Background: In MPACT, pts who received nab-P + G vs G had improved overall survival (OS; median 8.5 vs 6.7 mo; HR 0.72; p= 0.000015). Here we assessed potential PFs of OS. Methods: 861 pts with MPC were randomized 1:1, stratified by region, presence of liver metastases, and Karnofsky performance status (KPS), to nab-P + G or G. OS was described in subgroups. A step-wise multivariate analysis (with significance level for entry of 0.20 and for stay of 0.10) was performed to evaluate the treatment effect and identify possible predictors of OS. Results: Pts with poorer PFs had a shorter median OS, consistent with the literature, and OS consistently favored nab-P + G in pts with these PFs (Table). Region of Eastern Europe, age ≥ 65 years, poorer KPS, presence of liver metastases, and number of metastatic sites all predicted OS (increased risk of death). The treatment effect remained significant (HR 0.72; 95% CI, 0.605 - 0.849; p < 0.0001, Cox proportional hazards [CPH] model). In another multivariate analysis in which baseline CA19-9 was added to the final model described above, the treatment effect HR was 0.67 (95% CI, 0.573 - 0.794; p < 0.0001, CPH model). Baseline CA19-9, a predictor of OS by univariate analysis, was not predictive after correction for the above factors. Conclusions: In MPACT, the most important predictors of OS were KPS, age, presence of liver metastases, number of metastatic sites, and region. After correcting for these factors, assignment to nab-P + G was an independent significant predictor of improved survival. Clinical trial information: NCT00844649.