Distinct Progression and Efficacy of First-Line Osimertinib Treatment According to Mutation Subtypes in Metastatic NSCLC Harboring EGFR Mutations

Introduction Osimertinib (OSI), a third-generation EGFR tyrosine kinase inhibitor, is the standard treatment for patients with naive EGFR-mutant NSCLC. Nevertheless, information on how the mutation subtype affects disease progression after the failure of OSI treatment is scarce. Methods We retrospectively reviewed patients with EGFR-mutant NSCLC who received OSI as a first-line treatment between April 2015 and December 2021. Results This study included 229 patients. The objective response rate was 71%, with intracranial and extracranial response rates of 71% and 90%, respectively. The median progression-free survival was 23.3 mo (95% confidence interval [CI]: 19.6–26.7), and the median overall survival was 33.7 mo (95% CI: 31.3–58.6). Multivariate analysis revealed that the EGFR exon 21 L858R point mutation (L858R) (hazard ratio [HR] = 1.56, 95% CI: 1.04–2.34, p = 0.0328) and liver metastasis (HR = 2.63, 95% CI: 1.53–4.49, p = 0.0004) were significant predictors of progression-free survival in OSI treatment. The concomitant disease progression involving the central nervous system metastasis was significantly more common in patients with L858R (p = 0.048), whereas concomitant disease progression involving primary lesions was significantly more common in patients with exon 19 deletion mutation (p = 0.01). In addition, the probability of disease progression over time was higher for L858R compared with that for exon 19 deletion mutation, in patients with central nervous system metastasis (log-rank test, p = 0.027). Conclusions The mutation subtype had an impact not only on the clinical outcome of the first-line OSI treatment but also on progression patterns after OSI treatment in patients with NSCLC harboring EGFR mutations.

Osimertinib (OSI), a third-generation EGFR TKI, was developed for the treatment of patients with EGFR exon 20 T790M mutation-positive NSCLC, after the failure of the firstand second-generation EGFR TKIs. 9,10][13] Although OSI was found to have a marked therapeutic effect, almost all patients treated with OSI experience disease progression after approximately 15 to 20 months.
Several studies have suggested that the mutation subtype and metastatic site at the baseline are the predictive factors for the efficacy of EGFR TKIs. 14,15In contrast, there is little information on disease progression patterns after the failure of the OSI treatment, according to the mutation subtype.In this study, we evaluated the predictive factors of OSI response and the differences in progression patterns for OSI treatment in previously untreated NSCLC harboring EGFR mutations, on the basis of the type of mutation.

Study Design and Patients
Patients with EGFR-mutant NSCLC who received OSI (80 mg daily) as the initial therapy between April 2015 and December 2021 at a single-center, National Cancer Center hospital, in Japan, were included in this study.Patients who switched to other EGFR TKIs owing to adverse events were excluded.The cutoff date was March 30, 2022.Data including patient characteristics such as sex, age, cancer stage, histological type, smoking status, type of EGFR mutations, programmed cell deathligand 1 (PD-L1) expression, metastatic pattern, Eastern Cooperative Oncology Group performance status (ECOG PS), and clinically measurable laboratory biomarkers were retrospectively evaluated from the medical records.EGFR mutations were assessed after NSCLC diagnosis using liquid or tissue biopsies, which were subjected to either next-generation sequencing (Oncomine Dx Target test, Thermo Fisher Scientific, Waltham, MA), Cobas EGFR mutation test (Roche Molecular Systems, Pleasanton, CA), therascreen EGFR-mutant test (Qiagen, Hilden, Germany), and AmoyDx Pan Lung Cancer PCR Panel (Amoy Diagnostics Co., Ltd., Xiamen, People's Republic of China).Mutations other than 19del and L858R were defined as uncommon mutations.Tumor PD-L1 expression was analyzed through immunohistochemical analysis using the PD-L1 immunohistochemistry 22C3 pharmDx assay (Agilent, Santa Clara, CA).This study was approved by the Institutional Ethics Committee of the National Cancer Center Hospital (2015-355) and was performed in accordance with the Declaration of Helsinki.

Statistical Analysis
The tumor response to OSI was determined on the basis of the Response Evaluation Criteria for Solid Tumors, version 1.1. 16Comparisons of categorical variables were performed using Fisher's exact test or the chisquare test.PFS and OS were estimated using the Kaplan-Meier method, and comparisons were performed with the log-rank test.PFS was defined as the duration from the date of administration of the first dose of OSI to disease progression or death.OS was defined as the period between the administration of the initial dose of OSI and death.Both data were censored at the last follow-up visit.Cox proportional hazards models were used to analyze hazard ratios (HRs) and 95% confidence intervals (CIs) through univariate and multivariate analyses.8][19][20][21] Competing risk methods were used to analyze the cumulative incidence of CNS metastasis.Each progression event was defined as either CNS metastasis progression, any other progression, or death, and the data were censored when the earliest event occurred.A p value less than 0.05 was considered statistically significant.The analyses were performed using JMP Pro version 13.1.0(SAS Institute, Cary, NC).

Patient Characteristics
Baseline patient characteristics are summarized in Table 1.A total of 229 patients with advanced NSCLC harboring EGFR mutations received OSI treatment; the median age was 67 years (range: 28-87 y) and 154 (68%) of the patients were female.More than half of the patients were nonsmokers (68%) and most of the patients had adenocarcinoma (97%).ECOG PS greater than or equal to 2 was observed in 27 patients (12%).The most common EGFR mutation was 19del (n ¼ 125, 55%), followed by L858R (n ¼ 88, 38%).Uncommon EGFR mutations included the following: G719CþE709A (n ¼ 1), and G719XþS768I (n ¼ 1).Patient characteristics for each type of EGFR mutation assessed in this study are presented in Supplementary Table 1 (which reveals baseline patient characteristics by type of EGFR mutation).At the initiation of OSI treatment, 73 patients (32%) had CNS metastasis.Of these patients, 26 had prior local treatments in the brain; 11 had received whole-brain radiation therapy whereas 12 had received stereotactic radiotherapy.At the data cutoff, 118 patients were still receiving the OSI treatment.

Clinical Outcomes of OSI in Patients With Advanced NSCLC
The median follow-up duration was 19.7 mo (interquartile range: 13.8-27.5).In all patients who underwent OSI treatment, the objective response rate (ORR) was 71% (95% CI: 64.1-75.8), the disease control rate was 89% (95% CI: 84.4-92.5),and the intracranial (IC) ORR was 90% (95% CI: 80.2-95.4)(Supplementary Table 2, which reveals activity of OSI).The overall concordance between IC and extracranial (EC) response to OSI treatment was 89%.Among the patients with baseline CNS metastasis, 51 (84%) exhibited both IC and EC response whereas three (5%) exhibited neither.The patients with 19del had a significantly higher response than that in patients with L858R (79% versus 60%, p ¼ 0.003).IC ORR was 97% with 19del versus 79% with L858R (p ¼ 0.03).In the 19del, 42% of the patients exhibited an IC complete response, whereas only 25% of the patients in the L858R exhibited an IC complete response (Supplementary Table 3, which reveals activity of OSI by type of EGFR mutation).

OSI Activity According to the Metastatic Sites and EGFR Mutation Subtype
According to the metastatic sites, the median PFS in patients with liver, bone, or adrenal metastases was significantly shorter than in those without them, whereas there was no difference in the PFS between patients with and without brain metastasis (19.6 mo versus 24.1 mo; HR ¼ 1.380; 95% CI: 0.94-2.04,p ¼ 0.1015).Similar results were observed for pleural effusions (Fig. 2A to F).The multivariate analysis revealed that patients with L858R exhibited a significantly shorter PFS than those with 19del (HR ¼ 1.56, 95% CI: 1.04-2.34,p ¼ 0.0328) and patients with liver metastasis also exhibited a shorter PFS (HR ¼ 2.63, 95% CI: 1.53-4.49,p ¼ 0.0004) (Table 2).
In addition, we evaluated the impact of EGFR mutation subtypes (19del versus L858R) on PFS according to the metastatic spread at the baseline (Supplementary Fig 1A to F).Interestingly, in patients without brain metastasis at baseline, there was no difference in PFS between the patients with 19del and L858R (24.0 mo versus 26.7 mo; HR ¼ 1.26, 95% CI: 0.77-2.07,p ¼ 0.3479), whereas, in patients with brain metastasis, PFS was significantly shorter in patients with L858R compared with that in patients with 19del (12.8 mo versus 24.1 mo; HR ¼ 2.13, 95% CI: 1.10-4.12,p ¼ 0.0251).

Distinct Progression Patterns During OSI According to EGFR Mutation Subtype
Progression patterns are summarized in Figure 3A to C. During the course of the treatment, 107 patients exhibited progression of disease and 67 of them (63%) received subsequent treatment.The common sites of disease progression included intrathoracic lesions (60%), extrathoracic lesions (43%), and the CNS (21%).The concomitant disease progression involving CNS metastasis was significantly more common in patients with L858R (p ¼ 0.048), whereas concomitant disease progression involving the primary lesion was significantly more common in patients with 19del (p ¼ 0.01).Patients with CNS metastasis at the beginning of the treatment exhibited more frequent CNS metastasis progression than those without it (40% versus 6%, p < 0.0001).In patients with CNS metastasis (regardless of prior radiotherapy) at the baseline, the probability of CNS metastasis progression was higher over time for those with L858R compared with that in those with 19del (log-rank test; p ¼ 0.027) (Fig. 3D).In contrast, in patients without CNS metastasis, there was no significant difference in the probability of CNS metastasis progression between those with 19del and L858R (logrank test; p ¼ 0.15) (Fig. 3E).

Discussion
We evaluated the clinical outcomes of OSI treatment in previously untreated patients with NSCLC harboring EGFR mutations and identified the metastatic status and mutation subtypes that affected the clinical outcomes of OSI treatment.In addition, we observed that during OSI treatment, concomitant CNS metastasis progression was more common in patients with L858R, compared with that in patients with 19del.Our data suggest that the treatment strategy should be developed on the basis of the mutation subtype and metastatic spread.Our data reveal that the median PFS and OS in patients treated with OSI as first-line therapy were 23.3 and 33.7 mo, respectively, and PFS in OSI-treated patients with L858R and uncommon mutations was significantly shorter compared with that for patients with 19del.These results are in accordance with those of the FLAURA study (19del: 21.4 mo; L858R: 14.4 mo) 12,13 and other studies. 14,15,19In general, EGFR TKIs, including OSI, are more effective in patients with 19del than in those with L858R. 22One of the reasons for the difference in the efficacy of EGFR TKIs on 19del and L858R is the location of L858R at a distance from the ATP binding site 23 and 19del being a structural change resulting from the deletion of an essential residue in the alpha-helix of the tyrosine kinase domain. 24Second, although ligand-less dimerization in the kinase domain is common owing to oncogenic activity, the activity of L858R is dimerization dependent, whereas that of 19del is dimerization independent. 25Third, the postdimerization autophosphorylation sites in 19del and L858R are different, as are the subsequent downstream signaling pathways. 26ith respect to uncommon mutations, preclinical data revealed that OSI was active against most uncommon mutations, except for the exon 20 insertion mutation. 27Uncommon mutations have heterogeneous molecular alterations in exons from 18 to 21, 28 and differences in the clinical efficacy of OSI depend on the type of uncommon mutations. 29,30A recent detailed study identified that a structure and function-based approach predicts patient clinical outcome and drug sensitivity in EGFR mutations, including uncommon mutations. 31Our results on uncommon mutations are consistent with those of two previous studies (median PFS: 8.2 mo and 9.5 mo). 29,30e also evaluated how the metastatic sites at the baseline and progression sites affect the clinical outcomes of first-line OSI treatment in patients with naive EGFR-mutant NSCLC.The presence of liver metastasis affected the clinical outcomes of OSI treatment the most, compared with other metastatic sites, which is in accordance with previous reports. 14,15A poor clinical outcome in patients with EGFR-mutant NSCLC with liver metastasis has been reported not only in patients treated with OSI but also in those treated with other firstand second-generation EGFR TKIs.Previous reports have revealed that the tumor microenvironment of liver metastasis has an increased expression of the vascular endothelial growth factor, 32 and in EGFR-mutant NSCLC, activated EGFR signaling drives vascular endothelial growth factor expression in a hypoxia-independent manner. 33These contribute to resistance to EGFR TKIs in EGFR-mutant NSCLC with liver metastasis.
Progression patterns during EGFR TKI treatment of patients with advanced NSCLC harboring EGFR mutations are widely diverse.Importantly, approximately half of the patients with EGFR-mutant NSCLC develop brain metastasis, 34 whereas up to 40% develop CNS metastasis over the course of the disease after the initial response to EGFR TKI therapy. 35Therefore, we evaluated the efficacy of OSI treatment against CNS metastasis and distinct progression patterns on the basis of the type of mutation.Interestingly, in patients with CNS metastasis, OSI treatment was less effective against L858R than 19del (12.8 mo versus 24.1 mo, p ¼ 0.02).In contrast, in patients without CNS metastasis, there was no significant difference in the efficacy of OSI treatment between 19del and L858R (24.0 mo versus 26.7 mo, p ¼ 0.35).In addition, regardless of the presence or absence of CNS metastasis at baseline, CNS metastasis was a significantly more common progression site in patients with L858R compared with that in patients with 19del.In patients without CNS metastasis at baseline, there was no significant difference in the probability of progression in those with 19del and L858R.These results may have been influenced by the shorter PFS and lower IC response of OSI treatment in patients with NSCLC with L858R, compared with that in patients with 19del.In this study, patients with 19del received more brain radiotherapy for brain metastases compared with those with L858R (46% versus 11%, p ¼ 0.0023), suggesting a synergistic efficacy of brain radiotherapy and EGFR TKI as  another reason for the distinct progression patterns.Patients with 19del have more brain lesions and smaller brain edema than those of EGFR wild type, which would have potentially led to radiotherapy. 36Radiotherapy enhances drug penetration of the blood-brain barrier and consequently increases the effective concentration of EGFR TKI. 37Indeed, previous reports revealed that combination therapy with radiotherapy and EGFR TKI was associated with better outcomes than those achieved with TKI alone in EGFR mutation-positive NSCLC with brain metastases. 38,39The biological rationale for differences in CNS metastasis development between 19del and L858R remains unclear owing to the limited insight into the genetic makeup of metastasis.The variation in resistance mutations and differences in co-occurring mutations may define the risk of CNS metastasis progression.2][43] Furthermore, the occurrence of EGFR exon 20 T790M-resistant mutations may be less common in tumors with concurrent TP53 mutations, 44 and some of these concomitant mutations, including TP53 mutations, affect the inferior clinical outcome of EGFR TKI-treated patients. 45Our data provide a rationale for considering additional treatment strategies to prevent or delay the development of CNS metastasis, on the basis of the EGFR mutation subtype.This study had several limitations.First, it was a retrospective, single-center study on the Japanese population.Second, the intervals for assessing radiological treatment responses depended on the treating physician, allowing for potential bias.Nevertheless, all patients underwent scheduled follow-up and radiological assessment.Third, resistance mutations and co-occurring mutations could not be evaluated because of the difficulty in collecting the tissue after OSI treatment failure.Finally, we were unable to assess the initial resistance mutations involved in the efficacy of OSI treatment, nor were we able to evaluate the biology of the cancer resistance.
In conclusion, our findings elucidate the metastatic patterns and type of EGFR mutations associated with the clinical outcomes of first-line OSI therapy and suggest that distinct progression patterns dependent on the type of EGFR mutation may be important in determining the treatment strategies in patients with EGFR-mutant NSCLC.

Figure 1 .
Figure 1.Efficacy of osimertinib treatment in patients with untreated NSCLC harboring EGFR mutations; (A) PFS for all patients and (B) OS for all patients.(C) PFS and (D) OS by type of EGFR mutation.19del, exon 19 deletion; CI, confidence interval; L858R, L858R point mutation; OS, overall survival; PFS, progression-free survival.

Figure 3 .
Figure 3. (A) Osimertinib administration status and disease progression pattern after osimertinib treatment failure.Frequency of progression pattern in (B) intrathoracic and (C) extrathoracic sites.Cumulative incidence of disease progression by type of EGFR mutation (D) in patients without CNS metastasis and (E) in patients with CNS metastasis.The statistical test was the probability of CNS metastasis progression between the mutation types exon 19 deletions and L858R point mutations.CNS, central nervous system.

Table 1 .
Baseline Patient Characteristics

Table 2 .
Univariate and Multivariate Analyses of Progression-Free Survival Using Cox Regression Models