What question did this study set out to answer?

This research aims to identify factors influencing early mortality in acute myeloid leukemia patients and propose strategies for improvement.

April 5, 2026

Abstract 5429: REAL-AML: Real-world predictors of early mortality and treatment inequities in acute myeloid leukemia across a multi-institutional U.S. cohort.

Key Points

This research aims to identify factors influencing early mortality in acute myeloid leukemia patients and propose strategies for improvement.
Conducted a multi-institutional analysis using a federated electronic-health-record network.
Included adults diagnosed with acute myeloid leukemia from 2016 to 2024.
Utilized multilevel logistic regression to assess predictors and calculate adjusted odds ratios.
Evaluated adoption of targeted agents through interrupted time-series modeling.
Performed fairness analysis by demographic subgroups.
Overall 60-day mortality was 19.4% among 7,842 patients.
Independent predictors of mortality included age over 70 years and delayed treatment initiation beyond 5 days.
Adoption of targeted agents increased from 6% in 2018 to 29% in 2024, reducing mortality by 5.1 points.
Scenario modeling indicated that reducing treatment delays to 3 days could lower early mortality rates significantly.

Abstract

Abstract Despite therapeutic advances, early mortality (60 days) in newly diagnosed acute myeloid leukemia (AML) remains 15-25% in routine care, particularly among older and socioeconomically vulnerable patients. Randomized trials under-represent these groups, limiting generalizability. We conducted a multi-institutional, equity-focused analysis to identify modifiable determinants of early mortality and propose operational benchmarks for quality improvement.Adults (≥18 y) with newly diagnosed AML between 2016-2024 were identified from a federated, de-identified electronic-health-record network encompassing 35 U.S. health systems. Primary endpoint: 60-day mortality from diagnosis. Predictors included demographics, comorbidities, insurance type, urban/rural residence, facility type, cytogenetic risk, and receipt/timing of induction chemotherapy or hypomethylating-agent (HMA) regimens. Multilevel logistic regression with site random effects estimated adjusted odds ratios (aORs). Fairness analysis evaluated subgroup calibration by race, sex, and payer. Interrupted time-series modeling assessed 2020-2024 adoption of targeted agents (FLT3, IDH1/2 inhibitors). Only aggregate outputs were analyzed.Among 7,842 patients (median age 66; 47% female; 22% Black or Hispanic; 28% Medicaid/uninsured), overall 60-day mortality was 19.4% (95% CI 18.5-20.3). Independent predictors: age 70 y (aOR 2.43, 2.10-2.82), Medicaid/uninsured (aOR 1.54, 1.30-1.83), diagnosis at non-academic centers (aOR 1.39, 1.21-1.59), ECOG ≥2 (aOR 1.71, 1.50-1.95), and treatment initiation 5 days after diagnosis (aOR 1.46, 1.28-1.67). Adoption of targeted agents increased from 6% (2018) to 29% (2024), corresponding with a 5.1-point mortality reduction (p=0.003). Fairness analysis showed minimal calibration drift (ΔBrier 0.005) across demographic subgroups. Scenario modeling suggested that reducing treatment delays to ≤3 days could lower early mortality by ≈2.7% absolute (14% relative) nationwide.REAL-AML defines pragmatic, measurable drivers of early mortality in AML and identifies actionable levers-timely induction, equitable access to targeted agents, and improved insurance continuity-that could yield substantial survival benefit. This is the largest multi-system, equity-audited AML outcomes study to date and provides a template for real-world precision-implementation trials in hematologic malignancies. Citation Format: Randa Elzein, Safa Elzein, . REAL-AML: Real-world predictors of early mortality and treatment inequities in acute myeloid leukemia across a multi-institutional U.S. cohort abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 5429.

Mark Helpful

Bookmark

Relay