LCR1 and LCR2, two multi‐analyte blood tests to assess liver cancer risk in patients without or with cirrhosis

Summary Background No blood test has been shown to be effective in the prediction of primary liver cancer in patients without cirrhosis. Aim To construct and internally validate two sequential tests for early prediction of liver cancer. These tests enable an algorithm which could improve the performance of the standard surveillance protocol recommended (imaging with or without AFP), limited to patients with cirrhosis. Methods We performed a retrospective analysis in prospectively collected specimens from an ongoing cohort. We designed an early sensitive high‐risk test (LCR1) that combined (using Cox model) hepatoprotective proteins (apolipoproteinA1, haptoglobin) with known risk factors (gender, age, gammaglutamyltranspeptidase), and a marker of fibrosis (alpha2‐macroglobulin). To increase the specificity, we then combined (LCR2) these components with alpha‐fetoprotein. Results A total of 9892 patients, 85.9% without cirrhosis, were followed up for 5.9 years [IQR: 4.3‐9.4]. LCR1 and LCR2 time‐dependent AUROCs were not different in construction and validation randomised subsets. Among 2027 patients with high‐LCR1 then high‐LCR2, 167 cancers (113 with cirrhosis, 54 without cirrhosis) were detected, that is 12 patients needed to screen one cancer. The negative predictive value was 99.5% (95% CI 99.0‐99.7) in the 2026 not screened patients (11 cancers without cirrhosis) higher than the standard surveillance, which detected 113 cancers in 755 patients screened, that is seven patients needed to screen one cancer, but with a lower negative predictive value 98.0% (97.5‐98.5; Z = 4.3; P