Clinical and analytical validation of FoundationOne®CDx, a comprehensive genomic profiling assay for solid tumors

FoundationOne®CDx (F1CDx) is a United States (US) Food and Drug Administration (FDA)-approved companion diagnostic test to identify patients who may benefit from treatment in accordance with the approved therapeutic product labeling for 28 drug therapies. F1CDx utilizes next-generation sequencing (NGS)-based comprehensive genomic profiling (CGP) technology to examine 324 cancer genes in solid tumors. F1CDx reports known and likely pathogenic short variants (SVs), copy number alterations (CNAs), and select rearrangements, as well as complex biomarkers including tumor mutational burden (TMB) and microsatellite instability (MSI), in addition to genomic loss of heterozygosity (gLOH) in ovarian cancer. CGP services can reduce the complexity of biomarker testing, enabling precision medicine to improve treatment decision-making and outcomes for cancer patients, but only if test results are reliable, accurate, and validated clinically and analytically to the highest standard available. The analyses presented herein demonstrate the extensive analytical and clinical validation supporting the F1CDx initial and subsequent FDA approvals to ensure high sensitivity, specificity, and reliability of the data reported. The analytical validation included several in-depth evaluations of F1CDx assay performance including limit of detection (LoD), limit of blank (LoB), precision, and orthogonal concordance for SVs (including base substitutions [SUBs] and insertions/deletions [INDELs]), CNAs (including amplifications and homozygous deletions), genomic rearrangements, and select complex biomarkers. The assay validation of >30,000 test results comprises a considerable and increasing body of evidence that supports the clinical utility of F1CDx to match patients with solid tumors to targeted therapies or immunotherapies based on their tumor’s genomic alterations and biomarkers. F1CDx meets the clinical needs of providers and patients to receive guideline-based biomarker testing, helping them keep pace with a rapidly evolving field of medicine.

FoundationOne ® CDx (F1CDx) is a United States (US) Food and Drug Administration (FDA)-approved companion diagnostic test to identify patients who may benefit from treatment in accordance with the approved therapeutic product labeling for 28 drug therapies. F1CDx utilizes next-generation sequencing (NGS)-based comprehensive genomic profiling (CGP) technology to examine 324 cancer genes in solid tumors. F1CDx reports known and likely pathogenic short variants (SVs), copy number alterations (CNAs), and select rearrangements, as well as complex biomarkers including tumor mutational burden (TMB) and microsatellite instability (MSI), in addition to genomic loss of heterozygosity (gLOH) in ovarian cancer. CGP services can reduce the complexity of biomarker testing, enabling precision medicine to improve treatment decision-making and outcomes for cancer patients, but only if test results are reliable, accurate, and validated clinically and analytically to the highest standard available. The analyses presented herein demonstrate the extensive analytical and clinical validation supporting the F1CDx initial and subsequent FDA approvals to ensure high sensitivity, specificity, and reliability of the data reported. The analytical validation included several in-depth evaluations of F1CDx assay performance including limit of detection (LoD), limit of blank (LoB), precision, and orthogonal concordance for SVs (including base substitutions [SUBs] and insertions/deletions [INDELs]), CNAs (including amplifications and homozygous deletions), genomic rearrangements, and select complex biomarkers. The assay validation of >30,000 test results comprises a considerable and increasing body of evidence that supports the clinical utility of F1CDx to match patients with solid tumors to targeted therapies or immunotherapies based on their tumor's genomic alterations and biomarkers. F1CDx meets the clinical needs of providers and patients to receive a1111111111 a1111111111 a1111111111 a1111111111 a1111111111

Introduction
The advances in our understanding of cancer biology over the last two decades have underscored the genomic heterogeneity among tumors of the same type, as well as similar driver mechanisms in different cancer types, enabling both tailored approaches to individual patient care as well as pan-tumor therapeutic indications [1][2][3][4]. The rapidly expanding repertoire of targeted cancer therapies provides significant challenges for oncologists treating patients, as keeping up with companion diagnostic approvals, clinical trials, and evolving understanding of cancer biology can exceed the capacity of many physicians. Testing for genomic alterations is essential for selecting the most appropriate targeted treatments or immunotherapies and clinical trials, and for avoiding futile treatments where resistant tumors are present or actionable targets are absent [5,6].
The ability to treat cancer patients with precision medicine requires an approach to identify targetable alterations which may occur in many different genes, alteration types and complex genomic signatures. With the proliferation of highly effective targeted therapies, determining the best treatment may not always be possible by examining the patient or running a singleanalyte test, but rather requires a molecular diagnostic that is both time-and tissue-efficient while also delivering accurate and reproducible performance. Highly validated testing solutions that comprehensively cover actionable alterations and deliver the latest literature-based guidance on potential treatment options, along with relevant clinical trials and levels of evidence, offer a streamlined and comprehensive solution to oncologists and provide patients with reassurance that even rare but potentially treatable alterations can be identified [7][8][9][10][11][12][13]. Hot-spot and smaller-panel testing approaches require a-priori assumptions about the alterations that may be present and can miss actionable alterations that are beyond their narrow scope [14], including novel alterations that may be treatable or confer resistance, such as oncogene fusions [15][16][17][18][19][20]. In contrast, broader whole exome sequencing (WES; note all abbreviations/acronyms are available in the S1 Appendix, Table S1 in S1 Appendix [all supplemental tables are located in S1 Appendix]) or whole genome sequencing (WGS) raise different challenges for clinicians. WES evaluates all exonic protein-coding regions of the human genome, encompassing~30 million base pairs (~1% of the human genome), while WGS aims to sequence the entire genome. These efforts produce data from an excessive breadth of the genome, but generally deliver lower depth of sequencing that results in a higher LoD relative to broad-panel assays [21,22]. Further, WES and WGS approaches can entail higher cost, complex analytics, and longer turnaround time, but critically, they may also burden the report recipient with many variants of unknown significance (VUS) from genes not relevant to therapeutic decision-making for the patient [23][24][25][26][27].
CGP approaches are an increasingly valuable and important part of the molecular characterization of the tumor and subsequent selection of the most relevant treatment options for patients with advanced cancer, and can bridge the gap between the narrow focus of singlemarker or hot-spot tests and the lower sensitivity of WES or WGS. Clinical practice guidelines support use of CGP for an increasing number of cancer types including non-small cell lung cancer (NSCLC) [ patient genomic data, which contains potentially identifying or sensitive patient information and cannot be reported in a public data repository. FMI is committed to collaborative data analysis and we have well-established, and widely utilized mechanisms by which investigators can query our core genomic database of >400,000 de-identified sequenced cancers. Detailed data may be obtained by contacting the corresponding author or the Foundation Medicine Data Governance Council at data.governance.council@foundationmedicine. com.
critical safeguards to protect patients from potentially harmful tests and wasted resources [42]. It is crucial to apply rigorous standards consistently across tests with high-risk intended uses such as those used for cancer management. The rigor of the FDA's premarket review-a process grounded in the highest standards of scientific and medical integrity and transparencybuilds trust. By enforcing high evidentiary standards for these high-risk tests, regulatory authorities minimize the risks to patients and payers while enabling access to the most reliable test providers [43], as demonstrated by the FDA/Centers for Medicare & Medicaid Services (CMS) parallel review process [44]. The FDA's Center for Drug Evaluation and Research plays a crucial role in assuring access to safe and effective medicines based on a minimal standard of evidence that demonstrates a drug is likely to be beneficial, and not excessively harmful, to a given treatment population [45]. As treatment costs rise and criteria for selecting eligible treatment populations become more complex, the quality of methods for testing those criteria is under increased scrutiny. For complex predictive biomarkers like MSI, TMB, and gLOH, high quality standards for defining pre-analytics, sample handling, and validation are especially important, as a growing number of therapeutic decisions are influenced by these signatures [24,46,47]. Further, for complex biomarkers, such as TMB, it is essential to use a broad-panel test that comprehensively evaluates a large amount of genomic content and that is well-validated across tumor types. FDA approvals of companion diagnostics in a pan-tumor setting for TMB �10 mutations per megabase (mut/Mb) for pembrolizumab and NTRK-fusions for larotrectinib demonstrate the requirement for CGP in order to optimize therapeutics in oncology and ensure patients have the opportunity to access to potentially beneficial therapy [28][29][30][31][32][33][34][48][49][50][51][52][53].
The evidence used to demonstrate robustness of a diagnostic includes analytical validity, clinical validity, and clinical utility. Analytical validity of an NGS assay refers to how accurate and reliable the test is in identifying a particular analyte, such as a genomic alteration or genomic signature [54]. Clinical validity refers to a companion diagnostic's capability to predict, measure, or detect a clinical condition or status for its intended use [55]. While clinical validity refers to the relationship between an analyte and the presence or absence of a particular disease state, clinical utility refers to the ability of test results to predict improved health outcomes from an intervention [54]. The FDA requires robust evidence of analytical and clinical validation, exceeding standards set by Clinical Laboratory Improvement Amendments (CLIA), prior to approval or clearance of a diagnostic device, in order to assure consumers of the quality of results and ensure tests can be used to provide information to support the diagnosis, treatment, management, or prevention of a disease [56]. The FDA approvals of validated clinical trial assays (CTAs) as companion diagnostics support improved therapeutic results and directly demonstrate both clinical validity and clinical utility. The FDA also enables test providers to develop companion diagnostic claims through clinical bridging when trial specimens are available, or through non-inferiority studies when clinical trial specimens are not available. Through this pathway, companion diagnostic claims can be established for validated assays that have been demonstrated to be concordant or non-inferior to the clinical trial assays that demonstrated utility in pivotal trials. Further, a candidate companion diagnostic can demonstrate the ability to predict clinical efficacy of a corresponding therapeutic with comparable efficacy to that observed in the associated clinical trial. For example, given an observed therapeutic efficacy, if the candidate companion diagnostic were used to identify patients that would respond to treatment, instead of the clinical trial assay, the observed therapeutic efficacy will have been the same.
FDA's Center for Devices and Radiological Health [57] is responsible for establishing and ensuring quality standards by evaluating and regulating the performance of the companion diagnostics [58], particularly for high-risk testing services such as those for cancer genotyping for therapeutic decision-making [59]. Foundation Medicine profiles >100,000 tumor specimens each year across its laboratories in Cambridge, MA, USA, Research Triangle Park, NC, USA, and Penzberg, Germany. As Foundation Medicine is both the manufacturer and laboratory service provider, rigorous validation is performed to meet the requirements of both the FDA and CLIA.
The analyses presented here describe the broad analytical and clinical validation of F1CDx (Foundation Medicine; Cambridge, MA), a leading CGP platform. The validations of MSI, TMB, and gLOH have been further described in detail elsewhere [60-62].

FoundationOne 1 CDx assay
F1CDx is an FDA-approved NGS-based in vitro diagnostic device that targets 324 cancerrelated genes. F1CDx is performed exclusively by Foundation Medicine as a central laboratory testing service using DNA extracted from FFPE tumor samples from cancer patients [63]. The laboratory components are performed within Foundation Medicine's College of American Pathologists (CAP)-accredited laboratory that has been certified by CMS pursuant to CMS procedures for the certification of international laboratories in accordance with CLIA. The F1CDx assay is regulated by the FDA in accordance with FDA's standard of a reasonable assurance of safety and effectiveness, together with FDA's Quality System regulations (21 CFR Part 820). Additional clinical decision insights and genomic analyses are provided as a professional laboratory service under CLIA and CAP regulations.
F1CDx is the result of the evolution of Foundation Medicine's original FoundationOne 1 assay [64]. FoundationOne 1 testing has been used for clinical testing since 2012. The assay workflow is presented in Fig 1. The assay employs a single DNA extraction method from routine FFPE biopsy or surgical resection specimens; 50−1,000 ng of DNA undergo wholegenome shotgun library construction and hybridization-based capture of all coding exons from 309 cancer-related genes, including one promoter region, one non-coding (ncRNA), and select intronic regions from 34 commonly rearranged genes, 21 of which also include the coding exons (refer to Tables S2 and S3 in S1 Appendix for the complete list of genes included in F1CDx). Using an Illumina 1 NGS platform, hybrid capture-selected libraries are sequenced to high uniform depth (targeting >500X median coverage with >99% of exons at coverage >100X). Sequence data are then processed using a purpose-built analysis pipeline designed to detect all classes of genomic alterations, including SVs, CNAs (amplifications and homozygous gene deletions), and select genomic rearrangements (e.g., gene fusions). Genomic signatures including MSI and TMB are reported, in addition to positive homologous recombination deficiency (HRD) status (tumor BRCA-positive and/or gLOH high) in ovarian cancer.

Bioinformatic methods
Sequence data are analyzed using a proprietary software system developed by Foundation Medicine. Sequence data are mapped to the human genome (hg19) using Burrows-Wheeler Aligner (BWA; a program for aligning sequence reads with large-scale reference genomes) [65]. Polymerase chain reaction (PCR) duplicate read removal and sequence metric collection are performed using Picard and SAMtools [66]. Variant calling is performed only in genomic regions targeted by the test.
For the detection of SVs and rearrangements, a de novo assembly is performed. This is done using software to generate a de Bruijn graph including all k-mers in reads mapping to a particular locus [67]. For each variant, there is a set of k-mers supporting the variant and a set of kmers that would support the reference or another variant at the location. Each candidate variant is then scanned against reads in the locus to identify which support the candidate variant, a different variant, or the reference at the location. The supporting reads for each candidate variant are analyzed, and metrics used to evaluate the quality of the variant call are calculated. The final variant calls are made based on a series of quality control (QC) filters, which will reject a call based on the intrinsic sample noise, the expected noise level for the particular variant, and other known error modes (e.g., sequence homology).
Short variants are reported as a frequency relative to overall depth of coverage; this is denoted as mutation allele frequency (MAF). Alterations are classified as "known," "likely," or "unknown" based on their status in COSMIC. Alterations classified as "other" include truncating events in tumor suppressor genes (splice, frameshift, and nonsense) as well as variants that appear in hot-spot locations but do not have a specific COSMIC association. Variants may be classified as VUS when the significance and impact upon cancer progression is unknown due to a lack of reported evidence and conclusive change in function. It is recognized that some genetic alterations and variants will not impact functionality and do not increase cancer risk.
Genomic rearrangements are identified by analyzing chimeric read pairs. Chimeric read pairs are defined as read pairs for which reads map to separate chromosomes, or at a distance of over 2 kilobase pairs (kbp). Pairs are clustered by genomic coordinate of the pairs, and clusters containing at least five chimeric pairs (three for known fusions) are identified as rearrangement candidates. Filtering of candidates is performed by mapping quality (MQ >30) and distribution of alignment positions (standard deviation >10). Rearrangements are annotated for predicted function (e.g., creation of fusion gene).
CNAs are detected using a comparative genomic hybridization-like method [68]. First, a log-ratio profile of the sample is obtained by normalizing the sequence coverage obtained at all exons and genome-wide single-nucleotide polymorphisms (SNPs) against a process-matched normal control. This profile is segmented and interpreted using allele frequencies of sequenced SNPs to estimate tumor purity and copy number at each segment.
Tumor content and purity of a sample are derived separately. Board-certified pathologists will assess tumor content through the enumeration of nucleated tumor cells. The assay requires greater than 20% nucleated tumor cells to enter into the DNA extraction procedure. This upstream assessment is complemented by the downstream computational calculation of tumor purity just described above. The computational tumor purity assessment is calculated based off SNP allele frequencies and is also used to inform the accuracy of copy number modeling and the calling for several complex biomarkers. The underlying copy number modeling used in making these estimations is a simple 2-component system consisting of a mixture of normal diploid cells and aneuploid tumor cells, where the tumor purity corresponds to the fraction of tumor cells in the mixture. The aneuploidy of the tumor cells is modeled as an integral copy number level for each allele of each segment.
To determine MSI status, repetitive loci (minimum of five repeat units of mono-, di-, and trinucleotides) are assessed to determine what repeat lengths are present in the sample. A locus containing a repeat length not present in an internal database generated using >3,000 clinical samples is considered to be unstable. An MSI indicator is generated by calculating the fraction of unstable loci, considering only those loci that achieve adequate coverage for consideration for the sample. TMB is measured by counting coding SVs present at �5% allele frequency and filtering out potential germline variants according to published databases of known germline polymorphisms including Single Nucleotide Polymorphism Database (dbSNP) and Genome Aggregation Database (gnomAD). Additional germline alterations are assessed for potential germline status and filtered out using a somatic-germline/zygosity (SGZ) algorithm [69]. Furthermore, known and likely driver mutations are filtered out to exclude bias. The resulting mutation number is then divided by the coding region corresponding to the number of total variants counted, or approximately 790 kilobases (kb); the resulting number is reported in units of mut/Mb. The clinical validity of TMB defined by this panel has been established for TMB-high (TMB-H) as �10 mut/Mb as a qualitative status.
For ovarian tumor samples, gLOH is measured by the percentage of loss of heterozygosity (LOH) in the tumor genome. To compute gLOH for each tumor, LOH regions are inferred across the 22 autosomal chromosomes using the genome-wide copy number profile and minor allele frequencies of the germline SNPs. Certain LOH regions are excluded from analysis, including: 1) LOH regions spanning �90% of a whole chromosome or chromosome arm, as these LOH events are likely due to non-HRD mechanisms; and 2) regions in which LOH inference is ambiguous. For each tumor, the percentage of the genome with LOH is computed as 100 times the total length of non-excluded LOH regions divided by the total length of nonexcluded regions of the genome. gLOH �16 is defined as "LOH high," gLOH <16 is "LOH low," and an indeterminable result is "LOH unknown." In some cases, due to QC issues such as low tumor purity, noisy CNA data, and contamination that may affect copy number modeling, it is not possible to accurately calculate LOH. In such cases, LOH will be reported as "unknown."

Sample selection
In total, >30,000 test results have been evaluated in F1CDx assay validation studies. Samples used for analytical and clinical validation studies consisted of FFPE solid tumor specimens from patients with cancer, as well as DNA samples selected from an inventory of residual banked DNA isolated from FFPE tumor specimens, including >56 tumor tissue types. Common and rare alterations, including SUBs, INDELs, rearrangements, and CNAs, have been evaluated across the genomic breadth of the baitset, covering >300 genes. The extensive pantumor validation supports the clinical utility of the F1CDx test, which evaluates over 500 cancer disease ontologies. Institutional Review Board (IRB) approval was obtained from the New England IRB prior to use of samples in the described validation studies.
In some instances, such as the evaluation of CNAs or complex biomarkers, targeted allele frequencies were achieved by titrating DNA samples into fragmented genomic DNA derived from peripheral blood mononuclear cells.

Clinical and analytical validity of F1CDx
As presented in Table 1, F1CDx is an FDA-approved companion diagnostic across two pantumor indications (NTRK gene fusions and TMB-high [TMB-H]) as well as in NSCLC, breast, CRC, ovarian, prostate, melanoma, and cholangiocarcinoma. Importantly, as the companion diagnostic indications for F1CDx are constantly evolving, the FDA's list of cleared or approved companion diagnostic devices provides a complete list of the currently approved FDA companion diagnostic claims, including all approved companion diagnostic claims for F1CDx [70]. Each of the companion diagnostic claims for F1CDx were FDA-approved based upon the clinical validity as determined by use of F1CDx as the clinical trial assay, non-inferiority concordance testing against FDA-approved diagnostics for that indication, retrospective or prospective analyses, or clinical bridging studies.

Clinical validation: Non-inferiority
F1CDx was clinically validated through the performance of either clinical bridging or noninferiority. Using non-inferiority, the clinical validity of the F1CDx assay as a companion diagnostic was evaluated for identifying patients with specific disease indications to evaluate patient eligibility for treatment by targeted therapies for defined biomarkers. Concordance between F1CDx and a validated orthogonal comparator assay was assessed through the evaluation of negative percent agreement (NPA) and positive percent agreement (PPA). The orthogonal concordance assay was considered the gold standard within each non-inferiority analysis. Samples were tested across two replicates (denoted as CCD1 and CCD2) by the comparator test and one replicate by F1CDx, as described by Li et al [76]. Detailed concordance results and sample numbers are provided in supplementary tables. The clinical validation performed via the demonstration of non-inferiority is summarized in Table 2.

Clinical validation: Clinical bridging
In addition to non-inferiority, the F1CDx assay was clinically validated through clinical bridging analyses to establish clinical utility. In each scenario, the clinical efficacy as demonstrated by the local clinical trial assays (CTAs) was compared to the clinical efficacy as demonstrated by the F1CDx assay. The clinical bridging validation is summarized in Table 3. In addition, clinical utility of the F1CDx assay was also demonstrated in a pan-tumor setting for TMB-H   (Table 4). The results reported with F1CDx were similar to the clinical efficacy   F1CDx and FoundationFocus™ CDx BRCA assays are equivalent with the exception of an updated analysis pipeline in use for F1CDx and reporting software that allowed for comprehensive reporting of all relevant alterations detected by the F1CDx platform [84]. Comprehensive validation of the analysis pipeline, which included robust regression testing and reanalysis of FoundationFocus™ CDx BRCA clinical bridging sample data, was performed. The assays were determined to be concordant for determining HRD status for patients who benefited from rucaparib treatment (Table S29 in S1 Appendix).
ARIEL2 (NCT01891344) was a two-part, phase II, open-label study of oral rucaparib and Study 10 (NCT01482715) was a three-part, open-label, phase I/II study of oral rucaparib. The clinical validation of FoundationFocus™ CDx BRCA was based on clinical bridging retrospective analysis of 149 patients. ARIEL3 (NCT01968213) was a double-blind, placebo-controlled, phase 3 trial of rucaparib that enrolled 564 women with platinum-sensitive, high-grade ovarian, fallopian tube, or primary peritoneal cancer [85]. The clinical validation of the expanded version of test, FoundationFocus™ CDx BRCA LOH , which introduced the assessment of gLOH, was based on clinical bridging retrospective analysis of 518 of the 564 patients. Using the expanded test FoundationFocus™ CDx BRCA LOH, the median PFS of patients treated with rucaparib, who were determined to be BRCA1/2-positive, was 16.6 months as compared with 5.4 months for patients treated with placebo (HR: 0.23; 95% CI: 0.16, 0.34) (Table 5) [85,86].
SOLO1 was a phase III, randomized, double-blind, placebo-controlled, multicenter trial that compared the efficacy of olaparib with placebo in patients with advanced ovarian, fallopian tube, or primary peritoneal cancer harboring a BRCA mutation (documented mutation in BRCA1 or BRCA2) following first-line platinum-based chemotherapy [83]. A total of 391 patients were randomized (2:1) to receive olaparib tablets 300 mg orally twice daily (n = 260) or placebo (n = 131). The clinical validation of F1CDx was based on a clinical bridging retrospective analysis of 368 patients included in SOLO1 whose tumor samples were available for analysis, of which 313 patients were determined to have a documented mutation in BRCA1/2 by F1CDx. The median PFS in both the F1CDx-positive patients and the full analysis set from the clinical trial were not reached at the time of analysis for patients treated with olaparib, as compared with 11.9 months and 13.8 months for placebo-treated patients, respectively [87]. The percent of progression events was similar between the full analysis set and the F1CDx determined set (39% in each group). When comparing olaparib to placebo for progression or death in this patient population, the HR was similar for those with BRCA mutation status as determined in the full analysis set (HR: 0.3; 95% CI; 0.23, 0.41) and those determined by F1CDx (HR: 0.28; 95% CI: 0.20, 0.38) ( Table 5) [87].

Clinical validation: Prostate cancer
Clinical validation for detection of homologous recombination repair (HRR) mutations in prostate cancer. The clinical validity of F1CDx for determination of HRR status in metastatic castration-resistant prostate cancer (mCRPC) was determined with a prospective analysis of clinical data from the PROfound trial. PROfound was a phase 3, randomized, open-label, multicenter trial that assessed the efficacy and safety of olaparib monotherapy in patients with mCRPC who have qualifying HRR gene mutations that were predicted to be deleterious or suspected deleterious (known or predicted to be detrimental/lead to loss of function) and who have failed prior treatment with a novel hormonal agent (NHA) (investigators choice of NHA with either enzalutamide 160 mg orally once daily or abiraterone acetate 1000 mg orally once daily plus prednisone 5 mg orally twice daily [prednisolone was permitted for use instead of prednisone, if necessary]) [88]. Eligible patients were those with HRR gene mutation-positive (HRRm) mCRPC who had progressed following prior treatment with an NHA [88]. All patients must have had a qualifying HRR mutation assessed via the Foundation Medicine HRR CTA to be randomized [88]. Qualifying HRR gene mutations were in BRCA1, BRCA2, or ATM for Cohort A, and BARD1, BRIP1, CDK12, CHEK1, CHEK2, FANCL, PALB2, PPP2R2A, RAD51B, RAD51C, RAD51D, or RAD54L for Cohort B [88].
For the clinical utility results within Cohort A, the ORR for the full analysis set was similar to that of the F1CDx-determined HRRm patients (33.3% vs 33.8%) for olaparib-treated patients. Additional clinical efficacy results from Cohort A are reported in Table S30 in S1 Appendix. Further, the median radiological PFS (rPFS) was similar within this cohort for the full analysis set (7.4 months; 95% CI: 6.24, 9.33) and those HRRm patients as determined by F1CDx (7.4 months; 95% CI: 6.87, 9.33) for patients treated with olaparib, as compared to 3.6 months for those treated with NHA in both groups. The HR for the comparison of radiological progression or death for olaparib to placebo was also similar for the full analysis set (HR: 0.34; 95% CI: 0.25, 0.47) and for HRRm patients as determined by F1CDx (HR: 0.33; 95% CI: 0.24, 0.46). There was a statistically significant improvement in rPFS as assessed by the BICR for olaparib-treated patients compared with investigators choice of NHA-treated patients in Cohort A+B, with a 51% reduction in the risk of radiological disease progression or death and a prolongation of median progression-free interval of 2.3 months with olaparib vs investigator's choice of NHA (HR: 0.49; 95% CI 0.38, 0.63; P<0.0001) [87]. Median rPFS was 5.8 months for the full analysis set and 6.2 months for the confirmed F1CDx subgroup vs 3.5 months for investigator's choice of NHA (Table 6) [87].

Analytical performance validation
The performance characteristics of F1CDx were established using DNA from a wide range of FFPE tumor tissue types (Table S31 in S1 Appendix). Each study also included a broad range of representative alteration types for each class of alterations (SUBs, INDELs, CNAs, and rearrangements) in various genomic contexts across a broad selection of genes, as well as analysis of genomic signatures including MSI and TMB. The analytical validity of F1CDx was demonstrated across multiple analyses reporting the LoB, LoD, concordance with orthogonal method, and precision of the assay.
Limit of blank. The LoB describes the highest measurement result that is likely to be observed for a blank sample with a stated probability, α. Per industry standard, an α (type I error rate, false positive rate) of 0.05 was selected, corresponding to a percentage of false-positive <5%. The LoB of zero was confirmed using the mutation calls from mutation-negative FFPE samples (19 distinct samples with four replicates per sample). Seventy-five (75) samples (one sample failed to meet the Library Construction QC process specification) were used for the assessment of LoB. It was confirmed that each replicate of LoB sample was negative for the targeted variants defined in the LoD analysis, thus confirming the LoB of zero. A similar LoB study was conducted for each supplemental companion diagnostic claim; in each case, an LoB of zero was confirmed.
An additional study was conducted in a total of 218 evaluable test replicates for TMB-H status at a cut-off of �10 mut/Mb. The LoB was confirmed as zero. A supplementary statistical analysis was performed to assess the LoB for MSI-high (MSI-H). A total of 135 test replicates were evaluated confirming the LoB of zero for MSI-H.
Limit of detection. The LoD describes the precision of the assay at the lowest quantitative level at which an analyte (genomic variant) can be consistently detected. According to industry standard, "consistently detected" was defined as the level at which a 95% detection rate is observed. As defined by CLSI EP17A2E, the LoD is a "measured quantity value, obtained by a given measurement procedure, for which the probability of falsely claiming the absence of a measurand in a material is β, given a probability α of falsely claiming its presence" [91].
The LoD was determined for all F1CDx companion diagnostic claims. LoD was determined based on either allele frequency (as summarized in Table 7) or computational tumor purity (as summarized in Table 8). Allele frequency was the measurand for all SUBs and INDELs, wherein the frequency of an allele may be quantified. In contrast, the LoD for CNAs was based on the tumor purity of a given sample. DNA derived from FFPE tumor samples was evaluated for each companion diagnostic alteration or biomarker. For each LoD evaluation, five to six titrated levels of MAF or computational tumor purity, with 13 to 20 replicates per level, were prepared and assessed by F1CDx. Further, the LoD for each variant category assessed by F1CDx was evaluated to determine the sensitivity of tumor profiling. Several categories of alteration types were evaluated, as detailed in Table S32 in S1 Appendix, as well as Table S33 in S1 Appendix for SVs and INDELs, and Table S34 in S1 Appendix for CNAs and copy number rearrangements. As with the companion diagnostic claims, DNA derived from FFPE tumor samples was selected for each of the variant categories. For each sample, five to six levels of MAF or computational tumor purity, with 13 to 20 replicates per level, were evaluated for a total of 78 to 94 replicates per sample.
The tumor profiling LoD by variant type as determined by the hit rate method is reported in Table S32 in S1 Appendix LoD for SVs, including SUBs and INDELs, is based on MAF. LoD for structural variants (fusions, amplifications, homozygous deletions, rearrangements) and complex biomarkers is based on computational tumor purity (described in more detail in the S2 Appendix). For platform-wide LoD, the INDELs (other than homopolymer repeat context) are grouped together as they had similar LoD characteristics. The INDELs ranged from 42bp insertions to 276bp deletions. INDELs at homopolymer repeat context had higher LoD, with a dependency on the length of the repeat context. Tumor profiling LoDs for alterations based on allele frequency are further detailed in Fig 2A and Table S33 in S1 Appendix, and similarly, for alterations based on computational tumor purity in Fig 2B and Table S34 in S1 Appendix. Detailed evaluation of SUBs demonstrated that the driver status (e.g., known, likely, and unknown), as assigned by the F1CDx analysis pipeline, has an important impact upon the calculated LoD, particularly for SUBs. Given the specific reporting rules in the analysis pipeline, the LoD for known driver SUBs is lower than the LoDs for likely and unknown SUBs, as depicted in Fig 3. The LoD for TMB-H calling was evaluated using the hit rate method via two approaches: somatic TMB component variants by MAF; and tumor purity using cut-off at �10 mut/Mb. The resulting LoD (by MAF) for the TMB component variants within the interquartile range (IQR) ranged from 7.33% to 9.19% for SUBs, 7.08% to 11.74% for INDELs, and 7.32% to 9.24% overall. The LoD for TMB-H calling was determined empirically by the hit rate method (defined as the lowest level with 95% hit rate) as 28.16% computational tumor purity. The LoD results using �10 mut/Mb cut-off is reported in Table S35 in S1 Appendix. The LoD of MSI-H calling was evaluated using the hit rate method from five samples. Across the samples used for analysis, the highest observed tumor purity, with �95% of LoD ranges for (A) SVs and (B) amplifications and rearrangements a LoD calculations for the platform variants were based on the hit rate approach for variants with less than three levels with hit rate between 10% and 90% and probit approach for variants with at least three levels with hit rate between 10% and 90%. LoD from the hit rate approach is defined as the lowest level with 95% hit rate (worst-case scenario). b Data includes an alteration in the TERT promoter, 124C>T (LoD of 7.9%). TERT is the only promoter region interrogated and is highly enriched for repetitive context of poly-Gs, not present in coding regions. c Alterations classified as "known" are defined as those that are listed in COSMIC. d Alterations classified as "other" include truncating events in tumor suppressor genes (splice, frameshift, and nonsense) as well as variants that appear in hot-spot locations but do not have a specific COSMIC association or are considered VUS due to lack of reported evidence and conclusive change in function. e Sensitivity calculations for the platform variants were based on the hit rate approach for variants with less than three levels with hit rate between 10% and 90% and probit approach for variants with at least three levels with hit rate between 10% and 90%. LoD from the hit rate approach is defined as the lowest level with 95% hit rate (worst-case scenario). f Max represents VUS alteration at calling threshold. bp = base pair; CN = copy number; Mb = megabase; MSI = microsatellite instability; mut = mutation; TMB = tumor mutation burden; VUS = variants of unknown significance. replicates scored as MSI-H, was 15.67% tumor purity. The LoD for MSI-H is reported in Table S36 in S1 Appendix.
The LoD of gLOH calling in ovarian cancer samples was evaluated in four samples. The results validated a gLOH LoD (with 95% probability of detection) of 30% tumor purity, meeting the established acceptance criteria.
Precision: Reproducibility and repeatability. A total of 163 samples had alterations representative of companion diagnostic-associated alterations as well as exemplar alterations in a variety of genomic contexts. Each sample also included additional alterations that were included in the assessment. Across all samples, the pre-sequencing process failure rate was 1.5%, and the no-call rate was 0.18% for MSI-H, 6.38% for TMB-H (all samples), and 0.22% for TMB-H (samples with �10 mut/Mb). Within the assessment of repeatability and reproducibility for companion diagnostic variants, all variants from all samples demonstrated 100% agreement.
The reproducibility of calling hundreds of thousands of non-companion diagnostic platform alterations was evaluated across all sample replicates within the Precision study. The platform-level repeatability and reproducibility showed high overall agreement across alteration variants, and a high sample-level of reproducibility, as summarized in Tables 9 and 10   and negative call rates for the platform variants are reported in Table S37 in S1 Appendix. To supplement the previous study and further evaluate the precision for calling MSI-H and TMB-H in DNA derived from pan-tumor FFPE specimens, a supplementary analysis was  (Table 10). For the assessment of MSI, 100% repeatability and reproducibility was observed, with a lower limit of 99.7% and upper limit of 100%. Repeatability and reproducibility of MSI calling was robust. For MSI-H (using a cut-off at 0.0124), overall assessments of repeatability and reproducibility met acceptance criteria (i.e., �90% concordance). Repeatability and reproducibility across 46 samples were 96.02% (95% CI: 93.99%, 97.38%) and 97.01% (95% CI: 95.81%, 97.87%), respectively.
Analytical concordance. The detection of alterations by the F1CDx assay was compared to results of an externally validated NGS assay (evNGS) which was considered an orthogonal method for purposes of comparison. Concordance does not consider which method gives the "correct" result but only compares whether the methods used agree, and results can therefore be affected by technological and operational differences between the tests compared. The comparison between short alterations, including base SUBs and short INDELs, detected by F1CDx and the orthogonal method included 188 samples from 46 different tumors. The comparison of CNAs (including amplifications and homozygous deletions) and rearrangements detected between F1CDx and the orthogonal method included 168 samples. The samples analyzed for the orthogonal concordance for the companion diagnostic claims are further described in the Supplementary Appendix (S2 Appendix). A summary of PPA, NPA, and corresponding 95% two-sided exact CI is provided in Table 11.
The analysis for concordance of TMB-H (�10 mut/Mb) detection was performed using a laboratory validated WES assay. This analysis evaluated 218 samples, of which 89 were not pre-screened by F1CDx (Set A) and 129 were pre-screened by F1CDx (Set B). Concordance results between F1CDx and WES for TMB-H calling are summarized in Additionally, to support the use of retrospective data generated using FoundationOne 1 , a predecessor to F1CDx, a concordance study was conducted with F1CDx. This study evaluated a test set of 165 specimens. A total of 2,325 variants, including 2,026 SVs, 266 CNAs and 33 rearrangements, were included in the study. The study results are summarized in Table 13.

Discussion
In clinical application, CGP is designed to reduce the complexity of biomarker testing, enabling precision medicine to improve outcomes for patients with cancer. This improved outcome is predicated on test results being reliable, accurate, and validated to the highest standard available. The analyses presented here demonstrate the extensive analytical and clinical validation undertaken for F1CDx prior to initial FDA approval, and continuously over time to ensure reliability and relevance within an ever-changing clinical landscape.
Genomically-matched, biomarker-based precision therapies have been demonstrated to deliver significantly higher ORR across 13,203 patients in phase 1 and 32,149 patients in phase 2 trials, and in ORR, PFS, and overall survival (OS) in 38,104 patients in a meta-analysis of 112 registrational trials [92][93][94]. To assess CGP in informing the use of biomarker-based therapies, the clinical utility evidence base of Foundation Medicine CGP testing across tumor types has included assessment of actionability of test results and impact of test results on clinical decision-making (Table S38 in S1 Appendix) [4,41,[93][94][95][96][97][98][99][100][101]. The addition of companion diagnostic claims is expected to expand upon this actionability and decision impact, and should be included in assessment of clinical utility of the CGP approach. A growing body of evidence demonstrates clinical improvements in patient outcomes including PFS, OS, durable clinical response, and time to treatment failure (TTF) in patients whose treatments were informed by Foundation Medicine CGP, including F1CDx, in more recent studies (Table S39 in S1 Appendix) [4,6,94,97,[101][102][103][104][105].
Foundation Medicine has demonstrated direct clinical validity and clinical utility for F1CDx in a pan-tumor setting for TMB-H status and NTRK fusions, and for MET exon 14 skipping in NSCLC, PIK3CA mutations in breast cancers, BRCA1/2 alterations in ovarian  b PPA and NPA were adjusted using a prevalence of 9.6% to account for sampling differential. c Primary analysis: a sample was considered as positive if an NTRK1/2/3 rearrangement was detected; otherwise, it was considered as negative. Immune checkpoint inhibitors (ICPI) have transformed cancer care in select patient populations where these therapies are associated with clinical benefit. F1CDx can be supplemented by the incorporation of programmed death-ligand 1 (PD-L1) expression testing by immunohistochemistry in tumor types, which is relevant as a biomarker for ICPI treatment. Additional potential predictive biomarkers of ICPI resistance or hyperprogression that are evaluable by CGP include JAK/STAT pathway, KEAP, STK11, and MDM2/4 amplifications [44, 106]. MSI-H and TMB-H (�10 mut/Mb) have gained tumor-agnostic FDA approval as companion diagnostics for ICPI treatment. The importance of TMB-H was recently called into question as a biomarker for ICPI response [107]. However, as outlined in a Letter to the Editor, the failure to remove driver mutations during the in-silico cutoff analysis led to an overestimation of the TMB-H cohort, thereby diluting the clinical effect [56]. This highlights the importance of clinical and analytical validation of any assay intended to support therapy recommendations. TMB-H as a predictive biomarker for ICPI is supported by broad evidence, including a prospective cohort from the phase 2 KEYNOTE-158 study [57,73,108,109]. Foundation Medicine's broad-panel, CGP approach meets the clinical needs of providers and patients to receive guideline-based biomarker testing and improves the paradigm over sequential single-gene testing that often leads to tissue exhaustion and extended times to identify and initiate treatment, thus helping to keep pace with a rapidly evolving field of medicine. Foundation Medicine's approach to CGP has been demonstrated to deliver improved clinical decision-making, and the single, high-throughput platform approach across diverse tumor types has demonstrated robust analytical and clinical validity and utility data [6, 58, 94, 97, 101-105, 110, 111]. Highly informative and personalized F1CDx results are provided with an assurance of quality in accordance with FDA and CLIA regulations and within clinically acceptable turnaround times. In accordance with the product requirements, a 10-to 14-day calendar turnaround is achieved in �90% of the samples processed, as measured by specimen receipt to issue date of the report (10 days). In practice, the median turnaround time in 2021 was 10.9 calendar days, and 90.1% of tests demonstrate turnaround �12 calendar days, from specimen receipt to issue date of the report (since June 19, 2021). This aligns with the CAP/ International Association for the Study of Lung Cancer/Association for Molecular Pathology recommended turnaround time of 10 working days or two calendar weeks [112].
While F1CDx reports identified deleterious variants, the test report intentionally does not provide germline mutation calling, which requires genetic counseling in many countries and may risk delaying therapeutic decision-making [113]. However, testing DNA derived from a tumor may capture both germline and somatic variants predictive of outcomes to targeted therapies [114]. The F1CDx clinical report highlights potential germline alterations relevant to cancer that may warrant additional germline testing in the appropriate clinical context. Foundation Medicine's clinical reporting services also highlight genomic alterations that may have prognostic or diagnostic significance, such as NUTM1 fusions in NUT carcinomas, TMPRSS2-ERG in prostate cancers, and BRAF-KIAA1549 in gliomas, as well as context for low/intermediate MSI results in potential cases of Lynch Syndrome with high TMB [115][116][117][118]. Each specimen is reviewed by a licensed pathologist for tumor purity, suitability, and consistency with the stated diagnosis when provided, another crucial step in ensuring quality. Other Foundation Medicine decision support includes provision of molecular tumor board (MTB) services [63, 119,120], allowing clinicians to tap the wealth of knowledge that a team of expert genomicists, pathologists, and oncologists provides. In a study assessing the impact of the Foundation Medicine MTB program, of 54 cases presented, 81% had one or more potentially therapeutically relevant alterations, and genomically-matched therapy or clinical trial options were offered to 46% of patients based on the MTB discussion [63]. A well-curated gene list has also allowed for review of recent results to inform physicians and patients of newly approved therapies for which their patients may be eligible. For example, when sotorasib was approved for NSCLC patients with KRAS G12C mutations, physicians with patients whose Foundation Medicine reports included this mutation in the previous three months were notified of sotorasib's approved indications.
Therapeutic decision-making is an increasingly complex challenge for practicing oncologists. The validation data presented here are the backbone for the ultimate patient impact provided by CGP. In patients who do undergo non-CGP molecular testing methods, such as single-gene tests, hot-spot panels, or cancer-specific focused panels (which typically rely on PCR or FISH methodology), limitations include: incomplete information, as not all relevant genes and/or types of alterations nor genomic signatures are assessed [121][122][123][124][125][126][127][128]; inefficiency, as these methods may require sequential testing in certain cancer types [64,124,125,129,130]; and risk of re-biopsy, as multiple tests exhaust precious tissue [9,64,124,125]. For example, CGP identifies genomic alterations missed by other testing methods in 41% to 84% of previously tested patients [15,41].
As the number of actionable targets and associated targeted therapies expands, the capacity of most treating physicians to keep up with the precision oncology literature and guidelines will be exceeded. Therefore, to help support these crucial clinical choices, a streamlined testing process that allows a single test to identify all actionable alterations to be delivered together with a customized and timely literature review has already proven valuable to identify patients likely to respond to targeted therapy, and those patients bearing resistance markers who are not likely to respond [131]. This approach has already proven simpler and more sensitive than traditional sequential testing in screening CRCs for Lynch syndrome [132]. A test by itself does not directly impact patient response and survival outcomes; however, when properly validated, test results can inform appropriate treatment. Decision insights are significantly simplified in cancers with well-characterized mutations where companion diagnostics exist: a positive test result in the absence of contraindications indicates targeted therapy. This robust association of clinical benefit when therapeutic choice is predicated on a test has been achieved through hard-won clinical utility evidence from randomized controlled trials (RCTs) [49,58,83].
F1CDx has been assessed for FDA/CMS parallel review using the ACCE (Analytical validity, Clinical validity, Clinical utility, and Ethical, legal and social implications of genetic testing) model. FDA approval of F1CDx signified an acceptable evidence base for analytical and clinical validity requirements for coverage [133,134]. CMS then carefully evaluated clinical utility by examining clinical outcome measures including OS, PFS, PR, CR, ORR, stable disease, time to progression, and TTF in systematic evidence reviews and meta-analyses of clinical trial data, including data from RCTs, and from nonrandomized studies [134,135]. The CMS review resulted in Medicare coverage criteria for FDA-approved NGS-based in vitro companion diagnostic assays, like F1CDx, through a National Coverage Determination (NCD) [136]. F1CDx is covered under the NCD when the patient has: a. Either recurrent, relapsed, refractory, metastatic, or advanced stages III or IV cancer; and b. Either has not been previously tested using F1CDx for the same primary diagnosis of cancer or repeat testing using F1CDx only when a new primary cancer diagnosis is made by the treating physician; and c. Decided to seek further cancer treatment.
Patients may not undergo molecular testing for many reasons, including lack of physician knowledge about the benefits of results to help inform treatment decision-making, lack of access to or reimbursement for testing, or other factors specific to the clinical scenario for a given patient, which make molecular testing utilizing a tissue-based test impractical. In a retrospective analysis of genomic testing patterns in NSCLC patients seen at 15 US sites between 2013 to 2015, only 8% (63/814) of eligible patients received broad-panel genomic profiling [9]. In 2017, a nationally representative survey of US oncologists showed that fewer than 17% reported ordering four or more commercially available NGS tests to guide treatment decisions during the previous 12 months, even though most were affiliated with academic institutions. Despite over half of respondents having reported some training in genomic testing, over half also reported difficulty in interpreting the results "some" or "most" of the time [137]. Most US patients with cancer are treated in the community setting, and the majority of oncologists use some form of an NGS-based tumor testing to guide treatment decisions [137,138]. In a nationwide survey of US physicians, 40% confirmed they obtain pharmacogenetic testing information from package inserts [139]. However, in an internal market research survey conducted by Foundation Medicine, US oncologists indicated that while 75%−80% of their patients with advanced cancer had received some sort of molecular diagnostic testing, only 25%−30% had received broad-panel molecular profiling. These broad-panel molecular profiling tests are used most often in indications like NSCLC or prostate cancer, where the presence of multiple biomarkers with approved targeted therapies creates additional need for broadpanel molecular profiling [140].
The inclusion of companion diagnostic and biomarker information in the therapeutic drug labeling is an important source of information for treating physicians. Likewise, the continued increase in the number of FDA-approved targeted therapies and use of CGP in earlier lines of therapy will create additional inroads into CGP diagnostic clinical application. Among >191,000 F1CDx reports delivered in the US from 2018 to 2020, the largest increases in reporting frequency were observed for: cemiplimab, approved in 2018 for advanced cutaneous squamous cell carcinoma [129] and in 2021 for advanced basal cell and NSCLC [130]; alpelisib, approved in 2019 for HR-positive, HER2-negative, PIK3CA-mutated advanced breast cancer [131]; selumetinib, approved in 2019 for pediatric neurofibromatosis type 1 plexiform neurofibromas [132]; talazoparib, approved in 2018 for germline BRCA-mutated HER2-negative advanced breast cancer [133]; fam-trastuzumab deruxtecan, approved in 2019 for third-line HER2-positive advanced breast cancer [134] and in 2021 for HER2-positive previously treated advanced gastric cancer [135]. Other recent approvals include capmatinib (approved in May 2020) [141] and tepotinib (approved in February 2021) [142] for metastatic NSCLC patients whose tumors exhibit splice site alterations in MET exon 14, a family of somatic alterations whose diversity and prevalence across diverse tumor types was first published by Foundation Medicine in 2015 [143].
Companion diagnostic biomarkers include alteration types in which the presence or the absence (e.g., wild-type) of the biomarker can inform treatment decisions. In addition, if other identified genomic alterations or genomic signatures may be associated with treatment resistance, the F1CDx report will include a note notifying of potential resistance, providing actionability to help rule out potentially ineffective treatment. Even negative results (i.e., the absence of actionable alterations) from CGP can provide benefit to patients and physicians, so long as testing methods are sensitive across a broad range of actionable alterations. For those without potentially matching therapy or clinical trial options, CGP may confirm chemotherapy as the best option and/or inform discussions about palliative care [144].
Rare or novel rearrangements and gene fusions are detected by F1CDx in many common tumor types. Fig 4 illustrates the frequency of known/likely rearrangements involving targetable genes in common tumors from >191,000 US F1CDx results since 2018. These results show that using only DNA sequencing, F1CDx reports targetable rearrangements in common tumor types at rates consistent with those reported in the literature. Supplemental RNA sequencing is recommended for genomic profiling of hematologic, pediatric and rare malignancies and sarcomas, tumors in which fusions and gene rearrangements are more likely [145].
Clinical utility is challenging to assess in a rapidly evolving and complex field, and data from earlier studies (2014−2019) likely underestimate the potential therapeutic implications of F1CDx results today (Tables S38 and S39 in S1 Appendix). As evident in Fig 5, the number of targeted oncology therapy indications and approvals in oncology is growing rapidly [37], making the critical question in oncology care one of: "Which therapy is appropriate for the patient?". Foundation Medicine CGP provides results that allow the patient and the physician to make informed treatment decisions based on evidence-based interventions that improve health outcomes and help keep pace with a rapidly evolving field. F1CDx reports delivered potential therapeutic implications in >191,000 reports from 2018 to 2020, overall and by disease group rank-ordered by therapy options in-tumor type, other tumor type, matching clinical trials, and companion/complementary diagnostics (Fig 6). In over 36,000 reports for NSCLC,~50% included a companion diagnostic-associated therapy,~70% included an intumor type therapy,~80% included a therapy option in other tumor types, and nearly all included at least one relevant clinical trial. In over 20,000 breast cancer reports,~30% included a companion diagnostic-associated therapy,~76% included an in-tumor type therapy,~76% included a therapy option in other tumor types, and~90% include at least one relevant clinical trial. Refer to Table S40 in S1 Appendix for detailed data by tumor type.
Health economic outcomes data can provide additional evidence to demonstrate the value of CGP testing. Preliminary economic analyses have shown that the important clinical benefits of CGP may be accompanied by only a modest increase in cost, driven by longer duration of effective therapy for patients and meaningful prolongation of life which is the entire goal of cancer treatment [146][147][148]. As National Comprehensive Cancer Network (NCCN) states throughout its guidelines, "the best management of any patient with cancer is in a clinical trial" [28-34, 50-53]. Broad-panel genomic profiling has been associated with a 10% to 20% enrollment rate in clinical trials to date compared with a historical enrollment rate of �8% [41, 96,98,100,102,149], which may save payers up to $25,000 per patient through diversion of costs to study sponsors [18,41,96,98,100,102].
The FDA has established a transparent process to validate tests in the high-risk setting of therapy selection, where poor performance translates to inferior patient outcomes. Ongoing studies will demonstrate even greater impact on decision-making and patient outcomes as a result of the FDA approval of F1CDx as a regulated companion diagnostic for an increasing number of targeted therapies. Regrettably, only a small minority of providers have chosen to invest in transparently demonstrating the rigor of their services, and aside from CMS, many payers are still able to choose cost over quality. As with pharmaceuticals, patients and providers need a reliable source of transparent performance data to be confident of the quality of the healthcare services they are receiving. Enabling precision medicine through guideline-driven adoption of only high-quality, validated tests will emerge as a more cost-effective and holistically beneficial treatment paradigm compared with traditional empirical treatment.