To the Editor:

Recently, two well-designed trials investigated whether adding cetuximab to platinum-based doublets could improve survival in advanced metastatic non–small-cell lung cancer (NSCLC). Essentially the two trials found similar results, although only one was statistically significant.

The FLEX (First-Line Erbitux in Lung Cancer) study¹ added cetuximab to a standard cisplatin plus vinorelbine regimen and was powered to detect a difference of 2 months in favor of the experimental arm (hazard ratio [HR] = 0.8). With 1,125 patients, a less clinically relevant, but still statistically significant survival difference of 1.2 months was found (HR = 0.871; 95% CI, 0.762 to 0.996; P = .044). Even considering a small dilution of the effect caused by the less frequent use in the post-treatment phase of tyrosine kinase inhibitors in the cetuximab arm (17% v 27%), the observed benefit remains clinically unsatisfactory, particularly taking into account costs and increased toxicity.

The BMS-099 (Bristol-Myers Squibb 099) study² added cetuximab to carboplatin taxanes doublets and enrolled 676 patients. The study had a power of 90% to detect a difference of 25% in progression-free survival. The study did not show a statistically significant difference between treatment arms on the primary end point (HR = 0.902; 95% CI, 0.761 to 1.069; P = .236). The effect observed in overall survival (HR = 0.89; 95% CI, 0.754 to 1.051; P = .169) was consistent with that observed in the FLEX study. Overall, the combination of the estimates of these two trials yields an overall reduction of mortality of approximately 12% (HR = 0.878; 95% CI, 0.791 to 0.975), thus confirming the modest clinical impact of the addition of cetuximab to chemotherapy.

In this scenario, the optimization of patient selection becomes of increasing importance throughout the identification of subgroups with a different response to cetuximab. Unfortunately the article by Khambata-Ford et al³ focusing on potential predictive markers of cetuximab in advanced NSCLC failed to detect any association among treatment efficacy and KRAS mutation status as well as any other epidermal growth factor receptor–related biomarkers evaluated.

On the basis of their results, can we claim that the effect of cetuximab in NSCLC is not related to any tumor and patient characteristics, as opposed to what is observed in colon cancer? We think that the article by Khambata-Ford et al³ must be considered as inconclusive because the analysis is based on small numbers of patients (ie, the analysis on KRAS was made on 167 wild-type and 35 mutant patients), which allows us to detect with an adequate power only extremely large interaction effects. Even more inconclusive is the analysis of the impact of cetuximab in patients with EGFR mutations, which was done in only 15 patients (eight in one group and seven in the other). The recent literature on analysis of predictive factors in NSCLC basically consists of results from post hoc analyses of small subgroups of patients included in phase III studies, and researchers need to be aware that developing predictive factors in retrospective series may lead to biased results.⁴ Although the authors of the study in question stated that the analyses were preplanned, no information on planning and sizing of these analyses was reported in the article, thus not allowing the reader to fully understand the weight of the evidence shown. In our opinion, another opportunity for increasing the knowledge on predictive markers has been lost.

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The author(s) indicated no potential conflicts of interest.