Diagnostic Estimation of Noninvasive Tests for Hepatic Fibrosis in Chronic Hepatitis B Patients Without a Gold Standard.

BACKGROUND Assessment of hepatic fibrosis stage in patients with chronic hepatitis B (CHB) is indispensable for prognosis evaluation and therapeutic regime. Noninvasive tests are fast, safe and cheap and need low technical requirements for diagnosing hepatic fibrosis in CHB patients. OBJECTIVES Using the latent class model with a random-factor to estimate relative accuracy of noninvasive tests for the diagnosis of hepatic fibrosis without a gold standard in a large population with CHB. PATIENTS AND METHODS A total of 544 patients with CHB were assessed for fibrosis stage by four noninvasive tests containing liver stiffness measurement (LSM), aspartate aminotransferase-to-platelet ratio index (APRI), fibrosis index based on 4 factors (FIB-4) and globulin and platelet (GP). The diagnostic evaluation was made by the latent class method with random effect which analyzed the clinical data above to assess the accuracy of four ways of noninvasive diagnosis. RESULTS The latent class model with random effect permitted to conciliate the observed data and estimates of test performances. For significant fibrosis, the specificity/sensitivity were 83.24%/91.59% (APRI), 90.05%/95.57% (FIB-4), 75.11%/66.01% (LSM) and 71.13%/98.33% (GP), respectively. For cirrhosis, the specificity/sensitivity were 84.04%/17.91% (APRI), 89.86%/17.09 (FIB-4), 78.64%/37.07% (LSM) and 82.28%/37.07% (GP), respectively. CONCLUSIONS FIB-4 confirmed the best value for diagnosis of significant fibrosis. APRI had a sub-optimal diagnosis accuracy for significant fibrosis. LSM showed the most balance diagnosis value for cirrhosis with the highest sensitivity and moderate specificity.