Erlotinib is approved for the treatment of patients with EGFR mutation positive, metastatic NSCLC. It is also approved as second/third line therapy for EGFR mutation negative patients, but in this setting the benefit of erlotinib is modest and there is no validated biomarker for selecting EGFR wild-type patients who may benefit the most from the treatment. We retrospectively assessed EGFR and K-RAS mutational status, and EGFR, c-MET and IGF1-R expression in tumor samples of 72 patients with metastatic NSCLC treated with erlotinib after at least one prior line of chemotherapy, from 2008 to 2012. We analyzed the association between biomarkers and outcome (RR, PFS, and OS). EGFR mutated patients achieved a better RR (56% vs 8%, p = .002), PFS (10 vs 3 months, HR 0.53, p = 0.48) and OS (20 vs 6 months, HR 0.55, p = .07), compared to EGFR wild-type patients. Among 63 EGFR wild-type patients, those with EGFR high-expression had a better outcome in terms of RR (40% vs 2%, p = .002), PFS (7.5 vs 2 months, HR 0.45, p = .007) and OS (30 vs 5 months, HR 0.34, p < .001) compared to patients with EGFR intermediate or low/negative-expression. IGF1-R expression, c-MET expression and K-RAS mutational status did not significantly affect the outcome; however, no patients with K-RAS mutation or c-MET high-expression achieved an objective response. In patients with metastatic, chemo-refractory EGFR wild-type NSCLC, EGFR high-expression may represent a positive predictor of activity for erlotinib, whereas K-RAS mutation and c-MET high-expression may predict lack of activity. These findings deserve further prospective evaluation.