The use of serial serum measurements of the carbohydrate antigen 19-9 (CA19-9) to guide treatment decisions and serve as a surrogate end point in clinical trial design requires further validation. We investigated whether CA19-9 decline represents an accurate surrogate for survival and time to treatment failure (TTF) in a cohort of 76 patients with advanced pancreatic cancer receiving fixed-dose rate gemcitabine in three separate studies. Statistically significant correlations between percentage CA19-9 decline and both overall survival and TTF were found, with median survival ranging from 12.0 months for patients with the greatest degree of biomarker decline (> 75%) compared with 4.3 months in those whose CA19-9 did not decline during therapy (P < 0.001). Using specific thresholds, patients with > or = 25% decline in CA19-9 during treatment had significantly better outcomes than those who did not (median survival and TTF of 9.6 and 4.6 months vs 4.4 and 1.5 months; P < 0.001). Similar results were seen using both 50 and 75% as cutoff points. We conclude that serial CA19-9 measurements correlate well with clinical outcomes in this patient population, and that decline in this biomarker should be entertained for possible use as a surrogate end point in clinical trials for the selection of new treatments in this disease.