Matched Comparison Between Partial Nephrectomy and Radical Nephrectomy for T2 N0 M0 Tumors,a Study Based on the National Cancer Database

Introduction

The management of small renal masses entails two important principles: oncologic clearance and nephron preservation. Nephron preservation improves postoperative renal function, which is associated with reduced renal and cardiovascular complications and better overall survival (OS). ^1,2 Being able to preserve nephrons without compromising oncologic outcomes, partial nephrectomy (PN) has surpassed radical nephrectomy (RN) as the standard surgical option for T1 tumors whenever technically feasible. ^3,4

The role of PN for T2 tumors remains controversial. Studies based on the Surveillance, Epidemiology, and End Results (SEER) database confirmed similar cancer control between PN and RN for T2 tumors, but did not compare postoperative renal function or OS. ^5,6 The few studies that compared these outcomes showed significantly better postoperative renal function and similar OS after PN, but their sample sizes were relatively small. ^{7 –9} The PN and RN groups in these studies were also very different in key covariates like tumor size and histology. ^{7 –9} Theoretically, if PN maintains its ability in preserving renal function for T2 tumors, we should also see better OS among these patients when most other factors remain constant since the 7 cm cutoff between T1 and T2 disease is an arbitrary one. A large database with sufficient power to detect the difference in OS between PN and RN would help in this situation, and comparison should only be made after statistically matching the PN and RN groups to eliminate the confounding effects of other key covariates.

Therefore, we used the National Cancer Database (NCDB) with statistical matching to determine if PN maintains its advantage in OS for T2 tumors. We also aimed to identify significant factors for poor survival and to compare immediate postoperative outcomes between PN and RN.

Patients and Methods

The institutional review board approved this study (protocol number: 1611207381, approval date: December 2, 2016).

We used the NCDB participant user file (PUF) spanning from 2004 to 2014. To ensure that all cases for analysis had a minimum of 5 years of follow-up after PN, we only included cases between 2004 and 2009. Patients with renal-cell carcinoma (RCC) were identified by the code C649 for data item “Primary Site”, based on the International Classification of Diseases for Oncology, 3rd Edition. ¹⁰ Among them, we selected those with clinical T2 N0 M0 disease according to data items “AJCC Clinical T”, “AJCC Clinical N,” and “AJCC Clinical M,” based on the American Joint Committee on Cancer Stage Manual, 7th Edition. ¹¹ Cases were screened to identify those treated with PN or RN, according to data item “Surgical Procedure of the Primary Site at any CoC Facility” with codes 30 and 50 representing PN and RN, respectively, and were classified into two groups. Duration of follow-up was defined by data item “Last Contact or Death, Months from Dx” as the time from initial diagnosis till last contact or death of the patient, and survival status at the end of follow-up was defined by data item “PUF Vital Status”. Patients with missing data on survival status were excluded.

Primary endpoint of our study was OS between the PN and RN groups. Secondary endpoints included surgical margin status, postoperative length of stay, and 30-day unplanned readmission. Covariates included age, gender, race, Charlson-Deyo comorbidity score (CCI), tumor size, histology, and grade. Figure 1 summarizes the process of patient selection before statistical matching.

OPEN IN VIEWER

FIG. 1.

Diagram showing the process of patient selection for statistical matching. AJCC = American Joint Committee on Cancer; ICD-O-3 = International Classification of Diseases for Oncology, 3rd Edition; NCDB = National Cancer Database; PN = partial nephrectomy; PUF = participant user file; RCC = renal-cell carcinoma; RN = radical nephrectomy.

We recoded ratio covariates into clinically meaningful categories for statistical analysis. We performed propensity score matching with the fuzz factor set at 0, giving priority to exact matches and without replacement in sampling, to produce a PN and an RN group matching for age, CCI, tumor size, histology, and grade. Covariates were compared between groups by χ² test. Cox multiple regression was then used to estimate the hazard ratios (HR) of the covariates for all-cause mortality. Log-rank, Breslow, and Tarone-Ware tests were used to compare OS between matched groups.

We also compared immediate postoperative outcomes between the matched groups by χ² test and independent-samples t-test for categorical and ratio endpoints, respectively. All statistical analyses were done using SPSS version 24.0 (IBM Corp., Armonk, NY). A p-value of <0.05 defined statistical significance throughout the study.

Results

Altogether, 9254 patients with T2 N0 M0 RCC between 2004 and 2009 were treated with PN or RN. Among them, 535 underwent PN and 8719 underwent RN. One patient from the RN group did not have survival data and was excluded. Statistical matching on the remaining cases produced a PN and an RN group similar in age, CCI, tumor size, histology, and grade, with 527 patients in each group. Table 1 illustrates the descriptive statistics before and after matching.

After matching, there were 137 and 226 recorded all-cause mortality in the PN and RN group, respectively, over a median follow-up duration of 49.2 months. The median OS was 128.1 and 48.4 months for the PN and RN group, respectively. Cox multiple regression showed RN to be associated with a higher risk of all-cause mortality than PN (HR: 5.289; p < 0.001), after adjusting for all available covariates.

There was a trend showing increasing risks of all-cause mortality with increasing age when compared to the reference category of age below 50 years, with HR 2.071 among patients 50 to 59 years going up to 8.531 among patients 80 years and older. Compared to patients with CCI of 0, patients with CCI of 2 and above had a higher risk of all-cause mortality (HR: 1.740; p = 0.001), although this was not observed among patients with CCI of 1. Fuhrman grade 3 or 4 tumors were also associated with a higher risk when compared to Fuhrman grade 1 or 2 tumors (HR: 1.464; p = 0.001). On the other hand, papillary and chromophobe tumors were associated with lower risks (HR: 0.357, 0.260; p < 0.001, <0.001, respectively), when compared to clear cell RCC.

There were no significant differences in the risks of all-cause mortality between males and females, among the various categories of race, and between T2a and T2b tumors. Table 2 illustrates the result of Cox multiple regression estimating all-cause mortality.

The PN group had better OS than the RN group. Log-rank, Breslow, and Tarone-Ware tests consistently showed p-values <0.001. Figure 2 represents the OS curves for both groups, at mean values of other covariates.

OPEN IN VIEWER

FIG. 2.

OS curves for the matched PN and RN groups, at mean values of all other covariates. OS = overall survival.

A higher percentage of PN had positive surgical margins compared to RN (4.4% vs 2.5%, p = 0.01). There were no differences in postoperative length of stay and 30-day unplanned readmission between the two groups. Table 3 illustrates the immediate postoperative outcomes for both groups.

Discussion

The fundamental principle of PN involves tumor excision with an adequate margin, while leaving behind as much functional parenchyma as possible. This results in improved postoperative renal function without compromising oncologic outcomes, and translates into better OS due to decreased risks of renal failure and cardiovascular events. ^1,2 Currently, PN is the preferred option for T1 tumors whenever technically feasible. ^3,4

The larger size of T2 tumors makes it difficult to uphold the fundamental principle of PN. Removal of larger tumors means that the remaining parenchymal volumes are often much less than those in cases of T1 tumors, and such volumes are of paramount importance for preserving postoperative renal function. ¹² A large tumor also implies a technically more difficult operation, increasing ischemic time and complication rates. ^13,14

The proof for underutilization of PN in T2 tumors is to show better OS without compromising cancer-specific survival and perioperative safety. The current literature has already shown similar oncologic outcomes between PN and RN for T2 tumors, based on large national databases. ^5,6 The few studies with OS as an endpoint had small sample sizes, with key covariates such as age and tumor size being significantly different between their PN and RN groups. ^8,9 These studies might be underpowered for reliable conclusions. In fact, sample size was quoted as the “cornerstone limitation” by an early study comparing PN and RN for locally advanced tumors. ¹⁵ While the NCDB does not contain data on cancer-specific survival, it has information on OS and some immediate postoperative outcomes based on the majority of cancer cases diagnosed in the United States (US), thus overcoming the problem of small sample size.

The data from the NCDB were collected through daily clinical practices in the participating US centers; so patients with certain demographics and tumor characteristics might be preferentially treated with PN or RN. In fact, the PN and RN groups were significantly different in tumor size, histology, and grade before any statistical matching, and any of these covariates could act as confounders when making comparisons between the groups. Our use of statistical matching allowed for comparisons between PN and RN groups with high similarity in patient demographics and tumor characteristics.

Currently, the role of PN in T2 tumors remains controversial, and no guidelines have explicitly recommended its use. However, with advances in surgical equipment and techniques, tumors of increasing sizes are treated with PN, as evidenced by the changing recommendations in the European Association of Urology guidelines over time. ^4,16,17 It may only be a matter of time for PN use to cross the arbitrary 7 cm mark between T1 and T2 disease, and become the standard treatment for T2 tumors.

The key finding of our study was the significantly better OS associated with PN compared to RN, while controlling for other covariates. Even though the impact of cancer-related factors on survival tends to be diluted with time, the OS curves for the groups in our study appeared to increasingly diverge with longer duration of follow-up. Studies on T2 tumors from the SEER database between 1988 and 2011 showed similar cancer-specific survivals between PN and RN, ^5,6 while other studies showed better postoperative renal function after PN. ^7,9 Therefore, it is reasonable to infer that, at least to some extent, improved OS among PN patients in our study was attributed to nephron preservation.

Whether there is a threshold in the amount of nephron preservation necessary to impact survival could not be determined in this study, since the NCDB does not contain information on renal volumetrics and perioperative renal function. A recent study specifically looked at complex renal tumors that mandated significant renal volume excision even when treated with PN, and compared functional and survival outcomes between PN and RN. The authors found that a median renal volume preservation of 57% in the PN group was sufficient to decrease the risk of postoperative renal deterioration. With a median follow-up duration of 21 months, the cancer-specific survivals were similar for their PN and RN group (96.2% vs 98.1%), while the OS for the groups was more divergent (92.3% vs 82.7%), although statistical difference could not be achieved likely due to the small sample size of their study. ¹⁸

Besides the type of nephrectomy, we also identified several other covariates as significant factors for all-cause mortality. Some of these covariates, such as old age, poor comorbid status, and high tumor grade, are well known to have negative impacts on prognosis in renal cancer in general, ^4,19 and our results showed them as poor prognostic factors specifically for T2 tumors.

Our study confirmed histological subtype as a significant prognostic factor for T2 tumors. By comparing their respective HR for all-cause mortality, chromophobe tumors had better OS than papillary tumors, which in turn had better OS than clear cell tumors. A previous study investigating the prognostic value of histological subtypes showed a similar trend with univariate analysis, but not with multivariate analysis together with tumor stage, grade and performance status, despite a difference in median survival of 34 months between papillary and clear cell RCC. ²⁰ It is possible that histological subtypes indeed led to significant differences in OS, but these differences were statistically diluted in their multivariate model by other strong prognostic factors. Such suppression effect in multivariate analysis is a well-known source of bias, causing clinically meaningful findings to be masked by dominant covariates. ²¹ Our study involved T2 tumors only; thus no other T-stage categories were included in the multivariate model. Histological subtype remained a significant prognostic factor for OS, with every category significantly different in HR for all-cause mortality.

We also found a significantly higher positive margin rate for PN compared to RN. Whether positive margins after PN have any clinical impact is a controversy. A multicentered study by the Korean RCC Group found that positive margins after PN for T1 tumors did not affect recurrence-free survival, ²² while another US-based multicentered study had the opposite conclusion. ²³ A study based on the NCDB between 2003 and 2006 showed the association between positive margins and poorer OS. ²⁴ Nonetheless, despite higher positive margin rate, our study found that T2 tumors treated with PN still had better OS than RN.

Our study has several limitations. Retrospective analysis of the NCDB was inevitably subjected to selection bias. Statistical matching made the PN and RN groups highly similar in many key covariates; so any difference in OS between groups would be more likely attributed to the type of nephrectomy performed. However, matching could only account for covariates that were made available by the NCDB. Factors that affected patient assignment to PN or RN, yet unavailable in the NCDB, could not be accounted for during the process of matching. The validity of our findings was also largely dependent on the robustness of the NCDB. Tumor grade was the covariate with the highest proportion of missing data (17.0% and 9.9% in the PN and RN group, respectively), while most other covariates had missing data rates less than 1%, which we considered statistically very acceptable. Coding accuracy was another potential source of error beyond our control when analyzing an external database such as the NCDB. We assumed the quality of data from the NCDB to be high, since it has gained the reputation of being the largest clinical registry globally and a leading clinical surveillance and quality improvement mechanism for cancer programs in the US. ²⁵

The NCDB does not contain cancer-related outcomes such as cancer-specific and progression-free survivals. To circumvent this problem, we inferred from other studies based on the SEER database in a similar time frame that PN and RN for T2 tumors have similar oncologic outcomes. ^5,6 Since both the SEER and NCDB are considered representative of the majority of the US population, there should be sufficient clinical basis for such an inference.

Conclusions

Based on a contemporary national database, we found that T2 renal tumors treated with PN were associated with better OS than those treated with RN, despite higher positive margin rates. Age, CCI, tumor histology, and grade were significant factors affecting OS in patients with T2 tumors.