Nephrectomy improves the survival of patients with locally advanced renal cell carcinoma


Pierre I. Karakiewicz, Cancer Prognostics and Health Outcomes Unit, University of Montreal Health Center (CHUM), 1058, rue St-Denis, Montréal, Québec, Canada H2 X3J4.



To examine the cancer-specific survival of patients treated with nephrectomy and compared it to that of patients managed without surgery.


Of 43 143 patients with renal cell carcinoma (RCC) identified in the 1988–2004 Surveillance, Epidemiology and End Results database, 7068 had locally advanced RCC and with no distant metastasis. These patients had a nephrectomy (6786, 96.0%) or no surgical therapy (282, 4.0%). Multivariable Cox regression models, and matched and unmatched Kaplan-Meier survival analyses, were used to compare the effect of nephrectomy vs non-surgical therapy on cancer-specific survival. Also, competing-risks regression models adjusted for the effect of other-cause mortality. Covariates and matching variables consisted of age, gender, tumour size and year of diagnosis.


The 1-, 2-, 5- and 10-year cancer-specific survival of patients who had nephrectomy was 88.9%, 88.1%, 68.6% and 57.5%, vs 44.8%, 30.6%, 14.5% and 10.6% for non-surgical therapy. In multivariable analyses, relative to nephrectomy, non-surgical therapy was associated with a 5.8-fold higher rate of cancer-specific mortality (P < 0.001). Non-surgical therapy was also associated with a 5.1-fold higher rate of cancer-specific mortality in matched analyses (P < 0.001). Finally, competing-risks regression confirmed the statistical significance of the variable defining treatment type (nephrectomy vs non-surgical therapy) in multivariable and matched analyses (P < 0.001).


Relative to non-surgical treatment, nephrectomy improves the cancer-specific survival of patients with locally advanced RCC; our findings await prospective confirmation.


cancer/cause-specific survival


Surveillance, Epidemiology and End Results


International Classification of Diseases for Oncology.


The treatment options for patients with locally advanced RCC consist of surgical resection, systemic therapy or best supportive care. Surgical resection represents the treatment of choice [1], but the resection of locally advanced, especially T4 and/or N2 RCC, can be technically complex and in some instances can result in higher morbidity and mortality [2]. Furthermore, there are no level I and II data confirming a survival benefit in surgically treated patients with locally advanced RCC relative to other treatment options, e.g. systemic therapy or best supportive care [3–5].

Despite high-quality contributions on the fate of surgically treated patients with locally advanced RCC [6–8], no observational series have examined the cancer control outcomes of patients with locally advanced RCC, when surgery was compared with observation. Therefore, our intention was to examine the cancer-specific survival (CSS) of patients treated with nephrectomy for locally advanced, non-metastatic RCC and to compare it with that of patients managed without surgery. We decided to conduct the study using a large population-based registry, which approximates the population of the USA [9], with the aim of generalizing the most generally applicable findings that were not specific to a high-volume specialized care centre.


Patients diagnosed with primary invasive kidney cancer between 1988 and 2004 were identified within nine Surveillance, Epidemiology and End Results (SEER) cancer registries [9], including the Atlanta, Detroit, San Francisco-Oakland, Seattle-Puget Sound metropolitan areas, and the states of Connecticut, Hawaii, Iowa, New Mexico, and Utah. Characteristics of the SEER population are comparable with those of the general population of the USA [9]. Two kidney cancer diagnostic codes (International Classification of Disease for Oncology, Second edition, ICD-O-2, C64.9 code and the Ninth revision, ICD-O-9, 189.0 code) were used as inclusion criteria. The presence of both diagnostic codes resulted in the identification of 43 143 patients with RCC. This did not include upper tract TCC or ureteric, non-cortical renal tumours (i.e. melanomas, sarcomas and lymphomas). Exclusion of patients enrolled before 1988 was based on the unavailability of tumour size, which represents one of the key variables in the current analyses. Further exclusions consisted of patients with localized RCC (T1-2N0M0; 23 572, 54.6%), of those with overt distant metastases (TanyNanyM1; 8808, 20.4%) or of those with unknown stage (2969, 6.9%). Also, 585 patients (1.3%) with unknown tumour size were excluded from consideration, as were those who were autopsy-only cases, who died before planned surgery and who refused surgery (112, 0.2%). Finally, patients treated with local tumour destruction, thermal ablation, electrocautery or cryosurgery were also excluded (29, 0.07%). These exclusions resulted in 7068 patients with T3-4N0-2 RCC treated with either non-surgical therapy (282, 4.0%), radical nephrectomy (6575, 93.0%) or partial nephrectomy (211, 3.0%). For all nephrectomy patients, malignant histology was confirmed with the ICD-O-3 SEER histological codes. The cause of death was defined according to the SEER-specific cause of death (code 29020). Patients who did not die from RCC were considered as having died from other causes.

Independent-sample t-tests and chi-square tests were, respectively, used for comparing means and proportions between nephrectomy and non-surgical patients. The effect of treatment type (nephrectomy vs non-surgical therapy) on RCC-specific mortality was assessed in two ways. First, we used univariable and multivariable Cox regression analyses to test for the presence and to quantify the magnitude of the effect of treatment type (nephrectomy vs non-surgical therapy) on cause-specific mortality [10]. The covariates consisted of age, gender, tumour size and year of diagnosis. As the magnitude of the effect of treatment type might be overestimated, when one of the compared groups was substantially smaller than the other, we complemented the univariable and multivariable survival analyses with matched survival analyses [11]. In these we compared the actuarial cause-specific mortality rates between non-surgical therapy and nephrectomy. Each patient treated with non-surgical therapy was matched with up to four who had nephrectomy (partial or radical); assigning up to four controls maximizes the power of matched comparisons [11]. Exact matches were made for age, gender, year of diagnosis and for tumour size. For purpose of matching, decimals were not considered for age and tumour size, and the values rounded. After matching, the difference in the rates of cause-specific mortality was tested with the log-rank test and its magnitude assessed with the hazard ratio.

Kaplan-Meier analyses were used to graphically depict the univariable cause-specific mortality. Actuarial survival rates at various sample time after nephrectomy or various times after the diagnosis for non-surgical patients were calculated with life-table analyses. The date of diagnosis was considered as time 0 for non-surgical cases and the date of surgery represented the start of follow-up for surgery cases.

Given that a proportion of patients with RCC die from other causes, competing-risks regression, as described by Fine and Gray [12], was used to test the significance of the variable defining treatment type (nephrectomy vs non-surgical therapy) in predicting cause-specific mortality, after accounting for other-cause mortality. Competing-risks regression models were devised to test the statistical significance of variables of interest on time to event after controlling for other-cause mortality. A strong effect of competing mortality might result in extensive censoring due to cancer-unrelated deaths. This might in turn artificially reduce the pool of individuals at risk of RCC-specific events and overestimate the effect of cause-specific mortality. All tests were two-sided with a significance level set at 0.05.


The characteristics of the 282 non-surgical patients are described in Table 1. Of the 7068 assessable patients, 4703 (66.5%) were men and the mean age was 62.5 years; the median follow-up was 36.0 months (maximum 203).

Table 1.  The clinical characteristics of the study population of 7068 patients stratified according to treatment type (nephrectomy vs no surgery)
VariableTreatment typeP
No surgeryNephrectomyOverall
Total number28267867068 
Mean age, years 72.5  62.1 <0.001
Gender, n (%) men172 (61.0)4531 (66.8)4703 (66.5)0.046
Mean tumour size, cm  8.3   7.7 0.03
Year of diagnosis   0.16
 198812 (3.8)303 (96.2)  
 198910 (2.8)347 (97.2)  
 199017 (4.4)369 (95.6)  
 1991 6 (1.6)370 (98.4)  
 199222 (5.4)385 (94.6)  
 199313 (3.1)401 (96.9)  
 199410 (2.5)389 (97.5)  
 199520 (5.1)370 (94.9)  
 199617 (4.0)411 (96.0)  
 199713 (3.4)366 (96.6)  
 199821 (4.9)404 (95.1)  
 199921 (4.7)422 (95.3)  
 200021 (4.6)434 (95.4)  
 200118 (4.2)414 (95.8)  
 200223 (4.9)447 (95.1)  
 200316 (3.4)458 (96.6)  
 200422 (4.2)496 (95.8)  
Median (range) follow-up, months14.3 (0.1–187) 55.1 (0.1–203) <0.001
Patients with follow-up
 ≥2 years49 (17.3)4313 (63.5)  
 ≥5 years11 (3.9)2504 (36.9)  

After stratification according to treatment type, non-surgical patients were significantly older (72.5 vs 62.1 years, P < 0.001) and more frequently women (39.0% vs 33.2%, P = 0.046). Tumour size (8.3 cm for the non-surgical group vs 7.7 cm for nephrectomy, P = 0.03) and median follow-up (14.3 and 55.1 months, respectively, P < 0.001) were statistically different in the two groups. The proportion of patients who had non-surgical therapy did not change over time (chi-square trend test, P = 0.16).

At 1, 2, 5 and 10 years of follow-up the CSS rates were 88.9%, 88.1%, 68.6%, and 57.5% for nephrectomy, vs 44.8%, 30.6%, 14.5% and 10.6% for non-surgical therapy (hazard ratio 5.8, log-rank P < 0.001, Fig. 1a). In the multivariable analyses, after adjusting for age, gender, year of diagnosis and tumour size, the hazard ratio decreased to 5.0 but remained highly statistically significant (P < 0.001, Table 2). All variables, except for gender, achieved independent predictor status in multivariable competing-risks regression analyses, where other-cause mortality was accounted for (P < 0.001, Table 3).

Figure 1.

Kaplan-Meier survival curves depicting the cause-specific survival for: a, all 7068 patients; and b, the matched patients (200), stratified according to treatment type (nephrectomy vs no surgery).

Table 2. 
Univariable and multivariable Cox regression models predicting the rate of cause-specific mortality in the 7068 patients
VariableHazard ratio; P
Treatment type:
No surgery vs nephrectomy5.8; <0.0015.0; <0.001
Age, years
 50–59 vs <501.3; <0.0011.4; 0.002
 60–69 vs <501.4; <0.0011.6; <0.001
 70–79 vs <501.5; <0.0011.7; <0.001
 >80 vs <502.1; <0.0012.0; <0.001
Gender (male vs female)1.0; 0.41.1; 0.041
Tumour size, cm1.0; <0.0011.0; <0.001
Year of diagnosis0.992; 0.10.99; 0.048
Table 3.  Univariable and multivariable competing-risks regression models addressing all 7068 patients and the matched 200 patients for predicting CSS, after accounting for other-cause mortality
VariableAll, PMatched, P Univariable
Treatment type
No surgery vs nephrectomy<0.001<0.001<0.001
Age, years<0.001<0.001 
Gender (male vs female)  0.6  0.3 
Tumour size, cm<0.001<0.001 
Year of diagnosis<0.001<0.001 

In the second part of the analysis, we attempted to match the 282 patients who had no surgery with up to four treated surgically. Matching was possible for 88 of 282 patients (31.2%), for whom 112 surgical patients were identified. Matching was done for gender, age, year of diagnosis and tumour size. In this matched population, the 1-, 2-, 5- and 10-year CSS estimates were, respectively, 89.7%, 81.0%, 67.9% and 48.0% for the nephrectomy group, and at 1, 2 and 5 years the CSS was 46.7%, 31.9% and 12.8 for non-surgical patients (hazard ratio 5.1, log-rank P < 0.001, Fig. 1b). No data were available for assessing the 10-year survival estimates. Again, treatment type (nephrectomy vs non-surgical therapy, P < 0.001, Table 3) was statistically significant in the matched competing-risks regression models.


These results show a substantially better CSS in patients treated with nephrectomy than with no surgery in those with locally advanced kidney cancer. This survival disadvantage in the non-surgical group was apparent at 1, 2, 5 and 10 years of follow-up by CSS rates of 44.8%, 30.6%, 14.5% and 10.6%, vs 88.9%, 88.1%, 68.6% and 57.5% for the nephrectomy group. Overall, patients managed non-surgically had 5.8 times the rate of cancer-specific mortality than their surgical counterparts. As several biases can result in important differences in the characteristics of nephrectomy vs non-surgical patients, we used a multivariable analysis adjusted for age, gender, tumour size and year of diagnosis; the rate of cause-specific mortality remained 5.0 times higher for non-surgical patients (P < 0.001). Similarly, in matched analysis, where the matching variables consisted of age, gender, tumour size and year of diagnosis, patients who did not have a nephrectomy remained 5.1 times more likely to die than those treated with nephrectomy. The similarity of the results in both analysis types corroborates the validity of the survival disadvantage that is associated with the non-surgical management of locally advanced RCC.

The CSS data are consistent with other surgical series that addressed locally advanced RCC; e.g. the 5-year CSS was 67%[7], 56%[13] and 39.3%[14], respectively, in patients with pT3a, pT3b-c RCC and with pN1-2 RCC. The disadvantage in survival of patients not treated with surgery in the present series is also consistent with the survival of most of the patients with T3-4N1-2 RCC in the observational series of Baird et al.[5]. Five of the six patients with T3-4N1-2 RCC survived 4–22 months (mean 10.2, median 8). Only one patient with T3bN1 RCC survived 84 months, which increased the mean to 22.5 months and the median to 9.5 months. Unfortunately, no other non-surgical series have been reported to better substantiate the survival without nephrectomy.

Taken together, these data indicate that nephrectomy is associated with a substantially higher CSS in patients with T3-4N1-2 RCC. However, it might be argued that the reason for not using a nephrectomy might be related to multiple other comorbidities [15,16], which could predispose to early in deaths unrelated to cancer. Ideally, a comorbidity index could be used to adjust for other-cause mortality [17]; unfortunately, a comorbidity index is not available in the SEER database. To obviate this limitation and to validly address the potential confounding effect of other-cause mortality, we relied on the cause of death and used competing-risks regression models to adjust for this important confounding variable. In competing-risks regression, the effect of non-surgical management remained an independent predictor of cause-specific mortality in multivariable analysis. Similarly, the effect of surgical management remained highly statistically significant in matched competing-risks regression analyses. These findings indicate that surgical management has a protective effect on cause-specific mortality, even after adjusting for other-cause mortality. Consequently our data indicate that nephrectomy should be considered the treatment of choice for patients with locally advanced RCC.

The present study has some limitations; its design was not randomized and therefore the comparison might not be optimally adjusted for differences that might exist between the nephrectomy and the non-surgical groups. These differences might go beyond age, gender, year of diagnosis and tumour size, and could include comorbidities, renal function and the number of functional renal units. Unfortunately, the unavailability of variables accounting for these differences did not allow us to make the appropriate comparisons. It is unknown whether information on these differences would have obliterated or inflated the observed hazard rates, and only a randomized controlled trial that accounts for baseline comorbidities could validly confirm the present results. Moreover, our data only apply to patients from the USA; it is possible that different mechanisms are used for selecting non-surgical patients in Europe or other parts of the world. Our findings need to be corroborated in other datasets. The data type did not allow comparisons between laparoscopic vs open nephrectomy, but available data for these two methods do not suggest differences in cancer control rates between them [18,19]. Finally, our data could not account or adjust for the potential benefits of novel kinase inhibitors, as these molecules were not available and not approved during the study period. However, it could be argued that the neoadjuvant use of these agents might benefit those with unresectable locally advanced RCC. Moreover, several adjuvant protocols are evaluating the effect of these molecules on survival after nephrectomy in locally advanced and/or histologically unfavourable RCC [20].

In conclusion, our findings showed that non-surgical therapy is associated with a 44.1–57.5% worse survival than nephrectomy. Based on these findings, nephrectomy should be considered as the treatment of choice for patients with locally advanced RCC.


Pierre I. Karakiewicz is partially supported by the University of Montreal Health Center Urology Associated, Fonds de la Recherche en Santé du Quebec, the University of Montreal Department Of Surgery and the University of Montreal Health Center (CHUM) Foundation. Laurent Zini is partially supported by the Association Française de Recherche sur le Cancer, the Fondation de France- Fédération Nationale des Centres de Lutte Contre le Cancer, the Association Française d’Urologie and the Ministère Français des Affaires Etrangères et Européennes (Bourse Lavoisier).


None declared.