Interaction of treatment and biomarker in advanced non-small cell lung cancer

Purpose: There has been a long-standing interest in the investigation of interactions in science. The objective of the study is to evaluate interaction between Epidermal Growth Factor Receptor (EGFR) mutation and treatment from randomized, phase II study of chemotherapy versus chemotherapy plus erlotinib in patients with progressive Non-Small Cell Lung Cancer (NSCLC) following clinical benefit from erlotinib. Materials and Methods: Forty-six patients with advanced stage NSCLC and progression from erlotinib were randomized to receive chemotherapy (pemetrexed or docetaxel) or chemotherapy plus erlotinib between 2008 and 2012. Patient characteristics at baseline including age, gender, tumor stage, race, smoking history and EGFR mutation status along with the clinical outcomes, namely response, Progression- Free Survival (PFS) and Overall Survival (OS) were obtained. The effects of treatment, EGFR mutation and interaction between the two on survival outcomes were evaluated using Cox proportional hazards model with first-order interaction. Results: For PFS, there was a significant interaction between treatment (arm B) and EGFR mutation (mutant EGFR+) (p = 0.018), although the main effects of treatment (arm B vs. arm A) and EGFR mutation (mutant vs. wild-type EGFR) were statistically significant (with p = 0.03 and p = 0.088, respectively) favoring arm B and mutant EGFR+. Thus when taking the interaction between treatment and EGFR into account, the hazard ratio comparing arm B to arm A when EGFR is positive was 1.49 (95% CI: 0.72, 3.11); and the hazard ratio comparing arm B to arm A when EGFR is negative was 0.17 (95% CI: 0.04 - 0.84). Similarly, for OS, there was a significant interaction between treatment and EGFR mutation (p = 0.02), with significant main effects of treatment and EGFR favoring arm B and mutant EFGR+. Taking together, the hazard ratio comparing arm B to arm A when EGFR is positive was 1.61 (95% CI: 0.68 - 3.82); and the hazard ratio comparing arm B to arm A when EGFR is negative was 0.16 (95% CI: 0.03 - 0.9). Conclusion: The interaction identified by Cox model shows there was an antagonistic effect between chemotherapy + erlotinib and EGFR mutation, a situation that the whole is less than the sum of the parts, despite the prolonging-survival main effect of each factor from Cox model. As a result, the continuing erlotinib beyond progression adds no benefit in survival outcomes but leads to an increase in adverse events.