Optimal MRD-based end point to support response-adapted treatment cessation in newly diagnosed multiple myeloma

The therapeutic success of first-line quadruplet (QUAD) induction therapy and autologous stem cell transplantation (ASCT) has reinvigorated an interest in fixed-duration therapy, yet optimal short-term efficacy end point for treatment cessation is unknown. Using data from a phase 2 clinical trial and a prospective institutional database, we tested the predictive performance of 5 short-term efficacy end points among 221 patients who received QUAD + ASCT followed by treatment cessation if minimal residual disease (MRD) by next-generation sequencing negative for 2 consecutive time points. Efficacy end points tested were International Myeloma Working Group-defined stringent complete response, MRD <10-5 (single data point), MRD <10-6, sustained MRD (S-MRD; 2 consecutive assessments at least 1 year apart) <10-5, and S-MRD <10-6. We built 5 parallel Cox regression models for each efficacy end point with progression-free survival (PFS) as the outcome. Best fitting models were determined using the Akaike information criterion (AIC) and Heagerty and Zheng C-index. The best fitting model (AIC, 417.2; C statistic, 0.757) was based on S-MRD <10-5 (hazard ratio, 0.23; 95% confidence interval, 0.11-0.47). Similar results were seen for predicting the risk of progression/MRD resurgence among 121 patients undergoing MRD-guided treatment cessation. S-MRD <10-5 is the best predictor of PFS and yields the best predictive models for the risk of MRD resurgence or progression in the setting of fixed-duration therapy. This trial was registered at www.clinicaltrials.gov as #NCT03224507.