Abstract
BACKGROUND: To improve the early diagnosis of hepatocellular carcinoma (HCC), more effective diagnostic biomarkers are needed. A combination of biomarkers is reported to distinguish individuals with early-stage HCC from at-risk individuals. METHODS: Participants in this study were recruited from six hospitals in China. Literature review was used to choose 19 candidate proteins, a case-control study in the discovery stage was used to identify five proteins (P5) that constituted a diagnostic model. In the training and validation stages, the effectiveness of P5 for detecting early-stage HCC was tested (cross-sectional study). Finally, a nested case-control study independent of the other stages was set up to evaluate the P5 in the preclinical diagnosis of HCC. FINDINGS: Between February 2013 and June 2017, a total of 1396 participants were recruited. A panel of 5 proteins (P5: OPN, GDF15, NSE, TRAP5 and OPG) showed high diagnostic accuracy when differentiating the early-stage HCC from the at-risk group, with AUCs of 0·892, 0·907 and 0·852 for the training stage, validation cohort 1 and cohort 2 data sets, respectively. In the prediction set, the sensitivity of P5 for diagnosing preclinical HCC increased with time, starting from 12 months before to the time of definitive clinical diagnosis (range, 46·15% to 86·67%). INTERPRETATION: The P5 panel has the potential to screen populations at high risk of developing HCC and can enable the early diagnosis of HCC. FUNDING: Research supported by grants from eight funds. All sources of funding were declared at the end of the text.