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

Paper Details

Abstracts, authors, affiliations, and disclosures.

Sunday, December 13, 2026
06:00 p.m. - 08:00 p.m.
MCCNO - Hall F
Access, Representation, Population Health, Technology and Procedures, Assays, Imaging, Serologic Tests, Emerging Technologies, Profiling, Treatment Considerations, Drug Development, Biological Therapies, Antibody Therapy, Bispecific Antibody Therapy, Checkpoint Inhibitor, Immunotherapy, Adverse Events, Artificial Intelligence (AI), Machine Learning, Biological Processes, CHIP, Genomics, Immune Mechanism, Microbiome, Multi-Systemic Interactions, Diseases, Lymphoid Malignancies, Lymphomas, Plasma Cell Disorders, Myeloid Malignancies, Chronic Myeloid Malignancies, Measurable Residual Disease, Non-Hodgkin Lymphoma, Research, Translational Research, MPN, Clinical Research, Clinical Trials, Health Outcomes Research, Health Disparities Research, Real-World Evidence, Study Population, Human, Adult
PAPER

3966 Race-stratified cardiac and renal recovery in black/African American patients with light chain (AL) amyloidosis

Afua Ofori-Darko, MPH, John Hanna, MD2, Ran Zhao3, Faiz Anwer, MD4, Shahzad Raza, MD5, Jack Khouri, MD6, Christy Samaras, DO7, Beth Faiman, PhD, MSN, APN-BC, AOCN®, TCTCN®, FAAN, FAPO8, Diana Basali, MD9, Hossam M. Ali, MD10, Yara Shatnawi, MD11, Ali Mehdi, MD12, Hernan Rincon Choles, MD13, Mazen Hanna, Co-Director, Amyloidosis Center Section of Heart Failure and Cardiac Transplant14, Jason Valent, MD15, Sandra Mazzoni, DO16, Louis Samuel Williams, MD17
1 Case Western Reserve University School of Medicine, Cleveland, Ohio, United States
2 Cleveland Clinic Taussig Cancer Center, Cleveland, Ohio, United States
3 Cleveland Clinic Taussig Cancer Center, Cleveland, Ohio, United States
4 Cleveland Clinic Taussig Cancer Center, Cleveland, Ohio, United States
5 Cleveland Clinic Taussig Cancer Center, Cleveland, Ohio, United States
6 Cleveland Clinic Taussig Cancer Center, Cleveland, Ohio, United States
7 Cleveland Clinic Taussig Cancer Center, Cleveland, Ohio, United States
8 Cleveland Clinic Taussig Cancer Center, Cleveland, Ohio, United States
9 Cleveland Clinic Taussig Cancer Center, Cleveland, Ohio, United States
10 Cleveland Clinic Taussig Cancer Center, Cleveland, Ohio, United States
11 Cleveland Clinic Taussig Cancer Center, Cleveland, Ohio, United States
12 Cleveland Clinic Taussig Cancer Center, Cleveland, Ohio, United States
13 Cleveland Clinic Taussig Cancer Center, Cleveland, Ohio, United States
14 Cleveland Clinic Taussig Cancer Center, Cleveland, Ohio, United States
15 Cleveland Clinic Taussig Cancer Center, Cleveland, Ohio, United States
16 Cleveland Clinic Taussig Cancer Center, Cleveland, Ohio, United States
17 Cleveland Clinic Taussig Cancer Center, Cleveland, Ohio, United States

Background: Prediction of organ responses in AL amyloidosis remains a clinical challenge. Traditional logistic regression (LR) models are limited in capturing complex, non-linear interactions. This study evaluates the performance of extreme gradient boosting (XGBoost), an ensemble machine learning approach using successively refined decision trees for response classification, compared to LR in predicting cardiac and renal responses over time.

Methods: A retrospective chart review was conducted from January 1, 2017, to January 1, 2023, to identify 242 adult patients (pts) with newly diagnosed AL amyloidosis at a single multi-hospital tertiary care center. Median age at diagnosis was 67 years; 189 pts were White, 46 were Black/African American, and 7 were of other races. Cardiac involvement was present in 60% of pts and renal involvement in 50%. Initial therapy included CyBorD in 57% and daratumumab-CyBorD in 23%.

Hematologic and organ responses were assessed at 6, 12, 18, and 24 months according to International Society of Amyloidosis criteria. Predictive models were developed using LR and XGBoost, with performance evaluated by accuracy and area under the curve (AUC). Hyperparameters for XGBoost were optimized using grid search. For the purposes of our analysis, the initial dataset was split into training (80%) and testing (20%) subsets. SHAP (SHapley Additive exPlanations) values were used to interpret feature importance. Race-stratified comparisons focused on White and Black/African American pts because of small sample sizes in other racial groups.

Results: Black/African American pts had numerically higher baseline cardiac involvement than White pts (70% vs. 58%; p=0.18), whereas renal involvement was similar (48% vs. 50%; p=0.87); median involved free light-chain levels and baseline dFLC were also higher. Cardiac discordance declined similarly in Black/African American versus White pts: 35.0% vs. 32.6% at 6 months (p=1.00), 28.6% vs. 24.0% at 12 months (p=0.78), 23.5% vs. 23.3% at 18 months (p=1.00), and 17.6% vs. 18.9% at 24 months (p=1.00).

Renal discordance was consistently higher among Black/African American pts: 37.5% vs. 25.3% at 6 months (p=0.36), 47.1% vs. 27.2% at 12 months (p=0.15), 46.2% vs. 25.0% at 18 months (p=0.18), and 30.8% vs. 25.4% at 24 months (p=0.74). XGBoost showed higher discrimination than LR in the prediction of organ recovery for renal and cardiac recovery, most prominently for 6-month renal recovery (AUC 0.693 vs. 0.646). Shapley additive explanations (SHAP) analysis identified baseline 24-hour urine protein as the most influential predictor, followed by NT-proBNP, lower-extremity edema, hematologic response, and baseline dFLC.

Conclusions: Black/African American pts demonstrated numerically higher baseline cardiac involvement and a persistent renal discordance signal following treatment. Cardiac discordance declined over time in both racial groups, whereas renal discordance remained comparatively elevated among Black/African American pts through 18 months; however, none of the reported cardiac or renal involvement and discordance comparisons reached statistical significance. XGBoost demonstrated performance comparable to or better than LR across multiple organ-recovery endpoints. Analysis of larger data sets incorporating organ-specific biomarkers, genomic data, and interpretable machine learning are warranted to confirm these associations and assess their generalizability.

Afua Ofori-Darko, MPH: No relevant conflicts of interest to declare.