Development and validation of a noninvasive prediction model for nonalcoholic steatohepatitis resolution after lifestyle intervention

• Published in: Hepatology (Baltimore, Md.) Vol. 63; no. 6; pp. 1875 - 1887
• Main Authors: Vilar‐Gomez, Eduardo, Yasells‐Garcia, Ali, Martinez‐Perez, Yadina, Calzadilla‐Bertot, Luis, Torres‐Gonzalez, Ana, Gra‐Oramas, Bienvenido, Gonzalez‐Fabian, Licet, Villa‐Jimenez, Oscar, Friedman, Scott L., Diago, Moises, Romero‐Gomez, Manuel
• Format: Journal Article
• Language: English
• Published: United States Wiley Subscription Services, Inc 01-06-2016
• Subjects: Adult; Confidence intervals; Diabetes; Female; Forecasting; Hepatology; Humans; Intervention; Life Style; Lifestyles; Liver; Male; Middle Aged; Models, Theoretical; Multivariate Analysis; Non-alcoholic Fatty Liver Disease - therapy
• ISSN: 0270-9139; 1527-3350
• DOI: 10.1002/hep.28484

Liver biopsy is the gold standard method to assess nonalcoholic steatohepatitis (NASH) resolution after therapeutic interventions. We developed and validated a simple and noninvasive scoring system to predict NASH resolution without fibrosis worsening after 1 year of lifestyle intervention. This was a prospective cohort study conducted in 261 patients with histologically proven NASH who were treated with lifestyle changes for 52 weeks and underwent a second liver biopsy to confirm NASH resolution. We divided the data into development (140 subjects) and validation (121 individuals) sets. NASH resolution occurred in 28% (derivation group) and 27% (validation group). At the multivariable analysis, weight loss (odds ratio [OR] = 2.75, 95% confidence interval [CI] 1.65‐4.58; P < 0.01), type 2 diabetes (OR = 0.04, 95% CI 0.005‐0.49; P = 0.01), normal levels of alanine aminotransferase at the end of intervention (OR = 9.84, 95% CI 2.21‐44.1; P < 0.01), age (OR = 0.89, 95% CI 0.83‐0.97; P = 0.01), and a nonalcoholic fatty liver activity score ≥5 (OR = 0.08, 95% CI 0.01‐0.43; P < 0.01) were independent predictors of NASH resolution. The area under the receiver operating characteristic curve of the selected model was 0.956 and 0.945 in the derivation and validation cohorts, respectively. Using a score threshold of ≤46.15, negative predictive values were 92% in the derivation and validation groups, respectively. By applying a cutoff ≥69.72, positive predictive values were 92% and 89% in the derivation and validation groups, respectively. Using both cutoffs, a liver biopsy would have been avoided in 229 (88%) of 261 patients, with a correct prediction in 209 (91%) Conclusions: A noninvasive prediction model including weight loss, type 2 diabetes, alanine aminotransferase normalization, age, and a nonalcoholic fatty liver activity score ≥5 may be useful to identify NASH resolution in patients under lifestyle intervention. (Hepatology 2016;63:1875‐1887)