Radioembolization of hepatocellular carcinoma with ⁹⁰Y glass microspheres: an earlier administration day unexpectedly improves tumour control probability.

Abstract

⁹⁰Y glass microspheres have a shelf life of 12 days from the calibration date, allowing flexible administration after a variable decay interval. For a fixed intended activity, a longer interval results in a higher number of administered microspheres per GBq and in a lower activity per sphere a_s. This study aimed to demonstrate that, for a fixed Tumour Absorbed Dose (TAD), Tumour Control Probability (TCP) is higher when the decay interval is shorter (4 days vs. 8 days). In the second part of the study, we focused on fully perfused lesions, i.e. those showing matching perfused and radiological volumes, where calculating mean microsphere spatial density (ρ) is meaningful. We investigated which variable was associated with radiological response. We retrospectively analysed lesion-by-lesion response at the best response time using the mRECIST criterion. Two chronologically sequential cohorts of patients were compared. Both cohorts were planned and treated with lobar administration, with exactly the same methodology, except for the post-calibration decay interval (Administration day: day 4 vs day 8). Infiltrative lesions and those with additional treatments post-radioembolization were excluded. To compare efficacy at the same TAD, we plotted TCP curves as a function of TAD and stratified TCP by lesion mass (M), using an arbitrary 50 g cut-off (equivalent to a 4.6 cm diameter). On fully perfused lesions, TCP was analyzed also as function of ρ and mass. We then conducted univariate ROC and multivariate analysed to assess response impact based on TAD, ρ, a_s, and M, grouping CR + PR (Objective Response, OR) versus SD + PD, as well as CR versus PR + SD + PD. A total of 94 patients with 150 lesions were analyzed. TCP (for TAD up to 600 Gy) was significantly higher for the 59 lesions treated on Day 4 compared to the 91 lesions treated on Day 8, though the difference diminished at higher TAD values. In fully perfused lesions, TCP plateaued at 344 Gy and 160 Gy for lesions treated on Day 8 and Day 4 respectively. ROC analysis for fully perfused lesions showed poor AUC values for CR + PR versus SD + PD: 0.62, p = 0.01 for a_s, 0.63, p = 0.01 for TAD, and 0.60 p = 0.01 for M, with AUC for ρ being non-significant. When comparing CR versus PR + SD + PD classes, only M was significant, with a fair AUC value of 0.71, p = 0.01. Multivariate analysis showed that CR + PR was significantly associated only to a_s, with 79% higher response probability for administration on Day 4. When considering CR alone, significance was confirmed only for M, with an odd ratio of 0.19. Our findings on TCP confirmed our preliminary unpublished studies from a different lesion cohort assessed by two independent radiologists. Additionally, our results align with recent experimental histological studies on complete pathological necrosis (CPN) in explanted liver samples after neoadjuvant segmentectomy prior to liver transplantation. However, our data and the CPN findings contrast with previously published simulations based purely on physical parameters (dose-volume histograms at microscopic scale). No explanation is currently available for this discrepancy. When considering OR probability, administration of ⁹⁰Y glass microsphere on Day 8 is not recommended, favouring Day 4 instead. For tumours < 50 g treated on Day 4, increasing TAD beyond 160 Gy does not improve response probability but may increase toxicity risk in lobar administrations. Conversely, larger lesions benefit from a maximal tolerable activity approach. TRIAL REGISTRATION : NumberDose in TARE, INT 154/19. Registered on 8 August 2019.