Skip to main content

14-08-2018 | Non-small-cell lung cancer | Article

Making the first move in EGFR-driven or ALK-driven NSCLC: first-generation or next-generation TKI?

Nature Reviews Clinical Oncology

Authors: Gonzalo Recondo, Francesco Facchinetti, Ken A. Olaussen, Benjamin Besse, Luc Friboulet

Publisher: Nature Publishing Group UK


The traditional approach to the treatment of patients with advanced-stage non-small-cell lung carcinoma (NSCLC) harbouring ALK rearrangements or EGFR mutations has been the sequential administration of therapies (sequential treatment approach), in which patients first receive first-generation tyrosine-kinase inhibitors (TKIs), which are eventually replaced by next-generation TKIs and/or chemotherapy upon disease progression, in a decision optionally guided by tumour molecular profiling. In the past few years, this strategy has been challenged by clinical evidence showing improved progression-free survival, improved intracranial disease control and a generally favourable toxicity profile when next-generation EGFR and ALK TKIs are used in the first-line setting. In this Review, we describe the existing preclinical and clinical evidence supporting both treatment strategies — the ‘historical’ sequential treatment strategy and the use of next-generation TKIs — as frontline therapies and discuss the suitability of both strategies for patients with EGFR-driven or ALK-driven NSCLC.


The treatment of patients with lung cancer is rapidly evolving. In the past 20 years, the clinical management of these patients has shifted from a histology-based approach towards a molecularly driven approach, owing to the development of targeted therapies against the driver mutations of this disease, which affect a number of kinases13; this strategy has improved the outcomes for patients, which is important considering the high incidence and mortality of this disease4.
Approximately 50% of Asian patients with non-small-cell lung carcinoma (NSCLC) and 11–16% of patients in Western countries harbour mutations in EGFR, which affect the kinase domain of EGFR57. The majority of these alterations (>90%) are deletions within exon 19 or L858R point mutation8. Genomic rearrangements involving the ALK gene occur in 3–6% of patients with NSCLC9,10. Other genomic alterations (in MET, ROS1, HER2, BRAF, or RET) are less frequent.
In the past decade, the first-generation EGFR tyrosine-kinase inhibitors (TKIs) gefitinib, erlotinib, and icotinib, and the second-generation TKI afatinib were established as standard-of-care first-line therapies for patients with NSCLC harbouring activating mutations in EGFR11. Despite high initial response and disease control rates, virtually all the patients receiving these TKIs eventually experience tumour progression owing to the emergence of therapeutic resistance12. Resistance to TKIs is most commonly acquired de novo during treatment, but can also occur owing to the outgrowth of pre-existing resistant subclones13. In approximately 50% of patients, resistance was mediated by the acquisition of the ‘gatekeeper’ mutation T790M, which results in sterical blockade of first-generation or second-generation TKI binding and also increases the kinase affinity for ATP1417. Osimertinib is an irreversible third-generation EGFR TKI that is active against exon 19 deletions and L858R mutation, regardless of the presence of T790M mutation18. This TKI forms a covalent bond to the cysteine residue at position 797 and has lower activity than the aforementioned TKIs against wild-type EGFR protein. Osimertinib was initially approved by the FDA and European Medicines Agency (EMA) as the standard-of-care treatment for patients with tumours harbouring the EGFRT790M mutation after progression upon treatment with a first-line EGFR TKI1921.
Since 2011, the first-generation TKI crizotinib has been the frontline treatment for NSCLC harbouring translocations involving ALK22. As with EGFR TKIs, all patients ultimately develop resistance to this agent, and secondary point mutations in the kinase domain are responsible for drug resistance in approximately 20% of patients23. Unlike mutations causing EGFR resistance, a diverse range of mutations in ALK affect the kinase domain, and their incidence increases to 56% with sequential exposure to ALK TKIs23. Ceritinib, alectinib, and brigatinib are second-generation ALK inhibitors with activity against a wide spectrum of secondary resistance mutations affecting the ALK kinase domain2426. These TKIs were first developed in the setting of crizotinib resistance, in which they had shown potent activity in preclinical studies2426. Similarly, lorlatinib, a third-generation ALK inhibitor, has been developed to be administered after progression following treatment with first-generation and/or second-generation TKIs27. In this Review, ‘next-generation TKI’ refers to the third-generation EGFR TKI osimertinib, the second-generation ALK TKIs ceritinib, alectinib, and brigatinib, and the third-generation ALK TKI lorlatinib.
In the ‘historical’ sequential treatment approach, patients with NSCLC receive frontline therapy with a first-generation TKI and ‘switch’ to next-generation TKIs and/or chemotherapy upon disease progression. In 2017, however, next-generation inhibitors have emerged as treatment options in the first-line setting, on the basis of the increased efficacy observed when directly compared with historical first-line TKIs2830. The lack of comparative survival outcomes has hampered the elucidation of the most beneficial strategy for patients in the long term. Herein, we present the evidence currently available on the antitumour activity of EGFR and ALK TKIs, reported in both preclinical and clinical studies, and discuss the advantages and drawbacks of both strategies for patients with EGFR-driven or ALK-driven NSCLC.

Historical approach: sequential treatment


The publication of two studies in 2004 (refs31,32) describing the predictive value of sensitizing mutations in EGFR on the activity of EGFR inhibitors is a key landmark in the development of potent drugs to treat molecularly selected patients with NSCLC31,32. Multiple phase III trials comparing the first-generation EGFR TKIs erlotinib, gefitinib, or icotinib, as well as the second-generation TKI afatinib, with platinum-based chemotherapy as frontline therapies for patients with advanced-stage disease have been reported3350 (Table 1). A consistent benefit in favour of EGFR TKIs is observed across studies in terms of progression-free survival (PFS), response rates, and disease control rates. The median PFS with these compounds ranged from 8.0–13.1 months, compared with 4.6–6.9 months with chemotherapy (range of HRs 0.16–0.48). Given this impressive PFS benefit, an important overall survival benefit was expected51. Nevertheless, median overall survival durations were equivalent for both trial arms across studies (19.3–34.8 months), predominantly owing to the high rates of treatment crossover (54−95%). The findings of these studies also provided the first demonstration that, in the context of oncogene addiction, the clinical benefit derived from treatment with TKIs is independent of whether the patients were treated upfront or after first-line chemotherapy.
Table 1
Clinical trials testing EGFR TKIs in sequential strategy
Trial design (phase, primary end point and treatment arms, including number of patients harbouring EGFR mutations)a
Median follow-up duration (months)
Outcomes (ORR, median PFS and median OS)
First generation
• Gefitinib (n = 132) versus carboplatin + paclitaxel (n = 129)
• 71.2% versus 47.3%
• 9.5 mo versus 6.3 mo (HR 0.48; P < 0.001)
• 21.6 mo versus 21.9 mo (HR 1.00; P = 0.99)
33, 34
• OS
• Gefitinib (n = 26) versus cisplatin + gemcitabine (n = 16)
• 84.6% versus 37.5%
• 8 mo versus 6.3 mo (HR 0.54; P = 0.086)
• 27.2 mo versus 25.6 mo (HR 1.04)
• Gefitinib (n = 86) versus cisplatin + docetaxel (n = 86)
34 (59.1 for OS analysis)
• 62.1% versus 32.2%
• 9.2 mo versus 6.3 mo (HR 0.49; P < 0.0001)
• 34.8 mo versus 37.3 mo (HR 1.25)
36, 37
• Gefitinib (n = 114) versus carboplatin + paclitaxel (n = 114)
• 73.7% versus 30.7%
• 10.8 mo versus 5.4 mo (HR 0.30; P < 0.001)
• 27.7 mo versus 26.6 mo (HR 0.89; P = 0.48)
38, 39
• Erlotinib (n = 82) versus carboplatin + gemcitabine (n = 72)
• 83% versus 36%
• 13.1 mo versus 4.6 mo (HR 0.16; P < 0.0001)
• 22.8 mo versus 27.2 mo (HR 1.19; P = 0.27)
40, 41
• Erlotinib (n = 110) versus cisplatin + gemcitabine (n = 107)
28.9 (erlotinib arm) and 27.1 (chemotherapy arm)
• 62.7% versus 33.6%
• 11 mo versus 5.5 mo (HR 0.34; P < 0.0001)
• 26.3 mo versus 25.5 mo (HR 0.91; P = 0.61)
• Erlotinib (n = 86) versus platinum + gemcitabine or paclitaxel (n = 87)
18.9 (erlotinib arm) and 14.4 (chemotherapy arm)
• 63.6% versus 17.8%
• 9.7 mo versus 5.2 mo (HR 0.37; P < 0.0001)
• 19.3 mo versus 19.5 mo (HR 1.04; P = 0.87)
• II
• Erlotinib + bevacizumab (n = 109)
• 77%
• 13.2 mo whole cohort; 16.0 mo T790M+
• 28.2 months
• II
• Erlotinib + bevacizumab (n = 75) versus erlotinib (n = 77)
• 69% versus 64%
• 16 mo versus 9.7 mo (HR 0.54; P = 0.0015)
• NA
CTONG 0901
• Erlotinib (n = 128) versus gefitinib (n = 128)
• 56.3% versus 53.3%
• 13.0 mo versus 10.4 mo (HR 0.81, P = 0.11)
• 22.9 mo versus 20.1 mo (HR 0.84; P = 0.25)
• Icotinib (n = 148) versus cisplatin + pemetrexed (up to four cycles) eventually followed by pemetrexed maintenance (n = 137)
18 (icotinib arm) and 15.7 (chemotherapy arm)
• NR
• 11.2 mo versus 7.9 mo (HR 0.61; P = 0.006)
• 30.5 mo versus 32.1 mo (P = 0.89)
• PFS (non-inferiority in full data set)
• Icotinib (n = 29) versus gefitinib (n = 39)
• 62.1% versus 53.8%
• 7.8 mo versus 5.3 mo (HR 0.78; P = 0.32)
• 20.9 mo versus 20.2 mo (HR 1.1; P = 0.76)
Second generation
LUX-Lung 3
• Afatinib (n = 230) versus cisplatin + pemetrexed (n = 115)
• 56% versus 23%
• 11.1 mo versus 6.9 mo (HR 0.58: P = 0.001)
• Whole cohort: 28.2 mo versus 28.2 mo (HR 0.88; P = 0.39)
• Exon 19 deletion: 33.3 mo versus 21.1 mo (HR 0.54; P = 0.002)
48, 50
LUX-Lung 6
• Afatinib (n = 242) versus cisplatin + gemcitabine (n = 122)
• 66.9% versus 23%
• 11.0 mo versus 5.6 mo (HR 0.28; P < 0.0001)
• 23.1 mo versus 23.5 mo (HR 0.93; P = 0.61)
49, 50
LUX-Lung 7
• PFS, TTF and OS
• Afatinib (n = 160) versus gefitinib (n = 159)
• 70% versus 56%
• 11.0 mo versus 10.9 mo (HR 0.73; P = 0.017)
• 27.9 mo versus 24.5 mo (HR 0.86; P = 0.26)
53, 54
• IRC-assessed PFS
• Dacomitinib (n = 227) versus gefitinib (n = 225)
• 75% versus 70%
• 14.7 mo versus 9.2 mo (HR 0.59; P < 0.0001)
• 34.1 mo versus 26.8 mo (HR 0.76; P = 0.0438)
55, 56
Third generation
AURA (dose-escalation and expansion cohorts)
• I
• Safety and efficacy
• Osimertinib
• First line (n = 60 patients), second line or beyond (n = 193)
19.1 and NA
• 77% and 61%
• 20.5 mo and 9.6 mo
• NA
18, 120
AURA (extension cohort)
• Phase II
• Osimertinib (n = 201)
• Second line or beyond, prior treatment with erlotinib (58%), gefitinib (58%) and/or second-generation EGFR TKI (24%)
• 62%
• 12.3 mo
• Pooled analysis OS: 26.8 mo
• Median treatment exposure: 16.4 mo
57, 59
• Phase II
• Osimertinib (n = 210)
• Second line or beyond, prior treatment with erlotinib (57%), gefitinib (58%) and/or second-generation EGFR TKI (20%)
• 70%
• 9.9 mo
• Pooled analysis OS: 26.8 mo
• Median treatment exposure: 16.4 mo
58, 59
• Phase III
• Osimertinib (n = 279) versus platinum + pemetrexed (n = 140)
• Second line, prior treatment with gefitinib (59%), erlotinib (34%) or afatinib (7%)
• 71% versus 31%
• 10.1 mo versus 4.4 mo (HR 0.30; P < 0.001)
• NA
IRC, independent review committee; mo, months; NA, not available; NR, not reported; ORR, overall response rate; OS, overall survival; PFS, progression-free survival; T790M+, patients with NSCLC harbouring T790M mutation in EGFR; TKI, tyrosine-kinase inhibitor; TTF, time to treatment failure. aLine of treatment only stated for third-generation inhibitors; all the first-generation and second-generation inhibitors were tested in the first-line setting.
For the treatment of patients with the most frequent EGFR mutations (L858R and exon 19 deletions), the choice of a first-generation or second-generation EGFR inhibitor depends on the physician’s preference, the toxicity profile, and the local availability of each agent. No differences in the efficacy of erlotinib, gefitinib, or afatinib in terms of PFS and overall survival have been detected in comparative studies (CTONG 0901 (ref.52) and LUX-Lung 7 (refs53,54)). Icotinib has been demonstrated to be non-inferior to gefitinib, leading to its approval in 2014 in China as a frontline treatment for patients with advanced-stage EGFR-mutant NSCLC but its development in Western countries was not pursued45. In the ARCHER 1050 trial, dacomitinib, another second-generation irreversible EGFR TKI, was associated with longer PFS and overall survival durations than gefitinib (34.1 months versus 26.8 months, HR 0.76; P = 0.044)55,56 (Table 1). This improvement was achieved at the cost of higher toxicity (frequency of grade 3 adverse events 63% versus 41%) and a detrimental effect on quality of life (QOL)55. Similarly, the addition of erlotinib to bevacizumab extended PFS duration for an average of 6 months compared with erlotinib monotherapy47, although again at the expense of increased toxicity (frequency of grade 3 adverse events 91% versus 53%); the combination regimen was approved by the EMA in 2016 as a first-line treatment option46.
For patients treated with first-line EGFR TKIs, blood-based and/or tumour sampling analysis upon disease progression is mandatory to study the T790M mutational status, owing to the clinical benefits shown for patients in this subgroup who received sequential treatment with osimertinib in multiple studies18,21,5759 (Table 1). In the AURA 3 randomized phase III trial21, for example, osimertinib was associated with better median PFS durations and overall response rates (ORRs) than cisplatin plus pemetrexed in the second-line setting (Table 1). In comparison with the chemotherapy regimen, patients receiving osimertinib also had an improved QOL, with better scores for lung cancer symptoms and a lower incidence of grade ≥3 adverse events (23% versus 47%). At a median follow-up duration of 8.3 months, 71% of patients receiving chemotherapy had crossed over to receive osimertinib after disease progression, and the median overall survival had not been reached in either treatment arm. The extended benefit of the sequential administration of a first-generation EGFR TKI followed by osimertinib observed in this study21 drove the approval of this compound for patients with NSCLC harbouring the T790M mutation and disease progression after treatment with first-generation or second-generation EGFR TKIs.


Crizotinib is a first-generation TKI of ALK, MET, and ROS1, and was the first agent to be approved for the treatment of patients with NSCLC harbouring ALK translocations6063. Two randomized phase III trials established the superiority of crizotinib over chemotherapy in patients with advanced-stage NSCLC, either as a first-line therapy22 or in patients with disease progression after receiving a platinum-based regimen64. In the PROFILE 1014 study22, greater response rates and median PFS durations were achieved with crizotinib than with platinum-based therapy (Table 2). Again, no significant differences in overall survival were observed, with a 4-year survival of 56.6% with crizotinib and 49.1% with chemotherapy65. This effect was mostly due to the high crossover rates (84.2%) from crizotinib to the experimental arm. In an exploratory analysis, after adjusting for crossover, the median overall survival was 59.8 months with crizotinib and 19.2 months with chemotherapy.
Table 2
Clinical trials testing ALK TKIs in sequential strategy
Trial design (phase, primary end point and treatment arms, including number of patients and dosing schedule when relevant)a
Median follow-up duration
Outcomes (ORR, median PFS and OS)
First generation
• I
• ORR, DOR, TTR, PFS, 6–12 mo OS, and safety profile
• Crizotinib (n = 149)
• 60.8%
• 9.7 mo
• 1-year OS 74.8%
61, 62
• II
• Crizotinib (n = 1069)
• 54%
• 8.4 mo
• 21.8 mo
• Crizotinib (n = 173) versus pemetrexed or docetaxel (n = 174)
12.2 mo (crizotinib) and 12.1 mo (chemotherapy)
• 65% versus 20%
• 7.7 mo versus 3.0 mo (HR 0.49; P < 0.001)
• 20.3 mo versus 22.8 mo (HR 1.02; P = 0.54)
• Crizotinib (n = 172) versus platinum + pemetrexed (n = 171)
46 mo
• 74% versus 45%
• 10.9 mo versus 7.0 mo (HR 0.45; P < 0.001)
• NR (45.8 mo–NR) versus 47.5 mo (32.2 mo–NR; HR 0.76; P = 0.048)
22, 65
Second generation
• I
• Ceritinib (n = 246)
• First line (33%) or second line after crizotinib (66%)
11.1 mo
• 72% or 56%
• 18.4 mo or 6.9 mo
• NR or 16.7 mo
67, 68
• II
• Ceritinib (n = 140)
• Second line after crizotinib
11.3 mo
• 38.6%
• 5.7 mo
• 14.9 mo
• II
• Ceritinib (n = 124)
• First line
8.3 mo
• 63.7%
• 11.1 mo
• NA
• Ceritinib (n = 189) versus platinum + pemetrexed (n = 187)
• First line
19.7 mo
• 72.5% versus 26.7%
• 16.6 mo versus 8.1 mo (HR 0.55; P < 0.00001)
• NE (29.3 mo–NE) versus 26.2 mo (22.8 mo–NR; HR 0.73; P = 0.056)
• Ceritinib (n = 115) versus pemetrexed or docetaxel (n = 116)
• Second line after crizotinib
16.5 mo
• 39.1% versus 6.9%
• 5.4 mo versus 1.6 mo (HR 0.49; P < 0.0001)
• 18.1 mo versus 20.1 mo (HR 1.00; P = 0.5)
• I/II
• DLT and MTD (phase I) or ORR (phase II)
• Alectinib (n = 46)
• First line
36 mob
• 93.5%
• NR; 3-year PFS: 62%
• NE; 3-year OS: 78%
73, 74
• I/II
• Recommended phase II dose
• Alectinib (n = 47)
• Second line after crizotinib
4.2 mo
• 55%
• NA
• NA
• II
• Alectinib (n = 225; n = 189 evaluable for response)
• Second line after crizotinib
92.3 weeks
• 51.3%
• 8.3 mo
• 29.1 mo
76, 146
• Alectinib (n = 72) versus docetaxel or pemetrexed (n = 35)
• Second line after crizotinib
6.5 mo
• 37.5% versus 2.9%
• 9.6 mo versus 1.4 mo (HR 0.15; P < 0.001)
• 12.6 mo (9.7 mo–NR) versus NR (NR–NR; HR 0.89)
• Recommended phase II dose (phase I) or ORR (phase II)
• Brigatinib (n = 79)
• First line (10%, n = 8), second line after crizotinib (85%, n = 68) or third line after crizotinib and ceritinib (5%, n = 3)
>31 mob
First-line brigatinib (n = 8):
• 100%
• 34.2 mo
• NR (2-year OS 100%)
Brigatinib after crizotinib (n = 71):
• 73%
• 13.2 mo
• 30.1 mo (2-year OS 61%)
78, 79
• II
• Brigatinib 90 mg daily (n = 112) versus brigatinib standard dosec (n = 110)
• Second line after crizotinib
19.6 mo (90 mg daily) or 24.3 mo (standard dose)
• 46% versus 56%
• 9.2 mo versus 15.6 mo
• 29.5 mo (18.2 mo–NR) versus 34.1 mo (27.7 mo–NR)
80, 147
Third generation
• I
• Lorlatinib (n = 41)
• First line (2.4%), second line (34.2%), third line (56.1%) or fourth line (7.3%)
17.4 mo
• 46%
• 9.6 mo (whole cohort), 13.5 mo (second line), and 9.2 mo (third line and beyond)
• NA
66, 81
• II
• Lorlatinib (n = 228)
• First line (13.1%), second line or beyond, prior treatment with crizotinib only (11.8%), crizotinib + chemotherapy (14.1%), non-crizotinib ALK TKI (12.3%), any two ALK TKIs (28.5%), or any three ALK TKIs (20.2%)
• 90%, 69%, 33% or 39%
• NR, NR, 5.5 mo after treatment with ALK inhibitor other than crizotinib, and 6.9 mo after ≥2 lines of ALK TKIs
• NA
66, 81
DLT, dose-limiting toxicity; DOR, duration of response; mo, months; MTD, maximum tolerated dose; NA, not available; NE, not estimable; NR, not reported; ORR, overall response rate; OS, overall survival; PFS, progression-free survival; TKI, tyrosine-kinase inhibitor; TTR, time to treatment recurrence. aLine of treatment stated for second-generation and third-generation inhibitors; all the first-generation inhibitors were tested in the first-line setting. bUpdated presented data from the original publication. c90 mg daily for 7 days and then 180 mg daily.
Sequential treatment strategies with ALK inhibitors have been developed with the aim of extending overall survival durations6163,6681 (Table 2). Unlike EGFR inhibitors, a wide repertoire of ALK TKIs is available for patients with disease progression after treatment with crizotinib; the second-generation ALK TKIs ceritinib, alectinib, and brigatinib have been developed to overcome most resistance mechanisms2426. Treatment with ceritinib was associated with improved outcomes compared with second-line chemotherapy (ORR 39.1% versus 6.9%, and a median PFS gain of ~4 months) in patients with disease relapse after receiving crizotinib and platinum-based chemotherapy72 (Table 2). In the same disease setting, the results of the phase III ALUR trial77 and the phase II ALTA trial76 demonstrated beneficial outcomes with alectinib and brigatinib, respectively (Table 2). The third-generation ALK TKI lorlatinib has activity against resistance mutations arising after treatment with first-generation and/or second-generation TKIs, including the G1202R mutation27. Lorlatinib has been tested in a dose-escalation phase I study66 and in a phase II trial81 (Table 2).
One of the major concerns in the management of patients with ALK-translocated tumours is the high risk of developing brain metastases; 22–33% of patients present with central nervous system (CNS) involvement at diagnosis, and the prevalence of brain metastases increases to 45–70% upon progression on crizotinib treatment68,69,73,75,82. The improved CNS activity of second-generation and third-generation ALK TKIs results from both their higher CNS penetration and increased potency compared with crizotinib27. Intracranial responses have been observed in 45% of patients receiving ceritinib69, 64% of those receiving alectinib83, and 67% treated with brigatinib84. Brigatinib was associated with an intracranial PFS of 18.4 months with the standard dose84. Importantly, even 42% of patients in a heavily pretreated cohort (≥2 lines of ALK TKIs) had intracranial disease control with lorlatinib, and the cerebrospinal fluid concentration documented for lorlatinib was 75% of the plasma concentration66.

Translational studies of resistance

A number of ‘back-to-benchside’ studies have been conducted with the aim of characterizing the mechanisms underlying clinical resistance to EGFR or ALK TKIs. The results from these studies can provide a rationale for optimizing sequential treatment strategies, because a better understanding of the biological implications of therapeutic resistance can guide clinicians to provide the most adequate treatment upon disease progression (Fig. 1).
As discussed, the acquisition of the gatekeeper T790M mutation in EGFR is the most common mechanism of resistance to first-generation EGFR TKIs (detected in 50–60% of patients)12,14,85,86. The activation of ‘bypass’ signalling mechanisms is also relevant in this scenario, and involves potential therapeutic targets, such as MET, AXL, IGF1R, and other members of the EGFR family8790. Resistance to third-generation EGFR TKIs has also been described91: the most common tertiary mutation in EGFR is C797S in 24–40% of patients, which affects the covalent binding site of osimertinib9294. This tertiary mutation can be present in cis or trans with the T790M mutation95. The results of preclinical studies suggest that combinations of brigatinib or other novel EGFR inhibitors with anti-EGFR monoclonal antibodies are an effective treatment option when C797S is present in cis96,97. Resistance dependent on the presence of the tertiary mutation in trans can be overcome by combining first-generation and third-generation EGFR TKIs98,99.
A range of secondary mutations affecting the kinase domain of ALK confer resistance to different ALK TKIs. The following mutations have been implicated in resistance to crizotinib: G1269A, C1156Y, E1210K, I1171T, L1152R, S1206C/Y, I1151T/N/S, F1174C/L/V, V1180L, and L1196M23,100104. F1174C/L/V, 1151Tins, L1152P, and C1156Y mutations are associated with resistance to ceritinib24. Both V1180L and I1171T/N/S alterations confer resistance to alectinib, and double mutations in E1210K and S1206C or D1203N have been reported in patients with resistance to brigatinib23,105. G1202R is the most common resistance mutation emerging on treatment with second-generation ALK inhibitors and is only targetable with lorlatinib23,27,106,107. Interestingly, the acquisition of both the C1156Y and L1198F mutations upon lorlatinib treatment has been reported to resensitize the tumour to crizotinib108. After the description of this initial case report, the results of the first extensive preclinical and clinical study of mutations causing resistance to lorlatinib were published in 2018 by Yoda and colleagues109. Using N-ethyl-N-nitrosourea-generated mutagenesis screening to determine the secondary mutations in ALK that can arise upon lorlatinib treatment, these investigators found that single mutations in ALK cannot cause resistance to lorlatinib. Indeed, only double ALK mutations in cis were detected upon resistance to lorlatinib, both in preclinical experiments and in patient-derived samples. Thus, observations of the stepwise accumulation of resistance mutations in ALK suggest that upfront treatment with lorlatinib could markedly delay the onset of on-target resistance, leading to a more durable clinical benefit than the current sequential treatment approach.
Off-target resistance mechanisms, such as bypass pathway activation, have also been reported in patients with resistance to first-generation and second-generation ALK TKIs23,102,110. The results of preclinical studies revealed that treatment with second-generation ALK TKIs could overcome resistance to crizotinib that develops without the acquisition of secondary mutations in ALK24. This observation mainly suggests that crizotinib has lower inhibitory potency against ALK than do second-generation ALK TKIs, facilitating tumour growth upon modest activation of bypass signalling mechanisms. By contrast, treatment with lorlatinib does not overcome resistance to second-generation TKIs mediated by off-target mechanisms23,27. On the basis of these observations, bypass mechanisms involving robust oncogenic pathways, such as MAP2K1, SRC, EGFR, or PI3K, that are activated upon treatment with second-generation ALK TKIs have been proposed to also drive resistance to third-generation ALK TKIs23.
A series of laboratory studies have focused on the brain penetration of both EGFR and ALK TKIs. In studies using mouse models, alectinib was superior to crizotinib in controlling metastatic disease in the CNS111; moreover, responses to lorlatinib were observed even in mice with disease progression after alectinib treatment27. Importantly, evidence from several of these preclinical studies suggests that next-generation TKIs provide optimal long-term outcomes when used as frontline treatments19,26,27.
Other preclinical studies were aimed at providing a biological rationale to explain systematic relapse in patients treated with TKIs despite major initial responses. Several studies have shown that a small subpopulation of tumour cells (<5%) cultured in the presence of a TKI remain alive and are reprogrammed into a drug-tolerant state112117. These cells, with limited or no growth during months of TKI treatment, are referred to as ‘persister’ cells and provide a reservoir of cells from which drug-resistance mechanisms could emerge. Initial studies have suggested that epigenetic reprogramming of TKI-persister cells involves the histone demethylase KDM5A and thus could be selectively targeted by histone deacetylase inhibitors112. The results of preclinical studies indicate that persister cells can later cause tumour regrowth through the de novo acquisition of diverse genetically driven resistance mechanisms, such as secondary mutations or activation of bypass signalling113,115. Eradicating persister cancer cells early during the course of treatment might therefore block or drastically postpone the onset of resistance. Persister cells display an impaired apoptotic response to TKI (as assessed by annexin V staining)115, and, thus, treatment with inhibitors of the BCL-2 family anti-apoptotic proteins has been proposed to be a potentially effective therapeutic strategy; the combination of osimertinib and navitoclax is currently being tested in patients with NSCLC harbouring the EGFR T790M mutation (NCT02520778)115. Two studies with results published in 2017 revealed a common persister-cell-specific dependency on the lipid hydroperoxidase GPX4, targeting of which prevented tumour relapse in mice116,117.
Finally, tumour heterogeneity occurs early in the course of cancer progression: in patients with resectable NSCLC, a median of 30% of the somatic mutations detected are subclonal118. Tumour heterogeneity is an important factor contributing to the development of therapeutic resistance because it contributes to both the selective expansion of pre-existing resistant clones and the adaptive resistance of persister tumour cells115. In patients with NSCLC harbouring EGFR mutations and with disease progression after a first-generation or second-generation TKI, the allelic fraction of T790M mutations can, for instance, affect the therapeutic response to third-generation EGFR TKIs119. Observations in patients treated with osimertinib95 or lorlatinib109 indicate that clones resistant to third-generation TKIs can emerge upon sequential treatment with first-generation and second-generation EGFR or ALK TKIs, affecting the choice of the next optimal treatment strategy. In line with these observations, preclinical and clinical studies performed during first-line treatment with third-generation ALK and EGFR TKIs revealed that the emergence of resistance driven by on-target mutations can be delayed19,27,120. Mice bearing EGFR19 and ALK27 TKI-sensitive tumours treated with first-generation and third-generation inhibitors showed prolonged tumour responses and delay of resistance with third-generation TKIs. In two cohorts of patients with EGFR-mutated NSCLC treated with upfront osimertinib in the phase I AURA study, none of the evaluable patients had disease progression owing to T790M mutation120. Overall, in addition to enabling the interpretation of the outcomes of clinical studies, the studies discussed herein highlight the importance of characterizing the molecular mechanisms of resistance to TKIs during or after each line of treatment using blood or tissue sampling to inform clinical decision-making.

Paradigm shift for first-line therapy

The historical trend in the management of patients with cancer has been to move more-potent, more-specific, and possibly less-toxic drugs to the first-line treatment setting. Similarly to chemotherapy, the magnitude of efficacy of next-generation TKIs generally increases in accordance with an earlier administration during the course of treatment with targeted therapies21,28,29,71,72,77. Indeed, several single-arm early phase trials in patients with NSCLC who had not received any previous TKI showed prolonged disease control upon first-line treatment with osimertinib120, ceritinib70, or alectinib74 (in comparison with data available for first-generation and second-generation TKIs). In 2017, additional evidence of major PFS benefits emerged from three phase III trials, supporting the upfront use of next-generation TKIs over the standard first-line EGFR TKIs and crizotinib (Table 3).
Table 3
Clinical trials comparing first-generation and next-generation TKIs in the frontline setting
Trial design (phase, primary end point and treatment arms, including number of patients and dosing schedule when relevant)
Median follow-up duration
Outcomes (ORR, median investigator-assessed PFS, median IRC-assessed PFS, OS and grade ≥3 AEs)
• Investigator-assessed PFS
• Alectinib (n = 152; 600 mg b.i.d.) versus crizotinib (n = 151)
22.8 mo (alectinib arm) and 27.8 mo (crizotinib arm)a
• 82.9%a versus 75.5%
• 25.7 mo (95% CI 19.9 mo–NE) versus 10.4 mo (95% CI 7.7–14.6 mo; HR 0.50; P < 0.001); 34.8 moa versus 10.9 mo (HR 0.43; 95% CI 0.32–0.58)
• 1-year OS 84.3% versus 82.5% (HR 0.76; P = 0.24)
• 44.7%a versus 51%
29, 121
• IRC-assessed PFS
• Alectinib (n = 103; 300 mg b.i.d.) versus crizotinib (n = 104)
12 mo (alectinib arm) and 12.2 mo (crizotinib arm)
• 92% versus 79%
• NA; HR 0.34 (95% CI 0.21–0.55)
• Not reached (95% CI 20.3 mo–NE) versus 10.2 mo (95% CI 8.2–12.0 mo; HR 0.34; P < 0.0001)
• NA (immature data)
• 26% versus 52%
• Investigator-assessed PFS
• Osimertinib (n = 279) versus gefitinib or erlotinib (n = 277)
15 mo (osimertinib arm) and 9.7 mo (first-generation TKI arm)
• 80% versus 76%
• 18.9 mo versus 10.2 mo (HR 0.46; P < 0.001)
• 17.7 mo versus 9.7 mo (HR 0.45; P < 0.001)
• 18 mo OS 83% versus 71% (HR 0.63; P = 0.007, nonsignificant owing to immature data)
• 34% versus 45%
AE, adverse event; b.i.d., twice daily; mo, months; IRC, independent review committee; NA, not available; NE, not estimable; ORR, overall response rate; OS, overall survival; PFS, progression-free survival; TKI, tyrosine-kinase inhibitor.a Updated data.


In the randomized phase III FLAURA study28, osimertinib was compared as a frontline therapy with the standard choice of gefitinib or erlotinib in patients with NSCLC harbouring EGFR exon 19 deletions or L858R point mutation28. As expected, the median PFS was significantly prolonged by almost 9 months with osimertinib compared with first-generation TKIs (HR 0.46; P < 0.001), although the ORRs were similar between trial arms (Table 3). The median time to second-line treatment or death was 23.5 months with osimertinib and 13.8 months with first-line EGFR TKI, and the median time to third-line treatment was not reached and 25.9 months, respectively. Brain imaging was mandatory at study entry, as well as during the course of the study for patients with brain metastases; at study entry, 19% of patients in the osimertinib arm and 23% in the control arm had brain metastases. Fewer patients treated with osimertinib had disease progression in the CNS (6% versus 15%) or extracranial disease progression (38% versus 54%), compared with the control arm28. The benefit in PFS was maintained for patients with brain metastases (15.2 months with osimertinib versus 9.6 months with first-generation TKIs; HR 0.47; P < 0.001). Osimertinib was better tolerated than first-line TKIs (34% versus 45% of patients had grade 3 adverse events). Accordingly, the rate of treatment discontinuation was 13% in the osimertinib arm compared with 18% in the control arm. Of note, QT interval prolongations were more frequent with osimertinib than with first-line TKIs (10% versus 4%). In this trial, crossover to subsequent treatment with osimertinib was permitted in patients in whom the T790M mutation was detected after progression upon treatment with first-generation EGFR TKIs. Among the 129 patients who received treatment after disease progression in the control arm, 48 patients (37%) crossed over to receive treatment with osimertinib; data on the overall survival of patients who received treatment after disease progression are eagerly awaited.


Ceritinib was the first second-generation TKI approved as a first-line treatment option for patients with ALK-rearranged NSCLC on the basis of the superior efficacy over platinum-based chemotherapy observed in the ASCEND-4 study71 (Table 2). In this study, the incidence of grade 3–4 adverse events was higher with ceritinib than with chemotherapy (65% versus 40%), but treatment discontinuations owing to toxicity occurred in 5% of patients treated with ceritinib versus 11% in the control arm. This study was designed before crizotinib was established as standard first-line therapy in this disease setting; taking toxicities into consideration, ceritinib remains a valid option for first-line treatment. Encouraging results from a phase I/II trial of brigatinib (NCT01970865) include a median PFS of 34.2 months in 8 patients treated upfront with this agent78. In another phase II trial, the ORR was 90% in a cohort of 30 patients receiving frontline lorlatinib and the median PFS had not been reached at the time of reporting; mature results of this ongoing study will provide further insight into the clinical outcomes derived from lorlatinib treatment81.
Alectinib is the first ALK inhibitor that was compared against crizotinib in the first-line setting in two randomized studies: the phase III trials J-ALEX30, conducted in Japan, and the international ALEX trial29 (Table 3). None of the patients enrolled in J-ALEX had been previously treated with an ALK TKI, but 36% of them had received chemotherapy. Alectinib was associated with a significant PFS benefit (Table 3), as well as a more favourable toxicity profile than crizotinib: grade 3 adverse events were reported in 26% of patients receiving alectinib versus 52% of those receiving crizotinib, and fewer patients required dose interruptions (29% versus 74%) or toxicity-related treatment suspensions (9% versus 20%).
All the patients enrolled in the ALEX trial29 received alectinib in the frontline setting. The median PFS duration and ORR were higher with alectinib than with crizotinib; according to the last update121, median PFS was 34.8 months with alectinib and 10.9 months with crizotinib (HR 0.43; 95% CI 0.32–0.58 months). Crossover was not permitted in the study protocol, hampering the direct comparison of outcomes obtained by administering alectinib using sequential or upfront strategies. One strength of this study29, however, was the evaluation of CNS activity through mandatory brain MRI at study entry and every 8 weeks during treatment. Baseline brain metastases were detected in 42% of patients allocated to receive alectinib and in 38% of patients in the crizotinib group. Patients with measurable CNS metastases had an intracranial response rate of 81% (45% of them being complete responses) with alectinib and 50% (9% complete responses) with crizotinib. The median duration of CNS responses was 17.3 months with alectinib and 5.5 months with crizotinib, and the 12-month cumulative incidence of brain metastases was significantly lower with alectinib than with crizotinib (9.4% versus 41.4%), showing that alectinib provides superior control against the development of brain metastases compared with crizotinib. Interestingly, the difference in PFS between arms can be mainly attributed to the higher rates of CNS-related disease progression with crizotinib, because no significant differences in extra-CNS progression rates were observed between arms (24% and 22% with alectinib and crizotinib, respectively). Comparative trials of crizotinib with brigatinib (NCT02737501), lorlatinib (NCT03052608), or ensartinib (NCT02767804) will provide further information on the efficacy of all next-generation ALK TKIs in the first-line setting.

Choice of upfront treatment strategy

With the management of patients with advanced-stage EGFR-driven and ALK-driven NSCLC on the verge of a paradigm change, the risk–benefit balance of choosing between sequential treatment or next-generation upfront strategies needs to be taken into consideration when optimizing treatment strategies. Several arguments favour each strategy, and, thus, the choice remains complex (Box 1).

Box 1 | Arguments supporting different frontline treatment strategies

  • Arguments in favour of using first-generation tyrosine-kinase inhibitors (TKIs) upfront
  • Mature follow-up data available supporting long survival for patients treated with sequential TKIs
  • Multiple subsequent treatment options available in the event of resistance
  • Arguments in favour of using next-generation TKIs upfront derived from studies comparing with first-generation TKIs
  • In preclinical studies: longer disease control in mice
  • Reduced toxicity in most cases
  • Enhanced therapeutic activity in the central nervous system
  • Prolonged progression-free survival
  • Reduced need for subsequent molecular diagnostic

Traditional sequential approach

This approach has been in place for a longer time than the next-generation upfront strategy, and, thus, sufficient data support an impressive long-term survival with therapies involving sequencing TKIs. The long-term benefit of providing sequential therapies is based on the response rates and the duration of PFS that can be achieved with next-generation inhibitors upon resistance to first-generation TKIs. In patients with NSCLC harbouring mutations in EGFRT790M, a pooled analysis update of the AURA 2 and AURA extension studies59 revealed a median global overall survival of 26.8 months. The 2-year overall survival was 56% for the entire cohort. The mature survival outcomes of the AURA 3 study21 and data on treatment outcomes from the ASTRIS study122 have not yet been published; these results should provide insight into the clinical benefits derived from osimertinib treatment in patients with EGFRT790M-mutated NSCLC.
In patients with ALK-rearranged NSCLC, results from the PROFILE 1014 trial showed, at a median follow-up duration of 46 months, that median survival was not reached (95% CI 45.8 months–not reached) and that 4-year overall survival was 56.6% in patients treated with crizotinib, of whom 33% received subsequent next-generation TKIs65. The French national IFCT-1302 retrospective study123 analysed the survival outcomes of 318 patients with ALK-rearranged NSCLC involved in an expanded crizotinib access programme123. In this study, 31.9% of patients received the second-generation ALK inhibitors ceritinib or alectinib after disease progression on frontline crizotinib. The median overall survival duration from the first dose of crizotinib was not reached for patients who received sequential treatment, and 3-year survival was 59.2% (both ceritinib and alectinib analysed together). Impressively, the median overall survival from the time of diagnosis of metastatic NSCLC was 89.6 months. This duration is highly superior to that observed in patients with NSCLC not driven by alterations in EGFR or ALK and treated with chemotherapy in ‘real-world’ settings (~10 months)124.
The studies discussed support the notion that effective sequential strategies with upfront first-generation inhibitors can lead to impressive overall survival in some patients with NSCLC in which the driver alterations have been characterized; whether upfront next-generation inhibitors could provide a similar long-term benefit remains to be established. The available preclinical and clinical evidence suggests that no clinical benefit is derived from treatment with first-generation TKIs after disease progression on next-generation TKI treatment, with the exception of ALK L1198F108, MET amplification125, and EGFR C797S mutation in trans95, thus limiting the availability of targeted therapeutic options when next-generation inhibitors are used upfront.

Next-generation ALK and EGFR TKIs upfront

This therapeutic option is associated with prolonged PFS durations, improved disease control in the CNS, and a more favourable toxicity profile than treatment with first-generation TKIs — providing a major argument in favour of upfront treatment with next-generation TKIs. In the ALEX29 and FLAURA28 studies, the difference in the incidence of grade 3 adverse events with first-generation versus next-generation TKIs was ~10%, favouring the latter. With the upfront administration of next-generation TKIs, T790M or secondary ALK mutational screening does not need to be performed on a continuous basis, an approach that is convenient in centres where repeated molecular diagnosis is not available. Indeed, the medical practice environment needs to be considered in decisions of the best therapeutic strategy for patients. Close monitoring and timely access to molecular diagnostics and treatment options are essential to providing optimal care.
In the ALEX29 and FLAURA28 studies, alectinib and osimertinib showed greater efficacy in the treatment of brain metastases than first-generation TKIs; thus, these agents should be considered for patients in this setting126128. The prevention or delay of the onset of brain metastases is key to controlling morbidity and reducing the needs and costs for localized CNS therapies129. In this context, the results of the ongoing evaluation of responses to frontline lorlatinib are awaited81. Indeed, results from studies in mouse models suggest that frontline lorlatinib could dramatically delay the emergence of resistance, including those with brain metastases27. Despite having superior potency and the widest spectrum of activity against secondary mutations, lorlatinib might not replace alectinib as the standard-of-care ALK TKI in the first-line setting because of its association with an increased incidence of neurological adverse effects; lorlatinib, however, might represent the ideal second-line treatment option after disease progression on alectinib.
An important argument in favour of using next-generation upfront originates from the emerging evidence from studies of persister cells. An intuitive hypothesis is that a ‘hitting hard first’ strategy would help to limit the number of drug-tolerant cells that would later lead to disease progression; however, to our knowledge, direct comparisons of the persistence capacities of cancer cells treated with first-generation or next-generation TKIs have not been performed. Understanding the molecular mechanisms supporting the viability of these cells and how they can be targeted therapeutically are key questions that have not yet been solved.
Another key aspect that remains to be elucidated is whether frontline treatment with next-generation TKIs can decrease the emergence of subclonal heterogeneity involving TKI resistance mechanisms, either with a mutational or non-mutational component. Importantly, the existence of intratumour heterogeneity is evidenced by simultaneous oncogenic alterations that can mediate resistance to EGFR or ALK TKIs, including the co-occurrence of EGFR with ALK alterations or ALK with KRAS alterations, which present a challenge for treatment selection23,130134. To address this issue, multiple combinations of ALK or EGFR TKIs with other kinase inhibitors targeting MET (NCT02143466), MEK (NCT03392246, NCT03087448, NCT03202940, and NCT02143466), JAK (NCT02917993 and NCT03450330), mTOR (NCT02503722 and NCT02321501), SRC (NCT02954523), AXL (NCT03255083) or CDK4/6 inhibitors (NCT03455829 and NCT02292550), or apoptotic modulators, such as navitoclax (NCT02520778), are ongoing. The aim of these strategies is to revert, delay or prevent the onset of off-target resistance. In addition, several studies have intended to modulate the antitumour immune response by combining an EGFR or ALK TKI with anti-programmed cell death 1 (PD-1) and/or anti-programmed cell death 1 ligand 1 (PD-L1) monoclonal antibodies, which generally lack efficacy as single agents in patients with oncogene-addicted NSCLC135. Nevertheless, toxicity issues have already hampered the development of combinations of osimertinib with durvalumab and of crizotinib with nivolumab. In the phase Ib TATTON study, recruitment into the combination arm (osimertinib plus durvalumab) was closed owing to the occurrence of interstitial lung disease in 38% of patients136. In the multicohort phase I/II CheckMate 370 trial, the combination of nivolumab and crizotinib was associated with severe hepatic toxicity in 38% of patients, with two adverse-event-related deaths137. By contrast, preliminary data of the combination of crizotinib or lorlatinib with avelumab and of alectinib with atezolizumab have shown an acceptable safety profile138,139.

Integrative strategy

In the absence of survival data after disease progression from head-to-head comparative trials, investigators rely on the sum of PFS from studies held in different therapy lines to establish comparisons. This provocative approach is not supported statistically140 but can provide an estimation, in the absence of valid surrogates, of the theoretical benefit of sequential targeted therapies in patients with advanced-stage NSCLC (Fig. 2).
Relying on the results from clinical trials22,69,72,77,80, patients with ALK-translocated NSCLC would derive a median PFS of 16–25 months from frontline crizotinib followed by a next-generation ALK TKI, compared with 34.8 months with alectinib121. Likewise, patients with EGFR-mutated NSCLC would derive a PFS benefit ranging from 21–27 months21,36,41,47,53 with sequential treatment, a value close to the 18.9 months reported for frontline osimertinib in the FLAURA study28. Of note, chemotherapy is the standard treatment for patients with T790M-negative NSCLC with disease progression after receiving first-generation EGFR TKIs. For these patients, the median PFS with cisplatin-based chemotherapy after progression upon treatment with first-line EGFR TKIs was reported to be 5.4 months141; thus, frontline treatment with a first-generation TKI would provide a slightly inferior sum of PFS than frontline osimertinib.
In addition, a subset of patients treated with TKIs can develop oligoprogressive disease. In this scenario, and especially in the setting of brain metastasis, patients can benefit from a 6-month gain in PFS when local ablative treatments (such as surgery or radiotherapy) are applied142. These local ablative treatments are crucial because they enable the continuation of previously administered systemic therapies, delaying the switch to the next treatment line and prolonging systemic disease control.
The economic burden of novel drugs can also influence the choice of upfront TKIs — for example, osimertinib is more expensive than afatinib143. In the absence of definitive evidence of meaningful overall survival benefits, the prolonged administration of costly therapeutic agents might not be easily accepted by regulatory authorities.
In this new era, a growing need exists for the development of clinical trials to enable further understanding of the best sequential therapeutic strategy in the setting of advanced-stage NSCLC. Monitoring resistance onset using sequencing of circulating cell-free DNA can provide new insights into the effect of early treatment of subclinical resistance144. In the setting of EGFR-mutated NSCLC, the ongoing phase II APPLE trial145 will shed light on this matter, evaluating the overall survival outcomes of patients treated sequentially with a first-line EGFR TKI and switching to osimertinib upon progression, compared with treatment with osimertinib upfront.


At present, the optimal approach for the selection of a frontline EGFR or ALK TKI for patients with advanced-stage NSCLC remains a matter of debate, while results and post-progression survival analysis at longer follow-up durations from ongoing comparative trials are awaited. Both strategies have advantages and disadvantages that need to be carefully weighed (Box 1). The currently available evidence suggests that patients with EGFR-mutated NSCLC could benefit from frontline osimertinib over first-generation EGFR TKIs in terms of tolerability and efficacy, especially patients without targetable T790M mutations. Similarly, patients with ALK-rearranged NSCLC would derive a greater benefit from frontline alectinib than with first-line ALK TKIs in terms of tolerability, activity in the CNS, and PFS. For these patients, lorlatinib might be a favourable option for second-line treatment upon regulatory approval. Nonetheless, analysis of long-term survival outcomes of ongoing and future randomized trials, including the effect of post-progression treatments, will be key to settle what the most beneficial treatment strategy for patients with NSCLC according to the molecular profile of their tumours in order to adapt therapies to tumour dynamics.


The authors would like to thank T. Sourisseau for fruitful discussions and critical reading of the manuscript. The work of G.R. is supported by a grant from the Nelia & Amadeo Barletta Foundation. The work of L.F. is supported by a European Research Council (ERC) starting grant (agreement number 717034).

Competing interests

B.B. has received institutional grants for clinical and translational research from AstraZeneca, Boehringer-ingelheim, Bristol-Myers Squibb (BMS), Inivata, Lilly, Loxo, OncoMed, Onxeo, Pfizer, Roche-Genentech, Sanofi-Aventis, Servier, and OSE Pharma. All other authors declare no competing interests.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Related links
US National Institutes of Health database: https://​www.​clinicaltrials.​gov

Related topics

More on this topic