Annals of Hepatology 2008; 7(4): October-December: 350-357

Original Article

Annals
of
Hepatology AST to platelet ratio index (APRI)
for the noninvasive evaluation of liver fibrosis
Aurora Loaeza-del-Castillo;1 Francisco Paz-Pineda;1 Edgar Oviedo-Cárdenas;2 Francisco Sánchez-Ávila;1 Florencia
Vargas-Vorácková1

Abstract CHC, 0.564 in NAFLD, and 0.602 in AIH patients. For

advanced fibrosis, the AUCs were 0.803, 0.568 and
Liver biopsy is the recognized gold standard for liver 0.532 in CHC, NAFLD and AIH patients, respectively.
fibrosis staging. The aspartate aminotransferase to For cirrhosis, AUCs were 0.830 and 0.599 in CHC and
platelet ratio index (APRI) has been proposed as a AIH patients. In conclusion, APRI can be a useful
noninvasive and readily available tool for the assess- noninvasive alternative for the diagnosis of signifi-
ment of liver fibrosis in chronic hepatitis C (CHC). cant fibrosis and cirrhosis in our CHC patients. APRI
This study aimed to validate, in a Mexican tertiary values of d 0.3 and d 0.5 rule out significant fibrosis
health care setting, the diagnostic usefulness of APRI and cirrhosis, and a value of t 1.5 rules in significant
in CHC, nonalcoholic fatty liver disease (NAFLD) and fibrosis. In patients with NAFLD, APRI values tend to
autoimmune hepatitis (AIH). In an observational, increase with the degree of fibrosis, suggesting that it
cross-sectional, comparative and retrolective fashion, could be useful in this disease. APRI appears to be of
consecutive patients with CHC, NAFLD or AIH were no value in patients with AIH.
evaluated. Fibrosis was staged using the METAVIR
scale. Receiver operating characteristic ROC curves Key words: Fibrosis staging, chronic hepatitis C, non-
were constructed for significant fibrosis, advanced fi- alcoholic fatty liver disease, autoimmune hepatitis.
brosis and cirrhosis. One-hundred-sixty-four CHC, 30
NAFLD and 42 AIH patients were evaluated. For the Introduction
diagnosis of significant fibrosis, APRI values delimit-
ed an area under de ROC curve (AUC) of 0.776 in Liver fibrosis is the result of a chronic injury induced
by a variety of causes. Bridging fibrosis and regeneration
nodes are the clearest manifestation of this injury, being
1
Departments of Gastroenterology. cirrhosis the end stage of this process. In cirrhosis, the
2
Department of Pathology. molecular composition of fibrotic tissue is unique, irre-
spective of the etiology.1,2
Instituto Nacional de Ciencias Médicas y Nutrición Salvador
Zubirán, Mexico City, Mexico. Liver function decreases as extracellular matrix de-
posits in the subendothelial space. Despite the presence
Abbreviations: of cirrhosis, up to 40% of patients remain asymptomatic
APRI, aspartate aminotransferase to platelet ratio index for years. In these cases, liver biopsy is the most reliable
CHC, chronic hepatitis C
NAFLD, nonalcoholic fatty liver disease means for diagnosis.3
AIH, autoimmune hepatitis Evaluation of liver fibrosis in chronic liver disease has
ROC, receiver operating characteristic a dual purpose, treatment and prognosis. At the bedside,
AUC, area under de curve it allows assessment of therapeutic interventions and fol-
PLR, positive likelihood ratio
low-up of fibrosis progression. In chronic hepatitis C
Address for correspondence: (CHC) patients, the stage of liver fibrosis is a predictor of
Florencia Vargas-Vorácková response to interferon-based treatments. It is known that
Department of Gastroenterology the presence of advanced fibrosis is a predictor of non-re-
Instituto Nacional de Ciencias Médicas y Nutrición Salvador
Zubirán, Vasco de Quiroga 15, Colonia Sección XVI, Delegación sponse.1 Hepatitis C virus (HCV) infection evolves to
Tlalpan, Mexico City chronicity in 50-85% of cases, with the development of
14000, Mexico liver fibrosis, cirrhosis and its complications.
Telephone +52-55-5573-3418 To date, a variety of means are available for the stag-
Fax +52-55-5655-0942
E-mail: [email protected] ing of liver fibrosis. This staging is cross-sectional, and
does not reflect the dynamics of the liver fibrosis pro-
Manuscript received and accepted: 13 May and 20 August 2008 cess. Among these means, liver biopsy is the gold stan-

© 2019, Fundación Clínica Médica Sur, A.C. Published by Elsevier España S.L.U. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
 A Loaeza-del-Castillo et al. AST to platelet ratio index (APRI) for the noninvasive evaluation of liver fibrosis 351

dard. This procedure is invasive, and has known adverse documented liver disease were reviewed. Presence of liv-
events and limitations. In most cases, it is performed er disease was sustained with abnormal serum transami-
blindly by means of a percutaneous puncture, resulting nases, GGT, ultrasonography and/or liver biopsy.
in a small tissue sample that hardly represents the liver Diagnosis of CHC was defined as positive RNA-
(approximately 1/50,000th). This frequently leads to mis- HCV (quantitative PCR) and/or positive serum anti-
classification of liver fibrosis which involves, at least, HCV. Diagnosis of NAFLD was defined as the presence
one stage.4 Intra- and inter-observer variability contrib- of at least one feature of the metabolic syndrome,
utes to this misclassification with a 10 to 20%, depend- echogenic appearance on ultrasound consistent with
ing on the scoring system used.5 The most frequent com- steatosis, and a daily consumption of < 50 g of ethanol
plications of percutaneous liver biopsy are pain and in males or < 30 g in females. Metabolic syndrome was
bleeding, death occurs in 1/10,000 to 1/12,000 proce- defined according to the Adult Treatment Panel III (ATP
dures.6,7 III) criteria, i.e. i) triglycerides t 150 mg/dL or taking
For noninvasive, but indirect evaluation of liver fi- medication for hypertriglyceridemia, ii) fasting plasma
brosis, positron emission tomography, ultrasound and glucose t 110 mg/dL or taking medication for diabetes,
magnetic resonance imaging have shown to be promis- iii) blood pressure t 140/t 90 mmHg or taking medica-
ing.8 A novel method, the elastography (FibroScan), has tion for hypertension, iv) HDL cholesterol < 35 mg/dL
shown to be useful for the evaluation of liver fibrosis in in men or < 39 mg/dL in women, and v) abdominal obe-
CHC patients.9,10 However, high costs have limited a sity (waist circumference > 102 cm in men or > 88 cm in
broad use of this technology. women).13 AIH was defined as positive anti-nuclear and
Other noninvasive methods, such as serum markers anti-smooth muscle antibodies or positive anti-LKM1
have been proposed. Among them are enzymes that regu- antibodies. Clinical diagnosis was sustained in all cases
late extracellular matrix synthesis or degradation, and ex- by compatible histological findings in liver biopsies. In
tracellular matrix degradation products. None of these our setting, a diagnostically useful liver biopsy requires
markers has shown to be liver specific, sensible enough at least one cm of tissue length, and the presence of at
to diagnose different fibrosis stages, or easy to perform to least five portal triads.
justify routine clinical use. All patients had a liver biopsy and a platelet count
Combination of serum markers lead to the develop- performed the same day as the biopsy. They also had an
ment of indexes such as the Fibrotest, which includes D2- AST determination within a month before/after the biop-
macroglobulin, D2-globulin (haptoglobin), J-glutamil- sy. Patients were excluded if liver biopsy was not useful
transpeptidase, J-globulin, total bilirubin and apolipo- for fibrosis staging, or if liver injury was due to hepatitis
protein A1.4,11 This index, controlled for age and sex, has B virus, HIV, cholestasis or other cause of chronic liver
shown good correlation with the stage of liver fibrosis in inflammation. Patients were also excluded if having my-
CHC patients. However, the complex and costly analysis opathy or thrombocytopenia not related to portal hyper-
of most of its components limits its use in our country tension.
and, probably, in other developing environments. All liver biopsies were reviewed by a single patholo-
Recently, an index comprising routinely available gist (EOC). Fibrosis was staged according to the
laboratory tests was developed for the assessment of liver METAVIR scale, where F0 stands for no fibrosis, F1 for
fibrosis in CHC patients, namely the aspartate ami- portal fibrosis without septa, F2 for portal fibrosis with
notransferase to platelet ratio index (APRI).12 For bridg- rare septa, F3 for numerous septa without cirrhosis, and
ing fibrosis and cirrhosis, the proposed APRI cutoff val- F4 for cirrhosis.14
ues are > 1.5 and > 2, respectively, with areas under the The APRI was calculated according to the formula:12
receiver operating curve (ROC) of 0.80 and 0.89.
A noninvasive diagnostic test for liver fibrosis should AST level
be simple, available, inexpensive and accurate. All these ULN *
characteristics are fulfilled by the APRI. Therefore, we APRI= x100
Platelet counts (10 9 / L)
aimed to validate the APRI usefulness for the diagnosis
of different stages of liver fibrosis in CHC in our own set-
ting. Given that fibrosis and cirrhosis are not exclusive *ULN, AST upper level of normal (or 56 IU/L)
to CHC, our validation was extended to nonalcoholic fat-
ty liver disease (NAFLD), as well as to autoimmune hepa- Results are expressed as absolute frequencies (%),
titis (AIH). means ± standard deviations, and medians (minimum –
maximum). A Wilcoxon-type test for trend was used to
Material and methods test the trend of APRI values across fibrosis stages.15
Strength of association among APRI values and fibrosis
Observational, cross-sectional, comparative and ret- stages was estimated by means of Spearman correlation
rolective study. Charts of patients aged t 18 years, with indexes.
352 Annals of Hepatology 7(4) 2008: 350-357

Receiver operating characteristic (ROC) curves were graphic and laboratory characteristics of these patients
estimated for CHC, NAFLD and AIH, considering i) sig- are summarized in Table I.
nificant fibrosis (METAVIR t 2), ii) advanced fibrosis Significant fibrosis (METAVIR t 2) was observed in
(METAVIR t 3) and iii) cirrhosis (METAVIR 4). Areas 83 (51%) patients with CHC, in 15 (50%) patients with
under the ROC curves (AUC) were determined, as well as NAFLD, and in 29 (69%) patients with AIH. Advanced fi-
their 95% confidence intervals (95%CI). The APRI val- brosis (METAVIR t 3) was observed in 67 (41%) pa-
ues with both best sensitivity and specificity were con- tients with CHC, in 3 (10%) patients with NAFLD, and in
sidered as the optimal diagnostic threshold or cutoff 20 (47%) patients with AIH. Cirrhosis (METAVIR 4) was
point. At this thresholds, sensitivity, specificity, predic- observed in 47 (29%) CHC and in nine (21.5%) AIH pa-
tive values, accuracy and positive likelihood ratio (PLR), tients. No patient with NAFLD had cirrhosis (Table I).
with their respective 95%CI, were determined. PLRs Median APRI values in CHC patients increased ac-
weigh the relationship between two proportions, namely cording to fibrosis stage, namely 0.39 for F0, 0.505 for
the proportion of a positive test in disease (true positives) F1, 0.545 for F2, 0.745 for F3, and 1.64 for F4 (P < 0.001,
and the proportion of a positive test in non-disease (false Figure 1) For the same stages of fibrosis, median APRI
positives). The higher this ratio, the better the diagnostic values in NAFLD patients increased gradually from 0.28
performance of the APRI score. A PLR of three is consid- to 0.515, 0.585 and 0.67.(P < 0.44, Figure 2). In AIH pa-
ered clinically significant, and is interpreted as «the tients median APRI values were 8, 3.175, 13.3, 4.7 and
probability of a positive test (t APRI cutoff value) 9.26 for F0, F1, F2, F3 and F4, respectively(P < 0.40, Fig-
among the diseased (more fibrosis) is three times the ure 3). Correlations among APRI values and fibrosis stag-
probability of a positive test among the non-diseased es were 0.564 (P < 0.001) in CHC, 0.140 (P < 0.46) in
(less fibrosis)». Finally, and according to published evi- NAFLD, and 0.134 (P < 0.40) in AIH.
dence,16 a logistic regression was carried out to explore For the diagnosis of significant fibrosis (METAVIR t
sex and age effect on the performance of the APRI score 2) in CHC patients, APRI values delimited an AUC of
in the diagnosis of cirrhosis in CHC patients. 0.776 (95%CI 0.704-0.847; P < 0.001), with a threshold
A P value of < 0.05 was considered as statistically sig- of 0.6433. At this threshold, sensitivity was 74.7%
nificant. Stata v7.0 and SPSS v12.0 statistical packages (95%CI 64-83.6%), specificity 67.9% (56.6-77.8%), posi-
were used. tive predictive value 70.5% (59.8-79.7%), negative pre-
dictive value 72.4% (60.9-82%), accuracy 71.3% (63.8-
Results 78.1%), and PLR 2.33 (1.66-3.27)(Figure 4). For NAFLD
and AIH patients, the AUC was 0.564 (95%CI 0.347-
Two-hundred-thirty-six patients fulfilled de selection 0.782; P < 0.548) and 0.602 (0.418-0.787; P < 0.295), re-
criteria, 164 had CHC, 30 NAFLD and 42 AIH. Demo- spectively.

Table I. Distribution of demographic characteristics and laboratory results by diagnosis.

Characteristic CHC n = 164 NAFLD n = 30 AIH n = 42

Age, years 49.4 ± 12 43 ± 12 38 ± 11

Sex
females 105 (64%) 17 (57%) 40 (95%)
males 59 (36%) 13 (43%) 2 (5%)
Genotype (n = 124)
1 91 (73.4%) – –
2 27 (21.8%)
3 3 (2.4%)
4 3 (2.4%)
RNA-HCV 380,000 – –
(50-7’690,000)
AST, IU/L 74.5 72 201
(15-331) (16-703) (46-2,000)
Platelets, /mm3 184,000 262,000 158,500
(38,000-351,000) (150,000-410,000) (29,000-290,000)
Metavir
F0 31 (19%) 5 (17%) 3 (7%)
F1 50 (30%) 10 (33%) 10 (24%)
F2 16 (10%) 12 (40%) 9 (21.5%)
F3 20 (12%) 3 (10%) 11 (26%)
F4 47 (29%) 0 9 (21.5%)

CHC, chronic hepatitis C; NAFLD, nonalcoholic fatty liver disease; AIH, autoimmune hepatitis.
Data are expressed as mean ± standard deviation, median (minimum – maximum) and absolute frequencies (%).
 A Loaeza-del-Castillo et al. AST to platelet ratio index (APRI) for the noninvasive evaluation of liver fibrosis 353

For the diagnosis of advanced fibrosis (METAVIR t of 0.7532. At this threshold, sensitivity was 77.6%
3) in CHC patients, APRI values delimited an AUC of (95%CI 65.8-86.9%), specificity 75.3% (65.5-83.5%),
0.803 (95%CI 0.735-0.872; P < 0.001), with a threshold positive predictive value 68.4% (56.7-78.6%), negative
predictive value 83% (73.4-90.1%), accuracy 76.2% (69-
82.5%), and PLR 3.14 (2.17-4.54)(Figure 5). For NAFLD
and AIH patients, the AUC was 0.568 (95%CI 0.186-
0.95; P < 0.704) and 0.532 (0.353-0.711; P < 0.724), re-
spectively.
For the diagnosis of cirrhosis (METAVIR 4) in CHC
patients, APRI values determined an AUC of 0.830
(95%CI 0.765-0.895; P < 0.001), with a threshold of
10 0.7532. At this threshold, the sensitivity was 89.4%
(95%CI 76.9-96.5%), specificity 70.9% (61.8-79%), pos-
APRI

itive predictive value 55.3% (43.4-66.7%), negative pre-

dictive value 94.3% (87.2-98.1%), accuracy 76.2% (69-
82.5%), and PLR 3.08 (2.28-4.15)(Figure 6). For AIH,
this area was 0.599 (95%CI 0.398-0.800; P < 0.366).

1 Discussion

Prognosis and treatment of chronic liver diseases is re-

-0.1
lated to the degree of fibrosis. To evaluate fibrosis, liver
biopsy has been the gold standard for many years. How-
0 1 2 3 4 ever, invasiveness, complications, inter- and intra-ob-
Stage of fibrosis
server variations, and patient acceptability have limited
Figure 1. Distribution of APRI values according to fibrosis stage its use, leading to search for new and noninvasive meth-
in patients with CHC. Boxplots depict the median (heavy horizon- ods to diagnose liver fibrosis and cirrhosis.
tal line), the quartiles (lower and upper edges of the box), and the The APRI score correlates significantly to fibrosis
minimum and maximum values (vertical whiskers). Outliers are stage in patients with CHC.17 It is based on two routine
depicted as «o» and extreme values as «*».
laboratory tests and is, therefore, a promising tool with

10 10
APRI

APRI

1 1

-0.1 -0.1
0 1 2 3 0 1 2 3 4
Stage of fibrosis Stage of fibrosis

Figure 2. Distribution of APRI values according to fibrosis stage Figure 3. Distribution of APRI values according to fibrosis stage
in patients with NAFLD. Boxplots depict the median (heavy hori- in patients with AIH. Boxplots depict the median (heavy horizon-
zontal line), the quartiles (lower and upper edges of the box), and tal line), the quartiles (lower and upper edges of the box), and the
the minimum and maximum values (vertical whiskers). Outliers minimum and maximum values (vertical whiskers). Outliers are
are depicted as «o» and extreme values as «*». depicted as «o» and extreme values as «*».
 354 Annals of Hepatology 7(4) 2008: 350-357

1.0 1.0

APRI = 0.7532
0.8 APRI = 0.6433 0.8

0.6 0.6
Sensitivity

Sensitivity
AUC = 0.776
AUC = 0.803
0.4 0.4

0.2 0.2

0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
1- Specificity 1- Specificity

Figure 4. Receiver operating characteristic curve of APRI values Figure 5. Receiver operating characteristic curve of APRI values
for the diagnosis of significant fibrosis (METAVIR t F2) in pa- for the diagnosis of advanced fibrosis (METAVIR t F3) in pa-
tients with CHC. AUC, area under the ROC curve. tients with CHC. AUC, area under the ROC curve.

limited expense and widespread availability.16 It is 1.0

known that platelet counts decrease and AST levels in- APRI = 0.7532
crease with the progression of liver fibrosis. Platelet gen-
eration diminishes secondary to a decreased production 0.8
of thrombopoietin by hepatocytes.18,19 Also, platelets are
sequestered and destructed in the spleen as liver fibrosis
advances and portal hypertension develops.20 As to AST,
0.6
ongoing liver injury increases its release from mitochon-
Sensitivity

dria,21 and fibrosis decreases its clearance.22

We here-in evaluated the performance of the APRI AUC = 0.830
score in the diagnosis of fibrosis and cirrhosis in CHC, 0.4
NAFLD and AIH patients. Its diagnostic usefulness was
confirmed in our patients with CHC. In patients with
NAFLD, a tendency towards increasing APRI scores 0.2
across fibrosis stages was observed. APRI scores, howev-
er, did not show a pattern in our patients with AIH.
For the diagnosis of significant fibrosis in CHC pa- 0.0
tients, we obtained an AUC of 0.776. This AUC was very 0.0 0.2 0.4 0.6 0.8 1.0
close to the pooled AUC of 0.76 (95%CI 0.74-0.79), re- 1- Specificity
ported in a recent meta-analysis. In this same meta-analy-
sis, sensitivities of 81% (95%CI 76-86%) and 35% (30- Figure 6. APRI receiver operating characteristic curve of APRI
41%), and specificities of 50% (47-52%) and 91% (89- values for the diagnosis of cirrhosis in patients with CHC. AUC,
area under the ROC curve.
92%), were observed at APRI thresholds of 0.5 and 1.5,
respectively.16 Our sensitivities (86 and 41%) and speci-
ficities (53 and 90%) at the same thresholds were within power of 2.33, which contrasts with the 5.7 published in
the reported 95%CIs and/or similar. The cutoff point the above referred meta-analysis. This notorious differ-
showing the best sensitivity (74.7%) and specificity ence, can be explained with the somewhat lower sensitiv-
(67.9%) was, in our series, an APRI score of 0.6433, ity and higher specificity of the APRI score in our study,
which is close to the 0.5 value recommended by other au- as well as an inherent overstatement of the diagnostic
thors.12,16 In terms of PLR, a proxi of the diagnostic odds performance of the test by the DOR estimates obtained
ratio (DOR), the APRI score had, for the diagnosis of sig- by Shaheen et al.16,23 Assuming the 0.5 recommended
nificant fibrosis in our CHC patients, a discrimination threshold, the use of the APRI score in our setting renders
 A Loaeza-del-Castillo et al. AST to platelet ratio index (APRI) for the noninvasive evaluation of liver fibrosis 355

a limited 14% gain for the diagnosis of significant fibro- brosis and cirrhosis could be strongly included with
sis (51% pre- to 65% post-test probability). To exclude APRI scores t 2. Clinically, an APRI value of d 0.3 re-
significant fibrosis, however, the gain is twice as much, duces the pre-test probability of significant fibrosis from
or 29% (49% pre- to 78% post-test probability). There- 0.51 to d 0.20, which is sufficiently low to exclude sig-
fore, when < 0.5, the APRI score might be useful to ex- nificant fibrosis without resorting to a liver biopsy, thus
clude significant fibrosis, as proposed previously. avoiding 15% of these procedures. An APRI value of >
For the diagnosis of cirrhosis in CHC patients, we ob- 1.5, in turn, increases the pre-test probability of signifi-
tained an AUC of 0.830. This AUC was also very close to cant fibrosis from 0.51 to t 0.80, which is sufficiently
the pooled AUC of 0.82 (95%CI 0.79-0.86) reported in high to include significant fibrosis, avoiding 26% liver
Shaheen’s meta-analysis. At thresholds of 1.0 and 2.0, es- biopsies. Together, APRI values of d 0.3 and > 1.5 could
timated sensitivities were 76% (95%CI 68-82%) and avoid 41% biopsies in our patients with suspected sig-
49% (43-55%), and specificities 71% (69-73%) and 91% nificant fibrosis. As to cirrhosis, an APRI value of d 0.5
(90-93%), respectively.16 Our sensitivities at the same reduces the pre-test probability of this advanced stage of
thresholds were 77 and 38%, and specificities 77 and fibrosis from 0.29 to d 0.04, which is sufficiently low to
92%. The 95%CIs of all these parameters overlapped exclude cirrhosis without resorting to a liver biopsy,
those estimated in the meta-analysis. In our series, the avoiding 35% of these procedures. APRI values > 2, de-
cutoff point showing the best sensitivity (89.4%) and spite of providing a 35% diagnostic gain [(0.64-0.29) x
specificity (70.9%) was an APRI value of 0.7532, which 100], derive in post-test probabilities around 0.65, indi-
is close to the 1.0 threshold of the meta-analysis, and al- cating that other diagnostic resources are needed to rule
most identical to the one originally found by Wai.12 Our in this diagnosis. These findings are in agreement with
PLR of 3.08 was, again, notably lower than the pooled those obtained recently by Carvalho et al, in Brasil,25 and
one, i.e. 11.3. With this regard, and considering that a suggest that APRI thresholds are to be adjusted accord-
PLR above 10 (or less than 0.10) gives convincing diag-
nostic evidence, whereas one above 5 (or less than 0.20)
gives strong diagnostic evidence,24 the APRI score in our Table II. Positive likelihood ratios for the diagnosis of significant
environment apparently provides a less than strong diag- fibrosis and cirrhosis in patients with chronic hepatitis C at different
APRI thresholds.
nostic evidence at the recommended threshold values.
This might be due to the sex and age composition of our Post-test
CHC sample, which had a substantial proportion of fe- Diagnosis APRI PLR probability
males (64%) and a mean age of 49.4 years. Adjusting for
Significant fibrosis, < 0.2 0.00 0.00
sex, the APRI delimited area under the ROC curve did Pre-test 0.2-0.29 0.13 0.12
not show clinical difference but, statistically, tended to probability = 0.51 0.3-0.39 0.30 0.24
be higher in males than in females (0.825 vs 0.822, P < 0.4-0.49 0.65 0.40
0.065). Age also tended to affect the diagnostic accuracy 0.5-0.69 0.85 0.47
0.7-0.99 1.34 0.58
of the APRI score (P < 0.060) but, again, had no clinical- 1.0-1.49 1.59 0.62
ly significant effect. These findings are consistent with 1.5-1.99 3.58 0.79
those obtained in a published meta-regression analysis, t2 4.49 0.82
where the accuracy of the APRI score for the detection of
Cirrosis, < 0.5 0.10 0.04
cirrhosis was greater in studies with a higher proportion Pre-test 0.5-0.69 0.30 0.42
of males (63%), younger age (45.5 years) and, particular- probability = 0.29 0.7-0.99 1.15 0.32
ly, in patients with coexisting HIV infection.16 1.0-1.49 2.26 0.48
Given that, in terms of areas under the ROC curves, 1.5-1.99 3.32 0.58
2.0-2.49 4.36 0.64
the accuracy of the APRI score in our CHC patients was t 2.5 4.56 0.65
similar to the published ones, and in order to analyze its
clinical impact in our setting, we estimated PLRs for the Pre-test probability | Prevalence, Post-test probability | Positive predictive value.
PLR, positive likelihood ratio. Values in bold depict either rule-out or rule-in values.
diagnosis of significant fibrosis and cirrhosis at gradually
increasing thresholds (Table II). Here, the diagnostic per-
formance of the APRI score was adjusted to the pre-test Appendix 1.
probability (prevalence) of significant fibrosis and cir-
rhosis found in our study, namely 0.51 and 0.29. In other Pre - test probability
1. Pre - test odds =
settings, different pre-test probabilities are expected to 1 - (Pre - test probability)
derive in different post-test probabilities (positive pre-
dictive values) (Appendix 1). This analysis disclosed 2. Post - test odds = (Pre - test odds)x(Positive likelihood ratio)
that, in theory,24 an APRI score of < 0.2 is convincingly
exclusive of significant fibrosis in our setting, and a Post - test odds
3. Post - test probability =
score of < 0.5 is convincingly exclusive of cirrhosis. Fi- 1 + (Post - test odds)
356 Annals of Hepatology 7(4) 2008: 350-357

ing to the site of use or, as discussed above, stratified ac- the clinical course of this disease is characterized by con-
cording to the local CHC patient demographic and clini- tinuous exacerbations and remissions. When active, it
cal profile. presents with significant increases of both AST and ALT,
Another option to improve the non-invasive diagnosis which might further affect the accuracy of the APRI
of liver fibrosis or cirrhosis is the combined use of tests. score. The use of immunosuppressive treatments is anoth-
Here, besides de APRI score, the options are the Fi- er potential confounder of the accuracy. Unfortunately,
broTest and the FibroScan. In a recent evaluation of both activity and treatment were not controlled in this
these tests in a sample of 183 CHC patients, the area un- study, due to the small sample size. Our findings, howev-
der the ROC curve for the diagnosis of significant fibrosis er, are not encouraging enough to further evaluate the
was 0.78, 0.85 and 0.83, respectively. For the diagnosis APRI score in AIH. An unsatisfactory performance of the
of advanced fibrosis these areas were 0.84, 0.9 and 0.9. APRI score has also been observed in patients with alco-
Finally, the areas for the diagnosis of cirrhosis were 0.83, holic fibrosis or cirrhosis. Here, heavy alcohol intake in-
0.87 and 0.95. The highest diagnostic yield was ob- terferes affecting both AST and platelet count prior and
tained by the combined use of FibroTest and FibroScan. independently to the development of fibrosis.29 This sug-
Here, the areas under the ROC curve were 0.88, 0.95 and gests that the diagnostic usefulness of the APRI score
0.95 for the diagnosis of significant fibrosis, advanced fi- might not be generalizable to chronic liver diseases of all
brosis and cirrhosis, respectively.26 The combined use of etiologies justifying, in addition to a site-specific valida-
these two non-invasive tests in our setting is, however, tion, an etiology-specific one.
almost impossible due to their limited availability and In summary, in our CHC patients, the APRI score can
high cost. be a useful non-invasive alternative for the exclusion and
In patients with NAFLD, the APRI score seldomly inclusion of significant liver fibrosis, as well as for the
reached the value of 1, but a tendency towards higher exclusion of cirrhosis. At a threshold value of d 0.3 it
values in patients with advanced stages of fibrosis was rules out significant fibrosis, at a value of d 0.5 it rules
observed. This could be explained by gradual increases out cirrhosis, and at a threshold of t 1.5 it rules in signif-
of AST in the presence of normal platelet counts. Pa- icant fibrosis. The consistency of these findings needs to
tients with NAFLD characteristically present mild to be proven in other settings showing a similar CHC pa-
moderate increases in transaminase levels.27 Platelet tient demographic and clinical profile. The need of sex
counts, on the contrary, are usually normal, at least in the and age specific APRI thresholds warrants further evalu-
early stages of fibrosis. We should acknowledge that our ation. In patients with NAFLD, the APRI score appears to
ability to show a relationship among APRI values and increase with a higher METAVIR score, suggesting that
stage of fibrosis in this group was limited by the inclu- it might also be useful to diagnose fibrosis in this dis-
sion of subjects with a somewhat high cut-off value of ease. Probably due to the involvement of different mech-
alcohol consumption (50 g/d for men and 30 g/d for anisms of liver injury, the APRI score does not seem to
women), the absence of cirrhosis, and a small sample have diagnostic value in patients with AIH.
size. Alcohol consumption might have confounded our
association between APRI and stage of fibrosis given
that, itself, affects both AST and platelet count. Accord- References
ing to current recommendations, restriction of the sample 1. Friedman SL. Liver fibrosis - From bench to bedside. J Hepatol
to subjects with lower cut-off values of alcohol con- 2003; 38: S38-S53.
sumption, such as 30 g/d for men and 20 g/d for women 2. Schuppan D, Ruehl M, Somasundaram R, Hahn EG. Matrix as a
would, probably, control for this confounder. Last, but modulator of hepatic fibrogenesis. Semin Liver Dis 2001; 21:
351-72.
not least, METAVIR might not have been the most ap- 3. Fattovich G, Giustina G, Degos F, Tremolada F, Diodati G, Almasio
propriate staging score of fibrosis in this group, potential- P, Nevens F, et al. Morbidity and mortality in compensated cir-
ly leading to misclassification. The use of METAVIR in rhosis type C: a retrospective follow-up study of 384 patients.
all our study groups was a methodologically based deci- Gastroenterology 1997; 112: 463-72.
4. Manning DS, Afdhal NH. Diagnosis and quantitation of fibrosis.
sion, given that homogeneous criteria would allow more Gastroenterology 2008; 134: 1670-81.
valid comparisons among groups. Results are, however, 5. Regev A, Berho M, Jeffers LJ, Milikowski C, Molina EG,
encouraging enough to warrant further validation of the Pyrsopoulos NT, Feng ZZ, et al. Sampling error and intraobserver
APRI score in NAFLD, where all the above sources of variation in liver biopsy in patients with chronic HCV infection.
Am J Gastroenterol 2002; 97: 2614-8.
bias are overcome. 6. Castera L, Negre I, Samii K, Buffet C. Pain experienced during
As to AIH, there was a lack of association among percutaneous liver biopsy. Hepatology 1999; 30: 1529-30.
APRI scores and degree of fibrosis. In this disease, portal, 7. Lichtenstein DR, Kim D, Chopra S. Delayed massive hemobilia
periportal and lobular inflammation is notorious in all following percutaneous liver biopsy: treatment by
embolotherapy. Am J Gastroenterol 1992; 87: 1833-8.
stages of fibrosis.28 This, as observed in our study, can 8. Colli A, Fraquelli M, Andreoletti M, Marino B, Zuccoli E, Conte
derive in non-differentially high AST levels and, in turn, D. Severe liver fibrosis or cirrhosis: accuracy of US for detection
high APRI values irrespective of fibrosis stage. Besides, –analysis of 300 cases. Radiology 2003; 227: 89-94.
 A Loaeza-del-Castillo et al. AST to platelet ratio index (APRI) for the noninvasive evaluation of liver fibrosis 357

9. Sandrin L, Fourquet B, Hasquenoph JM, Yon S, Fournier C, Mal the pathogenesis of thrombocytopenia in patients with chronic
F, Christidis C, et al. staging Transient elastography: a new viral hepatitis. Br J Haematol 2001; 113: 590-5.
noninvasive method for assessment of hepatic fibrosis. Ultra- 20. Aster R. Pooling of platelets in the spleen: role in the pathogenesis of
sound Med Biol 2003; 29: 1705-13. hypersplenic thrombocytopenia. J Clin Invest 1996; 45: 645-57.
10. Saito H, Tada S, Nakamoto N, Kitamura K, Horikawa H, Hurita 21. Nalpas B, Vassault A, Le Guillou A, Lesgourgues B, Ferry N,
S, Saito Y, et al. Efficacy of non-invasive elastometry on of Lacour B, Berthelot P. Serum activity of mitochondrial aspartate
hepatic fibrosis. Hepatol Res 2004; 29: 97-103. aminotransferase: a sensitive marker of alcoholism with or with-
11. Imbert-Bismut F, Ratziu V, Pieroni L, Charlotte F, Benhamou Y, out alcoholic hepatitis. Hepatology 1984; 4: 893-6.
Poynard T, MULTIVIRC Group. Biochemical markers of liver 22. Kamimoto Y, Horiuchi S, Tanase S, Morino Y. Plasma clearance
fibrosis in patients with hepatitis C virus infection: a prospective of intravenously injected aspartate aminotransferase isozymes:
study. Lancet 2001; 357: 1069-75. evidence for preferential uptake by sinusoidal liver cells.
12. Wai CT, Greenson JL, Fontana RJ, Kalbfleisch JD, Marrero JA, Hepatology 1985; 5: 367-75.
Conjeevaram HS, Lok AS. A simple noninvasive index can pre- 23. Davies HT, Crombie IK, Tavakoli M. When can odds ratios mis-
dict both significant fibrosis and cirrhosis in patients with chronic lead? Br Med J 1998; 316: 989-91.
hepatitis C. Hepatology 2003; 38: 518-26. 24. Deeks JJ. Systematic reviews in health care systematic reviews of
13. Expert Panel on Detection, Evaluation, and Treatment of High evaluations of diagnostic and screening tests. Br Med J 2001;
Blood Cholesterol in Adults (Adult Treatment Panel III). Execu- 323: 157-62.
tive summary of the third report of the national cholesterol edu- 25. Carvalho-Filho RJ, Schiavon LL, Narciso-Schiavon JL, et al.
cation program (NCEP). J Am Med Assoc 2001; 285: 2486-97. Optimized cutoffs improve performance of the aspartate ami-
14. Bedossa P, Poynard T. An algorithm for the grading of activity in notransferase to platelet ratio index for predicting significant
chronic hepatitis C. The METAVIR Cooperative Study Group. liver fibrosis in human immunodeficiency virus/hepatitis C virus
Hepatology 1996; 24: 289-93. co-infection. Liver Int 2008; 28: 486-93.
15. Cuzick J. A Wilcoxon-type test for trend. Stat Med 1985; 4: 87-90. 26. Castera L, Vergniol J, Foucher J, Le Bail B, Chanteloup E, Haaser
16. Shaheen AA, Myers RP. Diagnostic accuracy of the aspartate M, Darriet M, et al. Prospective comparison of transient
aminotransferase-to-platelet ratio index for the prediction of hepa- elastography, Fibrotest, APRI, and liver biopsy for the assess-
titis C-related fibrosis: a systematic review. Hepatology 2007; 46: ment of fibrosis in chronic hepatitis C. Gastroenterology 2005;
912-21. 128: 343-50.
17. Kim BK, Kim SA, Park YN, Cheong JY, Kim HS, Park JY, Cho 27. Angulo P, Keach JC, Batts KP, Lindor KD. Independent predic-
SW, et al. Noninvasive models to predict liver cirrhosis in pa- tors of liver fibrosis in patients with nonalcoholic steatohepatitis.
tients with chronic hepatitis B. Liver Int 2007; 27: 969-76. Hepatology 1999; 30: 1356-62.
18. Kawasaki T, Takeshita A, Souda K, Kobayashi Y, Kikuyama M, 28. Nakaji M, Hayashi Y, Ninomiya T, et al. Histological grading
Suzuki F, Kageyama F, et al. Serum thrombopoietin levels in and staging in chronic hepatitis: its practical correlation. Pathol
patients with chronic hepatitis and liver cirrhosis. Am J Int 2002; 52: 683-90.
Gastroenterol 1999; 94: 1918-22. 29. Lieber CS, Weiss DG, Morgan TR, Paronetto F. Aspartate Ami-
19. Adinolfi LE, Giordano MG, Andreana A, Tripodi MF, Utili R, notransferase to platelet ratio index in patients with alcoholic
Cesaro G, Ragone E, et al. Hepatic fibrosis plays a central role in liver fibrosis. Am J Gastroenterol 2006; 101: 1500-8.