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

Abstract and Introduction


Background and Aims: Patients with nonalcoholic fatty liver disease (NAFLD) cirrhosis benefit from referral to subspecialty care. While several clinical prediction rules exist to identify advanced fibrosis, the cutoff for excluding cirrhosis due to NAFLD is unclear. This analysis compared clinical prediction rules for excluding biopsy-proven cirrhosis in NAFLD.

Methods: Adult patients were enrolled in the NASH Clinical Research Network (US) and the Newcastle Cohort (UK). Clinical and laboratory data were collected at enrolment, and a liver biopsy was taken within 1 year of enrolment. Optimal cutoffs for each score (eg, FIB-4) to exclude cirrhosis were derived from the US cohort, and sensitivity, specificity, positive predictive value, negative predictive value and AUROC were calculated. The cutoffs were evaluated in the UK cohort.

Results: 147/1483 (10%) patients in the US cohort had cirrhosis. All prediction rules had similarly high NPV (0.95–0.97). FIB-4 and NAFLD fibrosis scores were the most accurate in characterising patients as having cirrhosis (AUROC 0.84–0.86). 59/494 (12%) patients in the UK cohort had cirrhosis. Prediction rules had high NPV (0.92–0.96), and FIB-4 and NAFLD fibrosis score the most accurate in the prediction of cirrhosis in the UK cohort (AUROC 0.87–0.89).

Conclusions: This cross-sectional analysis of large, multicentre international datasets shows that current clinical prediction rules perform well in excluding cirrhosis with appropriately chosen cutoffs. These clinical prediction rules can be used in primary care to identify patients, particularly those who are white, female, and <65, unlikely to have cirrhosis so higher-risk patients maintain access to specialty care.


Nonalcoholic fatty liver disease (NAFLD) is an important public health problem, affecting an estimated 25%–30% of the world population,[1,2] with approximately 20% of patients having nonalcoholic steatohepatitis (NASH).[3] NAFLD can be progressive, particularly if NASH is present, with up to 30%–40% having progression of fibrosis and at risk for cirrhosis and its complications.[4–6] The increasing burden of NASH-associated cirrhosis is apparent, as it is the second leading indication for liver transplantation.[7] Fibrosis stage in NAFLD has been identified as the most important factor associated with mortality.[8–10] Identifying patients with cirrhosis due to NASH is critical so that appropriate surveillance for hepatocellular carcinoma (HCC) and clinical decompensation be performed. Given the NAFLD disease burden, triage of patients among primary care physicians and specialists is essential and simple, non-invasive tools to rule out cirrhosis in NAFLD patients are extremely useful. While noninvasive imaging studies such as ultrasound-based elastography and magnetic resonance elastography (MRE) have excellent ability to discriminate between cirrhosis and other stages of fibrosis with AUROC as high as ≥0.9,[11–13] several factors limit the ability to apply these tests to the large at-risk population: (1) the results of ultrasound-based elastography can be affected by obesity[14,15] or steatosis;[15] (2) testing requires available imaging technology and adequate experience to produce reliable results.[15] Liver biopsy, while considered the "gold standard" for diagnosis and staging of NAFLD, is fraught with problems when applying to the general population, due to cost, risk and potentially inadequate or misrepresentative specimens obtained.

Several clinical prediction rules to discriminate advanced fibrosis and cirrhosis (modified Brunt classification F3/4) from milder fibrosis (F0-2) have been published, including APRI, FIB-4 and NAFLD fibrosis score. Most of these rules were originally developed to assess for advanced fibrosis in non-NAFLD patients but have subsequently been well-validated for use in NAFLD patients.[16–19] Only the NAFLD Fibrosis Score and BARD scores were developed and validated for use in patients with NAFLD. The NAFLD fibrosis score has been demonstrated to be cost-effective in the risk stratification of patients with NAFLD in the primary care setting.[20] How well these tools separate non-cirrhotic (F0-3) vs. cirrhotic (F4) is also not well described. Given the myriad of clinical prediction tools available, a comprehensive evaluation of performance characteristics would be helpful to clinicians seeing patients with NAFLD.

Therefore, we aimed to evaluate seven clinical prediction rules in two large, well-defined, prospective clinical cohorts to determine appropriate cutoffs to rule out cirrhosis in NAFLD patients.