Introduction

Exon 19 deletions and exon 21 L858R mutations are 2 of the most common epidermal growth factor receptor (EGFR) gene mutations.1 Others are collectively treated as uncommon mutations.2-6 It has been generally accepted that the response rate to EGFR tyrosine kinase inhibitors (TKIs) in patients with uncommon mutations is poorer than in those with common mutations, and the progression-free survival (PFS) is short in this group.3,5 There are many types of uncommon mutations2-6 and, although rarely, some patients may simultaneously harbor both common and uncommon mutations.2-6 According to the largest open access database, the Catalogue of Somatic Mutations in Cancer (COSMIC), approximately 16 000 EGFR mutations have been registered and 594 types of EGFR mutations have been reported.2 EGFR compound mutations, defined as double nonsynonymous mutations in the EGFR gene,2,3 are a rare subtype of EGFR mutations. They are usually a combination of a common mutation and an uncommon mutation or 2 uncommon mutations.2-6 Due to the fact that there are many types of compound mutations and each occurs in a small number of patients, only a few studies examined the treatment response and survival of patients with compound mutations.2,3,5 However, next-generation sequencing (NGS) has recently made it possible to determine whether compound mutations coexist in patients treated with common mutations.7-10 In this study, using NGS, we aimed to clarify whether the number of compound mutations is related to prognosis in patients with EGFR-mutated non–small cell lung cancer (NSCLC). We compared the prognosis of patients with a single compound mutation in addition to the main EGFR mutation with those with 2 or more compound mutations.

Methods

Materials

We analyzed specimens collected from patients with the EGFR gene mutation treated in 2 tertiary hospitals (University of Tsukuba Mito Medical Center and Ryugasaki Saiseikai Hospital) since April 2009. Among them, a total of 109 patients were histopathologically diagnosed with EGFR-mutated NSCLC (including 63 [13.2%] of 476 and 46 [13.5%] of 341 NSCLC patients in each hospital, respectively). Of these, 47 consecutive patients with sufficient tissue specimens to enable a genetic analysis were included in this study. The performance status (PS) was assessed based on the Eastern Cooperative Oncology Group criteria, and a score of 2 or more was defined as poor PS.

Analysis of EGFR mutations

EGFR mutations were examined using a nonoverlapping integrated read sequencing system (DNA Chip Research Inc., Tokyo, Japan).7,8 Briefly, DNA was extracted from slices of formalin-fixed paraffin-embedded (FFPE) tissue blocks obtained from patients using a Maxwell RSC DNA FFPE kit (Promega, Madison, Wisconsin, United States). Double-stranded DNA was quantified with the Qubit dsDNA HS Assay (Thermo Fisher Scientific, Waltham, Massachusetts, United States) on the Qubit 2.0 Fluorometer (Thermo Fisher Scientific). A total of 50 ng of DNA was fragmented by a Covaris focused-ultrasonicator (Covaris Inc., Woburn, Massachusetts, United States) following the manufacturer’s instructions. The shearing time was optimized for FFPE material to obtain fragmented DNA with a peak around 200 bp. A molecular barcoded NGS library was constructed using the nonoverlapping integrated read sequencing system method, as described previously.7,8 A customized panel covering the entire region of EGFR tyrosine kinase domain (exons 18–21) was used to amplify the target regions. The constructed library was loaded on an Ion 540 chip using the Ion Chef System (Thermo Fisher Scientific). Sequencing was performed on the Ion S5 platform. Mutation variants with a P value lower than 0.01 were called somatic mutations. Common single-nucleotide polymorphisms deposited in the Human Genetic Variation Database were removed from the set of called variants. A CV78 filter7,8 was applied to remove artefactual substitutions with no entry in the COSMIC (v92) database. The content ratio of tumor cells in the tissue sample and the ratio of allele fraction (RAF)11,12 were also examined using this method.

Statistical analysis

Comparisons of patient background characteristics were evaluated using the χ2 test and the Mann–Whitney test. We investigated the association between patient characteristics and PFS based on the response to treatment with first-line TKIs and overall survival (OS). The effects of clinicopathological characteristics on survival were analyzed using the log-rank test and the Cox proportional hazards model. All analyses were performed using SPSS, version 23 (IBM Corporation, New York, United States). A P value lower than 0.05 was considered significant.

Ethics

This study was approved by the institutional ethics committee of each institute (approval NO18-46). Written, comprehensive informed consent for obtaining pathological specimens was provided by each patient at the time of admission.

Results

We had initially planned to exclude patients with exon 20 insertions from this study as their response to TKIs is clearly different to that of other patients with uncommon mutations, but none of the participants had this uncommon mutation. Therefore, no patients whose pathological specimens were examined were excluded from the study. Among the 47 patients examined, 19 (23.4%) had either a single compound mutation (n = 11) or 2 or more compound mutations (n = 8). The median follow-up period for these 19 patients was 24 months (range, 1–57). There were no significant differences in patient characteristics (age, sex, PS, stage, EGFR mutation, and type of TKI [afatinib or others]) between the 11 patients with a single compound mutation and the 8 patients with 2 or more compound mutations. As shown in Table 1, in both univariable (log-rank test) and multivariable analysis (Cox proportional hazards model), poor PS and 2 or more compound mutations were significant unfavorable predictors of PFS based on the response to treatment with first-line TKIs and OS.

Table 1. Univariable and multivariable analysis of progression-free survival and overall survival in patients with epidermal growth factor receptor gene mutations

Parameter

Univariable analysis (log-rank test)

Multivariable analysis

PFS using first-line TKIa

OSa

PFS using first-line TKI

OS

Hazard ratio

95% CI

P value

Hazard ratio

95% CI

P value

Age: ≥75 y vs <⁠75 y

0.47

0.40

–

–

–

–

–

–

Sex: male vs female

0.65

0.88

–

–

–

–

–

–

PS: 0–1 vs 2–3

0.06

0.01

9.06

1.74–47.10

0.009

13.42

2.09–86.23

0.006

Stage: IIIB vs IVA–B

0.14

0.20

0.30

0.07–1.39

0.125

–

–

–

EGFR mutation: exon 19 deletion vs others

0.17

0.10

3.88

1.19–12.69

0.025

2.68

0.85–8.44

0.092

TKI: afatinib vs others

0.46

0.23

–

–

–

–

–

–

No. of compound mutations: 1 vs ≥2

0.052

0.003

3.74

1.08–13.00

0.038

6.92

1.20–24.00

0.002

a In univariable analysis, only P values are displayed in the Table.

PFS was defined as the time from the initiation of first-line TKIs to disease progression or death from any cause, and overall survival (OS) was defined as the time from the initiation of first-line TKIs to death from any cause. In patients with a single uncommon mutation, median PFS was 23 months (range, 7–57) and median OS was 32 months (range, 11–72). In patients with 2 or more uncommon mutations, median PFS was 4 months (range, 1–53) and median OS was 13 months (range, 1–53).

Abbreviations: EGFR, epidermal growth factor receptor; OS, overall survival; PFS, progression-free survival; PS, performance status; TKI, tyrosine kinase inhibitor

In patients with a single compound mutation, the RAF of the compound mutation was calculated. In patients with 2 or more compound mutations, the sum of the RAFs was calculated. The relationship between RAF and PFS and OS was then investigated. Since RAF = 0.3 was previously reported as a cutoff value associated with poor prognosis,12 we examined cutoff values of 0.1 and 0.3. There was no difference in survival with either cutoff value (cutoff = 0.1: PFS, P = 0.51; OS, P = 0.92; cutoff = 0.3: PFS, P = 0.72; OS, P = 0.88).

Discussion

The majority of patients with EGFR gene mutations have common mutations, and many clinical trials focused on these patients.13 In contrast, due to the small number of patients and the variety of other (uncommon) mutations, it is difficult to evaluate the treatment response rate and duration of response for each uncommon mutation. Therefore, the prognosis associated with uncommon EGFR mutations has not been examined in a large numbers of patients.2-6 Furthermore, some patients have an uncommon mutation in addition to the main common mutation, and these individuals have been generally classified as patients with compound mutations.2-6 Due to the small number of such patients, there have been few reports investigating their prognosis.3,5 Recent advances in the technology that enables examination of the EGFR gene have made it possible to identify compound mutations more precisely.7,8 Therefore, we aimed to investigate how the number of compound mutations in these patients may affect their prognosis. Among the 47 patients included in this study, 19 (23.4%) had 1 or more compound mutations in addition to the common EGFR mutation. We compared the prognosis of 11 patients with a single compound mutation with the prognosis of 8 patients with 2 or more compound mutations. While there were no significant differences in patient characteristics between these 2 groups, both univariable and multivariable analysis showed that poor PS and 2 or more compound mutations were significant unfavorable predictors of PFS based on the response to treatment with first-line TKIs and OS. In the present study, the sum of RAF values11,12 was also investigated but its relationship with the prognosis was not clear. This result indicates that prognosis might not be defined simply by the RAF value. In this regard, an analysis of compound mutation amino acids with electrically charged chains could be useful.

Although interesting results were obtained, the study had some limitations. It was conducted retrospectively on a small number of patients, and even though we tried to avoid it, selection bias may be present. The effects of changes over time from the moment of diagnosis to the analysis of histopathological specimens should also be considered. In addition, various mechanisms, such as the KRAS gene mutation and amplification, cMET gene amplification, and PI3K gene mutation have been reported as factors contributing to resistance to EGFR TKIs;14,15 however, they were not examined in this study.

The present analysis highlighted the necessity to obtain detailed information on compound mutations in patients with common mutations in order to elucidate their effects on survival and the mechanism of drug resistance. The results of this study may provide suggestions for future research in this area.

Some patients with a common EGFR mutation also harbor compound mutations, and this study suggested that patients with 2 or more compound mutations might have a poorer prognosis than those with a single compound mutation. A more detailed examination of the EGFR gene might contribute to the understanding of the duration of TKI response and the mechanism of TKI resistance.