UROLOGICAL ONCOLOGY Prognostic Impact of Cytoreductive Nephrectomy in Patients with Metastatic Renal Cell Carcinoma: Data from a Large Population-Based Database Yanal Alnimer1*, Ayman Qasrawi2, Donglin Yan3, Peng Wang4 Purpose: Cytoreductive nephrectomy (CN) was considered a well-established treatment modality for patients with metastatic renal cell carcinoma (RCC) in the interferon era. However, its role after the introduction of multiple targeted therapies is less well established. Herein, We evaluated the effect of CN on overall survival (OS) on pa- tients with RCC who were identified through the Surveillance, Epidemiology, and End Results database (SEER). Materials and Methods: A total of 5,483 patients with metastatic RCC were identified from 2010 to 2016 using the SEER database. Factors pertaining to the following variables were collected: presence or absence of CN; age; gender; grade; status of metastasis to bone, liver, lung and brain; tumor stage; nodal status; histological subtypes; and chemotherapy status. Subjects who had CN were matched with those who did not in all previously mentioned covariates using inverse probability weighting. These weights were then used in adjusted Cox regression models to report doubly robust estimates. Results: CN was associated with 67% reduction in the hazards of death. Advanced T-stage, N1 disease, advanced tumor grade, non-clear histology and metastasis to bone, liver, lung or brain are independent risk factors for death. Patients with T4 disease benefited less of CN compared to those with T1 disease, while higher number of metastat- ic sites didn’t predict worse outcome among those who had CN. Conclusion: CN could provide a survival advantage in favorable risk patients with RCC in the era of targeted therapy. Keywords: cytoreductive nephrectomy; immunotherapy; interferon; metastatic renal cell carcinoma; overall survival INTRODUCTION Renal cell carcinoma (RCC) is an uncommon ma-lignancy that arises from the renal cortex. Patho- logically, RCC can be divided into several subtypes based on its morphology, molecular alterations, growth pattern, immunohistochemistry and cell of origin. Clear cell histology (CC) comprises the majority of RCC subtypes (75-85%) while papillary, chromophobe, on- cocytic and collecting duct tumors (of Bellini) account for the rest.(1) Cytoreductive nephrectomy (CN) was considered one of the main modalities of treatment in metastatic renal cell carcinoma (mRCC) in the era of interferon therapy. That was established after the publication of two ran- domized controlled trials; SWOG-8949 and EORTC; both have shown a survival advantage in patients who were treated with CN along with interferon compared to those who received interferon therapy alone. The median survival of the combined treatment modalities was 11.1 months compared to 8.1 months for the in- terferon therapy alone in the SWOG 8949 trial and 17 vs. 7 months in the EROTC trial.(2,3) Patients with good performance status (0-1) were included in these trials regardless of their tumor burden. 1Hospital Medicine, Virginia Commonwealth University, Richmond, Virginia, 23284, U.S.A. 2Division of Hematology, BMT and Cellular Therapy, University of Kentucky, Lexington, Kentucky 40536, U.S.A. 3Division of Medical Oncology, University of Kentucky, Lexington, KY 40536, U.S.A. 4Division of Medical Oncology, University of Kentucky, Lexington, KY 40536, U.S.A. *Correspondence: Hospital Medicine, Virginia Commonwealth University, , 3600 W. Broad Street Suite 115, Richmond, VA 23230 U.S.A.yanalalnimer@outlook.com. Received December 2020 & Accepted September 2021 However, the role of CN after the introduction of tar- geted therapies, which significantly improved survival in patients with mRCC, is still under debate. Several non-experimental studies demonstrated a survival ben- efit for CN. However, these studies were subjected to several biases.(4) In regards to experimental studies, the CARMENA trial demonstrated non-inferiority of sunitinib compared to the CN followed by targeted therapy arm in patients with mRCC. However, around 15% of the trial participants deviated from their treat- ment assignment, which could have contributed to the non-inferiority result of this trial. Also, the trial strati- fied patients based on their MSKCC prognostic indica- tors which has lower prognostic value in the targeted therapy era compared to international metastatic renal cell carcinoma database consortium (IMDC) prognos- tic indicators. Moreover, there was slight imbalance in the T stage between both groups and exclusion of pa- tients with low tumor burden which limits extrapolation of the trial results to this subgroup.(5) The SURTIME trial, a parallel randomized control trial that compared deferred CN after sunitinib to immediate CN. The trial did not show any difference in the progression free rate (PFR) between the groups who received an upfront CN Urology Journal/Vol 19 No. 2/ March-April 2022/ pp. 111-119. [DOI: 10.22037/uj.v18i.6593] followed by sunitinib compared to the one who deferred CN after sunitinib, but there was a statistically OS ad- vantage in the deferred CN arm.(6) The trial concluded that deferral of CN would help in identifying subjects with resistance to targeted therapy who are unlikely to benefit from CN. Nonetheless, the trial was limited by poor accrual rate; imbalance between the two arms in the proportion of subjects with three or more poor surgical risk factors and locally advanced disease; and early trial termination which could have biased the re- sult estimate away from the null. In another study using the IMDC, data from 1658 subjects with mRCC were retrospectively analyzed. The trial demonstrated 40% reduction in the hazard of death in mRCC patients who CN (Partial or total nephrectomy) (median, IQR, Proportions) No surgery) (median, IQR, Proportions) N 2991 (55%) 2492 (45%) Median age (95% CI) 60 (20-85) years 64 (18-85) years Gender Male 2115 (71%) 1753 (70%) Female 876 (29%) 739 (30%) Number of metastatic sites 0 199 (8%) 549 (18%) 1 1172 (47%) 1872 (62%) 2 785 (32%) 505 (17%) 3 282 (11%) 88 (3%) 4 38 (2%) 7 (0.2%) Given chemotherapy Yes 1596 (53%) 1477 (59%) No 1395 (47%) 1015(41%) Histology Clear cell 2436 (81%) 1909 (77%) Non-clear cell 555 (19%) 583 (23%) Grade Grade I / II 528 (18%) 312 (13%) Grade III / IV 2099 (70%) 378 (15%) Unknown 364 (13%) 1802 (72%) T stage T0 0 (0%) 33 (1%) T1/2 658 (22%) 869 (33%) T3 1518 (50%) 432 (17%) T4 282 (10%) 300(12%) TX 20 (1%) 483 (12%) Unknown 513 (17%) 408 (16%) Nodal positivity Yes 631 (21%) 692 (28%) No 1758 (59%) 1075 (43%) Unknown 602 (20%) 725 (29%) Table 1. Baseline characteristics for patients according to their cytoreductive nephrectomy (CN) status Figure 1. Weighted Kaplan Meier curves for the group who had CN and for those who didn’t. Cytoreductive nephrectomy in metastatic RCC-Alnimer et al. Urological Oncology 112 Vol 19 No 2 March-April 2022 113 underwent CN after adjustment for prognostic covari- ates including IMDC predictors. In this study, patients with estimated overall survival (OS) of less than 12 months and those with four or more IMDC prognos- tic indicators didn’t not benefit from CN.(7) On the oth- er hand, CN was associated with improvement in OS across all ranges of follow up in another retrospective study.(8) The latter study included subjects before the era of targeted therapy which limits drawing a firm causal conclusion of the effect of CN on survival after the era of targeted therapy. Due to the controversy in the literature regarding the role of CN, we identified patients with de novo mRCC using the SEER database after the approval of the tar- geted therapy. We studied the association between CN and OS in these patients. Also, we identified certain subgroups of patients who could benefit most from the CN. Since CN was demonstrated to have survival ben- efit in MRCC patients in 2009(9), we have included the SEER data between 2010-2016 to avoid any bias due to secular trend. MATERIALS AND METHODS Study population We used the SEER database to identify subjects who were diagnosed with mRCC (TNM Stage = M1) as their first malignancy from 2010 to 2016. The SEER Case Listing Session was used for analysis. Information was extracted from the database named “"Incidence - SEER 18 Regs Custom Data (with additional treatment Covariates that were strong predictors of treatment assignment. Reference category Estimate (OR) 95% CI P_value Age (years) 40-60 <40 1.28 0.55-2.78 0.55 > 60 0.69 0.30-1.49 0.37 Sex = male Female 1.09 0.84-1.41 0.50 Race White Black 1.14 0.74-1.72 0.54 Other 1.40 0.79-2.55 0.27 T stage T2 0.75 0.53-1.08 0.12 T3 T1 3.03 2.13-4.31 < 0.001 T4 0.69 0.46-1.04 0.07 Nodal status positive N0 0.36 0.28-0.48 < 0.001 Grade Grade II 4.00 2.21-7.44 < 0.001 Grade III Grade I 8.50 4.70-15.80 < 0.001 Grade IV 25.4 13.39-49.55 < 0.001 Histology = non-clear cell Clear Cell 0.80 0.58-1.11 0.17 Number of metastatic sites No liver, bone, lung or 0.43 0.37-0.50 < 0.001 brain metastasis Received chemotherapy No chemotherapy 0.62 0.48-0.80 < 0.001 Table 2. Results of the logistic regression comparing the variables that are associated with having CN. The following variables were included in the model to identify the variables that are strong confounders and are strongly associated with the exposure (probability of having CN): age; sex; race; T stage; nodal status; grade; histology; number of metastatic sites; and chemotherapy administration. Figure 2. Weighted Kaplan Meier curves for histological subtypes. Cytoreductive nephrectomy in metastatic RCC-Alnimer et al. Urological Oncology 114 fields), Nov 2018 Sub (1975-2016 varying)-Database ID: 30305". A total of 5,488 patients were identified as having microscopic confirmation of their metastatic disease. Inclusion and exclusion criteria Four patients with age of 18 years or less and one pa- tient with missing patient identification number were excluded from the study population. Therefore, a total Variables A) Regular Model B) T Stage Interaction (AIC 36198) C) Number of metastatic sites (AIC 36192) interaction (AIC 36194) Covariates that were Reference Estimate 95% CI P value Estimate 95% CI P value Estimate 95% CI P value strong predictors of category (HR) (HR) (HR) treatment assignment. Age (years) 40-60 <40 1.24 0.84-1.84 0.29 1.24 0.83-1.85 0.28 1.26 0.85-1.89 0.25 >60 1.33 0.90-1.98 0.16 1.33 0.90-1.98 0.15 1.36 0.92-2.01 0.13 Sex = male Female 0.87 0.73-1.03 0.11 0.88 0.74-1.04 0.14 0.87 0.74-1.03 0.10 Race White Black 1.16 0.86-1.57 0.32 1.16 0.86-1.57 0.33 1.19 0.88-1.60 0.26 Other 0.91 0.61-1.36 0.66 0.93 0.64-1.36 0.71 0.94 0.64-1.39 0.76 T stage T2 T1 0.97 0.76-1.25 0.82 0.82 0.55-1.22 0.34 0.96 0.74-1.23 0.72 T3 1.26 1.00-1.57 0.04 1.08 0.73-1.60 0.69 1.23 0.99-1.54 0.06 T4 1.51 1.14-2.00 0.004 1.17 0.73-1.87 0.51 1.49 1.12-1.98 0.005 Nodal status positive N0 1.47 1.23-1.76 < 0.001 1.47 1.24-1.75 < 0.001 1.51 1.26-1.81 < 0.001 Grade Grade II Grade I 1.61 1.02 -2.54 0.042 1.63 1.03-2.56 0.04 1.56 0.99-2.44 0.05 Grade III 2.06 1.30-3.27 0.002 2.06 1.30-3.28 0.002 2.02 1.29-3.17 0.002 Grade IV 2.58 1.62-4.11 < 0.001 2.57 1.61-4.11 < 0.001 2.54 1.61-4.00 < 0.001 Histology = non-clear cell Clear Cell 1.92 1.54-2.39 < 0.001 1.94 1.57-2.42 < 0.001 1.94 1.57-2.39 < 0.001 Number of metastatic sites to No liver, 1.62 1.45-1.80 < 0.001 1.62 1.46-1.80 < 0.001 1.79 1.48-2.16 < 0.001 bone, liver, lung or brain. bone, lung or brain metastasis Underwent cytoreductive No CN 0.33 0.28-0.40 < 0.001 0.24 0.17-0.33 < 0.001 0.42 0.30-0.59 < 0.001 nephrectomy (CN) Interaction term CN: T Stage CN: T2 CN: T1 1.49 0.93-2.38 0.09 CN: T3 1.41 0.91-2.19 0.12 CN: T4 1.78 1.06-2.99 0.03 Interaction term CN: 0.82 0.67-1.00 0.048 Number of metastatic sites Table 3. Results of the weighted Cox regression evaluating the effects of the following covariates on overall survival (OS): Age, sex, race, T stage, nodal status, grade, histological types, number of metastatic sites to bone, liver, lung and brain and cytoreductive nephrectomy. We adjusted for the chemotherapy variable in all models through stratification since the relationship between this variable and survival violates the proportional hazard assumption. A) Regular model. B) Second model with interaction with T stage. C) Third model with interaction with number of metastatic sites. Figure 3. Weighted Kaplan Meier curves for the groups who had N0 disease and those who had N1 disease. Cytoreductive nephrectomy in metastatic RCC-Alnimer et al. Vol 19 No 2 March-April 2022 115 of 5483 patients comprised the study population and were included in our analysis. Methods The main purpose of our study is to identify the effect of CN on OS on patients with mRCC. The following covariates of interest that should consider strong con- founders for the effect of CN on OS were extracted from the database: age at diagnosis, gender, TNM stage according the 7th American Joint Committee of Cancer (AJCC) edition, nodal status, sites of metastasis (bone, liver, brain, and lung), tumor size, grade, histology, du- ration of follow-up, date of death or loss to follow up, status of chemotherapy, and type of surgical resection. In order to study the causal effect of CN on OS, we matched subjects who had partial or total nephrectomy with those who didn’t using the weights that were cre- ated using the generalized boosted model (gbm). This method can achieve a good balance on the covariates of interest even in the absence of significant overlap in the propensity scores between the groups. We used the average treatment effect on the treated (ATT) estimates to study the potential outcome of patients who had CN Figure 5. Weighted Kaplan Meier curve for the groups with different number of metastatic sites. Figure 4. Weighted Kaplan Meier curves for the T-stage categories. Cytoreductive nephrectomy in metastatic RCC-Alnimer et al. Urological Oncology 116 if they did not have surgery. We matched patients who had CN with those who didn’t on the following covar- iates that we considered as a strong and intermediate confounder: age, gender, nodal status (N0, N1), chemo- therapy status, T-stage (T1, T2, T3, T4), grade (well differentiated, moderately differentiated, poorly differ- entiated, undifferentiated), number of organs involved by metastasis(1,2,3,4), race (white, Black, others) and his- tology (clear cell, non-clear cell). We categorized the age into three groups (less than 40, 40-60, more than 60), and chemotherapy was dichotomized as defined in the SEER database into (yes, no or unknown). We created a variable that we named “number of metastat- ic sites”. This indicates the number of organs (brain, bone, liver, and lung) involved by metastasis. We used the twang package in R to balance these covariates by creating weights using the gbm. We didn’t use the tu- mor size as a covariate in our matching process since we considered the T-stage. Stage and grade with val- ues of X in the SEER database were considered NA in our dataset. Nine patients with extreme weights were excluded from the final data analysis. Supp. Figure 1 (supplementary material) shows the propensity score distribution between patients who had CN and those who did not. Supp. Figure 2 (supplementary material) shows a significant decrease in the standardized mean difference (SMD) in the pretreatment covariates be- tween the group who had CN and those who did no’t. Kaplan Meier (KM) curves for CN pretreatment co- variates were created while considering the weights generated by the gbm. Results of weighted KM curves with log-rank test statistic are shown in Figures. 1 to 6 (Figures are reported only for the covariates that were significant predictor for survival in the Cox-regression models). Finally, we ran different models using the “svycoxph” command in R to account for the weights that were gen- erated. All models included CN as an outcome and the following covariates: age, number of metastatic sites, T stage, nodal status, grade, gender, race, histology, and chemotherapy status. Furthermore, different models were considered with an interaction term between CN and grade, CN and nodal status, CN and the number of metastatic sites, CN and the histological types and CN and T stage in addition to a model without any interac- tion term. Because the KM curves for the chemotherapy variable crossed each other, chemotherapy variable was stratified in Cox regression analysis using strata analy- sis and therefore we did not generate a hazard estimate for this variable. Models were compared using Akai- ka information criteria (AIC). Running cox-regression with weights under survey command precluded per- forming likelihood ratio test on nested models. Finally, we performed sensitivity analysis to assess our result sensitivity to an unobserved confounder. We re- lied on Rosenbaum approach to evaluate the strength of the association of an unobserved confounder (U) to our exposure of interest (CN nephrectomy) and to our outcome (OS) to change our results to non-significance. Because this approach relies on 1:1 matching, we creat- ed a matched dataset between subjects who had CN and those who didn’t using the MATCHIT package in R on the same covariates used in our primary analysis. We used a nearest matching method on propensity score with a caliper of 0.2 to achieve a good balance on the pretreatment covariates. In this analysis, 977 patients who had CN were matched to 977 patients who didn’t have CN (36% of the total cohort). Sensitivity analysis was ran using Rosenbaum spread sheet on survival out- come via Wilcoxon rank test. RESULTS Between 2010 and 2016, 5483 patients with mRCC were identified using the SEER registry. The patients’ median age was 62 years old (range 18-85). The ma- Figure 6. Weighted Kaplan Meier curves for the four levels of the tumor grade. Cytoreductive nephrectomy in metastatic RCC-Alnimer et al. Vol 19 No 2 March-April 2022 117 jority were males (70%) and 4345 (79%) had CC his- tology while the rest had nonnuclear cell (NCC). NCC included patients with chromophobe, papillary, collect- ing duct, medullary, oxyphilic, squamous, transitional and sarcomatoid RCC. There was a total of 2996 (54%) patients who underwent partial or radical nephrectomy. Baseline characteristics of the patients are summarized in Table 1. After a median follow-up time of 9 months (3-83 months), the median overall survival (OS) was 13 months. The median survival time for the cohort who had CN was 24 months (CI, 22-25) and 6 months (CI, 6-7) for patients who did not have surgery (P < 0.01, Figure 2). Grades III and IV, N1 disease, higher num- ber of metastatic sites, female gender, and non-clear cell histological types were associated with higher risk of death, while CN was associated with improvement in survival in weighted adjusted KM curves using log- rank test statistics. Also, there was a statistically signif- icant increase in the trend of using chemotherapy from 2000 to 2016 with a P value of 0.038 for the trend. On the other hand, there is a trend toward lower numbers of CN from 2010 to 2016 with a P value for a trend of 0.047 (Supplementary Material, Supp. Table 1). All potential confounder variables were included in a logistic regression model to identify variables that strongly predict treatment assignment (CN). Patients with T3 stage, Grade II-IV, lower number of organ in- volvement with metastasis and those who didn’t receive chemotherapy were more likely to have CN. Results of the logistic regression model are shown in Table 2. All pretreatment covariates that were used in weight- ed matching between subjects who had CN and those who didn’t were included in the Cox models to obtain a doubly robust estimate. All models with an interaction term have AIC values slightly above the model without an interaction term, therefore, the latter model with an Akika information criteria (AIC) value of 36192 was considered the primary model for our results. In this model, CN was associated with 67% reduction in the risk of death in patients with mRCC (HR 0.33, 95% CI 0.28-0.40). Non-clear cell histological subtype was in- dependently associated with higher risk of death (HR 1.9, 95% CI 1.54-2.40). Also, patients with nodal in- volvement and those with T4 disease had a statistically significant higher risk of death compared to those with N-0 and T0 disease (HR 1.50, 95% CI 1.23-1.75, and HR 1.5, 95% CI 1.14-2.0, respectively). Moreover, for each one-point increase in the number of metastat- ic sites, there was a statistically significant increase in risk of death with a HR 1.60, 95%CI 1.45-1.80. Final- ly, Grades II, III and IV were independently associated with higher risk of death compared to Grade I (HR 1.60, 95% CI 1.02-2.54; HR 2.1, 95% CI 1.3-3.3; and HR 2.58, 95% CI 1.62-4.11, respectively). Results of dou- bly robust Cox regression model are shown in Table 3. Only two models with an interaction term had a statis- tically significant P-value for the interaction. The test that included an interaction between tumor T stage and CN nephrectomy had a statistically significant interac- tion between CN and T-4 disease with a HR 1.78, 95% CI 1.10-2.99 (The AIC for the model is 36198). This result indicates that patients with T4 stage who had CN had 78% higher risk of death compared to those with T1 disease who had CN. If there is no interaction under the multiplicative interaction model, we would expect the HR to be 0.28 (1.17; the HR for the T4 stage; mul- tiplied by 0.24; the HR of CN). However, the current hazard of death for those who had CN and T4 disease is 0.5 (1.17; the HR for the T4 stage; multiplied by 0.24; the HR of CN, multiplied by 1.78; the HR for the inter- action term). In other words, patients with T4 disease would benefit less from CN compared to those with T1 disease. Similarly, the model that included an interac- tion term between surgery and number of metastatic sites showed a statistically significant estimate with a HR=0.82 (95% CI 0.67-0.997) indicating that patients with increased number of metastatic sites to bone, liver, lung or brain had 18% lower risk of death compared to those with no metastasis to bone, liver, lung or brain (The AIC for this model is 36194)). Under the assump- tion of no interaction and using the multiplicative in- teraction model between CN and number of metastatic sites, the HR of death for patients with higher metastatic sites who had CN should equal 0.75 (The HR for CN [0.42] multiplied by the HR for the increasing metastat- ic sites [1.79] under this model). However, when we multiply the estimates for patients who had CN (0.42) with the HR for the increased metastatic sites (1.79) and with the interaction term (0.82); the HR is 0.62. These results indicate that patients with increased metastatic sites would benefit more from CN compared to those with lower number of organ involvement. Results of doubly robust Cox regression model with interaction terms are shown in Table 3. Our sensitivity analysis showed an unobserved con- founder (U) that increase the odds of having CN by 42% and being a near perfect predictor of OS, would change our results toward the null. This U is plausible, since KPS which is not included in our model, could have such an association and it implies that our results could be sensitive to bias. However, this sensitivity analysis should be interpreted with caution since only 36% of the total cohort (1954 patients) were included in this analysis. The results of our sensitivity analysis are demonstrated in the supplementary material (Supp. Table 2). DISCUSSION The two large prospective randomized trials that eval- uated the role of CN in mRCC after the era of targeted have some limitations, mainly poor accrual, early trial termination and unbalanced randomization in regards to surgical risk factors and locally advanced disease in the SURTIME trial and deviation from the treatment assignment and inclusion of high risk patients for the CARMENA trial(Méjean et al., 2018).(5) This highlights the importance of non-experimental study designs in evaluating the role of CN in mRCC patients. Our data suggests a strong association between CN on OS. The median OS for the subjects who had CN was 24 months (CI, 22-25) compared to 6 months (CI, 6-7) for those who did not have CN with a HR of 0.33 (95%CI 0.28-0.40) in the doubly robust model in favor of CN. This is highly similar to an unadjusted weighted estimate (HR 0.38, 95%CI 0.33-0.44), indicating a great balance in all pretreatment covariates using an inverse probability weights on the treatment assigned (IPTW). Our results go in hand with the results from internation- al metastatic RCC consortium (IMRCC). In this study, subjects who had CN had an OS of 20.6 months com- pared to 9.5 months for those without CN.(7) Similarly, Cytoreductive nephrectomy in metastatic RCC-Alnimer et al. Urological Oncology 118 Vaishampayan et al and Pulmbo et al found that CN is associated with survival advantage in patients with lo- cally advanced or mRCC using the SEER database.(10,11) However, our study has a larger sample size with 2991 patients who underwent CN compared to 2483 patients who didn’t. Also, we applied a doubly robust method with inverse probability weighting and adjustment in regression analysis. This enabled us to effectively ad- just for the study confounders without excluding sub- jects from the CN group. This is important, because our inference measures the effect of CN on the whole sub- jects who had CN if they did not have surgery. Further- more, we conducted a sensitivity analysis to evaluate the strength of our association and the efficacy of CN on certain subgroups. In the SURTIME trial, patients with deferred CN had an OS advantage compared to those with immediate CN. However, the trial result should be interpreted with caution due to early termination which could have bi- ased the study estimate away from the null. Also, high- er number of subjects with surgical risk factors and T4 disease were assigned to immediate CN arm. The latter would support our result with a negative quantitative interaction between patients with T4 disease and CN and suggests that deferred CN could be the appropriate approach for patients with advanced T stage. We measured the average treatment effect on the treat- ed (ATT) in our analysis. Therefore, our results reflect the efficacy of CN on those who had CN if they did not have it. In our logistic regression model, subjects who didn’t receive chemotherapy by the time of CN, those with lower number of metastatic sites, and don’t have T4 disease were more likely to have CN and thus will have significant survival benefit from it. This supports the current practice, in which many centers perform CN for patients only with intermediate- or low-risk groups. These can be defined as the presence of four of less of the following risk factors: sarcomatoid features, low serum albumin, T3 or T4 disease, retroperitoneal or supradiaphragmatic lymphadenopathy and symptoms from metastatic sites such as bone and brain.(12) Other criteria could be used such as Karnofsky index (KPS) more than 80%, adequate organ function, ability to perform at least 75% tumor debulking and absence of extensive bone, liver and central nervous system me- tastasis.(13) Furthermore, data from IMRCC database revealed that patients who have four or more MSKCC risk factors (less than 1 year of diagnosis, KPS < 80%, serum lactate dehydrogenase >1.5, elevated serum cal- cium and low hemoglobin level) are unlikely to bene- fit from CN.(7) Our analysis shows that higher number of metastatic sites, advanced tumor grade and T stage and non-clear histology are independent risk factors for higher mortality. Although higher number of metastatic sites are associated with 60% higher hazard of death in our analysis, CN offers survival benefit among those subjects more than what would be expected taking into consideration their multiple organ metastases. This sup- port providing CN to patients with more than one met- astatic site if they are in good performance status and considered good surgical candidates. It is important to note here that there is a notable discrep- ancy between non-experimental studies that showed a significant survival benefit of CN in MRCC patients(7,10) and the randomized clinical trials (CARMENA and SURTIME clinical trials).(5,6) As noted earlier, In the SURTIME trial, quarter of the subjects who had imme- diate CN had three or more surgical risk factors. OS in those subjects could be negatively affected by immedi- ate CN. About half the subjects in the CARMENA trial had fallen into a poor risk category in the MSKCC mod- el. This could explain the discrepancy between these randomized trials and other non-experimental studies. This highlights the potential benefit of CN in low risk mRCC patients, and the need for a randomized control trial that is enriched with mRCC patients who have low risk features. Moreover, there is lack of enough data on the appropriate timing of CN relative to targeted thera- py. Further research on this area could identify the ap- propriate timing for CN that improves OS. Since we were unable to capture all the confound- ing covariates using the SEER database, we used the Rosenbaum approach to test the strength of the asso- ciation between CN and survival. Because this method and many other sensitivity analysis approaches rely on 1:1 matching, we performed a 1:1 matching analysis. In our sensitivity analysis, our gamma value was 1.421, which indicates an unobserved confounder that has at least 42% odds of association with CN and near prefect predictor of survival would change our estimate on CN toward the null. Usually, a gamma value of 4 or 5 will reflect a strong unbiased association. This indicates that our result is susceptible to bias and highlights the im- portance of conducting a randomized control trial with strict eligibility criteria to evaluate the effectiveness of CN after the introduction of targeted therapy. Our study has some limitations. Our analysis relies on an un-confounded assumption. However, given the limitations of the SEER database, some important co- variates were not included in our model such as KPS, comorbidities and covariates pertaining to the MSKCC risk model. This made our study susceptible to selection bias, namely confounding by indication. However, we used weights-based method to create a balanced data- set on all pretreatment variables that are strongly relat- ed to mRCC and to OS. Our new pseudo-dataset was well balanced on all pretreatment variables. Also, we matched subjects on the distribution of chemotherapy variable. Because patients with good performance sta- tus are more likely to receive chemotherapy, our analy- sis likely accounted for some confounding due to KPS. Metastatic disease was included as categorical variable for each site of metastasis (present or not) in the SEER database with no detailed information pertaining to the tumor burden. This could have resulted in residual con- founding that could have affected our study estimates. Also, no information regarding the extent of nodal in- volvement in the SEER database which precludes eval- uating the effect of CN in patients with extensive lym- phadenopathy. Dalong Cao et al. have used the SEER database to eval- uate the role of lymphadenectomy along with CN on survival using the SEER database from 2006-2015.(14) In addition, Vaishampayan et al. used the SEER data- base from 2000-2013 to evaluate the effect of CN on survival.(10) We included data only from 2010-2017 only to avoid the secular trend that have been observed since the introduction of targeted therapies in 2009. Many patients, particularly those without CN have missing values on tumor grade. This could possibly bias our results if these missing values were not missing at random (non-missingness). Nonetheless, we matched Cytoreductive nephrectomy in metastatic RCC-Alnimer et al. Vol 19 No 2 March-April 2022 100 both treatment groups on missing values which will de- crease the bias. CONCLUSIONS CN could provide a survival advantage to selected pa- tients with mRCC, regardless of histology. The risks and benefits of surgery should be discussed thoroughly and offered to patients on a case-by-case basis. Rand- omized trials with restricted inclusion criteria to low- risk patients is needed to fully disentangle the causal effect of CN on survival, especially with the new era of widespread use of immune checkpoint inhibitors in patients with mRCC. ACKNOWLEDGEMENT We would like to thank Russell-Simmons from the re- search communication office for her final editing and comments that greatly helped in improving the final manuscript. CONFLICT OF INTEREST The authors confirm that there is no conflict of interest regarding the publication of this paper. APPENDIX https://journals.sbmu.ac.ir/urolj/index.php/uj/libraryFiles/downloadPublic/37 REFERENCES 1. Thoenes W, Störkel S, Rumpelt HJ. Histopathology and classification of renal cell tumors (adenomas, oncocytomas and carcinomas). The basic cytological and histopathological elements and their use for diagnostics. Pathol Res Pract. 1986;181:125- 43. 2. Mickisch GH, Garin A, van Poppel H, de Prijck L, Sylvester R, Group EOfRaToCEG. Radical nephrectomy plus interferon-alfa-based immunotherapy compared with interferon alfa alone in metastatic renal-cell carcinoma: a randomised trial. Lancet. 2001;358:966-70. 3. Flanigan RC, Salmon SE, Blumenstein BA, et al. Nephrectomy followed by interferon alfa- 2b compared with interferon alfa-2b alone for metastatic renal-cell cancer. N Engl J Med. 2001;345:1655-9. 4. Graham J, Bhindi B, Heng DYC. The evolving role of cytoreductive nephrectomy in metastatic renal cell carcinoma. Curr Opin Urol. 2019;29:507-12. 5. Méjean A, Ravaud A, Thezenas S, et al. Sunitinib Alone or after Nephrectomy in Metastatic Renal-Cell Carcinoma. N Engl J Med. 2018;379:417-27. 6. Bex A, Mulders P, Jewett M, et al. Comparison of Immediate vs Deferred Cytoreductive Nephrectomy in Patients With Synchronous Metastatic Renal Cell Carcinoma Receiving Sunitinib: The SURTIME Randomized Clinical Trial. JAMA Oncol. 2019;5:164-70. 7. Heng DY, Wells JC, Rini BI, et al. Cytoreductive nephrectomy in patients with synchronous metastases from renal cell carcinoma: results from the International Metastatic Renal Cell Carcinoma Database Cytoreductive nephrectomy in metastatic RCC-Alnimer et al. Consortium. Eur Urol. 2014;66:704-10. 8. Hanna N, Sun M, Meyer CP, et al. Survival Analyses of Patients With Metastatic Renal Cancer Treated With Targeted Therapy With or Without Cytoreductive Nephrectomy: A National Cancer Data Base Study. J Clin Oncol. 2016;34:3267-75. 9. Motzer RJ, Hutson TE, Tomczak P, et al. Overall survival and updated results for sunitinib compared with interferon alfa in patients with metastatic renal cell carcinoma. J Clin Oncol. 2009;27:3584-90. 10. Vaishampayan U, George J, Vigneau F. Predictors of Cytoreductive Nephrectomy for Metastatic Kidney Cancer in SEER and Metropolitan Detroit Databases. J Kidney Cancer VHL. 2019;6:13-25. 11. Palumbo C, Mistretta FA, Knipper S, et al. Contemporary Cytoreductive Nephrectomy Provides Survival Benefit in Clear-cell Metastatic Renal Cell Carcinoma. Clin Genitourin Cancer. 2020;18:e730-e8. 12. Motzer RJ, Russo P. Cytoreductive Nephrectomy - Patient Selection Is Key. N Engl J Med. 2018;379:481-2. 13. Fallick ML, McDermott DF, LaRock D, Long JP, Atkins MB. Nephrectomy before interleukin-2 therapy for patients with metastatic renal cell carcinoma. J Urol. 1997;158:1691-5. 14. Cao D, Huang Y, Zhang C, et al. Adverse Effect of Lymph Node Dissection in Metastatic Renal Cell Cancer Patients Treated with Cytoreductive Nephrectomy: A Contemporary Analysis of Survival. J Cancer. 2019;10:4639- 46. Vol 19 No 2 March-April 2022 119