Prediction of oesophageal varices in patients with compensated cirrhosis: A novel scoring system

Introduction 

Portal hypertension commonly accompanies the presence of liver cirrhosis, and the development of oesophageal varices (OV) is one of the major complications of portal hypertension [1]. The prevalence of OV in patients with liver cirrhosis may range from 60% to 80%, and the reported mortality from variceal bleeding ranges from 17% to 57% [2], [3], [4]. Large OVs (LVs) are more likely to bleed than small OVs (SVs) [5].

The American Association for the Study of Liver Disease and the Baveno IV Consensus Conference on portal hypertension have recommended that all cirrhotic patients should be screened for the presence of OV when liver cirrhosis is diagnosed [6], [7].

Investigators have attempted to identify characteristics that non-invasively predict the presence of varices. These studies have shown that biochemical, clinical and ultrasonographic parameters alone, or together, have good predictive power for non-invasively assessing the presence of OV [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24].

Variceal size has been identified as one of the most important factors responsible for first variceal haemorrhage [25]. 10–20% of small varices progress in size during one year [26], revealing close to 20–30% risk of bleeding in the first two years after first detection [27]. It seems that recognising patients in initial diagnosis (or periodic intervals thereafter) with elevated risk of bleeding can allow on time interventions and hence reduce morbidity and cost.

As the incidence of chronic liver diseases is growing, it is of major interest to develop non-invasive predictive tools to narrow the scope for identification of cirrhotic patients with significant portal hypertension and varices [28].

Consensus based guidelines recommend endoscopic screening of all cirrhotic patients for the presence of varices at the time of diagnosis [7]. However, the relatively low risk of bleeding in compensated cirrhotic patients and the need to avoid invasive and avoidable procedures suggests performing an upper gastrointestinal endoscopy only on those patients with clinical evidence of portal hypertension [6].

Identification of this high-risk group and giving them the appropriate portal hypotensive agents minimises the risk of first variceal bleeding, which is associated with a mortality rate approaching 30–50% [15].

Several markers had been studied for the possibility of providing a non-invasive parameter for the prediction of oesophageal varices. Among these, platelet count has been reported to be a good predictor of varices. Garcia et al. [19] defined three independent parameters associated with the presence of significant portal hypertension and oesophageal varices: serum albumin, telangiectasis and thrombocytopenia. Zaman et al. [10] showed that patients with a platelet count of less than 88,000/mm3 had a fivefold increased risk of having large oesophageal varices compared with those with a higher platelet count. Ng [12] identified an association among ascites, thrombocytopenia, hyperbilirubinaemia and high-grade varices in Chinese population. Chalasani et al. [9] also identified thrombocytopenia and splenomegaly as significant predictors of large varices. Madhotra et al. [15] found that thrombocytopenia, splenomegaly and worsening of Child-Pugh score were significant predictors of oesophageal varices. Thomopoulos [29] found that thrombocytopenia, splenomegaly above 135 mm (longest axis) and ascites are independent factors for prediction of large varices, suggesting that endoscopy can be safely avoided in cirrhotic patients without the existence of these factors.

We therefore sought to identify a multivariate non-invasive scoring system which can predict non-invasively the presence and the size of oesophageal varices in patients with compensated cirrhosis. The surveillance of large oesophageal varices can help in the prediction of first variceal bleeding, of critical importance in preventing such bleeding and reducing its medical, social and economic costs.

Patients and methods 

This work was conducted in the wards, out-patient clinics, Endoscopy Unit of Internal Medicine and Radiology Departments of the Faculty of Medicine of Zagazig University, in the period from March 2007 to February 2009.

Seventy patients with compensated cirrhosis, whether due to hepatitis C or B viruses, bilharsiasis or co-infection, were enrolled in this retrospective study (39 males and 31 females, mean age of 43.5±9.9 years, range from 22 to 65 years).

The clinical parameters were age, gender, diagnosis (cirrhosis alone), Child-Pugh class (Table 1), absence of ascites, absence of portosystemic encephalopathy (PSE) according to the West Haven criteria [28], detectable splenomegaly, absence of previous history of GOVH, and haemodynamic instability.

Table 1. Criteria for Child-Pugh classification [45].

	Clinical and biochemical measurements				Points scored for increasing abnormality
	Haepatic encephalopathy (grade)	○	○	○	None	1 and 2	3 and 4
		1	2	3
	Ascites	○	○	○	Absent	Mild	Moderate
		1	2	3
	Total bilirubin (mg/dl)	○	○	○	<2.0	2.0–3.0	>3.0
		1	2	3
	Serum albumin (g/dl)	○	○	○	>3.5	2.8 - 3.5	<2.8
		1	2	3
	Prothrombin time (sec. prolonged) or prothrombin time INR	○	○	○	<4 or <1.7	4–6 or 1.7–2.3	>6 or >2.3
		1	2	3

The laboratory parameters assessed were haemoglobin, platelet count, prothrombin time, serum bilirubin, and albumin. All laboratory parameters were acquired before endoscopy. The ultrasonographic characteristics noted were splenic axis, any degree of ascites, and portal vein diameter. The spleen was considered enlarged if the splenic axis exceeded 130 mm, while the portal vein was considered enlarged if its diameter exceeded 11 mm [15], [30].

Diagnosis of cirrhosis was made on the basis of biochemical parameters and imaging studies and/or liver biopsy, whenever available.

Patients were selected according to the Child Turcott Pugh numerical scoring system for the evaluation of the severity of liver disease class (A). Criteria were determined according to:

(a)Clinical exclusion of features of liver cell failure: ascites, jaundice, spider angiomata, palmer erythema, flapping tremors, foetor haepaticus and haepatic encephalopathy. Clinical diagnosis of compensated cirrhosis was made on finding an enlarged liver with a sharp border and signs of portal hypertension (splenomegaly, dilated abdominal wall veins).

(b)Abnormal biochemical parameters indicative of compensated cirrhosis Child-Pugh class (A), including: serum bilirubin ⩽ 2 mg/dl, serum albumin ⩾3.5gm/dl, prothrombin time prolongation not exceeding 0–4s, and international normalised ratio <1.7 [31].

(c)Ultrasonographic criteria of liver cirrhosis including: coarse nodular cirrhotic echopattern, shrunken size, prominent caudate lobe [32].

All patients were kept fasting overnight prior to the procedure. Patients were examined in the supine position during quiet respiration, using a real time gray-scale device transducer with a frequency of 3.5 MHz scanning through several longitudinal, oblique and transverse projections visualising different organs. The liver was examined and its size noted. It was considered enlarged if its extent was below costal margin. Cirrhotic echopattern was determined from the coarse nodular appearance, shrunken size and prominent caudate lobe [32]. Echogenicity was compared to the cortex of the right kidney, portal vein diameter was measured at a point of crossing the inferior vena cava, with a diameter greater than 13 mm indicating portal hypertension [33]. The spleen was evaluated for its greatest axis at the midaxillary line; an axis greater than 130 mm suggesting enlargement [30]. The presence of portal-systemic collaterals was also recorded.

The following main Doppler parameters were always taken by real time ultrasound equipment (Toshiba, nemio 20) consisting of a colour Doppler and a pulsed Doppler device working at 3.5 MHz frequency and by the same operator: (1) portal vein flow velocity as time average maximal velocity in cm/s and portal vein diameter (mm) [13]; (2) congestive index of portal vein, calculated as portal vein cross sectional area in cm/portal vein mean flow velocity {cm/s}; (3) resistive index of intrarenal arteriolar circulation, calculated as peak systolic velocity – end systolic velocity/peak systolic velocity and (4) detection of flow pattern of splenic circulation: normal, reversed, or dual patterned [34].

All patients underwent upper digestive endoscopy to evaluate the presence of signs of portal hypertension (presence and degree of OV). All endoscopic examinations were performed in a single endoscopy unit using a video endoscope and all OV were classified according to size. Small OV were defined as varices that flattened with insufflation or minimally protruded into the oesophageal lumen, while large OV were defined as varices that protruded into the oesophageal lumen and touched each other (presence of confluence), or that filled at least 50% of the oesophageal lumen [25]. Gastric varices were classified according to their location in the stomach and relationship with oesophageal varices [15]. Portal hypertensive gastropathy was classified as mild when a mosaic-like pattern (MLP) was present, and severe when MLP was superimposed by red signs [1]. Patients with isolated gastric varices, and those who had undergone any endoscopic variceal ablative procedure in the past were excluded from the study.

Data were analysed using SPSS for Windows version 13. Descriptive statistics, including means, standard deviations and frequencies were computed. The chi-square test was used to compare differences, and student’s t-test was used to compare means of variables. Values were considered significant if p<0.05 (95% CI). A logistic regression equation was developed to predict presence and grade of OV. The sensitivity and specificity of the prediction rule were estimated by means of a receiver operating characteristic (ROC) curve, and area under the curve (AUC) was reported for independent predictors.

Results 

Seventy consecutive cirrhotic patients (39 men, 31 women) were enrolled in the study. Mean age (±SD) of the study population was 43.5 (±9.9) years. Table 2 shows the main clinical and laboratory characteristics of the 70 patients included in our study.

Table 2. The clinical and laboratory parameters of 70 cirrhotic patients [as n (%) or mean±SD (range)].

	Patients data	n (%) or mean±SD (range)
	Age (years)	43.5±9.9
	Range	22–65
	Gender
	Male	39 (55.7)
	Female	31 (44.3)
	Splenomegaly	36 (51.4)
	Ascites	0 (0)
	Encephalopathy	0 (0)
	Spider naevi	0 (0)
	Flapping tremors	0 (0)
	Lower limb oedema	0 (0)
	F. Hepaticus	0 (0)
	White blood cells (103/mm3)	5.6±2.4 (2.7–9)
	Red blood cells (106/mm3)	3.5±1.4 (3–4.5)
	Haemoglobin (gm/dl)	10.4±1 (8.4–12.3)
	Platelet (103/mm3)	97.8±38 (40–173)
	Prothrombin time	15.7±1 (13–19)
	Reticulocytic count (103/mm3)	0.679±0.3 (0.2–1.6)
	INR	1.146±0.16 (1–2.1)
	Bilirubin (mg/dl)	1.18±0.24 (0.7–1.8)
	Serum albumin (gm/dl)	3.27±0.3 (2.9–4.3)
	ALT (IU/L)	40.7±11.5 (17–76)
	AST (IU/L)	40.6±11 (25–71)
	Creatinine (mg/dl)	0.86±0.2 (0.3–1.5)
	Cirrhotics without varices	25 (35.7)
	Cirrhotics with varices	45 (64.3)
	Small varices	25 (35.7)
	Large varices	20 (28.6)
	Associated gastric varices–congestive gastropathy	30 (42.9)
	Portal vein Diameter (mm)	14.3±1.7 (9.4–17)
	Splenic axis (mm)	168.3±19.6 (130–200)
	Splenic vein diameter (mm)	10.77±1.9 (8–15)
	Platelet count/splenic axis ratio (PC/SD ratio)	601.2±283.9 (130.8–1071.4)
	PVCIx	0.13±0.18 (0.02–0.93)
	RRI	0.65±0.05 (0.48–0.75)

Twenty-five patients had small OV and 20 had large OV. Gastric varices and congestive gastropathy were detected in 30 patients (42.9%).

Table 2 shows the univariate analysis of the clinical, laboratory, and primary and derivative echo-Doppler factors in the study population according to the presence of oesophageal varices in any size.

Univariate analysis showed that there were highly significant differences between the two groups regarding to the presence of splenomegaly, lower white blood cell count, lower platelets count, hyperbilirubinaemia, lower albumin level, and more prolonged prothrombin time, which was more prevalent in the group with OV.

Univariate analysis showed that most of the echo-Doppler parameters were related to presence of OV (Table 3). Splenic axis, portal vein diameter, splenic vein diameter, portal vein congestive index and renal resistive index were significantly higher in patients with OV; these patients had lower platelet count/splenic axis ratio. Presence of LOV was related to most of the echo-Doppler parameters described in Table 4 (portal vein diameter (mm) {p=0.01}, splenic axis (mm) {p=0.001}, splenic vein diameter (mm) {p=0.001}, and platelet count/splenic axis ratio {p=0.01}).

Table 3. Mean±SD of the clinical, laboratory, and the primary and derivative echo-Doppler factors in study population according to the presence of oesophageal varices of any size.

		Without OV (25)	With OV (45)	t	P value
	Age	38.4±6.8	36.3±6.3	0.34	NS
	Splenomegaly	No.=3	No.=33	χ2=24.2	<0.001
	White blood cells	6.6±1.3	5.1±1.4	4.4	<0.001
	Red blood cells (106/mm3)	4±0.5	3.5±0.4	4.2	<0.001
	Haemoglobin (gm/dl)	11.4±0.5	9.8±0.7	9.7	<0.001
	Platelet count (103/mm3)	134.2±19.75	80.7±33.4	7.3	<0.001
	Prothrombin time	15.3±0.9	16.4±1.47	3.4	<0.001
	Reticulocytic count (103/mm3)	0.676±0.4	0.68±0.3	0.05	NS
	Bilirubin (mg/dl)	1.016±0.17	1.27±0.2	4.8	<0.001
	Albumin (gm/dl)	3.5±0.2	3.1±0.3	4.4	<0.001
	ALT (IU/L)	40±14.2	41±9.8	0.36	NS
	AST (IU/L)	40.7±12.5	40.5±10.2	0.08	NS
	Creatinine (mg/dl)	0.88±0.2	0.85±0.2	0.65	NS
	Portal vein diameter (mm)	13±1.2	15±1.6	5.16	<0.001
	Splenic axis (mm)	150±16.2	180±12.9	8.06	<0.001
	Splenic vein diameter (mm)	9±0.9	12±1.4	9.03	<0.001
	Platelet count/splenic axis ratio	900±146.2	445.1±210.6	9.5	<0.001
	Portal vein congestive index	0.056±0.01	0.14±0.14	k=36.7	<0.001
	Renal resistive index	0.62±0.05	0.686±0.04	t=5.95	<0.001
	Splenic venous flow pattern
	Normal {No. (%)}	25 (100)	38 (84.4)	χ2=4.32	0.03
	Reversed {No. (%)}	0 (0)	7 (15.6)

χ2=Chi-square test.

Table 4. Mean±SD of the clinical, laboratory, and the primary and derivative echo-Doppler factors in study population according to presence of small or large oesophageal varices.

		With SOV (25)	With LOV (20)	t	P value
	Age	43.6±11.7	49.6±7.1	0.512	NS
	Splenomegaly	No.=15	No.=18	χ2=5.11	0.02
	White blood cells (103/mm3)	5.5±1.4	4.6±1.3	2.1	0.04
	Red blood cells (106/mm3)	4.2±0.4	3.5±1	2.5	0.03
	Haemoglobin (gm/dl)	9.9±0.7	9.7±0.77	1.14	(NS)
	Platelet count (103/mm3)	93.1±31.4	65.2±29.6	3.02	0.004
	Prothrombin time	15.7±0.9	17.3±1.55	4.4	<0.001
	Reticulocytic count (103/mm3)	0.62±0.2	0.75±0.3	1.59	NS
	Bilirubin (mg/dl)	1.2±0.2	1.34±0.24	2	0.04
	Serum albumin (gm/dl)	3.2±0.26	3±0.33	1.88	NS
	ALT (IU/L)	40.3±10.5	41.9±9	0.52	NS
	AST (IU/L)	39.8±10.3	41.3±10.2	0.5	NS
	Creatinine (mg/dl)	0.83±0.2	0.87±0.25	0.58	NS
	Portal vein diameter (mm)	14.5±1.8	15.5±1	2.4	0.01
	Splenic axis (mm)	173.2±14.2	185±6.9	3.4	0.001
	Splenic vein diameter (mm)	11.2±1.3	12.55±1.27	3.5	0.001
	Platelet count/splenic axis ratio	517.9±217	354±165	2.8	0.01
	Portal vein congestive index	0.118±0.03	0.158±0.19	0.93	NS
	Renal resistive index	0.69±0.04	0.68±0.04	0.71	NS
	Splenic venous flow pattern
	Normal {No. (%)}	21 (84)	17 (85)	χ2=0.1	NS
	Reversed {No. (%)}	4 (16)	3 (15)

χ2=Chi-square test; SOV=small OV and LOV=large OV.

Table 5 shows a non-invasive scoring system proposed for the presence of oesophageal varices and the possibility of presence of large varices (with very high score). Here we chose the most important variables commonly associated with the presence of oesophageal varices. Of the total score of 15, 0–7=possibility of oesophageal varices is low; 8–12=possibility of OV presence cannot be excluded; and 13–15=possibility of varices is high.

Table 5. Proposed non-invasive scoring system to predict the presence of OV in patients with compensated cirrhosis.

	High-risk variable	Score
	Platelet count
	>150,000 cells/ml	0
	150,000–100,000 cells/ml	1
	100,000–80,000 cells/ml	2
	<80,000 cells/ml	3
	Portal vein diameter
	<13 mm	0
	13–15 mm	1
	>15 mm	2
	Splenic vein diameter
	<10 mm	0
	10–12 mm	1
	>12 mm	2
	Splenic axis
	<130 mm	0
	130–180 mm	1
	>180 mm	2
	Platelet count/splenic axis ratio
	>900	0
	900–445	1
	<445	2
	Portal vein congestive index
	<0.07	0
	0.07–0.14	1
	>0.14	2
	Renal resistive index
	<0.68	0
	0.68–0.7	1
	>0.7	2

Table 6 shows the validity of the new scoring system in detection of OV and large varices, where the OV presence at a total score above 7 showed sensitivity 80%, specificity 100%, positive predictive value of 100%, negative predictive value of 90%, and accuracy 87.1%. The scoring system for prediction of large varices presence at a total score above 13 shows sensitivity 100%, specificity 92%, positive predictive value of 91%, negative predictive value of 100%, and accuracy 95.6%.

Table 6. Validity of the new score in detection of presence OV or large OV.

		Sensitivity	Specificity	PV		Accuracy
				+ve	−ve
	Presence of OV	80	100	100	90	87.1
	Large OV	100	92	91	100	95.6

Discussion 

The aim of the present study has been to propose a multivariate non-invasive scoring system which can non-invasively predict the presence and size of oesophageal varices in patients with compensated cirrhosis. The surveillance of large oesophageal varices can help in the prediction of first variceal bleeding, of critical importance in preventing such bleeding and reducing its medical, social and economic costs.

This study included 70 patients regarded as compensated cirrhotics through clinical exclusion of signs of liver cell failure including ascites, jaundice, telangiectasia, haepatic encephalopathy, and laboratory criteria determined according to Child-Pugh score: serum bilirubin <2 mg/dl, serum albumin ⩾3.5 gm/dl, prothrombin time (0–4s prolonged), INR <1.7 [33]. The patients were divided into two groups: compensated cirrhotic patients without varices (25 patients), and compensated cirrhotic patients with varices (45 patients).

A highly significant statistical difference was found between the two groups as regards reduced leucocyte count, red cell count (RBCs), haemoglobin level, platelet count, and prolonged prothrombin time in the variceal group. This finding is consistent with Garcia et al. [19].

In patients with chronic liver disease, the decrease in platelets count may be attributed to several factors other than portal hypertension, including shortened mean life span, thrombopoietin deficiency, myelotoxic effect of alcohol, and viral infection of megakaryocytes in the bone marrow [35]. A splenomegaly present in cirrhotic patients is likely due to the vascular disturbance and immunologic disorders related to portal hypertension [36].

In this study we evaluated thrombocytopenia as a predictor for the presence of varices. Here a highly significant statistical difference was found in both groups, being lower in variceal group. The platelet count was found to be independently associated with presence of oesophageal varices. It showed a diagnostic sensitivity of 77.8% at a cut off value of <100,000/mm3, with specificity, accuracy, positive predictive value, negative predictive value at 92%, 82.85%, 94.6%, 69.7%, respectively. The platelet count showed significant negative statistical correlation with OV presence and negative correlation with the size of OV. The logical explanation for this lies in the fact that with worsening liver functions and thrombocytopenia (which is associated with advancing cirrhosis), the incidence of large OV will increase. This agrees with the study of Garcia et al. [19] who found that a platelet count cut off value of 68,000/mm3 was associated with significant portal hypertension and OV. Zaman et al. [10] found that a platelet count cut off value of less than 88,000/mm3 was associated with a fivefold increase in the risk of having large OV. Madhotra et al. [15] revealed that a platelet count <68,000/mm3 had the highest discriminating value. The differences in the cut off value between these studies and the current study may be attributed to the difference of aetiology of liver cirrhosis, malnutrition, viral infection and alcoholism.

The liver function tests, including serum bilirubin, serum albumin and prothrombin time showed a significant statistical difference between the two groups; serum bilirubin was higher while serum albumin was lower in variceal group, prothrombin time was more prolonged in the same group. The liver enzymes and serum creatinine were insignificant. This finding is supported by the results of Nakano et al. [37], who found that liver functions were worse in patients with OV, and Burton et al. [38], who showed that the incidence of large varices increases with worsening Child-Pugh class and raised a predictive model relying on thrombocytopenia and Child-Pugh class. Burton’s model revealed that patients with platelet count less than 90,000/mm3 will have a probability of 0.57, 0.81 and 0.82 of having any varices, if they are in Child A, B, or C classes, respectively. The probability of having any varices will decline to 0.36, 0.65 and 0.66, respectively if the count is more than 90,000/mm3.

For prediction of large varices, the probability of patients with platelet count less than 80,000/mm3 and lying in Child A versus Child B or C classes will be 0.24, or 0.46; this probability will decrease to 0.12 or 0.27, respectively, with platelet count more than 80,000/mm3. The predictive accuracy of Burton’s validation study was 0.67 for any varices (CI 95%; 0.583–0.759) and 0.74 for large varices (95% CI; 0.64–0.85).

We depended on ultrasonography during this study as a tool which can demonstrate many signs of portal hypertension and can non-invasively predict the presence of oesophageal varices. The ultrasonographic findings showed a highly significant statistical difference between the two groups as regards portal vein diameter, splenic axis and splenic vein diameter, being larger in the variceal group.

It was found that the use of platelet count alone as a predictor of varices may be a drawback, as this count can depend on factors other than portal hypertension and causes an increase in the false positive results. To avoid this bias, an index based on platelet count divided by splenic axis in (mm) was proposed by Giannini et al. [39], which normalises the platelet count to splenic sequestration. In this study, the platelet count/splenic axis ratio in cirrhotics with varices (445.1±210.6) was significantly reduced when compared to cirrhotics without varices (900±146.2), being lower in the variceal group. This ratio was independently associated with the presence of oesophageal varices with a diagnostic sensitivity (98%) at a cut off value of 900 with specificity, accuracy, positive predictive value, negative predictive value of 64%, 86%, 83% and 94% .

Platelet count/splenic axis ratio also showed a highly significant negative statistical correlation with the presence of oesophageal varices. This can be explained by the fact that when the ratio is below 900, the incidence of oesophageal varices increases; this is shown in the current study, in which a ratio less than 900 was detected in 44 patients with varices, and nine patients without varices. However, a ratio greater than 900 was encountered in one patient with oesophageal varices and in 16 patients without varices. Also, this ratio showed a significant negative statistical correlation with the size of varices.

Compared with the current study, the study done by Giannini et al. [39] proposed a platelet count/splenic diameter ratio with a diagnostic sensitivity 100% at a cut off value of 909 with specificity 93%, OR 0.527 (95% CI 0.427–0.65) (p<0.001).

The difference between these studies may be due to the larger number of patients in the study of Giannini et al. [39] which included 145 patients; it could also be due the fact that Giannini et al. also included different Child classes, in contrast to our study which evaluated patients with Child (A) cirrhosis only.

Thabut et al. [40] commented on Giannini’s results, suggesting that platelet count/splenic diameter ratio did not add significant discrimination to platelet count alone in their population under study (C-index of 0.92 versus 0.88), attributing these results to the surprisingly good performance of platelet count alone. This was in agreement with our study (C-index 0.64 versus 0.7).

Doppler ultrasound has resulted in a significant extension of the role of ultrasound in examining the abdomen, allowing determination of the direction of blood flow, velocity, and organ perfusion.

In the current study, duplex Doppler sonography was evaluated as a non-invasive tool for the prediction of oesophageal varices.

The portal vein congestive index (PVCIx) as revealed in our study was significantly higher in the variceal group, and was independently associated with presence of oesophageal varices with a diagnostic sensitivity 88.9% at a cut off value of 0.07 with specificity, accuracy, positive predictive value and negative predictive value of 96%, 91.4%, 97.6% and 82.8%, respectively.

As regards variceal size, PVCIx was found to be statistically non-significant. This may be explained by the fact that our patients belong to Child class A and the expected increase in portal vein dilatation and velocity were blunted with the increase of variceal size [39]. Also, the spontaneous portosystemic shunts drain and lessen the augmented portal circulation, and this affects the intravariceal pressure. Furthermore, 10% of patients with increased portal vein congestive index show no varices on endoscopy [41].

Duplex Doppler evaluation of renal perfusion represents a simple, inexpensive, safe method to identify a subset of symptomless patient with compensated cirrhosis. The renovascular impedance is increased in cirrhotic patients due to the imbalance between the renal vasoconstrictors and vasodilators which occur in favour of vasoconstrictors, mainly endothelins [42], and the Local Baroceptor theory in which the rise in sinusoidal pressure causes reflex renal artery vasoconstriction [43]. In the current study, renovascular impedance was evaluated by detecting Renal arteriolar Resistive Index (RRI), which showed a statistically highly significant difference, being higher in the variceal group. As regards OV size, the RRI was non-significant.

The RRI was independently associated with oesophageal varices, with a diagnostic sensitivity of 64.4% at a cut off value of 0.68 with specificity, accuracy, positive and negative predictive values of 100%, 77.1%, 100% and 61%, respectively. In agreement with our study, Agostino and Mirella [41] found that renal resistive index was consistent with increased impedance, with a value of ⩾0.7 in 18 of 50 studied patients (36%) and normal index in 32 patients (64%), and the proportion of varices present was significantly higher in the former group (14 patients; 78%) versus the later (10 patients; 31%) and the RRI in the detection of varices was better than portal vein congestive index. Further, Aydogdu et al. [44] showed that patients with cirrhosis are at high risk of renal deterioration which cannot be detected by serum urea, creatinine and glomerular filtration rate. The increase in renal resistive index is associated with progress of haepatocellular disease and the development of ascites and portal hypertension. The increased urinary N-acetyl-Beta-d glucosaminidase/creatinine and microalbuminuria might have prognostic value in patients with Child score ⩾6. This suggests that monitoring of RRI provides a non-invasive means of studying renal haemodynamics alteration in cirrhosis.

Splenic venous flow pattern was evaluated by means of duplex Doppler sonography. It was statistically significant and indicating portal hypertension, however this phenomenon was detected only in seven patients who belonged to the variceal group (10%), indicating that it is an uncommon finding and if present, it might be associated with a higher risk of variceal bleeding. This agrees with the findings of Barakat et al., [34] who conducted their study on 176 patients with portal hypertension. In this, a normal flow pattern was maintained in all except four patients (2.3%) who had either reversed flow or dual venous drainage patterns.

We have postulated a new scoring system, composed of selected variables, which can screen patients with compensated cirrhosis at high risk of OV development in order to narrow the spectrum of patients without a previous history of variceal bleeding or endoscopic documentation of oesophageal varices. In this system the total score is 15, and a score of 0–7 is associated with a high possibility of absent varices; a score of 8–12 is associated with a considerable possibility of variceal presence; and a score of 13–15 is associated with a high possibility of presence of large varices. This group of patients needs diagnostic work up and can be considered a high-risk group, and should be evaluated endoscopically. Patients who exhibit large varices should be treated with portal hypotensive agents as beta blockers.

This scoring system has a validity for the presence of varices with a diagnostic sensitivity of 80%, specificity, accuracy, positive predictive value and negative predictive value of 100%, 87.1%, 100% and 90%, respectively, and a validity for large varices presence with diagnostic sensitivity of 100%, specificity, accuracy, positive predictive value, negative predictive value of 92%, 95.6%, 91% and 100%, respectively.

Given these findings, we found that some laboratory parameters (serum bilirubin, serum albumin), some ultrasonographic parameters (portal vein diameter, splenic axis, splenic vein diameter and platelet count/splenic axis ratio) and duplex Doppler sonography parameters (PVCIx, IRR) may be important predictors for the presence of varices on univariate analysis. However, the strong confounding association between variables resulted in exclusion of some factors stepwisely from the multivariate model, leaving only platelet count, platelet count/splenic axis ratio, IRR, PVCIx as independent factors for the presence of oesophageal varices, and platelet count, platelet count/splenic axis ratio as independent factors for the presence of large varices.