Risk stratification in MASLD patients using non-invasive approaches: application of electronic multi-sensoring technologies

Background and aims: Fibrotic MASH, the inflammatory form of MASLD with significant activity and fibrosis (necroinflammatory activity -NAS score- ≥4 and fibrosis ≥2), and advanced fibrosis (fibrosis stage ≥3) represent the most treatment-requiring and prognostically relevant phenotypes of MASLD. For this reason, much research is focusing on identifying non-invasive tools capable of diagnosing this condition. In the last years, multisensory analysis techniques are emerging as diagnostic/prognostic tools potentially useful in various disorders, including liver diseases. This thesis aimed to evaluate the diagnostic performance of these technologies for risk stratification in MASLD, with particular focus on the identification of patients with fibrotic MASH and significant/advanced fibrosis. Methods: This was a prospective, exploratory feasibility study conducted at the Fondazione Policlinico Universitario Campus Bio-Medico di Roma. Consecutive patients with MASLD or metabolic risk factors were enrolled between December 2022 and December 2025. Two cohorts were defined: a non-invasive cohort, in which fibrotic MASH and fibrosis were assessed using the FAST score and liver stiffness measurement (LSM) by FibroScan®, and a histological cohort, in which disease was characterized by liver biopsy. All participants underwent breath, urine, and saliva sampling, analyzed using the BIONOTE e-nose and e-tongue systems, respectively. Diagnostic performance was evaluated using elastic net-regularized logistic regression with 10-fold stratified cross-validation, and reported as AUC, accuracy, sensitivity, specificity, PPV, and NPV. Results: A total of 215 patients were included (158 in the non-invasive and 57 in the histological cohort, mean age 56.6 ± 13.9 years; 65% male; median BMI 31.1 kg/m²). In the overall cohort assessed by non-invasive reference standards (FAST score and LSM), all three biological matrices showed limited diagnostic performance for both fibrotic MASH (AUC 0.54–0.60) and significant fibrosis (AUC 0.56–0.63). In the biopsy-proven cohort, the e-nose demonstrated moderate diagnostic accuracy for fibrotic MASH (accuracy 0.74; AUC 0.78) and significant fibrosis (accuracy 0.74; AUC 0.70), with moderate-to-high specificity and positive predictive value (0.87 and 0.89, respectively). E-tongue analyses showed lower and more variable performance across both matrices and outcomes. Within the histological cohort, the e-nose showed higher accuracy and PPV compared with LSM, whereas FAST retained higher sensitivity and negative predictive value (AUC 0.81 at the rule-out threshold), supporting complementary roles in risk stratification. Subsequent substudies, focused on integrated predictive modeling and metabolomic analysis, are ongoing. Conclusions: Multisensory electronic technologies showed modest-to-moderate performance in MASLD risk stratification, with the e-nose yielding the most promising results. Their non-invasiveness, rapidity, and scalability support the potential role of e-sensing-derived features as novel complementary components of integrated risk stratification strategies, rather than as stand-alone diagnostic tool.