Comparison of prognosis between patients undergoing radical nephrectomy versus partial nephrectomy for renal cell carcinoma ≤7 cm T3aN0/xM0: Survival benefit is biased toward partial nephrectomy

Abstract Background There is limited and controversial evidence on the prognosis of partial nephrectomy (PN) versus radical nephrectomy (RN) in patients with T3aN0/xM0 renal cell carcinoma (RCC) upstaged from clinical T1 RCC. In this study, we aimed to assess the prognosis difference following PN versus RN in patients with ≤7 cm T3aN0/xM0 RCC. Methods From the Surveillance, Epidemiology, and End Results database, a total of 3196 patients receiving treatment of PN/RN for ≤7 cm T3aN0/xM0 RCC with only extrarenal fat extension in 2010–2017 were identified. An inverse probability of treatment weighting (IPTW)‐adjusted cause‐specific Cox model with hazard ratio (HR) and 95% confidence interval (CI) was used for overall survival (OS) and cancer‐specific survival (CSS) analyses. Sensitivity analysis was based on the propensity score matching of PN and RN groups and from the dataset of 2010–2013. Results A total of 872 patients underwent PN, compared with 2324 undergoing RN. After IPTW adjustment, there was no significant difference in preoperative baseline characteristics between the PN and RN cohorts. Patients who underwent RN had worse OS (HRIPTW‐adjusted, 1.46; 95% CI, 1.16–1.84; p = 0.001) and comparable CSS (HRIPTW‐adjusted, 1.03; 95% CI, 0.64–1.66; p = 0.890) than those receiving PN in all cohorts and subgroups with T3a RCC of ≤4 cm and perinephric fat extension. Further, in patients with 4–7 cm T3a RCC with perinephric‐fat invasion and all sizes of T3a RCC with sinus/perisinus fat extension, PN led to comparable OS and CSS. Sensitivity analyses validated these results. Conclusion PN provides comparable CSS and OS or even better OS than RN for patients with RCC ≤7 cm T3aN0/xM0. Although our study has some limitations, our results indicated that PN might oncologically safe for clinical T1 RCC, even confirmed a pathologically T3a upstaging post‐PN.


| INTRODUCTION
Partial nephrectomy (PN) is the preferred treatment for T1 renal cell carcinoma (RCC). 1 Although three-dimensional reconstruction and virtual imaging techniques were reported to allow to facilitate the surgical planning preoperatively in complex renal masses evaluation, 2,3 there is still limited ability to accurately predict T3a stage disease preoperatively based on the current imaging modalities, it is not uncommon for T1 RCC to be upgraded to pathological stage T3a following PN, with reported incidence rates of 1.9%-14%. [4][5][6][7][8][9] This fact has led to a treatment strategy dilemma in clinical practice regarding whether it is preferable to keep PN unchanged, or change to radical nephrectomy (RN), since there is limited and controversial evidence of the surgical benefit of PN compared with RN for T3a RCC. 5,8,[10][11][12][13][14] Currently, there have been raised concerns about the oncological safety of PN for patients with T3a RCC. 5,8 In clinical practice, a transformation of surgery to RN from PN may not result in an excellent prognosis for patients with pathological T3a RCC upstaged from clinical T1 disease. Some prior studies have suggested that PN can yield a satisfactory prognosis for selected T3a RCC, [10][11][12] patients who underwent PN in comparison to RN experienced similar or even better oncological outcomes in pathological T3a RCC upstaged from a clinical T1 disease. 8 While some other studies have observed that RN is associated with a lower risk of recurrence than PN when conducted for clinical T1 tumors pathologically upgraded to T3a. 14 Notably, small sample sizes and with a short follow-up and had a selective bias resulting in confounding were the primary limitation for these studies. In addition, the influence of tumor diameter or extrarenal fat extension pattern on survival benefit from PN versus RN should be considered, since tumor diameter or extrarenal fat extension pattern have been identified as significant risk factors for survival outcomes in patients with T3a RCC. [15][16][17][18][19][20][21][22] In our present study based on a large sample size and a long follow-up time, we want to furtherly investigate whether PN compared with RN can provide a comparable prognosis for patients with T3a RCC and to investigate if T3a tumor diameter and extrarenal fat extension pattern could impact the survival outcomes between T3a RCC patients treated with PN and RN. Here, to address these important questions, we used the Surveillance, Epidemiology, and End Results (SEER) database to evaluate differences in prognosis between patients with T3a RCC undergoing PN versus RN, and to investigate the influence of tumor diameter and extrarenal fat extension type on the survival benefit from PN versus RN for T3a RCC patients.

| Database and patient selection
The SEER-18 registries database was screened to identify cases of pathological diagnosis of primary T3aN0/ xM0 RCC between 2010 and 2017; case lists were identified from SEER using *Stat software (version 8.3.9). This study was approved by the Ethics Review Board of Tongji Hospital of Huazhong University of Science and Technology based on the Declaration of Helsinki.
The data selected from the SEER database are presented in Figure 1. Briefly, all RCC cases included in the study were pathologically confirmed T3aN0/xM0 RCC of ≤7 cm diameter, which is equivalent to clinical T1 RCC. All patients included for analysis in our study were aged 18-85 years and underwent PN or RN surgery. Patients were excluded if: (1) no information on extrarenal fat extension pattern was recorded; (2) the specific cause of death and follow-up time were not recorded, or they were followed for <1 month after surgery or died within 30 days; or (3) the same patient identification numbers had multiple records in the case list.

| Survival outcome
In our present study, the two primary survival outcomes of interest were overall survival (OS) and cancer-specific survival (CSS). The Cause of Death Recode in the SEER was used to identify the cause of death. CSS was calculated according to "RCC-cause mortality," which was defined as patients who died from RCC, while patients who died from other causes were considered as "competing events" for CSS, and considered in the competing risk analysis. "All-cause mortality" included patients who died from any cause of death, and were included for OS analysis. "Time of survival" was calculated from the diagnosis date to the date of end event occurrence or censor.

| Statistical analyses
The continuous variables are described as mean (standard deviation) and were compared using a t-test. If the data meets the normal distribution; while it is expressed as median (interquartile range [IQR]) and were compared with the Kruskal-Wallis test if it did not meet the normal distribution. A Chi-square test was used for categorical variables comparison and is presented as frequencies (%).
Since this was not a randomized control study, treatment-selection bias could have impacted the analysis of the difference in prognosis between patients undergoing PN versus RN. The inverse probability of treatment weighting (IPTW) method was used to control for such bias and to balance preoperative confounding factors between the PN and RN groups. First, a F I G U R E 1 Flow chart showing the data screening and analysis workflow. FI, perinephric-fat invasion; PN, partial nephrectomy; RN, radical nephrectomy; SI, sinus/perisinus invasion propensity score (PS) was calculated for each individual using multivariate logistic regression analysis based on the probability of surgical treatment by PN compared with RN; the preoperative covariables included in the model were: year, age, gender, race, median household income, prior history of other system cancer, tumor diameter, and tumor laterality. Although Fuhrman grade, RCC histology, and sarcomatoid dedifferentiation were identified postoperatively, they may still have impacted the surgical intervention in the clinical practice; therefore, these covariables were also included in the analysis model. Then, weights were calculated as 1/PS for PN individuals and 1/(1 -PS) for RN individuals. Finally, the IPTW approach was used to balance the difference in perioperative confounding factors between the PN and RN cohorts.
The reverse Kaplan-Meier method was used to evaluate the median follow-up time. OS and CSS of the PN and RN cohorts were compared based on the IPTW-adjusted Kaplan-Meier method with the log-rank test. For the IPTW population, univariate and multivariable Cox proportional hazards regression models and cause-specific Cox regression models were constructed to evaluate risk factors for OS and CSS, respectively, and hazard ratio (HR) with 95% confidence interval (CI) values calculated to compare the influence of PN versus RN on OS and CSS. For the non-IPTW population, Fine and Gray competing risk regression models with a corresponding subdistribution hazard ratio were fitted to assess the risk of "RCC-specific mortality" and "other causes of mortality". Since tumor diameter and extrarenal fat extension pattern were significant predictors of surgery intervention, all analyses included all cohorts, with four subgroups: ≤4 cm T3a of perinephric fat extension, 4-7 cm T3a of perinephric fat extension, ≤4 cm T3a of sinus/perisinus extension, and 4-7 cm T3a of sinus/perisinus extension.
Furthermore, to maximize the accuracy of our hypothesis and ensure consistent results, we refitted the Cox and Fine and Gray models in two sensitivity analyses: (1) a model including unweighted data from 2010 to 2013, in which most patients were followed for >5 years; and (2) 1:1 propensity-score matched Cox proportional hazards model and Fine and Gray competing risk regression models to evaluate the influence of PN versus RN on OS and CSS. For the propensity-score matched analysis, year at diagnosis, age at diagnosis, gender, race, median household income, prior history of other system cancer, tumor diameter and tumor laterality, pathological T3a invasion type, Fuhrman grade, RCC histology, and sarcomatoid dedifferentiation were matched between PN and RN group. In addition, another 1:1 propensity-score matching analysis between PN and RN did not include the covariables of postoperative (pathological T3a invasion type, Fuhrman grade, RCC histology, and sarcomatoid dedifferentiation).

| Baseline characteristics and treatment comparison
In our present study, a small proportion of patients with T3a RCCs (27.3%) underwent PN, and the proportion undergoing PN did not increase with a year of diagnosis ( (detailed IPTW-adjusted data analyses are shown in Tables S1 and S2).

| Follow-up and survival outcomes
All patients had information of follow-up data. In our present study, patients dead within 1 month or followed <1 month were excluded ( Figure 1) Figure 2). Furthermore, we analyzed data from 2010 to 2013, with a median 6.75-year follow-up, to compare OS and CSS between patients who underwent PN and RN, and our findings validated the above results; 5-year OS rates were 83.6% and 74.2% for patients who received surgical treatment of PN and RN, respectively (Table 2). After IPTW adjustment, CSS was comparable in patients undergoing PN and RN in all cohorts and the four subgroups ( Figure 2).

| Subgroup analysis
Since tumor diameter and extrarenal fat extension are two important risk factors influencing survival outcomes of patients with RCC, 15,16,21 we further compared the prognosis of patients who received PN and RN in the four subgroups (  Table 3). Sensitivity analyses generated similar results, validating these findings ( Table 2). In addition, similar results were observed after propensity-score matching analysis which did not include the covariables of postoperative (data not shown).

| DISCUSSION
Early detection of small RCC (particularly T1a) tumors favors the adoption of treatment by PN. 1 Clinically, T1 RCC tumors can be identified as more aggressive T3a masses after final pathology analysis post-PN, with prior publications reporting a 1.9%-14% incidence rate. [7][8][9] In large RCC tumors, extrarenal fat extension is usually grossly visible on computed tomography or magnetic resonance imaging, whereas microscopic examination is generally required for small tumors. 23 Thus, the vast majority of pathological T3a RCC tumors, upstaged from the clinical  T1 stage, need a pathological diagnosis and cannot be detected by preoperative imaging. Beksac et al. 9 observed that larger tumor diameter and older age were two significant risk factors associated with clinical T1 RCC pathological upstaging to T3a RCC post-PN based on a high-volume multicenter cohort and national registry data. A recent meta-analysis by Veccia et al. 5 summarized the predictors of risk for pathological T3a upstaging from clinical T1 RCC and found that the characteristics, older age, larger tumor diameter, and higher RCC complexity, increased the risk of postoperative upstaging. Further, clear cell histology is also associated with a higher risk of histology upstage. 4 Another study indicated that RCC diameter, R.E.N.A.L. score, and systemic inflammatory response markers (such as lymphocyte to monocyte ratio) can be used to predict postoperative T3a upstaging 7 ; however, more accurate and better characterization of RCC is needed to predict preoperative T3a RCC upstaging.
Indeed, the dilemma generated by upstaging to a more aggressive pathological T3a tumor postoperatively could jeopardize patient survival and lead to challenges for treatment decisions after primary PN. 5 In our present study, compared to RN, PN can result in better, or at least a comparable, OS and CSS for T3a RCC. These results are consistent with those of a prior meta-analysis. 8 Besides, the majority of studies show that PN of small T3aN0M0 masses does not yield an inferior prognosis 4,10-12,14,24-32 (partial results presented in Table 3). Although the findings of our present study may not be useful for preoperative "surgical planning", our data support that it is oncologically safe to perform PN for small renal masses, even if the tumor is subsequently pathologically determined to be T3a, rather than T1.
Prior results from the SEER database showed that cardiovascular diseases were the main cause of non-RCC special mortality, especially for patients with localized RCCs after 5 years of treatment. 33 PN contributed to the prevention of the risk of cardiovascular diseases, which is associated with improved OS, 34 preservation of glomerular filtration rate related to PN might be considered as a potential explanation for the observation. 35,36 However, such effects might be based on the tumor size of RCC, the mortality by cardiovascular after RN was increased for RCC <2 cm, besides, 37 for small T1N0M0 RCC more than 4 cm, there was no significant difference in cardiovascular diseases incidence for RN versus PN. 33 Compared with large-diameter T3a RCC after PN, the small diameter T3a RCC obtains more normal renal parenchymal retention and has a more prominent role in the protection of renal function, and these patients benefit more. For complex small T3a RCCs, for example, T3 endophytic might require a longer operative time and deep suture, which  Univariate and multivariate Fine and Gray proportional risk regression analysis to calculate the (S)HR (95% CI) was used for RCC-special survival analysis in unweighted partial nephrectomy versus radical nephrectomy data and two sensitivity analyses; univariate and multivariate cause-specific Cox regression analysis to calculate the HR (95% CI) was used for RCC-special survival analysis in weighted partial nephrectomy versus radical nephrectomy data. b Adjust the covariables: age, sex, race, median household income, laterality, history of prior cancer, histology type, sarcomatoid dedifferentiation, and Fuhrman grade, regional lymph nodes removed. might impact kidney function. But because of the application of robot-assisted PN, the preoperative evaluation of imaging three-dimensional imaging technology, and the increased experience of the surgeon, the warm ischemia time is shortened compared with the past, which may not affect the postoperative renal function. 3,31,38,39 In complex RCC, to achieve trifecta after PN, surgical technique plays an important role. 40 Previous studies have reported that surgical technique is an independent predictor of trifecta outcome. The compliance rate of Trifecta outcome in open PN, laparoscopic-assisted PN, and robotassisted PN was 49%, 50.6%, and 69.9% (p = 0.003). A recent meta-analysis concluded that for complex RCC, compared to laparoscopic PN, robot-assisted PN can achieve a lower conversion rate of open surgery, shorter warm ischemia time, and better renal function protection than laparoscopic PN. 39 It is also necessary to pay attention to the influence of the methodological limitations of observational research on the results. For T3a RCC, it was previously reported that robot-assisted PN technology can achieve negative margins in T3a RCC patients with sinus fat extension, and achieve good tumor control in a short follow-up time. 31 However, a longer follow-up is needed, so as to better evaluate the effect of robot-assisted PN on the long-term prognosis of pT3a tumors. In addition, the study did not compare with other surgical methods, so it is difficult to describe the superiority of this technique. In addition, this study also emphasizes that the surgeon's experience and the complexity of RCC play an important role in predicting the outcome of surgery.
Prior studies indicate that patients with RCC tumors with a larger diameter, or classified as T3a with sinus/ perisinus extension, experience worse prognosis than those with small or T3a with perinephric fat extension tumors 15,16,21,41 ; therefore, analyses of survival outcomes should consider tumor diameter and extension pattern.
Here we report an important finding, that PN improved OS relative to RN for RCC patients with tumors ≤4 cm   10 Nevertheless, Chevinsky et al. 13 reported that pathological stage T3a significantly shortened recurrence-free survival; however, they included T3a pathologically upstaged from larger size T2 and T3a RCC with renal vein extension in their study. Notably, a recent meta-analysis proved that PN and RN are associated with similar recurrence-free survival. 8 The survival outcomes of patients with T3aN0/xM0 RCC may be determined mostly by inherent cancer biological features, rather than whether surgical PN or RN is applied. Hence, upstaging to T3aN0M0 RCC could not prevent surgeons from conducting PN.
In addition, it should be noted that positive surgical margins can increase the risk of local recurrence. 42 For T3a RCC, the incidence of positive surgical margins for PN is higher, with an average of about 18%. 31 Although the impact of positive surgical margins after PN on the prognosis is a matter of debate, some prior studies observed positive surgical margins were significantly associated with aggressive disease characteristics and low surgeon experience, 38 but not significantly impact the short-term survival functions. 38,43 Our study had some unavoidable limitations. Although the SEER database has the advantages of its large sample size, it also has some drawbacks, including the inherent limitation of its retrospective nature. The inability to adequately control the selection bias between the PN and RN cohorts is the main limitation of the study. Although the method of IPTW and propensity score matching was used in the current study, the SEER analysis does lack sufficient confounding information (such as patient performance status, body mass index, comorbidities, renal function, renal vein invasion, time from diagnosis to surgery, surgeon skill, and institutional case volume) to adequately control the selection factors that affect survival outcomes. Our study identified patients with pathologically confirmed ≤7 cm T3aN0/ xM0 RCC (assumed that those patients who received PN were clinical T1 but upgraded to T3aN0/xM0 RCC after pathological examination). However, we found that because PN is the preferred treatment for T1 RCC, those who receive RN may have a higher stage than T1. Therefore, those who receive RN are more likely to be at an advanced stage than those who receive PN. This imbalance between the two stages cannot be solved by statistical methods such as IPTW and propensity score matching. Because there is no higher level of evidence, a more rigorously designed randomized controlled study to prove the prognostic difference between the two surgical methods PN and RN in T3a RCC, our study is an addition to the previous literature, even though our study has limitations.

| CONCLUSIONS
Our present study shows that, relative to RN, PN provided comparable CSS, and even improved OS in patients with ≤7 cm T3aN0/xM0 RCC of only extrarenal fat extension. Our study indicated that the treatment of PN is a preferred treatment for clinical T1 renal mass, even where there is pathological upstaging to T3a RCC. The pathological characteristic of extrarenal fat extension pattern and tumors size does not shorten the survival benefit of PN. Nevertheless, there is no high-level evidence to validate our findings or prior similar studies; therefore, these results should be interpreted with caution, and further validation in multicenter randomized-design studies with large sample sizes and longer follow-up is warranted.