SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Aliment Pharmacol Ther 2012; 35: 92–104

Summary

Background  Preliminary data suggest that performance of non-invasive markers for liver fibrosis in hepatitis C may improve when combined. Three algorithms based on the combination of Fibrotest, Forns’ index and AST-to-platelet ratio (APRI) have been proposed: Sequential Algorithm for Fibrosis Evaluation (SAFE biopsy); Fibropaca algorithm; Leroy algorithm.

Aim  To compare three algorithms to diagnose significant fibrosis (≥F2 by METAVIR) and cirrhosis (F4).

Methods  A total of 1013 HCV monoinfected cases undergoing liver biopsy were consecutively enrolled in seven centres. Fibrotest, APRI and Forns’ index were measured at the time of liver biopsy, considered the reference standard.

Results  Overall, performance of combination algorithms was significantly higher than the single non-invasive methods (< 0.0001). SAFE biopsy and Fibropaca algorithm saved a significantly higher number of liver biopsies than the single methods (< 0.0001). For ≥F2, Fibropaca algorithm saved more biopsies than SAFE biopsy (51.7% vs. 43.8%, = 0.0003), but with lower accuracy (87.6% vs. 90.3%, = 0.05). Regarding F4, the number of saved liver biopsies did not differ between SAFE biopsy and Fibropaca algorithm (79.1% vs. 76.2%, = 0.12). However, SAFE biopsy showed a lower accuracy when compared with Fibropaca algorithm (91.2% vs. 94%, = 0.02). As to Leroy algorithm, although it showed a good performance for ≥F2 (93.5% accuracy), it saved less liver biopsies than SAFE biopsy and Fibropaca algorithm (29.2% vs. 43.8% and 51.7% respectively, < 0.0001).

Conclusions  SAFE biopsy and the Fibropaca algorithm have excellent performance for liver fibrosis in hepatitis C, allowing a significant reduction in the need for liver biopsies. They can be useful in clinical practice and for large-scale screening.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Liver fibrosis is the hallmark of disease progression in chronic hepatitis C (CHC).1 The precise stage of hepatic fibrosis is the most important predictor of disease progression and determines the need for antiviral therapy.2 Liver biopsy is still considered the standard of reference to stage liver fibrosis in CHC.2, 3 However, the accuracy of liver biopsy depends on the quality of liver specimen, the experience of liver pathologist and the histological staging system adopted.4–7 In addition, liver biopsy is invasive, costly and currently disliked by many patients and by several physicians.8–10 Its universal use in CHC is unpractical due to the huge number of chronically infected and often asymptomatic carriers.2, 11

These limitations have stimulated the search for non-invasive tools to diagnose and stage liver fibrosis.12, 13 A variety of methods have been proposed, including simple markers based on routine laboratory tests, such as the AST-to-platelet ratio index (APRI) and Forns’ index, and more complex scores such as Fibrotest.14–16 These three serum non-invasive markers are currently among the most validated.17–20

Recently, we and others have reported that combinations of these serum non-invasive markers for liver fibrosis may represent a rational approach to further improve the diagnostic accuracy of the single markers and to markedly reduce, rather than completely abolish, the need for liver biopsy.17, 21–24 Importantly, recent guidelines from the Asian Pacific Association for the Study of the Liver about liver fibrosis management concluded that a stepwise algorithm incorporating non-invasive markers of fibrosis may reduce the number of liver biopsies by about 30%.25

The Sequential Algorithms for Fibrosis Evaluation (SAFE biopsy) combines sequentially APRI and Fibrotest; Fibropaca algorithm is a synchronous combination of Fibrotest, APRI and Forns’ Index; Leroy algorithm is a synchronous combination of Fibrotest and APRI. In the original reports, these algorithms could avoid 30–80% of liver biopsies with high diagnostic accuracy.22, 23, 26

We here describe the results of an international, multicentre study aiming a large-scale, direct comparison of these three combination algorithms for identification of liver fibrosis in CHC. Our results show that, among these three combination algorithms, SAFE biopsy and Fibropaca algorithm may save up to 79% liver biopsies and preserve an excellent accuracy. These combination algorithms may significantly reduce the screening costs related to the staging of liver fibrosis in patients with CHC.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Objectives

The aim of this study was to compare three recently described algorithms that combine serum non-invasive markers (APRI, Fibrotest, Forns’ index) to detect significant fibrosis (≥F2 according to METAVIR classification) and cirrhosis (F4) in patients with CHC. These thresholds were selected as the first is generally considered an indication for antiviral therapy and the second requires a specific management and follow-up including screening for hepatocellular carcinoma and oesophageal varices.2

Participants

This was an international, multicentre retrospective study of patients with CHC seen between January 2005 and January 2008 in seven clinical centres across Europe. Of 1215 consecutive untreated patients with CHC who had a liver biopsy and serum non-invasive markers for liver fibrosis performed on the same day, we included 1013 patients mono-infected with HCV. All patients were positive for serum HCV-RNA by polymerase chain reaction (Amplicor HCV Monitor test; Roche Diagnostics, Indianapolis, IN, USA) and had well-compensated chronic HCV infection, all cirrhotic cases being in Child–Pugh class A. Information on patients’ demographics [gender, age, body mass index (BMI)], HCV genotype, liver biopsy characteristics (length and number of portal tracts) had been recorded in each Centre on the day of biopsy. The exclusion criteria were coinfection with HBV (56) or HIV (53) and co-morbidities that could confound the results of the non-invasive markers adopted, including current alcohol intake (>20 g/day), haemolysis, Gilbert’s syndrome and autoimmune thrombocytopenia (93 cases).

Histological assessment

Liver biopsies were analysed in each centre by the local pathologist and stage of fibrosis was reported according to the METAVIR classification.27 Significant fibrosis was defined as a METAVIR score ≥F2, while cirrhosis was defined as F4. To assess if and how the characteristics of liver specimens affect concordance between combination algorithms of non-invasive serum markers and liver biopsy, a subgroup analysis was performed according to the following criteria: liver biopsies longer than 2 cm and containing more than 11 complete portal tracts were considered as the ‘gold’ standard, as recently recommended.2

Non-invasive markers of liver fibrosis

The parameters [aspartate aminotransferase (AST), γ-glutamyl-transpeptidase (γGT), total bilirubin, α2-macroglobulin, apolipoprotein A1, haptoglobin, platelet count, cholesterol] allowing for calculation of Fibrotest, APRI and Forns’ index were determined on blood sampled the day of liver biopsy. APRI was calculated by dividing the AST level (IU/L), expressed as the number of times above the upper limit of normal (ULN), by platelet count (109/L): AST (/ULN) × 100/platelet count (109/L).14

Forns’ index was calculated by applying the following regression equation: 7.811 − 3.131 ln (platelet count (109/L)) + 0.781 ln (γGT (IU/L)) + 3.467 ln (age (y)) − 0.014 (cholesterol (mg/dL)).15

Fibrotest is a non-invasive blood test that combines five serum biochemical markers (α2-macroglobulin, haptoglobin, γGT, total bilirubin, apolipoprotein A1) with patient age and gender in a patented artificial intelligence algorithm to generate a measure of fibrosis in the liver.16 Fibrotest values were obtained through Biopredictive (Fibrotest; Biopredictive, Paris, France).

Algorithms combining serum non-invasive markers for liver fibrosis

Three recently described combination algorithms of serum markers were applied to the 1013 patients and the results were compared with liver histology, considered the standard of reference. For the purpose of this study, the coordinating Centre (VIMM-Padova) received the results of APRI, Forns’ index and Fibrotest blinded to any information about liver histology. One member of the coordinating centre (GS) applied the three algorithms and sent back the results to the different centres. Only at this point, the participating centres communicated the results of liver biopsy to the coordinating Centre.

SAFE biopsy

Two distinct algorithms for the detection of significant fibrosis and of cirrhosis based on sequential use of APRI, Fibrotest and liver biopsy were applied to the 1013 patients. The two algorithms use APRI as initial screening test, followed by Fibrotest as second step and limit the use of liver biopsy to those patients in whom the non-invasive markers have inadequate accuracy (Figures 1 and 2).26

image

Figure 1.  SAFE biopsy for significant fibrosis (≥F2 by METAVIR). The figure reports the cut-off used for APRI and Fibrotest in the decisional tree and the distribution of patients in the different directions when the algorithm was applied to the 1013 HCV patients of this study. SAFE, sequential algorithm for fibrosis evaluation; APRI, AST-to-platelet ratio index; HCV, hepatitis C virus.

Download figure to PowerPoint

image

Figure 2.  SAFE biopsy for cirrhosis (F4 by METAVIR). The figure reports the cut-off used for APRI and Fibrotest in the decisional tree and the distribution of patients in the different directions when the algorithm was applied to the 1013 HCV patients of this study. SAFE, sequential algorithm for fibrosis evaluation; APRI, AST-to-platelet ratio index; HCV, hepatitis C virus.

Download figure to PowerPoint

Fibropaca algorithm

An algorithm for detection of significant fibrosis and cirrhosis based on the synchronous use of APRI, Forns’ index, Fibrotest and liver biopsy was applied to the 1013 patients. The algorithm is based on the concordance between Fibrotest and APRI and/or Forns’ index, and it limits the use of liver biopsy to discordant cases (Figures 3 and 4).22

image

Figure 3.  Fibropaca algorithm for significant fibrosis (≥F2 by METAVIR). The figure reports the decisional tree of the algorithm, based on the concordance of Fibrotest, APRI and Forns’ index, and the distribution of patients in the different directions when the algorithm was applied to the 1013 HCV patients of this study. The following cut-off values were used to define concordance between the serum non-invasive markers: Fibrotest <0.49 (no significant fibrosis), Fibrotest ≥ 0.49 (significant fibrosis); APRI ≤ 0.5 (no significant fibrosis), APRI >1.5 (significant fibrosis); Forns’ index <4.2 (no significant fibrosis), Forns’ index >6.9 (significant fibrosis). APRI, AST-to-platelet ratio index; HCV, hepatitis C virus.

Download figure to PowerPoint

image

Figure 4.  Fibropaca algorithm for cirrhosis (F4 by METAVIR). The figure reports the decisional tree of the algorithm, based on the concordance of Fibrotest and APRI and the distribution of patients in the different directions when the algorithm was applied to the 1013 HCV patients of this study. The following cut-off values were used to define concordance between the serum non-invasive markers: Fibrotest <0.75 (no cirrhosis), Fibrotest ≥ 0.75 (cirrhosis); APRI ≤ 1 (no cirrhosis), APRI >2 (cirrhosis). APRI, AST-to-platelet ratio index; HCV, hepatitis C virus.

Download figure to PowerPoint

Leroy algorithm

An algorithm for detection of significant fibrosis based on the synchronous use of APRI, Fibrotest and liver biopsy was applied to the 1013 patients. The algorithm is based on the concordance between APRI and Fibrotest, and limits the use of liver biopsy to those cases with intermediate values of serum markers, which present with inadequate accuracy (Figure 5).23

image

Figure 5.  Leroy algorithm for significant fibrosis (≥F2 by METAVIR). The figure reports the cut-off used for Fibrotest and APRI in the decisional tree and also the distribution of patients in the different directions when the algorithm was applied to the 1013 HCV patients of this study. APRI, AST-to-platelet ratio index; HCV, hepatitis C virus.

Download figure to PowerPoint

Ethics

Written informed consent was obtained from all patients at the time of liver biopsy and the study was conducted according to the Declaration of Helsinki. The Ethics Committee for the Clinical Experimentation of the Province of Padova specifically exempted the study from ethic approval. Indeed, all the data collected for this study were part of a series of examinations, which are routinely performed in patients with CHC (blood exam and liver biopsy) and no additional procedure was required.

Discordance determination

Discordance was highly attributable to biopsy failure if the biopsy was of poor quality (size <1 cm). Discordance was considered highly attributable to Fibrotest, APRI or Forns’ score failure if liver biopsy was longer than 2 cm and contained more than 11 complete portal tracts. Discordance was considered undetermined if liver biopsy length was ranging between 1 and 2 cm. We considered that patients with F4 stage on Fibrotest, APRI or liver biopsy had cirrhosis if they exhibited at least two additional criteria of cirrhosis among radiological (liver morphology abnormalities at ultrasound), endoscopic (signs of portal hypertension) or biological criteria [low blood platelets (<150 g/L) or low PT ratio (<80%)].22

Statistical analysis

Descriptive results were expressed as mean ± SD (standard deviation) or number (percentage) of patients with a condition. The t-test or nonparametric Mann–Whitney test was used to compare quantitative data and the Chi-squared test was applied for comparison of frequency data. All tests were two-tailed and P-values < 0.05 were considered significant. The performance of the algorithms combining non-invasive markers for liver fibrosis was measured as sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, positive and negative likelihood ratio (LR). Sensitivity, specificity, PPV, NPV and accuracy were expressed as percentage. For the aim of this study, predictive values were considered clinically reliable for avoiding liver biopsy when they were >85%. The use of this cut-off as a clinical decision point is consistent with several publications in the field.14, 26, 28–30 The diagnostic value of the non-invasive methods was expressed using the area under the receiver operating characteristic curve (AUROC) and its corresponding 95% confidence intervals (CI). AUROCs were calculated including non-invasive markers quantitative values using empirical nonparametric method according to DeLong et al. and compared using the method of Hanley et al.31, 32

Standardisation of AUROCs according to the prevalence of fibrosis stages

As recently proposed by Poynard et al., to prevent spectrum biases, AUROCs were adjusted according to the prevalence of fibrosis stages using the Difference between advanced and non-advanced fibrosis (DANA) index.33 This DANA is an index for standardising comparisons to transform any different prevalence profile into a homogeneous distribution of fibrosis stages from F0 to F4, as defined by a prevalence of 0.20 for each of the five METAVIR stages (standard prevalence). DANA was calculated according to the following formula: [(prevalence F2 × 2 + prevalence F3 × 3 + prevalence F4 × 4)/(prevalence F2 + prevalence F3 + prevalence F4)] − [prevalence F1/(prevalence F0 + prevalence F1)]. The adjusted AUROCs (adjAUROCs) were calculated as follows: AdjAUROC = observed AUROC (obAUROC) + (0.1056) × (2.5 − DANA).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Characteristics of the 1013 HCV patients

The main demographic, laboratory and histological features of the 1013 patients with CHC are summarised in Table 1. Overall, there were 574 (56.7%) men and 439 (43.3%) women with mean age of 48.0 ± 12.0 years. All patients were treatment-naïve. Significant fibrosis was present in 552 (54.5%) patients, while cirrhosis was present in 113 (11.2%). The mean length of liver specimen was 20.0 ± 8.2 mm and the mean number of portal tracts was 11.0 ± 6.0. Biopsy length was greater than 20 mm (‘gold’ standard) in 459 (45.3%) cases.

Table 1.   Demographic, laboratory and histological features of 1013 HCV patients
  1. HCV, hepatitis C virus; SD, standard deviation; BMI, body mass index; AST, aspartate aminotransferase; ALT, alanine aminotransferase; γGT, γ-glutamyl-transpeptidase; ULN, upper limits of normal.

Gender (%)
 Males574 (56.7%)
 Females439 (43.3%)
Age (mean years ± SD)48.0 ± 12.0
BMI (mean Kg/m2 ± SD)24.6 ± 3.7
AST (mean IU/L ± SD)70 ± 66
AST/ULN ratio (mean ± SD)1.56 ± 1.47
ALT (mean IU/L ± SD)102 ± 99
ALT/ULN ratio (mean ± SD)2.27 ± 2.2
γGT (mean IU/L ± SD)72.4 ± 107.1
γGT/ULN ratio (mean ± SD)1.31 ± 1.95
Platelet count (mean 109/L ± SD)208.6 ± 19.4
Cholesterol levels (mean mmol/L ± SD)2.5 ± 1.4
HCV Genotype (%)
 HCV-1655 (64.7)
 HCV-2112 (11.1)
 HCV-3153 (15.1)
 HCV-471 (7.0)
 HCV-518 (1.8)
 HCV-64 (0.3)
Liver Fibrosis according to METAVIR (%)
 F0105 (10.4)
 F1356 (35.1)
 F2294 (29)
 F3145 (14.3)
 F4113 (11.2)

Fibrotest, APRI and Forns’ index for diagnosis of significant fibrosis

The performance of Fibrotest, APRI and Forns’ index to diagnose significant fibrosis is shown in Table S1. Fibrotest showed a good PPV (85.7%) to rule-in significant fibrosis, with an overall AdjAUROC of 0.77. Based on the high PPV value, liver biopsy could have been avoided in 350 (34.6%) of cases. For the diagnosis of significant fibrosis, APRI showed an overall AdjAUROC of 0.76. The cut-off to rule-out significant fibrosis (0.5) showed a only discrete NPV (76.7%), while the cut-off to rule-in significant fibrosis had a very good PPV (89.9%). Moreover, 448 (44.2%) cases fall in between 0.5 and 1.5 cut-offs and could not be classified. Based on the high PPV value of the 1.5 cut-off and considering the relatively low NPV of the 0.5 cut-off and the unclassified cases, liver biopsy could have been avoided in 158 (15.6%) patients. For the diagnosis of significant fibrosis, Forns’ index showed an overall AdjAUROC of 0.70. The cut-off to rule-out significant fibrosis (4.2) showed a only discrete NPV (73.7%), and the cut-off to rule-in significant fibrosis had a poor PPV (68.9%). Moreover, 398 (39.3%) cases fall in between 4.2 and 6.9 cut-offs and could not be classified. Based on the unsatisfactory PPV and NPV values of Forns’ index in our population, we could not use any cut-off to reliably rule-in or rule-out significant fibrosis. Approximately 15% of patients were infected by HCV-3 and showed significantly lower cholesterol levels as compared with cases infected with HCV non-3 (2.1 ± 1.2 vs. 2.54 ± 1.4 mmol/L, < 0.01). Although it has been suggested that performance of Forns’ index might be lower in HCV-3 cases,34 in our series, the ObAUROC of Forns’ index was similar in HCV-3 and non-3 cases (0.63 vs. 0.64), and this is consistent with previous observations.21

Fibrotest and APRI for diagnosis of cirrhosis

The performance of Fibrotest and APRI to diagnose cirrhosis is shown in Table S2. Fibrotest showed an excellent NPV (90.8%) to rule-out cirrhosis, with an overall AdjAUROC of 0.78. Based on the high NPV value, liver biopsy could have been avoided in 670 (68.1%) cases. With regard to APRI, the cut-off to rule-out cirrhosis (1) showed an excellent NPV (96.1%) and the cut-off value to rule-in cirrhosis (2) showed a poor PPV (55.2%), with an overall AdjAUROC of 0.83. Moreover, 165 (16.3%) cases fall in between 1 and 2 cut-offs and could not be classified. Based on the high NPV value of 1 cut-off and considering the low PPV of 2 cut-off and the unclassified cases, liver biopsy could have been avoided in 580 (57.3%) patients.

Combination algorithms for diagnosis of significant fibrosis

The performance of the three algorithms to diagnose significant fibrosis is described in Table 2. Figures 1, 3 and 5 describe the three algorithms for significant fibrosis, including related decisional tree and distribution of the study population.

Table 2.   Performance of the three algorithms for diagnosing significant fibrosis (≥F2 by METAVIR) in 1013 HCV patients
 SAFE biopsyFibropaca algorithmLeroy algorithm
  1. HCV, hepatitis C virus; SAFE, sequential algorithm for fibrosis evaluation; APRI, AST-to-platelet ratio index; PPV, positive predictive value; NPV, negative predictive value; LR, likelihood ratio; ObAUROC, observed area under the receiver operating characteristic curve; DANA, difference between advanced and non-advanced fibrosis; AdjAUROC, adjusted area under the receiver operating characteristic curve; CI, confidence interval.

Fibrotest (% of tests needed)43.2100100
APRI (% of tests needed)100100100
Forns’ index (% of tests needed)01000
Accuracy (%)90.387.693.5
Sensitivity (%)10085.589.6
Specificity (%)78.289.997.8
PPV (%)83.790.589.6
NPV (%)10084.797.8
LR+4.598.4740.72
LR−00.160.1
ObAUROC (95% CI)0.90 (0.85–0.95)0.88 (0.82–0.94)0.94 (0.89–0.99)
DANA1.91.91.9
AdjAUROC (95% CI)0.96 (0.91–1)0.94 (0.88–1)1 (0.95–1)
Saved liver biopsies (%)43.851.729.2

SAFE biopsy showed a significantly higher accuracy than the single non-invasive markers (< 0.0001). Moreover, it saved a significant higher number of liver biopsies than the single non-invasive markers (< 0.0001). Fibropaca algorithm showed a significantly higher accuracy than the single non-invasive markers (< 0.0001). Moreover, it saved a significant higher number of liver biopsies than the single non-invasive markers (< 0.0001). Leroy algorithm showed a significantly higher accuracy than the single non-invasive markers (< 0.0001). However, it saved a number of liver biopsies that was significantly inferior to that of Fibrotest alone (29.2% vs. 34.8%, = 0.008) and superior only to APRI (29.2% vs. 15.6%, < 0.0001). Overall, SAFE biopsy showed an AdjAUROC of 0.96% and 90.3% accuracy. Liver biopsy could have been saved in 444 (43.8%) patients (Figure 1). Fibropaca algorithm showed an AdjAUROC of 0.94% and 87.6% accuracy. Liver biopsy could have been saved in 524 (51.7%) patients (Figure 3). Leroy algorithm showed an AdjAUROC of 1 and 93.5% accuracy. Liver biopsy could have been saved in 296 (29.2%) patients (Figure 5). The number of saved liver biopsies was significantly higher using Fibropaca algorithm than SAFE biopsy (= 0.0003). Conversely, the accuracy of SAFE biopsy was significantly higher than that of Fibropaca algorithm (= 0.05). Leroy algorithm showed a significantly higher accuracy when compared to SAFE biopsy and Fibropaca algorithm (= 0.009 and < 0.0001, respectively). However, it saved significantly less liver biopsies as compared with SAFE biopsy and Fibropaca algorithm (< 0.0001).

Combination algorithms for diagnosis of cirrhosis

The performance of the two algorithms able to diagnose liver cirrhosis is described in Table 3. Figures 2 and 4 describe the SAFE biopsy and Fibropaca algorithm for cirrhosis, respectively, including related decisional tree and distribution of the study population. SAFE biopsy showed a significantly higher accuracy than the single non-invasive markers (< 0.0001). Moreover, it saved a significant higher number of liver biopsies than the single non-invasive markers (< 0.0001). Fibropaca algorithm showed a significantly higher accuracy than the single non-invasive markers (< 0.0001). Moreover, it saved a significant higher number of liver biopsies than the single non-invasive markers (< 0.0001). Overall, SAFE biopsy showed an AdjAUROC of 0.93% and 91.2% accuracy. Liver biopsy could have been saved in 801 (79.1%) patients (Figure 2). Fibropaca algorithm showed an AdjAUROC of 0.91% and 94.0% accuracy. Liver biopsy could have been saved in 772 (76.2%) patients (Figure 4). The number of saved liver biopsies did not differ between SAFE biopsy and Fibropaca algorithm (= 0.12). Accuracy of Fibropaca algorithm was significantly higher than that of SAFE biopsy (= 0.02).

Table 3.   Performance of SAFE biopsy and Fibropaca algorithm for diagnosing cirrhosis (F4 by METAVIR) in 1013 HCV patients
 SAFE biopsyFibropaca algorithm
  1. HCV, hepatitis C virus; SAFE, sequential algorithm for fibrosis evaluation; APRI, AST-to-platelet ratio index; PPV, Positive Predictive Value; NPV, negative predictive value; LR, likelihood ratio; ObAUROC, observed area under the receiver operating characteristic curve; DANA, difference between advanced and non-advanced fibrosis; AdjAUROC, adjusted area under the receiver operating characteristic curve; CI, confidence interval.

Fibrotest (% of tests needed)57.6100
APRI (% of tests needed)100100
Forns’ index (% of tests needed)NANA
Accuracy (%)91.294.0
Sensitivity (%)81.872.7
Specificity (%)92.496.7
PPV (%)57.473.4
NPV (%)97.695.6
LR+10.7622.0
LR−0.20.28
ObAUROC (95% CI)0.87 (0.81–0.93)0.85 (0.79–0.91)
DANA1.91.9
AdjAUROC (95% CI)0.93 (0.87–0.99)0.91 (0.85–0.97)
Saved liver biopsies (%)79.176.2

Influence of liver biopsy length

To assess if and how the characteristics of liver specimens affect the concordance between Fibrotest and liver biopsy, the performance of those cases with ‘gold’ standard liver biopsy was investigated separately and compared with the whole population of patients. Performance of the three algorithms was not significantly influenced by the length of liver biopsy (data not shown).

Combination algorithms in HCV elderly patients

To assess the performance of the combination algorithms in patients who may have a peculiar distribution of liver fibrosis stages, a dedicated analysis was performed on HCV elderly patients (age ≥ 65 years), who are expected to have a higher prevalence of advanced fibrosis stages. These patients represent a significant proportion of HCV carriers in the general population in most Western Countries. Moreover, the indication to perform a liver biopsy in these cases is often debated and controversial.35, 36

Overall, 91 (9.0%) cases were over 65 years of age, including 33 (36.3%) men and 58 (63.7%) women, with mean age of 69.6 ± 4.0 years. Significant fibrosis was present in 63 (69.2%) patients while cirrhosis was present in 14 (15.4%). Prevalence of significant fibrosis was significantly higher in the elderly patients than in the whole study population (significant fibrosis being present in 69.2% vs. 54.5%, = 0.006). No significant difference was observed in the prevalence of cirrhosis between HCV elderly patients and the whole study population (15.4% vs. 11.2%, = 0.22). The performance of the combination algorithms for significant fibrosis in HCV elderly patients was as follows: SAFE biopsy (AdjAUROC = 0.97, accuracy = 94.5%, PPV = 92.6%, NPV = 100%); Fibropaca algorithm (AdjAUROC = 0.95, accuracy = 91.5%, PPV = 93.5%, NPV = 82.2%); Leroy algorithm (AdjAUROC = 1, accuracy = 96.7%, PPV = 98.4%, NPV = 93.1%). For the diagnosis of significant fibrosis, liver biopsy could have been saved in 37 (59.3%) patients with SAFE biopsy, 51 (56%) patients with Fibropaca algorithm and 35 (38.5%) patients with Leroy algorithm. The performance of the combination algorithms for cirrhosis in HCV elderly patients was as follows: SAFE biopsy (AdjAUROC = 0.95, accuracy = 93.0%, PPV = 65.1%, NPV = 95.9%); Fibropaca algorithm (AdjAUROC = 0.98, accuracy = 91.2%, PPV = 65.0%, NPV = 93.2%). For the diagnosis of cirrhosis, liver biopsy could have been saved in 73 (80.2%) patients with SAFE biopsy and 57 (62.6%) with Fibropaca algorithm.

Even though there was a tendency for a higher performance of all the combination algorithms in the HCV elderly patients, especially for the diagnosis of significant fibrosis, the difference with respect to the whole study population was not statistically significant.

Other variables influencing the performance of the three algorithms

The performance of the three combination algorithms in relation to several demographic and serological variables was investigated. Gender, BMI, levels of transaminases, HCV genotype did not show any significant influence on the performance of the three algorithms for both significant fibrosis and cirrhosis, even when corrected for DANA. Moreover, the inter-centre variability was marginal (data not shown).

Analysis of discordant cases

Analysis of discordant cases between the combination algorithms and liver biopsy is shown in Table S3. For the diagnosis of significant fibrosis, discordance due to biopsy was observed in 32.7%, 21.4% and 34.8% with SAFE biopsy, Fibropaca algorithm and Leroy algorithm, respectively; discordance due to non-invasive markers was observed in 44.9%, 45.2% and 45.5% with SAFE biopsy, Fibropaca algorithm and Leroy algorithm, respectively. The rest of discordant cases were undetermined. For the diagnosis of cirrhosis, discordance due to biopsy was observed in 47.2% and 42.6% with SAFE biopsy and Fibropaca algorithm respectively; discordance due to non-invasive markers was observed in 33.7% and 36.0% with SAFE biopsy and Fibropaca algorithm respectively. The rest of discordant cases were undetermined.

Cost–benefit analysis

For the purpose of the cost–benefit analysis, we estimated the cost of an uncomplicated liver biopsy to be 700 Euros and the cost of a Fibrotest to be 100 Euros, while no extra costs were considered for APRI and Forns’ index, as patients with CHC usually undergo routinely the needed blood tests during their follow-up. On the same line, the price of total bilirubin and γGT, that are routinely performed, was not considered for Fibrotest. It should also be mentioned that Actitest, a non-invasive test that gives an estimation of the necroinflammatory activity in the liver, is calculated as free with Fibrotest.37 In this analysis, performing liver biopsy in all 1013 patients with CHC would cost 709 100 Euros for diagnosis of significant fibrosis (‘universal biopsy strategy’) (Table 4). The cost of ‘Fibrotest strategy’, ‘APRI strategy’, ‘SAFE biopsy strategy’, ‘Fibropaca algorithm strategy’ and ‘Leroy strategy’ would be 565 400 Euros, 598 00 Euros, 442 100 Euros, 443 600 Euros and 603 200 Euros, respectively. Thus, there would be a relevant reduction in the screening costs, especially with ‘SAFE biopsy strategy’ and ‘Fibropaca algorithm strategy’. Regarding the diagnosis of cirrhosis, performing liver biopsy in all patients with CHC would cost 709 100 Euros (‘universal biopsy strategy’) (Table 5). The cost of ‘Fibrotest strategy’, ‘APRI strategy’, ‘SAFE biopsy strategy’ and ‘Fibropaca algorithm strategy’ would be 341 400 Euros, 303 100 Euros, 206 700 Euros and 270 000 Euros respectively. Thus, there would be a relevant reduction in the screening costs, especially with ‘SAFE biopsy strategy’ and ‘Fibropaca algorithm strategy’.

Table 4.   Cost–benefit analysis of the ‘universal biopsy strategy’ vs. ‘single non-invasive markers’ and ‘combination algorithms strategy’ for the diagnosis of significant fibrosis (≥F2 by METAVIR) as for 1013 HCV patients
 Universal biopsy strategyFibrotest strategyAPRI strategySAFE biopsy strategyFibropaca algorithm strategyLeroy algorithm strategy
  1. HCV, hepatitis C virus; APRI, AST-to-platelet ratio index; SAFE, sequential algorithm for fibrosis evaluation.

Cost of liver biopsy (Euros)709 100464 100598 500398 300342 300501 900
Cost of Fibrotest (Euros)0101 300043 800101 300101 300
Total cost (Euros)709 100565 400598 500442 100443 600603 200
Saved cost vs. biopsy strategy (Euros)143 700110 600267 000265 500105 900
Misclassified cases (n)1451029812666
Table 5.   Cost–benefit analysis of the ‘universal biopsy strategy’ vs. ‘single non-invasive markers’ and ‘combination algorithms strategy’ for the diagnosis of cirrhosis (F4 by METAVIR) in 1013 HCV patients
 Universal biopsy strategyFibrotest strategyAPRI strategySAFE biopsy strategyFibropaca algorithm strategy
  1. HCV, hepatitis C virus; APRI, AST-to-platelet ratio index; SAFE, sequential algorithm for fibrosis evaluation.

Cost of liver biopsy (Euros)709 100240 100303 100148 400168 700
Cost of Fibrotest (Euros)0101 300058 300101 300
Total cost (Euros)709 100341 400303 100206 700270 000
Saved cost vs. biopsy strategy (Euros)367 700406 000502 400439 100
Misclassified cases (n)93408961

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

This study, based on a large consecutive population of patients with CHC, suggests that, among a number of combination algorithms using serum non-invasive markers for liver fibrosis, SAFE biopsy and Fibropaca algorithm may save up to 79% liver biopsies by preserving an excellent accuracy, thus leading to a significant reduction in the screening costs. Moreover, SAFE biopsy and Fibropaca algorithm showed a significantly higher performance and saved a significantly higher number of liver biopsies as compared with the single non-invasive methods adopted. This represents the first large-scale study, which compares three algorithms combining the most validated serum non-invasive markers for liver fibrosis in CHC, that are Fibrotest, APRI and Forns’ index.

Staging of liver fibrosis has always been considered of paramount importance for the definition of prognosis and urgency for antiviral treatment in patients with CHC.2 However, considering the huge number of HCV carriers worldwide, it is inconceivable to take a liver biopsy in all of them due to its cost and invasiveness. Indeed, a significant number of patients may refuse liver biopsy and the hepatologists themselves may have concerns about how to use it in clinical practice.9, 10 Although numerous serum non-invasive markers for liver fibrosis have been developed in the last decade, their implementation in clinical practice remains still limited by the scepticism shared by many clinicians on their diagnostic accuracy in substitution of liver histology.12 Indeed, in patients with CHC, the diagnostic accuracy of these markers does not overcome 75–85% when used individually, these values being in the lower range particularly for identification of significant fibrosis.17 More recently, it has been proposed that combinations of serum non-invasive markers may allow for reaching higher diagnostic accuracy.13, 17, 24, 38 The rationale for using such combinations of serum non-invasive markers is to reduce, rather than abolish, the need for liver biopsies. In this view, combinations of complementary serum non-invasive markers has been proposed as the initial step for disease staging, limiting liver biopsy to those cases in which the diagnostic accuracy of non-invasive markers appears unsatisfactory.

Recently, the Asian Pacific Association for the Study of the Liver has produced guidelines about liver fibrosis management. The guidelines concluded that a stepwise algorithm incorporating non-invasive markers of fibrosis may reduce the number of liver biopsies by about 30%.25 In the present study, we compared three algorithms that combine the most validated serum non-invasive markers for liver fibrosis in a large population of patients with CHC, using liver biopsy as the standard of reference. It should be noted that liver biopsy remains an imperfect gold standard due to intra- and inter-observer variability, which mainly depends on the quality of liver specimen.4–7 Interestingly, the group that patented Fibrotest has suggested that Fibrotest and biopsy may have a similar prognostic value and a similar risk of false positivity/negativity.39 Moreover, few reports indicate that serum fibrosis markers may be associated with portal hypertension, liver disease progression and the outcome of liver disease.40–42

Nevertheless, as liver biopsy remains the only direct way to assess liver histology, guidelines still recommend it for the staging of hepatic fibrosis.2 Indeed, the American Association for the Study of Liver Diseases has recently produced a position paper about liver biopsy, recommending a sample of at least 20 mm in length and containing at least 11 complete portal tracts. Moreover, in clinical practice, it was recommended the use of a simple (METAVIR) rather than complex (Ishak) scoring system for liver fibrosis.3 In our large-scale study, liver histological assessment was performed according to METAVIR system as recommended. On the other hand, liver biopsy samples were somehow suboptimal, around half of the specimens being longer than 20 mm. However, a subgroup analysis aimed to investigate the performance of the three algorithms only in those patients with a ‘gold’ standard liver biopsy did not show any significant difference with respect to the overall population studied. This is consistent with other reports suggesting that performance of Fibrotest does not depend on the size of liver specimen.43, 44 The lack of evaluation by a single Pathologist of all biopsies could be seen as a weakness of our study, but it better reflects what occurs in real life. Furthermore, the diagnostic performance of the three algorithms remained unchanged compared with the whole cohort when only patients having liver biopsy evaluated by a single Pathologist were considered (data not shown).

The overall accuracy for detection of significant fibrosis was very good for SAFE biopsy, Fibropaca algorithm and Leroy algorithm. However, Leroy algorithm saved much less liver biopsies than the other two algorithms and than Fibrotest alone (only 29.2%). SAFE biopsy had a slightly better performance with respect to Fibropaca algorithm. However, Fibropaca algorithm saved more liver biopsies. For diagnosis of cirrhosis, Fibropaca algorithm showed a significantly better performance than SAFE biopsy. The results of our study are consistent with the original studies in which the algorithms were proposed.21–23 SAFE biopsy has also been validated in a large-scale international study including more than 2000 patients with CHC. This study represents the largest independent study on serum non-invasive markers for liver fibrosis.26

Thanks to the large number of patients included in the study, we were also able to perform a subgroup analysis aimed to assess the performance of combination algorithms in HCV elderly patients, who may have a peculiar distribution of liver fibrosis stages. In these patients, liver biopsy is even more questionable.35, 36 Indeed, in elderly HCV patients, the greater prevalence of liver disease is linked to higher morbidity and mortality, which are excellent reasons for improving the evaluation of liver fibrosis and for a better identification of indication to antiviral therapies.35, 36 It has been suggested that the prevalence of advanced fibrosis and cirrhosis in the study population may significantly affect the performance of non-invasive markers in terms of PPV and NPV.45 In the present study, even though there was a tendency for higher accuracies and PPVs, and slightly inferior NPVs in HCV elderly patients as compared with the whole study population, these differences were not statistically significant. This is likely because in this population of patients, the prevalence of elderly patients was relatively limited (91 cases with age ≥ 65 years) and the prevalence of cirrhosis in this subgroup was not significantly higher than in the whole population study (15.4% vs. 11.2%). We can interpret this fact by considering that the study was retrospective and most of the patients underwent both liver biopsy and non-invasive markers as they were candidate for antiviral therapy.

Other studies have reported that combination algorithms of serum non-invasive markers may increase the performance of the single tests. An algorithm combining synchronously Fibrotest and Fibroscan® has also been proposed and we have recently compared it with SAFE biopsy in a prospective study of 302 patients with CHC, concluding that both algorithms are effective for assessment of liver fibrosis in CHC.46 An algorithm combining Hepascore, a patented test, and APRI was recently proposed.30, 47 The authors reported a high diagnostic accuracy (91%) with 45% saved liver biopsies to diagnose significant fibrosis. However, the main limitation of this algorithm is that Hepascore is not as validated as APRI, Fibrotest and Forns’ index. On the same line, Cales and colleagues reported that a combination of Fibrometer, a patented test, and Fibrotest may save 44.8% liver biopsies with an overall accuracy of 95.3%.48 On the basis of their study, Cales and colleagues have suggested that synchronous combination algorithms may be more effective than sequential algorithms such as SAFE biopsy. However, although Fibrometer had a good performance in studies coming from the patenting group, it has been validated less than Fibrotest and APRI.49, 50 Moreover, Fibrometer is not licensed in as many countries as Fibrotest.

Our study has several limitations. First, as already mentioned, this was a retrospective study. Second, most of the algorithms proposed here did not reach 100% predictive values, except for SAFE biopsy for significant fibrosis, therefore the cost of false positive (i.e. the cost of having unnecessary surveillance/investigations) and the cost of false negatives (i.e. the cost of missing fibrosis/cirrhosis) should be considered. The misclassified cases were mostly false positives. We might speculate that, for significant fibrosis, false positives would undergo an antiviral therapy that they would most likely need anyway in a short time and, for cirrhosis, false positives would start a surveillance programme that is not completely inappropriate as most of them had severe liver fibrosis (F3 and F3–F4 by METAVIR). An analysis of discordant cases between the combination algorithms and liver biopsy for significant fibrosis and cirrhosis was also carried out. Overall, when considering liver biopsies of poor quality, between one-third and 45% of the discordances could be highly attributable to liver biopsy failure.

Third, no information about fragmentation of liver biopsy specimens was available. Indeed, it has been demonstrated that the number of fragments may have a significant impact on the AUROC of non-invasive markers.51

Fourth, potential limitations of the use of predictive values for decision making due to their reliance on prevalence should be acknowledged. Nevertheless, most pertinent studies published to date have used predictive values.13–16, 23, 52 The relevance of the prevalence of fibrosis stages on the performance of non-invasive markers is here underscored by the correction of AUROCs by DANA.

In conclusion, based on the results of this multicentre, large-scale study, SAFE biopsy and Fibropaca algorithms are attractive methods in clinical practice for large-scale screening of liver fibrosis in hepatitis C. These algorithms may be particularly useful to screen HCV-infected individuals, where an immediate approach with liver biopsy is particularly problematic or questionable, such as elderly HCV carriers. The algorithm for significant fibrosis may be particularly indicated to screen HCV patients for indication to initiate antiviral therapy, while the algorithm for cirrhosis may be ideal for the follow-up of patients already known to have progressed to significant fibrosis based on previous histological evaluation.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Declaration of personal interests: GS has served as a speaker and consultant for Istituto Biochimico Italiano (IBI) and MSD, and has received research funding from Roche. PH has served as a speaker for MSD and has received research funding from Roche. PH owns patent for PCT protease inhibitors. LC has served as a speaker for Echosens and Ferrer. AM has served as a speaker, a consultant and an advisory board member for Jannssen, MSD, ROCHE and has received research funding from MSD, ROCHE. VD has served as a speaker, a consultant and an advisory board member for Roche, Gilead, BMS, MSD and Dohme). MP has served as a speaker for Bayer, Bristol-Myers-Squibb, Gilead Sciences, Merck, Roche. MV has nothing to declare. MB has served as a speaker and as a consultant for Roche, Schering-Plough, MSD, BMS, Gilead, GSK, Vertex. AA has served as a speaker, a consultant and an advisory board member for Roche, Gilead, Novartis, BMS, J&J, MSD, Schering-Plough and has received research funding from Gilead, MSD, BMS. Contributors: GS and AA contributed to the conception, study design, data, interpretation of the data and first draft of the article. PH, LC, AM, VDM, MP, MV, MB contributed with data. All authors approved the final version of the article. AA is the guarantor. Declaration of funding interests: The study was not funded.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Table S1. Performance of Fibrotest, APRI and Forns’ index in 1013 HCV patients to diagnose significant fibrosis (vs. liver histology).

Table S2. Performance of Fibrotest and APRI in 1013 HCV patients to diagnose cirrhosis (vs. liver histology).

Table S3. Discordant cases between combination algorithms and liver biopsy in 1013 HCV patients.

FilenameFormatSizeDescription
APT_4897_sm_tS1-3.doc78KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.