Alcoholic-related liver disease (ALD) accounts for >50% of all liver-related deaths. Sustained excess drinking is known to induce fatty liver and alcohol-related steatohepatitis, which can eventually lead to alcoholic liver fibrosis (ALF) and further to alcohol-related liver cirrhosis (ALC).

Identification of patients with early-stage ALD remains troublesome as these are largely asymptomatic, so the majority of them are only diagnosed by the time ALD has reached decompensated cirrhosis. Moreover, decompensated cirrhosis is accompanied by complications as bleeding and ascites.

A study discovered relevant upstream diagnoses to help understand the development of ALD and highlighted meaningful downstream diagnoses that represent its progression to liver failure. The study utilized the data from the Danish health registries covering the entire population of Denmark during nineteen years, to examine the possibility to identify patients likeliness to develop ALF or ALC according to their medical history.

High-dimensional statistical and machine learning techniques were utilized in the study to extract and analyze data from the Danish National Patient Registry. Late diagnoses of ALD, ALC is the most common form of ALD and ALC patients have a strong over-representation of diagnoses associated with liver dysfunction were the most common finding in the registry. Contrary, very few patients diagnosed with ALF, appeared to be less sick than those with ALC.

The researchers carried out a matched case-control study using patients with ALC (cases) and their matched patients with non-ALD (controls).

Machine learning models like SVM, RF, LightGBM and NaiveBayes, trained and tested-

On the group of ALC patients showed high performance for data classification

On the small set of ALF patients revealed a drop in performance

Thus it was concluded that the statistical and machine learning results tend to underscore small groups of upstream and downstream co-morbidities that aid in accurate detection of ALC patients and have a great prediction of ALF. These groups include conditions caused by alcohol or by malnutrition associated with alcohol overuse etc. There may also be co-morbidities related to trauma and lifestyle or complications to cirrhosis like oesophageal varices.

These findings determine the promising potential of this approach to uncover knowledge in registry data related to ALD.

Source: Grissa D, Nytoft Rasmussen D, Krag A, Brunak S, Juhl Jensen L (2020) Alcoholic liver disease: A registry view on comorbidities and disease prediction. PLoS Comput Biol 16(9): e1008244. https://doi.org/10.1371/journal.pcbi.1008244