Radical Nephrectomy Provides a Worse Prognosis Than Partial Nephrectomy in T3aN0M0 Renal Cancer of Small ( ≤ 4 cm) Size and No Invasion of Perisinus Fat

Background: Radical nephrectomy (RN) is the recommended treatment for T3aN0M0 renal cell carcinoma (RCC). However, it is not necessarily the best treatment for small T3aN0M0 RCCs. We evaluated the effect of tumor size combined with consideration of anatomic types of extrarenal-fat invasion on the surgical decision-making between partial nephrectomy (PN) vs. RN in T3aN0M0 RCC. Methods: Data were obtained from the Surveillance, Epidemiology, and End Results database (2004 to 2015) with 6125 patients suffering from T3aN0M0 RCC. Cox and Fine and Gray models were used for survival analyses. Propensity-score matching was used for PN vs. RN. Results: A larger T3aN0M0 RCC was associated with higher risk of mortality (hazard ratio (HR) all-cause mortality : 1.07, 95% condence interval (CI): 1.02–1.13, P = 0.011; HR RCC-cause mortality : 1.13, 95%CI: 1.06–1.21, P < 0.001) compared with a small T3aN0M0 RCC. After propensity-score matching, in T3aN0M0 ≤ 4 cm, RN compared with PN signicantly increased the risk of death (HR: 1.77; 95%CI: 1.14–2.74, P = 0.011) and offered no signicant difference in RCC-specic survival (HR: 1.57, 95%CI: 0.74–3.36, P = 0.240). However, RN and PN showed no signicant difference in overall survival in T3aN0M0 RCC >4 cm (HR: 0.98; 95%CI: 0.59–1.62, P= 0.929) or in T3aN0M0 RCC with sinus/perisinus-fat invasion (HR: 1.18; 95%CI: 0.61–2.27, P = 0.631). Conclusion: PN provided better overall survival compared with RN for small ( ≤ 4 cm) T3aN0M0 with a Wilcoxon rank-sum test. Categorical variables are presented as frequencies (%) and were compared using a chi-square test. Overall survival (OS) was compared using the Kaplan–Meier method for survival function along with the log-rank test. Multivariable Cox proportional hazards regression models were conducted. Fine and Gray competing risk proportional hazard regression models were tted to assess mortality caused by RCC and other competing events[11]. and follow-up time are as the The study design was a retrospective single-center study; tumors without lymph-node and distant metastasis and Kaplan–Meier and multivariate analysis were used for survival prediction.


Background
With the widespread use of partial nephrectomy (PN) in T1 renal cell carcinoma (RCC) [1], numerous patients with T1 RCC have undergone pathologic upgrades to T3aN0M0 after PN [2][3][4]. This action has led to a treatment-strategy dilemma (i.e., clinical observation or conversion to radical nephrectomy (RN) for such pathologic upstaging post-PN). In clinical practice, RN is preferred because there is not su cient evidence to show the value of PN in T3aN0M0 RCC.
However, some patients with RN might not obtain a better survival outcome. Besides, how to select patients suitable for PN is not known. Some studies have indicated that the prognosis of PN for small T3aN0M0 RCCs is satisfactory [4][5][6][7]. Those studies showed the limitation of heterogeneity and small sample sizes. Whether PN can provide a better prognosis in select cases of T3aN0M0 RCC and whether patients with small tumors and a certain type of fat invasion (perirenal fat vs. sinus/perisinus fat) can impact the choice of PN is not known [8][9][10]. Further studies are needed to address these important questions for surgical decision-making. We used data from the Surveillance, Epidemiology, and End Results (SEER) database and literature review to evaluate the impact of tumor size and fat-invasion types on the prognosis and factors associated with surgical decision-making in patients with T3aN0M0 RCC.

Patient cohorts
We screened 18 registries in the SEER database to identify cases of T3aN0M0 RCC with the kidney parenchyma as the primary site (ICD-O-3 code C64.9) between 2004-20105 after obtaining permission to access research data les. The SEER database covers ~28% of the population in the USA. The characteristics of the SEER population are comparable with those of the general population (https://seer.cancer.gov/). Supplementary Figure 1 shows a owchart with detailed information describing data selection from the SEER database. All tumors included in the present study were ≤15 cm in size. All patients were aged ≥18 years and had undergone PN or RN. Moreover, histology-con rmed clear cell, papillary, and chromophobe tumors were included. Patients were excluded if: (i) they had other types of primary malignancy or multiple RCCs; (ii) their information was missing or did not include the speci c cause of death. RCC with sarcomatoid dedifferentiation was also excluded because it carries a high risk of mortality by RCC, and RN was the best treatment. Also, we excluded patients who were followed for <3 months, and those who died within 30 days.

Outcome and variables for analyses
The primary outcomes of interest were all-and RCC-cause mortality. The SEER Cause of Death Recode was used to identify the cause of death. Patients who died from RCC were identi ed as "RCC-cause mortality," those who died from other causes were designated as "competing events before RCC-cause mortality," whereas any cause of death was considered "all-cause mortality." "Duration of survival" was de ned as the time from the date of diagnosis to the date of death or last contact. We collated a range of demographic variables for our study: year of diagnosis; age at diagnosis; sex; ethnicity (White and Others [Black, American Indian/Alaska Native, Asian Native, and Asian/Paci c Islander]). We also collated a range of data related to tumors: tumor size (cm) and histology cell type for RCC (clear cell, and nonclear cell [papillary and chromophobe]); tumor grade ("well-differentiated" [grade 1], "moderately differentiated" [grade 2], "poorly differentiated" [grade 3], and "undifferentiated" [grade 4]); invasion features (sinus/perisinus invasion and perinephric-fat invasion; in the SEER database, such characteristics were recoded after 2010).

Statistical analyses
Continuous variables are described as the mean (standard deviation [SD]) if they have a normal distribution compared with a Student's t-test or as the median (interquartile range [IQR]) if they did not have a normal distribution compared with a Wilcoxon rank-sum test. Categorical variables are presented as frequencies (%) and were compared using a chisquare test. Overall survival (OS) was compared using the Kaplan-Meier method for survival function along with the log-rank test. Multivariable Cox proportional hazards regression models were conducted. Fine and Gray competing risk proportional hazard regression models were tted to assess mortality caused by RCC and other competing events [11]. To evaluate the choice of surgical method (PN or RN) for T3aN0M0 RCC, we conducted matching of the propensity score using the "nearest-neighbor" method for the likelihood of carrying out PN [12]. Propensity scores were calculated using a multivariable logistic regression model for each patient based on all covariables for T3aN0M0 RCC (year and age at diagnosis, sex, ethnicity, tumor size, tumor grade, and tumor histology). Although tumor grade and tumor histology were histology features obtained postoperatively, they continued to impact the surgical choice in clinical practice for T3aN0M0 RCC. Therefore, they were also included in propensity-score matching. This practice balanced variables and reduced the risk of a selection bias, which may have exerted in uence over the association between the surgical method and outcomes. Patients were matched at a 1:1 ratio based on those propensity scores. A sensitivity analysis accounting for the invasion type of T3aN0M0 was carried out using the total dataset with information on the invasion type of T3aN0M0 after 2010 (because the feature of fat invasion beyond the renal capsule was only included in the SEER database then). All analyses were conducted using the R v.3.5.2 (R Foundation for Statistical Computing, Vienna, Austria; www.r-project.org). All P values are two-sided, and P < 0.05 was considered signi cant.
Tumor size is a predictor for survival from T3aN0M0 RCC Supplementary Outcomes of PN vs. RN based on tumor size Figure 1 presents the results of the Kaplan-Meier survival curve for PN vs. RN. In patients with T3aN0M0 (≤4 cm) RCC, PN was associated with improved OS compared with RN (P unmatched data <0.001; P matched data = 0.006). Table 2 shows that RN in patients with tumor size ≤4 cm was associated with a signi cantly higher risk for all-cause mortality adjusted for covariates (HR unmatched data : 1.65, 95%CI: 1.09-2.51, P = 0.019; HR matched data : 1.77, 95%CI: 1.14-2.74, P = 0.01); however, the type of surgical procedure was not an independent risk predictor of RCC-speci c mortality (HR unmatched data : 1.42, 95%CI: 0.66-3.05, P = 0.370; HR matched data : 1.57, 95%CI: 0.74-3.36, P = 0.24) for T3aN0M0 (≤4 cm) RCC. PN did not provide a signi cant outcome bene t compared with RN in patients with T3aN0M0 RCC of size >4 cm after adjustment for clinicopathological characteristics ( Table 2).

Outcomes of PN vs. RN based on different types of extrarenal fat invasion
The type of extrarenal fat invasion impacted the bene t of surgical decision-making signi cantly ( Figure 2

Discussion
T3a RCC is classi ed only by local extension (involvement of perirenal veins, fat invasion, or invasion of the pelvicalyceal system) regardless of tumor size. The potential role of tumor size and type of invasion for prognostication and treatment decision-making in patients with T3aN0M0 RCC is controversial but deserves increased attention [13,14]. Our study led to two important ndings. First, in patients with T3aN0M0 RCC, tumor size was identi ed as an independent prognostic predictor. Our data suggest that it is reasonable to stratify small and large T3aN0M0 RCC tumors in a manner that is not based solely on extrarenal invasion. Second, patients with small (≤4 cm) T3aN0M0 RCC without the presence of sinus-fat invasion could bene t more from PN than from RN. However, PN did not yield any signi cant bene ts to patients with T3aN0M0 RCC of tumor size >4 cm and besides, PN was not bene cial for RCC with sinus-fat invasion across all tumor sizes. These ndings may have critical implications for the next revision of the TNM staging system. Our ndings may also provide clinicians with important guidelines for prognostication and surgery decision-making (PN vs. RN) based on tumor size and different types of extrarenal-fat invasion in T3aN0M0 RCC (Figure 3).
Although tumor size is not considered a parameter for TNM staging in patients with T3aN0M0 RCC, resection is based mostly on preoperative tumor diameter as well as other factors (i.e., tumor diameter, tumor location, renal-vein invasion, and surgeon experience) in clinical practice. Wide use of PN in T1 RCC has led to an overall increase in the prevalence of T3aN0M0 upstaging from T1 after PN. The prevalence of pathologic upstaging from clinical T1 reported in contemporary studies is 5%-14% [4,9,15,16]. Pathologic upstaging post-PN poses a dilemma to surgeons regarding the choice of surgical approach (clinic observation following PN or conversion to RN), considering that conventional imaging methods are limited in terms of detecting features relating to perirenal-fat invasion.
RN is considered to be standard treatment for small T3aN0M0. However, treatment of small T3aN0M0 with PN has not yielded inferior results and has drawn increased attention [2,[5][6][7][8][9][10][15][16][17][18][19][20][21] (Table 3). It seems clear that the prognosis of T3aN0M0 disease may be determined by the biological characteristics of the tumor rather than just the type of surgical procedure. Hence, T3aN0M0 RCC should not deter surgeons from undertaking PN. Also, treatment of advanced RCC with drugs (e.g., targeted therapy and immunotherapy) has been shown to prolong progression-free survival [22], which further provides the possibility of PN in patients with T3aN0M0 RCC.
Smaller tumors (<4 cm) that have been upstaged to T3aN0M0 should be considered for PN because PN can provide comparable DSS, lower noncancer-speci c mortality, and better OS. In our study, PN was associated with longer OS in patients with T3aN0M0 RCC (≤4 cm) compared with that using RN. RN for small RCCs could increase the risk of chronic kidney diseases and threaten OS. However, this observation was not seen in patients with tumor size >4 cm; in addition, RN increased the risk of mortality by other competing events (sHR: 1.78, 95%CI 1.05-3.00, P = 0.032) for patients with RCCs <4 cm. Lee et al. [8] conducted a retrospective study in which 6.3% (215/3,431) of patients were upstaged from clinical T1 to T3aN0M0. Subgroup analyses of T3aN0M0 RCC in their study showed that the PN group did not show signi cant differences in terms of recurrence-free survival, DSS, or OS compared with those in the RN group. In a study conducted by Jong and colleagues [7], 63 and 142 patients were upstaged to T3aN0M0 from T1a and T1b, respectively. Those authors concluded that PN provided a recurrence-free survival outcome similar to that observed for RN in patients with small T3aN0M0 RCC.
While differentiating T3aN0M0 based on size seems logical for deciding on PN vs. RN, the propensity of invasion into sinus fat and/or the renal vein is important. Invasion into perirenal fat has been shown to not increase RCC-speci c mortality. Conversely, invasion into sinus fat portends a worse prognosis than perirenal-fat invasion even though both are classi ed as T3aN0M0 [23][24][25]. We found that RCC presenting with sinus/perisinus-fat invasion could not bene t from PN regardless of tumor size. Our results show that we must be alert to invasion of extrarenal sinus fat across all tumor sizes, and that RN should be suggested for patients with pT3a RCC of such histologic features because PN did not provide better OS for those patients.
The strengths of the SEER database are its sample size, established quality-assurance program, as well as internal and external validity, but there are some drawbacks to using the SEER database [26]. The data included in our study was obtained from the SEER database and this was a retrospective study. Hence, the retrospective nature of this investigation is an inherent limitation. Second, the SEER database does not detail the features of renal-vein invasion, and the latter portends a worse prognosis than perirenal-fat invasion. Tumors with renal-vein thrombi are, in general, treated with RN instead of PN, so our results might skew risk assessment. There was some missing information for one or more variables that were excluded from our study, which might have led to a selection bias. Therefore, a sensitivity analysis was conducted on the subset of cases diagnosed between 2010 and 2015 (the period in which the T3aN0M0 invasion type was included). Marked differences in results were not observed, which suggested that missing data did not cause a signi cant analytical bias.

Conclusion
For better clinical practice, selected patients with small T3aN0M0 RCCs (≤4 cm) without sinus/perisinus-fat invasion could receive PN, instead of RN only, because this practice may prolong OS. Tumor size is an independent prognostic predictor and should be considered for improved survival strati cation. Surgical decision-making when selecting PN or RN for T3aN0M0 RCC should be based on tumor size and presentation of fat-invasion types.    RCC, renal cell carcinoma; PN, partial nephrectomy; RN, radical nephrectomy; OS, overall survival; DSS, diseasespeci c survival; CSM, cancer-speci c mortality; RSF, recurrence-speci c survival; PSM, positive surgical margin; IQR = interquartile range. Tumor size and follow-up time are presented as the median (IQR). The study design was a retrospective single-center study; tumors without lymph-node and distant metastasis (N0M0), and Kaplan-Meier and multivariate analysis were used for survival prediction. †, data from two-institutions; ¶, data from multiple institutions; *, data are the mean or mean (standard deviation); †, Only the detailed information of all cohorts of PN and RN.