Long-term Prognosis of Patients With Alcohol-related Liver Disease or Non-alcoholic Fatty Liver Disease According to Metabolic Syndrome or Alcohol Use

Marie Decraecker; Dan Dutartre; Jean-Baptiste Hiriart; Marie Irles-Depé; Hortense Marraud des Grottes; Faiza Chermak; Juliette Foucher; Adèle Delamarre; Victor de Ledinghen


Liver International. 2022;42(2):350-362. 

In This Article

Patients and Methods


From January 2003 to December 2016, we prospectively collected data on consecutive patients presenting to our centre for non-invasive diagnosis of liver fibrosis (Hepatology Unit, University Hospital, Pessac, France). We included all consecutive patients older than 18 years with NAFLD or ALD of any severity. The diagnosis of NAFLD was based on the presence of steatosis assessed by ultrasonography without any other known cause of steatosis. The diagnosis of ALD was based on chronic biochemical disorders (elevated transaminases or hyperferritinaemia) associated with elevated alcohol consumption according to the physician (more than 21 units/week for men or 14 units/week for women). Alcohol consumption was assessed over the year prior to inclusion.

We excluded patients with other causes of chronic liver disease; in particular, those with chronic viral hepatitis.

Patient and Public Involvement Statement

The study protocol conformed with the ethical guidelines of the 1975 Declaration of Helsinki. Patients were enrolled after written informed consent. The cohort was registered at ClinicalTrials.gov (identifier: NCT01241227).

Clinical and Biological Parameters

For all patients, we registered weight, height, waist circumference, body mass index (BMI), T2DM, hypertension, alcohol consumption and statin use. Statins were used in 331 patients (24%) with ALD, and 619 patients (38%) with NAFLD. No patient was on clinical trial, and no patient received vitamin E or pioglitazone.

Biological parameters measured included aspartate aminotransferase, alanine aminotransferase, fasting plasma glucose, total and high-density lipoprotein (HDL) cholesterol and triglycerides.

The diagnosis of metabolic syndrome was made by the presence of at least three of the following parameters: impaired fasting blood glucose >1.26 g/L or T2DM,

hypertriglyceridaemia >1.6 mmol/L, a low level of HDL <1.04 mmol/L for men and <1.29 mmol/L for women, an increase in waist circumference (adjusted for ethnicity) ≥102 cm for men and ≥88 cm for women and high blood pressure ≥130/85 mmHg; and we calculated the number of components of metabolic syndrome (from 0 to 5).

Liver Stiffness Measurement

LSM using FibroScan® was performed on the right lobe of the liver, through intercostal spaces, with the patient lying in dorsal decubitus with the right arm in maximal abduction. Trained nurses, assisted by time-motion mode ultrasound imaging, located a liver portion at least 6-cm thick and free of large vascular structures. The measurement depth was between 25 and 65 mm using the M probe, or the XL probe in patients with BMI >30 kg/m2; results were expressed in kilopascals. Only procedures with at least ten validated measurements and an interquartile range (IQR) <30% of the median (M) value (IQR/M), or regardless of the IQR/M if the LSM was <7.1 kPa, were considered reliable.


Overall survival (irrespective of the cause of death and including liver transplantation) was the primary endpoint. Survival time was calculated from the date of LSM to the event date or to the endpoint date (31 December 2017).

Cirrhotic patients were followed in consultation every 6 months with abdominal ultrasound examination and blood sample. Other patients were followed every 12 months with blood sample.

We recorded at the end of follow-up eventual other secondary endpoints: cardiovascular events (myocardial infarction and stroke), cancers, liver complications (HCC, ascites, encephalopathy and variceal bleeding) and, if any, causes of death (liver, cardiovascular or cancer). As a secondary endpoint, we also assessed the prognostic value of LSM for predicting overall and cause-specific survival and cardiovascular events, cancers and liver events. Follow-up ended on 31 December 2017.

Statistical Analysis

The study population was described and compared with the chi-squared or Fisher test for qualitative comparisons, and the Mann Whitney test for quantitative comparisons. A P-value <.05 was considered significant.

Regarding survival, survival curves were calculated using the Kaplan Meier method for the primary endpoint and Fine & Gray regression for specific survival to take into account competitive risks into cause-specific competing risk models. We took into account cardiovascular-related and cancer-related mortality for the analysis of liver-related mortality, we took into account liver-related and cancer-related mortality for the analysis of cardiovascular-related mortality, and we took into account liver-related and cardiovascular-related mortality for the analysis of cancer-related mortality. We did not take into account unknown mortality causes because it concerned a small percentage of patients.

Second, all parameters (clinical, biological, histological and non-invasive method data) were initially analysed in univariable Cox models, with a P-value <.05 considered significant. Then, parameters significantly associated with overall survival were introduced to a multivariable Cox model to identify independent predictors. In a second step, a Cox proportional-hazard model analysis was conducted to identify independent predictors of overall survival.

Logistic regression was performed when temporal data were not available for secondary outcomes.

The prognostic accuracy of LSM was evaluated by calculating Harrell's C-index for overall and specific survival. This is an extension to the concept of the area under the receiver operating characteristic curve (AUC) for time-to-event (survival) data; it evaluates the concordance between the predicted risk of an event and the observed survival time. Its value varies from 0 to 1, with 1 indicating perfect concordance (and thus perfect discriminative power of the risk score), 0.5 indicating a prediction that is only as good as chance, and a value less than 0.5 indicating discrimination in the opposite direction to that expected. Three models were evaluated: model 1 included age, male sex, alcohol consumption and tobacco consumption; model 2 included only LSM and model 3 included LSM and age, male sex, alcohol consumption and tobacco consumption. The prognostic accuracy of LSM for morbidity was estimated by the AUC, which can vary from 0 to 1. Analyses were performed using R version 3.6.1 software (5 July 2019; Foundation for Statistical Computing, Vienna, Austria, https://www.R-project.org/).