Treatment Facility Volume and Survival in Patients with Metastatic Renal Cell Carcinoma: A Registry-Based Analysis

Abstract

Background:

Higher treatment facility (TF) volume has been linked with improved oncologic treatment outcomes.

Objective:

To determine the association between TF volume and OS in patients with metastatic RCC (mRCC).

Design, Setting, and Participants:

The National Cancer Data Base (NCDB) was queried for all patients with mRCC with survival data available (2004 to 2013, cohort A). We defined high volume TFs as those in the top 20^th percentile of mean number of mRCC patients treated per year, and also examined the association with continuous volume. Increasingly narrow inclusion criteria were used to confirm the cohort A association: Cohort B = mRCC patients with active treatment; cohort C = mRCC patients with systemic therapy; cohort D = mRCC patients with systemic therapy at the reporting institution; cohort E = mRCC patients with systemic therapy at the reporting institution with known liver and lung metastatic status. Survival analyses were also performed on sub-cohorts of mRCC who never underwent a nephrectomy (C1, D1, and E1).

Outcome Measurements and Statistical Analysis:

The effect of volume on OS was determined using unadjusted Kaplan-Meier curves and Cox regression models (MVA) after adjustment for multiple clinicopathologic factors.

Results and Limitations:

There were 41,836 mRCC patients treated at 1,222 TFs. The median age was 65. 66% were men and 79% had clear cell mRCC. Median TF volume was 2.2 patients/year, (range 0.1–44.3). High volume TFs (≥4.8pts/year) treated 54% of all mRCC patients. The unadjusted median OS of all mRCC patients treated at high vs. low volume TFs was 9.5 vs. 6.5 months (P<0.0001). This difference was maintained in all cohorts: Cohort B=14.2 vs. 11.3, cohort C=12.6 vs. 10.1, cohort D=12.7 vs. 9.2, and cohort E=14.0 vs. 10.6 months (P < 0.0001 for all). MVA confirmed that facility volume was associated with OS after adjustment; cohort A: HR=0.85, [95% CI 0.82–0.88], P < 0.001. These results were consistent regardless of the cohort examined, even after exclusion of all nephrectomy patients. For all TFs treating ≥ 5 pts/yr, OS improved as TF volume increased (20 pts/yr: HR=0.75). Limitations include the retrospective nature of NCDB analysis and the lack of information on treatment regimens used at specific facilities, which may explain mechanisms of effects.

Conclusions:

Patients with mRCC treated at higher volume facilities had a longer survival compared with those treated at lower volume facilities.

Patient Summary:

In this report we analyzed a large cancer database and found that patients with metastatic kidney cancer survived longer if they were managed at facilities that treated a higher volume of such patients. This information can be used as an aide to help find the best treatment environment for patients with metastatic kidney cancer.

Introduction

Although mortality rates for patients with renal cell carcinoma (RCC) have declined over several decades, survival following the diagnosis of metastatic RCC (mRCC) continues to be poor.[1] mRCC is often an aggressive disease that is poorly responsive to traditional cytotoxic systemic therapies, with the 5-year overall survival (OS) for patients as low as 8%, leading to over 14,000 deaths from RCC annually.[1, 2]

The landscape for systemic mRCC therapy has rapidly evolved over the last ten years, first with improvements in targeted therapies and more recently with the development of novel immunotherapies. Advanced knowledge of these ever-changing treatment options may be necessary to obtain optimal patient outcomes. Centers that manage higher volumes of cancer patients likely employ providers that have such advanced knowledge and treatment experience, as well as access to novel drugs via clinical trials. Indeed, treatment volume has historically been used as a surrogate marker for hospital and provider experience and may also indicate the presence of more streamlined care processes that can impact patient outcomes.

The volume-outcome relationship of various medical treatments and procedures has long been established, though the magnitude of this association varies greatly.[3, 4] This relationship also appears to hold true for cancer therapies, with mounting evidence to suggest that treatment facilities (TFs) that manage a higher volume of cancer patients might have improved survival outcomes.[5–9]

For RCC, the volume-outcome relationship has been explored for the treatment of localized disease. Many studies have demonstrated that high volume TFs lead to improved postoperative outcomes and fewer complications following renal cancer surgery.[10–12] Several studies showed improved in-hospital survival following high-risk nephrectomy for RCC, though it is unclear if surgery at high volume surgical centers necessarily translates to overall longer-term survival.[13, 14] However, there is little knowledge regarding the volume-outcome relationship for patients diagnosed and treated for mRCC. We therefore analyzed a large national cancer database to determine if there is a relationship between TF volume and survival outcomes for patients diagnosed with mRCC.

Material and Methods

Data Source

The National Cancer Database (NCDB), a program of the ACS® CoC (Commission on Cancer) and the American Cancer Society, is a national cancer registry and comprehensive clinical surveillance resource for cancer care in the United States. The NCDB compiles data from over 1,500 commission-accredited cancer programs in the United States and Puerto Rico, and captures approximately 70% of all newly diagnosed cancer cases.[15] The use of national de-identified registry data was exempt from IRB approval.

Study Population

Patients with metastatic renal cell carcinoma were identified in the NCDB based on ICD-O-3 site codes. All histologic subtypes of RCC were included (ICD-O-3 site code C649). Our study cohort included all patients who were diagnosed with primary RCC between 2004 and 2013. Only patients with metastatic (M1) disease at diagnosis were selected for analysis. Patients were excluded if survival data was unavailable, or if they did not receive any treatment at the reporting facility. To confirm any association between TF volume and survival, patients were divided into five cohorts defined by increasingly stringent inclusion criteria (Figure 1): Cohort A included all patients with mRCC and available survival data (N = 41,836); cohort B was restricted to mRCC patients who underwent some active treatment (surgery or systemic therapy, N = 27,557); cohort C was further restricted to mRCC patients whose treatment included systemic therapy (with our without primary surgery, N = 19,138); cohort D required treatment with systemic therapy at the reporting institution (N = 12,000); and cohort E was further subset to those with known sites of metastases (i.e. known if liver/lung metastases present, N=4,933).

Figure 1.

Study cohort eligibility CONSORT diagram and cohort criteria

In order to isolate the influence of volume effects not related to surgery, subgroups were generated excluding patients that had surgical intervention in addition to systemic therapy. Cohorts C1 (N = 10,489), D1 (N = 6,898), and E1 (N = 2,866) correspond to Cohorts C, D, and E, after excluding patients who had surgery (partial or radical nephrectomy) as part of their treatment (Figure 1).

Treatment Facility, Patient Volume, and Study Outcome

High volume TFs were defined a priori as those in the top 20^th percentile of mean number of mRCC patients treated per year, which was determined to be ≥4.8 patients per year. TF volume was then considered as a continuous value to more closely examine effects on survival, and we also used conditional inference trees to find a volume cut-point which best separates patients by outcome. The primary outcome was overall survival (OS).

Covariates

The regression models were adjusted along a set of covariates available in the NCDB. These included patient age, sex, race, Hispanic ethnicity, year of diagnosis, insurance type, income, education, location, Charlson-Deyo comorbidity score[16], and clinical characteristics (histology, presence or absence of liver and lung metastasis). Our main model did not adjust for facility type, due to potential collinearity with facility volume. It also did not adjust for treatment type (i.e. surgery) because treatment differences were assumed to be the mechanism by which volume affects outcomes. Treatment type, therefore, was not considered a confounding covariate. There was a small proportion of missing data for some of the (categorical) covariates, with the highest being 6.3% missingess for Hispanic ethnicity. For covariates with missing data, a missing category was added to the multiple regression models.

Statistical Analysis

The effect of high TF volume on OS was determined using unadjusted Kaplan-Meier analysis with log-rank tests and multivariable Cox regressions, using robust standard errors to account for clustering within facilities. Separate survival analyses were conducted on all treatment cohorts, including the subgroups C1-E1 that excluded surgical patients. We also modeled the effect of TF volume as a continuous variable using flexible cubic splines in the Cox regression models.[17] In a series of sensitivity analyses, we tested models in clear-cell only subgroups; controlling for facility type as a covariate; and using higher thresholds to define “high volume” facilities. We also tested whether time trends (of survival) differed by facility volume using models with an interaction between volume and year of diagnosis. Statistical analysis was done with SAS®, version 9.3 and R version 3.3 with p <0.05 considered statistically significant.

Results

Patient Characteristics

41,836 mRCC patients were treated at 1,222 TFs (Table 1). The median patient age was 65, and men comprised 66% of the cohort. 86% of patients were Caucasian and 10% were African American. 79% of patients had clear cell mRCC, followed by 16.6% “other mRCC”, 3.5% papillary mRCC, and 0.7% chromophobe mRCC. The majority of patients had a Charlson-Deyo comorbidity score of “0” at both high and low volume centers (72.4% at high-volume, 70.7% overall). Supplemental Table 1 demonstrates results of a multivariable analysis of patient characteristics with respect to treatment at a high volume facility.

Table 1.

Patient clinical and demographic characteristics (N = 41,836)

Descriptive Groups		Low volume	High volume	P-value
N		19,389	22,447
GENDER (%)				<0.0001
	Male	64.7	67.6
	Female	35.3	32.4
AGE (%)				<0.0001
	<50	8.6	12.7
	51–60	21.7	25.8
	61–70	28.7	30.5
	71+	41.0	31.0
RACE (%)				<0.0001
	Caucasian	87.3	84.5
	African American	9.0	11.0
	Other/Unknown	3.7	4.5
HISPANIC (%)				<0.0001
	No	87.7	86.7
	Yes	5.4	7.6
	Unknown	7.0	5.8
RCC HISTOLOGY				<0.0001
	Clear Cell	79.5	78.8
	Papillary	3.0	4.0
	Chromophobe	0.6	0.8
	Other	16.8	16.5
LOCATION (%)				<0.0001
	Large metropolitan	43.3	50.8
	Small metropolitan	34.0	27.8
	Suburban	11.3	10.5
	Rural	7.1	6.4
	Unknown	4.3	4.5
CHARLSON SCORE (%)				<0.0001
	0	68.7	72.4
	1	21.9	20.0
	2	9.4	7.6
FACILITY TYPE (%)				<0.0001
	Community	21.0	0.3
	Comprehensive Community	61.9	23.2
	Academic/Research	14.9	59.9
	Integrated Network	2.2	16.7
MEDIAN HOUSEHOLD INCOME (%)				0.0005
	Less than $38,000	17.9	19.8
	$38,000 – $47,999	26.2	25.1
	$48,000 – $62,999	28.0	27.6
	$63,000 +	27.9	27.5
SURGERY (%)				<0.0001
	Ablation	0.4	0.5
	Partial Nephrectomy	1.5	2.4
	Radical Nephrectomy	33.4	43.6
	Unknown	0.6	0.4
	None	64.1	53.1
SYSTEMIC THERAPY (%)				<0.0001
	Yes	43.0	48.1
	No	57.0	51.9

Treatment and Facility Characteristics

A large proportion of patients lived in large metropolitan areas (47.3%), followed by small metropolitan locations (30.7%). Many patients (39.0%) were treated in the South Atlantic or East North Central regions, with a relatively even distribution throughout the rest of the U.S. Supplemental Figure 1 demonstrates the distribution of hospitals by mRCC treatment volume. The median TF volume was 2.2 patients/year (range 0.1 – 44.3). High volume TFs treated 54% of all mRCC patients. Of patients receiving treatment at low volume TFs, the majority (61.9%) were managed at a facility designated as a Comprehensive Community Cancer program. Those patients receiving treatment at high volume TFs, the majority (59.9%) were managed at an Academic/Research program with 16.7% treated at an integrated cancer network program.

Surgical and systemic treatments were significantly different between high and low volume TFs. High volume TFs performed more radical nephrectomies than low volume TFs (43.6% vs. 33.4%, p < 0.0001). High volume TFs were also more likely to deliver systemic therapy (48.1% vs. 43.0%, p <0.0001).

Survival Outcomes

The unadjusted median OS of all mRCC patients (cohort A) treated at high vs. low volume TFs was 9.5 vs. 6.5 months (p < 0.0001). High volume TFs maintained a superior median OS when analyzed by each sub-cohort: cohort B (14.2 vs. 11.3 months), cohort C (12.6 vs. 10.1 months), cohort D (12.7 vs. 9.2 months), and cohort E (14.0 vs. 10.6 months) [p < 0.0001 for all cohorts]. Figure 2 demonstrates the Kaplan-Meier survival estimates for high and low volume TFs (cohort A and cohort E). Multivariate analysis confirmed that facility volume was independently associated with all-cause mortality (HR=0.85, [95% CI 0.82–0.88], p < 0.001), [Table 2]. These results were consistent when analyzed by each cohort (Table 3). For cohort E, the most discriminative group, TF volume continued to be significantly associated with all-cause mortality (HR=0.81 [95% CI 0.75–0.87], P<0.0001). To exclude the possible influence of volume effects attributable to surgery, subgroups C1-E1 were analyzed and were found to still associate with improved survival, though the benefit was attenuated (E1 HR = 0.87 [95% CI 0.80–0.95], p=0.002) [Table 3]. The best cutpoint chosen using conditional inference trees (cohort A) was 7.9 pts/yr (top 9% of hospitals), but using this alternative threshold did not result in substantive changes to the results (HR=0.83 [95% CI 0.80–0.86], p<0.0001) for ≥7.9 pts/yr vs. <7.9 pts/yr. When considering TF volume continuously using the non-linear spline approach, multivariate analysis demonstrated that increased volume was associated with improved OS, with clinically meaningful improvements at facilities treating at least 5 patients per year (5 pts/yr: HR=0.92; 10 pts/yr: HR=0.84; 20 pts/yr: HR=0.75, versus reference=1 pt/yr). Figure 3 graphically illustrates the variation in HR as a function of TF volume for cohorts A and E (truncated at 30 pts/year given that only 3 centers treated >30 pts/yr). In sensitivity analysis, limiting the cohorts to clear cell RCC cases yielded consistent results (Cohort A: HR=0.84, Cohort E: HR=0.81, all P<0.0001). Including facility type as a covariate resulted in attenuated, but still significant effects (Cohort A: HR=0.93, P=0.001, Cohort E: HR=0.87, P=0.003).

Figure 2.

Kaplan-Meier survival estimates for high and low volume TFs (Cohorts A and E)

Table 2.

Multivariate survival analysis for entire cohort (Cohort A) (N = 41,836)

Variable*		H.R.**	95% CI			P-value
Facility Volume	High	0.853	0.824	-	0.883	<.0001
	Low	ref
Sex	Male	0.970	0.947	-	0.992	0.0093
	Female	ref
Age	Per 10 year increase	1.208	1.192	-	1.224	<.0001
Race	African American	1.154	1.106	-	1.205	<.0001
	Other/Unknown	0.929	0.877	-	0.984	0.0129
	White	ref
Hispanic	Yes	0.906	0.862	-	0.952	<.0001
	Unknown	1.053	1.004	-	1.105	0.0347
	No	ref
Charlson Score	1	1.081	1.050	-	1.112	<.0001
	2	1.313	1.259	-	1.370	<.0001
	0	ref
Location Type	Small Metropolitan	1.021	0.989	-	1.054	0.1973
	Suburban	1.010	0.967	-	1.055	0.6502
	Rural	0.992	0.939	-	1.047	0.7581
	Unknown	0.972	0.906	-	1.043	0.4324
	Large Metropolitan	ref
RCC Histology	chromophobe	0.720	0.633	-	0.820	<.0001
	papillary	0.939	0.888	-	0.993	0.0279
	other	1.640	1.589	-	1.694	<.0001
	clear cell	ref
Diagnosis Year	Per 1-year increase	0.981	0.976	-	0.985	<.0001

^*Variables controlled for but not included in table: insurance status, income quartile, facility location, education (high school degree)

^**Refers to hazard of mortality (i.e. HR<1 interpreted as improved survival)

Table 3.

Hazard ratios for overall survival for high versus low volume facilities by cohort.

Cohort	HR	95% CI	P-value
Cohort A: All M1 RCC	0.853	0.824–0.883	<0.0001
Cohort B: M1 RCC undergoing active treatment	0.860	0.829–0.891	<0.0001
Cohort C: M1 RCC undergoing systemic treatment	0.866	0.834–0.901	<0.0001
Cohort D: M1 RCC undergoing systemic treatment at reporting facility	0.798	0.760–0.838	<0.0001
Cohort E: M1 RCC undergoing systemic treatment at reporting facility (liver or lung metastatic status known)	0.807	0.752–0.866	<0.0001
Cohort Subgroups	HR	95% CI	P-value
Cohort C1: M1 RCC undergoing systemic treatment [without surgery]	0.927	0.884 – 0.971	0.002
Cohort D1: M1 RCC undergoing systemic treatment at reporting facility [without surgery]	0.864	0.815 – 0.915	<0.0001
Cohort E1: M1 RCC undergoing systemic treatment at reporting facility (liver or lung metastatic status known) [without surgery]	0.870	0.797 – 0.950	0.002

Figure 3.

Variation in HR as a function of TF volume (Cohorts A and E)

Male sex was significantly associated with improved survival (HR 0.97 [95% CI 0.95 – 0.99], p = 0.0093), and African American race was associated with worse survival (HR 1.15 [95% CI 1.11 – 1.21], p <0.0001). As expected, older mRCC patients demonstrated worse survival (HR=1.21 per 10 year increase in age [95% CI 1.19 – 1.22, p <0.0001). There was a small but statistically significant decrease in hazard of death over time (HR=0.98/year, P<0.0001), but this effect did not differ by hospital volume (P=0.22).

Discussion

The management of mRCC is becoming increasingly complex, and understanding the optimal environment for care delivery and resource utilization for patients diagnosed with this aggressive and relatively uncommon disease is important. Our retrospective analysis of 41,836 patients in the NCDB finds that mRCC patients managed at high volume (top quartile, mean ≥4.8 patients/year) TFs experienced significantly improved OS compared to lower volume TFs (9.5 vs. 6.5 months, respectively). The OS benefit from management at high volume TFs persisted in all cohorts, regardless of selection criteria, including those who never underwent a nephrectomy as part of their care. The beneficial volume-outcome relationship strengthened as mean treatment volume increased, with a clinically meaningful HR of 0.75 at TFs treating 20 patients/year or more.

In locoregional RCC, the emerging consensus from retrospective databases is that high volume TFs confer a multitude of short-term benefits, such as lower overall complication rates, blood transfusions, and length of stay.[11, 12] In-hospital and short-term mortality following complex surgery is also thought to be lower at high volume TFs.[7, 10, 13] Surgery at high volume centers has recently been estimated to offer as much as a 15 month survival benefit in some Urologic malignancies.[18] The volume-outcome relationship for complex oncologic procedures has led to initiatives, such as Leapfrog, that advocate for volume-based referral to improve quality for technically challenging cases.[7, 19, 20]

The benefits of hospital volume on the treatment of metastatic RCC are less well understood. The lack of clarity on optimal treatment environment partly stems from the historically poor prognosis of patients diagnosed with advanced RCC, with typical 5-year survival rates <10%.[1] For example, high-dose interleukin-2 therapy is known to result in complete, durable responses in ~7% of mRCC patients, but the significant toxicity of this treatment restricts its use only to those with the highest performance status at select centers, thus leading to patient selection in favor of high volume TFs.[21] With the introduction of targeted therapies (VEGF-targeted TKIs, mTOR inhibitors, multi-kinase inhibitors) and, more recently, immune-targeted agents such as the recently-approved PD-1 inhibitor nivolumab and forthcoming CTLA-4 inhibitor ipilimumab, patients now have far greater options for systemic treatment following diagnosis with mRCC. As such, provider knowledge and experience regarding the appropriate use and sequencing of these agents, as well as prudent patient selection for consideration of surgical therapies (e.g. cytoreductive nephrectomies, metastasectomies), may impact treatment outcomes. For targeted therapies and checkpoint inhibitors in particular, treatment management is nuanced and decisions of when to continue treatment, even in light of mild progression, and how to manage treatment-related adverse effects are unique and potentially difficult to manage in certain patients.[22] Early diagnosis and management of adverse treatment effects can potentially avoid therapy discontinuation or failure.

We thus hypothesized that improvements in survival outcomes at high volume centers would be due to differences in care management. Indeed, there were some notable differences in treatment delivery between high and low volume TFs. High volume TFs were more likely to perform cytoreductive nephrectomy (43.6% vs. 33.4%, p < 0.0001) and deliver systemic therapy (48.1% vs. 43.0%, p <0.0001). Overall care management, though, is complex and multifactorial; and in the NCDB many relevant measures lack detail or are unavailable. Treatment (systemic therapy, surgery) is part of care management; we did not directly adjust for all treatment variables as we considered them to be part of the mechanism through which volume affects outcomes (and are therefore not confounders). Instead, we used treatment received to define a series of increasingly stringent cohorts, which demonstrated that the benefits of treatment at a high volume center are not solely attributable to patient selection effects (i.e. if patients healthy enough to receive active therapy are more likely to be seen at high volume centers). We also performed a survival analysis on cohorts that excluded patients who received surgery (C1-E1), and were thus only managed with systemic therapy. These cohorts also demonstrated a significant survival benefit, though the effects were attenuated after excluding surgical patients. Taken together, these findings suggest that both surgery and systemic therapy play a role in improving survival at high volume TFs.

The management of patients undergoing systemic treatment requires infrastructure that supports frequent treatment, inpatient and outpatient nursing vigilance, financial assistance, and prompt treatment- and disease-related medical management. Such infrastructure is difficult and costly to build and maintain. Recruitment to clinical trials, though present in varying degrees at lower volume centers, is likely more robust at high volume centers. A recent study out of England demonstrated an improved colorectal cancer survival for all patients treated at centers that have high clinical trial participation, regardless of individual level therapy.[23] In addition, many of the drugs that are now standard of care were likely still experimental during the treatment period for this NCDB cohort, and access may therefore have been limited at low volume TFs.

Although this study might suggest that regionalization of mRCC treatment can improve outcomes, concentrating care to fewer facilities could raise other unique difficulties. For instance, regionalization of treatment without redirecting funds and reallocating resources can overwhelm the treatment capacity of some high volume TFs. Treatment regionalization can also require some patients to travel long distances to obtain care, and can potentially exacerbate existing disparities in outcomes by socioeconomic status or race.[24]

Improving outcomes at low volume TFs is complex and likely requires multi-factorial solution planning. While some factors afforded to high volume TF are not reproducible (such as clinician experience and depth of resources), there may be opportunities for care improvement, such as inter- and intra-hospital patient co-management across specialties, access to resources owned or operated by centers of excellence, utilization of telehealth solutions, and improvements in the accessibility, visibility, and availability of clinical trials.

There are several important limitations to this study. First, there are inherent limitations to using the NCDB as a tool to study patient outcomes, a topic which has previously been described in depth.[25] This study is retrospective in nature and is limited to specific patient, treatment, and facility characteristics collected by the NCDB. There are multiple factors and selection biases that may explain differences between high and low volume TFs that cannot be controlled for with this data. The database also does not include specific information on which systemic therapy regimen was used for each patient, which limits our ability to determine if specific treatment regimens are responsible for the observed volume-outcome associations. We were also unable to determine if patients were managed on clinical trials due to the lack of comprehensive clinical trial information in the NCDB, but it is possible that patients treated at high volume facilities that offered clinical trials had better survival regardless of clinical trial participation. We did apply increasingly stringent cohort definitions to select for patients whose disease and treatment status were more homogeneous, and the volume-outcome relationship established for the entire cohort remained stable. Unmeasured confounders could have affected the outcome or mediators (management metrics), which would be difficult to ascertain from this data. Cancer-specific mortality cannot be measured in the NCDB, though most deaths in this high-risk demographic are likely attributable to mRCC.

Notably, there is currently no accepted threshold defining a high volume mRCC treatment center. For pragmatic reasons, we defined high-volume TFs using the top 20^th percentile of mean number of mRCC patients treated per year, but no evidence exists to suggest that this volume threshold is inherently superior to other thresholds. However, we did analyze a threshold of top 9% based on conditional inference trees and showed an increasing benefit in survival as treatment volume increased continuously, indicating the survival benefit may be stronger for very high volume centers.

Conclusion

Patients treated at high volume TFs demonstrated a significantly improved OS than patients treated at lower volume TFs. This association was strengthened as TF volume increased. These findings may help define the optimal treatment environment for the management of patients with mRCC, and should also be a call to improve treatment outcomes at lower volume facilities.

Supplementary Material

1

Supplemental Figure 1. Distribution of TFs by mRCC treatment volume. Dashed line at 4.8 cases/year.

Supplemental Table 1. Multivariate analysis of patient characteristics associated with treatment at a high volume facility (Cohort A) (N = 41,386)