We attempted to prospectively select individualized chemotherapy for 165 non-small cell lung cancer patients based on in vitro analysis of neuroendocrine (NE) markers and drug sensitivity testing (DST) using fresh tumor. The chemotherapy used for small cell lung cancer (SCLC) was selected when NE marker expression determined by L-dopa decarboxylase assay was documented. Selection of chemotherapy for other patients was guided by DST results using a modified dye exclusion assay when available; otherwise etoposide and cisplatin was administered. A total of 112 of 165 (68%) specimens were assayed for L-dopa decarboxylase and 36 patients (22%) had DST. In vitro data directed management for 27 of 96 (28%) patients given chemotherapy: 6 with NE markers were treated with the SCLC regimen; and 21 (58% of those with DST) received their DST-selected chemotherapy regimen. There were no significant differences in response rate among all 3 treatment arms (P = 0.076). However, response to chemotherapy for the patients treated prospectively with a SCLC regimen was 3 of 6 (50%), marginally better than patients given their DST-selected chemotherapy regimen (2 of 21; 9%; P = 0.056) or those treated with etoposide and cisplatin (10 of 69; 14%; P = 0.061). When patients whose NE markers were identified retrospectively are included, 4 of 9 (44%) responded to administered chemotherapy, compared to 7 of 55 (13%) with no NE markers present (P = 0.04). There were no differences in survival among the three treatment groups. Cisplatin and etoposide comprised the most active regimen in vitro for tumors from 16 of 36 (44%) patients, potentially limiting the benefit of DST since this is often the empiric therapy for non-SCLC. Furthermore, the correlation between in vitro and clinical response is nonsignificant for all drugs tested, highlighting the overall relative resistance of non-SCLC tumors to currently available chemotherapy.