# Adaptive reference ranges: From A to Z

## TL;DR

This paper compares two methods for creating adaptive reference ranges in personalized medicine and sports science, highlighting their strengths and trade-offs.

## Contribution

The study provides practical insights into selecting between Z-score and linear mixed-effects models for adaptive monitoring.

## Key findings

- The Z-score method favors specificity in adaptive reference ranges.
- Linear mixed-effects models prioritize sensitivity and handle missing data better.
- Linear mixed-effects models offer flexibility by incorporating population data and covariates.

## Abstract

Clinical reference ranges are fundamental in medical diagnostics, offering critical benchmarks for interpreting laboratory test results. Adaptive reference ranges, in particular, are essential for personalised monitoring, as they enable the detection of abnormal values by accounting for individual variability over time. This paper compares two key approaches for generating adaptive reference ranges: the Z-score method and the linear mixed-effects modelling framework. Through simulation studies and real data applications, we provide practical insights into selecting the most appropriate methods for adaptive monitoring in personalised medicine and sport science. Our findings highlight the trade-offs between these approaches, with the Z-score method favouring specificity, while the linear mixed-effects model prioritises sensitivity and offers greater flexibility by incorporating population-level data, accommodating covariates, and effectively handling missing data.