Prospective assessment of the clinical benefit of a tailored cancer gene set built on a next-generation sequencing platform in patients with recurrent or metastatic head and neck cancer

Abstract

We performed a prospective trial to assess the clinical benefit of a tailored gene set built on a next-generation sequencing (NGS) platform in patients with recurrent or metastatic head and neck squamous cell carcinoma (HNSCC). Archived tumor tissue obtained from patients with recurrent or metastatic HNSCC was analyzed for variants by a tailored Comprehensive Cancer Gene set of 40 genes (CCG-40) performed on a NGS platform. These data were provided to clinicians to inform treatment decisions. The primary endpoint was clinical benefit (disease control) that resulted from selection and administration of a targeted therapy based on results of the CCG-40. Barriers to performance and implementation of the assay were recorded. Forty patients enrolled. Primary tumor sites included oropharynx (14), larynx/hypopharynx (14), oral cavity (9), and nasopharynx (3). The CCG-40 assay was performed in 23 patients (57.5%), but not in 17 patients due inadequate financial coverage (12) or insufficient tumor tissue (5). Potentially actionable tumor variants were identified in 3 patients (7.5%); all were PIK3CA variants. Due to inability to obtain access to candidate drugs (2) or rapid decline in performance status (1), none of these patients received targeted therapy informed by the CCG-40 results. The CCG-40 assay did not provide clinical benefit to the patients on this trial. Identification of limitations of the assay and barriers to the test’s performance and application may be used to optimize this strategy in future trials.

Introduction

Precision oncology may be defined as the use of a cancer’s unique molecular characteristics to make informed decisions tailored to an individual patient [1, 2]. Increased understanding of the molecular biology of cancer, improvements in gene sequencing technology and increased availability of targeted therapies have made precision oncology a reality in several cancers [3-5]. However, limited progress has occurred toward development and implementation of precision oncology in head and neck squamous cell carcinoma (HNSCC). Biomarkers related to the human papillomavirus (HPV), Epstein–Barr virus and the programmed death-ligand 1 (PD-L1) are the only laboratory tests routinely used to make decisions in the clinic.

Foundational work reported by The Cancer Genome Atlas (TCGA) and others described the recurring genomic alterations in newly diagnosed HNSCC [6-11]. In contrast, limited information is available to describe the genomic alterations of recurrent or metastatic disease [12, 13]. Information in these reports provides an opportunity to select a panel of genetic alterations in cancer that are targetable. In contrast to whole-genome or whole-exome sequencing, these tailored gene sets are practical to perform in Clinical Laboratory Improvement Amendments (CLIA)-certified facilities with results reported in a timely way to inform decisions for an individual patient.

In HNSCC, the body of literature that evaluated the impact of tailored gene sets analyzed by a next-generation sequencing-based (NGS) clinical assay on treatment decisions and clinical benefit is limited. One report confirmed that the genetic alterations in formalin-fixed, paraffin-embedded (FFPE) tumor analyzed by a tailored NGS assay were comparable to those in frozen tumor analyzed by comprehensive research methods [14]. Three reports showed that tailored NGS clinical assay performed on tumor specimens identified candidate prognostic biomarkers, and illuminated the potential of a tailored gene set analyzed by NGS to impact treatment decisions in patients with recurrent or metastatic HNSCC [13, 15, 16]. However, only one report showed that administration of targeted therapy matched to the tailored NGS profile yielded clinical benefit to the patient [15].

The primary aim of this study was to determine the clinical benefit of a tailored Comprehensive Cancer Gene set of 40 genes (CCG-40) built on a NGS platform in a real-world cohort of patients with recurrent or metastatic HNSCC. We hypothesized that identification of variants in one or more of 40 genes commonly mutated in multiple cancers and tested in these patients’ tumors would lead to selection and administration of a matched targeted therapy that resulted in a measurable clinical benefit (disease control) to the patient. In this study, we report the results of this prospective study.

Methods

Patient selection criteria

Eligible patients were 18 years of age or older with recurrent or metastatic SCC of the oral cavity, oropharynx, nasopharynx, hypopharynx, or larynx. The protocol was approved by the Washington University (Human Research Protection Office) Institutional Review Board, and patients provided signed informed consent to participate. The study was conducted in accordance with the ethical principles of the Declaration of Helsinki.

Procedures and assessments

Archived FFPE tumor specimens were used to perform NGS. Archived tumor tissue from recurrence was used for testing if a sufficient specimen was available. In all other cases, archived tumor tissue from initial diagnosis was used. Genomic testing was performed by Genomics and Pathology Services at Washington University, a College of American Pathologists-accredited, CLIA-certified clinical laboratory. Representative slides were reviewed by the pathologist to confirm adequate tumor cellularity (≥ 10%), and from one to six FFPE tissue cores were obtained for DNA extraction, library preparation, and sequencing. The assay used Agilent SureSelect hybrid capture followed by sequencing performed on an Illumina HiSeq 2000, MiSeq, or HiSeq 2500 platform for exons of 40 cancer-related genes selected based on their potential actionable or prognostic significance in multiple cancers. The CCG-40 included ABL1, ALK, APC, ASXL1, ATM, BRAF, CEBPA, CTNNB1, DNMT3A, EGFR, ERBB2, ESR1, FGFR4, FLT3, IDH1, IDH2, JAK2, KIT, KRAS, MAP2K2, MAPK1, MET, MLL, MPL, MYC, MYD88, NOTCH1, NPM1, NRAS, PDGFRA, PIK3CA, PTEN, PTPN11, RB1, RET, RUNX1, TET2, TP53, VHL, and WT1. An additional set of 111 genes (CCG-111) related to cancer or drug metabolism and transport were similarly evaluated in a parallel assay with research intent (Supplementary Material). The 111 genes were not known to be actionable or prognostic when the protocol was initiated, but were performed with anticipation that they could be helpful in the future. Single-nucleotide variants and insertion/deletion events were reported. The assay was performed before methods were developed to detect rearrangements by NGS.

Variant annotation and reporting was previously described [17]. In brief, sequence variants were classified into 1 of 5 levels by an established bioinformatics classification system (Fig. 1). The system used was expanded from one previously described [17] to a set of 40 genes. The system stratified clinical actionability based on a locally available clinical-grade genomic database and genetic data published in databases including the Single Nucleotide Polymorphism database [18], the 1000 Genomes project [19], the c Variant Server of the National Heart, Lung, and Blood institute [20], and the Catalogue of Somatic Mutations in Cancer mutation database [21]. The classification was reviewed by a clinical genomicist in the pathology department.

Fig. 1.

Variant classification scheme and case distribution

Variants classified as levels 1–3 were reported as potentially clinically significant. The results of the CCG-40 clinical assay were provided to the treating physician, who then used the genomic information, along with the clinical scenario, to select therapy for the patient. The protocol did not assign specific targeted therapy to the results of the CCG-40 or automatically refer patients for matched trials.

The genetic variants identified in the CCG-40 and the CCG-111 gene sets were tabulated. The primary endpoint was the clinical benefit of the CCG-40 clinical assay in patients with recurrent or metastatic HNSCC. Clinical benefit was defined to occur when a CCG-40 assay identified a potentially actionable genetic variant that led to administration of a matched molecularly targeted therapy, which resulted in disease control. Disease control was defined as complete response, partial response, or stable disease using RECIST criteria [22]. A secondary endpoint was to describe the obstacles that occurred in the implementation of the CCG-40 assay and administration of the targeted therapy selected based on the CCG-40.

Statistical analysis

The primary hypothesis of this study was that identification of tumor variants in one or more genes from the CCG-40 assay would result in administration of a targeted therapy that yielded a measurable clinical benefit (disease control) to the patient. For the primary endpoint, we assumed a null hypothesis (H₀) of < 1% clinical benefit rate and we hypothesized (H_a) that a 10% clinical benefit rate would warrant further investigation. A Simon’s optimal two-stage design was used in the trial. At a significance level of 0.05, a sample of 39 patients with recurrent or metastatic HNSCC needed to be enrolled. If none of the 17 patients derived a clinical benefit during stage one, the null hypothesis would be accepted and the trial concluded. If at least one patient derived a clinical benefit, an additional 22 patients were to be enrolled to stage two. If 1 or fewer of 39 patients derived a clinical benefit, the null hypothesis would be accepted. If two or more patients derived a clinical benefit, the alternative hypothesis would be accepted and the conclusion would be that further investigation of the approach was warranted. However, the proportion of patients in stage one that had the CCG-40 assay performed (11 of 17; 65%) was lower than expected. As a result, we elected to proceed with enrollment to stage two of the trial. Therefore, this study functions as a one-stage design with a total sample size of 23 patients. A sample size of 23 achieves 81.4% power to detect a difference of 9% under the null hypothesis of 1% using a one-sided exact test at a significance level of 0.05.

Descriptive statistics were used to analyze the obstacles that occurred in the implementation of the CCG-40 assay and administration of the targeted therapy selected based on the results of the CCG-40. Variants discovered in the CCG-40 and the CCG-111 gene sets were described.

Clinical characteristics were summarized using descriptive statistics. Continuous and categorical variables between CCG performed and not performed groups were compared by a Kruskal–Wallis test and the Chisquare test, respectively. Overall survival (OS) was the time from the date of signed consent to the date of death. Patients were censored at the last known date of follow-up. Kaplan–Meier curves were generated that provide unadjusted survival estimates. All statistical tests were two-sided using an α = 0.05 level of significance.

Results

Patient and tumor characteristics

Forty patients enrolled into the trial from March to October, 2014 (Table 1). Most patients (75%) had a significant (≥ 10 pack-year) history of smoking. Primary tumor sites included oropharynx (35%), larynx and hypopharynx (35%), oral cavity (22.5%), and nasopharynx (7.5%). Nine of the fourteen cancers of the oropharynx were p16 positive. Median OS from trial enrollment was 1.24 years (95% CI 0.57–2.03) (Fig. 2).

Table 1.

Patient characteristics

Characteristic	All patients n = 40	CCG* performed n = 23	CCG not performed n = 17	p Value**
Age at diagnosis
Median (SD)	59 (8.8)	61 (8.7)	57 (8.3)	0.08
Range	39-77	44-77	39-70 (8.3)
Gender
Male	35 (87.5%)	20 (87%)	15 (88.2%)	1.00
Female	5 (12.5%)	3 (13%)	2 (11.8%)
Race
White	31 (77.5%)	17 (82.3%)	14 (73.9%)	0.71
Black	9 (22.5%)	6 (26.1%)	3 (17.7%)
Tumor site
Oral cavity	9 (22.5%)	6 (26.1%)	3 (17.6%)	0.71
Oropharynx	14 (35%)	7 (30.4%)	7 (41.2%)
Larynx	12 (30%)	7 (30.4%)	5 (29.4%)
Hypopharynx	2 (5%)	2 (8.7%)	0
Nasopharynx	3 (7.5%)	1 (4.3%)	2 (11.8%)
Smoking status
Current smoker	11 (27.5%)	5 (21.7%)	6 (35.3%)	0.46
Non-current smoker	23 (57.5%)	15 (65.2%)	8 (47.1%)
< 10 pack years	10 (25%)	6 (26.1%)	4 (23.5%)	1.00
≥ 10 pack years	30 (75%)	17 (73.9%)	13 (76.5%)
p16 expression (oropharyngeal only)
Positive	9 (64.3%)	5 (71.4%)	4 (57.1%)	1.00
Negative	4 (28.6%)	2 (28.6%)	2 (28.6%)
Unavailable	1 (7.1%)	0	1 (14.3%)

^*CCG Comprehensive Cancer Gene set

^**Comparison of CCG performed vs not performed subsets

Fig. 2.

Overall survival

Feasibility of performing the CCG-40 assay on clinical specimens

The CCG-40 assay was performed in 23 patients (57.5%), using archived tumor tissue obtained at recurrence (7) or initial diagnosis (16). The CCG-40 assay was not performed in 17 patients, due to inadequate financial coverage (12) or insufficient tumor tissue (5). Patient and tumor characteristics were similar in the cohort in which the CCG assay was performed compared to the cohort in which it was not performed (Table 1).

Variants identified on the CCG-40 assay

Variants were identified by the CCG-40 assay in 15 of 23 (65.2%) evaluable patients (Fig. 3). Potentially actionable (level 1) variants were identified in three patients (2 oral cavity, 1 laryngeal); all were activating mutations of PIK3CA. Variants classified as level 2 occurred in one patient (4%) and level 3 in 17 patients (74%). One patient had a BRAF variant identified which was thought to be insensitive to available BRAF inhibitors and therefore classified as level 3. Additionally, one PIK3CA variant was classified as a level 3 variant because it was a synonymous mutation. Variants in the following genes were identified: TP53 (n = 12), PIK3CA (4), NOTCH1 (2), BRAF (1), KRAS (1), PTEN (1), and ATM (1) (Fig. 3).

Fig. 3.

Results of the Comprehensive Cancer Gene (CCG)-40 clinical and CCG-111 research assays

Variants identified on the supplemental CCG-111 assay

The CCG-111 assay was performed on archived tumor tissue in the 23 patients who had the CCG-40 assay. Variants were identified by the CCG-111 assay in 17 (73.9%) of these patients (Fig. 3). The most common variants identified were in MAP3K1 (n = 8) and CDKN2A (5). These genes were tested on a research basis, and not available to clinicians.

Clinical benefit of the CCG-40 assay

Of the three patients with potentially actionable (level 1) variants identified, none were treated with targeted therapy selected based on the results of the CCG-40 assay. Reasons for this include inability to obtain access to matched candidate drugs (2) or rapid decline in performance status (1). Therefore, no clinical benefit was derived from the CCG-40 clinical assay.

Discussion

A key goal of precision oncology is to yield clinical benefit for patients by prescription of effective targeted therapy matched to driver alterations. With the exception of PD-L1 expression, no biomarker is predictive of clinical benefit from a targeted therapy in patients with HNSCC [23, 24]. The genomic landscape of HNSCC illuminated many alterations that could be targetable [10]. Examples include somatic alterations in components of cell signaling and cycling, angiogenesis, and differentiation. In this prospective trial, we hypothesized that identification of tumor variants in one or more genes commonly mutated in multiple cancers using the CCG-40 assay would result in selection and administration of a matched targeted therapy that yielded a measurable clinical benefit (disease control) to the patient. Among 40 patients enrolled, a potentially actionable variant was identified in only 3 patients (7.5%), and none of these patients were treated with a targeted therapy matched to the actionable variant identified by the CCG-40 assay. The results of the trial showed that the CCG-40 assay did not yield clinical benefit in these patients.

Several large-scale tailored genomic testing trials with multiple cancer types have been reported. Some of these trials included patients with HNSCC. In one study, 6 of 28 patients (21%) with HNSCC had actionable mutations identified by a 50-gene set [25]. The variants detected in these patients were similar to those identified by the CCG-40 assay. In the MOSCATO 01 trial, 15 of 111 patients (13.5%) with HNSCC had actionable alterations identified by targeted sequencing, CGH, WES and RNAseq, and were treated with matched targeted therapy [26]. Eight of these patients (7% overall) achieved clinical benefit from the targeted therapy. Other large-scale trials included very few patients with HNSCC [27-29].

A logical next step is to perform a randomized trial comparing a targeted NGS panel-guided approach to standard of care in HNSCC. The relatively low proportion of patients with HNSCC who have targetable mutations (7.5–21%) [13, 30] is a major obstacle to conducting a randomized trial specific to HNSCC.

A meta-analysis of phase 2 clinical trials across a broad range of cancers (including HNSCC) showed that a personalized cancer treatment strategy resulted in better efficacy compared to a non-personalized approach [29]. However, the randomized SHIVA trial, which included a limited number of patients with HNSCC, failed to show a clinical benefit of targeted therapy based on tumor profiling compared to conventional therapy in patients with advanced cancer [31].

Several barriers to demonstrating a clinical benefit of the CCG-40 assay were identified. These included inadequate financial resources to pay for the assay, insufficient tumor tissue for testing and lack of assess to candidate drugs matched to the actionable variant identified in the CCG-40 assay. The CCG-40 assay included 40 genes, of which variants in only 19 could be matched to candidate targeted agents. Addition of the CCG-111, which included 111 more genes, revealed more potentially actionable variants, including MAP3K1, CDKN2A, ERBB3, FGFR2, FLT1, and FLT4. However, none of these variants are validated predictive biomarkers in HNSCC.

There are several limitations of this trial. The CCG assay identified single-nucleotide variants and insertions or deletions; however, rearrangements and copy number variations could not be identified due to test methodology in use at the time. Copy number variation and rearrangements known to be important in HNSCC include EGFR, CCND1, PIK3CA, FGFR, ERBB3, MYC amplifications, and rearrangements involving FGFR3 [10]. Although several of these alterations are targetable, none are validated predictive biomarkers in HNSCC. The number of patients in our trial was limited to 40, and only 23 patients underwent genome testing. Under these conditions, and appreciating the large number of genetic alterations that occur in HNSCC, our trial had limited power to detect a clinical benefit in small targetable subsets. Additionally, we did not have a pre-defined pathway to link actionable alterations to targeted therapy or clinical trial enrollment.

The landscape of personalized medicine has changed since our trial was performed. Development of better and cheaper assays, alternative sources of tumor (cell-free DNA), and a larger variety of targeted agents and clinical trials offer optimism that personalized cancer treatment can be realized in patients with HNSCC.

The CCG-40 assay did not provide clinical benefit to these patients with recurrent or metastatic HNSCC. Identification of limitations of the CCG-40 assay and barriers to the test’s performance and application may be used to optimize this strategy in future trials.

Supplementary Material

Electronic supplementary material The online version of this article (https://doi.org/10.1007/s12032-019-1336-3) contains supplementary material, which is available to authorized users.