To assess “predictors” of esophageal varices (EV) and variceal bleeding using non-invasive markers in Albanian patients diagnosed with liver cirrhosis.

One hundred thirty-nine newly diagnosed cirrhotic patients without variceal bleeding were included in this analysis. Model for end-stage liver disease (MELD), aspartate aminotransferase (AST) to alanine aminotransferase (ALT) ratio (AST/ALT), AST to platelet ratio index (APRI), platelet count to spleen diameter (PC/SD), fibrosis-4-index (FIB-4), fibrosis index (FI) and King’s Score were measured for all participants. All patients underwent endoscopic assessment within two days of hospitalization. The major end point was the first esophageal variceal bleeding (EVB) event. The diagnostic performance of “predictors” for the presence of EV and EVB were assessed by sensitivity and specificity values obtained from the receiver operating characteristics procedure.

FIB-4 was the only strong and significant “predictor” of esophageal varices (multivariable-adjusted OR = 1.57 for one unit increment; 95%CI: 1.15-2.14). Furthermore, a cut-off value of 3.23 for FIB-4 was a significant predictor of esophageal varices, with a sensitivity of 72%, a specificity of 58% and a proportion of area under the curve (AUC) of 66% (P = 0.01). During the follow-up (median: 31.5 mo; interquartile range: 11-59 mo), 34 patients (24%) experienced a first EVB. FIB-4 was a poor predictor of EVB (the AUC was only 51%) for a cut-off value of 5.02. Furthermore, the AUC of AST/ALT, APRI, PC/SD, FI, MELD and King's Score ranged from 45% to 55%. None of the non-invasive markers turned out to be a useful predictor of EVB.

Despite the low diagnostic accuracy, FIB-4 appears the most efficient non-invasive liver fibrosis marker which can be used as an initial screening tool for cirrhotic patients.

Liver cirrhosis is the 13^th leading cause of death globally, with increasing mortality rate worldwide[1]. Portal hypertension is a frequent consequence in the progression of liver cirrhosis and plays a crucial role in the clinical manifestations of disease[2,3]. One of the most serious complications of portal hypertension is the development of esophageal varices (EV) caused by increased hepatic vascular resistance related to hepatic fibrosis and regenerative nodules[3]. In addition, the variceal bleeding due to the rupture of varices is still the most common lethal complication of cirrhosis[4]. Therefore, assessing the presence of esophageal varices in cirrhotic patients is clinically important in prevention of their bleeding. To date, the upper gastrointestinal endoscopy remains the golden diagnostic methods for EV and the recent Baveno VI Meeting Consensus recommends the endoscopy screening for all cirrhotic patients at the time of their diagnosis and periodical endoscopy examination in patients with EV[4]. However, routine endoscopy screening may not be cost-effective, as less than 50% of all patients with cirrhosis have EV[5]. Furthermore, there is a low prevalence of varices which requires primary prophylaxis[5]. Also, the upper endoscopy is an invasive and uncomfortable procedure which may not be acceptable for the patients. Hence, predicting the presence of EV through non-endoscopic and non-invasive markers is important in order to identify the patients who benefit from routine endoscopy screening and may reduce considerably the number of avoidable endoscopies[6].

Recently, various non-invasive markers, such as model for end-stage liver disease (MELD), aspartate aminotransferase (AST) to alanine aminotransferase (ALT) ratio (AST/ALT), AST to platelet ratio index (APRI), platelet count to spleen diameter (PC/SD), fibrosis-4-index (FIB-4), fibrosis index (FI) and King’s score, have been demonstrated as a simple, non-invasive and easier practical alternative to predict the presence of EV in cirrhotic patients[7-14]. However, the findings of these previous studies are controversial and their utility in clinical practice is uncertain. The conclusions of these studies vary in different populations and in different etiologies of liver cirrhosis[13-16]. In this context, our prospective study aimed to assess the non-invasive markers, composed on routine laboratory parameters that could predict the presence of EV in liver cirrhosis patients in Albania, an endemic area of hepatitis B virus infection in Southeastern Europe and a Mediterranean country with a high consumption of domestic alcoholic beverages[17,18].

Despite the large number of studies, to date, no single or combination of non-invasive markers has been widely evaluated for predicting the first variceal bleeding in patients with cirrhosis[19,20]. Therefore, the second objective of our study was to identify the cirrhotic patients with a high risk for variceal bleeding using non-invasive markers.

Table 1 presents the distribution of baseline characteristics in patients with (n = 113) and without (n = 26) esophageal varices. Among patients with esophageal varices, 19 (17%) presented red signs compared with 94 (83%) patients without red signs. There was no sex-difference in the prevalence of esophageal varices between the two groups (upper panel). An alcoholic and/or viral etiology of cirrhosis was significantly more prevalent among patients with esophageal varices compared with their counterparts without esophageal varices. Conversely, grade A of CTP was significantly more prevalent among patients without varices than those with esophageal varices (50% vs 27%, respectively).

Mean age was not significantly different between patients with and without esophageal varices (Table 1 - lower panel). Likewise, there were no group differences regarding mean levels of hemoglobin, creatinine, AST and ALT, GGT and gamma globulin. Conversely, mean value of INR was significantly higher among patients with varices than in those without varices (1.7 vs 1.4, respectively; P < 0.001). Similarly, mean levels of most of the non-invasive markers were significantly higher in patients with varices than in those without variances (FIB-4: 6.3 vs 4.0, respectively; fibrosis index: 3.2 vs 2.8, respectively; King’s score: 87 vs 67, respectively; APRI: 2.7 vs 2.5, respectively; and MELD index: 16.1 vs 13.6, respectively). On the other hand, mean level of PL/SD was lower in patient with varices than in those without varices (1089 vs 1395, respectively; P = 0.03) (Table 1 - lower panel).

In crude (unadjusted) binary logistic regression models (Table 2 - left panel), there was a significant association between esophageal varices and FIB 4 (OR = 1.18 for one unit increment; 95%CI: 1.01-1.38), and MELD score (OR = 1.11 for one unit increment; 95%CI: 1.01-1.22). On the other hand, there was no evidence of any significant association of esophageal varices with the other non-invasive markers (fibrosis index, King’s score, APRI, PL/SD and AST/ALT). Upon adjustment for demographic characteristics, clinical parameters and all the biomarkers in a backward stepwise elimination procedure (right panel), FIB 4 was the only strong and significant “predictor” of esophageal varices (OR = 1.57 for one unit increment; 95%CI: 1.15-2.14). Furthermore, a cut-off value of 3.23 for FIB 4 was a significant predictor of esophageal varices (Table 3), with a sensitivity of 72%, a specificity of 58% and a proportion of area under the curve of 66% (P = 0.01).

During follow-up (median: 31.5 mo; interquartile range: 11-59 mo), 34 patients (24%) experienced first esophageal variceal bleeding. Of these, 4 (11.8%) patients were without EV upon enrollment, but experienced EV in the course of the follow-up. There was no evidence of any significant differences in the distribution of baseline characteristics in patients with (n = 34) and those without (n = 105) hemorrhagic events (Table 4).

The performance of the non-invasive markers for prediction of esophageal variceal bleeding is presented in Table 5, as an additional objective of our analysis was to identify the cirrhotic patients with high risk for variceal bleeding employing these non-invasive markers. Yet, none of the non-invasive markers turned out to be a useful predictor of esophageal variceal bleeding in this sample of Albanian patients. The area under the curve was pretty close to 50% for all of the non-invasive markers. For PL/SD there was evidence of a poorer prediction than even by chance (the area under the curve was 45%). FIB-4, which was shown a powerful predictor of esophageal varices, was nevertheless a very poor predictor of esophageal variceal bleeding (the area under the curve was only 51%) for a cut-off value of 5.02. The analysis was also restricted to individuals with variceal prophylactic therapy (VPT) (n = 83); similarly though, none of the non-invasive markers appeared to be a useful predictor of esophageal variceal bleeding (Table 6).

Nowadays, clinicians are interested to identify some “ideal” non-invasive biochemical markers which may be cheaper and easy to obtain, but with a high sensitivity and specificity for reducing the number of upper endoscopy needed for screening and management of EV in patients with liver cirrhosis. Such non-invasive tools are especially needed in developing countries with shortage of endoscopy, limited resources and a strong appeal for rationalization of funding. Therefore, we tried to explore whether any of the non-invasive markers (including AST/ALT, APRI, PC/SD, FIB-4, FI and King’s Score) could predict the presence of EV in cirrhotic patients in Albanian, a transitional Mediterranean country. We found that among all clinical and biochemical features assessed in univariate analyses, FIB-4 and MELD score significantly correlated with the presence of EV. However, in multivariable-adjusted logistic regression models, only FIB-4 remained a significant independent predictor of EV. It is interesting that only FIB-4, a fibrosis marker based on parameters linked to liver dysfunction or advanced diseases (AST and ALT), linked to portal hypertension (platelet count) and age could predict the EV despite the non-invasive markers that we included in our analysis. One plausible explanation for the lack of prediction of the other markers may consist of the age of the patients included in our study. Considering that liver fibrosis is a dynamic process and there is a close relationship between liver fibrosis, portal hypertension and development of EV, older patients are more likely to have EV. Indeed, mean age of our patients was relatively high and most of the patients were admitted in an advanced liver failure stage. Furthermore, previous studies investigating FIB-4 as a predictor of EV in liver cirrhosis patients showed similar results. In another prospective studies, for a cut-off value of 3.98 and 2.8, the AUC for predicting the EV was 62% and 78%, respectively[22,23], whereas Sebastiani et al[10] showed an AUC of 64% for the prediction of EV at a cut-off value of 3.5 in a retrospective study. However, their sensitivities and specificities were relatively low, ranging from 66% to 76% and from 54% to 80%, respectively. Similarly, a recent meta-analysis demonstrated that AUC of FIB-4 for prediction the presence of EV was 77%[24]. In our study, a cut-off value of 3.23 of FIB-4 was proposed for diagnosis of esophageal varices. At this cut-off, the sensitivity was 72%, specificity was 58%, PPV was 88%, NPV was 32% and the AUC was 66%.

Esophageal variceal bleeding is the most important complication of liver cirrhosis and the major cause of death in cirrhotic patients. Therefore, prevention of esophageal variceal bleeding by using prophylactic treatment is an important goal to be achieved in cirrhotic patients with high risk esophageal varices. In our study, the prophylactic treatment with non-selective beta blockers was introduced in almost 60% of cirrhotic patients based on endoscopic criteria and liver dysfunction. Nonetheless, during the five years of follow-up, 11 patients (20%) among 56 cases without VPT and 23 patients (28%) among 83 cases who underwent VPT experienced upper gastrointestinal hemorrhage from variceal bleeding confirmed by endoscopy. Thus, these patients may require other prophylactic treatments such as band ligations of esophageal varices. Furthermore, identifying the non-invasive markers to predict EVB during the follow-up may be an important tool for a better management of cirrhotic patients. From this point of view, we conducted the present study under the assumption that progression of EV and ultimately EVB is caused directly by portal hypertension which correlates with liver fibrosis and may be assessed by non-invasive markers. Our study demonstrated insufficient accuracy of AST/ALT, APRI, PC/SD, FIB-4, FI and King’s score for the prediction of EVB which ranged from 0.45 to 0.55. These findings suggest that none of the non-invasive markers is a useful predictor of esophageal variceal bleeding in this sample of Albanian patients.

To our knowledge, this is the first study exploring the role of non-invasive biomarkers as predictors of EVB in a longitudinal study. These non-invasive biomarkers are based on regular laboratory data, require no extra cost, specialized devices or additional biochemical tests and differ from other noninvasive markers which may not be easily accessible[25]. This fact may be very important in developing and transitional countries with rather limited resources. However, only FIB-4 turned out to be a useful predictor in this sample of Albanian patients and, therefore, the usefulness and applicability of these noninvasive markers should be considered cautiously. Furthermore, our data were collected at a single center in which the standard of care did not change substantially during the period under study. Also, the study participants were representative of the population of cirrhotic patients with different etiologies of liver cirrhosis avoiding selection bias pertinent to different etiologies or subgroups. In addition, the relatively large number of participants and the reasonable follow-up period constitute other strength of this study.

Multivariate modeling is frequently used to predict the risk or prognosis of diseases and treatment’s

response. In our study, the multivariate model allowed us to assess important variables most of which, nevertheless, did not predict the presence of EV and the risk of first variceal bleeding in cirrhotic patients. However, multi-center prospective studies are needed to confirm our findings and ensure that they are applicable to diverse populations with different etiologies of liver cirrhosis. In addition, we cannot exclude the possibility that other predictors could induce hematological changes such as antiviral treatment or continued consumption of alcohol during the follow-up period - factors which were not accounted for in our analysis. Also, assessment of the sensitivity and specificity upon a second-time measurement would have provided an additional insight into the predictive power of the non-invasive parameters included in our analysis.

In conclusion, our results, based on cirrhotic patients with different etiologies, suggest that the FIB-4 is the most reliable predictor of esophageal varices in liver cirrhosis patients. Despite the low diagnostic accuracy, FIB-4 is the most efficient non-invasive liver fibrosis marker which can be used as an initial screening tool for cirrhotic patients in the areas with lack of endoscopy facilities. However, none of the non-invasive markers assessed in this sample of Albanian patients was a useful predictor of esophageal variceal bleeding. Thus, these markers may not be adequate for the replacement of upper endoscopy. Future large prospective studies are warranted to further define the diagnostic accuracy of non-invasive markers in the diagnosis of EV and prediction of EVB in countries where there is a shortage of endoscopy.