The first 2 authors contributed equally.
Cytoreductive nephron-sparing surgery does not appear to undermine disease-specific survival in patients with metastatic renal cell carcinoma
Article first published online: 16 OCT 2007
Copyright © 2007 American Cancer Society
Volume 110, Issue 11, pages 2428–2433, 1 December 2007
How to Cite
Hutterer, G. C., Patard, J.-J., Colombel, M., Belldegrun, A. S., Pfister, C., Guille, F., Artibani, W., Montorsi, F., Pantuck, A. J. and Karakiewicz, P. I. (2007), Cytoreductive nephron-sparing surgery does not appear to undermine disease-specific survival in patients with metastatic renal cell carcinoma. Cancer, 110: 2428–2433. doi: 10.1002/cncr.23054
- Issue published online: 19 NOV 2007
- Article first published online: 16 OCT 2007
- Manuscript Accepted: 31 MAY 2007
- Manuscript Revised: 29 MAY 2007
- Manuscript Received: 10 APR 2007
- nephron-sparing surgery;
- renal cell carcinoma;
The role of nephron-sparing surgery (NSS) showed promise in patients with metastatic renal cell carcinoma (MRCC). The disease-specific survival of patients with MRCC was compared according to the type of surgery, NSS (N = 45) versus radical nephrectomy (RN) (N = 732), in unmatched and matched analyses.
Kaplan-Meier, life tables, log-rank test, and univariate as well as multivariate Cox regression analyses addressed disease-specific survival of NSS versus RN patients. Subsequently, up to 4 RN cases were matched with each NSS case for TNM stage, Fuhrman grade, and histology. Then, disease-specific survival differences were tested with the log-rank statistic. Finally, the sample size necessary to achieve 80% power in survival analyses between the 2 groups (NSS vs RN) was calculated.
Of 45 NSS cases, 38 were matched with 99 of 732 RN cases. First, in multivariate unmatched analyses RN predisposes to 1.7-fold higher RCC-specific mortality rate; second, in matched analyses RN predisposes to 1.5-fold higher RCC-specific mortality rate; and third, both analyses failed to demonstrate statistically significant differences. Based on these findings it could be postulated that until further data become available, NSS does not appear to undermine RCC-specific survival in carefully selected patients with MRCC. The power analyses demonstrated that at least 146, 48, and 76 observations per arm are necessary at 1, 2, and 3 years, respectively, to confirm survival equivalence.
Although the data were limited in size and completeness, they may indicate that RCC-specific survival may not be undermined if NSS is performed in properly selected cases. Cancer 2007. © 2007 American Cancer Society.
Renal cell carcinoma (RCC) accounts for 3% of cancers in adults as well as 85% of all primary malignant kidney tumors.1 The 5-year survival rate for all stages of RCC has improved in recent years due to important stage migration, whereby the majority of patients are diagnosed with localized disease.2 Despite this stage shift, nearly 25% of contemporary patients are diagnosed with advanced disease, which includes patients with distant metastases.2 In this subset of patients, cytoreductive nephrectomy followed by interferon-based immunotherapy represents the current standard of care and was substantiated by 2 randomized controlled trials.3, 4
There is consensus regarding the role and the benefit of cytoreductive nephrectomy in patients with good performance status (ECOG 0–1) and distant metastases.3, 4 However, there are no studies addressing the optimal surgical approach to nephrectomy in metastatic RCC (MRCC) patients. A recent, highly controversial report addressed the effect of the surgical approach on disease-specific survival.5 It indicated virtual equivalence of survival between nephron-sparing surgery (NSS) (N = 16) and radical nephrectomy (RN) (N = 404). Based on survival and complications, the authors concluded that NSS represents a valid alternative to RN in patients with MRCC. The benefits of NSS in such patients include lower risk of renal insufficiency and associated need for dialysis, as well as better ability to treat or palliate the systemic manifestations of metastatic disease or its treatment.5 We addressed the same hypothesis in a larger cohort, where we performed 2 types of analyses. First, we addressed disease-specific survival in a standard multivariate Cox regression model. Then we assessed disease-specific survival after matching for tumor characteristics. Matching better controls for selection biases than multivariate analyses (MVA), especially when a small group (NSS) is compared with a much larger cohort (RN). Finally, we performed power calculations to determine the number of patients that are required to confirm survival equivalence between NSS and RN.
MATERIALS AND METHODS
Patient records were retrieved from institutional databases of 17 participating institutions and yielded 5167 patients treated with either partial or RN between 1984 and 2001 (Table 1). Within this cohort 796 patients had MRCC. Exclusions consisted of 14 cases for missing N stage, 4 cases for missing age, and 1 case for missing follow-up time. Of the remaining 777 patients, 45 underwent open cytoreductive NSS, whereas 732 had an open RN. The indications for NSS consisted of renal insufficiency of various causes and of a solitary kidney. In other cases, NSS was performed on an elective basis. Exact matches were performed for T and N stages, Fuhrman grade, and primary tumor histology. Up to 4 RN patients were matched with each NSS patient.
|Variables||Overall cohort||Nephron-sparing surgery||Radical nephrectomy||Nephron-sparing surgery||Radical nephrectomy|
|For M1 RCC patients||Matched cases||Matched controls|
|Mean (median)||60.8 (62.0)||63.7 (62.0)||60.6 (61.0)||62.5 (62.0)||62.4 (62.0)|
|Men||563 (72.5%)||32 (71.1%)||531 (72.5%)||26 (68.4%)||71 (71.7%)|
|T1||114 (14.7%)||36 (80.0%)||78 (10.7%)||29 (76.3%)||64 (64.7%)|
|T2||92 (11.8%)||3 (6.7%)||89 (12.1%)||3 (7.9%)||12 (12.1%)|
|T3||502 (64.6%)||6 (13.3%)||496 (67.8%)||6 (15.8%)||23 (23.2%)|
|T4||69 (8.9%)||—||69 (9.4%)||—||—|
|I||31 (4.0%)||5 (11.1%)||26 (3.6%)||4 (10.5%)||14 (14.1%)|
|II||207 (26.6%)||25 (55.6%)||182 (24.8%)||20 (52.7%)||46 (46.5%)|
|III||359 (46.2%)||14 (31.1%)||345 (47.1%)||13 (34.2%)||36 (36.4%)|
|IV||180 (23.2%)||1 (2.2%)||179 (24.5%)||1 (2.6%)||3 (3.0%)|
|N1||155 (19.9%)||7 (15.6%)||148 (20.2%)||3 (7.9%)||9 (9.1%)|
|N2||86 (11.1%)||3 (6.7%)||83 (11.3%)||1 (2.6%)||4 (4.0%)|
|Clear cell||693 (89.2%)||41 (91.2%)||652 (89.1%)||37 (97.4%)||98 (99.0%)|
|Papillary||46 (5.9%)||1 (2.2%)||45 (6.1%)||—||1 (1.0%)|
|Chromophobe||10 (1.3%)||2 (4.4%)||8 (1.1%)||1 (2.6%)||—|
|Collecting duct||9 (1.2%)||—||9 (1.2%)||—||—|
|Unclassified||19 (2.4%)||1 (2.2%)||18 (2.5%)||—||—|
|Overall study follow-up time, y|
|Mean (median)||1.8 (1.0)||1.8 (1.2)||1.8 (0.9)||1.9 (1.4)||2.9|
|Median actuarial cause-specific survival, y||1.4||5.1||1.3||5.1||3.3|
Clinical and Pathologic Evaluation
The TNM stages were retrospectively assigned according to the 2002 AJCC/UICC classification. Tumor size definition was based on pathological specimens and was defined as the greatest tumor diameter in centimeters. Histologic subtypes were stratified according to the 2002 AJCC/UICC classifications.6 Patients were staged preoperatively with computed tomography (CT) of the abdomen and pelvis, chest CT or chest x-ray, serum electrolytes, and liver function tests. The presence of nodal metastases was defined according to lymphadenectomy findings. In all cases a hilar lymphadenectomy was performed and included all lymph nodes on the ipsilateral side of the great vessels. In select cases, based on surgeon preference, more extensive lymphadenectomies were performed. In all cases the presence of nodal metastases was confirmed pathologically. The presence of distant metastases was confirmed based on radiographic findings.
Clinical and radiologic follow-up consisted of 1 postoperative baseline visit. Subsequently, the minimum follow-up consisted of at least 3 annual visits. At each visit CT of the chest or chest radiography accompanied a CT of the abdomen. Various forms of systemic therapy were delivered according to center protocols. These ranged from interferon (IFN-α) through high-dose interleukin (IL-2) regimens to chemotherapy in select cases. The timing, type, and dosage of systemic therapy varied between centers and could not be included in the analyses. The cause of death was obtained from medical charts and death certificates. RCC-specific mortality (RCC-SM) included deaths that were directly attributable to RCC.
Kaplan-Meier plots graphed cancer-specific survival and life table analyses determined actuarial survival predictions. The log-rank statistic compared cancer-specific survival rates. Univariate (UVA) and MVA Cox regression models tested the time to RCC-SM according to surgery type: RN versus NSS. Covariates included age, T stage, N stage, Fuhrman grade, and histology. Power analyses at 1, 2, and 3 years addressed the required number of subjects to achieve 80% power in a log-rank test given an alpha of 0.05 and a 2-sided comparison. All statistics were performed using S-PLUS Professional v. 1 (MathSoft, Seattle, Wash). Power analyses were performed with nQuery Advisor v. 4.0 (Statistical Solution, Saugus, Md).
The study cohort characteristics (Table 1) are stratified according to surgery type (NSS vs RN), as well as to analysis type (unmatched vs matched). Of all 777 MRCC patients treated with nephrectomy, 45 (5.8%) underwent NSS. Figure 1 shows the overall RCC-SS of individuals. The median follow-up of the entire cohort ranged from 0.1 to 20.0 years (mean, 1.8). At 1, 2, and 3 years the RCC-SS rates were, respectively, 58.2, 40.9, and 30.4% and the median survival was 1.4 years.
Stratification of the population between NSS and RN resulted in important differences in pathologic T stage (NSS 80.0% pT1 vs 10.7% for RN), Fuhrman grade (NSS 55.6% Fuhrman grade 2 vs 24.8% for RN), and histology (NSS papillary 2.2% vs 6.1% for RN). The overall length of follow-up time (NSS median 1.0 vs 1.2 years for RN) was virtually identical between the 2 populations. After matching, the distribution of T stages, N stages, Fuhrman grade, and histology was virtually the same between the 2 groups, as anticipated.
In the unmatched comparison the life table analyses demonstrated 1, 2, and 3 year disease-specific survival rates of 86.6 versus 73.3, 86.6 versus 60.2 and 75.0 versus 52.7%, respectively for NSS versus RN MRCC patients. The median actuarial survival (Fig. 1) of the RN versus NSS patients was 1.3 versus 5.1 years (rate ratio: 3.0; P < .001). In MVA Cox regression models (Table 2) that addressed the nonmatched population, when disease-specific survival was adjusted for age, T and N stages, as well as primary tumor Fuhrman grade and histology, RN was associated with a 1.7-fold higher, albeit statistically nonsignificant, rate of RCC-SM (P = .1) compared with NSS.
|Univariate analyses||Multivariate analyses|
|Rate ratio; P||Rate ratio; P|
|RN vs NSS||3.0; <.001||1.7; .1|
|Age||1.0; .2||1.0; .7|
|Histologic type (others vs clear cell)||1.7; <.001||1.4; .04|
|Fuhrman grade||—; <.001||—; <.001|
|II vs I||2.1; .02||1.6; .1|
|III vs I||2.9; <.001||2.1; .03|
|IV vs I||5.6; <.001||3.1; .001|
|T classification||—; <.001||—; <.001|
|T2 vs T1||1.9; .001||1.4; .1|
|T3 vs T1||2.2; <.001||1.5; .02|
|T4 vs T1||5.8; <.001||2.9; <.001|
|N classification||—; <.001||—; <.001|
|N1 vs. N0||2.1; <.001||1.6; <.001|
|N2 vs N0||1.3; .09||1.1; .8|
Of 45 NSS patients, 38 (84.4%) were matched with 99 (13.5%) of the 732 RN cases. Seven cases could not be matched with any RN patients. The matched analysis allowed us to examine RCC-SS according to the type of surgical procedure (NSS vs RN) after adjustment for all available tumor characteristics. Differences in tumor characteristics according to surgery type may confound unmatched analyses, especially when important sample size discrepancies exist between groups that are being compared.
The life table analyses that addressed the matched NSS and RN cases demonstrated 1, 2, and 3 year disease-specific survival rates of 86.6 versus 73.3, 86.6 versus 60.2, and 75.0 versus 52.7%, respectively, for NSS versus RN MRCC patients (Fig. 2). The mean and median follow-up times were, respectively, 1.9 and 1.4 years (range, 0.1–8.8 years) for NSS patients and 2.9 and 1.5 years (range, 0.1–20.0 years) for RN controls. The median actuarial survival of the NSS versus RN patients was 5.1 versus 3.3 years, which resulted in a 1.5-fold higher, albeit statistically nonsignificant, rate of RCC-SM (P = .2) when RN was compared with NSS.
Finally, power analyses performed for the observed survival rates in the matched population demonstrated that at 80% power the log-rank-based test would require 146, 48, and 76 patients per group at, respectively, 1, 2, and 3 years to confirm the equivalence in disease-specific survival.
The role of NSS has been recently examined in the metastatic setting.5 The indications for NSS in patients with localized RCC are well established and consist of pT1a disease, solitary kidney, bilateral renal masses, renal insufficiency, presence of hypertension, diabetes, or of hereditary RCC syndromes.7 Most recent reports have also confirmed the equivalence of cancer control outcomes of patients with T1b and T2 RCC treated with NSS relative to RN.7
Despite the direct and indirect benefits related to renal function preservation after NSS in patients with pT1–2 N0 M0 RCC, the effect of NSS on cancer control rates has only been examined in 1 series of MRCC patients.5 Besides the preservation of renal function, in MRCC additional benefits of NSS may consist of improved performance status, elimination of paraneoplastic symptoms, and eradication of the source of new metastases. However, the effect of NSS versus RN on cancer control has not yet been tested in the metastatic setting, except for 1 recent report.5 In this elegant yet provocative article, Krambeck et al. compared disease-specific survival according to surgical approach, namely, NSS versus RN. The comparison relied on 16 M1 renal cell carcinoma patients treated with partial nephrectomy and 404 M1 patients treated with RN. The survival curves demonstrated comparable outcomes, which prompted the authors to conclude that, relative to RN, NSS does not undermine disease-specific survival. However, the Krambeck et al. study was limited by sample size.
Sample size limitations may undermine the validity of statistical comparisons due to power limitations and in consequence due to spuriously high P-values. For example, in Krambeck et al.'s series the authors compared a large series of 404 MRCC RN-treated patients, which corresponded to 96% of the overall cohort, to a relatively small (3.8%) NSS patient cohort. Their comparison solely relied on UVA, which demonstrated a statistically nonsignificant survival benefit (P = .097) when NSS was performed relative to RN. Despite the novelty related to the use of NSS in MRCC patients, the methodology used to test the hypothesis of equivalence might have been limited by important selection biases. The criteria used for performing NSS consisted of either solitary kidney or bilateral disease. One NSS was performed on an elective basis.5 These criteria resemble ours and may result in the selection of patients in whom a partial nephrectomy is technically, biologically, and oncologically possible. Therefore, it is likely that patients who fulfilled these 3 criteria have in general lower T, N, and M tumor burden than their counterparts who do not qualify for NSS. These selection biases might have contributed to the apparently better survival of NSS cases versus RN patients.
We attempted to circumvent this limitation by performing a matched analysis. However, before matching we analyzed the entire cohort of MRCC patients and found that in UVA analyses RN does predispose to 3.0-fold higher RCC-SM in a statistically significant fashion (P < .001). Nonetheless, after MVA adjustment RN was associated with a higher, albeit statistically nonsignificant, rate of RCC-SM (rate ratio: 1.7; P = .1).
In matched analyses, 38 NSS cases were matched with 99 RN controls (1:3.8 ratio of NSS to RN). The life table analyses of the matched subgroups, as well as the Kaplan-Meier plots, demonstrated noninferiority of RCC-SS in NSS patients. Interestingly, despite matching, a 1.5-fold, albeit statistically nonsignificant, increase in RCC-SM was noted for RN cases (P = .2). This more balanced analysis has some limitations, which mainly consist of limited power to confirm or refute differences in disease-specific survival. Sample size limitations are important, as insufficient sample size may preclude the confirmation of statistically significant differences. To illustrate the sample size requirement for equivalence in survival curves, we performed power analyses, which demonstrated that at least 146, 48, and 76 patients per arm would be required to confirm the equivalence of mortality rates at respectively 1, 2, and 3 years. Therefore, our NSS sample size was 108, 10, and 38 observations too small to achieve this endpoint.
These results indicate that the study of Krambeck et al., as well as our study, are unable to statistically significantly confirm the equivalence of the survival patterns of the 2 groups. Based on this observation it appears premature to state that survival after NSS is definitely equivalent to RN in patients with MRCC. Nonetheless, our study confirms the noninferiority of RCC-SS with NSS that was previously reported by Krambeck et al..
Taken together, our results indicate the following: 1) in MVA unmatched analyses RN predisposes to a 1.7-fold higher RCC-SM rate, 2) in matched analyses RN predisposes to a 1.5-fold higher RCC-SM rate, and 3) both analyses failed to demonstrate statistically significant differences. Based on these findings it could be postulated that, until further data become available, NSS does not appear to undermine RCC-SS in carefully selected patients with MRCC.
Our findings can be explained by the observation that in carefully selected patients complete tumor removal can be achieved with either NSS or RN. Provided the completeness of tumor removal, the RCC-SS can be expected to be the same. Actually, our data indicate that NSS patients have more favorable prognosis than RN patients. This is consistent with lower stage, Fuhrman grade, rate of nodal metastases, and possibly with less extensive metastatic disease. Unfortunately, our database did not allow us to control for the extent of metastases. However, it is tempting to make this assumption.
A number of limitations apply to our analysis. These consist of lack of central pathology as well as of interinstitutional differences in surgical and adjuvant treatment of included patients. Similarly, indications and the selection of various systemic regimens may also have differed between centers. However, before the advent of kinase inhibitors, systemic interferon and IL-2 were virtually the same as medroxyprogesterone or placebo, at least in intermediate risk patients.8 Multiinstitutional data without central pathology review tend to introduce interobserver biases. For example, tumor grades may be interpreted differently by different pathologists.9 However, central review does not reflect the everyday practice patterns, where numerous pathologists stage, grade, and assign histologic variants to surgical specimens. Moreover, from a practical perspective it would be impossible to review the stage and grade of the primary tumor in a dataset of 5167 patients. The indications for NSS versus RN may have differed between individual surgeons who contributed to this study. Some may have had lower tumor size cutoff criteria than others when elective NSS was performed. We corrected for these differences by matching for T and N stages, Fuhrman grade, and histology of the primary tumor, which removed some but certainly not all of the population differences. Although we attempted to match for as many variables as possible, some variables were not measured and could not have been accounted for, such as the administration of systemic therapy, performance status, hemoglobin, calcium, and lactate dehydrogenase, which represent important predictors of survival in MRCC patients.10 The above limitations preclude definite conclusions about whether NSS can be used ‘electively’ in the cytoreductive setting.
Based on these limitations it is important to emphasize that neither our findings nor those of Krambeck et al.'s are final and that additional prospective studies will be required to confirm or reject these hypotheses.
Although our data were limited in size and completeness, they may indicate that RCC-SS may not be undermined if NSS is performed in properly selected cases. However, the sample size was insufficient to statistically significantly confirm the equivalence of outcomes between the 2 approaches.
- 6American Joint Committee on Cancer Staging Manual. 6th ed. New York: Springer; 2002., , , et al.