Divergent Kinetics: Physiologically Based Versus Population Pharmacokinetic Dosimetry for 177lu-PSMA In Xcat Phantom Virtual Patients
Abstract
Purpose
To compare the predictions of two pharmacokinetic models for 177Lu-PSMA radiopharmaceutical therapy on a virtual patient cohort and evaluate how their differences in predicted time-activity curves propagate to the organ-at-risk self-dose estimations across anatomically heterogeneous patients.
Methods
A population-based pharmacokinetic (popPK) model by Siebinga et al. was compared against a physiologically based (PBPK) model from Fele-Paranj et al on a virtual patient cohort consisting of 33 male 4D eXtended CArdiac-Torso (XCAT) phantoms. Reflecting typical usage, the PBPK model was personalized using XCAT-derived parameters, whereas the popPK model maintained standard population-average priors without augmentation. Time-activity curves (TACs) were estimated using both models (total administered dose: 3000 MBq), and key features: peak activity, time to peak, and cumulative exposure, were extracted. Organ-at-risk (OAR) self-doses were then estimated using MIRD S-values, mass-scaled at the organ level to each individual XCAT phantom.
Results
Across the cohort, the PBPK model consistently predicted earlier and sharper activity peaks, whereas the popPK model exhibited delayed peaks and prolonged retention (tumor peak(mins): PBPK mean ±stddev is 304±20.8 vs popPK is 1438). This resulted in higher cumulative exposure: 2.0x higher in tumors (popPK:16943 vs PBPK:8493±122.4MBq·s) and 2.5x higher in salivary glands, leading to higher self-dose estimations in salivary glands (popPK:36.6±3.7b4 vs PBPK:14.8±0.17 mGy), kidneys, and liver. Furthermore, popPK-derived self-doses had higher variance (3.74 vs. 0.17, 1.51 vs. 0.02, 5.68 vs 0.79 mGy in salivary glands, liver and kidneys respectively) .
Conclusion
Divergent kinetics and dose estimations highlight how model architecture affects 177Lu-PSMA pharmacokinetic predictions. The higher variance in the popPK model’s self-dose estimation reveal a limitation in scaling activity with anatomy, implying reduced robustness when population-average priors are applied to heterogeneous patients. This underscores the need for personalized approaches which combine explainability and flexibility of PBPK models with the empirical credibility of population-based approaches.