Objective: The need for personalization of the reference values of thyroid function tests has been previously suggested. We aimed at determining TSH reference values in a large cohort of children according to age, sex, BMI, and ethnicity. Design: A population-based cohort study. Methods: The study cohort included 75 549 healthy children aged 5-18 years. Data analyzed included age, gender, TSH, FT4 levels, BMI and ethnicity. Multivariate logistic regression analysis examined the associations between the study parameters. Results: TSH in the Jewish population is lower than in the non-Jewish population (median: 2.1 IU/L (IQR: 1.5) vs 2.2 IU/L (IQR: 1.5), P < 0.0001). TSH is significantly affected by BMI for children defined as underweight, normal weight, overweight or obese, levels increased as weight diverged from the normal range (median levels: 2.1 IU/L (IQR: 1.4), 2.0 IU/L (IQR: 1.3), 2.1 IU/L (IQR: 1.4), 2.4 (IQR: 1.5), respectively, P < 0.001). The 2.5 percentile is affected by gender and BMI (P < 0.02 and P < 0.001, respectively), while the 97.5 percentile is affected by ethnic origin and BMI (P < 0.001 for both). New TSH reference intervals (RI) adjusted according to BMI and ethnicity are suggested. Comparison of the old and new RI demonstrate the significance of RI personalization: 25.1% of the children with TSH levels above the old RI are within the new RI, while 2.3% of the children who were in the old RI are below the new RI. Conclusions: TSH reference values in children are affected by BMI and ethnicity. Reference values should be individualized accordingly to improve future clinical decision-making and treatment.