Fax: (206) 543-3272
Article first published online: 28 MAR 2008
Copyright © 2008 American Cancer Society
Volume 112, Issue 11, pages 2401–2408, 1 June 2008
How to Cite
Wright, J. L., Lin, D. W. and Porter, M. P. (2008), The association between extent of lymphadenectomy and survival among patients with lymph node metastases undergoing radical cystectomy. Cancer, 112: 2401–2408. doi: 10.1002/cncr.23474
See related editorial on pages 2331-3, this issue.
Presented at the 82nd Annual Meeting of the Western Section of the American Urologic Association, Maui, Hawaii, October 22–27, 2006 and at the Annual Meeting of the American Urologic Association, Anaheim, California, May 19–24, 2007.
- Issue published online: 19 MAY 2008
- Article first published online: 28 MAR 2008
- Manuscript Accepted: 30 OCT 2007
- Manuscript Revised: 1 OCT 2007
- Manuscript Received: 20 JUN 2007
- lymph node density;
- overall survival;
- surgical extent;
- transitional cell carcinoma
Long-term survival in patients with lymph node-positive bladder cancer who undergo cystectomy suggests a therapeutic role for lymphadenectomy. The objective of this study was to describe the association between extent of lymphadenectomy and survival in lymph node-positive patients who underwent radical cystectomy.
The cohort consisted of patients from the Surveillance, Epidemiology, and End Results registry with transitional cell carcinoma who underwent cystectomy with lymphadenectomy and had at least 1 positive lymph node and no distant metastases. The Kaplan-Meier method and multivariate Cox proportional-hazards regression analyses were used to estimate differences in survival among different lymphadenectomy variables.
In total, 1260 patients had at least 1 positive lymph node. A median of 9 lymph nodes were removed (range, 1–48 lymph nodes) with a median of 2 positive lymph nodes (range, 1–18 positive lymph nodes), and the median lymph node density was 22%. In multivariate analysis controlling for patient demographics, tumor classification, and year of diagnosis, the number of positive and total lymph nodes removed remained independent predictors of survival. There was an inverse association between the number of lymph nodes removed and the risk of death for all quartiles. Removal of > 10 lymph nodes was associated with increased overall survival (hazard ratio, 0.52; 95% confidence interval, 0.43–0.64). In addition, with a lymph node density from 0.1% to 12.5% as the referent group, each higher quartile experienced worse survival.
An increased number of lymph nodes removed at the time of cystectomy was associated with improved survival in patients with lymph node-positive bladder cancer. Improved survival was observed at a lower lymph node density threshold than previously reported. The current findings support performing a more extensive lymphadenectomy at the time of cystectomy. Cancer 2008. © 2008 American Cancer Society.
Bladder cancer is the 1 of the most common cancer diagnoses in the United States with an estimated 61,420 new cases in 2006.1 Of these patients, from 20% to 40% will have or will develop muscle-invasive disease for which radical cystectomy is indicated. Among the patients who undergo cystectomy, approximately 25% will have lymph node involvement.2 Long-term survival in patients who have lymph node-positive bladder cancer suggests that there may be a therapeutic role for lymphadenectomy.3–7 However, debate exists with regard to the boundaries of lymphadenectomy required to achieve this survival benefit, with proponents of either a standard dissection8, 9 (cephalad boundary at the common iliac bifurcation) or an extended dissectoin3, 5, 6, 10 (cephalad boundary at the aortic bifurcation or takeoff of the inferior mesenteric artery).
Efforts to stratify patients with lymph node-positive bladder cancer into different prognostic groups have reported different thresholds for the number of positive lymph nodes,11, 12 the total number of lymph nodes removed,6, 13, 14 and the lymph node density11, 15 (calculated as the number of positive lymph nodes divided by the total number of lymph nodes removed). These previous single-institution studies, however, were limited by their small sizes. In the current study, we used population-based data to examine the relation between the extent of lymph node dissection, as defined by the number of lymph nodes removed, and survival in patients with lymph node metastases who underwent cystectomy for bladder cancer.
MATERIALS AND METHODS
The Surveillance, Epidemiology, and End Results (SEER) Program database was used to identify the cohort of patients for this study.16 SEER collects cancer incidence and survival data from 17 population-based cancer registries that account for approximately 26% of the United States population. Data from 1988 to 2003 from 17 SEER registries were used (San Francisco-Oakland Standard Metropolitan Statistical Area, Connecticut, Metropolitan Detroit, Hawaii, Iowa, New Mexico, Seattle [Puget Sound], rural Georgia, Utah, Metropolitan Atlanta, Alaska, San Jose-Monterey, Los Angeles, Kentucky, Louisiana, New Jersey, and Greater California). Data from 1973 to 1987 were not included, because detailed stage data were not available from those years. Cases were available only for the years 1992 through 2002 from Alaska, San Jose-Monterey, Los Angeles, and rural Georgia and for the years 2000 through 2002 from Kentucky, Louisiana, New Jersey, and Greater California.
Potential patients were identified initially by using International Classification of Diseases for Oncology, third edition (ICD-O-3) site codes for the urinary bladder (codes C670-C679) and ICD-O-3 histology codes for transitional cell carcinoma. Then, patients who underwent cystectomy with lymphadenectomy were selected. Patients who had at least 1 lymph node removed were included in the analysis. From this subset, the final cohort was limited to patients who had histologically confirmed positive lymph nodes. Patients with visceral or skeletal metastasis were excluded.
Data Collection and Coding
Demographic data included patient age, race, sex, and tumor registry. Because age was a potentially strong confounder, it was coded and entered into the regression model in several different forms: categorized into 5-year age groups, categorized into 10-year age groups, as a continuous spline, and as a quadratic spline. The final model contained age categorized into 10-year age groups. Race was categorized as white, black, or other based on SEER coding. The year of treatment and the tumor registry also were ascertained. Pathologic data included the total number of lymph nodes removed, the total number of positive lymph nodes, the primary tumor (T) classification, tumor grade, and whether or not patients received radiation therapy. Chemotherapy, previous intravesical therapy, and comorbidity data are not available in SEER. The number of negative lymph nodes was calculated by subtracting the number of positive lymph nodes from the total number of lymph nodes. Lymph node density was calculated by dividing the number of positive lymph nodes by the total number of lymph nodes. Each of the lymph node variables (total, positive, negative, and density) was converted to a categorical variable based on quartiles. All covariates entered in the regression model were entered as naturally coded categorical variables. Given the low prevalence of lymph node disease in patients with Ta (0.9%) and T1 (4.4%) bladder cancer, these patients were combined in the analysis with T2 cancers comprising a group with “organ-confined tumors.” Survival was calculated starting at the date of diagnosis to the date of death. If death was not observed, then patients were censored at the date of last follow-up.
Demographic and pathologic data are reported for the cohort. Kaplan-Meier survival curves were generated to compare the unadjusted survival experience by the different lymph node variables (positive lymph nodes, total lymph nodes, and lymph node density). Multivariate Cox regression was performed to evaluate differences in mortality risk based on the lymph node categories. Because lymph node density was derived from the total lymph node and positive lymph node variables, 2 separate regression models were used (because of collinearity). The first model included positive and total lymph nodes, and the second model included lymph node density. Age, stage, radiotherapy, race, sex, and year of treatment were included in both of the final multivariate models as potential confounders based on the a priori analysis plan. Adjustment for tumor registry did not change the hazard ratios (HRs) for the primary predictors of interest and was not retained in the final model. Interaction between age and the lymph node variables was assessed by using nested models and the likelihood-ratio test. A subanalysis that excluded patients with only positive lymph nodes (ie, a lymph node density of 100%) also was performed.
The proportional hazards assumption for the Cox regression was evaluated with Schoenfeld residuals. Individual variables and the combined model were tested and met the proportional hazard assumption (P = .38, with P < .05 indicating evidence of nonproportionality). Collinearity between final model variables was assessed with bivariate correlation coefficients. No variables demonstrated significant collinearity (largest correlation coefficient [R] = 0.29 for positive lymph nodes and total lymph nodes). HRs are presented along with their 95% confidence intervals (95% CIs). To fully characterize the relation between extent of lymphadenectomy and the risk of death, we plotted the log HRs and corresponding 95% CIs for the uncategorized form of total lymph nodes removed and lymph node density from the multivariate models. All statistical analyses were conducted using Stata software, version 8 (Stata, Inc., College Station, Tex).
In total, 5201 patients had lymphadenectomy data available and 1260 patients with nonmetastatic transitional cell carcinoma of the bladder (24.3%) underwent cystectomy between 1998 and 2002 and had positive lymph nodes identified. Of the patients with lymph node data available, 2.4% (n = 129) had no T classification data and were excluded, and 95.8% had poorly or undifferentiated grade.
Table 1 lists the patient demographics. Greater than 70% of patients were white men aged >60 years. More patients were identified during 1998 through 2002 because of the inclusion of additional registries. Table 1 also lists the pathologic characteristics of the tumors. Extravesical primary tumor stage was common and was present in >75% of patients. The median number of total lymph nodes removed was 9 (range, 1–75 total lymph nodes removed), the median number of positive lymph nodes was 2 (range, 1–38 positive lymph nodes), the median number of negative lymph nodes was 6 (range, 0–70 negative lymph nodes), and the median lymph node density was 25%.
|Characteristic||No. of patients (%)|
|Year of diagnosis|
|Total no. of lymph nodes|
|No. of positive lymph nodes|
|No. of negative lymph nodes|
|Lymph node density, %|
The median survival for the entire group was 18 months. Figure 1 illustrates the Kaplan-Meier survival curve according to the number of positive lymph nodes. Patients who had only 1 positive lymph node had the best survival (P = .0027). Figure 2 illustrates survival according to the quartile of total lymph nodes removed (P = .0001). The median survival by total lymph node category was 13 months (1- 5 total lymph nodes removed), 17 months (6–9 total lymph nodes removed), 23 months (10- 16 total lymph nodes removed), and 23 months (>16 total lymph nodes removed). The Kaplan-Meier curves stratified by lymph node density quartiles are shown in Figure 3 (P < .0001). The median survival by lymph node density was 32 months (0.1–12.5%), 21 months (12.6–25%), 17 months (25.1–50%), and 11 months (>50%).
The univariate and multivariate HRs from the Cox regression analyses are shown in Table 2. Compared with the lowest quartile of total lymph nodes removed (1–5 total nodes removed), more lymph nodes removed were associated with a lower risk of death (P < .001). In addition, compared with removing 6 to 9 lymph nodes, the removal of 10 to 16 lymph nodes (HR, 0.73; 95% CI, 0.60–0.90) and the removal of >16 lymph nodes (HR, 0.76; 95% CI, 0.62–0.92) were associated with improved survival. There was no difference in the risk of death between the highest 2 quartiles of total lymph nodes removed (HR, 1.03; 95% CI, 0.84–1.25). For lymph node density, compared with the lowest quartile (0%–12.5%), each higher level of lymph node density was associated with a higher risk of death, including a 30% higher risk for those with a density from 12.6% to 25%. The risk of death in the >50% quartile was worse compared with the other quartiles, including the 12.6% to 25% lymph node density group (HR, 1.84; 95% CI, 1.52–2.22) and the 25.1% to 50% lymph node density group (HR, 1.57; 95% CI, 1.31–1.87). There was a trend toward worse survival in the 25.1% to 50% lymph node density quartile compared with the 12.6% to 25% lymph node density quartile (HR, 1.17; 95% CI, 0.98–1.41). Subanalysis that excluded those patients with a lymph node density of 100% (n = 162 patients) produced similar results for all categories of total lymph nodes removed and lymph node density (data not shown). There was no interaction between age and the number of positive lymph nodes (P = .69), the number of total lymph nodes (P = .21), or lymph node density (P = .98).
|Variable||Overall survival||Disease-specific survival|
|Unadjusted HR (95% CI)||p†||Adjusted HR (95% CI)||p†||Adjusted HR (95% CI)||p†|
|No. of positive lymph nodes|
|2||1.18 (0.99–1.40)||1.12 (0.93–1.34)||1.17 (0.95–1.43)|
|3||1.24 (1.01–1.51)||1.30 (1.06–1.60)||1.33 (1.05–1.68)|
|>3||1.34 (1.14–1.58)||1.59 (1.33–1.90)||1.62 (1.33–1.99)|
|Total no. of lymph nodes|
|6–9||0.68 (0.57–0.81)||0.71 (0.59–0.86)||0.76 (0.62–0.94)|
|10–16||0.54 (0.45–0.65)||0.52 (0.43–0.64)||0.53 (0.42–0.67)|
|>16||0.54 (0.45–0.64)||0.54 (0.45–0.65)||0.55 (0.44–0.69)|
|Lymph node density, %‡|
|12.6–25||1.40 (1.16–1.70)||1.31 (1.08–1.60)||1.24 (0.99–1.56)|
|25.1–50||1.70 (1.41–2.05)||1.54 (1.27–1.85)||1.62 (1.30–2.00)|
|>50||2.80 (2.32–3.37)||2.40 (1.98–2.92)||2.47 (1.97–3.08)|
|T3||1.61 (1.34–1.93)||1.57 (1.30–1.88)||1.70 (1.37–2.11)|
|T4||2.06 (1.72–2.48)||1.87 (1.55–2.26)||2.11 (1.70–2.63)|
|Received radiation||1.43 (1.16–1.78)||.001||1.31 (1.05–1.63)||.017||1.42 (1.11–1.81)||.005|
|Women||1.11 (0.96–1.27)||.147||1.06 (0.91–1.22)||.455||1.15 (0.72–1.52)||.809|
|51–59||0.97 (0.70–1.35)||0.87 (0.62–1.21)||1.04 (0.72–1.52)|
|60–69||1.29 (0.95–1.76)||1.09 (0.80–1.49)||1.05 (0.73–1.50)|
|70–79||1.67 (1.23–2.26)||1.34 (0.98–1.83)||1.23 (0.86–1.77)|
|≥80||2.77 (1.97–3.89)||2.27 (1.55–3.23)||2.30 (1.54–3.44)|
|Black||0.98 (0.75–1.28)||.873||1.01 (0.77–1.33)||.939||0.88 (0.64–1.23)||.459|
|Other||0.76 (0.56–1.03)||.077||0.73(0.54–0.99)||.044||0.67 (0.47–0.97)||.032|
Figure 4 shows the results of the log HR plots for total number of lymph nodes removed and lymph node density. The risk of death appears to be decreased until approximately 10 to 14 total lymph nodes are removed (Fig. 4a). The lymph node density plot suggests an increased risk of death across all values of increasing lymph node density (Fig. 4b).
This population-based study of patients with lymph node metastases who underwent cystectomy for bladder cancer has provided several important findings. First, the results are consistent primarily with single-institution case series that reported the association between improved survival and removal of more lymph nodes in patients with lymph node metastases who undergo cystectomy. Second, we have demonstrated that there is an incremental benefit to more total lymph nodes removed up to approximately 10 to 14 lymph nodes. Finally, the current study suggests that the cutoff point for lymph node density that is associated with improved survival may be lower than previously suggested.
It has been demonstrated in several studies that the removal of more lymph nodes in all patients who undergo cystectomy is associated with improved survival, and several groups have suggested different thresholds.5, 6, 11, 13, 14, 17 A collaborative effort from several high-volume bladder cancer centers proposed radical cystectomy guidelines, recommending that adequate lymph node dissection is defined by the removal of at ≥10 to 14 lymph nodes.18 In patients who do not have lymph node metastases identified, the role of adequate lymph node dissection and its association with improved survival may be a result of more accurate staging. However, the removal of more lymph nodes in patients with lymph node metastases is less clear in terms of both the biologic mechanism and the definition of adequate dissection.
Three large, single-institution studies have reported on their experience with lymph node-positive patients.6, 11, 15 Leissner et al. evaluated 79 patients with <5 positive lymph nodes, and the unadjusted Kaplan-Meier analysis demonstrated improved survival when >16 lymph nodes were removed, although a multivariate analysis was not reported.6 Herr reviewed a lymph node-positive cystectomy series of 162 patients and observed that the removal of >13 total lymph nodes was not a significant predictor of survival on multivariate analysis (P = .56).15 Finally, Stein et al. described the largest lymph node-positive series with 244 patients and observed no recurrence-free survival advantage with the removal of ≥15 lymph nodes (P = .21).11 In the current population-based study of 1260 lymph node-positive patients who underwent cystectomy, we observed that the removal of more lymph nodes was associated with improved survival. Furthermore, our findings are consistent with the recommendations from the Bladder Cancer Collaborative Group that from 10 to 14 lymph nodes should be removed at the time of cystectomy.18 However, the recommendations of the Collaborative Group largely are consensus-based and reflect the limited data available on lymphadenectomy and bladder cancer.
The concept of lymph node density was introduced by Stein et al. in 2003 to account for both the tumor burden and the extent of lymphadenectomy in a single measure.11 Both Stein et al.11 and Herr15 proposed a 20% cutoff for lymph node density and observed that it was an independent predictor of survival in their series. A more recent study suggested a lymph node density cutoff of 25%.19 Our findings support the role of lymph node density in stratifying lymph node-positive patients after cystectomy. However, our data suggest that the previously proposed threshold may be too high, because a 12.5% cutoff point for lymph node density strongly predicted survival, even compared with the 12.6% to 25% cutoff point. In addition, from the graph of the log HRs (Fig. 4B), there is not a clear cutoff point for lymph node density after which the risk of mortality plateaus.
Another recent article used the SEER database to examine bladder cancer patients and provided some data on lymphadenectomy and survival in both lymph node-negative and lymph node-positive patients.20 It is important to highlight the differences between that study and ours. First, the primary focus of Konety and Joslyn was on the factors that determined whether patients underwent cystectomy or received radiation therapy. Their subset analysis on patients with lymph node metastasis also included patients with visceral metastasis, whereas patients with distant metastasis were excluded in our analysis. Second, Konety and Joslyn included patients who did not undergo lymphadenectomy in their analysis, and they used those patients as the referent group in analyzing the role of lymphadenectomy. We limited our investigation only to those patients who underwent lymphadenectomy with cystectomy. Furthermore, their analysis was restricted to 11 of the SEER registries over 7 years, and they identified 554 patients with positive lymph nodes. Our analysis included 17 SEER registries over 12 years, and we identified 1260 lymph node-positive patients, increasing our analytic power.
The current study not only provides clinically useful information regarding number of lymph nodes removed at lymphadenectomy and risk of death in this subset of patients; it also provides a possible explanation for this observed association. In the multivariate analysis, both the number of positive lymph nodes and the number of total lymph nodes were included in the multivariate model. It can be seen from Table 2 that more positive lymph nodes removed were associated with a higher risk of death. This finding is not surprising, because larger numbers of involved lymph nodes indicate higher tumor burden and more advanced metastatic disease. Because the number of total lymph nodes is simply the combined sum of negative and positive lymph nodes, and because the model is adjusted for positive lymph nodes, the HRs for total lymph nodes provide information solely on the effect of removing more negative lymph nodes. In fact, when we generated a model that included negative lymph nodes instead of total lymph nodes, the HRs for negative lymph nodes were almost identical to the model with total lymph nodes. We chose to present the model with total lymph nodes, because this is a more clinically useful indicator of extent of dissection and is in line with previous literature.
In patients with negative lymph nodes, the beneficial effect of extended dissection often has been linked to more accurate staging. Removing more lymph nodes probably is associated with more accurate final pathologic lymph node staging, and it is well established that patients without lymph node metastases have a much better survival than patients with lymph node involvement. However, in patients who have metastatic involvement of their lymph nodes, this may not be the entire explanation. An alternate explanation is that removing more negative lymph nodes results in clearance of micrometastatic foci of cancer that are not identified on pathologic review. With this alternate explanation, a portion of the observed improved survival with extended dissections in patients without lymph node metastases also may be because of the clearance of micrometastases.
Limitations of this study were related primarily to database analysis and observational cohort studies in which the potential for random error may bias the results.21 Miscoding, incomplete data, or missing data could not be accounted for in the analysis. No data were available in SEER on the use of chemotherapy or comorbid conditions. We also could not investigate whether extracapsular extension of lymph node metastasis was present, and extracapsular extension has been proposed as an important prognostic factor in survival for this population.22 We were unable to characterize the anatomic borders of lymphadenectomy, and we could not address whether lymph nodes were submitted en bloc or in separate packets, which is known to affect the number of lymph nodes.23, 24 However, any misclassification of the number of lymph nodes removed, whether from pathologist variability in counting or from inaccurate data entry, is likely to be nonselective, the result of which would be to bias our results toward the null and attenuate the estimates of risk. Finally, it is possible that some of these patients did not undergo lymphadenectomy with curative intent, ie, they may have undergone extremely limited lymph node dissections or directed lymph node excisional biopsies to provide lymph node staging information during radical cystectomy. Furthermore, if patients with significant comorbidities were more likely than healthier patients to undergo this type of limited diagnostic lymphadenectomy, then part or all of the observed survival advantage of more extensive lymphadenectomy may be the result of selection bias through the effect of comorbid disease on survival.25 However, in the subanalysis that excluded patients (n = 162) who only had positive lymph nodes removed (ie, 100% lymph node density), the beneficial effects of removing more lymph nodes were not different from the analysis of the entire cohort.
Despite these limitations, the findings of this study are consistent with primarily single-institution series of patients with lymph node-positive bladder cancer. The current study adds to the existing evidence by using a large population-based cohort to support the role of a more extended lymphadenectomy in patients with lymph node-positive bladder cancer who undergo cystectomy.
- 1American Cancer Society, Inc. Bladder Cancer Statistics [revised August 8, 2006]. Atlanta, Ga: American Cancer Society, Inc.; 2006.
- 16Surveillance, Epidemiology, and End Results (SEER) Program. SEER Program Public-Use Data, 1973–2003 (released April 2006; based on the November 2005 submission.). Bethesda, Md: National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch; 2005.