Development of predictive relapse indicators for myeloma T cell engagers (PRIME) model for myeloma patients receiving BCMA- and GPRC5D-targeting T cell engagers

Abstract

Abstract Introduction: T cell engagers (TCEs), particularly those targeting B-cell maturation antigen (BCMA) and G-protein coupled receptor class C group 5 member D (GPRC5D), have transformed the treatment landscape of relapsed/refractory multiple myeloma (RRMM), achieving unprecedented response rates. Despite these advances, many patients either relapse early or exhibit only transient responses, highlighting the urgent need for reliable predictors of treatment outcomes. To address this, we developed a predictive model that estimates the risk of progression at the time of TCE initiation: Predictive Relapse Indicators for Myeloma T cell Engagers (PRIME). The model will be accessible as an online calculator, allowing clinicians to deliver personalized risk assessments and tailor treatment strategies accordingly. Methods We retrospectively identified 325 RRMM patients treated with BCMA and GPRC5D-targeting TCEs within the Mount Sinai Health System to evaluate the prognostic impact of all available clinical variables. To ensure accuracy of the model, we only included 231 patients with minimal missing clinical data. We developed and internally validated a multivariable Cox model to predict progression-free survival (PFS) following TCE therapy in 231 patients. Thirteen pre-specified clinical and laboratory predictors (21 degrees of freedom) were modeled using restricted cubic splines and categorical contrasts. The cohort experienced 125 PFS events, with a median PFS of 17.6 months [95% CI: 12.4, 24.3] and median follow-up of 23.8 months [95% CI: 21.2, 27.6]. To mitigate overfitting, parameter-wise shrinkage was applied post-estimation using Riley's shrinkage methodology. The final global model demonstrated an optimism-adjusted C-index of 0.659 [95% CI: 0.629, 0.749], and the nomogram derived from this model provides individualized 6-, 12-, 24- and 48-month PFS estimates. Continuous predictors were modeled flexibly using restricted cubic splines; hazard ratios (HRs) for these variables represent contrasts between clinically relevant values (e.g., LDH 600 vs 200) and not categorical thresholds. Results In this cohort, the median age was 67.6 years, 42.4% had high-risk cytogenetics and 22.1% had extramedullary disease (EMD). Around 20% of patients received a chimeric antigen receptor (CAR) T cell therapy prior to the TCE and nearly 60% of the cohort had received ≥5 prior lines of therapy with 85.3% being triple class refractory and 39.8% being penta-drug refractory. Around 55% of patients received a BCMA-targeting TCE and 10% of all patients received a TCE in combination with daratumumab and/or IMiDs. At baseline, 69% of patients had adequate renal function and the cohort had a median ferritin level of 507.4 ng/mL, C-reactive protein (CRP) of 11.97 mg/dL, absolute lymphocyte count (ALC) of 1.05x103/uL, absolute neutrophil count (ANC) of 3.67x103/uL, hemoglobin of 10.59 g/dL, platelet count of 168x109/uL, and lactate dehydrogenase (LDH) of 289.1 IU/L. In the shrinkage-adjusted global model, five predictors showed meaningful associations with PFS: ALC, high-risk cytogenetics, EMD, ferritin and LDH. Patients with ALC ≥0.9 vs. <0.9 x103/uL had a 38% lower risk of progression or death (HR = 0.62 [0.43, 0.90], p=0.011) while those with high-risk cytogenetics had a 59% higher risk of progression or death compared to those without (HR = 1.59 [1.10, 2.30], p=0.015). Patients with EMD had a 54% increased hazard compared to those without (HR = 1.54 [1.02, 2.33], p=0.041). For continuous markers modeled with restricted cubic splines, the HR comparing LDH 600 vs 200 U/L was 1.39 [1.06, 1.81], p=0.015, while a ferritin level of 1,000 ng/mL vs 100 ng/mL was associated with a 54% increased hazard (HR = 1.54 [1.01, 2.34], p=0.045). Conclusion We developed and internally validated PRIME, a clinically applicable model that integrates key clinical and laboratory parameters to predict PFS in RRMM patients treated with BCMA- or GPRC5D-targeting TCEs. PRIME demonstrated moderate discrimination and provides individualized 6-, 12-, 24- and 48-month PFS estimates through an accessible online calculator. Finally, we are planning to validate this model with external cohorts from other centers.