Unravelling signal escape through maintained EGFR activation in advanced non-small cell lung cancer (NSCLC): new treatment options.

The discovery of activating epidermal growth factor receptor (EGFR) mutations has opened up a new era in the development of more effective treatments for patients with non-small cell lung cancer (NSCLC). However, patients with EGFR-activating mutated NSCLC treated with EGFR tyrosine kinase inhibitors (TKIs) ultimately develop acquired resistance (AR). Among known cases of patients with AR, 70% of the mechanisms involved in the development of AR to EGFR TKI have been identified and may be categorised as either secondary EGFR mutations such as the T790M mutation, activation of bypass track signalling pathways such as MET amplification, or histologic transformation. EGFR-mutant NSCLC tumours maintain oncogenic addiction to the EGFR pathway beyond progression with EGFR TKI. Clinical strategies that can be implemented in daily clinical practice to potentially overcome this resistance and prolong the outcome in this subgroup of patients are presented.

Introduction

Lung cancer is the third most frequent cancer (1.82 million cases) and the leading cause of cancer-related death (1.59 million deaths) worldwide according to the GLOBOCAN 2012 database.1 The current SEER database shows that 16% of non-small cell lung cancers (NSCLCs) are diagnosed at localised stage (for which surgery remains the standard treatment for fit patients), 22% of patients are diagnosed with locally advanced disease and up to 60% with advanced stage.2 Historically, platinum-based doublet chemotherapy has been the standard first-line treatment for non-selected patients with advanced NSCLC who have a good performance status.3 The knowledge of genomic alterations in lung cancer has implications for the management of this disease.4 The introduction of targeted therapies, based on the recognition of the significance of acquired genetic driver mutations, has changed the treatment paradigm of patients with lung cancer, establishing tumour genotyping as an essential routine diagnostic tool in clinical practice, notably in cases of adenocarcinoma histology.5 Different types of mutations have been reported in lung adenocarcinomas,4 but only 20–25% of them are actionable oncogenic driver mutations.6 It has been reported that patients with NSCLC with a tumour harbouring known oncogenic drivers and who receive a matched targeted agent live significantly longer than those who have a driver mutation but do not receive personalised treatment (HR, 0.69; 95% CI 0.53 to 0.90; p=0.006),7 supporting the clinical benefit of this policy.6

In Caucasian patients, the most frequent genetic alterations in advanced adenocarcinoma lung cancer are the KRAS mutation ∼29%, epidermal growth factor receptor (EGFR) mutations ∼11%, ALK rearrangements ∼5%6 and MET mutations (exon 14) in 4%.8 Other less frequent mutations include BRAF and PIK3CA mutations in ∼2% each, the HER2 mutation in 1% of tumours6 and ROS1 rearrangements9 in 1% of patients with NSCLC. These oncogenic drivers are almost always mutually exclusive in this patient population.7

There are several classes of activating somatic EGFR mutations, with in-frame deletions in exon 19 (ELREA, Del19) and single-point mutations in exon 21 (L858R) being the most frequent. These mutations are markedly more frequent in the Asian than in the Caucasian population (∼50% vs 10%).10 In this review, we summarise the state of play for current treatments in the EGFR-mutant population for different treatment settings following acquired resistance (AR) and present the rationale behind new approaches being investigated following development resistance.

Failure to first-line treatment in patients with advanced EGFR-mutant NSCLC

EGFR mutations predict sensitivity to the first-generation reversible EGFR tyrosine kinase inhibitors (TKIs) erlotinib and gefitinib and to the second-generation irreversible EGFR TKIs such as afatinib and dacomitinib. All four of these drugs have demonstrated improvement in the response rate (RR), progression-free survival (PFS) and quality of life over standard first-line platinum-doublet chemotherapy in at least nine randomised phase III trials in patients with advanced EGFR-mutant NSCLC. To date, no differences in overall survival (OS) versus standard platinum-doublet chemotherapy have been reported in these trials, possibly due to the high crossover rate to the TKI arm following disease progression in chemotherapy-treated patients.11 A pooled analysis of two randomised phase III trials (LUX-Lung 3 and LUX-Lung 6), which compared afatinib versus first-line chemotherapy, reported that although afatinib did not improve OS in the whole population of either trial, OS was improved with the drug for patients with Del19 EGFR mutations.12 In a recent meta-analysis of seven trials (1649 patients) evaluating EGFR TKIs, the EGFR TKI benefit over standard first-line chemotherapy was 50% greater for tumours with the Del19 mutation (HR, 0.24; 95% CI 0.20 to 0.29) than for those with the exon 21 L858R substitution (HR, 0.48; 95% CI 0.39 to 0.58; p<0.001).13 In a second meta-analysis, patients with the Del19 EGFR mutation had a significant OS benefit under TKI treatment (HR, 0.72; 95% CI 0.60 to 0.88; p=0.02) but not with the L858R EGFR mutation (HR, 1.15; 95% CI 0.95 to 1.39; p=0.07).14 These results suggest that EGFR-mutant tumours are not equal, with different treatment impacts according to the mutation subtype in first-line setting.

There is currently no consensus as to which inhibitor maximises therapeutic efficacy in patients with EGFR-mutant NSCLC. The phase III study CTONG 0901 (NCT01024413) comparing erlotinib with gefitinib in 256 patients with advanced NSCLC harbouring common EGFR mutations presented no significant difference in outcomes (OS, PFS and ORR) or toxicity between these drugs.15 The results from the LUX-Lung 7 study, a randomised phase IIb trial (NCT01466660) with 319 patients comparing afatinib to gefitinib as first-line therapy for patients with EGFR-mutant tumours, showed a modest but significant PFS prolongation (11.0 vs 10.9 months, HR, 0·73; 95% CI 0.57 to 0.95; p=0·017) and ORR (70.0% vs 56.0%; p=0.008) in favour of afatinib compared to gefitinib, independently of EGFR mutation subtype.16 In a recent pooled analysis from two randomised trials, dacomitinib was an active drug with comparable outcomes to erlotinib in patients with NSCLC with common EGFR mutations.17 The phase III trial ARCHER1050 (NCT01774721) comparing dacomitinib with gefitinib in patients with treatment-naive EGFR-mutant NSCLC has finished recruitment.

New strategies are needed to further improve PFS in patients with EGFR-mutant NSCLC. In a phase II trial, 154 Asian patients with EGFR-mutant NSCLC were randomised to erlotinib plus bevacizumab (15 mg/kg every 3 weeks) or erlotinib as first-line treatment. The combination significantly improved median PFS compared with monotherapy (16 vs 9.7 months, HR, 0.54; 95% CI 0.36 to 0.79, p=0.0015), without differences in serious adverse events between both treatments.18 In a single-arm phase II trial in 109 Caucasian patients with EGFR-mutant NSCLC, erlotinib and bevacizumab reported an overall 1-year PFS of 56.7% and a median PFS of 13.8 months (95% CI 10.3 to 21.3). The median PFS was 16.0 months for patients with T790M at diagnosis (34% of patients) and 10.5 months for those without de novo T790M.19 These data led to EMA approval for erlotinib and bevacizumab combination in this population in 28 April 2016. In another phase II study in an Asian population treated with gefitinib and bevacizumab, PFS differed significantly according to EGFR mutation subtype (18.0 vs 9.4 months, p=0.006, for Del19 vs L858R, respectively).20 A phase III trial (NCT02411448) of ramucirumab (anti-VEGFR2) and erlotinib is ongoing to validate this strategy.

Another combination strategy has been developed with pemetrexed. A phase II trial in 191 East Asian patients with EGFR-mutated NSCLC achieved a significant improvement in PFS for the combination compared with gefitinib monotherapy (15.8 vs 10.9 months; HR, 0.68; 95% CI 0.48 to 0.96; p=0.014); however, the incidence of grade 3–4 study related adverse events was significantly higher for the combination than for monotherapy (42.1% vs 18.5%; p=0.001).21 This strategy has only been tested in the Asian population but is not considered acceptable elsewhere as a first-line strategy.

Intercalated administration of an EGFR TKI and chemotherapy has been postulated as a feasible option providing pharmacodynamic separation of the two drugs.22 The intercalated strategy was evaluated in the randomised phase III FASTACT-2 trial in 451 unselected patients with advanced NSCLC. Among patients with EGFR-mutant NSCLC, an improvement was resulted in PFS (HR, 0.25; p<0.0001) and OS (HR, 0.48; p=0.0092) compared with chemotherapy plus placebo.23 It is as yet unclear if such a strategy will be more effective than a first-line EGFR TKI followed by an adapted treatment in the event of resistance.

NSCLC tumours invariably develop AR after a median of 9–11 months from the start of the treatment.11 Several potential mechanisms of AR have been identified: (1) development of secondary EGFR mutations such as the gatekeeper T790M point mutation in exon 20 of the EGFR gene; (2) bypass signalling pathways such MET amplification, PIK3CA mutation, BRAF mutation and HER2 amplification (figure 1); and (3) phenotypic changes, specific to small cell lung cancer or to NSCLC with evidence of epithelial-to-mesenchymal transformation. However, ∼20% of AR mechanisms are still unknown.24 There is no well-defined strategy for EGFR TKI AR, and the patients are managed according to known mechanisms of AR or disease progression patterns.25

Figure 1

Main mechanisms of AR to EGFR TKI in EGFR-mutant NSCLC cells. AR, acquired resistance; EGFR, epidermal growth factor receptor; TKI, tyrosine kinase inhibitor; M, mutated; NSCLC, non-small cell lung cancer; P, phosphorylation.

Persisting oncogenic addiction to the EGFR pathway beyond progression

Persistent inhibition of tumour-dependent pathways beyond progression is a therapeutic strategy, which can be exploited in cancers in which there is a reliance on a single pathway for growth and proliferation. This ‘addiction’ is seen in various cancers such as advanced prostate cancer, which remains almost uniformly dependent on androgen receptor despite acquired resistant to hormonal therapies such as castration,26 or HER2-positive advanced breast cancers, which can achieve an outcome at different HER2 inhibitors.27 Similarly, EGFR-mutant NSCLC maintains dependence on EGFR signalling even after the development of AR,28 29 and many trials have studied the impact of intensifying EGFR inhibition continuing the same EGFR TKI beyond progression, or using a second-generation irreversible EGFR TKI as a monotherapy or in combination with anti-EGFR monoclonal antibodies.

Continuing single-agent EGFR TKIs beyond progression

Some patients with EGFR-mutant NSCLC have indolent progression, and the patients may be treated with an EGFR TKI beyond RECIST progression if there is no clinical evidence of deterioration or intolerable toxicity.30 In the retrospective SLADB study,31 the median PFS according to the physician criteria was 14.0, whereas in the prospective EURTAC trial,32 the same group had a median PFS of 9.7 months according to the RECIST criteria. Continuing EGFR TKI beyond RECIST progression may delay salvage systemic therapy for a year or more.33 Recently, the phase II ASPIRATION trial conducted in 208 Asian patients with NSCLC, whose tumour harbours common EGFR mutations, and treated with first-line erlotinib reported that continuing erlotinib beyond RECIST progression improved PFS by 3.9 months (from 11.0 to 14.9 months).34 However, the lack of an optimal control arm, the fact that the decision to continue erlotinib at progression was at the investigators' or patients' discretion and the unknown local therapies administered diminish strength of the study. Taken together, the results of these studies suggest that continuing EGFR TKI beyond RECIST progression is an adequate strategy for patients with good performance status, progression in previously identified lesions,35 a longer time to progression on EGFR TKI33 and no more than one metastatic site.36

Switching to second-generation irreversible EGFR TKIs

Second-generation irreversible EGFR TKIs such as afatinib12 and dacomitinib37 are effective in the treatment of untreated EGFR-mutant lung cancers, although, as a monotherapy, they failed to overcome T790M-mediated resistance.38 Two phase II trials have tested the efficacy of afatinib in patients with advanced NSCLC who progressed after one or two lines of chemotherapy and with at least 12 weeks of erlotinib and/or gefitinib. In the LUX-Lung 1 trial, 697 patients were randomised to afatinib or placebo. A 2-month PFS improvement was seen with afatinib (3.3 vs 1.1 months; HR, 0.38; 95% CI 0.31 to 0.48; p<0.0001), although there was no OS benefit (10.8 vs 12.0 months; HR, 1.08; 95% CI 0·86 to 1·35; p=0·74).39 The LUX-Lung 4,40 a single-arm trial, reported a median PFS of 4.4 months with afatinib in this pretreated population. Neither of these trials supports switching to second-generation EGFR TKIs, but rather continuing inhibition of EGFR pathway beyond EGFR TKI progression.

Combining second-generation irreversible EGFR TKIs and monoclonal antibodies

In a phase Ib clinical trial, the combination of afatinib and cetuximab (an anti-EGFR monoclonal antibody) in 126 heavily pretreated patients with advanced EGFR-mutant lung cancer and AR to erlotinib/gefitinib reported encouraging results in terms of RR, independently of T790M status (32% vs 25%, for T790M positive vs T790M negative, p=0.341), with a median PFS of 4.7 months. However, the 44% rate of grade 3 toxicity might limit its applicability in daily clinical practice.41 A further phase I study was designed to find a more tolerable combination (NCT02020577), with preliminary results giving 11% grade 3 toxicity.42 In another phase Ib/II trial with 50 pretreated patients with EGFR-mutant NSCLC, the combination of afatinib and nimotuzumab, a humanised anti-EGFR monoclonal antibody, gave similar outcomes for RR (23%), median PFS (4 months) and OS (11.7 months) with 16% grade 3 toxicity. No differences in RR (p=0.628) or PFS (p=0.720) according to the T790M status were reported.43 Two randomised phase II trials (NCT 02438722 and IFCT1503 ACE) are currently evaluating afatinib with or without cetuximab as first-line treatment in patients with EGFR-mutant NSCLC.

Despite the efficacy outcomes with this combination approach, its toxicity and the efficacy of third-generation EGFR TKI as first- and second-line treatment may limit its widespread implementation in daily clinical practice.

Third-generation EGFR TKIS

The substitution of threonine to methionine at amino acid position 790 (T790M) in exon 20 of the EGFR gene is the most frequent mechanism of AR, accounting for 49–63% of cases depending on the detection method.44–46 The T790M mutation enhances the ATP affinity of the kinase domain of the EGFR-mutant receptor restoring its affinity for ATP to that of wild-type EGFR. Given that EGFR TKIs are competitive inhibitors with ATP, their ability to bind to the kinase domain is decreased by this mutation.47 AR via the T790M mutation defines a subset of EGFR-mutant lung cancers with indolent growth in preclinical models28 and clinical studies.48 49 However, as not all authors confirm this finding,50 its prognostic implication remains undetermined. Moreover, a study focusing on the discrepancies seen between tissue and plasma samples when identifying the T790M mutation is underway using the amplification-refractory mutation system (ARMS) and droplet digital PCR methods (NCT02418234).

Recently, de novo T790M mutations in EGFR TKI-naïve patients were described with a highly frequency, ranging from <1% to 80%, according to the detection method,51–53 and it predicts shorter outcome to reversible EGFR TKI.54 55

Osimertinib (AZD9291),56 rociletinib (CO1686)57 and olmutinib (HM61713)58 are third-generation oral EGFR TKIs, mutant-selective covalent inhibitors of commonly mutated and resistant (T790M) forms of EGFR (eg, in H1975 (L858R/T790M) and PC-9 VanR (Del19/T790M) resistant cell lines, osimertinib reported activity with a mean IC50 potency of <15 nmol/L compared to 6073 nmol/L with first-generation EGFR TKI), which do not affect wild-type EGFR. They cause less frequent and less severe gastrointestinal and skin toxicity compared to first-generation or second-generation EGFR TKIs56 (table 1).

Table 1

IC50 values for four different EGFR-mutant, T790M-resistant cancer cell lines treated with reversible (gefitinib, erlotinib) and irreversible (afatinib, dacomitinib, AZD9291) TKIs (data from Cross D56)

	H1975 (L858R/T790M) IC50 potency (nmol/L)	PC-9 VanR (Del19/T790M) IC50 potency (nmol/L)	PC-9 (Del19)	H3255 (L858R)*	WT
Osimertinib	15	6	17	60 to 49	480
Dacomitinib	40	6	0.7	1.2 to 1.3	12
Afatinib	22	3	0.6	1 to 0.8	15
Gefitinib	3102	741	7	11 to 12	59
Erlotinib	6073	1262	6	8 to 11	91

*95% CI.

TKIs, tyrosine kinase inhibitors; WT, wild-type.

In the phase I AURA trial, osimertinib was tested in 253 patients with advanced NSCLC who had radiologically documented disease progression after previous treatment with first-generation EGFR TKI. EGFR T790M was detected in 62% of patients, not detected in 28%, and the status was unknown in 10%. The overall RR was 51% with no differences by ethnicity, and the median PFS was 8.2 months. Patients with a centrally confirmed T790M mutation had better RR (61% vs 21%) and PFS (9.6 vs 2.8 months). Among all doses tested (20 up to 240 mg daily), the dose of 80 mg daily was considered optimal to maximise efficacy and minimise skin and gastrointestinal toxicity observed at higher doses.59 In the updated results of this study, 80 mg osimertinib provided a 71% of RR and a median PFS of 9.7 months60 (table 2). The FDA approved osimertinib in November 2015 for patients with acquired EGFR T790M NSCLC, based on data from the two AURA phase II studies (AURA extension61 and AURA262), which demonstrated efficacy in 411 patients with T790M EGFR-mutant NSCLC that had progressed on or after an EGFR TKI with a 59% of RR. The EMA followed suit in April 2016. Phase III clinical trials with osimertinib, as first-line treatment (FL-AURA, NCT02296125) in patients with common EGFR mutations and as second-line treatment (AURA3, NCT02151981) in patients with tumours harbour T790M mutation, are ongoing.

Table 2

Efficacy of third-generation EGFR TKI in patients with EGFR-mutant NSCLC after acquired resistance to EGFR TKIs

	Osimertinib60*	Rociletinib†65	Olmutinib58 66
T790M positive
ORR (%)	At 80 mg*:71	45	At 800 mg: 56
PFS (months)	At 80 mg*:9.7	6.1	At 800 mg: NR
T790M-negative
ORR (%)	26*	17	12
PFS (months)	3.4*	1.8	2.5

*Updated results reported in last ELCC congress (Amsterdam 2016).

†Updated results reported by Sequist et al65 in NEJM 2016.

NR, not reached; ORR, overall response rate; PFS, progression-free survival.

Rociletinib was evaluated in a phase I/II TIGER trial of 130 patients with EGFR-mutant NSCLC with AR to first-generation or second-generation EGFR TKIs. The RR among patients with centrally confirmed T790M-positive tumours was 59%, independently of the EGFR mutation subtype, with an estimated median PFS of 13.1 months, whereas the RR was 27% and the median PFS was 5.7 months among T790M-negative disease.63 The updated results in 270 patients with T790M-positive NSCLC confirmed an overall RR of 53% and a median PFS of 8 months (10.3 months for those patients without brain metastasis) with rociletinib at 500 mg two times a day, whereas among the T790M-negative cohort the RR was 37%64 (table 1). However, updated data from a pooled cohort of patients from TIGER-X and TIGER-2 (another phase 2 study of rociletinib) reported an RR of only 28–34%. A retrospective, independent review of data from the phase I TIGER trial reported a 45% RR and 6.1 months PFS versus 17% and 1.8 months for T790M-positive versus negative tumours, respectively,65 and rociletinib development was halted.

Finally, olmutinib has been tested in a phase II trial among 76 patients with T790M-positive NSCLC as second-line treatment. Olmutinib achieved an RR of 56% with a median duration of response of 8.3 months.66 This promising efficacy has led the first approval for this drug among the patients with EGFR T790M mutation-positive NSCLC in South Korea in 2016.

Addiction beyond third-generation EGFR TKI

EGFR oncogenic addiction can even persist after AR to third-generation EGFR TKI. Acquisition of C797S mutation in exon 20 of EGFR is the main mechanism of resistance to osimertinib with or without the T790M mutation.67 The context in which the C797S mutation develops with respect to other EGFR alleles affects the efficacy of subsequent treatments. In preclinical models, if the C797S and T790M mutations are in trans nature (on a different allele), cells are resistant to third-generation EGFR TKIs, but are sensitive to a combination of first-generation and third-generation TKIs. If the mutations are in cis nature (on the same allele), no EGFR TKIs alone or in combination can suppress activity.68 If this preclinical approach is confirmed in the clinic, treatment-making decisions will require sequencing biopsies on progression with third-generation EGFR TKI to determining the cis or trans nature of C797S with respect to T790M mutation.

It has recently been reported that patients progressing on rociletinib achieved a response with osimertinib. The clinical benefit of this strategy may be explained by incomplete target inhibition with rociletinib, because in rociletinib-resistant tumours, a C797S resistance mutation has not been identified.69 Clonal selection due to cancer cell heterogeneity in response to drug treatment pressure might explain the efficacy of sequential treatment with different EGFR TKIs (figure 2). This postprogression efficacy with sequential EGFR TKI strategies mirrors crizotinib-resistant ALK tumours, which can respond to sequential ALK TKI therapies based on different resistance-mutational profiles with different ALK TKIs.70

Figure 2

Mechanisms of AR after treatment pressure. AR, acquired resistance; CT, chemotherapy; PFS, progression-free survival.

Additional data will help determine the optimal strategy for using third-generation EGFR TKIs in patients with EGFR-mutant NSCLC, notably their use as first-line therapy or after failure of first-generation or second-generation TKIs, and the optimal sequence.68 69 Prescription of these agents requires a positive T790M mutation in a new tumour tissue biopsy. However, non-invasive assessment and monitoring of the T790M mutation in plasma samples (liquid biopsies) offer a promising alternative to tissue-based biopsies and could complement tumour testing by identifying T790M mutations missed because of tumour heterogeneity or biopsy inadequacy.71 72 Moreover, recent results reported equal efficacy of rociletinib independently of whether the T790M analysis was performed in plasma or tissue;64 this suggests that liquid biopsies could become new standard tests in the near future and may even be used as a dynamic marker of treatment efficacy.

Oncogenic addiction: by-pass mechanisms and MET amplification

The EGFR oncogenic addiction also persists in cases of AR due to by-pass mechanisms. Amplification of the MET is the second most common mechanism of AR to EGFR TKIs (∼20%)73 irrespective of the T790M mutational status.74 Targeting MET in combination with EGFR TKI to overcome resistance is a viable option from the biological standpoint (figure 1). In a phase II trial, the MET inhibitor cabozantinib administered in combination with erlotinib in patients with EGFR-mutant NSCLC with AR to EGFR TKI resulted in an ORR of 8.1%, a PFS of 3.7 months and an OS of 9.1 months. MET gene amplification was not found in any of those with postprogression biopsies (available for 41% of patients).75 INC280, another MET inhibitor, gave a 15% of RR in combination with gefitinib in patients with EGFR-mutant and MET-positive (amplification (FISH ≥5 CN) or overexpression (IHC 2/3+)) NSCLC.76 An INC280 and erlotinib combination is also feasible, and expansion cohorts in EGFR-mutant, treatment-naïve and pretreated patients are ongoing.77 MET inhibitors in combination with EGFR inhibitors further evaluation in selected patients, but toxicity can be an issue.

Switching to chemotherapy

For T790M-negative patients at the time of progression after a first-line EGFR TKI, other than participating in a clinical trial, second-line, platinum-based chemotherapy is a rational option, especially for those patients with systemic progression. NSCLC harbouring EGFR-activating mutations are more likely to express low excision repair cross-complementing 1 (ERCC1) mRNA levels,78 which may justify the enhanced efficacy of first-line, platinum-based chemotherapy among patients with EGFR-mutant NSCLC reported in some clinical trials.79 However, it remains unclear whether prior EGFR TKI impacts the efficacy of subsequent chemotherapy. Results from two retrospective analyses were inconsistent,80 81 and a third study reported that pemetrexed as second-line treatment significantly prolonged PFS compared to second-line, platinum-doublet chemotherapy with similar RRs for the two strategies in patients with EGFR-mutant NSCLC with AR EGFR TKI.82 However, the analysis from the phase III NEJ002 trial (comparing gefitinib with carboplatin plus paclitaxel) showed that second-line, platinum-based chemotherapy followed at progression by gefitinib was similar to first-line, platinum-based chemotherapy in terms of RR (partial response with chemotherapy second-line treatment vs first-line treatment: 25.4% vs 30.7%; OR 1.45; 95% CI 0.75 to 2.81; p=0.345) and OS (28.9 vs 27.6 months; HR, 0.77; 95% CI 0.52 to 1.14; p=0.188) with no influence of EGFR mutation subtype in the efficacy of second-line, platinum-based chemotherapy.83

Given the potential heterogeneity of cancer cells during AR to EGFR TKI, with some clones remaining sensitive to the original EGFR inhibitor,28 and the risk of disease flare (a phenomenon of rapid disease progression during a washout period after EGFR TKI cessation84) in up to 25% of patients, chemotherapy combination and continuing EGFR TKI are likely to have impact outcome. Four randomised phase III studies in molecularly unselected populations failed to show a better outcome with concurrent combination of EGFR TKI and chemotherapy over chemotherapy alone as first-line treatment in patients with advanced NSCLC.85 In patients with AR to afatinib, the randomised phase III LUX-Lung 5 trial compared continuing afatinib with paclitaxel (40 mg/day; 80 mg/m2 weekly) versus the investigator's choice of chemotherapy alone in 202 patients. The combination regimen demonstrated a significant improvement in RR (32.1% vs 13.2%; p=0.005), evaluated by the investigator, and PFS (5.6 vs 2.8 months; p=0.003) compared with chemotherapy alone with no differences in OS.86 Benefit of continued EGFR TKI with chemotherapy in a cohort of 78 patients with EGFR-mutant NSCLC with AR to an EGFR TKI was evaluated in an institutional database. This also showed that continuing the EGFR TKI along with chemotherapy improved RR over chemotherapy alone (41% vs 18%; p=0.02), although no improvement in PFS and OS was observed between the groups.87

On the other hand, the randomised phase III IMPRESS trial compared maintenance gefitinib combined with pemetrexed and cisplatin versus chemotherapy alone in 265 patients with AR to gefitinib. No PFS difference was reported in both arms (median PFS 5·4 months in both groups; HR, 0·86; 95% CI 0.65 to 1.13; p=0.27) and had a deleterious effect on OS compared with the chemotherapy alone (14.8 vs 17.2 months; HR, 1.62; 95% CI 1.05 to 2.52; p=0.03, immature data),88 suggesting that the EGFR TKI should be discontinued in AR patients in combination with doublet second-line chemotherapy.

Two ongoing phase II trials (NCT 02098954, NCT 02064491) may help to develop a definitive recommendation about combining chemotherapy and erlotinib in patients with erlotinib-resistant, EGFR-mutant NSCLC. The biomarker analyses of the IMPRESS trial addressed the question whether T790M status affected the benefit of continuing an EGFR TKI with chemotherapy. T790M status was tested using plasma circulating cell-free, tumour-derived DNA (centrally detected using a quantitative emulsion BEAMing digital Sysmex PCR assay with positivity defined as ≥ 0.02% mutant DNA fraction). Patients without T790M mutation (40% of patients) had a non-significant trend towards benefit with combined treatment compared with chemotherapy alone (PFS: 6.7 vs 5.4 months; HR, 0.67; 95% CI 0.43 to 1.03; p=0.07).89 This hypothesis thus requires further confirmation in a prospective randomised study.

Taken together, these results suggest that second-line, platinum-based doublet chemotherapy will remain the standard of care in patients without the T790M mutation or other targetable resistance mechanisms, as viable evidence does not support maintenance of an EGFR TKI when switching to second-line chemotherapy.

Local strategies beyond EGFR TKI progression

Based on the pattern of relapse, patients with EGFR-mutant NSCLC can be classified into three categories with different prognostic and therapeutic implications: dramatic progressors who will be treated with chemotherapy, gradual progressors and oligometastatic progressors (figure 3).25 In cases of rapid and symptomatic progression, a biopsy is recommended to rule out phenotypic transformation to small cell lung cancer, in which chemotherapy should be adapted according to the pathological report.

Figure 3

Patterns of clinical relapse and algorithm for the therapeutic strategy when AR to EGFR TKI occurs in patients with EGFR-mutant NSCLC. *After discussion with the patients. AR, acquired resistance; EGFR, epidermal growth factor receptor; TKI, tyrosine kinase inhibitor; NSCLC, non-small cell lung cancer.

For patients with oligometastatic progression, local therapies such as radiotherapy, surgery and stereotactic ablative radiotherapy in conjunction with continued EGFR TKI can give long-term survival. For gradual progressors patients, continued EGFR TKI treatment is recommended in asymptomatic patients.90 91

The true incidence of brain metastases in EGFR-mutant NSCLC is unknown, and whether EGFR TKIs are effective in these patient subgroups is unknown.92 However, some patients develop brain metastasis during EGFR TKI treatment, in which local therapy (surgery or radiotherapy) and continued EGFR TKI may be suitable, extending disease control by over 6 months.90 Moreover, concurrent EGFR TKI (erlotinib) plus concurrent whole brain radiation therapy is safe and well tolerated impacting the outcome of EGFR-mutant patients.93 However, current first-generation EGFR TKIs generally have poor properties for penetrating across the blood–brain barrier at recommended doses.94–96 Second-generation EGFR TKIs such as afatinib have reported higher efficacy on central nervous system metastases.97 A novel EGFR TKI, AZD3759, which was designed to effectively cross the blood–brain barrier, has recently demonstrated promising efficacy among EGFR-mutant patients with brain metastases.98

Immunotherapy and EGFR TKIs

The ability of cancer cells to evade antitumour T cell activity in the microenvironment is a hallmark of cancer progression.99 One means of evading immune destruction is through the expression of endogenous immune checkpoints, such as programmed cell death ligand 1 (PD-L1), that terminate immune responses after antigen activation.100

Three randomised phase III trials have reported a statistically significant improvement in RR and OS with checkpoint inhibitors such as nivolumab and pembrolizumab over standard second-line docetaxel chemotherapy in NSCLC.100–102 However, the efficacy of these checkpoint inhibitors among EGFR-mutant patients was lower than in the wild-type population (OS HR EGFR mutant vs EGFR wild-type: 0.88 vs 0.66 and 1.18 vs 0.66 with pembrolizumab102 and nivolumab in non-squamous cancers,101 respectively). Recently, benefit from immunotherapy has been associated with tumours bearing high levels of somatic mutations.103 Low level of mutational load in EGFR-mutant tumours could explain the lower efficacy of immune checkpoint therapies among this NSCLC subpopulation.

EGFR-mutant NSCLC expresses higher PD-L1 levels than wild-type, and in vitro and in vivo experiments have shown that gefitinib can reduce PD-L1 expression by inhibiting KF-KB, suggesting that combined strategies of EGFR TKI and immunotherapy warrant further evaluation.104 Preliminary clinical results of combined nivolumab plus erlotinib showed clinical benefit (ORR 19%) and an acceptable safety profile in patients with EGFR-mutant advanced NSCLC with AR to EGFR TKI.105 Pembrolizumab and gefitinib also reported clinical activity in heavily pretreated (up to four prior therapies) patients with EGFR-mutant NSCLC.106 These results support further evaluation of checkpoint inhibitors in patients with EGFR-mutant NSCLC, and several phase I/II trials are ongoing in EGFR TKI-naïve and pretreated patients (NCT 02013219: erlotinib and atezolizumab, NCT 02630186: rociletinib and atezolizumab, NCT 02323126: EGF816 and nivolumab, NCT 02364609: afatinib and pembrolizumab). However, in the phase I TATTON trial (NCT02143466) and in the randomised phase III CAURAL trial (NCT02454933) the combination of osimertinib and durvalumab (anti-PD-L1 monoclonal antibody) in patients with EGFR- mutant NSCLC with AR to EGFR TKI and T790M positivity have been halted due to an excess of pulmonary toxicity with the combination, suggesting that toxicity can limit the effectiveness of this combination.

Conclusions

Despite the efficacy of first-generation and second-generation EGFR TKIs, all patients develop AR to the treatment. Optimal postprogression therapy should not be systematically always tailored according to the RECIST progression criteria, with delaying the switch from first-line therapy for months after progression with maintenance EGFR and the addition of some local therapies being the optimal option in some patients. While combination targeted therapies offer promising alternatives in many AR settings, some recent studies have also raised the issue of a balancing toxicity against their potential efficacy benefits. For patients with T790M-positive NSCLC, third-generation EGFR TKIs are the most appropriate strategy. In this case, genomic guidance of treatment could be performed in the near future through a liquid biopsy. Chemotherapy remains the standard of care for all patients, particularly for T790M-negative patients and other targetable resistance mechanisms. EGFR TKIs should be interrupted during chemotherapy doublet treatment. Immunotherapy is a promising strategy in EGFR-mutant patients currently being addressed for implementation in the clinic in the future.