Hybrid Multi-timepoint/Single-timepoint Imaging for Multi-cycle Radiopharmaceutical Therapy Dosimetry: Performance in [177Lu]Lu-DOTATATE PRRT

ABSTRACT Background Reducing multi-timepoint (MTP) imaging requirements for estimating time-integrated activity (TIA) for radiopharmaceutical therapy (RPT) dosimetry, while maintaining accuracy, is of much interest. Inherent to the widely used single time point (STP) imaging methods for TIA estimation is the assumption of population kinetics. The hybrid MTP/STP imaging method assumes constant kinetics across cycles but benefits from patient-specific kinetics determined from MTP imaging in one cycle to estimate TIA in subsequent cycles using only STP imaging. We aimed to comprehensively evaluate the assumptions and performance of TIA estimation with the hybrid method and two standard STP methods. Furthermore, we propose and test a data-driven scaling approach to reduce bias associated with these analytical methods. Methods Multi-timepoint (up to 4) quantitative SPECT/CT was available for 22 patients imaged after four [ 177 Lu]Lu-DOTATATE cycles. Imaging timepoints (TP) were 2–6 hours, day 1–2, day 3–5, and day 6–8 post-therapy. For kidney and tumor (N=74), mono or bi-exponential fitting was performed on the extracted time–activity data at each cycle to determine the reference TIA and compared with estimates from STP and MTP/STP methods with and without applying the data-driven bias correction. Results Data was analyzed from 82 cycles with at least 3 imaging timepoints. Biokinetics, characterized by the effective clearance half-life (T 1/2eff ), varied substantially between patient structures but was relatively stable across cycles (total between-structure SD: 7.3 h for kidney and 25.2 h for tumor; corresponding within-structure SD across cycles: 3.3 h and 12.1 h). Intraclass correlation coefficients indicate that more than 83% (81% for tumor) of the total variability in T 1/2eff came from differences between kidneys (or tumors) rather than within-kidney (or within-tumor) changes across cycles, supporting the assumptions underlying the hybrid approach over that of STP methods. The hybrid method showed the most stable performance and the least sensitivity to timepoint selection, with median percentage bias across cycles ranging from -2 to 6% (TP1), 0 to 4% (TP2), -4 to -2% (TP3), and 3 to 11% (TP4) for kidney and from -6 to 4% (TP1), -2 to 2% (TP2), -3 to 0% (TP3), and 1 to 2% (TP4) for tumor. With this method, 94–100% of kidneys showed an absolute percentage bias ≤20% across cycles when STP imaging in subsequent cycles was performed at day 1 to 5 post therapy (TP2-3). In internal cross validation, the scaling correction provided the greatest improvement for the Hänscheid approach at all time points and the Madsen method at TP1-2, with minimal improvement for the hybrid method (scale factor ∼1), which already demonstrated low bias relative to the MTP reference. Conclusion Data-driven scaling offers improvements in bias associated with STP methods, but external validation is needed for use in a clinical setting. Hybrid MTP/STP is superior to methods that solely use STP imaging for multi-cycle dosimetry, and it is a practical alternative (when MTP imaging across all cycles is not feasible) that can help advance dosimetry-guided RPT to clinical practice.