Background: NMAs of cancer treatments are usually based on the proportional hazard assumption (PHA) which implies a constant hazard ratio as well as treatment effect on only scale parameter and therefore not reflecting clinical dynamics of survival outcomes over time. Multivariate NMA which does not rely on the PHA is often constrained by lack of individual patient-level data (IPD). We applied a novel method of reconstructing pseudo-IPD from published survival curves that enabled application of a dynamic method of NMA of survival curves without PHA constraint to compare the efficacy of 6 later-line therapies for metastatic colorectal cancer (mCRC), compared to placebo. Methods: We identified 8 phase II/III trials reporting the efficacy of 6 later-line therapies for mCRC from PubMed, Embase, Scopus and Cochrane Library. Trial survival curves were reconstructed based on pseudo-IPD generated using the Guyot method, and the estimated survival proportions and hazard ratios (HRs) over 12 months (12m) of follow up were calculated using a random effects lognormal distribution model with placebo as the common comparator. Results: Compared to placebo, 12m PFS HRs varied from 0.29 (95%CrI = 0.19-0.44) for TAS+bevacizumab to 0.72 (95%CrI = 0.62-0.85) for atezolizumab. Similarly, OS HRs ranged from 0.55 (95%CrI = 0.26-0.92) for TAS+bevacizumab to 1.01 (95%CrI = 0.79-1.20) for atezolizumab (Table). Conclusions: In this NMA of survival curves, all therapies except atezolizumab were superior to placebo in PFS and OS over 12m follow up period. TAS+bevacizumab provided the highest survival advantage over placebo, with fruquintinib, regorafenib, and TAS-102 also showing survival benefits. TAS+bevacizumab may be the preferred choice for later-line mCRC treatment.