SUPPLEMENTAL DIGITAL CONTENT 2

Results

Efficacy and quality of life in the overall Japanese study population

Pre-planned analyses of the subset of 233 IPASS patients randomized in Japan (gefitinib [n=114], carboplatin/paclitaxel [n=119]; 19.1% of the overall study population) showed that efficacy and safety datafor patients recruited in Japan,as presented previously at ASCO and WCLC in 2009,16,17were generally consistent with the overall study population.6 PFS was significantly longerwith gefitinib than carboplatin/paclitaxel (HR, 0.69; 95% CI, 0.51–0.94;p=0.019; 176/233 progressed, 76% maturity)(Supplemental Digital Content 3, Figure). The effect was not constant over time, with the probability of being progression-free initially favoring carboplatin/paclitaxel, and later being in favor of gefitinib, likely driven by the differing effects in EGFR mutation subgroups. The PFS treatment by country interaction test (Japan vs. other countries) was not significant (p=0.47), showing that statistically, the treatment effect for PFS did not differ between patients recruited in Japan and patients recruited in other countries. The ORR was numerically higher with gefitinib than carboplatin/paclitaxel (41.2% vs. 34.5%, respectively; OR, 1.34; 95% CI, 0.78–2.30; p=0.30). Quality-of-life (QoL) results were also similar to that of the overall study population, with significantly more patients in the gefitinib group than in the carboplatin/paclitaxel group having a clinicallyrelevant improvement in QoL, as assessed by scores on the Trial Outcome Index (TOI) (OR, 1.92; 95% CI, 1.11–3.34; p = 0.02), and similar rates of reduction in symptoms, as assessed on the basis of Lung Cancer Subscale (LCS) scores (OR, 0.85; 95% CI, 0.50–1.43; p = 0.53); however, unlike in the overall study population, gefitinib did not show an advantage over carboplatin/paclitaxel in QoL as assessed by the Functional Assessment of Cancer Therapy – Lung (FACT-L) (OR, 0.94; 95% CI, 0.56–1.60; p=0.83).

Tolerability in the overall Japanese study population

The median duration of treatment was 5.6 months for gefitinib and 3.3 months for carboplatin/paclitaxel (compared with 5.6 months and 4.1 months, respectively, in the overall study population). The median number of carboplatin/paclitaxel cycles was 4 (compared with 6 in the overall study population). This is consistent with medical practice and previous reports in this clinical setting in Japan. The pattern of adverse event (AE) reporting was similar to that for the overall study population, with fewer CTC grade 3/4 AEs in the gefitinib group compared with the carboplatin/paclitaxel group (33.3% vs. 62.7%). However, among patients in Japan there was a higher incidence of dose modifications due to toxicity (33.3% vs. 16.1% with gefitinib and 56.4%/57.6% vs. 35.2%/37.5% with carboplatin/paclitaxel) and of AEs leading to discontinuation of treatment (13.2% vs. 6.9% with gefitinib and 29.7% vs. 13.6% with carboplatin/paclitaxel) compared with the overall study population for both treatments. No patients in Japan had an AE leading to death on either treatment. The most common AEs were consistent with previous clinical experiences with gefitinib and carboplatin/paclitaxel;20,21 although some were reported more frequently among patients in Japan compared with the overall study population.16,17 The safety profile of gefitinib in patients recruited in Japan was consistent with the prescribing information for gefitinib in Japan. As for the overall study population, gefitinib had a more favorable safety profile than carboplatin/paclitaxel.

Tumor EGFR mutation status and clinical outcomes

In contrast with the cfDNA EGFR mutation results, the subset of patients with known tumor EGFR mutation statuswas not representative of the overall Japanese study population and the overall study population as shown by inconsistent PFS and ORR results (Supplemental Digital Content 4, Table).6 This was thought to be due to the imbalance between the treatment groups, with a lower proportion of EGFR M+ patients in the gefitinib group compared with the carboplatin/paclitaxel group, and was a result of the small number of evaluable samples (n = 91).

For this reason the analyses by EGFR mutation status were difficult to interpret. Nevertheless, as for the overall study population, the PFS hazard ratio was more favorable for gefitinib for the EGFR M+ subgroup (HR, 0.70 [95% CI, 0.36–1.35; p= 0.28]) and more favorable for carboplatin/paclitaxel for the EGFR M- subgroup (HR, 1.45 [95% CI, 0.67–3.15; p = 0.35]). ORR was higher for gefitinib than carboplatin/paclitaxel in the tumor M+ subgroup (69.6% [16/23] and 48.5% [16/33], respectively; OR, 3.25; 95% CI, 0.95–11.06; p = 0.06) and slightly higher for carboplatin/paclitaxel than gefitinib in the tumor M- subgroup (7.1% [1/14] and 0% [0/21], respectively; OR not calculated).

Comparison of EGFR mutation status in pre-treatment pleural effusion, cfDNA, and tumor tissue

All nine patients with a pre-treatment pleural effusion EGFR mutation status also had a known pre-treatment cfDNA EGFR mutation status and the results matched in eightcases (three positive, five negative); for one patient, the cfDNA status was negative and the pleural effusion status was positive. All four patients with an EGFRM+ pleural effusion sample hadonly exon 19 deletions by pleural effusion. Three of these patients wereEGFR M+ by both pleural effusion and cfDNA; two with the same mutation type and one patient with exon 20 T790M only by cfDNA (all three were unknown by tumor). The fourth patient was found to have an exon 19 deletionby pleural effusion and tumor but wasEGFR M- by cfDNA. The EGFR mutation status matched for all three patients with known status in tumor and pleural effusion samples (one positive and two negative). The analysis results of cfDNA from the supernatant and precipitant of pleural effusion samples were also concordant. Interpretation of these results is, however, limited by the low number of samples.

Comparison of EGFR mutation status in pre-treatment and post-progression cfDNA

A total of 144 patients had samples evaluable for post-progression cfDNA EGFR mutation analysis: 62 in the gefitinib group and 82 in the carboplatin/paclitaxel group. Of the 144 patients who had a known mutation result by cfDNA both in a pretreatment and post-progression serum sample, 18.8% were cfDNA EGFR M+ at pre-treatment: 21 were cfDNA EGFR M+ at pre-treatment and post-progression, 6were EGFR M+ at pre-treatment but not at post-progression, and 13 were EGFR M- at pre-treatment but EGFR M+ at post-progression. The proportion of concordance (the number of times that the pre-treatment and post-progression results agree) was 125/144 (86.8%; kappa coefficient [95% CI]: 0.61 [0.45–0.77], corresponding to moderate-to-substantial agreement).22 The mutations in pre-dose vs. post-progression samples are shown in the Supplemental Digital Content 4, Table; no patients gained T790M mutations. It is not known whether the changes in EGFR mutation status are true changes or are related to methodology issues, e.g., low levels of cfDNA and sampling issues.

1