Analysis of acquired resistance mechanisms to osimertinib in patients with EGFR-mutated advanced non

A number of small-scale studies have reported candidate resistance mechanisms to osimertinib at the point of disease progression when it is used as a second- or later-line treatment13,14,15,16, and more recently, when used in the first-line setting, including from the Phase III FLAURA study17,18. Functional studies for many pathways of acquired resistance to osimertinib have been reported previously19,20,21. However, understanding resistance mechanisms is important to help define appropriate combination therapies for patients with EGFRm advanced NSCLC following acquired resistance to EGFR-TKIs or to prevent the development of resistance. Tumor-specific molecular characteristics can be tested using circulating tumor DNA (ctDNA) isolated from the plasma of some cancer patients, thus providing a potentially valuable biomarker status that can be obtained in a minimally invasive manner22,23,24. The data available for resistance mechanisms to osimertinib are limited and collected from across different studies using diverse methodologies, with the majority of resistance studies focusing on ctDNA versus tissue analysis14,16,21. A heterogenous mixture of resistance mechanisms to osimertinib have been detected including EGFR mutations and MET amplification13,14,16. A loss of detectable T790M has also been reported in 42–68% of patients13,14,21. Here we report data on the plasma ctDNA genomic profile of patients with EGFR T790M advanced NSCLC, whose disease progressed on second-line osimertinib treatment during the Phase III AURA3 study.

Results

Demographics

In AURA3, 279 patients were randomized to osimertinib and 140 to platinum-pemetrexed; 83 (30%) and 30 (21%), respectively, had paired (baseline sample and sample at disease progression and/or treatment discontinuation) plasma samples analyzed by NGS (Fig. 1). Among patients with treatment discontinuation samples, 54 (84%) in the osimertinib arm and 16 (94%) in the platinum-pemetrexed arm discontinued due to disease progression. Other reasons for treatment discontinuation included adverse event (n = 5 [8%]), subject decision (n = 2 [3%]), and other (n = 3 [5%]) in the osimertinib arm, and maximum cycle of chemotherapy reached (n = 1 [6%]) in the platinum-pemetrexed arm.

Among patients with paired plasma samples, 103/113 (91%) had baseline detectable plasma EGFR mutations (Ex19del/L858R and/or T790M), and were included in the acquired resistance analysis subset: 78/83 (94%) in the osimertinib arm and 25/30 (83%) in the platinum-pemetrexed arm (Fig. 1). Within this subset, most patients (75/103, 73%) had plasma samples only available at treatment discontinuation, compared with patients with samples only available at disease progression (22/103, 21%). A small number of patients had samples available at both disease progression and treatment discontinuation (6/103; 6%); results from both time points were in the analysis for these patients.

Generally, baseline demographics and clinical characteristics for patients included in the resistance analysis subset were consistent with those reported for all patients randomized in AURA3; baseline demographics for the osimertinib arm are shown in Table 1. Slightly more patients in the osimertinib arm received prior treatment with erlotinib and slightly fewer received gefitinib compared with all patients randomized to osimertinib.

Acquired resistance mechanisms by treatment arm (plasma ctDNA analysis): Osimertinib arm

In the osimertinib arm acquired resistance analysis subset, 32/78 (41%) patients had at least one detectable acquired resistance mechanism; 46 (58%) had no detectable candidate mechanism of resistance (Fig. 2A). EGFR mutations and MET amplification were the most common acquired resistance mechanisms detected, occurring in 17 (22%) and 14 (18%) patients, respectively. Acquired HER2 amplification, MAPK/PI3K alterations and oncogenic fusions (FGFR3-TACC3, NTRK1-TMP3, RET-ERC1, and RET-CCDC6) were each detected in four patients (5%); PIK3CA amplification was detected in three patients (4%). Acquired EGFR mutations included C797X in 14 patients (18%), two patients with C797X co-occurring with L792X, and one patient each with G796S, L718Q, and exon 20 insertion. A total of 39 (50%) patients had a loss of detectable plasma EGFR T790M at progression and/or treatment discontinuation, of which 10/39 (26%) had acquired alterations (Fig. 2A). Ten patients (13%) had undetectable EGFR T790M at baseline. Among patients with MET amplification and detectable plasma EGFR T790M at baseline, 6/12 (50%) lost EGFR T790M detection at progression and/or treatment discontinuation (Fig. 2A). In contrast, none of the patients with acquired EGFR mutations and detectable baseline plasma EGFR T790M lost EGFR T790M detection at progression and/or treatment discontinuation.

More than one acquired resistance mechanism was detected in 15 (19%) patients (Fig. 2A), meaning 47% of all 32 patients with an acquired resistance mechanism had multiple mechanisms detected. MET amplification co-occurred with EGFR C797X (C797S or C797G) in five patients, including two patients with cell cycle gene alterations, one of whom also had BRAF V600E, and one patient with KRAS G12D. Further co-occurrence with MET amplification was detected in three other patients with cell cycle gene alterations, detected in one patient with TPM3-NTRK1 fusion, and in one patient with EGFR G796S + HER2 amplification. Two further patients with HER2 amplification had co-occurring PIK3CA amplification and cell cycle gene alterations, including one patient with co-occurring EGFR C797G/L792H/L792F. One patient with FGFR3-TACC3 fusion had co-occurring EGFR C797S + BRAF V600E.

Non-genomic mechanisms of resistance, such as small cell transformations may have been present, but could not be detected using NGS; histological assessment of tumor tissue at progression/treatment discontinuation was not available to compare these resistance mechanisms.

Acquired resistance mechanisms by treatment arm (plasma ctDNA analysis): platinum-pemetrexed arm

In the platinum-pemetrexed arm acquired resistance analysis subset (n = 25), acquired alterations included loss of amplification in MET (n = 2), HER2 (n = 1), and PIK3CA (n = 1). One patient had a PTEN truncating mutation. There were no EGFR C797S mutations, no acquired mutations in H/N/KRAS, BRAF, FGFR1, PIK3CA, no acquired MET, HER2, or FGFR1 amplifications, and no oncogenic fusions. In total, 4/25 patients (16%) lost detectable EGFR T790M at progression/treatment discontinuation; five patients (20%) had undetectable EGFR T790M at baseline (Fig. 2B).

Osimertinib duration of treatment by candidate resistance mechanisms

Among patients in the osimertinib arm acquired resistance analysis subset (n = 78), duration of treatment was variable across patients with different acquired mechanisms of resistance and loss or retention of detectable EGFR T790M at progression (Fig. 3). Due to the heterogeneity of acquired resistance mechanisms, there was no clear correlation between type or co-occurrence of acquired mechanisms of resistance and duration of treatment.

Discussion

The data presented in this exploratory analysis of the AURA3 study represent, to the best of our knowledge, the largest study of resistance mechanisms in a single cohort of patients with similar baseline characteristics from a randomized controlled study who received second-line osimertinib. Numerous acquired resistance mechanisms were detected with osimertinib, the most common being acquired EGFR mutations and MET amplification, occurring in 22% and 18% of patients, respectively. The most common EGFR mutation was C797S, reported in 14% of patients; less frequently reported EGFR mutations included, C797G, L792H/F, G796S, L718Q, and exon 20 insertion. Other resistance mechanisms observed with osimertinib included; HER2 amplification, PIK3CA amplification, cell cycle gene alterations, and oncogenic fusions FGFR3-TACC3, NTRK1-TPM3 RET-ERC1, and RET-CCDC6.

The acquired resistance mechanisms detected in this study are consistent with previous reports for osimertinib in later-line settings, where MET amplification and EGFR C797S have also been reported as among the most common resistance mechanisms14,16,17,21. EGFR C797S was detected in 22–29% of patients in these studies, higher than the frequency reported here, while higher frequencies of up to 50% have been reported for acquired MET amplification. It should be noted that there is no consensus on the criteria for defining MET amplification with NGS. Furthermore, MET polysomy cannot be effectively accounted for with NGS25. A similar pattern of acquired resistance mechanisms has been reported following first-line osimertinib treatment in the phase 3 FLAURA study18,26 and in a recent large study comparing acquired resistance between first- and second/later-line osimertinib17. There is a degree of overlap between the mechanisms of resistance reported here and after treatment with first- and second-generation EGFR-TKIs. Although EGFR T790M is the most common mechanism of resistance following first- or second-generation EGFR-TKIs, amplification of EGFR, MET, HER2, and PIK3CA mutations have also been identified3,27. Of note, no new resistance mechanisms were identified in this study with second-line osimertinib that could lead to a more aggressive disease.

Loss of detectable EGFR T790M was reported in approximately half of the AURA3 patients studied here and in these patients, no clear association with a shorter duration of treatment was observed compared with patients who retained EGFR T790M. Although the possibility that EGFR T790M was undetected due to the assay’s limit of detection cannot be discounted, the baseline EGFR-TKI sensitizing mutation was detectable in the majority of patients at progression, indicating sufficient levels of ctDNA. Approximately a third of those patients with a loss of detectable EGFR T790M had at least one detectable acquired resistance mechanism, including activation of pathways either downstream or parallel to EGFR, including MET, HER2, and PIK3CA amplifications or cell cycle gene alterations. In a previous analysis of patients with T790M NSCLC and acquired resistance to osimertinib, patients with loss of detectable EGFR T790M had a shorter median time to treatment discontinuation compared with patients who retained EGFR T790M (6.1 months versus 15.2 months)14.

Co-occurrence of acquired resistance mechanisms was common in this study and could have clinical implications when determining subsequent treatments, highlighting the need for combination therapies to overcome multiple resistance mechanisms; for example, MET amplication co-occurred with HER2 amplification and NTRK fusions. Data from the phase 1b TATTON study (NCT02143466) provided early evidence for the combination of osimertinib and savolitinib, a MET inhibitor, in patients with MET-amplified advanced NSCLC who progressed after receiving ≥1 first-, second-, or third-generation EGFR-TKIs (median duration of response 7.1 months, objective response rate (ORR] 52%)28,29, highlighting the importance of maintaining EGFR inhibition in subsequent lines of therapy. Another study addressing EGFR and MET inihibiton was the Phase Ib/II study of capmatinib plus gefitinib, which demonstrated preliminary clinical activity (ORR 47%) in patients with EGFR-mutated NSCLC and MET-amplified tumors after progression on EGFR-TKI therapy10.

To address the acquired EGFR C797S mutation, a potential therapeutic option could be to combine osimertinib with a first-generation EGFR-TKI which does not require EGFR C797 for activity. The ongoing ORCHARD platform study is investigating osimertinib plus gefitinib in patients with acquired EGFR C797X following progression on first-line osimertinib30. In addition, preclinical data have suggested that osimertinib combined with gefitinib, which is active against EGFR C797S, may delay the emergence of acquired resistance31. Preliminary clinical evidence also supports the potential for this combination, as demonstrated in an ongoing Phase I/II study where concurrent osimertinib plus gefitinib for the first-line treatment of EGFR-mutated NSCLC resulted in an ORR of 85% with rapid plasma clearance of the EGFR mutation32.

For other acquired mutations, interesting results have also been obtained in preclinical studies. For example, co-treating tumor cells that are EGFR T790M and BRAF V600E positive with osimertinib and the BRAF V600E inhibitor encorafenib increased tumor sensitivity compared with encorafenib treatment alone33. Further research is needed to elucidate the mechanisms of resistance to second-line osimertinib, and the therapeutic strategies to address them.

Caution should be taken when interpreting these data due to the exploratory nature of this analysis. Amplification events may be underestimated due to a high false-negative rate of plasma NGS for amplification compared with fluorescent in situ hybridization (FISH)25. As plasma NGS only detects genomic alterations in ctDNA, other non-genomic mechanisms of resistance including histological transformation (e.g., small cell lung cancer [SCLC]) were not evaluated, though it would be possible to study these potential mechanisms of resistance in the future trials that involve tissue samples17. Additionally, as no paired tissue biopsies were available for analysis, the plasma mutations could not be compared with tissue.

In conclusion, multiple mechanisms of resistance to second-line osimertinib were observed, similar to those observed in previous studies, with no predominating single mechanism identified. The most frequent resistance mechanisms were MET amplification and the EGFR C797S mutation and approximately half of the patients had a loss of detectable EGFR T790M. Importantly, no new mutations that lead to more aggressive cancer biology were detected. The results identify the need for tissue samples to be taken in order to further investigate non-genomic mechanisms of resistance including histological transformation and for continued investigation into combination therapy approaches to prevent or overcome emergent resistance.