Summary
Precision oncology is predicated on optimal molecular profiling that is “fit for purpose” to identify therapeutic vulnerabilities. Liquid biopsies may compensate for inadequate genotyping, but remain less sensitive and specific compared with tissue biopsies. The liquid biopsy toolbox is poised to expand through novel assays and insights from longitudinal profiling.
In this issue of Clinical Cancer Research, Sugimoto and colleagues present one of the largest concordance studies under the auspices of LC-SCRUM, a National lung cancer profiling study in Japan, comparing tissue-based next-generation sequencing (NGS) analysis against circulating free DNA (cfDNA) profiling using Guardant360 (1). In paired analysis of 1,062 non–small cell lung cancer (NSCLC) tissue-plasma samples, targetable gene alterations were identified in 44.5% and 51.6% of plasma and tissue samples, respectively. When examining the eight oncogenic drivers including EGFR, KRAS, BRAF, HER2, MET exon14 skipping mutations and fusions of ALK, ROS1, RET, positive percent agreement (sensitivity) was 72% between tissue and liquid NGS, and negative percent agreement (specificity) 87% (Table 1). The positive percent agreement was expectedly higher for mutations (77%) than fusions (47%), with very low concordance observed for ROS1 fusions (18%), likely due to the heterogeneity of ROS1 fusion partners. It is important to evaluate and consider these observations in the context that tissue-based NGS was performed in a central laboratory, where DNA and RNA were extracted from largely fresh-frozen tissue (90%) and subjected to Oncomine Comprehensive Assay v1 and v3, in a setting with optimal preanalytical conditions for successful tissue-based molecular profiling.
Publication . | Samples analyzed . | Population . | cfDNA-based assay . | Prevalence of actionable alterations (cfDNA) . | Tissue-based assay . | Prevalence of actionable alterations (Tissue) . | Sensitivity . | Specificity . |
---|---|---|---|---|---|---|---|---|
Sugimoto et al. Clin Cancer Res 2022 (1) | 1,062 | Japan | Guardant360 | 44.5% | Oncomine Comprehensive Assay v1/v3 | 51.6% | 72% | 87% |
Cui et al. Eur J Cancer 2022 (3) | 243 | UK (Caucasian 90%, Asian 5%, Others 4%) | Guardant360 | 38% | NHS SoC tests (NGS, EGFR Cobas, ALK IHC, ROS1 IHC) | 27% | 75% | 75% |
Schwartzberg et al. JTO Clin Res Rep 2022 (4) | 131 | US (White 73.3%, African 5.34%, Others 12.2%, Unknown 9.16%) | FoundationOne Liquid/Liquid CDx | 43% | FoundationOne CDx | 66% | 69% | 96% |
Palmero et al. JCO Precis Oncol 2021 (13) | 186 | Spain (Caucasian 97%, Others 3%) | Guardant360 | 80.7% | SoC tissue genotyping (NGS, FISH, IHC, PCR, RT-PCR) | 57.1% | 66.7% (EGFR) 40% (ALK) | 100% (EGFR) 99.2% (ALK) |
Park et al. Cancer 2021 (6) | 262 | Korea | Guardant360 | 46.6% | Oncomine Focus Assay | 62.6% | 67.7% | 88.8% |
Leighl et al. Clin Cancer Res 2019 (2) | 282 | US (White 81.9%, African 6.4%, Asian 6.0%, Others 3.2%, Unknown 2.5%) | Guardant360 | 21.3% | SoC tissue genotyping (NGS, PCR, FISH, IHC, Sanger) | 27.3% | 80% | NR |
Publication . | Samples analyzed . | Population . | cfDNA-based assay . | Prevalence of actionable alterations (cfDNA) . | Tissue-based assay . | Prevalence of actionable alterations (Tissue) . | Sensitivity . | Specificity . |
---|---|---|---|---|---|---|---|---|
Sugimoto et al. Clin Cancer Res 2022 (1) | 1,062 | Japan | Guardant360 | 44.5% | Oncomine Comprehensive Assay v1/v3 | 51.6% | 72% | 87% |
Cui et al. Eur J Cancer 2022 (3) | 243 | UK (Caucasian 90%, Asian 5%, Others 4%) | Guardant360 | 38% | NHS SoC tests (NGS, EGFR Cobas, ALK IHC, ROS1 IHC) | 27% | 75% | 75% |
Schwartzberg et al. JTO Clin Res Rep 2022 (4) | 131 | US (White 73.3%, African 5.34%, Others 12.2%, Unknown 9.16%) | FoundationOne Liquid/Liquid CDx | 43% | FoundationOne CDx | 66% | 69% | 96% |
Palmero et al. JCO Precis Oncol 2021 (13) | 186 | Spain (Caucasian 97%, Others 3%) | Guardant360 | 80.7% | SoC tissue genotyping (NGS, FISH, IHC, PCR, RT-PCR) | 57.1% | 66.7% (EGFR) 40% (ALK) | 100% (EGFR) 99.2% (ALK) |
Park et al. Cancer 2021 (6) | 262 | Korea | Guardant360 | 46.6% | Oncomine Focus Assay | 62.6% | 67.7% | 88.8% |
Leighl et al. Clin Cancer Res 2019 (2) | 282 | US (White 81.9%, African 6.4%, Asian 6.0%, Others 3.2%, Unknown 2.5%) | Guardant360 | 21.3% | SoC tissue genotyping (NGS, PCR, FISH, IHC, Sanger) | 27.3% | 80% | NR |
Abbreviation: NR, not reported.
Tissue-based NGS, while regarded as the preferred assay for molecular profiling, has its challenges. First, not all patients have lesions amenable to biopsy or may have medical contraindications for invasive procedures. Second, the longer turnaround time for tissue-based assays, particularly in the presence of organ-threatening disease, can lead to delay in commencing effective therapy. Several studies, including the current study by Sugimoto and colleagues, have demonstrated that compared with tissue biopsies, cfDNA was associated with significantly shorter turnaround time due to bypassing steps involved in scheduling interventional procedures as well as tissue fixation and processing (1–4).
Where tissue is obtained through a needle biopsy, specimen attrition and low tumor purities can compromise the accuracy of results. Indeed, as reported by Sugimoto and colleagues, 6% of patients had unavailable tissue or no mutations identified with the tissue NGS, yet the liquid biopsy was able to identify actionable alterations in these patients. Previous studies have highlighted the issue of inadequate tissue genotyping in the real-world setting. Leighl and colleagues demonstrated that with standard-of-care (SoC) tissue-based profiling, only 18.1% of patients with NSCLC completed genotyping for eight guideline-recommended biomarkers, of which only two-thirds were utilizing panel-based NGS (2). In contrast, 95% of patients had requisite biomarkers evaluated by cfDNA, with identification of a biomarker in 27.3% in plasma versus 21.3% with SoC testing, highlighting the value of liquid biopsies in compensating for inadequate genotyping. Moreover, prospective studies have confirmed the positive predictive value of blood-based ALK fusion detection using FoundationACT, where overall response rate was 87.4% [95% confidence interval (CI), 78.5–93.5] and 12-month progression-free survival rate of 78.4% (95% CI, 69.1–87.7; ref. 5). This was comparable with previous trials using tissue-based ALK detection with IHC or FISH.
Liquid biopsies are gaining traction in molecular profiling due to specimen accessibility. However, multiple studies have highlighted the limitations of liquid biopsies. When comparing the same Guardant 360 liquid biopsy assay against a smaller Oncomine tissue panel, detection of National Comprehensive Cancer Network–recommended biomarkers in Korean patients with metastatic NSCLC was 46.6% in cfDNA versus 62.6% in tissue, with a sensitivity of 67.7% and specificity of 88.8% (summarized in Table 1; ref. 6). Similarly, liquid biopsies in this study only detected 18% (3/16) of ROS1 fusions and just under half the ALK fusions (46%, 21/45) when compared with RNA-based tissue NGS analysis. In a study utilizing a tissue and liquid NGS panel (FoundationOne/FoundationACT) in 131 paired samples, adding liquid to comprehensive tissue NGS did not result in incremental yield, with up to 23% of guideline-recommended biomarkers missed by liquid testing alone (4). Low cfDNA burden and variant allele frequencies observed in most plasma samples further complicates interpretation of the results. In addition, when the plasma sample is analyzed without a matching normal sample, some mutations may turn out to be false positives attributable to clonal hematopoiesis in peripheral blood mononuclear cells (7). Taken together, these studies to date underscore the importance and continuing role for tissue-based testing.
With rapid developments in the liquid biopsy space, we can also anticipate increased dimensionality in information gleaned from plasma NGS. This can include longitudinal tracking of disease, dynamic readouts of response or resistance, as well as depicting molecular portraits of resistance. The latter was illustrated in resistance to KRAS G12C inhibitors recently, where cfDNA could reveal multiple parallel resistance mutations in downstream BRAF, KRAS, MEK pathways after adagrasib treatment (8). Finally, to make timely treatment recommendations in patients with lung cancer, it would be necessary to consider other actionable biomarkers, such as PDL1 expression, tumor mutational burden, and copy-number amplifications (e.g., MET).
Emerging computational and NGS-based technologies are also bringing new tools and capabilities to the liquid biopsy toolbox. Whole-genome sequencing at low sequencing depth, in combination with artificial intelligence–based error correction, may offer a highly sensitive approach for detection of minimal residual disease in plasma (9). Plasma-based methylation profiling is a versatile approach that could enable both sensitive detection and localization of cancer (10). In addition, the fragmentation patterns of cfDNA could provide a window into the epigenetic state of cancer cells (11), enabling joint cancer genotyping and epigenetic profiling across time (12). Some of the key considerations when selecting fit for purpose cfDNA assays are shown in Fig. 1.
In summary, Sugimoto and colleagues reinforce the complementary role of liquid biopsies in enabling precision oncology approaches, where access to some form of comprehensive liquid profiling would be preferred to suboptimal profiling. The challenge remains how to determine the minimal burden of proof required to demonstrate clinical utility as a dynamic predictive biomarker for changing therapies, for example, combinatorial approaches or treatment de-escalation. Subsequently, how best to implement these technologies will require close dialogue between key stakeholders, including academics, industry partners, patient advocacy groups, regulatory agencies and payers, to chart a path into routine clinical use.
Authors' Disclosures
D.S.W. Tan reports non-financial support from C2i Genomics and personal fees from Guardant during the conduct of the study, as well as grants and personal fees from AstraZeneca, Takeda, and Pfizer and personal fees from Novartis, Genentech, Bayer, DKSH, Merck, MSD, and BMS outside the submitted work. No disclosures were reported by the other authors.
Acknowledgments
A. Jacobsen Skanderup is supported by Ministry of Health-National Medical Research Council, Singapore OFIRG21nov-0083. D.S.W. Tan is supported by Ministry of Health-National Medical Research Council, Singapore NMRC/OFLCG/002/2018 and MOH-CSAINV-19nov-0005, and Duke-NUS Medical School, Singapore 08/FY2021/EX(SL)/102-A156(a).