Establishing and evaluating pediatric thyroid reference intervals on the Roche Modular Analytics E 170 using computational statistics and data-mining techniques

Abstract

Objectives This study determined age specific, free thyroxine and thyroid stimulating hormone reference intervals for a pediatric population. Methods The data set was partitioned into age groups and the Tukey method was used to identify outliers for removal. The Harris and Boyd algorithm was used to determine if groups could be merged. Reference intervals were calculated using both non-parametric and robust methods. Boot-strapping was used to determine the 90% confidence intervals surrounding the upper and lower limits of each reference interval. These population based reference intervals were then compared to the manufactures suggested reference intervals. Results This study provides non-parametric and robust reference intervals with associated 90% confidence intervals for free thyroxine and thyroid stimulating hormone as measured on the Roche Modular Analytics E 170. Conclusions The results emphasize the importance of establishing population-based reference intervals for the clinical laboratory.