Risk prediction of hepatobiliary and pancreatic cancers in elderly Chinese: The Dongfeng-Tongji cohort.

Abstract

Hepatobiliary and pancreatic (HBP) cancers are among the deadliest malignancies. The objective of the study is to build cost-effective models to identify high-risk individuals for early diagnosis and substantially to reduce the burden of HBP cancers. Based on the prospective Dongfeng-Tongji cohort with ∼6 years follow-up, we identified 162 incident cases of hepatocellular carcinoma (HCC), 53 of biliary tract cancer (BTC), and 58 of pancreatic cancer (PC). We matched three controls to each case by age, sex, and hospital. We applied conditional logistic regression to identify predictive clinical variables, from which we constructed clinical risk scores (CRSs). We evaluated the utility of CRSs in stratifying high-risk individuals by 10-fold cross-validation. Among 50 variables we screened, 6 were independent predictors of HCC, with the top ones being hepatitis (OR= 8.51, 95% CI (3.83, 18.9)), plateletcrit (OR= 0.57, 95% CI (0.42, 0.78)), and alanine aminotransferase (OR= 2.06, 95% CI (1.39, 3.06)). Gallstone (OR= 2.70, 95% CI (1.17, 6.24)) and direct bilirubin (OR= 1.58, 95% CI (1.08, 2.31)) were predictive of BTC, while hyperlipidemia (OR= 2.56, 95% CI (1.12, 5.82)) and fasting blood glucose (OR= 2.00, 95% CI (1.26, 3.15)) were predictive of PC. The CRSs achieved AUCs of 0.784 for HCC, 0.648 for BTC, and 0.666 for PC, respectively. When applying to the full cohort with age and sex included as predictors, the AUCs were increased to 0.818, 0.704, and 0.699, respectively. Disease history and routine clinical variables are predictive of incident HBP cancers in elderly Chinese.