While therapy-related (t)-MDS has worse outcomes than de novo MDS (d-MDS), some t-MDS patients have an indolent course. Most MDS prognostic models excluded t-MDS patients during development. The performance of the International Prognostic Scoring System (IPSS), revised IPSS (IPSS-R), MD Anderson Global Prognostic System (MPSS), WHO Prognostic Scoring System (WPSS), and t-MDS Prognostic System (TPSS) was compared among patients with t-MDS. Akaike information criteria (AIC) assessed the relative goodness of fit of the models. We identified 370 t-MDS patients (19%) among 1950 MDS patients. Prior therapy included chemotherapy alone (48%), chemo-radiation (31%), and radiation alone in 21%. Median survival for t-MDS patients was significantly shorter than for d-MDS (19 v. 46 months, P<0.005). All models discriminated survival in t-MDS (P<0.005 for each model). Patients with t-MDS had a significantly higher hazard of death relative to d-MDS in every risk model, and had inferior survival compared to patients with d-MDS within all risk group categories. AIC Scores (lower is better) were 2316 (MDSS), 2343 (TPSS), 2343 (IPSS-R), 2361 (WPSS) and 2364 (IPSS). In conclusion, subsets of t-MDS patients with varying clinical outcomes can be identified using conventional risk stratification models. The MPSS, TPSS, and IPSS-R provide the best predictive power.Leukemia accepted article preview online, 23 January 2017.