Single‐time‐point renal dosimetry for [ 177 Lu]Lu‐PSMA‐617: Incorporating multi‐cycle population data using nonlinear mixed‐effects models
Indra Budiansah, Steffie M. B. Peters, Bastiaan Privé, James Nagarajah, Mark W. Konijnenberg, Gerhard Glatting, Deni HardiansyahAbstract
Background
Single‐time‐point (STP) dosimetry using nonlinear mixed‐effects (NLME) modeling has emerged as a promising approach, with recent studies demonstrating high accuracy. This framework enables estimation of the time‐integrated activity coefficient (TIAC) from a single imaging time point and, consequently, prediction of absorbed dose for subsequent therapy cycles. However, the use of multi‐cycle population data to support NLME‐based STP dosimetry has not yet been systematically evaluated.
Purpose
This study aimed to evaluate whether incorporating multi‐cycle population data into an NLME framework improves agreement between STP‐derived TIACs and all‐time‐point (ATP)‐derived model‐based reference TIACs in renal dosimetry.
Methods
Renal biokinetic data [ 177 Lu]Lu‐PSMA‐617 were collected from 10 patients over two therapy cycles, with both kidneys analyzed, yielding 40 renal datasets. In both cycles, SPECT/CT imaging was performed at five time points. A literature‐based structural function was used in the NLME modeling to incorporate inter‐individual variability, inter‐cycle variability, and variability between left and right kidney positions. Goodness‐of‐fit was evaluated based on visual inspection, relative standard errors (RSEs) < 50%, and the absolute value of the off‐diagonal elements in the correlation matrix < 0.8. TIACs derived from ATP fitting were used as the reference (rTIACs). STP fitting was performed using the leave‐one‐out method across defined scenarios. The TIACs derived from STP fitting were designated as STP TIACs (sTIACs). The accuracy of sTIACs in predicting rTIACs was assessed using relative deviations (RDs) and mean absolute percentage error (MAPE).
Results
The model fits satisfied all predefined goodness‐of‐fit criteria. STP dosimetry using the multi‐cycle dataset offered greater flexibility in time‐point combinations, leading to MAPEs < 10%. Additionally, an improvement in early‐time‐point STP dosimetry was observed when using multi‐cycle data.
Conclusions
Incorporating multi‐cycle data into NLME‐based STP dosimetry reduced extreme RDs, increased flexibility in feasible time‐point combinations, and improved early‐time‐point STP performance in this small two‐cycle cohort.