Expert opinion on NSCLC small specimen biomarker testing — Part 2: Analysis, reporting, and quality assessment

Biomarker testing is now essential for guiding treatment decisions in advanced non-small cell lung cancer (NSCLC), with European Society for Medical Oncology (ESMO) guidelines suggesting that “all patients with advanced, possible, probable or definite, adenocarcinoma should be tested for oncogenic drivers” [1]. Additionally, molecular testing is recommended in cohorts of patients with non-adenocarcinoma histology (e.g. squamous cell carcinoma) who are < 50 years of age [2] and those who are never-smokers, long-time ex-smokers, or light-smokers (< 15 pack-years) [1]. This strategy is driven by the relative probability of finding an actionable alteration. As noted in Part 1 [3], there is a growing body of evidence indicating that patients with actionable oncogenic driver mutations who receive targeted therapy have improved clinical outcomes versus those without actionable driver mutations who receive chemotherapy [4‐6].

The clinical armamentarium for advanced NSCLC currently comprises seven European Medicines Agency (EMA)–approved targeted agents with associated biomarkers (excluding programmed death ligand 1 [PD-L1]; see Table 1) [7, 8]. These biomarkers for actionable genetic targets now include sensitising mutations in exons 18, 19, and 21 of the epidermal growth factor receptor (EGFR), Kirsten rat sarcoma viral oncogene homolog (KRAS) p.G12C point mutation, B-Raf proto-oncogene V600E point mutation (BRAF p.V600E), and rearrangements involving anaplastic lymphoma kinase (ALK), ROS proto-oncogene 1 (ROS1), neurotrophic tyrosine receptor kinase (NTRK1, 2, and 3), and rearranged during transfection (RET) [7, 8]. As noted in ESMO guidelines, testing for EGFR mutations and rearrangements involving ALK and ROS1 is now considered mandatory in most European countries [1]. As first-line B-Raf/mitogen-activated protein kinase (MEK) inhibitors become more widely approved, BRAF p.V600E mutation testing is also mandated in many oncology services [1]. KRAS p.G12C is now an actionable genetic target in Europe following approval of sotorasib by the European Commission in January 2022 [9]. NTRK is a target with approved treatments in many European countries, while the Erb-B2 receptor tyrosine kinase 2/human epidermal growth factor receptor 2 (ERBB2/HER2) and hepatocyte growth factor receptor (MET) exon 14 skipping mutations are evolving targets/biomarkers [1]. An ESMO Precision Medicine Working Group developed the ESMO Scale for Clinical Actionability of molecular Targets (ESCAT) to help clinicians prioritise actionability of the various genetic targets [10]. An ESCAT level I alteration means that a drug has been validated in clinical trials and, therefore, the alteration should drive treatment decisions in daily clinical practice. ESMO recommends that all level I alterations are profiled in patients with lung adenocarcinoma using next-generation sequencing (NGS).

In addition to the biomarkers for actionable genetic targets, PD-L1 expression by immunohistochemistry (IHC) is mandatory to inform treatment selection with immune checkpoint inhibitors (Table 1) [1]. ESMO guidelines specify a mandatory threshold of a tumour proportion score of ≥ 50% in first-line treatment [1]; the tumour proportion score is defined as the number of PD-L1 + tumour cells divided by the total number of viable tumour cells, multiplied by 100% [11]. However, definitions and thresholds for biomarker analyses are not standardised; in the case of PD-L1, at least five assays are available that have specific scoring systems and tumour site indications [12]. Studies have indicated a high level of concordance in the results of some of these assays for NSCLC [13]. Nevertheless, assay standardisation for emerging biomarkers is a challenge that will require the coordinated efforts of all stakeholders to ensure the future success of biomarker-guided targeted therapy [14].

The number of EMA-approved biomarkers for actionable targets is set to increase over coming years, owing to a rich pipeline of targeted therapeutic agents. These include MET exon 14 skipping mutations and gene amplifications, ERBB2/HER2 mutations and amplifications, neuregulin-1 (NRG1) rearrangements, fibroblast growth factor receptor 1 (FGFR1), and EGFR exon 20 insertions [8]. Agents targeting these genes are under investigation and some of these have already received approval in certain countries (see Table 1).

The rapid pace of innovation in targeted drug development, which is epitomised by the current and future state of precision oncology in NSCLC, makes it challenging for clinical guidelines and associated practice to keep pace. Figure 1 demonstrates the increasing gap between current recommendations and approved therapies across the main international guidelines for biomarker testing.

The gap between real-world practice and technical innovation is further increased by variation in national guidelines and reimbursement decisions, which often differ considerably from the initial EMA approvals in terms of timelines and outcomes. The significant variability between European countries in terms of real-world biomarker testing practice was illustrated in a recent review by Kerr and colleagues [8] (Fig. 2), highlighting that implementation of biomarker testing for patients with NSCLC continues to be suboptimal across certain regions and countries.

The availability of seven EMA-approved, biomarker-directed NSCLC therapies (excluding checkpoint inhibitors) and emerging targeted therapies suggests an imperative for multiplexed, massively paralleled sequencing technology (i.e. NGS) over multiple single-gene tests as the standard of care for patients with advanced NSCLC. NGS enables simultaneous testing of multiple oncogenic drivers [1, 2, 10] and provides a method to cope with increasing numbers of actionable targets, and limited volumes of available tissue (as discussed in Part 1). NGS may allow the analysis of clinically relevant co-mutations such as serine/threonine kinase 11 (STK11), kelch-like ECH associated protein 1 (KEAP1) and tumour protein p53 (TP53), and DNA damage response pathway alterations involving breast cancer type 1/2 (BRCA1/2). Other emerging predictors of neoantigen burden and immunotherapy response, such as tumour mutational burden (TMB), comprehensive genomic profiling (CGP), and DNA methylation, may also be analysed by NGS. As the number of evaluable biomarkers continues to increase, running multiple standalone assays in parallel or sequentially becomes increasingly inefficient in terms of time and cost, eventually tipping the balance in favour of NGS. Accordingly, recent ESMO guidelines state that the use of NGS for molecular testing is preferable for certain tumour types (e.g. level I alterations in lung adenocarcinoma, ovarian cancer, prostate cancer, and cholangiocarcinoma) [10], and NGS is rapidly being adopted as the standard approach to identify lung adenocarcinomas with oncogenic targets [1]. However, despite current guideline recommendations, there remain significant discrepancies in access to/use of NGS across Europe [15], where reimbursement constraints are a key limitation for adoption of best practice in biomarker testing.

In the previous article in this series, we explored the challenges and evidence-based recommendations related to obtaining sufficient quality tissue to undergo biomarker testing. In this review (Part 2), we summarise evidence-based recommendations relating to the analysis, reporting, and quality assessment of biomarker testing in small specimens from patients with advanced NSCLC. Where no guidelines or literature explicitly describe best practice, we report our recommendations for best practice according to the experience of the author group.

Biomarker testing methodologies fall into two categories: single-gene or multiplex assays (i.e. NGS or multiplex polymerase chain reaction [PCR]) of DNA and/or RNA [1, 2]. Single-gene testing approaches include DNA sequencing by real-time quantitative PCR (qPCR), pyrosequencing or sanger sequencing, RNA sequencing by reverse transcriptase (RT)-PCR, detection of cell protein expression by IHC, and detection of gene fusions/amplifications by (fluorescence) in situ hybridisation ([F]ISH). The appropriate diagnostic modality depends on the molecular target of interest, as illustrated in Table 2. To cover testing of all the biomarkers in Table 2, broad-panel NGS sequencing is more cost-effective than multiple standalone biomarker tests using combinations of IHC, FISH, and PCR (acknowledging that IHC is currently the only reliable method for PD-L1 assessment and is the method of first choice for ALK, with equivocal results confirmed by FISH [1, 2]). In agreement with this expectation, studies have shown that NGS is more cost-effective than single-gene testing when multiple targets need to be tested [16‐18], and increasing use of NGS versus single-gene testing correlates with an increase in life-years gained for patients with advanced NSCLC [18]. Overall, NGS represents an efficient alternative to single-gene testing [19].

While current DNA-based NGS can, theoretically, be used to detect sensitising mutations (point mutations, deletions, and insertions), copy number variations, and structural rearrangements (gene fusions), reliance on DNA-based fusion detection carries a risk of false negatives related to missing relevant fusions when large intronic regions stand between the fusion partners [20, 21]. The sensitivity of DNA-based NGS assays may be limited by the size of the intronic regions for genes such as NTRK, as the breakpoints usually occur within large intronic regions [20]. However, differences exist between the various systems used for library preparation in terms of the false negative error rate. In contrast to DNA, RNA sequencing is not affected by intronic regions that are spliced out during transcription. Therefore, the authors recommend using RNA-based NGS in parallel to DNA-based NGS to help improve sensitivity for the detection of gene fusions. The choice of the technology is also important, as hybrid capture assay and anchored multiplex technology allow broader fusion analysis but require a larger amount of material than amplicon-based methods [22]. Furthermore, RNA-based NGS allows identification of gene transcripts, permitting conclusions regarding in-frame gene fusions that are fully functioning, as well as the identification of gene fusion partners. In terms of the potential impact on available tissue, one-step co-extraction of RNA and DNA and simultaneous NGS of both DNA and RNA can help reduce tissue consumption [23, 24].

It should be acknowledged that, particularly in the context of fusion gene testing, IHC may be complementary to, and/or an alternative to, sequencing or FISH testing; however, in the authors’ experience, the expense and tissue consumption of these approaches should also be considered. Sometimes, elevations in protein levels are observed in tumour cells when driven by an oncogenic fusion gene. Detection of gene-product overexpression by IHC is a useful screening tool for assessing ALK, ROS1, and NTRK fusions in NSCLC. This approach is recommended in ESMO guidelines [1], and the US Food and Drug Administration has approved the Roche VENTANA ALK (D5F3) CDx IHC assay as a primary therapy-determining test for ALK kinase inhibitors [25]. For ROS1 and NTRK IHC + cases, confirmation by another molecular method (e.g. FISH, qPCR, NGS) is mandatory [1]. For RET fusions, IHC is not recommended as a screening tool, as false positive and negative cases have been reported [26]. Taken together, combined DNA/RNA NGS, using appropriately validated assays and processed by suitably qualified operators, is a reliable and efficient approach for comprehensive detection of all approved and emerging biomarkers in advanced NSCLC (excluding PD-L1 detection by IHC). There may be other roles for IHC in the context of fusion gene testing. IHC may be possible in samples with few tumour cells or with high non-neoplastic cell contamination and where NGS fails or is not feasible. Strong IHC staining may be directly clinically actionable (e.g. for ALK fusions) or strongly indicative of the presence of a fusion gene (e.g. ROS1 and NTRK), in the appropriate histological context. There is also evidence that presence of the protein (positive IHC) may be indicative of greater probability of clinical response to therapy [27, 28], suggesting that IHC may be complementary to molecular methods for fusion gene identification/detection.

Sequencing of plasma-circulating cell-free DNA (cfDNA) via liquid biopsy is a complementary approach to tissue-based biomarker testing, particularly when tissue samples are insufficient or unsuitable/inadequate for biomarker testing, or if re-biopsy cannot be performed safely [29]. In the authors’ experience, cfDNA sequencing analysis can be conducted using as little as 6 mL of peripheral whole blood stored at room temperature in ethylenediaminetetraacetic acid (EDTA) tubes. Ideally, blood collected in EDTA tubes requires centrifugation within 3 h (to reduce degradation of cfDNA and the risk of a false negative result), yielding 3 mL of plasma, which subsequently undergoes cfDNA extraction using commercially available kits. A variety of sequencing methods may then be applied to the extracted DNA including qPCR, droplet digital PCR, and NGS [30]. Analytical techniques must be highly sensitive to detect tumour-specific cfDNA, which represents only a small fraction of total circulating cfDNA.

While plasma is most commonly used for liquid biopsy, all biological fluids can potentially represent a source of tumour DNA for testing; however, limited data exist on the use of these alternative sources in the genomic characterisation of NSCLC for guiding therapy. Nevertheless, evidence suggests that cerebrospinal fluid testing may be more sensitive than that of plasma for detection of genomic alterations in patients with NSCLC and leptomeningeal metastases [31, 32]. It has also been suggested that the combination of plasma and urine testing can increase the sensitivity of EGFR mutation testing in NSCLC [33].

Liquid biopsies may also overcome tumour heterogeneity sampling bias associated with tissue biopsy and/or permit longitudinal studies of tumour evolution and response to therapy [8, 30]. A further advantage of liquid biopsy is the avoidance of invasive procedures for tissue acquisition [30]. However, there are concerns that overreliance on liquid biopsy could lead to poorer tissue pathology services in some laboratories, and the technique is not without limitations. For example, there is a lack of consensus on optimal pre-analytical procedures or consistently validated thresholds, and a scarcity of reporting guidelines. Nevertheless, new recommendations on liquid biopsy are emerging [30, 34], most notably with the updated consensus statement from the International Association for the Study of Lung Cancer (IASLC), published in 2021 (see Fig. 3) [34]. Additionally, there remains a risk of false negatives (sensitivity ~ 87%) as not all tumours shed sufficient cfDNA for detection, and cfDNA sequencing cannot distinguish morphological transition in the context of disease relapse with kinase inhibitor therapy [30]. Negative results from cfDNA analysis should therefore be confirmed by tissue testing (including a tissue re-biopsy if necessary). In the authors’ opinion, the issue of specificity will probably represent another significant limiting factor for new markers detected in cfDNA. Unlike EGFR mutations, which are highly specific for NSCLC, other mutations, such as BRAF p.V600E, are seen in different human malignancies. For these mutations, liquid biopsy can provide vital additional information to aid decision-making and may lead to the identification of a different tumour than expected, or a second tumour. Clonal haematopoiesis may also result in the expansion of mutations in peripheral blood cells, which can cause false positives if the liquid biopsy results are misinterpreted [29]. Finally, challenges limit liquid biopsy for gene fusion analysis by means of RNA-based NGS. Tumour cell-free RNA (cfRNA) can be found in the circulation but studies evaluating cfRNA as a diagnostic tool have been hampered by poor reproducibility and specificity due to issues with isolation procedures and background noise from healthy cells. Novel strategies to preserve, extract, and sequence extracellular mRNAs from plasma may help to overcome these obstacles in the future [35]. Given the limitations at present, it is recommended to pursue tissue-based testing whenever possible, and a detailed protocol for tissue utilisation and liquid biopsy should be established in each laboratory for evaluation of predictive biomarkers [36].

Cytology specimens have been demonstrated to be suitable for genomic profiling of patients with lung cancer [37, 38]. However, as visual verification of cellularity is not always possible, in the authors’ experience, potential false negative results should be considered in the absence of detected variants. Several factors may limit the accuracy of biomarker testing from cytological specimens including the potential small number of tumour cells analysed that may not recapitulate tumour heterogeneity, low DNA/RNA input, and a low ratio of neoplastic cells to non-transformed cells.