Identification and validation of radiographic enhancement for reliable differentiation of CD117(+) benign renal oncocytoma and chromophobe renal cell carcinoma.

Abstract

PURPOSE:

The diagnostic differential for CD117/KIT(+) oncocytic renal tumor biopsies is limited to benign renal oncocytoma (RO) vs. chromophobe renal cell carcinoma (ChRCC), however further differentiation is often challenging and requires surgical resection. We investigated clinical variables that might improve preoperative differentiation of CD117(+) RO vs. ChRCC to avoid the need for benign tumor resection.

EXPERIMENTAL DESIGN:

124 nephrectomy patients from a single institute with 133 RO or ChRCC tumors were studied. Patients from 2003–2012 comprised a retrospective cohort to identify clinical/radiographic variables associated with RO versus ChRCC. Prospective validation was performed among consecutive RO/ChRCC tumors resected from 2013–2017.

RESULTS:

Tumor size and younger age were associated with ChRCC, and multifocality with RO, however the most reliable variable for ChRCC vs. RO differentiation was the tumor:cortex peak early-phase enhancement ratio (PEER) using multiphase computer tomography (CT). Among 54 PEER-evaluable tumors in the retrospective cohort (19 CD117(+), 13 CD117(−), 22 CD117-untested), PEER classified each correctly as RO (PEER >0.50) or ChRCC (PEER ≤0.50), except four misclassified CD117(−) ChRCC variants. Prospective study of PEER confirmed 100% accuracy of RO/ChRCC classification among 22/22 additional CD117(+) tumors. Prospective inter-observer reproducibility was excellent for PEER scoring (intra-class correlation coefficient, ICC=0.97) and perfect for RO/ChRCC assignment (ICC=1.0).

CONCLUSIONS:

In the largest clinical comparison of RO vs. ChRCC to our knowledge, we identified and prospectively validated a reproducible radiographic measure that differentiates CD117(+) RO from ChRCC with potentially 100% accuracy. PEER may allow reliable biopsy-based diagnosis of CD117(+) RO, avoiding the need for diagnostic nephrectomy.

INTRODUCTION

Historically, renal tumors have been treated as renal cell carcinoma (RCC) until proven otherwise by surgical resection using partial or radical nephrectomy. Yet at least 10% of resected renal masses are benign.[1,2] The majority of resected benign tumors are asymptomatic renal oncocytomas (RO) with typically small size amenable to partial nephrectomy. However, partial nephrectomy is associated with overall and high-grade complication rates of 25–30% and 10–15%, respectively, in addition to a small decline in renal function and an estimated healthcare cost of $24,000 per resection (>$30,000 if complicated).[3–7] Based on >60,000 new kidney tumor diagnoses in the U.S. annually[8] of which >90% are RCC[9], and a 5% incidence (range 3–7%) of RO among resected renal cell tumors[10,11], RO resection can be estimated to incur an annual health care cost in the U.S. of nearly 100 million dollars. Reliable preoperative RO diagnosis would allow significant reductions in surgical resections, patient morbidity and healthcare costs.

Currently, preoperative differentiation of RO from RCC is aided by renal tumor biopsy, which has a high diagnostic rate, a low incidence of predominantly low grade complications, and a miniscule (<1:10,000) estimated risk of tumor seeding in contemporary studies.[12–17] Yet biopsy is still commonly deferred because RCC cannot be definitively ruled out when the biopsy favors RO.[18] The most common diagnostic challenge in this regard is the differentiation of RO from chromophobe RCC (ChRCC).[18–21] RO and ChRCC share classic oncocytic histology and also a remarkably similar molecular biomarker profile.[19,20,22,23] In addition, eosinophilia that is characteristic of RO can be mimicked by ChRCC, which has predominant eosinophilia in 40% of cases and at least some eosinophilic foci in most remaining cases.[20] Both tumor subtypes are relatively indolent and appropriate for active surveillance management. However whereas RO is benign, ChRCC has occasional metastatic potential and resection is warranted for select patients.[24]

Biopsy differentiation of renal tumor subtypes has been aided by contemporary immunohistochemical biomarker panels. Most (≥70%) but not all RO and ChRCC tumors express the protein biomarker, CD117/KIT, which is absent in other RCC subtypes and therefore narrows the differential diagnosis to RO vs. ChRCC.[23] Similarly, both RO and ChRCC lack classic biomarkers (e.g., CAIX, vimentin, AMACR) of clear cell RCC (CcRCC) and/or papillary RCC (PRCC).[22,23] Accordingly, while the diagnostic differential for an oncocytic renal tumor includes many renal cell tumor subtypes,[18] contemporary immunostain panels incorporating CD117 and other biomarkers reliably differentiate RO and ChRCC from other oncocytic RCC variants, including CcRCC and PRCC. When the diagnostic differential has been narrowed to RO versus ChRCC, the CK7 protein biomarker is commonly used to support ChRCC diagnosis. However, CK7 is occasionally inaccurate in this role,[22,25,26] and more reliable approaches are needed.[21] Radiographic approaches, such as computer tomography (CT) contrast enhancement measurement, have not yet shown utility in aiding preoperative RO diagnosis. However most prior radiographic studies have focused on the differentiation of RO from RCC subtypes collectively, overlooking the specific comparison of RO with the ChRCC subtype and the potential impact of occasional CD117(−) variants.[27–29]

Here we report the largest study to our knowledge comparing clinical features of RO and ChRCC, and the first study to also consider the impact of CD117 positivity. We hypothesized that certain clinical and/or radiographic features might allow reliable preoperative differentiation of these two renal tumor subtypes, particularly among classic CD117(+) tumors for which the ChRCC/RO question on biopsy is most relevant. Using both a retrospective discovery cohort and prospective validation cohort, we identified a simple CT enhancement measure that differentiated RO from ChRCC with 100% accuracy and perfect inter-observer reproducibility whenever CD117 was expressed; occasional inaccuracy was limited strictly to CD117(−) tumor variants. This generalizable approach may facilitate RO vs. ChRCC diagnosis among CD117(+) oncocytic tumor biopsies and negate the need for CD117(+) RO resection to rule out malignancy.

MATERIALS AND METHODS

Retrospective Patient Cohort

Institutional review board approval at Roswell Park Cancer Institute, a National Comprehensive Cancer Network institute, was obtained for this study. A prospectively maintained nephrectomy database was queried to identify all patients who underwent partial or radical nephrectomy for pathologically confirmed RO or ChRCC between August 2003 and December 2012. One patient with four hybrid RO-ChRCC tumors was excluded from analyses due to indeterminate malignant versus benign pathology.[30,31] CD117 and CK7 immunostain results were obtained from pathology reports, and classified as positive for moderate-to-strong/diffuse membranous stain, or as negative for absent or weak/focal (<10%) stain.[25] Hematoxylin- and eosin-stained slides from a subset of ChRCC tumors were scored retrospectively for eosinophilia tissue percentage (0–100%) and eosinophilia intensity (0–3+) by a genitourinary pathologist (BX).

Radiographic tumor measurements

Preoperative cross-sectional radiographic imaging for each tumor in the retrospective cohort was evaluated by two members of a committee comprised of three radiologists (CH, MP, CR) and one urologic oncologist with kidney cancer specialization (EK) while blinded to tumor histologic subtype. Scans were acquired from several different radiology facilities and no standardized imaging protocol was used. CT signal intensity (Hounsfield units, HU) of each tumor was measured in an early (arterial/cortical or venous/parenchymal) contrast phase and a delayed/excretory contrast phase for the peak enhancing portion of each tumor using a 1.0-cm diameter region-of-interest (ROI) circle; or a 1.0-cm × 0.5-cm elliptical ROI for small heterogeneous tumors to avoid hypo-enhancing portions. When multiple early CT phases were available for a patient, the earliest phase was used. Signal intensity of renal cortex adjacent to the tumor was measured using an elliptical ROI without specified dimensions. Tumor heterogeneity was estimated using a subjective 3-point scale (1- homogenous, 2- mild heterogeneity, 3- high heterogeneity).

Prospective validation

Clinical/radiographic variables that were significantly associated with RO or ChRCC in the retrospective analysis were evaluated prospectively between January 2013 and February 2017. All resected tumors from this period were stained prospectively for CD117 and CK7. Preoperative radiographic imaging for each resected tumor was evaluated independently by three observers (EK, MP, CR), who assigned ChRCC or RO classification based on radiographic assessment while blinded to the actual tumor subtype histology.

Statistical analyses

Patient and tumor characteristics were reported using the mean/range for continuous variables, and frequencies/relative frequencies for categorical variables. Comparisons were made using the Wilcoxon rank sum and Fisher’s exact tests for continuous and categorical variables, respectively. Inter-rater reliability was assessed using the intra-class correlation coefficient (ICC), where 0.90 is considered a priori as sufficient demonstration of reliability. All analyses were completed in SAS v9.4 (Cary, NC) at a significance level of 0.05.

RESULTS

Retrospective evaluation

87 patients (93 tumors) comprised the retrospective cohort, which included 53 RO patients (59 tumors) and 34 ChRCC patients (34 tumors). RO and ChRCC patients were similar for most preoperative features (Table 1). Younger age (<50 years) was significantly associated with ChRCC (80% specific), but only 24% sensitive for this subtype (Table 1a). Multifocality was significantly associated with RO (100% specific), but only 27% sensitive for this subtype (Table 1b). Smaller tumor sizes were enriched for RO, whereas larger tumor sizes and higher nephrometry scores were enriched for ChRCC (Table 1b).

Table 1. Clinicopathologic features of the retrospective cohort.

A)Patient features. B)Tumor features.

A.
	ChRCC	RO	P-value
Total patients, n (%)	34 (39%)	53 (61%)
Age, mean (range)	61.2 (28–84)	66.2 (35–89)	0.16
< 50, n (%)	8 (24%)	2 (4%)	0.012
Gender, n (%)
Male	20 (59%)	37 (70%)	0.36
Female	14 (41%)	16 (30%)
Race, n (%)
White	32 (94%)	49 (92%)	1.00
Black	1 (3%)	3 (6%)
Other	1 (3%)	1 (2%)
BMI, n (%)
≥ 30	16 (47%)	32 (62%)	0.27
≥ 35	10 (29%)	10 (19%)	0.31
Smoking History
Never smoker, n (%)	23 (68%)	27 (51%)	0.18
Any smoking, n (%)	11 (32%)	26 (49%)
Pack Years, mean (range)	24.9 (1–100)	25.6 (5–50)	0.31
Serum Creatinine, mean (range)	1.0 (0–1.9)	1.1 (0.8–2.4)	0.37
Charlson Index, mean (range)	2.5 (0–8)	2.5 (0–11)	0.98

B.
	ChRCC	RO	P-value
Total tumors, n (%)	34 (37%)	59 (63%)
Laterality, n (%)
Left	22 (65%)	35 (59%)	0.66
Right	12 (35%)	24 (41%)
Multifocality, n (%)	0 (0%)	16 (27%)	<0.001
Synchronous RCC	0 (0%)	3 (19%)
Synchronous RO	0 (0%)	11 (69%)
Synchronous AML	0 (0%)	2 (12%)
Tumor Diameter, cm
Mean (range)	6.4 (1.8–20.0)	3.4 (0.9–12.0)	<0.001
≤ 4, n (%)	13 (38%)	44 (75%)	<0.001
4.1–7, n (%)	11 (32%)	12 (20%)
> 7, n (%)	10 (29%)	3 (5%)
RENAL score
Total, mean (range)	8.3 (5–11)	7.3 (4–11)	0.022
R, mean (range)	1.9 (1–3)	1.3 (1–3)	0.001
E, mean (range)	1.6 (1–3)	1.7 (1–3)	0.47
N, mean (range)	2.6 (1–3)	2.2 (1–3)	0.022
A, n (%)
Anterior	13 (38%)	27 (46%)	0.65
Posterior	11 (32%)	14 (24%)
Neither	10 (29%)	18 (31%)
L, mean (range)	2.2 (1–3)	2.2 (1–3)	0.76
Endophytic %, mean (range)	51.1 (5–100)	56.2 (10–100)	0.40
Kidney Pole, n (%)
Upper	7 (21%)	13 (22%)	0.77
Middle	16 (47%)	31 (53%)
Lower	9 (26%)	14 (24%)
N/A	2 (6%)	1 (2%)
Stellate Scar, n (%)
Yes	2 (6%)	5 (8%)	1.00
No	26 (76%)	46 (78%)
Unknown	6 (18%)	8 (14%)
Heterogeneity score, mean (range)	2.0 (1–3)	1.9 (1–3)	0.62
CT Enhancement (HU), mean (range)
Tumor
Early phase	99.1 (53.5–186)	133.4 (70–250)	0.002
Net early phase	63.7 (17.5–162)	110.4 (48–225)	<0.001
Delayed phase	74.6 (39–146)	94.9 (41–180)	0.043
Net delayed phase	42.7 (16–122)	70.5 (15.5–155)	0.009
Tumor:Cortex
Early phase	0.63 (0.32–1.00)	0.80 (0.55–1.04)	0.001
Net early phase	0.45 (0.18–0.96)	0.77 (0.50–1.00)	<0.001
Delayed phase	0.53 (0.36–0.86)	0.70 (0.48–0.93)	<0.001
Net delayed phase	0.39 (0.16–0.80)	0.62 (0.23–0.95)	<0.001

The variable most reliably associated with RO vs. ChRCC diagnosis was the CT peak signal intensity within a tumor, particularly when expressed as a tumor:cortex ratio (Table 1b). Differences in tumor:cortex peak signal intensity between RO and ChRCC were greater when measured using net enhancement (i.e., contrast phase signal intensity minus non-contrast phase signal intensity) (Figure 1, Supplementary Figure 1a) compared to absolute signal intensity using the contrast phase only (Supplementary Figure 1b, c); and when measured using an early (arterial/cortical or venous/nephrogenic) contrast phase (Figure 1, Supplementary Figure 1b) compared to a delayed/excretory contrast phase (Supplementary Figure 1a, c). Combining these observations, we identified the most discriminating radiographic measurement to be the peak early-phase enhancement ratio (PEER) of the tumor:cortex (Figure 1). All RO tumors were relatively hyperenhancing with a tumor:cortex PEER >0.50 (Figure 1, 2a-d). In contrast, all but four ChRCC tumors were relatively hypoenhancing with a tumor:cortex PEER <0.50 (Figures 1, 2e-h). Intriguingly, all four ChRCC tumors with a tumor:cortex PEER >0.50 were CD117(−) variants (Figure 1, Supplementary Figure 2). Thus, a tumor:cortex PEER threshold value of 0.50 reliably separated RO (PEER >0.50) from ChRCC (PEER <0.50) for all tumors that were either CD117(+) or CD117-untested (Figure 1).

Figure 1. Peak early contrast phase enhancement (PEER) in RO versus ChRCC tumors of retrospective cohort patients according to CD117 expression.

The contrast CT peak signal intensity (HU) of each tumor calculated as (tumor_contrast-tumor_noncontrast)/(cortex_contrast-cortex_noncontrast) is plotted according to RO versus ChRCC histology and stratified by CD117 immunostain result. An early contrast phase includes either an arterial/cortical phase or venous/nephrogenic phase.

Figure 2. Representative early phase contrast CT images of CD117(+) RO/ChRCC tumors.

(A-D) Four representative CD117(+) RO tumors with tumor hyperenhancement that is similar to the enhancement level of the adjacent cortex (tumor:cortex PEER scores = 0.96, 0.92, 0.93, 0.71, respectively). (E-H) Four representative CD117(+) ChRCC tumors with hypoenhancement relative to the adjacent cortex (tumor:cortex PEER scores = 0.28, 0.27, 0.32, 0.27, respectively). Dotted circles and ellipses (ROI) denote the peak enhancing areas of the tumors and representative regions of renal cortex, respectively, used for tumor:cortex PEER calculation.

Prospective validation

Variables of age, tumor size, multifocality and tumor:cortex PEER were next evaluated prospectively over a four-year period for ability to classify 40 additional tumors (20 RO, 20 ChRCC) from 37 consecutive nephrectomy patients with RO or ChRCC. A total of 28/40 (14 RO, 14 ChRCC) tumors had preoperative multiphase CT scans available for PEER measurement, 22 of which were CD117(+) and 6 of which were CD117(−). All 22 CD117(+) tumors were correctly classified using a PEER threshold value of 0.50 by each of the three independent reviewers; whereas 3/6 CD117(−) tumors were misclassified by each reviewer using this approach (Figure 3a,b). Inter-observer reliability was excellent for PEER scoring (ICC= 0.97), and perfect for RO vs. ChRCC classification (ICC= 1.0) (Figure 3a). Multifocality correctly classified 6/6 CD117(+) tumors as RO; age <50 years correctly classified 4/5 CD117(+) tumors as ChRCC; size <2 cm correctly classified 4/5 CD117(+) tumors as RO; and size >7 cm correctly classified 4/5 CD117(+) tumors as ChRCC.

Figure 3. Prospective validation of tumor:cortex PEER score for assignment of RO versus ChRCC histology.

(A) Comparison of PEER scores from three independent physician reviewers for 28 consecutive RO/ChRCC tumors in the prospective cohort, indicating high inter-observer agreement (ICC =0.97) and perfect agreement for the assigned histology (ICC =1.0). (B) The mean tumor:cortex PEER score for each tumor in the prospective cohort is plotted according to CD117 expression. (C) PEER scores for all tumors in the combined retrospective and prospective cohorts are plotted according to CD117 expression. (D) PEER scores from part (C) after exclusion of tumors measured using a venous/nephrogenic rather than arterial/cortical contrast phase show improved separation of CD117(+) RO and CD117(+) ChRCC tumors. .

In a combined analysis of the prospective and retrospective cohorts, PEER classified RO vs. ChRCC correctly in 63/63 (100%) tumors without evidence of CD117(−) staining, including 41/41 cases confirmed to be CD117(+) (Table 2, Figure 3c). In contrast, the CK7 biomarker occasionally misclassified CD117(+) tumors, including 17% of CD117(+) ChRCC tumors (Table 2, Supplementary Figure 3). Neither PEER nor CK7 performed well among CD117(−) variants (Table 2). Although CD117(+) RO and CD117(+) ChRCC tumors were reliably separated by PEER using either the arterial/cortical or venous/nephrogenic phase, the former phase allowed greater separation (Figure 3d). Among CD117(+) ChRCC tumors evaluated by PEER, eosinophilia of moderate-to-strong intensity was present diffusely (≥80% of tissue) in close to one third of cases and at least focally in most cases; however, eosinophilia tended to be more prevalent among CD117(−) tumors (Supplementary Table 1). PEER values of four tumors from the excluded hybrid patient ranged from 0.44 to 0.47, each of which was CD117(+).

Table 2.

Overall accuracy of PEER and CK7 for ChRCC/RO classification in the combined retrospective and prospective cohorts.

	ChRCC		RO		Total
CD117 level	PEER (%)	CK7 (%)	PEER (%)	CK7 (%)	PEER (%)	CK7 (%)
Not tested	5/5 (100)	2/2 (100)	17/17 (100)	4/4 (100)	22/22 (100)	6/6 (100)
Positive	15/15 (100)	15/18 (83)	26/26 (100)	28/28 (100)	41/41 (100)	43/46 (93)
Negative	3/10 (30)	7/10 (70)	9/9 (100)	7/10 (70)	12/19 (63)	14/20 (70)

DISCUSSION

Currently, most benign RO tumor patients continue to undergo unnecessary partial or radical nephrectomy because biopsy alone is inadequate to definitively rule out cancer. Improved preoperative diagnosis of benign RO tumors would avoid a need for diagnostic RO resection and thereby allow significant reductions in patient morbidity and healthcare costs.[3–7] The greatest challenge for non-surgical diagnosis of RO has been the difficult differentiation of RO from ChRCC on percutaneous needle-core biopsy, due to significant overlap in histology and biomarker profile, including common expression of CD117.[19–23] Although both tumor types confer excellent candidacy for active surveillance management, the latter has occasional lethal potential and warrants resection in select patients.

The current study investigated clinical variables which might facilitate non-operative diagnostic differentiation of RO from ChRCC, including the potential impact of CD117 expression status. In the largest clinical comparison of ChRCC versus RO to our knowledge, we retrospectively identified a simple CT enhancement measure (tumor:cortex PEER) that differentiated RO from ChRCC whenever CD117(+) with 100% accuracy. All RO tumors in the retrospective cohort had hyperenhancement (PEER >0.50) regardless of CD117 positivity, whereas ChRCC tumors typically had hypoenhancement (PEER <0.50) except for a minority of cases which, to our surprise, were all CD117(−) variants. To validate this finding, all resected RO or ChRCC tumors in the past four years at our institution were evaluated prospectively by CD117/CK7 expression and PEER, as measured by three independent reviewers blinded to whether each tumor was RO or ChRCC. All CD117(+) tumors in this prospective cohort were correctly classified as RO or ChRCC by each reviewer using PEER; and as in the retrospective cohort, only CD117(−) ChRCC variants were misclassified. In combined retrospective and prospective cohort analyses, PEER classified RO vs. ChRCC correctly when not CD117(−) in 63/63 (100%) evaluable cases, including 41/41 confirmed CD117(+) cases. These findings indicate PEER is an accurate tool for differentiating CD117(+) RO from ChRCC.

Currently, the CD117 biomarker is used in contemporary pathology to narrow the diagnostic differential of an oncocytic renal tumor to either RO or ChRCC.[22,23] Histomorphology is the gold-standard for RO vs. ChRCC differentiation but may be inadequately represented in small biopsy tissues. Furthermore, eosinophilia present to some degree in most ChRCC tumors and predominant in 40% of cases,[20] can mimic classic RO eosinophilia when captured on biopsy. In this scenario, the CK7 clinical biomarker can be helpful to support ChRCC diagnosis, however some ChRCC variants lack CK7 expression.[22,25,26] Accordingly, definitive RO diagnosis cannot be confirmed based on CK7 absence alone in a biopsy, and is often deferred to surgical resection pathology.[20,22,26] In the current study, PEER outperformed CK7, which misclassified 17% of CD117(+) ChRCC cases as RO, consistent with prior reports.[22,25,26] Although neither PEER nor CK7 was very accurate among CD117(−) tumors, this scenario has less relevance to the clinical challenge of ChRCC vs. RO, since the diagnostic differential of a CD117(−) oncocytic tumor biopsy is not limited to these two subtypes and includes several other RCC subtypes as well.[18] Only when an oncocytic tumor is CD117(+) does the RO vs. ChRCC distinction become critical, and in this scenario PEER achieves high if not perfect accuracy.

Figure 4 illustrates a proposed algorithm that combines PEER with CD117 immunostain in the management of oncocytic renal tumor biopsy patients. PEER does not negate the need for biopsy in preoperative RO diagnosis, but may improve biopsy accuracy so that diagnostic resection can be avoided for CD117(+) RO tumors, which account for most RO cases. For CD117(−) RO tumors, PEER does not facilitate diagnosis and other novel diagnostic approaches are needed to avoid surgery. In our prospective series, CD117(−) tumors accounted for 21% of all RO/ChRCC cases, which is consistent with reported incidences.[23]

Figure 4. Management algorithm for an oncocytic renal tumor biopsy patient using CD117 immunostain and tumor:cortex PEER.

An oncocytic renal tumor biopsy should undergo CD117 and CK7 immunostain; multiple biopsy cores should be obtained to differentiate diffuse versus focal expression. If CD117(−), the differential is large and includes CcRCC and PRCC, in addition to less common RCC subtypes, and resection is favored if tumor size and patient health are appropriate. If CD117(+), the differential is limited to ChRCC and RO, and a tumor:cortex PEER score can be measured using CT imaging that includes an early contrast phase and an unenhanced phase. For tumor:cortex PEER ≤0.50, the CD117(+) tumor is predicted to be ChRCC and can be resected or surveyed depending on tumor size and patient health/life expectancy. A PEER score of 0.51–0.55 was not observed in this study for any CD117(+) tumor; a hypothetical CD117(+) tumor with PEER score falling within this range based on a venous/nephrogenic phase should undergo repeat CT imaging with a true arterial/cortical phase, which may yield a higher PEER value if the tumor is RO. If PEER remains between 0.51–0.55 using an arterial/cortical phase, consider CK7 immunostain result and other clinical features (patient age, tumor size/focality) or defer to diagnostic resection.

*active surveillance may be appropriate in select patients with small tumors and/or limited life expectancy

**this scenario was not encountered in the current study

Additional study is needed to validate the management algorithm in Figure 4 and confirm the added value of PEER in the clinic. Goals of this future research should include verification of the optimal PEER cutoff value for ChRCC vs. RO differentiation. The cut-off PEER value of 0.50 in our prospective cohort was arbitrarily selected based on our retrospective observations, however a cut-off PEER value of 0.55 too would have achieved 100% diagnostic accuracy in both the retrospective and prospective cohorts. Because PEER measurements never fell within the 0.51–55 range for any CD117(+) tumor in this study, additional investigation is needed to determine if this scenario ever occurs; and if so, whether such tumors are RO (in which case a 0.50 cut-off would be optimal) or ChRCC (in which case a cut-off of 0.55 would be optimal). In the interim, for a hypothetical CD117(+) tumor with PEER value of 0.51–0.55 based only on a venous/nephrogenic phase, a repeat CT scan with an arterial/cortical phase should be considered since this phase augmented slightly the separation RO vs. ChRCC tumors in this study when CD117(+) (Figures 3D, 4).

An association between CD117 positivity and radiographic enhancement level is not previously reported to our knowledge. Our discovery that CD117 absence correlates with frequent ChRCC hyperenhancement is unexpected based on known CD117 function. CD117 is a tyrosine kinase membrane receptor for the ligand, stem cell factor (SCF), and positively regulates a diversity of cell processes, including proliferation, differentiation, migration and angiogenesis.[32,33] CD117 interacts with several cell signaling pathways implicated in cancer development, such as those of PI3K/Akt, Src family kinases and Ras-Erk.[34] The role of CD117 in ChRCC/RO tumorigenesis, however, is unknown. Given its known function in increasing microvessel density and angiogenesis,[33] absence of CD117 expression might be expected to decrease vascularity/enhancement, which is opposite to observations in this study. CD117 is a differentiation marker for the nephron distal renal tubule,[35] from which RO and ChRCC tumors are believed to originate, and it is possible that its loss in ChRCC tumors might reflect dedifferentiation to a hypervascular and potentially more aggressive variant. Also worthy of note is our observation that CD117(−) tumors have a high rate of predominant eosinophilia. Together, these findings suggest the existence of a ChRCC biologic variant characterized by CD117 negativity, frequent atypical hypervascularity, and common eosinophilia. Study is currently ongoing to characterize the clinicopathologic features and prognostic significance of these CD117(−) ChRCC variants.

Prior studies that compared clinical features of RO and RCC have generally overlooked the direct RO vs. ChRCC comparison,[1,36,37] which is most relevant to the contemporary diagnostic challenge of a CD117(+) oncocytic tumor biopsy.[28] Furthermore, no study comparing radiographic features of RO and ChRCC has considered CD117 expression. Bird et al studied CT enhancement in various renal cell tumor types, which included 12 RO and 5 ChRCC.[29] RO enhancement was three times higher than ChRCC enhancement, with greatest difference in the arterial phase, as in the current study. Gorin et al. has reported that a sestamibi scan has high sensitivity for RO but cannot reliably rule out ChRCC, which occasionally gives a positive result.[38] Canvasser et al. reported recently an algorithm using magnetic resonance imaging (MRI) signal intensity to predict CcRCC histology.[39] 9 ChRCC and 7 RO tumors of unspecified CD117 status were studied and showed overlap in MRI enhancement.

Novel molecular biomarkers may help differentiate RO and ChRCC on biopsy, but have not yet translated clinically.[21,40–44] Kauffman et al reported CD82/KAI1 as a sensitive ChRCC marker that is typically absent in RO, a finding validated subsequently by Truong et al.[42,45] Additional promising biomarker candidates to distinguish RO and ChRCC include S100A1 and cyclin D1[46,47]; whereas other potential biomarkers involve more complicated definitions of positivity or molecular approaches.[40,41,43,44] In contrast, a PEER ratio can be determined rapidly and cheaply using routine multiphase CT imaging, which has already been obtained for many patients prior to the CD117(+) oncocytic tumor biopsy. A strength of this study is that CT scans were from different facilities without protocol standardization. Although arterial phases gave the best separation of RO from ChRCC when CD117(+), either the arterial or venous phase was 100% accurate. Hence, the specific CT protocol and precise timing of the early contrast phase does not appear to be critical, which should facilitate generalized PEER usage.

A historical concern with active surveillance of biopsy-favored RO is that the tumor might simultaneously harbor ChRCC, as in rare hybrid oncocytic tumors (HOCT).[48] However, to our knowledge, HOCT metastatic behavior has never been confirmed, raising the assertion by some investigators that these hybrid tumors might be benign.[30] Consistent with this possibility, Pote et al recently concluded that HOCT is genetically more similar to RO than ChRCC and may therefore be a RO variant.[31] Regardless, given its indolent clinical behavior and lack of known metastatic potential, the rare possibility of HOCT should not justify routine resection of tumors with RO diagnosis favored on biopsy, since HOCT too is appropriate for surveillance.[49]

In addition to PEER outcomes, we found that younger age and larger tumor size enriched for ChRCC, while multifocality and smaller tumor size enriched for RO. Although an age cutoff of <50 years had high specificity for ChRCC, the optimal age cutoff to discern these two subtypes needs to be better defined in larger studies with statistical analyses designed to identify this cutoff. The association of ChRCC with young age is consistent with recent observations from Daugherty et al, who noted ChRCC tumors to be the most common non-clear RCC subtype in young white females.[50] In contrast to PEER, however, each of these other variables (age, size, focality) was poorly sensitive for either subtype; and with the exception of multifocality, none was 100% specific. Nevertheless, these variables might be clinically helpful to corroborate PEER findings among CD117(+) tumors with a PEER value near the 0.50 threshold (Figure 4). Caution must be exercised with multifocal tumors, since RO diagnosis for one tumor does not exclude malignancy in another.[48] Additional biopsy should be considered for any synchronous tumor with radiographic features (e.g., PEER, size, growth rate) that are discordant from the biopsied RO tumor.

Consistently high PEER values for RO, regardless of CD117 positivity, suggest that a tumor:cortex PEER <0.50 might rule out RO without the need for biopsy. Fat-poor angiomyolipoma (AML) represents the other main benign renal tumor besides RO that often undergoes unnecessary diagnostic resection; although unlike RO, AML diagnosis by percutaneous biopsy is uncomplicated. Interestingly, AML is characterized by hypervascularity that might be expected to yield uniformly high PEER values, similar to RO. Future studies are therefore warranted to determine the accuracy of a low tumor:cortex PEER value to rule out any common benign histology (i.e., RO and fat-poor AML) without the need for biopsy.

A limitation of this study is that CD117 staining was determined with surgically resected tumors, whereas PEER is proposed to guide clinical management based on the CD117 status of biopsy tissues (Figure 4). Additional investigation is therefore needed to confirm concordance between CD117 staining on biopsy tissue and resected specimens. In the interim, a CD117(+) stain result based on only a single biopsy core should be interpreted with caution since this result may not rule out a CD117(−) tumor with focal positivity. To avoid this scenario, multiple tumor biopsies are recommended to ensure adequate tumor sampling that reliably differentiates focal vs. diffuse CD117 expression.

Additional study limitations include a single institute setting and lack of external validation. Although the study is the largest clinical comparison of ChRCC and RO to our knowledge, the cohort is nevertheless limited in size, and many patients were lacking preoperative multiphase CT scans necessary to measure PEER. Furthermore, MRI enhancement was not evaluated in this study due to a limited number of available cases at our institute and warrants future investigation. Finally, many tumors in the retrospective cohort were not tested for CD117 expression; however, all CD117-untested tumors were classified correctly by PEER, so revealing their CD117 status would not have changed the 100% accuracy of PEER among CD117(+) tumors.

CONCLUSIONS

A historical challenge to reliable preoperative RO diagnosis using biopsy has been differentiation from ChRCC, which shares common CD117 biomarker expression with RO but warrants resection in select cases due to occasional lethal potential. We identified and prospectively validated a reproducible and easily generalizable radiographic measure (tumor:cortex PEER) that uses standard multiphase CT imaging to differentiate CD117(+) RO from CD117(+) ChRCC with potentially 100% accuracy. Age, tumor size, focality and CK7 immunostain may corroborate RO/ChRCC diagnosis but are alone not 100% reliable. In contrast, a high PEER value may reliably confirm RO diagnosis whenever an oncocytic renal tumor biopsy is CD117(+), and thereby allow diagnostic resection to be avoided. The number of patients in this study with known CD117(+) tumors that were evaluable with PEER was limited (41 cases), and external validation in larger cohorts will be helpful to confirm the clinical value of PEER for CD117(+) RO diagnosis.

Supplementary Material

2

Supplemental Figure 1. Non-PEER CT peak signal intensity/enhancement approaches in RO versus ChRCC tumors according to CD117 expression.(A) Net peak enhancement using the delayed/excretory contrast phase relative to the unenhanced phase; (B) Absolute peak signal intensity using an early (arterial/cortical or venous/nephrogenic) contrast phase without subtraction of the unenhanced phase signal intensity; (C) Absolute peak signal intensity using the delayed/excretory contrast phase without subtraction of the unenhanced phase signal intensity.

3

Supplemental Figure 2. Representative CT imaging of CD117(−) RO/ChRCC tumors using an early contrast phase. (A) This representative CD117(−) RO tumor has relative hyperenhancement (tumor:cortex PEER 0.79) that is similar to CD117(+) RO tumors. (B) This representative CD117(−) ChRCC tumor has relative hyperenhancement (tumor:cortex PEER 0.76) that is similar to RO tumors and in contrast to CD117(+) ChRCC tumors.

4

Supplemental Figure 3. Representative examples of CD117 and CK immunostaining in ChRCC and RO tumors. Representative staining images are shown at 100X magnification. (A-B) Hematoxylin & Eosin staining for (A) ChRCC and (B) RO. (C-D) Typical strong/diffuse CD117 staining in (C) ChRCC and (D) RO. (E-F) Atypical focal/weak CD117 staining in (E) ChRCC and (F) RO. (G-H) Typical CK7 staining that is (G) strong/diffuse in ChRCC and (H) negative in RO. (I-J) Atypical CK7 staining that is (I) focal/weak in ChRCC and (J) strong/diffuse in RO.

Statement of Translational Relevance:

Current inability to reliably differentiate benign renal oncocytomas (RO) from renal cell carcinoma (RCC) using needle core biopsy remains the primary reason behind routine surgical resection of these benign tumors, which is associated with frequent perioperative morbidity and an estimated annual health care cost approaching 100 million dollars in the U.S. alone. Historically, the greatest challenge for preoperative RO diagnosis using biopsy has been the difficult differentiation of this benign tumor from the chromophobe RCC (ChRCC) subtype due to significant histologic and molecular overlap, including common expression of the CD117 biomarker. Here we retrospectively identified and prospectively validated a simple radiographic measurement (“PEER”) that achieved 100% (41/41 cases) accuracy in differentiating CD117(+) RO from CD117(+) ChRCC. In the setting of a CD117(+) renal tumor biopsy, a high PEER value may reliably confirm RO diagnosis without the need for surgical resection, allowing significant reductions in patient morbidity and healthcare costs.