Impact of somatic mutations on prognosis in resected non‐small‐cell lung cancer: The Japan Molecular Epidemiology for lung cancer study

Abstract Background To report the follow up data and clinical outcomes of the JME study (UMIN 000008177), a prospective, multicenter, molecular epidemiology examination of 876 surgically resected non‐small‐cell lung cancer (NSCLC) cases, and the impact of somatic mutations (72 cancer‐associated genes) on recurrence‐free survival (RFS) and overall survival (OS). Methods Patients were enrolled between July 2012 and December 2013, with follow up to 30th November 2017. A Cox proportional hazards model was used to assess the impact of gene mutations on RFS and OS, considering sex, smoking history, age, stage, histology, EGFR, KRAS, TP53, and number of coexisting mutations. Results Of 876 patients, 172 had ≥2 somatic mutations. Median follow‐up was 48.4 months. On multivariate analysis, number of coexisting mutations (≥2 vs 0 or 1, HR = 2.012, 95% CI: 1.488‐2.695), age (≥70 vs <70 years, HR = 1.583, 95% CI: 1.229‐2.049), gender (male vs female, HR = 1.503, 95% CI: 1.045‐2.170) and pathological stage (II vs I, HR = 3.386, 95% CI: 2.447‐4.646; ≥III vs I, HR = 6.307, 95% CI: 4.680‐8.476) were significantly associated with RFS, while EGFR mutation (yes vs no, HR = 0.482, 95% CI: 0.309‐0.736), number of coexisting mutations (≥2 vs 0 or 1, HR = 1.695, 95% CI: 1.143‐2.467), age (≥70 vs <70 years, HR = 1.932, 95% CI: 1.385‐2.726), and pathological stage (II vs I, HR = 2.209, 95% CI: 1.431‐3.347; ≥III vs I, HR = 5.286, 95% CI: 3.682‐7.566) were also significant for OS. Conclusion A smaller number of coexisting mutations, earlier stage, and younger age were associated with longer RFS and OS, while EGFR mutations were significantly associated with improved OS.


| INTRODUCTION
Lung cancer is the leading cause of cancer-related morbidity and death worldwide and is one of the most molecularly complex cancers. [1][2][3] Driver mutations in cancers have been intensively examined and identified over a decade using advanced and robust tools, namely, next-generation sequencing (NGS), and these serve as the basis for the precision therapy. 4 Certainly, some of these somatic mutations play a critical role in cancer development, and a molecular epidemiological approach has been helpful to uncover the mechanisms of the disease and provide a strategy for cancer prevention. However, a number of genetic changes may not have functional importance, and it seems a few meaningful driver mutations have a prognostic or predictive value.
Among the genes responsible for somatic mutations in non-small-cell lung cancer (NSCLC), the most frequent driver oncogenes were epidermal growth factor receptor (EGFR), v-Ki-ras2 Kirsten rat sarcoma (KRAS), and tumor protein p53 (TP53). Sensitizing EGFR mutations were first reported in 2004 5 and have become the most important somatic mutations for precision therapy for advanced NSCLC because of their high prevalence and the striking treatment efficacy of EGFR tyrosine kinase inhibitors (TKIs). [6][7][8] KRAS, a member of the RAS family, was one of the first oncogene to have been identified in NSCLC. 9,10 KRAS mutations occur frequently in codons 12 and 13, 11 are usually found in nonsquamous carcinoma and in patients who smoke, and are associated with a poor prognosis. 12,13 TP53 (in relation to mutations in the tumor suppressor gene that encodes p53 protein) has a high detection rate in all subtypes of lung cancer, with reported mutation incidence of approximately 40%-80%. 14 Although TP53 plays multiple roles in prevention and suppression of abnormal cell growth through cell cycle arrest, the prognostic or predictive effect of TP53 in NSCLC is limited. 15,16 Furthermore, the frequency of multiple driver mutations, including the three gene mutations mentioned, in NSCLC has not been reported, and the prognostic and predictive effects have not been well studied.
We had previously reported molecular profiling as a primary endpoint in a prospective, multicenter, molecular epidemiology research by collecting samples from 876 patients with NSCLC who had undergone surgical resection and examining the somatic mutations in 72 cancer-associated genes using next-generation sequencing (Japan Molecular Epidemiology for lung cancer study [JME]). 17 In this report, we have demonstrated the incidence of somatic mutation status in resected NSCLC, the mutational spectrum associated with a unique signature of exposure to smoking and body mass index (BMI), and the noteworthy effect of smoking on developing driver mutations.
The secondary endpoints, as per the present research, were overall survival (OS) and recurrence-free survival (RFS) analyses (UMIN 000008177). Therefore, to clarify the impact of somatic mutations on RFS and OS for resected NSCLC, the follow up data and clinical outcomes of the JME study were collected prospectively, and the impact of somatic mutations, including EGFR, KRAS, and TP53 and coexisting multiple mutations, on RFS and OS was analyzed.

| Patients
Eligible patients had pathologically NSCLC with clinical stage I, II, IIIA or IIIB disease (TNM classification version 7 18 ) and had undergone surgery with curative intent. The projected sample size was 900 (450 smokers and 450 nonsmokers) as reported earlier. 17 Patients with prior radiotherapy and/or chemotherapy were excluded, as were patients with other prior malignancies except for adequately treated basal cell or squamous-cell skin cancer or in situ cervical cancer. Other criteria for inclusion were the availability of a surgical specimen and written informed consent. All informed consents were obtained before surgery.

| Statistical considerations
Clinical data, including sex, smoking history, age, stage, histology, mutations in EGFR, KRAS, and TP53 genes, and other minor mutations were used for the this study, as well as additional post hoc analysis on the number of coexisting mutations. Kaplan-Meier (K-M) plots were used for RFS and OS analyses and for determination of median and 95% CI values. A P value of less than .05 was considered significant. Multivariate logistic regression model and Cox proportional hazards models were used to assess the impact of the mutations on RFS and OS. Statistical analysis was conducted using JMP software (version 12, SAS Institute Inc). This study was registered (UMIN 000008177).

| Patients' characteristics
Between July 2012 and December 2013, 957 patients were enrolled from 43 institutions, and, upon performing molecular analyses, and 876 samples were successfully examined for gene mutations by NGS with a mean coverage of 4253×, as reported previously. 17 All 876 patients' clinical and prognostic data were prospectively collected. The data cut-off date for this study was November 30th, 2017, and the median follow up time was 48.4 months. The incidence of tumor gene mutations indicated in this study has previously been reported (JME study). 17 The characteristics of the patients are shown in

| Impact of patients' characteristics on RFS
Upon performing univariate analysis, it was found that age (≥70 years), sex (male), histology (squamous carcinoma), pathological (p-) stage (III-IV > II > I), smoking history (smoker), EGFR mutation (negative), TP53 mutation (positive), and the number of coexisting mutations (≥2) were factors related to shorter RFS. On the other hand, KRAS status and ALK rearrangement did not affect RFS (data not shown). Figure 1 shows an RFS curve in the overall population (a) and RFS curves stratified by p-stage (b and c), EGFR (d), KRAS (e), and TP53 genes(f), and the number of coexisting mutations (g

| Impact of patients' characteristics on OS
Upon performing univariate analysis, it was found that age (≥70 years), sex (male), histology (squamous carcinoma), and pathological stage (III-IV > II > I), smoking history (smoker), EGFR mutation (negative), KRAS mutation (positive), and TP53 mutation (positive), and the large number of coexisting mutations were factors related to shorter OS. On the other hand, ALK rearrangement had no effect on OS (data not shown). Figure 2 shows an OS curve in the overall population (a), and OS curves stratified by pstage (b and c), EGFR (d), KRAS (e), TP53 (f), and the number of coexisting mutations (g

| DISCUSSION
In this prospective analysis, a smaller number of coexisting mutations were associated with longer RFS and OS, implying that multiple mutations are indicative of cancer aggressiveness, thereby resulting in a high relapse rate. EGFR mutation positivity was associated with longer OS, suggesting that the prognosis of EGFR mutation-related lung cancer could be improved by EGFR-targeted therapy, which may also apply to incidences of resected NSCLC.
To the best of our knowledge, the present research is one of the first report prospectively showing that the number of coexisting mutations affects RFS and OS, and the EGFR mutation status has a significant impact on OS in resected NSCLC.
In previous reports, KRAS mutation was not a significant prognostic factors in resected early-stage NSCLC, 21 with similar results recently reported for TP53. 16 Although EGFR mutation is associated with longer survival in advanced diseases, some previous study reported that such results have been inconsistent in surgical series. 22 Furthermore, in a prior report, EGFR, KRAS, and EGFR/KRAS plus TP53 co-mutations were not significant  prognostic markers in early-stage resected NSCLC. 23 The study retrospectively analyzed the impact of EGFR, KRAS, and TP53 mutations on OS using four key trials of early-stage resected NSCLC. The reason why these factors did not affect OS might be that there was no chemotherapy effectively targeting KRAS and TP53 mutations, and EGFR-TKIs were not yet in clinical use at the time.
Here, we have shown that EGFR mutation positivity was associated with longer OS, likely reflecting that EGFR-TKIs were used as a standard therapy in patients harboring EGFR mutations during the period of the present study. [5][6][7] Although erlotinib, used as an adjuvant agent in adjuvant therapy, did not improve OS even in EGFR mutation-positive subgroup of 161 (16.5%) patients in the prior trial (RADIANT) 24 ; it was interpreted that EGFR-TKI therapy after recurrence effectively prolonged survival. On the other hand, a smaller number of coexisting mutations were associated with longer RFS and OS in the present study. Recently, it has been found that, in the evolution of tumors, early founder (clonal or trunk) somatic mutational events that drive tumorigenesis develop as clonal mutations, genome doubling events often occur early in tumor evolution in the trunk of the evolutionary tree, and subclonal driver events may follow after genome doubling in the branches of the evolutionary tree of the tumor. 25 Zhang et al applied multiregion, whole-exome sequencing to specimens from eleven patients with early stage lung adenocarcinoma, and they demonstrated associations of the numbers of subclones in the tumor with relapse. 26 They showed that larger subclonal mutation fractions may be correlated with an increased likelihood of postsurgical relapse in localized lung adenocarcinoma patients. Although larger prospective trials are needed to confirm the results, the present trial, a large-scale prospective study that analyzed somatic mutations in early stage NSCLC, provides robust support for these observations.
In this study, pathological earlier stage and younger age were correlated with longer RFS and OS, and the stage of cancer was the most crucial factor affecting RFS and OS. These results demonstrate that progression of cancer, including metastasis, is more influential than the status of coexistent mutations. Figures 1 and 2(b and c) show that the pathological stage is associated not only with OS but also, clearly, with RFS. TNM classification version 7 18 is based on a retrospective examination that accumulated a large quantity of data, 27 and it was reported by Chansky et al (2009) that the analysis of the classification confirmed age and sex as important prognostic factors, while histology was less important in surgically resected NSCLC. Finally, they showed pathologic TNM category as the most significant prognostic factor. This study is the first to have proven the accuracy of TNM classification version 7 in the prospective manner and to have demonstrated that the pathological stage is the independent prognostic factor irrespective of patient background and somatic mutations. In addition, age was also confirmed as an important prognostic factor. Although older age was related to higher recurrence rate, the relationship may be the same as the correlation between age and morbidity.
Limitations of the present research include a small number of recurrence events and death events, and the relatively short observation period (4 years). There were a considerable number of stage I patients whose prognosis has become better, that is, the relapse rate is quite low due to the progress of diagnostic techniques and developments in improved surgical techniques, resulting in decreased incidences of relapse and death. Although the present study is the largest prospective trial to analyze correlations between prognosis and somatic mutations, a considerable part of the results lacks enough statistical power, and they should be interpreted with caution and need further validation in a prospective study. However, the prognostic information was collected exactly as planned with no missing data, and we have successfully demonstrated the correlations between coexisting mutations and prognosis on multivariate analysis. Another limitation is that the entire genome sequences were not examined, and not all somatic mutations were tested. However, major somatic mutations in genes related to cancer incidence were covered in our analyses, and clinically meaningful somatic mutations were sufficiently tested and their impacts and those of coexisting mutations on prognosis were also effectively examined.
In conclusion, this prospective, observational study showed that a smaller number of coexisting mutations, earlier stage, sex (female), and younger age were associated with longer RFS, while EGFR mutation positivity was significantly associated with improved OS, as well as earlier stage, a smaller number of coexisting mutations, and younger age, in resected NSCLC. The outcomes of the JME study provide valuable information on the impact of somatic mutations and coexisting mutations on RFS and OS, and further prospective studies are warranted to better understand the impacts of somatic mutations and coexisting mutations.