Abstract
RATIONALE & OBJECTIVE: We aimed to find predictive plasma and urine metabolites for nondiabetic chronic kidney disease (CKD), and to validate these biomarkers in a diabetic kidney disease (DKD) population, using data of the population-based multiethnic Healthy Life in an Urban Setting study. STUDY DESIGN: Cross-sectional metabolome study. SETTING & PARTICIPANTS: From the Healthy Life in an Urban Setting population-based cohort, we included 124 participants with nondiabetic CKD, 45 with DKD and 200 healthy controls. PREDICTORS: Plasma and urine metabolites were measured using ultra-high-performance liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS) with an untargeted approach. OUTCOMES: (Nondiabetic) CKD. ANALYTICAL APPROACH: We used machine learning models to predict nondiabetic CKD from metabolite profiles and used logistic regression models with adjustment for potential confounders to verify our results in the best predicting metabolites. In addition, we assessed the associations between the best predicting metabolites and DKD. RESULTS: Urine metabolites were more predictive of nondiabetic kidney disease than plasma metabolites. In plasma, the best predicting metabolites for nondiabetic CKD included many amino acids, including N-acetylated amino acids, histidine, and indolepropionate. In urine, the highest-ranked metabolites were predominantly lipids, including sphingomyelins and phosphatidylcholines. There was limited overlap among the top-ranked metabolites in predicting nondiabetic CKD between plasma and urine. Almost all associations with nondiabetic CKD could be translated to DKD. No interactions were observed with ethnicity. LIMITATIONS: The cross-sectional design limits causal inference. CONCLUSIONS: Our analyses revealed that urine metabolites were strongly associated with CKD than plasma metabolites in this multiethnic population. The finding that specific associations of plasma and urine metabolites could be translated to subjects with DKD suggests a shared pathophysiological background.