Erlotinib for Frontline Treatment of Advanced Non–Small Cell Lung Cancer: a Phase II Study

There are ∼1.2 million new cases of lung cancer diagnosed every year worldwide, and 1.1 million patients die of the disease (1). Non–small cell lung cancer (NSCLC) represents 80% of lung cancers. Surgery is the most important curative modality in the treatment of early-stage NSCLC; however, only ∼20% to 30% of patients are diagnosed at an operable stage.

Platinum-based doublet chemotherapy, the standard treatment for advanced NSCLC, achieves mostly partial responses in only 30% to 40% of patients and a modest survival increase (2, 3); in general, single agents are inferior to combinations. Median survival and 1-year survival after first-line chemotherapy are 8 to 10 months and 30% to 40%, respectively (4). These regimens are associated with considerable toxicity, particularly myelosuppression. More effective and less toxic treatments are urgently needed for advanced NSCLC.

Among the new drugs that have been introduced in the treatment of advanced NSCLC after failure of first-line chemotherapy are docetaxel, pemetrexed, and, recently, epidermal growth factor receptor (EGFR) tyrosine kinase (TK) inhibitors. All of these agents produce a low response rate (below 10%) and have different toxicity profiles. In a large randomized study, erlotinib improved survival versus best supportive care in second-line and third-line treatment (5). Erlotinib is a targeted agent that inhibits EGFR by competing with ATP at the intracellular TK domain of the receptor. Erlotinib is given orally continuously and is devoid of bone marrow and major organ toxicities (6). Common side effects are diarrhea and skin toxicity.

Despite promising preclinical data, the combination of small-molecule inhibitors of EGFR with first-line chemotherapy failed to improve survival, compared with chemotherapy alone in four large randomized studies that investigated erlotinib or gefitinib in combination with either carboplatin-paclitaxel or cisplatin-gemcitabine (6).

We did a phase II study of single-agent erlotinib as first-line therapy in patients with advanced NSCLC. We also did biological studies on tumor samples of patients entered in the study to identify patients who benefited most from treatment.

Patients with a histologic or cytologic diagnosis of advanced NSCLC not amenable to radical surgery or radiotherapy were eligible for the study. No prior chemotherapy or other systemic treatment was allowed. Other eligibility criteria were measurable disease according to the Response Evaluation Criteria in Solid Tumors, ages ≥ 18 years, performance status of 0 to 2, life expectancy ≥ 12 weeks, at least 4 weeks since any prior surgery or radiotherapy, granulocyte count ≥1.5 × 10⁹/L, platelet count ≥100 × 10⁹/L, bilirubin and transaminases ≤1.5 × upper limit of normal, and serum creatinine ≤1.5 × upper limit of normal, or creatinine clearance ≥60 mL/min. Females of childbearing potential had to have a negative pregnancy test. Patients were excluded if they had unstable systemic disease (active infection, uncontrolled hypertension, unstable angina, congestive heart failure, myocardial infarction within the previous year, and serious cardiac arrhythmia requiring medication), any other malignancies within 5 years except for adequately treated carcinoma in situ of the cervix, or basal or squamous cell skin cancer. Patients with brain metastases were only allowed in the study if there was no evidence of progression in the brain and in the absence of corticosteroid treatment. Patients with significant eye disorders were excluded (severe dry eye syndrome, Sjogren syndrome, severe exposure keratinitis, and any other eye disorder likely to increase the risk of corneal epithelial lesions). Signed informed consent was required for all patients. The study was conducted according to the latest version of the Declaration of Helsinki, and the protocol was approved by the independent Ethics Committee and the Review Board from each participating institution.

Erlotinib 150-mg tablets were given orally daily, in the morning with 200 mL of water, and at least 1 hour before or 2 hours after ingestion of food or medication. Caution was taken to avoid medications that interact with CYP 450-3A4 enzyme, and concomitant warfarin was not permitted. Assessment of toxicity was according to National Cancer Institute Common Toxicity Criteria version 2.0. Dose adjustments, in 50-mg decrements, were done if grade 3 or 4 toxicity was encountered.

Before starting treatment, patients were assessed by physical examination, electrocardiogram, disease measurement by appropriate radiological techniques, complete blood cell counts, serum chemistries, and urine pregnancy test in females of childbearing potential. Patients were evaluated every 3 weeks, and hematology and blood chemistry were done; tumor measurements were assessed every 6 weeks. An electrocardiogram was repeated after 18 weeks of treatment and as clinically indicated afterwards. Response assessment was according to the Response Evaluation Criteria in Solid Tumors (7). All patients received the Lung Cancer Symptoms Score patient scale questionnaire at baseline and again at 3 and 6 weeks and every 6 weeks thereafter.

Mutation analysis of EGFR, PIK3CA, and K-ras. Paraffin-embedded tumor material was cut into 4-μm-thick sections and placed onto glass slides, stained with H&E, and the presence of tumor cells was verified by a pathologist. Tumor cells were microdissected, and genomic DNA was isolated using the QIAamp DNA Micro kit (Qiagen, Venlo, The Netherlands). Nested PCRs were carried out using primers (Table 1) to amplify exons 18 to 21 of EGFR, exons 9 and 20 of PIK3CA, and exons 1 to 2 of K-ras. To facilitate sequencing, internal primers incorporated an M13Tag. Sequencing of PCR products was done with the ABI PRISM 310 Genetic analyzer (Applied Biosystems, Foster City, CA). Mutations were confirmed by sequencing independent PCR products.

Because of concerns about the sensitivity of direct sequencing, DNA from nine independent samples from Dana-Farber Cancer Institute and 13 from the Vrije Universiteit Medical Center were analyzed by the other institution, in a blinded fashion, for EGFR mutations in exons 18 to 21. At the Dana-Farber Cancer Institute, specimens were analyzed using Surveyor DNA endonuclease, combined with the WAVE HS DHPLC system (Transgenomic, Inc., Omaha, NE) as described (8). DNA variants detected using Surveyor were subjected to fractionation and sequencing.

EGFR gene copy number. Chromogenic in situ hybridization (CISH) was used to study EGFR gene copy number. Tissue sections were incubated at 64°C for 2 to 4 hours, deparaffinized, and rehydrated. Samples were boiled in 1 mmol/L EDTA/Tris (pH 9) and digested with 0.01% pepsin/0.2 N HCl at 37°C. After dehydration in alcohol, the EGFR probe was added (Zymed Spotlight EGFR amplification probe 84-1300; Zymed, San Francisco, CA). Samples were denatured at 80°C for 10 minutes and incubated overnight in a humidifying chamber at 37°C. The next day, samples were washed for 5 minutes in 0.5 × SSC at 75°C to 80°C and incubated in 3% H₂O₂/PBS for 10 minutes to block endogenous peroxidase activity. After blocking the slides in pre-immune normal goat serum for 10 minutes, slides were incubated with mouse-α-digoxigenin antibody (1:100 dilution) for 60 minutes. CISH signals were visualized using the DAKO Envision system (DakoCytomation, Heverlee, Belgium). CISH signals were counted in 200 cells by two observers who were blinded to the patient's clinical characteristics. The scoring system was as described by Cappuzzo et al. for fluorescence in situ hybridization (9).

The primary end point of the first part of the study was the rate of nonprogression at 6 weeks, defined as the number of patients whose tumor did not progress (<19% tumor increase to total disappearance of tumor, according to the Response Evaluation Criteria in Solid Tumors). Deterioration of performance status was considered progression. This end point was considered an early conservative estimate of biological activity of the compound and was based on results obtained with single-agent first-line chemotherapy (10, 11).

Sixteen patients were to be entered, and of these, at least eight were to be nonprogressive in order for the study to continue accrual to the second part. If nine or more patients progressed during the first 6 weeks, the study was to be stopped (Simon's optimal design: p₀ = 40%, p₁ = 60%, α = 0.05, β = 0.20). The second part of the study had a targeted total accrual of an additional 30 patients, with a total accrual to the study of 46 patients. If at least 24 patients did not progress, the study would be considered of interest and promising.

Secondary objectives of the study were objective response rate, disease control rate, duration of response, time to disease progression or death, survival, and safety. Patients who did not receive at least 6 weeks of treatment were considered as progressors. Exploratory studies included quality-of-life assessment and the correlation of biological markers with treatment outcome. Survival, time to progression, and duration of response were calculated from day 1 of treatment. Survival curves were constructed using the Kaplan-Meier method, and survival was compared using the log-rank test. The SPSS statistical package for Windows version 9.01 was used.

From January to July 2004, 54 patients were enrolled in the study from three institutions in Europe. One patient was not eligible because of unstable brain metastases and did not start treatment. Table 2 summarizes the main characteristics of the eligible patients.

Response. Of 16 patients entered into the first part of the study, 10 had not progressed after 6 weeks of treatment; therefore, the study was opened to further accrual. Thirteen patients had <6 weeks of treatment; 12 were classified as early progressions, and one refused further treatment. The response rate based on intent to treat was 22.7% (95% confidence interval, 12.3-36.2%; Table 3). The percentage of nonprogressors at 6 weeks was 52.8% (28 of 53; 95% confidence interval, 38.6-66.7%). The median duration of response was 333 days.

Responses were significantly correlated with histology (six responses in adenocarcinomas, four in bronchioloalveolar carcinomas, two in squamous carcinomas, and none in large cell and other histologies; Pearson χ², P = 0.025), smoking history (seven responses in never smokers and five in former/current smokers; Pearson χ², P = 0.016), but not gender, performance status, stage of disease, or age.

When patients progressed, they were offered chemotherapy. Chemotherapy was given to 31 patients, 28 of whom received carboplatin or cisplatin in combination with gemcitabine; the response rate was 28.6% in these patients. Three patients received single-agent chemotherapy, and one patient was rechallenged with erlotinib after a partial response and interruption of treatment for 8 months at the patient's request; this patient again experienced a partial response. No further therapy was given to 13 patients who progressed on first-line erlotinib (24.5%), and four had only palliative radiotherapy. The median survival of progressive disease patients was 234 days versus 592 days for those with disease control (P = 0.004). The 13 patients who did not receive any form of treatment at progression had a significantly worse performance status at start (likelihood ratio χ², P = 0.020). The median survival of the 28 patients who received platinum-based chemotherapy was 350 days, calculated from the date of progression on erlotinib.

Toxicity. The most frequent drug-related side effects were mild to moderate skin toxicity and diarrhea (Table 4). All responders had some degree of rash, whereas only seven nonresponders developed rash (likelihood ratio χ², P = 0.048). Emesis was usually mild. Liver toxicity (transaminase and bilirubin increase) was observed in just >10% of patients and was of mild or moderate severity and reversible. Dose reduction to 100 mg/d was necessary in five cases due to increased bilirubin and in one due to skin toxicity. Dose interruptions for >1 day were necessary in 13 patients; in five patients, this was due to hyperbilirubinemia.

Survival. Median follow-up was 518 days. At the time of this analysis, 28 patients were dead; two were lost to follow-up; and 23 were alive. Median survival time for the 53 eligible patients was 391 days. The 1-year survival rate was 54%.

By univariate analysis, the following variables were significantly correlated with survival: smoking history, age, histology, and response to erlotinib (Table 5). The severity of skin toxicity was also significantly correlated with survival.

Median time to progression was 84 days, with three patients who had not yet progressed. Two patients remain on treatment (after 22 and 20 months, respectively), and one patient is being rechallenged with erlotinib. One patient underwent an R0 resection after a response on erlotinib. The percentage of patients not progressing at 1 year was 16%.

Quality of life. The Lung Cancer Symptoms Score was analyzed by assessing the visual analogue averages to the first six questions, concerning disease-related symptoms (appetite, fatigue, coughing, shortness of breath, hemoptysis, and pain). A decrease of 25% of the score compared with basal level was considered symptomatic progression. Although there was a difference in time to symptomatic progression between responding patients and stable disease and progressing patients, this was not significant (not shown).

EGFR, K-ras, and PIK3CA mutations and EGFR copy number. Histologic material from 29 patients was available to perform mutation analyses. EGFR mutations were detected in seven cases, five of which were deletions in exon 19; four of these cases responded to therapy, and one had long-lasting stable disease. One major response was detected in a patient with wild-type EGFR. Furthermore, two point mutations were detected in exon 20: one not previously described (V802I) and one that has been described in resistant disease (T790M; refs. 12, 13). In the latter sample, another point mutation in exon 21 was also recognized (R832H). These point mutations, which were only detectable using the WAVE system, but not direct sequencing, were observed in progressing patients and in tumors also bearing a K-ras mutation. The exon 19 deletions were only observed in females (Fisher's exact test, P = 0.047), mainly in never smokers (Fisher's exact test, P = 0.030). No patients with K-ras mutations had an exon 19 deletion (likelihood ratio, P = 0.026). There was no association between classic EGFR mutations and histology, performance status, stage, age, or appearance of rash upon treatment.

K-ras mutations were detected in 10 patients, all smokers (Fisher's exact test, P = 0.061), and nonresponders. There was a trend for a correlation between K-ras mutations and lack of development of rash (Fisher's exact test, P = 0.052) and no association with gender, histology (although more frequent in adenocarcinomas), performance status, stage, or age.

The presence of classic EGFR mutations was significantly correlated with response to treatment (Fisher's exact test, P = 0.001), and K-ras mutations were more commonly associated with lack of response (Fisher's exact test, P = 0.125). One PIK3CA mutation was detected in a patient who had stable disease on treatment. A summary of the mutation analyses is reported in Table 6. Survival was longer in patients with classic EGFR mutations than in the other patients, although the difference did not reach statistical significance (Table 5; Fig. 1A). Conversely, survival of patients with K-ras mutations was significantly shorter than that of patients with wild-type K-ras (Table 5; Fig. 1B). Of the two patients still on treatment, one did not have tumor available for mutation analysis; the other patient's tumor contained an exon 19 deletion of EGFR.

EGFR CISH analysis could only be done in 13 tumors due to insufficient tumor tissue remaining after mutation analyses. Amplification was detected in two cases and high polysomy in four cases. One of the two samples with amplified EGFR also contained an exon 19 deletion, whereas the other sample contained a K-ras mutation. No clear correlation was observed between EGFR copy number and response to treatment (Fisher's exact test, P = 0.192) or survival.

EGFR TK inhibitors are active drugs in advanced NSCLC, but in unselected patients, the response rate in second-line and third-line for both erlotinib and gefitinib is in the range of ≤10% (6). However, response to EGFR TK inhibitors is higher in women, never smokers, East Asians, adenocarcinoma and bronchioloalveolar carcinomas, and those who develop rash (14). Furthermore, activating mutations in the EGFR TK domain and EGFR amplification have been shown to be predictors of response and survival in several reports (9, 15–17).

In the present study, the primary end point was met, with >50% of patients showing no progression after 6 weeks of treatment. The response rate of 22.7% is similar to that obtained with most single chemotherapy agents. Although the protocol did not prescribe any type of patient selection, a higher percentage of women, never smokers, and patients with bronchioloalveolar carcinoma histology were included in the study than one would expect in a typical sample of patients with advanced NSCLC. This was a result of the growing evidence, accumulating during the conduct of the study, that these characteristics are more often associated with benefit from EGFR TK inhibitors.

Although response to erlotinib was more common in patients with adenocarcinoma (including bronchioloalveolar carcinoma histology) and in never smokers, like shown in other studies (5, 18), responses were also observed in patients with tumors of other histology and in former/current smokers. For example, a complete response was reported in a male patient who was a smoker. These results are consistent with those observed in the BR.21 study, in which erlotinib produced a significant survival benefit compared with best supportive care in most of the patient subgroups analyzed (5).

In the present study, the median overall survival (∼13 months) and the 1-year survival rate of 54% were better than those reported for several platinum-based doublet chemotherapy regimens. The use of erlotinib in first-line did not seem to have a detrimental effect on the response to subsequent chemotherapy. Our results are similar to those obtained with first-line gefitinib in Japanese patients (19). Promising results with erlotinib 150 mg/d monotherapy have also been achieved in an ongoing phase II study of patients ages >70 years with previously untreated advanced NSCLC (20).

Erlotinib was well tolerated in this study with the main toxicities being rash and diarrhea. Both response and survival were correlated with the grade of rash, in agreement with several other trials (18).

Several studies have assessed integration of EGFR inhibitors into first-line treatment of advanced NSCLC. Although the concomitant administration of erlotinib (21, 22) and gefitinib (23, 24) with chemotherapy failed to improve survival in the overall population, a survival benefit was observed in never smokers who received erlotinib and chemotherapy (25). This suggests that patient selection may identify groups of patients who may derive a large benefit from erlotinib treatment in combination with chemotherapy.

Certain mutations in the TK domain of EGFR are associated with increased sensitivity to EGFR TK inhibitors (16, 17, 26). In our study, four of five responding patients had EGFR mutations represented by well-documented deletions in exon 19. However, the tumor of one responding patient did not possess an EGFR mutation, and a mutation was found in a patient with long-lasting stable disease. In addition to exon 19 deletions, two point mutations were identified in patients who progressed, including the known resistant mutant T790M and a novel point mutation. Interestingly, in these samples, K-ras mutations were also found, suggesting that some EGFR point mutations can be associated with K-ras mutations and are related to resistance. The T790M EGFR mutation has been described in patients who were initially sensitive to EGFR TK inhibitors as a result of an exon 19 deletion and who later became resistant (12, 13). However, the T790M mutation has also been described in a patient with primary resistance (27), and germ line inheritance of this variant has been proposed in one family (28). The use of irreversible EGFR TK inhibitors or monoclonal antibodies directed to the extracellular domain of EGFR might circumvent resistance in these cases (12, 29–32). In our study, the presence of classic EGFR mutations was correlated with longer survival, although, possibly due to the small number of samples analyzed, this did not reach statistical significance. We also analyzed the EGFR copy number using CISH but were unable to confirm the predictive value of this variable in our series, although the number of samples analyzed was too small to draw any firm conclusion.

In the BR.21 study, several techniques were used to assess EGFR status: mutation analysis, fluorescence in situ hybridization, and immunohistochemistry (33). In addition, fluorescence in situ hybridization seemed to be the strongest predictor of response and survival, although by multivariate analysis, none of the techniques were able to significantly predict outcome. In a retrospective analysis of patients treated with gefitinib, alone or in combination with chemotherapy, EGFR amplification (detected by a PCR method) and EGFR mutations identified two distinct groups of patients, and mutations were better predictors of outcome than amplification (34).

Technological issues are still very important for both the detection of mutations and the assessment of EGFR copy number (14). In a cross-validation analysis (data not shown), it was apparent that simple direct sequencing does not provide sufficient sensitivity for detection of all mutations; however, the use of more sensitive methods may identify mutations of uncertain significance (8, 14). Tumor heterogeneity and possibly the presence of mutations in a proportion of tumor cells may in fact potentially play a role in the level of detectability and biological relevance.

K-ras mutations seem to be associated with primary resistance to EGFR TK inhibitors (35). In our study, K-ras mutations were noted only in patients who did not respond to erlotinib, and survival was also significantly shorter. Poorer clinical outcomes were shown also for K-ras mutant patients treated with erlotinib plus chemotherapy (36), suggesting that K-ras mutations not only confer a poor prognosis but may also constitute a negative factor in patients treated with EGFR TK inhibitors. Interestingly, the addition of erlotinib to gemcitabine chemotherapy has been shown to significantly increase survival in patients with advanced pancreatic cancer (37), a tumor type in which K-ras mutations are very common, raising the question of whether the negative findings in NSCLC are limited to this tumor type. Potentially, the simultaneous inhibition of EGFR and ras signaling may be of interest in NSCLC (38).

The results of several studies show that EGFR mutations and K-ras mutations are mutually exclusive (39, 40). Interestingly, in our study, classic EGFR mutations and K-ras mutations were mutually exclusive; however, tumors from two refractory patients possessed both EGFR point mutations and K-ras mutations, indicating that the mutual exclusivity may only apply to activating mutations of EGFR. In the attempt to select patients for EGFR TK inhibitor, the exclusion of patients with K-ras mutations should be considered.

Given the difficulty of combining erlotinib and gefitinib with concomitant chemotherapy in first-line, other strategies need to be developed. The promising efficacy of erlotinib monotherapy in this study suggests that such therapy is a potential option for the first-line treatment of advanced NSCLC. Studies investigating the optimal sequence of erlotinib and chemotherapy are warranted. Selection of patients based on clinical and biological characteristics will allow a significant improvement of overall results. Combinations of targeted therapies also warrant investigation in this setting: erlotinib plus bevacizumab produced promising antitumor activity in a phase I/II trial involving patients with previously treated advanced NSCLC (41), and this combination is now also being investigated in first line.

In summary, the results of this study show that erlotinib has significant antitumor activity in the first-line treatment of advanced NSCLC, which may be comparable with single-agent chemotherapy, and is devoid of significant side effects. However, these results have to be seen in the context of the patient population entered, which is weighed toward individuals with higher chances of having EGFR mutations. Although responses to erlotinib were more common in some subgroups of patients, clinical benefit was not restricted to these subgroups. Further investigation of erlotinib monotherapy in this setting is warranted and the identification of patients who most benefit from the treatment is of paramount importance.

We thank Dr. Gerrit Meijer and Karijn Floor for the CISH analysis, Alison J. Holmes for the cross-validation of EGFR mutations, and Dr. M. Smith (Thomson Gardiner-Caldwell Communications) for his assistance in drafting the article.