A cohort electron blood self-dose fraction and its propagated uncertainty for ¹²⁴I PET pre-therapy dosimetry in radioiodine therapy of differentiated thyroid cancer.

Abstract

ObjectivePre-therapy blood-marrow dosimetry for ¹³¹I treatment of differentiated thyroid cancer (DTC) requires serial imaging and blood sampling, limiting clinical adoption. When blood data are unavailable, operators fall back to fixed or imaging-only prescribing, usually without an explicit uncertainty bound. We characterise the cohort electron blood self-dose fraction (βfrac) in a single-centre ¹²⁴I PET cohort and quantify the uncertainty incurred by substituting that reference for patient-specific blood data.MethodsRetrospective single-centre analysis of 18 pre-therapy ¹²⁴I PET courses in 15 DTC patients with paired serial venous blood-counting, all prepared by thyroid-hormone withdrawal. Time-activity curves were fitted in the ¹³¹I therapy frame following the EANM Standard Operational Procedure, separating Dblood into its electron (β, blood self-dose) and photon (γ, body cross-fire) components. The substitution uncertainty was propagated analytically, then validated against patient-level errors by leave-one-patient-out cross-validation.ResultsThe cohort β fraction was 84.3 ± 4.8% across 18 courses (mean ± SD; median 85.4%; 95% CI 82.0-86.7%), and 84.4 ± 4.6% over the 15 patients. The β fraction decreased with body weight (- 0.113 pp/kg, p = 0.019), although not significantly when reduced to one value per patient (p = 0.118); the association is therefore exploratory. Substitution gave a per-patient 1-σ uncertainty on the imaging-only Dblood estimate of ≈ 30%. Under leave-one-patient-out cross-validation the mean absolute percentage error was 26.3% and the RMSE 32.7%, with 12 of 18 courses within ± 30% of their reference.ConclusionsA cohort β fraction of 84.3 ± 4.8% is proposed as a fall-back reference for blood-free prescribing in DTC radioiodine therapy. Because the estimator's error is multiplicative on the estimate, a prescribing cap must come from the lower tail of that error, giving approximately 1.0-1.2 Gy in this cohort. No 95% coverage is demonstrated at this sample size, and the cap is offered as an exploratory observation requiring local confirmation rather than a validated threshold. The decomposition is imaging-modality agnostic for ¹³¹I clearance studies and therefore potentially applicable to low-dose ¹³¹I SPECT, although no SPECT data were analysed here.