This work is part of the doctoral thesis of C.G.S.
Cancer-specific survival after radical cystectomy and standardized extended lymphadenectomy for node-positive bladder cancer: prediction by lymph node positivity and density
Version of Record online: 11 FEB 2009
© 2009 THE AUTHORS. JOURNAL COMPILATION © 2009 BJU INTERNATIONAL
Volume 104, Issue 3, pages 331–335, August 2009
How to Cite
Wiesner, C., Salzer, A., Thomas, C., Gellermann-Schultes, C., Gillitzer, R., Hampel, C. and Thüroff, J. W. (2009), Cancer-specific survival after radical cystectomy and standardized extended lymphadenectomy for node-positive bladder cancer: prediction by lymph node positivity and density. BJU International, 104: 331–335. doi: 10.1111/j.1464-410X.2009.08403.x
- Issue online: 9 JUL 2009
- Version of Record online: 11 FEB 2009
- Accepted for publication 14 November 2008
- bladder cancer;
- radical cystectomy;
- extended lymph node dissection;
- lymph node density
To investigate the associations between different overall or topographically restricted lymph node (LN) variables and cancer-specific survival (CSS) after radical cystectomy (RC) and extended LN dissection (LND) with curative intent in patients with LN-positive bladder cancer.
PATIENTS AND METHODS
Between 2001 and 2006, 152 patients had RC with standardized extended LND for bladder cancer with curative intent. Patients with positive LNs were stratified according to the median of the LN variables (LNs removed, number of positive LNs, LN density). CSS was related to overall and topographically restricted LN variables, e.g. different levels of LND, and relationships were tested by univariate and multivariate analyses. Level 1 LND comprised the regions of the external and internal iliac LNs and of the obturator LNs, level 2 the templates of common iliac and presacral LNs, and level 3 the para-aortic and paracaval LNs up to the inferior mesenteric artery. The mean (range) follow-up was 22 (1–84) months.
LN metastases were diagnosed in 46 of the 152 patients (30%) with extended LND. In these 46 patients, the median number of removed LNs was 33 (level 1, 15.5; level 2, 9.0; level 3, 7.0), the median number of positive LNs was 3 (1.5, 0.5 and 0.0, respectively) and the median LN density was 0.11 (0.10, 0.02 and 0.0, respectively). The CSS was 76% at 1 year and 23% at 3 years. There were significant correlations between the 3-year CSS and the overall LN density (≤0.11 vs >0.11; 34% vs 8%, P = 0.008), and the total number of positive LNs (≤3 vs >3; 33% vs 8%; P = 0.05). Overall LN density (hazard ratio 0.33, 95% confidence interval 0.15–0.72; P = 0.006) was an independent predictor for CSS in multivariate analysis.
Overall LN density is an independent predictor of survival after RC and extended LND with curative intent. Evaluation of topographically restricted LN positivity and density for different regions and levels of LND does not improve the prediction of CSS compared with overall LN positivity and density. A low incidence of level 3 LN positivity questions the clinical relevance of removing para-aortic and paracaval LNs. However, our data need to be confirmed by a prospective randomized trial.
lymph node (dissection)
transurethral resection of bladder tumour
Bladder cancer is the second most common genitourinary malignancy; ≈90% of patients have TCC and up to 40% initially present or develop invasive disease . Radical cystectomy (RC) with bilateral pelvic lymph node (LN) dissection (LND) is the standard procedure for high-grade invasive bladder cancer [1,2]. Tumour extension to the LNs is recognized in up to 35% and is positively correlated with the tumour stage in the RC specimen [3,4]. The LN status serves as the most powerful predictor for cancer-specific survival (CSS) after surgery . However, to date the necessary extent of LND remains unclear . Nevertheless, there is increasing evidence that a minimum number of LNs must be removed at RC [6,7]. During the past few years, the extent of LND, the number of positive LNs and the LN density (defined as the number of positive LNs divided by the number of removed LNs) have been shown to be related to recurrence-free and overall survival [6,8]. However, there is so far neither agreement on the required minimum number of LNs that must be removed nor on the threshold value for LN density as predictors of different outcomes. Furthermore, due to the lack of randomized prospective data it remains unclear whether the prognostic advantage of an extended LND for different subgroups is a therapeutic effect (removal of positive LNs), a staging effect (higher specificity of N-categories) or a ‘Will Rogers’ phenomenon .
The goal of the present study was to determine the prognostic value of LN variables in LN-positive patients who had RC with a standardized extended LND with curative intent for TCC of the bladder. Previous LN mapping studies showed that LN positivity varies for regions and levels of LND that are closer to the bladder than more distant locations . LN density in patients with positive nodes has been shown to be a good predictor of prognosis . However, as LN regions distant from the bladder generally have a lower LN positivity, and as standardization of the extent of LND is lacking, LN density might be ‘diluted’ when a more extended LND results in removal of more distant and consequently more negative LNs. Thus we aimed to further differentiate LN positivity (number of removed/positive LNs, LN density) and its relationship to CSS by analysing these variables for topographic restrictions such as regions and levels of LND.
PATIENTS AND METHODS
Between 2001 and 2006, 152 patients (116 men, 36 women; median age 65.0 years, range 38–84) had a RC with a prospectively standardized extended LND with curative intent for TCC of the bladder. In this selected group of patients treated with RC of curative intent, CT of the abdomen and a chest X-ray were undertaken in all before RC, and had suggested local disease. Patients with LN enlargement on CT that was suspected to indicate tumour spread to LNs, and those with neoadjuvant systemic chemotherapy, were excluded from this analysis. All patients had undergone at least one transurethral resection of bladder tumour (TURBT, range 1–17) before RC. The median (range) interval between the first TURBT and RC was 2 (0–170) months. Adjuvant chemotherapy was given to 27 of the 46 patients (59%) in whom LN metastases were diagnosed. The median (range) follow-up was 22 (1–76) months.
The limits of our standardized LND are; cranially, the inferior mesenteric artery; laterally, the genitofemoral nerve; caudally, the femoral ring; and dorsally, the deep obturator and levator ani muscles, the sacrotuberous ligament and the sacral bone. The LNs were sent to the pathologist separately for 14 different regions according to our sampling template (Fig. 1A). Different LN variables were analysed for three levels of LND: level 1 comprised the regions of the external and internal iliac LNs and of the obturator LNs; level 2 the regions of the common iliac and presacral LNs; and level 3 the para-aortic and paracaval LNs up to the inferior mesenteric artery. LNs were allocated to the described regions by the surgeon immediately after tissue removal on a sterile laminated version of our sampling template.
The histopathology of the RC specimens was classified according to the TNM system . The removed LNs were evaluated histopathologically and prospectively according to a standardized protocol. No fat-clearing solutions were used. The lymphatic tissue was macroscopically cleaned from its surrounding fatty tissue. LNs were separately paraffin-embedded and dissected in 3-mm slices. The total number of removed LNs and the number of positive LNs were recorded separately for each region.
The prevalence of LN metastases was correlated with gender, age, histopathological tumour stage and grade of the RC specimen, presence of tumour upstaging, number of TURBTs, interval between first TURBT and RC and number of removed LNs. Patients with LN metastases were stratified according to thresholds that were defined by the median values for each LN variable in the series, e.g. number of removed LNs, number of positive LNs and LN density. CSS was analysed and related to overall LN variables and to topographically restricted LN variables of different levels of LND.
The prevalence of LN metastases was analysed using Fisher’s exact test. CSS was estimated using the Kaplan-Meier method, with differences in CSS assessed according to the studied LN variables by Cox regression analysis. A multivariate analysis by the Cox regression hazard model was used to assess independent prognostic factors. In all test, the statistical significance was set at P ≤ 0.05.
The patients’ demographics, histopathology of the RC specimens, clinical variables and their correlation with the prevalence of LN metastases are summarized in Table 1. Histopathological tumour stage (organ confined vs extravesical) and tumour upstaging in the RC specimen (yes vs no, both P < 0.001) showed a strong correlation with the prevalence of LN metastases. Tumour grade, number of TURBTs, interval between first resection and RC and number of removed LNs were not significantly correlated with the prevalence of LN positivity. In all, 46 of the 152 patients (30%) had LN metastases at RC. The 3-year CSS was significantly higher in patients without LN metastases than in those with positive LNs (91% vs 23%, P < 0.001, Fig. 1B).
|No. of patients||106||46|
|Tumour stage in RC|
|Tumour grade in RC specimen||0.6|
|Number of removed LNs|
|Number of TURBTs|
|Interval TURBT to RC|
|Upstaging at RC|
Patients who presented with LN metastases had a median (range) of 33.0 (15–77) removed LNs, with LNs per level of: level 1, 15.5 (2–37); level 2, 9.0 (1–24); level 1 + 2, 24.5 (10–48); level 3, 7.0 (1–29); and a median of 3.0 (1–28) positive LNs, with LNs per level of: level 1, 1.5 (0–10); level 2, 0.5 (0–12); level 1 + 2, 2.5 (0–22); level 3, 0.0 (0–13). The median LN density was 0.11 (0.01–0.73), with a LN density per level of 0.10, 0.02, 0.13 and 0.0, respectively. Of the 46 patients with LN metastases, 10 (22%) had positive LNs in levels 2 and 3 but negative LNs in level 1. The median number of removed LNs was 43.5 (18–77) and the median number of LN+ was 1 (1–28). Solitary LN metastases were diagnosed in six of the 10 patients. In three cases they were exclusively in level 2 and in another three they were diagnosed in level 3 only.
The CSS of patients with LN metastases was 76% and 23% at 1 and 3 years, respectively. Among the evaluated LN variables significant correlations were detected between the 3-year CSS and overall LN density, and overall number of positive LNs, by univariate analysis. For an overall LN density of ≤11% (median LN density 0.11), the 3-year CSS was 34%, vs 8% for a LN density of >11% (P = 0.008; Fig. 1C). The 3-year CSS was 33% for ≤3 positive LNs (median of overall positive LNs, three) vs to 8% for >3 positive LNs (P = 0.05; Fig. 1D). Also, the 3-year CSS was significantly longer in patients who had a solitary LN metastasis than in those with at two or more positive LNs (3-year CSS 47% vs 8%; P = 0.01, Fig. 1E). Overall, and to separate level restricted number of removed LNs, to separate level restricted number of positive LNs and to separate level restricted LN density were not significantly correlated to the 3-year CSS. Patients who received adjuvant chemotherapy had a 3-year CSS rate of 46%, vs 8% when no adjuvant chemotherapy was administered (P = 0.03). Multivariate analysis showed that overall LN density was the only independent predictor for CSS (hazard ratio 0.33, 95% CI 0.15–0.72, P = 0.006). The results of univariate and multivariate analyses are shown in Table 2.
|LNs removed (≤33 vs >33)||0.9||ns|
|LN+ (≤3 vs >3)||0.05||ns|
|LN + (≤1 vs >1)||0.01||ns|
|LN density (≤0.11 vs > 0.11)||0.003||0.006|
|Hazard ratio (95% CI)||0.33 (0.15–0.72)|
|Level 3 restricted|
|LNs removed (≤7 vs >7)||0.6||ns|
|LN+ (0 vs >0)||0.08||ns|
|LN density (0 vs >0)||0.08||ns|
|Level 2 restricted|
|LNs removed (≤9 vs >9)||0.6||ns|
|LN+ (≤0.5 vs >0.5)||0.4||ns|
|LN density (≤0.02 vs >0.02)||0.4||ns|
|Level 1 restricted|
|LNs removed (≤15.5 vs >15.5)||0.6||ns|
|LN+ (≤1.5 vs >1.5)||0.08||ns|
|LN density (≤0.10 vs >0.10)||0.3||ns|
|Level 1 + 2 restricted|
|LNs removed (≤24.5 vs >24.5)||0.8||ns|
|LN+ (≤2.5 vs >2.5)||0.02||ns|
|LN density (≤0.13 vs >0.13)||0.04||ns|
Of patients with bladder cancer, 15–35% have LN metastases at RC; the diagnosis of positive LNs decreases CSS significantly [1,11–13]. Adequate LND is not only important for an accurate assessment of tumour burden, but might also increase survival. Leissner et al. found that removing at least 20 LNs by pelvic LND identified ≈80% of their patients presenting with LN metastases, in a retrospective follow-up study. Herr  reported a series of 162 patients with LN-positive bladder cancer where the total number of removed LNs was correlated with the prevalence of positive LNs. However, currently the numbers of removed LNs that are required as a criterion for a complete pelvic or extended LND is still under dispute. Poulsen et al. reported that extending pelvic LND from the level of the iliac bifurcation to the aortic bifurcation increased the number of removed LNs from 15 to 25. In a prospective randomized multicentre study, a mean of 43 LNs were removed, when the LND was cranially extended to the inferior mesenteric artery, and LN metastases were diagnosed in 28%. Stein et al. reported a median of 30 retrieved LNs, a median of two positive LNs and LN metastases in 23% when extended LND was done up to a level somewhere above the aortic bifurcation. There was a comparable number of retrieved and positive LNs (33 and three) in the present series, suggesting a similar extent of LND. The extent of LND is known to influence the accuracy of LN staging. Dhar et al. found 26% vs 13% LN metastases for extended vs limited LND; in that study, 20% of LN metastases were located outside the boundaries of a limited LND. In the present series, 10 of 46 patients (22%) had LN metastases that were exclusively located above the common iliac bifurcation level, including three who had a solitary positive LN that was even located above the aortic bifurcation level. The present 30% prevalence of positive LNs is comparable to the 26% with extended LND reported by Dhar et al.
In the present series, the CSS rate for LN-positive patients was 76% at 1 year, decreasing to 23% at 3 years. Several studies suggested that LN variables were useful clinical predictors of CSS in LN+ disease [6,8,11,14,17]. The number of removed and of positive nodes, and more recently assessment of LN density, were reported to be strong indicators of outcome. The concept of LN density was established by Stein et al. to relate tumour burden (LN+) to the individual extent and quality of LND (LNs removed). A multivariate regression model for categorized variables showed that the total number of removed LNs (≤15 vs >15), number of positive LNs (≤8 vs >8) and LN density (≤20% vs >20%) were independent prognostic predictors for survival in 244 LN+ patients. When LN density was up to 20%, the overall 5-year survival rate was 43%, vs 8% when the LN density was >20%. Considerable limitations of the study are the arbitrarily defined thresholds for the investigated LN variables, and the lack of recording of the anatomical location of positive LNs. Also, with a large range of removed nodes (1–96) in that series, the number of LN metastases might potentially be underestimated in patients with a few removed nodes only. Consequently, by removing many nodes, the advantage in survival of the subgroups could reflect a more precise staging of the lymphatic tumour burden (staging effect) rather than a therapeutic effect of removing positive LNs. The improvement in survival of the subgroups (e.g. lower vs higher total number of removed LNs) is virtual when there is no improvement of the whole population by the studied intervention, which is referred to as the Will Rogers phenomenon . Fleischmann et al. recognized that the number of positive nodes (≤5 vs >5) and LN density (≤20% vs >20%) were associated with CSS in univariate analysis but failed to be independent prognostic predictors in multivariate analysis . The authors noted critically that the LN density is not only a function of tumour burden divided through the extent or quality of LND, but also depends on the anatomical variation of the total number of LNs, which theoretically should not influence the outcome. Recently, Herr  questioned that LN density is ready to be introduced into clinical practice before systematic nodal ratio analyses from multi-institutional patient datasets are available. Another concern about the reported 20% threshold value of LN density is the function of para-aortic/paracaval cranial extension of LND. When the results of different mapping studies indicate that LN positivity decreases with distance from the bladder, and ‘skip’ lesions at para-aortic/paracaval regions are extremely rare, including these regions in the LN density evaluation implies a ‘dilution effect’ for pelvic LND [4,18]. We stratified our patients according to thresholds that reflected the median values of each LN variable. We also recorded the exact location of removed LNs by allocating them to defined regions immediately after removal. By contrast with the study of Fleischmann et al., we found the overall LN density (≤11% vs >11%) to be an independent predictor of CSS by multivariate analysis in the present series. By analysing the LN variables at an overall scale and at different topographically restricted subscales (levels of LND), we tried to address the question of whether their prognostic power could be increased, and if a critical level for the minimum extent of LND can be identified. However, LN variables at the separately restricted levels of LND failed to show any association with CSS, assuming a lack of clinical relevance for the evaluation of LN variables at these single subscale levels even when using a standardized extended LND with histopathological evaluation for each topographical region. However, the retrospective character of our series and the relatively few patients examined must be considered in the interpretation of these preliminary data. Despite a high standardization of the LND in this single-centre study, there was still a wide range (8–80) retrieved LNs, which highlights the natural variability of available LNs in the individual, and strongly supports Herr’s call for a multi-institutional nodal-ratio analysis to achieve a validated threshold value for the LN density. Positive LNs were exclusively located above the common iliac bifurcation level in 22% of the present patients with LN metastases, which supports the usefulness of extended LND to ensure an accurate assessment of lymphatic tumour burden. However, with a low incidence of level 3 LN involvement in our series, the clinical relevance of removing para-aortic and paracaval LNs up to the inferior mesenteric artery remains unclear. Prospective randomized trials comparing pelvic (aortic bifurcation) and extended (para-aortal/paracaval) LND are needed to validate the present preliminary data.
In conclusion, determining LN variables at an overall scale is superior to their evaluation at their topographically restricted subscales. Overall LN density is independently correlated with CSS in patients with LN metastases. With a low incidence of level 3 LN involvement, the clinical relevance of removing paraaortic and paracaval LNs remains unclear, although ‘skip’ lesions might occur. Our data need to be validated in prospective randomized trials.
CONFLICT OF INTEREST
- 10TNM Classification of Malignant Tumors, 5th edn. New York: Wiley-Liss, 1997,