AI-Generated Synthetic Patient Data Helps in Evaluating Daratumumab Treatment Benefit in Multiple Myeloma Subgroups

Introduction: Advances in AI have enabled the generation of synthetic individual patient data (SynthIPD) that mimics the statistical properties of the real IPD, which can help generate subgroup-specific insights in the absence of subgroup-specific IPD. We apply this technology in evaluating Daratumumab treatment benefit in high-risk cytogenetic multiple myeloma subgroup. The addition of daratumumab to standard multiple myeloma (MM) regimens has been associated with improved response rates and clinical outcomes. A systematic review (meta-analysis) enables further evaluation of the effectiveness of daratumumab for MM patients with high-risk cytogenetic factors. However, concerns have been raised about meta-analyses due to the statistical model assumptions they impose. This study aims to address these limitations by employing SynthIPD to provide a more robust analysis.

Methods: This study focuses on newly diagnosed multiple myeloma (NDMM) and relapsed or refractory multiple myeloma (RRMM) patients with high-risk cytogenetic profiles. We included six published studies: ALCYONE, CASSIOPEIA, and MAIA for NDMM, and CANDOR, CASTOR, and POLLUX for RRMM. Large language models were used to extract up-to-date clinical evidence from the literature. SynthIPD including cytogenetic risk profile were generated via digitization and an innovative optimization algorithm. Meta-analyses based on random effects models were then conducted, and the results were compared with those from pooled analyses using synthetic data.

Results: SynthIPD for cytogenetic high-risk patients were generated for three studies for NDMM: ALCYONE (treatment vs control: n=53 vs n=45), CASSIOPEIA (n=82 vs n=86), MAIA (n=48 vs n=44); and three studies for RRMM: CANDOR(n=48 vs n=26), CASTOR (n=40 vs n=35), POLLUX (n=35 vs n=35). The generated hazard ratios (HRs) for each study closely aligned with the reported HRs for high-risk patients, with a maximum relative error of less than 2%. By pooling these datasets, we obtained SynthIPD datasets for NDMM and RRMM patients. Stratified Cox regression analyses yielded HRs of 0.764 (95% CI 0.529-1.104) for NDMM and 0.431 (95% CI 0.294-0.613) for RRMM. Random-effects meta-analyses showed HRs of 0.764 (95% CI 0.529-1.104) for NDMM and 0.425 (95% CI 0.287-0.621) for RRMM. Compared with traditional meta-analysis techniques, the synthetic data approach provides additional clinical insights that were not available in the original publications. For instance, the clinical results for the six studies within high-risk group (e.g., KM curves in high-risk subgroup) can be reproduced, which were not reported in the original studies.

Conclusion: AI-generated synthetic data enhances the robustness of conventional meta-analysis outcomes. The identification of HRs and corresponding confidence intervals through SynthIPD enhances the credibility of our proposed method, offering a reliable approach to evaluating the efficacy of daratumumab in high-risk patients. Since SynthIPD includes individual covariate information, any statistical results derived from the original IPD can be effectively approximated using SynthIPD. Using the SynthIPD, we show that the addition of daratumumab to standard regimens improves PFS for RRMM patients with high-risk cytogenetic factors, although its benefit for NDMM patients is not statistically significant. This analysis can be extended to other multiple myeloma high risk subgroups.

Bar:Bristol Myers Squibb: Current Employment. Cowan:Sanofi: Consultancy, Research Funding; Adaptive Biotechnologies: Consultancy, Research Funding; Harpoon: Research Funding; IgM biosciences: Research Funding; HopeAI: Consultancy, Current holder of stock options in a privately-held company; Caelum: Research Funding; Abbvie: Research Funding; Regeneron: Research Funding; Juno/Celgene: Research Funding; BMS: Consultancy, Research Funding; Nektar: Research Funding; Janssen: Consultancy, Honoraria, Research Funding; Sebia: Consultancy. Shi:Yiviva Inc: Consultancy, Membership on an entity's Board of Directors or advisory committees; Regeneron Pharmaceuticals, Inc.: Consultancy, Membership on an entity's Board of Directors or advisory committees; Hoosier Cancer Research Network: Consultancy, Membership on an entity's Board of Directors or advisory committees; Kronos Bio: Consultancy, Membership on an entity's Board of Directors or advisory committees; Mirati Therapeutics Inc: Consultancy, Membership on an entity's Board of Directors or advisory committees; Genmab: Consultancy, Membership on an entity's Board of Directors or advisory committees; BMS: Research Funding; Roche/Genetech: Research Funding; Janssen: Research Funding; Novartis: Research Funding; MPAACT: Research Funding. Ma:HopeAI, Inc.: Current Employment, Current equity holder in private company.