Establishing continuous reference intervals for thyroid function tests

This study aimed to establish age- and gender-specific reference intervals for thyroid function tests-fT3 (pmol/L), fT4 (pmol/L), and TSH (μIU/L)-in both pediatric and adult populations, and for anti-TPO (kIU/L) only in adults. It also compared partitioned and continuous modeling approaches to determine the most appropriate method for clinical interpretation. In this retrospective observational study, laboratory data from individuals who visited Sivas Cumhuriyet University Hospital between 2017 and 2022 were analyzed. Partitioned reference intervals were determined using the refineR method. Continuous reference intervals were modeled using Generalized Additive Models for Location, Scale, and Shape (GAMLSS), including LMS (Box-Cox Normal), LMST (Box-Cox t), and LMSP (Box-Cox Power Exponential) distributions. Model selection was based on Generalized Akaike Information Criterion (GAIC) values, and gender differences were evaluated using parametric and non-parametric tests.

In pediatric subjects, hormone distributions largely overlapped between sexes during early childhood, whereas significant gender differences in fT3 appeared after approximately age 12 (p < 0.001). In adults, males generally showed higher fT3 concentrations than females, while sex differences in fT4 were less consistent across ages. TSH levels showed minimal gender-related variation in both pediatric and adult groups. According to GAIC values, the LMSP model provided the best fit for thyroid function parameters in both pediatric and adult datasets. Continuous reference interval modeling captured gradual age-related changes and provided a flexible framework for evaluating thyroid function.