BACKGROUND: ML-NMR provides a framework for defining an individual-level regression model, which is integrated over aggregate populations to form an aggregate-level model. ML-NMR avoids aggregation bias and produces estimates in any target population. ML-NMR has been published and applied to binary outcomes, but not time-to-event outcomes.

OBJECTIVES: To evaluate the implementation of ML-NMR using a case study evaluating the comparative efficacy of ide-cel, selinexor + dexamethasone (Sd), belantamab mafodotin (BM), and conventional care (CC) for the treatment of TCE RRMM in terms of overall survival as time-to-event outcome.

METHODS: Three single-arm clinical trials (KarMMa [ide-cel], STORM-2 [Sd], and DREAMM-2 [BM]) and two CC studies (KarMMa-RW and MAMMOTH) were identified by a systematic literature review. ML-NMR requires a connected network of randomized controlled trials (RCTs); therefore, we generated artificial-RCTs (aRCTs) based on a naïve comparison of KarMMa vs KarMMa-RW (two-arm aRCT based on individual patient data (IPD)) and a naïve comparison of the DREAMM-2 intention-to-treat (ITT) vs STORM-2 ITT vs MAMMOTH treated populations (three-arm aRCT based on aggregate data (AD)). The ML-NMR assumed a proportional hazards Weibull, Gompertz, or exponential model, adjusted for number of prior lines, age, and triple-class refractory (TCR) status.

RESULTS: Using Weibull distribution, we estimated hazard ratios (HRs) and 95% credible intervals (CrIs) for the aRCT-IPD target population, which suggested ide-cel was more efficacious than Sd (HR 0.28 [95%CrI: 0.17, 0.47]), BM (HR 0.50 [95%CrI: 0.30, 0.82]), and CC (HR 0.45 [95%CrI: 0.29, 0.69]). The CrIs included the null effect for Sd vs CC and BM vs CC. The CrIs for the covariate-related parameters (main effect and effect modifier) all included zero, except for the TCR main effect.

CONCLUSIONS: ML-NMR provides a useful method to adjust for between-study differences using IPD and AD RCTs. ML-NMR analysis suggests ide-cel to be more efficacious than Sd, BM, and CC.