From the Journals

Biomarker algorithm may offer noninvasive look at liver fibrosis

 

Key clinical point: A serum biomarker–based algorithm may provide a noninvasive method of detecting advanced hepatic fibrosis in patients with nonalcoholic fatty liver disease (NAFLD).

Major finding: The area under the receiver operator characteristic (AUROC) curve for a combination of validation samples was 0.856.

Study details: A retrospective study of liver fibrosis serum markers and clinical data from 396 patients with NAFLD and various stages of fibrosis.

Disclosures: The study was supported by Prometheus Laboratories. Authors not employed by Prometheus Laboratories were employed by Duke University or the University of California, San Diego; each institution received funding from Prometheus Laboratories.

Source: Loomba R et al. Clin Gastroenterol Hepatol. 2018 Nov 15. doi: 10.1016/j.cgh.2018.11.004.


 

FROM CLINICAL GASTROENTEROLOGY AND HEPATOLOGY

Serum biomarkers may enable a noninvasive method of detecting advanced hepatic fibrosis in patients with nonalcoholic fatty liver disease (NAFLD), according to results from a recent study.

Nonalcoholic fatty liver disease Nephron/Wikimedia/Creative Commons License

An algorithm created by the investigators distinguished NAFLD patients with advanced liver fibrosis from those with mild to moderate fibrosis, reported lead author Rohit Loomba, MD, of the University of California at San Diego and his colleagues.

“Liver biopsy is currently the gold standard for diagnosing NASH [nonalcoholic steatohepatitis] and staging liver fibrosis,” the investigators wrote in Clinical Gastroenterology and Hepatology. “However, it is a costly and invasive procedure with an all-cause mortality risk of approximately 0.2%. Liver biopsy typically samples only 1/50,000th of the organ, and it is liable to sampling error with an error rate of 25% for diagnosis of hepatic fibrosis.”

Existing serum-based tests are reliable for diagnosing nonfibrotic NAFLD, but they may misdiagnosis patients with advanced fibrosis. Although imaging-based techniques may provide better diagnostic accuracy, some are available only for subgroups of patients, while others come with a high financial burden. Diagnostic shortcomings may have a major effect on patient outcomes, particularly when risk groups are considered.

“Fibrosis stages F3 and F4 (advanced fibrosis) are primary predictors of liver-related morbidity and mortality, with 11%-22% of NASH patients reported to have advanced fibrosis,” the investigators noted.

The investigators therefore aimed to distinguish such high-risk NAFLD patients from those with mild or moderate liver fibrosis. Three biomarkers were included: hyaluronic acid (HA), TIMP metallopeptidase inhibitor 1 (TIMP-1), and alpha2-macroglobulin (A2M). Each biomarker has documented associations with liver fibrosis. For instance, higher A2M concentrations inhibit fibrinolysis, HA is associated with excessive extracellular matrix and fibrotic tissue, and TIMP-1 is a known liver fibrosis marker and inhibitor of extracellular matrix degradation. The relative strengths of each in detecting advanced liver fibrosis was determined through an algorithm.

The investigators relied on archived serum samples from Duke University, Durham, N.C., (n = 792) and University of California at San Diego (n = 244) that were collected within 11 days of liver biopsy. Biopsies were performed with 15- to 16-gauge needles using at least eight portal tracts, and these samples were used to diagnose NAFLD. Patients with alcoholic liver disease or hepatitis C virus were excluded.

Algorithm training was based on serum measurements from 396 patients treated at Duke University. Samples were divided into mild to moderate (F0-F2) or advanced (F3-F4) fibrosis and split into 10 subsets. The logical regression model was trained on nine subsets and tested on the 10th, with iterations 10 times through this sequence until all 10 samples were tested. This process was repeated 10,000 times. Using the median coefficients from 100,000 logistical regression models, the samples were scored using the algorithm from 0 to 100, with higher numbers representing more advanced fibrosis, and the relative weights of each biomarker measurement were determined.

A noninferiority protocol was used to validate the algorithm, through which the area under the receiver operating characteristic (AUROC) curve was calculated. The AUROC curve of the validation samples was 0.856, with 0.5 being the score for a random algorithm. The algorithm correctly classified 90.0% of F0 cases, 75.0% of F1 cases, 53.8% of F2 cases, 77.4% of F3 cases, and 94.4% of F4 cases. The sensitivity was 79.7% and the specificity was 75.7%.

The algorithm was superior to Fibrosis-4 (FIB-4) and NAFLD Fibrosis Score (NFS) in two validation cohorts. In a combination of validation cohorts, the algorithm correctly identified 79.5% of F3-F4 patients, compared with rates of 25.8% and 28.0% from FIB-4 and NFS, respectively. The investigators noted that the algorithm was unaffected by sex or age. In contrast, FIB-4 is biased toward females, and both FIB-4 and NFS are less accurate with patients aged 35 years or younger.

“Performance of the training and validation sets was robust and well matched, enabling the reliable differentiation of NAFLD patients with and without advanced fibrosis,” the investigators concluded.

The study was supported by Prometheus Laboratories. Authors not employed by Prometheus Laboratories were employed by Duke University or the University of California, San Diego; each institution received funding from Prometheus Laboratories.

SOURCE: Loomba R et al. Clin Gastroenterol Hepatol. 2018 Nov 15. doi: 10.1016/j.cgh.2018.11.004.

Next Article: