The accuracy of the FIB-4 index for the diagnosis of mild fibrosis in chronic hepatitis B

Authors


Dr V. Mallet, Hépatologie, Hôpital Cochin, 27 rue du faubourg Saint Jacques, 75014 Paris, France.
E-mail: vincent.mallet@cch.aphp.fr

Summary

Background  The Fib-4 index is a simple and inexpensive biomarker to delineate liver fibrosis in chronic hepatitis C.

Aim  To assess the accuracy of the FIB-4 index in chronic hepatitis B.

Methods  We compared the FIB-4 index with 138 synchronous liver biopsies and with 372 synchronous FibroTest performed either in France or in an endemic area (Mayotte, an overseas collectivity of France).

Results  The FIB-4 index allowed the correct identification of patients with nil-to-moderate fibrosis with an area under the receiving operating characteristic curve of 0.81 (P < 0.001), increasing as a function of the length of the liver biopsy (up to 0.94 for liver biopsies ≥20 mm). A cut-off value ≤1.45 differentiated moderate fibrosis from severe fibrosis with a negative predictive value of 86%, a sensitivity of 71.1% and a specificity of 73.1%. Beyond 1.45, the FIB-4 index was not informative. The FIB-4 index was more precise than the AST-to-platelet ratio index and correlated with the FibroTest in 89% of the cases (κ = 0.27, P < 0.001) to exclude severe fibrosis.

Conclusion  The FIB-4 index is a simple, accurate and inexpensive method to exclude significant liver fibrosis in chronic hepatitis B, a major advantage in HBV-endemic developing countries.

Introduction

Liver biopsy (LB) is the gold standard for the assessment of liver fibrosis. LB is, however, hampered by morbidity and mortality, sampling error with approximately 24% of false negatives for cirrhosis,1 limited usefulness for dynamic surveillance and follow-up, poor intra- and interobserver concordance between nonspecialized pathologists2 and cost considerations, especially in developing countries. Non-invasive biomarkers such as the AST-to-platelet ratio index (APRI) test,3 the Forns test,4 the Hepascore, the FibroTest5 have been developed to replace LB and other scores are being developed. Transient elastography (FibroScan; Echosens, Paris, France), a morphological method which measures liver stiffness has been developed with interesting results and good correlation with the LB and the FibroTest in the evaluation of hepatitis C virus (HCV)-related chronic hepatitis.6 These non-invasive tools have a rather good predictive positive value for the diagnosis of both nil or minimal fibrosis and extended fibrosis.7 Although being a potential alternative to LB, routine use of these non-invasive tests is hampered either by the cost of the device (FibroScan) or by the need of a specific laboratory agreement standardization of the assays for performing the test (FibroTest). Besides, all these techniques have been evaluated for the most part in chronic hepatitis C and few studies are available for the evaluation of fibrosis in chronic hepatitis B (HBV).8–12

Investigators in charge of the Apricot protocol,13 a research programme on the efficacy of peg-interferon and ribavirin in patients coinfected by the human immunodeficiency virus (HIV) and the HCV, have proposed a simple non-invasive test (called FIB-4) of liver fibrosis, using the formula: age (years) × AST [U/L]/(PLT [109/L] × (ALT [U/L])1/2).14 We have thereafter shown that the FIB-4 index is also accurate for assessing moderate or severe liver fibrosis in chronic HCV mono-infection.15

The aim of our study was to evaluate the performance of the FIB-4 index in the semi-quantitative diagnosis of fibrosis associated with chronic HBV infection. We first evaluated the performance of the accuracy of the FIB-4 index in HBV mono-infected patients in comparison with the LB and then compared the FIB-4 index with the FibroTest.

Patients and methods

Patients

Comparison of the FIB-4 index with the LB.  We reviewed our computerized files of HBV-infected patients attending either the Hepatology unit of the Centre Hospitalier Universitaire Cochin Port Royal, Paris or the Gastroenterology unit of the Centre Hospitalier Saint Joseph, Marseille (both in France) and selected all cases corresponding to the following criteria: (i) chronic hepatitis B defined by: HBsAg positivity for more than 6 months; detectable HBV-DNA with a level >105 copies/mL; persistent or intermittent elevation of ALT levels and biopsy-proven chronic hepatitis;16 (ii) no previous or concomitant anti-HBV therapy; (iii) available LB allowing a confident scoring according to the METAVIR score; (iv) laboratory assessments allowing FIB-4 calculation (ALT, AST, platelet count) performed on the same day as LB or on the preceding day; (v) absence of liver comorbidity including hepatitis delta superinfection, HCV coinfection, chronic ethanol consumption (<40 g of pure alcohol per day), haemochromatosis, Wilson disease, alpha-1-antitrypsin deficiency, auto-immune hepatitis or non-alcoholic steatohepatitis; (vi) absence of HIV coinfection; (vii) absence of immune suppression.

We used the FIB-4 index for semi-quantitative evaluation of fibrosis and compared the results with those of a LB performed at the same time.

Comparison of the FIB-4 index with the APRI and with the FibroTest.  We first compared the accuracy of the FIB-4 index and the APRI to delineate fibrosis in the previous sample.

We then selected all the HBV-infected patients attending either the Hepatology unit of the Centre Hospitalier Universitaire Cochin Port Royal, Paris, or the Centre Hospitalier de Mayotte, Mayotte (a French island in the Indian ocean), for whom FibroTest (Biopredictive SAS, Paris, France) results were available using the following criteria: (i) chronic hepatitis B defined by: HBsAg positivity for more than 6 months; detectable HBV-DNA with a level >105 copies/mL; persistent or intermittent elevation of ALT levels;16 (ii) laboratory assessment allowing FIB-4 calculation (AST, ALT, platelet count) performed on the same day as FibroTest; (iii) absence of HIV, delta and HCV infection; (iv) absence of heavy alcohol consumption (<40 g pure alcohol per day); (v) absence of other liver comorbidity including haemochromatosis, Wilson’s disease, alpha-1-antitrypsin deficiency, autoimmune hepatitis or non-alcoholic steatohepatitis; (vi) absence of immune suppression.

We compared the FIB-4 index and the FibroTest results performed on the same day.

Methods

The FIB-4 values were calculated automatically with the published formula;14 age of the patient was the age at the time of LB for the first comparison and the age at the time of FibroTest for the second comparison.

The APRI was computed the same day as the FIB-4 index with the previously published formula.3

FibroTest was computed as previously described with the following biochemical values: gamma glutamyl transferase, haptoglobin, alpha-2-macroglobulin, bilirubin and apolipoprotein A1.17

A senior pathologist analysed all the liver biopsies. The degree of activity and the extent of fibrosis were assessed using the METAVIR index composed of a two-letter and two-number coding system: A = histological activity (A0 = no activity, A1 = mild activity, A2 = moderate activity, A3 = severe activity) and F = fibrosis (F0 = no fibrosis, F1 = portal fibrosis without bridges, F2 = portal fibrosis with rare bridges, F3 = numerous bridges without cirrhosis and F4 = cirrhosis).2 The local ethics committee approved the study.

Statistical analysis

Results are presented as means (±standard deviation), counts and percentages. The FIB-4 results were compared with the METAVIR scores and the FibroTest values using contingency tables, kappa and area under the receiver operating characteristic (AUROC) curves. All Calculations were made using stata (V 10) (StataCorp LP, College Station, TX, USA). A P-value <0.05 was considered statistically significant.

Results

Comparison of the FIB-4 index and LB in chronic HBV infection

One hundred and thirty-eight patients residing in the French Metropole fulfilled the selection criteria. There were 40 women and 98 men, with a mean age of 42 (±15) years. Mean serum ALT was 105 (±116) IU/L (range: 15–790) and mean serum AST was 63 (±62) IU/L (range: 15–394); platelet counts ranged from 55 000 to 329 000 with a mean value of 178 000 (±51 000)/mm3. All patients underwent LB. The mean length of liver fragments was 17.6 (±6.8) mm (median: 17 mm, 25th percentile: 13 mm, 75th percentile: 20 mm). Regarding the distribution of fibrosis scores, 50.7% of the sample displayed a nil-to-moderate (F0–F1) fibrosis, 19.6% an intermediate fibrosis (F2) and 29.7% a severe fibrosis (F3–F4).

On the whole sample, FIB-4 values ranged from 0.21 to 12.32. Mean values increased as a function of the fibrosis score from 0.96 ± 0.39 in F0 cases to 2.97 ± 2.13 in F4.

The area under the ROC curve was 0.81 for fibrosis ≥F3 [95% confidence interval (CI) 0.74–0.88] (P < 0.001) and 0.79 (95% CI 0.7–0.88) for cirrhosis. The area under the ROC curve for fibrosis ≥F3 increased as a function of the size of the liver biopsies from 0.66 (95% CI 0.46–0.85) (P = 0.177) for the first quartile (LB length below 13 mm) to 0.94 (95% CI 0.84–1) (P < 0.001) for the last quartile (LB length over 20 mm) (Figure 1).

Figure 1.

 Receiving operating characteristic curve of the FIB-4 index for the diagnosis of severe fibrosis (F3–F4) according to the size of liver biopsy (LB) samples.

The area under the ROC curve for fibrosis ≥F3 decreased as a function of the ALT level from 0.914 (95% CI 0.82–1) (P < 0.0001) for the first quartile (ALT level below 40 IU/L) to 0.7 (95% CI 0.53–0.88) (P = 0.05) for the last quartile (ALT level above 122 IU/L).

The FIB-4 values were placed in two classes (FIB-4 ≤ 1.45, FIB-4 > 1.45) and METAVIR fibrosis scores were grouped into two categories (F0–F1–F2, F3–F4). The results presented in Tables 1 and 2 show the diagnostic performances of the FIB-4 index with respect to fibrosis. Using the threshold value of 1.45 to distinguish between severe (F3–F4) and no, moderate or intermediate fibrosis (F0–F2), the predictive value to exclude significant fibrosis was 86.25% with a sensitivity of 71.1% and a specificity of 73.1%. For FIB-4 ≤ 0.7, no patient had severe fibrosis; therefore, the negative predictive value of severe fibrosis regarding a result in this group was 100%. Beyond 1.45, the FIB-4 index was not informative.

Table 1.   Comparison of FIB-4 index and liver biopsy results
 Fibrosis score (METAVIR), n (%)Total
F0–F2F3–F4
  1. * All these patients had a low platelet count and were young.

FIB-4 ≤ 1.4569 (86.2)11 (13.8)*80
FIB-4 > 1.4528 (48.3)30 (51.7)58
Total9741138
Table 2.   Diagnostic performances of the FIB-4-index to exclude significant liver-fibrosis in chronic HBV-infection when compared to liver biopsy
 Estimate95% Confidence interval
  1. AUROC, area under the receiving operating characteristic curve.

Sensitivity0.710.61–0.79
Specificity0.730.58–0.84
Positive predictive value0.860.77–0.92
Negative predictive value0.520.39–0.64
Likelihood ratio (positive)2.651.57–4.47
Likelihood ratio (negative)0.40.27–0.57
AUROC0.810.74–0.88

We focused on the cases misclassified by the FIB-4 index. Eleven patients (8.6% of all patients) who had severe fibrosis on the LB (F3–F4) had an underestimated FIB-4 index (≤1.45). Five patients had, as expected, a low platelet counts [mean value 139 600/mm3 (range: 118 000–163 000)] and were young [mean age 35.2 years (range: 29–49)]. The six others had unexplained elevated platelet count [mean 219 800/mm3 (range: 189 000–280 000)] despite severe fibrosis on the LB. Twenty-eight patients (20.3% of all patients) had nil-to-moderate fibrosis (F0–F2) and a discordant FIB-4 index (>1.45) related to a low platelet count [mean 132 600/mm3 (range: 55 000–162 000)] in 22 patients. Of note, the size of the LB was below 15 mm in 13 of these cases.

Comparison of the FIB-4 index and the APRI score in chronic HBV infection

In the previous population of patients, the APRI ranged from 0.15 to 8.78. When compared to the LB, mean values of the APRI increased as a function of the fibrosis score from 0.46 ± 0.16 in F0 cases to 1.29 ± 1.05 in F4. The area under the ROC curve of the APRI for fibrosis ≥F3 was 0.73 (95% CI 0.64–0.81); P < 0.001. Using the threshold value of 0.5 to distinguish between severe (F3–F4) and no, moderate or intermediate fibrosis (F0–F2), the predictive value to exclude significant fibrosis was 88.5% with a sensitivity of 47.4% and a specificity of 85.4%. As shown in Figure 2, the APRI was less efficient than the FIB-4 index to differentiate mild fibrosis from severe fibrosis (P < 0.02).

Figure 2.

 Receiving operating characteristic curve of the FIB-4 index and AST-to-platelet ratio index (APRI) score for the diagnosis of severe fibrosis (F3–F4); P < 0.02 for the difference.

Comparison of the FIB-4 index and the FibroTest in chronic HBV infection

Three hundred and seventy-two FibroTest and FIB-4 results corresponding to 329 patients (177 men and 152 women) were available for this analysis. The sample comprised 145 patients who were followed up in Mayotte and 184 in Paris (Table 3). Patients were younger and there were more women in the sample from Mayotte. Apolipoprotein A1 was higher and haptoglobin was lower in the sample from Paris. On the whole sample, FIB-4 values ranged from 0.20 to 8.19 (mean: 1.22). FibroTest values ranged from 0.03 to 0.97 (mean 0.32).

Table 3.   Characteristics of patients from Paris and from Mayotte
 Paris (n = 184)Mayotte (n = 145)P
Age (years)44 ± 1430 ± 9<0.001
Male gender, n (%) 123 (67)54 (37.2)<0.001
Aspartate aminotransferase (IU/L)39 ± 4244 ± 620.392
Alanine aminotransferase (IU/L)53 ± 7159 ± 1100.560
Platelets (/mm3)196 ± 53200 ± 710.49
Alpha-2-macroglobulin (g/L)2.4 ± 0.82.4 ± 0.20.775
Apolipoprotein A1 (g/L)1.4 ± 0.31.3 ± 0.3<0.001
Haptoglobin (g/L)0.9 ± 0.51 ± 0.50.003
Bilirubin (μmol/L)11.7 ± 5.710.7 ± 8.40.201

From the 282 cases with a FIB-4 score ≤ 1.45, 251 (89%) were in agreement with the FibroTest results to exclude severe fibrosis (F3–F4) (κ = 0.27, P < 0.001). Beyond 1.45, there was no relation between the FIB-4 index and the Fibrotest.

The FIB-4 index underestimated significant fibrosis (Fib-4 index ≤1.45; FibroTest F3–F4) in 31 patients. The post-hoc review of the files showed that these discordant cases were from Mayotte in 13 cases (9%) and from Paris in 18 cases (10%) (P = 0.849) and could be imputed to a FibroTest failure in 19 cases (comprising nine patients from Mayotte; 6% of those patients) because of an isolated elevation of nonconjugated bilirubin (n = 9; four from Mayotte), a low haptoglobin (n = 7; five from Mayotte) and a high alpha-2-macroglobulin (n = 3; all from Paris). In five cases, the discordances could be imputed to a FIB-4 index failure because of young age (<35 years old). In these patients, the platelet count was always below 165 000/mm3. In the remaining seven cases, there was no clear explanation for the discordance.

Discussion

Non-invasive methods have been proposed to overcome the limits of the LB to assess liver-fibrosis. Most of these non-invasive tools have a rather good negative predictive value to exclude extended fibrosis.3–6 Ideally, a non-invasive test should be easy to perform, inexpensive and reproducible. None of the available biomarkers corresponds to such a definition and most of them have been evaluated in chronic hepatitis C.

Some have been tested in chronic hepatitis B.8, 10, 12 The FibroTest, the most widely available technique, has an AUROC curve of 0.77 to identify F2–F4 fibrosis stages with a negative predictive values of 92%.8 A marker combining age, gamma-glutamyltransferase (GGT), alpha-2-macroglobuline and hyaluronic acid reported an AUROC curve comprised between 0.77 and 0.84 to differentiate minimal fibrosis from moderate-to-severe fibrosis with a negative predictive value to exclude severe fibrosis comprised between 86% and 90.9%.10 Another marker combining body mass index, platelet count, serum albumin and total bilirubin levels reported a negative predictive value of 92% to exclude significant fibrosis.11 Transient elastography has an AUROC curve of 0.96 for the diagnosis of cirrhosis. With a cut-off value of 17.6 kPa, patients with cirrhosis are detected with a positive predictive value and a negative predictive value of 90%.12 All of these techniques are ever expensive or rely on nonroutine laboratory tests barely available in developing countries.

In this work, we show that a very simple test combining age, AST, ALT and platelet levels, the FIB-4 index, can easily exclude significant fibrosis in chronic hepatitis B. In the first step of our study, we compared the results of the FIB-4 index with those of the LB, the current gold standard for assessing liver fibrosis in chronic hepatitis B. We found an AUROC curve of 0.81 to exclude severe fibrosis (F3–F4), a value comparable to previously published studies dealing with the non-invasive assessment of liver fibrosis in chronic hepatitis B.8, 10, 11 The negative predictive value to exclude severe fibrosis was excellent for FIB-4 values of ≤1.45 and even better for values <0.7. Of note, the FIB-4 index has to be interpreted with caution in young patients with a low platelet count as it can, in this setting, underestimate significant fibrosis.

To assess the efficiency of the FIB-4 index, we used the LB as a reference for fibrosis evaluation. However, LB can over- or underestimate the degree of liver fibrosis and has to be considered as an imperfect comparator.1, 18 For example, it has been shown that discordances between LB and FibroTest results can be explained by the technical difficulties involved in the examination of the LB samples, which make it unreliable.19 Sampling variation can explain most of these difficulties: only 65% of biopsies relying on 15 mm samples afford correct fibrosis assessment and 75% of biopsies relying on 25 mm liver samples are correct.1 In our study, we observed an increase in the performance of the FIB-4 index in parallel with the size of the LB suggesting, as previously reported, misinterpretation of small LB.20, 21 Thus, the performances of the FIB-4 index presented in Table 2 are probably underestimated. Similarly, we found that the FIB-4 index overestimated fibrosis in patients with flares of their disease, suggesting that the FIB-4 index should be interpreted with caution as patients with nil to mild fibrosis could have a FIB-4 index beyond 1.45 in this situation. The FIB-4 index should be repeated at distance of a bout of cytolysis.

Comparisons between the FIB-4 index, the APRI and FibroTest were performed in the second and the third part of this study. The FIB-4 was more precise than the APRI to exclude severe fibrosis and showed a concordance of 89% with the FibroTest for a FIB-4 index lower than 1.45. The analysis of the discordant cases showed that more than half of them were linked to a FibroTest failure either because of a low haptoglobin level or an elevation of bilirubin. This could be explained by haemolysis (sickle cell anaemia, chronic malarial infection) in our patients from Mayotte.

Finally, results were inconclusive for a FIB-4 index of more than 1.45, mostly because of nonliver related thrombopenia, and clearly in these cases, other non-invasive and/or invasive methods must be proposed to evaluate liver fibrosis with accuracy.

In conclusion, we have found that the FIB-4 index was accurate in the evaluation of nil-to-moderate fibrosis in chronic hepatitis B. A FIB-4 value ≤1.45 (and especially 0.7) enabled the correct identification of patients with nil-to-moderate fibrosis. Therefore, the FIB-4 index is an easy and inexpensive method to define those patients with nonsignificant liver fibrosis (FIB-4 value ≤1.45) and those who need (FIB-4 value >1.45) other non-invasive and/or invasive methods for assessing liver fibrosis to define therapeutic indications. The comparison with FibroTest showed a very high concordance to exclude severe fibrosis suggesting that the FIB-4 index could replace in most cases other expensive or unavailable methods, especially in developing countries. Other studies are now required for validating this new score alone or in combination with other non-invasive tests to enhance the diagnostic performance, especially for a FIB-4 index beyond 1.45.

Acknowledgements

Declaration of personal interests: Honoraria were received by V. Mallet from Schering-Plough and Bristol-Myers Squibb; and by S. Pol from Bristol-Myers Squibb, Schering-Plough, Wyeth, Roche, Novartis, Gilead, Tibotec and Boehringer Ingelheim. Declaration of funding interests: This study was funded by the French Agence Nationale de la Recherche (ANR). The funding source had no role in the design, analysis, or interpretation of the study or in the decision to submit the manuscript for publication.

Ancillary