EASL Clinical Practice Guidelines

Methodological considerations when using non-invasive tests

The performance of a non-invasive diagnostic method is evaluated by calculation of the area under the receiver operator characteristic curve (AUROC), taking liver biopsy as the reference standard. However, biopsy analysis is an imperfect reference standard: taking into account a range of accuracies of the biopsy, even in the best possible scenario, an AUROC >0.90 cannot be achieved for a perfect marker of liver disease [15]. The AUROC can vary based on the prevalence of each stage of fibrosis, described as spectrum bias [16]. Spectrum bias has important implications for the study of non-invasive methods, particularly in comparison of methods across different study populations. If extreme stages of fibrosis (F0 and F4) are over-represented in a population, the sensitivity and specificity of a diagnostic method will be higher than in a population of patients that has predominantly middle stages of fibrosis (F1 and F2). Several ways of preventing the “spectrum bias” have been proposed including the adjustment of AUROC using the DANA method (standardization according to the prevalence of fibrosis stages that define advanced (F2–F4) and non-advanced (F0–F1) fibrosis) [[17], [18]] or the Obuchowski measure (designed for ordinal gold standards) [19]. What really matters in clinical practice is the number of patients correctly classified by non-invasive methods for a defined endpoint according to the reference standard (i.e. true positive and true negative).

General statements

  • Even though liver biopsy has been used as the reference method for the design, evaluation and validation of non-invasive tests, it is an imperfect gold standard. In order to optimize the value of liver biopsy for fibrosis evaluation, it is important to adhere to the following recommendations: (i) sample length >15 mm by a 16G needle; (ii) use of appropriate scoring systems according to liver disease etiology; and (iii) reading by an experienced (and if possible specialized) pathologist.
  • Non-invasive tests reduce but do not abolish the need for liver biopsy; they should be used as an integrated system with liver biopsy according to the context.