Short-Term Outcomes

The manner in which perioperative outcomes or complications are reported is a significant confounder when trying to assess differences in complication rates between RARP and ORP. Such confusion has led to efforts to standardize the reporting of complications after surgery. Specifically, Martin and colleagues developed 10 criteria for the evaluation of studies reporting postoperative complications ( Table 1 ). These include methods of data accrual, definition of complications, outpatient information, severity grading, procedure-specific complications, length of stay, mortality rates and cause of death, duration of follow-up, and data on preoperative risk factors. These criteria were subsequently modified and adapted for urologic surgery by Donat. Although many notable studies have adopted the Martin-Donat criteria for standardized reporting of complications, these criteria are not routinely applied in most settings. For example, a recently published systematic review comparing the perioperative outcomes of RARP and ORP identified only 1 publication that fulfilled all of the 10 Martin criteria. Regardless, retrospective comparative assessments critically need to fulfill these criteria to be considered valid and relevant.

The cornerstone of the Martin-Donat criteria is using a standardized grading system for complications. The most commonly used grading system is based on the work by Clavien and colleagues. In their pioneering investigation, the investigators systemically categorized postoperative complications into 4 grades according to their severity. In 2004, this grading system was updated by Dindo and colleagues, who modified these criteria to improve their accuracy and applicability to the surgical community.

The current grading system uses the following definitions:

• •
  

  Grade 0: absence of any complications

  
  
• •
  

  Grade 1: presence of any deviation from the normal postoperative course

  
  
• •
  

  Grade 2: management that includes not more than intravenous blood transfusion

  
  
• •
  

  Grade 3: complications that require surgical, endoscopic, or radiologic intervention

  
  
• •
  

  Grade 4: life-threatening complications requiring intensive care management

  
  
• •
  

  Grade 5: complications that cause the death of the patient

This system is notable for recording any deviation from the regular postoperative course as a complication. Previous studies showed that this grading system is easily applicable and reproducible in patients treated with RP. Many comparative assessments between ORP and RARP do not consider blood transfusions as complications but as a separate endpoint. As such, many investigators have argued that if blood transfusions were considered complications, most if not all evidence would show lower complication rates with RARP compared with ORP. Regardless, the adoption of these standardized evaluation tools in more recent publications facilitates the comparison of short-term outcomes of RARP and ORP. For example, Agarwal and colleagues demonstrated the safety of robotic surgery in a large cohort of patients with clinically localized PCa treated at a single referral tertiary center; this report represents one of the first efforts to use the standardized Martin-Donat criteria to examine morbidity and mortality after RARP.

Functional Outcomes

The use of clear definitions for potency and continence recovery is essential for comparing functional outcomes between patients treated with RARP and ORP. Indeed, the adoption of validated questionnaires, such as the International Index of Erectile Function (IIEF), has been widely advocated for the assessment of erectile function after surgery. Briganti and colleagues showed that an erectile function domain of the IIEF greater than or equal to 22 represents a reliable score for defining a satisfactory erectile function after radical prostatectomy. Therefore, such a definition should be applied when assessing the rates of erectile function recovery after ORP and RARP.

Similarly, when evaluating postoperative urinary continence, previous studies have demonstrated that the definition of incontinence substantially affected the rates of continence recovery after surgery. In an effort to define more stringent criteria for satisfactory continence, Liss and colleagues recently observed that patients reporting the use of 1 pad or more per day had a significant decrease in postoperative quality of life compared with their counterparts using no pad. Consequently, many investigators have advocated that urinary continence recovery after surgery should be strictly defined as the use of no pad. Additionally, the odds of functional outcomes recovery significantly vary over time; however, improvements in urinary continence and erectile function recovery after 36 months of follow-up are trifling. In consequence, a median follow-up of at least 3 years should be considered compulsory when comparing functional outcomes between open and minimally invasive surgery.

Oncologic Outcomes

When evaluating oncologic outcomes, the short follow-up in series of patients treated with RARP prevents investigators from comprehensively comparing cancer-specific mortality rates between the 2 surgical approaches. Given the indolent natural history of clinically localized PCa, a long-term follow-up is needed to assess important postoperative oncologic outcomes, such as metastasis-free survival and cancer-related mortality. In consequence, secondary endpoints have been considered.

Biochemical recurrence (BCR) represents one of the most frequently reported surrogate endpoints. It needs to be taken into consideration, however, that the rates of BCR are conditional to several confounders, such as preoperative and pathologic characteristics, length of follow-up, and use of adjuvant hormonal or radiation therapies. The pioneering study by Menon and colleagues was the first to report the 5-year BCR-free survival rates in a large cohort of patients treated with RARP alone (without adjuvant therapies), supporting the safety of this approach.

The presence of positive surgical margins at final pathology has also been proposed as a proxy for cancer control. Again, caution should be used when considering this endpoint. The impact of positive margins on the long-term risk of BCR and cancer-specific mortality remains controversial and is likely dependent on the presence of other adverse pathologic features at RP as well as patient life expectancy. Additionally, the rates of positive margins may depend more on surgical expertise and/or surgical technique (ie, aggressiveness of nerve sparing) rather than represent the quality of the approach (RARP vs ORP). The learning curve phenomenon may play a more significant role in patients treated with minimally invasive surgery, given the recent introduction of the robotic technique and the low cumulative case volume of early adopters. Consequently, the use of this endpoint may not be adequate to comprehensively assess the oncologic safety of robot-assisted versus open surgery.

Similar limitations apply when comparing the use of postoperative cancer-related therapies between patients treated with RARP or ORP. The administration of adjuvant radiotherapy and hormonal therapy strongly depends on disease characteristics at final pathology. For example, evidence from randomized trials supports the use of adjuvant androgen deprivation therapy after RP in patients with node-positive PCa. Because the likelihood of lymph node invasion at RP depends on the extent of the pelvic lymph node dissection performed and because patients treated with RARP are less likely to receive a lymph node dissection at RP, the use of postoperative cancer therapies may be higher in ORP patients when it really reflects more precise nodal staging. Moreover, given the lack of consensus on the benefits of adjuvant radiotherapy after RP, the selection of patients for adjuvant therapies relies immensely on patient-physician perceptions and preferences. Such considerations undermine the validity of this endpoint as a proxy of cancer control after RP.

Costs and Expenditures

In an era of heightened scrutiny for health care spending and resource allocation, treatment-associated expenditures represent an important endpoint when comparing RARP and ORP. One of the purported disadvantages related to the adoption of minimally invasive surgery is the substantially higher costs associated with the purchase of robotic equipment and the use of disposables. On the other hand, several investigators have postulated that shorter length of stay and lower rates of transfusions may result in lower costs of RARP compared with ORP in the early postoperative setting. The absence of prospective studies, however, comparing the costs and expenditures associated with RARP and ORP limits the ability to fully grasp these interesting hypotheses. Moreover, the proposed savings from a shortened length of hospitalization and lower transfusion rates rely on health care provider reimbursement policies, which may vary from one country to another. The potential benefits of RARP with regard to lower rates of positive surgical margins and use of additional cancer therapies would also result in substantial savings for the health care system in the long term ; however, prospective studies incorporating these endpoints are needed to fully address this.

Retrospective Studies from Tertiary Referral Centers

Several retrospective studies from high-volume tertiary referral centers showing better short-term postoperative outcomes for patients undergoing RARP fueled the initial enthusiasm for minimally invasive approaches for PCa surgery. These data should be interpreted with caution, however. First, results obtained from high-volume hospitals and surgeons may not be applicable to the broader general population, because most patients are treated at community hospitals. Second, the introduction of a robotics training program at high-volume centers is associated with more stringent patient selection with regard to preoperative disease and patient characteristics. Specifically, patients with more favorable disease and health are more likely to be selected to the novel approach (ie, minimally invasive surgery). Such selection bias may result in better short- and long-term outcomes for patients treated with minimally invasive approaches compared with their open counterparts. Some of these selection biases are often unrecorded in retrospective observational studies, thus may undermine the validity of retrospective comparative investigations, because these unmeasured confounders may exert strong influence on the outcomes, to an extent that any statistical adjustment would not appropriately mitigate.

Several systematic reviews and meta-analyses based on these retrospective data from high-volume centers showed significant benefits for RARP with regard to perioperative outcomes, functional results, and oncologic endpoints. Unfortunately, the aforementioned limitations apply also to these investigations, in addition to the usual publication bias for positive studies. Consequently, despite the high number of patients evaluated, these meta-analyses do not provide a definitive and compelling answer to the question, Which surgical approach is best for the treatment of patients with clinically localized PCa? If it is assumed that adjustment for case mix is appropriately performed (which is not a given), the best interpretation of these data may be that the outcomes of the best RARP series (or surgeons) are better than those of the best ORP series (or surgeons).

Population-Based Studies

As discussed previously, data from high-volume referral centers may not be generalizable to the overall population. In the absence of randomized controlled trials, several investigators give credence to population-based analyses evaluating the comparative effectiveness of RARP versus ORP in large contemporary cohorts of patients with clinically localized PCa. Such studies allow for comparison of competing therapies across a broad range of health care settings.

Results obtained from these data differ somewhat from those originating from high-volume referral centers. In their assessment of a large population-based cohort of patients within the Nationwide Inpatient Sample, Trinh and colleagues were able to demonstrate superiority of RARP over ORP for virtually all perioperative outcomes. In a landmark investigation by Hu and colleagues, however, no differences were observed between ORP and RARP with regard to perioperative outcomes and long-term functional results when evaluating a population aged 65 years or older enrolled in Medicare. Gandaglia and colleagues examined postoperative complications and use of additional cancer treatments in a more contemporary cohort of Medicare beneficiaries and corroborated the results of the Hu and colleagues’ study. Similarly, Barry and colleagues compared the odds of problems with continence and sexual function after RARP and ORP in Medicare beneficiaries treated between 2008 and 2009 using rigorous survey instruments and validated questionnaires. They observed that robotic surgery was associated with a nonsignificant trend toward greater problems with urinary continence. Additionally, the adoption of RARP was not associated with better erectile function recovery rates at a median follow-up of 14 months.

The reasons for such discrepancy may reside in preoperative case mix, where individuals included in these population-based studies may be older and sicker compared with their counterparts treated in referral centers. Many of these studies originate from Medicare enrollees, who are by definition older than 65 years of age. This may also explain why Trinh and colleagues’ study showed a benefit for RARP with regard to perioperative outcomes, whereas the other studies did not. Additionally, the learning curve phenomenon may be more influential at a population-based level. From a practical prospective, when results of institutional series conflict with large population-based studies, these findings highlight the importance of the surgeon rather than the surgical approach. For example, relying on Nationwide Inpatient Sample data that showed that RARP was associated on average with better perioperative outcomes than ORP, Sammon and colleagues nevertheless observed that ORP performed at high-volume hospitals had better outcomes than RARP performed at low-volume hospitals.

Statistical Methodology Applied in Retrospective Studies Comparing the Two Techniques

Several efforts have been made to limit the effect of selection bias in retrospective observational studies comparing RARP with ORP. Several advanced statistical tools have been applied to minimize the effect of confounders. For example, propensity score matching represents a commonly used approach in observational retrospective investigations. This method allows the selection of control subjects matched with treated subjects for readily available covariates, which, if unaccounted for, lead to biased estimates of treatment effects. When matching is performed, the covariates in the control and treatment groups are balanced after the matching process. Thus, analyses performed on the postpropensity score–matched population should lead to theoretically unbiased comparisons between postoperative outcomes of the 2 surgical techniques. Many investigators think, however, that although the effect of measured confounders is minimized with propensity score matching, the effect of unmeasured confounders may be amplified.

Another statistical method gaining traction in the field is the instrumental variable analysis. This approach claims to perform pseudorandomization by accounting for both measured and unmeasured confounders. By definition, an instrumental variable should be associated with the odds of receiving the treatment of interest (eg, RARP) but should not be associated with the analyzed endpoint (eg, cancer-specific survival) except through the choice of treatment. Examples of instrumental variables that could be used to compare RARP with ORP are the density of RARP cases performed in a given area, the distance to the closest hospital performing RARP, or even physician-level preference for RARP. These instruments are conceptually sound; however, the quality of the instrument must always be verified using statistical calculations, such as the F -statistic. For example, the density of RARPs performed in a given area are expected to influence the choice of treatment; however, that variable is not expected to affect the endpoint (for example, postoperative complications) except through the choice of treatment (RARP vs ORP). The instrumental variable is subsequently used for pseudorandomization, thereby allowing estimation of the effect of a certain treatment on the marginal population (eg, individuals for whom the likelihood of undergoing the treatment is based on the instrumental variable).

Although these statistical tools may limit the effect of selection bias when applied correctly, data from observational studies will never be as compelling as evidence from well-designed prospective randomized trials.