Diabetes Mellitus Impacts on the Performance of Hepatocellular Carcinoma Risk Scores in Chronic Hepatitis B Patients

Abstract

We examined whether changing clinical characteristics and presence of diabetes mellitus (DM) impact the performance of hepatocellular carcinoma (HCC) risk scores. Adult patients with chronic hepatitis B (CHB) on ≥6 months of entecavir/tenofovir treatment between January 2005 and March 2020 were identified using a territory-wide electronic database in Hong Kong. DM was defined by antidiabetic agents, hemoglobin A1c ≥6.5%, fasting glucose ≥7mmol/L, and/or diagnosis codes. PAGE-B, modified PAGE-B (mPAGE-B), and aMAP scores were assessed by area under the time-dependent receiver operating characteristic curves (AUROCs) and compared to CAMD and REAL-B scores with DM as a component. Of 48,706 patients, 2,792, 11,563, 15,471, and 18,880 started entecavir/tenofovir treatment between 2005-2008, 2009-2012, 2013-2016, and 2017-2020 respectively; DM prevalence rose from 15.5% in 2005-2008 to 24.3% in 2017-2020. AUROCs were comparable across the four periods in the five HCC risk scores (AUROCs ranged between 0.75-0.81). At a median follow-up of 4.4 years, 1,512 (4.0%) non-diabetic and 645 (6.2%) diabetic patients developed HCC. AUROC of all five scores were lower in diabetic patients than in non-diabetic patients (AUROCs ranged between 0.67-0.71 vs. 0.78-0.82; all P<0.001). REAL-B score achieved an AUROC of 0.71 in diabetic and 0.82 in non-diabetic patients. Both diabetic and non-diabetic patients in the low-risk group by REAL-B score had a low HCC incidence below the threshold of cost-effective HCC surveillance, i.e. 0.2% annually. REAL-B score is accurate and preferred in entecavir/tenofovir-treated CHB patients given the increasing prevalence of DM.