Background: BL-cfDNA combined with plasmatic changes in tumor-specific mutations after 14 days of therapy are explored as independent determinants of outcome in pts with aCRC. Methods: Archival tumor tissue and plasma samples at BL and D14 after regorafenib initiation were prospectively collected in aCRC pts (n=141) in a multicentric trial (NCT01929616). Tumor-specific mutations, selected on their allelic frequency on CRC-oriented targeted sequencing of tumor tissue, were analyzed for circulating tumor DNA (ctDNA) at BL and D14 via droplet digital PCR (Bio-Rad QX200 ddPCR system) using a 12 % cutoff to assess changes in ctDNA levels. Results: The cfDNA’s optimal cutoff at BL (Contal & O’Quigley method) is 1 µg/ml in 134/141 evaluable pts. Hazard ratio (HR) of cfDNA ≥1 vs <1 is 2.50 (95% CI, 1.73-3.63) P<0.001 for progression-free survival (mPFS) and 3.83 (95% CI, 2.57-5.71) P<001 for overall survival (mOS). The most common mutated genes (96 evaluable pts for ctDNA) are APC (73%), TP53 (72%), KRAS (66%), and PI3KCA (23%). On average 2 (1-4) mutations are followed per pt. An increase of ≥1 tumor-specific mutation as compared to none is associated with a significantly worse outcome: mPFS 1.3 vs 3.0 months (mo) (HR 1.88, P=0.002, 95% CI 1.24-2.85) and mOS 3.9 vs 8.5 mo (HR 2.04, P<0.001, 95% CI 1.33-3.12). BL cfDNA level and ctDNA changes between BL & D14 are not statistically correlated (P=0.23), and combined, define 4 subgroups with different prognosis. (Table 1) Conclusions: BL cfDNA and early (D14) changes in ctDNA during treatment with regorafenib are independent predictive markers for pts’ outcome and appear as potential robust tools for treatment personalization. Clinical trial information: NCT01929616