2026 ASH®
Annual Meeting & Exposition December 12-15, 2026 ▪ New Orleans, Louisiana

Paper Details

Abstracts, authors, affiliations, and disclosures.

Monday, December 14, 2026
06:00 p.m. - 08:00 p.m.
MCCNO - Hall F
Access, Representation, Population Health, Technology and Procedures, Serologic Tests, Treatment Considerations, Supportive Care, Non-Biological Therapies, Anticoagulant Drugs, Coagulant Drugs, Women’s Health, Adverse Events, Artificial Intelligence (AI), Deep Learning, Machine Learning, Natural Language Processing, Clinical Practice (Health Services and Quality), Clinical Guidelines, Health Economics, Patient Safety, Practice Models, Quality Improvement, Quality of Care, Systems-Based Hematology, Telemedicine, Value, Diseases, Bleeding and Clotting, Bleeding Disorders, Hemophilia, Thromboembolism, Thrombocytopenias, Thrombotic Disorders, Bone Marrow Failure Syndromes, Acquired Marrow Failure Syndromes , Inherited Marrow Failure Syndromes, Aplastic Anemia, Paroxysmal Nocturnal Hemoglobinuria, Genetic Disorders, Immunology, Autoimmune Disorders, Metabolic Disorders, Iron Deficiency, Research, Clinical Research, Health Outcomes Research, Real-World Evidence, Study Population, Human, Adult, Pediatric
PAPER

6144 Blood cell signatures of platelet reactivity predict cardiovascular, thrombotic, neurodegenerative, and mortality risk

Tianshe He, PhD, Elliot Luttrell-Williams, MSc2, Matthew Muller, MS3, Omonigho M. Bubu, MD, MPH, PhD, CPH4, Ricardo S Osorio, MD5, Nhan Do, MD6, Ang Li, MD, MS7, Nathanael Fillmore, PhD8, Jeffrey Berger, MD9, Jaime Ramos-Cejudo, PhD10
1 NYU Grossman School of Medicine, New York, New York, United States
2 NYU Grossman School of Medicine, New York, New York, United States
3 NYU Grossman School of Medicine, New York, New York, United States
4 NYU Grossman School of Medicine, New York, New York, United States
5 NYU Grossman School of Medicine, New York, New York, United States
6 VA Boston Healthcare System, Boston, Massachusetts, United States
7 Baylor College of Medicine, Houston, Texas, United States
8 VA Boston Healthcare System, Boston, Massachusetts, United States
9 NYU Grossman School of Medicine, New York, New York, United States
10 NYU Grossman School of Medicine, New York, New York, United States

Background. Platelets play a central role in atherothrombosis, and increased platelet activity has been associated with mortality and cardiovascular (CV) events. However, large-scale study of platelet function is limited by the technical demands and cost of light transmission aggregometry (LTA). A composite Platelet Aggregation Score (PLAS) was recently developed from LTA measurements. Because platelet function is closely linked to circulating leukocyte and erythroid cell phenotypes, we hypothesized that routinely collected complete blood counts (CBC) can be used to estimate platelet hyperreactivity and identify downstream clinical risk in large-scale medical record data.

Methods. PLAS derivation was performed across four internal cohorts of healthy controls (n=31-134 per cohort). Platelet-rich plasma was isolated and stimulated with platelet agonists, and reactivity was quantified by LTA. PLAS was defined as a composite of submaximal aggregation responses to epinephrine, collagen, adenosine diphosphate, and arachidonic acid.

We then derived WASP (Weighted Approximating Sum of PLAS) as a CBC-based proxy for PLAS. Using partial least squares regression, we related the 12 CBC parameters to PLAS. We then applied 2-view kernel clustering to group the CBC variables based on how they contributed to this relationship, weighing each variable by its strength within its cluster. This approach produced interpretable, directionally weighted CBC clusters that approximate PLAS.

Models were trained and validated sequentially across derivation sub-cohorts, and CBC and PLAS measures were harmonized across cohorts using ComBat.

For downstream validation, we analyzed two large cohorts from the US Veterans Affairs (VA) and NYU Langone health systems. WASP was calculated using the first available CBC obtained between 2011 and 2020, outcomes were assessed through 2024.

The 2-component WASP was evaluated for major adverse cardiovascular events (MACE; myocardial infarction (MI), stroke, or CV death), venous thromboembolism (VTE), Alzheimer's disease and related dementias (AD/ADRD), and all-cause mortality. Analyses were restricted to individuals age ≥50 (55 for AD/ADRD) with a 30- (90- for AD/ADRD) day landmark after CBC collection. Hazard ratios for WASP components were estimated using Cox models adjusted for age, sex, race, hypertension, diabetes, obesity, and prior MI. Generalizability was further assessed by training mortality prediction models in VA and externally validating in NYU; comparator models included covariates alone and all 12 CBC variables plus covariates.

Results. Among candidate clusters, K=2 provided the best internal validation in the harmonized 4-cohort dataset (n=263). Sub-cohort PLAS correlations ranged from r=.38 to .42, and leave-one-out validation yielded r=.34. The 2 clusters comprised:

WM (myeloid): RBC, HGB, HCT, MCH, MCHC (weights: .15, .19, .24, .20, .23 respectively) WE (erythroid-platelet): monocytes, platelets, RDW-SD, MPV, neutrophils (weights: .16, .25, .18, .20, .18).

In downstream models (VA n=44,931; NYU n=121,313), hazards (h) for WM and WE were stronger across all outcomes than for established markers such as the systemic inflammation index.

For MACE:  h(WM) = .94 [95% CI: .89−.98] and h(WE) = 1.41 [1.33–1.50] in VA; h(WM) = .90 [.86−.93] and h(WE) = 1.27 [1.21–1.34] in NYU.  

For VTE: h(WM) = .83 [.70−.99] and h(WE) = 1.50 [1.21–1.86] in VA; h(WM) = .79 [.73−.86] and h(WE) = 1.33 [1.19–1.49] in NYU.   

For AD/ADRD: h(WM) = .85 [.78−.92] and h(WE) = 1.16 [1.04–1.29] in VA; h(WM) was not significant and h(WE) = 1.21 [1.19–1.49] in NYU.

For Death: h(WM) = .75 [.73−.78] and h(WE) = 1.62 [1.57−1.68] in VA; h(WM)= .66 [.64−.68], h(WE) =1.54 [1.48−1.59] in NYU.   

For external validation of mortality prediction, a VA-trained logistic model incorporating WM,W E, and covariates yielded AUCs of .72, .73, and .74 for 6-, 12-, and 18-month mortality in NYU, respectively, improving upon the covariate-only model (.68, .69, .71) and the 12-measure CBC plus covariates model (.71, .71, .73).

Conclusion. A 2-component CBC-derived signature associated with platelet aggregation was reproducibly linked to cardiovascular, thrombotic, neurodegenerative, and mortality outcomes across two large health systems. WASP generalized across settings and outperformed a full CBC model in mortality prediction. Routine hemogram data may provide scalable information on platelet-related blood cell biology relevant to clinical risk stratification.

Tianshe He, PhD: No relevant conflicts of interest to declare.