Study on the Effect of Pharmacokinetic Model Structure on the Calculation of Time-Integrated Activity (TIA) in Prostate Cancer Treatment Using [177Lu]Lu-PSMA-617 Mohammad Sidik Cahyana (a*),Raden Ayu Nurfadhillah Rifqah (a), Bisma Barron Patrianesha (a), Deni Hardiansyah (a)
Departemen of Physics, Faculty of Mathematics and Natural Sciences, Universitas Indonesia
Abstract
This study evaluated the effect of pharmacokinetic (PK) model structure on time-integrated activity (TIA) estimation and tumor volume prediction in [177Lu]Lu-PSMA-617 therapy using three PK models integrated with a pharmacodynamic (PD) framework. Biokinetic data from six patients treated with two cycles of [177Lu]Lu-PSMA-617 therapy (3 GBq in the first cycle and 6 GBq in the second cycle, administered at 8-week intervals) were analyzed using quantitative SPECT/CT imaging acquired at five time points per cycle, while CT-based tumor volume was assessed 24 weeks after the second cycle. The investigated models included a mono-exponential model with individual fitting (Mono-I), and a minimal physiologically based pharmacokinetic model with individual fitting (mPBPK-I). For Mono-I and mPBPK-I, tumor and kidney release rates, receptor densities in tumors and kidneys, and intrinsic radiosensitivity were individually estimated. The comparison were performed is Mono-I versus mPBPK-I using Mono-I as the reference. Evaluated outputs included kidney and tumor TIA and tumor volume. Inter-model deviation was assessed using mean square error (RMSE) and mean absolute percentage error (MAPE). Model selection among the three models was performed using Akaike weights. Relative to Mono-I, mPBPK-I further reduced discrepancies in kidney TIA (30%/26%), tumor TIA (42%/31%), and tumor volume (75%/62%). Akaike weights indicated stronger statistical support for mPBPK-I than Mono-I. This study demonstrates that pharmacokinetic model structure significantly affects TIA estimation and tumor response prediction in [177Lu]Lu-PSMA-617 therapy, with the individualized mPBPK model showing stronger statistical support and a greater ability to represent patient biokinetics than the individualized mono-exponential model.
