Objective: To develop a risk prediction model for early discrimination between transient and permanent congenital hypothyroidism (CH).
Design and setting: In a retrospective cohort, 1047 confirmed CH neonates, from 15 randomly selected provinces in Iran, were entered to the study. Clinical and biochemical information of transient and permanent cases, distinct at the age of 3 years were retrospectively gathered.
Results: Among CH neonates, the overall prevalence of permanent CH was 57.1%. Using forward stepwise multivariable logistic regression analysis, confirmatory venous TSH, total T4 < 8.2 ng/dl, requiring levothyroxine dosage increase, venous TSH ≥ 10 mU/l between 6 and 12 months of age, parental consanguinity and family history of thyroid diseases were associated with increased risk of permanent CH. The prediction model achieved a very good power in discriminating patients with transient and permanent CH with an optimism-corrected area under the ROC curve of 0.86 (95% CI:0.84-0.88) with a very good calibration. Integrated discrimination improvement (IDI) test indicated significantly greater diagnostic performance of the model compared to serum TSH alone.
Conclusions: Using several potential predictors for permanent CH, we developed a relatively powerful risk prediction model as a cost-saving screening tool in order to avoid unnecessary long-term treatment of transient cases which might empower clinicians for prognostication of the CH course and tailoring treatment up to 1 year of age.
Keywords: Clinical prediction model; Congenital hypothyroidism; Diagnosis; Endocrinology-pediatric; Screening; Thyroid function tests.