MolDX: Transcriptional Biomarkers for Therapeutic Decision-Making in Renal Carcinoma

Title XVIII of the Social Security Act, §1862(a)(1)(A) allows coverage and payment for only those services that are considered to be reasonable and necessary.

42 CFR §410.32(a) Diagnostic x-ray tests, diagnostic laboratory tests, and other diagnostic tests: Conditions

CMS Internet-Only Manual, Pub. 100-02, Medicare Benefit Policy Manual, Chapter 15, §80 Requirements for Diagnostic X-Ray, Diagnostic Laboratory, and Other Diagnostic Tests, §80.1.1 Certification Changes

Molecular and proteomic transcriptional biomarkers for therapeutic decision-making in renal carcinoma are currently non-covered by this contractor. However, should evidence be further developed to meet reasonable and necessary standards, such tests may be considered for coverage when ALL the following criteria are met:

1. The patient is being managed for a diagnosis of Renal Cell Carcinoma (RCC) according to national or international consensus guidelines (e.g., National Comprehensive Cancer Network (NCCN); American Urological Association (AUA) guidelines).
2. The patient is a candidate for multiple potential treatments and at least one of the following is true:
1. The therapies under consideration have varied levels of intensity based on a consensus guideline, and the physician and patient must decide among these treatments OR
2. The test has shown that it predicts response to a specific therapeutic intervention among accepted therapy options based on nationally recognized society consensus guidelines (i.e., National Comprehensive Cancer Network [NCCN], American Society of Clinical Oncology [ASCO], Society of Urologic Oncology [SUO], or American Urological Association [AUA]).
3. The patient has not been previously tested with the same test or other comparable test for the same intended use.
4. The test demonstrates significant improvement in accuracy for identifying risk of recurrence or metastasis or for predicting a therapeutic response compared to available information in the clinical medical record.
5. Testing is performed according to the intended use of the test in the intended patient population for which the test was developed and validated.
6. If the test relies on an algorithm, the algorithm must be validated in a cohort that is not a development cohort for the algorithm.
7. The analytes measured have demonstrated clinical validity and clinical utility in the peer-reviewed published literature, establishing a clear and significant biological basis for stratifying patients and subsequently selecting (either positively or negatively) their clinical management decision within a clearly defined population.
8. Analytical validity, clinical validity, and clinical utility are assessed as part of a successful technical assessment (TA) by the Molecular Diagnostic Services Program (MolDX^®).
9. The test demonstrates equal or superior performance in risk stratification or therapy response prediction to any other biomarker test that has already met the criteria above.

Note: Diagnostic tests using next-generation sequencing (NGS) for tumor mutation profiling are not within scope of this policy and are governed according to the criteria in LCD L38067, MolDX: Next-Generation Sequencing for Solid Tumors.

Introduction

There are about 80,000 new diagnoses of kidney and renal pelvis cancers annually (3-5% of all cancers) in the United States, with a male-to-female ratio of 2:1.¹ The five-year overall relative survival is 79.2%.¹ Approximately 70% of patients present with stage I disease at diagnosis, with an excellent (~94%) five-year survival rate; however, for the ~15% who present with distant metastases, five-year survival decreases to 20%.^1,2

The incidence of new kidney and renal pelvic masses is increasing,¹ largely driven by small renal masses (SRMs), which are tumors <4 cm. The majority of these are detected incidentally, with the widespread use of abdominal imaging leading to incidental cancer detection in 37% to 61% of cases.^2,3Approximately 26% of SRMs are benign and would likely never require therapy. Notably, the “classic triad” of symptoms (hematuria; flank pain; abdominal mass) occurs in <10% of patients and is associated with locally advanced or metastatic cancer.²

The divergence of a decreasing mortality rate alongside rising kidney cancer incidence¹ suggests the possibility of overdiagnosis and overtreatment for some patients, particularly regarding SRMs. Definitively identifying whether a renal mass is malignant remains difficult with current diagnostic tools, resulting in documented overtreatment.⁴ Notably, benign masses account for approximately 15-30% of partial and 5% of radical nephrectomies; 8.5% of inpatient renal surgical admissions are performed for benign masses.^5-7

The prognosis of patients with early-stage, localized disease is primarily determined by tumor stage, with tumor size, grade, and histology also contributing substantially.^2,8-10 Early-stage, low grade kidney cancer has an estimated recurrence rate of ~5% after nephrectomy.² However, 20-30% of patients with high-grade localized tumors experience relapse after surgical excision.⁹ High risk features across tumor stages additionally include the presence of invasive disease, nodal involvement, and sarcomatoid differentiation.²

Over the past decades, several risk models and nomograms using clinical and pathological factors have been developed to risk-stratify patients with localized renal cell carcinoma, including the Stage, Size, Grade and Necrosis (SSIGN) score, UCLA Integrated Staging System (UISS), and the Leibovich and Karakiewicz models. A recent post hoc analysis utilizing prospective data evaluated these risk models in patients from the ASSURE trial¹¹ (a large study assessing the benefit of select targeted tyrosine kinase inhibitors (TKIs) compared with placebo in the adjuvant setting in patients with intermediate- or high-risk resected localized RCC), revealing a decline in their discriminatory performance. When evaluated within this contemporary trial population, the models’ concordance statistics (C-indices) were markedly lower than the estimates reported in previous retrospective datasets.¹² Among the models evaluated, those that included tumor biology factors such as tumor stage and grade (e.g. Leibovich and SSIGN) tended to outperform those that included clinical symptoms at presentation (e.g. UISS, Cindolo, Karakiewicz, Yaycioglu).¹² The SSIGN score performed best, with a concordance statistic (C-index) for survival outcomes of 0.688; 95% confidence interval [CI] 0.686- 0.689), whereas the UISS performed worst (C-index 0.556; 95% CI, 0.555 - 0.557).¹² Notably, prior published C-indices for SSIGN and UISS were 0.76-0.88 and 0.64-0.86, respectively, demonstrating a substantial decrease in each model’s discriminatory performance in this contemporary clinical trial cohort.¹² Further, some of the models only marginally outperformed the standard Tumor size, Node involvement, and Metastasis cancer staging system (TNM) and all varied in their discriminatory performance over time, showing the best performance within the first 2 years after diagnosis. This time-dependent variation is likely driven by biological factors that contribute to early vs. late recurrences as well as a post-resection event distribution in RCC that is heavily skewed toward early recurrence.¹² The authors of the study caution that many of the adjuvant therapy trials have relied on these models for study design and patient eligibility. Notably, many of these models were developed using retrospective data limited to patients with only clear cell RCC histology and include outdated TNM staging criteria.¹² Both the AUA and NCCN recommend risk stratification based on TNM staging, and do not endorse these prognostic risk models in early-stage, localized RCC.^8,9 The AUA specifically classifies patients who have been treated surgically into four classes of risk, from “Low” to “Very High” on the basis of stage, grade, and histology; this risk assignment serves as the basis for follow-up surveillance and management protocols.⁸ NCCN guidelines also endorse consideration of a more rigorous approach to surveillance and management for patients with high-grade tumors.⁹

Conversely, in metastatic RCC (mRCC), a hetergeneous disease with varying rates of progression and response to therapy, NCCN guidelines stratify treatment recommendations based on histology as well as risk group assignment based on the use of specific prognostic models, specifically the International mRCC Database Consortium (IMDC) and Memorial Sloan Kettering Cancer Center (MSKCC).⁹ These models share several of the same clinical variables (Karnofsky score, hemoglobin, corrected calcium, and diagnosis to treatment time) and categorize patients with mRCC into favorable, intermediate, or poor risk. Further, IMDC may also be predictive of immunotherapy response.⁹ The prospective analysis described above that evaluated prognostic risk models using patients from the ASSURE trial found MSKCC’s prior published vs. current study C-indices to be 0.79-0.82 and 0.652 (95% CI, 0.650-0.653), respectively.¹² MSKCC underestimated the recurrence rates of high-risk patients, although it accurately predicted 5-year progression-free survival (PFS) for low- and intermediate-risk individuals.¹²

Approximately 80% of kidney cancers are renal cortical tumors known as renal cell carcinoma (RCC), the most common subtype (~70%) being clear cell RCC (ccRCC).^2,9 ccRCC is generally considered more aggressive than papillary (pRCC) (~15%) and chromophobe (chRCC) (~5%) RCC, though a large contemporary analysis of pooled data from the SORCE (n = 1689) and ASSURE (n = 1853) phase 3 trials found that survival of patients with pRCC and ccRCC was similar.¹³ Further, this analysis of pooled data found that risk of relapse was influenced by degree of risk according to the 2003 Leibovich criteria (i.e., patients with intermediate and high-risk pRCC relapsed earlier and exhibited poorer prognosis than previously reported for the subtype), though the authors note that their study focused on higher-risk patients who underwent radical nephrectomy (RN), differentiating it from some earlier studies that included more patients with T1 tumors who underwent partial nephrectomy (PN).¹³

While most cases of RCC are sporadic, an inherited predisposition should be considered for patients with RCC diagnosed at age less than 50 years, bilateral RCC, or multiple tumors in one kidney. Loss of the von Hippel Lindau (VHL) tumor suppressor gene occurs in 45-90% of ccRCC, though it is also seen in other types of RCC as well as in benign renal oncocytomas.² Genomic alterations in VHL are also a prominent feature of VHL Syndrome. Although it is the most common hereditary RCC syndrome, VHL Syndrome is rare overall, and only 5-16% of stage III or IV cases are secondary to any hereditary renal cancer syndromes.²

Diagnosis and decision making is currently dependent on standard modalities. Imaging remains the mainstay for RCC diagnosis and is used by physicians to guide the decision between patient intervention or surveillance.² While it has been limited in its ability to differentiate benign from aggressive malignant tumors and ccRCC from other RCC subtypes (e.g. differentiation by multiphase computed tomography (CT) of ccRCC from other RCC subtypes has been reported with an accuracy of 75%, sensitivity of 64%, and specificity of 87%),¹⁴ particularly among SRMs, the literature around optimization of imaging for this purpose continues to develop, and involves the optimization of different modalities (e.g. CT vs. magnetic resonance imaging (MRI)).^2,4 Additionally, renal mass biopsy (RMB) is an important adjunct for patients with a suspicious renal mass detected by imaging (generally these are masses that are solid rather than cystic and often ≥4 cm in size) that allows for risk stratification and tailoring of management options. However, despite high diagnostic accuracy (sensitivity 96-99%, specificity 94-96%, and diagnostic rate ~86%), its use has been limited for various reasons including concerns regarding sampling error and the risk of complications such as bleeding, though the overall complication rate is low (<10%). Contemporary data support a greater use of RMB in select circumstances.^3,6,15-17 Further, the use of RMB has been associated with fewer RNs for benign or indolent disease, particularly for certain renal masses; specifically, patients with clinical stage T1b solid renal masses identified by imaging who had a RMB had a lower likelihood of RN (odds ratio (OR): 0.47, CI: 0.31-0.72, P < .0001) than those without a RMB. The risk-adjusted RN rate for T1b renal masses was 41.4% without RMB vs 27.8% with RMB; therefore, 7.4 RMBs would be needed to avoid 1 RN for benign or indolent disease.¹⁸ Moreover, benign pathology in resection specimens was significantly (p < .0001) more common when RMB was not performed compared to when it was performed prior to surgery: 14.8% vs 7.2% of PNs and 10.2% vs 1.7% of RNs.¹⁸ Another study found a significantly lower benign resection rate (3.2%) after use of core needle biopsy (CNB) compared to the national average (> 30%).¹⁶

Multiple options are available for the management of localized renal masses including surgery, thermal ablation (TA), and active surveillance (AS). Surgery is considered one of the preferred approaches for renal tumors confined to the kidney (with PN favored over RN when possible, particularly for SRMs). Although AS is a recognized management option for some patients and is supported by AUA guidelines, the absence of reliable clinical and imaging-based predictive markers of tumor aggressiveness has limited its use, particularly given the concurrent limited use of RMB in routine practice.^8,19 Adjuvant therapy is recommended after nephrectomy for localized high-risk RCC,^3,5 and systemic therapy is recommended for advanced and metastatic tumors.⁹

Biomarkers

Inclusion of genetic and immune signatures have been studied to further refine prognostic and predictive risk in RCC.^12,20 Some have been evaluated in localized vs. advanced disease, some to predict recurrence post-nephrectomy, and others to predict response to various systemic therapies in the adjuvant or metastatic settings.²⁰

Transcriptional biomarkers have been developed to help risk stratify RCC for purposes of optimizing patient management. One such test, based on gene expression microarray, used consensus clustering data to identify two subtypes of ccRCC with different prognostic implications for disease-specific survival (DSS): “good risk” clear cell type A (ccA) and “poor risk” type B (ccB).²¹ Using this foundation, a 34-gene classifier (ClearCode34) was developed to further define and validate these subtypes in formalin-fixed paraffin-embedded (FFPE) tissues from patients with non-metastatic ccRCC tumors (primarily TNM stages I-III).²² Patients with ccB tumors experienced relapse after nephrectomy more frequently (hazard ratio (HR): 2.1; 95% CI, 1.3–3.4; p = 0.001) and had higher risk of both cancer-specific mortality (HR: 3.0; 95% CI, 1.3–7.0; p = 0.005) and overall mortality (HR: 2.2; 95% CI, 1.3–3.6; p = 0.001) compared to patients classified as ccA; there were 7 deaths (10%) in ccA and 25 (28%) in ccB signatures.²² While the ccA/ccB signature remained a prognostic factor even after adjustment for Fuhrman grade and stage, classification using these standard clinicopathologic (CP) variables were the most highly significant independent variables for predicting recurrence-free survival (RFS): Fuhrman grade (I/II vs. greater; p < 0.0001) and stage (stage I vs. greater; p =0.0007).²² A model incorporating the gene classifier with the CP variables better predicted disease-specific events (RFS and cancer-specific survival (CSS)) and was additive independently of both the UISS and SSIGN prognostic models.²² However, compared to prior evaluations, this study found an under-performance of the established clinical risk predictors/nomograms; to that effect, the authors acknowledged that their cohort of patients had different demographic features compared to prior cohorts.²³ For example, all of the patients not only underwent surgery but also had tumors large and solid enough to contribute tissue for molecular analyses, thereby excluding those with small or cystic tumors that otherwise would be expected to contribute to the clinical risk assessment. Consequently, the study investigators expressed caution regarding the use of such classifiers without further validation.²³

In an independent validation using a cohort of 350 ccRCC patients from The Cancer Genome Atlas (TCGA) consortium, both tumor stage and the ccA/ccB signature remained significant in a multivariable analysis (MVA).²⁴ The ccB signature was associated with a worse prognosis in patients with stage I (HR > 10; p < 0.001), stage II/III (HR: 3.03; p = 0.003), and stage IV ccRCCs (HR: 2.15; p = 0.015).²⁴ The signature was also significant in the MVA with the established SSIGN prediction model in a subgroup of patients.²⁴ However, tumor stage was also an independent predictor of CSS in MVA (tumor grade was also significant in the univariate analysis) and importantly, the signature could not be compared with other clinical nomograms because data on essential parameters were not available for most of the patients.²⁴ Further, only 1% of patients had Grade 1 tumors. Importantly, the study found heterogeneous expression patterns, with ccA and ccB signatures coexisting in 8 of 10 cases of stage II–IV ccRCCs evaluated across multiple tumor regions; only two tumors homogeneously expressed the ccA signature.²⁴ The authors acknowledge the lack of ability to interpret these findings, as it remains unknown whether a tumor with a small ccB component has a similarly poor prognosis to an identically sized tumor dominated by the ccB signature, or whether the absolute size of the poor-risk clone is a more important indicator. Finally, the authors acknowledged that prognostic markers were of limited clinical utility in ccRCC due to the absence (at the time of the study) of effective adjuvant strategies.²⁴

The cell cycle progression (CCP) score is another tissue-based RNA expression signature comprising genes involved in the cell cycle and implicated in tumor biology. The score is calculated based on the unweighted average expression of 31 cell cycle genes normalized to the expression of 15 housekeeping genes, and ranges from –3 to 3, with a one-unit increase representing doubling of expression.²⁵ In a multi-institutional study by Michigan Medicine evaluating 565 patients with localized pT1-T3 RCC (including pRCC, chRCC and ccRCC histologies) following RN, a broad range of CCP scores was observed within each pathological stage. In an MVA, the score was found to be an independent predictor of recurrence (HR per interquartile range (IQR) 1.60; 95% CI 1.17–2.19, p < 0.001).²⁵ However, several CP variables also retained independent prognostic significance on multivariate analysis (tumor stage (HR 4.87 [95% CI 2.19-10.85], p < 0.001); tumor size (HR 1.18 [95% CI 1.09-1.27], p < 0.001); lymphovascular invasion (HR 3.38 [95% CI 1.86-6.15], p < 0.001); Karakiewicz Score (HR 8.20 [95% CI 4.84-23.61], p < 0.001).²⁵ Furthermore, while the combined R-CCP score (integrating both CCP and Karakiewicz scores) had a c-index of 0.87, this represented a marginal increase in discriminatory performance over the baseline Karakiewicz nomogram (c-index 0.84) for stratifying DSM at 5 years.²⁵ Further, the event rate was low (only 68 patients (12%) recurred and 32 (5.6%) died of RCC within 5 years of nephrectomy) and all patients included underwent RN; thus, any potential use of the score to make management decisions in untreated patients would require interrogation of RMB specimens. A retrospective, multi-institutional cohort study (also by the group at Michigan Medicine) of patients who underwent RMB followed by surgery (PN or RN) found that the CCP score obtained from biopsy specimens was significantly associated with adverse pathology (AP) when modeled both as a binary (OR: 2.44 for CCP score >0, p = 0.02) and a continuous (OR: 1.72 per one unit increase, p = 0.04) variable, when added to a baseline model including age, sex, race, lesion size, biopsy grade, and histology; notably, in the baseline model, AP was also significantly associated with male sex, increased lesion size on imaging, and high tumor grade.²⁶ Area under the curve (AUC) slightly improved from 0.73 in the baseline model to 0.75 and 0.76 in models incorporating the CCP score.²⁶ In a sub-analysis of patients with low-grade tumors on biopsy (n = 175), a biopsy CCP score of >0 was associated with 2.52-fold increased OR of AP (95% CI 1.18–5.66, p = 0.02), and the continuous CCP score was again associated with AP (OR 1.64 per unit, 95% CI 0.95–2.90, p = 0.08) but did not meet conventional levels of statistical significance.²⁶ As acknowledged by the authors, a major limitation of the study is that AP is not the best predictor of long-term oncologic outcomes, and there were too few events in this cohort to assess these endpoints.²⁶ Similar to the group’s previous study, clinical utility could also not be assessed here.

Given the above limitations, the group at Michigan Medicine developed yet another score, this time a 15-gene prognostic signature (15G) from whole transcriptome sequencing performed on RNA isolated from archived RN specimens from localized (pT1-3) ccRCC tumors.²⁷ In a discovery cohort, the 15G signature was independently associated with worse disease-free survival (DFS) and DSS (DFS: HR 11.08 [95% CI 4.9 - 25.1]; DSS: HR 9.67 [95% CI, 3.4 - 27.7]) in an MVA adjusting for CP parameters (including the SSIGN score, MSKCC nomogram, and CCP score).²⁷ In the validation data sets, a high 15G score was also independently associated with worse DFS and overall survival (OS) (DFS: HR 2.11 [95% CI, 1.24 -3.6], OS: HR 3 [95% CI, 1.64 - 5.7]); however, standard CP parameters were also independently associated with worse DFS and OS: tumor stage T3-4 (DFS: HR 3.62 [95% CI, 2.17-6.1]; OS: HR 3.2 [95% CI, 1.84-5.5], both p < 0.001) and Fuhrman grade G3-4 (DFS: HR 1.96 [95% CI, 1.19-3.2], p = 0.008; OS: HR 1.4 [95% CI, 0.81-2.4)], p = not significant).²⁷ Further, the presence of dedifferentiation (e.g., sarcomatoid) and other relevant CP variables for computing the SSIGN score and MSKCC nomograms were not available for the validation cohorts, limiting the study’s assessment of utility.

The 15G score was also evaluated in metastatic ccRCC across six treatment groups: atezolizumab plus bevacizumab, sunitinib, atezolizumab, avelumab plus axitinib, nivolumab, and everolimus. A high 15G score was associated with significantly worse PFS in 4 of the 6 treatment groups though the overall response rate (ORR) was only significantly worse in 15G-high compared with 15G-low results among sunitinib-treated patients (43% vs. 29%, Fisher’s p = .002).²⁸ Additionally, as seen in the previous studies, multiple additional parameters (including IMDC and MSKCC scores, as well as tumor mutations, PD-L1 status, sarcomatoid histology and male sex) were also significantly different in tumors with 15G high vs. low scores.²⁸ In an MVA, MSKCC risk group was also predictably associated with worse PFS (HRs 1.3 - 5.2 for poor risk and 1.3-2.2 for intermediate risk).²⁸ Further, greater differences in survival were noted between 15G high and low when derived from metastatic (HR 3.0 [95% CI, 2.0 - 4.5]) as opposed to primary (HR 1.5 [95% CI, 1.2 - 1.8]) tumors, underscoring the importance of addressing concordance between original and secondary sites to determine the optimal implementation of the 15G score in the metastatic setting.²⁸ The authors noted that the 15G score was developed from nephrectomy specimens on the basis of genes that were differentially expressed in primary tumors that recurred; that threshold was carried over to the present study, though it might not be universally applicable.

A 16 gene recurrence score (16G) (comprised of 11 cancer-related genes and 5 housekeeping genes) was developed using RNA from archived FFPE tissue specimens from an observational cohort study of 942 patients with localized (stage I–III) ccRCC who underwent nephrectomy between 1985 and 2003 at the Cleveland Clinic.²⁹ Most patients (68%) had stage 1 disease; 221 recurrences (23%) occurred during a median follow-up time of 6.2 years. According to the Leibovich classification, 93% of 540 low-risk patients, 78% of 263 intermediate-risk patients, and 36% of 128 high-risk patients were recurrence-free at 5 years, findings that were consistent with previously published data.²⁹ A large number of patients were excluded from the discovery set because of histologic and/or clinical reclassification after central review. The gene signature was then validated using RNA from archived FFPE tissue from an independent cohort of 626 patients with stage I–III ccRCC who underwent nephrectomy between 1995-2007.²⁹ In an MVA, the 16G score was significantly associated with the risk of tumor recurrence (HR of 3.37 per 25-unit increase in the score [95% CI 2.23–5.08], p<0.0001), after stratification by stage and adjustment for tumor size (which was also significantly associated with recurrence with an HR of 2.09 [95% CI 1.07-4.08, p = 0.02]), grade, and Leibovich score.²⁹ The addition of the 16G recurrence score to the Leibovich score improved the C statistic for recurrence from 0.74 to 0.81.²⁹ However, the authors acknowledged that, in this population with a focus on patients with low and intermediate risk according to traditional measures (Leibovich score 0–6) and a consequent narrow range of risk, the C statistic is limited in its ability to represent clinically meaningful risk discrimination.²⁹ Further, there were many limitations in this study including that the assessment of the effect of tumor heterogeneity was tested in only eight patients, branch renal vein invasion status was not recorded, follow-up for every patient was not standardized, and comparison with MSKCC or Karakiewicz was not possible because the necessary information was not consistently collected. A subsequent analysis of the 16G score performed in high-risk stage III (T3) patients randomized to placebo or adjuvant sunitinib did not find an interaction of the score and sunitinib treatment.³⁰ Notably, <50% of eligible and consented patients had available tumor tissue and, while the assay seemed to provide prognostic information in the placebo arm, the authors acknowledged that the power to test for interaction was low (<40%), highlighting the need for future studies with sufficient power to detect an interaction of the score and treatment to determine whether the test might predict differential benefit from adjuvant therapy.³⁰ However, since the time of that study, sunitinib is no longer a recommended first-line adjuvant therapy option for non-metastatic RCC patients due to its toxicity profile and lack of an overall survival benefit.⁹

The 16G recurrence score was further investigated in a retrospective study of patients with non-metastatic ccRCC using data obtained from The Cancer Genome Atlas (TCGA). In an MVA, both SSIGN score (sub-distribution hazard ratio (sHR) 1.35 [95% CI 1.21–1.50], p <0.001) and 16G (sHR 1.43 per 25 16G score [95% CI 1.00–2.04], p <0.050) were significantly associated with recurrence when the latter was analyzed as a continuous variable.³¹ However, in the MVA of SSIGN and categorical 16G risk groups, both the SSIGN intermediate- and high-risk groups remained significantly associated with recurrence (sHR 3.80 [95% CI 1.85–7.84], p <0.001, and sHR 7.04 [95% CI 3.28–15.10], p <0.001, respectively), while only the high-risk 16G group remained significantly associated with recurrence (sHR 1.84 [95% CI 1.03–3.26], p = 0.040).³¹ Notably, tumor stage, tumor size, and lymph node involvement were also significantly associated with recurrence.³¹ The 16G score was not significantly associated with recurrence in low- or high-risk SSIGN patients but was found to further risk stratify the SSIGN intermediate-risk group (sHR 2.22 [95% CI, 1.10–4.50], p =0.03).³¹ SSIGN low-, intermediate-, and high-risk groups demonstrated 2.7%, 15.2%, and 27.5% 3-year recurrence risk, respectively; SSIGN intermediate-risk patients with low and high 16G scores had 3-year recurrence rates of 8.0% and 25.2%, respectively.³¹ However, for intermediate-risk SSIGN patients meeting key eligibility requirements for adjuvant treatment, differences in recurrence rates were not statistically significant on MVA.³¹ Notably, this retrospective study using a TCGA cohort included small numbers of patients within each subgroup; as such, the data may not be representative of the general RCC population. Moreover, it was limited to evaluating prognostic information with a median of only 43 months follow-up time and did not evaluate predictive response to therapy. In fact, the score no longer provided statistically significant information when specifically applied to patients who would be considered for adjuvant therapy based on the Keynote-564 study, the findings of which serve as the basis for recommendations regarding adjuvant immunotherapy treatment in current guidelines.^9,32

Other gene expression profile studies have suggested that RCC tumors expressing angiogenic molecular signatures may demonstrate greater response to TKI therapies, while immunogenic molecular signatures may confer responsiveness to immunotherapies including immune checkpoint inhibitors (ICIs).^33,34 However, ICI therapies have become the “near-ubiquitous choice” for first line adjuvant and metastatic treatment for ccRCC, rendering some of these studies, particularly those evaluating comparisons with TKIs such as sunitinib, out of date.³⁴ While tests using such biomarkers remain largely exploratory, future research may demonstrate clinical utility in some subgroups, such as RCC patients who are refractory to first-line ICI-containing regimens.

Prognostic information in RCC may be available from tumor mutational profiles obtained by genomic sequencing, rather than from transcriptional biomarkers. A large multinational study evaluated the association of somatic mutations and outcomes in a post-nephrectomy cohort of patients with ccRCC. In an MVA adjusted for testing multiple genes, mutations in 12 driver genes associated with RCC were not significantly associated with DFS in the validation cohort. However, sequenced tumors containing a VHL mutation alone showed significantly improved outcomes in comparison with tumors containing a VHL plus clinically significant additional mutations. In the validation cohort (n=474), five-year DFS rates were 61.5%, 73.7%, 84.7%, and 90.4% for patients with VHL+≥3, VHL+2, VHL+1, and VHL+0 additional mutations, respectively.³⁵ Notably, these genomically defined groups were independent of overall tumor mutational burden (TMB). DFS rate at 5 years among the 397 patients not considered eligible for adjuvant therapy was 90.6% (95% CI, 88%–94%) versus 63.6% (95% CI, 57%–71%) for the 196 patients eligible for adjuvant therapy.³⁵ Patients defined as being eligible for adjuvant therapy could be further stratified by risk of relapse based on the genomic classification of their tumors. Five-year DFS rates were 79.3% (95% CI, 69%–91%) amongst the 56 (29%) patients with VHL+0 tumors, 69.4% (95% CI, 60%–81%) amongst the 77 (39%) patients with VHL+1 tumors, 45.6% (95% CI, 33%–63%) amongst the 46 (23%) patients with VHL+2 tumors, and 35.3% (95% CI, 19%–67%) amongst the 17 (9%) patients with VHL+≥3 tumors; the VHL+2 and VHL+≥3 groups had significantly poorer survival compared with the VHL+0 group (p = 0.00055 and p < 0.0001, respectively).³⁵ On the basis of these findings, the authors suggest that patients with VHL+0 additional mutations could potentially be spared from further treatment, while patients with VHL+2, and VHL+≥3 tumors should consider adjuvant therapy.³⁵ Moreover, this tumor mutational status retained the ability to meaningfully sub-stratify patients within patients categorized into risk groups by the Leibovich score. Notably, other genomic features, such as copy-number alterations, were not evaluated though they may allow further refinement of genomic groups. Further, while VHL+2, and VHL+≥3 tumors are associated with the highest risk of disease recurrence, the benefit of adjuvant ICI in these patients requires further study.

While gene mutations may assist with risk stratification, some studies suggest caution may be warranted as there is significant risk of false positive findings due to intratumor heterogeneity (ITH). In one study of stage T2-T4 RCC tumors, approximately 75% of driver alterations (copy number variants (CNVs) and sequence mutations) were found to be subclonal and not observed throughout all regions of the tumor.³⁶ Another study evaluated both small (<4cm) and large (>7cm) ccRCC tumors (with 3 regions sampled from each) using CNVs, gene expression analysis (for ccA/ccB profiles) and the CCP score.³⁷ Total CNVs and subclonal CNV events were less frequent in small tumors (p<0.001). However, significant gene expression heterogeneity was observed for both CCP scores and ccA/ccB classifications, and larger tumors had more variance in CCP scores (p=0.026). The distribution of ccA/ccB differed between small and large tumors with mixed ccA/ccB tumors occurring more frequently in the larger tumors (p=0.024).³⁷ Finally, in a study evaluating patients who had both a nephrectomy and metastasectomy, ITH of ccA/ccB subtypes was observed in 22% (95% CI 3–60%) of metastatic tumors.³⁸ Subtype differed across longitudinal metastatic tumors from the same patient in 23% (95% CI 10–42%) of cases and across patient-matched primary and metastatic tumors in 43% (95% CI 32–55%) of cases, suggesting that the primary tumor is not a good surrogate for the metastatic tumor.³⁸

As with other cancers, the management of RCC relies on accurate prognostic and predictive tools for effective patient counseling, surveillance, and choice of therapy. To this end, physicians have relied on the guideline-endorsed combination of CP parameters, including tumor size, stage, histology, and grade, as well as predictive risk models and nomograms (particularly for mRCC).

Biomarker classifiers have the potential to improve risk stratification as well as inform treatment selection for patients with RCC, such that patient management could be optimized. In general, most low-risk patients with localized RCC will not recur and most high-risk patients have great enough risk such that they will be offered adjuvant therapy regardless of a genomic or proteomic classifier. Therefore, a genomic score may have the largest role in discriminating between patients in an intermediate risk group (as determined by CP criteria and nomograms), given the wider range of clinical outcomes. However, even among low- and high-risk patients, the recurrence rates vary and there may be clinical utility for a biomarker that can further refine risk, providing information above what can be obtained from current CP parameters. For example, since current guidelines support the use of adjuvant immunotherapy in patients with high-risk RCC after nephrectomy, a test that can further refine risk may be of use in minimizing overtreatment. In metastatic RCC, there may also be utility for a biomarker that can predict response to therapy, thereby supporting physician choice among multiple possible treatment options.

However, the current evidence does not support the use of biomarker classifiers for any of these purposes. The studies evaluating such classifiers in RCC have been limited by outdated patient cohorts, incomplete CP information, limited and non-standardized follow-up, lack of centralized pathology or radiology review, underrepresentation of histologies other than ccRCC, and insufficient power for confident subgroup and interaction analyses. On the whole, their performance has not been evaluated in large prospective trials (though prospectively designed observational studies have been performed using archived tissue samples) and in the available literature, they have not been demonstrated to provide clinically meaningful information above that which can be obtained from available CP parameters (including nomograms), such that patient management and outcomes were improved. Further, tumor stage and grade, as well as other CP parameters, are heavily utilized in RCC risk stratification and have been shown to be the some of the most highly significant predictors of risk, often surpassing biomarker classifiers in comparative prognostic analyses. To this end, we note that a marginal increase in a concordance index does not necessarily equate to meaningful clinical information. As such, these classifiers have yet to be widely deployed clinically or recognized by national or international consensus guidelines.^9,20 Additionally, while certain studies have found some classifiers to be prognostic, future research is required to define their role in guiding management decisions and improving outcomes.

Additionally, treatment options have changed in the last few years, rendering many of the available studies outdated and not generalizable to modern therapeutic interventions, as their outcomes were based on former therapies that are no longer used or not routinely used as first-line (e.g., sunitinib in the adjuvant setting). Specifically, immunotherapies have become the current mainstay of first-line systemic therapy in RCC. The performance and potential use of these biomarkers for predicting response in the current landscape of recommended therapies is unknown.

The classifiers evaluated to date have also had limited applicability, given their general lack of evaluation in tumors other than ccRCC as well as the significant confounding issues around tumor heterogeneity. Important questions remain regarding the clinical interpretation of subclonal abundance and how heterogeneous tumors can be better profiled and utilized in precision medicine. For example, it is unknown whether a tumor with a small ccB component has a similarly poor prognosis to an identical size tumor dominated by the ccB signature. It is additionally important to consider that primary RCC tumors are often genetically divergent from metastases, a significant proportion of which harbor ITH, thereby presenting a challenge in biomarker tools for predicting the behavior of metastatic disease.

In sum, additional validation, ideally using prospective clinical data with sufficient patient numbers (particularly regarding the various risk subsets of RCC and accounting for ITH) and follow-up time, is needed to better understand the performance and utility of transcriptomic classifiers for risk stratification or therapy selection in RCC. Meanwhile, additional prognostic information (above existing clinical parameters, including nomograms) may already be obtained from diagnostic sequencing tests/tumor mutational profiling, rendering any further testing by gene expression analysis unnecessary unless supplementary information is provided. It is unclear whether transcriptomic classifiers will demonstrate clinical utility in the future, such that their use impacts patient management with a resultant improvement in outcomes.

1. National Cancer Institute. Cancer Stat Facts: Kidney and Renal Pelvis Cancer. https://seer.cancer.gov/statfacts/html/kidrp.html. Accessed 7/17/2026.
2. Rose TL, Kim WY. Renal cell carcinoma: a review. Jama. 2024;332(12):1001-1010. doi:10.1001/jama.2024.12848
3. Ali SN, Tano Z, Landman J. The changing role of renal mass biopsy. Urol Clin North Am. 2023;50(2):217-225. doi:https://doi.org/10.1016/j.ucl.2023.01.002
4. Silverman SG, Pedrosa I, Schieda N, Margulis V, Kapur P, Davenport MS. In pursuit of KI-RADS: toward a single, evidence-based imaging classification of renal masses. Radiology. 2025;314(3):e240308. doi:10.1148/radiol.240308
5. Nguyen KA, Brito J, Hsiang W, et al. National trends and economic impact of surgical treatment for benign kidney tumors. Urol Oncol. 2019;37(3):183.e9-183.e15. doi:10.1016/j.urolonc.2018.11.019
6. Lounová V, Študent V, Jr., Purová D, Hartmann I, Vidlář A, Študent V. Frequency of benign tumors after partial nephrectomy and the association between malignant tumor findings and preoperative clinical parameters. BMC Urol. 2024;24(1):175. doi:10.1186/s12894-024-01543-3
7. Kim JH, Li S, Khandwala Y, Chung KJ, Park HK, Chung BI. Association of prevalence of benign pathologic findings after partial nephrectomy with preoperative imaging patterns in the United States from 2007 to 2014. JAMA Surg. 2019;154(3):225-231. doi:10.1001/jamasurg.2018.4602
8. Campbell SC, Uzzo RG, Karam JA, Chang SS, Clark PE, Souter L. Renal mass and localized renal cancer: evaluation, management, and follow-up: AUA guideline: part II. J Urol. 2021;206(2):209-218. doi:10.1097/ju.0000000000001912
9. National Comprehensive Cancer Network. NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines). Kidney Cancer. Version 2.2026. https://www.nccn.org/professionals/physician_gls/pdf/kidney.pdf. Accessed 7/17/2026.
10. Costantini M, Poeta ML, Pfeiffer RM, et al. Impact of histology and tumor grade on clinical outcomes beyond 5 years of follow-up in a large cohort of renal cell carcinomas. Clin Genitourin Cancer. 2021;19(5):e280-e285. doi:10.1016/j.clgc.2021.07.003
11. Haas NB, Manola J, Uzzo RG, et al. Adjuvant sunitinib or sorafenib for high-risk, non-metastatic renal-cell carcinoma (ECOG-ACRIN E2805): a double-blind, placebo-controlled, randomised, phase 3 trial. Lancet. 2016;387(10032):2008-16. doi:10.1016/s0140-6736(16)00559-6
12. Correa AF, Jegede O, Haas NB, et al. Predicting renal cancer recurrence: defining limitations of existing prognostic models with prospective trial-based validation. J Clin Oncol. 2019;37(23):2062-2071. doi:10.1200/jco.19.00107
13. Oza B, Frangou E, Eisen T, et al. Determining the impact of histology on the incidence, pattern, and timing of recurrences in patients with renal cell carcinoma: a pooled analysis from the SORCE and ASSURE Trials. Eur Urol Open Sci. 2025;79:19-26. doi:10.1016/j.euros.2025.07.003
14. Lee-Felker SA, Felker ER, Tan N, et al. Qualitative and quantitative MDCT features for differentiating clear cell renal cell carcinoma from other solid renal cortical masses. AJR Am J Roentgenol. 2014;203(5):W516-W524. doi:10.2214/AJR.14.12460
15. Marconi L, Dabestani S, Lam TB, et al. Systematic review and meta-analysis of diagnostic accuracy of percutaneous renal tumour biopsy. Eur Urol. 2016;69(4):660-673. doi:10.1016/j.eururo.2015.07.072
16. Gao H, Nowroozizadeh B, Zepeda JP, et al. The success rate of small renal mass core needle biopsy and its impact on lowering benign resection rate. BMC Urol. 2023;23(1):189. doi:10.1186/s12894-023-01363-x
17. Mansour H, Tran-Dang MA, Walkden M, et al. Renal mass biopsy - a practical and clinicopathologically relevant approach to diagnosis. Nat Rev Urol. 2025;22(1):8-25. doi:10.1038/s41585-024-00897-5
18. Boynton DN, Mirza M, Van Til M, et al. Renal mass biopsy is associated with fewer radical nephrectomies for benign or indolent disease, particularly for T1b renal masses. Urol Pract. 2025;12(1):148-156. doi:10.1097/upj.0000000000000710
19. Campbell SC, Clark PE, Chang SS, Karam JA, Souter L, Uzzo RG. Renal mass and localized renal cancer: evaluation, management, and follow-up: AUA guideline: part I. J Urol. 2021;206(2):199-208. doi:10.1097/ju.0000000000001911
20. Cotta BH, Choueiri TK, Cieslik M, et al. Current landscape of genomic biomarkers in clear cell renal cell carcinoma. Eur Urol. 2023;84(2):166-175. doi:10.1016/j.eururo.2023.04.003
21. Brannon AR, Reddy A, Seiler M, et al. Molecular stratification of clear cell renal cell carcinoma by consensus clustering reveals distinct subtypes and survival patterns. Genes Cancer. 2010;1(2):152-163. doi:10.1177/1947601909359929
22. Brooks SA, Brannon AR, Parker JS, et al. ClearCode34: a prognostic risk predictor for localized clear cell renal cell carcinoma. Eur Urol. 2014;66(1):77-84. doi:10.1016/j.eururo.2014.02.035
23. Rathmell WK, Brooks SA, Parker JS, Nielsen ME. Reply to Alexander S. Parker, Brad C. Leibovich, Jeanette E. Eckel-Passow, John C. Cheville's letter to the editor re: Samira A. Brooks, A. Rose Brannon, Joel S. Parker, et al. ClearCode34: a prognostic risk predictor for localized clear cell renal cell carcinoma. Eur Urol 2014;66:77-84. Eur Urol. 2014;66(5):e92. doi:10.1016/j.eururo.2014.07.002
24. Gulati S, Martinez P, Joshi T, et al. Systematic evaluation of the prognostic impact and intratumour heterogeneity of clear cell renal cell carcinoma biomarkers. Eur Urol. 2014;66(5):936-48. doi:10.1016/j.eururo.2014.06.053
25. Morgan TM, Mehra R, Tiemeny P, et al. A multigene signature based on cell cycle proliferation improves prediction of mortality within 5 yr of radical nephrectomy for renal cell carcinoma. Eur Urol. 2018;73(5):763-769. doi:10.1016/j.eururo.2017.12.002
26. Tosoian JJ, Feldman AS, Abbott MR, et al. Biopsy cell cycle proliferation score predicts adverse surgical pathology in localized renal cell carcinoma. Eur Urol. 2020;78(5):657-660. doi:https://doi.org/10.1016/j.eururo.2020.08.032
27. Mehra R, Nallandhighal S, Cotta B, et al. Discovery and validation of a 15-gene prognostic signature for clear cell renal cell carcinoma. JCO Precis Oncol. 2024;8:e2300565. doi:10.1200/po.23.00565
28. Monda SM, Nallandhighal S, Vaishampayan U, et al. Validation of a 15-gene prognostic signature in metastatic clear cell renal cell carcinoma. JCO Precis Oncol. 2025;9:e2500213. doi:10.1200/po-25-00213
29. Rini B, Goddard A, Knezevic D, et al. A 16-gene assay to predict recurrence after surgery in localised renal cell carcinoma: development and validation studies. Lancet Oncol. 2015;16(6):676-85. doi:10.1016/s1470-2045(15)70167-1
30. Rini BI, Escudier B, Martini JF, et al. Validation of the 16-gene recurrence score in patients with locoregional, high-risk renal cell carcinoma from a phase III trial of adjuvant sunitinib. Clin Cancer Res. 2018;24(18):4407-4415. doi:10.1158/1078-0432.Ccr-18-0323
31. Patel N, Hakansson A, Ohtake S, et al. Transcriptomic recurrence score improves recurrence prediction for surgically treated patients with intermediate-risk clear cell kidney cancer. Cancer Med. 2023;12(5):6437-6444. doi:10.1002/cam4.5399
32. Choueiri TK, Tomczak P, Park SH, et al. Adjuvant pembrolizumab after nephrectomy in renal-cell carcinoma. N Engl J Med. 2021;385(8):683-694. doi:10.1056/NEJMoa2106391
33. McKinnon MB, Rini BI, Haake SM. Biomarker-informed care for patients with renal cell carcinoma. Nat Cancer. 2025;6(4):573-583. doi:10.1038/s43018-025-00942-1
34. Bakouny Z, Hakimi AA, Reznik E, Motzer RJ. Biomarkers for renal cell carcinoma - a pragmatic approach. Nat Rev Urol. 2026;23(3):151-153. doi:10.1038/s41585-025-01073-z
35. Vasudev NS, Scelo G, Glennon KI, et al. Application of genomic sequencing to refine patient stratification for adjuvant therapy in renal cell carcinoma. Clin Cancer Res. 2023;29(7):1220-1231. doi:10.1158/1078-0432.Ccr-22-1936
36. Gerlinger M, Horswell S, Larkin J, et al. Genomic architecture and evolution of clear cell renal cell carcinomas defined by multiregion sequencing. Nat Genet. 2014;46(3):225-233. doi:10.1038/ng.2891
37. Ueno D, Xie Z, Boeke M, et al. Genomic heterogeneity and the small renal mass. Clin Cancer Res. 2018;24(17):4137-4144. doi:10.1158/1078-0432.Ccr-18-0214
38. Serie DJ, Joseph RW, Cheville JC, et al. Clear cell type A and B molecular subtypes in metastatic clear cell renal cell carcinoma: tumor heterogeneity and aggressiveness. Eur Urol. 2017;71(6):979-985. doi:10.1016/j.eururo.2016.11.018

• Provider Education/Guidance

• Transcriptional Biomarkers
• Renal Carcinoma