Spectrochemical approach combined with symptoms data to diagnose fibromyalgia through paper spray ionization mass spectrometry (PSI-MS) and multivariate classification

This study performs a chemical investigation of blood plasma samples from patients with and without fibromyalgia, combined with some of the symptoms and their levels of intensity used in the diagnosis of this disease. The symptoms evaluated were: visual analogue pain scale (VAS); fibromyalgia impact questionnaire (FIQ); Hamilton anxiety rating scale (HAM); Tampa Scale for Kinesiophobia (TAMPA); quality of life Questionnaire—physical and mental health (QL); and Pain Catastrophizing Scale (CAT). Plasma samples were analyzed by paper spray ionization mass spectrometry (PSI-MS). Spectral data were organized into datasets and related to each of the symptoms measured. The datasets were submitted to multivariate classification using supervised models such as principal component analysis with linear discriminant analysis (PCA-LDA), successive projections algorithm with linear discriminant analysis (SPA-LDA), genetic algorithm with linear discriminant analysis (GA-LDA) and their versions with quadratic discriminant analysis (PCA/SPA/GA-QDA) and support vector machines (PCA/SPA/GA-SVM). These algorithm combinations were performed aiming the best class separation. Good discrimination between the controls and fibromyalgia samples were observed using PCA-LDA, where the spectral data associated with the CAT symptom achieved 100% classification sensitivity, and associated with the VAS symptom achieved 100% classification specificity, with both symptoms at the moderate level of intensity. The spectral variable at 579 m/z was found to be substantially significant for classification according to the PCA loadings. According to the human metabolites database, this variable can be associated with a LysoPC compound, which comprises a class of metabolites already evidenced in other studies for fibromyalgia diagnosis. This study proposed an investigation of spectral data combined with clinical data to compare the classification ability of different datasets. The good classification results obtained confirm this technique is as a good analytical tool for the detection of fibromyalgia, and provides theoretical support for other studies about fibromyalgia diagnosis.