Variation in serum biomarkers with sex and female hormonal status: implications for clinical tests.

Brenda W. J. H. Penninx,Jordan M. Ramsey,Sabine Bahn,Jason D. Cooper

doi:10.1038/srep26947

Brenda W. J. H. Penninx, Jordan M. Ramsey + Show 2 more

Open Access

https://doi.org/10.1038/srep26947

Copy DOI

Abstract

Few serum biomarker tests are implemented in clinical practice and recent reports raise concerns about poor reproducibility of biomarker studies. Here, we investigated the potential role of sex and female hormonal status in this widespread irreproducibility. We examined 171 serum proteins and small molecules measured in 1,676 participants from the Netherlands Study of Depression and Anxiety. Concentrations of 96 molecules varied with sex and 66 molecules varied between oral contraceptive pill users, postmenopausal females, and females in the follicular and luteal phases of the menstrual cycle (FDR-adjusted p-value <0.05). Simulations of biomarker studies yielded up to 40% false discoveries when patient and control groups were not matched for sex and up to 41% false discoveries when premenopausal females were not matched for oral contraceptive pill use. High accuracy (over 90%) classification tools were developed to label samples with sex and female hormonal status where this information was not collected.

Highlights

Few serum biomarker tests are implemented in clinical practice and recent reports raise concerns about poor reproducibility of biomarker studies
We used serum samples collected from the Netherlands Study of Depression and Anxiety (NESDA) to investigate molecular concentrations in males and females taking the oral contraceptive pill (OC), in the follicular and luteal phases of the menstrual cycle, and after menopause
Similar classification accuracies were achieved for this reduced classifier (91.4% overall accuracy in the validation cohort; Fig. 5A and Supplementary Table 6C) compared to the first classifier constructed from the full set of 189 predictors. This investigation represents a substantial contribution to our understanding of sex and female hormonal status (including use of the oral contraceptive pill (OC), menopause, and menstrual cycle phase) as sources of variation in serum molecular concentrations and their considerable impact on biomarker studies

Summary

Introduction

Few serum biomarker tests are implemented in clinical practice and recent reports raise concerns about poor reproducibility of biomarker studies. Failing to account for sex and female hormonal status as important sources of variability in the concentrations of serum molecules may lead to confounding of these variables with disease status and reduce power to detect differences, contributing to the poor performance of biomarker tests in initial and follow-up studies. We further evaluated the implications of not accounting for sex-based differences in biomarker studies by determining the potential number of false positive findings caused by confounding of disease status with sex or female use of OCs. we developed a classifier to label samples as males, OC users, postmenopausal females, or females with a menstrual cycle for retrospective analysis of biomarker studies where these variables were not initially recorded. ® Human DiscoveryMap platform were examined, which is a pre-selected multiplex immunoassay panel that has www.nature.com/scientificreports/

Methods

Results

Conclusion