Impact of the introduction of a robotic training programme on prostate cancer stage migration at a single tertiary referral centre

Authors


Correspondence: Alberto Briganti, Urological Research Institute, Department of Urology, Vita-Salute San Raffaele University, 20132, Via Olgettina 60, Milan, Italy.

e-mail: briganti.alberto@hsr.it

Abstract

What's known on the subject? and What does the study add?

  • Previous studies have retrospectively compared functional and oncological outcomes between robot-assisted radical prostatectomy (RARP) and open retropubic radical prostatectomy (RRP), reporting non-inferior or superior outcomes associated with the use of the robotic approach.
  • The study demonstrates that baseline selection biases makes comparisons of RARP and RRP problematic. Patients treated with RARP may have significantly better baseline characteristics than patients treated with RRP. Indeed, when both facilities are available within the same centre, patients with the most favourable clinical and cancer profile are inherently selected to RARP. In the absence of a prospective randomized trial, the decision to prefer one approach over another should be tailored according to each single patient and surgeon.

Objective

  • To evaluate the trend in robot-assisted radical prostatectomy (RARP) and open retropubic radical prostatectomy (RRP) use over time and to compare preoperative and pathological characteristics of patients treated with RARP or RRP at a single centre.

Patients and Methods

  • Between 2006 and 2010, 2511 consecutive patients treated with RP, with or without pelvic lymph node dissection (PLND), for prostate cancer (PCa) at a single tertiary care centre were analysed.
  • Baseline patient characteristics and PCa risk distribution were compared according to treatment type (RRP vs RARP) in the overall population, as well as in three surgeons' initial 50 RARP and three surgeons' initial 50 RRP cases (n = 300).
  • We used a chi-squared trend test to evaluate the differences in treatment type administration over time according to PCa characteristics. Logistic regression analyses focused on the prediction of PLND and adjuvant radiotherapy (RT) use.

Results

  • Overall, 1873 (74.6%) and 638 (25.4%) patients underwent RRP and RARP, respectively.
  • Men treated with RARP were younger (mean age: 62 vs 65 years), less obese (mean BMI: 24.8 vs 26.4 kg/m2), healthier (Charlson comorbidity index = 0: 68.7 vs 53.3%) and more likely to harbour clinical low-risk PCa (51 vs 30%) than their RRP counterparts (all P < 0.001). Similar findings were observed in sub-analyses focusing on six surgeons' 50 initial patients (all P ≤ 0.02).
  • A significant increase in the rate of patients with low-risk PCa treated with RARP vs RRP was reported over time (5 vs 95% and 66 vs 34% in 2006 and 2010, respectively). Conversely, 76% of patients with high risk PCa were still treated with RRP in 2010.
  • Patients treated with RARP were less likely to receive PLND at RP and adjuvant RT (all P ≤ 0.01), even after adjusting for clinical and PCa characteristics.

Conclusions

  • The introduction of a robotic training programme at a high volume centre led to significant patient selection in terms of clinical and PCa characteristics.
  • When both RRP and RARP facilities are available within the same centre, patients with the most favourable clinical and cancer profile are selected to undergo RARP.
  • Use of RARP negatively influenced the rates and the extent of PLND as well as the use of adjuvant RT after surgery. Thus, baseline patient selection, surgical and treatment biases make any comparisons of RARP with RRP problematic.

Introduction

Recently, a stage migration towards more aggressive prostate cancer (PCa) was reported for patients treated with open retropubic radical prostatectomy (RRP), both in European and North-American series [1, 2]. Specifically, the rate of patients treated with RRP and harbouring intermediate- and high-risk disease has been increasing during the most recent years [1, 2]. The most plausible explanation for this phenomenon may be the paradigm shift in the management of low-risk PCa. In this context, alternatives to standard RRP including both active surveillance and minimally invasive approaches have been increasingly implemented over time [3, 4].; however, while the long-term results of active surveillance are still under investigation, robot-assisted radical prostatectomy (RARP) has quickly become an unwritten standard of care in the treatment of low-risk disease, even in the absence of robust, long-term oncological comparative data. This is mainly because of more favourable peri-operative and early functional outcomes of patients treated with RARP compared with those receiving RRP [5-17]. Nevertheless, the vast majority of studies comparing RARP and RRP are based on retrospective series [5]. Moreover, these reports are limited by the lack of at least one of the following analyses: (i) extensive comparison of baseline clinical and functional characteristics between RARP and RRP patients; (ii) evaluation of the impact of surgical volume on patient outcome; and (iii) assessment of adjuvant therapy administration patterns between patients undergoing RARP and those undergoing RRP. If all these factors are not simultaneously accounted for, a major potential source of bias might be introduced, since each of the mentioned points strongly influences functional and oncological outcomes after surgery. For these reasons, several retrospective comparisons might be biased because of inherent, not measured factors. To address these limitations, we analysed the trend in use of RARP and RRP after the introduction of a robotic training programme at a single, high-volume centre.

Patients and Methods

Study Population

Between 2006 and 2010, 2511 consecutive patients surgically treated for PCa at a single tertiary care centre were considered. Of those, 1873 (74.6%) and 638 (25.4%) patients underwent RRP and RARP, respectively.

Patient Characteristics

We assessed the following patient characteristics: age at surgery, body mass index ([BMI] continuously coded) and Charlson comorbidity index ([CCI] either tested as a continuously coded variable or stratified according to three groups: 0 vs 1 vs ≥2). Preoperative erectile function was evaluated using the International Index of Erectile Function (IIEF), which was administered at the time of hospital admission. Patients with an IIEF score ≥ 26 were considered preoperatively potent [18].

Tumour and Treatment Characteristics

Tumour characteristics included serum PSA level at surgery (continuously coded), clinical stage (cT1 vs cT2 vs cT3), biopsy Gleason score (≤6 vs 7 vs 8–10), 2002 TNM pathological stage (pT2 vs pT3a vs pT3b vs pT4), pathological Gleason score (≤6 vs 7 vs 8–10), and surgical margin status (negative vs positive). For the purpose of these analyses, patients were divided into low- (cT1, PSA < 10 ng/mL, biopsy Gleason score ≤ 6; n = 889), high- (cT3 or PSA > 20 ng/mL or biopsy Gleason score ≥ 8; n = 489) and intermediate-risk (all the remaining patients; n = 1133) PCa, as previously described [19]. Other variables considered in our analyses included pelvic lymph node dissection (PLND) rate, as well as number of lymph nodes removed (continuously coded) and administration of adjuvant radiotherapy ([RT] yes vs no).

Statistical Analyses

In the first part of our analysis, we examined baseline patient characteristics and PCa risk distribution according to treatment type (RRP vs RARP). Descriptive statistics focused on frequencies and proportions for categorical variables. Means, medians and interquartile ranges (IQRs) were reported for continuously coded variables. Chi-squared and independent-sample t-tests were used to compare the statistical significance of differences in proportions and means, respectively. Subsequently, a chi-squared trend test was used to evaluate the differences in treatment type over time. The same analyses were then performed within patients harbouring low-, intermediate- and high-risk disease.

In the second part of our analysis, the same methodology was used to compare all patient characteristics of three surgeons' initial 50 RARP and three surgeons' initial 50 RRP cases (n = 300).

In the third part of our analysis, univariable and multivariable logistic regression analyses focused on prediction of administration of adjuvant RT, after accounting for clinical and cancer characteristics. The same analyses were repeated in patients with pT3a and pT3 disease separately. Finally, a chi-squared trend test was used to evaluate the differences in pelvic PLND rates, as well as in the number of lymph nodes removed between the two approaches over time.

We used univariable and multivariable logistic regression models to test the association between surgical approach (namely RARP vs RRP) and PLND rates. Covariates consisted of age at surgery, PCa risk group and CCI. All analyses were conducted using the R statistical package (version 2.13.2). All tests were two-sided, and a P value <0.05 was considered to indicate statistical significance.

Results

Baseline characteristics of the entire population are shown in Table 1. Men treated with RARP were younger (mean age: 65 vs 65 years), had a lower mean BMI (24.8 vs 26.4 kg/m2) and were healthier (CCI = 0: 68.7 vs 53.3%) than those treated with RRP (all P < 0.001) Moreover, patients treated with RARP were more likely to harbour clinical low-risk PCa (51 vs 30%, respectively; P < 0.001), and less likely to receive PLND (pNx: 49 vs 7%, respectively; P < 0.001). Among patients receiving PLND, the number of lymph nodes removed was significantly lower in patients who underwent RARP than in those who underwent RRP (mean number of nodes: 9 vs 20, respectively; P < 0.001). Finally, patients treated with RARP were less likely to receive adjuvant RT (7.8 vs 23.9%; P < 0.001), regardless of pT3 stage or surgical margin status (Table 2). Similarly, in sub-analyses considering only the initial 50 cases of six different surgeons performing either RARP (n = 3; 150 patients) or RRP (n = 3; 150 patients), RRP-treated patients were again significantly older (mean age: 66 vs 62 years), had a higher BMI (27 vs 26 kg/m2), and were more likely to have baseline illnesses (CCI ≥ 1: 52 vs 31.1%), and to harbour aggressive PCa (high risk PCa: 22 vs 5%; all P ≤ 0.02 [Table 3]).

Table 1. Descriptive characteristics of patients undergoing RRP or RARP for PCa between 2006 and 2010.
VariablesOverallOverall population
RRPRARPP
  1. *Analyses performed only in patients who underwent PLND.
No. of patients, n (%)2511 (100.0)1873 (74.6)638 (25.4)
Age at surgery, years   <0.001
Mean (median)64.4 (64.8)65.1 (65.7)62.4 (62.6)
IQR59.7–69.460.2–70.158.1–67.4
BMI, kg/m2   <0.001
Mean (median)26.2 (25.5)26.4 (26)24.8 (25)
IQR24.1–27.824.2–28.123.8–27.3
CCI   <0.001
01433 (57.1)998 (53.3)438 (68.7)
1829 (33.0)671 (35.8)156 (24.4)
≥2249 (9.9)204 (10.9)44 (6.9)
Preoperative IIEF erectile function domain score, n (%)   0.8
<261481 (59.0)1101 (58.8)379 (59.4)
≥261030 (41.0)772 (41.2)259 (40.6)
Risk groups, n (%)   <0.001
Low889 (35.4)566 (30.2)323 (50.6)
Intermediate1133 (45.1)855 (45.6)278 (43.6)
High489 (19.5)452 (24.1)37 (5.8)
PLND, n (%)   <0.001
No2079 (82.5)125 (6.7)314 (49.2)
Yes432 (17.5)1748 (93.3)324 (50.8)
No. of lymph nodes removed*   <0.001
Mean (median)18.3 (17)20.0 (19)9.2 (8)
IQR11–2414–255–12
Adjuvant RT, n (%)   <0.001
No2004 (79.8)1425 (76.1)588 (92.2)
Yes507 (23.9)448 (23.9)50 (7.8)
Table 2. Rates of adjuvant RT administration in patients with pT3a or pT3b PCa, stratified according to surgical approach.
VariablespT3apT3b
RRP, n (%)RARP, n (%)PRRP, n (%)RARP, n (%)P
Overall166 (52.0)26 (34.0)0.02211 (78.6)15 (50.0)0.004
Surgical margin status      
Negative73 (41.3)9 (20.0)0.0480 (73.5)10 (66.7)0.5
Positive87 (70.8)16 (50.0)0.04119 (82.1)4 (25.0)<0.001
Table 3. Descriptive characteristics of 300 patients undergoing RRP (n = 150) or RARP (n = 150) for PCa, representing the first 50 cases of six different surgeons, between 2006 and 2010.
VariablesFirst 300 cases
RRPRARPP
  1. *Analyses performed only in patients who underwent PLND.
No. patients, n (%)150 (50.0)150 (50.0)
Age at surgery, years  <0.001
Mean (median)66.2 (66.2)62.4 (63.2)
IQR61.9–70.457.5–67.7
BMI, kg/m2  0.02
Mean (median)26.6 (26.4)25.6 (24.8)
IQR24.5–28.723.6–27.3
CCI, n (%)  0.008
072 (48.0)103 (68.9)
158 (38.7)42 (27.8)
≥220 (13.3)5 (3.3)
Preoperative IIEF erectile function domain score, n (%)  <0.001
<26111 (73.7)87 (57.8)
≥2639 (26.3)63 (42.2)
Risk groups, n (%)  <0.001
Low-39 (26.0)84 (56.0)
Intermediate-78 (52.0)58 (38.7)
High-33 (22.0)8 (5.3)
PLND, n (%)  <0.001
No0 (0)60 (40.0)
Yes150 (100)90 (60.0)
No. of lymph nodes removed*  <0.001
Mean (median)14 (13)7 (7)
IQR8–174–8
Adjuvant RT, n (%)  <0.001
No118 (78.7)140 (93.1)
Yes32 (21.3)10 (6.9)

In temporal trend analyses, we examined the use of RARP after its introduction at our institution in 2006. A significant increase in the use of RARP was observed, mainly in patients with favourable preoperative PCa characteristics. The rate of patients with low-risk disease was 95 and 5% for RRP and RARP in 2006 vs 34 and 66% in 2010, respectively (Fig. 1A). A similar trend was also observed in patients with intermediate-risk disease (Fig. 1B). Conversely, 76% of patients with high-risk PCa were still treated with RRP in 2010 (Fig. 1C).

Figure 1.

Temporal trend analyses addressing the use of RRP and RARP over time in patients with A, low- B, intermediate- and C, high-risk PCa.

Finally, we assessed whether the use of RARP may have had an impact on the postoperative clinical decision-making process. After adjusting for all pathological characteristics, patients who underwent RARP were less likely to receive adjuvant RT (odds ratio [OR]: 0.41, CI: 0.25–0.66, P < 0.001). This observation was also confirmed among patients with pathological T3a (OR: 0.39, CI: 0.19–0.79, P = 0.01) and T3b disease only (OR: 0.26, CI: 0.10–0.68, P = 0.006). Within the latter groups, surgical approach was the only significant predictor of adjuvant RT administration (Table 4). Moreover, despite a significant increase in the rate of PLND among patients undergoing RARP (20% in 2006 vs 66% in 2010, P < 0.001; Fig. 2 and B), men treated with RARP had a significantly lower chance of receiving PLND, even after accounting for confounding factors (P < 0.001; Table 5). In addition, although an increasing trend was also observed when assessing the extent of PLND performed in men receiving RARP, the number of lymph nodes removed remained significantly lower in patients treated with RARP than in those receiving RRP (mean number of nodes removed in 2010: 11 vs 19, P < 0.001; Fig. 3).

Figure 2.

Temporal trend analyses addressing the rates of PLND over time among RRP and RARP groups, A, overall and B, according to PCa characteristics (namely low- vs intermediate-/high- risk disease).

Figure 3.

Temporal trend analyses addressing the mean number of lymph nodes removed over time according to surgical approach (namely RRP vs RARP).

Table 4. Multivariable logistic regression analyses predicting administration of adjuvant RT in the overall population, and within patients harbouring pT3a or pT3b disease, respectvely.
VariableOverallpT3apT3b
OR (95% CI)POR (95% CI)POR (95% CI)P
Age0.99 (0.96–1.01)0.30.98 (0.94–1.02)0.50.98 (0.93–1.03)0.4
Pathological stage      
pT21.0 (ref.)    
pT3a14.13 (9.43–21.18)<0.001
pT3b32.03 (19.45–52.76)<0.001    
pT46.17 (1.83–20.77)<0.001    
Pathological Gleason score1.0 (ref.)1.0 (ref.)1.0 (ref.)
≤62.94 (1.68–5.14)<0.0013.21 (1.21–8.44)0.026.82 (0.58–79.65)0.1
75.78 (2.97–11.25)<0.0017.04 (2.38–20.86)<0.08.89 (0.76–104.12)0.1
8–10   01  
Surgical margin status1.0 (ref.)1.0 (ref.)1.0 (ref.)
Negative4.34 (3.06–6.16)<0.0013.82 (2.23–6.56)<0.01.0 (0.49–2.03)0.9
Positive   01  
Surgical approach      
RRP1.0 (ref.)1.0 (ref.)1.0 (ref.)
RARP0.41 (0.25–0.66)<0.0010.39 (0.19–0.79)0.010.26 (0.10–0.68)0.006
Table 5. Multivariable logistic regression analyses predicting PLND in the overall population, and within patients harbouring low-, and intermediate-/high-risk PCa, respectively.
 OverallLow-risk PCaIntermediate- /High-risk PCa
OR (95% CI)POR (95% CI)POR (95% CI)P
Variable      
Age1.01 (0.99–1.03)0.31.00 (0.97–1.03)0.91.02 (0.99–1.05)0.2
Risk groups      
Low1.0 (ref.)
Intermediate/High5.60 (4.28–7.33)<0.001    
CCI      
01.0 (ref.)1.0 (ref.)1.0 (ref.)
11.24 (0.85–1.82)0.31.59 (0.96–2.64)0.070.93 (0.52–1.65)0.8
≥20.89 (0.50–1.58)0.70.82 (0.39–1.74)0.61.02 (0.40–2.62)0.9
Surgical approach      
RRP1.0 (ref.) 1.0 (ref.)
RARP0.08 (0.06–0.11)<0.0010.07 (0.05–0.10)<0.0010.10 (0.07–0.15)<0.001

Discussion

In the absence of a prospective, randomized trial, retrospective observational comparisons of RARP vs RRP on postoperative outcomes are abundant [5-17, 20-24]. These studies add insight to the topic and underline the potential benefits associated with the use of the robotic approach [5-17], but results from these reports should be interpreted with caution. Differences in clinical and cancer characteristics among patients who undergo RRP and RARP, as well as in surgical expertise among surgeons, may represent inherent limitations that could bias these comparative analyses. To examine the extent of these potential biases, we analysed the patterns of use of RARP and RRP over time at a single high-volume centre.

Several results of the present study are noteworthy. First, the results confirm a significant shift towards the use of RARP in contemporary patients diagnosed with clinically localized PCa. This shift was especially evident in men with low-risk disease (Fig. 1A). However, despite an observed trend towards an increased use of RARP also in patients with unfavourable clinical characteristics in the most recent years, 76% of patients with high risk PCa were still treated with RRP in 2010 (Fig. 1C). Moreover, patients treated with RRP were more likely to have comorbidities, to be older, and to have a higher BMI than their RARP counterparts. These results were also confirmed when considering only the first 50 cases of different treating physicians performing their initial cases using either RARP or RRP (Table 3). The present results, therefore, seem to suggest that when both facilities are available within the same centre, patients with the most favourable clinical and cancer profile are inherently selected to RARP. Conversely, those patients treated with RRP were those with more aggressive disease and poorer performance status. Such patients are by definition at higher risk of cancer recurrence, and less likely to completely recover their functional status after surgery, regardless of the type of surgery used [25]. For these reasons, any comparison between RRP and RARP which does not account for baseline clinical and cancer characteristics might be misleading and biased in favour of RARP. Having said this, one might argue that the effect of these differences might be accounted for in multivariable analyses, but some key variables strongly associated with patient outcome, such as surgical expertise, are hardly quantifiable in this setting. For example, beginner RARP surgeons often come with significant baseline experience since they represent seniors or staff urologists who recently converted to the robotic approach, as was the case in the present study. Conversely, beginner RRP surgeons are most often urologists in training, with an average of only 1–2 cases annually in the USA [26]; therefore, those less experienced surgeons performing RRP are challenged upfront with more advanced and aggressive cases. By contrast, many patients with low-risk disease who might not even need any treatment for PCa are operated on by highly experienced, well trained surgeons. Such paradoxical clinical scenarios might contribute to an even greater inflation of the differences between RARP and RRP outcomes, given the well-known association between surgical volume and cancer control, especially in the high risk setting [27].

Second, the present data seem to support another potential risk associated with the introduction of RARP: the type of surgical approach (instead of evidence-based data) significantly influenced the clinical decision-making process. For example, patients treated with RARP were less likely to undergo PLND during surgery (51 vs 93%, respectively, P < 0.001). Although this is to be expected owing to the higher rate of patients with low-risk disease treated robotically, these results were confirmed at multivariable analyses even after accounting for PCa characteristics. The use of RARP, therefore, was independently associated with lower rates of PLND. Moreover, even when PLND was performed, a significantly lower number of lymph nodes removed during RARP was found as compared with RRP (mean nodes removed: 9 vs 20, P < 0.001). Such low numbers of lymph nodes retrieved during RARP suggests the use of limited PLND. A possible explanation for this might be a certain reluctance by beginner RARP surgeons to perform extended PLND because of the potentially higher risk of vascular complications. This conflicts with the recommendations given by virtually all PCa guidelines, which indicate that whenever a staging PLND is indicated, this should be extended [19, 28]. Having said this, the present data also show that, with increased robotic experience, this limited approach tended to disappear. Nevertheless, despite the observed trend towards more extended PLNDs during RARP in the most recent years, the difference in the number of lymph nodes removed between RRP and RARP in 2010 was still significant (19 vs 11 nodes removed, respectively, P < 0.001). While there is no doubt that anatomical extended PLND is absolutely feasible using a robotic approach [29], time is still needed before the two techniques can be compared without incurring surgical selection biases. Similarly, RARP-treated patients were also less likely to receive adjuvant RT, even after adjusting for PCa characteristics (OR 0.41, P < 0.001). In patients with pT3 disease, surgical approach emerged as a significant predictor of adjuvant RT administration, despite the presence of robust prospective randomized data supporting the role of adjuvant RT in this patient group [30]. Although these observations may certainly be worrying, it is possible that RARP surgeons might be reluctant to refer younger and healthier men to adjuvant RT because of concerns related to post-RT functional impairments. As such, it is possible that accurate PSA monitoring after surgery with administration of early salvage RT at the first PSA rise would represent the first choice for most of these men. Even though such an approach can be considered reasonable, prospective randomized data supporting the equivalence of adjuvant and early salvage RT on PCa outcomes are still awaited. Moreover, these differences in post-surgical RT administration rates might also further inflate differences in functional outcomes between RARP and RRP, in favour of the former. Despite this potential effect, the impact of use of adjuvant RT according to surgical approach has not been accounted for in any of the previous comparative trials [5]. While the endpoint of the present study was not to demonstrate the benefit of PLND or adjuvant RT on cancer outcome, the results certainly show a significant discrepancy in the treatment of PCa according to the type of surgical approach used, even within the same centre.

Despite its novelty, the present study has some limitations. First, we cannot demonstrate that the selection biases observed in our patient population are applicable to other PCa series. Second, we were unable to perform a per-surgeon analysis in order to determine whether such differences in PCa management were applied by each single surgeon when performing either RRP or RARP. Despite these limitations, the data strongly support the need for a prospective, randomized trial before a fair comparison between RRP and RARP can be performed. Otherwise, it is at least necessary that at tertiary care high-volume centres, the observed trend towards a similar case mix distribution between RRP and RARP reaches a plateau. According to the data, it is conceivable that this will happen in the next few years, if it has not already happened at institutions with even larger robotic series; however, until these data are available, any realistic comparison of RARP and RRP remains problematic.

In conclusion, the introduction of a robotic training programme at a high-volume centre led to significant patient selection in terms of clinical and PCa characteristics. Based on the present results, it seems that when both facilities are available within the same centre, patients with the most favourable clinical and cancer profile are selected to undergo RARP. Furthermore, use of RARP negatively influenced the rate and the extent of PLND as well as the use of adjuvant RT after surgery, regardless of disease characteristics. For all these reasons, retrospective comparisons not accounting for all these data might be biased owing to inherent unmeasured factors.

Conflict of Interest

None declared.

Abbreviations
RARP

robot-assisted radical prostatectomy

RRP

open retropubic radical prostatectomy

PLND

pelvic lymph node dissection

PCa

prostate cancer

RT

radiotherapy

IQR

interquartile range

BMI

body mass index

CCI

Charlson comorbidity index

IIEF

International Index of Erectile Function

OR

odds ratio

Ancillary