Key points

Introduction

Discovery of epidermal growth factor receptor (EGFR) sensitizing mutations in 2004 changed the treatment paradigm for advanced non–small cell lung cancer (NSCLC). At that time, combination chemotherapy resulted in modest improvements in patient outcomes and had reached a therapeutic plateau with a median survival of approximately 8 to 10 months. Although most patients present with advanced stage disease, one study of the National Cancer Database showed that 25% of all patients diagnosed with stage IV NSCLC from 2000 to 2008 did not receive any cancer-directed treatment. Many of these patients who are diagnosed at a median age of 70 years have multiple comorbidities or reduced functional status and are not offered therapy because of concerns that they will not tolerate it in light of its limited benefits.

The molecular characterization of NSCLC has provided novel therapeutic targets that are amenable to targeted therapies. The development of EGFR tyrosine kinase inhibitors (TKIs) resulted from the observation that malignant cells overexpress EGFR compared with benign neighboring cells. Activation of EGFR on the cell surface was found to be associated with cell proliferation, angiogenesis, invasion, metastasis, and an ability to escape apoptosis. However, only 10% to 20% of unselected patients responded to EGFR TKIs after chemotherapy in initial studies. Of these patients, east Asians, women, and never smokers with adenocarcinoma were more likely to achieve partial response with EGFR inhibitor therapy. It was later elucidated that ∼80% to 90% of these responses were related to the 2 most common activating mutations: exon 19 deletion (del 19) and exon 21 L858R point mutation, which affect the ATP (Adenosine triphosphate) EGFR binding sites. Although EGFR mutated NSCLCs are exquisitely sensitive to TKIs and have led to improvements in progression-free survival (PFS) compared with standard first-line chemotherapy, patients inevitably progress. Several resistance mechanisms have been identified, with the most common being the emergence of secondary T790M mutations and bypass pathways. In an effort to overcome resistance, second-generation/third-generation TKIs have been developed and have led to the US Food and Drug Administration (FDA) approval of the first T790M-targeted TKI, osimertinib. However, as understanding of the complexity of resistance mechanisms improves, so does the need for better techniques to detect clinically relevant targets and to treat them. This article discusses the evidence to support current clinical recommendations for EGFR mutated advanced NSCLC, emerging resistance mechanisms, and strategies to treat them.

Introduction

Discovery of epidermal growth factor receptor (EGFR) sensitizing mutations in 2004 changed the treatment paradigm for advanced non–small cell lung cancer (NSCLC). At that time, combination chemotherapy resulted in modest improvements in patient outcomes and had reached a therapeutic plateau with a median survival of approximately 8 to 10 months. Although most patients present with advanced stage disease, one study of the National Cancer Database showed that 25% of all patients diagnosed with stage IV NSCLC from 2000 to 2008 did not receive any cancer-directed treatment. Many of these patients who are diagnosed at a median age of 70 years have multiple comorbidities or reduced functional status and are not offered therapy because of concerns that they will not tolerate it in light of its limited benefits.

The molecular characterization of NSCLC has provided novel therapeutic targets that are amenable to targeted therapies. The development of EGFR tyrosine kinase inhibitors (TKIs) resulted from the observation that malignant cells overexpress EGFR compared with benign neighboring cells. Activation of EGFR on the cell surface was found to be associated with cell proliferation, angiogenesis, invasion, metastasis, and an ability to escape apoptosis. However, only 10% to 20% of unselected patients responded to EGFR TKIs after chemotherapy in initial studies. Of these patients, east Asians, women, and never smokers with adenocarcinoma were more likely to achieve partial response with EGFR inhibitor therapy. It was later elucidated that ∼80% to 90% of these responses were related to the 2 most common activating mutations: exon 19 deletion (del 19) and exon 21 L858R point mutation, which affect the ATP (Adenosine triphosphate) EGFR binding sites. Although EGFR mutated NSCLCs are exquisitely sensitive to TKIs and have led to improvements in progression-free survival (PFS) compared with standard first-line chemotherapy, patients inevitably progress. Several resistance mechanisms have been identified, with the most common being the emergence of secondary T790M mutations and bypass pathways. In an effort to overcome resistance, second-generation/third-generation TKIs have been developed and have led to the US Food and Drug Administration (FDA) approval of the first T790M-targeted TKI, osimertinib. However, as understanding of the complexity of resistance mechanisms improves, so does the need for better techniques to detect clinically relevant targets and to treat them. This article discusses the evidence to support current clinical recommendations for EGFR mutated advanced NSCLC, emerging resistance mechanisms, and strategies to treat them.

Characteristics of epidermal growth factor receptor mutated subsets

Adenocarcinoma of the lung is the most common subtype, representing about 50% of all NSCLCs. The US Lung Cancer Mutation Consortium showed that 64% of these patients have an oncogenic mutation, most of which are mutually exclusive. EGFR represents the second most common mutation at ∼15% after Kirsten rat sarcoma (KRAS), but it is the most clinically relevant because of the availability of FDA-approved targeted drugs. In Asians, EGFR mutations are even more common and represent 30% to 40% of the population. EGFR mutated NSCLC represents 40% to 60% of never smokers and 15% to 30% of former or current smokers. Less than 5% of squamous cell cancers have EGFR mutations, which are more common in adenosquamous carcinomas. Therefore, all patients with nonsquamous cancers, and never/light smokers, should have molecular profiling.

EGFR mutations are located in exons 18 to 21, which encode the ATP binding site of the tyrosine kinase domain. At present, 2 reversible ATP competitive TKIs (gefitinib, erlotinib) and 1 irreversible TKI (afatinib) are approved in the first-line setting to treat EGFR mutated NSCLC. However, not all mutations seem to have the same sensitivity to TKIs. Patients with del 19 treated with EGFR TKIs seem to have a better outcome compared with those with exon 21 mutation. Some of the other less common EGFR mutations also seem susceptible to EGFR TKIs. The most prevalent of these rarer mutations include exon 18 (G719X), exon 19 insertions, exon 20 insertions (20 ins), de novo exon 20 T790M, exon 20 Ser768I, exon 21 (L861Q), and combined mutations. In an analysis of 123 Chinese patients with NSCLC and uncommon mutations, it was noted that although TKI therapy prolonged overall survival (OS) in these patients, the PFS and overall response rate (ORR) were highest for del 19 or L858R combinations either with other mutations (9.53 months, 55%) or each other (9.79 months, 71.4%) and poor with 20 ins (2 months, 8.3%) or de novo T790M even if combined with del 19/L858R (1.94 months, 22%). T790M and most of the Exon 20 ins tend to confer resistance with poorer outcomes. Although numerous mutations have been identified, the numbers of patients in these reports are too small to make broad statements about efficacy of TKIs for each of them.

Epidermal growth factor receptor tyrosine kinase inhibitor monotherapy as first-line therapy

Nine randomized phase III trials have shown that EGFR TKIs surpass standard first-line platinum chemotherapy in objective response rate, PFS, and quality-of-life measures for patients with activating EGFR mutations, but have failed to show a difference in OS ( Table 1 ). The lack of OS benefit has been attributed to postprogression therapy with TKIs for patients who receive first-line chemotherapy. In these studies, OS for patients treated with a TKI ranged from 19.3 to 35.5 months. Although most of these trials evaluated patients with del 19 and L858R mutations, the IPASS ((IRESSA (gefitinib) Pan-Asia Study), NEJ002 (North-East Japan Study), LUX-LUNG 3, and LUX-LUNG 6 studies also included patients with uncommon mutations, although the proportion of these was low (∼10%).

The IPASS study was the first phase III study to prospectively show the predictive role of EGFR mutations in determining therapeutic outcomes. The trial enrolled 1217 east Asian patients who were light/never smokers with adenocarcinoma, 60% of whom had EGFR mutations, and randomized them to gefitinib or carboplatin/paclitaxel. Despite an initial benefit to chemotherapy, the Kaplan-Meier curves crossed and gefitinib significantly improved PFS at 12 months (25% vs 7%). The preplanned subgroup analysis revealed striking differences in the efficacy of gefitinib, with an ORR of 71% in EGFR mutated patients versus 1% in EGFR mutation–negative patients. Among patients with EGFR mutations, 95% had del 19 or L858R, and 4% had other less common mutations. In EGFR mutated patients, gefitinib prolonged PFS (9.6 vs 6.3 months; hazard ratio [HR], 0.48; P <.001), but in wild-type (wt) patients the PFS was inferior compared with chemotherapy (1.5 vs 6.5 months; HR, 2.85; P <.001). In addition, quality of life was significantly better for those who received gefitinib, with the main side effects being rash, diarrhea, and increase in liver transaminase levels compared with more nausea/vomiting, hematologic toxicities, and neuropathy in the carboplatin/paclitaxel group.

Three other east Asian phase III studies of gefitinib compared with chemotherapy showed similar results. Though the First-Signal trial (First-line single-agent Iressa (gefitinib) versus gemcitabine and cisplatin trial in never-smokers with adenocarcinoma of the lung) also enrolled patients based on clinical characteristics, both the WJTOG 3405 (West Japan Thoracic Oncology Group) and NEJ02 included only patients with EGFR mutations. The magnitude of benefit in favor of EGFR TKIs was consistently higher regardless of whether cisplatin-based or carboplatin-based regimens were used in the comparator arm.

OPTIMAL (erlotinib vs chemotherapy in the first-line treatment of patients with advanced EGFR mutation-positive) was the first phase III study to evaluate erlotinib versus first-line chemotherapy. One-hundred and fifty-four Chinese patients with either del 19 or L858R mutations were randomized to erlotinib versus carboplatin/gemcitabine and showed a significant improvement in PFS (HR, 0.16; P <.0001) in favor of erlotinib. These results were confirmed in a European patient population (EURTAC [European Tarceva (erlotinib) versus Chemotherapy]) showing that the superiority of TKIs compared with chemotherapy is not exclusive to east Asians. Similar to gefitinib, the most common adverse events (AEs) with erlotinib were rash, diarrhea, and an increase in transaminase levels.

Afatinib, a second-generation TKI, binds covalently to the EGFR and inhibits the receptor irreversibly in addition to targeting the HER2 receptors. Afatinib was compared with platinum-based chemotherapy in 2 large phase III (LUX-LUNG 3 and LUX-LUNG 6) studies of treatment-naive EGFR mutated patients. LUX-LUNG 3 was unique in that it was the only phase III trial that compared a TKI with cisplatin/pemetrexed as a comparator arm, a regimen that is considered to be a preferred treatment option for lung adenocarcinoma. At the time, maintenance chemotherapy was not allowed because it was not considered standard of care until after accrual completed. In conjunction, the LUX-LUNG 6 study randomized EGFR mutated Asian patients to afatinib or cisplatin/gemcitabine. Common side effects of afatinib included rash, diarrhea, stomatitis, and paronychia and were generally tolerable, although grade 3/4 toxicities were more frequently reported than with first-generation TKIs. Although both studies affirmed that afatinib was superior to chemotherapy in terms of ORR (56%–67% vs 23%) and PFS (11–11.1 months vs 5.6–6.9 months), there was once again no OS advantage in either study individually or combined. However, the pooled analysis of LUX-LUNG 3 and LUX-LUNG 6 showed a significant improvement in OS with afatinib compared with chemotherapy for the del 19 subgroup. Although it was noted that a lower proportion of patients received subsequent therapies and may have contributed to this finding, no difference in postprogression therapy was observed between the mutation subtypes.

Comparison of epidermal growth factor receptor tyrosine kinase inhibitors

Although direct comparisons of gefitinib, erlotinib, and afatinib in the clinic are limited, 3 meta-analyses suggested that efficacy between these agents were not significantly different. Recently, 2 large prospective studies have added some clarity by comparing afatinib with gefitinib in treatment-naive patients and gefitinib with erlotinib as second-line therapy after chemotherapy ( Table 2 ).

In an international phase IIB study, 319 EGFR mutated patients were randomized to afatinib or gefitinib and revealed significant improvements in ORR (70% vs 56%; P = .0083), time to treatment failure (13.7 vs 11.5 months; P = .0073), and PFS (11 vs 10.9 months; HR, 0.73; P = .017) in favor of afatinib. Although median PFS was similar, the difference in PFS was more pronounced over time (47.4% vs 41.3% at 12 months and 17.6% vs 7.6% at 24 months) suggesting a more sustained response. These findings persisted irrespective of mutation subtype. Severe AEs (grade ≥ 3) were higher with afatinib (31% vs 17%). Diarrhea, rash, and fatigue were the most prominent for those taking afatinib as opposed to increased liver transaminase levels, rash, and interstitial lung disease (ILD) (3%) for those on gefitinib. Although afatinib showed modest improvement in outcomes compared with gefitinib, this was at the expense of increase toxicity.

The phase III study comparing gefitinib with erlotinib was initially designed as a noninferiority trial for unselected patients in the second-line setting, and hence it did not meet its primary PFS end point (6.5 vs 7.5 months; HR, 1.125; 95% confidence interval [CI], 0.94–1.35) with an OS of 22.8 versus 24.5 months (HR, 1.04; 95% CI, 0.83–1.29). Of the 516 patients enrolled, 401 had EGFR mutations and revealed a similar PFS for the 2 TKIs (8.3 vs 10 months; P = .424) in this subgroup. Although patients with del 19 seemed to have a longer PFS compared with patients with L858R, gefitinib and erlotinib proved equally effective in both mutation subtypes. Side effects for both drugs were manageable, but grade 3/4 increases in liver transaminase levels were more frequent with gefitinib as opposed to more frequent grade 3/4 rash with erlotinib. The incidence of ILD was similar (4%). Overall, gefitinib fared better in terms of worst grade per patient toxicity ( P <.01). In one of the largest studies to date, gefitinib and erlotinib had comparable efficacy with a side effect profile that slightly favored gefitinib. A recent phase II study comparing gefitinib with erlotinib in EGFR mutated patients showed similar results and comparable toxicities between the two agents. However, note that erlotinib significantly prolonged PFS in the subset of patients treated second line or beyond (11.4 vs 7.9 months; HR, 0.58; P = .015).

In a pooled analysis of 2 randomized studies, dacomitinib, a second-generation irreversible TKI, was compared with erlotinib in previously treated patients after chemotherapy. Among the 101 patients with advanced NSCLC that had a common EGFR sensitizing mutation, PFS for dacomitinib was 14.6 months versus 9.6 months for erlotinib (HR, 0.77; P = .146). ORR was similar (67.9%; CI, 53.7%–80.1% vs 64.6%; CI, 49.5%–77.8%), but seemed to be slightly higher for those patients with del 19 treated with dacomitinib compared with patients with L858R. Dacomitinib had a higher incidence of diarrhea, paronychia, mucositis, and rash. Despite a PFS trend in favor of dacomitinib, further randomized studies are needed to evaluate its role in treatment of EGFR mutated patients. A phase II trial comparing gefitinib with erlotinib, a phase III trial comparing gefitinib with second-generation TKI (dacomitinib), and 2 phase III trials comparing third-generation TKIs (AZD9291, ASP8273) with erlotinib/gefitinib are ongoing.

Acquired resistance mechanisms and predictive biomarkers

Although first-line TKIs have proved effective in blocking an oncogenic addicted driver mutation, progression generally occurs at approximately 1 year in most patients. Escape mechanisms include newly acquired gatekeeper EGFR mutations (T790M), bypass pathways (MET/HER2 [MNNG HOS Transforming gene/human epidermal growth factor receptor] amplification), activation of downstream pathways (PIK3CA [phosphatidylinositol 3′-kinase], BRAF (v-raf murine sarcoma viral oncogene homolog B) mutations), and histologic transformation to small cell. Of those who progress on a first-generation TKI, 50% to 60% have an acquired exon 20 T790M mutation that enhances EGFR’s affinity for ATP more than TKIs. Although tumorigenesis may develop from an activating mutation, recent evidence suggests that, in contrast, resistance often develops from emerging subclones, which leads to more tumor heterogeneity. As such, EGFR sensitizing mutations have consistently been found across biopsy sites, but detection of T790M seems more discordant. Piotrowska and colleagues found that T790M wt and T790M-positive clones coexisted in a pretreated specimen compared with post–TKI treatment cells that still retained their original EGFR activating mutation despite cell variability in T790M expression. Furthermore, they also noted that those patients who had higher proportions of T790M cells compared with T790Mwt on pretreatment biopsies had a better tumor response to third-generation TKIs. However, others have suggested that, in addition to subclones, drugs can still promote the development of novel T790M mutations. Ultimately, it is likely that multiple mechanisms are at play within an individual, accounting for intertumor and intratumor variability, as seen when autopsy specimens were analyzed.

Tumor heterogeneity can have broad clinical implications for determining the next steps in a patient’s treatment and for future clinical trials. With osimertinib, a third-generation TKI now available for T790M mutated patients, it has become standard of care to rebiopsy patients on progression to determine their T790M status. However, it is clear that a single biopsy may not fully reflect the resistance mechanisms at play. Researchers have shown that detection of T790M may vary with sequential biopsies with no clear correlation with TKI therapy and can subsequently be negative despite an initial positive biopsy. In other cases, a biopsy may not even be feasible or tissue may be inadequate for testing. Therefore, alternative noninvasive methods are being investigated in the form of circulating cell free DNA (cfDNA) and circulating tumor cells (CTCs).

cfDNA is used to detect cell fragments in the blood and hence it may be more representative of the disease burden as a whole. Recently, the cobas EGFR mutation test v2 was FDA approved to prescreen plasma for exon del 19 and L858R mutations based on the ENSURE trial (First-line erlotinib versus gemcitabine/cisplatin in patients with advanced EGFR mutation-positive), which showed a sensitivity of 77% and specificity of 98% to tissue testing. A meta-analysis of 3110 patients determined that cfDNAs have a high specificity (0.96; CI, 0.93–0.98), but a moderate sensitivity (0.62; CI, 0.51–0.72) for detecting EGFR mutations compared with tissue. Although the false-negative rate improved in patients with advanced-stage disease, it seemed that chemotherapy also affected its accuracy. A separate concurrent analysis of T790M detection in cfDNA, CTCs, and tissue biopsies from 40 patients who had progressed on a prior TKI showed that cfDNA and CTCs each missed 20% to 30% of the mutations. However, both tests combined had an even higher rate of detection than the matched tissue biopsies, implying that multimodality testing may be warranted for a complete assessment. Such clinically relevant pretreatment differences between plasma and tissue detection of T790M were noted in patients treated with a third-generation inhibitor, rociletinib. Fifty-eight percent of patients with wt T790M and 69% of patients with inadequate tissue were found to have cfDNA-positive T790M. In contrast, 19% of patients were found to have positive tissue testing that was not confirmed by cfDNA. Of those patients with urine specimens available, 68% of T790Mwt and 72% of patients with inadequate tissue were found to have urine positive for T790M, whereas 18% of patients had T790M-positive tissue testing that was not confirmed by the urine. Overall, 96% of patients were T790M positive by at least 1 method. Detection by blood or urine was better for stage IV patients who had extrathoracic metastatic disease, although the difference was only statistically significant for plasma testing. Regardless of the method of T790M detection used, the ORR (34% by tissue, and 32% by plasma, 37% by urine), duration of response (DOR) (8.7 by tissue, 7.2 by plasma, 8.7 by urine), and PFS (5 months by tissue, 4.1 months by plasma, 4.3 months by urine) were similar.

Using cfDNA as a method to monitor therapy may be just as important as initial detection of predictive biomarkers. In the FAST-ACT 2 trial (First-line Asian Sequential Tarceva (erlotinib) and Chemotherapy), cfDNA EGFR mutation–positive patients who had undetectable cfDNA mutation level after 3 cycles of chemotherapy combined with erlotinib had an improvement in PFS (12 vs 7.2 months; HR, 0.31; P <.0001) and OS (31.9 vs 18.2 months; HR, 0.51; P = .0066). Other reports have not only observed similar correlations between response and plasma levels but have also noted early markers of resistance such as T790M before radiologic progression. One report of cfDNAs from 57 patients revealed that suppression of activating mutations and the increase of T790M mutations correlated with response and progression, respectively. Because these cfDNA platforms are now becoming widely available, its role for patient care needs to be better defined.