Virtually all staging schemes aimed at predicting the prognosis of surgically treated patients diagnosed with metastatic renal cell carcinoma (MRCC) omit the use of lymph node stage. In the current study, the authors tested the prognostic significance of lymph node stage in patients with MRCC within a population-based cohort of patients treated with cytoreductive nephrectomy to assess whether the inclusion of lymph node stage could improve the accuracy of cancer-specific mortality predictions.
Within the Surveillance, Epidemiology, and End Results database, the authors identified 1153 patients who were treated with cytoreductive nephrectomy for MRCC, with (negative lymph nodes [N0] vs positive lymph nodes [N1-2]) or without (unknown lymph node stage [Nx]) lymphadenectomy. Of 797 patients treated with lymphadenectomy, 42.9% were found to have lymph node metastases. Kaplan-Meier plots and univariate and multivariate Cox regression analyses tested the statistical significance and the independent predictor status of lymph node stage, Fuhrman grade, tumor size, year of surgery, race, sex, and age in patients who underwent lymphadenectomy at the time of cytoreductive nephrectomy.
At 3 years after cytoreductive nephrectomy, the cancer-specific mortality-free rates of N1-2 versus N0 versus Nx patients were 14.4% versus 34.7% versus 34.0%, respectively. Lymph node stage represented the most informative variable and achieved independent predictor status in all multivariate models (P < .001). Consideration of lymph node stage added 3.2% accuracy to other predictors of cancer-specific mortality.
The TNM staging system as well as 4 scoring algorithms provide prognosis for patients with metastatic renal cell carcinoma (MRCC). The TNM system1-3 assigns the stage and indicates the prognosis according to the characteristics of the primary tumor for patients without lymph node or distant metastases. Conversely, the N category stratifies the prognosis in patients who have established lymph node metastases but not distant disease. Finally, the M category defines the prognosis for patients with distant metastases, regardless of the T and N classifications. Consequently, the TNM system does not account for the presence of lymph node metastases in patients with MRCC. Among the existing prognostic algorithms for MRCC patients, the most contemporary set of the Motzer criteria are comprised of 3 variables.4 Among these, the presence of low Karnofsky performance status (<80) and/or elevated corrected serum calcium (>10 mg/dL) and/or low hemoglobin (<13 g/dL in males and <11.5 g/dL in females) are predictive of poor prognosis. Again, the presence of lymph node metastases is not considered. The Leibovich scoring system also does not account for the lymph node stage, despite relying on 9 variables,5 which consist of the presence of constitutional symptoms at the time of nephrectomy, metastases to the bone, metastases to the liver, metastases in multiple simultaneous sites, metastases at the time of nephrectomy or within 2 years of nephrectomy, complete resection of all metastatic sites, tumor thrombus levels I to IV, tumor nuclear grade 4, and the presence of histologic tumor necrosis. In the subset of patients with MRCC, the University of California at Los Angeles Integrated Staging System accounts for Fuhrman grade and Eastern Cooperative Oncology Group performance status to stratify prognosis.6 However, similar to all other schemes, it does not rely on lymph node stage. To the best of our knowledge, only the prognostic nomogram of Karakiewicz et al. accounts for the presence of lymph node metastases in MRCC as well as other RCC stages.7 However, only a small population of patients within this nomogram series had MRCC. Therefore, the importance of lymph node stage in this tool may very well stem from patients with nonmetastatic RCC.
On the basis of the limited significance assigned to the effect of lymph node stage in virtually all prognostic models that apply to MRCC patients, we decided to test and quantify the prognostic significance of this variable on a large population-based cohort of MRCC patients who were treated with cytoreductive nephrectomy. Our hypothesis was that the inclusion of lymph node stage may better predict the probability of cancer-specific mortality in this patient group.
MATERIALS AND METHODS
Patients diagnosed with RCC between 1988 and 2004 were identified within 9 Surveillance, Epidemiology, and End Results (SEER) cancer registries.8 Those included the Atlanta, Detroit, San Francisco-Oakland, and Seattle-Puget Sound metropolitan areas, as well as the states of Connecticut, Hawaii, Iowa, New Mexico, and Utah. The characteristics of the SEER population are comparable to those of the general population of the United States.8 Two kidney cancer diagnostic codes (International Classification of Diseases for Oncology, 2nd edition C64.9 code and 9th revision 189.0 code) were used as inclusion criteria. The presence of both diagnostic codes resulted in the identification of RCC patients, excluding patients with upper tract transitional carcinoma or noncortical renal tumors (ie, melanomas, sarcomas, and lymphomas). Only MRCC patients treated with cytoreductive nephrectomy were included (n = 2695). Further exclusion criteria consisted of patients with unknown tumor size and unknown Fuhrman grade (n = 1542). This resulted in 1153 assessable patients who were treated with cytoreductive nephrectomy for MRCC.
The cause of death was defined according to the SEER specific cause of death code (code 29020). For the purpose of this analysis, deaths from RCC were coded as cause-specific events. All other deaths were considered as other-cause mortality. Fuhrman grades were dichotomized between low1, 2 and high.3, 4 This Fuhrman classification results in the most sparse use of the degrees of freedom and maximizes the efficiency of statistical analyses.9
Patients were divided into 3 groups according to lymph node status: lymph node positive (N1-2), lymph node negative (N0), and unknown lymph node stage (Nx). Analysis of variance and chi-square tests were, respectively used for comparisons of means and proportions between the 3 groups. Kaplan-Meier plots graphically explored the effect of lymph node stage (N0 vs N1-2 vs Nx), Fuhrman grade (low vs high), and tumor size quartiles (≤6.0 cm vs 6.1-9.0 cm vs 9.1-11.5 cm vs >11.5 cm) on cancer-specific mortality. Thereafter, univariate and multivariate Cox regression analyses were used to test the significance and to quantify the magnitude of the effect of lymph node stage (N0 vs N1-2) on cancer-specific mortality. The covariates were comprised of age, sex (women vs men), race (other vs Caucasian), year of surgery (coded continuously in years), tumor size (coded continuously in cm), and Fuhrman grade (low vs high). Cox regression coefficients were used to quantify the univariate and multivariate accuracy of individual variables, as well as their combined effect, in the prediction of cancer-specific mortality. In accuracy analyses, a value of 100% indicates perfect prediction; 50% is equivalent to the toss of a coin. Predictive accuracy is usually quantified with receiver operating characteristic-derived area under the curve (AUC) estimates. In Cox regression models, the AUC is substituted with the Harrell concordance index,10 which was used in the current analyses. Two hundred bootstrap resamples were used to reduce overfit bias and for internal validation of all accuracy estimates.
All statistical tests were performed using S-PLUS Professional statistical software (version 1; Mathsoft, Seattle, Wash) or the Statistical Package for Social Science (version 15.0; SPSS Inc, Chicago, Ill). Moreover, all tests were 2-sided, with a significance level set at .05.
Of 1153 assessable patients with MRCC who were treated with cytoreductive nephrectomy, 39.5% had no evidence of lymph node metastases (N0), 29.7% were found to harbor lymph node metastases (N1-2), and 30.9% did not undergo a lymph node dissection (Nx). Table 1 shows the characteristics of the cohort. Overall, 66.5% of patients were men. The majority (85.4%) were Caucasian. The mean and median size of the primary tumor was 92.2 mm and 90 mm, respectively. The majority of patients had high Fuhrman grade tumors (67.7%), and the majority of surgeries (39.4%) were performed between the years 2001 and 2004. After stratification according to lymph node stages N1-2 versus N0 versus Nx, statistically significant differences were identified between the 3 groups. Those differences were comprised of mean patient age (59.1 years vs 60.8 years vs 62.4 years; P = .001), mean tumor size (96.6 mm vs 92.3 mm vs 87.9 mm; P = .01), and the proportion of patients with a high Fuhrman grade (77.8% vs 65.7% vs 60.7%; P < .001). The mean follow-up of censored cases was found to be significantly longer in patients with N0 and Nx disease compared with those with N1-2 disease (31.1 months and 41.7 months, respectively, vs 19.0 months; P < .001). The potential effect of those differences was adjusted for in multivariate Cox regression models.
Table 1. Descriptive Characteristics of 1153 Patients Diagnosed With Metastatic Renal Cell Carcinoma and Treated With Cytoreductive Nephrectomy*
Data were stratified according to lymph node stage (N0 vs N1-2 vs Nx).
Total no. of patients
Mean age, y (median), range
60.8 (61.0), 25-95
60.8 (61.0), 29-95
59.1 (60.0), 25-89
62.4 (63.0), 26-84
Year of surgery
Mean tumor size, mm (median), range
92.2 (90.0), 9-250
92.3 (90.0), 9-250
96.6 (91.5), 10-230
87.9 (85.0), 10-220
Mean follow-up of censored cases, mo (median), range
31.6 (14.0), 0.1-200
31.1 (11.0), 0.1-200
19.0 (9.0), 0.1-168
41.7 (26.0), 0.1-187
The median cancer-specific mortality-free survival was 16 months in the overall population (Fig. 1A). The median cancer-specific mortality-free survivals according to lymph node stage N1-2 versus N0 versus Nx were, respectively, 10 months versus 22 months versus 18 months (Fig. 1B). A significant difference was found to exist between the N1-2 and N0 categories (P < .001, log-rank test). Conversely, no significant difference was observed between the N0 and Nx categories (P = .3, log-rank test).
Cox regression models were restricted to patients with pathologically confirmed lymph node stage (N1-2 and N0). In univariate Cox regression models, lymph node stage (P < .001), Fuhrman grade (P < .001), tumor size (P = .001), year of surgery (P = .02), and race (P = .03) were found to achieve statistical significance (Table 2). Patients with lymph node metastases demonstrated a 2-fold higher rate of cancer-specific mortality than their counterparts without lymph node metastases (hazards ratio [HR], 2.0; P < .001). High versus low Fuhrman grade demonstrated a similar effect (HR, 1.8; P < .001). In multivariate Cox regression models (Table 2), the presence of lymph node metastases determined a 1.9-fold increase rate of cancer-specific mortality (P < .001) versus 1.8-fold for high versus low Furhman grade (P < .001). In addition, tumor size (P = .001), year of surgery (P = .001), and race (P = .002) achieved independent predictor status.
Table 2. Univariate and Multivariate Analyses Predicting the Probability of Cancer-Specific Mortality in Patients Treated With Cytoreductive Nephrectomy and Lymphadenectomy (n = 797)*
Multivariable Analysis, Full Model, HR (95% CI); P
HR (95% CI); P
AUC of Individual Predictor Variables
HR indicates hazards ratio; 95% CI, 95% confidence interval; AUC, area under the curve; N1-2, positive lymph nodes; N0, negative lymph nodes.
The AUC reflects the prognostic value of individual variables (columns), as well as of multivariate models in predicting cancer-specific mortality.
Lymph node stage N1-2 vs N0
1.97 (1.65-2.35); <.001
1.86 (1.56-2.23); <.001
Fuhrman grade, high (3-4) vs low (1-2)
1.79 (1.47-2.18); <.001
1.74 (1.42-2.13); <.001
1.003 (1.00-1.01); .001
1.004 (1.00-1.01); .001
Year of surgery
0.98 (0.96-1.00); .03
0.97 (0.95-0.99); .001
Race, other vs white
0.74 (0.57-0.95); .02
0.75 (0.58-0.97); .03
Sex, women vs men
0.99 (0.82-1.19); .9
0.99 (0.82-1.19); .9
1.00 (0.99-1.01); .8
1.00 (0.99-1.01); .2
AUC of multivariate models
In accuracy analyses based on AUC values, of all the variables, lymph node stage achieved the highest accuracy (59.5%); Fuhrman grade ranked second (57.0%). When all variables were considered simultaneously, the exclusion of lymph node stage resulted in a 3.2% decrease in accuracy versus a 2.5% decrease when Fuhrman grade was removed from the model.
In the current study, we tested the prognostic significance of lymph node stage in patients with MRCC who were treated with cytoreductive nephrectomy and lymphadenectomy. The lack of consideration of lymph node stage in the vast majority of prognostic schemes is surprising, especially in the light of the current findings, which confirmed our initial hypothesis that lymph node stage can discriminate between poor and favorable-risk, surgically treated MRCC patients. These findings demonstrated that the median survival of patients with N0 MRCC who are treated with cytoreductive nephrectomy was 22 months versus 10 months in patients with established lymph node metastases. The independent predictor status of lymph node stage was confirmed in multivariate Cox regression models and indicated a 1.9-fold increase in the cancer-specific mortality rate (P < .001). Moreover, the exclusion of lymph node stage from the multivariate analyses resulted in a 3.2% decrease in accuracy. This decrease was more significant than the decrease related to the removal of Fuhrman grade (2.5%). It is interesting to note that lymph node stage surpassed the accuracy of Fuhrman grade to predict cancer-specific mortality. Fuhrman grade represents a predictor variable in the University of California at Los Angeles Integrated Staging System,6 as well as in the prognostic nomograms of Karakiewicz et al.7
To the best of our knowledge, this is the first report that demonstrated and quantified the prognostic significance of lymph node stage in a population-based cohort of patients treated with cytoreductive nephrectomy for MRCC. Only 1 previous report by Vasselli et al.11 confirmed the bivariate statistical significance of lymph node stage. The only other variable included in the reported multivariate analysis was Eastern Cooperative Oncology Group performance status.11 Conversely, several multi-institutional reports demonstrated that lymph node metastases have major prognostic significance in patients with nonmetastatic RCC.12-14 The findings of the current study suggest that lymph node stage, which represents a readily identifiable variable, should be given more consideration in novel prognostic schemes devised for patients with MRCC.
In previous analyses, lymph node stage failed to reach independent predictor status. For example, in the study by Leibovich et al.,5 lymph node stage was not found to be an independent predictor of survival. This lack of independent predictor status was most likely based on the amount of detail provided by the 9 variables selected for inclusion in the final model, as well as on the limited sample size (n = 285). It is conceivable that overfit bias spuriously inflated the accuracy of some variables and lowered that of others. To obviate the problem related to overfit bias, we internally validated the regression coefficients and the AUC estimates. Therefore, in the current study, we reduced the chance of overfitting and ensured the reporting of the most bias-free estimates.
In 2 survival analyses that focused on patients with MRCC,15, 16 Pantuck et al. demonstrated that the presence of lymph node metastases was associated with more unfavorable tumor stage and worse survival. However, in 1 of these studies, multivariate tests were not performed.15 In the other, when multivariate analyses were performed, lymph node stage failed to reach independent predictor status.16 Limited sample size (n = 322) represents a possible explanation for this observation.15, 16
There are several implications related to our results. First, our findings suggest that the TNM-based staging and assignment of prognostic risk groupings with various schemes should rely on lymph node stage. Future studies should further address the value of lymph node stage, because it was the most informative prognostic variable (59.5%; P < .001) in the current study. Second, the TNM staging of MRCC, which classifies patients into a single category, may be overly simplistic. The current analysis demonstrated that at least 1 readily available staging variable, lymph node stage, can discriminate between poor-risk and favorable-risk MRCC patients. The accuracy gain related to the consideration of lymph node stage was 3.2%. To the best of our knowledge, few other prognostic variables can improve accuracy by that amount. For example, in a previous study, performance status and symptom score failed to improve the prognostic discrimination of nomograms addressing the cancer-specific mortality.17 Third, from a therapeutic perspective, the detrimental value of lymph node metastases on RCC-specific mortality may be interpreted as a confirmation of the existing consensus regarding the importance of complete resection of radiographically or palpably abnormal lymph nodes. Unfortunately, to the best of our knowledge, randomized data supporting the survival benefit of lymphadenectomy in MRCC patients do not exist.18
In addition to the main findings related to the importance of lymph node stage in patients with MRCC, we also observed that year of surgery and race affect cancer-specific mortality in patients treated with cytoreductive nephrectomy and lymphadenectomy. More favorable cancer-specific mortality rates were recorded in more recent years. This observation may be because of stage migration that may have been triggered by better imaging and earlier detection of metastases in patients with MRCC. Unfortunately, our study cannot offer a valid and conclusive explanation concerning the effect of race on cancer-specific mortality. Nonetheless, these results should prompt other studies that rely on more detailed databases to better define, corroborate, or refute the significance of race.
Several limitations apply to the current study. First, we lacked detailed information regarding the number and site of metastases, as well as information regarding other clinical and pathologic variables that were considered by Leibovich et al.5 or by Motzer et al.4 The availability of either the criteria of Leibovich et al.5 or Motzer et al.4 might have undermined the prognostic significance of the lymph node stage in the current analysis. Consequently, the importance of considering lymph node stage applies particularly to patients for whom these detailed variables are not available, or who fall into the favorable category of Motzer et al.4 Second, we could not adjust for the extent and completeness of the resection of lymph node metastases. Those parameters were shown to have a prognostic impact on survival rates.11 Third, approximately 30.9% of patients treated with cytoreductive nephrectomy did not undergo a lymphadenectomy. In Kaplan-Meier analyses of cancer-specific mortality, these patients behaved exactly like patients without lymph node metastases. However, the exact pathologic stage of these individuals is not known. It is conceivable that some may have fallen into the N1 category, in which a single positive lymph node is found at pathologic evaluation of the lymphadenectomy specimen. It is possible that such patients behave exactly like the counterparts without pathologically confirmed lymph node metastases. However, this assumption is speculative. Therefore, we decided to restrict the Cox regression analyses to patients with pathologically proven lymph node stage. Fourth, we have no details available regarding the handling of pathologic specimens and the tumor evaluation, which have certainly varied between centers. Fifth, because our regression and predictive accuracy analyses are restricted to patients with pathologically confirmed lymph node stage, our findings are also limited to patients who undergo a lymph node dissection. We hope that the importance of lymph node stage will encourage urologic surgeons to routinely perform lymphadenectomies at the time of cytoreductive nephrectomy. Sixth, we have no information regarding adjuvant and/or salvage treatment regimens. Some patients received adjuvant immunotherapy, whereas others received immunotherapy at the time of disease recurrence. Some were treated with experimental chemotherapy, whereas others received only best supportive care. Nonetheless, our survival findings might have been contaminated by the effect of immunotherapy in some individuals. During the study period, most patients received interferon, which was demonstrated to have a modest impact on survival.19 Therefore, therapies other than cytoreductive nephrectomy and lymphadenectomy were unlikely to affect cancer-specific mortality rates.20 Finally, it should be emphasized that the group of patients staged with a lymphadenectomy likely represents a heterogeneous group, because the decision to perform a lymphadenectomy might be based on idiosyncratic beliefs and situational factors.
Despite these limitations, the current study represents, to the best of our knowledge, the largest analysis performed to date of the prognostic significance of lymph node stage in patients with MRCC who are treated with cytoreductive nephrectomy. Moreover, its population-based nature makes these findings highly generalizable.
Our population-based analysis of the prognostic variables demonstrated that lymph node stage (N0 vs N1-2) can discriminate between poor and favorable-risk MRCC patients treated with cytoreductive nephrectomy and lymphadenectomy. Unfortunately, virtually all staging schemes do not consider the role of lymph node stage for risk stratification of surgically managed MRCC patients. Consequently, lymph node stage warrants consideration in future prognostic schemes.
Conflict of Interest Disclosures
Dr. Karakiewicz is partially supported by the University of Montreal Urology Associates, Fonds de la Recherche en Santé du Québec, the University of Montreal Department of Surgery, and the University of Montreal Foundation.