year and 0.7 year for ORP and RALP, respectively; 288 patients were followed for at least 2 years without BCR. In total, 97 patients developed BCR during the study period. Figure 1 shows the unadjusted cumulative incidence of BCR as a function of time from surgery by procedure type, suggesting similar outcomes between groups. Figure 2 shows the predicted 2-year BCRFS for each procedure type with a density plot demonstrating the distribution of nomogram risk for our cohort. The overall adjusted 2-year probability of recurrence was 4.1% and 3.3% for ORP and RALP, respectively. If there were any differences between ORP and RALP, these might be expected to occur amongst highest risk patients; however, the density plot demonstrates that the large majority of patients had <20% risk of BCR at 5 years. The hazard ratio for RALP compared with ORP on multivariable Cox regression adjusting for preoperative risk was 0.88; 95% CI 0.56–1.39; P = 0.6. When added to the model, the interaction term between nomogram risk and procedure type was not statistically significant (P = 0.6). We saw no evidence of an overall difference in outcome between procedure types or differences that depended on baseline risk (e.g. ORP was not more effective for higher risk patients).
When NCCN risk was used in place of nomogram risk the results were similar. For ORP, the 2-year probability of recurrence was 0.7%, 4.4% and 22.4% for low, intermediate and high risk, respectively. The comparable estimates for RALP were 0.5%, 2.9% and 15.2%. Using NCCN risk as the covariate in a Cox model gave virtually identical results to the main analysis (hazard ratio 0.74; 95% CI 0.47–1.17; P = 0.2). The interaction term between NCCN risk and procedure type was also non-significant (P = 0.9). Finally, analysing the results for each of the four surgeons separately the 2-year probability of recurrence, again adjusted for nomogram risk, was 2.5%, 2.9%, 3.9% and 4.8%.
RALP was first described in 2001  and has become the most common surgical approach for performing RP in the USA . Despite this rapid adoption, there are only limited data comparing oncological outcomes of open vs robotic RP in contemporaneous cohorts. In this study we compared the outcomes of four experienced surgeons, two performing ORP and two performing RALP, over a concurrent time period at a single centre and demonstrated no significant differences with respect to positive margin rates or BCRFS even in a patient population that is enriched for high risk disease . Ideally, a comparison of oncological outcomes in patients undergoing RP using two different surgical approaches would be made through a randomized trial. While no such trial exists, several investigators have used single institutional databases to establish the equivalence of these approaches [10–12]. Our results are consistent with these previous investigations; however, there are several unique considerations that make our findings of additional value. First, we utilized several measures to reduce potential biases that may have impacted previous investigations. Second, the higher risk nature of our cohort compared with other studies demonstrates that these surgical approaches can result in similar outcomes even in this population.
In the present study, in order to reduce any impact of the learning curve, which has previously been established to affect oncological outcomes, we only included data from contemporary experience of urological oncologists who had performed well in excess of 250 RPs using their predominant technique [13,14]. Regardless of surgical approach, for all operations the primary focus was to adhere to strict oncological principles. Seminal vesicles were completely excised in all cases. A full PLND was performed for all patients with ≥2% risk of nodal invasion based on previously validated nomograms and selected patients whose risk fell below this number. This resulted in a large and similar median nodal yield in each group [15,16]. This contrasts with studies that demonstrate that RALP surgeons are up to five times more likely to omit PLND than ORP surgeons, even for high risk cancers [17,18].
For the current cohort, all pathological tissue was collected, processed and analysed by the same group of dedicated genitourinary pathologists in identical fashion. Surgical approach was determined by surgeon and patient, and there was no specific selection bias based on disease risk. In contrast to many large robotic series, low risk patients at MSKCC are routinely encouraged to consider active surveillance. This results in a generally high risk surgical cohort with similar demographics and oncological characteristics between the ORP and RALP groups .
Additionally, to control for any remaining differences in case mix, we performed a multivariable regression model using a validated nomogram to adjust for any differences in Gleason score, volume of disease, PSA or clinical stage between the two cohorts. We then repeated this analysis using NCCN risk groups to demonstrate the outcomes across risk groups. In both of these analyses we found that surgical approach had no impact on the risk of BCR. Despite suggestions that ORP may provide better cancer control for men with high risk cancers , we found no evidence to support this belief (Fig. 2). Within the limits of the power of our study to detect any significant difference and with relatively short follow-up, we found that RALP, when performed by highly experienced surgeons, was not associated with lower rates of BCRFS or higher rates of positive surgical margins.
Not surprisingly the proportion of patients with positive surgical margins was greater for both ORP and RALP as pathological stage or NCCN risk group increased. While there were differences between surgical approaches by NCCN risk group, they were largely ameliorated when patients were stratified by pathological T stage, and reflective of the limitations of risk groupings . The ability to achieve negative surgical margins depends on both the disease and the plane of dissection. Surgeons must achieve a balance between preservation of the neurovascular bundles and wider resections of tumours in cases suspected of having extracapsular extension. In only 2% of the cases were both neurovascular bundles completely excised; complete or partial bilateral nerve sparing was performed in 91% of the cohort.
Even with evidence suggesting that RP may be more effective than other treatments at controlling high risk disease , there is growing concern that surgery is under-utilized in this population . As fewer ORPs are being performed in the USA  and fewer trainees are gaining familiarity with the open approach, it is particularly important to demonstrate that for higher risk patients equivalent outcomes can be achieved using robotic assistance. Our findings demonstrate no worse outcomes for higher risk patients receiving RALP when emphasis is placed on strict adherence to oncological surgical principles.
It would be difficult to demonstrate superiority in cancer control of one surgical approach compared with another, since such differences, if they exist, are likely to be dwarfed by large heterogeneity among even experienced surgeons . Within this small group of highly experienced surgeons, we observed larger differences between surgeons than between approaches. This suggests that surgical approach should be based primarily on the skill and confidence of the surgeon to perform the appropriate and complete operation rather than on the technique.
Our findings are consistent with prior investigations comparing outcomes of ORP and RALP that have not found that surgical approach had an impact on BCRFS. Barocas et al.  compared 491 patients who underwent ORP to 1413 patients who underwent RALP performed by several surgeons at a single institution over a concurrent time frame. The case mix was such that the ORP cohort had higher PSA values, Gleason scores and positive margin rates and more palpable disease. Despite these disparities they found no significant difference in BCRFS. Krambeck et al.  compared 294 RALP patients with 588 matched ORP patients and noted that 3-year BCRFS rates and margin positivity rates were not significantly different between groups. The study may have been influenced by the heterogeneity of the surgeons, as the majority of the RALPs were performed by one surgeon whereas 17 different surgeons performed ORPs. Di Pierro et al.  performed a prospective study comparing oncological outcomes of two consecutive series of patients treated by ORP or RALP. They found no significant difference in BCRFS between groups but the study was limited by its consecutive cohort design and because all RALPs were performed by one surgeon with only 6 months of experience.
Our study has limitations, particularly the relatively short follow-up. RALP is a relatively new procedure, not commonly performed at MSKCC until 2007. With longer follow-up, our results with regard to BCR may be different. Furthermore, we did not distinguish between patterns of PSA recurrence, such as those associated with early appearance of metastases or the more indolent slowly rising PSA levels often associated with local recurrence .
In conclusion, we found no evidence to suggest that RALP results in worse oncological outcomes, even for patients with high risk cancer. Differences between surgeons were larger than differences between surgical approaches. Our findings suggest that, if the operation is performed optimally, RALP need not be limited to patients with low to intermediate risk cancer.
This work was supported by the Sidney Kimmel Center for Prostate and Urologic Cancers, by funds provided by David H. Koch through the Prostate Cancer Foundation and by the National Cancer Institute T32 CA082088-12 training grant.
CONFLICT OF INTEREST