Background: The optimal treatment for mCRC beyond 2nd line is still questioned. Recently, REG and TAS-102 showed to improve survival compared to BSC. While in real-world practice CTr/r is often considered in this setting, supporting evidences are limited. In absence of studies comparing all these strategies, we aimed to compare the prognostic performance of CTr/r, REG and TAS-102 in mCRC treated beyond 2nd line. Methods: mCRC pts progressing after at least 2 lines of CT, treated with CTr/r, REG or TAS-102 between Jan-10 and Jan-19 were considered eligible. The primary endpoint was OS; secondary endpoints were PFS and RR. Cox’s proportional hazard models for survivals were estimated. A propensity score (PS) adjustment for baseline characteristics was further accomplished for survival analysis. Results: The clinical data of 341 pts (CTr/r 133, REG 150, TAS-102 58) were retrospectively collected. At multivariate analysis type of treatment, ECOG PS, number of metastatic sites and treatment line independently correlated with OS (p< .001, p .001, p< .001 and p .029, respectively). The mOS was 18.5 (95% CI, 14.3–22.7), 6 (95% CI, 5.6–9.5) and 7.6 months (95%CI, 5.6–9.5), for CTr/r, REG and TAS-102 group, respectively (log-rank p< .0001). mOS was significantly longer for pts receiving CTr/r than for those treated with REG/TAS-102 (15.8 vs 7.1 months; adjusted HR 1.96, 95% CI 1.44-2.66; p< .0001) at the PS analysis, adjusted for ECOG PS, number of metastatic sites and treatment line; 2-yrs OS was 34% and 11.6% for CTr/r and REG/TAS-102, respectively. PFS was significantly longer for pts receiving CTr/r than for those treated with REG/TAS-102 (5.5 vs 3.9 months; HR 1.45, 95% CI 1.11-1.91; p .006) at the PS analysis. Accordingly, RR was higher in pts receiving CTr/r compared to REG/TAS-102 (29.0 vs 1.5%; Chi-square p< .00001). Conclusions: Our analysis, although underpowered, generates the hypothesis of a superiority of CTr/r in comparison to REG or TAS-102, in both efficacy and activity. Given the retrospective nature of our analysis, and the potential role of selection bias in treatment assignment, a prospective validation is mandatory.