Based on the limitations of liver biopsy and the present overview of the diagnostic value of FT-AT, it seems that these non-invasive markers should be used as a first line assessment of liver injury in patients with chronic hepatitis C.
Liver biopsy has three major limitations, which are the risk of adverse events [2, 3, 7], sampling error [4–6], and inter- and intra- pathologist variability . An overview of published studies summarizes the risks of liver biopsy as pain (around 30%), severe adverse events (3/1,000) and death (3/10,000) [2, 3, 7]. Sampling variation is the major cause of variability [4–6]. In a study of patients with chronic hepatitis C that included only good quality biopsies, 30 of 124 patients (24.2%) had a difference of at least one grade, and 41 of 124 patients (33.1%) had a difference of at least one stage between the right and left lobes . In 18 patients (14.5%), an interpretation of cirrhosis was made in one lobe, whereas stage 3 fibrosis was made in the other . Recently, Bedossa et al.  observed very high coefficients of variation (55%) and high discordance rates (35%) for fibrosis staging in biopsies measuring 15 mm in length. The variability significantly improved in biopsies measuring 25 mm in length but was still very high with a 45% coefficient of variation and 25% discordance rate; the minimal variability was reached for biopsies, which were 40 mm in length .
Liver biopsy has also potential advantages. Biopsy could be of diagnostic value for other unrecognized liver disease. These events are probably rare in practice, as we observed no such a case in a prospective study of 537 consecutive patients with chronic hepatitis C . For FT-AT it must be realized that the same predictive values were observed for patients coinfected with HIV , and in patients with other causes of liver fibrosis such as chronic hepatitis B , alcoholic liver disease  or non-alcoholic steato-hepatitis .
It is possible that biochemical markers such as those described here may provide a more accurate (quantitative and reproducible) picture of fibrogenic and necrotic events occurring within the liver than hepatic biopsy. The greater accuracies of FT-AT, when assessed with biopsy specimens greater than 15 mm versus smaller biopsies, suggest that some discordance between FT-AT and histology were due to biopsy specimen sampling error . Several case reports have observed false negatives of liver biopsy versus biochemical markers [8, 9, 11]. The error was attributable to biopsy because there were overt clinical signs of cirrhosis such as esophageal varices, low platelet counts or a dysmorphic liver on ultrasound. In a recent prospective study we estimated that 18% of discordances between FT-AT and histology were attributable to biopsy failure (mostly due to small length) and 2% to FT-AT failure .
The present work allowed frequently asked questions to be answered, the first being whether the diagnostic values of FT-AT had been confirmed in all studies performed to date. A major strength of the studies pertaining to FT-AT is that they were carried out on a large number of patients with chronic hepatitis C, and the results were reproducible in different populations, including patients coinfected with HIV. There was a small variability in the AUROCs, both for the diagnosis of significant fibrosis (0.73 to 0.87) and significant necrosis (0.75 to 0.86).
A weakness of this study was that the same group, which developed these tests, performed most of the published studies. However the independent published studies found the same significant diagnostic values than non-independent or multicentre studies. Several recent independent studies confirmed the predictive value of FT-AT [26, 30].
The second question concerned the comparison of FT-AT to other tests. In their recent review, Gebo et al.  concluded that panels of markers might have the greatest value in predicting the absence or no more than minimal fibrosis on biopsy, and in predicting the presence of cirrhosis on biopsy (Evidence Grade B). They pointed out that five studies [11, 32–35] used large panels of markers and achieved the greatest predictive values. Among these 5 studies were the first FT-AT study  and another study developed by the same group (combining age and platelets) . A recent study compared FT-AT to the age and platelets index in the same patients and found that FT-AT was significantly better . Three studies directly compared FT-AT, to hyaluronic acid , the Forns index  and the Wai index  in the same patients. FT-AT had higher diagnostic values (the AUROC was significantly higher). FT was in particular more sensitive for discriminating between F1 and F2, and more linearly correlated to stages when compared to those 3 other markers [12, 16, 17]. An additional weakness of the Forns index is the inclusion of cholesterol, which varies greatly in patients with genotype 3 . The limitations of these three comparisons [12, 16, 17] are that they were retrospective and were performed by the same group. These comparisons, however, had no evident sources of bias. The comparison with the Forns Index  included all patients of the Imbert-Bismut et al. study (n = 323) , as the parameters belong to the routine biochemical tests. The comparison with the APRI index included 249/323 patients (77%) without any difference between included or non-included patients when all characteristics were compared . The comparison with hyaluronic acid  included a total of 165 out of the 244 (68%) randomized patients pre-included. The 165 included patients did not differ from the 79 non-included patients according to the main characteristics. Among the 165 patients, the fibrosis index was assessed in 461 samples and hyaluronic acid in 457 samples .
Recently, a study using profiles of serum protein N-glycans found that a profile has a similar AUROC than FT for the diagnosis of compensated cirrhosis. When combined with FT this marker had 100% specificity and 75% sensitivity for the diagnosis of compensated cirrhosis, which is not significantly different from the 92% specificity and 67% sensitivity of the FT . This study was independent and prospectively designed for taking FT as the comparison test. Only 24 patients with cirrhosis were included and no details were given concerning the causes of discordance between biopsy and biochemical markers.
However FT-AT is the only panel of markers identified by an independent overview , which has been compared in the same patients with most of the other proposed markers. No studies were found that compared FT-AT with a panel of extra-cellular matrix markers . Compared to other panels, FT-AT also allowed an estimation to be made not only of the fibrosis stage but also the necroinflammatory (histological) activity.
The present analysis of the integrated database demonstrated that the diagnostic value of FT-AT did not depend on HCV genotype or viral load. However, because of the small number of patients included, studies in genotype 4, 5 and 6 would be useful.
The present analysis also answered another frequently asked question concerning the predictive values for the intermediate stages of fibrosis. Contrary to the initial hypothesis, the diagnostic values of FT-AT for consecutive stages of fibrosis and grades of necroinflammatory activity were the same for both moderate and extreme stages and grades. Our interpretation is that the same overlap exists between all stages, which is mainly related to the sampling error of the biopsy. It is very reassuring that the medians of FT-AT are linearly associated with stages and grades (Figures 3A,3B). The linearity of this association became even more evident as a larger number of patients were included (data not shown).
Finally, the integrated database allowed a simple conversion system to be proposed to clinicians between liver injury as estimated by the FT-AT and that as estimated by liver biopsy (Figure 4). One conventional way to express the diagnostic values of FT-AT was summarized using the cutoffs of the distribution by stages and grades (Tables 2 and 4). The negative predictive value of FT for excluding significant fibrosis was excellent for the 0.31 cutoff (91%), as was the negative predictive value for excluding significant activity at the 0.36 cutoff of AT (85% negative predictive value). The positive predictive value of the 0.72 cutoff of FT for significant fibrosis was also high at 76%. This, however, may appear lower than the negative predictive value. There is a technical explanation owing to the prevalence of significant fibrosis, which was only 0.31 in this population. According to the excellent specificity (above 0.95), the positive predictive value increased rapidly in populations with more fibrosis (data not shown). We recently observed that the main reason for this was probably because most of the so-called false positives of the FT were in fact false negatives due to the small sampling size of liver biopsies [5, 9]. The same comments can be made concerning the positive predictive value of AT for significant necrosis with 77% at the 0.60 cutoff. Again, it is probable that a large proportion of so-called false positives of AT were in fact false negatives due to liver biopsies which were too small. The ideal study would be one using biopsies measuring 40 mm in length, as two samples of 20 mm each during laparoscopy. Only this very high quality biopsy can be considered as a true gold standard. Obviously this type of biopsy cannot be performed routinely as first line, but it could be recommended for clinical research.