The Impact of Lymph Node Dissection on Survival in Intermediate- and High-Risk Prostate Cancer: A Population-Based Study

Zhen Zhou Shanghai Tenth People's Hospital Shi-Yong Xin Henan University of Science and Technology Ting-Shuai Zhai Shanghai Tenth People's Hospital Liang Jin Shanghai Tenth People's Hospital Xiang Liu Shanghai Tenth People's Hospital Jia-Xin Zhang Shanghai Tenth People's Hospital Xian-Chao Sun Shanghai Tenth People's Hospital Wang-Li Mei Shanghai Tenth People's Hospital Bi-Hui Zhang Shanghai Tenth People's Hospital Xu-Dong Yao Shanghai Tenth People's Hospital Lin Ye (  ericyelin@163.com ) Shanghai Tenth People's Hospital


Introduction
Prostate cancer (PCa) is a serious disease that is harmful to men's health worldwide, ranking rst in cancer incidence and second in cancer mortality for males in the United States [1] . To present, radical prostatectomy (RP) remains the main treatment option for D'Amico intermediate-and high-risk PCa according to European Association of Urology (EAU) guidelines [2][3][4] . Besides, the guidelines recommend pelvic lymph node dissection (PLND) in patients with a risk of nodal metastases over 5% [5,6] . Moreover, PLND refers speci cally to extended PLND (ePLND) [7] . However, the curative effect of PLND is controversial. It is widely believed that PLND provides important staging and prognosis information that is unmatched by any other currently available procedure [8] . Moreover, several reports demonstrated a potential therapeutic effect of PLND in select patient with presence of lymph node involvement (LNI) [9,10] . Conversely a recent systematic review has shown that operating PLND during RP can't improve oncological outcomes, including survival [8] . Besides, the disadvantages of PLND are obvious, which refer to longer surgery time and more importantly, greater morbidity, such as formation of lymphoceles, thromboembolic, or neurovascular events [11,12] .
In intermediate-risk PCa, the estimated risk of having positive lymph nodes (LNs) is between 3.7-20.1%, and the risk for positive LNs in high-risk localisted PCa is 15-40% [13] . In locally advanced PCa, a PLND is considered the standard procedure during RP, but clinical nodal involvement (cN + ) was not a signi cant predictor of cancer-speci c survival (CSS) [14] . Besides, there is no report about the effect of PLND on survival in locally advanced PCa patients. According to a latest article [15] of American Urological Association (AUA), it analyzed 9,742 patients from 4 centers and demonstrated that there was no signi cant difference in CSS, biochemical recurrence (BCR) and LNs metastasis in patients with or without PLND at RP.
In the absence of prospective, randomized trials and studies on the role of PLND in contemporary patients, we sought to elucidate its potential curative value of PLND by retrospective analysis.
Speci cally, we tried to analyze which clinical or pathological factors might be associated with LNI.
Moreover, we compared the oncologic outcomes in patients between limited PLND (lPLND) and ePLND according to different risk stages.

Database
The data on PCa patients over seven years (2010)(2011)(2012)(2013)(2014)(2015)(2016) were selected from SEER database. The SEER*Stat software program (version 8.3.7) was used to collect all data. Data on about 28% of the U.S. population is stored in the SEER database. The data of this work came from the following resources available in the public domain: SEER database. There was no direct information about characteristics of D'Amico intermediate or high risk PCa in the SEER database, but we got that information indirectly by ltering it according to EAU guideline [16] . All the AJCC TNM stage was clinical diagnoses. In our study, due to the defects of SEER database, ePLND refers to dissection of more than three lymph nodes, and lPLND means the dissection of one to three lymph node(s). Patients Selection 97,924 patients who underwent RP between 2010 and 2016 were extracted in this study from SEER database. Only patients with D'Amico intermediate-risk stage and patients at high-risk stage were included in analysis. The information about age at diagnosis, sex, race, marital status, pathological grade, state of radiotherapy, state of chemotherapy, prostate-speci c antigen (PSA), Gleason score (GS), derived AJCC TNM stage (7th edition, 2010-2016) and pathological LNI condition was available. We excluded patients with unknown race, marital status, grade, PSA, GS or LNI. Eventually, 75,583 patients were enrolled in the cohort study (Fig. 1).

Statistical analysis
We utilized SPSS v25.0 (SPSS Inc., Chicago, IL, USA) for all of the statistical analyses of the data. The χ2 test was used to compare clinical characteristics between patient groups. Logistic regression analysis was used to investigate the in uences of different clinical and pathological factors on LNI. P values < 0.05 were considered statistically signi cant. Variables with P < 0.05 in univariate analysis were included in the nal multivariate analysis model. The multivariate Cox regression analysis was used to determine the association with OM rate and CSM. In order to account for potential important differences between patients with vs without PLND performed during RP, we relied on 1:1 nearest neighbor propensity score matching (PSM) [17] . We used a caliper of 0.1 in order to achieve a standardized mean difference in all relevant variables. Therefore, propensity-score-matched cohort was balanced according to clinical and pathological characteristics. Kaplan-Meier analysis was done to graphically depict overall survival (OS) and cancer-speci c survival (CSS) before and after PSM.  Abbreviations: PLND, pelvic lymph node dissection; PSA, prostate-speci c antigen; OM, overall mortality; CSM, cancer-speci c mortality; HR, hazard ratio; CI, con dence interval; Ref, reference.

Pelvic Lymph Node Dissection Effects
The impact of PLND on the general population After 1:1 PSM, Kaplan Meier analysis con rmed that the patients who received PLND had a poorer prognosis than those patients without PLND (P = 0.010 in OS, P = 0.038 in CSS, Fig. 2). Similar outcomes can be obtained in the cohort before adjustment (P < 0.001 in OS & CSS, Fig. 2). Bedsides, in the Kaplan-Meier analysis of the cases before PSM, we found that patients without PLND at RP had a better survival outcome when compared to patients who underwent ePLND or lPLND (All P < 0.001, Both OS and CSS, Fig. 2). But there was no signi cant difference between patients with ePLND and lPLND (All P > 0.05, Both OS and CSS). However, in multivariable Cox regression models after adjustment for clinical and pathological characteristics, PLND was not an independent predict factor for OS and CSS (OM: HR 1.03, 95% CI 0.90-1.19, P = 0.667; CSM: HR 0.75, 95% CI 0.55-1.03, P = 0.075. Table 1).  Figure 3). However, in locally advanced PCa, the patients with PLND at RP had a poorer survival outcome than patients without PLND (All P < 0.001, Both OS and CSS. Figure 3). When compared to patients without PLND at RP, the multivariable Cox regression models showed that PLND was an independent risk factor of locally advanced PCa (OS: HR 1.67, 95% CI 1.36-2.06, P < 0.001; CSS: HR 2.26, 95% CI 1.63-3.12, P < 0.001. Table 2). When No PLND, ePLND, and lPLND were compared in pairs, the K-M analysis determined that patients without PLND had a better survival prognosis than patients with ePLND or lPLND (P(NvsE) < 0.001, P(NvsL) < 0.001, Both OS and CSS. Figure 3), and that there was no signi cant difference between patients treated with expanded and limited lymph node dissection (P = 0.262 in OS and P = 0.692 in CSS). The multivariable Cox regression models showed the same results in locally advanced PCa when compared to patients without PLND (All P < 0.001, Both OM and CSM. Table 3 and Fig. 1).

Discussion
There is no deny that PLND plays an important role in PCa staging, but its potential curative value is controversial, and prospective trials of PLND are still missing. In view of the current research status, this paper had made some achievements, but also found some problems in it.
PLND is recommended as a standard surgical procedure at RP in patients with a risk of nodal metastases over 5%. Several studies showed a better CSS in patients treated with more ePLND at RP [18] . Other studies showed that PLND has no effect on patients' survival outcomes in intermediate-risk PCa. This conclusion is consistent with the results shown in our study only after balancing variables according to PLND and No PLND. One recent study showed that ePLND and lymph node yield were found no statistical link to BCR in patients with intermediate-risk PCa. However, some scholars reported that higher lymph node yield was associated with biochemical free recurrence survival mainly because of identi cation of an increased number of positive LNs [19] . In fact, due to the heterogeneity of intermediaterisk PCa, more work needs to be done to further characterize patients within this group and identify the minority of patients with a more favorable prognosis for which expectant treatments would be effective.
High-risk PCa actually includes any patient with a PSA > 20, Gleason score 8 or higher, clinical T2c, or a locally advanced cancer (≥ T3). In this context, we would like to call it high-risk PCa and locally advanced PCa, respectively. The EAU guidelines recommend that all patients with high-risk PCa should receive PLND when RP is planned. In high-risk stage of localised PCa, our data suggested that PLND or No PLND had no effect on OS and CSS in patients at RP. This is the rst report about the survival outcomes of highrisk stage of localised PCa that we know of so far. Surgical treatment has been traditionally discouraged in locally advanced PCa. But increasing recent evidence in literature push urologists to operate for RP in cT3 PCa patients assessing no LNI is shown [20] . In terms of whether performing PLND at RP in patients with locally advanced PCa, this is even less reported. In the present study, patients treated with PLND at RP had a worse survival outcome than patients without PLND. This may be due to the fact that patients in the advanced stage are inherently inoperable, especially when combined with PLND. Furthermore, it may be due to surgical complications associated with lymph node dissection. But at the same time, locally advanced PCa had a high risk of lymph node invasion. Therefore, in these speci c cases, individual based management must be discussed with the patients and tailored as part of a multiplex therapy. Due to the difference in recommendations within this patient category, urologists would be the center of decision making rather than following clear cut guidelines.
According to the relationship between the number of nodes removed and oncologic outcomes, a recent study reported that a large number of PLND was associated with a signi cant improvement in time to BCR [21] . Unfortunately, our ndings suggested that there was no signi cant difference between ePLND and lPLND. This may be due to limitations of SEER database, that we have to narrow the de nition of ePLND and lPLND.
Moreover, we assessed the impact of PLND on different populations according to age, sex, race, marital status, pathological grade, state of radiotherapy, state of chemotherapy, PSA, GS, T stage and N stage (Table 3). We found that in the majority of the population, patients with different subtypes did not bene t signi cantly from PLND compared to patients without PLND, and they had a worse OS and CSS. Although PLND has been reported to be useful for survival biochemical recurrence-free survival, its effect on actual survival of patients remains to be determined.
Finally, we investigated the predictive function of different clinical and pathologic characteristics for LNI, and found that the sensitivity and speci city of clinical lymphatic diagnosis were 10.9% and 99.9%, respectively. This result indicated the inadequacy of current clinical work, leading us to have to stage PCa by lymphatic biopsy or PLND. In addition, the phenomenon that PSA > 20ng/ml and GS 8-10 were signi cantly associated with LNI was consistent with clinical experience. We should be highly alert to the possibility of lymphatic invasion in in this group of patients.
At present, PLND at RP is performed blind, without knowledge of the presence of metastases. The conventional PLND template only covers 50-60% of the entire pelvic lymph node back ow, and the tumor cells in positive pelvic lymph nodes may be in hibernation and not lethal. Second, sometimes the initial metastatic lymph nodes are outside the pelvic cavity, such as sigmoid colon, mesentery, para-aortic lymph nodes, subclavian lymph nodes, and even the lung [22] . Therefore, whether to perform PLND at RP should take all the information of the patients into consideration and combine with the patients' will.
This article also has some inevitable limitations. First, our results came from retrospective observational data. Therefore, our ndings required prospective randomized validation. However, to our knowledge, no such trials are currently being recruited or conducted. Moreover, since the SEER database lacks post-operative follow-up information such as the time of BCR, it is di cult to fully assess the e cacy of PLND with a single indicator in intermediate-and high-risk PCa. Nevertheless, OS and CSS are also one of the best indicators to evaluate the prognosis of PCa patients. Finally, the determination of tumor's characteristics largely depends on the clinician's expertise, and the reasons for performing or not performing PLND are not clear. The possibility of these differences may further confuse our results.

Conclusions
In intermediate-or high-risk localised PCa, there was no signi cant difference in survival outcome in patients with or without PLND at RP. Although locally advanced PCa has a higher risk of lymph node involvement, patients treated with PLND has a higher overall mortality and cancer-speci c mortality risks compared to patients without PLND. Thus, for patients with locally advanced PCa, if radical prostatectomy is necessary, PLND is worthy of serious discussion or even been avoided.

Figure 1
The ow chart describes the steps taken to identify 75,583 patients in the SEER database.