(S-089) Use of population time-activity curve modeling and simulation to support the reduced tumor dosimetry collection timepoints in radioligand therapy (RLT) clinical trials
Senior Principal Pharmacometrician Novartis, United States
Disclosure(s):
Larissa Lachi Silva: No financial relationships to disclose
Objectives: Dosimetry is an essential measure to deliver safe and efficacious radioligand therapy (RLT) treatments to patients with a favorable benefit-risk enabling dosage and regimen selection. However, the dosimetry acquisition is burdensome to patients both in terms of imaging time in a SPECT/CT scanner and number of visits to the clinical site. The objective of this work is to demonstrate the application of NLME and design optimization to reduce the number of dosimetry imaging timepoints for clinical trials using Phase I dose escalation tumor time-activity data.
Methods: A NLME model was developed in Monolix 2023R1 to describe the tumor time-activity curves (TACs) from 28 lesions for the Phase I [177Lu]Lu-NeoB (NCT03872778) clinical trial where 6 dosimetry timepoints (1, 6, 24, 48, 72, 168 h) in Cycle 1 were collected. The estimated model was fed into PopED for design optimization[1], to assess the feasibility of reducing the dosimetry collection to 4 timepoints for Cycle 1. A sensitivity analysis was conducted to assess the impact of sampling window on the selected designs. Moreover, assuming that subsequent cycles follow the same kinetics as Cycle 1, the TACs were simulated in Simulx 2023R1 using combinations of 2 timepoints and compared to the full profile in Cycle 1 using the percentage error of area under the curve (%errorAUC). The combination of 2 timepoints that minimized %errorAUC was selected as the optimal collection timepoints for the subsequent cycles of a Phase I trial or for all cycles of later phase trial with prior experience of dosimetry.
Results: A bi-exponential decay function including IIV on all parameters and proportional error was the base model for the tumor dosimetry data (167 observations from 11 patients, including different tumor types). Dose level (cohort) and tumor mass effect were added as covariates in the model. All final model parameters were estimated with reasonable precision with RSE of 56.6% at most. The design optimization with PopED led to two optimal 4 timepoints designs 2, 24, 48, 168 h and 2, 24, 72, 168 h for Cycle 1 based on the precision of estimates, lower RSEs compared to other tested designs. The sensitivity analysis showed no significant changes in efficiency and precision of parameter estimates for the tested sampling windows of ± 6 h for intermediate timepoints or ± 24 h for the last timepoint. For the selection of reduced 2 timepoints for subsequent cycles of a Phase I or later phase trails, simulation of TACs using 2 and 48 h provided a median %errorAUC of 4.16% with less variability than the combination of 2 and 72 h (%errorAUC of 5.14%), while combinations using 2 and 24 h or 2 and 168 h showed large difference and variability compared to the observed data.
Conclusions: A reduced timepoint dosimetry approach was developed and evaluated for tumor TACs using Phase I dose escalation data. NLME modeling and covariate analysis from multiple timepoint imaging in Cycle 1 can be used for design optimization and reduction of imaging timepoints from 6 to 4 in the first cycle or 2 in subsequent cycles in Phase I and later phase trials while maintaining reasonable accuracy in determining AUCs of individual TACs, making collection of dosimetry operationally practical for the clinical sites and less burdensome to patients.
Citations: [1] Nyberg J, Ueckert S, Stroemberg EA, Hennig S, Karlsson MO, Hooker AC (2012). “PopED: An extended, parallelized, nonlinear mixed effects models optimal design tool.” Computer Methods and Programs in Biomedicine, 108.
