Jason F. Goldberg

BACKGROUND: Machine learning (ML) models have been used to evaluate one-year post-transplant mortality in donor-recipient pairs. Previous modeling utilizing non-interpretable ML methods (deep neural networks [DNN] and XGBoost) showed modest gains in area under the receiver operating curve (AUC) beyond logistic regression, but suffered a significant drop in predictive AUC applied to the subsequent years’ data and lacked statistical significance in validating previously identified risk factors. METHODS: Using balanced SRTR datasets, we evaluated non-interpretable models against interpretable ML models (adaptive logistic regression with interaction terms [aLR], Classification and Regression tree [CART], and conditional inference tree [CIT]) for one-year mortality, including comprehensive clinician-supervised data curation and inclusion of variables describing pre- and post-2018 listing status changes. Models were trained/tested using rolling-window validation across years and further analyzed with repeated ten-fold cross-validation. Interaction terms were obtained via Adaptive Best-Subset Selection (ABESS). RESULTS: Predictive validation before the listing policy change in 2018 showed similar AUCs between DNN (0.579), aLR (0.642), CART (0.579), and CIT (0.584), with XGBoost having a higher (0.763) AUC. However, in the post-2018 predictive analysis, aLR outperformed XGBoost (AUC 0.613 vs. 0.586). The interpretable ML models confirm the significance of previously reported risk factors (recipient bilirubin and creatinine) and identify risk factors not previously reported (donor pH, potential recipient distance, and recipient transfusion), as well as clinically relevant interaction terms. CONCLUSIONS: Carefully developed interpretable ML models of one-year transplant mortality have similar predictive performance to black-box models, while identifying novel risk factors, and showing improved performance after recent listing policy changes. With appropriate validation and additional data, interpretable ML modeling may allow real-time data-driven donor selection.

BACKGROUND: Circulating microRNAs are promising biomarkers of acute cellular rejection (ACR) and antibody-mediated rejection (AMR) in heart transplantation. The study objective was to assess the characteristics and diagnostic performance of previously identified microRNAs and clinical rejection scores (CRS) in distinct blood samples obtained at the time of an endomyocardial biopsy (EMB). METHODS: In the 5-center, prospective, longitudinal cohort study, GRAfT (Genomic Research Alliance for Transplantation), microRNA sequencing was performed on blood samples. The previously identified microRNAs associated with ACR (n=12) and AMR (n=17) were used to fit a logistic regression model to the current cohort, and the scores were scaled from 0 to 100. Diagnostic performance of ACR and AMR microRNA panels was assessed by the area under the receiver-operating characteristic curve. An adjusted Cox proportional hazard model evaluated the effect of CRS on long-term outcomes. RESULTS: In 173 heart transplant recipients (29% female sex, 41% Black race, median follow-up 374 days after transplant), 922 blood samples were sequenced, 720 paired with EMB. Among 14 episodes of ACR, the median ACR CRS was 78 compared with 42 without ACR, P P P P =0.003). CONCLUSIONS: Circulating microRNAs reliably identified ACR and AMR on EMB. An elevated microRNA CRS was associated with an increased risk of subsequent rejection, allograft dysfunction, or death. This biomarker, with further validation, can serve as a noninvasive test to screen and diagnose ACR and AMR without the need for an EMB. REGISTRATION: URL: https://www.clinicaltrials.gov ; Unique identifier: NCT02423070.

BACKGROUND: Heart transplant recipients with donor-specific antibodies (DSAs) have an increased risk for antibody-mediated rejection. However, many patients with graft dysfunction and DSA do not have evidence of antibody-mediated rejection by endomyocardial biopsy (EMB). METHODS: Participants from this prospective, multicenter study underwent serial EMB, echocardiogram, DSA, and donor-derived cell-free DNA evaluations. Outcomes were defined as pAMR+ (pAMR≥1) or DSA+/left ventricle (LV) dysfunction (DSA presence+LVEF drop ≥10% to an LVEF≤50%). Cox regression evaluated the association between antibody-mediated rejection categories and death or sustained (for 3 months) reduction of LVEF to P =0.0002), or DSA+/LV dysfunction (0.40% [interquartile range, 0.36–1.24]; P P =0.043); DSA+/LV dysfunction (hazard ratio, 26.2 [95% CI, 9.6–71.3]; P P P =0.031). CONCLUSIONS: Expanding the characterization of antibody-mediated rejection to include patients with DSA and concurrent allograft dysfunction identified DSA+ patients at risk for death and prolonged LV dysfunction.

Background: Childhood cancer survivors have increased risk of dyslipidemia and atherosclerotic cardiovascular disease (CVD). The aim of this study was to evaluate the prevalence and associated cardiovascular risks of specific lipid abnormalities among childhood cancer survivors. Methods: Comprehensive lipid panel measurements were obtained from 4115 5-year survivors, with 3406 (mean age at evaluation = 35.2 years, SD = 10.4 years) not having previous dyslipidemia diagnosis, as well as 624 age, sex, and race and ethnicity matched community controls. Results: Previously undiagnosed dyslipidemia with abnormal low-density lipoprotein (LDL) cholesterol (>160 mg/dL), non–high density lipoprotein (HDL) cholesterol (>190 mg/dL), HDL cholesterol (<40 mg/dL for men, 150 mg/dL) were identified in 4%, 6%, 30%, and 17%, respectively. Survivors without previous dyslipidemia diagnosis had higher LDL cholesterol and non-HDL cholesterol and lower HDL cholesterol than community controls. Cranial radiotherapy (relative risk [RR] = 2.2, 95% confidence interval [CI] = 1.6 to 3.0 for non-HDL cholesterol) and total body irradiation for hematopoietic cell transplantation (RR = 6.7, 95% CI = 3.5 to 13.0 for non-HDL cholesterol; RR = 9.9, 95% CI = 6.0 to 16.3 for triglycerides) were associated with greater risk of dyslipidemia. Diagnoses of low HDL cholesterol (hazard ratio [HR] = 2.9, 95% CI = 1.8 to 4.7) and elevated triglycerides (HR = 3.1, 95% CI = 1.9 to 5.1) were associated with increased risk for myocardial infarction, and diagnoses of high LDL cholesterol (HR = 2.2, 95% CI = 1.3 to 3.7), high non-HDL cholesterol (HR = 2.2, 95% CI = 1.3 to 3.7), low HDL cholesterol (HR = 3.9, 95% CI = 2.8 to 5.4), and elevated triglycerides (HR = 3.8, 95% CI = 2.7 to 5.5) were associated with increased risk for cardiomyopathy. Conclusions: Previously undiagnosed dyslipidemia among childhood cancer survivors was associated with increased risk for myocardial infarction and cardiomyopathy. Comprehensive dyslipidemia evaluation and treatment are needed to reduce cardiovascular morbidity in this population.

Survival for pediatric patients diagnosed with cancer has improved significantly. This achievement has been made possible due to new treatment modalities and the incorporation of a systematic multidisciplinary approach for supportive care. Understanding the distinctive cardiovascular characteristics of children undergoing cancer therapies has set the underpinnings to provide comprehensive care before, during, and after the management of cancer. Nonetheless, we acknowledge the challenge to understand the rapid expansion of oncology disciplines. The limited guidelines in pediatric cardio-oncology have motivated us to develop risk-stratification systems to institute surveillance and therapeutic support for this patient population. Here, we describe a collaborative approach to provide wide-ranging cardiovascular care to children and young adults with oncology diseases. Promoting collaboration in pediatric cardio-oncology medicine will ultimately provide excellent quality of care for future generations of patients.

Background: Limited data exist regarding left ventricular remodeling patterns observed in adult survivors of childhood cancer after therapy.Methods: Among 1190 adult survivors diagnosed with childhood cancer (median age at diagnosis, 9 years [interquartile range (IQR), 3.8‐14.4 years]; age at evaluation, 35.6 years [IQR, 29.5‐42.8 years]), treatment exposures included anthracyclines (n = 346), chest radiotherapy (n = 174), both (n = 245), or neither (n = 425). Prospective echocardiographic assessment compared survivors with 449 noncancer controls classified according to left ventricle geometric patterns. Associations between left ventricle geometric patterns and decreased exercise tolerance were assessed.Results: Overall, 28.2% of survivors (95% confidence interval [CI], 25.6%‐30.8%) exhibited concentric remodeling, 2.4% (95% CI, 1.6%‐3.5%) exhibited eccentric hypertrophy, and 1.1% (95% CI, 0.6%‐1.9%) exhibited concentric hypertrophy. A greater proportion of survivors who received only chest radiotherapy (41%) had concentric remodeling compared with those who received only anthracyclines (24%), both (27%), or neither (27%; all P < .001), and all were greater than the proportions in noncancer controls (18%; all P < .05). Concentric remodeling was associated with radiation exposure, but not with anthracycline exposure, in multivariable models. Survivors who had concentric remodeling were more likely to have a maximal oxygen uptake peak <85% compared with those who had normal geometry (81.0% vs 66.3%; odds ratio, 1.75; 95% CI, 1.15‐2.68).Conclusions: Chest radiation therapy, but not anthracycline therapy, increased the risk for concentric remodeling in survivors of childhood cancer. The presence of concentric remodeling was associated with increased exercise intolerance.

Cancer survivors who have undergone hematopoietic cell transplantation (HCT) are at risk for myocardial dysfunction. Children who receive allogenic HCT encounter systemic inflammation resulting in tachycardia and hypertension. The effect of these abnormalities on myocardial function is not known. The aim of this study was to determine whether cardiac dysfunction early after HCT can be predicted by tachycardia or hypertension, within a retrospective single-center sample of pediatric HCT recipients. Early tachycardia or hypertension was defined as a majority of values taken from infusion date to 90 days post-infusion being abnormal. Ejection fraction p = 0.024) and diastolic dysfunction (OR = 3.9 [1.3–11.5], p = 0.013). Tachycardia and cardiac dysfunction are common and associated with one another in the early period after pediatric HCT. Future studies may elucidate the role of tachycardia and myocardial dysfunction early after HCT as important predictors of future cardiovascular dysfunction.

Elevated tricuspid regurgitant velocity (TRV) ≥2.5 m/s is a predictor of disease severity in adults and children with sickle cell anemia (SCA), but how disease-modifying therapies (DMTs) affect this biomarker is incompletely understood. We investigated the effect of DMTs on TRV elevation in children. In a prospective single-center study, 204 subjects with HbSS or HbSβ0 thalassemia (mean age, 10.6 years; range, 5-18) had echocardiograms with assessment of TRV, with repeat evaluations after 2 years of observation. One-hundred and twelve participants received DMTs (hydroxyurea, n = 72; monthly erythrocyte transfusions, n = 40), 58 did not receive any DMT, and 34 were begun on hydroxyurea during this observation period. In the entire cohort, an increase in hemoglobin of 1.0 g/dL was associated with a 0.03-m/s decrease in TRV (P = .024), and a decrease in absolute reticulocyte count of 1.0 × 106/mL was associated with a 0.34-m/s decrease in TRV (P = .034). Compared with baseline, hydroxyurea exposure (continuous or newly started) was associated with an average 5% decline in mean TRV at the 2-year evaluation. Among participants newly started on hydroxyurea (mean treatment duration 1.2 ± 0.6 years), an increase in hemoglobin of 1.0 g/dL was associated with a 0.06-m/s decrease in TRV (P = .05). We conclude that hydroxyurea therapy may mitigate TRV elevation in children with SCA, possibly as a result of a reduction in hemolysis and improvement in anemia.

Background: Survivors of childhood cancer have an increased risk of therapy-related cardiovascular disease. It is not known whether family history of cardiovascular disease further increases risk of adverse cardiovascular outcomes among survivors. Methods: Family history of cardiovascular disease was collected from 1,260 survivors [median age at diagnosis, 8 years (range, 0–23); age at last follow-up, 35 years (range, 18–66)] of childhood cancer in the St. Jude Lifetime Cohort Study. Multivariable risk models evaluated associations with cardiovascular disease (Common Terminology Criteria for Adverse Events grade 2–4 events) and cardiovascular risk factors. Results: Among survivors exposed to chest-directed radiation and/or anthracycline chemotherapy (n = 824), 7% reported a first-degree family history of heart failure, 19% myocardial infarction, 11% stroke, 26% atherosclerotic disease (myocardial infarction and/or stroke), 62% hypertension, and 31% diabetes mellitus. Eighteen percent of exposed survivors developed heart failure, 9% myocardial infarction, 3% stroke, 11% atherosclerotic disease, 30% hypertension, and 9% diabetes mellitus. Having a first-degree family history of atherosclerotic disease was independently associated with development of treatment-related heart failure [RR, 1.38; 95% confidence interval (CI), 1.01–1.88; P = 0.04] among exposed survivors. Risk for hypertension was increased among exposed survivors with a first-degree family history of hypertension (RR, 1.55; 95% CI, 1.26–1.92; P < 0.0001) or of any cardiovascular disease [myocardial infarction, stroke, or heart failure (RR, 1.30; 95% CI, 1.06–1.59; P = 0.01)]. Conclusions: Family history of cardiovascular disease and cardiovascular risk factors independently increased risk of heart failure and hypertension among survivors of childhood cancer exposed to cardiotoxic therapies. Impact: These data show the importance of cardiovascular family history as a risk factor for cardiovascular disease in survivors of childhood cancer.