Divergent neuropsychiatric and systemic toxicity profiles of abemaciclib and palbociclib: a triangulation study integrating pharmacovigilance, genetic epidemiology, and multi-omics profiling.

Cyclin-dependent kinase 4/6 inhibitors improve outcomes in hormone receptor-positive, human epidermal growth factor receptor 2-negative advanced breast cancer, but their toxicity profiles may differ in clinically meaningful ways. We aimed to compare the neuropsychiatric and systemic toxicity patterns of abemaciclib and palbociclib and to explore pharmacokinetic and molecular features that might contribute to these differences. We conducted a triangulation study integrating pharmacovigilance, Mendelian randomization, explainable machine learning, transcriptomics, and molecular docking. We analyzed 15,215 reports from the US Food and Drug Administration Adverse Event Reporting System and retained 5524 matched patients after propensity score matching. In the matched cohort, abemaciclib showed a toxicity profile more consistent with central nervous system involvement, with reporting enrichment for headache, dizziness, and memory impairment, and a shorter median time to onset than palbociclib (27.5 days vs 37.0 days). By contrast, palbociclib showed a greater reporting burden of fatigue and anxiety and a higher fatal outcome reporting rate than abemaciclib (12.46% vs 6.52%). Mendelian randomisation showed an association between genetically predicted lower CDK6 levels and reduced cognitive performance (odds ratio 0.991, 95% CI 0.983-1.000; p=0.040), but no association with C-reactive protein (odds ratio 0.999, 95% CI 0.987-1.012; p=0.914). The machine learning model achieved an area under the curve of 0.641 for prediction of central nervous system toxicity. Transcriptomic analysis showed broader transcriptional reprogramming with abemaciclib than with palbociclib, with 5450 versus 265 differentially expressed genes, while docking analysis identified more favorable predicted binding of abemaciclib than palbociclib to GSK-3β and ABCB1 in this in silico setting, providing structural support for but not proving the proposed mechanistic interpretation. Abemaciclib and palbociclib were associated with distinct neuropsychiatric and systemic toxicity patterns. These findings argue against a uniform class effect and support further prospective evaluation of drug-specific survivorship toxicity profiles.