Single-time-point tumor dosimetry using population-based model selection and nonlinear mixed-effects modeling in...

Molecular radiotherapy with [¹⁷⁷Lu]Lu-PSMA-617 is an effective treatment for metastatic castration-resistant prostate cancer. Accurate dosimetry is essential for maximizing therapeutic efficacy while minimizing toxicity. However, standard dosimetry requires multiple imaging sessions, posing logistical challenges. Single-time-point (STP) dosimetry offers a practical alternative but remains challenging for tumor kinetics due to high inter-patient variability. Nonlinear mixed-effects (NLME) modeling, combined with population-based model selection (PBMS), has demonstrated potential for improving STP dosimetry accuracy.

The purpose of this study was to evaluate the accuracy of STP tumor dosimetry using SPECT/CT data, PBMS, and an NLME model in a large population with diverse biokinetic measurements for [¹⁷⁷Lu]Lu-PSMA-617 therapy.

Biokinetic data for [¹⁷⁷Lu]Lu-PSMA-617 in tumors were obtained from forty-nine patients with metastatic castration-resistant prostate cancer using SPECT/CT at time points (1.80 ± 0.80), (18.67 ± 0.90), (42.63 ± 1.03), (66.27 ± 0.96), and (159.02 ± 23.35) h after injection. Ten different functions, derived from various parameterizations of two- to four-exponential functions, were fitted to the data using the NLME framework. Each function's parameters were defined as a combination of fixed and random effects. A PBMS approach was employed, using goodness-of-fit tests and Akaike weights to identify the function best supported by the data. The selected function from the NLME fitting of all time points with the leave-one-out method was used to calculate the reference time-integrated activities per volume (TIAVs). The parameters from STP fitting were used to calculate the STP TIAVs. Additionally, STP dosimetry was performed using the Hänscheid method to calculate the TIAVs. Relative deviations (RDs) and root-mean-square errors (RMSEs) were used to analyse the accuracy of the calculated STP TIAVs and Hänscheid method TIAVs compared with the reference TIAVs.

The function $f_{4b}\left(t\right)=A_1{e}^{-\left({\lambda }_1+{\lambda }_{\mathrm{phys}}\right)t}+A_2{e}^{-\left({\lambda }_{\mathrm{phys}}\right)t}-\left(A_1+A_2\right)e^{-\left({\lambda }_2+{\lambda }_{\mathrm{phys}}\right)t}$ was selected as the fit function most supported by the data with an Akaike weight of 93%. For STP dosimetry, a single SPECT/CT measurement at time point 4 (66.3 ± 0.9 h) after injection showed a relatively low mean RD of 0.5% ± 11.0% and median RD of -0.6%. The RMSEs for the STP TIAVs RDs for time points 1-5 were 53%, 29%, 17%, 11%, and 38%, respectively. The STP tumor dosimetry using the PBMS NLME method outperformed the Hänscheid method for all investigated time points.

Our findings demonstrate that a single SPECT/CT measurement at 3 days after injection may be used to estimate tumor TIAVs in [¹⁷⁷Lu]Lu-PSMA-617 therapy using the NLME method and PBMS.