Prognosis of patients with pelvic lymph node (LN) metastasis after radical prostatectomy: Value of extranodal extension and size of the largest LN metastasis

Authors


  • N.M.P. and H.F. contributed equally.

Abstract

Objective

  • To assess the prognostic role of extranodal extension (ENE) and the size of the largest lymph node (LN) metastasis in predicting early biochemical relapse (eBCR) in patients with LN metastasis after radical prostatectomy (RP).

Patients and Methods

  • We evaluated BCR-free survival in men with LN metastases after RP and pelvic LN dissection performed in six high-volume centres.
  • Multivariable Cox regression tested the role of ENE and diameter of largest LN metastasis in predicting eBCR after adjusting for clinicopathological variables.
  • We compared the discrimination of multivariable models including ENE, the size of largest LN metastasis and the number of positive LNs.

Results

  • Overall, 484 patients were included. The median (interquartile range, IQR) follow-up was 16.1 (6–27.5) months. The median (IQR) number of removed LNs was 10 (4–14), and the median (IQR) number of positive LNs was 1 (1–2).
  • ENE was present in 280 (58%) patients, and 211 (44%) had their largest metastasis >10 mm. Patients with ENE and/or largest metastasis of >10 mm had significantly worse eBCR-free survival (all P < 0.01).
  • On multivariable analysis, number of positive LNs (≤2 vs >2) and the diameter of LN metastasis (≤10 vs >10 mm), but not ENE, were significant predictors of eBCR (all P < 0.003).
  • ENE and diameter of LN metastasis increased the area under the curve of a baseline multivariable model (0.663) by 0.016 points.

Conclusions

  • The diameter of the largest LN metastasis and the number of positive LNs are independent predictors of eBCR.
  • Considered together, ENE and the diameter of the largest LN metastasis have less discrimination than the number of positive LNs.

Introduction

The prognosis of patients with prostate cancer and lymph node (LN) involvement after radical prostatectomy (RP) is highly variable [1]. Risk stratification could improve clinical decision-making and personalised post-surgical management, i.e. adjuvant androgen-deprivation therapy (ADT) and/or radiotherapy [2, 3]. Indeed, the only randomised clinical trial available to date, showed that adjuvant ADT improved survival in patients with mostly gross nodal disease and adverse pathological features [2]. However, the recent push to more extended pelvic LN dissection (PLND) has increased the number of patients with microscopic or limited nodal involvement [4, 5], and such patients might be spared unnecessary adjuvant therapy and its side-effects. To address this question, several studies have suggested that patients with LN metastases can be stratified according to readily available and well established pathological data, e.g. the number of positive LNs [4-10]. Other parameters including the dimension of the LN metastasis and presence of extranodal extension (ENE) remain controversial [11-16]. However, to date, we do not yet know the best way to classify these patients. The most recent TNM classification of 2010 excluded sub-stratification of LN-positive patients according to the size of the metastatic LN, previously adopted in the 1992 version [17]. However, recent data suggests that in patients with positive LNs after RP the metastatic nodal burden is a crucial prognostic factor, while the role of ENE still has a conflicting role [11-16]. Furthermore, in a contemporary series from a tertiary care centre, ENE and the diameter of the largest LN metastasis did not significantly increase the discrimination of a model predicting recurrence [18]. The aim of the present study was to investigate the prognostic roles of the diameter of the largest LN metastasis and ENE, and then compare these variables with a classification based on the number of positive LNs.

Patients and Methods

The present study was conducted following Institutional Review Board approval and all participating centres provided the necessary data and agreements. We collected data on 571 patients with pathologically confirmed positive LNs after RP and bilateral PLND that underwent surgery in six centres between February 2002 and September 2011. We decided to exclude from our study population 87 patients that underwent adjuvant ADT due to lack of accurate treatment regimen. The final cohort included 484 patients. Overall, 376 men (78%) underwent open RP, while the remaining 108 (22%) underwent minimally invasive RP. Preoperative staging included abdominopelvic CT and bone scan, both negative for metastatic disease or extrapelvic LN involvement. The extent of the PLND was conducted at the surgeon's discretion. We included only patients that did not receive any preoperative neoadjuvant hormonal or radiation therapy. After RP, patients were generally followed with PSA measurement and physical examination at 6 weeks, then every 6 months for 5 years and annually thereafter. Overall, 76 (16%) patients underwent adjuvant radiotherapy after RP. Radiation therapy was undertaken after an extensive discussion of all possible risks and benefits between the attending physician and the patient. Radiation doses generally consisted in 50 Gy to the prostatic fossa and periprostatic tissue, with an associated boost of 10–14 Gy to the prostatic fossa alone. There were no statistically significant differences in clinical and pathological characteristics between men that underwent adjuvant radiotherapy and those that did not.

Pathological Analysis

All LN specimens were separately sent for permanent section pathological analysis; frozen-section analysis was not used. After fixation in 10% neutral buffered formalin, the LNs were dissected and manually counted by the pathologists. Each identified LN was cut when appropriate, embedded in paraffin, sectioned at 5 μm, stained with haematoxylin and eosin, and examined by light microscopy. No immunohistochemical stains for keratin or PSA were used. ENE was defined as clear-cut perforation of the LN capsule by tumour cells infiltrating peri-nodal tissue. Furthermore, the diameter of the largest LN metastasis was recorded, that is the greatest diameter of the largest LN affected by metastatic disease.

Statistical Analysis

The statistical analyses were conducted in different steps. First, the Kaplan–Meier method was used to assess early biochemical relapse (eBCR)-free survival rate at 1, 2 and 5 years after RP. eBCR was defined as two consecutive PSA values of >0.2 ng/mL. The number of positive LNs was dichotomised as ≤2 vs >2, as previously described [4]. The diameter of the largest positive LN metastasis was dichotomised as ≤10 vs >10 mm based on the results of a previous study [14]. The log-rank test was used to compare eBCR-free survival rates of patients sub-divided according to the dichotomised number of positive LNs, presence of ENE, dichotomised diameter of largest positive LN, and a combination of ENE and the largest diameter of positive LNs (both ENE and diameter >10 mm, vs either ENE or diameter >10mm, vs no ENE and diameter ≤10 mm). Second, we tested the association between eBCR, ENE, and positive LN diameter in a multivariable Cox regression model, after having accounted for the effects of age at RP, preoperative PSA level, pathological stage, pathological Gleason score (dichotomised as ≤7 vs 8–10), surgical margins, number of positive LNs (dichotomised as ≤2 vs >2) and adjuvant radiation therapy. As only 13 patients had a pathological Gleason score of 6, we incorporated them with those with a Gleason score of 7. Finally, using Harrel's concordance index, we calculated the discrimination of a baseline multivariable model predicting eBCR and compared it with a model including ENE and diameter of largest positive LN and with a model containing the dichotomised number of positive LNs. The concordance index analysis allows us to test whether or not the addition of a marker to a baseline model increases the predictive accuracy of the model itself [19]. The baseline model included age at surgery, preoperative PSA level, pathological stage, pathological Gleason score (dichotomised as ≤7 vs 8–10), surgical margins and adjuvant radiation therapy. We corrected for optimism using 10-fold cross validation. Additionally, we explored the role of ENE and LN metastasis diameter with multivariable analysis in patients with ≤2 (410 patients) and in those with >2 positive LNs (74). All statistical tests were performed with the use of STATA version 12 (Stata Corp, College Station, TX, USA). All tests were two-sided with a significance level at 0.05.

Results

Descriptive Statistics

Clinical and pathological characteristics are shown in Table 1. Being a high-risk population, the vast majority of men (87%) had extraprostatic extension, 52% had involvement of the seminal vesicles, and 44% had a pathological Gleason score of ≥8. The median (interquartile range, IQR) number of removed LNs was 10 (4–14), and half of the patients had only one positive LN (IQR 1–2). Stratifying patients according to the number of positive LNs, 85% of them had ≤2 metastatic LNs, while only 15% had ≥3. Overall, 58% of the men had ENE and 44% had a diameter of their LN metastasis of >10 mm.

Table 1. Clinical and pathological characteristic of the population
VariableValue
Number of patients484
Median (IQR) age at surgery, years61 (56, 66)
Median (IQR) preoperative PSA level, ng/mL9.0 (5.8, 17.0)
N (%): 
Clinical stage: 
T1 b,c106 (22)
T2 a,b,c249 (51)
T382 (17)
T42 (0)
Unknown45 (9)
Biopsy Gleason score: 
≤6194 (40)
7186 (38)
8–10104 (21)
Pathological stage: 
pT262 (13)
pT3a170 (35)
pT3b252 (52)
Pathological Gleason score: 
≤7270 (56)
8–10214 (44)
Surgical margins: 
Negative313 (65)
Positive171 (35)
Median (IQR) number of removed LNs10 (4, 14)
Median, (IQR) number of positive LNs1 (1, 2)
N (%): 
Positive LN threshold: 
≤2 positive LNs410 (85)
>2 positive LNs74 (15)
ENE: 
No204 (42)
Yes280 (58)
Diameter of largest positive LN, mm: 
<10273 (56)
≥10211 (44)
Adjuvant radiotherapy: 
No408 (84)
Yes76 (16)

eBCR-free Survival Rates

Overall, 275 patients (57%) developed eBCR after RP. The median (IQR) follow-up was 16.1 (6–27.5) months. eBCR-free survival rates at 6, 12, 24 and 36 months were 68%, 59%, 39% and 25%, respectively (95% CI at 6, 12, 24 and 36 months were 63–72%, 54–63%, 34–44% and 20–31%, respectively; Fig. 1). Overall, 11 patients had detectable PSA levels at 1 month after RP. Figures 2 and 3 show Kaplan–Meier curves predicting eBCR-free survival according to the dichotomised number of positive LNs, ENE, and diameter of LN metastasis. Patients with one or two positive LNs had better survival rates than those with ≥3 (64% vs 29% at 12 months; P < 0.001; Fig. 2). Similarly, patients without ENE (at 12 months: 65% vs 53%; P < 0.01; Fig. 3a) or in whom the largest LN metastasis had a diameter of <10 mm (at 12 months: 64% vs 52%; P < 0.001; Fig. 3b) had better survival rates than their respective counterparts. The combination of ENE and LN metastasis size stratified patients into statistically significant different risk groups for eBCR-free survival: patients with no ENE and a smaller metastasis had better survival rates than patients with either ENE or a larger nodal burden, who, in turn, had better survival rates than patients with both ENE and a larger nodal burden (at 12 months 72% vs 59% vs 47%; P < 0.001; Fig. 3c). A further stratification, including the number of positive LNs and all the other pathological features discussed above, identified a good prognosis category defined as patients with ≤2 positive LNs, no evidence of ENE and metastases size of <10 mm (at 12 months: 73% vs 55%; P = 0.007; Fig. 3d).

Figure 1.

Kaplan–Meier curve showing overall BCR-free survival.

Figure 2.

Kaplan–Meier curve showing overall BCR-free survival according to the dichotomised number of positive LNs (≤2 vs >2).

Figure 3.

Kaplan–Meier curve showing overall BCR-free survival (a) according to ENE status, (b) according to dichotomised diameter of largest LN metastasis (<10 vs ≥10 mm), (c) according to ENE status and dichotomised diameter of largest LN metastasis (<10 vs ≥10 mm), and (d) according to number of positive LNs (≤2 vs >2), ENE status and dichotomised diameter of largest LN metastasis (<10 vs ≥10 mm). Patients were categorized according to best pathological features (≤2 positive LNs, no ENE and metastases size <10 mm) vs all other features.

Multivariable Analyses

Table 2 shows the result of a multivariable Cox regression analysis predicting eBCR. Three or more positive LNs and a LN metastasis of >10 mm were independent risk factors of eBCR (all P < 0.003), but ENE was not. Other independent predictors included age at RP, preoperative PSA level, a pathological Gleason score of 8–10, seminal vesicle invasion, positive surgical margins and adjuvant radiotherapy (all P < 0.05). There was no interaction between the number of positive LNs, ENE, and metastasis size. In a sub-analysis, among patients with ≤2 positive LNs (410 patients), the size of the largest LN metastasis was a significant predictor of eBCR (P = 0.03), while ENE was not (P = 0.3). Similarly, the size of the LN metastasis retained a role as an independent predictor of eBCR in the group of patients with ≥3 LN metastases (P = 0.02), while ENE did not reach statistical significance (P = 0.17).

Table 2. Multivariable Cox regression analysis predicting BCR
 HR (95% CI)P
Age at surgery0.97 (0.96, 0.99)0.002
Preoperative PSA level1.02 (1.01, 1.02)<0.001
Pathological Gleason score  
≤7 vs 8–101.54 (1.18, 2.00)0.001
Pathological stage  
pT2Ref.
pT3a1.66 (1.00, 2.77)0.051
pT3b2.26 (1.37, 3.72)0.001
Positive surgical margins1.45 (1.11, 1.89)0.006
>2 positive LNs2.80 (1.99, 3.93)<0.001
ENE1.29 (0.99, 1.68)0.062
Positive LN diameter >10 mm1.48 (1.16, 1.89)0.002
Adjuvant radiotherapy1.40 (1.00, 1.95)0.050

Table 3 shows the comparison of the discrimination between different multivariable models predicting eBCR. The discrimination of a baseline model including age at RP, preoperative PSA level, pathological stage, pathological Gleason score, and positive surgical margins was 0.647. Adding ENE and diameter of the largest LN metastasis to the model increased the discrimination to 0.663, while adding the dichotomised number of positive LNs (≤2 vs >2) increased the discrimination to 0.686. The concordance index for the full final model, including ENE, metastasis size and number of positive LNs was 0.697.

Table 3. Harrel's C-index of multivariable models predicting eBCR
ModelHarrel's C-indexDifference with baseline model
  1. *Baseline model: age at surgery, preoperative PSA level, pathological stage, pathological Gleason score, positive surgical margins and adjuvant radiotherapy.
Baseline*0.647
Baseline + ENE + LN diameter0.663+0.016
Baseline + positive LNs (≤2 vs >2)0.686+0.039
Baseline + positive LNs (≤2 vs >2) + ENE + LN diameter0.697+0.05

Discussion

Outcomes of patients with LN metastasis after RP are not by definition poor [1]. Indeed, we found that outcomes vary largely among men with LN metastases. Using the largest series to date including size of LN metastasis and ENE, we show that patients with LN-positive prostate cancer can be stratified according to readily available pathological information into statistically and prognostically different risk groups. Furthermore, we confirmed previous results [4-10] that patients with one or two positive LNs have better survival rates than men with ≥3 LN metastases. In multivariable analyses, the diameter of the largest LN metastasis and the dichotomised number of positive LNs remained independent predictors of eBCR. On the other hand, ENE did not reach statistical significance as an independent predictor. ENE and the diameter of the LN metastasis were able to increase the discrimination of the standard multivariable model for predicting eBCR by 1.6%. However, the discrimination of the baseline model was increased to a greater extent by adding the number of LNs as a predictor (3.9%), suggesting that the dichotomised number of LNs carries more prognostic information and could better stratify patients than ENE and LN metastasis diameter. Indeed, patients with ≥3 positive LNs have 2.7-times the probability of having eBCR than men with one or two positive LNs. However, the size of the largest LN metastasis contributed to the multivariable model to the same extent as pathological Gleason score and surgical margins status (hazard ratio 1.46 vs 1.49 vs 1.45, respectively). When all three variables were included in the model, discrimination increased by 5%, indicating a benefit of combining these prognostic factors rather than considering each factor alone. Indeed, even if only the number of positive LNs and diameter of the metastasis were independent factors of eBCR, it is worth analysing all three of them together, as markers should be judged on their ability to increase predictive accuracy and not only on their P-value [19]. While the discrimination of the final model is not optimal, these features can be used to better identify patients who are at risk of eBCR after RP and thus more likely to benefit from adjuvant therapy. In this fashion, unnecessary adjuvant therapy and its potential side-effects can be avoided for patients who are likely to be cured with RP alone.

While the number of positive LNs has been proven as a reproducible and accurate predictor of long-term outcomes among men with LN metastasis after RP [4-10], the role of ENE and LN metastasis diameter are still controversial. ENE has been shown to accurately stratify patients with positive LNs in various epithelial cancers [13]. While one early study suggested a prognostic role for ENE in prostate cancer [11], others have not [13-16]. Griebling et al. [11] reported that ENE was a strong predictor of survival, second only to pathological Gleason score. However, their population consisted of many advanced cancers, mostly composed of pre-PSA era patients, and only 27% of them underwent RP. Cheng et al. [13] also analysed a pre-PSA population, but all of the patients were treated with RP and extended PLND. Despite the fact that, unlike the present study, all their patients underwent adjuvant ADT, ENE predicted neither development of distant metastases nor cancer-specific death after accounting for the effects of Gleason score, DNA ploidy and LN cancer volume. The only predictive factor of long-term disease-free survival and cancer-specific survival was the total cancer volume in the LNs. In addition, they found that the number of LN metastases did not impact outcomes. Interestingly, they found that ENE did not impact survival even when considering patients with only one LN metastasis. This is consistent with the present findings, where ENE was not associated with eBCR among patients with one or two positive LNs, suggesting that these men already represent a category of patients with a favourable prognosis. Fleischmann et al. [14] also failed to find any value to ENE for prognosticating BCR-free, disease-specific or overall survival. However, similar to the present study, the diameter of the largest LN metastasis was a significant prognosticator of all three endpoints. Also, none of the patients in their study underwent adjuvant treatment after RP. Boormans et al. [16] recently reported that ENE is not useful in predicting cancer-specific survival while diameter of the largest LN metastasis and the Gleason score of the LN tumour are. Their results are concordant with ours, even though their population consisted mostly of high-risk patients who underwent a limited PLND for staging only (i.e. to confirm LN involvement). In a recent series reporting the outcome of LN-positive patients, Carlsson et al. [18] conducted a sub-analysis on patients for whom they had available data on ENE and size of the metastasis. They concluded that the strongest predictor of BCR was the number of positive LNs, while ENE and size of largest positive LN did not contribute significantly to the discrimination of the model. However, such analysis was conducted on a small population of 96 patients.

The size of the largest LN metastasis is a strong predictor of clinical outcomes in patients with prostate cancer with LN metastasis. In the present series, size of the largest LN metastasis, apart from the number of positive LNs, was the only other LN variable to reach independent predictor status in multivariable analyses. Fleischmann et al. [15] have previously shown that having a LN metastasis of >10 mm quadruples a patients' risk of cancer-specific death. Furthermore, they found that patients with micrometastasis, (defined by a diameter of <2 mm) had the most favourable outcome among LN-positive patients. In addition to decreasing the number of patients with positive LNs at time of RP, the widespread use of PSA screening has changed the pathological features of the LN metastasis, such as reducing the tumour burden [4-10]. Despite this evidence, the 1997 TNM revision has not been updated and patients with LN metastasis are grouped together. The previous edition of the 1992 TNM categorised patients with positive LNs according to the diameter of their LN metastases. LN-positive patients were divided into three categories: pN1, a single positive LN with a diameter of ≤20 mm; pN2, a single positive LN of >20 mm but <50 mm, or multiple LNs none of >50mm; pN3 a positive LN with diameter of >50 mm [17]. The simplification of staging introduced in 1997 represents the concept that once prostate cancer has metastasised to the LNs it should be considered as a poor outcome. However, independent of TNM revisions, we feel that a combination of pathological features, namely the number of positive LNs and the size of the greatest tumour-affected LN, could stratify patients with LN-metastatic prostate cancer to provide optimal personalised counselling about adjuvant therapy and follow-up scheduling.

The present study is not devoid of limitations. First, due to the short follow-up (median 16.1 months) we were only able to report eBCR-free survival outcome. Even if this is of less value when compared with more concrete and long-term outcomes, such as development of metastases and cancer-specific mortality, the present data can be used to identify patients with a higher risk of developing eBCR in the short-term. An eBCR of ≤12 months of RP is indeed a harbinger of poor clinical outcomes suggesting systemic metastatic disease at the time of RP [20, 21]. One could imagine a risk stratification of patients with LN metastasis for adjuvant hormonal treatment based on the number of positive LNs, ENE and diameter of the LN metastasis. Another limitation is the multicentric nature of the present cohort. Even if all patients underwent PLND following similar anatomical templates, intra- and inter-surgeon variations have caused significant variation in the LN yield with a relatively low median number of LNs removed (10) in our series [1]. To be considered an accurate staging procedure, PLND must be extended [22, 23], reducing the rate of false-negatives. It is important to remember that the chances of detecting LN metastases increase with the number of dissected LNs [24, 25], and hence, an anatomically extended PLND significantly improves the detection rate of positive LNs [1, 26]. Furthermore, the present study lacked central pathological review. Also, 108 underwent minimally invasive RP, and thus, the retrieval of LNs through trocars might have artificially altered the macrostructure of the LN, therefore affecting measurement of the size and evaluation of ENE. In addition, pathological analysis of ENE could be ameliorated with the aid of immunohistochemical stains, as extension of tumour cells into the peri-nodal fatty tissue could be hard to analyse due to the fatty nature of pelvic LNs [13]. Moreover, we did not register the Gleason score of the metastatic LNs. There is strong evidence that patients in whom the LN Gleason score was higher than the one in the prostatic specimen had higher rates of systemic progression [27], and that the behaviour of the disease is linked to the worst (i.e. highest) Gleason score recorded among the metastatic LNs, more than the prostatic one [17, 28]. All of these weaknesses could only be addressed by a prospective study with centralised pathological review and standardised surgical decision-making and technique.

In conclusion, the diameter of the largest LN metastasis and the number of positive LNs are the two powerful predictors of BCR. ENE, on the contrary, is not an independent risk factor for eBCR. The number of positive LNs added more prognostic information to standard clinicopathological features than ENE or the largest positive LN. However, together they performed better than any factor alone. Taken together, one could envision a shared decision-making process based on these three readily available pathological features. Patients with two or less positive LNs, no ENE and with the largest LN metastatic focus <10 mm constitute a subgroup with a better prognosis, and thus may be spared adjuvant therapy. The present data, if confirmed in a prospective study with centralised pathological review and extended PLND templates, could give a further impetus for a change in the N staging of the TNM classification. Furthermore, ENE and the size of the metastasis should be routinely recorded during pathological examination, especially due to the facility of assessment of such measurements.

Acknowledgments

The Authors would like to acknowledge Dr Karim Touijer for help in the conception of the tested hypothesis.

Conflict of Interest

None declared.

Abbreviations
ADT

androgen-deprivation therapy

eBCR

early biochemical relapse

ENE

extranodal extension

IQR

interquartile range

LN

lymph node

PLND

pelvic LN dissection

RP

radical prostatectomy

Ancillary