Exhibit 99.3

Abstract 3259

Relationship of Bone Marrow Blast (BMBL) Response to Overall Survival (OS) in Patients with Higher-risk Myelodysplastic Syndrome (HR-MDS) Treated with Rigosertib After Failure of Hypomethylating Agents (HMAs)

Lewis R. Silverman, MD(1), Pierre Fenaux, MD(2), Aref Al-Kali, MD(3), Maria R. Baer, MD(4), Mikkael Sekeres, MD(5), Gail Roboz, MD(6), Gianluca Gaidano, MD(7), Bart Scott, MD(8), Peter Greenberg, MD(9), Uwe Platzbecker, MD(10), David P. Steensma, MD(11), Suman Kambhampati, MD(12), Karl-Anton Kreuzer, MD(13), Lucy Godley, MD(14), Robert Collins, Jr, MD(15), Ehab Atallah, MD(16), Shyamala C Navada, MD(1), Francois Wilhelm, MD, PhD(17), Nozar Azarnia, PhD(17), Guillermo Garcia-Manero, MD(18)

(1) Division of Hematology/Oncology, Icahn School of Medicine at Mount Sinai, New York, NY

(2) Hospital St Louis, Paris, France

(3) Mayo Clinic, Rochester, MN

(4) University of Maryland, Baltimore, MD

(5) Cleveland Clinic, Cleveland, OH

(6) Weill Cornell Medical College, New York, NY

(7) Amedeo Avogadro University of Eastern Piedmont, Novara, Italy

(8) Fred Hutchinson Cancer Research Center, Seattle, WA

(9) Stanford Medical School, Stanford, CA

(10) Universitätsklinikum Dresden, Dresden, Germany

(11) Dana-Farber Cancer Institute, Boston, MA

(12) University of Kansas Medical Center, Westwood, KS

(13) Universitätsklinikum Köln, Köln, Germany

(14) University of Chicago Medical Center, Chicago, IL

(15) Univ of Texas, Southwestern Medical Center at Dallas, Dallas, TX

(16) Froedtert Hospital and Medical College of Wisconsin, Milwaukee, WI

(17) Onconova Therapeutics, Inc., Newtown, PA

(18) MD Anderson Cancer Center, Houston, TX

Background: Patients (pts) with HR-MDS have a median OS of 4 to 6 months (mo) after HMA failure (Prebet et al, J Clin Oncol 2011) and no approved salvage therapy. Development of new therapeutics for this population will benefit from the availability of surrogate endpoints and markers that can predict survival. Gore et al established response to azacitidine (Vidaza®) in first-line therapy for HR-MDS as a reasonable surrogate to predict survival (Gore et al, Haematologica 2013). Rigosertib, a novel dual PI3K/PLK pathway inhibitor, has been shown to reduce bone marrow blasts (BMBL) in these pts (Seetharam et al, Leuk Res 2012). Silverman et al described complete or partial bone marrow (BM) response, or stabilization after 4-8 weeks (wks) of treatment with rigosertib as a potential surrogate for predicting survival in pts with HR-MDS after failure of primary HMA therapy (Silverman et al, Hematol Oncol 2014). We tested this hypothesis in the context of a randomized Phase III trial.

Methods: Pts with HR-MDS were randomly assigned 2:1 to rigosertib or best supportive care (BSC) after progressing on, failing to respond to, or relapsing after HMA treatment. BM aspirates were assessed pretreatment, at 4 weeks and at 8-week intervals thereafter. Central slide review was undertaken in a representative population of samples. The BMBL response at each

time point was assessed using the following definitions:  bone marrow complete response (mCR) = BMBL < 5% and decrease of > 50% from baseline; bone marrow partial response (mPR) = BMBL decrease from baseline of > 50%, but BMBL still > 5%; stable disease (SD) = BMBL decrease or increase from baseline of < 50%; progressive disease (PD) = BMBL increase from baseline of > 50% by an absolute minimum of 5%; Not evaluable (NE).

Results: Bone marrow assessment was carried out in 156 patients (pts) on the rigosertib arm and 24 pts on the BSC arm at 4 wks after enrollment, and in 86 and 20 pts, respectively, at 12 wks. The invasive BM procedure was optional on the BSC arm, which accounts for the low number of assessments in this group. BM responses at the 2 time points are presented in Table 1.

Since no difference in overall survival was noted between pts who had objective BM response and those who did not progress (ie, stable disease), a landmark analysis was conducted that separated pts who were alive at the 4-wk landmark time into two 4-wk response categories: BM response + SD vs. PD. Results of this analysis in rigosertib-treated patients were statistically significant at p = 0.011, with a hazard ratio (HR) of 0.62 and a median OS (from 4 wks onward) of 9.8 months in the mCR + mPR + SD group vs. 4.6 months in the PD group (Figure 1).

Another landmark analysis was conducted at 12-wks. Results of this analysis were also significant (p < 0.001) in rigosertib-treated patients, with an HR of 0.39 and a median OS (from 12 wks onward) of 10.4 months in the mCR + mPR + SD group vs.7.5 months in the PD group (Figure 2).

A time-dependent Cox regression of OS by 4-wk BMBL response reinforced the validity of the 4-wk and 12-wk BM assessments as surrogate biomarkers for survival (Table 2).

Conclusions: These data suggest that BMBL response at 4 or 12 weeks was correlated with OS in this population of pts with HR-MDS treated with rigosertib after HMA failure and are consistent with previous observations in Phase II studies.

Table 1

4- and 12-week Bone Marrow Blast Response (Intention-to-Treat Population)

* Bone marrow assessment was not required on the BSC arm.

Table 2

Time-dependent Cox Regression of Overall Survival by Bone Marrow Blast Response

*Stratified by pretreatment BMBL: 5%-19% vs. 20%-30%