We read with interest the review by Martínez et al. on noninvasive assessment of liver fibrosis,1 but some important points were not discussed.
First, the authors mention that the relatively high misclassification of the “gold standard” (liver biopsy) makes it impossible for noninvasive tests to achieve high concordances. This is indeed true for noninvasive tests whose development was independent from liver histology (e.g., elastography). However, serum markers have been calibrated with direct reference to sets of liver biopsies. Therefore, the perfect serum marker would replicate even the misclassifications of the “golden” histological standard and could theoretically reach an AUROC (area under the receiver operating characteristic curve) of 1.
Second, the explosive development and overenthusiastic acceptance of noninvasive markers is not always supported by sufficient evidence and validation. In our meta-analysis on elastography, we exposed issues such as invalidated stiffness cutoffs for specific liver disease stages and low methodological quality in the vast majority of published studies.2 In numerous studies, the maximum interval between elastography and liver biopsy was >3-6 months; in such cases, there is an erroneous assumption that fibrosis remains stable over these periods of time. Furthermore, from all publications, only six studies had both optimal histological and elastography measurements.2
Third, the authors correctly state that liver biopsy is more of a reference standard than a gold standard for assessing fibrosis. As we have pointed out before, histological “scores” of fibrosis are ordinal categories that incorporate both architectural changes and fibrosis, and have no quantitative relationship between them.3 Therefore, using them as continuous variables is inappropriate.3 Validation of noninvasive markers of “fibrosis” should ideally use quantitative histological measures. We have described such a measure, namely, collagen proportionate area, and correlated it with hepatic venous pressure gradient.4 More importantly, we evaluated its prognostic value with respect to patient outcome.5 Notably, collagen proportionate area performed better than Ishak staging and hepatic venous pressure gradient for predicting decompensation (AUROC = 0.97).5
In conclusion, the way forward involves carefully designed studies that validate noninvasive fibrosis markers against quantitative histological measures and/or clinical outcomes. Because we are ultimately treating patients, the clinical consequences of false positive and false negative classifications should be incorporated in the validation processes.