Mid-infrared spectroscopy of serum, a promising non-invasive method to assess prognosis in patients with ascites and cirrhosis

• Published in: PloS one 2017-10, Vol.12 (10), p.e0185997-e0185997
• Main Authors: Le Corvec, Maëna, Jezequel, Caroline, Monbet, Valérie, Fatih, Nadia, Charpentier, Frédéric, Tariel, Hugues, Boussard-Plédel, Catherine, Bureau, Bruno, Loréal, Olivier, Sire, Olivier, Bardou-Jacquet, Edouard
• Format: Article
• Language: English
• Subjects: Algorithms; Analgesics; Analysis; Area Under Curve; Ascites; Ascites - blood; Ascites - diagnosis; Ascites - mortality; Biology and Life Sciences; Biomarkers - blood; Cancer; Cirrhosis; Computer simulation; Diagnosis; Feasibility studies; Female; Fibrosis - blood; Fibrosis - diagnosis; Fibrosis - mortality; Fourier transforms; Health aspects; Humans; Hypertension; Infrared spectra; Infrared spectroscopy; Liver cirrhosis; Liver diseases; Liver transplants; Male; Mathematics; Medical diagnosis; Medical prognosis; Medicine and Health Sciences; Metabolism; Methods; Middle Aged; Mortality; Patients; Physical Sciences; Prognosis; Regression analysis; Reproducibility of Results; Research and Analysis Methods; Risk Assessment; Risk factors; ROC Curve; Sensitivity and Specificity; Serum - metabolism; Spectrophotometry, Infrared - methods; Spectroscopy; Spectrum analysis; Statistics; Survival; Transplants & implants
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0185997

Prognostic tests are critical in the management of patients with cirrhosis and ascites. Biological tests or scores perform poorly in that situation. Mid-infrared fibre evanescent wave spectroscopy (MIR-FEWS) which allows for global serum metabolic profiling may provide more relevant information by measuring a wider range of metabolic parameters in serum. Here we present the accuracy of a MIR-FEWS based predictive model for the prognosis of 6 months survival in patients with ascites and cirrhosis. Patients with ascites were prospectively included and followed up for 6 months. MIR-FEWS spectra were measured in serum samples. The most informative spectral variables obtained by MIR-FEWS were selected by FADA algorithm and then used to build the MIR model. Accuracy of this model was assessed by ROC curves and 90%/10% Monte Carlo cross-validation. MIR model accuracy for 6 months survival was compared to that of the Child-Pugh and MELD scores. 119 patients were included. The mean age was 57.36±13.70, the MELD score was 16.32±6.26, and the Child-Pugh score was 9.5±1.83. During follow-up, 23 patients died (20%). The MIR model had an AUROC for 6 months mortality of 0.90 (CI95: 0.88-0.91), the MELD 0.77 (CI95: 0.66-0.89) and Child-Pugh 0.76 (CI95: 0.66-0.88). MELD and Child-Pugh AUROCs were significantly lower than that of the MIR model (p = 0.02 and p = 0.02 respectively). Multivariate logistic regression analysis showed that MELD (p