Medication Risks of Blood Cancers during Pregnancy and Drug Toxicity in Medical Oncology

Abstract

This dissertation comprises three studies examining medication risks in two settings: the risk of childhood blood cancers (hematologic malignancies) following prenatal exposures during pregnancy, and the safety of CDK4/6 inhibitors in patients with metastatic breast cancer. Together, these studies address gaps where clinical trial data are limited, leverage large administrative claims databases to generate and evaluate safety signals, and provide real-world evidence to inform clinical practice and future research. Chapter 1 examines whether maternal infection and systemic antibiotic use during pregnancy are associated with the risk of childhood hematologic malignancies in the offspring. Using two large U.S. administrative claims databases with mother-child linkages (2001–2023), a population-based cohort of 4,799,406 singleton livebirths was assembled. The 5-year cumulative incidence of childhood hematologic malignancies was 77 per 100,000 children, consistent with national surveillance estimates, with leukemia accounting for nearly three-quarters of cases. Cumulative incidence was broadly similar regardless of maternal infection or antibiotic exposure, with no consistent pattern by infection type, antibiotic class, or trimester. A gradient boosted machine learning model (XGBoost) was trained to benchmark the predictive relevance of these exposures against established risk factors. Feature importance was dominated by chromosomal abnormality (33%), insurance type (17%), and maternal history of cancer (9%), while all antibiotics combined contributed only 18% of predictive gain, with no individual antibiotic emerging as a strong or consistent predictor. Modest risks observed for metronidazole and ampicillin should be interpreted as hypothesis-generating, potentially reflecting confounding by indication rather than a causal effect. These findings provide a methodological framework for signal-generation research in rare pediatric outcomes. Chapter 2 evaluates major adverse cardiovascular effects of CDK4/6 inhibitors (CDK4/6i) compared with endocrine monotherapy (EMT) in patients with hormone receptor-positive, human epidermal growth factor receptor 2-negative (HR+/HER2−) metastatic breast cancer (mBC). Using two large U.S. commercial claims databases, a propensity score-matched cohort study compared the 1-year incidence of major cardiovascular events between treatment groups. CDK4/6i initiation was not associated with an increased 1-year risk of major adverse cardiovascular events. Heart failure, ischemic stroke, myocardial infarction (MI), and ventricular arrhythmias were all below 1% and broadly comparable between groups. Pericarditis was more frequent among CDK4/6i initiators, particularly with abemaciclib, and abemaciclib was associated with a higher risk of MI (RR: 2.38; 95% CI: 1.09, 5.21), a finding that warrants further investigation. These findings provide reassuring real-world evidence for the short- to intermediate-term cardiovascular safety of CDK4/6i, while highlighting the need for clinical awareness of pericarditis as a potential adverse effect. Chapter 3 addresses the risk of venous thromboembolism (VTE) associated with CDK4/6i versus EMT in the same patient population. After 1:1 propensity score matching of 8,020 CDK4/6i initiators to EMT initiators, the 18-month cumulative incidence of VTE was 3.3% (95% CI: 2.8%, 3.8%) with CDK4/6i versus 1.5% (95% CI: 1.1%, 1.9%) with EMT, corresponding to a risk ratio of 2.13 (95% CI: 1.60, 2.84). Deep vein thrombosis accounted for approximately two-thirds of events (RR: 2.33; 95% CI: 1.67, 3.26), and pulmonary embolism was also more frequent (RR: 1.98; 95% CI: 1.23, 3.20). The median time to VTE onset was approximately 90 days, indicating risk is concentrated early in treatment. Elevated VTE risk was observed across all three agents—palbociclib (RR: 2.07), abemaciclib (RR: 2.04), and ribociclib (RR: 1.34)—though the ribociclib estimate was imprecise due to smaller numbers. These findings support heightened clinical vigilance for VTE during CDK4/6i initiation and underscore the need for prospective evaluation of thromboprophylaxis strategies in this population. Taken together, this dissertation demonstrates that large administrative claims databases, combined with rigorous design and analytical methods, can generate actionable pharmacoepidemiologic evidence across rare and complex clinical settings—from pediatric cancer etiology to oncology drug safety in adults.