Work

Novel blood signature for hepatocellular carcinoma screening

Year: 2025

Type: article

Abstract: Background Alpha-fetoprotein is commonly used for hepatocellular carcinoma (HCC) screening in at-risk populations, but its effectiveness is limited. Routine blood tests offer insights into cancer-related conditions and improve detection in other cancers. This study explores the postulated changes in routine blood tests of HCC patients, allowing the development of routine blood-based artificial intelligence for early HCC detection. Patients and methods This population-based retrospective study analyzed patient records from 2000 to 2018 from the Hong Kong Hospital Authority Data Collaboration Laboratory. Patients with chronic liver disease (CLD), both with and without HCC, were identified using ICD codes, antiviral drug history, virology tests, and radiology reports. Those with decompensated CLD were excluded. Routine blood tests included complete blood count, liver function test, renal function test, and clotting profiles, with records collected within 1 month before HCC diagnosis. Statistical analyses included descriptive statistics and the Mann–Whitney U (MWU) test. Results The cohort comprised 223 862 patients, including 31 149 with HCC (13.9%). Statistical analysis revealed a distinct blood signature for HCC patients, characterized by significant liver function derangement (elevated alanine aminotransferase, alkaline phosphatase, bilirubin, aspartate aminotransferase; decreased albumin), signs of systemic inflammation (lower lymphocyte count, red cell distribution width), bleeding tendencies (prolonged prothrombin time, activated partial thromboplastin time; low platelet count), and indications of cachexia (lower albumin, creatinine, urea)—all statistically significant ( P < 0.05). Conclusions This study presents a novel blood signature for HCC detection based on extensive clinical data. The unique spectral characteristics effectively differentiate HCC from CLD controls, supporting the potential for machine learning models in HCC detection.

Source: ESMO Gastrointestinal Oncology

DOI: 10.1016/j.esmogo.2025.100185