The effectiveness of partial versus radical nephrectomy for pT3aN0M0 renal cell carcinoma: A propensity score analysis.

BACKGROUND
The survival benefit of partial nephrectomy (PN) in pT3a RCC patients is controversial. Here we aimed to explore the potential benefit of PN for pT3aN0M0 renal cell carcinoma (RCC).


MATERIAL AND METHODS
Data of patients with pT3aN0M0 RCC who were diagnosed between 2010 and 2012 in the National Cancer Institute Surveillance, Epidemiology, and End Results (SEER) database were retrospectively collected. Overall survival (OS) and cancer specific survival (CSS) were compared using a Cox proportional hazards model between PN and radical nephrectomy (RN) in pT3aN0M0 RCC. Propensity score (-adjusted, -stratified, -weighted, and -matched) analyses were performed to control for imbalances in individual risk factors.


RESULTS
A total of 1277 patients with pT3aN0M0 RCC were identified, of whom 200 patients were treated with PN and 1077 patients were RN. PN showed favorable OS and CSS in 0-4 cm pT3aN0M0 RCC (P < 0.05), and similar OS and CSS in 4-7 cm pT3aN0M0 RCC, compared with RN using un-adjusted analyses. The Propensity score analyses further demonstrated the survival benefit of PN compared with the RN in 0-4 cm pT3aN0M0 RCC (P < 0.05).


CONCLUSIONS
In this retrospective study, PN was associated with improved survival compared with RN in 0-4 cm pT3aN0M0 RCC. Moreover, survival was comparable between PN and RN in 4-7 cm pT3aN0M0 RCC. These data provided evidence that PN could be an alternative choice for T3aN0M0 RCC less than 7 cm. Particularly, patients with 0-4 cm pT3aN0M0 RCC might benefit from PN.


Conclusions
In this observational study, PN was associated with improved survival compared with RN in 0-4cm pT3aN0M0 RCC. Moreover, survival was comparable between PN and RN in 4-7cm pT3aN0M0 RCC. These data provided evidence that PN is an alternative choice for T3aN0M0 RCC less than 7 cm. Patients with 0-4cm pT3aN0M0 RCC may bene t from PN.

Background
Renal cell carcinoma (RCC) represents 2.2% estimated new cases and 1.8% estimated deaths of all cancers worldwide [1]. Surgery is the most important curative intention in patients with localized RCC, including partial nephrectomy (PN) and radical nephrectomy (RN). Though PN has comparable overall survival (OS) with RN in clinical T1 stage RCC, PN demonstrated better preserved kidney function, thereby potentially lowering the risk of development of cardiovascular disorders. According to the European Association of Urology Guidelines, PN is strongly recommended for patients with cT1 RCC (< 7 cm) [2].
Majority of cT1 RCC has consistent pathological T1 stage after surgery, while there is a rate of 3.2%-31% cT1 RCC upstaged to pT3a in the PN treatment [3]. Moreover, pT3a RCC upstaged from cT1 RCC was signi cantly associated with poor recurrence-free survival, overall survival, and cancer-speci c survival [3]. According to the Eighth Edition of the Tumor-Node-Metastasis Staging Classi cation System, pT3a is de ned based on anatomical tumor expansion including either vein, Renal pelvis or fatty in ltration, regardless of tumor size [4]. Several studies highlighted the importance of upstaging and underlying tumor characteristics such as tumor grade, size and complexity, and invasion site as predictors of recurrence and survival for pT3aRCC, while the surgery type is controversial [5][6][7].
This study aimed to investigate the survival bene t of PN compared with RN in pT3aN0M0 RCC.

Patients and Study Design
The data of patients with RCC from the Surveillance, Epidemiology and End Results (SEER) database were obtained through SEER*Stat 8.3.6 software (National Cancer Institute, Bethesda, MD, USA) (Accession number: 14558-Nov2018). A total of 1277 RCC patients who underwent partial or radical nephrectomy from 2010 to 2012 were enrolled by meeting the following including criteria: ( ) histologically con rmed renal cell carcinoma; ( ) pT3a stage; ( ) absence of lymph node and distant metastasis; (IV) age ≥ 18; (V) treated with PN or RN; ( ) single primary RCC.
The renal cancer coded as Primary Site (C64.9) and ICCC site recode ICD-O-3 (VI (b) Renal carcinomas). Partial surgery coded as surgery of primary site (30) and Radical surgery coded as surgery of primary site (50). The invasion site of perirenal fat invasion (PFI) coded as CS site-speci c factor 1 (10) with CS extension (450 or 460), sinus fat invasion (SFI) coded as CS site-speci c factor 1 (20) with CS extension (450 or 460), PFI and SFI coded as CS site-speci c factor 1 (30) with CS extension (450 or 460), PFI and renal vein invasion (RVI) coded as CS site-speci c factor 1 (10) with CS extension (601 or 605), SFI and RVI coded as CS site-speci c factor 1 (20) with CS extension (601 or 605), and all the three factors coded as CS sitespeci c factor 1 (30) with CS extension (601 or 605). For each patient, collected data includes demographic (age, gender. race, year of diagnosis) and clinicopathologic information (tumor size, grade, laterality, pathology, invasion site for T3a, the 7th TNM classi cation, the surgical type, follow-up data and so on). The clinical data collected from the SEER database, which is a public research resource that does not require patient consent and ethical consent.

Outcomes
In this study, the outcomes included: OS and cancer-speci c survival (CSS). They were coded by SEER.

Statistical Analysis
Categorical variables were presented as frequency and percentage, which were compared by Chi-squared test. Univariable logistic regression models were used to estimate crude odds ratios and 95% CIs to evaluate the association between each variable and surgical type (partial or radical nephrectomy). A Cox proportional hazards model was used to compare OS and CSS of patients treated with PR and NR for all patients, patients with tumor size ≤ 4 cm, and patients with tumor size 4-7cm, respectively. The hazard ratios (HR) were calculated using four methods: stratifying on the propensity score in which patients were classi ed into quintiles by their propensity scores, 1: 1 ratio matching patients in the PR and NR groups by propensity score with a caliper of 0.1, using inverse probability weighting by the inverse of the propensity score of the treatment received, and adjusting for the propensity score by including it as a continuous covariate in the Cox model. We assessed heterogeneity of treatment effects with tests of interaction and subgroup analyses of age, gender, invasion site, tumor size, grade, laterality pathology. Propensity scores were estimated with a multivariable logistic regression model in which treatment assignment was regressed on age, gender, invasion site, tumor size, grade, and laterality pathology for all patients, patients with tumor size ≤ 4 cm, and patients with tumor size 4-7cm, respectively. All hypothesis tests were two-tailed with a signi cance level of 0.05. Statistical analyses were conducted using R version 3.0.2 and SPSS software version 24.0 (IBM, Armonk, NY, USA).

Results
Factors associated with the PN and RN Baseline patient characteristics are listed in Table 1. Patients who received PN were more likely to have small tumor size, well/moderately differentiation and PFI invasion, and less likely to have poor differentiation, more than one site invasion and clear cell adenocarcinoma type (P < 0.05 for all comparisons). Based on tumor size, pT3aN0M0 RCC were divided into four subgroups: 0-4cm, 4-7cm, 7-10cm and > 10 cm.
OS and CSS comparing PN and RN in all subgroups were further analyzed. The CSS and OS of PN are better than RN in 0-4cm subgroup, and they are similar with RN in the 4-7cm, 7-10cm and > 10 cm pT3aN0M0 RCC subgroups. (Fig. 1).
Propensity score survival analysis The subgroup of 0-7cm pT3aN0M0 RCC was further analyzed (     While for the 4-7cm pT3aN0M0 RCC group, the analyses adjusted, strati ed, inverse-probability-weighted and -matched by propensity score quintile showed that PN has no bene t in OS and CSS, compared with RN (Table 3).

Discussion:
The role of PN in the treatment of cT1 RCC has been well studied. While Chen et al. showed that the rate of cT1 upstaged to pT3a ranged from 3.2-31%, median 5.5% [3]. Moreover, upstaged cT1/pT3a RCC was associated with poor recurrence-free survival, overall survival, and cancer-speci c survival in the most of the reports [3]. It has been showed that several factors would affect the prognostic of pT3a RCC, such as tumor size, invasion site, tumor grade and pathology type. Tumor size over 7 cm is an important poor factor for predicting the outcome of patients with pT3a RCC with fat invasion [8,9]. Furthermore, there is a trend of worse outcomes with tumor size increases from 0-4cm to 4-7cm and over 7 cm in pT3a RCC [10]. We previously showed that the PFI and SFI has similar prognosis, and single fat invasion site invasion has better survival than multiple sites invasion in pT3aN0M0 RCC [11]. There are also some studies showed that tumor grade and pathology type could affect the survival of pT3a RCC [12,13]. These studies suggested tumor size, invasion site, tumor grade and pathology type were the prognostic factors of pT3a RCC.  PN vs 91 RN), in which the positive margin rate was approximately 15% (7/49), and more than half (4/7) of positive margin patients recurred in PN [18]. Although the survival bene t is inconsistent, most of the studies showed that PN patients had a higher postoperative eGFR than RN patients [19]. Actually, there are some shortcomings in previous studies, such as small sample size, short follow-up time, lacking invasion site and tumor size information, and including Nx/N1 patients. Moreover, single or inadequate analysis model is also an important factor.
To our knowledge, this observational study contains the largest cohort of pT3aN0M0 RCC to analyze the survival of PN (200 cases) and RN (1077 cases). And factors including age, gender, tumor size invasion site, grade, and pathology type, which may affect survival, were all analyzed. As it is rare to perform PN in pT3a RCC over 7 cm, thus we focus on pT3aN0M0 RCC less than 7 cm, comparing the survival differences between PN and RN. Unadjusted data showed that the characteristics of pT3aN0M0 patients are signi cantly different in tumor size, invasion site, tumor grade, pathology type and surgery type, therefore it is necessary to balance the bias by using analytic models. Previous studies did not balance the bias or just used the match model, which might result in unreliable conclusions. Propensity score analyses were more reliable to control imbalances in individual risk factors [20].
There are inconsistent conclusions of PN and RN for T3a RCC from previous studies, and majority of these studies showed similar survival outcomes between PN and RN in T3aN0M0 RCC. Our data showed that PN is associated with better survival compared to RN in 0-4cm pT3aN0M0 RCC. It is clear that the OS/CSS of PN is comparable to RN in cT1 RCC from the prospective and retrospective clinical studies, and PN decreases chronic kidney disease [21][22][23]. The survival bene t of PN in 0-4cm pT3aN0M0 RCC may be explained by Hamilton, Z.A., which showed that most recurrences were distant with a recurrence rate of 92.5% (99/107), and loco-regional recurrence was only 11.2% (12/107) in upstaged pT3a RCC, which was even lower than loco-regional recurrence of 17.8% (30/169) in non-upstaged cT1-2 RCC [10]. Furthermore, Patel et al. reported that in pathologically upstaged pT3a RCC, PN did not adversely affect risk of recurrence and provided functional bene t [24].Although, Shah et al. reported that cT1/pT3a upstaged patients (49 PN/91 RN), the positive margin rate was approximately 15% (7/49) in PN, and more than half (4/7) of positive margin patients recurred [18]. The positive margin of PN signi cantly increased in cT1/pT3a RCC, compared with cT1/pT1 [25]. While, positive margin was not associated with local recurrence of pT3a RCC [10]. As distant recurrence is the major risk of locally advanced T3a RCC patients, therefore, local positive margin may not be a key prognostic factor in this cohort patients. Single, rather than multiple sites invasion, is the major invasion characteristic of 0-4cm pT3aN0M0 RCC, which is possible to be completely removed by PN surgery. Thus, the effect of the local disease control may be similar in 0-4cm pT3aN0M0 RCC patients between PN and RN. In this study, the results have further been con rmed by propensity score (-adjusted, -strati ed,weighted, and -matched) analyses, which may attenuate the possibility of bias. The cohort of 4-7cm pT3aN0M0 RCC, PN has similar survival outcomes compared with RN, which indicated that tumor size is still the key prognostic factor for pT3aN0M0 RCC.
PN is a technically demanding surgery, especially for patients with tumors > 7 cm, Fuhrman grade III-IV, and pT3a. RN is the rst choice for T3a RCC. While PN can preserve additional renal function and lead less cardiovascular disorders, which may let pT3aN0M0 RCC patients have better tolerance to adjuvant therapy with targeted drug, and it is bene c for longer disease-free survival of advanced RCC patients [26]. Therefore, PN is not a contraindication for T3a RCC patients. Especially for T3a RCC patients with smaller size 0-4cm, whom even can bene t from PN.
Although this is the largest cohort used to assess the survival bene t of PN in pT3aN0M0 RCC, it has several limitations. Limitations of the present study include the retrospective nature of the analysis. Selection bias is another important limitation which is intrinsic to the retrospective design of the study. Though we analyzed several prognostic factors of pT3aN0M0 RCC including tumor size, grade, pathology type, and invasion site for the rst time, some prognostic factors including positive margin, perioperative complication, renal function, metastasis and adjuvant therapy were not analyzed here. Propensity score-based approaches are su cient for minimizing the impact of observed confounders, such methodology does not address unobserved confounders (ie, unmeasured patient selection factors for PN/RN that are also associated with survival such as hypertension and diabetes). Furthermore, there is no central pathology review to recon rm pathological characteristics, as this was a national data set study, Lastly, follow-up was relatively short and endpoint events only occurred in a small part of patients.

Conclusion
The present study is the largest evaluation of PN in pT3aN0M0 RCC. Our data indicate that PN is associated with favorable survival in 0-4cm pT3aN0M0 RCC subgroup, even when controlling for known prognostic imbalances between PN and RN. And the survival is similar between PN and RN in 4-7cm pT3aN0M0 RCC subgroup. Given the fact that PN is the preferred choice for cT1 RCC, while PN to RN still showed comparable prognosis of cT1/pT3a RCC, therefore it is not a contraindication to perform PN in selective cT3a RCC patients with tumors less than 7 cm. Availability of data and materials All data generated or analysed during this study are included in this published article. More details are available from the corresponding author upon request.