Eligibility

Patients were required to have histologically proven, stage IIIB (by pleural effusion) or IV BAC. Pathologic eligibility was based on institutional assessment, so that a histologic diagnosis consistent with BAC or its subsets by the participating institution was sufficient; specimens subsequently underwent central review through the SWOG Lung Correlative Science Subcommittee. Cytologic specimens were not accepted.

Patients with a SWOG performance status (PS) of 0, 1, or 2 were eligible. All patients had evidence of disease by computed tomography (CT) of the chest; patients were required to meet RECIST criteria for measurable or assessable disease.

No prior biologic therapy targeting the EGFR axis was allowed. Prior radiation was permitted if at least 4 weeks had elapsed from radiation and disease was present outside of the radiation port. Patients enrolled on the chemo-naïve stratum could not have received prior chemotherapy, while the pretreated stratum had no limit on number or duration of prior regimens. Chemotherapy recipients had to have completed therapy at least 4 weeks before enrollment on the present trial.

All patients were informed of the investigational nature of this study and signed a written informed consent in accordance with local institutional review board and federal guidelines.

Study Design

Treatment consisted of gefitinib administered at a daily dose of 500 mg orally, and cycles were defined as 28 days. Delays in treatment and dose modification were performed for grade 3/4 toxicity, for which gefitinib could be withheld for a maximum of 2 weeks. Treatment was restarted on improvement of toxicity to less than or equal to grade 1, or grade 2 with a dose reduction to 250 mg daily. A single dose reduction was permitted; patients who required additional dose reduction or were unable to be retreated after a 2-week delay were removed from the study.

Patients underwent repeat history and physical examination before each treatment period. Repeat CT imaging to assess response was performed following every two cycles. Baseline CT imaging and the first two follow-up scans were requested to be performed with the same CT scanner and technique, and electronic or film copies were to be submitted for central review using a computer-assisted image analysis system. Patients off treatment were observed at least every 6 months for 2 years and then annually until death or data cutoff. Patients were removed from protocol treatment for disease progression, unacceptable toxicity as assessed by the investigator, development of intercurrent, noncancer related illnesses that prevented continuation of treatment, prolonged treatment delay, or on patient request for any reason.

Study Evaluation and Statistical Methods

Toxicities were graded according to National Cancer Institute Common Toxicity Criteria Version 2.0. Response assessments (complete or partial response or stable disease) were made by RECIST criteria,¹² but response was not required for completion of the study protocol because the primary end point was OS.

The primary objective of S0126 was to assess the 1-year survival rate in both chemotherapy-naive and pretreated patients with advanced BAC receiving daily gefitinib. Based on the results from S9714,⁴ a prospective definition was established that this regimen would be of considerable interest if 1-year survival rates in the chemotherapy-naïve cohort exceeded 60%, but would be of no further interest if the true 1-year survival was 40% or less. A target population of 55 patients accrued at a rate of 24 patients per year would provide a power of 0.89 to detect an improvement in 1-year survival from 40% to 60%, with a one-sided α of .05.

While there were no established benchmark survival data for the previously treated cohort, a prospective definition was established that this regimen would be of considerable interest if 1-year survival in this cohort exceeded 50%, but would be of no further interest if the true 1-year survival was 25% or less. A target accrual of 35 patients in this patient population was established, with a power of 0.91 to detect an improvement in true 1-year survival rates from 25% to 50%, with a one-sided α of .05.

OS and PFS were determined based on the method of Kaplan and Meier.¹³ Confidence intervals for the median overall and PFS were calculated according to the method of Brookmeyer and Crowley.¹⁴ Hypothesis tests for differences in OS and PFS based on predefined covariates of patient sex, development of rash, smoking status, and patient performance status were performed using the Wald χ² Test in Cox regression.¹⁵

Patient Characteristics

S0126 was activated in December 2001; accrual to the previously untreated stratum was completed in February 2003, and for the previously treated stratum in July 2003. The accrual rate in the previously untreated arm considerably exceeded expectations; in light of the potential value of exploratory analyses for clinically and/or molecularly defined patient subsets, enrollment remained open beyond the initial target number.

Of 145 patients registered (104 chemotherapy-naïve, 41 pretreated), eight patients were ineligible, primarily due to an institutional pathology report not consistent with a diagnosis of BAC subsets. One patient is not assessable due to death before initiation of therapy. Characteristics of 136 assessable patients are presented divided by strata in Table 1. Reviewing the demographic data of the present study as an aggregate across strata, the median age was 68, with a range from 34 to 88 years of age. Sixty-seven patients (49%) were male and 69 (51%) were female. PS was 0 or 1 in 121 patients (89%) and 2 in 15 patients (11%). Stage IV disease was present in 127 patients (93%), while nine patients (7%) had stage IIIB NSCLC with a pleural effusion. Measurable disease was present and response determinable in 91 of 136 patients (67%).

Treatment Delivery

Of the 123 patients who have discontinued treatment at the time of this report, 20 (16%) were due to toxicity, 17 based on prospectively specified toxicity criteria (failure of toxicity to resolve to grade ≤ 1 after 2 weeks off of therapy or recurring transaminase elevations), and three due to patient preference. Based on calculation of medication delivered, 52 patients (38%) had a dose reduction to 250 mg daily.

Time to Progression and Survival Analysis

With a median follow-up of 22 months, the median PFS was 4 (95% CI, 3 to 6) and 3 months (95% CI, 2 to 5), for the previously untreated and treated groups, respectively (Fig 1). Median OS, illustrated by Kaplan-Meier plot in Figure 2, is 13 months for previously untreated (95% CI, 8 to 18) and 12 months for the previously treated subgroup (95% CI, 6 to 17). One-year survival was 51% for both strata. Two-year survival for previously untreated patients was 39%, and for the previously treated patients was 27%. Three-year OS was 23% and 22% for untreated (95% CI, 12% to 34%) and previously treated patients (95% CI, 7% to 37%), respectively. Among 87 patients with follow-up data submitted, 55 have received additional therapy, with no further details available.

Response

Best radiographic response among the 91 patients with measurable disease is shown in Table 2. Among 69 previously untreated patients, four (6%; 95% CI, 2% to 14%) achieved a complete response (CR) and eight (12%; 95% CI, 5% to 22%) achieved a partial response (PR). Among 22 patients who had received prior chemotherapy, two (9%; 95% CI, 1% to 29%) achieved a PR, and no CRs were observed. Stable disease (SD) was observed as best response among 22 (32%; 95% CI, 21% to 44%) previously untreated and eight (36%; 95% CI, 17% to 59%) previously treated patients, for a total disease control rate (CR + PR + SD) of 49% (95% CI, 37% to 62%) among chemotherapy-naïve patients and 45% (95% CI, 24% to 68%) among pretreated patients. Progressive disease (PD) was the best response in 22 previously untreated patients (33%; 95% CI, 22% to 46%) and eight previously treated patients (36%; 95% CI, 17% to 59%). Twelve patients (17%) and four patients (18%) in the previously untreated and treated groups, respectively, did not have reassessment adequate for determining response and are assumed to be nonresponders. Results of response by computer-assisted image analysis are beyond the scope of this article and will be reported elsewhere.

Survival and Response by Clinically Defined Subsets

Exploratory analyses were performed for survival and response among patient subsets defined herein. As shown in Figure 3, median OS among the 69 females was 19 months, significantly better than the 8-month median OS among the 67 males (P = .0252). The median OS of 16 months among the 112 patients who developed a rash was significantly superior to the median OS of 5 months among the 24 patients who did not develop a rash (P = .003; Fig 4A). There was no difference between patients who developed a grade 1 rash versus those who developed a more severe (grade 2 or 3) rash, with a median OS of 16 months for patients who developed a grade 2 or 3 rash versus 15 months for those who developed a grade 1 rash (Fig 4B). As illustrated in Figure 5, the 39 never-smokers (29%; fewer than 100 cigarettes) had a significantly superior median OS of 26 months (95% CI, 12 to 34), compared with a median OS of 10 months (95% CI, 7 to 16) for the 97 former or current smokers (71%; P = .049). The 121 patients (89%) with a PS of 0 or 1 had a median OS of 15 (95% CI, 10 to 18) months, significantly better than the 5 month (95% CI, 1 to 6) median OS seen among the 15 patients (11%) with a PS of 2 (P = .023; Fig 6). While each of these factors was associated with a significant hazard ratio for survival in univariate analysis, multivariate Cox regression analysis revealed that all of these factors except for smoking status remained statistically significant (Table 3).

As illustrated in Table 4, exploratory analysis of potential differences in response rate as a function of the clinical variables noted above demonstrated no significant differences except development of rash, in which a 19% response rate (95% CI, 11% to 30%) was seen among patients with rash, compared with 0% response rate (95% CI, 0% to 18%) among patients who did not develop a rash (P = .036, Fisher's exact test).

Toxicity

Maximum toxicities of gefitinib in 136 eligible patients did not differ between previously untreated and pretreated patients, with data across strata listed in Table 5. Rash was the most commonly reported adverse event (82% of patients) and was grade 3 or 4 in 12%. Diarrhea was seen in 69% of patients and was grade 3 of 4 in 21%. Fatigue (46%; grade 3 of 4 in 5%), nausea (28%; grade 3 or 4 in 4%), and elevated transaminases (31%; grade 3 or 4 in 6%) were also reported. Three patients experienced progressive lung infiltrates of unclear etiology. Two of these patients developed symptoms within the first month (at days 5 and 14 days after starting gefitinib); the third patient developed symptoms and radiographic changes after more than 2 months on therapy. All three of these patients died after treatment with supplemental oxygen, antibiotics, and steroid therapy. In each case, gefitinib-associated interstitial lung disease was considered the most likely etiology, although pneumonia and progressive BAC were also possible. Lung biopsy to provide a histologic diagnosis was not performed in these cases.