Feasibility and Target AUC Accuracy of Personalized High Dose Melphalan in Myeloma

Background: Autologous transplant remains standard first line treatment in multiple myeloma (MM) yet melphalan 200 mg/m² leads to 500% variation in systemic exposure (Nath CE et al, 2010, PMID 20573084), as measured by area under the curve (AUC). Unfortunately, PK-targeted melphalan dosing has been unsuccessful with test dose strategies failing to reliably estimate melphalan AUC for doses >100 mg/m². We developed a PK-based dosing regimen whereby the first 100 mg/m² is used to generate PK data in real-time, allowing us to adjust the second dose within 48 hours (Sweiss K et al, 2020, PMID 32285957). Here we report final results of the Phase A portion of our multicenter Phase 1 trial aimed at demonstrating feasibility and success of this first ever personalized melphalan dosing strategy (NCT04483206).

Methods: Phase A was a pre-planned run-in required by the FDA to show feasibility and accuracy of achieving narrow AUC targets of 13-14 (cohort 1) and 14.1-15 (cohort 2) mg*h/L. Eligible patients who had received two or more lines of therapy (Rajkumar V et al, 2015, PMID 26272048) received day-3 melphalan 100 mg/m² and 7 blood samples were collected and shipped overnight for LC-MS/MS analysis. Using noncompartmental analysis (NCA), day -3 AUC estimation was used to determine day -1 dosing to achieve the target AUC. For comparison, we performed a Bayesian post-hoc estimation of clearance (CL) and AUC with nonlinear mixed effects modeling (NLME) using our developed population PK model.

Results: Target enrollment of 20 patients was achieved. Median age was 64 (49-75) years, with 6 (30%) ≥70 years. 14 (70%) were male and 15 (75%) were Black. Median day -3 CrCl was 97.6 (44.8-185.8) ml/min and median body weight was 91.7 (65-142.4) kg. Our real-time workflow was feasible and adopted at both centers, with 273 (97.5%) of the prescribed blood samples collected, processed, and shipped overnight successfully to the PK laboratory, and dose adjustment was performed in all patients. 7 samples not collected on day -1 were due to unexpected overnight shipping delays which resulted in delayed drug administration. After day -3 melphalan 100 mg/m², the median day -3 CL and AUC was 22.1 (16.3-38.2) L/hr and 9.4 (7.0-11.8) mg ´ h/L in Cohort 1, and 21.4 (17.8-25.5) L/hr and 9.8 (7.1-10.8) mg ´ h/L in Cohort 2, respectively, resulting in all but 1 (95%) patient needing a lower day -1 adjusted dose for both cohorts. Based on the PK data observed, a BSA-based dose, 200 mg/m², would have resulted in a median total melphalan AUC of 18.2 (14.2-22.8) mg ´ h/L, with a CV of 12.6%. PK-guided dose adjustment, however, resulted in less variability with a median total AUC of 13.5 mg ´ h/L (CV = 4.20%) in Cohort 1 and 14.4 mg ´ h/L (CV = 5.62%) in Cohort 2. 8 (88.9%) of 9 deviated from the target AUC window by a median of -1.83% (-4.55% - -0.19%). Only 1 patient exceeded the target window by 10.1%.

We then performed posthoc analyses to compare NLME to NCA and found that day -3 CL estimates correlated well (r=0.98). The median posthoc total AUC using NLME was 13.3 (12.6-13.9) mg ´ h/L in Cohort 1 and 14.3 (13.9-15.9) mg ´ h/L in Cohort 2. Of the 4 and 5 patients who did not achieve the target AUC using NCA in Cohorts 1 and 2, respectively, NLME would have brought 2 (50%) and 3 (60%) patients within the target AUC range. 3 patients who had missing samples on day -1 were outside the target windows using NCA but moved to within the target window ranges when NLME was used, highlighting the decreased vulnerability of this method to missing samples when estimating CL. To reduce the number of blood samples, we used an optimal design ($DESIGN) option in NONMEM7, version 7.5.0 and tested post-hoc AUC-targeting performance of 2 sampling schemas including time points at 0.5, 0.75, 1.75 and either 3 or 4.5-hour post start of infusion. CL estimates were similar to that from CL estimates from NLME with full 7 samples.

Conclusion: For the first time in the clinic we are able to demonstrate feasibility of targeting melphalan AUC with a high degree of accuracy and rapid turnaround of sample and data analyses. NLME further improves our ability to achieve the target AUC, even with limited samples. Given this, Phase B of this trial will proceed to identify the maximum tolerated systemic exposure (MTSE) of up to 5 AUC cohorts in a 5+5 design using our developed NLME model and a reduced 4-time point sampling schedule. Clinical response and safety follow up will be presented at the meeting.