Clemence Moreau, Marine Roux, Jeremie Riou, Hannes Hagström, Huapeng Lin, Hye Won Lee, Terry Cheuk-Fung Yip, Emmanuel Tsochatzis, Salvatore Petta, Elisabetta Bugianesi, Masato Yoneda, Ming-Hua Zheng, José Luis Calleja, George Boon-Bee Goh, Wah-Kheong Chan, Rocio Gallego-Durán, Arun J Sanyal, Victor de Lédinghen, Philip N Newsome, Jian-Gao Fan, Laurent Castéra, Michelle Lai, Céline Fournier-Poizat, Grace Lai-Hung Wong, Grazia Pennisi, Angelo Armandi, Atsushi Nakajima, Wen-Yue Liu, Ying Shang, Marc de Saint-Loup, Elba Llop, Kevin Kim Jun Teh, Carmen Lara-Romero, Amon Asgharpour, Sara Mahgoub, Mandy Sau-Wai Chan, Manuel Romero-Gomez, Seung Up Kim, Vincent Wai-Sun Wong, Jérôme Boursier
By automatically integrating and interpreting the dynamics of non-invasive tests of liver fibrosis, JLCM enables the personalised prediction of the risk of hard outcomes, rather than the imperfect evaluation of histological surrogates.
BACKGROUND AND AIMS: The evolution of non-invasive tests of liver fibrosis during follow-up of patients with metabolic dysfunction-associated steatotic liver disease (MASLD) remains difficult to interpret in clinical practice. We aimed to translate the dynamics of non-invasive tests into a personalized prediction of liver-related events (LRE) in MASLD.
METHODS: We used the international multicentre VCTE-Prognosis cohort including adult patients with MASLD who underwent liver stiffness measurements (LSM) by vibration-controlled transient elastography. The study outcome was LRE, a composite endpoint including cirrhosis decompensation or hepatocellular carcinoma. A joint latent class model (JLCM) was used to compute dynamic predictions resulting in a personalized estimation of the risk of LRE (0-100% at chosen time horizons).
RESULTS: 13,627 patients were included, with 238 LRE occurring during the median follow-up of 4.0 years (IQR: 2.1-5.9). LSM trajectory adjusted on FIB-4 and sex (longitudinal part) and age with platelets (survival part) were selected by the JLCM for LRE prediction. Calibration plots showed very good agreement between the predicted and the observed risk of LRE. The JLCM provided excellent discrimination for LRE with integrated AUROCs increasing from 88.2% at baseline to 92.2% at the 5-year follow-up visit. At the different study visits, 61-80% of the patients who experienced LRE were identified as high risk by the JLCM, versus 39-56% with LSM and 32-60% with FIB-4.
CONCLUSION: By automatically integrating and interpreting the dynamics of non-invasive tests of liver fibrosis, JLCM enables the personalised prediction of the risk of hard outcomes, rather than the imperfect evaluation of histological surrogates.