Talha Munir, MBChB PhD MRCP FRCPath
DISCLAIMER
The contents of the presentation are designed for educational and scientific exchange purposes and are not promotional. They may contain information on investigational products or investigational uses of approved products. No conclusions regarding safety and/or efficacy for such investigational products or uses may be made.
SUMMARY
This study developed a generative machine-learning model using individual patient data (IPD) from phase 3 zanubrutinib trials to simulate complete patient trajectories in chronic lymphocytic leukemia (CLL) and evaluated its ability to reproduce outcomes and support comparative effectiveness research across patient subgroups. A latent ordinary-differential-equation generative model was trained using IPD from the SEQUOIA (treatment-naive CLL) and ALPINE (relapsed/refractory CLL) studies. Validation was conducted in two steps: first, synthetic and observed Kaplan-Meier curves were compared across clinically relevant patient subgroups, including those used during model fine-tuning (del(17p)/TP53, immunoglobulin heavy chain variable region [IGHV], complex karyotype, prior treatment, and β2-microglobulin status) and subgroups defined by characteristics not used for fine-tuning (age, Binet stage, bulky disease, and gender); second, synthetic cohorts generated from the SEQUOIA study were used for comparison with the AMPLIFY study. The median absolute difference in progression-free survival (PFS) was 0.016, with approximately 92% of each synthetic 95% CI contained within the bounds of the corresponding observed CI for both PFS and overall survival; agreement was maintained in non–fine-tuned subgroups (median PFS absolute difference, 0.014; approximately 94% CI containment). At 36 months, the model-generated bendamustine-rituximab arm closely reproduced the AMPLIFY chemoimmunotherapy control (66.9% vs 65.9% PFS, respectively). In an exploratory model-based indirect comparison, synthetic zanubrutinib and observed acalabrutinib plus venetoclax cohorts yielded an estimated PFS hazard ratio (HR) of 0.34 (95% CI, 0.21-0.55), consistent with a previously published network meta-analysis (HR, 0.46; 95% CI, 0.28-0.76) and a matching-adjusted indirect comparison (HR, 0.45; 95% CI, 0.23-0.88 without age adjustment; HR, 0.26; 95% CI, 0.13-0.54 with age adjustment), while producing narrower CIs with the synthetic cohort (CI width: 0.34) vs the matching-adjusted indirect comparison (range, 0.41-0.65) and the network meta-analysis (0.48).
ClinicalTrials.gov ID: NCT03336333 (SEQUOIA), NCT03734016 (ALPINE)
Haitao Wang, PhD
Alvaro Alencar, MD
Nakhle S. Saba, MD
Are you a returning user?
LOGINFor new users, please select your country/region
This website is intended exclusively for healthcare professionals and is provided for scientific exchange and educational purposes only. It is not intended for patients or the general public and does not constitute advertising, promotional activity or medical advice.
Content may include information about products, indications or uses that are not approved in all countries or regions, and may be subject to local regulatory requirements.
By continuing, you confirm that you are a healthcare professional and acknowledge the above.
This link is provided for information purposes only. You’ll be leaving beonemedaffairs.com and accessing a third-party website over which [BeOne Medicines | GmbH] has no control or influence.
This link will take you to a third-party website outside of beonemedaffairs.com.
[BeOne Medicines | GmbH] has no control over this website and makes no representations regarding its content or availability in your location. [BeOne Medicines | GmbH] declines any responsibility for third-party information, data usage policies, or personal data processing.
By proceeding, you acknowledge that you are part of the intended audience of the third-party website and confirm your compliance with applicable laws.
Do you want to leave beonemedaffairs.com
You are now leaving the website beonemedaffairs.com and entering another BeOne website with different terms of use. Information included on this website may not be specific to your country. Click Continue to proceed