Impact of Smoking and Brain Metastasis on Outcomes of Advanced EGFR Mutation Lung Adenocarcinoma Patients Treated with First Line Epidermal Growth Factor Receptor Tyrosine Kinase Inhibitors

Objectives This purpose of this study was to examine clinical-pathologic factors – particularly smoking and brain metastases – in EGFR mutation positive (M+) lung adenocarcinoma (ADC) to determine their impact on survival in patients treated with first line EGFR TKI. Methods A retrospective review of EGFR mutation reflex testing experience for all ADC diagnosed at a tertiary Asian cancer centre from January 2009 to April 2013. Amongst this cohort, patients with advanced EGFR M+ ADC treated with first line EGFR TKI were identified to determine factors that influence progression free and overall survival. Results 444/742 (59.8%) ADC reflex tested for EGFR mutations were EGFR M+. Amongst never-smokers (n=468), EGFR M+ were found in 74.5% of females and 76.3% of males, and amongst ever smokers (n=283), in 53.3% of females and 35.6% of males. Exon 20 mutations were found more commonly amongst heavy smokers (> 50 pack years and > 20 pack years, Pearson’s chi square p=0.044, and p=0.038 respectively). 211 patients treated with palliative first line TKI had a median PFS and OS of 9.2 and 19.6 months respectively. 26% of patients had brain metastasis at diagnosis. This was significantly detrimental to overall survival (HR 1.85, CI 1.09-3.16, p=0.024) on multivariate analysis. There was no evidence that smoking status had a significant impact on survival. Conclusions The high prevalence of EGFR M+ in our patient population warrants reflex testing regardless of gender and smoking status. Smoking status and dosage did not impact progression free or overall survival in patients treated with first line EGFR TKI. The presence of brain metastasis at diagnosis negatively impacts overall survival.


Introduction
EGFR tyrosine kinase inhibitors (TKI) such as gefitinib and erlotinib, are now established first line treatment options for EGFR mutation positive (EGFR M+) lung adenocarcinoma (ADC), demonstrating significant improvement in progression free survival (PFS) over platinumbased doublet chemotherapy [1][2][3][4][5][6][7]. Previous studies examining the impact of smoking history on TKI response often reflect surrogacy for EGFR mutations and majority of phase III studies were enriched for never smokers. A recent retrospective study suggested that smoking history and smoking dosage may be associated with significantly poorer response rates and survival outcomes in EGFR mutation positive non-small cell lung cancer (NSCLC) [8]. However, this finding is confounded by the fact that a greater proportion of smokers had received EGFR TKI beyond the second and third line setting, and the impact of smoking on survival in EGFR mutation positive NSCLC patients receiving first line EGFR TKI remains unclear [9].
Due to the high incidence of EGFR mutations in Asian ADC compared to the West [10][11], many academic hospitals, including our centre, have adopted reflex testing for EGFR mutations. As cost effectiveness of EGFR TKI is driven by patient selection based on EGFR mutation status [12], it is important to define the prevalence of the mutation in both smokers (current and ex-smokers) as well as never smokers through systematic testing of consecutive cases. Clinical pathologic factors such as smoking status [8], location of EGFR mutation [13], and presence of brain metastases [14] may impact on treatment outcomes. Of particular interest, brain metastasis in EGFR mutation positive NSCLC is a common site of involvement at diagnosis and treatment failure-occurring in up to 23% of newly diagnosed patients [15]. Elucidating prognostic factors in EGFR mutant ADC treated with first line TKI will facilitate improved stratification and identify therapeutically challenging patient subgroups.
In this study, we report our reflex EGFR testing experience on consecutive lung adenocarcinomas seen in an Asian tertiary cancer centre and determine the prevalence of EGFR mutations by gender and smoking status. Relationships between mutation spectra and clinical characteristics such as age, gender, ethnicity and smoking status were also explored. Further, in those who had received first line treatment with an EGFR TKI, we examined clinical pathologic characteristics that had an impact on survival.
Prior to 1st June 2010, EGFR mutation testing in our centre for patients with newly diagnosed ADC was ordered as per physician discretion. From 1st June 2010 all ADC samples identified by the pathologists were reflex tested for EGFR mutations, regardless of stage and smoking status.
Smoking status for patients was obtained from electronic medical records and Lung Cancer Consortium Singapore, where patients' lifestyle factors were captured through interviews by research coordinators. Patients were classified as never smokers (NS), and ever smokers (exsmokers [quit 1 year] and current smokers) (ES). NS were defined as those who had smoked less than 100 sticks in their lifetime.
Patients with Stage 4 EGFR M+ disease who had received TKI therapy were evaluated. Patients who had already been commenced on TKI based on their phenotypic traits (Asian, nonsmoker), but subsequently underwent confirmatory EGFR mutation testing were also included. Baseline imaging was evaluated for presence or absence of brain metastases, and response evaluation scans were performed according to physician discretion, ranging from within 6 weeks from start of TKI for the first scan, to 8-12 weekly for subsequent scans. First line TKI treated EGFR M+ ADC patients were analysed for progression free survival (PFS) and overall survival (OS), and an exploratory analysis was done for factors that influenced survival.

EGFR Mutational Analysis
EGFR mutation analysis was carried out by Sanger sequencing on genomic DNA that had been extracted from formalin fixed paraffin embedded (FFPE) tissue samples using Qiagen FFPE DNA extraction kit. Extracted DNA was then subject to polymerase chain reaction (PCR) amplification of exons 18, 19, 20 and 21, and products were analysed by direct Sanger sequencing.

Statistical Analysis
Continuous clinical variables were summarized using median and range, while categorical variables were summarized by frequency and percentage. Pearson's chi-squared test (or Fisher's exact test if there were expected cell frequencies less than 5) was used to test for associations between patient characteristics and mutation type, smoking status and presence of brain metastasis at diagnosis. The strength of the association was estimated using Cramer's V.
PFS was calculated as the duration from start of TKI to the PFS date defined by clinico-radiological progression of disease on CT imaging as per RECIST 1.1. OS was calculated as the duration of time from start of TKI to date of demise. The Kaplan-Meier method was used to estimate survival functions. The log-rank test was used to determine if there was a difference in survival curves between different groups. A 2-sided p-value of less than 0.05 was taken as statistically significant. Univariate and multivariate analysis was performed using the Cox proportional hazards model. All analyses were performed in Stata (Version 12.1; StataCorp, Texas, USA).

Ethics
This research was approved by SingHealth CIRB (CIRB 2010/516/B) and all clinical investigation was conducted according to the principles expressed in the Declaration of Helsinki. The data was collected and analysed anonymously then reported.

Reflex Testing Experience
A total of 994 cases of ADC were seen in our centre between 1st January 2009 and 18th April 2013. Reflex testing was adopted from 1 st June 2010. A 6 monthly review from 1 st June 2010 to 19 th April 2013 revealed that 87-94% of all ADC cases were tested, with an EGFR M+ rate of 50-68% (S1 Table). A total of 742 patients were tested for EGFR mutations. 444 (59.8%) were positive and 289 (38.9%) negative for mutations. 9 cases were unsuccessfully profiled due to incomplete sequencing (n = 5: unsuccessful for Exon 18, n = 1; Exon 20, n = 2; Exons 20/21, n = 1 and Exons 19/20/21, n = 1), tumor content of less than 15% raising the possibility of a false negative result (n = 1) or insufficient tissue for any analysis (n = 4). None of these patients underwent a repeat biopsy and they were treated as for ADC that was EGFR mutant negative. Hence the ascertainment rate with EGFR sequencing in our centre was 98.8%.
The respective EGFR M+ rates amongst female NS, female ES, male NS and male ES were 75.1%, 53.3%, 76.9%, and 35.7% (Fig 1). ES were further stratified by pack years based on available information regarding their smoking histories. EGFR M+ rates were at least 20 percent in both male and female heavy smokers but further interpretation was limited by the small numbers in these sub-groups (Fig 2).

Impact of smoking dosage and EGFR mutations
Heavy smokers with more than 50 pack years appeared to have more mutations in exon 20 than those with less pack years or who had never smoked (p = 0.044, Cramer's V = 0.145; weak association). A less stringent definition for heavy smokers (>20 pack years) still showed a significant association (18% vs. 6%, p = 0.038, Cramer's V = 0.155; weak association). There was no significant association between type of mutation (indel vs. point) and heavy smoking status. chemotherapy, 28 (9.5%) received a combination of TKI and other agents including chemotherapy and combination targeted therapy in a phase I setting. 46 patients were censored as they were either lost to follow up, on best supportive care, declined treatment or passed away prior to commencing TKI. Of the 55 patients who did not receive TKI in the first line, 33 went on to receive TKI as second line therapy, 5 as third line, and 2 as fourth line. The remaining 15 never received EGFR TKIs in their lifetime. In total, 279 out of 294 (94.9%) EGFR M+ ADC received systemic therapy with a TKI in their lifetime.

Clinical characteristics of EGFR TKI treated Stage IV NSCLC
A homogenous population of 211 patients with Stage IV disease treated with first line TKI were evaluated for survival on TKI and smoking history and other clinical characteristics were analysed for impact on survival. The median age of the cohort was 62 (33-84). 128 (60.7%) were females. 166 (78.7%) had never smoked, 32(15.1%) were ex-smokers and 13 (6.2%) were smokers. Amongst ever smokers with accurate smoking histories (n = 32), the median number of pack years was 30 pack years (ranging from 0.9 to 102). 26% had proven brain metastasis at time of diagnosis. Other key characteristics are summarised in Table 1.
Baseline characteristics were compared between patients who were never smokers vs. ever smokers and those without brain metastasis at diagnosis vs. those with brain metastasis at  diagnosis (S2 and S3 Tables). There was a strong correlation between sex and smoking status (p<0.001, Cramer's V = 0.576). 93.3% of ever smokers were male. Significant but weak associations were also observed between age and smoking status (p = 0.012, Cramer's V = 0.173), type of mutation and smoking status (p = 0.044, Cramer's V = 0.173), age and presence of brain metastasis (p = 0.014, Cramer's V = 0.169), and ECOG performance status and presence of brain metastasis (p = 0.017, Cramer's V = 0.181).
The progression free and overall survival of this specific population follows (Tables 2 and  3). Since the ever smokers were mostly male, smoking status was combined with sex into a single variable for multivariate analysis. Males were sub-categorised into never smokers and ever smokers.

Progression Free Survival of EGFR TKI treated Stage IV Disease
The median PFS of all patients was 9.2 months (95% CI 8.1-11.2 months). Patients with single non-exon 19/21 mutations and double mutations had inferior PFS (p = 0.013) as compared to those with exon 19/21 mutations.
There was no significant difference in the PFS of ever (current and ex-), and never smokers (p = 0.844) nor between heavy smokers and others (p = 0.510 for 50 pack years, and p = 0.895 for 20 pack years) ( Table 2).
On multivariate analysis, there was a trend for an inferior PFS in patients with brain metastasis at diagnosis (HR 1.56, CI 0.99-2.45, p = 0.053) ( Table 3). Kaplan-Meier plots suggested that the presence of brain metastasis did not worsen progression free survival in patients who were ECOG 2-4 (Fig 4). This interaction between brain metastasis and ECOG status was incorporated into the multivariate model.

Overall Survival of EGFR TKI treated Stage IV Disease
The overall median OS for all patients was 19.6 months (95% CI 16.4-23.3). The presence of brain metastasis at diagnosis was associated with a significantly worse OS in both univariate (p = 0.029), and multivariate (HR 1.82, CI 1.07-3.11, p = 0.028) analysis. Similar to findings on analysis for progression free survival, Kaplan-Meier plots suggested that the presence of brain metastasis did not worsen overall survival in patients who were ECOG 2-4, and this was incorporated into the multivariate model. The median OS of patients with no brain metastasis at diagnosis was 22.1 months while that of patients with brain metastasis at diagnosis was 17.9 months. Age, sex, ethnicity, smoking status (never vs. ever; never vs ex vs current; pack year < 50 vs pack year 50; pack year < 20 vs pack year 20), functional status, location and type of mutations had no significant impact on overall survival (Table 2). Multivariate analysis showed that smoking had no significant impact on overall survival (Table 3).

Subset analyses in defined patient cohorts
Multivariate analysis of progression free survival and overall survival was performed amongst 2 defined patient cohorts to further investigate the effect of confounding factors. The first subset comprised 82 males (40 never smokers and 42 ever smokers) and was examined for the possible impact of smoking status on survival. Due to small sample size, the analysis was adjusted for ECOG status only. There was no evidence that smoking status was related to survival in this analysis (Table 4). A second subset comprised 70 female never smokers aged 65 with ECOG 0-1 at diagnosis (48 had no brain metastasis at diagnosis and 22 had brain metastasis at diagnosis) and was studied for the impact brain metastasis at diagnosis had on survival. After adjusting for type of mutation, brain metastasis at diagnosis had a significant impact on both progression free survival (HR 2.06, CI 1.03-4.11, p = 0.041) and overall survival (HR 2.86, CI 1.20-6.84, p = 0.018) ( Table 5) in this cohort.

Discussion
In our study, systematic testing of 762 lung adenocarcinoma for EGFR mutations revealed a prevalence of 61%-regardless of smoking status. Notably, we found that as high as 36% of male and 53% of female patients with a smoking history harboured EGFR mutations, and similar prevalence in males and females amongst never smokers (76.9% vs 75.1%)-underscoring the inadequacy of EGFR mutation testing based on clinical phenotype alone. This is especially relevant as cost-effectiveness of EGFR TKI is driven by patient selection based on mutation status, supporting reflex testing in our patient population regardless of smoking status and gender. Our cohort of 211 EGFR mutant patients with stage IV NSCLC treated with first line EGFR TKI represents one of the largest to date examining prognostic factors. On multivariate analysis, taking into account EGFR mutations (exon 19 deletions versus L858R) and smoking status, the presence of brain metastasis at diagnosis-comprising 26% of our cohort-was an independent risk factor for worse overall survival, corroborating a recently reported retrospective series [16]. Although clinical activity of EGFR TKIs against intracranial disease has been previously described [17][18][19][20], most studies have comprised of modest-sized, heterogeneous and selected    [21]. In contrast, majority of patients in our study (91%) received gefitinib. It remains to be elucidated if erlotinib might exhibit superior control of intracranial disease due to higher CNS penetration and drug concentrations achieved compared to gefitinib [22]. Additionally, alternate treatment approaches such as pulsed high-dose strategies are actively being explored [23]. Further prospective studies addressing CNS disease are warranted, and should focus on improved delineation of leptomeningeal and cerebral metastases, quantity and quality of CNS disease (extent and symptoms), as well as optimal sequencing of EGFR TKI and radiation. Nevertheless, in the era of TKIs, EGFR M+ ADC patients with brain metastasis at diagnosis achieve a median survival of 17.9 months.
To date, there is limited data on the impact of smoking on progression free survival in patients treated with first line EGFR TKI. In three phase III trials that included up to 38% of ever smokers, subgroup analysis suggested no PFS benefit with TKI over conventional chemotherapy in contrast to those who had never smoked [5][6][7]. Retrospective studies evaluating the impact of smoking on the survival in EGFR M+ ADC [8,[24][25][26] have also yielded mixed results with only 1 study suggesting that smoking dosage has an impact on survival (Table 6). In our  [8]. We acknowledge that our sample size was limited and that smoking status was confounded with gender, which we have tried to address. It is noteworthy that majority of smokers in the Korean study received EGFR TKI in the setting of second, third line or beyond, where cumulative toxicities and poor functional status may confound outcomes. The presence or absence of brain metastasis was also not reported in that study.
The retrospective nature of our study limits our ability to evaluate on-treatment CNS activity for EGFR TKI, due to irregular frequency and thoroughness of follow-up radiological evaluation. In addition, there may yet be unaccounted bias between those with and without brain metastases, although the commonly reported factors such as performance status, smoking status and type of EGFR mutation were addressed. To our knowledge, this is the first study examining these clinical parameters in a large EGFR M+ patient cohort predominantly treated with first line gefitinib.
In conclusion, our study highlights the importance of reflex molecular testing, where as high as 36% of ever smoker males were found to harbour an EGFR mutation. Furthermore, despite the higher mutational burden in smokers [27], the observation that smoking status did not impact on PFS or OS in patients treated with first line EGFR TKI, suggests a hierarchical relationship in genetic alterations, where dominant truncal events-such as EGFR mutationsremain therapeutically tractable. On the other hand, the presence of brain metastases emerged as an independent negative prognostic factor, and continues to be a therapeutic challenge in EGFR M+ NSCLC.
Supporting Information S1 Table. Reflex testing of all newly lung adenocarcinoma cases for mutations in EGFR was commenced from 1 st June 2010, with current rates of testing between 87 to 94%. EGFR mutations were found in 50-68% of cases. Prior to the implementation of reflex testing, patients were tested for EGFR mutations based on physician discretion. A total of 742 patients underwent EGFR mutation testing. 444 (59.8%) were positive and 289 (38.9%) negative for mutations. 9 cases were unsuccessfully profiled. Hence the ascertainment rate with EGFR sequencing in our centre was 98.8%. (DOCX) S2  Table. Baseline characteristics of patients without brain metastasis at diagnosis vs. those with brain metastasis at diagnosis amongst 211 patients treated with 1 st line TKI.