Diagnosing fibrosis in hepatitis C: Is the pendulum swinging from biopsy to blood tests? (see Editorial below by Nezam Afdahl, MD)
Noninvasive prediction of fibrosis in patients with chronic hepatitis C
Hepatology, May 2003, Volume 37, Number 5. Thierry Poynard, M.D., Ph.D. et al.
To the editor:
In a cohort of 476 untreated patients with chronic hepatitis C virus (HCV), Forns et al. (Forns X, Ampurdanes S, Llovet JM, Aponte J, Quinto L, Martinez-Bauer E, et al. Identification of chronic hepatitis C patients without hepatic fibrosis by a simple predictive model. Hepatology 2002;36:986-992) recently developed an index that accurately identified patients with minimal hepatic fibrosis (METAVIR stage < F2). This index, including age, platelets, -glutamyl-transpeptidase, and cholesterol had good discriminative power, as shown by an area under the receiver operating characteristic (ROC) curve of .81 in the validation cohort (n = 125). As discussed by the authors, our group also has derived and validated a fibrosis index (Fibrotest; Biopredictive, Houilles, France) that includes 5 biochemical markers (apolipoprotein A1, haptoglobin, 2-macroglobulin, -glutamyl transpeptidase, and total bilirubin) and shows high predictive values.
We sought to compare these indices by using data collected retrospectively in 249 patients from our original population with complete biochemical data. In
this group, the prevalence of F2-F4 fibrosis was 38%, and HCV genotypes (n = 136 with available data) were 1 (63%), 2 (13%), 3 (19%), 4 (4%), and 5 (2%). The
areas under the ROC curves (±SE) for the identification of F2-F4 fibrosis were 0.78 ± 0.03 for the Forns index versus 0.84 ± 0.02 for the Fibrotest (P = .02),
suggesting that the latter has greater discriminative power.8 Moreover, an index including the 4 variables of Forns et al. derived in our population by using stepwise forward multiple logistic regression yielded nearly identical results (area under the ROC curve, 0.78 ± 0.03). There was a continuous, almost linear, relationship between Fibrotest and the fibrosis stage, with significant differences between stages that were not observed for the Forns et al.
In addition to superior diagnostic power, the Fibrotest has several advantages over the index of Forns et al.1 First, the Fibrotest is not genotype dependent,
whereas the Forns et al.1 index includes serum cholesterol, which varies with HCV genotype. Patients with genotype 3 have lower cholesterol levels, which
increase after viral eradication.9,10 Although 85% of the Spanish cohort had genotype 1 (genotype 3 prevalence was not reported), the prevalence in most
populations, including our own, is lower. In countries where genotype 3 is more common (e.g., in Asia), use of this index may not be advisable.
Second, the Forns et al.1 index includes the platelet count, which is poorly standardized between laboratories. We did not include platelets in the Fibrotest so as to avoid this limitation and develop an index that is easily automated using minimal equipment. Apolipoprotein A1, haptoglobin, and 2-macroglobulin, the nonroutine components of the Fibrotest, can be measured using a single autoanalyzer and have minimal interlaboratory variability (coefficients of variation < 6%).
Third, the Fibrotest has been validated in several cohorts, including patients with human immunodeficiency virus/HCV-coinfection, who often have human
immunodeficiency virus- or medication-related thrombocytopenia. Moreover, the Fibrotest is responsive to changes in fibrosis attributable to interferon-based therapy. Finally, a combination of alanine aminotransferase levels and the Fibrotest (Actitest; Biopredictive) accurately predicts the severity of
necro-inflammatory activity, another important histologic feature. Both tests are easy to perform and are available on the internet at www.biopredictive.com.
In conclusion, this retrospective comparison suggests that the Fibrotest provides a more accurate estimate of HCV-related fibrosis than the index of Forns et al. Prospective comparisons in additional centers are warranted to confirm this finding.
Nezam H. Afdhal, M.D.
Liver Center, Beth Israel Deaconess Medical Center, Boston, MA
One of the major clinical problems facing the hepatology and gastroenterology community is how best to evaluate and manage the increasing numbers of
patients identified with hepatitis C virus (HCV). In the last decade, advances in serologic and virologic testing for HCV and improvements in therapy have led
more patients to be identified and to seek treatment. However, little progress has been made in improving either our ability to determine the degree of hepatic injury, particularly fibrosis, or to predict the risk of disease progression for the individual patient. This information still requires an old fashioned liver biopsy. The clinician relies on the biopsy results for both prognostic and therapeutic decision making, which can have a major impact on the patient's life. While we certainly are daunted at the prospect of performing 3 million liver biopsies with the associated cost, manpower issues, and risk for patient injury, we need to have a reliable alternative that can just as effectively guide our decision analysis.
Multiple histologic scoring systems have been proposed for the grading of HCV inflammation and the staging of HCV fibrosis, including the commonly used
METAVIR and Ishak scoring systems. The pathologist is able to use these systems to grade the inflammatory component of HCV and stage the degree of
fibrosis. Repeat liver biopsies can determine the effect of therapy on improving liver inflammation and fibrosis and the true rate of disease progression in
individuals who decline therapy. Perhaps the first question we should ask is, how good is our gold standard of liver biopsy for staging HCV-induced liver fibrosis? A single-pass liver biopsy is able to correctly diagnose the stage of fibrosis or presence of cirrhosis in 80% of patients.5 Factors that improve the diagnostic accuracy of liver biopsy include the presence of a uniform disease throughout the liver such as HCV, multiple passes, use of a Trucut 15-gauge needle rather than Menghini type needles, and an unfragmented biopsy core of 2 cm or greater in length. Even with experienced physicians performing the liver biopsy and expert pathologists interpreting the biopsy, our gold standard has up to a 20% error rate in staging disease.
In addition, the actual role of liver biopsy in HCV is controversial. The liver biopsy is not necessary to make the diagnosis and rarely is it of value in excluding secondary diagnoses such as coexistent, autoimmune, iron, nonalcoholic steatohepatitis or alcohol-induced injury. The major clinical utility of the index biopsy in HCV is to enable the clinician to determine the need for therapy. Because of the complexity and side effects of interferon-based therapies for HCV, the liver biopsy has taken an increasing role in the clinician's decision whether to treat a patient. Patients with significant liver fibrosis (METAVIR >F2 or Ishak >F3) or moderate inflammation are all considered as suitable for therapy, whereas patients with milder disease are often not as aggressively offered treatment. However, as therapy for HCV improves, the clinical need for a biopsy may be less apparent. For example, genotype 2 and 3 patients have a greater than 70% sustained virologic response with pegylated interferons and ribavirin, and in uncomplicated cases a rationale can be made to treat all these patients without liver biopsy and only to biopsy those who fail treatment. As newer, better tolerated, and more efficacious therapies are developed, the need for biopsying all HCV patients to grade and stage disease may become redundant. Therefore, the development of noninvasive tests that can differentiate between patients with mild disease (METAVIR F0 or F1) versus those with more significant fibrosis (METAVIR F2-F4) could have a widespread clinical utility in managing HCV patients in the future.
This concept of attempting to use noninvasive serum tests to classify patients as having mild or significant liver fibrosis was addressed in a recent publication
by Forns et al. and in several letters in this issue of HEPATOLOGY. Forns examined a selected cohort of patients with HCV and, by using multiple logistic
regression, identified age and 3 commonly performed serum tests that could be used to predict patients with F0/F1 disease. They derived a formula giving a
numerical score in an initial population of 351 HCV patients, which was based on the patient age combined with single estimations of glutamyl transpeptidase (GGT), cholesterol level, and platelet count. This score was then validated in a prospective group of 125 patients, in which 92 patients had mild fibrosis and 33 patients had significant fibrosis. A score of less than 4.2 was seen in 47 patients of whom 45 (96%) had stage F0/F1 giving an overall accuracy of 96% in identifying patients with mild fibrosis. A score of greater than 6.9 was taken as the high cutoff point for significant fibrosis. Only 10 of 33 patients with stage F2 to F4 were identified by this scoring system, and 15 patients with F0 to F1 were incorrectly classified as having advanced disease. The authors concluded that this relatively simple, inexpensive fibrosis scoring system could prevent the need for a liver biopsy in one third of patients with mild disease.
Certainly, on preliminary analysis this looks promising but there are several important considerations. First, the patient population was highly selected to
exclude patients over 65 years of age, those with regular alcohol consumption, obese patients, and patients coinfected with HBV or human immunodeficiency
virus. More importantly, only 51% of the patients studied could actually be classified, and the scoring system was indeterminate for the remainder of the
patients who scored between 4.2 and 6.9. Therefore, the widespread applicability of such a fibrosis scoring system might be limited.
Interestingly, two separate groups have validated the findings of Forns et al. in separate cohorts of HCV patients. Patel et al. in their letter in this edition of
HEPATOLOGY applied the Forns scoring system to 110 HCV patients. A score of less than 4.2 had a negative predictive value of 89%, but 59% of patients were
again in the indeterminate group and could not be classified. Thabut et al. in their letter compare the Forns score to the previously published FIBROTEST.12
The FIBROTEST utilizes 5 less commonly used biomarkers for fibrosis including apolipoprotein A1, haptoglobin, alfa 2 macroglobulin, GGT, and total
bilirubin. In their retrospective analysis, the FIBROTEST appeared slightly better at diagnosing significant fibrosis and showed a more linear correlation with degree of fibrosis than the Forns scale. Both FIBROTEST and the Forns scale were excellent at excluding fibrosis, but the FIBROTEST had a greater positive predictive value (90%) for diagnosing advanced fibrosis. However, it is critical to point out that neither of these tests can truly distinguish between the different histologic stages of fibrosis. In addition, the FIBROTEST has similar issues with a large percentage of patients falling into an indeterminate group. The inability to classify large numbers of patients is not surprising because neither of these systems really measures fibrosis but they both rather reflect alterations in hepatic function associated with progressive disease. More recently, combining markers of extracellular matrix (ECM) such as tissue inhibitor of metalloproteinase and hyaluronic acid with alfa 2 macroglobulin, a marker of hepatic function, has been evaluated by Patel et al. This combination of fibrosis markers has a similar ability to differentiate F0/F1 from F2 to F4 and reduces the number of patients (20%-30% depending on fibrosis prevalence) falling into the indeterminate range.
First, all of these investigators should be congratulated for pursuing clinically relevant noninvasive methods to measure liver fibrosis. There is little doubt that we need such surrogate fibrosis markers or scoring systems for HCV, particularly as we start to develop a new generation of antifibrotic therapies. Several large studies are already in progress to evaluate both alfa and gamma interferons as antifibrotic agents in HCV and all are relying on serial liver biopsies as the primary outcome determinant. However, biopsies remain a static determinant of the mass of fibrosis at a given time, and an antifibrotic agent could be highly effective in ECM remodeling but show little change on biopsy. This is where the utility of true fibrosis markers is critical. An ideal noninvasive surrogate marker of liver fibrosis should be liver specific; not influenced by alterations in liver, renal, or reticuloendothelial cell function; reproducible; and easy to perform.
These markers should also reflect the stage of fibrosis for diagnostic purposes and the dynamic balance of matrix deposition and removal. Finally, they should
be specific for liver fibrosis from any cause and be independent of any associated hepatic inflammation. Advances in our understanding of the hepatic
ECM has led to many potential markers being identified, but none have sufficient merit for clinical use as solitary markers. A large multicenter European study combining multiple markers that measure ECM production and breakdown has shown promise in the diagnostic staging of liver disease, and further longitudinal studies from this cohort are awaited.
So, are surrogate markers of liver fibrosis ready for clinical primetime? We could argue from the study by Forns et al. and the FIBROTEST that we can, to some extent, categorize almost a third of patients into those with mild disease and use this information for decision analysis without a liver biopsy. Unfortunately, there is somewhat of a rush to commercialize these so-called tests of liver fibrosis without rigorous scientific validation. I would propose that the clinical acceptance of fibrosis markers will depend on their fulfilling the ideal characteristics listed above. We need to focus basic and clinical research efforts on identifying better fibrosis markers and exploring how to incorporate them into clinical practice. The study of Forns et al. and others are a good beginning in our quest for surrogate markers, but for the moment, the pendulum still favors biopsy over blood.