Correlation of tumor response and survival in advanced NSCLC patients treated with paclitaxel plus carboplatin (PC) versus paclitaxel plus carboplatin plus gemcitabine (PCG)
7650 Background: We showed that PCG significantly increases both response rate (RR) (43.6% vs 20%) and median survival (10.8 mo vs 8.3 mo) over PC and that at Cox analysis, the only independent prognostic factors were PS and treatment (Paccagnella et al, J Clin Oncol 2006;24: 681–687). According to the Prentice criteria (Stat Med 1989;8: 431–440), to directly relate response and survival it is necessarily that responding patients (and non responding) of both arms have a similar survival and that the survival difference between the two arms disappear when the response factor is included in the multivariate analysis. Methods: Out of 324 pts included in the original analysis, 26 pts were not evaluable for response (early death, toxicity, refusal) before the planned response evaluation at two months and were excluded (15 pts from PC arm and 11 pts from PCG arm). The analysis however was also performed considering the non evaluable patients as non responders. Results: Overall, Responder patients had a median Survival that nearly doubled that of no responders: 14.73 mo vs 7.67 mo (HR: 0.49; CI: 0.31–0.54; P=0.000). No responder pts from PC and PCG arms had a similar survival (median 7.53 mo and 8.07 mo respectively; P= 0.96) as well as responder (CR + PR) patients (median 14.13 mo and 15.40 mo respectively; P=0.38). The principal difference between the two arms was that more than the double of patients in PCG arm responded (43.6% vs 20%) and consequently had a survival advantage of clinical relevance in comparison to patients in PC arm. When tumor response was introduced in the Cox model (as a four level variable), the difference in Overall Survival between PCG and PC changed from a significant level (HR=1.28; CI 1.00–1.63; P=0.049) to a not significant level (HR=0.99; CI: 0.76 - 1.28; P=0.97). Conclusions: To our knowledge this is the first report showing a significant direct correlation between response and survival in advanced NSCLC according to Prentice criteria. No significant financial relationships to disclose.