Clinical Course of Patients with Non–Small Cell Lung Cancer and Epidermal Growth Factor Receptor Exon 19 and Exon 21 Mutations Treated with Gefitinib or Erlotinib

Materials and Methods

Patients and clinical characteristics. We have included all patients with NSCLC on whom successful EGFR genotyping was done in the Laboratory of Diagnostic Molecular Pathology at Memorial Sloan-Kettering Cancer Center. The results of the EGFR mutational analysis, tumor stage, pathologic subtype, and survival were obtained by retrospective chart review. The reported age is the age at diagnosis with NSCLC. Sex and race/ethnicity were determined by patient self-report at the time of initial registration at this institution. Stage at presentation was determined using current American Joint Committee on Cancer staging criteria (18). Histology was obtained from the pathology report corresponding to the specimen sent for EGFR genotype analysis. Smoking history was determined by review of the medical records, including both physician description of smoking history and an institutional smoking history questionnaire completed by the patient. Patients described as receiving treatment with erlotinib or gefitinib were determined by review of the medical record. Those patients who had smoked <100 cigarettes in their lifetime were categorized as never smokers. Those patients who smoked cigarettes within 1 year of the diagnosis were categorized as current smokers. The review of records was done under a waiver of authorization approved by the MSKCC Privacy Board. Survival time and progression-free survival were calculated from the date of initiation of erlotinib or gefitinib. Patients who were not deceased were censored at the date of last contact with this institution. This date was verified by inpatient and outpatient medical records and/or confirmation with the patient's primary physician. Included within this analysis are 60 patients whose mutation status has been previously reported (19).

Mutational analysis. Tumor specimens were not routinely microdissected before mutational analysis. In some cases where only a small focus of tumor was present in the block, manual microdissection was done. EGFR exon 19 deletions were studied by length analysis of fluorescently labeled PCR products on a capillary electrophoresis device, using the following primers: EGFR-Ex19-FWD1, 5′-GCACCATCTCACAATTGCCAGTTA-3′ and EGFR-Ex19-REV1, 5′-Fam-AAAAGGTGGGCCTGAGGTTCA-3′. The cases were also screened for the exon 21 L858R mutation by a PCR RFLP assay, based on a new Sau96I restriction site created by the L858R mutation (2573T → G). The Sau96I-digested fluorescently labeled PCR products were analyzed by capillary electrophoresis, and the following primers were used: EGFR-Ex21-FWD1, 5′-CCTCACAGCAGGGTCTTCTCTGT-3′ and EGFR-Ex21-REV1, 5′-Fam-TCAGGAAAATGCTGGCTGACCTA-3′ (17). These assays reliably detect mutations in mixtures containing as little as 5% to 10% tumor DNA (17). Direct sequencing of EGFR was also done in some cases, as previously described (7). In such cases, all sequencing reactions were done in both forward and reverse directions, and all mutations were confirmed by PCR amplification of an independent DNA isolate.

Statistical Analysis. The variables measured in the study were investigated for association using the Fisher's exact test when two categories existed for a variable and Pearson's χ² test when three or more categories existed for a variable. Statistical significance was determined by a two-tailed P < 0.05. Median duration of overall survival and progression-free survival were calculated using the Kaplan-Meier method. Comparison of survival between groups was made using the log-rank test. Survival analysis was done using Stata version 8 for Windows.

Results

Relationship between presence of EGFR mutations and clinico-pathologic features. NSCLC tumors from 291 patients were examined for the presence of the two most common EGFR mutations associated with response to erlotinib or gefitinib. Seventy patients (24%) were found to have either EGFR exon 19 deletion or EGFR L858R. The clinical characteristics of patients tested for EGFR mutations are listed in Table 1. Univariate analyses showed an association between EGFR mutation and both never smoker status (P = 0.0001) and self-reported Asian race/ethnicity (P = 0.0023). There was no association with female gender (P = 0.31). No correlation of EGFR mutation with pathologic subtype was found, but the large majority of patients tested for EGFR mutation had adenocarcinoma or adenocarcinoma with bronchioloalveolar carcinoma features (20). One patient with squamous and another with adenosquamous tumor histology were found to have an EGFR exon 19 deletion. Of the 70 patients with mutations, 32 patients were former or current smokers. These smokers had a median smoking history of 9 pack-years (range, 0-75 pack-years). Three patients with ≥70 pack-year history of smoking were found to have EGFR mutations.

Correlation of EGFR genotype with clinicopathologic features. Of the 70 patients with an EGFR mutation, 43 (61%) had an EGFR exon 19 deletion and 27 (39%) had an EGFR L858R mutation. The clinical characteristics of this group are shown in Table 2. When comparing patients whose tumors had an EGFR exon 19 deletion to EGFR L858R, there was no significant difference in the proportion of women (26 of 43 versus 18 of 27; P = 0.62), Asians (4 of 43 versus 3 of 27; P = 1.0), African Americans (4 of 43 versus 2 of 27; P = 1.0), never smokers (24 of 43 versus 14 of 27; P = 0.8), or adenocarcinoma histology (37 of 43 versus 27 of 27; P = 0.07).

Correlation of EGFR genotype with clinical course during treatment with gefitinib or erlotinib. Thirty-four of the 70 patients with EGFR exon 19 deletion or EGFR L858R had been treated with either gefitinib or erlotinib. The remaining 36 patients either were not available for follow-up (n = 6) or did not receive treatment because all sites of disease were surgically resected (n = 30). The median time from initial lung cancer diagnosis to beginning either erlotinib (n = 12) or gefitinib (n = 22) was 17 months (95% confidence interval, 4-25 months). Patients who presented with early-stage disease and subsequently experienced disease recurrence (n = 9) had a median time to recurrence of 15 months. From the diagnosis of recurrent/metastatic disease, the median time to treatment was 9 months for erlotinib-treated patients and 10 months for those patients treated with gefitinib. Among these patients with EGFR mutations, the median progression-free survival after initiation of gefitinib or erlotinib was 12 months (95% confidence interval, 8 months to not reached; range, 3 to ≥29 months), and the median overall survival after initiation of gefitinib or erlotinib was 20 months (95% confidence interval, 16 months to not reached; range, 4 to ≥34 months). Comparison of patients treated with erlotinib to those treated with gefitinib revealed no difference in overall survival (log-rank P = 0.54).

To explore whether the two major EGFR mutation genotypes respond differently to treatment with gefitinib and erlotinib, we determined the progression-free survival and overall survival of patients with these genotypes from the time of initiation of treatment (Fig. 1). In this subset of patients, there was no significant difference in proportion of women, never smokers, adenocarcinoma histology, or bronchioloalveolar carcinoma features (Table 3). At the time of initiation of treatment, whereas patients with EGFR exon 19 deletion were more likely to have a site of at least one extrathoracic metastasis (14 of 23 versus 2 of 11; P = 0.03), there was no significant difference in the incidence of metastases to individual sites of bone, brain, or liver. There was no significant difference in the baseline Karnofsky performance status of these patients. Patients with EGFR exon 19 deletion had a median progression-free survival of 12 months compared with 5 months for patients with EGFR L858R (P = 0.01). Patients with EGFR exon 19 deletion had a median overall survival of 34 months compared with 8 months for patients with EGFR L858R (P = 0.01).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Overall survival by EGFR mutation.

Correlation of overall survival with smoking history in patients with EGFR mutations. Of the patients with EGFR mutations who had been treated with gefitinib or erlotinib and were former or current smokers, the median smoking history was 8 pack-years. For these patients, we compared the overall survival of patients who were never smokers (n = 16), of former smokers who had smoked less than the median smoking history of <8 pack-years (n = 8), and of those who had smoked ≥8 pack-years (n = 10; Fig. 2). The median overall survival for never smokers was 20 months. The median survival of former smokers who smoked <8 pack-years has not been reached, whereas the median survival of former smokers with ≥8 pack-year history of smoking was 8 months.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Overall survival by smoking history.

Discussion

Our retrospective review of the results of EGFR mutation analysis done as part of clinical care of patients with NSCLC represents the largest group of patients from North America for which tumor EGFR genotyping has been reported. In addition to correlating EGFR mutations with clinical factors predictive of response to erlotinib or gefitinib, we found that patients with EGFR exon 19 deletions had longer overall survival when compared with patients with EGFR L858R after treatment with erlotinib or gefitinib (P = 0.01).

We report a relatively high proportion (24%) of patients with one of the two most common EGFR mutation genotypes. As clinical factors predictive of sensitivity to EGFR TKIs antedated identification of EGFR mutations, clinicians tended to request mutational studies of patients with such factors. The group of patients studied is enriched for never smokers, women, patients with adenocarcinoma, and patients of Asian ethnicity. Only 8% (24 of 291) patients tested had nonadenocarcinoma histology. Additionally, this series has a relatively large proportion of patients who presented with metastatic disease. Many of the previous series were made up almost solely of patients who had resections for early-stage NSCLC.

Multiple prior international reports have confirmed the association of mutations with those clinical characteristics known to predict response to gefitinib or erlotinib (i.e., Asian ethnicity, history of never smoking, and adenocarcinoma histology). In this large group of patients from North America, we confirmed the association of EGFR mutation with some of the clinical predictors of response, including history as a never smoker and Asian ethnicity. Women were not more likely to have tumors with EGFR mutations in this series that included a disproportionate number of women (51% of patients with NSCLC referred to our medical oncology service are women). This observation has been made by other groups as well. However, some have found gender to be predictive in multivariate analyses.

Since the initial reports of partial responses with gefitinib, researchers have noted a significant geographic and ethnic variability in the sensitivity of patients to this class of agents. After the identification of somatic mutations in EGFR and their association with response, it became clear that these two factors were linked. A significantly higher proportion of patients from East Asia have mutations in EGFR based on multiple reports from Japan, China, Taiwan, and Europe, as well as smaller studies from North America. Accordingly, obtaining accurate information about each patient's country of origin, race, and ethnicity is vital. In contrast to a previous report that noted a relatively low incidence of EGFR mutations in patients of African-American ethnicity (2%, 1 of 41), 43% (6 of 14) of African-American patients in our series had EGFR mutations. Clearly, a variety of factors could modify this incidence with the most obvious being smoking history. Fifty-seven percent (8 of 14) of African-American patients in our series were never smokers.

These data extend the observation that a heavy smoking history in patients with NSCLC is associated with a poorer survival to this subpopulation of NSCLC patients with EGFR mutations (21, 22). The cause of this difference is not fully understood, although repeated exposure to tobacco-related carcinogens is thought to lead to a more genetically complex tumor in patients with heavy smoking history. Although EGFR mutations and KRAS mutations, the most well-studied genetic lesion in smokers, are thought to be mutually exclusive, it remains possible that other mutations caused by tobacco-related carcinogens may be present in heavy smokers and lead to decreased overall survival (23). More comprehensive molecular analysis of tumors from patients with NSCLC and EGFR mutations will be helpful to clarify this issue.

Many recent series have compared NSCLC patients with EGFR mutations to those patients with no detectable EGFR mutations to explore the clinical relevance of EGFR mutations for sensitivity to gefitinib or erlotinib. Some series show an improvement in overall survival for patients with EGFR mutations who have been treated with gefitinib or erlotinib (10, 24). Mitsudomi et al. studied 59 patients with disease recurrence after surgery who had been treated with gefitinib (25) and found that patients with EGFR mutations had a prolonged survival (P = 0.005). Similarly, Chou et al. examined 54 patients, 61% of whom had mutations in EGFR and showed longer progression-free survival (P = 0.011) and overall survival (P = 0.046) after treatment with gefitinib when compared with patients with wild-type EGFR (26). Cappuzzo et al., in examining a group of 89 patients who had been treated with gefitinib, found 15 patients with mutations in EGFR with a median survival of 21 months as opposed to a median survival of 8 months for the 74 patients without mutations (P = 0.09; ref. 27). Similarly, Tokumo et al. looked at 21 gefitinib treated patients, 9 with mutations in EGFR, and noted a 25-month survival for patients with mutations and 14 months for those without mutations (P = 0.15; ref. 15). A variety of pitfalls are associated with comparing patients with EGFR mutations to those without detectable mutations, including potential sensitivity issues in the use of direct sequencing to detect EGFR mutations (whereby patients with EGFR mutations are included in the wild-type group) and the uncertainty of association of EGFR mutations other than exon 19 deletions and L858R with response to erlotinib or gefitinib. To avoid these problems, our survival analysis included only patients with detectable mutations and compared the two most common subtypes of mutations, EGFR L858R and EGFR exon 19 deletion.

Our results suggest that specific EGFR mutation genotypes may be predictive of survival after treatment with EGFR TKI in patients with NSCLC, showing that EGFR exon 19 deletion patients treated with erlotinib or gefitinib lived longer than patients with EGFR L858R (P = 0.01). This survival benefit was observed in spite of the higher frequency of extrathoracic metastases for patients with EGFR exon 19 deletions. A previous report of patients treated with gefitinib or erlotinib noted a difference in response rate, which favored patients with EGFR exon 19 deletion, but differences in survival were not detected (25). Mitsudomi et al. noted a 62% (8 of 13) response rate in patients with EGFR point mutations compared with 100% (16 of 16) response rate in patient with EGFR exon 19 deletion (P = 0.0108). In contrast, Shigematsu et al. reported data from patients, never treated with EGFR TKI, who had surgically resected NSCLC (14). They compared 62 patients with EGFR mutations (31 with EGFR L858R and 31 with EGFR exon 19 deletion) and noted a relatively prolonged survival for those patients with L858R (P = 0.05). If prospectively confirmed, our data suggest that treatment with gefitinib or erlotinib alters the natural history of patients with EGFR-mutated NSCLC, converting the EGFR exon 19 deletion subtype of NSCLC from a disease with a worse prognosis to one with a more favorable prognosis.

Differences in response rate determined by different mutations and prolonged survival for TKI therapy has also been observed with imatinib in the treatment of patients with gastrointestinal stromal tumors. After the initial identification of KIT mutations and their association with response to imatinib, it was noted that particular mutations conferred greater sensitivity to imatinib (28–30). Recently, an analysis of tumor specimens from an intergroup trial of imatinib showed a longer time to treatment failure and overall survival in patients with KIT exon 9 mutations compared with those patients with KIT exon 11 mutations (31).

Our data highlight the need for a prospective analysis of EGFR genotype in a larger study of patients treated with erlotinib or gefitinib. Unfortunately, recent results of EGFR genotyping from the largest trial of erlotinib to date suffered from low levels of tissue acquisition, obtaining usable tissue for EGFR sequence analysis on only 177 patients of 731 patients randomized, and a relatively low total number of mutations identified (n = 40, of which only 21 were L858R mutations or exon 19 deletions, the mutations previously associated with sensitivity to erlotinib), making analysis of these data problematic (32). Future analyses should incorporate the prospective collection of tumor blocks and an attempt to correlate other factors, which may be predictive of response to gefitinib or erlotinib, including EGFR amplification, EGFR expression, KRAS mutations, and phospho-AKT expression (27).