In non-small cell lung cancer (NSCLC) mutations in the epidermal growth factor receptor (EGFR) gene are common driver mutations. In the metastatic setting, EGFR tyrosine kinase inhibitors (EGFR-TKIs) are generally more efficient than chemotherapy [1]. Comparative studies such as LUX-Lung 7, ARCHER and FLAURA have demonstrated better outcomes with second- and third-generation EGFR-TKIs (afatinib, dacomitinib and osimertinib) compared to first-generation EGFR-TKIs with median progression-free survival (mPFS) of 11.7, 14.7 and 18.9 months, respectively [2,3,4]. Despite missing comparative trials, osimertinib is considered the standard first-line option in the metastatic setting due to favourable outcomes and safety profile [1]. Improvements in detection methods have shown that so-called uncommon or atypical mutations can constitute up to 30% of the cases, including exon 20 insertions (ex20ins), major uncommon mutations (G719X, S768I, L861Q including compounds of those), exon 19 insertions (e.g. K745_E746insTPVAIK), exon 18 deletions and many others [5, 6]. However, most comparative studies only included the classical EGFR mutations L858R and exon 19 deletions, neglecting atypical mutations.

This short review aims to provide an overview of available clinical data for atypical EGFR mutations.

Exon 20 insertions

Insertions predominately occur between amino acids (AA)762 and AA775. With few exceptions (e.g. A763_Y764FQAE), patients with ex20ins generally exhibit poor response to first- and second-generation EGFR-TKIs [6,7,8]. Platinum-based chemotherapy has been the recommended treatment option [1]. Currently, amivantamab, a bispecific MET and EGFR antibody, is the only treatment approved by the US Food and Drug Administration (FDA) and European Medicines Agency (EMA) other than chemotherapy, based on data from the CHRYSALIS study [9]. Here, overall response rate (ORR) was 40% and mPFS was 8.3 months for patients who progressed after platinum-containing chemotherapy. Thirty-five percent of the patients experienced ≥ grade 3 adverse events (AE). Most common AEs (all grades) were rash (86%), paronychia (45%) mostly related to EGFR inhibition, and hypoalbuminaemia (27%) and peripheral oedema (18%) related to MET inhibition. Five patients (4%) were reported with interstitial lung disease [9].

The recent PAPILLON phase 3 study compared the combination of amivantamab plus chemotherapy with chemotherapy alone in advanced-stage, treatment-naive NSCLC patients. Median PFS was significantly longer with the amivantamab combination, 11.4 months vs. 6.7 months for chemotherapy alone (hazard ratio 0.4). However, 75% of the patients treated with the combination reported ≥ grade 3 AEs vs. 54% ≥ grade 3 AEs in the chemotherapy alone arm. The most common treatment-emergent AEs for the combination were neutropenia, paronychia, rash, anaemia, infusion-related reactions and hypoalbuminaemia. The rate of treatment discontinuation was also higher in the combination arm (24%) than in the chemotherapy alone arm (10%) [10]. While the combination of amivantamab plus chemotherapy may become the first-line standard of care for patients with ex20ins, side effects could become unpleasant and should be closely monitored and treated as soon as they appear.

Significant progress has also been made in development of ex20ins-specific EGFR-TKIs, with many currently under investigation in clinical trials. One example, sunvozertinib (DZD9008) showed an ORR of 61% (n = 59) in a phase II study conducted in China, including patients who progressed from chemo(immuno)therapy, with a promising safety profile. Grade ≥ 3 AEs were mostly elevated blood creatinine phosphokinase levels (18%) and diarrhoea (8%) [11]. However, another promising contender, mobocertinib, was recently withdrawn from the market because the confirmatory phase 3 study (EXCLAIM2) did not meet its primary endpoint, although it was initially granted accelerated approval by the FDA in 2021 [12]. The full data of this study have not been released yet. However, one might speculate that the overall lack of efficacy could be due to the high degree of heterogeneity of ex20ins. In this regard, insertions into the regulatory αC-helix domain (AA762-AA766) and those in the loop following it may be considered distinct subgroups. Furthermore, loop insertions can be divided into near-loop (AA767–AA772) and far-loop (AA773–775) insertion leading to potentially different responses with specific TKIs [5, 13].

Major uncommon EGFR mutations

The first landmark publication demonstrating efficacy of afatinib in patients with atypical EGFR mutations was published by Yang and colleagues in 2015 [8]. In a post hoc analysis, the authors investigated the outcomes of patients with atypical EGFR mutations from LUX Lung studies 2, 3 and 6. Of the 100 identified patients, 25 had received chemotherapy, and 75 patients had been treated with afatinib. Atypical EGFR mutations were categorized into three groups: ex20ins, de novo T790M mutations and a third group that included point mutations and duplications, or both, in exons 18–21. The latter group mainly consisted of L861Q, G719X and S768I mutations, either individually or in combination with each other or classical EGFR mutations. This set of mutations is now often referred to as uncommon or major uncommon EGFR mutations. While this study revealed the general ineffectiveness of afatinib in ex20ins (mPFS 2.7 months) and T790M de novo mutations (mPFS 2.9 months), it notably reported a mPFS of 10.7 months for afatinib in patients with major uncommon mutations, compared to 8.5 months for platinum-based chemotherapy (Table 1). This initial study has been expanded over time and was last updated in 2022, now including data from over 1000 patients taken from clinical trials, early access and compassionate use programs, and case reports from the literature [14]. In the EGFR-TKI-naive cohort (n = 587 patients), the time to treatment failure (TTF) was 10.7 months and ORR was 49.8% (Table 1). Of these mutations, 52% (305/587) were major uncommon mutations, 23% were ex20ins, and 10% were T790M mutations. For major uncommon mutations (G719X, L861Q and S768I including compounds of such) mTTF was 12.6 months and 10.7 months for other atypical mutations (excluding T790M and ex20ins). The efficacy of afatinib in atypical EGFR mutations was also confirmed by other studies, some of which are summarized in Table 1.

Table 1 Overview of selected studies investigating outcomes of atypical EGFR mutations

The Korean multicentre, phase II KCSG-LU15-09 study evaluated the efficacy of osimertinib in patients with mostly major uncommon EGFR mutations. Thirty-six EGFR-TKI-naive patients with recurrent or metastatic NSCLC were included. The overall mPFS was 8.2 months, and the ORR was reported to be 50% (Table 1; [15]). These data were corroborated by the prospective UNICORN study conducted in Japan (n = 40 patients, mPFS 9.4 months, first-line EGFR-TKI [16]), a prospective study from the USA (n = 17 patients, mPFS 10.5 months, ORR 47%, EGFR-TKI-naive [17]) and other retrospective studies summarized in Table 1.

Despite slight variations in patient cohorts and reported outcome parameters, together these data provide evidence for the efficacy of EGFR-TKI in patients with major uncommon EGFR mutations. It is noteworthy that the efficacy is comparatively lower than that observed for classical EGFR mutations. Both afatinib and osimertinib are now recommend as preferred first-line treatment options by international guidelines for patients with major uncommon EGFR mutations [1, 18].

Other uncommon EGFR mutations and future directions

A recent retrospective, multicentre analysis from the national Network Genomics Medicine in Germany, including 856 patients and 276 different atypical EGFR aberrations (excluding T790M mutations), highlighted the fact that next to ex20ins and major uncommon EGFR mutations, the very heterogenous group of very rare atypical EGFR mutations was the largest group (45%, 382/856) [6]. This group included exon 18 deletions, exon 19 insertions, different compound mutations not included in the group of major uncommon mutations, and a very large group of very rare point mutations (26.1% of the total cohort, 223/856). While confirming the benefit of EGFR-TKIs in major uncommon EGFR mutations, this study surprisingly also reported a numerical benefit for EGFR-TKIs compared to chemotherapy for very rare EGFR mutations in the first-line setting: mPFS was 6.7 months for EGFR-TKIs compared to 5.5 months for chemotherapy. However, this difference was not statistically significant, but detailed analyses of subgroups and individual mutations in this large group of very rare atypical EGFR mutations revealed high variability in responses to different EGFR-TKIs.

These data raise the question of how we can accumulate more evidence for this large group of patients with very rare atypical EGFR mutations, especially since the frequency of each individual very rare mutation is most likely far too low for evaluation in standard clinical trials. Is there a better way to classify atypical EGFR mutations than simply by their location in the gene or by their frequency?

In this context, Robichaux and colleagues investigated a structure-based approach [5]. Employing a combination of in vitro sensitivity analyses and in silico modelling, they identified four subgroups characterized by distinct structural features: (1) Classical-like EGFR mutations, situated distant from the ATP binding pocket, with sensitivity to first-, second- and third-generation EGFR-TKI. (2) So-called PACC (P-loop and alpha-helix compressing) mutations, located at the interior face of the ATP binding pocket with sensitivity primarily to second-generation EGFR-TKI. (3) Exon20 loop insertions and (4) T790M-like mutations with increased affinity to ATP compared to classical EGFR mutations. They validated their hypothesis regarding PACC mutations by comparing time to treatment failure (mTTF) of first-, second- and third-generation EGFR-TKIs in patients with PACC mutations in a retrospective analysis. Their results demonstrated better efficacy of second-generation EGFR-TKIs (mostly afatinib), with an mTTF of 21.7 months (n = 25). In contrast, mTTF for first-generation EGFR-TKI was 10.0 months (n = 17), and only 4.1 months for third-generation EGFR-TKI (n = 11; Table 1).

S768I and G719X are considered PACC mutations according to Robichaux’s classification, and patients with these mutations exhibited more favourable outcomes than with L861Q in afatinib studies highlighted in Table 1. Conversely, L861Q is considered a classical-like mutation, and patients with this mutation tend to have better outcomes than those with S768I and G719X mutations in most osimertinib studies (Table 1). This observation may provide additional support for Robichaux’s classification.

While this strongly indicates that structure-based approaches may enhance a better classification and subsequent treatment of patients with rare EGFR mutations, some questions remain unanswered. For instance, the precise criteria for classifying a mutation into one of the four groups lacks clarity. Consequently, the course of action for mutations not yet classified by Robichaux and colleagues remains uncertain. Although this approach does not yet present a ready-to-use solution for routine clinical oncology practice, it undeniably provides a promising initial strategy that warrants further development. The dataset can serve as a valuable resource by experts in molecular tumour boards aiding clinical decisions making.

In summary, recent data substantiate the activity for afatinib and osimertinib, particularly in major uncommon EGFR mutations, albeit with overall responses lower than what was observed for classical EGFR mutations. This information has been incorporated into clinical recommendations and guidelines [1, 18]. Amivantamab in combination with chemotherapy may become the first-line standard of care treatment for ex20ins, but toxicities are high. Other atypical EGFR mutations should be discussed in molecular tumour boards on a case-by-case basis. Structure-based approaches, estimated binding energies and in vitro sensitivities as elucidated by Robichaux and other researchers [5, 19] could be used in molecular tumour boards for clinical decision making. Moreover, these approaches may contribute to generating more robust evidence for rare variants of driver mutations even beyond EGFR mutations in the future.