The type 2 deiodinase and its Thr92Ala-DIO2 polymorphism have been linked to clinical outcomes in acute lung injury and coronavirus disease 2019 (COVID-19). The objective was to identify a potential association between Thr92Ala-DIO2 polymorphism and body composition (appendicular muscle mass, myosteatosis, and fat distribution) and to determine whether they reflect the severity or mortality associated with the disease. In this prospective cohort study (June-August 2020), 181 patients hospitalized with moderate-to-severe COVID-19 underwent a non-contrast-enhanced computed tomography (CT) of the thorax to assess body composition, laboratory tests, and genotyping for the Thr92Ala-DIO2 polymorphism. In total, 181 consecutive patients were stratified into three subgroups according to the genotype: Thr/Thr (n = 64), Thr/Ala (n = 96), and Ala/Ala (n = 21). The prevalence of low muscle area (MA) (< 92 cm²) was 52.5%. Low MA was less frequent in Ala/Thr patients (44.8%) than in Thr/Thr (60.9%) or Ala/Ala patients (61.9%) (P = 0.027). Multivariate logistic regression analysis confirmed that the Thr/Ala allele was associated with a reduced risk of low MA (41% to 69%) and myosteatosis (62% to 72%) compared with Thr/Thr + Ala/Ala (overdominant model). Kaplan-Meier curves showed that patients with low muscle mass and homozygosity had lower survival rates than the other groups. Notably, the heterozygotes with MA ≥92 cm² exhibited the best survival rate. Thr92Ala-DIO2 heterozygosity is associated with increased skeletal MA and less myosteatosis in patients with COVID-19. The protective effect of Thr92Ala-DIO2 heterozygosity on COVID-19 mortality is restricted to patients with reduced MA.