Comparison of Clinical Prediction Rules for Ruling Out Cirrhosis in Nonalcoholic Fatty Liver Disease (NAFLD)

Danielle Brandman; Marie Boyle; Stuart McPherson; Mark L. Van Natta; Arun J. Sanyal; Kris Kowdley; Brent Neuschwander-Tetri; Naga Chalasani; Manal F. Abdelmalek; Norah A. Terrault; Art McCullough; Ricki Bettencourt; Cyrielle Caussy; David E. Kleiner; Cynthia Behling; James Tonascia; Quentin M. Anstee; Rohit Loomba


Aliment Pharmacol Ther. 2022;55(11):1441-1451. 

In This Article


Study Design and Participants: US Cohort

The current study on predictive clinical rules for cirrhosis was a cross-sectional analysis based primarily upon data collected prospectively from the large U.S. Multicenter NASH Clinical Research Network (CRN) cohort. Data from the enrolment visit from all participants in the NAFLD Database study and all participants enrolled up to the close of the analysis database on November 2015 in NAFLD Database 2 study were included.[21] The details of the inclusion and exclusion criteria and study designs have previously been published.[21] Briefly, the NAFLD Database 2 study is an extension of the NAFLD Database 1 study and uses similar inclusion and exclusion criteria, except for requiring histological proof of NAFLD as an inclusion criterion.[22] The NASH CRN studies are sponsored by the National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, and all patients provided written informed consent for participation.

Information on demographic characteristics, medical history, clinical tests and liver biopsy results were collected at the baseline visit, as previously described.[21] All eligible adults met the following diagnostic criteria for NAFLD: (1) histologic diagnosis of NAFLD or histologic diagnosis of cryptogenic cirrhosis; (2) alcohol use history of <70 g/week for females or < 140 g/week for males; and (3) exclusion of liver disease of other aetiologies, including viral or autoimmune hepatitis, drug-induced liver disease and cholestatic or metabolic liver disease.

Liver Histology

All liver biopsy slides were stained with haematoxylin and eosin and Masson's trichrome, and were reviewed and scored centrally by the NASH CRN pathology committee as previously reported;[23] the review was performed blindly without knowledge of local pathology evaluation or clinical or laboratory characteristics of the patients.[23,24] The NAFLD activity score (NAS) was graded from 0 to 8 and is the sum of scores for steatosis (0–3), lobular inflammation (0–3) and hepatocellular ballooning (0–2). Definite NASH was deemed present as judged by the majority of the local NASH CRN pathologist and two additional NASH CRN pathologists.[24] Fibrosis stage was assessed according to the modified Brunt classification; 0 = no fibrosis, 1a = mild, zone 3 perisinusoidal fibrosis (requires trichrome), 1b = moderate, zone 3 perisinusoidal fibrosis (does not require trichrome), 1c = portal/periportal fibrosis, 2 = zone 3 perisinusoidal and periportal fibrosis, 3 = bridging fibrosis, 4 = cirrhosis.[23–26] Patients with stage F3 or F4 fibrosis were considered to have advanced fibrosis, and those with F4 were considered to have cirrhosis. All biopsies were taken within 1 year (NASH CRN database 1) or 90 days (NASH CRN database 2) of the enrolment visit.

Study Design and Participants: UK Cohort

This study included consecutive patients with biopsy-proven NAFLD who attended the specialist fatty liver clinic at the Freeman Hospital, Newcastle-upon-Tyne, UK, enrolled prospectively. The clinical data were sourced ethically following receipt of informed consent from each patient and their research use was in accordance with the terms of the informed consents under an IRB/EC approved protocol. All data were obtained in a similar manner as for the US cohort. Liver biopsies were conducted as per routine clinical care for the investigation of abnormal liver function tests (raised ALT, AST or GGT) or to stage disease severity in patients with imaging evidence of fatty liver. Clinical and laboratory data were collected prospectively from the time of liver biopsy. Patients with evidence of other liver diseases (autoimmune hepatitis, viral hepatitis, drug-induced liver injury, haemochromatosis, cholestatic liver disease or Wilson's disease) were excluded. In addition, subjects consuming excessive amounts of alcohol (alcohol intake >20 g/day for women; >30 g/day for men) at the time of biopsy or in the past were excluded. Patients with incomplete data to calculate all the non-invasive scores based on liver enzymes and clinical data were excluded.

Percutaneous liver biopsies were performed as per unit protocol and prepared in the same manner as in the US cohort within 90 days of enrolment, and were assessed by an experienced local hepatopathologist. Patients with liver biopsies specimens less than 15 mm in length were excluded. Histological scoring was performed according to the NASH Clinical Research Network criteria (CRN)[23] using the same criteria as the US cohort.

Statistical Analysis Plan

Seven clinical prediction rules (AST:ALT ratio, AST to platelet ratio (APRI), BMI AST/ALT ratio diabetes (BARD) score, FIB-4 index, NAFLD fibrosis score, Bonacini Cirrhosis discriminant score (CDS), and Lok index) were calculated using previously published formulas (Table S1).[27–33] These scores have well-validated cutoffs to predict advanced fibrosis (bridging fibrosis or cirrhosis). In this study, new optimal cutpoints for diagnosis of cirrhosis in seven clinical prediction rules were identified using Youden's index as the metric to obtain the maximum sum of sensitivity and specificity, using the data from the US cohort. These cutoffs were then evaluated in the UK cohort. Each clinical prediction rule was evaluated for sensitivity, specificity, positive predictive value, negative predictive value and AUROC. The results adhere to the Standards for Reporting of Diagnostic Accuracy (STARD) for diagnostic tests, with the STARD checklist included as Table S2.

Data are presented as means with standard deviations or as percentages unless otherwise specified. Variables were compared in patients with cirrhosis to those without cirrhosis using chi-square test for categorical variables and t-test for continuous variables. The data analysis for this paper was generated using both SAS (SAS version 9.4, SAS Institute Inc.) and Stata software (StataCorp. 2017. Stata Statistical Software: Release 15.1; StataCorp LLC).