RATIONAL: Risk strAtificaTIon Of Nonalcoholic fAtty Liver

NAFLD, paralleling the pandemic of obesity and diabetes, is the most common and emerging liver disease in Western countries, leading to both hepatic and extrahepatic -mostly cardiovascular and nonliver cancers- complications. An effective management and treatment of NAFLD patients is limited by several unmet needs. First, only a minority of patients progress towards more severe forms of diseases and complications. Second, NASH and fibrosis are the target of novel pharmacological approaches; in particular, liver fibrosis is the main determinant of prognosis. Third, prediction of the natural history and prognosis of individuals with NAFLD is very difficult. The lack of robust and widely applicable noninvasive methods to stratify the risk of liver disease severity and predict events is a key hurdle currently hampering advancements in clinical management and the validation of effective therapeutic approaches. The overall aim of the project is to develop and validate novel algorithms that closely map those defined by the BIPED system and enabling to 1. Disease risk stratification of NASH and liver fibrosis (NASD >=F2; F2-F4;F3-F4), 2. Prediction of liver-related and nonrelated events, and 3. identification of clinically relevant patient sub-groups with different prognosis.
The project builds up on a large retrospective/prospective cohort of >1500 patients with histologically diagnosis of NAFLD or with clinical diagnosis of NAFLD compensated cirrhosis with 1) well characterized and homogeneous dataset at baseline of clinical, biochemical, anthropometric variables and imaging data (liver stiffness measurement - LSM); 2) novel biomarkers - including collagen biomarkers, miRNA, CK-18 and IL-32, and genetic risk variants; and 3) a median follow-up of clinical events of 60 months. By exploiting artificial intelligence, we will develop accurate noninvasive algorithms for disease risk stratification, and prediction of hepatic as well as extrahepatic morbidity/mortality that will allow us to identify clinically relevant patient sub-groups with different prognosis across the spectrum of NAFLD. The use of Al will allow to overcome limitations of classical statistical approaches by rethinking the model design as prediction problems, by building non-linear models that can account for covariate/biomarkers interactions, and by evaluating potential interactions among different groups using clustering and Topological Data Analysis. The new obtained algorithms will be validated in external cohorts of >1000 NAFLD cases as expression of international collaborations of the Pl, and in the large prospective under recruitment NAFLD cohort form the RESIST-NASH Sicilian network.
The results of this project will provide novel data for a personalized approach to NAFLD patients and for an optimization of economical resources. Finally, the obtained results will be disseminated in the main national and international meetings, in the main scientific journals, and across the website of the Health Care Systems and Social media.