1459-P: NMR-Based Metabolomics for the Screening of Sleep-Disordered Breathing in Type 2 Diabetes

Abstract

Sleep-disordered breathing (SDB) is common in type 2 diabetes (T2D). This study aimed to develop a screening tool to detect SDB in T2D. Method: SDB was screened in 95 patients by an overnight monitor and diagnosed if apnea-hypopnea index ≥ 5. Anthropometric and HbA1c data, and SDB risk by questionnaire were obtained. Quantitative NMR experiment was performed on a Bruker AVANCE 400 Ascend NMR spectrometer. Carr−Purcell−Meiboom−Gill (CPMG) pulse sequence at 310 K was applied to serum samples to enhance signal-to-noise ratio and hence detect low-molecular weight metabolites. Models based on six machine learning methods (e.g., learned vector quantization, stochastic gradient boosting model, support vector machines, decision tree, partial least squares, generalized linear model, and logistic regression) were constructed to predict SDB, tuned with the CARET R package, and evaluated using three repeats of 10-fold cross validation. Results: Seventy-six patients (80.0%) had SDB. Compared with those without, those with SDB were more likely to be male, have SDB risk, higher BMI and neck circumferences, longer diabetes duration, and higher HbA1c. NMR-based metabolic profiling identified significantly lower levels of branched-chain amino acids (BCAAs) (e.g., L-leucine and L-isoleucine), lactate and threonine metabolites in those with SDB, while level of D-glucose-6 phosphatase was higher than those without SDB. Accuracies for predicting SBD using clinical data alone were ∼80% for each machine learning model. L-leucine and D-glucose-6-phosphatase were identified as being independently associated with SDB using a logistic regression, in addition to clinical factors, with an improved accuracy of 86.3%, and a Cohen’s kappa of 0.429. Conclusions: Due to its high prevalence, clinical factors alone were not sufficient in predicting SDB in T2D. The discovery of metabolic profiles, including BCAAs and those involved in glucose metabolism, could facilitate a more precise and specific SDB prediction. Disclosure K. Lertdetkajorn: None. P. Poungsombat: None. S. Khoomrung: None. J. Phetcharaburanin: None. S. Hongthong: None. C. Kuhakarn: None. N. Siwasaranond: None. A. Manodpitipong: None. H. Ninitphong: Speaker’s Bureau; Self; AstraZeneca, Boehringer Ingelheim Pharmaceuticals, Inc., Merck Sharp & Dohme Corp., Novo Nordisk Inc., Takeda Pharmaceutical Company Limited. S. Saetung: None. B. Ongphiphadhanakul: None. V. Reutrakul: None. S. Reutrakul: None. Funding Endocrine Society of Thailand; Ramathibodi Hospital; Mahidol University; Thailand Center of Excellence for Innovation in Chemistry