Statistical analysis

SPSS 22 was used for statistical analysis, with patient as the unit of analysis. Demographic and baseline characteristics were compared between groups using one-way ANOVA, Kruskal–Wallis ANOVA and chi-square test. Live birth rate, clinical and ongoing pregnancy and pregnancy loss were determined for the patients’ first six cycles of IUI or fewer cycles in case the first ongoing pregnancy was achieved earlier.

Logistic regression was used with the natural logarithm of TSH as a continuous predictor. The first model contained only a linear term. In a second analysis, a quadratic term was added to account for a non-linear relationship. In a final analysis, patients were categorized based on TSH levels, using the quartiles of the distribution (0.3–1.21 mIU/L; 1.22–1.75 mIU/L; 1.76–2.34 mIU/L; 2.35–4.5 mIU/L). Differences in pregnancy outcomes were tested using logistic regression analysis with pregnancy outcome as the dependent variable and TSH group as predictor.

Adjusted analyses were performed in which we corrected for: age, BMI, use of alcohol and tobacco (Knudsen et al., 2005; Hornstein, 2016), use of gonadotrophins, semen count, presence of diminished ovarian reserve defined as FSH > 10 IU/L with an ovulating cycle, presence of unexplained subfertility, and primary or secondary subfertility. Categories were as follows: BMI < 20.9/21.0–28.9/29–34.9/>35 kg/m² (van der Steeg et al., 2008); use of alcohol: no alcohol (0 units a week)/ moderate (1–8 units a week)/ heavy (>8 units a week) (Rachdaoui and Sarkar, 2013; Health Council of the N, 2015; Donald et al., 2018); use of tobacco yes/no (Wiersinga, 2013); use of gonadotrophin yes/no, washed semen count 0.5–2.9 × 10⁶/3–4.9 × 10⁶/5– 29.9 × 10⁶/30–50 × 10⁶, diminished ovarian reserve yes/no, unexplained subfertility yes/no and primary/secondary subfertility.

Odds ratios were reported as effect–size together with their 95% CI and P-values. With the given group size of 920 patients, a pairwise comparison between the four groups with an alpha of 0.05 and a beta of 0.2 allowed proportions of > 7.25% to be distinguished.