Bridging Trials and Real Life in Fixed-Duration BTKi-Venetoclax for CLL: A Delphi-Enhanced Synthesis Incorporating Artificial Intelligence (AI) Benchmarks.

[BACKGROUND] Optimal patient selection for the most effective BTK inhibitor (BTKi) partner of venetoclax in fixed-duration (FD) BTKi-venetoclax regimens for chronic lymphocytic leukemia (CLL) remains uncertain. This uncertainty largely reflects trial populations that are younger and fitter than many real-world patients, which may limit generalizability to older individuals or those with multiple comorbidities. The present study aims to provide harmonized, pragmatic guidance that bridges trial evidence and routine clinical practice for physicians caring for patients within a regional Italian area served by an established CLL network.

[METHODS] An adapted Delphi process engaged nine hematologists from Campania, a Southern Italian region, to evaluate the impact of BTKi-venetoclax upfront decision making on three key CLL decision domains: (1) specific comorbidities, (2) genomics (TP53/del(17p), IGHV status, complex karyotype), and (3) logistics (caregiver support, distance to center, and venetoclax ramp-up capacity). A preregistered systematic literature review (2014-2025) informed 12 statements, with consensus defined as ≥ 75% agreement. The two-round Delphi included structured discussion and anonymous voting. Additionally, artificial intelligence (AI) benchmarking with ChatGPT-4 assessed concordance with evidence briefs.

[RESULTS] Consensus was achieved on 83.3% of statements. The panel recommended FD BTKi-venetoclax regimens for fit, younger patients with low- or intermediate-genomic risk. For patients with low-intermediate cardiovascular risk, continuous second-generation BTKi monotherapy or FD therapy combining a second-generation BTKi acalabrutinib with venetoclax is preferable to FD ibrutinib-venetoclax. Adverse genomic features and logistical challenges were key factors favoring continuous second-generation BTKi therapy. AI benchmarking indicated potential for hybrid human-AI decision support, although concordance was fair (58.3%; Cohen's κ = 0.32).

[CONCLUSION] This study provides a pragmatic, consensus-driven framework for FD BTKi-venetoclax therapy in CLL, emphasizing patient selection and logistical feasibility. The work lays the groundwork for more rigorous evaluation of human-AI collaboration in the still-complex landscape of first-line CLL therapy.