Performance of the aspartate aminotransferase-to-platelet ratio index for the staging of hepatitis C-related fibrosis: An updated meta-analysis§

Authors

  • Zhong-Hua Lin,

    1. Medical College of Qingdao University, Qingdao, Shandong Province, China
    2. Department of Gastroenterology, Qingdao Municipal Hospital, Qingdao, Shandong Province, China
    Search for more papers by this author
    • *These authors contributed equally to this work.

  • Yong-Ning Xin,

    1. Department of Gastroenterology, Qingdao Municipal Hospital, Qingdao, Shandong Province, China
    2. College of Medicine and Pharmaceutics, Ocean University of China, Qingdao, Shandong Province, China
    Search for more papers by this author
    • *These authors contributed equally to this work.

  • Quan-Jiang Dong,

    1. Department of Gastroenterology, Qingdao Municipal Hospital, Qingdao, Shandong Province, China
    Search for more papers by this author
  • Qing Wang,

    1. Department of Gastroenterology, Qingdao Municipal Hospital, Qingdao, Shandong Province, China
    Search for more papers by this author
  • Xiang-Jun Jiang,

    1. Department of Gastroenterology, Qingdao Municipal Hospital, Qingdao, Shandong Province, China
    Search for more papers by this author
  • Shu-Hui Zhan,

    1. Department of Gastroenterology, Qingdao Municipal Hospital, Qingdao, Shandong Province, China
    Search for more papers by this author
  • Ying Sun,

    1. Department of Gastroenterology, Qingdao Municipal Hospital, Qingdao, Shandong Province, China
    Search for more papers by this author
  • Shi-Ying Xuan

    Corresponding author
    1. Department of Gastroenterology, Qingdao Municipal Hospital, Qingdao, Shandong Province, China
    2. College of Medicine and Pharmaceutics, Ocean University of China, Qingdao, Shandong Province, China
    • Department of Gastroenterology, Qingdao Municipal Hospital, Qingdao 266021, Shandong Province, China
    Search for more papers by this author
    • fax: +86-532-82836421


  • Potential conflict of interest: Nothing to report.

  • Partially supported by grants from the Natural Science Foundation of Shandong Province, China (No. ZR2009CQ031).

Abstract

The aspartate aminotransferase-to-platelet ratio index (APRI), a tool with limited expense and widespread availability, is a promising noninvasive alternative to liver biopsy for detecting hepatic fibrosis. The objective of this study was to update the 2007 meta-analysis to systematically assess the accuracy of APRI in predicting significant fibrosis, severe fibrosis, and cirrhosis stage in hepatitis C virus (HCV) monoinfected and HCV / human immunodeficiency virus (HIV) coinfected individuals. Studies comparing APRI versus biopsy in HCV patients were identified via a thorough literature search. Areas under summary receiver operating characteristic curves (AUROC), sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were used to examine the APRI accuracy for the diagnosis of significant fibrosis, severe fibrosis, and cirrhosis. Heterogeneity was explored using meta-regression. Twenty-one additional studies were eligible for the update and, in total, 40 studies were included in this review (n = 8,739). The summary AUROC of the APRI for the diagnosis of significant fibrosis, severe fibrosis, and cirrhosis were 0.77, 0.80, and 0.83, respectively. For significant fibrosis, an APRI threshold of 0.7 was 77% sensitive and 72% specific. For severe fibrosis, a threshold of 1.0 was 61% sensitive and 64% specific. For cirrhosis, a threshold of 1.0 was 76% sensitive and 72% specific. Moreover, we found that the APRI was less accurate for the identification of significant fibrosis, severe fibrosis, and cirrhosis in HIV/HCV coinfected patients. Conclusion: Our large meta-analysis suggests that APRI can identify hepatitis C-related fibrosis with a moderate degree of accuracy. Application of this index may decrease the need for staging liver biopsy specimens among chronic hepatitis C patients. (HEPATOLOGY 2011)

Hepatitis C virus (HCV) infection, with an estimated prevalence of more than 170 million worldwide, is a major public healthcare problem.1 Chronic hepatitis C (CHC) is the most common cause of cirrhosis and hepatocellular carcinoma (HCC), and the leading indication for liver transplantation in the United States and many Western countries. Cirrhosis and its disease-related complications are responsible for more than 40,000 deaths annually in the United States.2 HCV chronic infection develops into chronic hepatitis in more than 70% of patients and in about 20% of them progresses to cirrhosis and eventually HCC.3 In HCV monoinfected patients with compensated cirrhosis, the cumulative incidences of HCC, ascites, bleeding, and encephalopathy at 5 and 10 years were 7.8%/7%/2.5%/0% and 28%/20%/5%/2.5%, respectively.4 Early diagnosis of cirrhosis is important in patients with CHC not only because it prompts screening for HCC and esophageal varices, but also because it is the important factor for initiation of treatment in patients with hepatitis C infection.5

At present, liver biopsy is still the most commonly used reference standard for the assessment of liver fibrosis. However, it is an invasive method that is associated with patient discomfort and in rare cases with serious complications.6 In addition, the accuracy of liver biopsy is limited as a result of intra- and interobserver variability and sampling errors.7 Furthermore, the dynamic process of liver fibrosis resulting from progression and regression cannot be easily quantified by liver biopsy. Therefore, much research has focused on the evaluation of noninvasive methods for the assessment of liver fibrosis. Ideally, such a test should be simple, readily available, inexpensive, and reliable and accurate in predicting liver fibrosis. To date, several laboratory tests and scores have been proposed for the noninvasive prediction of cirrhosis in patients with CHC, including direct biochemical markers of hepatic fibrosis (collagen, hyaluronic acid, laminin, and YKL-40), indirect biochemical markers of hepatic fibrosis (PGA index, Forns' index, Fibrotest, and Hepascore), radiological imaging, and transient elastography. Although several noninvasive direct and indirect serum markers (such as Fibrotest and Hepascore) have exhibited good diagnostic accuracy in some studies, most of these markers require complicated calculations, the use of a specialized set of biochemical markers, and are costly. These limit their application in clinical practice.

Recently, a novel index by combining aspartate aminotransferase (AST)-to-platelet ratio index (APRI) was reported to identify patients with hepatic fibrosis.8 APRI is an indirect biochemical marker of hepatic fibrosis, based on routine laboratory parameters, reflecting alterations in hepatic function. Since the initial report in 2003 by Wai et al.,8 an increasing number of studies have evaluated APRI for the diagnosis of liver fibrosis in a multitude of liver diseases with inconsistent results. In 2007, Shaheen and Myers9 published a meta-analysis that included 4,266 patients to assess the accuracy of APRI in predicting HCV-related significant fibrosis and cirrhosis stage. However, it did not assess the value of APRI in predicting severe fibrosis stage. Furthermore, the results that APRI was more accurate for the identification of cirrhosis in HIV/HCV coinfected patients lacked sufficient data to support it.

In view of the uncertain clinical value of APRI in HCV/HIV coinfection and the limitations of the previous meta-analysis, we conduct an updated systematic review and meta-analysis to comprehensively assess the overall performance of APRI for the diagnosis of hepatitis C-related fibrosis and to analyze the heterogeneity between the available studies to date before its wide application in clinical practice.

Abbreviations

APRI, aspartate aminotransferase-to-platelet ratio index; AUROC, area under the receiver operating characteristic curve; DOR, diagnostic odds ratio; HCV, hepatitis C virus; NPV, negative predictive value; PPV, positive predictive value; QUADAS, the Quality Assessment of Studies of Diagnostic Accuracy Included in Systematic Reviews; SROC, summary receiver operating characteristic curves.

Materials and Methods

Search Strategy.

The objective of our search was to identify published articles of studies examining the APRI for the prediction of HCV-related fibrosis. An electronic search was completed on PubMed, EMBASE, and the Cochrane Library (01/2003-04/2010) including the following search terms: APRI, AST-to-platelet ratio index, AST, platelet, hepatitis C, and noninvasive fibrosis markers, and serum markers of liver fibrosis.10 The language was limited to English only. Additional studies were identified via a manual review of the reference lists of identified studies and review articles.

Selection Criteria.

Studies were included if they met the following inclusion criteria: (1) The study evaluated the performance of the APRI for the prediction of fibrosis in HCV-infected patients. Studies including patients with other causes of liver disease were included if data for HCV-infected patients could be extracted. (2) Liver biopsy was used as the reference standard for assessing fibrosis. (3) Data could be extracted to allow the construction of at least one 2×2 table of test performance, based on some cutoff point of the APRI for a fibrosis stage; If data were not available in the publication, corresponding authors were contacted to provide supplemental data. (4) They assessed the diagnostic accuracy for fibrosis stage F≥2, F≥3, or F≥4 according to METAVIR or a comparable staging system. (5) The study included at least 30 patients. Smaller studies were excluded because of poor reliability.

Data Extraction and Quality Assessment.

Two reviewers (Lin and Xin) independently evaluated study eligibility, graded quality, and extracted outcome data. Disagreements were resolved by consensus. These parameters included study publication year, region, method, patient sex, age, number, author, underlying chronic liver disease etiology, histological scoring system, average liver biopsy length, duration of time between biopsy and performance of APRI, prevalence of the fibrosis stage, as well as cutoff values to identify the fibrosis stage. To assess the quality of the studies included in the meta-analysis, the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) questionnaire was used.11 This validated tool was designed to assess the internal and external validity of diagnostic accuracy studies included in systematic reviews.

The primary outcome was the identification of significant fibrosis, defined as METAVIR,12 Batts and Ludwig,13 or Scheuer14 stages F2 through F4 or Ishak et al.15 stages F3 through F6. This outcome was chosen because it is often considered a threshold for the initiation of antiviral therapy.16 We also examined the identification of severe fibrosis (METAVIR, Batts and Ludwig, or Scheuer F3-4, or Ishak F4-6) and cirrhosis (METAVIR, Batts and Ludwig, or Scheuer F4, or Ishak F5-6).

Statistical Analysis and Data Synthesis.

Where data were available, 2×2 tables were constructed to derive sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) at each threshold value. To provide clinically meaningful results, three measures of diagnostic test accuracy were used to examine the APRI accuracy for the diagnosis of significant fibrosis, severe fibrosis, and cirrhosis: the area under the summary receiver operating characteristic curve (AUROC), summary sensitivities and specificities, and summary PPV and NPV based on the prevalence of fibrosis.

The Meta-Disc software (v. 1.4) and Stata 8.0 (College Station, TX) were used to analyze the reports and tests for sensitivity, specificity, and area under the summary receiver operating characteristic curves (SROC), as well as meta-regression approaches. Studies with a larger sample size and therefore a smaller standard error received more weight when calculating the mean AUROC.17 The diagnostic odds ratio (DOR) describes the odds of a positive test in disease cases compared with noncases. As these analyses require a single measure of accuracy for each study and many reported multiple APRI thresholds, we calculated the average DOR among all thresholds per study.18 Because of a priori assumptions about the likelihood for heterogeneity between primary studies, the random-effects model was used for pooled analyses. Wherever zero counts occurred for 2×2 tables, the value of 0.5 was added to all cells containing the value 0 to facilitate analysis. Heterogeneity of accuracy estimates across studies was evaluated using the I2 statistic, which describes the percentage of the variability in estimates that is due to heterogeneity rather than sampling error (chance). A value >50% may be considered substantial heterogeneity.19

A meta-regression technique was used to explore the factors that may induce the heterogeneity, according to the following predefined characteristics: (a) study design (retrospective versus prospective); (b) etiology (HCV versus HCV/HIV); (c) blinded interpretation of APRI and reference standard (yes versus no); (d) liver biopsy length (≥15 mm or not); (e) liver biopsy scoring system (METAVIR, Ishak, Batts Ludwig, and Scheuer); (f) QUADAS score; (g) sample size; (h) median age; (i) percentage of males; (j) location of study (North America, Europe, and other); (k) prevalence of significant fibrosis / severe fibrosis / cirrhosis. To assess possible publication bias, we examined for asymmetry of funnel plots of the accuracy for detecting fibrosis (using the natural logarithm of the DOR) versus the inverse of the square root of the effective sample size.20

Results

Search Results.

A total of 113 studies were retrieved based on the described search strategies. In all, 57 eligible studies were identified for evaluation. Ultimately, 17 studies were excluded for insufficient data (n = 11), mixed etiology (n = 4), or failure to use biopsy as the reference test (n = 2) (Fig. 1). Thus, our final dataset for the meta-analysis included 40 studies.8, 21-59

Figure 1.

Flow diagram of study identification.

The main features of the studies included in the meta-analyses are shown in Table 1. A total of 8,739 patients (median age, 46 years; 66% male) were included. The overall prevalence of significant fibrosis, severe fibrosis, and cirrhosis were 46% (range, 9%-79%), 28% (range, 9%-59%), and 19% (range, 4%-33%), respectively. The fibrosis staging system used to classify the histology varied. Eighteen studies used a METAVIR score, 12 studies used an Ishak score, six studies used a Batts and Ludwig score, and four studies used a Scheuer score. Thirty studies included HCV monoinfected patients (n = 6,891), and 10 included HIV/HCV coinfected patients (n = 1,848). According to the QUADAS scale, the methodological quality of the included studies was excellent (Table 2).

Table 1. Characteristics of the 40 Studies Included in the Meta-Analysis
 Author, Year, RegionStudy/Center DescriptionNInterval Between Biopsy & APRIMedian/Mean Age, yr (% male)EtiologyLiver Biopsy Scoring SystemBlindLiver Biopsy Median LengthPrevalence of Significant Fibrosis, Severe Fibrosis, CirrhosisQUADAS Score
1Cals, 2010, FranceProspective, multicenter169≤3 months41 (65%)HCV+HIVMETAVIRYes25±12 mm66%, 33%, 20%13
2Macas, 2010, SpainRetrospective, multicenter519Unclear43 (79%)HCV+HIVMETAVIRNo15 (12-20) mm51%, 26%, 12%11
3Boursier, 2009,FranceRetrospective, multicenter1056≤3 months46 (60%)HCVMETAVIRYes21±8 mm52%, 25%, 11%13
4Castera, 2009, FranceRetrospective, one center298Same time52 (57%)HCVMETAVIRYes19.5±7.8 mm75%, 42%, 23%14
5Corradi, 2009, ItalyRetrospective, one center36Same time58 (83%)HCVMETAVIRYes29 (16-55) mm36%, 14%, 3%13
6Schiavon, 2009,BrazilRetrospective, one center102≤6 months44 (60%)HCVMETAVIRYes13.9±3.9 mm20%, NA, NA14
7Tural, 2009, Spaincohort, one center324Unclear38 (72%)HCV+HIVScheuerYes1.8±0.9 mm48%, 29%, 6%12
8Carvalho-Filho, 2008, BrazilRetrospective, one center111≤6 months40 (73%)HCV+HIVMETAVIRYes14.5±4.0 mm41%, 25%, 18%14
9Cheung, 2008, USAProspective, multicenter490Same time49 (98%)HCVBatts LudwigNoUnclear66%, 38%, 14%11
10Dinesen, 2008, GermanyRetrospective, one center96Unclear48 (57%)HCVBatts LudwigYesUnclear91%, 59%, 28%11
11Khan, 2008, PakistanRetrospective, one center120Unclear37 (69%)HCVMETAVIRYesUnclear54%, 26%, 8%12
12Loko, 2008, FranceRetrospective, one center200Unclear40 (67%)HCV+HIVMETAVIRYes15.7±7.5 mm79%, 36%, 20%13
13Paggi, 2008, ItalyProspective, multicenter430Unclear53 (55%)HCVMETAVIRYesUnclear70%, 37%, 20%13
14Schiavon, 2008, BrazilRetrospective, one center185≤6 months45 (64%)HCVMETAVIRYes13.7±4.9 mm24%, NA, NA14
15Silva, 2008, BrazilRetrospective, one center50≤4 months50 (68%)HCVMETAVIRYesUnclear56%, 36%, 26%12
16Trang, 2008, USARetrospective, one center81≤6 months47 (84%)HCV+HIVBatts LudwigNo22.5 mm61%, 35%, 23%12
17Halfon, 2007, FranceRetrospective, multicenter356≤1 week45 (53%)HCVMETAVIRYes22.0±7.1 mm41%, 15%, 4%13
18Leroy, 2007, FranceRetrospective, one center180Same time44 (62%)HCVMETAVIRNo23 (6-60) mm51%, 28%, 14%11
19Schiavon, 2007, BrazilRetrospective, one center203≤6 months45 (64%)HCVMETAVIRYes13.7±4.8 mm24%, 9%, 3%14
20Toniutto, 2007, ItalyRetrospective, one center102Same time56 (61%)HCVIshakNoUnclear68%, NA, NA10
21Bourliere, 2006, FranceProspective, multicenter235Same time46 (55%)HCVMETAVIRYes16±7.5 mm42%, 24%, 7%13
22Chrysanthos, 2006, GreeceRetrospective, tertiary center284Same time49 (51%)HCVIshakYes≥15 mm51%, NA, 20%14
23Fabris, 2006, ItalyProspective, one center30Unclear38 (65%)HCVIshakYesUnclear13%, NA, NA11
24Lieber, 2006, USARetrospective multicenter133Unclear46 (97%)HCV+ alcoholicIshakNoUnclear44%, NA, NA9
25Liu, 2006, TaiwanProspective, tertiary center79Unclear43 (35%)HCVMETAVIRYes19±1 mm27%, 9%, 0%12
26Macias, 2006, SpainRetrospective, 5 centers263≤1 month37 (84%)HCV+HIVScheuerYes≥15 mm58%, NA, 15%13
27Parise, 2006, BrazilProspective, one center206≤3 months47 (56%)HCVBatts LudwigYesUnclear42%, NA, 21%12
28Romera, 2006, SpainRetrospective, tertiary center131Same time40 (60%)HCVScheuerNoUnclear47%, 17%, 12%10
29Schneider, 2006, GermanyProspective, one center83Unclear49 (49%)HCVIshakNoUnclear57%, NA, 23%9
30Sene, 2006, FranceProspective, tertiary center138Median 1 month (range 0.5-3.5)58 (50%)HCVMETAVIRNo67%≥15 mm47%, NA, 14%11
31Snyder, 2006, USARetrospective, tertiary center339≤4 months45 (72%)HCVBatts LudwigNo23±8 mm49%, 20%, 2%12
  Prospective, tertiary center151Same time48 (70%)HCVBatts LudwigYes22±8 mm52%, 33%, 17% 
32Testa, 2006, ItalyProspective, tertiary center75≤1 day50 (68%)HCVIshakYes≥15 mm49%, 20%, NA13
33Wilson, 2006, USAProspective, multicenter119≤45 days42 (82%)HCVIshakYes11 mm9%, NA, 0%13
34AlMohri, 2005, CanadaRetrospective, 2 centers46≤3 months42 (89%)HCV+HIVBatts LudwigNoUnclear72%, 41%, 20%11
35Islam, 2005, SwedenRetrospective, tertiary center179Same time43 (55%)HCVIshakYes≥10 mm44%, NA, 12%13
36Kelleher, 2005, USARetrospective, tertiary center95Same time45 (63%)HCV+HIVIshakYes10 mm27%, 20%, 16%14
37Lackner, 2005, AustriaRetrospective, two centers194≤1 month48 (57%)HCVIshakNo19±8 mm50%, 26%, 16%11
38Nunes, 2005, USAProspective, 2 centers40≤6 months47 (77%)HCV+HIVIshakYes15 mm48%, NA, 33%13
39Berg, 2004, GermanyRetrospective, multicenter484Unclear46 (59%)HCVScheuerNoUnclear52%, 26%, 13%10
40Wai, 2003, USAProspective, tertiary center192≤4 monthsTraining: 48 (64%)HCVIshakYesUnclear47%, NA, 15%13
  Prospective, tertiary center78≤4 monthsValidation: 48 (66%)HCVIshakYesUnclear50%, NA, 17% 
Table 2. Quality Assessment of Included Studies
 Author, Year, RegionQ1Q2Q3Q4Q5Q6Q7Q8Q9Q10Q11Q12Q13Q14
 Spectrum CompositionSelection CriteriaAppropriate Reference StandardDisease Progression BiasPartial Verification BiasDifferential Verification BiasIncorporation biasTest execution detailsReference execution detailsTest review biasDiagnostic review biasClinical review biasIntermediate resultsWithdrawals
1Calès, 2010, FranceYesYesYesYesYesYesYesYesYesYesYesYesNoYes
2Macías, 2010, SpainYesYesYesUnclearYesYesYesYesYesUnclearUnclearYesYesYes
3Boursier, 2009, FranceYesYesYesYesYesYesYesYesYesYesYesYesNoYes
4Castéra, 2009, FranceYesYesYesYesYesYesYesYesYesYesYesYesYesYes
5Corradi, 2009, ItalyYesYesYesYesYesYesYesYesYesYesYesYesNoYes
6Schiavon, 2009, BrazilYesYesYesYesYesYesYesYesYesYesYesYesYesYes
7Tural, 2009, SpainYesYesYesUnclearYesYesYesYesYesYesYesYesNoYes
8Carvalho-Filho, 2008, BrazilYesYesYesYesYesYesYesYesYesYesYesYesYesYes
9Cheung, 2008, USAYesYesUnclearYesYesYesYesYesYesUnclearUnclearYesYesYes
10Dinesen, 2008, GermanyYesYesUnclearUnclearYesYesYesYesYesYesYesYesNoYes
11Khan, 2008, PakistanYesYesUnclearUnclearYesYesYesYesYesYesYesYesYesYes
12Loko, 2008, FranceYesYesYesUnclearYesYesYesYesYesYesYesYesYesYes
13Paggi, 2008, ItalyYesYesUnclearUnclearYesYesYesYesYesYesYesYesNoYes
14Silva, 2008, BrazilYesYesUnclearYesYesYesYesYesYesYesYesYesNoYes
15Schiavon, 2008, BrazilYesYesYesYesYesYesYesYesYesYesYesYesYesYes
16Trang, 2008, USAYesYesYesYesYesYesYesYesYesUnclearUnclearYesYesYes
17Halfon, 2007, FranceYesYesYesYesYesYesYesYesYesYesYesYesNoYes
18Leroy, 2007, FranceYesYesYesYesYesYesYesYesYesUnclearUnclearYesNoYes
19Schiavon, 2007, BrazilYesYesYesYesYesYesYesYesYesYesYesYesYesYes
20Toniutto, 2007, ItalyYesYesUnclearYesYesYesYesYesYesUnclearUnclearYesNoYes
21Bourliere, 2006, FranceYesYesYesYesYesYesYesYesYesYesYesYesNoYes
22Chrysanthos, 2006, GreeceYesYesYesYesYesYesYesYesYesYesYesYesYesYes
23Fabris, 2006, ItalyYesYesUnclearUnclearYesYesYesYesYesYesYesYesNoYes
24Lieber, 2006, USAYesYesUnclearUnclearYesYesYesYesYesUnclearUnclearYesNoYes
25Liu, 2006, TiawanYesYesYesUnclearYesYesYesYesYesYesYesYesNoYes
26Macias, 2006, SpainYesYesYesYesYesYesYesYesYesYesYesYesNoYes
27Parise, 2006, BrazilYesYesUnclearYesYesYesYesYesYesYesYesYesNoYes
28Romera, 2006, SpainYesYesUnclearYesYesYesYesYesYesUnclearUnclearYesNoYes
29Schneider, 2006, GermanyYesYesUnclearUnclearYesYesYesYesYesUnclearUnclearYesNoYes
30Sene, 2006, FranceYesYesYesYesYesYesYesYesYesUnclearUnclearYesNoYes
31Snyder, 2006, USAYesYesYesYesYesYesYesYesYesUnclearUnclearYesYesYes
32Testa, 2006, ItalyYesYesYesYesYesYesYesYesYesYesYesYesNoYes
33Wilson, 2006, USAYesYesYesYesYesYesYesYesYesYesYesYesNoYes
34Al- Mohri, 2005, CanadaYesYesUnclearYesYesYesYesYesYesUnclearUnclearYesYesYes
35Islam, 2005, SwedenYesYesYesYesYesYesYesYesYesYesYesYesNoYes
36Kelleher, 2005, USAYesYesYesYesYesYesYesYesYesYesYesYesYesYes
37Lackner, 2005, AustriaYesYesYesYesYesYesYesYesYesUnclearUnclearYesNoYes
38Nunes, 2005, USAYesYesYesYesYesYesYesYesYesYesYesYesNoYes
39Berg, 2004, GermanyYesYesUnclearUnclearYesYesYesYesYesUnclearUnclearYesYesYes
40Wai, 2003, USAYesYesUnclearYesYesYesYesYesYesYesYesYesYesYes

Diagnostic Accuracy for the Prediction of Significant Fibrosis.

Thirty-three studies in 6,259 patients assessed the APRI for the prediction of significant fibrosis. The average prevalence of significant fibrosis in these studies was 46% (range, 9%-79%). When combined, the area under the AUROC was 0.77 (SE = 0.012) (Fig. 2). The summary DOR was 6.19 (5.13-7.49), and heterogeneity was not significant in the analysis of significant fibrosis stage (Q = 54.93, I2 = 39.9%). The summary sensitivities and specificities of the APRI at various thresholds for the identification of significant fibrosis are listed in Table 3. At the lower threshold of 0.5 recommended by Wai et al., the summary sensitivities and specificities were 74% (95% confidence interval [CI], 73%-76%) and 49% (47%-51%), respectively. At the higher recommended cutoff of 1.5, the summary sensitivities and specificities were 37% (95% CI, 35%-39%) and 93% (91%-94%), respectively. At the optimal threshold of 0.7, the summary sensitivities and specificities were 77% (95% CI, 72%-81%) and 72% (66%-77%), respectively. Based on these values, and assuming a 46% prevalence of significant fibrosis (as observed in the 33 included studies), the estimated PPV and NPV of the 0.5 cutoff were 55% and 69%, respectively. At the 1.5 cutoff, the estimated PPV and NPV were 82% and 63%, respectively. At the 0.7 cutoff, the estimated PPV and NPV were 70% and 79%, respectively.

Figure 2.

SROC curve of the APRI for significant fibrosis. AUC, area under the SROC curve. The size of the dots for 1-specificity and sensitivity of the single studies in the ROC space is derived from the respective sample size.

Table 3. Summary Sensitivities and Specificities of the APRI at Various Diagnostic Thresholds for Prediction of Significant Fibrosis, Severe Fibrosis and Cirrhosis
Test ThresholdNumber of Studies (Patients)Summary Sensitivity (95% CI)Summary Specificity (95% CI)
Significant Fibrosis   
 0.45 (836)0.88 (0.84-0.92)0.54 (0.50-0.58)
 0.523 (4,595)0.74 (0.73-0.76)0.49 (0.47-0.51)
 0.63 (531)0.76 (0.71-0.81)0.60 (0.54-0.66)
 0.74 (609)0.77 (0.72-0.81)0.72 (0.66-0.77)
 13 (821)0.62 (0.58-0.66)0.45 (0.40-0.51)
 1.23 (571)0.48 (0.42-0.54)0.89 (0.85-0.93)
 1.523 (4,502)0.37 (0.35-0.39)0.93 (0.91-0.94)
 <0.511 (2,052)0.89 (0.86-0.91)0.50 (0.47-0.53)
 0.6-1.013 (2,424)0.68 (0.65-0.71)0.67 (0.65-0.70)
 1.1-1.46 (1,048)0.46 (0.42-0.51)0.89 (0.86-0.91)
 >1.54 (862)0.27 (0.23-0.31)0.95 (0.92-0.97)
Severe Fibrosis   
 0.55 (1,484)0.60 (0.55-0.64)0.43 (0.40-0.46)
 16 (2,111)0.61 (0.57-0.65)0.64 (0.61-0.66)
 1.54 (1,125)0.50 (0.44-0.55)0.87 (0.84-0.89)
 25 (1,908)0.36 (0.32-0.40)0.93 (0.91-0.94)
Cirrhosis   
 113 (2,636)0.76 (0.71-0.80)0.72 (0.70-0.74)
 211 (2,429)0.46 (0.41-0.51)0.91 (0.90-0.93)

According to the meta-regression analysis, APRI accuracy for detecting significant fibrosis was affected by blinding (P = 0.008), with a mean AUROC of 0.80 for studies in which the pathologists were blinded for blood tests, and 0.74 for studies in which the pathologists were not blinded for blood tests. Part of the heterogeneity was explained by liver biopsy scoring system (AUROC of 0.77 for METAVIR; AUROC of 0.76 for Ishak; AUROC of 0.79 for Batts Ludwig; and AUROC of 0.77 for Scheuer). In addition, the AUROC of APRI for detecting HCV monoinfection and HCV/HIV coinfection-related significant fibrosis were 0.79 and 0.75, respectively. However, this difference was not statistically significant in meta-regression analysis. Moreover, the other covariates were not significant (data not shown). An analysis for funnel plot asymmetry suggested possible publication bias for the prediction of cirrhosis (P = 0.002) (Supporting Information Fig. A1).

Diagnostic Accuracy for the Prediction of Severe Fibrosis.

Thirteen studies in 4,441 patients assessed the APRI for the prediction of severe fibrosis. The average prevalence of severe fibrosis in these studies was 28% (range, 9%-59%). When combined, the AUROC was 0.80 (SE = 0.023) (Fig. 3). The summary DOR was 2.24 (1.84-2.73), and heterogeneity was not significant in the analysis of severe fibrosis stage (Q = 5.09, I2 = 0). The summary sensitivities and specificities of the APRI at various thresholds for the identification of severe fibrosis are listed in Table 3. At the optimal threshold of 1, the summary sensitivities and specificities were 61% (95% CI, 57%-65%) and 64% (61%-66%), respectively. Based on these values, and assuming a 28% prevalence of severe fibrosis (as observed in the 13 included studies), the estimated PPV and NPV of the 1 cutoff were 40% and 81%, respectively.

Figure 3.

SROC curve of the APRI for severe fibrosis. AUC, area under the SROC curve. The size of the dots for 1-specificity and sensitivity of the single studies in the ROC space is derived from the respective sample size.

According to the meta-regression analysis, APRI accuracy for detecting severe fibrosis was not affected by the covariates. In addition, the AUROC of APRI for detecting HCV monoinfection and HCV/HIV coinfection-related severe fibrosis were 0.80 and 0.76, respectively. However, this difference was not statistically significant in meta-regression analysis. According to the regression-based analysis of funnel plot asymmetry, there was no evidence of publication bias (P = 0.361) (Supporting Information Fig. A2).

Diagnostic Accuracy for the Prediction of Cirrhosis.

Eighteen studies in 4,548 patients assessed the APRI for the prediction of cirrhosis. The average prevalence of cirrhosis in these studies was 19% (range, 4%-33%). When combined, the AUROC was 0.83 (SE = 0.013) (Fig. 4). The summary DOR was 2.19 (1.77-2.72), and heterogeneity was not significant in the analysis of cirrhosis stage (Q = 3.78, I2 = 0). At the lower recommended threshold of 1.0, the summary sensitivities and specificities were 76% (95% CI, 71%-80%) and 72% (70%-74%), respectively (Table 3). At the higher recommended cutoff of 2.0, the summary sensitivities and specificities were 46% (95% CI, 41%-51%) and 91% (90%-93%), respectively. Based on these values, and assuming a 19% prevalence of cirrhosis (as observed in the 18 included studies), the estimated PPV and NPV of the 1.0 cutoff were 55% and 69%, respectively. At the 2.0 cutoff, the estimated PPV and NPV were 82% and 63%, respectively.

Figure 4.

SROC curve of the APRI for cirrhosis. AUC, area under the SROC curve. The size of the dots for 1-specificity and sensitivity of the single studies in the ROC space is derived from the respective sample size.

According to the meta-regression analysis, APRI accuracy for detecting cirrhosis was affected by blinding (P = 0.001), with a mean AUROC of 0.83 for studies in which the pathologists were blinded for blood tests, and 0.82 for studies in which the pathologists were not blinded for blood tests. APRI accuracy for detecting cirrhosis was also affected by the research methods (P = 0.001), with a mean AUROC of 0.81 for the retrospective studies, and 0.86 for the prospective studies. APRI accuracy for detecting cirrhosis was also affected by quantitative factors, such as QUADAS score (P = 0.002), median age (P = 0.007), and the prevalence of cirrhosis (P = 0.002). Moreover, the AUROC of APRI for detecting HCV monoinfection and HCV/HIV coinfection-related cirrhosis were 0.83 and 0.79, respectively. However, this difference was not statistically significant in meta-regression analysis. The other covariates were not significant (data not shown). According to the regression-based analysis of funnel plot asymmetry, there was no evidence of publication bias (P = 0.093) (Supporting Information Fig. A3).

Discussion

Liver fibrosis is the excessive accumulation of extracellular matrix (ECM) resulting from chronic liver diseases. Factors associated with matrix deposition or degradation and some cytokines involved in fibrosis may be used as individual markers or as a combination of markers to generate an algorithm to evaluate the stage of fibrosis. Also, the stage of fibrosis may be predicted using indirect markers such as a single routine laboratory test or multicomponent indirect fibrosis tests. Considering the limitations and risks of biopsy, as well as the improvement of diagnostic accuracy of noninvasive biochemical markers, there is great interest in developing and validating noninvasive methods to detect hepatic fibrosis among patients with chronic liver disease, and liver biopsy should no longer be considered mandatory. APRI is a novel index of liver fibrosis initially validated in patients with CHC, and then in the other common fibrotic liver diseases. It showed great value in detecting liver fibrosis, based on routine laboratory parameters.

In this systematic review and meta-analysis, we identified and evaluated 40 studies from the published literature comparing APRI with liver biopsy for detecting HCV-related fibrosis. Our meta-analysis showed that that the accuracy of the APRI is perhaps less than initially described. In the original Wai et al.8 study, the AUROC for significant fibrosis and cirrhosis were 0.80 to 0.88 and 0.89 to 0.94, respectively. In our meta-analysis the summary AUROC of the APRI for the diagnosis of significant fibrosis was 0.77. Moreover, the 0.5 threshold was 74% sensitive and 49% specific. Assuming a 46% prevalence of significant fibrosis (as observed in the included studies), this translates into an estimated PPV of 55% and NPV of 69%. On the contrary, a cutoff of 1.5 was more specific (93%) but less sensitive (37%). Assuming a 46% prevalence of significant fibrosis, this translates into an estimated PPV of 82% and NPV of 63%. Optimal cutoff values for APRI were chosen to maximize the sum of sensitivity and specificity, thereby optimizing the diagnostic performance (the sum of true positives and true negatives over the total number of patients). The 0.7 threshold appears promising, maximizing the sum of sensitivity and specificity (sensitivity, 77%; specificity, 72%). Assuming a 46% prevalence of significant fibrosis, this translates into an estimated PPV of 70% and NPV of 79%. With respect to severe fibrosis, the summary AUROC was 0.80. Moreover, the 1.0 threshold was 61% sensitive and 64% specific. Assuming a 28% prevalence of severe fibrosis (as observed in the included studies), this translates into an estimated PPV of 40% and NPV of 81%. With respect to cirrhosis, the summary AUROC was 0.83. Moreover, the 1.0 threshold was 76% sensitive and 72% specific. Assuming a 19% prevalence of severe fibrosis (as observed in the included studies), this translates into an estimated PPV of 55% and NPV of 69%. On the contrary, a cutoff of 2.0 was more specific (91%) but less sensitive (46%). Assuming a 19% prevalence of severe fibrosis, this translates into an estimated PPV of 82% and NPV of 63%.

Compared with the previous meta-analysis,9 our results showed similar performance of the APRI for the staging of significant fibrosis and cirrhosis. Moreover, APRI tended to show less accurate results for the identification of significant fibrosis, severe fibrosis, and cirrhosis in HIV/HCV coinfected patients than HCV monoinfected patients, which was different from the previous meta-analysis. This finding was in accord with our hypothesis that its accuracy may be diminished in coinfected patients because of HIV-related or antiretroviral-related thrombocytopenia.60 However, this difference was not statistically significant in meta-regression analysis. A diagnostic tool is defined as perfect if the AUROC is 100%, excellent if the AUROC is greater than 90% and good if the AUROC is greater than 80%. According to these results, APRI can be used in clinical practice as a good tool for the confirmation of severe fibrosis and cirrhosis when other clinical signs and examinations are nondecisive.

Based on these results, APRI shows less value to identify HCV-related fibrosis than some other noninvasive methods. With respect to FibroTest, a meta-analysis by Shaheen and Myers9 showed that the AUROC of FibroTest to detect HCV-related significant fibrosis and cirrhosis was 0.81 and 0.90, respectively. With respect to transient elastography, it showed that the AUROC of FibroScan to detect HCV-related significant fibrosis and cirrhosis was 0.83 and 0.95, respectively.61 Although APRI shows less diagnostic accuracy than FibroTest and FibroScan to identify HCV-related significant fibrosis and cirrhosis, APRI, a tool with limited expense and widespread availability, is still an attractive first-line estimate of liver fibrosis, particularly in regions with limited healthcare resources, where the prevalence of HCV tends to be the highest. According to World Health Organization estimates, over 85% of the 170 million HCV patients worldwide reside outside of the Americas and Europe, the majority in developing countries.

A strength of our review is that meta-regression analyses have been used for exploring factors that may be responsible for heterogeneity. We selected the following predefined characteristics as potential covariates that might contribute heterogeneity: (a) study design (respective versus prospective); (b) etiology (HCV versus HCV/HIV); (c) blinded interpretation of APRI and reference standard (yes versus no); (d) liver biopsy length (≥15 mm or not); (e) liver biopsy scoring system (METAVIR, Ishak, Batts Ludwig, and Scheuer); (f) QUADAS score; (g) sample size; (h) median age; (i) percentage of males; (j) location of study (North America, Europe, and other); (k) prevalence of significant fibrosis/severe fibrosis/cirrhosis. Among patients with significant fibrosis, blinding and liver biopsy scoring systems were found to provide heterogeneity to summary test results. Among patients with severe fibrosis, neither of these variables was found to provide heterogeneity to summary test results. Among patients with cirrhosis, blinding, research methods, QUADAS score, median age, and the prevalence of cirrhosis were found to provide heterogeneity to summary test results. The other advantage of the present study is the large number of studies included, as well as the opportunity to analyze an integrated database. This permitted taking into account the variability factors associated with APRI diagnostic value.

Our systematic review and meta-analysis also have several limitations. One limitation is that we focused our analysis on HCV-infected patients only. The APRI has been used to examine chronic hepatitis B (CHB),42, 62 alcoholic liver disease (ALD),44 and nonalcoholic fatty liver disease (NAFLD),63, 64 but the published studies suggest reduced accuracy. If the APRI diagnostic value was indeed less in patients with HCV than in patients with the three other frequent fibrotic diseases, it warrants further confirmation. Because there were few published studies related to the above chronic liver diseases-related fibrosis, we restricted our analysis to HCV. Another limitation is that we included English studies only, so the language bias may influence the results to some extent.

In summary, our meta-analysis suggests that the APRI has moderate diagnostic utility for the prediction of fibrosis in HCV-infected patients. Although APRI shows less diagnostic accuracy than some other noninvasive methods, APRI is still the first choice for HCV patients to identify hepatitis C-related fibrosis in regions with limited healthcare resources. Future studies of novel fibrosis markers should demonstrate improved accuracy and cost-effectiveness compared with this simple, economical, and widely available index.

Acknowledgements

The authors thank Professor Tian-Song Zhang, Senior Medical Statistician, Jing'An District Centre Hospital, Shanghai, China, and Professor An-Jin Chen, Senior Medical Statistician, Qingdao Municipal Hospital, Qingdao, China, for their valuable statistical assistance.

Ancillary

Advertisement