Diagnosis of post-surgical fine-needle aspiration biopsies of thyroid lesions with indeterminate cytology using HRMAS NMR-based metabolomics

• Published in: Metabolomics 2018-10, Vol.14 (10), p.141-10, Article 141
• Main Authors: Rezig, Lamya, Servadio, Adele, Torregrossa, Liborio, Miccoli, Paolo, Basolo, Fulvio, Shintu, Laetitia, Caldarelli, Stefano
• Format: Article
• Language: English
• Subjects: Alanine; Benign; Biochemistry; Biomedical and Life Sciences; Biomedicine; Biopsy; Biopsy, Fine-Needle; Cell Biology; Cellular biology; Citric acid; Cytology; Developmental Biology; Diagnosis; Female; Humans; Inositol; Leucine; Life Sciences; Male; Malignancy; Metabolites; Metabolomics; Molecular Medicine; Multivariate Analysis; NMR; Nuclear magnetic resonance; Nuclear Magnetic Resonance, Biomolecular; Original Article; Phenotypes; Phenylalanine; Serine; Surgery; Thyroid cancer; Thyroid gland; Thyroid Neoplasms - diagnosis; Thyroid Neoplasms - metabolism; Thyroid Neoplasms - surgery; Thyroid Nodule - diagnosis; Thyroid Nodule - metabolism; Thyroid Nodule - surgery; Thyroidectomy; Tumors; Tyrosine; Ultrasound
• ISSN: 1573-3882; 1573-3890
• DOI: 10.1007/s11306-018-1437-6

Introduction Ultrasound examination coupled with fine-needle aspiration (FNA) cytology is the gold standard for the diagnosis of thyroid cancer. However, about 10–40% of these analyses cannot be conclusive on the malignancy of the lesions and lead to surgery. The cytological indeterminate FNA biopsies are mainly constituted of follicular—patterned lesions, which are benign in 80% of the cases. Objectives The development of a FNAB classification approach based on the metabolic phenotype of the lesions, complementary to cytology and other molecular tests in order to limit the number of patients undergoing unnecessary thyroidectomy. Methods We explored the potential of a NMR-based metabolomics approach to improve the quality of the diagnosis from FNABs, using thyroid tissues collected post-surgically. Results The NMR-detected metabolites were used to produce a robust OPLSDA model to discriminate between benign and malignant tumours. Malignancy was correlated with amino acids such as tyrosine, serine, alanine, leucine and phenylalanine and anti-correlated with myo-inositol, scyllo-inositol and citrate. Diagnosis accuracy was of 84.8% when only indeterminate lesions were considered. Conclusion These results on model FNAB indicate that there is a clear interest in exploring the possibility to export NMR metabolomics to pre-surgical diagnostics.