Kordonouri et al. (1) provide interesting information regarding the frequency of thyroid autoimmunity in pediatric-aged patients with type 1 diabetes. Their recommendations of “yearly examinations of thyroid antibodies” and “in cases of thyroid antibody positivity, thyroid function tests and ultrasound assessment” “to minimize the risk of undiagnosed hypothyroidism in young patients with type 1 diabetes” are not supported by the data. If the goal is to detect subjects with hypothyroidism, then as shown in Table 1 of their study, 15.8% (241) of 1,530 patients who were thyroid antibody positive had an elevated thyroid stimulating hormone (TSH) and would be considered true positives; 7.8% (434) of 5,567 thyroid antibody–negative patients had an elevated TSH and would be considered false negatives. Thus, the sensitivity [true positives/(true positives + false negatives)] for thyroid antibody testing equals 35%. There were 5,133 antibody-negative patients with normal TSH values (true negatives) and 1,289 thyroid antibody–positive TSH-normal patients. Thus, the specificity [true negatives/(true negatives + false positives)] for thyroid antibody testing in their study was 80%. In regard to patients requiring thyroxine treatment, 10.6% (162) of the antibody-positive patients were true positives (antibody positive and thyroxine treated) and 0.6% (33) of the antibody-negative patients were false negatives (antibody negative and thyroxine treated). In addition, there were 5,534 true negatives (antibody negative, no treatment) and 1,388 false positives (antibody positive, no treatment). Thus, antibody testing was 83% sensitive and 80% specific.
Since there is no proven benefit in treating antibody-positive patients with normal TSH levels (2), and since screening tests should be highly sensitive, the data actually support yearly primary TSH screening with possible secondary antibody testing.
Address correspondence to Robert P. Hoffman, MD, Children’s Hospital ED541, 700 Children’s Dr., Columbus, OH 43205. E-mail: firstname.lastname@example.org.