Abstract
Early diagnosis of thyroid disorders is key to further treatment. We assessed the ability of a high-throughput proton NMR metabolomic profile to distinguish disease type amongst of Graves’ disease (n=87), Hashimoto’s thyroiditis (n=17), toxic goiter (n=11), and autoimmune thyroiditis [i. e., subacute thyroiditis (n=4), postpartum thyroiditis (n=1)]. This observational study was conducted investigating patients presenting with a thyroid disorder at a Swiss hospital endocrine referral center and an associated endocrine outpatient clinic. The main outcome was diagnosis of thyroid disorder based on classical parameters. Blood draws took place as close as possible to treatment initiation. We performed one-way ANOVA and partial least squares discriminant analysis (PLS-DA) as multivariate classification and feature ranking method. One-way ANOVA analysis yielded following significantly different metabolites, triglycerides in small VLDL, triglycerides in very small VLDL, and triglycerides in large LDL (FDR=0.04). There was no distinct separation of any of the 4 diagnoses by PLS-DA. We did not find a metabolomic biomarker combination capable of predicting diagnosis. Preanalytical issues might have influenced our results. We strongly suggest replicating our work in another cohort.
Key words
metabolomics - Graves’ disease - Hashimoto’s disease - diagnosis
