AbstractFirst-line selective internal radiation therapy (SIRT) showed promising outcomes in patients with uveal melanoma liver metastases (UMLM). Patient survival depends on liver’s disease control. SIRT planning is essential and little is known about dosimetry. We investigated whether 99mTc-MAA-SPECT/CT dosimetry could predict absorbed doses (AD) evaluated on 90Y-PET/CT and assess the dose–response relationship in UMLM patients treated with first-line SIRT. This IRB-approved, single-center, retrospective analysis (prospectively collected cohort) included 12 patients (median age 63y, range 43–82). Patients underwent MRI/CT, 18F-FDG-PET/CT before and 3–6 months post-SIRT, and 90Y-PET/CT immediately post-SIRT. Thirty-two target lesions were included. AD estimates in tumor and non-tumor liver were obtained from 99mTc-MAA-SPECT/CT and post-SIRT 90Y-PET/CT, and assessed with Lin’s concordance correlation coefficients (ρc and Cb), Pearson’s coefficient correlation (ρ), and Bland–Altman analyses (mean difference ± standard deviation; 95% limits-of-agreement (LOA)). Influence of tumor characteristics and microsphere type on AD was analyzed. Tumor response was assessed according to size-based, enhancement-based and metabolic response criteria. Mean target lesion AD was 349 Gy (range 46–1586 Gy). Concordance between 99mTc-MAA-SPECT/CT and 90Y-PET/CT tumor dosimetry improved upon dose correction for the recovery coefficient (RC) (ρ = 0.725, ρc = 0.703, Cb = 0.969) with good agreement (mean difference: − 4.93 ± 218.3 Gy, 95%LOA: − 432.8–422.9). Without RC correction, concordance was better for resin microspheres (ρ = 0.85, ρc = 0.998, Cb = 0.849) and agreement was very good between predictive 99mTc-MAA-SPECT/CT and 90Y-PET/CT dosimetry (mean difference: − 4.05 ± 55.9 Gy; 95%LOA: − 113.7–105.6). After RC correction, 99mTc-MAA-SPECT/CT dosimetry overestimated AD (− 70.9 ± 158.9 Gy; 95%LOA: − 382.3–240.6). For glass microspheres, concordance markedly improved with RC correction (ρ = 0.790, ρc = 0.713, Cb = 0.903 vs without correction: ρ = 0.395, ρc = 0.244, Cb = 0.617) and 99mTc-MAA-SPECT/CT dosimetry underestimated AD (148.9 ± 267.5 Gy; 95%LOA: − 375.4–673.2). For non-tumor liver, concordance was good between 99mTc-MAA-SPECT/CT and 90Y-PET/CT dosimetry (ρ = 0.942, ρc = 0.852, Cb = 0.904). 99mTc-MAA-SPECT/CT slightly overestimated liver AD for resin (3.4 ± 3.4 Gy) and glass (11.5 ± 13.9 Gy) microspheres. Tumor AD was not correlated with baseline or post-SIRT lesion characteristics and no dose–response threshold could be identified. 99mTc-MAA-SPECT/CT dosimetry provides good estimates of AD to tumor and non-tumor liver in UMLM patients treated with first-line SIRT.