Early warning signs of thyroid autoantibodies seroconversion: A retrospective cohort study

Serum anti-thyroid peroxidase antibody (anti-TPO) and anti-thyroglobulin antibody (anti-Tg) levels are key indicators for the diagnosis of autoimmune diseases, especially autoimmune thyroiditis. Before the thyroid autoantibodies turn from negative to positive, it is unknown whether any clinical indicators in the body play a warning role. To establish an early prediction model of seroconversion to positive thyroid autoantibodies. This retrospective cohort study collected information based on clinical laboratory data. A logistic regression model was used to analyse the risk factors associated with a change in thyroid autoantibodies to an abnormal status. A machine-learning approach was employed to establish an early warning model, and a nomogram was used for model performance assessment and visualisation. Receiver operating characteristic (ROC) curves, calibration curves, and decision curve analyses were used for internal and external validation. Logistic regression analysis revealed that albumin to globulin ratio, triglyceride levels, and Glutamic acid levels among liver function and some metabolism-related indicators, high density lipoprotein C among metabolism-related indicators, and cystatin C among renal function indicators were all risk factors for thyroid antibody conversion (P < 0.05). In addition, several indicators in the blood count correlated with thyroid conversion (P < 0.05). Changes in the ratio of free thyroxine to free triiodothyronine were a risk factor for positive thyroid antibody conversion (ORfT4/fT3 = 1.763; 95% confidence interval 1.554–2.000). The area under the curve (AUC) of the early warning model based on the positive impact of clinical laboratory indicators, age, and sex was 0.85, which was validated by both internal (AUC 0.8515) and external (AUC 0.8378) validation. The early warning model of anti-TPO and anti-Tg conversion combined with some clinical laboratory indicators in routine physical examination has a stable warning efficiency.