5056 Background: Although PFS is considered the standard primary end-point in EPC, the correlation with OS has never been explored in RCTs randomizing patients (pts) to HT plus radiotherapy (RT) versus RT. Given the relatively long prognosis in this disease setting, the correlation between PFS and CSS should be investigated as well. Methods: All phase III trials reporting all outcome’ data were considered eligible. The correlation has been explored according to a linear regression model considering both each single outcome pair (PFS, OS and CSS rates) for all arms, and each reported Hazard Ratio (HRs). The correlation was estimated according to both the Pearson- (r) and R2-coefficient (parametric) and the Spearman coefficient (Rho, non-parametric). A sensitivity analysis in 2 subgroups (long- and short-term HT) to test for effect robustness has been accomplished as well. A model to determine the target sample size to determine CSS benefit of 3%, 4%, 6% and 7% months, respectively, was calculated as well. Results: Six RCTs (4,212 pts) were collected (follow-up range: 4.5–7.6 years). In the overall population, when considering the crude rates, a linear stronger correlation was found between PFS and CSS (r=0.71, R2=0.51, p=0.003; Rho=0.75, p=0.005), rather than with OS (r=0.55, R2=0.30, p=0.06; Rho=0.78, p=0.11). Again, when considering HRs, a linear stronger correlation was found between PFS and CSS (r=0.87, R2=0.76, p=0.02; Rho=0.94, p=0.005), rather than with OS (r=0.75, R2=0.56, p=0.08; Rho=0.77, p=0.07). Similar correlations were found whatever subgroups was explored. The sample size model (on the basis of the beta-coefficient=0.71), calculate 4,575, 2,006, 1,115 and 700 pts to improve PFS of 4%, 6%, 8%, and 10% months, which means to improve CSS of 2.8%, 4.3%, 5.7% and 7.1%, respectively. Conclusions: The correlation between PFS and CSS in RCTs exploring the benefit of adding HT to RT for EPC is significant, and suggests its further investigation as surrogate end-point. The natural history of the disease clearly explains the stronger correlation of PFS with CSS rather than with OS. No significant financial relationships to disclose.