Identification of liver fibrosis in HIV/HCV-coinfected patients using a simple predictive model based on routine...

We constructed noninvasive models to predict significant fibrosis (F > or = 2) and advanced fibrosis (F > or = 3) among human immunodeficiency virus (HIV)/hepatitis C virus (HCV)-coinfected patients, naïve for anti-HCV treatment. A total of 296 patients with liver biopsy were randomly assigned to an estimation group (EG = 226; 70%) and a validation group (VG = 70; 30%). We developed the Hospital Gregorio Marañón (HGM)-1 index, based on platelet count, aspartate aminotransferase (AST) and glucose, to predict F > or = 2 and the HGM-2 index, based on platelet count, international normalized ratio, alkaline phosphatase and AST to predict F > or = 3. The area under the receiver operating characteristic curves (AUROCs) of the HGM-1 index for the EG and the VG were 0.807 and 0.712 respectively. The AUROCs of the HGM-2 index for the EG and the VG were 0.844 and 0.815 respectively. With the HGM-1 index applied to the VG, using best cutoff scores, the negative predictive value (NPV) to exclude F > or = 2 was 54.5% and the positive predictive value (PPV) to confirm F > or = 2 was 93.3%. With the HGM-2 index applied to the VG, using best cutoff scores, the NPV to exclude F > or = 3 was 92.3, and the PPV to confirm F > or = 3 was 64.3%. Thus, HGM-2 accurately predicted F > or = 3 among HIV/HCV-coinfected patients. HGM-1 was less accurate at predicting F > or = 2.