Liver Biopsy Analysis Has a Low Level of Performance for Diagnosis of Intermediate Stages of Fibrosis

BACKGROUND & AIMS: There is controversy about the performance of noninvasive tests such as FibroTest in diagnosing intermediate stages of fibrosis. We investigated whether this controversy results from limitations of biopsy analysis for intermediate-stage fibrosis and inappropriate determination of the standard area under the receiver-operator characteristic curve (AUROC). METHODS: To determine whether biopsy has a lower diagnostic performance for fibrosis stage F2 (few septa) vs F1 (fibrosis without septa), compared with its performance for F1 vs F0 or F4 vs F3, we determined the fibrotic areas of large surgical samples collected from 20 consecutive patients with chronic liver disease or normal liver tissue that surrounded tumors. We analyzed digitized images of 27,869 virtual biopsies of increasing length and also analyzed data from 6500 patients with interpretable FibroTest results who also underwent biopsy analysis. RESULTS: The overall performance of biopsy analysis (by Obuchowski measure) increased with biopsy length from 0.885 for 5-mm to 0.912 for 30-mm samples (P < .0001). The performance of biopsy was lower for the diagnosis of F2 vs F1 samples (weighted AUROC [wAUROC] = 0.505) than for F1 vs F0 (wAUROC = 0.773; 53% difference; P < .0001) or F4 vs F3 (wAUROC = 0.700; 39% difference; P < .0001), even when 30-mm biopsy samples were used. The performance of FibroTest was also lower for the diagnosis of F2 vs F1 samples (wAUROC = 0.512) than for F1 vs F0 samples (wAUROC = 0.626; 22% difference; P < .0001) or F4 vs F3 (wAUROC = 0.628; 23% difference; P < .0001). However, the FibroTest had smaller percentage differences among wAUROC values than biopsy. CONCLUSIONS: Biopsy has a low level of diagnostic performance for fibrosis stages F2 and F1. The recommendation for biopsy analysis, instead of a validated biomarker panel such as FibroTest, for the diagnosis of intermediate stages of fibrosis is therefore misleading.