Subjects were divided equally into four groups (
quartiles) based on thyroid laboratory testing as follows: thyroid-stimulating hormone (TSH, [mu]IU/mL)
quartiles, Q1 (TSH < 1.30), Q2 (1.30 [less than or equal to] TSH <1.90), Q3 (1.90 [less than or equal to] TSH < 2.75), and Q4 (TSH [greater than or equal to] 2.75); free thyroxine (free T4, ng/dL)
quartiles, Q1 (free T4 < 1.15), Q2 (1.15 [less than or equal to] free T4 < 1.26), Q3 (1.26 [less than or equal to] free T4 < 1.38), and Q4 (free T4 [greater than or equal to] 1.38); and triiodothyronine (T3, pg/mL)
quartiles, Q1 (T3 < 2.87), Q2 (2.87 [less than or equal to] T3 < 3.12), Q3 (3.12 [less than or equal to] T3 < 3.40), and Q4 (T3 [greater than or equal to] 3.40).
Metformin reduced the rate of developing diabetes only among patients in the highest risk
quartile, whose mean 3-year diabetes risk averaged 60% (placebo rate=59.6%; metformin rate=38.2%; mean absolute risk reduction [ARR]=21.4%; 3-year number needed to treat [NNT]=4.6).