A Generative AI Framework for Pharmacokinetic Clinical Study Report Authoring

Clinical Study Reports (CSRs) constitute the final consolidation of findings from clinical studies and routinely include Pharmacokinetic (PK) results. To assist with PK results authoring, we developed a generative Artificial Intelligence (AI) based method that employs a hierarchical, chained large language model (LLM) framework with in-context learning to draft PK results directly from study Tables, Listings, and Figures (TLFs), with optional human input. Unlike traditional fine-tuning approaches, our method does not require large datasets or extensive compute, while producing outputs closely aligned with established CSR structure, tone, and analytical conventions using fewer than a dozen example reports. To assess performance, AI-generated reports and manually expert-written CSRs were evaluated in two blinded review sessions by clinical pharmacologists and pharmacometricians, focusing on relative bioavailability (rBA) and drug-drug interaction (DDI) studies. The AI-generated reports achieved an average reporting quality score of ~90% relative to the manually written CSRs. Together, these results demonstrate a practical, scalable solution for assisting PK report authoring in clinical studies, potentially reducing authoring time while maintaining high-quality standards.