Endocrine Abstracts

Background: In bariatric surgery (BS) patients, nonalcoholic fatty liver disease and fibrosis are present in up to 90% and 24%, respectively. Several fibrosis risk scores have been developed and have been validated in cohorts different from morbid obesity (MO). We aimed to determine whether established liver fibrosis scores were accurate in predicting fibrosis in MO subjects.

Material and Methods: Cross-sectional analysis in a cohort of MO patients undergoing BS in a Hospital in Spain over a 2 – year period. Demographics, anthropometric, clinical and laboratory features were assessed. The APRI, FIB-4, Forns, NAFLD-fs, BARD, BAAT and Hepamet fibrosis scores were calculated.

Liver biopsies were performed during BS. The NAS score and Kleiner scale were used to assess steatohepatitis (NASH) and fibrosis, respectively.

Student t-test, Fisher-Pitman, Pearson’s chi-squared or Fisher’s exact tests were used. A P < 0.05 was considered significant. The AUROC was calculated, as well as measures of diagnostic accuracy based on established thresholds. Modified cutoff values for differentiating F2–4 (SF: significant fibrosis) disease were calculated. Logistic regression analysis was performed to find predictors of significant fibrosis.

Results: 50 patients were included. Nine participants (18%) had SF. Sixteen (32%) had NASH. Proportion of patients with NASH was higher in the SF group (88.9 vs 19.5%, P < 0.05). BMI, HbA1c, ALT, AST an GGT levels were significantly different in patients with SF. Basal glucose, HbA1c, AST and GGT were identified as independent predictor of SF.

APRI, FIB-4, Forns and Hepamet fibrosis scores were significantly higher in the SF group (P < 0.05). When thresholds were modified to optimize detection of significant fibrosis, they were considerably lower than those described in the literature, this allowed to identify a greater proportion of fibrosis, improving sensitivity and consequently increasing negative predictive value (NPV).

BARD (AUC 0.76) and Forns (AUC 0.67) scores had the best performance considering the cutoff suggested by the ROC analysis, both with sensitivity of 88.9%, specificity of 51.2%, NPV of 95.5% and efficiency of 58%, which makes them appropriate to exclude fibrosis.

Conclusions: Basal glycemia, HbA1c, AST and GGT are independent predictors of SF in our population. Existing scoring systems are unable to stratify fibrosis risk in MO patients using established threshold. It is necessary to modify these cutoffs to improve accuracy. By doing so, the BARD and the Forns scores had a good global performance with NPP of 95.5% and permits to predict the absence of SF.