LAUSR

We read with interest the letter that the editor of JMRI received regarding our article, “Differentiation of focal nodular hyperplasia (FNH) from hepatocellular adenoma (HCA): Role of the quantitative analysis of gadobenate dimeglumine-enhanced hepatobiliary phase MRI.” In our study, the mean Liver to Lesion Contrast Enhancement Ratio (LLCER) of FNH—corresponding to the quantification of relative Gd-BOPTA uptake within a selected region of interest (ROI) compared to that of adjacent liver—was significantly higher than that of HCA (P < 0.0001). The area under the ROC curve (AUC) for the differentiation of FNH from HCA with LLCER was 0.98 for both observers. With a cutoff value of –0.3% for observer 1 and 1.2% for observer 2, LLCER assessment provided respective sensitivity and specificity values of 100% and 96.2% for the differentiation of FNH from HCA. The intraclass correlation coefficient (ICC) was 0.8. In an external validation of our findings, the authors found that LLCERs were significantly higher in FNH (P < 0.001), but only for observer 1. Using our method, they could not reproduce our optimal cutoffs: Optimal cutoff values for LLCER and their corresponding sensitivity and specificity were –3.4%, 75.0%, and 89.6% for observer 1, and –15.3%, 75.0%, and 47.9% for observer 2. ICC was 0.58. Using their alternative method, measuring only regions of maximal enhancement, LLCERs were significantly higher in FNH than HCA (P < 0.001) for both observers. The AUC for differentiating HCA and FNH was 0.90 for observer 1 and 0.87 for observer 2. Optimal cutoff values for LLCER and their corresponding sensitivity and specificity were –3.0%, 85.7%, and 91.7% for observer 1, and –6.7%, 82.1%, and 83.3% for observer 2. The ICC was 0.84. The precise methodology used by these authors is not mentioned in their letter. Yet several causes could explain these different results.