A good step toward low-cost prognostication of liver-related outcome awaits more validation

We read the article “Development and validation of a model to predict incident chronic liver disease in the general population: the CLivD score” with great interest. The study demonstrates the possibility of prognosticating the risk of liver-related outcomes using easily accessible information, and highlights potential room for further validation and optimization.

Although the CLivD model was developed using European cohorts, we observed similar overall risk-stratification capacity in the Third National Health and Nutrition Examination Survey (NHANES III). NHANES III was a study conducted by the US Center of Disease Control in 1988-1994 to obtain nationally representative information on the health and nutritional status of the US population.^, The study used a complex, multistage, probability sampling design to select participants representative of the civilian, non-institutionalized US population. The mortality data of NHANES III participants were last collected in 2015. We calculated CLivD Model_lab and Model_non-lab in 5,783 participants aged 40-70 years in NHANES III, representing an estimated population of 187 million. We used weighted Cox proportional hazard analysis to examine the effectiveness of CLivD models to predict liver-related mortality, which included chronic liver disease, cirrhosis, and liver cancer as the underlying cause of death (UCOD 024, 093, 094, and 095). The NHANES III data showed that both Model_lab and Model_non-lab effectively risk-stratified liver-related mortality (Fig. 1A,B). In both models, no liver-related death was seen in the minimal risk category (green), and the highest mortality was seen in the high-risk category (red). Between the 2 models, Model_lab performed better than Model_non-lab (C-statistics 0.733 vs. 0.637). The high-risk category by Model_lab had a hazard ratio (HR) of 7.7 (1.4–43.5, p = 0.02) over combined low- and minimal-risk categories in Cox proportional hazard models, while Model_non-lab had an HR of 4.8 (1.2–20.3, p = 0.03). Of note, the incidence of liver-related mortality underestimates the burden of advanced liver disease compared to the hospitalization data captured by Åberg and colleagues.

The NHANES III data also revealed the need to further validate this predictive model in populations with different baseline risks of liver diseases. In NHANES III, Hispanics were at a 4.2-fold (1.2–15.1, p = 0.03) increased risk of liver-related death compared to non-Hispanic Caucasian counterparts. This observation mirrors a high prevalence of PNPLA3 I148M polymorphism in Hispanics, which has been linked to the development of non-alcoholic fatty liver disease, alcohol-related liver disease, as well as liver cancer.^{,  ,}  The CLivD Model_lab captured this heightened incidence of liver-related mortality among high-risk groups with an overall C-statistic of 0.751 (Fig. 1C). High- and intermediate-risk categories (red and orange) in Hispanics were associated with HRs of 46.3 and 8.6 (p <0.001 and 0.04, respectively), compared to combined low- and minimal-risk categories. In comparison, the performance of the CLivD model was inadequate among African Americans, with a C-statistic of 0.554 (Fig. 1D). NHANES III did not provide sufficient statistical power for the analysis among Asians or other ethnic groups, which will require further studies.

Identifying patients at risk of liver-related outcomes at an asymptomatic phase of the disease remains a major challenge. This challenge falls mostly on primary care physicians, who often need to make decisions based on cost-effectiveness. In this regard, the CLivD score is a useful tool that offers an easy and flexible approach to risk-stratify patients using simple measurements. Yet, improvement is still possible. GGT has been replaced by ALT and AST in many parts of the world. Waist-circumference is not always measured in routine clinical visits, while BMI could be more accessible. We have recently shown that a point-based SODA-2B score, using Sex, Other history of liver disease, Diabetes, Age and BMI, can risk-stratify individuals with high liver stiffness. Following the advent of many effective tools to measure liver fibrosis developed by the liver community, hepatologists have a responsibility to find strategies to combine these tools for risk stratification in a cost-effective manner.