Anty et al developed a model for predicting nonalcoholic fatty liver disease in morbidly obese patients. This can help to identify patients to target for more aggressive management. The authors are from the Institut National de la Sante et de la Recherche Medicale (INSERM) in Nice, Centre Hospitalier Universitaire of Nice, University of Nice-Sophia-Antipolis and centre Hospitalier Princess Grace in Monaco.


Patient selection: morbidly obese referred for bariatric surgery


Endpoint: activity score >= 5 in the NASH classification of Kleiner et al



(1) serum ALT concentration (using a Roche assay with pyridoxal phosphate on a Hitachi Modular)

(2) serum cytokeratin 18 fragment concentration (by ELISA kit M30-Apoptosense)

(3) clinical evidence for the metabolic syndrome using the International Diabetes Foundation (IDF) criteria


X =

= (0.0378 * (serum ALT in IU/L)) + (0.002215 * (serum CK18 in IU/L)) + (1.825 if metabolic syndrome present) - 5.654


probability of endpoint =

= 1 / (1 + EXP((-1) * X))


The prevalence of the endpoint was 12%. The best cutoff value was 0.14 (presumably the probability value). This had a sensitivity of 84% and specificity of 86%. A value >= 0.14 had a positive predictive value of 0.44, while a value < 0.14 had a negative predictive evalue of 0.98.



• The normal reference ranges for the assays were not listed.


To read more or access our algorithms and calculators, please log in or register.