Partial and radical nephrectomy provide comparable long-term cancer control for T1b renal cell carcinoma

Authors


Abstract

Objectives

To examine utilization rates of partial nephrectomy relative to radical nephrectomy for T1b renal cell carcinoma in contemporary years, to identify sociodemographic and disease characteristics associated with partial nephrectomy use, and to compare effectiveness of partial versus radical nephrectomy with respect to cancer control.

Methods

Using the Surveillance, Epidemiology, and End Results database, 16 333 patients treated with partial or radical nephrectomy for T1bN0M0 renal cell carcinoma between 1988 and 2008 were identified. Logistic regression models were carried out to identify determinants of partial nephrectomy. Subsequently, cumulative incidence rates of cancer-specific and other-cause mortality between partial and radical nephrectomy were assessed, within the matched cohort. Furthermore, competing-risks regression analyses were used for prediction of cancer-specific mortality, after adjusting for other-cause mortality, and vice versa.

Results

The utilization rate of partial nephrectomy increased from 1.2% in 1988 to 15.9% in 2008 (P < 0.001). Younger individuals, smaller tumors, persons of black race, as well as men, were more likely to be treated with partial nephrectomy in the current cohort (all P ≤ 0.002). In the post-propensity cohort, the 5- and 10-year cancer-specific mortality rates were 4.4 and 6.1% for partial versus 6.0 and 10.4% for radical nephrectomy, respectively (P = 0.03). Competing-risks regression analyses showed that nephrectomy type was not statistically significantly associated with cancer-specific mortality, even after adjusting for other-cause mortality (hazard ratio 0.89, P = 0.5).

Conclusions

Despite providing a comparable cancer control, the use of partial over radical nephrectomy for T1b renal cell carcinoma in USA has remained limited in recent years.

Abbreviations & Acronyms
CI

confidence interval

CSM

cancer-specific mortality

HR

hazard ratio

ICD

International Classification of Disease for Oncology

IQR

interquartile range

OCM

other-cause mortality

OR

odds ratio

PN

partial nephrectomy

RN

radical nephrectomy

RCC

renal cell carcinoma

Ref

reference

SEER

Surveillance, Epidemiology and End Results

SES

socioeconomic status

Introduction

PN is currently the preferred treatment modality for patients with localized RCC, whenever technically feasible.[1-3] This recommendation was made after the emergence of numerous studies that showed improved functional and oncological outcomes relative to RN.[4-6] However, many of such reports predominantly focused on T1a RCC. Having recognized the expanding indications of nephron sparing, several investigators sought to show that the benefits of PN in the setting of T1a RCC are equally applicable in the setting of T1b RCC.[7, 8] Indeed, according to data from centers of excellence, PN can result in comparable cancer control outcomes relative to RN. However, previous studies using population-based data showed that utilization of PN remained low in the community setting (5–9%).[9, 10]

The reluctance of nephron-sparing consideration amongst individuals with T1b RCC might stem from the paucity of data confirming the safety of PN relative to RN in contemporary years. Whereas a recent study using a population-based cohort originating from the USA showed that PN resulted in comparable oncological outcomes after RN, the authors omitted to account for the effect of baseline comorbidities at surgery, which represent an important predictor of OCM.[10]

To address this limitation, we sought to compare and evaluate CSM after accounting for OCM, and vice versa, between PN and RN using a retrospective population-based cohort of patients diagnosed with T1bN0M0 RCC. We hypothesized that PN is associated with comparable cancer control outcomes relative to RN.

Methods

Data source

The current study relied on the SEER database reported by the National Cancer Institute statistics program between 1988–2008. The SEER routinely collects patient demographics, and publishes cancer incidence and survival data from population-based cancer registries that cover approximately 26% of the USA population.

Study population

Patients with a primary diagnosis of T1bN0M0 RCC treated with either PN or RN were abstracted using ICD diagnostic codes C67.0–C67.9. Only patients with clear cell, papillary, chromophobe RCC, as well as those with cystic, sarcomatoid and collecting duct histological subtypes were included (histological type ICDO3). To eliminate most childhood renal tumors, as well as to account for SES, we excluded patients aged younger than 25 years. Furthermore, death certificate and autopsy cases were removed from our analysis (n = 61).

Description of variables

Variables evaluated included: age at diagnosis, sex (male vs female), race (white vs black vs others vs unknown), marital status (married vs others [separated, divorced, widowed, never married] vs unknown), SES (high vs low), year of surgery, tumor size, tumor grade, histological subtype (clear cell vs papillary vs chromophobe vs others [cystic, sarcomatoid and collecting duct]) and hospital region (east vs Pacific vs south vs north). Patients' age at diagnosis, year of surgery and tumor size were examined as continuous variables. Tumor grade was categorized into: low (grade I–II), high (III–IV) and unknown. SES was defined according to three county-attribute variables: education level as the percentage of persons who did not complete high school, poverty as the percentage of persons living below the poverty line and the median family income. These county attribute variables were measured using the census 2000 documentation files available at: http://www.census.gov/main/www/cen2000.html. We created a composite variable of SES using the three aforementioned variables based on the methods of previous methodology.[11] First, we recoded the variables individually to ensure that low values represent low SES, and vice versa. Second, we transformed all values into standardized scores. Third, we computed the sum of these scores and categorized the total scores into two groups according to the median, which resulted in our low and high SES.

Statistical analysis

Descriptive statistics focused on frequencies and proportions for categorical variables. Means, medians and interquartile ranges were reported for continuously coded variables. The Mann–Whitney test and χ2-test were used to compare the statistical significance of differences in medians and proportions, respectively.

In the first part, we focused on identifying sociodemographic and disease characteristics that were most associated with the use of PN for patients treated between the years 2000 and 2008. The decision to focus on more contemporary years of diagnosis was based on the fact that the importance of nephron sparing was less in more historical years. This was assessed using logistic regression analyses. Covariates comprised of patient age, sex, race, marital status, SES, year of diagnosis, tumor size, tumor grade and hospital region.

In the second part, we examined the comparative effectiveness of oncological outcomes between PN and RN. Because of inherent baseline differences between individuals undergoing PN or RN, we relied on a 4:1 propensity-score matched analyses to reduce such differences to a minimum. The reliance on propensity-score matching eliminates the customary bias associated with the conventional multivariable modeling approach.[12, 13] Propensity scores were computed by modeling a logistic regression with the dependent variable as the odds of undergoing a PN, and the independent variables as patient age at diagnosis, sex, race, marital status, year of surgery, SES, tumor size, tumor grade and hospital region. The standardized difference measure was used to test how well the PN patients matched to the RN patients. Subsequently, covariate balance between the matched groups was examined.[13]

Within the post-propensity matched cohort, unadjusted survival outcomes between PN and RN were assessed at 5 and 10 years using cumulative incidence analyses. Finally, given that the selection of patients considered for PN will likely depend on tumor characteristics, such as tumor size, and sociodemographic characteristics, such as age, we carried out multivariable competing-risks regression analyses to account for such known confounders. In all multivariable models, covariates comprised of patient age, sex, race, marital status, SES, tumor size, tumor grade, year of surgery, histological subtype and hospital region, as well as OCM. The latter could be a reliable proxy for baseline comorbidities, which when accounted for, can reduce overestimation of CSM.

All statistical analyses were carried out using the R statistical package system (R Foundation for Statistical Computing, Vienna, Austria), with a two-sided significance level set at P < 0.05.

Results

Pre-propensity matching

Overall, 16 333 individuals treated with PN or RN for T1bN0M0 RCC between the years 1988 and 2008 were identified (Table 1). The proportion of patients treated with PN was 9.3% during the entire study span. The utilization rate of PN rose from 1.2% in 1988 to 15.9% in 2008, (+0.9% per year, P < 0.001; Fig. 1). In general, PN-treated individuals were younger (median 60 vs 62 years, P < 0.001), more frequently male (68 vs 60%, P < 0.001) and of black race (12.7 vs 9.7%, P < 0.001) compared with their RN counterparts. With respect to disease characteristics, PN patients had smaller tumors at surgery (median 5.0 vs 5.5 cm, P < 0.001) and more frequently harbored low-grade disease (I–II: 57.4 vs III–IV: 55.3%, P < 0.001). No clinically meaningful differences were observed between PN and RN patients in terms of hospital region, socioeconomic status and marital status at nephrectomy.

Figure 1.

Utilization rate of partial nephrectomy in T1bN0M0 RCC patients, SEER 1988–2008.

Table 1. Pre-propensity and post-propensity baseline descriptive characteristics of 16 333 T1b RCC patients treated with PN or RN, SEER 1988–2008
VariablesOverallPre-propensityPost-propensity
PNRNStd. mean diff.PNRNStandard mean differences
Patients, n (%)16 333 (100)1 526 (9.3)14 807 (90.7) 1 526 (20)6 104 (80) 
Age at diagnosis (years)   −0.128  −0.011
Mean (median)61.7 (62)60.3 (61)61.9 (62)60.3 (61)60.6 (61)
IQR53–7252–6953–7252–6952–70
Sex   −0.158  −0.012
Male, n (%)9 963 (61)1 033 (67.7)8 930 (60.3)1 033 (67.7)4 138 (67.8)
Female, n (%)6 370 (39)493 (32.3)5 877 (39.7)493 (32.3)1 966 (32.2)
Race   0.016  −0.003
White, n (%)13 671 (83.7)1 249 (81.8)12 422 (83.9)1 249 (81.8)5 102 (83.6)
Black, n (%)1 626 (10)194 (12.7)1 432 (9.7)194 (12.7)610 (10)
Others, n (%)943 (5.8)74 (4.8)869 (5.9)74 (4.8)345 (5.7)
Unknown, n (%)93 (0.6)9 (0.6)84 (0.6)9 (0.6)47 (0.8)
Marital status   −0.014  −0.004
Married, n (%)10 545 (64.6)996 (65.3)9 549 (64.5)996 (65.3)3 999 (65.5)
Others, n (%)5 192 (31.8)479 (31.4)4 713 (31.8)479 (31.4)1 857 (30.4)
Unknown, n (%)596 (3.6)51 (3.3)545 (3.7)51 (3.3)248 (4.1)
SES   −0.051  −0.016
Low, n (%)8 115 (49.7)764 (50.1)7 351 (49.6)764 (50.1)2 999 (29.1)
High, n (%)8 218 (50.3)762 (49.9)7 456 (50.4)762 (49.9)3 105 (50.9)
Year of surgery0.810.015
Tumor size   −0.443  −0.007
Mean (median)5.45 (5.4)5.14 (5)5.48 (5.5)5.14 (5)5.15 (5)
IQR4.7–64.5–5.54.8–64.5–5.54.5–5.5
Tumor grade   −0.029  0.007
Low, n (%)9 062 (55.5)876 (57.4)8 186 (55.3)876 (57.4)3 506 (57.4)
High, n (%)3 261 (20)352 (23)2 910 (19.7)351 (23)1 393 (22.8)
Unknown, n (%)4 010 (24.6)299 (19.6)3 711 (25.1)299 (19.6)1 205 (19.7)
Hospital region   −0.019  −0.002
East, n (%)5 258 (32.2)567 (37.2)4 691 (31.7)567 (37.2)2 256 (37)
Pacific, n (%)7 877 (48.2)725 (47.5)7 152 (48.3)725 (47.5)2 904 (47.6)
North, n (%)2 364 (14.5)172 (11.3)2 192 (14.8)172 (11.3)698 (11.4)
South, n (%)834 (5.1)62 (4.1)772 (5.2)62 (4.1)246 (4)

Determinants of PN use

In multivariable analyses, men (OR 1.37, P < 0.001), persons of black race (OR 1.29, P = 0.003) and those treated in more contemporary years (OR 1.16, P < 0.001) were more likely to undergo PN (Table 2). In contrast, older patients (OR 0.99, P < 0.001), persons of other race (OR 0.74; P = 0.04) and persons with larger tumors (OR 0.95, P < 0.001) were less likely to undergo PN.

Table 2. Pre-propensity univariate and multivariate logistic regression analyses focusing on determinants of PN use in patients with T1b RCC treated with either PN or RN, within SEER between years 2000 and 2008
PredictorsRates of PN
UnivariateMultivariate
OR (95% CI)P-valueOR (95% CI)P-value
Age at diagnosis0.99 (0.985–0.993)<0.0010.99 (0.984–0.993)<0.001
Sex    
FemaleRef Ref 
Male1.38 (1.23–1.55)<0.0011.39 (1.23–1.56)<0.001
Race    
WhiteRef Ref 
Black1.29 (1.09–1.52)0.0031.28 (1.08–1.53)0.005
Others0.82 (0.64–1.07)0.140.75 (0.57–0.99)0.04
Unknown0.95 (0.48–1.90)0.90.84 (0.42–1.71)0.6
Marital status    
MarriedRef Ref 
Others0.96 (0.85–1.08)0.50.99 (0.88–1.12)0.9
Unknown0.79 (0.58–1.07)0.10.77 (0.57–1.06)0.1
SES    
HighRef Ref 
Low0.93 (0.83–1.04)0.20.91 (0.82–1.02)0.1
Year of diagnosis1.13 (1.10–1.15)<0.0011.13 (1.10–1.16)<0.001
Tumor size0.95 (0.94–0.96)<0.0010.95 (0.94–0.96)<0.001
Hospital region    
PacificRef Ref 
East1.04 (0.92–1.17)0.50.98 (0.87–1.11)0.8
North0.93 (0.77–1.12)0.50.89 (0.73–1.08)0.2
South0.95 (0.72–1.27)0.70.95 (0.71–1.27)0.7

Post-propensity matching

To reduce to a minimum treatment selection biases, 4:1 matching was carried out using propensity-score methodology. Specifically, 6104 RN individuals were matched to the 1526 PN individuals (Table 1). After matching, baseline sociodemographic and disease characteristics between the two groups were now comparable, as shown by the standardized mean differences of ≤10%, signifying a high degree of similarity in the distribution of both populations.

Cancer control outcomes

In the post-propensity matched cohort, the 5- and 10-year CSM rates were 4.4 and 6.1% for PN versus 6.0 and 10.4% for RN, respectively (P = 0.03; Fig. 2a). For the same time-points, OCM rates were 7.0 and 23.8% for PN versus 9.1 and 20.3% for RN (P = 0.3, Fig. 2b). After adjusting for all covariates, including OCM, in competing-risks analyses predicting CSM, PN was not statistically significantly different with respect to RN (HR 0.89, P = 0.5; Table 3). Similarly, in multivariable analyses predicting OCM, nephrectomy type remained non-statistically significantly different (HR 0.95, P = 0.6).

Figure 2.

Cumulative incidence plots showing (a) CSM and (b) OCM rates at 5 and 10 years in T1bN0M0 RCC patients treated with PN or RN, SEER 1988–2008. image, PN; image, RN.

Table 3. Post-propensity univariable and multivariable competing-risks regression analyses for prediction of cancer-specific mortality among 16 333 T1b RCC patients, SEER 1988–2008
VariablesUnivariableMultivariable
HR (95% CI)P-valueHR (95% CI)P-value
Treatment type    
RNRef Ref 
PN0.91 (0.65–1.26)0.60.89 (0.64–1.24)0.5
Age at diagnosis1.03 (1.02–1.04)<0.0011.03 (1.02–1.05)<0.001
Sex    
FemaleRef. Ref. 
Male1.07 (0.81–1.42)0.61.04 (0.78–1.39)0.8
Race    
WhiteRef. Ref. 
Black1.41 (0.94–1.56)0.061.69 (1.13–2.51)0.01
Others0.6 (0.3–1.23)0.20.47 (0.2–1.07)0.07
Unknown
Marital status    
MarriedRef. Ref. 
Others0.94 (0.71–1.24)0.70.92 (0.68–1.23)0.6
Unknown0.85 (0.4–1.81)0.70.86 (0.4–1.84)0.7
SES    
HighRef. Ref. 
Low1.21 (0.94–1.56)0.11.23 (0.94–1.61)0.1
Tumor size1.04 (1.02–1.05)<0.0011.04 (1.02–1.05)<0.001
Tumor grade    
LowRef. Ref. 
High2.3 (1.7–3.1)<0.0012.23 (1.63–3.04)<0.001
Unknown1.23 (0.89–1.69)0.21.08 (0.77–1.5)0.7
Histological subtype    
Clear cellRef. Ref. 
Papillary0.93 (0.59–1.46)0.80.92 (0.57–1.47)0.7
Chromophobe0.92 (0.47–1.97)0.80.95 (0.48–1.87)0.9
Others0.5 (0.12–2.04)0.30.52 (0.13–2.15)0.4
Hospital region    
PacificRef. Ref. 
East0.8 (0.6–1.07)0.10.78 (0.57–1.07)0.1
North1.19 (0.82–1.72)0.41.08 (0.73–1.59)0.7
South0.96 (0.5–1.83)0.90.92 (0.47–1.78)0.8
Year of surgery0.95 (0.92–0.99)0.0070.95 (0.91–0.98)0.004

Discussion

In the past decades, the incidence of RCC has steadily increased worldwide.[14] Because of increased access and utilization of imaging, the majority of newly diagnosed cases are small renal masses (∼60%), resulting in a downward stage migration of the disease over time.[15, 16] In the context of such paradigm shift, PN, historically reserved for a tumor in a solitary kidney, bilateral renal tumors or renal function impairment, emerged as a viable alternative to RN.

There is currently a strong body of literature supporting the use of PN over RN for T1a RCC. Benefits of nephron sparing include better renal function outcomes,[17] lower rates of cardiovascular-related outcomes[18] and overall survival.[5, 18] Under such premise, several investigators have challenged the 4-cm tumor size cut-off for indication of PN, and hypothesized that provided technical feasibility, it could be reasonable to consider PN for larger tumors. In consequence, we sought to examine the utilization rates of PN amongst patients with T1b RCC, and to identify sociodemographic and disease characteristics most associated with the use of PN in the community. In as well, we sought to examine and compare cancer control outcomes between PN and RN-treated individuals, using propensity-based matched analyses and competing-risks regression analyses.

Several findings were noteworthy. Primarily, albeit an increasing trend, utilization of PN remained low during the entire study span. As of the year 2008, just 16% of patients with T1b disease underwent a nephron-sparing approach, suggesting that PN is uncommonly considered in patients with tumors greater than 4 cm. A few hypotheses could be advanced for the slow adoption of PN. First, reluctance of surgeons to carry out PN in patients with T1b disease might be related to concerns regarding postoperative complications. Indeed, the risk of postoperative morbidity as a result of the technical complexity of PN relative to RN represents a strong argument against a nephron-sparing approach for many.[19] Second, the advent of laparoscopy has led many surgeons to lean towards laparoscopic RN in lesions that would have otherwise been amenable to open PN.[20] Finally, many patients simply do not have access to PN, for reasons that could pertain to traveling time, inability to be treated at a hospital that offers PN, insurance, racial disparity or socioeconomic status.[21-25] Indeed, in analyses focusing on characteristics most associated with PN treatment, the present results showed that younger individuals and males were more likely to be treated with PN than their older and female counterparts. Interestingly, the present results also showed that persons of black race were more likely to be treated with the nephron-sparing approach relative to whites, contrary to results of previous reports that suggested a treatment disparity for blacks. This positive finding should nonetheless be interpreted with the consideration that within the current cohort, T1b patients of black race were much younger than whites. Given that older individuals, frequently harboring several baseline comorbidities, might not be offered a PN because of the increased risk of complications, it might have resulted in an artificial effect where whites seemed less likely to be treated with PN. Additionally, the present findings showed that increasing tumor size was associated with a lower rate of PN utilization. Such results plausibly reflect difficulty in carrying out PN with increasing tumor size, as well as a cautious attitude towards the consideration of a nephron-sparing approach amongst surgeons treating more complex cases.

Importantly, the present results showed that PN resulted in comparable cancer control outcomes relative to RN after adjusting for confounders. Furthermore, in contrast to previous data focusing on T1a RCC that showed a significant protective effect of PN relative to RN,[26] patients treated with PN for T1b RCC in the current study were not associated with more favorable OCM compared with RN. One possible explanation stems from the fact that resection of T1b tumors compared with T1a tumors might be associated with a decreased amount of functional renal parenchyma preserved, which is likely to be significantly associated with postoperative renal function.[27] In addition, PN in T1b might be associated with higher rates of postoperative bleeding, urinary fistula and ischemia time, which could comprise the residual renal function.[28, 29] However, the lack of an apparent survival difference between nephrectomy types with respect to CSM confirms the relative safety of PN with respect to oncological or functional outcomes postoperatively. These findings are in line with previous reports. For example, using the SEER database, Crépel et al. tested the effect of treatment type (RN vs PN) on CSM using competing-risks analyses amongst patients with T1b lesions, treated between 1988 and 2004.[30] After matching, a non-statistically significant difference for CSM between PN and RN was recorded (P = 0.3). Given that those results could be outdated, Badalato et al. sought to assess comparative effectiveness between PN and RN in T1b lesions using a more contemporary version of the SEER (1998–2007).[10] In that study, no difference with respect to CSM was recorded between PN and RN. However, the lack of adjustment for baseline comorbidities or OCM might have introduced a bias. Consequently, the current study represents the most up-to-date and conceptually complete report for comparison of PN and RN for T1b RCC.

Taken together, the current results support findings from previous studies that showed that PN is relatively safe compared with RN in patients with lesions >4 cm. Presumably, provided an adequate learning curve and technical feasibility, the benefits of PN related to functional and oncological outcomes that have been shown for T1a can be equally applicable to patients with T1b lesions. However, the current results also suggest a notable slow adoption of PN for T1b over time. Relative to the reported adoption of PN at centers of excellence, community acceptance of PN has not yet occurred. Furthermore, disparate use of PN was recorded with respect to several sociodemographic characteristics. These issues should ideally be addressed in upcoming years.

The present study was not devoid of limitations. First, unavailable information on comorbidities represents a limiting factor in the evaluation of the PN use. This limitation was shared with previous studies.[10, 30] Accounting for comorbidities could have altered the current findings. However, we adjusted for OCM, which might represent a relative estimate for comorbidities. Second, the SEER database contains no information about the type (open vs laparoscopic vs robotic) of PN or RN that was carried out. Third, the SEER database does not provide any information about tumor location and renal function. These criteria distinguish between relative and absolute PN indication. Furthermore, we were not able to adjust for hospital characteristics, such as hospital volume. Finally, it is a retrospective design like all previous studies that assessed oncological outcome in patients treated with PN or RN for T1b.[7, 10, 30]

In conclusion, PN for T1b RCC is safe compared with RN. That being said, no superior OCM advantage was recorded for PN relative to RN in such patients, which could render the indication of PN over RN less appealing and questionable if functional outcomes are considered.

Acknowledgments

Pierre I Karakiewicz is partially supported by the University of Montreal Health Centre Urology Specialists, Fonds de la Recherche en Santé du Québec, the University of Montreal Department of Surgery and the University of Montreal Health Centre (CHUM) Foundation.

Conflict of interest

None declared.

Ancillary