Pb2110: predictors of frontline (fl) doublet or triplet regimen use among transplant-ineligible (tie) patients (pts) with newly diagnosed multiple myeloma (ndmm) using a machine learning approach

Topic: 14. Myeloma and other monoclonal gammopathies - Clinical Background: Progress in drug development for NDMM in recent years has led to improved outcomes and a shift toward guideline-based utilization of triplet over doublet regimens as standard of care for the vast majority of pts. Nevertheless, FL doublets continue to be used in NDMM pts and may lead to suboptimal outcomes. Aims: This study aimed to identify predictors for initiation of FL doublet or triplet therapy among pts with TIE NDMM, and understand the direction and magnitude of each predictor, using real-world data. Methods: TIE NDMM pts who initiated a FL doublet or triplet regimen between 1/1/2015-11/5/2021 were analyzed using the US nationwide Flatiron Health electronic health record-derived de-identified database. Pts with concurrent malignancy, amyloidosis, insufficient follow-up (<28 days), incomplete therapy data, and quadruplet or stem cell transplant use before the end of FL were excluded. Pt characteristics (i.e., demographics, clinical characteristics, comorbidities, medications, cytogenetics, vitals, and laboratory tests) were evaluated during the 12-month period prior to FL initiation. A random forest model was used to identify predictors of FL doublet or triplet use and a multivariate logistic regression was used to determine the magnitude and direction of each predictor. Results: A total of 1,195 pts initiated a doublet regimen (median age: 76 years, 48.8% female) and 2,512 initiated a triplet regimen (median age: 70 years, 45.5% female). The random forest model identified 43 features predictive of regimen choice; the top features included age, year of FL initiation, frailty index, detectable M protein, living in an urban state, and high-risk cytogenetics (HRCG). The model had good predictive power (AUROC = 0.74). Pts treated with a FL doublet were older (≥80 vs <60 years old: odds ratio [OR]=6.56), more frail (frail vs fit: OR=1.36), and more likely to have IgG MM (OR=1.24, all P<0.05). Predictors of FL triplet use included detectable M protein (OR=1.33), ≥1 form of trisomy (OR=1.39), del(17/17p) (OR=1.87), International Staging System (ISS) (stage 3 vs 1: OR=1.35), urban vs rural state (OR=1.38), Black race (OR=1.31), and having musculoskeletal (MSK) disease (OR=1.45, all P<0.05). Triplet use was also more likely in later years: 44.6% in 2015 vs 80.2% in 2021 (OR=6.76; P<0.05). Summary/Conclusion: Increasing use of triplets for TIE NDMM was observed from 2015 to 2021. However, doublet use persisted in a significant fraction of pts (20% in 2021). Doublets were more commonly used in older and more frail pts, while pts with HRCG, detectable M protein, MSK disease, living in an urban state or of Black race were more likely to get triplets. As FL triplet therapies have shown improved efficacy with a tolerable safety profile, there is an opportunity to improve pt outcomes, especially among frail and older pts. Keywords: Multiple myeloma, Elderly, Triple therapy, Real world data