Background the tear fluid is a valuable non-invasive source of biomarkers for objective analysis of systemic diseases. 

Aim to investigate whether specific metabolites were differentially expressed in tears of allergic asthmatic children and controls (CTRLs), and make a comparison with their urinary metabolome.

Methods a prospective study including 29 prepubertal children with allergic asthma and 12 CTRLs was carried out. Tear and urine samples were analyzed by targeted metabolomics for the determination of amino acids, carnitines and other specific metabolites, and classifiers were built using machine-learning algorithms. Results feature selection methods showed that six carnitines (C) (C5.1, C6-DC, C10.1, C10.2, C18, C22) in tears and three (C5.1, C6-DC, C24) in urine were relevant. In tears, a negative correlation between short-chain carnitines (C5.1, C6-DC) and forced expiratory volume in 1 s (FEV1)/forced vital capacity (FVC) ratio (rho=-0.32, p=0.040; rho=0.32, p=0.041, respectively), and a positive correlation between C18 and FEV1/FVC (rho=0.38, p=0.016) were found. Differently, no correlation was found between metabolites in urine and respiratory parameters. Finally, random forest models built including metabolomic and clinical data were able to classify the asthmatics with an 95.8% accuracy in tears and an 97.6% accuracy in urine. 

Conclusions abnormal mitochondrial and peroxisomal fatty acid oxidation might be pathways involved in asthma. Metabolomics in tears might be suitable for discriminating between asthmatics and CTRLs. Future studies are needed to elucidate the aetiological of carnitine in asthma as well as its potential therapeutic use.