Objective: Adequate thyroid hormone availability is essential for fetal growth and development. Selecting pregnant women at risk for thyroid dysfunction is complicated by changes in maternal physiology. Reference intervals (RI) should be pregnancy-specific in order to identify women with thyroid disease. A recent meta-analysis from our group has shown major methodological differences in published papers, but the consequences of these varying methodologies are largely unknown.

Methods: The study was performed in the Consortium on Thyroid and Pregnancy. In line with international guidelines, cohort-specific RIs based on the 2.5^th and 97.5^th percentiles were calculated after exclusion of participants with known thyroid disease or medication, twin pregnancy or TPOAb positivity. To evaluate the effect of current recommendations and additional exclusion criteria, the RIs were also calculated using the above mentioned methods 1) using the 5^th to 95^th percentiles, 2) without excluding TPOAb positivity and using additional exclusion criteria defined as 3) exclusion of TgAb 4) diabetes mellitus, 5) chronic hypertension, 6) obesity 7) active smoking and 8) any pregnancy complications.

Results: The final study population was N=69236 from 20 cohorts. The use of the 95^th percentile led to a considerable decrease in the upper limits, varying from -0.13 to -1.24 mU/L for TSH and up till -5.22 ng/dL for FT4. Not excluding TPOAb positivity led to an increase in the upper limit of TSH in most cohorts, ranging from -0.07 to +1.32. The additional exclusion of TgAb positivity caused minor changes in the TSH upper limit, with a small majority showing a decrease in the upper limit of TSH. All other additional exclusion criteria led to minor changes in the reference limits.

Conclusion: Additional exclusions frequently encountered in literature do not affect TSH or FT4 RIs during pregnancy. This indicates that the majority of published studies can be implemented into clinical practice despite methodological differences and future studies can adapt simplified study setups to define valid RIs.