Gefitinib has been available in the market for 20 years, but its pharmacokinetic mechanism of response is little known. In this study, we examined the pharmacokinetic and metabolomic profiles in non-small cell lung cancer (NSCLC) patients with sensitive EGFR mutations. A total of 216 advanced NSCLC patients were enrolled, and administered gefitinib at the standard dosage of 250 mg/day, which was established in heterogeneous subjects with non-sensitive mutations. We identified and quantified three main metabolites (named as M1, M2 and M3) in the plasma of patients, the correlations between the concentration of gefitinib/metabolites and efficacy were analyzed. In exploratory and validation set, gefitinib concentration was not correlated with clinical effects. Considering the result that the therapeutic effects of 250 mg/2-day was better than that of 250 mg/day in a multiple center clinical trial, the standard dose might be higher than that for maximal efficacy according to the hypothetical dose-response curve. Among the three metabolites, the IC50 of M2 in HCC827 and PC9 cell lines was significantly lower, and Conc.(brain)/Conc.(plasma) of M2 in mice was significantly higher than those of gefitinib, suggesting its higher potential to penetrate blood-brain barrier and might be more effective in the treatment of brain metastatic tumor than gefitinib. Consistently and attractively, higher M2 plasma concentration was found to be correlated with better clinical outcome in patients with brain metastases (the median PFS of C-M2 < 12 ng/mL and C-M2 >= 12 ng/mL were 17.0 and 27.1 months, respectively, P = 0.038). The plasma concentration of M2 >= 12 ng/mL was a strong predictor of the PFS of NSCLC patients. In conclusion, for NSCLC patients with EGFR sensitive mutations, the standard dose is suspectable and could be decreased reasonably. M2 plays an important role in efficacy and may be more effective in the treatment of metastatic tumor than gefitinib.