AUTHORS
Curtis Lachowiez, MD, Ying Huang, MA, MS2, Tara L. Lin, MD3, Shivani Handa, MD4, Ann-Kathrin Eisfeld, MD5, Ronan Swords, MD6, Elie Traer, MD, PhD7, Wendy Stock, MD8, Olatoyosi Odenike Jr, MD9, Anand Ashwin Patel, MD10, Maria R. Baer, MD11, Vu Duong, MD12, Eytan M Stein I, MD13, William Blum, MD14, Colin Vale15, Emily Curran, MD16, Yazan F. Madanat, MD17, Rebecca L Olin, MD18, Gary J. Schiller, MD19, Angela Nichols, NP20, Lacey Williams, MD21, Nyla Heerema, MD22, Timothy Chen, PhD23, Sonja Marcus, MPH24, Molly Martycz25, Mona Stefanos, MD, MBChB26, John Byrd, MD27, Brian Druker, MD28, Ross L. Levine, Prof. Dr.29, Ashley Yocum, PhD30, Alice Scott Mims, MD31, Joshua F. Zeidner, MD32, Uma Borate, MBBS, MSc33
AFFILIATIONS
1 Oregon Health & Science University, Portland, Oregon, United States
2 The Ohio State University, Columbus, Ohio, United States
3 University of Kansas Medical Center, Kansas City, Kansas, United States
4 The Ohio State University, Columbus, Ohio, United States
5 The Ohio State University, Columbus, Ohio, United States
6 Oregon Health & Science University, Portland, Oregon, United States
7 Oregon Health & Science University, Portland, Oregon, United States
8 University of Chicago Comprehensive Cancer Center, Chicago, Illinois, United States
9 University of Chicago Comprehensive Cancer Center, Chicago, Illinois, United States
10 University of Chicago Comprehensive Cancer Center, Chicago, Illinois, United States
11 University of Maryland Greenebaum Comprehensive Cancer Center, Baltimore, Maryland, United States
12 University of Maryland Greenebaum Comprehensive Cancer Center, Baltimore, Maryland, United States
13 Memorial Sloan Kettering Cancer Center, New York City, New York, United States
14 Winship Cancer Institute of Emory University, Atlanta, Georgia, United States
15 Winship Cancer Institute of Emory University, Atlanta, Georgia, United States
16 University of Cincinnati, Cincinnati, Ohio, United States
17 University of Texas Southwestern Medical Center, Dallas, Texas, United States
18 University of California San Francisco, San Francisco, California, United States
19 David Geffen School of Medicine at University of California, Los Angeles, Los Angeles, California, United States
20 University of North Carolina, Chapel Hill, North Carolina, United States
21 University of North Carolina, Chapel Hill, North Carolina, United States
22 The Ohio State University, Columbus, Ohio, United States
23 The Ohio State University, Columbus, Ohio, United States
24 Blood Cancer United, Washington, District of Columbia, United States
25 Blood Cancer United, Washington, District of Columbia, United States
26 The Ohio State University, Columbus, Ohio, United States
27 University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania, United States
28 Oregon Health & Science University, Portland, Oregon, United States
29 Memorial Sloan Kettering Cancer Center, New York City, New York, United States
30 Blood Cancer United, Washington, District of Columbia, United States
31 The Ohio State University, Columbus, Ohio, United States
32 University of North Carolina, Chapel Hill, North Carolina, United States
33 The Ohio State University, Columbus, Ohio, United States
ABSTRACT
Introduction
ELN 2024 “Less-Intensive” risk classification is the current standard for prognostication in patients (pts) with newly diagnosed acute myeloid leukemia (ND-AML) receiving azacitidine (AZA) plus venetoclax (VEN). Prognostic Risk Integration for Survival Modeling (PRISM) in AML is a recently developed clinico-genomic risk model – encompassing 17 pre-treatment clinical, cytogenetic, and mutational variables – that demonstrated robust prognostic discrimination in large real-world cohorts treated with hypomethylating agents (HMA)+VEN. Whether PRISM retains predictive performance in pts treated on prospective clinical trials is unknown. We therefore performed a post hoc evaluation of PRISM in pts enrolled on the randomized phase II OPTI-AML trial.
Methods
OPTI-AML was an open-label, randomized phase II non-inferiority study comparing CR rates following two cycles of AZA plus 14 vs. 28 days of VEN in pts ≥60y with ND-AML (Borate et al, Blood 2026).
The primary objective of this analysis was to evaluate the PRISM risk model’s ability to discriminate overall survival (OS) in a prospective trial cohort. Secondary endpoints included progression-free survival (PFS) and comparison of PRISM and ELN 2024 in terms of prognostic performance. The PRISM risk score divided all pts into “PRISM-3” risk groups (low, moderate, high) using published thresholds.
Median survival was estimated using the Kaplan-Meier method. Cox proportional hazards models were fit to estimate global Wald χ2 test p-values, Harrell's concordance (C) index (higher values indicate better discrimination), and Akaike information criterion (AIC; lower values indicate better balance of fit and parsimony). Median follow-up was estimated using the reverse Kaplan-Meier method.
Results
Among 169 pts with ND-AML, 86 vs. 83 were randomized to receive 14 vs. 28 days of VEN, respectively. Median age was 73 years; ELN 2024 classified 110 (65%), 27 (16%), and 32 (19%) pts as favorable-, intermediate-, and adverse-risk, respectively, whereas PRISM classified 70 (41%), 60 (36%), and 39 (23%) pts as low-, moderate-, and high-risk. PRISM reclassified 37% of pts overall, including 38% (n=42) of ELN 2024 favorable, 48% (n=13) of intermediate, and 13% (n=4) of adverse risk pts.
After a median follow-up of 15.6 months, 76 deaths occurred. Median OS for the full trial cohort was 17.1 months. No significant OS difference was observed between pts randomized to 14 vs. 28 days of VEN (median: 13.6 vs. 21.4 months, p=0.07). Median OS was 23.2, 13.1, and 9.0 months for ELN 2024 favorable-, intermediate-, and adverse-risk pts, respectively (P=0.0002; C-index: 0.608).
Median OS was not reached (12-month OS: 75%), 18.0, and 9.0 months for PRISM low-, moderate-, and high-risk pts, respectively (P<0.0001; C-index: 0.644), consistent with the original PRISM cohorts. PRISM effectiveness was maintained after adjustment for trial treatment arm (C-index: 0.663). OS was not significantly different between 14 vs. 28 days of VEN in pts classified as PRISM low- (p= 0.200), moderate- (p= 0.300), or high-risk (p=0.600). PRISM also effectively stratified PFS across low, moderate, and high risk groups (median: 21.4, 9.9, and 5.7 months, respectively, P<0.0001; C: 0.630).
Modeling OS, PRISM had enhanced goodness-of-fit (AIC 673.9 vs 681.4) and discrimination (C-index: 0.644 vs 0.608) compared to ELN 2024. PRISM also improved OS discrimination vs. other published classifiers including the refined ELN 2024 (C-index: 0.638; AIC: 675.1), Mayo (C-index: 0.638; AIC: 680.9), and Beat AML risk models (C-index: 0.594; AIC: 682.1).
Importantly, within the ELN 2024 favorable-risk group, PRISM identified substantial prognostic heterogeneity. Median OS was not reached (12-month OS: 77%) for PRISM low-risk pts (n=65) compared with 18.6 months and 1.5 months for PRISM moderate- (n=42) and high-risk (n=3) pts, respectively (P=0.0009), highlighting clinically meaningful risk refinement within pts classified as ELN 2024 favorable risk.
Conclusion
In this post hoc analysis of a randomized phase II trial, PRISM-AML demonstrated comparatively high prognostic discrimination in pts with ND-AML treated with AZA+VEN. PRISM reclassified more than one-third of pts and identified marked survival heterogeneity within ELN 2024 favorable-risk disease, supporting its potential utility for risk stratification in future clinical trials and for identifying pts who may benefit from treatment intensification with novel therapeutic approaches.
DISCLOSURES
Curtis Lachowiez, MD: AbbVie - Research Funding; COTA, Inc. - Consultancy (Includes expert testimony); Syndax - Consultancy (Includes expert testimony); Servier - Consultancy (Includes expert testimony); Abbvie/Genentech - Consultancy (Includes expert testimony); Astellas Pharma - Consultancy (Includes expert testimony); Rigel - Consultancy (Includes expert testimony)