Identification of Actionable Genomic Alterations Using Circulating Cell-Free DNA

Nora S. Sánchez, PhD; Michael P. Kahle, PhD; Ann Marie Bailey, PhD; Chetna Wathoo, MD; Kavitha Balaji, PhD; Mehmet Esat Demirhan, MD, Dong Yang, PhD; Milind Javle, MD; Ahmed Kaseb, MD; Cathy Eng, MD; Vivek Subbiah, MD; Filip Janku, MD, PhD; Victoria M. Raymond, MS; Richard B. Lanman, MD; Kenna R. Mills Shaw, PhD; Funda Meric-Bernstam, MD


JCO Precis Oncol. 2019;2019(3) 

In This Article


Patient Characteristics

We enrolled 575 patients into the study. Patient characteristics and disease types are listed in Table 1. The major diseases represented are those with 10 or more patients enrolled.

Overall Utility of cfDNA in Identifying Actionable Alterations in Patients With Advanced Cancers

Of the 575 patients enrolled, 76.2% (n = 438) had at least one alteration reported in any gene; of these 438 patients, 81.5% (n = 357) had at least one alteration reported within an actionable gene, and of these 357, 54.7% (n = 205) had at least one HPCA alteration (see Patients and Methods). Thus, 35.6% of the sample (205 of 575 patients) had at least one HPCA alteration (Figure 1).

Figure 1.

Flow diagram that depicts the overall utility of cell-free DNA (cfDNA) in identifying actionable alterations in patients with advanced cancers. A total of 575 patients was enrolled, of whom 76.2% (n = 438) had at least one alteration reported. Of these 438, 81.5% (n = 357) had at least one alteration reported within an actionable gene, of whom in turn, 54.7% (n = 205) had at least one high potential for clinical action alteration. Hence, of the original 575 patients tested, 35.6% (n = 205) had at least one alteration classified as having a high potential for clinical action.

Overall Landscape of Molecular Alterations

The panel used offered extensive point mutation coverage for 70 to 73 genes (amplifications, 18 genes; fusions, six genes; indels, 23 genes). Consistent with this, single nucleotide variants were the most commonly reported alteration type, followed by amplifications, indels, and fusions. The results reflected the genomic landscape reported across tumor types through solid tumor testing[19,20] and is shown in Figure 2A. Amplification levels detected ranged from positive, strongly positive, to very strongly positive, as previously defined.[9–14] Genes most frequently reported with very strong positive amplification levels were CCNE1, FGFR1, MET, ERBB2, and AR. The remaining genes usually reported positive or strongly positive amplifications levels (Figure 2B). Aneuploidy/polysomy versus focal amplification can be ascertained when multiple genes reside on the same chromosome. For example, BRAF, CDK6, EGFR, and MET all reside on chromosome 7, so when all are amplified, aneuploidy is assumed. Similarly, more than four genes are included on chromosome 17, so focal amplification of ERBB2 (HER2) can be determined.

Figure 2.

Landscape of molecular alterations identified. (A) Commonly reported alterations by gene and alteration type. (B) Amplification levels of genes with copy number alterations. (C) Most frequently altered genes with consideration of each variant's potential for clinical action. Shown are the most frequently altered genes in 205 patients with at least one alteration classified as high potential for clinical action. Indel, insertion/deletion alteration; SNV, single nucleotide variant.

Molecular Landscape in the Context of Potential for Clinical Action

Among the overall molecular landscape for the 357 patients with at least one alteration in an actionable gene, the most frequently altered genes (> 40 patients) were KRAS (22.7%; n = 81), PIK3CA (21%; n = 75), NF1 (16.5%; n = 59), EGFR (15.1%; n = 56), MET (15.1%; n = 54), BRAF (14.6%; n = 52), and ERBB2 (12.9%; n = 46). However, if the specific alteration's potential for clinical action is factored in, the number of patients with HPCA alterations decreased from 357 to 205. Likewise, the order of most actionable genes within the context of these 205 patients shifted, with PIK3CA (18%; n = 37) being the most represented followed by BRAF (13.7%; n = 28), KRAS (10.2%; n = 21), NF1 (9.8%; n = 20), CDKN2A (7.8%; n = 16), ERBB2 (7.3%; n = 15), CCNE1 (7.3%; n = 15), and FGFR1 (7.3%; n = 15). Again, these results highlight the importance of assessing each alteration's biologic FS individually and in the disease context (Figure 2C).

Proportion of Patients With Alterations Identified, by Disease Type

Given the diverse set of advanced cancers (37 types) of patients enrolled in this study, we investigated whether differences existed in the suitability of cfDNA testing. Among diseases with 10 or more patients enrolled, most tumor types had at least one alteration reported within an actionable gene, as follows: colorectal, 95%; breast, 91%; gall bladder, 81%; prostate, 80%; lung, 74%; cholangiocarcinoma, 68%; hepatocellular, 63%; pancreatic, 61%; head and neck, 52%; ovarian, 50%; neuroendocrine, 46%; sarcoma, 40%; and appendiceal, 27.6%. For all other disease types combined, 46.3% of patients had at least one alteration that occurred within an actionable gene (Figure 3A).

Figure 3.

Identification of clinically actionable alterations and clinical trials by disease type. (A) Distribution of clinically actionable alterations by disease type. (B) Identification of clinical trials across all disease types in the context of variant's potential for clinical action.

The actual variant actionability, or potential for clinical action, was used to stratify patients by disease type into three categories: HPCA, low potential for clinical action, and not recommended for clinical action. Among diseases with 10 or more patients enrolled, the proportion of patients who had at least one HPCA alteration was as follows: breast, 67.6%; colorectal, 60%; lung, 56.5%; gall bladder, 56.3%; ovarian, 50%; head and neck, 43.5%; cholangiocarcinoma, 43.3%; prostate, 40%; hepatocellular, 25.4%; pancreatic, 19.5%; neuroendocrine, 15.4%; sarcoma, 14.9%; and appendiceal, 10.6% (Figure 4A).

Figure 4.

(A) Clinical management outcomes of patients with at least one variant of high potential for clinical action identified. (B) Reasons why genomic alterations were not acted upon in patients with at least one variant of high clinical potential identified. a2. b3. c5. d6. e7. f11. g26. h27. Mets, metastases; PS, performance status.

Overall, these findings demonstrate that cfDNA testing is informative across all tumor types included in this study. Specifically, among tumor types with greater representation (≥ 10 patients enrolled), 50% or more patients had at least one HPCA alteration. Tumors with slightly lower proportions of patients having at least one HPCA alteration were neuroendocrine tumors (15.4%), sarcoma (14.9%), and appendiceal tumors (10.6%).

Comparison With Previous Solid Tumor Testing

This study was not designed to compare cfDNA testing with tissue testing, and thus, patients did not have acquisition and testing of tumor and plasma samples simultaneously. The median time between the dates plasma and the archival tumor tissue that was tested were obtained was 20 months (range, 19 days to 13.6 years). When we compared cfDNA testing and previous tumor testing in 184 patients with previous solid tumor testing, regardless of the time interval but accounting for panel coverage among alterations in actionable genes detected in previous testing, there was a 36% concordance rate (56 of 157 alterations). If only including cfDNA panels that detected one or more alterations (143 patients), the concordance rate was 44% (56 of 127 alterations). The concordance rate for mutations was 42% (41 of 98 mutations). There were 135 alterations in actionable genes reported on cfDNA testing that were not covered by the previous, usually more limited, tumor testing. In addition, there were 182 alterations detected in actionable genes on cfDNA testing that were not detected in previous panels, although they were covered by the assay. These alterations potentially represent genomic evolution and mechanisms of acquired resistance and included detection of a PIK3CA mutation after treatment with a BRAF inhibitor and BRAF and MET mutations after treatment with AKT/mTOR inhibitors.

On the other hand, among 391 patients with no previous solid tumor testing attempted, 529 alterations in actionable genes were reported on cfDNA testing. In addition, of 14 patients who had previous solid tumor testing attempted but the tumor genomic assays failed (eg, because of an insufficient amount of DNA obtained from the sample), cfDNA testing reported 17 alterations in actionable genes, with 10 patients having any alteration detected, seven having at least one alteration in an actionable gene, and two having at least one specific alteration that was actionable at the time of annotation.

Identification of Clinical Trials for Variants Identified Through cfDNA

The utility of cfDNA testing depends on the ability to match patients to relevant targeted therapy delivered through clinical trials, standard of care, or off-label use. In addition to variant annotation, for patients with HPCA variants and/or select low potential for clinical action variants (ie, FS, unknown; AVC, potentially; actionable for call, therapeutic intervention), the decision support team identified genotype-matched trials for the patient's tumor type, which accounted for additional biomarker inclusion/exclusion criteria, where applicable. The portfolio of clinical trials available for matching consisted of those available across MD Anderson at the time of a patient's annotation. On the basis of these criteria, trials were identified in 286 (80%) of 357 patients (Figure 3B). Among patients who had at least one HPCA alteration, identification of trials was even higher, with potential trials identified for 188 (92%) of 205 patients (Figure 3B).

Clinical Course of Action After cfDNA Testing

Of the 205 patients with at least one HPCA variant, 118 had reached the 6-month follow-up period at the time of data freeze (March 6, 2018). At this time, 11% of patients (13 of 118) had been treated with genotype-matched targeted therapy on the basis of cfDNA testing, where a match included enrollment into a clinical trial (n = 9), off-label drug use (n = 3), or standard of care (n = 1). Of note, 7.6% of patients (nine of 118) who were not matched had already acted on this alteration on the basis of prior molecular testing (Figure 4A). Thus, we sought to understand why the genomic alterations in the rest of the patients (n = 96) were not acted upon. We found that most unmatched patients received standard of care (38.5%) or did not begin a new treatment regimen (36.5%), whereas a smaller subset of patients continued on a treatment regimen established either on or before the ordering of cfDNA testing (9.4%), enrolled in another clinical trial (genomic and non-genomically driven; 12.5%), had surgery (2.1%), or had additional tumor testing ordered (1%; one of 96; Figure 4A). Moreover, a major contributor to not acting on genomic alterations was the overall PS, with 28% of patients (27 of 96) reporting brain metastasis and/or other comorbidities at the time of trial consideration (Figure 4B).