Combination of Whole-Brain Radiotherapy with Epidermal Growth Factor Receptor Tyrosine Kinase Inhibitors Improves Overall Survival in EGFR-Mutated Non-Small Cell Lung Cancer Patients with Brain Metastases

Brain metastases (BM) cause morbidity and mortality in patients with non-small cell lung cancer (NSCLC). The use of upfront epidermal growth factor receptor (EGFR) tyrosine kinase inhibitors (TKIs) and withholding of whole-brain radiation therapy (WBRT) is controversial. We aim to investigate the impact of WBRT on overall survival (OS). After screening 1384 patients, a total of 141 EGFR-mutated patients with NSCLC and BM were enrolled. All patients received EGFR-TKIs between 2011 and 2015. Ninety-four patients (66.7%) were treated with WBRT (TKI + WBRT group). With a median follow-up of 20.3 months (95% confidence interval (CI), 16.9–23.7), the median OS after the diagnosis of BM was 14.3 months (95% CI, 9.5 to 18.3) in the TKI + WBRT group and 2.3 months (95% CI, 2 to 2.6) in the TKI alone group. On multivariate analysis, WBRT (p < 0.001), female, surgery to primary lung tumor, and surgery to BM were associated with improved OS. The 1-year OS rate was longer in the TKI+WBRT group than that in the TKI alone group (81.9% vs. 59.6%, p = 0.002). In conclusion, this is the first study to demonstrate the negative survival impact from the omission of WBRT in patients with EGFR-mutant NSCLC.

WBRT on BBB permeability. To address this issue, we assessed the effectiveness of TKI given alone or in combination with WBRT to patients with EGFR-mutated NSCLC and newly diagnosed BM.

Ethics Approval Statement
The present study (KMUHIRB-E(II)-20180185) was approved by the ethical and research committee of the Kaohsiung Medical University Hospital. This study was conducted in compliance with institutional review board regulations in accordance with the Helsinki Declaration of 1975 as revised in 1983. All patients provided written informed consent for treatment; patient information was anonymized and de-identified before analysis; consequently, all data were analyzed anonymously.

Patients
Of 1384 NSCLC patients in the database of the tertiary hospital, we identified and analyzed 141 consecutive patients with pathologically proven lung adenocarcinoma who had received TKIs between 3 January 2011 and 29 December 2015. Their BM were diagnosed by either cytology or brain neuroimaging studies. The inclusion criteria for this study were pathologically proven positive EGFR mutations, the diagnosis of BM, and the use of TKI. The exclusion criteria were a history of prior brain RT, or a history of malignancies other than lung cancer, or EGFR-TKI resistance mutation, or incapability to receive EGFR-TKI. Patient follow.ups were conducted by clinic visits or telephone calls until June 2018.
The following variables were collected: age, gender, initial clinical tumor and nodal classification, time from initial diagnosis to BM, extracranial metastasis, histological grading, EGFR mutation, operation to primary lung tumor, number of lines of chemotherapy, name of EGFR-TKI, number of lines of TKI, mean duration of TKI use, Eastern Cooperative Oncology Group (ECOG) performance status at the time of BM, number of BM, smoking history, whether the patient was symptomatic from BM, size of the largest BM, number of BM and disease-specific Graded Prognostic Assessment (dsGPA). The dsGPA was calculated for each patient to determine whether the cohorts shared similar prognostic features [28].

Target Therapy
All patients underwent pretreatment workups comprising a physical examination, a history review, chest radiography, bronchoscopy with a tumor biopsy, chest computed tomography (CT), brain magnetic resonance imaging (MRI) or CT, and routine laboratory studies. The tumor stage was classified according to the seventh edition of the Cancer Staging Manual and Handbook of the American Joint Committee on Cancer (AJCC) [29]. All patients started taking EGFR-TKI once the diagnosis of stages IIIB -IV lung cancer with EGFR mutation was established. TKIs included gefitinib, erlotinib, afatinib and osimertinib. The first generation was used and then shifting to the second or third generation might be chosen at the discretion of the thoracic oncologist.

WBRT
Ninety-four patients had WBRT once the diagnosis of BM was confirmed. For WBRT, each patient was simulated in the supine position in a customized thermoplastic immobilization mask. Three-dimensional conventional radiotherapy (3D-CRT) was delivered using a 2100 C/D linear accelerator (Varian Medical Systems, Palo Alto, CA, USA) for 57 patients. The remaining 37 patients were treated by intensity-modulated radiotherapy (IMRT) either with a Hi-Art helical tomotherapy unit, version 2.2.4.1 (TomoTherapy, Inc., Madison, WI, USA), or Eclipse, version 8.6 (Varian Medical Systems Inc., Palo Alto, CA, USA). For the 37 patients who had a boost dose to their BM, the tumor and boost beams were combined in one integrated treatment plan. Fractionation schemes were as follows: 30 Gy in 10 fractions with or without a simultaneous boost to the brain of 45 Gy, or 37.5 Gy in 15 fractions with or without a simultaneous boost to the brain of 45 Gy. The decision whether to give a RT dose boost to the BM sites was at the discretion of each radiation oncologist. The mean radiation dose was 3781 ± 749 cGy to BM.

Statistical Analysis
The primary end points were overall survival (OS) and the OS after a diagnosis of BM (OSm). OS was defined as the time from the date of lung cancer diagnosis to the date of death from any cause or until the date of the last follow-up. OSm was defined as the time from the date of BM diagnosis to the date of death from any cause or until the date of the last follow-up. OS and OSm rates were assessed by Kaplan-Meier methods and the log-rank test was used to compare time-to-event distributions. The data set was stratified and outcomes were compared by t-test or chi-squared test. Univariate analyses and a multivariate Cox proportional hazards regression were used to inspect all collected variables. Estimated risks of death were calculated using hazard ratios (HR) with 95% confidence intervals (CIs). The level of statistical significance was set at p < 0.05; all reported p values were two-tailed. The analyses were performed using the SPSS software package, version 19.0 for Windows (SPSS, Chicago, IL, USA).

Patient Characteristics
One hundred forty-one patients out of 1384 patients were retrospectively enrolled after the aforementioned inclusion and exclusion criteria were applied. Gender difference existed in terms of smoking status (never vs. ever); 98.9% of the female patients and 37.7% of the male patients had never smoked (p < 0.001). The median duration of TKI use was 13.2 months (95% confidence interval (CI), 10.1 to 16.2) in TKI + WBRT group and 10 months (95% CI, 7.3 to 12.8) in the TKI alone group. The mean durations of TKI use were 18.1±15.1 months and 15.4 ± 16.4 months for patients with and without WBRT, respectively (p = 0.327). In this cohort of 141 patients, 52 patients had more than one line of TKIs due to intolerance or disease progression. Table 1 summarizes the clinical characteristics of the 141 patients, divided by whether they had WBRT (TKI + WBRT group vs. TKI alone group).
All of them had EGFR-TKI. The mean and median age of this retrospective cohort was 64.5 years and 62 years respectively. Ninety-four patients (66.7%) received WBRT, and 47 patients (33.3%) did not. Patients who received WBRT were more likely to have surgery to their BM (38.3% in the TKI + WBRT group and 14.9% in the TKI alone group; p = 0.004); neurological symptoms (76.6% in the TKI + WBRT group and 53.2% in the TKI alone group; p = 0.005); larger BM (70.2% over 1 cm in the TKI + WBRT group and 53.2% in the TKI alone group; p = 0.046); and more BM (p = 0.043). No significant differences were observed in terms of age, gender, stage, initial clinical tumor and nodal classification, extracranial metastases, histological grading, EGFR mutation, primary lung surgery, number of lines of chemotherapy, name of EGFR-TKI, number of lines of TKI, mean duration of TKI use, smoking history, ECOG performance status at the time of BM, and dsGPA (all p > 0.05; Table 1).

OS and OSm
The median OS was 20.3 months (95% CI, 16.9 to 23.7) for the entire cohort. Seventeen and two patients were still alive in the TKI+WBRT group (18.1%) and TKI alone group (4.3%), respectively. The mean OS was longer for patients with WBRT (27.2 ± 16.7 vs. 21.6 ± 20.4 months, p = 0.033) The median OSm was 10.5 months (95% CI, 7.2 to 13.9) for the entire cohort. The combination group survived much longer after the diagnosis of BM. The median OSm was 14.3 months (95% CI, 9.5 to 18.3) in the TKI + WBRT group and 2.3 months (95% CI, 2 to 2.6) in the TKI alone group. The mean survival after BM was 18 ± 15.2 months and 7.1 ± 10.8 months for patients with and without WBRT, respectively (p < 0.001).
The 1-year OS rates were 81.9% and 59.6% with and without WBRT (p = 0.002). WBRT (p = 0.002), younger age (p = 0.003), female gender (p = 0.029) and surgery to primary lung cancer (p = 0.03) were favorable prognostic factors for longer 1-year OS rate ( Table 2). WBRT was a favorable prognostic factor for longer OS (p = 0.034; Figure 1A). To investigate the prognostic factors, we included five factors with p < 0.025 (WBRT, female gender, surgery to primary lung tumor, surgery to BM and smoking status) in a multivariable model (Table 3). WBRT was a strong favorable prognostic factor for longer survival (p < 0.001; Figure 1B).

Subgroup Analyses
In identifying potential differences in the benefits of WBRT for patients by the dsGPA score, there was a trend toward improved OS in the group of TKI + WBRT (p = 0.091, Figure 1C); furthermore, WBRT significantly improved OSm regardless of dsGPA score (p < 0.001, Figure 1D), while the mean BM-free survival rates were similar in both groups (9.2 ± 13.6 months vs. 14.5 ± 17.8 months, p = 0.312). As a result, OSm caused survival difference, and longer OSm contribute to longer OS. WBRT was a strong favorable prognostic factor for longer survival.

Discussion
We now routinely use molecular selection to identify patients with NSCLC who would benefit from target therapy. The Bureau of National Health Insurance of Taiwan reimburses TKIs prescribed after a diagnosis of stage IIIB or IV lung cancer. Target therapies have resulted in major shifts in the treatment paradigm for lung cancer [30]. Fifteen years ago, Omuro et al. reported that the incidence of the central nervous system as an initial failure site reached 33% in EGFR-TKI responders with advanced NSCLC regardless of disease control in the lungs [31]. Intrinsic resistance of metastatic clones, incomplete TKI penetration of the BBB and longer survival are possible explanations for this high incidence [31]. One retrospective study in Taiwan reported that more patients with advanced EGFR-mutated NSCLC died of BMs than did those with wild-type (44.8% vs. 8.3%, p < 0.001) [32]. This change in the causes of death was noted after the era of EGFR-TKI treatment. The present study found that WBRT prolonged OS in patients with EGFR-mutated NSCLC who developed BM.
Xu et al. stated that aggressive local ablative therapy including surgery or RT to all metastatic sites improved OS compared with local ablative therapy to partial sites or observation alone [33].
Magnuson et al. performed a retrospective study on the topic of the optimal sequence of stereotactic radiosurgery, WBRT, and EGFR-TKIs in patients with EGFR-mutated NSCLC who developed BM. They reported that upfront brain RT resulted in longer OS compared with upfront EGFR-TKIs (stereotactic radiosurgery with 46 months versus WBRT with 30 months versus EGFR-TKI with 25 months, p < 0.001) [34]. Li et al. also confirmed the use of upfront WBRT for patients with EGFR-mutated NSCLC and multiple BM improved OS [35]. Although the timing of WBRT was not involved in the present study, we demonstrated worsened OS without WBRT and that WBRT contributed to the addition of approximately one year of survival after the diagnosis of BM.
However, Ke et al. reported no statistically significant difference in the OS between the First-line EGFR-TKI-alone group and First-line EGFR-TKI plus WBRT [36]. It is worth noting that first-generation EGFR-TKIs hardly penetrate across the BBB at the recommended doses [24]. In their study [36], the performance status, dsGPA, surgery to primary or metastatic sites were not documented, and these factors might affect OS. He et al. reported that concurrent EGFR-TKI and WBRT significantly improved the median intracranial progression-free survival compared with EGFR-TKI alone (17.7 vs. 11.0 months, p = 0.015); however, there was no significant OS difference (28.1 vs. 24.0 months, p = 0.756) [37]. In their study, they prescribed three types of different TKIs (erlotinib, gefitinib and icotinib) and 20 patients in the group of 48 patients who were given EGFR-TKI alone initially received salvage WBRT upon BM progression. This group was not purely without WBRT. Lee [38]. Ng et al. found that one of the favorable prognostic factors was female gender (p <.001) in patients with NSCLC receiving WBRT [39]. In the present study, the median OSm was 14.3 months (95% CI, 9.5 to 18.3) in the TKI + WBRT group and 2.3 months (95% CI, 2 to 2.6) in the TKI alone group. On multivariate analysis, WBRT (p < 0.001) and female (p = 0.003) were associated with improved OS.
WBRT is associated with the risks of acute and late toxicities. Cognitive deficits attributed to RT were first reported in children treated for leukemia or brain tumors [40], and this bias was partly caused by greater susceptibility of the developing brain in youngsters [41]. BM by itself negatively affects cognitive function; additionally, baseline cognitive decline from aging in the cancer patients may also impact cognition [42]. Cognitive dysfunction can be caused by brain tumors, psychological distress, comorbidities such as vascular risk factors and diabetes, or by tumor-related epilepsy and its treatment (surgery, RT, anticonvulsants, chemotherapy, or corticosteroids) [40]. It can be difficult to differentiate from the effects of the tumor itself or RT complication [43]. Even though several recent publications have brought into question the role of WBRT and the possible risk of long-term neurotoxicity, WBRT curbed neurological decline [44]. A prospective study showed that the BBB permeability of gefitinib increased in accordance with escalated dose of WBRT [24]. An analysis from Radiation Therapy Oncology Group (RTOG) Study 91-04 showed that WBRT improved the scores on Mini-Mental State Examination in the patients with BM [45]. The optimization of WBRT with pharmacological and technical innovations to selectively spare organs involved in the memory process may decrease the potential long-term neurotoxicity [33].
At present, the treatment selection based on driver mutation status improves survival. Given the advancement of systemic therapy for extracranial lesions of metastatic NSCLC, patients now live long enough to develop BM [32]. WBRT, however, may be deferred and even omitted after the emergence of TKI by some clinicians. Based on prospective cohort studies, recently the European Society for Medical Oncology (ESMO) Clinical Practice Guidelines for metastatic NSCLC recommended the use of next-generation TKI for patients with a druggable oncogene driver (EGFR, ALK) and clinically asymptomatic BM [46]. Contradictory results were offered. Another retrospective study in North America reported that First-line WBRT for BM from EGFR/ALK-driven NSCLC was associated with longer time to intracranial progression than was radiosurgery or TKI alone [47]. For patients with ECOG 0-2 in the present study, the absence of WBRT was detrimental to their survival. In terms of different subgroups, even those with favorable dsGPA scores had survival benefit from the addition of WBRT compared with TKI alone.
The results of this study should be interpreted with caution, owing to the heterogeneity of patient characteristics and possible intrinsic bias related to the retrospective design. We intended to minimize bias by using multivariate analyses. There were some pitfalls of the present study. Firstly, radiosurgery was in general not used due to the regulations of National Health Insurance reimbursement. Secondly, we used OS rate to measure the clinical benefits, which might not represent the tumor response. Thirdly, cognitive evaluation was not fully documented. The precise roles of WBRT need to be validated in a randomized control trial. Moreover, Osimertinib is a third-generation EGFR-TKI developed specifically to treat patients with T790M mutation, and only 3.5% of the patients in the study cohort used Osimertinib.

Conclusions
The present study suggested that WBRT significantly prolonged OS in patients with EGFR-mutated NSCLC who developed BM. The combination of WBRT and TKI improved OS compared with TKI alone. To the best of our knowledge, this study is the first to demonstrate the negative survival impact from the omission of WBRT in patients with targetable driver mutation. A longer follow-up studying the role of multi-modality treatment in EGFR-mutated NSCLC with BM is urgently warranted.