If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Multiple Biomarker Testing Tissue Consumption and Completion Rates With Single-gene Tests and Investigational Use of Oncomine Dx Target Test for Advanced Non–Small-cell Lung Cancer: A Single-center Analysis
First-line targeted therapies have been developed for advanced non–small-cell lung cancer (NSCLC). However, small biopsy samples pose a challenge to testing all relevant biomarkers. The present study characterized clinician-ordered single-gene lung cancer testing and evaluated tissue stewardship and the ability to successfully determine mutation status with single-gene testing or investigational use of the Oncomine Dx Target Test.
Materials and Methods
Clinician-submitted orders for 3659 single-gene tests (EGFR, ALK, ROS1, BRAF, KRAS, ERBB2, MET, RET, FGFR1) across 1402 samples at a large US-based commercial reference laboratory and 169 investigational Oncomine Dx Target Tests were retrospectively evaluated. The testing success rates and tissue consumption were evaluated by sample type, test type, and number of single-gene tests per sample.
Results
The large majority of lung tissue samples submitted for clinical testing were small (70.5% core needle biopsies; 10.0% fine needle aspirations). With single-gene testing, mutation status was successfully reported for ≥ 1 biomarker for 88.4% of the clinical samples. The success rates decreased and tissue consumption increased with testing of additional biomarkers. Investigational Oncomine Dx Target Tests were permitted 1 tissue slide each and demonstrated success rates similar to single-gene testing for ≥ 5 biomarkers on core needle biopsies, ≥ 4 biomarkers on fine needle aspirations, and ≥ 2 biomarkers on surgical resection specimens.
Conclusion
Tissue stewardship is important to enable successful completion of genetic testing and informed NSCLC treatment decisions. Preliminary assessment of the investigational Oncomine Dx Target Test suggests it could facilitate access to multiple biomarker testing using small tissue samples to support therapy decisions for patients with advanced NSCLC.
For patients with stage IIIB or IIIC NSCLC at diagnosis, the 5-year cancer-specific survival has ranged from 26% or 24% (for clinically staged) to 46% or 52% (for pathologically staged).
The 5-year overall survival (OS) rates for patients with advanced NSCLC (aNSCLC) patients have been reported at 26% for stage IIIB, 13% for stage IIIC, 10% for stage IVA, and 1% for stage IVB.
For patients with activating genetic mutations, targeted therapies have been shown to increase progression-free survival (PFS), and, more recently, OS compared with nontargeted chemotherapy.
Erlotinib versus standard chemotherapy as first-line treatment for European patients with advanced EGFR mutation-positive non-small-cell lung cancer (EURTAC): a multicentre, open-label, randomised phase 3 trial.
The small lung cancer tissue samples available for testing can make it difficult for patients and physicians to access the mutation status across multiple relevant genes to guide treatment-decisions using traditional, sequential, single-gene testing. A study at the University of Pittsburgh Medical Center found that only 67% of computed tomography-guided core needle biopsies (CNBs) and 46% of fine needle aspirations (FNAs) had sufficient tumor to successfully determine the EGFR, ALK, and KRAS mutation status.
With the continuing development and approval of targeted therapies for NSCLC (eg, crizotinib for ROS1+ patients, dabrafenib plus trametinib for BRAF+ patients), physicians and patients selecting appropriate treatments face the challenge of the increasing number of genes that must be tested using small amounts of available tumor tissue.
The feasibility of next generation sequencing (NGS) for multiple biomarker testing and targeted treatment guidance for lung cancer and other solid tumors is being investigated in clinics and cancer trials. The ongoing NCI-MATCH (National Cancer Institute Molecular Analysis for Therapy Choice) trial has used a customized NGS assay and the Ion Torrent platform (Thermo Fisher Scientific, Waltham, MA) and reported a testing completion rate of 85% at the interim analysis across all tumor biopsy samples. This success rate increased to 94% in later updated reports, when adjustments to the protocol required that supplemental cytology specimens be submitted with the biopsy tissue sample.
The Oncomine Dx Target Test (Ion Torrent PGM Dx Sequencer; Thermo Fisher Scientific) is an NGS panel for NSCLC testing approved by the US Food and Drug Administration (FDA) in June 2017. It is a qualitative, in vitro diagnostic test that uses high-throughput parallel sequencing technology to detect sequence variations in 23 genes on DNA and RNA isolated from formalin-fixed paraffin-embedded (FFPE) specimens. The diagnostic is validated against standard references, including clinical concordance, limit of blank, limit of detection, and variant-level detection accuracy and reproducibility.
The results from completed clinical studies have supported the concordance and clinical efficacy of the Oncomine Dx Target Test for 3 companion diagnostic genetic targets: BRAF, EGFR, and ROS1 (overall clinical concordance, 100%, 99%, and 96.5%, respectively), with corresponding accuracy and reproducibility across standard measures. Likewise, analytical performance has been established for 4 additional variants, and the Oncomine Dx Target Test’s performance for outstanding gene targets has been validated using a representative method.
As more gene targets are identified and recognized in clinical practice guidelines for aNSCLC, physicians and patients need to be able to test more biomarkers on small amounts of available tissue to make fully informed treatment decisions. As such, tissue stewardship is critical to enable testing; however, little evidence has been reported on the rates of successful determination of mutation status across multiple biomarkers for a single patient comparing single-gene testing or NGS methods or on the amounts of tissue needed to support these methods. The present study characterized the current single-gene testing paradigm in the molecular assessment for lung cancer patients and evaluated the success rates and tissue consumption with clinician-ordered single-gene testing and investigational use of the Oncomine Dx Target Test across different types of lung tumor tissue samples.
Materials and Methods
The present retrospective study analyzed the records from a large, US-based, Clinical Laboratory Improvement Amendments–certified, commercial reference laboratory. This laboratory conducted clinical molecular assessment of physician-submitted FFPE lung tissue samples and performed investigational testing using the Oncomine Dx Target Test before FDA approval on archival FFPE lung tissue samples. Investigational use of the Oncomine Dx Target Test was conducted in accordance with the FDA-approved protocol, except that only 1 tissue slide was permitted per test. The laboratory data contained the status (successful completion or reason for failure) of a test procedure but not the patient’s mutation status. Because the laboratory data were accessed retrospectively and provided to the researchers such that the subjects could not be identified, directly or through identifiers linked to the subject, no patient consent forms or institutional review board approval was necessary.
The present study evaluated the proportion of tests that were successful (able to report results) and the tissue stewardship (number of slides cut for testing). Both outcomes were assessed by test type, sample type, and, for samples evaluated using single-gene testing, the number of tests per sample.
Clinical Single-gene Testing
The clinical single-gene testing paradigm was characterized using data from physician-ordered genetic tests on lung tumor samples submitted to the laboratory from September 2015 through October 2016. Single-gene tests ordered for activating mutations relevant to first-line treatment for aNSCLC were included: therascreen EGFR RGQ PCR Kit (QIAGEN Manchester Ltd, Manchester, UK); Vysis IntelliFISH for ALK (Abbott Laboratories, Abbott Park, IL); cobas 4800 BRAF V600 Mutation Test (Roche Molecular Systems, Pleasanton, CA); and laboratory-developed tests (LDTs) for BRAF, KRAS, MET amplification, RET, ERBB2, FGFR1, and ROS1. The BRAF and KRAS LDTs used real-time, or quantitative, polymerase chain reaction. All other LDTs used fluorescent in situ hybridization.
Testing for CD274 gene expression (ie, programmed cell death ligand 1 expression), was not offered at this laboratory until nearly halfway through the study period (earliest test was in March 2016). Therefore, these tests were not included in the present study. Clinical samples were excluded if any type of NGS panel test was initiated or if a CD274 gene expression test was initiated, because the use of sample tissue to attempt these tests this could have affected the ability to complete other single-gene tests using the sample.
Data on each clinical test order included the type of sample submitted for testing: CNB, FNA, surgical resection, or cell block. The single-gene tests ordered per sample, the status of each ordered test showing whether test results were successfully reported, and the reason for not reporting results, if applicable, were also available from the data.
Lung tissue samples with ≥ 1 study-eligible single-gene test were included in the present analysis. Cancelled tests were excluded from the present study. A test was considered cancelled if the test status showed that the reason for not reporting results was cancellation by the client, duplication of the order, or conversion of the order to NGS. A test was considered unsuccessful if the test status was any of the following: PCR amplification failure, insufficient DNA quantity, poor DNA quality, high background, no hybridization, decalcification failure, wrong tumor type, insufficient tumor quantity, or poor tumor quality. This included tests that were ordered by physicians but were not initiated, which could have occurred if the sample’s tumor content did not meet the minimum testing requirements as defined by the laboratory’s protocol for single-gene tests. Although numerical data on the tumor content percentage were missing for some clinical samples, all samples included in the present analysis had > 0% tumor content. A clinical single-gene test was considered successfully completed if the test status indicated that results were reported.
The number of slides cut from the sample for each single-gene test was used to inform the analyses of tissue consumption. In some cases, these data fields were blank, which could indicate that no slides had been cut (ie, sample rejection) or that the data were missing. The tests for which these data was not reported—whether that was because of sample rejection or missing data—were excluded from the tissue consumption analyses.
Investigational Use of the Oncomine Dx Target Test
The investigational use of the Oncomine Dx Target Test was conducted on archival FFPE lung cancer samples as part of a previous study conducted by the laboratory from April 2016 to July 2016. The tests were conducted using the Ion Torrent PGM Dx platform. The panel included 23 genes: AKT1, ALK, BRAF, CDK4, DDR2, EGFR, ERBB2, ERBB3, FGFR2, FGFR3, HRAS, KIT, KRAS, MAP2K1, MAP2K2, MET, MTOR, NRAS, PDGFRA, PIK3CA, RAF1, RET, and ROS1 (the list of specific gene variants tested is provided in Supplemental Table 1; available in the online version). The investigational protocol was identical to the protocol for the FDA-approved Oncomine Dx Target Test, except for the number of tissue slides used per test.
The pre-established protocol for the investigational study allowed only 1 slide per test, and the protocol for the FDA-approved test recommended ≥ 2 slide-mounted 5-μm sections of surgical resection samples or 9 slide-mounted 5-μm sections of CNB samples.
Investigational Oncomine Dx Target Tests were included in the present analysis if the archival sample was a CNB, FNA, or surgical resection sample. The laboratory data included the type of sample tested and the date stamps for completion of the procedural steps, if successful. Investigational Oncomine Dx Target Tests were considered successfully completed if a date stamp for completion of sequencing was available. The lack of a date stamp for sequencing completion indicated an unsuccessful Oncomine Dx Target Test. A successfully completed Oncomine Dx Target Test sequenced and reported the genetic results for all the target genes; however, an unsuccessful test would not be able to report the genetic results for any target genes.
Statistical Analysis
The success rates and tissue consumption in clinical single-gene testing were evaluated by type of single-gene test and by the total number of single-gene tests ordered per sample. The success rate was defined as the percentage of physician-ordered tests or physician-submitted samples for which the biomarker status could be reported, regardless of test initiation or sample rejection. Each success rate was evaluated only if ≥ 10 applicable tests or samples were available in the data set.
Analyses stratified by the number of tests per clinical sample were conducted to evaluate both the resources required to test increasing numbers of single-gene tests and the ability of the current single-gene testing paradigm to provide information to support first-line treatment decisions for aNSCLC. The total tissue consumption per sample was evaluated by the number of tests on that sample with data available on the number of slides cut. The average number of slides required to attempt a certain number of single-gene tests was assessed across samples that had slides cut for that exact number of tests. The probability of successfully completing a certain number of single-gene tests was evaluated among samples on which at least that many included tests had been ordered. For example, the probability that 5 single-gene tests could be successfully completed on a single sample was evaluated among the clinical samples for which ≥ 5 single-gene tests had been ordered.
Overall analyses were conducted across all clinical samples to describe the success rates and tissue consumption of the single-gene testing paradigm in general. Differences between the success rates for increasing numbers of single-gene tests were evaluated using the χ2 test if the number of samples was > 5 and the Fisher exact test otherwise. The success rates for > 1 single-gene test were compared to the success rate of completing 1 single-gene test. The differences in sequential success rates were also evaluated (eg, 2 vs. 3 tests, 3 vs. 4 tests, etc.). Statistical significance was considered present at P < .05. Statistical analyses were conducted using SAS, version 9.4 (SAS Institute, Cary, NC).
The success rates and tissue consumption with the investigational Oncomine Dx Target Test were not assessed overall across all the archived samples but were assessed for different lung tumor sample subcategories. Subanalyses by sample type and tumor content were conducted for both the clinical single-gene testing and the investigational use of the Oncomine Dx Target Test.
Subanalyses were conducted for the sample types present among both the clinical and the archival samples: CNB, FNA, or surgical resection. Because most tissue samples submitted for aNSCLC testing were CNBs, further subanalyses were conducted to evaluate both the clinical and the archival CNB samples stratified by tumor content. CNB samples for which tumor content data were available were evaluated in 2 subgroups: < 25% and ≥ 25% tumor content. CNB samples for which data on tumor content were missing were evaluated as a separate “unknown” category. Just as with the overall analysis of clinical single-gene testing, the success rates in the subanalyses were only evaluated if ≥ 10 tests or samples were available.
Results
Data from the reference laboratory on 3659 physician-ordered clinical single-gene tests were included in the present analysis. The clinical single-gene tests were ordered on 1402 samples from 1368 patients. The clinical samples for testing were predominantly from biopsies, with CNBs comprising 70.5% of the submitted samples and FNAs comprising an additional 10.0% (Table 1). The range in tumor content was as high as 100% for all sample types and as low as 2% among CNBs, 5% among FNAs, 5% among surgical resection samples, and 1% among cell blocks. Of the CNB samples, 21.2% had < 25% tumor content recorded.
Table 1Tissue Sample Characteristics
Characteristic
Sample Type
All Sample Types
CNB
FNA
Resection
Cell Block
Any
Tumor Content
< 25%
≥ 25%
NA
Clinical samples submitted for single-gene testing
Samples, n (% of total)
988 (70.5)
209 (14.9)
685 (48.9)
94 (6.7)
140 (10.0)
167 (11.9)
107 (7.6)
1.402 (100.0)
Patients, n
971
209
677
92
140
165
104
1,368
Tumor content of samples
Samples with tumor content data, n
894
209
685
0
112
166
82
1,254
Tumor proportion, %
Mean
45.2
18.3
53.4
NA
53.2
56.2
42.3
47.2
Range
2-100
2-22
25-100
NA
5-100
5-100
1-100
1-100
Tests ordered, n (% of samples)
EGFR therascreen
947 (95.9)
201 (96.2)
663 (96.8)
83 (88.3)
131 (93.6)
153 (91.6)
102 (95.3)
1,333 (95.1)
ALK Vysis
745 (75.4)
160 (76.6)
542 (79.1)
43 (45.7)
113 (80.7)
128 (76.6)
71 (66.4)
1,057 (75.4)
ROS1 LDT
480 (48.6)
100 (47.8)
345 (50.4)
35 (37.2)
61 (43.6)
73 (43.7)
49 (45.8)
663 (47.3)
BRAF cobas
66 (6.7)
13 (6.2)
47 (6.9)
6 (6.4)
3 (2.1)
5 (3.6)
5 (4.7)
79 (5.6)
BRAF LDT
16 (1.6)
2 (1.0)
14 (2.0)
0 (0.0)
2 (1.4)
2 (1.2)
1 (0.9)
21 (1.5)
KRAS LDT
146 (14.8)
33 (15.8)
105 (15.3)
8 (8.5)
12 (8.6)
41 (24.6)
10 (9.3)
209 (14.9)
MET LDT
101 (10.2)
18 (8.6)
79 (11.5)
4 (4.3)
5 (3.6)
16 (9.6)
7 (6.5)
129 (9.2)
RET LDT
99 (10.0)
17 (8.1)
78 (11.4)
4 (4.3)
5 (3.6)
13 (7.8)
8 (7.5)
125 (8.9)
FGFR1 LDT
29 (2.9)
5 (2.4)
22 (3.2)
2 (2.1)
2 (1.4)
4 (2.4)
3 (2.8)
38 (2.7)
ERBB2 LDT
4 (0.4)
0 (0.0)
3 (0.4)
1 (1.1)
1 (0.7)
0 (0.0)
0 (0.0)
5 (0.4)
Archival samples for investigational use of Oncomine Dx Target Test
Samples, n
69 (40.8)
41 (24.3)
28 (16.6)
0 (0.0)
13 (7.7)
87 (51.5)
0 (0.0)
169 (100.0)
Tumor, %
NA
NA
Mean
29.0
15.0
49.5
28.9
42.4
35.9
Range
1-90
1-20
25-90
1-90
5-90
1-90
Abbreviations: CNB = core needle biopsy; FNA = fine needle aspiration; LDT = laboratory developed test.
The laboratory data also included 169 investigational use cases of the Oncomine Dx Target Test. Compared with the clinical samples, the archival samples used for the investigational Oncomine Dx Target Tests included a greater proportion of surgical resection samples (11.9% of clinical samples vs. 51.5% of archival samples) and a lower proportion of CNBs (70.5% of clinical samples vs. 40.8% of archival samples). Regardless of sample type, the archival samples tended to have lower tumor content than the clinical samples. The range in tumor content among the archival samples was ≤ 90% across all sample types and as low as 1% among the CNBs or FNAs and 5% among the surgical resection samples.
Across all clinical samples submitted for testing, the single-gene tests most commonly ordered by physicians were EGFR therascreen (95.1%), ALK Vysis (75.4%), and ROS1 LDT (47.3%; Figure 1). Among the clinical samples on which single-gene tests for ≥ 3 biomarkers were ordered, a large majority included tests for EGFR, ALK, and ROS1. When single-gene tests for ≥ 4 biomarkers were ordered on a single sample, the tests for MET, RET, KRAS, or BRAF were commonly included.
Figure 1Probability of Test Order Stratified by Number of Genes
The proportion of physician-ordered single-gene tests for which mutation status was successfully reported varied by test type from 62.4% for RET LDT to 89.1% for ALK Vysis (Figure 2A). Evaluations by sample type found greater success rates on surgical resection samples and lower success rates on FNA samples across nearly all the single-gene tests (Figure 2B). The exception was RET LDT, for which the success rates were lower from the surgical resection samples than from the CNB samples.
Figure 2Clinical Testing Success Rates Across Single-Gene Tests: (A) Overall and (B) Stratified by Sample Type. *ERBB2 Was Not Evaluated Owing to the Small Number of Samples. Success Rates Were Evaluated Only if ≥ 10 Tests Were Available
Tissue consumption analyses included 3314 single-gene tests on 1258 samples. The number of slides needed to run a single-gene test was greatest for the BRAF LDT, with a mean tissue consumption of 7.1 slides per test (n = 20 tests). The KRAS LDT was the second-most tissue-intensive test, consuming 6.8 slides per test on average (n = 180). The EGFR therascreen consumed 2.7 slides per test on average (n = 1107). The MET LDT, RET LDT, ROS1 LDT, FGFR1 LDT, and ALK Vysis tests consumed an average of 2.0 (n = 117), 1.6 (n = 112), 1.3 (n = 614), 1.3 (n = 32), and 1.1 (n = 979) slides per test, respectively. BRAF cobas (n = 67) and ERBB2 LDT (n = 2) each consumed 1.0 slide per test.
Some clinical samples did not have sufficient tissue to allow slides to be cut for all the single-gene tests requested by the ordering physician. For those samples and tests for which slides could be cut, the average number of slides consumed per sample increased steadily with the number of biomarker tests attempted (Figure 3A). This likely reflects histotechnologist standard operating procedures for uniformity of sectioning and not pathologic evaluation of the amount of tumor present in the different types of samples. When slides could be cut to run tests, tissue consumption increased from 2.4 slides to run 1 single-gene test, to 8.7 to run 4 single-gene tests, to 17.0 slides to run 8 single-gene tests on a single sample.
Figure 3Multiple Biomarker Testing on Clinical Samples Stratified by Number of Single-Gene Tests. (A) Tissue Consumption Stratified by Number of Tests Run. (B) Testing Success Rates Stratified by Number of Single-Gene Tests Ordered. Lines Show (A) the Number of Tissue Slides Cut to Attempt Exactly Each Number of Single-Gene Tests or (B) the Testing Success Rates for at Least Each Number of Single-Gene Tests. Columns Show the Number of Samples Available for Each Evaluation. *Success Rates Were Evaluated Only if ≥ 10 Samples Were Available
When multiple single-gene tests were ordered for the same clinical sample, the probabilities of successfully reporting the molecular status for multiple biomarkers decreased as the number of biomarkers increased (Figure 3B). Across all submitted clinical samples, 88.4% had successfully reported mutation status results from ≥ 1 single-gene test. Among the samples on which ≥ 4 tests were ordered, the probability of having successfully reported the mutation status for ≥ 4 biomarkers was 76.6%. The success rates for ≥ 2 tests through ≥ 7 tests were all significantly lower than the success rate for ≥ 1 test (P < .05 for ≥ 2 tests; P < .0001 for ≥ 3-7 tests). Furthermore, the probability of successfully completing ≥ 3 tests was significantly lower than the probability of successfully completing ≥ 2 tests (P < .0001). Subsequent sequential differences in the success rates (eg, ≥ 3 vs. ≥ 4 tests) did not reach statistical significance, although the difference in the success rates for ≥ 4 versus ≥ 5 tests did show marginal statistical significance (P = .052). Even so, the observed success rates continued to show a trend downward for greater numbers of single-gene tests.
The evaluations stratified by sample type found that the numbers of slides cut to run multiple single-gene tests on clinical samples were similar across CNBs, FNAs, and surgical resection samples, again likely reflecting histotechnologist standard operating procedures for uniformity of sectioning and not pathologic evaluation of the amount of tumor present in the different sample types (Figure 4A). Investigational use of the Oncomine Dx Target Test was artificially restricted to 1 slide per test, regardless of the sample type, in accordance with the pre-established protocol specific to the investigational use testing study for which they were conducted. Similar to clinical single-gene testing, the success rates with the Oncomine Dx Target Test were lowest for the FNA samples and greatest for the surgical resection samples (Figure 4B). The success rates with the investigational use Oncomine Dx Target Test was 69.2% for the FNA samples, 75.4% for the CNB samples, and 98.9% for the surgical resection samples. These success rates decreased between the clinical single-gene testing success rates for 3 and 4 biomarkers on FNA samples (75.0% and 40.0%, respectively), 4 and 5 biomarkers on CNB samples (77.9% and 70.8%, respectively), and 1 and 2 biomarkers on surgical resection samples (100.0% and 98.5%, respectively).
Figure 4Multiple Biomarker Testing Stratified by Sample Type for Clinical Single-Gene Testing and Investigational Use of the Oncomine Dx Target Test. (A) Tissue Consumption and (B) Testing Success Rates. Lines Show (A) the Number of Tissue Slides Cut to Attempt Exactly Each Number of Single-Gene Tests and (B) the Testing Success Rates for At Least Each Number of Single-Gene Tests. The Numbers of Samples Available for Each Evaluation Are Graphed as Columns and Shown in Parentheses in the Table. *Success Rates Were Evaluated Only if ≥ 10 Samples Were Available
Further subgroup analyses of the CNB samples stratified by tumor content found similar numbers of slides consumed to run multiple single-gene tests on samples with < 25% tumor content compared with samples with ≥ 25% tumor content (Figure 5A). When multiple single-gene tests were ordered, the rates of successfully reporting the mutation status for multiple biomarkers were consistently greater among the samples with ≥ 25% tumor content than among the samples with < 25% tumor content (Figure 5B). The tumor content was not reported for 9.5% of the clinical CNB samples (n = 94 of 988). The success rates were very low among these CNB samples with unknown tumor content. The success rates with the investigational use Oncomine Dx Target Test were 70.7% for CNB samples with < 25% tumor content and 82.1% for CNB samples with ≥ 25% tumor content. Just as with the overall analysis of the success rates on CNB samples, these success rates were between the clinical single-gene testing success rates for 4 and 5 biomarkers.
Figure 5Multiple Biomarker Testing on Core Needle Biopsy (CNB) Samples Stratified by Tumor Content: Clinical Single-Gene Testing and Investigational Use of the Oncomine Dx Target Test: (A) Tissue Consumption and (B) Testing Success Rates. Lines Show (A) the Number of Tissue Slides Cut to Attempt Exactly Each Number of Single-Gene Tests, and (B) the Testing Success Rates for at Least Each Number of Single-Gene Tests. The Numbers of Samples Available for Each Evaluation Are Graphed as Columns and Are In Parentheses in the Table. *Success Rates Were Evaluated Only if ≥ 10 Samples Were Available
Most of the 1402 lung cancer patients’ tumor tissue samples submitted by physicians for clinical testing at this large US-based reference laboratory were small samples. More than three fourths were CNB samples and another one tenth were FNA samples. Among the CNB samples, more than one fifth had < 25% tumor content. These findings demonstrate the small amounts of tissue available for most aNSCLC patients who attempt to use genetic testing to inform treatment decisions.
The amount of tissue required to attempt a single-gene test for these patients varied widely across test types. Although the BRAF LDT had the greatest average tissue consumption among the single-gene tests (7.1 slides), the laboratory switched to the FDA-approved BRAF cobas test (1.0 slide) during the study period, reducing the tissue consumption necessary to test for BRAF-activating mutations. KRAS has been acknowledged by the recommended guidelines as having prognostic value in lung cancer. However, running the KRAS LDT required an average of 6.8 slides per test. The EGFR therascreen required 2.7 slides per test on average to run, and the other single-gene tests required < 2 slides per test on average. For some samples in this real-world analysis, some of the physician-ordered single-gene tests were not attempted and no slides were cut, likely owing to insufficient or depleted tumor tissue.
Across all physician-ordered tests in the data set, the success rates for different types of single-gene tests were all < 90%, including those for activating mutations addressed by FDA-approved targeted therapies. The ALK Vysis testing success rate was greatest at 89.1%. The success rates for EGFR therascreen or BRAF cobas were both ∼83%, and the success rate for the ROS1 LDT was only 76.6%. The subanalyses showed that the success rates were greatest with the surgical resection samples but lowest for the FNA samples, showing that lung cancer patients with smaller samples have greater challenges in accessing the genetic information needed to guide targeted treatment decisions.
Physicians ordered > 1 single-gene test for nearly 80% of the submitted clinical samples. The number of slides needed to attempt multiple single-gene tests increased steadily with the number of biomarkers to be tested. At the same time, the probability of successfully reporting the mutation status results from the additional single-gene tests decreased. Only 88.4% of the clinical samples submitted for single-gene testing had ≥ 1 biomarker successfully reported, suggesting that > 1 in 10 of these patients (11.6%) had no genetic information available to inform their treatment decisions. The success rates for determining the mutation status decreased significantly for 2 biomarkers and 3 biomarkers and continued to show a trend downward after that. Currently, FDA-approved targeted therapies are available for activating mutations in 4 genes in aNSCLC. The success rate for reporting the mutation status for 4 biomarkers in the present study was 76.6%, suggesting that the necessary genetic information to guide appropriate selection for current targeted therapies might be unavailable for 23.4% of aNSCLC patients in the current single-gene testing paradigm.
The subanalyses stratified by sample type and tumor content included both clinical single-gene testing and investigational use of the Oncomine Dx Target Test. Although tissue consumption for multiple single-gene tests did not vary widely by sample type or tumor content, the testing success rates showed more variation. The success rates were consistently lower among CNB than among surgical resection samples and were lower still among FNA samples. Among the CNB samples, the success rates were consistently lower among those samples with < 25% tumor content compared with those with ≥ 25% tumor content. The lower success rates underscore the difficulties in molecular assessment for patients with aNSCLC whose samples are mostly small and for whom the number of therapeutically relevant biomarkers is only increasing.
Investigational use of the Oncomine Dx Target Test was artificially restricted to a single tissue slide based on a previous study’s protocol but still showed testing success rates that were at least comparable to the clinical success rates for ≥ 2 single-gene tests on surgical resection samples, ≥ 4 single-gene tests on FNA samples, and ≥ 5 single-gene tests on CNB samples. Running that number of single-gene tests required a mean of 4.5 slides, 9.8 slides, and 8.5 slides per surgical resection, FNA, and CNB sample, respectively. Although only 1 slide was used for each investigational Oncomine Dx Target Test, the FDA-approved protocol has recommended 2 slides when testing a surgical resection sample and 9 when testing a CNB sample. Therefore, in the real-world clinical testing setting, in which the number of tissue slides is limited only by the amount of sample available, the success rates and tissue consumption with the Oncomine Dx Target Test might be greater than that reported in the present retrospective study.
A previously reported prospective study conducted from July 2011 to June 2013 at the University of Pittsburgh Medical Center (UPMC) conducted multiple biomarker testing on CNB and FNA lung tumor tissue samples for EGFR, ALK, and KRAS, specifically. The mutation status for all three genes could be successfully reported for 67% of the CNB samples and 46% of FNA samples.
The present retrospective study found testing success rates for 3 biomarkers of 76.2% among CNB samples and 75.0% among FNA samples. The differences in testing success rates for the 3 biomarkers between these 2 studies might be attributable to the specific genes tested. The study at UPMC specifically aimed to test EGFR, ALK, and KRAS on all their samples. In contrast, although the orders in the present study for ≥ 3 single-gene tests on a single sample overwhelmingly included EGFR and/or ALK testing (98.7% and 97.9%, respectively), only 23.8% included KRAS testing. The present study showed that evaluating KRAS mutation status required the largest number of tissue slides of all the included biomarkers. The rate of success for reporting the mutation status of 3 biomarkers might have been greater in the present study than in the UPMC study owing to the inclusion of less tissue-intensive single-gene tests. This suggests that the success rates for single-gene testing for multiple biomarkers might be even lower than reported in the present study for patients and physicians interested in the prognostic information provided by KRAS.
Study Limitations
The retrospective clinical testing data contained some clinical samples for which data on the number of slides cut for single-gene tests or the tumor content of the sample were missing. Analyses of tissue consumption excluded the single-gene tests in which the number of slides cut was missing. However, these analyses included 90% of all single-gene tests and samples in the data set, suggesting that the results are representative of single-gene testing at this large, US-based, Clinical Laboratory Improvement Amendments–certified, commercial, reference laboratory.
Samples for which tumor content was unknown were included and evaluated separately in the subanalyses of CNB samples stratified by tumor content. The subanalyses found that the tissue consumption and success rates among these samples were very low compared with that from the samples for which the tumor content was known. This suggests that these samples might have been smaller or might have had lower tumor content. Therefore, the true success rates for CNB samples with low tumor content (< 25%) might be lower than reported in the present study.
The clinical samples included in the present analysis were submitted from September 2015 to October 2016. At that time, the only FDA-approved, National Comprehensive Cancer Network guideline-recommended, first-line targeted therapies for aNSCLC patients were for EGFR or ALK activating mutations.
As expected, the EGFR therascreen and ALK Vysis were consistently the most commonly ordered single-gene tests in the present study, and ≥ 3 single-gene tests were ordered for only 14.6% (n = 205 of 1402) of the clinical samples. Since then, targeted therapies for aNSCLC patients with ROS1 or BRAF activating mutations have been approved by the FDA and recommended by the National Comprehensive Cancer Network guidelines.
With these developments, the genetic testing ordering patterns in clinical practice are likely to include single-gene tests for ROS1 and BRAF more frequently than was described in the present study. Furthermore, with more genes identified as therapeutically relevant to aNSCLC, it is more likely that additional biomarkers will be ordered on a single sample. Therefore, the distribution of genetic testing order patterns described in the present study might no longer be representative of the current single-gene testing paradigm, and the results among samples with larger numbers of single-gene tests ordered might be more applicable to the molecular assessment needs for targeted treatment selection for patients with aNSCLC in current clinical practice.
The sample characteristics available for both sets of samples were limited to sample type and tumor content. Other potentially relevant characteristics were unknown and could limit the comparability of these findings on clinical single-gene testing and the investigational Oncomine Dx Target Test. The clinical single-gene testing was run on a large number of real patient samples submitted for testing and is therefore likely to be reflective of clinical practice and real-world genetic testing. In comparison, relatively fewer Oncomine Dx Target Tests were included in the present study, and the archival tissue samples were purchased commercially, were more likely from surgical resections, and had lower tumor content regardless of sample type. Therefore, the results with the Oncomine Dx Target Test were presented by sample type and, for CNBs, by tumor content group. However, other unobserved characteristics could have also influenced the success rates. Future prospective studies might benefit from evaluating more runs of the FDA-approved Oncomine Dx Target Test ordered by physicians as part of real-world clinical testing and collecting and controlling for more sample characteristics.
Finally, the retrospective data on the investigational use of the Oncomine Dx Target Test were from tests run using a previous study’s protocol which limited each test to using only a single slide. This differs from the FDA-approved protocol, which has recommended 2 tissue slides to test surgical resection samples and 9 tissue slides to test CNB samples. Therefore, the potential mean tissue consumption of the newly FDA-approved Oncomine Dx Target Test for clinical use cannot be estimated from our results. Both tissue consumption and success rates might be greater in clinical settings in which additional slides can be used.
Conclusion
The large majority of lung cancer samples submitted for genetic testing were small tissue samples. Sequential single-gene testing can require large numbers of tissue slides and might not be able to determine the mutation status for all relevant biomarkers for aNSCLC, especially for patients with smaller sample types or lower tumor content samples. This highlights the need for tissue stewardship and efficient molecular assessment methods for lung cancer, especially as new targeted therapies are developed for activating mutations in additional genes. The present preliminary assessment of the investigational use of the Oncomine Dx Target Test suggests that it could provide a viable option for testing multiple biomarkers with fewer slides required for patients with aNSCLC. Additional studies of this test in the clinical setting would be useful to inform aNSCLC genetic testing decisions and strategies to inform treatment selection.
Clinical Practice Points
•
Several targeted aNSCLC treatments have shown the potential to improve outcomes by delaying disease progression for patients with activating mutations.
•
Directing patients to appropriate treatment rests on the ability to complete multiple biomarker genetic testing using the small tissue samples available for lung cancer patients.
•
Tissue stewardship is essential as more treatments are developed and more genes need to be tested.
•
The present retrospective analysis of clinical testing for lung cancer patients showed that sequential single-gene testing required increasing tissue consumption and had decreasing rates of successful tests completion as the number of genes tested increased.
•
Investigational use of the Oncomine Dx Target Test on 1 tissue slide per test demonstrated success rates comparable to, or better than, those with single-gene testing for 2 genes on a surgical resection sample, 4 genes on an FNA sample, and 5 genes on a CNB sample, suggesting that it might facilitate multiple biomarker assessments on small tissue samples to inform treatment decisions for aNSCLC patients.
Disclosure
T.M.Y., A.T., and A.J.L. are employees of Navigant Consulting, Inc. C.M. is an OmniSeq employee and stockholder and an employee of Roswell Park Cancer Institute. The remaining author declares that he has no competing interests.
Acknowledgments
The authors acknowledge and thank Dr Raj Stewart for his insights and assistance in the writing of this paper. The authors further thank Thermo Fisher Scientific for providing funding for this research to Quorum Consulting, Inc, now Navigant Consulting, Inc.
Supplemental Data
Supplemental Table 1Gene Variants Included in the Oncomine™ Dx Target Test
Erlotinib versus standard chemotherapy as first-line treatment for European patients with advanced EGFR mutation-positive non-small-cell lung cancer (EURTAC): a multicentre, open-label, randomised phase 3 trial.
30204-3&id=gr1.jpg){kind=link}
30204-3&id=gr2.jpg){kind=link}
30204-3&id=gr3.jpg){kind=link}
30204-3&id=gr4.jpg){kind=link}
30204-3&id=gr5.jpg){kind=link}