Establishment of Reference Intervals for Eight Steroid Hormones in Korean Adults Using LC-MS/MS: Impact of Outlier Exclusion and Statistical Methods.

OBJECTIVE: Reference intervals for steroid hormones can vary significantly depending on the statistical approach used. METHODS: We established reference intervals for eight steroid hormones in 111 Korean females and 113 Korean males aged 18-49 years. Outlier detection was performed using both Tukey's fence criteria and the Reed-Dixon method. Sex- and age-stratified reference intervals were determined using a robust method after exclusion of outliers identified by Tukey's fence criteria. These intervals were compared with those derived using a nonparametric approach. RESULTS: Depending on the hormone, the proportion of outliers identified by Tukey's fence criteria ranged from 0% to 17.6%, whereas the Reed-Dixon method rarely detected extreme outliers. The reference intervals derived using Tukey's fence criteria were generally consistent with previously reported values in Caucasian populations. A total of 12.0% of Korean males had testosterone concentrations below 10.4 nmol/L (300 ng/dL), the cutoff value for low testosterone according to the American Urological Association guidelines. In contrast, the nonparametric method with Reed-Dixon outlier detection yielded a lower reference limit of 0.80 nmol/L for testosterone, which would be considered pathologically low in adult males. CONCLUSIONS: Our study highlights the importance of careful selection of the reference population, outlier exclusion method, and statistical approach when establishing reference intervals for steroid hormones.