Use of data mining in the establishment of age-adjusted reference intervals for parathyroid hormone.

Background-aim: Appropriately defined reference values are of vital importance for the correct interpretation of laboratory results. However, the implementation of in-house reference intervals by traditional methods is rare due to difficulties in the selection of reference individuals, so indirect methods are often alternatively used. Parathyroid hormone (PTH) is a crucial hormone for the metabolism of calcium and phosphorus whose result can lead to erroneous diagnoses and medical actions if adequate reference intervals are not readily available. Our goal was to obtain reference values ​​for adult population for PTH by using data mining. Based on possible new reference intervals, we also aimed to perform a retrospective evaluation of the individuals mistakenly classified.

Methods: Retrospective observational study between 2014 and 2019. All requests from individuals aged ≥ 18 years were considered if PTH was requested together with serum creatinine, 25-hydroxyvitamin D, calcium, albumin and phosphate (within their respective reference intervals). PTH was measured on the Cobas e411 platform (Roche). Requests were grouped by age and sex. Differences among partitions were assessed by the Harris-Boyd's test and reference intervals were determined by the non-parametrical calculation of percentiles 2.5 and 97.5.

Results: A total of 2279 laboratory requests were included. Reference intervals for PTH were seen to be sex-independent, but age-dependent reference intervals were needed: 18-39, 40-59 and >60 years. Based on the reference intervals obtained, up to 20.1% of individuals were misclassified.

Conclusions: Appropriateness of PTH reference intervals would lead to a reduction in the number of additional tests and avoid wrong medical actions, thus improving not only patient safety, but also the healthcare system as a whole.