Single-Scan Renal TIAC Modeling During Lu-177 PSMA-617 Therapy

Key Takeaways

• Among patients with low-volume metastatic hormone-sensitive prostate cancer receiving [177Lu]Lu-PSMA-617, several multi-cycle single-scan renal TIAC pairings had MAPE below 10% against modeled reference estimates.
• A midrange first-cycle scan paired with a later second-cycle scan produced the lowest average errors in the two-cycle single-scan comparison.
• Using all first-cycle scans improved earlier second-cycle estimation relative to the evaluated monoexponential cross-cycle method, but still required first-cycle serial imaging.
• Substantial deviations in individual kidneys persisted despite low average error across some scan pairings.

Repeated single-photon emission computed tomography/computed tomography (SPECT/CT) acquisitions for renal biokinetic sampling add visits and measurements during successive [177Lu]Lu-PSMA-617 treatments. Renal time-integrated activity coefficients (TIACs) depend on activity over time, yet serial patient-specific sampling is demanding.

Investigators evaluated imaging from a prospective clinical study in a retrospective renal dosimetry analysis. Ten patients with low-volume metastatic hormone-sensitive prostate cancer received first and second cycles of [177Lu]Lu-PSMA-617; their kidneys yielded 40 correlated kidney-cycle datasets, not 40 participants. Each cycle included five SPECT/CT acquisitions after injection. Nonlinear mixed-effects (NLME) modeling used a literature-derived sum-of-exponentials function and, in its final form, a combined inter-individual/inter-cycle random effect. Fitting all-time-point (ATP) observations and analytically integrating the fitted curves produced model-derived reference TIACs, not independently measured true values. Leave-one-patient-out comparisons assessed single-time-point (STP) estimates by mean absolute percentage error (MAPE) and relative deviation (RD). The comparisons covered separate single-cycle NLME fits; a first-cycle full-scan monoexponential half-life carried into second-cycle STP fitting; two-cycle NLME with one scan per cycle; and NLME with all first-cycle scans plus one second-cycle scan. The monoexponential comparator used the investigated patient’s complete first-cycle scans to estimate an individual half-life. The last NLME setting also retained that patient’s complete first-cycle scans, alongside one second-cycle scan. Other patients supplied population-model ATP data.

The lowest average errors for the two-cycle STP/STP pairing occurred with imaging at approximately 23 hours in the first cycle and 71 hours in the second. First-cycle MAPE was 4.5 ± 1.7% against the ATP model-derived reference. Model fits met preset criteria. Separate single-cycle NLME analyses numerically favored the corresponding imaging windows, although differences from neighboring time points were not statistically significant.

For the same paired scans, second-cycle MAPE was 6.5 ± 3.4%. With all first-cycle scans available instead, the three intermediate second-cycle scans fell in the low-error range, while the earliest second-cycle scan yielded 13.9 ± 5% MAPE. The NLME approaches showed better average agreement at earlier second-cycle scans than the evaluated monoexponential cross-cycle comparator, an in-study comparison. Average performance did not capture every kidney-level estimate: two pairings involving the earliest first-cycle observation produced individual RDs up to 109% in one patient’s kidneys.

These retrospective, patient-based comparisons measured agreement with ATP references derived under the specified model, rather than independent true TIACs, verified absorbed doses, toxicity, or patient outcomes. Kidneys and successive cycles remained correlated within patients. Because the small cohort did not support separate estimates of inter-individual and inter-cycle variability, the combined random effect cannot establish an isolated cycle effect. The setting with complete first-cycle imaging still depended on serial scans in that cycle; it was not a wholly single-scan treatment course. The authors noted no consensus on acceptable clinical dosimetry error and identified external-cohort comparisons and alternative reference fits as unresolved assessments.

Multi-cycle NLME offered more scan-timing choices with relatively close agreement to model-derived renal TIACs in this cohort. Individual reliability and generalizability beyond the analyzed patients remain unresolved.

Clinician Questions

Why was late monoexponential renal TIAC estimation inaccurate after [177Lu]Lu-PSMA-617 therapy?

For second-cycle renal TIAC estimation after [177Lu]Lu-PSMA-617 therapy, the evaluated monoexponential approach transferred an individual effective half-life estimated from all first-cycle scans into a one-scan second-cycle fit. The authors attributed very large errors at the late, low-activity scan mainly to extrapolation failure and noted that holding the half-life constant across cycles could also bias estimates.

Were baseline kidney-function markers included in [177Lu]Lu-PSMA-617 multi-cycle renal TIAC modeling?

Baseline renal-function markers were not covariates in the NLME renal TIAC modeling of [177Lu]Lu-PSMA-617 for patients with low-volume metastatic hormone-sensitive prostate cancer. Investigators could not rule out an effect of individual renal function on renal uptake or clearance and proposed evaluating such markers in larger cohorts with prespecified covariate analyses.

Recommended Reading

• For more on PSMA-targeted prostate cancer therapies: Improved Outcomes in mCRPC with PSMA-Directed Diagnostics and Therapies