Updated Molecular Testing Guideline for the Selection of Lung Cancer Patients for Treatment With Targeted Tyrosine Kinase Inhibitors: Guideline From the College of American Pathologists, the International Association for the Study of Lung Cancer, and the Association for Molecular Pathology

The CAP, IASLC, and AMP convened an expert panel consisting of practicing pathologists and oncologists with expertise and experience in lung carcinoma. The CAP, IASLC, and AMP approved the appointment of the project cochairs and expert panel members. In addition, a methodologist experienced in systematic review and guideline development consulted with the panel throughout the project.

Prior to acceptance on the expert panel, potential members completed a joint conflict of interest disclosure process, whose policy and form require disclosure of material financial interest in, or potential for benefit of significant value from, the guideline's development or its recommendations. The potential members completed the conflict of interest disclosure form, listing any relationship that could be interpreted as constituting an actual, potential, or apparent conflict. Potential conflicts were managed by the cochairs. All expert and advisory panel members were required to disclose conflicts prior to beginning and continuously throughout the project's timeline. Disclosed conflicts of the expert panel members are listed in the Appendix. The CAP, IASLC, and AMP provided funding for the administration of the project; no industry funds were used in the development of the guideline. All panel members volunteered their time and were not compensated for their involvement. Please see the supplemental digital content (SDC) at www.archivesofpathology.org in the March 2018 table of contents for full details on the conflict of interest policy.

The expert panel was charged with the review and update of the CAP-IASLC-AMP molecular testing guideline for selection of lung cancer patients for EGFR and ALK tyrosine kinase inhibitors. The panel reviewed any new studies that would change or refute the statements from the 2013 guideline. In addition, the panel also addressed additional key questions:

1. Which new genes should be tested for lung cancer patients?

2. What methods should be used to perform molecular testing?

3. Is molecular testing appropriate for lung cancers that do not have an adenocarcinoma component?

4. What testing is indicated for patients with targetable mutations who have relapsed on targeted therapy?

5. What is the role of testing for circulating cell-free DNA for lung cancer patients?

Key questions 1 through 3 relate to patients diagnosed with nonsquamous non–small cell lung cancer (NSCLC) of all stages. The key questions are included in full detail in the SDC.

A detailed account of the methods used to create this guideline can be found in the SDC, including additional scope questions.

A systematic literature review was completed with 2 comprehensive searches. The first search was designed to assess the 2013 guideline statements and was based on the original search strategy. It included medical subject headings and keywords to address the concepts lung cancer, tumor biomarkers, and laboratory testing and was run in Ovid MEDLINE (Ovid Technologies, Inc, New York City, New York) on May 17, 2015, to locate studies published in English with publication dates from January 1, 2012 through May 17, 2015. Publication filters were applied to identify guidelines, systematic reviews, meta-analyses (MAs), and randomized clinical trials. The search was rerun on June 27, 2016, to identify relevant new literature published since May 17, 2015.

The second search was based on new key questions that focused on additional biomarkers not included in the 2013 guideline, with specific search strategies designed for each key question. All searches were performed in Ovid MEDLINE and PubMed (US National Library of Medicine, Bethesda, Maryland) (June 28, 2015) and were limited to English-language studies. Supplemental searches were run in Scopus (Amsterdam, Netherlands) (June 25, 2015) to identify relevant publications not indexed in MEDLINE. A search for relevant clinical trials was completed using the clinicaltrials.gov Web site, and focused searches on guideline repository sites (eg, guideline.gov, g-i-n.net) and organizations' Web sites were undertaken to identify relevant publications. Further detail about the systematic literature search, including the Ovid search strings, can be found in the SDC.

Studies were not limited to randomized controlled trials but also included other study types, including cohort designs, case series, evaluation studies, and comparative studies. Letters, commentaries, editorials, narrative reviews, case reports, studies in mouse models, in vitro studies, consensus documents, abstracts, and non-English articles were excluded a priori.

Published studies were selected for inclusion in the systematic review of evidence if they were peer-reviewed full-text articles that met the following criteria:

1. The study population consisted of patients with nonsquamous, non–small cell lung adenocarcinoma, small cell lung carcinoma, or squamous cell lung cancer of any stage.

2. The study evaluated, prospectively or retrospectively, sensitivity, specificity, negative predictive value, or positive predictive value of EGFR, ALK, KRAS, ROS1, RET, MET, BRAF, or ERBB2 (HER2) tests for detection of gene-specific mutation, rearrangement, translocation, amplification, overexpression, or response to a targeted gene-specific therapy.

3. The study examined potential testing algorithms for NSCLC molecular testing.

4. The study examined the correlation of EGFR, ALK, KRAS, ROS1, RET, MET, BRAF, or ERBB2 (HER2) status in primary or metastatic tumors from the same patients.

5. The study included primary outcomes such as accuracy, sensitivity, specificity, positive predictive value, and negative predictive value of tests and concordance across platforms to determine EGFR, ALK, KRAS, ROS1, RET, MET, BRAF, or ERBB2 (HER2) status or treatment response, alone or in combination.

An assessment of the quality of the evidence was performed for all retained studies following application of the inclusion and exclusion criteria. Using this method, studies deemed low quality would not be excluded from the systematic review, but would be retained and their methodologic strengths and weaknesses discussed where relevant. Each guideline statement includes a rating of the strength of the evidence as described in Table 1 (also in SDC Table 1). The process used to assess the quality of the evidence base is fully detailed in the SDC.

In order to articulate recommendation statements that were clearly written and easy to implement, the expert panel used GLIDES (Guidelines Into Decision Support) methodology and accompanying BridgeWiz software (Yale University, New Haven, Connecticut).²⁷  This methodology prioritizes the use of active language; however, in some situations, the person responsible for ensuring guidance is implemented is dependent on the organization of the clinic and/or laboratory. To ensure clarity of guidance in these situations, the expert panel used passive-voice language to emphasize the recommended action. Development of recommendations required that the panel review the identified evidence and make a series of key judgments (using procedures described in the SDC). This guideline uses a 3-tier system to rate the strength of recommendations, as well as a “no recommendation” category when there is insufficient evidence to support a recommendation. Table 2 (also in SDC Table 2) summarizes the strength of evidence and net benefits and harms, as well as obligatory language that was used for each of the recommendation types.

This guideline will be reviewed every 4 years or earlier in the event of publication of substantive and high-quality evidence that could potentially alter the original guideline statements. If necessary, the entire panel will reconvene to discuss potential changes and, if indicated, recommend revision of the guideline to CAP, IASLC, and AMP.

Practice guidelines and consensus statements reflect the best available evidence and expert consensus supported in practice. They are intended to assist physicians and patients in clinical decision making and to identify questions and settings for further research. With the rapid flow of scientific information, new evidence may emerge between the time a practice guideline or consensus statement is developed and when it is published or read. Guidelines and statements are not continually updated and may not reflect the most recent evidence. Guidelines and statements address only the topics specifically identified therein and are not applicable to other interventions, diseases, or stages of diseases. Furthermore, guidelines and consensus statements cannot account for individual variation among patients and cannot be considered inclusive of all proper methods of care or exclusive of other treatments. It is the responsibility of the treating physician or other health care provider, relying on independent experience and knowledge, to determine the best course of treatment for the patient. Accordingly, adherence to any practice guideline or consensus statement is voluntary, with the ultimate determination regarding its application to be made by the physician in light of each patient's individual circumstances and preferences. The CAP, IASLC, and AMP make no warranty, express or implied, regarding guidelines and statements and specifically exclude any warranties of merchantability and fitness for a particular use or purpose. The CAP, IASLC, and AMP assume no responsibility for any injury or damage to persons or property arising out of or related to any use of this statement or for any errors or omissions.

For the reaffirmation of the 2013 guideline recommendations, a total of 610 studies met the search term requirements. Following a review of the 610 abstracts, the full texts of 77 studies that met the inclusion criteria and were likely to refute or change the 2013 recommendations were reviewed. A total of 21 articles were included for data extraction. Excluded articles were available as discussion or background references.

For the new key questions, 1654 articles met the search term requirements. Based on review of these abstracts, 488 articles met the inclusion criteria and continued to full-text review. Articles that addressed any of the new key questions were moved to a second-level full-text–review phase. A total of 118 articles were included for data extraction. Excluded articles were available as discussion or background references.

The panel convened 5 times (3 times by teleconference and 2 face-to-face meetings) to develop the scope, draft recommendations, review and respond to solicited feedback, and assess the quality of evidence that supports the final recommendations. A nominal group technique was used by the panel for consensus decision making to encourage unique input with balanced participation among group members. An open comment period was held from June 28 to August 2, 2016, during which the 2013 guideline statements and new draft recommendations and statements were posted for public comment. The public comment period was posted on the AMP Web site at www.amp.org. All 2013 recommendations received strong agreement (95%–99%) from the open comment period participants. There were 20 new draft statements with strong agreement, ranging from 86% to 97%, from the open comment period participants (refer to Outcomes in the SDC for full details). The expert panel members were assigned to review the public comments in small groups. The panel modified the draft statements and recommended the deletion of 1 expert consensus opinion and a no recommendation statement based on the feedback during the considered judgment process. The final recommendations were approved by the expert panel with a vote. The panel considered benefits and harms, required resources, feasibility, and acceptability throughout the entire process, although neither cost nor cost-effectiveness analyses were performed. A description of the benefits and harms of implementing the guideline statements is included in the SDC (SDC Table 3).

Each organization instituted a review process to approve the guideline. For the CAP, an independent review panel representing the Council on Scientific Affairs was assembled to review and approve the guideline. The independent review panel was masked to the expert panel and vetted through the conflict of interest process. The IASLC approval process required review and approval by the IASLC Board of Directors. The AMP approval process required content review by an independent subject matter expert panel, led by the Publications & Communications chair, with representation from the Clinical Practice Committee and Solid Tumors Subdivision leadership, and organizational approval by the AMP Executive Committee.

The 2013 guideline recommended universal testing of lung cancer patients with advanced-stage cancers with an adenocarcinoma component, using molecular diagnosis for activating “hot-spot” mutations in EGFR exons 18 to 21 with at least 1% prevalence (ie, codons 709 and 719, exon 19 deletion 768, and exon 20 insertions 790, 858, and 861), and using fluorescence in situ hybridization (FISH) for rearrangements involving ALK. Any methodology or testing algorithm with suitable analytic sensitivity (ability to detect mutations in formalin-fixed samples with 50% or more malignant cells) and turnaround time (10 days between sample receipt and reporting of all results), with appropriate validation and deployment under the Clinical Laboratory Improvement Act of 1988, was acceptable.

The 2013 guideline recommended against applying clinical parameters (eg, tobacco exposure, age, sex, ethnicity) to select patients for testing, testing pure squamous carcinomas, using KRAS negativity as a determinant of anti-EGFR therapy, using IHC for EGFR or ALK testing, and using FISH for EGFR testing.

The 2013 guideline left several decisions open to each institution to set policy, such as whether or not to test early-stage patients, whether or not to use clinical predictors to select patients with minimally sampled squamous carcinoma biopsies such that a mixed adenosquamous carcinoma could not be excluded, and whether or not to use a simultaneous or sequential testing approach. Of these, the question concerning testing early-stage disease remains open, and awaits data from more clinical trials before an evidence-based recommendation can be made. Although the American Society of Clinical Oncology Clinical Practice Guidelines Committee highlighted consideration of molecular testing for early-stage lung cancer patients,⁴  our opinion remains that each institution should set its own policy regarding testing patients with early-stage disease, balancing the benefit of having results on record from testing a high-quality resection sample for alterations that are likely to become necessary at a time of future progression when a high-quality sample could be hard to obtain against the cost of testing patients for whom a subset will be surgically cured and never need the test result. Accordingly, the testing recommended below applies to patients with advanced-stage (stages IIIB and IV) lung cancer.

Following review of literature published since 2013, the original recommendations are largely reaffirmed. Several statements have gained strength with the publication of additional supporting evidence (SDC Tables 4a, 4b, and 5). Some warranted a complete reevaluation in this revision, and will appear subsequently (Table 3); these include the use of IHC for ALK, the use of multigene NGS panels, and the question of testing nonadenocarcinoma samples.

Of the remaining 2013 recommendations, the following changes are made:

The original recommendation preferred cell blocks over smears. A recent systematic review²⁸  identified by the literature search has indicated that numerous studies have been published showing excellent performance of smear preparations, such that this preference is no longer appropriate. It is incumbent upon any laboratory that tests cytopathology specimens to perform appropriate validation studies of these as separate sample types, distinct from tissue and blood samples.

Although the original studies demonstrating response of EGFR-mutated lung cancers to treatment with EGFR inhibitors used unmodified Sanger sequencing with a sensitivity limit of 50% tumor cellularity, this is insufficient in practice because many lung cancer samples are small and comprise a majority of benign stromal cells, and most of the larger phase III clinical trials that confirmed the clinical utility of EGFR mutation testing used polymerase chain reaction (PCR)–based methods that were more sensitive than unmodified Sanger sequencing. Given the widespread availability of technology capable of reliably detecting lower-frequency mutational events in small samples, it is no longer appropriate to offer a low-sensitivity test that cannot test tumors with 20% to 50% tumor content and requires patients to undergo more procedures, and potentially more invasive procedures, solely to procure a tissue sample with high tumor content.

There is no role whatsoever for IHC against total EGFR protein as a determinant of treatment with an EGFR kinase inhibitor. The targetable mutations lead to activation of the cytoplasmic kinase of this transmembrane protein, but that has no bearing on the extent of expression at the cell surface, which is what is detected by the total EGFR immunostain. Although EGFR expression by IHC was performed for some of the very early studies of EGFR kinase inhibitors in the start of this century, clinical responses were seen in patients with mutations but absent/weak IHC expression, and poor responses were seen in patients with strong IHC expression but no mutations.

Following the discovery of EGFR mutations, antibodies were developed for IHC directed at the most common mutated forms of the protein, most notably the L858R substitution and the 746 to 750 ELREA deletion. The original guideline allowed for the use of the mutant-specific EGFR antibodies by IHC in a setting with extremely limited material. Although published evidence for these antibodies shows good accuracy for the L858R activating mutation and for some of the exon 19 deletions, these antibodies have poor sensitivity for other exon 19 deletions, insensitivity to less common mutations (eg, codon 719 mutations), and false-positive results with exon 20 insertions.²⁹  Overall, the performance is suboptimal for reliable detection of EGFR mutations. Given that advances in molecular diagnostic technology now enable analysis of very limited samples as well as circulating tumor DNA (see below), at this time there is no role for routine use of mutant-specific IHC in selecting anti-EGFR treatment for lung cancer patients.

Question 1: Which New Genes Should Be Tested for Lung Cancer Patients?—In the 2013 guideline, genes fell into 1 of 2 categories: testing is necessary (EGFR, ALK), or testing is investigational. One gene, KRAS, was considered conditionally necessary in the context of sequential testing algorithms because of its ease of analysis and mutual exclusivity with EGFR and ALK. By 2018, however, we believe that there are now 3 categories into which genes should be placed. One set of genes must be offered by all laboratories that test lung cancers, as an absolute minimum: EGFR, ALK, and ROS1. A second group of genes should be included in any expanded panel that is offered for lung cancer patients: BRAF, MET, RET, ERBB2 (HER2), and KRAS, if adequate material is available. KRAS testing may also be offered as a single-gene test to exclude patients from expanded panel testing. All other genes are considered investigational at the time of publication.

In this context, institutions providing care for lung cancer patients have a choice: (1) offer a comprehensive cancer panel that includes all of the genes in the first 2 categories (EGFR, ALK, ROS1, BRAF, MET, ERBB2 [HER2], KRAS, RET) for all appropriate patients, or (2) offer targeted testing for the genes in the must-test category (EGFR, ALK, ROS1) for all appropriate patients and offer as a second test an expanded panel containing the second-category genes (BRAF, MET, ERBB2 [HER2], and RET) for patients who are suitable candidates for clinical trials, possibly after performing a single-gene KRAS test to exclude patients with KRAS-mutant cancers from expanded panel testing. Table 4 includes a list of the recommendation statements with the strength of recommendations.

ROS1 testing must be performed on all lung advanced-stage adenocarcinoma patients, irrespective of clinical characteristics.

The strength of evidence was convincing to support the use of ROS1 molecular (ie, reverse transcription PCR [RT-PCR] or sequencing) or cytogenetic (ie, FISH or other in situ hybridization) testing to identify ROS1 rearrangements. The strength of evidence supporting the use of any clinical characteristic to identify patients who should receive ROS1 testing was adequate. This recommendation is evidence-based and supported by 9 studies,^30–38  6 of which informed on the association between ROS1 rearrangement and patient or tumor characteristics^{30,31,34–37}  and consisted of 1 prospective cohort study (PCS),³⁵  1 prospective-retrospective cohort study (PRCS),³¹  and 4 retrospective cohort studies (RCSs).^30,34,36,37  The 3 remaining studies assessed clinical outcomes of patients treated with the ROS1-targeted therapy crizotinib^32,33,38  and included 1 nonrandomized clinical trial³³  and 3 RCSs.^32,38  All included studies were assessed for quality and none were found to have methodologic flaws that would raise concerns about the studies' findings (SDC Table 6). Refer to SDC Table 7 for a summary of findings from studies supporting the use of ROS1 molecular or cytogenetic testing to enable selection of patients for ROS1-targeted therapy.

Although relatively rare, accounting for less than 2% of non–small cell lung carcinomas^30,31,34  and 2% to 3% of lung adenocarcinomas,^30,34,35  structural rearrangements involving the ROS1 gene generate an oncogenic fusion that can be treated successfully with targeted inhibitors. A single phase I clinical trial of 50 NSCLC patients demonstrated that the presence of a ROS1 rearrangement by FISH or RT-PCR predicts response to targeted inhibition using crizotinib, with a response rate of 72% and median progression-free survival of 19.2 months.³³  Based on this trial, the FDA approved the expanded use of crizotinib in patients with ROS1-rearranged NSCLC in 2016. A European multi-institutional retrospective study of 32 patients with ROS1-rearranged NSCLC treated with crizotinib demonstrated an 80% response rate and 9.1-month progression-free survival.³²  Overall survival for patients with ROS1-rearranged tumors irrespective of use of targeted therapy appears longer than that for patients with other molecular alterations undergoing targeted therapy.^38,39 

As with ALK, ROS1 activation is driven by structural variants, with multiple different partners fusing to the C-terminal portion of ROS1 containing the cytoplasmic tyrosine kinase and driving downstream signaling through MAPK, JAK/STAT, and PI3K pathways. Common fusion partners include SLC34A2, CD74, and TPM3, among others. The role of wild-type ROS1 is still being elucidated, but it shares similar structure with ALK, albeit with significant differences, notably absence of a dimerization domain, an extracellular domain with some resemblance to cell adhesion molecules, and no clear ligand.

As with EGFR mutations and ALK rearrangements, light to never smoking history has been associated with an increased incidence of ROS1 rearrangements in patients with lung adenocarcinoma.^30,37  However, this association has not been consistently observed across studies.³⁴  Other clinical characteristics, such as younger age, female sex, and non-Asian ethnicity, have been associated with ROS1 rearrangement in isolated studies only.^30,31,35  Therefore, clinical characteristics should not be used to either select or exclude patients from testing for ROS1 rearrangements. ROS1 rearrangements occur in a mutually exclusive fashion with other oncogenic driver alterations (such as EGFR and KRAS mutation and ALK rearrangement). In recognition of the rarity of ROS1 rearrangement, it may be reasonable to perform sequential testing of EGFR and ALK followed by ROS1 testing. Indeed, the frequency of ROS1 rearrangements is enriched to 5% to 10% in otherwise driver (ie, EGFR, ALK, KRAS, BRAF)-negative lung adenocarcinomas.^31,37 

Notably, in the United States in 2016, crizotinib therapy in ROS1-rearranged tumors does not require the use of an FDA-approved companion diagnostic. Published methods that have established clinical utility of testing ROS1 in order to choose ROS1-targeted therapy have relied primarily upon FISH and RT-PCR. Outside the United States, a diagnostic test using RT-PCR was used for an international phase II clinical trial,⁴⁰  involving mainly East Asian countries, for selection of tumors with ROS1 rearrangement. This assay has been approved as an in vitro diagnostic in Europe and China, and may be recognized as a companion diagnostic test in some countries. Although targeted RT-PCR assays may be challenging because of variation in ROS1 break points (typically introns 31–35) and partner genes, capture-based sequencing strategies for RNA or DNA may be used, provided they are properly validated on known positive samples. Within the United States, FISH methods have been published more frequently. Fluorescence in situ hybridization testing should be performed with a break-apart probe design given the multiple fusion partners, and should show rearrangement, defined as signals split by at least 1 probe diameter, in 15% or more of tumor cells.⁴¹ 

ROS1 IHC may be used as a screening test in advanced-stage lung adenocarcinoma patients; however, positive ROS1 IHC results should be confirmed by a molecular or cytogenetic method.

The strength of evidence is inadequate supporting the use of IHC as a screening assay for ROS1 molecular testing. This statement is evidence-based and supported by 6 studies,^42–47  consisting of 2 PCSs,^43,44  1 PRCS,⁴²  and 3 RCSs.^45–47  Five studies compared ROS1 IHC with a FISH reference test^42–45,47  and 1 study compared ROS1 IHC with an RT-PCR reference test.⁴⁶  Using reported true-positive, false-positive, true-negative, and false-negative data from studies comparing IHC with FISH, an MA was conducted to determine a pooled estimate of sensitivity and specificity for ROS1 IHC (Figure 1). All included studies were assessed for quality and none were found to have methodologic flaws that would raise concerns about the studies' findings (SDC Table 8). Refer to SDC Table 9 for a summary of findings from studies supporting the use of IHC as a screening assay for ROS1 molecular testing.

In light of the relative rarity of ROS1 rearrangement events in NSCLC, screening by IHC may be preferable to FISH or molecular techniques in some settings. Interpretation of ROS1 IHC is challenging, however, as expression can be seen in a patchy pattern, typically at weak intensity, in up to a third of tumors that do not have an underlying rearrangement.^44,45,48  Although some studies suggest that ROS1 expression in the absence of a rearrangement may have prognostic significance,⁴⁸  focal or patchy expression in tumor cells is rarely associated with a ROS1 rearrangement and therefore is unlikely to predict response to ROS1-targeted therapy. Moreover, the pattern of staining can vary among fusion types, including granular to globular staining in CD74-ROS1 fusions, weak membranous staining in EZR-ROS1 fusions, and vesicular localization staining in GOPC-ROS1 fusions.⁴⁵ 

A single commercially available antibody clone (D4D6, Cell Signaling Technology, Danvers, Massachusetts) has been used in studies published to date. Most retrospective studies of ROS1 IHC using the D4D6 antibody demonstrate a sensitivity of 100% relative to FISH or RT-PCR.^35,42–47  Tumors lacking ROS1 expression can be safely interpreted as lacking a ROS1 fusion. However, the specificity of ROS1 IHC is more variable, ranging from 92% to 100% using different methods and interpretive cutoffs.^35,42–47  Meta-analysis of 5 studies identified by the literature search determined a pooled sensitivity of 96% (95% CI, 71%–99%) and specificity of 94% (95% CI, 89%–96%) for IHC compared with FISH when the D4D6 antibody with a staining intensity of at least 2+ (as defined within the study) was used (Figure 1). Several cutoffs have been proposed using intensity alone or H score (intensity × percentage of tumor cells staining). In most studies, FISH- or molecularly confirmed ROS1-rearranged tumors have at least moderate-intensity ROS1 protein expression, but published evidence is insufficient to recommend one specific cutoff or scoring system,^42,45  and each laboratory must validate its own interpretive cutoff from known positive and negative samples.

Because of imperfect specificity and challenges related to interpretation of nonspecific expression, we recommend that all ROS1 IHC positive results undergo confirmation by FISH or a molecular method (ie, RT-PCR, NGS) prior to considering a patient for ROS1-targeted therapy. Given the high sensitivity of IHC, however, tumors that clearly lack ROS1 staining can be interpreted as negative for ROS1 fusion.

Of the genes newly included in this guideline, only ROS1 testing must be offered to all appropriate lung cancer patients. Testing for the following genes should be included with any expanded multigene panel testing performed for lung cancer patients, whether or not the panel is offered for all lung cancer patients, or if the panel is reserved as a second-line test for EGFR/ALK/ROS1 wild-type patients seeking clinical trials.

BRAF molecular testing is currently not indicated as a routine stand-alone assay outside the context of a clinical trial. It is appropriate to include BRAF as part of larger testing panels performed either initially or when routine EGFR, ALK, and ROS1 testing is negative.

The strength of evidence was inadequate to support the use of BRAF molecular testing. This statement was evidence-based and supported by 9 studies: 4 PCSs^49–52  and 3 RCSs,^53–55  all of which informed on the association between BRAF mutation and patient or tumor characteristics,^49–55  and 2 additional nonrandomized clinical trials that assessed the activity of a BRAF inhibitor in p.V600E mutant patients.^56,57  All included studies were assessed for quality and none were found to have methodologic flaws that would raise concerns about the studies' findings (SDC Table 10). Refer to SDC Table 11 for a summary of findings from studies supporting the use of BRAF molecular testing.

Activating mutations in BRAF, especially p.V600E, lead to oncogenic signaling through MAPK, and are rare recurrent alterations in lung adenocarcinoma, seen in 0.5% to 4.9% of tumors.^49–52,54  In lung cancer, data from a 2016 phase II single-arm clinical trial⁵⁶  showed that (1) single-agent dabrafenib given in second line to stage IV BRAF p.V600E mutant NSCLC had a partial response rate of 33% and disease control rate of 58% and (2) combination dabrafenib-trametinib therapy given in second line to stage IV BRAF p.V600E mutant lung adenocarcinoma had a partial response rate of 63% and disease control rate of 75%.

Based on these data, the FDA conferred a breakthrough therapy designation for the combination treatment in BRAF p.V600E mutation–positive NSCLC, and FDA approval was granted in 2017. Hence, this was the most controversial of all recommendations among the working panel. Although there was a strong opinion in the working group that BRAF mutation analysis should be performed at the time of initial molecular testing in lung adenocarcinoma, the published evidence available at the time of publication lacked controlled prospective trials, and therefore lacked the strength to warrant an international recommendation for single-gene testing for BRAF for all lung adenocarcinoma patients. We anticipate the publication of stronger evidence supporting the utility of BRAF inhibition in BRAF-mutant lung cancer, and our opinion is that BRAF testing will be proven necessary. We expect that the next revision of this guideline will include a recommendation to include single-gene testing for BRAF alongside EGFR, ALK, and ROS1, but we are unable to make that recommendation in the spring of 2017. Although stand-alone single-gene testing for BRAF is not currently recommended, if a panel strategy is used, either initially or for patients who are known wild type for EGFR, ALK, and ROS1, then BRAF should be included.

As with EGFR and KRAS mutations, selected hot-spot mutations in BRAF exert an oncogenic effect. The V-raf murine sarcoma homolog b (BRAF) gene encodes for a nonreceptor serine-threonine kinase in the MAPK kinase signaling pathway, between RAS and MEK. The most common BRAF mutation in NSCLC is the c.1799T>A (p.V600E) point mutation that is the predominant mutation in many other cancers, including melanoma, papillary thyroid cancer, colorectal cancer, hairy cell leukemia, and ganglioglioma. However, in contrast to other cancers with BRAF mutations, lung cancers frequently have non-p.V600E BRAF mutations, including other mutations at codon 600 (eg, p.V600K) and nearby codons in exon 15, and substitutions at codons 466 and 469 in exon 11.

Like many other targetable oncogenes in lung cancer, BRAF mutations are more frequent in adenocarcinomas than in squamous cell carcinomas. BRAF p.V600E mutation is more frequent in females^52,54  and never smokers⁵⁴  in some studies, but several studies failed to show these associations.^49,50,53,58  One distinction between BRAF mutations and other targetable oncogenes is that non-p.V600E BRAF mutations (particularly the exon 11 mutations) may coexist with mutations in KRAS,^49,52,53,59  whereas the p.V600E mutations are mutually exclusive of KRAS, EGFR, or ALK alterations.

Single-gene assays for BRAF are in wide use for other cancer types, particularly for melanoma patients being considered for targeted therapy, but most of these methods cannot detect the exon 11 mutations that are seen in lung cancer. Although the evidence supporting utility of BRAF testing was restricted to the p.V600E mutations, our opinion is that testing for BRAF, done as part of a large panel or for clinical trial enrollment, should use a method that evaluates at a minimum exons 11 and 15.

A similar challenge arises concerning the use of mutation-specific IHC using antibodies against the p.V600E mutant protein (VE1), which have been widely used in melanoma diagnosis. Reported data on small numbers of lung cancer cases^58,60  demonstrate the VE1 clone can stain between 90% and 100% of p.V600E-mutant adenocarcinomas. In 1 of these studies, all non-p.V600E cases were negative on IHC testing,⁶¹  whereas in another, a single non-p.V600–mutated case out of 21, with a unique 599 insertion T mutation, showed positive staining. There is currently insufficient evidence to support a recommendation either for or against BRAF p.V600E IHC (VE1) testing in NSCLC.

RET molecular testing is not recommended as a routine stand-alone assay outside the context of a clinical trial. It is appropriate to include RET as part of larger testing panels performed either initially or when routine EGFR, ALK, and ROS1 testing is negative.

The strength of evidence to support the use of RET molecular testing was inadequate. This statement is evidence-based and supported by 3 studies,^37,62,63  consisting of 1 PCS⁶²  and 2 RCSs.^37,63  All included studies were assessed for quality and none were found to have methodologic flaws that would raise concerns about the studies' findings (SDC Table 12). Refer to SDC Table 13 for a summary of findings from studies supporting the use of RET molecular testing.

Structural variants causing RET fusions are rare, being found in 0.6% to 0.9% of NSCLCs and in 1.2% to 2% of adenocarcinomas.^62,64–67  The potential to treat RET-positive lung cancers with inhibitors of the RET kinase is being explored in phase II clinical trials,^68,69  although small series and case reports have shown promise.^70,71  Given the rarity of RET rearrangements and limited evidence of therapeutic benefit, testing for RET alterations is not recommended as a stand-alone test for all lung adenocarcinoma patients. However, any large multigene panel test developed for lung cancer patients, either for initial workup or for patients who are wild type for EGFR, ALK, and ROS1, should include RET.

As with ALK and ROS1 rearrangements, RET is activated by rearrangements that fuse the tyrosine kinase domain of RET with coiled-coil dimerization domains of one of a variety of recurring partner genes, including KIF5B (the most common, at 90%),^64,72,73  CCDC6,^65,74  NCOA4,⁶²  and TRIM33.⁷²  RET rearrangement is mutually exclusive with aberrations in EGFR, KRAS, ALK, HER2, and BRAF in lung cancer.^62,64,65 

RET fusion occurs more frequently in never smokers than ever smokers.^{37,62,64,66,72}  Patients with RET fusion harboring tumors are usually younger than patients with an EGFR mutation and have an equal sex distribution.⁶⁷  RET fusion proteins have been detected in adenocarcinoma^37,62,64  and in adenosquamous carcinoma.⁶²  Histologic subtypes in adenocarcinomas include those with mucinous/signet ring cells and those with a cribriform^37,62,65  or solid growth pattern.^37,62  However, no clinical or histologic features (other than excluding from testing pure squamous histology cases) should be used to select a patient for RET testing.

Multiple methods have been applied for RET analysis, including break-apart FISH analysis,⁷⁵  IHC,³⁷  RT-PCR,⁷⁵  and NGS.³⁷  RET FISH is particularly challenging, however, because of the narrow spacing between the split probe signals seen in the common fusion types, and a pattern of split RET signals separated by as little as 1 signal diameter distance is interpreted as positive.³⁷  Similar to ALK rearrangement testing by FISH, the threshold for RET FISH positivity for rearrangement is 15% of cells with split signals or single 3′ probe signals. In another study, a 4-colored RET FISH assay was used⁶² ; samples were positive for RET rearrangement or KIF5B-RET fusion if more than 20% of tumor cells exhibited split red-green signals or touching golden-green signals, respectively.

One recent retrospective study used RET IHC (anti-RET antibody ab134100, Abcam, Cambridge, United Kingdom) showing diffusely granular cytoplasm staining and occasionally membranous or perinuclear staining, with moderate to strong intensity. A sensitivity of 100% and specificity of 88% were reported,³⁷  although corroborating evidence is not strong enough to warrant a recommendation.

Although multiplex RT-PCR may be successful for common fusions involving KIF5B-RET and CCDC6-RET,⁷⁵  as with ALK and ROS1, targeted RT-PCR alone is usually insufficient to detect new partners or isoforms. However, although the diversity of treatable rearrangements in ALK and ROS1 has matured sufficiently through years of testing and clinical trials, such that targeted RT-PCR assays for these genes can be designed with adequate clinical sensitivity, the diversity of treatable RET rearrangements is earlier in evolution. A capture-based sequencing approach, involving DNA or RNA, may be more sensitive and more readily integrated into a large multigene panel.⁷⁶ 

ERBB2 (HER2) molecular testing is not indicated as a routine stand-alone assay outside the context of a clinical trial. It is appropriate to include ERBB2 (HER2) mutation analysis as part of a larger testing panel performed either initially or when routine EGFR, ALK, and ROS1 testing is negative.

The strength of evidence was inadequate to support the use of ERBB2 (HER2) molecular testing. This recommendation was evidence-based and supported by 10 studies, 9 that reported on the association between ERBB2 (HER2) and patient or tumor characteristics^49,77–84  and 1 that assessed the use of ERBB2-targeted therapy (dacomitinib)⁸⁵  in patients with ERBB2 (HER2) mutations and amplifications. All included studies were assessed for quality and none were found to have methodologic flaws that would raise concerns about the studies' findings (SDC Table 14). Refer to SDC Table 15 for a summary of findings from studies supporting the use of ERBB2 (HER2) molecular testing.

Alterations in the human epidermal growth factor receptor 2 gene (HER2, ERBB2) have emerged as oncogenic drivers and as potential therapeutic targets in lung cancer.^81,83,84,86  Sequence alterations and gene amplification occur in this setting and constitute approximately 2% to 3% and 2% to 5% of reported recurrent alterations, respectively. Therapeutic targeting of HER2 (the protein product of the ERBB2 gene) remains an area of active investigation at this time. Earlier clinical trials selecting patients based on protein expression by IHC or ERBB2 amplification by FISH did not demonstrate a clear benefit.^87,88  An additional phase II trial using ERBB2 mutation and ERBB2 amplification for patient selection demonstrated durable responses to dacomitinib, but only in patients with specific HER2 mutations.⁸⁵ 

In-frame insertions in exon 20 and substitutions at S310 are the most common mutations seen, and are typically mutually exclusive with other recurrent alterations, including mutations in EGFR, KRAS, and BRAF, as well as rearrangements involving ALK and ROS1. Insertions in exon 20 are variable, with most being a 12–base pair duplication of codons 775–778 encoding amino acids YVMA,⁸¹  and are more commonly observed in younger patients and patients with no smoking history. De novo ERBB2 amplification may occur with or without ERBB2 mutation,^82,84,86  with highly variable reported rates of co-occurrence from 0% to 87%.^81,84,86  Although differences in methods and criteria defining amplification levels may be responsible for these observed discrepancies and require standardization, the higher prevalence of ERBB2 amplification independent of ERBB2 mutation suggests that mutation and amplification could represent distinct markers and therapeutic targets in lung cancer.⁸⁹  ERBB2 amplification has also been reported rarely as a secondary event in patients with sensitizing EGFR mutations and as a potential mechanism of resistance following treatment with EGFR inhibitors.⁹⁰ 

In this context and with current evidence, routine stand-alone testing for ERBB2 mutations is not indicated outside a clinical trial. Nevertheless, when broader testing is performed through a multiplex assay or NGS, it is appropriate to include ERBB2 as part of the testing, as it may identify patients to be directed to clinical trials—in this context, testing for sequence alterations in ERBB2, particularly insertions/duplications in exon 20, which have been associated with response to treatment with targeted inhibitors of ERBB2 in case reports and small series.^85,91 

KRAS molecular testing is not indicated as a routine stand-alone assay as a sole determinant of targeted therapy. It is appropriate to include KRAS molecular testing as part of larger testing panels performed either initially or when routine EGFR, ALK, and ROS1 testing is negative.

The strength of evidence was adequate to support the use of KRAS molecular testing when selecting patients for targeted therapy. The strength of evidence supporting the use of any clinical characteristic to identify patients who should receive KRAS testing was inadequate. This statement is evidence-based and supported by 7 studies,^{49,51,52,92–95}  comprising 2 MAs,^93,94  4 PCSs,^49,51,52,92  and 1 RCS.⁹⁵  Five studies attempted to identify associations between patient or tumor characteristics and KRAS mutational status.^{49,52,92,93,95}  Two MAs^93,94  reported on overall survival and EGFR-TKI response rates when KRAS mutation–positive patients were treated with standard care. All included studies were assessed for quality and none were found to have methodologic flaws that would raise concerns about the studies' findings (SDC Table 16). Refer to SDC Table 17 for a summary of findings from studies supporting the use of KRAS molecular testing.

KRAS mutations are reported in 20% to 30% of lung adenocarcinomas. KRAS mutations are encountered more frequently in people with tobacco exposure, but have been reported in approximately 5% of lung cancer patients who have never used tobacco. Most studies indicate an increased incidence in males and those of white or African ancestry, in comparison with females and those of Asian ancestry. KRAS mutations occur most frequently in codon 12 and 13, much less commonly in codon 61, and rarely in codon 146, and can readily be detected by quick targeted hot-spot assays (ie, real-time PCR, droplet digital PCR, or pyrosequencing) interrogating these codons, as well as incorporated into larger panel tests. They are typically mutually exclusive with other driver mutations such as EGFR mutations and ALK rearrangements.^{49,51,92,93,95–101} 

Therapies directed against mutated KRAS have not been proven clinically effective. For example, although promising results (37% objective response rate) were obtained in a phase II study of selumetinib, an inhibitor of MEK1 (downstream of KRAS), plus docetaxel¹⁰²  in KRAS-mutant advanced lung cancer, this combination failed to demonstrate an outcome benefit in the Selumetinib Evaluation as Combination Therapy-1 (SELECT-1) phase III trial,¹⁰³  and a phase II study of selumetinib + erlotinib in KRAS-mutant lung cancers failed to show response to selumetinib independent of erlotinib.¹⁰⁴  Hence, intense research investigation into therapeutic strategies against this common mutation continues, and it is appropriate to include KRAS in a larger testing panel used for directing patients to investigational therapies.

Another application of KRAS mutation testing is in a sequential testing algorithm, with a positive result greatly diminishing the likelihood of another, targetable oncogenic alteration. If the KRAS test is performed prior to EGFR, ALK, or ROS1 testing, however, the laboratory must ensure that sufficient tumor is available for EGFR, ALK, and ROS1 testing within the recommended time frame, particularly in the event of a negative KRAS result. Similarly, the presence of a KRAS mutation renders unlikely the other oncogenes recommended for larger panels, such as RET, ERBB2 (HER2), and BRAF. In this context, a rapid, targeted assay for KRAS may have value in helping to determine whether or not an EGFR/ALK/ROS1 wild-type patient would benefit from expanded panel testing, in that panel testing would be less likely to benefit KRAS-mutant cancer patients. This model may, however, change as technology evolves, as newer ultrasensitive methods have shown co-occurrence of driver oncogenes, including KRAS, in subpopulations within tumors that previously had not been detected by less sensitive methods.^105,106 

MET molecular testing is not indicated as a routine stand-alone assay outside the context of a clinical trial. It is appropriate to include MET as part of larger testing panels performed either initially or when routine EGFR, ALK, and ROS1 testing is negative.

The strength of evidence is inadequate supporting the use of MET molecular testing. This statement was evidence-based and supported by 7 studies,^107–113  comprising 1 MA,¹⁰⁷  1 phase II randomized controlled trial,¹⁰⁹  1 PCS,¹¹⁰  and 4 RCSs.^{108,111–113}  All included studies were assessed for quality and none were found to have methodologic flaws that would raise concerns about the studies' finding (SDC Table 18). Refer to SDC Table 19 for a summary of findings from studies supporting the use of MET molecular testing.

Initially reported as a mechanism of secondary resistance to EGFR TKI therapy in EGFR-mutant lung cancer,^114,115  both the understanding of the mechanism of activation of MET and the utility of MET testing in lung cancer have gone through several phases. MET copy gain was initially recognized in association with secondary resistance to EGFR inhibitors,¹¹⁴  prompting the development of targeted therapies that showed disappointing results.¹¹⁶  More recently, interest in targeting MET has been rekindled by the discovery of activating mutations that may respond to targeted inhibition.

The MET gene encodes for the receptor for hepatocyte growth factor (HGFR), and its activation has pleotropic functions in promoting cell survival, proliferation, motility, invasion, and epithelial-mesenchymal transition.^117–120  HGFR can become activated and drive oncogenesis through several different mechanisms, including (1) amplification resulting in high expression of the receptor,^121,122  (2) tyrosine kinase domain mutations resulting in constitutive activation of the receptor,¹²³  and (3) splicing mutations resulting in skipping of exon 14 and loss of Y1003, the Casitas B-lineage lymphoma proto-oncogene (CBL) binding site required for the ubiquitin-mediated degradation of the protein.¹²⁴  Although most of the exon-skipping mutations involve canonical splice sites, some are located further into the intronic sequence, and thus can be difficult to interpret or may be missed by assays examining only exons and the immediately adjacent 5′ and 3′ splice acceptor and donor sites.

Activating MET alterations are inhibited by crizotinib, a treatment for ALK- and ROS1-rearranged lung cancers. Despite the reported association of the MET gene amplification and high protein expression as a poor prognostic marker^107,125  and recent reports that patients with MET amplification or MET exon 14 mutation are sensitive to crizotinib in some cases, there is as yet no approved targeted therapy to treat patients whose tumor harbor these MET genomic aberrations.^126–130  In this context, a routine stand-alone testing for these MET genomic aberrations or HGFR protein level is not indicated outside a clinical trial. Nevertheless, when multiplex testing for putative oncogenic driver mutations is applied to lung cancer patients, either initially when testing for EGFR/ALK/ROS1 or after they are found to be negative, these MET gene aberrations should be included in the test panels.

To date, more than 100 somatic splice site alterations resulting in MET exon 14 skipping have been described. Mutations exhibit a highly diverse sequence composition, encompassing small insertions, deletions, complex indels, and single-nucleotide variants, which are primarily located in splice donor and acceptor sites. Point mutations deeper in the introns, up to 25 base pairs into the intronic noncoding regions, adjacent to the splice acceptor sites, have also been reported at a lower frequency, although many assays do not interrogate this region. In general, the overall incidence and the effect of these less common mutations in exon splicing have not been defined.

Given the wide variability and complexity of mutations affecting MET exon 14, comprehensive diagnostic testing could prove challenging depending on the method used. Targeted NGS-based assays interrogating MET as part of a wider gene panel are preferred for screening purposes. For DNA-based testing, assay design should be such as to allow accurate and full sequencing of exon 14 and its flanking introns. Novel mutations, particularly those alterations affecting regions adjacent to splice sites but deeper into the introns, may require confirmation of exon 14 skipping using an RNA-based assay. Alternatively, up-front RNA-based testing interrogating MET as part of a wider gene panel designed for comprehensive assessment of structural variants or gene expression may also be used.

Fluorescence in situ hybridization has been traditionally used for detection of gene amplification in clinical practice. Currently, there is no guideline for cutoff of MET positivity in lung cancer specimens. MET amplification has been classified by using MET:CEP7 ratio as low (≥1.8 to ≤2.2), intermediate (>2.2 to <5), and high (≥5).¹³¹  Other examples of MET FISH-positive criteria include 5 or more MET signals per cell¹³²  and a MET:CEP7 ratio of 2 or higher (PathVysion, Abbott Park, Illinois). Low and intermediate levels of MET amplification can occur synchronously with other oncogenic mutations and gene rearrangements (KRAS, EGFR, BRAF, ERBB2 [HER2], ALK, ROS1, RET) in up to 63% of lung carcinomas.¹³³  However, this overlap has not been observed in high–MET-amplification tumors (MET:CEP7 ratio ≥5), suggesting that MET amplification is probably a true oncogenic driver.¹³³  This group of tumors, with high-level amplification, showed a response to crizotinib, whereas no response was seen in tumors with low- or intermediate-level amplification. Importantly, about 20% of lung adenocarcinomas with MET exon 14–skipping mutations have concurrent high-level MET amplification, confounding the interpretation of each.^125,126,129  Regarding the significance of amplification alone, case reports have shown response to crizotinib.^134,135 

The same challenges defining a clinically valid cutoff of MET amplification positivity exist in the setting of acquired resistance to EGFR TKI. A recent phase II study showed 40% response rate in patients with acquired EGFR TKI resistance and a MET copy number of 5 or higher when treated with a combination of gefitinib and capmatinib; no response was observed in a group with a MET copy number less than 5.¹³⁶ 

Immunohistochemistry for MET protein expression performed on formalin-fixed, paraffin-embedded tissue samples has been the most frequently used method in lung cancer specimens. A number of commercially available monoclonal and polyclonal antibodies are directed against various epitopes of MET, with different sensitivities and specificities for both total and phosphorylated MET. Immunohistochemistry procedures and scoring methods for MET assessment have not been standardized. As a result, MET protein overexpression in unselected NSCLC cases has been reported to range from 20% to 70%.^137,138  An MA has found that MET expression by IHC in NSCLC is a negative prognostic factor in patients with surgically resected NSCLC.¹⁰⁷  A frequently used commercially available antibody, particularly in clinical trials, is the CONFIRM anti-total MET (SP44) rabbit monoclonal primary antibody (Ventana Medical Systems, Tucson, Arizona) directed against a membranous and cytoplasmic epitope of MET.¹⁰⁹  At the time of publication, it remains unclear whether total MET or phosphoMET protein overexpression represents a reliable indicator of MET activation. Both MET IHC and FISH are not predictive of efficacy of onartuzumab combined with erlotinib in advanced NSCLC patients.¹¹⁶ 

The spectrum of recurring alterations in lung cancer continues to evolve, and several promising alterations have been reported that were not included in this recommendation. This includes fusions involving genes in the NTRK and FGFR families, both of which have experimental targeted inhibitors with supporting in vitro data and case reports. No guideline can be completely up to date, and practitioners of lung cancer care are advised to keep abreast of these and other developments. Table 5 includes a list of emerging biomarkers for molecular testing in lung cancer.

Question 2: What Methods Should Be Used to Perform Molecular Testing?

Immunohistochemistry is an equivalent alternative to FISH for ALK testing.

The strength of evidence supporting the use of IHC for ALK testing was adequate. This recommendation is evidence-based and was supported by 20 studies,^{61,111,139–156}  comprising 6 PCSs,^{61,139,141–143,154}  3 PRCSs,^140,146,155  and 11 RCSs.^* Of the 20 studies, 19 used FISH as the reference standard when assessing the diagnostic potential of IHC.^{111,145,147–153,156}  The remaining study used IHC as the reference standard and FISH as the index test.¹⁴⁴  Using reported true-positive, false-positive, true-negative, and false-negative data from 14 studies using FISH as the reference standard, an MA was conducted to determine a pooled estimate of sensitivity and specificity for ALK IHC (Figure 2). All included studies were assessed for quality and none were found to have methodologic flaws that would raise concerns about the studies' findings (SDC Table 20). Refer to SDC Table 21 for a summary of findings from studies supporting the use of the IHC assay for ALK testing.

At the time of the original guideline, the only assay that had evidence of clinical utility from prospective studies to select patients for crizotinib therapy was an ALK FISH break-apart assay, which is interpreted as positive if at least 15% of tumor cells show signals separated by at least 2 probe diameters or a single 3′ signal (deleted 5′). The ALK FISH assay can be technically challenging, particularly with tumors showing positive signals near the cutoff of 15%. FISH assays in general are limited by high cost, need for specialized personnel for interpretation, and limited availability of equipment, space, and testing personnel. However, ALK IHC can also show variation in staining among antibodies, protocols, and interpretation.

For all of these reasons, different assay approaches have been proposed for identification of ALK rearrangements in lung carcinoma. Many studies have focused on IHC as a widely available and cost-effective screening assay. In 2013, IHC had been shown comparable with FISH in some studies, but significant variations among antibodies and methods, and limited market availability of some of the more accurate antibodies, precluded making an evidence-based recommendation at that time.

By 2016, however, numerous publications^{61,111,139–156}  had established the technical performance of several ALK IHC assays and their correlation with ALK FISH results. An important early observation was that the amount of ALK fusion protein expression in the ALK rearranged non–small cell carcinomas tends to be lower than is found in anaplastic large cell lymphoma, from which the gene gets its name and for which the first IHC antibodies were developed.¹⁵⁷  The ALK1 antibody (mouse monoclonal anti-human CD246, clone ALK1) typically used to diagnose anaplastic large cell lymphoma failed to identify a significant number of the ALK rearranged NSCLCs using standard techniques.^{140,158–161}  To overcome this problem, several technical steps have been applied, such as tyramide amplification and enhanced polymer-based detection systems. Despite these advances, although the ALK1 antibody has good specificity (91%–99%), sensitivity is still poor, ranging from 67% to 100%, and therefore the ALK1 antibody is not recommended for ALK rearrangement screening in lung carcinoma.

Subsequently, 2 commercially available clones, mouse monoclonal 5A4 (Novocastra, Leica Biosystems, Buffalo Grove, Illinois)– and rabbit monoclonal D5F3 (Ventana)–based assays, showed clinically acceptable sensitivities and specificities, ranging from 95% to 100%, when compared with ALK FISH results.^{61,111,139,141–156}  Studies also showed that IHC-positive ALK protein expression correlates with tumor response to ALK inhibitors even in ALK FISH–negative cases.¹⁶²  In the United States, an assay using the D5F3 antibody (Ventana) is now approved by the FDA for selection of lung cancer patients to receive treatment with crizotinib.

Based on published evidence with 5A4 and D5F3 monoclonal antibodies, properly validated IHC assays are an equivalent alternative to ALK FISH. A meta-analysis that pooled 14 studies using FISH as a reference standard determined a pooled sensitivity of 97% (95% CI, 93%–98%) and a pooled specificity of 99% (95% CI, 97%–100%) for ALK IHC for both the 5A4 or D5F3 assays (Figure 2). The laboratory may choose which antibody to use based on analytic precision, clinical sensitivity, and clinical specificity in accordance with published standards. ALK IHC is an acceptable alternative to FISH and treatment decisions can be made when IHC results are clearly positive, as manifested by strong granular cytoplasmic staining with/without membrane accentuation, or negative; however, weak staining can be challenging to interpret, and the specificity of weak staining relative to FISH should be determined in each laboratory during validation. Occasional cases may be difficult to interpret because of heterogeneous fixation/preservation and/or nonspecific staining artifacts, such as light cytoplasmic staining in alveolar macrophages, neural cells, extracellular mucin, necrosis, and glandular epithelium. In these settings, these cases should also be tested by a validated method (eg, ALK FISH, RT-PCR, NGS).^141,146 

Discordant results between ALK FISH and IHC assays have been described in rare cases. ALK IHC–negative cases were reported in association with ALK FISH–positive assays that showed a lower percentage of tumor cells with rearrangement (15%–20%). Technical errors cannot be reliably excluded in a case with a lower percentage of nuclei positive for rearrangement. Recent studies suggest that a 5′ deletion FISH pattern may more commonly represent a false-positive result with discrepant IHC results than cases with a split FISH signal.^163–165  Importantly, however, clinical outcomes of patients with discrepant FISH and IHC results have not shown a consistent pattern of superiority of one method over the other.^156,166 

Although at the time of writing RT-PCR and NGS are not approved by the FDA in the United States as first-line methods for determining ALK status in selection of patients for ALK inhibitor therapy, these approaches have shown comparable performance with IHC^163–165  when designed to detect the majority of fusions, and are standard practice in many other countries.^{163–165,167}  These methods are highly specific for most fusions,^97,168,169  and patients with positive results should be treated with an ALK inhibitor, although patients with negative results may benefit from a more sensitive method to exclude the possibility of a variant fusion. Similarly, amplicon-based NGS assays of DNA may likewise fail to detect all fusion variants, and therefore a capture-based DNA or RNA approach is preferred for NGS testing for ALK fusions. Current data are still too limited to develop a specific recommendation either for or against the use of NGS for ALK fusions as a sole determinant of ALK-TKI therapy.

Multiplexed genetic sequencing panels are preferred over multiple single-gene tests to identify other treatment options beyond EGFR, ALK, and ROS1.

The strength of evidence is inadequate supporting the use of multiplexed genetic sequencing panels compared with single-gene tests. The statement is evidence-based and supported by 5 studies,^169–173  comprising 1 PCS,¹⁷²  2 PRCSs,^171,173  and 2 RCSs.^169,170  All included studies were assessed for quality and none were found to have methodologic flaws that would raise concerns about the studies' findings (SDC Table 22). Refer to SDC Table 23 for a summary of findings from studies supporting the use of multiplexed genetic sequencing panels.

The rapid recent emergence of so-called NGS, or massively parallel sequencing, has changed the practice of molecular diagnostics considerably, in lung cancer and in other contexts. This technology involves spatial separation of individual DNA molecules from a sample, PCR amplification (either with or without a preceding hybridization capture step) of predetermined regions of the genome, and parallel sequencing by synthesis of each of the amplified single DNA molecules on a massive scale, followed by computational data processing to recombine and identify the sequences and provide a digital display of each sample's genomic features. This technology enables sensitive and specific assessment of multiple genomic regions at once, up to and including the entire genome. The resources necessary for clinical NGS implementation are substantial and the assays are complex in design, performance, and interpretation. Consequently, this technology is not universally deployed. Nonetheless, multiple academic institutions and private companies have used this technology at scale, providing a relatively DNA- and cost-efficient method for assessment of cancer gene mutations as compared with performing multiple single-gene analyses. In the first guideline, this was a new and unproven technology, but it is now well established. In early 2017, professional practice guidelines were published regarding the validation¹⁷⁴  and interpretation/reporting¹⁷  of NGS assays for cancer.

NGS enables the simultaneous assessment of all 3 of the “must-test” genes in lung cancer—EGFR, ALK, ROS1—as well as each of the genes suggested for inclusion in larger panels—BRAF, RET, ERBB2 (HER2), KRAS, MET—and hundreds to thousands of other genes that may have potential roles in cancer development. In addition to small mutations, NGS assays are able to detect fusions/rearrangements and copy number changes in the targeted genes, if designed with these alterations in mind.

Numerous studies^169–173  have demonstrated the excellent sensitivity of NGS methods relative to single-gene targeted assays, particularly for single-nucleotide–substitution mutations. Next-generation sequencing methods typically require less input DNA and can accommodate smaller samples with lower concentrations of malignant cells, and, although typically slower than 1 single-gene assay, can often be performed more rapidly than sequential multiple single-gene assays.^175,176  A reduced need for repeat biopsy is an additional benefit of panel testing.

Two basic designs of NGS assays are used in molecular oncology: amplicon based and hybrid capture based. Both of these designs generate a “library” of amplified DNA that is sequenced as single molecules, using one of several sequencing platforms. Amplicon sequencing uses multiple PCR reactions to generate the library, which is generally easier, faster, and able to detect mutations at lower allele frequencies (ie, subclonal populations). However, it is more suitable for simpler assays of fewer genes, and typically used for analysis of oncogene hot spots or small, selected regions of selected genes of interest. In its basic form, it cannot reliably detect fusions or copy number variations. The greater analytic sensitivity of this method makes it suitable for very small or heterogeneous samples. Capture-based methods, by contrast, use hybridization to generate the library and are more complex and involve more steps, resulting in a longer turnaround time, but are better for larger set of genes or genomic regions to be analyzed; however, these methods typically are less sensitive in highly heterogeneous or small samples. In general, capture-based methods may be preferable for initial testing of lung cancer samples in order to detect rearrangements, such as in ALK and ROS1, as well as a broader range of potential genetic markers. For monitoring during secondary clinical resistance (ie, EGFR), where a more narrow range of possible mutations is needed and smaller or more heterogeneous samples may be desired, amplicon sequencing may be preferred; however, either method may be designed and deployed, successfully, for either application.

Laboratories should ensure test results that are unexpected, discordant, equivocal, or otherwise of low confidence are confirmed or resolved using an alternative method or sample.

The strength of evidence supporting the routine use of orthogonal methods to confirm results for any of the molecular markers is insufficient. However, our opinion is that good laboratory practice for somatic alterations is to perform confirmatory testing for results that are unusual, suboptimal, or inconsistent with other laboratory findings or clinical information.

All assays should be appropriately validated before being offered for clinical use. This typically includes an assessment of sensitivity, specificity, and reproducibility, in addition to other performance characteristics, as required by most laboratory-certifying authorities. The performance characteristics of sequencing-based assays can be readily determined for the more common alterations. The reliability of infrequently encountered mutations or specific categories of chromosomal alterations may be more challenging to document, and laboratories using such technologies should have procedures in place to verify any results that are unexpected, discordant with other results, equivocal, or of compromised confidence in order to provide an optimal result for patient care and to better understand any intrinsic assay limitations. Corroboration of such questionable results might be sought by evaluating a separate specimen from the same lesion, confirmatory testing in another laboratory, or evaluation using an orthogonal methodology.

Question 3: Is Molecular Testing Appropriate for Lung Cancers That Do Not Have an Adenocarcinoma Component?

Physicians may use molecular biomarker testing in tumors with histologies other than adenocarcinoma when clinical features indicate a higher probability of an oncogenic driver.

The strength of evidence supporting the use of molecular biomarker testing in lung cancers that lack an adenocarcinoma component is insufficient. This statement is based on expert consensus opinion.

Upon systematic review, no new evidence established the routine molecular testing of any genes for typical squamous cell carcinoma, small cell carcinoma, or other neuroendocrine lung tumors. Although small studies have reported rare EGFR mutations in squamous cell carcinoma biopsies, these may have represented partial sampling of adenosquamous cancers and have not been borne out in fully resected samples with confirmed squamous histology.¹⁷⁷  Evidence from controlled and well-powered studies, supporting the clinical utility of molecular testing of lung cancers for selection of targeted therapies, remains confined to nonsquamous non–small cell lung carcinomas, predominantly adenocarcinomas or mixed cancers with an adenocarcinoma component. However, strict reliance upon adenocarcinoma histology may occasionally exclude some patients who do not have a definitive diagnosis of adenocarcinoma (eg, non–small cell lung carcinoma, not otherwise specified), and might benefit from targeted therapy, particularly for small biopsies that partially sample a larger tumor. Although actionable mutations have been reported in biopsies with nonadenocarcinoma non–small cell types, the frequency of such findings is low enough that it is not recommended to test all small biopsy samples with nonadenocarcinoma histology. In this context, molecular testing is appropriate to perform in lung cancers with non–small cell histologies other than adenocarcinoma when clinical features are atypical and/or consistent with a higher likelihood of a targetable mutation.

Chief among these clinical factors that may indicate a higher probability of a targetable oncogenic driver in the setting of a nonadenocarcinoma histology are young age and absence of tobacco exposure. In nonadenocarcinoma non–small cell histologies, the finding of EGFR, ALK, or ROS1 alterations has been most commonly reported in situations in which patients had a minimal (1–10 packs per year) or no history of tobacco exposure.^{30,159–161,178–214}  Thus, across the spectrum of lung carcinomas, light or absent tobacco exposure should be sufficient rationale to prompt testing, regardless of sampling methodology or completeness of exclusion of adenocarcinoma component.

Similarly, some studies have suggested associations between the presence of ALK or ROS1 alterations and younger patient age.^{30,35,181,215,216}  Although other studies have indicated that these findings may reflect testing bias,²¹⁷  the documentation of an association between younger patient age and an actionable biomarker is another consideration in selecting patients for testing. The boundary between young and not young is not well defined, however, and a clear evidence-based cutoff for this guideline cannot be established. Systematic review from the original (2013) guideline¹  demonstrated that adenocarcinoma patients with EGFR mutations had a significantly lower mean age than patients without mutations (56 versus 63, P = .03), although the difference in mean ages for patients with and without ALK fusions was not significant (60 versus 66), nor is the difference in mean ages for patients with and without ROS1 fusions from this review (65 versus 62), and the difference in means does not completely capture the distribution of ages and, accordingly, the sensitivity and specificity of any given age cutoff. In the absence of published evidence, our opinion is that a reasonable strategy would be to test patients younger than 50 years with nonadenocarcinoma histology.

Of note, reflex testing algorithms initiated and/or managed by pathologists need to accommodate intricacies of clinical management, which can be challenging as sufficient clinical information is often not available for pathologists to incorporate into their evaluation. Establishing a program for reflex molecular testing of lung cancer samples should be an institutional decision, and should include an open dialogue between pathologists and oncology teams, in order to put in place an optimal strategy. Once those practices are established by the team, however, reflex testing initiated by the pathologist is reasonable.

Lastly, in the context of increasing use of panel-based/NGS-based testing, it may become unnecessary to identify specific analytes of interest in specific clinical situations, instead identifying clinical situations in which panel-based testing may be beneficial.

Question 4: What Testing Is Indicated for Patients With Targetable Mutations Who Have Relapsed on Targeted Therapy?

In lung adenocarcinoma patients who harbor sensitizing EGFR mutations and have progressed after treatment with an EGFR-targeted TKI, EGFR T790M mutational testing should be used to guide selection of treatment with third-generation EGFR inhibitors.

The strength of evidence was adequate to support the use of EGFR T790M mutation testing when selecting patients for third-generation EGFR-targeted therapy. This recommendation is evidence-based and supported by 5 studies,^218–222  including 1 MA,²²²  2 single-arm phase I nonrandomized clinical trials,^220,221  1 PCS,²¹⁹  and 1 RCS.²¹⁸  All studies were assessed for quality and none were found to have methodologic flaws that would raise concerns about the studies' findings (SDC Table 24). Refer to SDC Table 25 for a summary of findings from the studies supporting the use of EGFR T790M mutation testing when selecting patients for third-generation EGFR-targeted therapy.

Laboratories testing for EGFR T790M mutation in patients with secondary clinical resistance to EGFR-targeted kinase inhibitors should deploy assays capable of detecting EGFR T790M mutations in as little as 5% of EGFR alleles.

The strength of evidence supporting this recommendation is insufficient. This recommendation is based on the analytical sensitivity of the allele-specific real-time PCR assay used in clinical trials that established the utility of third-generation EGFR inhibitors directed against the T790M mutant protein.

The major mechanism of secondary clinical resistance to the first-generation EGFR TKIs erlotinib and gefitinib is the development of T790M mutation on the same EGFR allele that harbors the original sensitizing mutation, blocking inhibition of the mutant protein by these TKIs. The detection of the EGFR T790M mutation in this setting has become clinically necessary because of the development of third-generation EGFR TKIs, such as osimertinib, which are active in the presence of this mutation.^223,224  However, although rare responses have been reported to third-generation inhibitors in EGFR T790M–negative disease,²²⁵  such cases may harbor other resistance mechanisms such as MET or ERBB2 amplification^90,114,226  that may be more effectively targeted by other agents. Therefore, determining appropriate therapy in the setting of secondary clinical resistance to an EGFR inhibitor requires knowledge of the presence or absence of the T790M mutation.

Importantly, genetic mechanisms of secondary clinical resistance arise in a subclonal fashion because they confer resistance to the subpopulation of cells in which they are present, and this subclone expands gradually under the selective pressure of the EGFR TKI. Experimental studies have shown that the presence of an EGFR T790M mutation in small proportion of a bulk tumor cell population (as little as 5%), often undetectable by Sanger sequencing, can lead to increased growth in spite of treatment with EGFR TKI.^225,227  Detection is made even more challenging when biopsies contain a high proportion of nontumor cells. These considerations lead us to recommend that laboratories should have a high sensitivity assay available for the detection of the EGFR T790M mutation in posttreatment biopsies from patients who demonstrate progression or relapse after an initial response to EGFR TKI.

The clinical trials^{221,228–230}  that established the clinical utility of T790M testing in predicting response to osimertinib used a commercial allele-specific real-time PCR assay with a reported lower limit of detection of 5% mutant allele fraction.²³¹  Several studies^232,233  have shown comparable, if not superior, analytical sensitivity with droplet digital PCR, and NGS methods^234,235  can also provide this level of sensitivity, given the appropriate assay design. It is important, however, to recognize that unmodified Sanger sequencing, which was an acceptable method in the original 2013 guideline,¹  does not provide adequate sensitivity for this application. Regardless of the method chosen, careful validation must be performed to establish appropriate sensitivity.

Finally, it should be noted that trials are currently underway to assess the value of these third-generation EGFR inhibitors as first-line treatment of EGFR-mutant lung adenocarcinoma.^220,221,236  Data are emerging²³⁷  that a second acquired resistance mutation, C797S, can arise in tumors that have progressed after osimertinib treatment for T790M disease, but these cases are so far rare and this mutation is poorly studied and not currently treatable, so testing for C797S is not recommended for routine management at this time.

There is currently insufficient evidence to support a recommendation for or against routine testing for ALK mutational status for lung adenocarcinoma patients with sensitizing ALK mutations who have progressed after treatment with an ALK-targeted TKI.

The evidence was insufficient to inform a recommendation on the association between pretreatment or rebiopsy discovery of ALK secondary resistance and clinical outcomes.

Several groups have reported a diverse set of secondary mutations in ALK that confer acquired resistance to crizotinib (eg, L1152R, C1156Y, F1174L, L1196M, L1198P, D1203N, and G1269A).²³⁸  For second-line ALK inhibitors, other acquired mutations have been reported (G1202R, G1202del, V1180L, S1206Y, E1201K) as well.

However, although some studies²³⁸  have suggested that different secondary ALK mutations may show sensitivity or resistance to different ALK inhibitors, these data are still limited and insufficient to guide selection of treatment in the setting of acquired resistance. Moreover, second-generation ALK inhibitors also show activity in NSCLC without ALK resistance mutations, suggesting that a significant proportion of ALK-rearranged lung carcinomas become resistant to crizotinib because of inadequate suppression of ALK. Accordingly, current practice is to administer one of several second-generation ALK inhibitors (ceritinib, brigatinib, lorlatinib, and alectinib) that have received FDA approval for the treatment of crizotinib-refractory, ALK-rearranged NSCLC, without testing for secondary ALK mutations. As more patients experience resistance and receive second-generation inhibitors, we anticipate maturation of data to strengthen the association between secondary mutation and sensitivity/resistance to different inhibitors.

For now, however, we believe there is insufficient clinical utility to warrant routine testing for secondary ALK mutations in patients who have relapsed after initial response to an ALK inhibitor.

Question 5: What Is the Role of Testing for Circulating cfDNA for Lung Cancer Patients?—Numerous recent studies^99,100,239  have demonstrated that lung cancer cells shed their DNA into the circulation at levels that are detectable with several modern technologies, such as droplet digital PCR, allele-specific PCR, and NGS. This event enables testing of plasma cfDNA obtained from peripheral blood samples, at least in some instances, as an alternative to a biopsy sample, to identify mutations occurring in lung cancer both at diagnosis and during the course of disease.

A theoretical advantage of these assays is the derivation of circulating tumor DNA from multiple disease sites, and therefore it may represent an integrative measure of all sites of disease. Although not formally proven, this potential advantage of cfDNA is particularly important in the setting of secondary clinical resistance,²³⁶  to enable broad sampling of different tumor subclones.^232,240,241 

Analytical methods for cfDNA have high analytical specificity, with very low (<5%–20%) false-positive rates,^{232,234,235,242,243}  such that demonstration of a mutation, in the proper clinical context, can be used to guide treatment with a targeted inhibitor. However, sensitivity of cfDNA analysis is lower (60%–70%),^{232,234,235,242,243}  such that the absence of mutation finding does not exclude the possibility of a mutation.

It is also important to understand that, despite the promise afforded by this technology, much is still unknown about the dynamics of release of DNA from cancer cells. Factors that increase or decrease the release of DNA from cells, and its half-life in circulation and mechanisms of elimination, are poorly understood.^244–246  There is, however, an overall correlation between burden of disease (both volume and number of metastatic sites) and prevalence of mutations in cfDNA.

Finally, other methods of analysis can be applied to blood samples. Circulating tumor cells can be isolated from the blood, as can exosomes bearing DNA that have been released by cancer cells. Analysis of these latter 2 samples is more challenging technically, and has not been sufficiently studied in lung cancer to warrant consideration in this guideline. Similarly, data are emerging regarding the analysis of cfDNA in other body fluids, particularly urine, but are similarly insufficient to warrant a recommendation at this time. Most of the data to date, and the subject of the comments that follow, apply to plasma cfDNA.

There is currently insufficient evidence to support the use of circulating plasma cfDNA molecular methods for establishing a primary diagnosis of lung adenocarcinoma.

The evidence was insufficient to inform a recommendation on the use of cfDNA for diagnosis of primary lung adenocarcinoma.

Theoretically, because sensitizing mutations in EGFR are characteristic and specific alterations in lung cancers, one may question whether the combination of a cfDNA result showing such a mutation in an appropriate clinical context, with radiographic evidence of a lung lesion, could enable a diagnosis of EGFR-mutant lung cancer without requiring an anatomic pathology diagnosis. However, no studies in the medical literature have rigorously evaluated this approach in a prospective manner.

In some clinical settings in which tissue is limited and/or insufficient for molecular testing, physicians may use a cfDNA assay to identify EGFR mutations.

The strength of evidence supporting the use of cfDNA to determine EGFR mutation status in situations where tissue is limited or insufficient is adequate. This recommendation is evidence-based and supported by 6 studies,^† comprising 2 MAs,^243,247  2 PCSs,^235,242  and 2 PRCSs.^232,234  The identified studies used various EGFR detection methods, but all verified the results from cfDNA with results from tumor tissue. Using reported true-positive, false-positive, true-negative, and false-negative data from 4 studies, a meta-analysis was conducted to determine a pooled estimate of sensitivity and specificity for cfDNA detection of EGFR mutation (Figure 3). All included studies were assessed for quality and none were found to have methodologic flaws that would raise concerns about the studies' findings (SDC Table 26). Refer to SDC Table 27 for a summary of findings from studies supporting the use of cfDNA to determine EGFR mutational status in situations where tissue is limited or insufficient.

Analysis of cfDNA for EGFR mutations has intermediate sensitivity (66.4%; 95% CI, 62.7%–69.9%) and high specificity (95.6%; 95% CI, 83.3%–99.0%) in lung adenocarcinoma (Figure 3). In some clinical settings in which tissue biopsy material is unavailable or insufficient and tissue rebiopsy is not feasible, so that a tissue-based EGFR analysis cannot be performed, then a cfDNA assay for an activating EGFR mutation may be conducted as an alternative molecular diagnostic procedure. Because the sensitivity of this assay is less than 80% in all reports, it should be recognized that not all lung adenocarcinoma patients with EGFR mutation–positive disease will have the mutation detected in their cfDNA, so that a negative result by cfDNA analysis is not reliable evidence that there is not an EGFR mutation in a given patient's cancer. In this context, physicians should renew efforts to obtain an adequate tissue sample for analysis.

Physicians may use plasma cfDNA methods to identify EGFR T790M mutations in lung adenocarcinoma patients with progression or secondary clinical resistance to EGFR-targeted TKIs; testing of the tumor sample is recommended if the plasma result is negative.

The strength of evidence supporting the use of cfDNA methods to identify EGFR T790M mutation is inadequate. This statement is evidence-based and supported by 4 studies,^{233,236,248,249}  comprising 2 PCSs^233,236  and 2 RCSs.^248,249  All included studies were assessed for quality and none were found to have methodologic flaws that would raise concerns about the studies' findings (SDC Table 28). Refer to SDC Table 29 for a summary of findings from studies supporting the use of cfDNA methods to identify EGFR T790M mutation.

Molecular testing for EGFR T790M mutation should be performed in lung adenocarcinoma patients with sensitizing EGFR mutations whose disease progresses on, or shows secondary clinical resistance to, EGFR TKI. Such testing is particularly appropriate because third-generation EGFR inhibitors (eg, osimertinib) are proven to have significant benefit for T790M mutant cancers²²¹  and have been approved for this indication by health authorities around the world. Analysis of cfDNA or analysis of a new tissue biopsy is appropriate for EGFR T790M detection. Cell-free DNA may be preferred for patients unwilling or unable to undergo a biopsy at the time of progression; moreover, as progression may represent subclonal processes, cfDNA testing may represent a more global sampling of disease as compared with a tissue biopsy. However, cfDNA analysis for EGFR T790M has intermediate sensitivity (0.40–0.78) and high specificity.^{233,236,248,249}  Hence, a negative result does not rule out the possibility that EGFR T790M mutation is the mechanism of resistance to TKI therapy, and a new tissue biopsy from a site of progressive disease should be considered if the cfDNA result is wild type.

Note that acquired T790M mutations are often subclonal, and a cfDNA sample could show the original sensitizing EGFR mutation (eg, L858R) and still have a false-negative result for T790M, requiring a tissue or cytology sample for confirmation; implicit in this is the benefit of including the original sensitizing mutation (eg, exon 19 deletion, L858R) in the assay to confirm that tumor DNA is being shed into the circulation, although this may not be practical for less common sensitizing mutations. On the other hand, because of the high specificity of cfDNA, a positive T790M finding in cfDNA is equivalent to a tissue biopsy finding of T790M,²³⁶  and can be used to guide therapy with osimertinib.

There is currently insufficient evidence to support the use of circulating tumor cell molecular analysis for the diagnosis of primary lung adenocarcinoma, the identification of EGFR or other mutations, or the identification of EGFR T790M mutations at the time of EGFR TKI resistance.

The evidence was insufficient to inform a recommendation on the use of circulating tumor cells for diagnosis of primary lung adenocarcinoma.

Tissue should be preserved to enable testing for immunomodulatory therapies.

Subject of upcoming guideline.

Since publication of the last guideline, immunotherapy in lung cancer has rapidly evolved to become a part of standard of care for many patients with advanced NSCLC. Recent studies^250–254  have shown significant benefits in a subset of advanced lung cancer patients when treated with these agents. Government regulatory agencies have approved immunomodulatory therapies as second-line agents for advanced lung cancer patients,^255–257  as well as first-line therapy for patients with high level of PD-L1 expression and absence of sensitizing EGFR mutations or ALK rearrangements. For some of these agents, selection of patients with a companion diagnostic is required,²⁵⁸  for others a complementary diagnostic is recommended,^259–261  and for some no biomarker selection is indicated. Putative biomarkers to predict response to these agents, and the methods used to assess them, are varied and not yet standardized.

The principle of the immunomodulatory therapies is their ability to disrupt inhibitory signaling between tumor cells and immune cells (typically T cells), which occurs when tumor cells express proteins that induce immunologic tolerance and prevent the immune system from attacking the tumor. Normally this mechanism is used to control the immune response and prevent autoimmune disease. Several such inhibitory signaling processes exist, although the greatest progress in clinical therapy in lung cancer involves the interaction between PD-L1 on tumor cells and programmed death receptor-1 (PD-1) on T cells. This interaction effectively silences the T-cell response to a tumor. By blocking PD-1 with so-called immune checkpoint inhibitors, T cells become enabled to recognize and respond to foreign antigens presented on the cancer cells.

Because most lung cancer cells contain many mutations beyond their oncogenic drivers, they typically express a large number of mutant proteins, some of which can be displayed on the cell surface by human leukocyte antigen molecules as “foreign” neoantigens. The more mutations in a cell, the more neoantigens are probably expressed, and the more likely the immune system is to destroy the cells, provided that the tolerance mechanisms, such as PD-L1/PD-1, are not activated. Expression of PD-L1 by tumor cells (or local macrophages), expression of PD-1 by tumor-infiltrating lymphocytes, number of mutations and neoantigens, and evidence of an immune infiltrate (“inflamed tumor”) are all candidates to predict response to these treatments. Other inhibitory signaling pathways may also be involved, such as the interaction between cytotoxic T-lymphocyte–associated protein 4 (CTLA-4) and CD80/86, which is targeted by the immunomodulatory agent ipilimumab, typically given to patients with melanoma.

The results of treatment with these agents have been impressive,^250–254  with some patients experiencing durable responses that have lasted years. However, unlike the targeted therapies, the frequency with which patients respond, even in biomarker-selected populations, is closer to 20% to 30% in the second line and 50% in the first line (as opposed to 80% for targeted therapies),^252–254  although radiographic response may not be the best indicator of effectiveness of an immunomodulatory therapy because of the impact of active inflammation on the size of a tumor as assessed by standard imaging. In this regard, better biomarkers are needed, as the companion diagnostic biomarkers do not fully address the clinical need to determine, with sensitivity and specificity, who should receive these drugs. Complicating this further are the multitude of different biomarkers, with different methodologies developed in parallel by competing companies for competing drugs, such that no clear message emerges, with evidentiary basis, regarding which biomarkers should be tested or how to test for them. Studies are currently underway to compare a variety of candidate markers and to attempt to harmonize different assays targeting each biomarker (eg, different IHC methods for PD-L1).

Because of the lack of firm evidence supporting specific methodology or agents, we cannot make evidence-based recommendations regarding testing for these drugs in this guideline. A subsequent practice guideline is being planned to focus specifically on evidence-based assessment of methods for selecting patients to receive immunomodulatory therapies.

Despite the exclusion of this question from our systematic literature review, it is our opinion that samples should be preserved for assessment of biomarkers that predict response to immunomodulatory therapies, in accordance with the labeling requirements of the drugs under consideration. Importantly, often the ideal section of a tumor for this application is different from the ideal section of tumor required for molecular testing. For molecular testing, especially sequencing, an ideal sample of cancer has nearly “pure cancer,” with little intervening or adjacent stroma or inflammation. Assessment of biomarkers for immunomodulatory therapies, theoretically, should be performed on a section of tumor containing intervening and adjacent stroma, particularly if rich in infiltrating T cells; however, there are no guideline recommendations and this has not been prospectively studied. Response to some drugs²⁵⁰  has been shown to associate with the nature of the inflammatory cells within a tumor more than with the tumor cells themselves. This distinction, with the operational implications of needing, potentially, to identify and recut 2 sections from each cancer (1 for molecular testing, 1 for the IHC assessment of immune-regulatory molecules) is essential for surgical pathologists and histology laboratories to understand.

For most applications, the necessary analysis is IHC for PD-L1, although the exact antibody and staining protocol used and the interpretive criteria vary for different treatments. Other candidate IHC biomarkers may be required, however, as well as characterization of reactive cell populations within a tumor. The potential validity and utility of mutational burden calculations (mutations/base pair of total genomic sequence) as assessed by NGS panels are being explored as an investigational biomarker, as are neoantigen prediction algorithms derived from whole-exome sequencing data.

The importance of molecular diagnostics for the care of lung cancer patients continues to develop at a rapid pace. Many of the guideline recommendations for clinical practice that were published in 2013 were subject for reconsideration by 2015. Similarly, this guideline will likely need revision within a relatively short period of time. Relying upon published prospective studies to serve as a basis for practice recommendations will always lag behind the latest discoveries and advancing edge of care as presented in meetings of professional societies. Practitioners today, and those of us making recommendations, are faced with the daunting challenge of balancing precision oncology, the notion that each patient has a unique combination of factors that should be incorporated into determining individual treatment plans, with evidence-based medicine, the notion that appropriate treatment decisions should be based upon large interventional studies of otherwise identical patients. Compounding this is the increasingly small sizes of the populations being defined by large-scale genomic analyses today, which makes designing effective large controlled interventional studies exceptionally difficult. When a dramatic clinical response is seen in 0.5% of patients with a condition—even a common condition such as lung cancer—how do we recruit enough patients to study to prove that the dramatic response is a general truth that should change practice, and how do we decide that everyone in the world should get the test that can determine if they are in that 0.5%?

We have updated the 2013 recommendations to recognize the changes that have passed this threshold—the importance of ROS1 testing, the value of IHC for ALK, and the importance of testing for T790M mutations in patients who progress on anti-EGFR therapy. In addition, we have laid out the emerging and promising molecular alterations that are “1 step lower”—alterations in BRAF, MET, ERBB2 (HER2), and RET—which we anticipate will pass this bar in a short time, and which we believe should be included in the expanded analyses that are possible because of the emergence of NGS technology. We also see the promise and results obtained with immunomodulatory therapies and await the systematic review that will be conducted to identify and recommend best practices to select patients for these therapies. We look forward to the continuing evolution in diagnostics and care for lung cancer patients as technology, scientific understanding, and clinical practice evolve.

The authors gratefully acknowledge the following: advisory panel members Timothy Craig Allen, MD, JD, Julia A. Bridge, MD, Lukas Bubendorf, MD, Paul Bunn, MD, Federico Cappuzzo, MD, Edmund S. Cibas, MD, Dhananjay A. Chitale, MD, Marileila Varella Garcia, PhD, A. John Iafrate, MD, PhD, Pasi A. Janne, MD, PhD, Mark G. Kris, MD, Antonio Marchetti, MD, PhD, Tetsuya Mitsudomi, MD, PhD, Tony Mok, MD, Marina Nikiforova, MD, Suresh Ramalingam, MD, Natasha Rekhtman, MD, PhD, Charles Powell, MD, Juan-Sebastian Saldivar, MD, Yi-Long Wu, MD, David Yankelevitz, MD, and patient advocates Bonnie J. Addario and Kim Norris, BA, for their review of the key questions, recommendations, and draft manuscript; and Lisa A. Fatheree, SCT(ASCP), Pia Hirsch, Mrudula Pullambhatla, MS, and Mary Steele Williams, MNA, MT(ASCP)SM, CAE, for their support throughout the guideline development process.