Virtual diagnosis of diabetic nephropathy using metabolomics in place of kidney biopsy: The DIAMOND study

Aims: To explore the clinical factors and urinary metabolites that predict biopsy-confirmed diabetic nephropathy (DN) in patients with type 2 diabetes mellitus (T2DM).Methods: Data from the medical records of 126 patients with T2DM who underwent kidney biopsy between January 2010 and October 2020 at a single-center were retrospectively reviewed to investigate the clinical factors that predict DN. Urine samples were collected to perform urine metabolomics in patients with T2DM divided by biopsy-confirmed DN, immunoglobulin A, and membranous nephropathy, and a control group of healthy participants. Each group comprised 11 age- and sex-matched participants. A prediction model was developed using a combination of clinical factors and urinary metabolites, and a multivariate receiver operating characteristic (ROC) analysis was conducted.Results: Age, presence of proliferative diabetic retinopathy, T2DM duration, and hemoglobin A1c levels were clinical factors predictive of DN. Four urinary metabolites (alanine, choline, N-phenylacetylglycine, and trigonelline) had variable importance in projection scores > 1 and were predictive of DN. When conducting multivariate ROC analysis with a combination of clinical factors and urinary metabolites, the area under the curve was 1.000. Conclusions: The combination of clinical factors and urinary metabolites is highly valuable for predicting biopsyconfirmed DN in patients with T2DM.