Statistical methods for assessing the reproducibility of biomarkers.

Tryposkiadis, Konstantinos (2023). Statistical methods for assessing the reproducibility of biomarkers. University of Birmingham. Ph.D.

Preview

Tryposkiadis2023PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (7MB) | Preview

Abstract

Introduction
Biomarkers are often subject to measurement error, affecting their reproducibility. Statistical methods for examining the reproducibility of continuous biomarkers are well-developed. However, methods for performing systematic reviews and meta-analyses of biomarker reproducibility, and examining the reproducibility of count-based biomarkers are under-researched.

Aims
To propose methods for performing meta-analyses of reproducibility of continuous measurements, and for examining the reproducibility of count measurements.

Methods
Current methods for performing systematic reviews and meta-analyses of reproducibility of continuous measurements were systematically identified and critiqued. Meta-analytic methods were developed and evaluated in a case study examining the reproducibility of grip strength measured in different populations. Methods for count outcomes were evaluated in a case study of a biopsy-based biomarker for Sjorgren’s syndrome. Simulation extended the case study to examine how reproducibility varied for larger numbers of biopsied samples.

Results
Methods for meta-analysis performed well, indicating that grip strength measurements are reproducible. Count-based methods outperformed continuous-based methods when examining the reproducibility of biopsy-based count measurements. The methods indicated low reproducibility, with simulation suggesting improved reproducibility for larger samples.

Conclusion
The proposed methods provide robust summary and primary evidence of the reproducibility of continuous and count-based biomarkers, respectively, allowing better decisions regarding their use in practice.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):