ABDA Score: A Non-invasive Model to Identify Subjects with Fibrotic Non-alcoholic Steatohepatitis in the Community

Abstract

Background: Non-alcoholic fatty liver disease and non-alcoholic steatohepatitis (NASH) are prevalent in the community, especially among those with metabolic syndrome. Patients with fibrotic NASH are at increased risk of liver-related-events. Currently available non-invasive tests have not been utilized for screening for fibrotic NASH among the community. We aimed to develop a screening tool for fibrotic NASH among community members.

Methods: We included two large cohorts aimed at assessing cardiovascular disease among community members. Fibrotic NASH was defined using the FibroScan-aspartate aminotransferase score of ≥0.67 that identifies ≥F2 fibrosis and a non-alcoholic fatty liver disease activity score ≥4 with a specificity of 90%. Metabolic parameters, biochemical tests and anthropometry were used to develop a multivariate model.

Results: The derivation cohort (n = 1660) included a population with a median age 45 years, 42.5% males, metabolic syndrome in 66% and 2.7% (n = 45) with fibrotic NASH. Multivariate analysis identified the four significant variables (Age, body mass index , Diabetes and alanine aminotransferas levels) used to derive an ABDA score. The score had high diagnostic accuracy (the area under receiver-operating characteristic curve, 0.952) with adequate internal validity. An ABDA score ≥-3.52 identified fibrotic NASH in the derivation cohort with a sensitivity and specificity of 88.9% and 88.3%. The score was validated in a second cohort (n = 357) that included 21 patients (5.9%) with fibrotic NASH, where it demonstrated a high area under receiver-operating characteristic curve (0.948), sensitivity (81%) and specificity (89.3%).

Conclusions: ABDA score utilizes four easily available parameters to identify fibrotic NASH with high accuracy in the community.

Keywords: NAFLD; diabetes; fatty liver; fibrosis; lipids.

Figure 1

(A) The Hosmer–Lemeshow calibration plot, with Loess smoothing, for the prediction model in the overall derivation cohort showing satisfactory fit. EO, expected/observed ratio; CITL, calibration-in-the-large; AUC, area under the curve; Loess, locally estimated scatterplot smoothening (B) Receiver operating characteristic curve for the ABDA score in predicting fibrotic non-alcoholic steatohepatitis in the derivation cohort.

Figure 2

Line diagrams showing the sensitivity and specificity of various ABDA scores in predicting subjects with fibrotic non-alcoholic steatohepatitis.

Figure 3

(A) The Hosmer–Lemeshow calibration plot, with Loess smoothing, for the prediction model in the validation cohort showing satisfactory fit. EO, expected/observed ratio; CITL, calibration-in-the-large; AUC, area under the curve; Loess, locally estimated scatterplot smoothening (B) Receiver operating characteristic curve for the ABDA score in predicting fibrotic non-alcoholic steatohepatitis in the validation cohort.

Figure 4

Receiver operating characteristic curve for the ABDA score, non-alcoholic fatty liver disease fibrosis score, fibrosis-4 and body mass index, aspartate aminotransferase to alanine aminotransferase ratio and diabetes score in predicting fibrotic non-alcoholic steatohepatitis. The ABDA score showed a better discriminative ability than all other predictive scores (P < 0.05, Delong test).

Figure S1