Immune checkpoint inhibitor treatment in patients with oncogene-addicted non-small cell lung cancer (NSCLC): summary of a multidisciplinary round-table discussion

Abstract

The introduction of targeted treatments and more recently immune checkpoint inhibitors (ICI) to the treatment of metastatic non-small cell lung cancer (NSCLC) has dramatically changed the prognosis of selected patients. For patients with oncogene-addicted metastatic NSCLC harbouring an epidermal growth factor receptor (EGFR) or v-Raf murine sarcoma viral oncogene homologue B1 (BRAF) mutation or an anaplastic lymphoma kinase (ALK) or ROS proto-oncogene 1, receptor tyrosine kinase (ROS1) gene alteration (translocation, fusion, amplification) mutation-specific tyrosine kinase inhibitors (TKI) are already first-line standard treatment, while targeted treatment for other driver mutations affecting MET, RET, human epidermal growth factor receptor (HER) 2, tropomyosin receptor kinases (TRK) 1–3 and others are currently under investigation. The role of ICI in these patient subgroups is currently under debate. This article summarises a round-table discussion organised by ESMO Open in Vienna in July 2018. It reviews current clinical data on ICI treatment in patients with metastatic oncogene-addicted NSCLC and discusses molecular diagnostic assessment, potential biomarkers and radiological methods for response evaluation of ICI treatment. The round-table panel concluded ICI should only be considered in patients with oncogene-addicted NSCLC after exhaustion of effective targeted therapies and in some cases possibly after all other therapies including chemotherapies. More clinical trials on combination therapies and biomarkers for ICI therapy based on the specific differing characteristics of oncogene-addicted NSCLC need to be conducted.

Video abstract.

Download video file ^{(197.4MB, mp4)}

DOI: 10.1136/esmoopen-2019-000498.video01

Introduction

The management of patients with metastatic non-small cell lung cancer (NSCLC) underwent significant transformation in the last 10–15 years by the development of precision medicine based on molecular characterisation. Molecular analysis revealed distinct targetable driver mutations in about 10%–20% of patients with metastatic NSCLC.1 The most frequently observed targetable mutations are aberrations in the epidermal growth factor receptor (EGFR) gene (about 10%–15% in Caucasians), followed by gene rearrangements/gene fusions in the anaplastic lymphoma kinase (ALK) gene (about 5%) and ROS proto-oncogene 1, receptor tyrosine kinase (ROS1) (about 1%–3%). In addition, mutations in v-Raf murine sarcoma viral oncogene homologue B1 (BRAF) (about 4%) can be observed and targeted therapeutically with mutation-specific tyrosine kinase inhibitors (TKI).1 Besides these targetable genetic alterations, also the expression of programmed death-ligand 1 (PD-L1), an immune suppressive molecule, needs to be considered for therapeutic decision-making. The programmed cell death protein 1 (PD-1) immune checkpoint inhibitor (ICI) pembrolizumab was shown to have higher efficacy as first-line treatment compared with platin-based chemotherapy in patients without the presence of a driver oncogene alteration but PD-L1 expression in more than 50% of tumour cells2 and to increase efficacy of platin-based chemotherapy in patients with a lower PD-L1 expression level.3

Therefore, NSCLC is a molecularly heterogenous disease and initial molecular diagnosis forms the basis for systemic treatment decisions, given the clinical superiority of TKI over chemotherapy in patients harbouring a predictive molecular alteration. However, resistance to targeted therapies and progression occurs in almost all patients and especially brain metastases can be observed frequently as an area of progression.4 Treatment options on exhaustion of targeted therapies and chemotherapy are few, underscoring the need to explore the new and promising treatment category of ICI, also in patients with oncogene-addicted metastatic NSCLC.

ICI target the inhibitory T cell co-receptors and thereby increase the capacity of the tumour-specific immune response. This new category of immune-modulating therapies has revolutionised oncology as in comparison to targeted therapies and chemotherapy long lasting and durable responses can be achieved in a subfraction of patients.5 Here, PD-L1 inhibitors, PD-1 inhibitors and cytotoxic T-lymphocyte-associated protein 4 inhibitors have been investigated in patients with metastatic NSCLC.6

In the context of NSCLC, patients with oncogene addiction were frequently excluded from registration trials, resulting in so far limited clinical knowledge on the efficacy of ICI in the subcohort of molecularly altered NSCLC.7–14 This article aims to review the available clinical data on ICI treatment in patients with metastatic oncogene-addicted NSCLC and discusses molecular diagnostic assessment, potential biomarkers and radiological methods for response evaluation of ICI treatment.

Molecular diagnostic assessment in NSCLC

Personalised cancer therapy comes with increased and complex diagnostic testing. In clinical practice and as recommended by clinical guidelines,15 selection of targeted therapies for NSCLC requires testing for EGFR and BRAF mutations, rearrangements or fusion protein expression involving the ALK and ROS1 genes and the expression of PD-L1. Therefore, molecular testing should be carried out in all patients who have a definite, probable or possible diagnosis of adenocarcinoma, for whom this diagnosis cannot be reasonably excluded, and for patients with non-small cell carcinoma or for patients with squamous cell carcinoma who have a high risk of a target mutation or rearrangement (never or light smokers, very long-term ex-smokers or young women).2 16

Given the high amount of analysis to be made on often sparse tumour material, strong recommendations on tissue preservation for biomarker studies have been outlined by several guidelines.2 It is critical that pathology laboratories develop policies for integrating biomarker testing into their routine tissue-processing workflows to minimise the number of ancillary stains performed for the diagnosis and classification. The time point of molecular testing, right after pathology diagnosis as indicated by the pathologist (reflex testing) or only after additional claim by the treating clinician (bespoke testing), is currently a topic of debate and organised differently throughout centres.17 Molecular testing initiated by the pathologist immediately after diagnosis of cancer (reflex testing) provides results in 5–10 working days, in contrast to bespoke testing requested by the oncologist or the multidisciplinary team only when the test is needed. Reflex testing has the advantages of a quicker molecular profiling for clinical decisions and a higher efficiency in the diagnostic process in the laboratory. However, it increases needed resources and potentially results in costly testing in patients without therapeutic consequence18 19 (figure 1).

Figure 1.

Molecular testing parallel algorithm without next generation sequencing (adapted from Kerr and López-Ríos17). ALK, anaplastic lymphoma kinase; EGFR, epidermal growth factor receptor; FISH, fluorescence in situ hybridisation; ICI, immune checkpoint inhibitor; MDT, multidisciplinary team; NSCLC, non-small cell lung cancer; PD-1, programmed cell death protein 1; PD-L1, programmed death-ligand 1; TKI, tyrosine kinase inhibitor

Testing of driver mutations can be performed by targeted sequencing, a combined sequencing and immunohistochemistry/immunofluorescence approach or next generation sequencing (NGS). EGFR and BRAF testing are conducted by DNA sequencing, while in several laboratories due to cost-effectiveness, ALK and ROS1 testing are mostly performed by immunohistochemistry (IHC) and/or fluorescence in situ hybridisation (FISH). Currently, the approved method for PD-L1 testing is IHC.20 NGS is rapidly emerging as an option for the delivery of multiplexed genomic testing in lung cancer, especially in academic centres. NGS testing potentially provides more data on genetic alterations than the treating clinicians would usually include in their decision-making. Alterations for which no treatment is available or for which treatment is available only through a clinical trial could therefore also be detected. Moreover, NGS approaches are becoming available for the identification of uncommon fusion genes involving ALK and ROS1, but experience of the clinical significance of these aberrations is still limited in the absence of IHC or FISH alterations.17 NGS is still relatively costly and its use will depend on whether it is considered cost-effective compared with doing several single-gene tests (figure 2).

Figure 2.

Molecular testing algorithm when NGS is commonplace (adapted from Kerr and López-Rios17). MDT, multidisciplinary team; NSCLC, non-small cell lung cancer; NGS, next generation sequencing; PD-L1, programmed death-ligand 1; TMB, tumour mutational burden.

Liquid biopsies

Liquid biopsy is a broad term that refers to the analysis of biomarkers that can be isolated from body fluid of patients with cancer. However, the analysis of cell-free DNA (cfDNA) isolated from the plasma fraction of peripheral blood is the only approach that entered clinical practice so far.21 In patients with advanced NSCLC, cfDNA testing can provide information on the presence of driver alterations at the time of therapy decision or on the mechanisms of acquired resistance to TKI in those with driver genetic alterations. Indeed, liquid biopsy is the preferred approach for the assessment of the p.T790M EGFR variant in EGFR-mutant patients who progress after treatment with first-generation or second-generation TKI.22 Additional tissue rebiopsy is usually reserved to patients whose results are negative at liquid biopsy testing. Evidence from trials with third-generation TKI in EGFR T790M-mutant patients suggests that tissue and liquid biopsy might provide complementary information.23 24 A negative liquid biopsy T790M test in patients with tumour positive for T790M is associated with a better prognosis compared with the prognosis of patients with both tissue and tumour positive. This finding most likely reflects the correlation between cfDNA levels and tumour burden and/or aggressiveness of the disease—the higher the tumour load, the higher is the amount of cfDNA. On the other hand, patients with a positive blood T790M test and negative tissue have an intermediate outcome as these patients are likely to carry a heterogeneous expression of the T790M leading to a mixed response to third-generation TKI.22 24

NGS-based analysis of liquid biopsy revealed that approximately 50% of T790M-positive resistant patients also carry additional genetic alterations.25 The presence of multiple resistance mechanisms has been associated with resistance to treatment with third-generation TKI.25–28 This highlights that the genetic background of EGFR-mutant lung cancer might significantly change over time. In fact, the molecular complexity of the disease is likely to increase after each line of treatment because of the emergence of multiple clones of resistant cells. In consequence, liquid biopsy testing with NGS-based techniques might better recapitulate the genetic landscape of the disease compared with tissue biopsy in resistant patients.29

The emergence of resistance against EGFR targeting TKI in precision treatment of NSCLC

Patients with NSCLC who harbour mutations in the EGFR gene are candidates to receive treatment with TKI. After a mean time of treatment of 10–14 months, patients usually stop responding to first-generation and second-generation TKI and in consequence show tumour progression which might be systemic, oligoprogression or restricted to the central nervous system (CNS).4 Mechanisms involved in resistance development have been extensively studied not only for first-generation or second-generation inhibitors but also for third-generation EGFR TKI.30

Resistance against first-generation and second-generation TKI

Emergence of resistance to first-generation and second-generation TKI may be due to alterations in the target gene EGFR or to the acquisition of alterations in other genes. The most frequent resistance mechanism is the acquisition of the mutation affecting the amino acid threonine located at position 790 of the EGFR protein.31 32 This mutation increases the binding of the ATP molecule, compared with the inhibitor and therefore compensates the inhibition of the EGFR. The mutation p.T790M is found in more than 50% of EGFR-mutated patients at the time of progression. It may be detected alone or simultaneously to the amplification of the EGFR gene, or to other resistance mechanisms. Other mutations affecting the EGFR gene have also been found in a limited number of patients, such as EGFR p.L747S, p.D761Y and p.T854A.33–35 Mechanisms of resistance involving genes different from EGFR have been also detected, although to a lesser extent. Among these the most recurrently found are: MET and human epidermal growth factor receptor 2 (HER2) amplification, PIK3CA and BRAF mutations and small cell histologic transformation.36 37 More recently, CDKN2A loss, MTOR mutations and FGFR3 alterations including translocations have also been implicated in mediating EGFR TKI resistance.38 39 The specific third-generation EGFR TKI osimertinib that targets the mutation p.T790M has been developed and demonstrates high efficacy in most patients.40

Resistance against third-generation TKI

Third-generation TKI show even higher effectiveness in EGFR-mutant patients with a median response rate of 80% in untreated patients, including those bearing the p.T790M mutation.41 However, resistance development also eventually occurs, although with 18.9 months significantly later than with first-generation and second-generation TKI.41 The mechanisms of resistance, as with first-generation and second-generation inhibitors, include EGFR-dependent and EGFR-independent mechanisms.40 Among the EGFR-dependent resistance mechanisms, mutations affecting the binding of the drug such as the p.C797S or the p.E709K, p.L692V and p.L798I mutations have been observed. Of note, regarding resistance to third-generation TKI, it seems that the appearance of resistance mutations, depending on the location—either in cis or trans—has different implications: if C797S and p.T790M mutations are in trans, cells will be resistant to third-generation TKI but remain sensitive to a combination of first-generation and third-generation EGFR-TKIs.42 Tumours with C797S and p.T790M mutations in cis are greatly resistant to EGFR-TKI and their combinations.42 If C797S mutation develops in T790 wild-type cells after administration of third-generation EGFR-TKI, the cells retain their sensitivity to first-generation TKI, underscoring the preclinical evidence of TKI sequencing.42 Among EGFR-independent mechanisms of resistance to third-generation TKI, MET amplification as well as amplification of genes involving receptors (such as Insulin-like growth factor 1 receptor (IGF1R)) and mutations or amplifications of genes involved in the signalling cascades (such as BRAF) have been reported.43

Molecular biomarkers for response to ICI therapy

Increasing PD-L1 levels, tumour mutational burden (TMB), CD8 T cell infiltration have been associated with increasing benefit from ICI. However, patient selection by predictive biomarkers remains controversial as no absolute predictive markers reliably differentiating between responding and not responding patients were identified yet.44 45

Expression of PD-L1

PD-L1 expressed by the tumour cells can be induced not only by the oncogenic pathways, which induce tumour development, but also by the immune response itself, especially after induction by the interferon-gamma pathway.46 Thus, in case of immune attack, the tumour cell defends itself and upregulates PD-L1. Accordingly, PD-L1 expression on tumour cells was extensively studied as a predictive marker for PD-1 axis targeting ICI. The expression by the tumour cells is heterogenous with areas of expression alternated with areas of absent expression. In line, cut-off values between 1% and 50% of PD-L1 expression tumour cells were investigated as predictive marker for PD-1 axis targeting ICI.47 48 Analysis of biopsies is challenging in this context as samples could be rated as false negative, potentially inhibiting an effective treatment option.

Furthermore, PD-L1 expression has an imperfect negative predictive value, as also a PD-L1 negative tumour can present with clinically relevant response. In addition, PD-L1 analysis might be challenging due to the expression of PD-L1 not only of tumour cells but also by cells of the microenvironment including macrophages and T cells. In addition to this variability in location, PD-L1 expression was shown to also vary over time, as treatments including chemotherapy and radiotherapy can impact the expression level.49

Importantly, oncogenes were shown to not only impact on tumour growth but also on the expression of immunosuppressive molecules including PD-L1 (figure 3). Here, the presence of ALK translocation was shown to be associated with PD-L1 expression, while the presence of EGFR mutation is inversely correlated with PD-L1 expression.50–52 In patients harbouring an EGFR mutation, those with rare EGFR mutations and not harbouring the specific T790M mutation are more likely to express PD-L1 on the tumour cells.50 53 54

Figure 3.

Relationship of programmed death-ligand 1 (PD-L1) expression with oncogene alterations.

Tumour mutational burden

TMB was recently shown to be an important predictive marker for ICI response, as cancer entities with higher TMB like melanoma or smoking-associated NSCLC were shown to have higher response rates55 56 (figure 4). The high TMB and the resulting high rate of pathological or foreign folded proteins is associated with a higher fraction of neo-epitopes potentially triggering an immune response. It is important that most responding patients have mainly clonal mutations, while non-responders have mutations present in tumour subclones, allowing immune evasion and lower response to ICI.57 Currently, the most suitable cut-off value to define ‘high’ and ‘low’ TMB as well as a uniformed method for detection still need to be defined. Recently, patients with NSCLC with a high TMB, defined as ≥10 mutations per megabase using the FoundationOne CDx assay, treated with a combination of the ICI nivolumab and ipilimumab, showed improved progression-free survival (PFS) compared with those treated with conventional chemotherapy in the first-line setting.58

Figure 4.

Relationship between mutational load and response to immunotherapies targeting PD-1/PD-L1.56 Reprinted by permission from Springer Nature. Yarchoan M, Johnson III BA, Lutz ER et al. Targeting neoantigens to augment antitumour immunity. Nat Rev Cancer 2017. NSCLC, non-small cell lung carcinoma; PD-1, programmed cell death protein 1; PD-L1, programmed death-ligand 1; RCC, renal cell carcinoma; SCLC, small cell lung carcinoma.

Infiltration by CD8 lymphocytes and transcriptomic signature

Another potential biomarker is CD8+ T cell infiltration, which defines the notion of ‘hot’ tumour and ‘cold’ tumour based on the density of tumour-infiltrating lymphocytes (TIL). Particularly in metastatic melanoma, CD8+ T cell infiltration is associated with better response to immunotherapy.59 In lung cancer, T cell infiltrate and tertiary lymphoid structures were reported to be associated with a good outcome in chemotherapy patients.60 Beyond the presence of T cells, the analysis of the transcriptomic signature of the whole tumour and the deconvolution of the signals allow to identify transcriptomic signatures of activated T cells. Strong expression of effector T cell and interferon gene signatures was shown to be associated with a better response to immunotherapies targeting the PD-1/PD-L1 checkpoint in patients with NSCLC.61

The association of higher CD8+ TIL density and non-synonymous mutation burden was verified in a small cohort of patients with oncogene-addicted NSCLC.51 However, in general, the EGFR-mutated NSCLC harbour a much less inflamed tumour microenvironment compared with EGFR wild-type cancers.62 This might be induced by the expression of the immunosuppressive molecule CD73 in EGFR-mutated NSCLC.63 Indeed, given that most patients with EGFR-mutated NSCLC are never smokers, the rate of passenger mutations is lower and in consequence the TMB and the immunogenicity.

Tumour molecular alterations

Certain molecular alterations affecting the tumour cell such as the β-catenin and phosphatase and tensin homologue (PTEN) pathways have been recently linked to response to immunotherapy in melanoma. When the β-catenin pathway is activated in patients with melanoma, there is little infiltration of CD8+ T cells, whereas this infiltration is important when the β-catenin pathway is not activated.64 Similarly, the loss of anti-tumour protein PTEN is associated with a lack of response to immunotherapy in patients with melanoma.65 However, currently no data exists on the relevance of these pathways in oncogene-addicted NSCLC.

Clinical efficiency of ICI in patients with oncogene-addicted NSCLC

Beyond the already approved indications in the first-line and second-line setting of advanced NSCLC, only relatively little data is available regarding anti-PD-1 and anti-PD-L1 efficacy in patients with oncogene-addicted NSCLC. Most of the available data is for patients harbouring an EGFR mutation or ALK rearrangement, while data for the other even more rare NSCLC subtypes is mostly lacking.

The phase 1 study CA209-012 of PD-1 inhibitor nivolumab (n=52) as first-line treatment in patients with metastatic NSCLC reported an impaired overall efficacy in patients with EGFR mutation patients as compared with patients with EGFR wild-type NSCLC. The overall response rate (ORR) was only 14% in patients with EGFR mutation (ie, one of seven) versus 30% in patients with wild-type NSCLC (ie, nine of 30). Further, the PFS at 24 weeks was only 14% in patients with EGFR mutation versus 51% in wild-type NSCLC, respectively.9

Study CA209-153 is a phase 3b/4 safety trial of nivolumab in patients with advanced or metastatic squamous or non-squamous NSCLC who received at least one prior line. The EGFR mutation status was available in 549 patients and 103 patients presented with an activating EGFR mutation. However, partial response rate was 11% (n=55 patients available for response assessment) in the EGFR-mutated cohort compared with 16% (n=300 patient available for response assessment) in the EGFR wild-type cohort.10

The phase 1 KEYNOTE-001 trial found that PD-1 inhibitor pembrolizumab provides promising long-term OS benefit with a manageable safety profile for PD-L1-expressing treatment-naive advanced NSCLC, with greatest efficacy observed in patients with PD-L1 tumour proportion score (TPS) ≥50%.66 The best objective response rate based on mutation status was 16% in patients with EGFR mutation (n=19) versus 37% without mutation (n=89) and 60% in patients (n=5) with unknown EGFR status. Across all PD-L1 subgroups, patients with EGFR-mutant had a lower objective response rate than patients with EGFR wild-type tumour.11 In the 3-year follow-up of the KEYNOTE-001 trial, the median overall survival (OS) in patients with EGFR mutation was 6 months (95% CI 4.4 to 8.8) compared with 12 months (95% CI 9.2 to 14.3) in wild-type patients.67 A recent phase 2 trial investigated pembrolizumab in TKI-naive patients with PD-L1 positive EGFR-mutant NSCLC. No responses were observed in the first 11 included patients and the trial had to be discontinued.68

The phase 2 BIRCH study on PD-L1 inhibitor atezolizumab12 showed higher response rates in patients with higher expression of PD-L1 (on tumour cells (TC) ≥50%) (ORR 35% vs 26% (for all treated patients)). Overall, 13 patients with activating EGFR mutation were included and the ORR was 31% for mutant EGFR versus 23% for wild-type EGFR patients (n=103).

The ImPOWER 150 study compared the use of bevacizumab plus atezolizumab plus carboplatin plus paclitaxel versus carboplatin plus paclitaxel plus bevacizumab.13 Eighty patients (35 in the intervention and 45 in the control arm) with EGFR mutation and 34 patients (13 in the intervention and 21 in the control arm) with ALK rearrangement were included after progression on established TKI treatment. PFS was also longer in those patients with oncogene addicted NSCLC in the intervention arm containing atezolizumab compared with the standard arm (median, 9.7 months vs 6.1 months; unstratified HR, 0.59; 95% CI 0.37 to 0.94)

ATLANTIC is a phase 2, open-label, single-arm trial studying the efficacy of durvalumab, a PD-L1 inhibitor, in pretreated NSCLC including 111 patients with EGFR or ALK alteration.14 Eligible patients had advanced NSCLC with disease progression following at least two previous systemic regimens, including platinum-based chemotherapy. Patients with EGFR or ALK alteration had received standard treatment with TKI before. Among the 111 oncogene-addicted patients, 77 presented with PD-L1 expression in at least 25% of tumour cells. The objective response rate was 12.2% (95% CI 5.7 to 21.8) in the oncogene-addicted patients with PD-L1 expression >25% of tumour cells, while patients with <25% PD-L1 expression the objective response rate was only 3.6% (95% CI 0.1 to 18.3). PFS was not different according to PD-L1 expression in the EGFR or ALK altered patients (1.9 months). In summary, the proportions of patients who achieved a response were generally lower in patients with EGFR or ALK positive NSCLC than in those with EGFR negative and ALK negative NSCLC and higher PD-L1 expression appears to enrich for response. The figure 5 shows response after durvalumab treatment in cohort 1 of the ATLANTIC trial (EGFR+/ALK+)

Figure 5.

Response after durvalumab treatment in cohort 1 of the ATLANTIC trial (EGFR+/ALK+).14 Reprinted with permission from Elsevier. DOR, duration of response; OS, overall survival; PFS, progression-free survival; TTR, time to response.

A recent systematic review and meta-analysis of five randomised trials comparing ICI (nivolumab, pembrolizumab and atezolizumab) versus docetaxel in the second-line setting after chemotherapy showed an OS benefit for EGFR wild-type NSCLC with an impressive HR of 0.67 (p<0.001) for ICI compared with chemotherapy.8 However, in contrast, no OS advantage was observed for patients with EGFR-mutant NSCLC, although the small sample size needs to be considered (HR of 1.11, p=0.54). Only 12% (n=271) of the included patients indeed presented with an EGFR mutation. The major limitation of the study is that EGFR mutation was not determined by centralised testing and in 764 patients (25%) EGFR status was not assessed. The different types of mutations are also unknown.

Table 1 gives an overview on efficacy of anti-PD-1 and anti-PD-L1 in patients with wild-type NSCLC versus patients with EGFR-mutated NSCLC.

Table 1.

Anti-PD-1 and anti-PD-L1 efficacy in patients with wild-type NSCLC versus patients with EGFR-mutated NSCLC

Study	Treatment	EGFR-mutated n (%)	HR OS wild-type versus mutated patients who received anti-PD-1 or PD-L1
Check Mate 05744	Nivolumab versus docetaxel	82 (14%)	0.66 (0.51–0.85) versus 1.18 (0.45–2.07)
KEYNOTE-01082	Pembrolizumab versus docetaxel	86 (8%)	0.66 (0.55–0.79) versus 0.88 (0.45–1.72)
OAK83	Atezolizumab versus docetaxel	85 (10%)	0.69 (0.57–0.83) versus 1.24 (0.71–2.18)
POPLAR61	Atezolizumab versus docetaxel	18 (6%)	0.70 (0.47–1.04) versus 0.99 (0.29–3.40)
ImPOWER 15013	Atezolizumab plus bevacizumab and chemotherapy versuschemotherapy plus bevacizumab	80 (10%)	0.62 (0.52–0.74) versus 0.41 (0.22–0.78)

NSCLC, non-small cell lung cancer; OS, overall survival; PD-1, programmed cell death protein 1; PD-L1, programmed death-ligand 1.

The retrospective multicenter Immunotarget Cohort study reviewed data on the efficacy of ICI in 527 patients with stage IV NSCLC harbouring various activating molecular alterations including KRAS (n=252), EGFR (n=110), BRAF (n=38), MET (n=36), HER2 (n=23), ALK (n=18), RET (n=14), ROS1 (n=5) and multiple drivers (n=31).7 Outcomes by molecular subtypes are shown in table 2. Overall, EGFR-mutant patients presented with a shorter PFS after ICI-based therapy compared with KRAS-mutant patients (p<0.001). The EGFR p.T790M mutation was associated with a shorter PFS than other EGFR mutations (p=0.0001). Among patients with MET alterations, exon 14 mutations were shown to present with the highest response rate to ICI. Also, among patients harbouring an oncogene alteration, smoking status (p=0.003) and PD-L1 expression (p=0.02) were associated with PFS.

Table 2.

Immune checkpoint inhibitor efficacy outcomes in various molecular alterations

Driver	n	Best response (%)			PFS			OS
									CR/PR	SD	PD	Median (months)	6 month PFS (%)	1 year PFS (%)	Median (months)
									BRAF	38	28.1	28.1	43.8	3	35	19	13.6
		KRAS	252	27.2	23.1	49.8	3.2	39	26	13.5
ROS1	5	20	0	80	NA	NA	NA	NA
MET	36	15.6	34.4	50	3.4	33	23	18.4
EGFR	110	11	18	71	2	16	6	8.8
HER2	23	9.5	28.6	61.9	3.5	34	17	10
RET	14	7.1	21.4	71.4	2.2	16	8	6.5
ALK	18	0	21.4	78.6	2.1	16	8	17

Reprinted with permission from American Society of Clinical Oncology. Copyright 2018. All rights reserved. Mazieres J, Drilon AE, Mhanna LJ et al. Efficacy of immune-checkpoint inhibitors in non-small cell lung cancer patients harbouring activating molecular alterations (ImmunoTarget). J Clin Oncol 2018; 36, suppl:abstr 9010.

CR, complete response; NA, not available; PFS, progression-free survival; PR, partial response.

Radiological methods for response evaluation during ICI treatment in NSCLC

Response Evaluation Criteria in Solid Tumours (RECIST) and immune-related response criteria (irRC) are size-based response assessment methods. However, during ICI treatment, lesions can initially increase in size due to an influx of immune cells. When subsequent radiological follow-up shows a decrease in tumour size after initial increase in size or even lesion frequency, this pattern of response is called ‘pseudoprogression’. It is associated with a favourable response to immunotherapy. However, early during treatment pseudoprogression cannot be radiologically discriminated from tumour progression. The resolution limitation of CT, ranging typically around 1 mm,69 can even cause the appearance of ‘new lesions’ as a result of pseudoprogression. Lesions just below the resolution limitation of the CT can grow due to immune cell influx and become large enough to be visible on a CT. With RECIST, this would be classified as progressive disease, while with irRC, the diameters of the new lesion(s) are added to the sum of all diameters and when this sum remains below 20% increase as compared with the baseline value, the response will be classified as stable disease (figure 6). In melanoma, it was estimated that conventional RECIST underestimates the benefit of single agent PD-1 treatment with pembrolizumab in approximately 15% of patients and that the use of irRC better classifies patients according to survival benefit and prevents premature cessation of a potentially successful treatment.70

Figure 6.

Comparison of key differences in Response Evaluation Criteria in Solid Tumours (RECIST) V.1.1 and immune-related response criteria (irRC). Reprinted with permission from American Society of Clinical Oncology, Copyright 2016. All rights reserved.70

Positron emission tomography (PET)-CT using 18F-fluorodeoxyglucose (¹⁸F-FDG) (FDG-PET) visualises and quantifies glucose metabolism of tumour lesions. Metabolic responses after chemotherapy have been associated with favourable outcome in terms of PFS and can precede size-based responses.71 72 A FDG-PET study in patients with NSCLC treated with atezolizumab showed that FDG-PET 6 weeks after treatment initiation is able to classify patients according to survival benefit.73 However, FDG-PET did not seem to outperform the predictive value of CT. This could be due to the nature of FDG-PET, not discriminating between metabolic activity of immune cells and tumour cells. After 6 weeks of PD-L1 directed treatment, it is likely that tumour FDG uptake is the result of a mixture of an increase in FDG metabolism due to influx of immune cells and a decrease due to tumour cell death, hampering response evaluation. Therefore, an earlier time point after treatment initiation aiming at quantifying the metabolic activity of the influx of immune cells and preceding tumour cell death might be a better discriminator of responders and non-responders.

The balance between the a priori probabilities of progression and pseudoprogression should guide treatment decision. For patients with molecularly driven NSCLC, response rates are generally low to very low and therefore progression according to RECIST and irRC is expected to represent real tumour progression in most cases and rarely pseudoprogression. In patients with molecularly driven NSCLC, RECIST-based response evaluation is expected to be the best discriminator between patients who derive benefit from PD-L1 directed therapy and those who do not. Treatment beyond progression is therefore currently not advised, as well as the use of irRC.

Sequence of treatments in molecularly mutated NSCLC: is there a place for ICI?

The introduction of mutation-specific TKI revolutionised the treatment of patients with NSCLC harbouring an oncogene addiction. However, at some point, resistance occurs in almost all patients. Durable responses can be observed in patients treated with ICI; however, the response in patients with oncogene-addicted NSLCC was shown to be impaired compared with wild-type patients.8–11 To offer patients with oncogene addiction the chance of immune induced long-term control of their disease, new combinations are most certainly the only clinical possibility. Here, the ATLANTIC trial proposed that the quartet combination therapy could be a promising approach in patients’ progression on all available generations of oncogene-specific TKI.14 Given the potential negative impact of the oncogene on the activity of the inflammatory tumour microenvironment, the combination of TKI with ICI is of potential clinical interest. Certainly, side effects would increase as seen by the combination of ICI with TKI in other entities like melanoma.74–77 However, in theory, the rapid antigen release through dying tumour cells by the TKI could enhance the inflammatory response. First phase I trials combining ICI with EGFR mutation directed TKI show acceptable side effect profiles.78–81 Several clinical trials investigating the combination of TKI with ICI are currently recruiting (eg, NCT02364609, NCT01454102, NCT01998126).

Conclusion

Lung cancer is becoming a more diverse disease with regard to management with a wide range of targets and treatment options. Recent clinical data on ICI in NSCLC harbouring activating mutations reviewed at the round-table discussion and summarised in this article shows overall low efficacy, although interpretations have to be drawn carefully due to the limited amount of data available. However, promising subgroups needing further clinical investigation were identified. For example, the clinical activity of durvalumab late line salvage therapy in patients with EGFR-mutated NSCLC showing PD-L1 expression in ≥25% of tumour cells is encouraging.14 Here, the clinical challenge is to further understand the biological drivers of inflammation in NSCLC and to identify subgroups driving the benefit as well as defining the optimal treatment sequence with established TKI. In addition, further research is needed to address the heterogeneity of EGFR-mutant lung cancer and to assess whether changes in the biology of the disease following different lines of therapy might increase the sensitivity to ICI.

To offer patients with oncogene addiction the chance of immune induced long-term control of their disease, new combinations are probably the only clinical possibility. A combination strategy as recently analysed in the ImPOWER 150 study13 including platin-based chemotherapy, bevacizumab and atezolizumab showed a clinically meaningful efficacy also in patients with oncogene-addicted NSCLC progressing on all available generations of TKI. However, several other trials reported no significant increase in response rate or PFS in patients with oncogene-addicted NSCLC.9–11 67 Given the potential negative impact of the oncogene on the activity of the inflammatory tumour microenvironment, the combination of TKI with ICI is of potential clinical interest. Several clinical trials investigating the combination of TKI with ICI are currently recruiting (eg, NCT02364609, NCT01454102, NCT01998126). Due to the overall low clinical efficiency of ICI in patients with oncogene-addicted NSCLC based on the so far available data from prospective clinical trials, the round-table panel concluded that ICI should currently only be considered after exhaustion of targeted therapies including standard and salvage chemotherapies in these patients.