Prognostic factor analysis of overall survival (OS) in gastric cancer from two phase III studies of second-line ramucirumab (RAM) (REGARD and RAINBOW) using pooled individual patient (pt) data.

4028 Background: From 2009-2012, 1020 pts were enrolled in two phase III, randomized, double-blind studies of RAM in metastatic gastroesophageal junction and gastric adenocarcinoma following progressive disease (PD) on first-line platinum- and/or fluoropyrimidine-containing therapy: REGARD (N = 355, RAM + best supportive care [BSC] vs placebo [PL] + BSC) and RAINBOW (N = 665, RAM + paclitaxel [PTX] vs PL + PTX). Methods: Individual pt data were pooled, and 41 key baseline covariates, common in both studies, were examined (19 clinical characteristics; 22 lab parameters). Lab tests were parameterized in two ways based on local lab abnormality assessments (high/normal/low): high vs normal or low; low vs normal or high. To identify prognostic factors for OS, univariate Cox models were first used to select covariates with p ≤ 0.05. For these covariates, a multivariate Cox model was used to make stepwise selection with both entry and exit p=0.01. All models were stratified by treatment and geographic region. Re...