The authors previously demonstrated that never-smokers with stage IIIB/IV nonsmall cell lung cancer (NSCLC) lived 50% longer than former/current smokers. This observation persisted after adjusting for age, performance status, and sex. In this study, the authors hypothesized that smoking-dependent differences in the distribution of driver mutations may explain differences in prognosis between these subgroups.
In total, 293 never-smokers and 382 former/current smokers with lung adenocarcinoma who underwent testing for epidermal growth factor receptor (EGFR) mutations and v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS) mutations and rearrangements in anaplastic lymphoma kinase (ALK) between 2009 and 2010 were investigated. Clinical outcomes and patient characteristics were collected. Survival probabilities were estimated using the Kaplan-Meier method. Group comparison was performed with log-rank tests and Cox proportional hazards methods.
Although the overall incidence of these mutations was nearly identical (55% never-smokers vs 57% current/former smokers; P = .48), there were significant differences in the distribution of mutations between these groups for EGFR mutations (37% never-smokers vs 14% former/current smokers; P < .0001), KRAS mutations (4% never-smokers vs 43% former/current smokers; P < .0001), and ALK rearrangements (12% never-smokers vs 2% former/current smokers; P < .0001). Among never-smokers and former/current smokers, the prognosis differed significantly by genotype. Patients who had KRAS mutations had the poorest survival. Smoking status, however, had no influence on survival within each genotype.
We have witnessed a sea change in the management of patients with nonsmall cell lung cancer (NSCLC) in the past decade. Before 2004, biologic diversity was observed largely through the lens of histology, with patients grouped into adenocarcinoma, squamous, and large cell carcinoma subtypes. The predictive and prognostic power of this schema was limited, however. It was not until the initial efficacy studies of the epidermal-growth factor receptor (EGFR) tyrosine kinase inhibitors (TKIs) in patients with advanced NSCLC that evidence for a mechanistic model of disease heterogeneity emerged.1, 2 Although the overall response rate to EGFR TKIs in unselected patients was a modest 10%, subgroup analyses revealed comparatively higher response rates and survival in women, never-smokers, and Asians.1-3 The subsequent identification of mutations4-6 in the tyrosine kinase domain of EGFR that predicted for response to EGFR TKIs made it clear that the limited efficacy observed in those early studies was a by-product of the relatively low frequency of EGFR mutations in NSCLC as a whole. Conversely, it was determined that those subpopulations with higher response rates had a greater proportion of EGFR mutations, approaching 50% in never-smokers with lung adenocarcinoma. In the wake of these findings, genotyping efforts have identified driver mutations in the majority of lung adenocarcinoma specimens.7 Mutations in v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS) account for approximately 25% of specimens8; mutations in EGFR account for 15% of specimens; rearrangements of anaplastic lymphoma kinase (ALK) account for 3% to 7% of specimens9; and mutations in human epidermal growth factor receptor 2 (HER2), v-raf murine sarcoma viral oncogene homolog B1 (BRAF), and phosphoinositide-3-kinase, catalytic, alpha polypeptide (PIK3CA) account for 1% to 2% of specimens.
In addition to smoking-dependent variations in the incidence of EGFR mutations,6, 10 differences in the frequency of KRAS mutations between subgroups also have been described. As few as 8% of never-smokers and as high as 57% of former/current smokers with lung adenocarcinoma harbor KRAS mutations.11 Although 3% to 5% of unselected patients with lung adenocarcinoma have ALK rearrangements, the frequency appears to be much greater in never-smokers.12 There is a suggestion that the prognosis associated with these mutations also varies. Although treatment with EGFR TKIs likely has altered the prognosis of patients with EGFR mutations,13-15 the prognostic relevance of KRAS mutations is less well characterized. The prognostic significance of ALK rearrangements remains largely unknown.
We recently reported data on the prognostic impact of smoking history in patients with stage IIIB/IV NSCLC.16 We observed that patients who were never-smokers had a longer overall survival (OS) compared with patients who smoked <15 pack-years and ≥15 pack-years. This relationship persisted even after adjusting for differences in age, performance status, and sex on multivariate analysis.
We hypothesized that tumor biology was the principle driver of the differences in survival between these subgroups. We therefore reviewed the mutation status, clinical characteristics, and survival of never-smokers and former/current smokers with lung adenocarcinomas at our institution.
MATERIALS AND METHODS
Study Design and Patients
Patients with lung adenocarcinomas who were tested for mutations in EGFR (exon 19 deletions, exon 21 leucine to arginine codon 858 [L858R] substitutions) and KRAS as well as rearrangements in ALK at Memorial Sloan-Kettering Cancer Center and who had tissue obtained between May 2009 and May 2010 were reviewed. Testing was performed under a reflex molecular profiling program for patients with lung adenocarcinoma histologies (Lung Cancer Mutation Analysis Program) without selection by any specific pathologic or clinical feature. Smoking history was determined through the use of a prospectively administered questionnaire. Never-smokers were defined as those patients who smoked <100 cigarettes in their lifetime. Medical records were reviewed to determine sex, ethnicity, age, Karnofsky performance status (KPS), stage, and treatment history. All chart review/tissue collection was approved by the Memorial Sloan-Kettering Cancer Center Institutional Review Board/Privacy Board. For the purposes of the current analysis, patients were grouped into those with early stage disease (stages I-IIIA) and those with advanced stage disease (stages IIIB/IV and recurrent disease) according to the International Association for the Study of Lung Cancer seventh edition TNM staging system.
EGFR exon 19 deletions and exon 21 L858R point mutations were detected using a polymerase chain reaction (PCR)-based assay.17 A 207-base pair (bp) genomic DNA fragment encompassing exon 19 was amplified using the following primers: forward 1 (FW1), 5′-GCACCATCTCACAATTGCCAGTTA-3′; and reverse 1 (REV1), 5′-Fam-AAAAGGTGGGCCTGAGGTTCA-3′. A 222-bp genomic DNA fragment spanning exon 21 was amplified using the following primers: FW1, 5′-CCTCACAGCAGGGTCTTCTCTGT-3′; and REV1, 5′-Fam-TCAGGAAAATGCTGGCTGACCTA. PCR products were subjected to capillary electrophoresis on an ABI 3730 Genetic Analyzer (Applied Biosystems, Foster City, Calif). KRAS exon 2 mutations were identified through direct sequencing,18 and rearrangements in ALK were identified through fluorescence in situ hybridization (FISH) using a dual-color, break-apart probe.19 Positive cases were defined as the presence of a split signal indicating rearrangement of the ALK locus at region 2 of the short arm of chromosome 23 (2p23) or the presence of a single red signal indicating loss of the 5′ DNA sequence in ≥15% of cells. When tissue was available, PCR for specific echinoderm microtubule-associated protein-like 4-ALK (EML4-ALK) transcript variants was performed to confirm the presence of an EML4-ALK translocation. On the basis of past data demonstrating the nonoverlapping nature of mutations in EGFR and KRAS and rearrangements in ALK,20, 21 only patients who had wild-type (WT) EGFR and KRAS underwent testing for ALK rearrangements.
For patients with advanced-stage disease, OS was measured from the date of diagnosis of stage IIIB/IV or recurrent disease until the date of death. Patients who did not die during the study period were censored at the time of last available follow-up. Survival and follow-up data were obtained through medical records or the Social Security Death Index and were updated as of June 2011. Survival probabilities were calculated using the Kaplan-Meier method. Group comparison was performed with log-rank tests (for univariate analyses) and Cox proportional hazards methods adjusted for age, sex, and KPS (for multivariate analyses). We examined the effect of smoking history on OS within each genotype (EGFR, KRAS, ALK, and other/uncharacterized), and conversely, the effect of genotype within each smoking subgroup (never-smokers and former/current smokers). Because of the small number of patients in some subgroups defined by the combination of genotype/smoking history, it was not feasible to fit a comprehensive model that examined the interaction between the 2 variables.
Patients became eligible for the study at the time of their molecular diagnosis. To account for the potential length-time bias associated with differences between the date of diagnosis of advanced disease and the date of molecular testing, all analyses were performed using left truncation (or delayed entry) techniques. With this method, survival probabilities and hazard ratios are calculated conditional on patients having survived until the date of their molecular diagnosis. Statistical analyses were performed using SAS statistical software (SAS Institute, Inc., Cary, NC) and the “survival” package in R (R Foundation for Statistical Computing, Vienna, Austria; available at: http://www.r-project.org/; [Accessed January 2012]).
Patient Characteristics and Mutation Frequency
Of the 675 patients with lung adenocarcinoma who we analyzed between May 2009 and May 2010, 293 (43%) were never-smokers, and 382 (57%) were former/current smokers. Patient characteristics are summarized in Table 1. There were no significant differences in sex, age, stage, or KPS between never-smokers with EGFR and KRAS mutations. Never-smokers who harbored an ALK rearrangement were more likely to be men compared with patients who had EGFR mutations (49% vs 30%; P = .02) or KRAS mutations (49% vs 8%; P = .02). Never-smokers with ALK rearrangements were significantly younger than patients with EGFR mutations (median age, 57 years vs 63 years; P = .006). Among former/current smokers, no significant differences in clinical characteristics were present between patients with EGFR mutations, KRAS mutations, and ALK rearrangements.
Table 1. Patient Clinical Characteristics by Genotype and Smoking History
The relative distribution of mutations between never-smokers and former/current smokers is illustrated in Figure 1. The overall incidence of EGFR and KRAS mutations and rearrangements in ALK was virtually identical in never-smokers (55%; 95% confidence interval [CI], 48%-60%) and former/current smokers (57%; 95% CI, 52%-63%; P = .43). Never-smokers had a significantly higher incidence of EGFR mutations (38% vs 14%; P < .0001) and ALK rearrangements (12% vs 2%; P < .0001) than former/current smokers. In contrast, KRAS mutations were more common in former/current smokers than in never-smokers (41% vs 5%; P < .0001). The majority of patients had an identifiable mutation in 1 of these genes.
There were no significant differences in the frequencies of EGFR exon 19 deletions versus L858R substitutions in never-smokers (56% vs 43%; P = .22) or former/current smokers (44% vs 56%; P = .50). Glycine-to-aspartic acid codon 12 (G12D) transition mutations (43%) were the single most common KRAS variant in never-smokers. G12C and glycine-to-valine codon 12 (G12V) transversion mutations were most common in former/current smokers (64%). Rearrangements in ALK were detected predominantly through the presence of a split FISH signal in both groups.
The median number of treatments received by patients within each genotype is listed in Table 2 and did not vary by smoking status. Evaluation of treatment response was not possible given the absence of uniform pretreatment and post-treatment imaging. Most patients with an EGFR mutation, regardless of smoking history, received treatment with an EGFR TKI (never-smokers vs former/current smokers, 88% vs 76%; P = .57). Treatment with an EGFR TKI also was received by 7% to 27% of patients without a documented EGFR mutation, the frequency of which was greatest in patients with ALK rearrangements. There was no significance difference between never-smokers and former/current smokers in the proportion of patients with KRAS mutations or ALK rearrangements who received an EGFR TKI. Forty percent of never-smokers and former/current smokers with ALK rearrangements had received crizotinib (Xalkori; Pfizer, New York) as part of a clinical trial, with a median duration of therapy of 7 months in each group.
Table 2. Treatment by Genotype in Patients With Advanced Stage Lung Adenocarcinoma
Of the 440 patients with advanced-stage disease who were included in this analysis, 228 patients died during the analysis period. The median OS of never-smokers versus former/current smokers with advanced disease was 20 months (95% CI, 16-27 months) versus 12 months (95% CI, 11-13 months), respectively. This difference was significant on univariate (P < .001) and multivariate analyses adjusting for age, sex, and KPS (adjusted hazard ratio [HR], 1.9; 95% CI, 1.4-2.5; P < .001). The median OS for all patients with EGFR mutations, KRAS mutations, and ALK rearrangements was 23 months (95% CI, 19-36 months), 11 months (95% CI, 8-13 months), and 35 months (95% CI, 20 months to not reached), respectively. There were significant differences in survival between patients with KRAS mutations and EGFR mutations (adjusted HR, 2.3; 95% CI, 1.5-3.6; P < .001) and patients with KRAS mutations and ALK rearrangements (adjusted HR, 2.9; 95% CI, 1.4-5.8; P < .001). There was no significant difference in survival between patients with ALK rearrangements and EGFR mutations (adjusted HR, 0.8; 95% CI, 0.4-1.6; P = .42).
These differences persisted within the never-smoker and former/current smoker subgroups. Among advanced-stage never-smokers, patients who had KRAS mutations had a higher risk of death compared with patients who had EGFR mutations (adjusted HR, 2.7; 95% CI, 1.1-6.7; P = .04) and patients with ALK rearrangements (adjusted HR, 4.6; 95% CI, 1.5-14.1; P = .008). There was no significant difference in survival between never-smokers with ALK rearrangements versus EGFR mutations (adjusted HR, 0.6; 95% CI, 0.2-1.4; P = .23).
Among smokers, patients with KRAS mutations had significantly worse survival relative to patients with EGFR mutations (adjusted HR, 3.3; 95% CI, 1.6-5.7; P < .0001). Meaningful survival comparisons with smokers who harbored ALK rearrangements were not possible given the small numbers detected, although there were trends in survival similar to those observed among never-smokers and the overall cohort. A planned analysis of disease-free survival among early stage patients could not be performed given the absence of events to date.
Despite these differences in genotype-specific survival, there were no differences in survival between never-smokers and former/current smokers who harbored a given mutation. The median OS of never-smokers versus former/current smokers with EGFR mutations was 25 months (95% CI, 19 months to not reached) versus 19 months (95% CI, 16-37 months; P = .33). The median OS of never-smokers versus former/current smokers with KRAS mutations was 11 months (95% CI, 8-13 months) versus 10 months (95% CI, 6 months to not reached; P = .77). The median OS survival of never-smokers versus former/current smokers with ALK rearrangements was 26 months (95% CI, 9 months to not reached) versus 44 months (95% CI, 20 months to not reached; P = .41). The associated Kaplan-Meier survival curves are provided in Figures 2A-C.
Multivariate analysis adjusted for age, sex, and KPS also revealed no significant differences between former/current smokers and never-smokers with EGFR mutations (adjusted HR, 0.9; 95% CI, 0.4-2.0; P = .80) and KRAS mutations (adjusted HR, 1.1; 95% CI, 0.5-2.5; P = .80). Multivariate analysis for patients with ALK rearrangements could not be performed because of small sample sizes in both groups.
Although mutations in EGFR and KRAS and rearrangements in ALK comprise the majority of driver mutations in both never-smokers and former/current smokers, some 40% of patients in each group lacked these genetic aberrations. These other/uncharacterized patients were the only subgroup to exhibit a significant difference in OS when grouped by smoking status (median OS, 18 months vs 12 months; unadjusted HR, 2.0; 95% CI, 1.33-2.98; P < .001) (Fig. 3). Similar findings were observed on multivariate analysis (adjusted HR, 1.8; 95% CI, 1.2-2.7; = .006).
In an attempt to explain the survival advantage that never-smokers exhibit over smokers with lung adenocarcinomas, we hypothesized that a unique distribution of driver mutations in never-smokers, enriched for those mutations with better prognoses, leads to an improved outcome compared with smokers.
Our data demonstrate that, whereas the overall incidence of driver mutations is identical, the distribution of mutations in patients with lung adenocarcinomas differs significantly based on smoking status. Never-smokers had a significantly higher proportion of EGFR mutations and ALK rearrangements, totaling 50%. Conversely, former/current smokers had a significantly higher proportion of KRAS mutations, totaling 41%.
It is noteworthy that we observed no significant differences in the OS of never-smokers and former/current smokers who had identical genotypes. To our knowledge, this observation is the first of its kind to suggest that former/current smokers, independent of genotype, do not have a poorer prognosis compared with never-smokers. Although our analysis was relatively large, the low incidence of ALK rearrangements was a limiting factor in the survival analysis of these patients. Although there was a numerical difference in survival among never-smokers and former/current smokers with ALK rearrangements, the difference was not significant. Ultimately, a larger sampling of patients will be needed to clarify our observation.
We note that former/current smokers with EGFR mutations and ALK rearrangements in our analysis were not light tobacco consumers, with median pack-years smoked of 18 and 15, respectively. These data are particularly compelling when placed beside those from the recent whole genome sequencing of a smoking-related lung adenocarcinoma tumor, suggesting that the high somatic mutation rate caused by smoking may not substantially alter the course of tumors driven by certain oncogenes, such as EGFR and KRAS.22
It is important to note that this study was not sufficiently equipped to resolve the impact of targeted therapy versus standard chemotherapy in these subpopulations. Data suggest that both EGFR TKIs and ALK-directed therapies alter the clinical courses of patients who harbor sensitizing EGFR mutations and ALK rearrangements relative to chemotherapy.13, 23 This study does not address the untreated natural histories of the genotypes tested. The prognoses of each mutation subtype undoubtedly were affected by the proportion who received an appropriate targeted therapy: this has bearing on the current study only insofar as significant disparities in the delivery of a targeted therapy may exist between never-smokers and smokers. This was not the case. Indeed, as new and improved treatments emerge for patients with specific genotypes, we anticipate that the prognostic differences that currently exist between smokers and never-smokers likely will change. The promise of this is perhaps greatest for patients whose tumors harbor KRAS mutations, for whom an effective, directed therapy has not yet been identified.
Forty percent of patients with lung adenocarcinomas do not harbor mutations in EGFR or KRAS or rearrangements in ALK. Our data indicate the presence of an as-yet-unknown oncogenic driver event or series of events that is present differentially among never-smokers and former/current smokers in this subgroup of patients. To date, mutations in BRAF (3%, predominantly in former/current smokers),24HER2 (4.8% overall, up to 8% in never-smokers),25PIK3CA (2%),26 and AKT1 (1%) have been identified in addition to those tested in the current analysis. The prognoses associated with these mutations, however, are unclear.
Although we report herein that all EGFR and KRAS mutations were mutually exclusive, the data presented here cannot comment rigorously on the mutual exclusivity of ALK rearrangements and mutations in EGFR and KRAS based on the mutational testing protocol we used. Once either an EGFR or a KRAS mutation was detected, ALK testing was not pursued routinely. Consequently, all patients with ALK rearrangements had WT EGFR and KRAS; however, we did not test all patients with EGFR and KRAS mutations for the coincident presence of ALK rearrangements. Our routine testing procedures were based on past series and our own internal series (article in preparation) that have demonstrated mutual exclusivity of these oncogenic events.20, 21 Recent data from the Lung Cancer Mutation Consortium demonstrated overlapping EGFR and KRAS mutations in 3 of 38 (8%) tumor samples that harbored rearrangements in ALK.27 We observed that the frequency of coincident EGFR or KRAS mutations in the 44 ALK rearranged tumors in our series was 0% (95% CI, 0%-10%).
From a practical standpoint, although we advocate stratifying patients by genotype in clinical studies of targeted therapies, we recognize that this is sometimes not possible, particularly when a predictive biomarker has not been well characterized. Furthermore, despite state-of-the-art molecular testing, a known mutation will not be identified in many patients. Our data suggest that stratification of patients by smoking history should be performed in all trials in which genotype is not a prespecified eligibility criterion, given the significant prognostic difference that we observed. Various proportions of never-smokers and former/current smokers within treatment arms may confound survival results and should be accounted for at study inception rather than post hoc. Finally, we note that, although comorbidity, either in range or severity, was not captured by us as part of this study, its influence on survival apart from performance status was previously demonstrated by Toh et al to be nonsignificant on both univariate and multivariate analysis.28
In conclusion, our data demonstrate that never-smokers and former/current smokers with lung adenocarcinomas are not homogeneous subgroups. Each group is made up of individuals with a set of disparate mutations that, in sum, generates an overall prognosis. Never-smokers carry a higher proportion of EGFR mutations, but this should not lead to reflexive treatment of never-smokers with an EGFR TKI. Conversely, EGFR mutations do occur in former/current smokers, who exhibit similar survival outcomes as never-smokers when they receive treatment with EGFR TKIs. All patients with lung adenocarcinoma, regardless of smoking history, should undergo testing for EGFR mutations and rearrangements in ALK in an effort to match patients with an appropriate targeted therapy. Patients who are WT for EGFR and KRAS and who do not exhibit ALK rearrangements should have their tumors prioritized for future molecular profiling efforts. Clinical trials of unselected patients with lung adenocarcinoma and trials of patients who do not harbor mutations in EGFR/KRAS or rearrangements in ALK should be stratified by smoking history.
This work was supported by a grant from the National Cancer Institute (P01-CA129243).
CONFLICT OF INTEREST DISCLOSURES
Mark G. Kris is a consultant to Pfizer, Boehringer-Ingelheim, and Roche-China; Vincent A. Miller is a consultant to Clovis, Astellis, Genentech, Arqule, and Boehringer-Ingelheim.