Background: Identifying non-responders to expensive salvage therapies with modest benefits and substantial treatment related adverse events (TRAEs) (e.g. regorafenib [Reg] or TAS102 [Tas] in mCRC) is key to precision care. Retrospective studies suggest that ctDNA changes at timepoints (4-10 weeks [wks] into therapy) before radiographic assessment may predict treatment outcomes. However, prospective studies assessing early ctDNA changes are lacking and clinical utility remains uncertain. Methods: TACT-D is a randomized study to validate early dynamic changes in ctDNA (ΔctDNA: change in either maximum variant allele frequency (maxVAF) or mean VAF in predicting treatment response/resistance. Patients (pts) with mCRC clinically eligible for Reg/Tas were randomly assigned 2:1 to either standard of care (SOC) or ctDNA arm. All pts had ctDNA sequencing by Guardant360 CDx assay on cycle 1 day 1 (C1D1) and C1D15. On SOC arm, pts were restaged at 8 wks. On ctDNA arm, ΔctDNA (C1D15 - C1D1) was run in real-time and increase in ctDNA (ΔctDNA > 0) triggered early radiographic restaging. Therapy was continued for responders (RECISTv1.1 stable disease/response) and stopped for progression (non-responders). Co-primary endpoints were: 1) comparison of TRAEs among study arms and 2) association of ΔctDNA and objective response rate (RR). Key secondary endpoints were progression free (PFS) and overall (OS) survival. Study was powered (82%; 2-sided α = .05) to detect 30% decrease in toxicity. Results: Between 4/2019 and 8/2023, 100 pts were randomized; 80 evaluable had median age of 56 years, 46% were females, 44% and 56% received Tas and Reg, respectively. Baseline ctDNA levels (ρ 0.90) and ΔctDNA (ρ 0.68) using maxVAF (reported below) and mean VAF showed strong correlation (P < .001). Median ΔctDNA for entire cohort was -47% with no significant difference by treatment arms (SOC -45% v Exp -58%, P = .79) and therapy (Tas -71 v Reg -44, P = .19). ΔctDNA increased in 18% pts. Grade 3/4 TRAEs (32% v 40%, P = .62) did not differ significantly between arms. No significant association was seen between ΔctDNA and RR (OR .88, P = 1.0), PFS (HR .99, P = .88) and OS (HR 1.00, P = .64). Notably, higher baseline maxVAF was strongly associated with response (median maxVAF: 21.8% in non-responders v 3.4% in responders), PFS (HR 1.02) and OS (HR 1.03) (all P < .001). After adjusting for baseline maxVAF, ΔctDNA was found to be associated with both PFS (HR 1.87, P = .038) and OS (HR 3.55, P = .001). Conclusions: In the first prospective study of clinical utility of monitoring ctDNA in mCRC, baseline ctDNA was strongly prognostic for clinical benefit from salvage therapies in mCRC. Adjusted for this prognostic impact, ΔctDNA between C1D1 and C1D15 was predictive of clinical outcomes. Efforts are needed to establish novel signatures, optimal cutoffs/intervals for assessing ctDNA response in mCRC, tailored to pts and their therapy. Clinical trial information: NCT03844620.