13534 Background: The oral fluoropyrimidine X has superior efficacy and improved safety compared with bolus 5-FU/LV in CRC. Twice-daily oral X is replacing 5-FU as the backbone of combination regimens. The combination of X and 3-weekly oxaliplatin (XELOX regimen) has demonstrated similar efficacy to FOLFOX with some safety advantages and enhanced convenience in phase II/III clinical trials. The objective of this retrospective study was to investigate the use of CEA as an early predictor of outcome. Methods: The data used in this analysis came from a previously published large phase II study (n=96) of XELOX as first-line treatment for patients with MCRC [Cassidy et al. J Clin Oncol 2004]. A population-based pharmacodynamic analysis of CEA and tumor size progressions was performed. The resulting data were linked to hazard models for the secondary clinical endpoints: time to disease progression (TTP) and overall survival. As we were interested in early prediction, model-predicted time courses of CEA and tumor size (up to 10 weeks) were examined as covariates in the hazard models. The resulting predictions for TTP and survival based on either measure were tested and compared using stochastic simulations. Results: An indirect effect model provided the best description of CEA progression. For tumor size, a similar indirect effect model with an added resistance term gave the best description. CEA performed marginally better than tumor size in predicting survival and less well in predicting TTP. For survival the misclassification rate decreased from 0.426 to 0.404 when using tumor size and to 0.396 when using CEA as predictors. Conclusions: CEA and tumor size were found to have clinically equivalent power as early phase predictors of clinical outcome. [Table: see text]