icon-folder.gif   Conference Reports for NATAP  
  The Liver Meeting
Digital Experience
November 13 - 16 - 2020
Back grey_arrow_rt.gif
Simple Tool Tells Primary Care Docs When to Refer People With NAFLD
  AASLD, The Liver Meeting, November 12-15, 2021
Mark Mascolini
A SAFE score built from easily gathered values could tell primary care physicians which nonalcoholic fatty liver disease (NAFLD) patients they can manage themselves and which run a high risk of progression and should be referred to a specialist [1]. Researchers from Stanford Medicine developed the tool and applied it to 12,000 US residents in the National Health and Nutrition Examination Survey III (NHANES) to see how well it predicts death.
The study team noted that NAFLD remains extremely prevalent among patients seen in primary care practice. Research links significant liver fibrosis (F2 or greater) to poor long-term outcomes, but primary care providers lack a reliable tool based on noninvasive variables that can rule out F2 fibrosis in people with NAFLD. Existing noninvasive markers can detect advanced fibrosis, but the value of referral to a specialist is largely lost once a patient has advanced fibrosis.
The investigators began with a training set of 635 people in the NASH CRN observational study who had biopsy-proved NAFLD and fibrosis stage ranging from F0 to F4. They used traditional logistic regression and machine-learning techniques (random forest, gradient boosting machine, and generalized additive models) to build a prediction model that could discriminate between F0/F1 fibrosis and F2 or greater fibrosis.
The researchers validated predictive models in two independent groups-280 people in the FLINT trial with biopsy-verified nonalcoholic steatohepatitis (NASH), and 130 people in a local Stanford NAFLD cohort with magnetic resonance elastography data. The Stanford team used area under the receiver operating characteristic (AUROC) curves to test the diagnostic performance of the training set against the two validation sets.
Among the several models considered, the investigators got the best performance from multivariable logistic regression analysis. The final model, called the Steatosis-Associated Fibrosis Estimator or SAFE score, performed the best. SAFE includes a short list of easily obtained values: age, body mass index, diabetes (yes/no), platelets, serum globulin concentration, and aspartate and alanine aminotransferase. SAFE appears online at https://medcalculators.stanford.edu/safe.
AUROC comparisons showed that SAFE significantly outperformed two other noninvasive markers-FIB-4 and NAFLD Fibrosis Score-in all three datasets: the training set and the two validation sets. The researchers used sensitivity and specificity thresholds determined before their analysis began to scale the model so that a score below 0 means low risk of progression to death, a score above 100 means high risk, and a score between 0 and 100 means intermediate risk.
Next the researchers applied SAFE to 11,953 NHANES III participates with or without NAFLD to gauge SAFE's ability to predict long-term outcomes. While 64% of NHANES III participates had no NAFLD, 36% had NAFLD, defined as steatosis on ultrasound without other chronic liver disease. In this 36% with NAFLD, SAFE determined that 54% had a low risk of progression, 31.6% an intermediate risk, and 14.4% a high risk.
The investigators used Cox regression adjusted for age, sex, race/ethnicity, smoking, and hypertension to analyze hazard ratios for long-term mortality in each SAFE stratum. This analysis showed that survival through 25 years in people called low risk by SAFE was statistically similar to survival in NHANES members without NAFLD (adjusted hazard ratio [aHR] 0.95, 95% confidence interval [CI] 0.86 to 1.06, P = 0.384). In contrast, survival of the SAFE intermediate group was significantly worse than the no-NAFLD group (aHR 1.10, 95% CI 1.00 to 1.20, P = 0.042). And survival was still worse in the SAFE high-risk group (aHR 1.53, 95% CI 1.38 to 1.71, P < 0.001).
The Stanford investigators stated their key findings that a SAFE score below 0 indicates patients with a low risk of significant fibrosis who can be safely managed by primary care providers, whereas a SAFE score at or above 100 signals a high risk of significant fibrosis and calls for referral to a specialist.
1. Sripongpun P, Mannalithara A, Kim D, Kwong AJ, Kim WR. The steatosis-associated fibrosis estimator (SAFE) score: a tool for primary care to triage low-risk non-alcoholic fatty liver disease patients. AASLD, The Liver Meeting, November 12-15, 2021. Parallel session 10: Diagnostics and Biomarkers of NAFLD.