Different Survival Benet of Osimertinib in Different Sequences: A Real ‐ World Outcome of Osimertinib Treatment in Pretreated T790M-Positive Advanced NSCLC in Taiwan

Background: To investigate the relationships among the clinical characteristics, different EGFR-TKIs, and osimertinib treatment in different treatment lines. Methods: We retrospectively screened a total of 3807 patients diagnosed between 2013 and 2019 at Kaohsiung Chang Gung Memorial Hospital. Furthermore, 98 patients after re-biopsy or liquid with EGFR T790M mutation who received osimertinib were enrolled for analysis. Results: Among all 98 patients, the median PFS of those who received osimertinib therapy was 10.48 months, and the median OS of those who received osimertinib therapy was 42.21 months. The OS of those who received osimertinib therapy after previous getinib, afatinib, or erlotinib therapy was 87.93, 49.00, and 42.00 months, respectively (P=0.006). There was a signicant difference in disease control rate between those who received osimertinib treatment after previous chemotherapy (Group A) or immediately following EGFR-TKI therapy (Group B) (93.3% vs. 77.4%, P=0.029). There was also a signicant difference in PS between those who received osimertinib as a second-line treatment and those who received it as a third-line treatment (10.83 vs. 17.33 months, P=0.044). In addition, COPD tended to be a poor prognostic factor for PFS and OS. Conclusion: In this retrospective real ‐ world analysis, it was determined that pretreatment with getinib and previous chemotherapy could affect the treatment outcomes of NSCLC patients treated with osimertinib. Furthermore, COPD tended to a poor prognostic factor for PFS and OS in such patients.

those who received osimertinib therapy after previous ge tinib, afatinib, or erlotinib therapy was 87.93, 49.00, and 42.00 months, respectively (P=0.006). There was a signi cant difference in disease control rate between those who received osimertinib treatment after previous chemotherapy (Group A) or immediately following EGFR-TKI therapy (Group B) (93.3% vs. 77.4%, P=0.029). There was also a signi cant difference in PS between those who received osimertinib as a second-line treatment and those who received it as a third-line treatment (10.83 vs. 17.33 months, P=0.044). In addition, COPD tended to be a poor prognostic factor for PFS and OS.
Conclusion: In this retrospective real-world analysis, it was determined that pretreatment with ge tinib and previous chemotherapy could affect the treatment outcomes of NSCLC patients treated with osimertinib. Furthermore, COPD tended to a poor prognostic factor for PFS and OS in such patients.

Background
Lung cancer is the cancer with high prevalence and high mortality worldwide. Non-small cell lung cancer (NSCLC) accounts for about 80%-85% of all cases of lung cancer. According to the results of history and molecular biology tests, the treatment of lung cancer is personalized. Among various target oncogenes, epidermal growth factor receptor (EGFR) mutations are earliest and key genetic drivers of NSCLC. EGFR mutations are present in 10% of the Caucasian population, but in 40%-50% of the Asian population, including the population of Taiwan. [1][2][3] Previous clinical trials and studies have shown that compared with platinum-based chemotherapy regimens, EGFR-tyrosine kinase inhibitors (TKIs) produce better response rates and fewer adverse reactions. The objective response rate of the rst and second generation EGFR-TKI is around between 60% and 80%, and the median progression-free survival (PFS) duration is around 10 to 13 months. [4][5][6][7][8][9][10][11][12] When these patients experienced disease progression (PD), newly acquired resistant EGFR p.Thr790Met (T790M) point mutations were developed in about 50%-70% of patients. [13][14][15] These acquired resistant mutations enhance the binding a nity of adenosine triphosphate to the EGFR kinase domain, thereby reducing the e cacy of the rst and second generation EGFR-TKIs.
Osimertinib, a third-generation EGFR-TKI, was designed to and is active in non-small cell lung cancers harboring the EGFR T790M mutation. [16][17][18][19] Published reports of clinical trials have shown that osimertinib has better e cacy in patients who undergo disease progression after the rst and secondgeneration EGFR-TKI treatments.[16-19] AURA 3, a phase 3 clinical trial regarding osimertinib, also reported an better PFS associated in osimertinib compared to standard chemotherapy for NSCLC patients with acquired T790M mutations. [19] Therefore, re-biopsy or liquid biopsy is needed to prove the mechanism of acquired drug-resistance when EGFR mutations patients with PD after EGFR-TKI treatment.
In this study, we evaluated the response rate, progression-free survival (PFS), and overall survival (OS) of patients who received osimertinib treatment after a rst-generation EGFR-TKI (ge tinib or erlotinib) or a second-generation EGFR-TKI (afatinib). The main objective of this study was to investigate the relationships among the clinical characteristics, different EGFR-TKIs, and osimertinib treatment in different treatment lines.

Methods
The study retrospectively screened a total of 3807 patients who were diagnosed with pathologicallycon rmed lung cancer between January 2013 and April 2019 at Kaohsiung Chang Gung Memorial Hospital. Among these patients, there were 879 patients with inoperable EGFR mutation-positive adenocarcinoma who had received a rst-generation EGFR-TKI (ge tinib or erlotinib) or a secondgeneration EGFR-TKI (afatinib) as the rst-line therapy. Furthermore, 267 of these 879 patients who were resistant to rst-or second-generation EGFR-TKIs had received a re-biopsy (including bronchoscopy, chest computed tomography guided biopsy, or video-assisted thoracoscopic surgery) and/or liquid biopsy (the Department of Pathology of Kaohsiung Chang Gung Memorial Hospital was in charged for the detection of the EGFR T790M mutation in cell-free plasma DNA) between March 2015 and December 2018. Of those patients, there were 98 patients with EGFR T790M mutation-positive adenocarcinomas who had received osimertinib therapy (80 mg per day) for at least 2 weeks since March 2016. Among these 98 patients, 91 patients were provided with treatment through the expanded access programs supported by AstraZeneca until the occurrence of disease progression or the unacceptable adverse effects. All of the 98 patients who received osimertinib treatment were enrolled for analysis.
Each of these 98 patients regularly received a chest CT scan in initially start of the osimertinib treatment and every three months thereafter to evaluate their tumor responses. Brain MRI imaging and Tc-99m MDP bone scans would also be performed if there were related symptoms. Progression-free survival (PFS), overall survival (OS), overall response rate (ORR), and disease control rate (DCR) were calculated to evaluate their e cacy. The PFS was calculated from the time of starting osimertinib until the time of radiological progression based on RECIST (according to the Response Evaluation Criteria in Solid Tumors) v1.121 or death; with censoring at the time of the last follow-up in the event which the patient was not disease progression. The ORR was de ned as the percentage of patients who presented a complete response or partial response in the rst follow-up image study after the starting osimertinib treatment, while the DCR was calculated as the percentage of patients who exhibited a complete response, partial response, or stable disease. Furthermore, the duration of overall survival was calculated the duration from the starting osimertinib treatment until the patient expired.

Results
The demographic and clinical characteristics of the 98 patients with EGFR T790M mutation-positive adenocarcinomas who received osimertinib therapy are described in Table 1 Table 2 shows the responses to osimertinib treatment after previous therapy with a different rst-line EGFR-TKI. There was no signi cant difference in response rate to osimertinib after previous therapy between the patients treated with the different rst-line EGFR-TKIs. The median PFS of those who received osimertinib therapy after previous therapy with ge tinib, afatinib, or erlotinib was 12.83, 11.87, and 10.90 months, respectively (P=0.293) (Supplementary Figure 1) . The median OS of those who received osimertinib therapy after previous therapy with ge tinib, afatinib, or erlotinib was 87.93, 49.00, and 42.00 months, respectively (Supplementary Figure 1); there was a signi cant difference in OS between the patients treated with the different rst-line EGFR-TKIs (P=0.006). Table 3 shows the response, PFS, and OS results of the patients who received osimertinib treatment after previous chemotherapy (Group A) or immediately following treatment with another EGFR-TKI (Group B NA months, P=0.274) between these two groups.
Furthermore, we compared the response results for the patients treated with osimertinib as the secondline, third-line, or ≥ fourth-line therapy (Table 4). There was partial signi cant difference in median PFS between the patients treated with osimertinib as the second-line, third-line, or ≥ fourth-line therapy (10.83, 17.33, and 9.33 months, respectively, P=0.077) (Supplementary Figure 2), but there was a signi cant difference in median PFS between the patients treated with osimertinib as the second-line or third-line therapy (10.83 vs. 17.33 months, hazard ratio=0.51, 95% CI=0.26-0.99), P=0.044) (Supplementary Figure  3). There was no signi cant difference in OS among these patients.  Figure 5), and without brain metastasis before osimertinib treatment (Patients were without brain metastasis at the time osimertinib was initiated) (P=0.029, hazard ratio=0.56, 95% CI= 0.33-0.95) (Supplementary Figure 6). In terms of OS, there was a signi cant difference only in the patients without COPD (P=0.031, hazard ratio=0.45, 95% CI= 0.21-0.95). Using a Cox proportional hazards regression, we determined that brain metastasis before osimertinib treatment was a poor prognostic factor for PFS and that ge tinib as a rstline therapy and inclusion in Group A (osimertinib treatment after previous chemotherapy) were better prognostic factors for OS (Table 6). Furthermore, COPD tended to be a poor prognostic factor for PFS and OS (Table 6).

Discussion
In this study, we evaluated the response to osimertinib among NSCLC patients with T790M EGFRresistant mutations following treatment with rst-or second-generation EGFR-TKIs. We found that the ge tinib group had better OS (Table 2), that osimertinib treatment after previous chemotherapy (Group A) had a better response rate (Table 3), that osimertinib as the third-line treatment had better PFS than osmertinib as the second-line treatment (Table 4), that brain metastasis noted during osimertinib treatment was a poor prognostic factor for PFS, that ge tinib as a rst-line therapy and inclusion in Group A (osimertinib treatment after previous chemotherapy) were better prognostic factors for OS, and that COPD tended to be a poor prognostic factor for PFS and OS (Table 6).
As shown in Table 7. In group A, 12 (26.67%) patients were with brain metastasis before osimertinib; the PFS was 11.07 months in patients with brain metastasis before osimertinib versus 21.13 months in patients without brain metastasis before osimertinib, respectively. In group B, 20 (37.7%) patients were with brain metastasis before osimertinib; the PFS was 10.27 months in patients with brain metastasis before osimertinib versus 11.87 months in patients without brain metastasis before osimertinib, respectively. So, this could explain osimertinib as the third-line treatment had better PFS than osimertinib as the second-line treatment.
In the LUX-Lung 3 and LUX-Lung 6 trials, OS was signi cantly longer for patients with EGFR Del19positive tumors in the afatinib group than in the chemotherapy group in both trials: in LUX- Lung  (AF mEGFR ) and T790M (AF T790M ) after acquiring resistance between the rst and second generation EGFR-treated. In Kuo's study, the AF T790M /AF mEGFR ratio of the rst-generation EGFR-TKIs treatment group was signi cantly higher than that of the second-generation EGFR-TKIs treatment group. In addition, there was a highly signi cant correlation between AF T790M and AF mEGFR. This could explain why osimertinib tended to have a better PFS following pretreatment with ge tinib than with afatinib in this study. In our study, these data regarding AF T790M /AF mEGFR ratio was not available due to its retrospective study. So Kuo's data cannot explain a better PFS following pretreatment with ge tinib than with erlotinib.
In Taiwan To detect T790M resistance mutations, in most studies, re-biopsy was performed when the disease progressed [13,37], and the results showed that T790M accounted for 50-60% of the resistance mechanism. Since the cancer genome is heterogeneous, it can evolve over time, and it can also interact with different treatments[38], It is unclear whether the timing of a re-biopsy or liquid biopsy will affect the detection rate of T790M. However, in one previous study [39], The results provide evidence that there is no signi cant association between the timing of re-biopsy and the detection rate of T790M. In addition, this study also shows that T790M can exist for a long time after the progression of EGFR-TKI treatment, and it is also an important carcinogenic driving factor. In our study, the time interval between biopsies was 25.95±16.56 (1.33-99.10) months (Table 1); furthermore, the patients treated with ge tinib had a longer time interval between biopsies than those for the patients treated with erlotinib and afatinib. As previous description, ge tinib (since November 2007) was covered by national reimbursement earlier than erlotinib (since June 2008) and afatinib (since May 2014) in Taiwan. So, this could explain why the ge tinib group had a longer PFS than the erlotinib and afatinib groups. Furthermore, osimertinib was approved with second-line use since 2016 and rst-line use since 2019, but covered by national reimbursement since April 2020. The timing difference of approval and national reimbursement time difference could affect outcome between these three rst-line EGFR-TKIs.
Non-small cell lung cancer is the main cause of brain metastases.[40, 41] Amongst these with recurrent/advanced NSCLC, brain metastases are a common cause for cancer-related morbidity and mortality. As targeted therapy continues to improve the prognosis of NSCLC patients with target oncogene,[8] The deterrence of brain metastases has become an increasingly relevant treatment problem. The rst and second generation EGFR-TKIs (ie ge tinib, erlotinib, and afatinib) cannot effectively cross the intact complete blood-brain barrier which the ratio of the patient's cerebrospinal uid to plasma is as low as 0.01 to 0.003. In the AURA 3 and FLAURA studies [19,42], the PFS bene t of osimertinib was observed in patients with or without known or treated brain metastases at trial entry. Patients with brain metastases tended to have a worse PFS bene t (PFS = 15.2, 95% CI = 12.1-21.4 months) than those without brain metastases (PFS =19.1, 95% CI = 15.2-23.5 months) in EGFR mutation NSCLC patients in the FLAURA study [42]. It seems that this could explain why initial brain metastasis did not in uence the osimertinib PFS but brain metastasis during osimertinib treatment did in uence the osimertinib PFS in our study.
This retrospective study has several limitations. First, this study was conducted at a single medical center, such that the patient population may be biased by patient selection and referral patterns. Second, this study was a retrospective survey, which not only resulted in incomplete data for some patients, but also did not control for laboratory examinations. Third, the multiple lines of treatment before administering osimertinib may have confounded the effects. Another limitation was that any genomic alterations beyond EGFR mutations were not measured in this study. Only rst-generation EGFR-TKIs were enrolled for analysis in AURA 3 trial. Although both rst-and second-generation EGFR-TKIs were enrolled for analysis, but it still is a retrospective analysis. In the future, further randomized controlled trial should be conducted to evaluate PFS and OS bene t between different sequences of EKFR-TKIs.

Conclusion
We found that the ge tinib group had better OS, that osimertinib treatment after previous chemotherapy (Group A) had a better response rate, that osimertinib as the third-line treatment had a better PFS than osimertinib as the second-line treatment, that brain metastasis noted before osimertinib treatment was a poor prognostic factor for PFS, that ge tinib as a rst-line treatment and inclusion in Group A (osimertinib treatment after previous chemotherapy) were better prognostic factors for OS, and that COPD tended to be a poor prognostic factor for PFS and OS. But, osimertinib is still neither easily available nor covered by national reimbursement in many countries. In our study, an alternative sequence (using chemotherapy rst when initially osimertinib not available) still s better PFS bene t. Furthermore,

Funding
This study was supported by grants from the Chang Gung Memorial Hospital (CMRPG8E1661~1663, CMRPG8F1351, CMRPG8F1491~1493, and CMPRG8H1201 to Chin-Chou Wang. CMRPG8F1441 to Chia-Cheng Tseng). The funding body had no role in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript.

Competing interests statement
The authors state that that there no potential con icts of interest.  * Excluding the data of osimertinib use immediately after previous EGFR-TKI therapy (n=53, Group B). ** All patients (n=98).    PFS= median progression free survival, OS= median overall survival. Group A= osimertinib treatment after previous chemotherapy, Group B= osimertinib treatment immediately following treatment with another EGFR-TKI. Brain Metastasis (OSI) = brain metastasis was noted before osimertinib treatment. Brain Metastasis (OSI) = brain metastasis was noted before osimertinib treatment.