Noninvasive predictors of large esophageal varices: is there an emerging role of aspartate aminotransferase-to-platelet ratio index in hepatocellular carcinoma?

Background and aim Variceal size has been identified to be closely related to variceal bleeding. Repeated endoscopic examinations have a great burden on endoscopic units and cost-implication issues. Our aim was to evaluate the role of AST to platelet ratio index (APRI) in predicting the existence of large esophageal varices (EV) in hepatitis C virus-related liver cirrhotic patients. Patients and methods Seventy four patients with liver cirrhosis were prospectively recruited. Laboratory data, CTP, MELD and APRI, also ultrasonographic and endoscopic findings are performed and investigated whether associated with the size and bleeding of EV. Results Patients were divided into two groups; group 1 with small varices and group 2 with large varices. Group 2 had significantly prolonged prothrombin time, splenomegaly, ascites, higher Child score compared to group 1. CTP was associated with variceal bleeding (P = 0.028). While APRI was a poor predictor both for the presence of LVs and bleeding yet it revealed favorable results with bleeding EVs in patients with HCC with AUC (0.61). APRI was a good predictor for the presence of HCC and number of focal lesions with AURC (0.651, 0.61 respectively). Conclusion Splenomegaly, CTP, ascites could be used as noninvasive predictors for large EVs. However, at the moment, these tests could not substitute for endoscopy. Although APRI is a poor predictor for the size and bleeding of EV, yet it might have a role in prediction of HCC and number of focal lesions.

As the aspartate aminotransferase-to-platelet ratio index (APRI) is a predictor of fi brosis, it is reasonable to explore whether it could be considered a noninvasive marker for detecting EV [9].
Th e aim of this study was to determine the ability of APRI in predicting the existence of LEVs in hepatitis C virus (HCV)-related liver cirrhotic patients.

Patients and methods
Th is study was a cross-sectional comparative trial that was conducted at the Internal Medicine Department of Kasr Al-Ainy Hospital from August 2012 to July 2013.
It was approved by the ethical committee of the Faculty of Medicine, Cairo University. Patients were fully informed about possible complications of the diagnostic procedures, and written informed consent was obtained from the patients or from a responsible family member. Seventy-four patients were included in this prospective study. Th e included patients were above 18 years of age with HCV-related liver cirrhosis based on clinical, biochemical, and ultrasonographic fi ndings.

Exclusion criteria
Patients with previous variceal hemorrhage, prior variceal bleeding prophylaxis or treatment, (including nonselective β-blocker use, nitrates, endoscopic variceal ligation, or sclerotherapy), and those with gastrointestinal (GI) ulcerations were excluded. None Noninvasive predictors of large esophageal varices: is there an emerging role of aspartate aminotransferase-to-platelet ratio index in hepatocellular carcinoma?

Background and aim
Variceal size has been identifi ed to be closely related to variceal bleeding. Repeated endoscopic examinations have a great burden on endoscopic units and cost-implication issues. Our aim was to evaluate the role of AST to platelet ratio index (APRI) in predicting the existence of large esophageal varices (EV) in hepatitis C virus-related liver cirrhotic patients.
of the patients were treated with NSAIDs, antiplatelets, or anticoagulants. Patients with liver cirrhosis due to causes other that HCV and those with renal failure, portal or splenic vein thrombosis, or any associated malignancies other than hepatocellular carcinoma (HCC) were excluded as well.
Physical and clinical characteristics recorded included age, sex, and symptoms and signs of liver cell failure, with special emphasis on presentation with fi rst attack of GI bleeding. Laboratory workup was carried out, which included measurement of hemoglobin, platelet count, alanine aminotransferase and aspartate aminotransferase (AST), prothrombin time, and international normalized ratio (INR), serum bilirubin, albumin, creatinine, and α-fetoprotein. Severity of liver disease was assessed by the Child-Pugh-Turcotte classifi cation (CPT) [10]. Model for end-stage liver disease (MELD) [11] was reviewed.
Ultrasonographic characteristics were recorded, especially splenic size and degree of ascites, which was graded as none, mild (detectable only on ultrasound), moderate (visible as a moderate symmetrical abdominal distension), or severe (marked abdominal distension). For HCC, the size and number of the hepatic focal lesions were determined.
Upper GI endoscopy was performed for all patients using a videoscope (Olympus 240, Tokyo, Japan) to detect the presence and grading of EV. All endoscopies were performed in a single endoscopy unit by an experienced endoscopist. Accordingly, EVs were classifi ed as small (veins minimally elevated above the esophageal mucosal surface), medium (tortuous veins occupying less than one-third of the esophageal lumen), and large (those occupying more than one-third of the esophageal lumen) [12]. Presence of fundal varices and portal hypertensive gastropathy were also recorded. Control of bleeding was done either with injection sclerotherapy or with band ligation. In this study, patients were divided into two groups according to the size of EVs: group 1 included patients with no or small EVs and group 2 included patients with medium-sized or large-sized varices and an APRI score: Th e APRI score was calculated as follows [9]: (AST /ULN)/platelet count (10 9 /l)] ×100, in which ULN is the upper limit of normal.

Statistical methodology
Data were analyzed on a n IBM computer usin g SPSS (Statistical Program for Social Science, version 12) software. Quantitative variables were presented as mean and SD and qualitative variables as number and percentag e. Th e χ 2 -test was performed to compare qualitative variables between groups. Th e unpaired t-test was used to compare quantitative variables in parametric data (SD <50% mean). One-wa y analysis of variance was calculated to compare more than two groups as regards quantitative variables. Spearman's correlation coeffi cient was used to rank variables versus each other positively or inversely . Th e receiver operating characteristic (ROC) curve was generated to fi nd the bes t cutoff value, and the validity of certain variables (P > 0.05, insignifi cant; P < 0.05, signifi cant; P < 0.01, highly signifi cant).

Comparison between group 1 and group 2
According to the result of the upper GI endoscopy, patients were classifi ed into two groups according to the size of EVs: group 1 included 32 patients with small varices (SVs) and group 2 included 42 patients with large varices. Table 2 shows a comparison between the two groups as regards the studied variables.
LVs were associated with elevated INR and splenomegaly. Patients with LVs were associated with elevated INR. Distribution of CPT was evaluated. Frequency of Child A score was significantly higher in group 1 compared with group 2 (40.6 vs. 7.1%, respectively). Similarly, there was a significantly lower frequency of Child B and C scores in group 1 (28.1 and 31.3%, respectively) compared with group 2 (42.9 and 50.0%, respectively) (P = 0.002).
When the ROC curve for CPT and MELD sores was determined, it showed that CPT was superior to MELD in the prediction of large varice s. Area under the curve (AUC) was 0.663 for CPT (P = 0.017) compar ed with 0.615 (P = 0.092) for MELD score. Th e best threshold for CPT was above 7.5. Sensitivity was 73.8% and specifi city was 50% (Fig. 1).
To demonstrate the association between diff erent studied parameters including laboratory, imaging, and endoscopic fi ndings and their relation to both bleeding EV and the presence of HCC, the whole population was reclassifi ed according to the presence of both bleeding EV and HCC.

Predictors of upp er gastrointestinal tract bleeding
Th e whole population was reclassifi ed according to the presence of gastrointestinal tract (GIT) bleeding into two groups: group A, which presented with GIT bleeding, and group B, which presented without bleeding (Table 3). Comparison between group A and group B in relation to laboratory, imaging, and endoscopic fi ndings was performed. Th is analysis revealed that 62/74 patients (83.7%) had splenomegaly, 29/62 patients (80.6%) had no bleeding, and 33/62 (86.8%) presented with variceal bleeding (P = 0.538). However, the probability of GIT bleeding was 1.5 times among those with splenomegaly [odds ratio 1.59; 95% confi dence interval (CI) 0.456-5.568] (results not shown). CPT score was higher with variceal bleeding, and at a threshold greater than 7.5 points it was signifi cantly associated with LVs (P = 0.017), with a sensitivity of 73.8% and a specifi city of 50% (Fig. 1).   Association between aspartate aminotransferase-toplatelet ratio index and variceal size and bleeding APRI had no predictive ability for LEVs or bleeding (Tables 1 and 2). Table 5 summarizes the diff e rent cutoff values of APRI to assess its predictive value in LEVs. When the ROC curve was determined, at a cutoff of 1.4, APRI was not associated with variceal size (AUC = 0.540; 95% CI 0.401-0.679) (Fig. 2). Figure 3 illustrates that APRI was a poor predictor of bleeding EV with AUC 0.51, sensitivity 53%, specifi city 46%, positive predictive value (PPV) 55%, and negative predictive value (NPV) 50%. Above a cutoff value of 1.2, sensitivity was 75%, specifi city was 52%, PPV was 53% , and NPV was 57%. On performing logistic regression analysis, APRI was insuffi cient in providing diagnostic accuracy for EV bleeding (P = 0.248).

Correlation between aspartate aminotransferase-toplatelet ratio index and the studied variables
An unexpected fi nding is that APRI was a good predictor of bleeding EV in patients with HCC only: AUC 0.55, sensitivity 70%, specifi city 51%, PPV 49% , and NPV 64%, at a cutoff value of 1.5.
Th is was further investigated at diff erent cutoff values of APRI. APRI was a good positive marker for prediction of HCC with AUC 0.65. Table 6 and Fig. 4 showed a favorable sensitivity and specifi city of APRI in predicting the presence of HCC. At a lower cutoff value sensitivity was high, whereas specifi city was low (51.1%).

Figure 2
Receiver operating characteristic curves of APRI for prediction of large EVs (AUC = 0.540, 95% CI 0.401-0.679). APRI, aspartate aminotransferase-to-platelet ratio index; AUC, area under the curve; CI, confi dence interval; EV, esophageal varices. AFP, α-fetoprotein; APRI, aspartate aminotransferase-to-platelet ratio index; INR, international normalized ratio; MELD, model for end-stage liver disease; PLT, platelet; Bold denotes signifi cant (P less than 0.05) and highly signifi cant value (P less than 0.01). Validity of APRI in predicting the number of focal lesions was studied. When the ROC curve was ascertained we found that an APRI cutoff of 1.4 was better positive than negative in the prediction of more than three focal lesions if the actual value was greater than the cutoff value and of less than three lesions if the value was less than the cutoff value, at an AUC of 0.61 (Table 6).

Discussion
Th e mortality rate from variceal bleeding is about 20% when patients are treated optimally in hospital [1]. Prevention of bleeding from EV is crucial and remains at the forefront of long-term management of cirrhotic patients. Th e standard diagnostic screening tool for EV is endoscopy [3]; however, endoscopy is invasive and costly. Egypt as a developing country with limited resources and a high incidence of liver cirrhosis mainly due to CHC infection has to limit the frequency of unnecessary endoscopic examinations. Whereas there is ample evidence indicating the value of noninvasive variables for the presence or absence of EV [4][5][6][7][8][13][14][15][16], there are few data to predict the presence of large varices in decompensated patients [7,17].
In this study, we considered simple, commonly available, reproducible, and inexpensive laboratory and imaging parameters as predictors of variceal size and bleeding.
In order not to miss the presence of EV, which can be extremely hazardous to the health of patients, a good NPV of these studied variables is needed.
We included 74 patients with liver cirrhosis due to CHC who were divided into two groups according to the size of varices: group 1 with SVs and group 2 with LVs. Th e reason for including CHC only is that APRI was fi rst prescribed to predict fi brosis among patients with CHC [9], and thus it could be possible that it would have a better performance in this population. Both groups were comparable with respect to clinical and laboratory fi ndings; however, upper GI bleeding showed a higher frequency in patients with LVs compared with those with SVs, with a statistically signifi cant diff erence (P = 0.018). LVs were associated with elevated INR, higher CPT score, and splenomegaly. Patients with LVs were associated with elevated INR, higher CPT score, splenomegaly, and ascites. Also, upper GI bleeding was more frequent in patients with LVs (P = 0.018).
No association between platelet count and size of varices was detected in our patients, which was similar to some published data [13,18]. Th is could be partly attributed to the matched platelet count in the two studied groups. Moreover, it is worth noting that portal Receiver operating characteristic curves of APRI for prediction of bleeding EVs. APRI, aspartate aminotransferase-to-platelet ratio index; EV, esophageal varices.  Receiver operating characteristic curves of APRI for prediction of HCC (AUC = 0.651, P = 0.029). AUC, area under the curve; APRI, aspartate aminotransferase-to-platelet ratio index; HCC, hepatocellular ca rcinoma.
hypertension is not the sole factor responsible for thrombocytopenia. Furthermore, diff erent mechanisms have been implicated in thrombocytopenia, including shortened mean life span, thrombopoietin defi ciency, myelotoxic eff ect, and bone marrow suppression [17].
Although previous studies have found that platelet count below 100 000 nearly always predicts the presence of LVs [3,6], there is still no consensus as to the best cutoff point for platelets for predicting EV [19,20].
Prothrombin time and serum albumin are considered markers of hepatocellular dysfunction. Patients with LVs had higher INR compared with patients with SVs.
Although serum albumin was lower in patients with LVs than in those with SVs, no statistical diff erence was found (P = 0.058). However, it was a good marker for variceal bleeding (P = 0.031), which indicated that these biomarkers, consistent with previous studies [14], could have a good diagnostic performance. On the other side, other investigators showed discordant results [15,18].
Decompensated liver cirrhosis (Child B and C) is deemed to be the appropriate parameter for predicting variceal bleeding [12,18]. Our results demonstrated that CPT score was higher with variceal bleeding, and at a threshold more than 7.5 points it was signifi cantly associated with LVs (P = 0.017). Our data were in accordance with previous results [21][22][23][24].
It is notable that inclusion of prothrombin time and albumin within the CPT score gave a favorable association between Child score and these parameters. In contrast, CPT might be a poor predictor of EVs because of the subjectivity of its clinical parameters and limited discriminant ability. In accordance with previous studies [7,25], we detected no correlation between MELD score and the size and bleeding of EV (P = 0.092 and 0.265, respectively).
Evaluation of diff erent sonographic fi ndings revealed that enlarged spleen and presence of ascites were able to discriminate the variceal size. Th e predictive value of splenomegaly in variceal bleeding was poor. However, the detailed analysis of our data demonstrated that the probability of GIT bleeding was 1.5 times among those with splenomegaly (odds ratio 1.59; 95% CI 0.456-5.568).
Our results were in general agreement with these studies [7,17,18,25]. Nevertheless, discordant results were reported [14,26]. A possible explanation for these variable results might be the diff erent etiologies of liver cirrhosis, as splenomegaly is more frequently found in posthepatitic cirrhosis than in alcoholic cirrhosis [27].
Th e divergence of the results between studies could be attributed to the fact that both endoscopy and ultrasonography are operator-dependent techniques with a lack of interobserver agreement [19].
Increasing size of varices is associated with an increase in variceal wall tension to a critical level, at which varices rupture an d cause life-threatening bleeding.
SV progress to LV at a rate of 5-10% per year [16]. Th e current results revealed a strong association between variceal size and bleeding. We found that 71.1% of patients presenting with bleeding EV had LEVs, whereas 58.3% of those with no bleeding had SEVs (P = 0.018). Th is was supported by the North Italian Endoscopic Club [12].
Wai et al. [9] validated for the fi rst time an index known as the APRI, which establishes the relationship between this score and liver fi brosis.
APRI has good accuracy in predicting fi brosis, the major cause of portal hypertension. Th is index is feasible and simple as it uses two easily obtained parameters; thus, it can be applied in every cirrhotic patient without a cost burden. An earlier study [28] reported the association of APRI with the presence of EV; however, they included compensated cirrhotic patients. A possible explanation for that association may be that fi brosis indicates more severe hepatic parenchyma architectural distortion and increased intrahepatic circulatory resistance, resulting in portal hypertension, variceal formation, and fi nally EV bleeding [22].
In accordance with a good performance of the liver stiff ness measurement in assessing both liver fi brosis and EVs [29], further studies may be necessary to evaluate the potential utility of serum liver fi brosis markers. However, it remains unclear whether serum liver fi brosis markers can be useful for screening the presence of EVs [29][30][31].
Because of the aforementioned reasons, we propose here that APRI may be a potential valuable marker for predicting the presence of LEVs and their risk of bleeding. Th ere are convincing data showing that patients with cirrhosis and HCC are more likely to suff er from variceal hemorrhage [7].
Our data revealed that APRI was insuffi cient in providing diagnostic accuracy for discrimination between SEVs and LEVs even at diff erent cutoff values. When the ROC curve was determined, we found that at a cutoff of 1.4 APRI was not associated with variceal size [AUC = 0.540 (P = 0.559), 95% CI 0.401-0.679]. Further, the predictive ability of APRI was poor regarding bleeding EV, with an AUC of 0.51.
Diff erent cutoff values for APRI as a predictor of EV were tested in previous research, which showed similar results to ours [31][32][33].
HCC develops in a multistage process involving chronic liver injury and local infl ammation, progressive liver fi brosis and cirrhosis, initiation of neoplastic niches, and malignant transformation. It is well known that advanced fi brosis and cirrhosis are high-risk conditions that prompt intensive surveillance for HCC. It could be hypothesized that APRI might have a role in predicting the progression of advanced fi brosis and cirrhosis into HCC. To investigate this hypothesis, we investigated the association of APRI with HCC at diff erent cutoff values. Higher APRI levels were detected in patients with HCC with statistically signifi cant diff erence compared with patients without HCC (P = 0.029). At cutoff more than 1.39, APRI was found to be a good positive marker for predicting the presence of HCC. On applying the ROC curve at the same cutoff point, AUC was 0.65. Th ere was signifi cant association between APRI and number of focal lesions, with AUC 0.61. Moreover, signifi cant correlation was found between APRI and α-fetoprotein level.
APRI has been evaluated as a prognostic biomarker in HCC patients secondary to hepatitis B virus after radiofrequency ablation [34]. A recent publication by Hann et al. [35] concluded that APRI might be a marker o f HCC in hepatitis B virus patients. Despite the inconsistent results, these studies substantiated the potential usefulness of APRI in the evaluation of liver diseases.
To the best of our knowledge, our study is the fi rst to investigate the role of APRI in the prediction of HCV-related HCC, especially advanced cases, as indicated by the presence of EV and GIT bleeding. Extensive fi brosis indicates more severe hepatic parenchyma architectural distortion and increased intrahepatic circulatory resistance, resulting in portal hypertension, variceal formation, and fi nally EV bleeding [32]. Th ere are convincing data to show that patients with cirrhosis and HCC are more likely to suff er from variceal hemorrhage [7]. Th e current results showed that APRI was a good predictor for bleeding EVs in a cohort of patients with HCC with good discriminative value (AUC = 0.55). Th ese data point at the role of APRI in advanced cirrhosis that has progressed to HCC. Additional studies should be conducted to confi rm whether or not the condition is cirrhosis dependent.

Conclusion
Although we found an association between LEV and CPT, splenomegaly, and ascites, at present these parameters cannot be advocated as a surrogate for endoscopy. In addition, APRI cannot be used as a predictor of LEV.
However, our study highlighted the proposed role of APRI in HCC, its association with the size and number of focal lesions, and its predictive role in variceal bleeding in this cohort of patients. Yet, further studies are warranted to validate the clinical applicability of APRI to predict HCC and its clinicopathological parameters.

Financial support and sponsorship
Nil.

Confl icts of interest
Th ere are no confl icts of interest.