Personalised Reference Intervals for integrated Biomarkers through nEw statistical methods and unique longitudinal data'. (R-10165)

Clinicians use reference intervals (RIs) for interpreting various test results of patients, be it clinical biochemistry (e.g. HDL cholesterol) or physiological (e.g. BMI) results. Whether test results lie within or outside an established RI determines diagnostic and treatment decisions. The current RI have an interpretation that refers to the distribution of the test results within a large population. However, from the precision medicine perspective, current intervals reflect population level averages and is not suited to be used in a preventive medicine setting. The interpretation of these tests could be made more customized if we can implement RIs that reflect the test result distribution of the patient in question. In a longitudinal dataset, collected by VITO, we see indeed large variability between subjects and rather small variability within subjects. In this project, we therefore propose to construct individual RIs (IRI). We propose robust estimation methods based on quantile regression models and empirical Bayes methods. The latter will allow for borrowing strength between subjects, even for the estimation of subject-specific IRIs. We will provide methodological and software solutions that will allow researchers and clinicians to better understand the variability of biomarker data and use the IRIs especially to follow-up patients in a preventive medicine setting. The unique longitudinal dataset from VITO will be used in the implementation and evaluation phases.

Keywords:ASYMPTOTIC THEORY

Disciplines:Statistics, Development of bioinformatics software, tools and databases, Protein diagnostics, Biomarker discovery and evaluation not elsewhere classified