Prediction of Declining Renal Function and Albuminuria in Patients With Type 2 Diabetes by Metabolomics

Context: Renal disease in type 2 diabetes mellitus (T2DM) is associated with excess morbidity/mortality. Although estimated glomerular filtration rate (eGFR) and albuminuria are routine for assessing renal impairment, novel biomarkers could improve risk stratification and prediction. Objective: To identify specific biomarkers of progression of renal dysfunction. Design: Prospective observational. Setting: Academic diabetes clinics. Patients: A total of 286 T2DM patients (age, 62 ± 8 y; glycosylated hemoglobin, 7.2 ± 0.9%; eGFR, 85 ± 20 mL · min−1 · 1.73 m2). Interventions: None. Main Outcome Measures: Progression of eGFR and albuminuria. Results: We performed screening metabolomics in serum and urine samples by gas chromatography/mass spectroscopy (MS) and ultra-high performance liquid chromatography/MS/MS. Biomarker identification was performed by random forest using an eGFR cutoff of < 60 mL · min−1 · 1.73 m2 or an albumin/creatinine ratio (ACR) cutoff ≥ 30 mg/g as response variables. At follow-up, eGFR had declined by 16 [9] (median [interquartile ratio]) mL · min−1 · 1.73 m2, and ACR had increased by 41 [135] mg/g in patients in the respective top quartile of changes from baseline. Clinical parameters (gender, age, fasting glucose, and baseline eGFR) predicted outcome, with receiver operator characteristics curve (ROC) = 0.671. The five serum metabolites best correlated with either eGFR < 60 or ACR ≥ 30 at baseline were tested for their ability to improve clinical prediction. The sum of C-glycosyl tryptophan, pseudouridine, and N-acetylthreonine (MetIndex) raised the ROC to 0.739 (P < .0001). eGFR decline was predicted by the top MetIndex quartile (odds ratio = 5.48 [95% confidence interval, 2.23–14.47]). MetIndex also predicted an ACR increase with an odds ratio of 2.82 [1.20–7.03] and a ROC of 0.750. Top urine metabolites did not add significant predictivity. Conclusions: A limited number of circulating intermediates of amino acid and nucleotide pathways carry clinically significant predictivity for deterioration of renal function in well-controlled T2DM.