Metabolic phenotypes and vitamin D response in the critically ill: A metabolomic cohort study.

Background & Aims: Although vitamin D deficiency is common in critically ill patients, randomized controlled trials fail to demonstrate benefits of supplementation. We aimed to identify distinct vitamin D responsive metabolic phenotypes prior to trial intervention of high-dose vitamin D by applying machine learning clustering method to metabolomics data from the Correction of Vitamin D Deficiency in Critically Ill Patients (VITdAL-ICU) trial.

Methods: In the randomized, placebo-controlled VITdAL-ICU trial, critically ill adults received placebo or high-dose vitamin D. To distinguish vitamin D responsive metabolic phenotypes prior to intervention, we implemented consensus clustering with partitioning around medoids algorithm to the plasma metabolome data before randomization. Individual metabolite differences were determined utilizing linear mixed-effects regression models stratified for metabolomic phenotypes with false discovery rate adjustment. The association between vitamin D supplementation and 180-day mortality was evaluated in each metabolic phenotype, applying multivariable logistic regression analysis.

Results: In 453 critically ill adults, the study identified 4 distinct metabolic phenotypes (clusters A. N = 134; B. N = 123; C. N = 92; D. N = 104). We found differential metabolic pathway patterns in the four clusters. Specifically, branched chain amino acid catabolic metabolites, long-chain acylcarnitines and diacylglycerol species are significantly increased in a specific metabolic phenotype (cluster D) following high-dose vitamin D. Further, in cluster D high-dose vitamin D supplementation had a significantly lower adjusted odds of 180-day mortality after controlling age, sex, Simplified Acute Physiology Score II, admission diagnosis, and baseline 25-hydroxyvitamin D (OR 0.28 (95%CI, 0.09-0.89); P = 0.03). In metabotype A, B, and C, high-dose vitamin D supplementation was not significantly associated with lower 180-day mortality following multivariable adjustment.

Conclusion: In this post-hoc cohort study of the VITdAL-ICU trial, the clustering analysis of plasma metabolome data identified biologically distinct metabolic phenotypes. Among clusters, we found the different associations between high-dose vitamin D supplementation and specific metabolite pathways as well as 180-day mortality. Our findings facilitate further research to validate metabolic phenotype-targeted strategies for critical illness treatments.