Unlocking the Potential of Artificial Intelligence in Acute Myeloid Leukemia and Myelodysplastic Syndromes.

PURPOSE OF THE REVIEW: This review aims to elucidate the transformative impact and potential of machine learning (ML) in the diagnosis, prognosis, and clinical management of myelodysplastic syndromes (MDS) and acute myeloid leukemia (AML). It further aims to bridge the gap between current advances of ML and their practical application in these diseases. RECENT FINDINGS: Recent advances in ML have revolutionized prognostication, diagnosis, and treatment of MDS and AML. ML algorithms have proven effective in predicting disease progression, optimizing treatment responses, and in the stratification of patient groups. Particularly, the use of ML in genomic and epigenomic data analysis has unveiled novel insights into the molecular heterogeneity of MDS and AML, leading to better-informed therapeutic strategies. Furthermore, deep learning techniques have shown promise in analyzing complex patterns in bone marrow biopsy images, providing a potential pathway towards early and accurate diagnosis. While still in the nascent stages, ML applications in MDS and AML signify a paradigm shift towards precision medicine. The integration of ML with traditional clinical practices could potentially enhance diagnostic accuracy, refine risk stratification, and improve therapeutic approaches. However, challenges related to data privacy, standardization, and algorithm interpretability must be addressed to realize the full potential of ML in this field. Future research should focus on the development of robust, transparent ML models and their ethical implementation in clinical settings.

Classification, risk stratification and response assessment in myelodysplastic syndromes/neoplasms (MDS): A state-of-the-art report on behalf of the International Consortium for MDS (icMDS).

The guidelines for classification, prognostication, and response assessment of myelodysplastic syndromes/neoplasms (MDS) have all recently been updated. In this report on behalf of the International Consortium for MDS (icMDS) we summarize these developments. We first critically examine the updated World Health Organization (WHO) classification and the International Consensus Classification (ICC) of MDS. We then compare traditional and molecularly based risk MDS risk assessment tools. Lastly, we discuss limitations of criteria in measuring therapeutic benefit and highlight how the International Working Group (IWG) 2018 and 2023 response criteria addressed these deficiencies and are endorsed by the icMDS. We also address the importance of patient centered care by discussing the value of quality-of-life assessment. We hope that the reader of this review will have a better understanding of how to classify MDS, predict clinical outcomes and evaluate therapeutic outcomes.

Thirty-Day Unplanned Hospital Readmissions in Patients With Cancer and the Impact of Social Determinants of Health: A Machine Learning Approach.

PURPOSE: Develop a cancer-specific machine learning (ML) model that accurately predicts 30-day unplanned readmissions in patients with solid tumors. METHODS: The initial cohort included patients 18 years or older diagnosed with a solid tumor. Two distinct cohorts were generated: one with and one without detailed social determinants of health (SDOHs) data. For each cohort, data were temporally partitioned in 70% (training), 20% (validation), and 10% (testing). Tree-based ML models were developed and validated on each cohort. The metrics used to evaluate the model's performance were receiver operating characteristic curve (ROC), area under the ROC curve, precision, recall (R), accuracy, and area under the precision-recall curve. RESULTS: We included 13,717 patients in this study in two cohorts (5,059 without SDOH data and 8,658 with SDOH data). Unplanned 30-day readmission occurred in 21.3% of the cases overall. The five main non-SDOH factors most highly associated with an unplanned 30-day readmission (R, 0.74; IQR, 0.58-0.76) were: number of previous unplanned readmissions; higher Charlson comorbidity score; nonelective index admission; discharge to anywhere other than home, hospice, or nursing facility; and higher anion gap during the admission. Neighborhood crime index, neighborhood median home values, annual income, neighborhood median household income, and wealth index were the main five SDOH factors important for predicting a high risk for an unplanned hospital readmission (R, 0.66; IQR, 0.56-0.72). The models were not directly comparable. CONCLUSION: Key drivers of unplanned readmissions in patients with cancer are complex and involve both clinical factors and SDOH. We developed a cancer-specific ML model that with reasonable accuracy identified patients with cancer at high risk for an unplanned hospital readmission.

Whole-genome sequencing identifies novel predictors for hematopoietic cell transplant outcomes for patients with myelodysplastic syndrome: a CIBMTR study.

Recurrent mutations in TP53, RAS pathway and JAK2 genes were shown to be highly prognostic of allogeneic hematopoietic cell transplant (alloHCT) outcomes in myelodysplastic syndromes (MDS). However, a significant proportion of MDS patients has no such mutations. Whole-genome sequencing (WGS) empowers the discovery of novel prognostic genetic alterations. We conducted WGS on pre-alloHCT whole-blood samples from 494 MDS patients. To nominate genomic candidates and subgroups that are associated with overall survival, we ran genome-wide association tests via gene-based, sliding window and cluster-based multivariate proportional hazard models. We used a random survival forest (RSF) model with build-in cross-validation to develop a prognostic model from identified genomic candidates and subgroups, patient-, disease- and HCT-related clinical factors. Twelve novel regions and three molecular signatures were identified with significant associations to overall survival. Mutations in two novel genes, CHD1 and DDX11, demonstrated a negative impact on survival in AML/MDS and lymphoid cancer data from the Cancer Genome Atlas (TCGA). From unsupervised clustering of recurrent genomic alterations, genomic subgroup with TP53/del5q is characterized with the significant association to inferior overall survival and replicated by an independent dataset. From supervised clustering of all genomic variants, more molecular signatures related to myeloid malignancies are characterized from supervised clustering, including Fc-receptor FCGRs, catenin complex CDHs and B-cell receptor regulators MTUS2/RFTN1. The RSF model with genomic candidates and subgroups, and clinical variables achieved superior performance compared to models that included only clinical variables.

Prognostic landscape of mitochondrial genome in myelodysplastic syndrome after stem-cell transplantation.

Despite mitochondrial DNA (mtDNA) mutations are common events in cancer, their global frequency and clinical impact have not been comprehensively characterized in patients with myelodysplastic neoplasia (also known as myelodysplastic syndromes, MDS). Here we performed whole-genome sequencing (WGS) on samples obtained before allogenic hematopoietic cell transplantation (allo-HCT) from 494 patients with MDS who were enrolled in the Center for International Blood and Marrow Transplant Research. We evaluated the impact of mtDNA mutations on transplantation outcomes, including overall survival (OS), relapse, relapse-free survival (RFS), and transplant-related mortality (TRM). A random survival forest algorithm was applied to evaluate the prognostic performance of models that include mtDNA mutations alone and combined with MDS- and HCT-related clinical factors. A total of 2666 mtDNA mutations were identified, including 411 potential pathogenic variants. We found that overall, an increased number of mtDNA mutations was associated with inferior transplantation outcomes. Mutations in several frequently mutated mtDNA genes (e.g., MT-CYB and MT-ND5) were identified as independent predictors of OS, RFS, relapse and/or TRM after allo-HCT. Integration of mtDNA mutations into the models based on the Revised International Prognostic Scores (IPSS-R) and clinical factors related to MDS and allo-HCT could capture more prognostic information and significantly improve the prognostic stratification efforts. Our study represents the first WGS effort in MDS receiving allo-HCT and shows that there may be clinical utility of mtDNA variants to predict allo-HCT outcomes in combination with more standard clinical parameters.

Patient-specific comorbidities as prognostic variables for survival in myelofibrosis.

Treatment decisions in primary myelofibrosis (PMF) are guided by numerous prognostic systems. Patient-specific comorbidities have influence on treatment-related survival and are considered in clinical contexts but have not been routinely incorporated into current prognostic models. We hypothesized that patient-specific comorbidities would inform prognosis and could be incorporated into a quantitative score. All patients with PMF or secondary myelofibrosis with available DNA and comprehensive electronic health record (EHR) data treated at Vanderbilt University Medical Center between 1995 and 2016 were identified within Vanderbilt's Synthetic Derivative and BioVU Biobank. We recapitulated established PMF risk scores (eg, Dynamic International Prognostic Scoring System [DIPSS], DIPSS plus, Genetics-Based Prognostic Scoring System, Mutation-Enhanced International Prognostic Scoring System 70+) and comorbidities through EHR chart extraction and next-generation sequencing on biobanked peripheral blood DNA. The impact of comorbidities was assessed via DIPSS-adjusted overall survival using Bonferroni correction. Comorbidities associated with inferior survival include renal failure/dysfunction (hazard ratio [HR], 4.3; 95% confidence interval [95% CI], 2.1-8.9; P = .0001), intracranial hemorrhage (HR, 28.7; 95% CI, 7.0-116.8; P = 2.83e-06), invasive fungal infection (HR, 41.2; 95% CI, 7.2-235.2; P = 2.90e-05), and chronic encephalopathy (HR, 15.1; 95% CI, 3.8-59.4; P = .0001). The extended DIPSS model including all 4 significant comorbidities showed a significantly higher discriminating power (C-index 0.81; 95% CI, 0.78-0.84) than the original DIPSS model (C-index 0.73; 95% CI, 0.70-0.77). In summary, we repurposed an institutional biobank to identify and risk-classify an uncommon hematologic malignancy by established (eg, DIPSS) and other clinical and pathologic factors (eg, comorbidities) in an unbiased fashion. The inclusion of comorbidities into risk evaluation may augment prognostic capability of future genetics-based scoring systems.

Personalized model to predict seizures based on dynamic and static continuous EEG monitoring data.

BACKGROUND/OBJECTIVE: Early recognition of patients who may be at risk of developing acute symptomatic seizures would be useful. We aimed to determine whether continuous electroencephalography (cEEG) data using machine learning techniques such as neural networks and decision trees could predict seizure occurrence in hospitalized patients. METHODS: This was a single center retrospective cohort analysis of cEEG data in patients aged 18-90 years who were admitted and underwent cEEG monitoring between 2010 and 2019 limited to 72 h excluding those who were seizing at the onset of recording. A total of 41,491 patients were reviewed; of these, 3874 were used to develop the static model and 1687 to develop the dynamic model (half with seizure and half without seizure in each cohort). Of these, 80% were randomly selected as derivation cohorts for each model and 20% were randomly selected as validation cohorts. Dynamic and static machine learning models (long short term memory (LSTM) and Extreme Gradient Boosting algorithm (XGBoost)) based on day-to-day dynamic EEG changes and binary static EEG features over the prior 72 h or until seizure, which ever was earlier, were used. RESULTS: The static model was able to predict seizure occurrence based on cEEG data with sensitivity and specificity of 0.81 and 0.59, respectively, with an AUC of 0.70. The dynamic model was able to predict seizure occurrence with sensitivity and specificity of 0.72 and 0.80, respectively, and AUC of 0.81. CONCLUSIONS: Machine learning models could be applied to cEEG data to predict seizure occurrence based on available cEEG data. Dynamic day-to-day EEG data are more useful in predicting seizures than binary static EEG data. These models could potentially be used to determine the need for ongoing cEEG monitoring and to prioritize resources.

A machine learning model of response to hypomethylating agents in myelodysplastic syndromes.

Hypomethylating agents (HMA) prolong survival and improve cytopenias in individuals with higher-risk myelodysplastic syndrome (MDS). Only 30-40% of patients, however, respond to HMAs, and responses may not occur for more than 6 months after HMA initiation. We developed a model to more rapidly assess HMA response by analyzing early changes in patients' blood counts. Three institutions' data were used to develop a model that assessed patients' response to therapy 90 days after the initiation using serial blood counts. The model was developed with a training cohort of 424 patients from 2 institutions and validated on an independent cohort of 90 patients. The final model achieved an area under the receiver operating characteristic curve (AUROC) of 0.79 in the train/test group and 0.84 in the validation group. The model provides cohort-wide and individual-level explanations for model predictions, and model certainty can be interrogated to gauge the reliability of a given prediction.

The ABNL-MARRO 001 study: a phase 1-2 study of randomly allocated active myeloid target compound combinations in MDS/MPN overlap syndromes.

BACKGROUND: Myelodysplastic/myeloproliferative neoplasms (MDS/MPN) comprise several rare hematologic malignancies with shared concomitant dysplastic and proliferative clinicopathologic features of bone marrow failure and propensity of acute leukemic transformation, and have significant impact on patient quality of life. The only approved disease-modifying therapies for any of the MDS/MPN are DNA methyltransferase inhibitors (DNMTi) for patients with dysplastic CMML, and still, outcomes are generally poor, making this an important area of unmet clinical need. Due to both the rarity and the heterogeneous nature of MDS/MPN, they have been challenging to study in dedicated prospective studies. Thus, refining first-line treatment strategies has been difficult, and optimal salvage treatments following DNMTi failure have also not been rigorously studied. ABNL-MARRO (A Basket study of Novel therapy for untreated MDS/MPN and Relapsed/Refractory Overlap Syndromes) is an international cooperation that leverages the expertise of the MDS/MPN International Working Group (IWG) and provides the framework for collaborative studies to advance treatment of MDS/MPN and to explore clinical and pathologic markers of disease severity, prognosis, and treatment response. METHODS: ABNL MARRO 001 (AM-001) is an open label, randomly allocated phase 1/2 study that will test novel treatment combinations in MDS/MPNs, beginning with the novel targeted agent itacitinib, a selective JAK1 inhibitor, combined with ASTX727, a fixed dose oral combination of the DNMTi decitabine and the cytidine deaminase inhibitor cedazuridine to improve decitabine bioavailability. DISCUSSION: Beyond the primary objectives of the study to evaluate the safety and efficacy of novel treatment combinations in MDS/MPN, the study will (i) Establish the ABNL MARRO infrastructure for future prospective studies, (ii) Forge innovative scientific research that will improve our understanding of pathogenetic mechanisms of disease, and (iii) Inform the clinical application of diagnostic criteria, risk stratification and prognostication tools, as well as response assessments in this heterogeneous patient population. TRIAL REGISTRATION: This trial was registered with ClinicalTrials.gov on August 19, 2019 (Registration No. NCT04061421).

Role of allogeneic transplantation in chronic myelomonocytic leukemia: an international collaborative analysis.

To determine the survival benefit of allogeneic hematopoietic cell transplantation (allo-HCT) in chronic myelomonocytic leukemias (CMML), we assembled a retrospective cohort of CMML patients 18-70 years old diagnosed between 2000 and 2014 from an international CMML dataset (n = 730) and the EBMT registry (n = 384). The prognostic impact of allo-HCT was analyzed through univariable and multivariable time-dependent models and with a multistate model, accounting for age, sex, CMML prognostic scoring system (low or intermediate-1 grouped as lower-risk, intermediate-2 or high as higher-risk) at diagnosis, and AML transformation. In univariable analysis, lower-risk CMMLs had a 5-year overall survival (OS) of 20% with allo-HCT vs 42% without allo-HCT (P < .001). In higher-risk patients, 5-year OS was 27% with allo-HCT vs 15% without allo-HCT (P = .13). With multistate models, performing allo-HCT before AML transformation reduced OS in patients with lower-risk CMML, and a survival benefit was predicted for men with higher-risk CMML. In a multivariable analysis of lower-risk patients, performing allo-HCT before transformation to AML significantly increased the risk of death within 2 years of transplantation (hazard ratio [HR], 3.19; P < .001), with no significant change in long-term survival beyond this time point (HR, 0.98; P = .92). In higher-risk patients, allo-HCT significantly increased the risk of death in the first 2 years after transplant (HR 1.46; P = .01) but not beyond (HR, 0.60; P = .09). Performing allo-HCT before AML transformation decreases life expectancy in lower-risk patients but may be considered in higher-risk patients.

Why Machine Learning Should Be Taught in Medical Schools.

The rapid development of machine learning (ML) applications in healthcare promises to transform the landscape of healthcare. In order for ML advancements to be effectively utilized in clinical care, it is necessary for the medical workforce to be prepared to handle these changes. As physicians in training are exposed to a wide breadth of clinical tools during medical school, this offers an ideal opportunity to introduce ML concepts. A foundational understanding of ML will not only be practically useful for clinicians, but will also address ethical concerns for clinical decision making. While select medical schools have made effort to integrate ML didactics and practice into their curriculum, we argue that foundational ML principles should be taught broadly to medical students across the country.

Relapse and Disease-Free Survival in Patients With Myelodysplastic Syndrome Undergoing Allogeneic Hematopoietic Cell Transplantation Using Older Matched Sibling Donors vs Younger Matched Unrelated Donors.

IMPORTANCE: Matched sibling donors (MSDs) are preferred for allogeneic hematopoietic cell transplantation (allo-HCT) in myelodysplastic syndrome even if they are older. However, whether older MSDs or younger human leukocyte antigen-matched unrelated donors (MUDs) are associated with better outcomes remains unclear. OBJECTIVE: To investigate whether allo-HCT for myelodysplastic syndrome using younger MUDs would be associated with improved disease-free survival and less relapse compared with older MSDs. DESIGN, SETTING, AND PARTICIPANTS: This retrospective cohort study assessed data reported to the Center for International Blood and Marrow Transplant Research database from 1761 adults 50 years or older with myelodysplastic syndrome who underwent allo-HCT using an older MSD or younger MUD between January 1, 2011, and December 31, 2017, with a median follow-up of 48 months. Data analysis was performed from January 8, 2019, to December 30, 2020. INTERVENTIONS/EXPOSURES: Allo-HCT from an older MSD (donor age ≥50 years) or a younger MUD (donor age ≤35 years). MAIN OUTCOMES AND MEASURES: The primary outcome was disease-free survival. Secondary outcomes were overall survival, relapse, nonrelapse mortality, acute graft-vs-host disease (GVHD), chronic GVHD, and GVHD-free relapse-free survival. RESULTS: Of 1761 patients (1162 [66%] male; median [range] age, 64.9 [50.2-77.6] years in the MSD cohort and 66.5 [50.4-80.9] years in MUD cohort), 646 underwent allo-HCT with an older MSD and 1115 with a younger MUD. In multivariable analysis, the rate of disease-free survival was significantly lower in allo-HCTs with older MSDs compared with younger MUDs (hazard ratio [HR], 1.17; 95% CI, 1.02-1.34; P = .02), whereas the difference in overall survival rate of allo-HCT with younger MUDs vs older MSDs was not statistically significant (HR, 1.13; 95% CI, 0.98-1.29; P = .07). Allo-HCT with older MSDs was associated with significantly higher relapse (HR, 1.62; 95% CI, 1.32-1.97; P < .001), lower nonrelapse mortality (HR, 0.76; 95% CI, 0.59-0.96; P = .02), lower acute GVHD (HR, 0.52; 95% CI, 0.42-0.65; P < .001), chronic GVHD (HR, 0.77; 95% CI, 0.64-0.92; P = .005), and a lower rate of GVHD-free relapse-free survival beyond 12 months after allo-HCT (HR, 1.42; 95% CI, 1.02-1.98; P = .04). CONCLUSIONS AND RELEVANCE: This cohort study found higher disease-free survival and lower relapse for allo-HCT in myelodysplastic syndrome using younger MUDs compared with older MSDs. The risk of nonrelapse mortality and GVHD was lower with older MSDs. These results suggest that the use of younger MUDs should be considered in the donor selection algorithm for myelodysplastic syndrome, in which it is pivotal to minimize relapse given limited treatment options for managing relapsed disease.

Outcomes of Allogeneic Hematopoietic Cell Transplantation in T Cell Prolymphocytic Leukemia: A Contemporary Analysis from the Center for International Blood and Marrow Transplant Research.

T cell prolymphocytic leukemia (T-PLL) is a rare, aggressive malignancy with limited treatment options and poor long-term survival. Previous studies of allogeneic hematopoietic cell transplantation (alloHCT) for T-PLL are limited by small numbers, and descriptions of patient and transplantation characteristics and outcomes after alloHCT are sparse. In this study, we evaluated outcomes of alloHCT in patients with T-PLL and attempted to identify predictors of post-transplantation relapse and survival. We conducted an analysis of data using the Center for International Blood and Marrow Transplant Research database on 266 patients with T-PLL who underwent alloHCT between 2008 and 2018. The 4-year rates of overall survival (OS), disease-free survival (DFS), relapse, and treatment-related mortality (TRM) were 30.0% (95% confidence interval [CI], 23.8% to 36.5%), 25.7% (95% CI, 20% to 32%), 41.9% (95% CI, 35.5% to 48.4%), and 32.4% (95% CI, 26.4% to 38.6%), respectively. In multivariable analyses, 3 variables were associated with inferior OS: receipt of a myeloablative conditioning (MAC) regimen (hazard ratio [HR], 2.18; P < .0001), age >60 years (HR, 1.61; P = .0053), and suboptimal performance status, defined by Karnofsky Performance Status (KPS) <90 (HR, 1.53; P = .0073). Receipt of an MAC regimen also was associated with increased TRM (HR, 3.31; P < .0001), an elevated cumulative incidence of grade II-IV acute graft-versus-host disease (HR, 2.94; P = .0011), and inferior DFS (HR, 1.86; P = .0004). Conditioning intensity was not associated with relapse; however, stable disease/progression was correlated with increased risk of relapse (HR, 2.13; P = .0072). Both in vivo T cell depletion (TCD) as part of conditioning and KPS <90 were associated with worse TRM and inferior DFS. Receipt of total body irradiation had no significant effect on OS, DFS, or TRM. Our data show that reduced-intensity conditioning without in vivo TCD (ie, without antithymocyte globulin or alemtuzumab) before alloHCT was associated with long-term DFS in patients with T-PLL who were age ≤60 years or who had a KPS >90 or chemosensitive disease.

Simplified Pediatric Index of Mortality 3 Score by Explainable Machine Learning Algorithm.

Pediatric Index of Mortality 3 is a validated tool including 11 variables for the assessment of mortality risk in PICU patients. With the recent advances in explainable machine learning algorithms, we aimed to assess feasibility of application of these machine learning models to simplify the Pediatric Index of Mortality 3 scoring system in order to decrease time and labor required for data collection and entry for Pediatric Index of Mortality 3. DESIGN: Single-center, retrospective cohort study. Data from the Virtual Pediatric Systems for patients admitted to Cleveland Clinic Children`s PICU between January 2008 and December 2019 was obtained. Light Gradient Boosting Machine Regressor (a gradient boosting decision tree algorithm) was used for building the machine learning models. Variable importance was analyzed by SHapley Additive exPlanations. All of the 11 Pediatric Index of Mortality 3 variables were used as input variables in the machine learning models to predict Pediatric Index of Mortality 3 risk of mortality as the outcome variable. Mean absolute error, root mean squared error, and R-squared were calculated for each of the 11 machine learning models as model performance parameters. SETTING: Quaternary children's hospital. PATIENTS: PICU patients. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Five-thousand sixty-eight patients were analyzed. The machine learning models were able to maintain similar predictive error until the number of input variables decreased to four. The machine learning model with five input variables (mechanical ventilation in the first hour of PICU admission, very-high-risk diagnosis, surgical recovery from a noncardiac procedure, low-risk diagnosis, and base excess) produced lowest mean root mean squared error of 1.49 (95% CI, 1.05-1.93) and highest R-squared of 0.73 (95% CI, 0.6-0.86) with mean absolute error of 0.43 (95% CI, 0.35-0.5) among all the 11 machine learning models. CONCLUSIONS: Explainable machine learning methods were feasible in simplifying the Pediatric Index of Mortality 3 scoring system with similar risk of mortality predictions compared to the original Pediatric Index of Mortality 3 model tested in a single-center dataset.