Thyroid autoimmunity and pregnancy loss: reconciling epidemiological risk, mechanistic heterogeneity, and the failure of universal treatment

Thyroid autoimmunity (TAI), defined principally by thyroid peroxidase or thyroglobulin antibodies, is associated with pregnancy loss in several observational settings. Its interpretation remains uncertain because antibody phenotype, thyroid reserve, reproductive setting, maternal age, embryonic aneuploidy, and outcome ascertainment are incompletely separated. We conducted a structured narrative review of PubMed and the Web of Science Core Collection (through 12 September 2026), prioritising guidelines, systematic reviews, randomised trials, longitudinal cohorts, human tissue studies, and directly relevant experimental evidence. Evidence was organised across spontaneous conception, recurrent pregnancy loss, and assisted reproduction and categorised by proximity to human pregnancy-loss outcomes. Associations were most consistent in clinically enriched populations or when thyroid reserve was limited, but effect estimates varied with assay thresholds, gestational timing, thyroid-function criteria, and analytical denominators. Human mechanistic evidence remains incomplete: immune and endometrial findings are mainly surrogate, whereas direct trophoblast pathways rely largely on animal or in-vitro models. TABLET, POSTAL, and T4LIFE did not show clinically important benefits of routine levothyroxine in broadly defined euthyroid TPOAb-positive populations. We therefore propose a multidimensional, time-indexed model integrating antibody phenotype, longitudinal thyroid function, reproductive load, immune context, and embryo or placental vulnerability. Current evidence supports surveillance and treatment of established thyroid dysfunction, not universal antibody-directed therapy. Future studies require preconception recruitment, harmonised phenotyping, serial endocrine and immune measurements, assessment of embryonic cause, tissue-linked validation, and prospectively enriched trials.