Introduction:

The current guidelines of the European Association of Urology recommend an extended pelvic lymph node dissection (external iliac nodes, obturator nodes and internal iliac nodes up to the ureter crossing) in patients with a nomogram-based risk of nodal metastasis of higher than 5%. However, this recommendation is mainly based on the staging benefit of extended pelvic lymph node dissection (PLND) and not on any oncological outcome improvements as recently shown by a systematic review. The majority of the currently available literature compares patients who underwent PLND (extent often insufficiently classified) to patients that did not undergo PLND. Studies incorporating the anatomical extent of PLND are scarce and mainly underpowered due to a limited number of patients and/or short follow-up time. Furthermore, several studies lack a sufficient analytical framework to draw causal inference regarding the oncological benefit of extended versus non-extended PLND. Therefore, the aim of this study was to quantify the oncologic benefit of extended versus non-extended PLND in patients undergoing radical prostatectomy for localized prostate cancer. Specifically, we aimed to identify the direct effect of extended over non-extended PLND (i.e. the removal of occult micro-metastases) that is not mediated through the detection of nodal disease and potential adjuvant therapy (indirect effect).

Methods:

A retrospective bi-center cohort study involving a consecutive series of patients who underwent radical prostatectomy and PLND for localized prostate cancer between January 2006 and December 2016 in two tertiary referral centers with differing PLND templates (extended: University Hospital of Zurich, Zurich, Switzerland, non-extended: University Health Network, Toronto, Ontario, Canada) was performed. Patients were followed until April 2018 for the occurrence of either biochemical recurrence or secondary therapy (composite outcome). To compare typical extended PLND patterns of the Swiss center to typical non-extended PLND patterns of the Canadian center, we excluded all surgeons who performed less than ten PLNDs during the study period as well as surgeons whose PLND template is not reflective of the center. Furthermore, we restricted the cohort to patients whose pelvic lymph node yield was between the 20th and the 80th percentile of the surgeon who performed the procedure. Time to biochemical recurrence or secondary therapy between extended and non-extended PLND was compared by Kaplan-Meier analysis/log-rank test and Cox proportional hazards regression, both unweighted and weighted by inverse probability weights. Balance of the weighted pseudo-cohort was assessed by the standardized difference of the mean (less than 0.2). The direct effect of extended over non-extended that is not mediated through the detection of nodal disease and potential adjuvant therapy was investigated by causal mediation analysis (natural effects Cox model with 5,000 bootstrap replications).

Results:

During the study period of eleven years, 3,923 patients with available follow-up data were identified (Switzerland: 958 [24.4%]; Canada: 2,965 [75.6%]). After exclusion of 271 patients with missing covariates and 1,942 patients who did not receive PLND, a preliminary cohort of 1,710 males was available for the exploration of PLND patterns. Applying our above-mentioned PLND definitions yielded a final cohort of 1008 patients (extended PLND: 368 [36.5%]; non-extended PLND: 640 [63.5%]) and also confirmed our hypothesis regarding differential PLND patterns between two centers (see Figure 1). Unweighted and weighted survival analysis demonstrated results in favor of extended PLND (unweighted hazard ratio (HR): 0.77 [95% confidence interval: 0.59-1.01], p=0.056; weighted HR: 0.75 [0.56-0.99], p=0.044; see Figure 2). We observed a well-balanced weighted pseudo-population with standardized differences of the mean of less than 0.2. The causal mediation analysis confirmed the total effect of 0.77 [0.71-0.82]. After disentangling this total effect into an indirect effect (via detection of nodal disease and potential adjuvant therapy) and a direct effect (via removal of occult micro-metastases), we an even more protective direct effect of 0.69 [0.63-0.75].

Conclusion:

Among patients undergoing radical prostatectomy and PLND for localized prostate cancer, extended versus non-extended PLND seems to be beneficial from an oncologic perspective. In our causal mediation analysis, the protective effect was even more pronounced after accounting for therapeutic impacts of improved staging. To our knowledge, this is the first study that used a purely causal inference-driven approach including causal mediation analysis to assess the oncologic benefit of extended versus non-extended PLND. Our results not only indicate the limited utility of non-extended PLND but also that the effect of extended PLND is not restricted to a staging benefit and probably involves a therapeutic benefit mediated through the removal of occult micro-metastases. This study is obviously limited by its retrospective nature, unmeasured residual confounding, and the lack of power for mortality outcomes. Our results need to be confirmed in prospective studies sufficiently powered for mortality outcomes.

Funding: N/A