Original article 1335
Diagnostic accuracy of abdominal ultrasound in the screening of esophageal varices in patients with cirrhosis Pau Sorta, Magdalena Muelasb, Alvaro Isavaa, Jordina Llaóa, Francesc Portaa, Ignasi Puiga, Claudia Domínguez-Curellc, Enrique Esteveb, Carles Yanguasd and Francesc Vidaa Background Abdominal ultrasound (US) may provide data on the presence of esophageal varices in cirrhosis. We assess the diagnostic accuracy of this procedure. Patients and methods Retrospective recording of clinical data was carried out in cirrhotic patients who underwent abdominal US and upper gastrointestinal endoscopy. We compared patients with and without large varices and assessed the value of US in predicting the presence of these lesions as well as other significant variables. Results Of the 353 patients included, 123 (35%) had esophageal varices. The presence of US signs of portal hypertension independently predicted the existence of esophageal varices with a sensitivity of 87.9%, a specificity of 34.9%, a positive predictive value of 40.6%, and a negative predictive value of 85.1%, which could increase to 91.5% if the patient presented plasma albumin and platelet concentrations above the mean values (3.1 g/dl and 122 × 109 cells/l, respectively). Plasma albumin and platelet concentrations were the two other variables with independent predictive capacity. Applying these selection criteria, up to 30% of screening endoscopies may not be necessary, and up to 43% in patients with compensated
Introduction Cirrhosis is an irreversible disease of the liver. In the Child–Pugh classification, it is divided into three stages according to the deterioration of liver function: stage A, in which patients are mainly asymptomatic; stage B, the intermediate stage; and stage C, the terminal stage of the disease [1]. Upper gastrointestinal bleeding because of gastroesophageal varices is a frequent complication of cirrhosis. It is estimated that 50% of patients with cirrhosis present varices, ranging from 40% in patients with compensated cirrhosis to 85% in those with a highly deteriorated liver function [2]. The 1-year probability of bleeding (5–15%) is especially high in patients with large varices (LVs) with red wale markings on the surface [3]. For this reason, the current recommendation is to identify cirrhotic patients with a major risk of bleeding by upper gastrointestinal endoscopy (UGE) at regular periods: at the time of diagnosis, yearly in patients with decompensated 0954-691X © 2014 Wolters Kluwer Health | Lippincott Williams & Wilkins
cirrhosis. In patients with decompensated cirrhosis, however, US does not have predictive capacity. The results obtained are comparable with those reported for transient elastography. Conclusion Abdominal US is a highly reliable technique for detecting patients with a low risk of presenting esophageal varices. Its use may avoid up to 43% of screening endoscopies in patients with compensated cirrhosis. The results obtained are similar to those observed using transient elastography. Eur J Gastroenterol Hepatol 26:1335–1341 © 2014 Wolters Kluwer Health | Lippincott Williams & Wilkins. European Journal of Gastroenterology & Hepatology 2014, 26:1335–1341 Keywords: esophageal varices, liver cirrhosis, portal hypertension, ultrasound a Digestive Diseases Service, bInternal Medicine Service, cIntensive Care Service and dDiagnostic Imaging Service, Althaia Xarxa Hospitalaria i Universitària de Manresa, Universitat Internacional de Catalunya, Manresa, Spain
Correspondence to Pau Sort, PhD, Hospital Sant Joan de Deu, c/Dr Joan Soler 1-3, 08243 Manresa, Barcelona, Spain Tel: + 34 93 874 21 12; fax: + 34 93 8744554; e-mail: [email protected] Received 29 April 2014 Accepted 30 June 2014
cirrhosis, and every 2–3 years in the case of compensated cirrhosis [4–6]. However, endoscopic screening programs are costly and UGE is not always well received by patients [7]; therefore, it seems sensible to limit its use to patients with a higher risk of bleeding. Some studies have evaluated alternative methods for predicting the presence of LVs without performing UGE [8–13]. However, their results are inconclusive either because the main alternative diagnostic tool [transient elastography (TE)] is not available at many centers, because the population studied did not reflect the heterogeneity of cirrhotic patients (e.g. only Child–Pugh group A or patients awaiting liver transplantation were included), or because the results did not reach statistical significance. The aim of this study is to establish whether abdominal ultrasound (US), a technique available at all health facilities, can identify cirrhotic patients who are less likely to present gastroesophageal varices and who therefore need not be included in screening programs. DOI: 10.1097/MEG.0000000000000174
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
1336
European Journal of Gastroenterology & Hepatology 2014, Vol 26 No 12
Objectives The aims of this work were to study the accuracy of abdominal US in discriminating between cirrhotic patients with and without LV, to define the variables that identify patients with the lowest risk of bleeding in both compensated and decompensated cirrhosis, and to establish whether their application might help to reduce the number of endoscopies performed in the screening of LV.
Patients and methods Patients
Clinical, laboratory, radiological, and endoscopic data were retrospectively collected from 469 consecutive cirrhotic patients who had consulted our outpatient clinic for any reason or had been admitted to our hospital, a secondary-level health facility, between January 2011 and June 2013. Cirrhosis was diagnosed when liver biopsy performed for any reason indicated typical features of this disease, when macroscopically cirrhotic liver was incidentally found during abdominal surgery, or when US or computed tomographic scan findings (heterogeneous liver parenchyma with hypertrophic caudate lobe and nodular margins) coupled with analytical (low platelet or leukocyte count, anemia, abnormal prothrombin time, bilirubin or albumin serum concentration) or clinical data (spider naevi, ascites, hepatic encephalopathy) suggested this disease. Abdominal US was performed at the time of evaluation in all cases. All patients who, in addition to the abdominal US, had undergone UGE within 3 months of the US were included in the study. Patients with suspected or confirmed alcoholic hepatitis or end-stage hepatocarcinoma were excluded. Finally, the study sample comprised 353 patients. The reasons for not performing UGE in the remaining 101 patients were as follows: severe comorbidity (such as invalidating cardiac or pulmonary failure; 35 cases), advanced age [21 cases; mean age: 84.56 years, 95% confidence interval (CI): 82.90–86.11], patients who did not return to the outpatient unit (9 cases), refusal (6 cases), and others (30 cases). Fifteen patients were excluded from the study because of endstage hepatocarcinoma (10 cases) and acute alcoholic hepatitis (5 cases). Abdominal ultrasound
All the radiology studies were performed by five radiologists with considerable experience with US of the liver and who were blinded to the results of the complementary examinations. Following previous publications, we considered that the presence of a portal diameter higher than 12 mm, a spleen size larger than 12 cm, perigastric or perisplenic collateral circulation, or ascites indicated the existence of portal hypertension and, when present, two or more of these US findings were associated in the vast majority of patients [14–16].
Previous studies have shown that the sensitivity of US in detecting portal hypertension, as estimated by hepatic vein catheterization, is about 40% with a sensitivity higher than 80% considering the presence of portosystemic venous collaterals [17]. In cirrhosis, ascites occurs only in the presence of portal hypertension [18]. Upper gastrointestinal endoscopy
UGE was performed by five gastroenterologists with wide experience in digestive endoscopy and who were blinded to the results of the complementary examinations. The examinations were performed under topical anesthesia with lidocaine or light sedation with propofol, and the presence or absence of gastroesophageal varices and their size was recorded. Varices were staged in accordance with the criteria published elsewhere [3]. According to the UGE results, patients were classified into two groups: those with medium-sized or large gastric or esophageal varices with signs of risk of bleeding (group LV + ) and those with small varices without signs of risk of bleeding or only portal hypertensive gastropathy, or who did not present endoscopic lesions that could be attributed to portal hypertension (group LV − ). Statistical analysis
The study compares the characteristics of all the patients divided according to the presence or absence of LV, and of the subgroups of patients with compensated and decompensated cirrhosis. The variables analyzed were age, sex, etiology of the cirrhosis, presence of ascites, prothrombin time, platelet count, albumin, bilirubin, Child–Pugh score, presence of splenomegaly, presence of US signs indicating portal hypertension, and presence of LV. Qualitative variables were compared using Pearson’s χ2-test. Quantitative variables were compared using the Student t-test if the variances between the two groups did not show statistically significant differences on the Levene test and also showed a normal distribution in the Kolmogorov–Smirnov test; if these two conditions were not fulfilled, a nonparametric test (the Mann–Whitney U-test) was used. Parameters that reached statistical significance in the univariate analysis were included in a forward stepwise logistic regression model. Quantitative variables that independently predicted the existence of LV were dichotomized according to their mean values. We could thus calculate the sensitivity, specificity, and positive and negative predictive values, and odds ratio (OR) for the existence of LV for both qualitative and quantitative variables. To obtain an accuracy of 7% for estimation of the sensitivity of abdominal US to detect LV in cirrhotic patients, with a 95% CI, and assuming a sensitivity of 84% [10] compared with UGE as the benchmark, at least 106 cirrhotic patients with LV had to be included. Assuming
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Accuracy of ultrasound in varices Sort et al. 1337
a prevalence of large esophageal varices of 30%, the minimum sample size required was 353 cirrhotic patients. The software used was SPSS (SPSS Inc., Chicago, Illinois, USA). Quantitative variables are expressed as means and 95% CIs. Differences were considered significant when the P value was less than 0.05. The protocol was approved by our hospital’s independent Clinical Research Ethical Committee. The study fulfilled all the legal requirements and local regulations and those of the Declaration of Helsinki.
Results Patients’ characteristics
Table 1 shows the characteristics of the patients included in the study. The majority of patients were men older than 60 years of age. Alcoholic and viral etiologies each accounted for close to 40% of cases, with the remaining 21% because of other causes. The majority of patients were Child–Pugh stage A (49%), followed by stage B (41%); 10% of patients were stage C (i.e. highly deteriorated liver function). At the time of inclusion, 35% of the patients had ascites, 73% showed US signs of portal hypertension, and 35% had LV. Ability of ultrasound to predict the existence of large varices in cirrhotic patients
Patients were divided into two groups (Table 2): group LV + (123 patients; 35%) and group LV − (230 patients; 61%). The groups were similar with respect to age, etiology of cirrhosis, and patients’ sex; however, the LV + group had significantly worse hepatic function, as estimated by serum bilirubin, prothrombin time, serum albumin, presence of ascites, and Child–Pugh score. Platelet count was also significantly lower in this group and a higher proportion of LV + patients showed US signs of portal hypertension (88 vs. 65% respectively;
Table 1
P < 0.000). The multivariate analysis showed that only three of the variables that had been significant in the univariate analysis were independent predictors of the presence of LV: US signs of portal hypertension (OR: 3.91; 95% CI: 2.01–7.28; P < 0.0001), platelet count (OR: 2.76; 95% CI: 1.66–4.60; P < 0.0001), and plasma albumin (OR: 1.81; 95% CI: 1.14–2.87; P < 0.012). The sensitivity, specificity, positive predictive value, negative predictive value, positive likelihood ratio, and negative likelihood ratio of US in the prediction of the presence or absence of esophageal varices are shown in Table 3. The results indicate that if abdominal US were used to exclude patients at a low risk of LV from the endoscopic screening program, the number of endoscopies with this indication would decrease by 27%, but 14.9% of patients would be misclassified. However, if patients with US signs of portal hypertension or platelets less than 120 × 109 cells/l or plasma albumin below 3.14 mg/dl are excluded from endoscopic screening, the sensitivity increases to 96.5% and the negative predictive value decreases to 91.5%. This would lead to a 14% reduction in the endoscopies performed to screen the esophageal varices, but would leave 8.5% of patients misclassified (OR: 6.41; 95% CI: 2.24–18.35; P < 0.0005). Differences in the utility of abdominal ultrasound in compensated patients
The results for patients with compensated cirrhosis (Child A) showed that the lack of US signs of portal hypertension predicted the absence of LV in a very high percentage of patients (negative predictive value 93%) (OR: 5.16; 95% CI: 2.13–12.50; P < 0.0003). The use of US as a criterion for excluding low-risk patients from endoscopic screening would reduce the number of endoscopies performed by 43%. The other variable that presented an independent predictive value in the univariate analysis was plasma
Patients’ baseline characteristics
Age (years) Male sex [n (%)] Etiology [n (%)] Alcohol Viral Others Prothrombin time (%) Serum bilirubin (mg/dl) Serum albumin (mg/dl) Platelets (×109 cells/l) Ascites [n (%)] Child–Pugh [n (%)] A B C Splenomegaly [n (%)] PHT-US [n (%)] LV + [n (%)]
Overall (n = 353)
Child A (n = 172)
Child B + C (n = 181)
62.54 (61.19–63.88) 236 (67)
61.6 (59.64–63.61) 112 (65)
63.07 (61.20–64.94) 124 (68)
153 140 60 73.20 1.78 3.14 121.8 123 172 145 36 207 257 123
(43) (40) (17) (71.19–75.21) (1.60–1.97) (3.05–3.23) (114.0–129.7) (35) (49) (41) (10) (59) (73) (35)
57 80 35 83.01 1.14 3.59 127.9 0
(33) (47) (20) (80.88–85.15) (1.04–1.23) (3.50–3.68) (117.5–138.2) (0)
– – – 93 (54) 99 (58) 46 (27)
96 60 25 65.11 2.45 2.73 118.5 123
(53) (33) (14) (62.57–67.65) (2.11–2.78) (2.60–2.86) (106.6–130.4) (68)
– – – 114 (63) 158 (87) 77 (43)
LV + , patients with medium-sized or large esophageal varices; PHT-US, ultrasound signs of portal hypertension, as described in the Patients and methods section.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
1338
European Journal of Gastroenterology & Hepatology 2014, Vol 26 No 12
Table 2 Variables presenting statistically significant differences in the univariate analysis between patients with esophageal varices (LV + ) and without (LV − ) LV + (n = 123) All patients Prothrombin time (%) Serum bilirubin (mg/dl) Serum albumin (mg/dl) Platelets (×109 cells/l) Ascites [n (%)] Child–Pugh [n (%)] A B C Splenomegaly [n (%)] PHT-US [n (%)] Child A Prothrombin time (%) Serum bilirubin (mg/dl) Platelets (×109 cells/l) Splenomegaly [n (%)] PHT-US [n (%)] Child B + C Platelets (×109 cells/l) Splenomegaly (n = 111) [n (%)]
69.23 2.14 2.97 97.7 55 46 56 22 94 108 79.83 1.26 118.4 35 42
LV − (n = 230)
(66.50–71.93) (1.75–2.53) (2.82–3.12) (87.5–108.0) (45)
75.10 1.61 3.23 134.1 68
(37) (46) (18) (77) (88)
126 89 14 113 149
(76.70–82.96) (1.04–1.48) (100.50–136.30) (76) (91)
85.27 0.90 141.2 59 66
101.1 (85.7–116.47) 59 (72)
P
(72.41–77.79) (1.41–1.81) (3.11–3.34) (123.7–144.4) (29)
0.002 0.016 0.008 0.000 0.011 0.000
(55) (38) (6) (51) (65)
0.000 0.000
(79.37–91.16) (0.79–1.00) (120.5–162.00) (47) (52)
0.047 0.036 0.001 0.003 0.000
153.9 (119.47–188.37) 54 (57)
0.000 0.024
LV + , presence of medium-sized or large esophageal varices; LV − , absence of medium-sized or large esophageal varices; PHT-US, presence of ultrasound signs of portal hypertension.
Table 3 Predictive capacity of the presence of medium-sized/large esophageal varices of the variables presenting significant differences in the multivariate analysis
Test All patients (n = 345) PHT-US 87.9 Platelets (cutoff 77.2 9 120 × 10 cells/l) Plasma albumin (cutoff 44.8 3.14 mg/dl) 96.5 Most favorable groupa Child A patients (n = 172) PHT-US 83.3 Plasma albumin (cutoff 54.8 3.6 mg/dl) 92.9 Most favorable groupb Child B + C patients (n = 181) 76.5 Platelets (cutoff 117 × 109 cells/l)
NPV (%)
LR +
LR −
Patients with negative tests [n (%)]
(80.1–93.2) 34.9 (28.8–41.5) 40.64 (34.5–497.0) (68.4–88.5) 44.5 (38.4–51.7) 41.3 (34.5–48.2)
85.1 (76.3–91.6) 79.7 (71.7–86.3)
1.4 (1.2–1.5) 1.4 (1.2–1.6)
0.4 (0.2–0.6) 0.5 (0.4–0.8)
95 (27) 128 (37)
(35.6–54.3)
69 (62.6–74.9)
42.28 (33.4–51.5)
71.2 (64.7–77.0)
1.5 (1.1–1.9)
0.8 (0.7–1.0)
222 (64)
(91.3–99.2)
18.8 (13.9–24.5)
37.4 (31.9–43.2)
91.5 (79.6–97.6)
1.2 (1.1–1.3)
0.2 (0.1–0.5)
47 (14)
(68.6–93.0) 50.8 (41.7–59.8) (38.7–70.2) 68.3 (54.4–76.3)
36.1 (26.6–46.5) 36.5 (24.5–49.6)
90.1 (80.0–95.9) 81.9 (73.9–88.7)
1.7 (1.35–2.1) 0.3 (0.2–0.7) 1.7 (1.2–2.5) 0.7 (0.5–0.9)
72 (43) 105 (63)
(80.5–98.4) 32.5 (24.5–41.5)
31.5 (23.4–40.4)
93.2 (81.3–98.49) 1.4 (1.2–1.6)
(64.6–85.9)
48.2 (38.4–58.0)
S (%)
Sp (%)
42.9 (32.9–53.3)
PPV (%)
72.41 (59.1–83.3)
1.3 (1.1–1.7)
0.2 (0.1–0.7)
62 (37)
0.6 (0.3–0.9)
58 (33)
LR − , negative likelihood ratio; LR + , positive likelihood ratio; NPV, negative predictive value; PHT-US, presence of ultrasound signs of portal hypertension; PPV, positive predictive value; S, sensitivity; Sp, specificity. a Absence of US-portal hypertension, albumin, and platelets above the mean. b Absence of US-portal hypertension and albumin above the mean.
albumin concentration. At levels above 3.6 mg/dl, plasma albumin had a negative predictive value for the presence of LV of 81.9%, which could reduce the number of screening endoscopies by 63% (OR: 2.6; 95% CI: 1.27–5.31; P < 0.0087). The absence of signs of portal hypertension together with a plasma albumin level above 3.6 mg/dl had a negative predictive value of the presence of LV of 93% (OR: 6.56; 95% CI: 2.42–17.80; P < 0.0002), which would exclude 37% of patients from the endoscopic screening program, but would misclassify 71 patients. In this subgroup, platelets did not have independent predictive value.
Differences in the usefulness of abdominal ultrasound in decompensated patients
In decompensated patients, the only variable with an independent predictive value of the presence of LV was platelet count: values above 117 × 109 cells/l (OR: 2.44; 95% CI: 1.22–4.85; P < 0.0093) had a negative predictive value of 72%, which would allow 35% of decompensated cirrhotic patients to be excluded from the endoscopic screening program. In this patient subgroup, neither abdominal US nor any other variable could predict the presence or absence of LV.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Accuracy of ultrasound in varices Sort et al. 1339
Discussion Our study shows that abdominal US has a high sensitivity and a high negative predictive value for the detection of LV in patients with liver cirrhosis. These results can be improved by combining this radiological examination (which is available at almost all medical centers) with basic analytical parameters. In patients with cirrhosis as a whole, the use of US can reduce the need for screening endoscopies by 30%. The reduction is achieved mainly in patients with compensated cirrhosis, in whom 43% of the examinations may be avoided; in decompensated cirrhosis, it does not seem to yield any benefit. Significantly, this approach does not increase costs because abdominal US has to be performed every 6 months in patients with liver cirrhosis as screening for the appearance of hepatocellular carcinoma [19], and the US results are as reliable as those obtained with other more innovative diagnostic methods such as TE. Since the development of effective drugs for the prevention of upper digestive tract bleeding because of portal hypertension, it has been proposed that patients with liver cirrhosis should undergo periodical endoscopic screening to detect the cases most likely to present LV. In fact, in 1996, the American Association for the Study of Liver Diseases (AASLD) recommended endoscopic screening in all patients with decompensated cirrhosis or with compensated cirrhosis and a portal diameter greater than 13 mm or thrombocytopenia (< 140 × 109 cells/l) [20]. The results of our study clearly support these criteria. In addition, cirrhotic patients should also be screened for varices at diagnosis of the disease according to the Baveno V consensus workshop; although not specifically designed to confirm this recommendation, the very high negative predictive value of US obtained in our study may not support this approach [6]. The AASLD’s recommendation led to a series of investigations to determine whether the requirement for endoscopic screening could be restricted even further in these patients. Some of these studies have used US data, but none have specifically studied the usefulness of abdominal US. We reviewed three studies designed to identify cirrhotic patients who would not require UGE for screening [8,9,21]. All three were carried out in patients
Table 4
with compensated cirrhosis or who had never presented bleeding. They found that between 20 and 45% of patients had LV, in agreement with the rate of 35% observed in our study. These studies conclude that platelet concentration is an independent predictive variable of the presence of LV, but differ with respect to the importance of other factors such as splenomegaly [8], the presence of spiders [9], or liver function parameters [9,21]. Although none of these studies was designed to assess the diagnostic accuracy of US, the study by Schepis et al. [21] found that a portal diameter above 13 mm was associated independently with the presence of LV. These results are in agreement with ours; however, our study assessed the dichotomous variable presence/ absence of portal hypertension on US and did not evaluate separately each of the US findings that indicate the presence of portal hypertension (portal diameter, splenomegaly, ascites, and collateral circulation). The study of Schepis and colleagues also included only compensated patients. The results of these three studies are in agreement with ours in terms of the possibility of reducing the number of screening endoscopies performed in patients with liver cirrhosis by between 30 and 40% (Table 4). However, it has been suggested that these values could be improved even further by the use of more modern diagnostic methods such as TE, a noninvasive technique based on the measurement of liver stiffness as a marker of cirrhosis progression. Although the results obtained with TE are very promising, in its clinical guide for 2007 [22], the AASLD did not consider that the procedure was sufficiently consolidated to modify the indications of endoscopic screening of LV currently in place. Several studies have confirmed the high diagnostic accuracy of TE for screening for the existence of LV [10–13]; its predictive value ranges between 93% [10] and 100% [13] in different patient groups, and it was found to have a higher diagnostic value than splenomegaly, platelet concentration, or other analytical procedures [10]. However, these studies analyzed highly heterogeneous groups of patients (including compensated and decompensated patients, with or without previous upper digestive tract bleeding, patients coinfected with HIV); the cutoff points used do
Main results of the studies in the literature
References This study Chalasani et al. [8] Pilette et al. [9] Schepis et al. [21] Kazemi et al. [10] Bureau et al. [11] Castéra et al. [12] Montes Ramirez et al. [13]
Quoted
n
LV (%)
Predictive variables
Unnecessary UGE (%)
– 7 8
453 346 143
Type of patients Child A/B/C No UGIB No UGIB
35 20 45
27 28 –
15 9 10 11 12
143 165 150 70 85
Compensated Compensated Child A/B/C Compensated HIV +
20 28 48 19 20
US, platelets, albumin Splenomegaly, platelets Platelets, prothrombin time, spiders Platelets NPV 93% – NPV 84% NPV 100%
43 30 26 15–25
LV, large varices; No UGIB, no upper gastrointestinal bleeding; NPV, negative predictive value; UGE, upper gastrointestinal endoscopy; US, ultrasound.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
1340
European Journal of Gastroenterology & Hepatology 2014, Vol 26 No 12
not coincide, varying from 19 kPa [10] to 30.5 kPa [12], and in fact, the percentage of endoscopies avoided is no higher than that reported in our study. Our study shows that abdominal US has a high negative predictive value in screening for LV in patients with cirrhosis. Although other studies had already shown this capacity, the use of the dichotomous variable presence/ absence of signs of portal hypertension may have made the variable more robust and may have increased its negative predictive value, which reached 93.2% in the most favorable group. Our results also show that there is little difference between the use of abdominal US and TE, and that, as abdominal US is performed routinely every 6 months in patients with liver cirrhosis, the economic cost involved will be lower. Finally, almost all the studies indicate that ∼ 30% of screening endoscopies could be avoided if a method was used to detect patients at a low risk of having LV. In our study, this proportion increased to 43% in the subgroup of patients with compensated liver cirrhosis, although in patients with decompensated liver cirrhosis, US was not effective. The aim of all these studies is to reduce the number of endoscopic examinations. The impact of the application of combined criteria (US and analytical data) on the reduction in the total number of endoscopies performed in our study is limited (a reduction of 14%), but if we apply US criteria alone, the number of explorations avoided increases to 27%. The rate of misclassification would increase from 8.5 to 14.9%, but if we bear in mind the estimates that each year at most 15% of patients with LV will present digestive tract bleeding, the actual clinical impact of the increase in patients classified incorrectly is very low. Our study presents a number of limitations. First, the definition of cirrhosis is based in most cases on clinical and US data, although it is well known that some patients with liver cirrhosis may present normal US and analytical tests. Therefore, the number of patients with Child stage A cirrhosis may be undervalued. However, other studies also show similar (or even lower) proportions of patients with preserved liver function [7,10], and so, our study population appears to be comparable. Second, the study evaluates the presence of portal hypertension according to US findings, but, as is well known, some patients with portal hypertension measured by liver hemodynamic studies do not present US alterations [17]. There is no doubt that a hemodynamic study would make the results more robust, but not all medical centers have the necessary equipment. Finally, the retrospective nature of the study gives rise to a range of limitations: (a) the evaluations performed by the different endoscopists of the size of the varices were not compared; inter-rater variability in the classification of varices is known to be large, and this may have influenced the results. (b) The effect of an US-based screening program on the natural
evolution of the disease cannot be assessed, nor its effect on the costs deriving from this screening. (c) We cannot directly compare US and TE; at present, we can only speculate about the results and costs of the two approaches. Conclusion
The results of our study indicate that (a) the absence of US signs of portal hypertension is a highly reliable indicator of a low risk of presenting LV. (b) The expected reduction in the number of screening endoscopies when US criteria are used is ∼ 30% in patients with cirrhosis as a whole and over 40% in patients with compensated disease. (c) The negative predictive value of US in patients with compensated cirrhosis increases when it is combined with analytical data, but this combination markedly reduces the number of endoscopies avoided: with the current pressure on health services to reduce costs, it may be acceptable to use US as the sole screening technique as the expected number of episodes of bleeding is very low. (d) As the results obtained with US and TE are similar, comparative studies of their cost effectiveness are needed that also include a comparison with endoscopy as the gold standard in the screening of LV. (e) In compensated cirrhosis, the timing of endoscopic screenings should be flexible and should be guided by the results of the analytical tests and abdominal US.
Acknowledgements The authors thank Dr Rafael Fernández and Dr Jesús Montesinos for their critical review of the manuscript. They also thank Anna Arnau for her statistical support. Conflicts of interest
There are no conflicts of interest.
References 1
2
3
4
5
6
7
D’Amico G. Natural history of cirrhosis and prognosis. In: Arroyo V, Abraldes JG, Gines P, Sánchez-Tapias J, Forns X, Bataller R, et al., editors. Treatment of liver diseases. Barcelona: Grupo Ars XXI de comunicacion SL; 2009. pp. 143–152. Pagliaro L, D’Amico G, Pasta L, Politi F, Vizzini G, Traina M, et al. Portal hypertension in cirrhosis: natural history. In: Bosch J, Goszmann RJ, editors. Portal hypertension. Pathophysiology and treatment. Oxford, UK: Blackwell Scientific; 1994. pp. 72–92. North Italian Endoscopic Club for the Study and Treatment of Esophageal Varices. Prediction of the first variceal hemorrhage in patients with cirrhosis of the liver and esophageal varices. A prospective multicenter study. N Engl J Med 1988; 319:983–989. De Franchis R. Updating consensus in portal hypertension: report of the Baveno III consensus workshop on definitions, methodology and therapeutic strategies in portal hypertension. J Hepatol 2000; 33:846–852. D’Amico G, Garcia-Tsao G, Cales P, Escorsell A, Nevens F, Cestari R, et al. Diagnosis of portal hypertension: how and when. In: De Franchis R, editor. Portal hypertension III. Proceedings of the third Baveno international consensus workshop on definitions, methodology and therapeutic strategies. Oxford, UK: Blackwell Science; 2001. pp. 36–64. De Franchis R. Baveno V Faculty. Revising consensus in portal hypertension: report of the Baveno V consensus workshop on methodology of diagnosis and therapy in portal hypertension. J Hepatol 2010; 53:762–768. Calès P, Desmorat H, Vinel JP, Caucanas JP, Ravaud A, Gerin P, et al. Incidence of large esophageal varices in patients with cirrhosis. Application to prophylaxis of first bleeding. Gut 1990; 31:1298–1302.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Accuracy of ultrasound in varices Sort et al. 1341
8 Chalasani N, Imperiale T, Ismail A, Sood G, Carey M, Wilcox M, et al. Predictors of large esophageal varices in patients with cirrhosis. Am J Gastroenterol 1999; 94:3285–3291. 9 Pilette C, Oberti F, Aube C, Rousselet MC, Bedossa P, Gallois Y, et al. Noninvasive diagnosis of esophageal varices in chronic liver diseases. J Hepatol 1999; 31:867–873. 10 Kazemi F, Kettaneh A, N’kontchou G, Pinto E, Ganne-Carrie N, Trinchet JC, Beaugrand M. Liver stiffness measurement selects patients with cirrhosis at risk of bearing large esophageal varices. J Hepatol 2006; 45:230–235. 11 Bureau C, Metivier S, Peron JM, Selves J, Robic MA, Gourrauds PA, et al. Transient elastography accurately predicts presence of significant portal hypertension in patients with chronic liver disease. Aliment Pharmacol Ther 2008; 27:1261–1268. 12 Castéra L, Le Bail B, Roudot-Thoraval F, Bernard PH, Fourcher J, Merrouche W, et al. Early detection in routine clinical practice of cirrhosis and esophageal varices in chronic hepatitis C: comparison of transient elastography (fibroscan) with standard laboratory tests and non-invasive scores. J Hepatol 2009; 50:59–68. 13 Montes Ramirez M, Pascual-Pareja J, Sánchez-Conde M, Bernardino JI, Zamora Vargas FX, Miralles P, et al. Transient elastography to rule out esophageal varices and portal hypertensive gastropathy in HIV-infected individuals with liver cirrhosis. AIDS 2012; 26:1807–1812. 14 Cottone M, D’Amico G, Maringhini A, Amuso M, Sciarrino E, Traina M, et al. Predictive value of ultrasonography in the screening of non-ascitic cirrhotic patients with large varices. J Ultrasound Med 1986; 5:189–192. 15 Rector WG Jr, Campra J, Ralls PW, Charms M. Utility and limitations of splanchnic venous ultrasonography in diagnosis of portal hypertension. J Clin Ultrasound 1986; 14:689–696.
16
17
18
19
20
21
22
Macías-Rodríguez MA, Rendón-Unceta P, Martínez-Sierra MC, TeyssiereBlas I, Diaz-García F, Martín-Herrera L. Prognostic usefulness of ultrasonographic signs of portal hypertension in patients with Child–Pugh A liver cirrhosis. Am J Gastroenterol 1999; 94:3595–3600. Vilgrain V, Lebrec D, Menu Y, Scherrer A, Nahum H. Comparison between ultrasonographic signs and the degree of portal hypertension in patients with cirrhosis. Gastrointest Radiol 1990; 15:218–222. Schrier RW, Arroyo V, Bernardi M, Epstein M, Henriksen JGH, Rodés J. Peripheral arteriolar vasodilation hypothesis: a proposal for the initiation of renal sodium and water retention in cirrhosis. Hepatology 1988; 8:1151–1157. European Association For The Study Of The Liver; European Organisation For Research And Treatment Of Cancer. EASL-EORTC: clinical practice guidelines; management of hepatocellular carcinoma. J Hepatol 2012; 56:908–943. Grace ND, Groszmann RJ, Garcia-Tsao G, Burroughs AK, Pagliaro L, Makuch RW, et al. Portal hypertension and variceal bleeding: an AASLD single topic symposium. Hepatology 1998; 28:868–880. Schepis F, Cammà C, Niceforo D, Magnano A, Pallio S, Cinquegrani M, et al. Which patients with cirrhosis should undergo endoscopic screening for esophageal varices detection? Hepatology 2001; 33:333–338. Garcia-Tsao G, Sanyal A, Grace N, Carey W. Practice Guidelines Committee of the American Association for the Study of Liver Diseases; Practice Parameters Committee of the American College of Gastroenterology. Prevention and management of gastroesophageal varices and variceal hemorrhage in cirrhosis. Hepatology 2007; 46:922–938.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Diagnostic accuracy of abdominal ultrasound in the screening of esophageal varices in patients with cirrhosis Pau Sorta, Magdalena Muelasb, Alvaro Isavaa, Jordina Llaóa, Francesc Portaa, Ignasi Puiga, Claudia Domínguez-Curellc, Enrique Esteveb, Carles Yanguasd and Francesc Vidaa Background Abdominal ultrasound (US) may provide data on the presence of esophageal varices in cirrhosis. We assess the diagnostic accuracy of this procedure. Patients and methods Retrospective recording of clinical data was carried out in cirrhotic patients who underwent abdominal US and upper gastrointestinal endoscopy. We compared patients with and without large varices and assessed the value of US in predicting the presence of these lesions as well as other significant variables. Results Of the 353 patients included, 123 (35%) had esophageal varices. The presence of US signs of portal hypertension independently predicted the existence of esophageal varices with a sensitivity of 87.9%, a specificity of 34.9%, a positive predictive value of 40.6%, and a negative predictive value of 85.1%, which could increase to 91.5% if the patient presented plasma albumin and platelet concentrations above the mean values (3.1 g/dl and 122 × 109 cells/l, respectively). Plasma albumin and platelet concentrations were the two other variables with independent predictive capacity. Applying these selection criteria, up to 30% of screening endoscopies may not be necessary, and up to 43% in patients with compensated
Introduction Cirrhosis is an irreversible disease of the liver. In the Child–Pugh classification, it is divided into three stages according to the deterioration of liver function: stage A, in which patients are mainly asymptomatic; stage B, the intermediate stage; and stage C, the terminal stage of the disease [1]. Upper gastrointestinal bleeding because of gastroesophageal varices is a frequent complication of cirrhosis. It is estimated that 50% of patients with cirrhosis present varices, ranging from 40% in patients with compensated cirrhosis to 85% in those with a highly deteriorated liver function [2]. The 1-year probability of bleeding (5–15%) is especially high in patients with large varices (LVs) with red wale markings on the surface [3]. For this reason, the current recommendation is to identify cirrhotic patients with a major risk of bleeding by upper gastrointestinal endoscopy (UGE) at regular periods: at the time of diagnosis, yearly in patients with decompensated 0954-691X © 2014 Wolters Kluwer Health | Lippincott Williams & Wilkins
cirrhosis. In patients with decompensated cirrhosis, however, US does not have predictive capacity. The results obtained are comparable with those reported for transient elastography. Conclusion Abdominal US is a highly reliable technique for detecting patients with a low risk of presenting esophageal varices. Its use may avoid up to 43% of screening endoscopies in patients with compensated cirrhosis. The results obtained are similar to those observed using transient elastography. Eur J Gastroenterol Hepatol 26:1335–1341 © 2014 Wolters Kluwer Health | Lippincott Williams & Wilkins. European Journal of Gastroenterology & Hepatology 2014, 26:1335–1341 Keywords: esophageal varices, liver cirrhosis, portal hypertension, ultrasound a Digestive Diseases Service, bInternal Medicine Service, cIntensive Care Service and dDiagnostic Imaging Service, Althaia Xarxa Hospitalaria i Universitària de Manresa, Universitat Internacional de Catalunya, Manresa, Spain
Correspondence to Pau Sort, PhD, Hospital Sant Joan de Deu, c/Dr Joan Soler 1-3, 08243 Manresa, Barcelona, Spain Tel: + 34 93 874 21 12; fax: + 34 93 8744554; e-mail: [email protected] Received 29 April 2014 Accepted 30 June 2014
cirrhosis, and every 2–3 years in the case of compensated cirrhosis [4–6]. However, endoscopic screening programs are costly and UGE is not always well received by patients [7]; therefore, it seems sensible to limit its use to patients with a higher risk of bleeding. Some studies have evaluated alternative methods for predicting the presence of LVs without performing UGE [8–13]. However, their results are inconclusive either because the main alternative diagnostic tool [transient elastography (TE)] is not available at many centers, because the population studied did not reflect the heterogeneity of cirrhotic patients (e.g. only Child–Pugh group A or patients awaiting liver transplantation were included), or because the results did not reach statistical significance. The aim of this study is to establish whether abdominal ultrasound (US), a technique available at all health facilities, can identify cirrhotic patients who are less likely to present gastroesophageal varices and who therefore need not be included in screening programs. DOI: 10.1097/MEG.0000000000000174
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
1336
European Journal of Gastroenterology & Hepatology 2014, Vol 26 No 12
Objectives The aims of this work were to study the accuracy of abdominal US in discriminating between cirrhotic patients with and without LV, to define the variables that identify patients with the lowest risk of bleeding in both compensated and decompensated cirrhosis, and to establish whether their application might help to reduce the number of endoscopies performed in the screening of LV.
Patients and methods Patients
Clinical, laboratory, radiological, and endoscopic data were retrospectively collected from 469 consecutive cirrhotic patients who had consulted our outpatient clinic for any reason or had been admitted to our hospital, a secondary-level health facility, between January 2011 and June 2013. Cirrhosis was diagnosed when liver biopsy performed for any reason indicated typical features of this disease, when macroscopically cirrhotic liver was incidentally found during abdominal surgery, or when US or computed tomographic scan findings (heterogeneous liver parenchyma with hypertrophic caudate lobe and nodular margins) coupled with analytical (low platelet or leukocyte count, anemia, abnormal prothrombin time, bilirubin or albumin serum concentration) or clinical data (spider naevi, ascites, hepatic encephalopathy) suggested this disease. Abdominal US was performed at the time of evaluation in all cases. All patients who, in addition to the abdominal US, had undergone UGE within 3 months of the US were included in the study. Patients with suspected or confirmed alcoholic hepatitis or end-stage hepatocarcinoma were excluded. Finally, the study sample comprised 353 patients. The reasons for not performing UGE in the remaining 101 patients were as follows: severe comorbidity (such as invalidating cardiac or pulmonary failure; 35 cases), advanced age [21 cases; mean age: 84.56 years, 95% confidence interval (CI): 82.90–86.11], patients who did not return to the outpatient unit (9 cases), refusal (6 cases), and others (30 cases). Fifteen patients were excluded from the study because of endstage hepatocarcinoma (10 cases) and acute alcoholic hepatitis (5 cases). Abdominal ultrasound
All the radiology studies were performed by five radiologists with considerable experience with US of the liver and who were blinded to the results of the complementary examinations. Following previous publications, we considered that the presence of a portal diameter higher than 12 mm, a spleen size larger than 12 cm, perigastric or perisplenic collateral circulation, or ascites indicated the existence of portal hypertension and, when present, two or more of these US findings were associated in the vast majority of patients [14–16].
Previous studies have shown that the sensitivity of US in detecting portal hypertension, as estimated by hepatic vein catheterization, is about 40% with a sensitivity higher than 80% considering the presence of portosystemic venous collaterals [17]. In cirrhosis, ascites occurs only in the presence of portal hypertension [18]. Upper gastrointestinal endoscopy
UGE was performed by five gastroenterologists with wide experience in digestive endoscopy and who were blinded to the results of the complementary examinations. The examinations were performed under topical anesthesia with lidocaine or light sedation with propofol, and the presence or absence of gastroesophageal varices and their size was recorded. Varices were staged in accordance with the criteria published elsewhere [3]. According to the UGE results, patients were classified into two groups: those with medium-sized or large gastric or esophageal varices with signs of risk of bleeding (group LV + ) and those with small varices without signs of risk of bleeding or only portal hypertensive gastropathy, or who did not present endoscopic lesions that could be attributed to portal hypertension (group LV − ). Statistical analysis
The study compares the characteristics of all the patients divided according to the presence or absence of LV, and of the subgroups of patients with compensated and decompensated cirrhosis. The variables analyzed were age, sex, etiology of the cirrhosis, presence of ascites, prothrombin time, platelet count, albumin, bilirubin, Child–Pugh score, presence of splenomegaly, presence of US signs indicating portal hypertension, and presence of LV. Qualitative variables were compared using Pearson’s χ2-test. Quantitative variables were compared using the Student t-test if the variances between the two groups did not show statistically significant differences on the Levene test and also showed a normal distribution in the Kolmogorov–Smirnov test; if these two conditions were not fulfilled, a nonparametric test (the Mann–Whitney U-test) was used. Parameters that reached statistical significance in the univariate analysis were included in a forward stepwise logistic regression model. Quantitative variables that independently predicted the existence of LV were dichotomized according to their mean values. We could thus calculate the sensitivity, specificity, and positive and negative predictive values, and odds ratio (OR) for the existence of LV for both qualitative and quantitative variables. To obtain an accuracy of 7% for estimation of the sensitivity of abdominal US to detect LV in cirrhotic patients, with a 95% CI, and assuming a sensitivity of 84% [10] compared with UGE as the benchmark, at least 106 cirrhotic patients with LV had to be included. Assuming
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Accuracy of ultrasound in varices Sort et al. 1337
a prevalence of large esophageal varices of 30%, the minimum sample size required was 353 cirrhotic patients. The software used was SPSS (SPSS Inc., Chicago, Illinois, USA). Quantitative variables are expressed as means and 95% CIs. Differences were considered significant when the P value was less than 0.05. The protocol was approved by our hospital’s independent Clinical Research Ethical Committee. The study fulfilled all the legal requirements and local regulations and those of the Declaration of Helsinki.
Results Patients’ characteristics
Table 1 shows the characteristics of the patients included in the study. The majority of patients were men older than 60 years of age. Alcoholic and viral etiologies each accounted for close to 40% of cases, with the remaining 21% because of other causes. The majority of patients were Child–Pugh stage A (49%), followed by stage B (41%); 10% of patients were stage C (i.e. highly deteriorated liver function). At the time of inclusion, 35% of the patients had ascites, 73% showed US signs of portal hypertension, and 35% had LV. Ability of ultrasound to predict the existence of large varices in cirrhotic patients
Patients were divided into two groups (Table 2): group LV + (123 patients; 35%) and group LV − (230 patients; 61%). The groups were similar with respect to age, etiology of cirrhosis, and patients’ sex; however, the LV + group had significantly worse hepatic function, as estimated by serum bilirubin, prothrombin time, serum albumin, presence of ascites, and Child–Pugh score. Platelet count was also significantly lower in this group and a higher proportion of LV + patients showed US signs of portal hypertension (88 vs. 65% respectively;
Table 1
P < 0.000). The multivariate analysis showed that only three of the variables that had been significant in the univariate analysis were independent predictors of the presence of LV: US signs of portal hypertension (OR: 3.91; 95% CI: 2.01–7.28; P < 0.0001), platelet count (OR: 2.76; 95% CI: 1.66–4.60; P < 0.0001), and plasma albumin (OR: 1.81; 95% CI: 1.14–2.87; P < 0.012). The sensitivity, specificity, positive predictive value, negative predictive value, positive likelihood ratio, and negative likelihood ratio of US in the prediction of the presence or absence of esophageal varices are shown in Table 3. The results indicate that if abdominal US were used to exclude patients at a low risk of LV from the endoscopic screening program, the number of endoscopies with this indication would decrease by 27%, but 14.9% of patients would be misclassified. However, if patients with US signs of portal hypertension or platelets less than 120 × 109 cells/l or plasma albumin below 3.14 mg/dl are excluded from endoscopic screening, the sensitivity increases to 96.5% and the negative predictive value decreases to 91.5%. This would lead to a 14% reduction in the endoscopies performed to screen the esophageal varices, but would leave 8.5% of patients misclassified (OR: 6.41; 95% CI: 2.24–18.35; P < 0.0005). Differences in the utility of abdominal ultrasound in compensated patients
The results for patients with compensated cirrhosis (Child A) showed that the lack of US signs of portal hypertension predicted the absence of LV in a very high percentage of patients (negative predictive value 93%) (OR: 5.16; 95% CI: 2.13–12.50; P < 0.0003). The use of US as a criterion for excluding low-risk patients from endoscopic screening would reduce the number of endoscopies performed by 43%. The other variable that presented an independent predictive value in the univariate analysis was plasma
Patients’ baseline characteristics
Age (years) Male sex [n (%)] Etiology [n (%)] Alcohol Viral Others Prothrombin time (%) Serum bilirubin (mg/dl) Serum albumin (mg/dl) Platelets (×109 cells/l) Ascites [n (%)] Child–Pugh [n (%)] A B C Splenomegaly [n (%)] PHT-US [n (%)] LV + [n (%)]
Overall (n = 353)
Child A (n = 172)
Child B + C (n = 181)
62.54 (61.19–63.88) 236 (67)
61.6 (59.64–63.61) 112 (65)
63.07 (61.20–64.94) 124 (68)
153 140 60 73.20 1.78 3.14 121.8 123 172 145 36 207 257 123
(43) (40) (17) (71.19–75.21) (1.60–1.97) (3.05–3.23) (114.0–129.7) (35) (49) (41) (10) (59) (73) (35)
57 80 35 83.01 1.14 3.59 127.9 0
(33) (47) (20) (80.88–85.15) (1.04–1.23) (3.50–3.68) (117.5–138.2) (0)
– – – 93 (54) 99 (58) 46 (27)
96 60 25 65.11 2.45 2.73 118.5 123
(53) (33) (14) (62.57–67.65) (2.11–2.78) (2.60–2.86) (106.6–130.4) (68)
– – – 114 (63) 158 (87) 77 (43)
LV + , patients with medium-sized or large esophageal varices; PHT-US, ultrasound signs of portal hypertension, as described in the Patients and methods section.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
1338
European Journal of Gastroenterology & Hepatology 2014, Vol 26 No 12
Table 2 Variables presenting statistically significant differences in the univariate analysis between patients with esophageal varices (LV + ) and without (LV − ) LV + (n = 123) All patients Prothrombin time (%) Serum bilirubin (mg/dl) Serum albumin (mg/dl) Platelets (×109 cells/l) Ascites [n (%)] Child–Pugh [n (%)] A B C Splenomegaly [n (%)] PHT-US [n (%)] Child A Prothrombin time (%) Serum bilirubin (mg/dl) Platelets (×109 cells/l) Splenomegaly [n (%)] PHT-US [n (%)] Child B + C Platelets (×109 cells/l) Splenomegaly (n = 111) [n (%)]
69.23 2.14 2.97 97.7 55 46 56 22 94 108 79.83 1.26 118.4 35 42
LV − (n = 230)
(66.50–71.93) (1.75–2.53) (2.82–3.12) (87.5–108.0) (45)
75.10 1.61 3.23 134.1 68
(37) (46) (18) (77) (88)
126 89 14 113 149
(76.70–82.96) (1.04–1.48) (100.50–136.30) (76) (91)
85.27 0.90 141.2 59 66
101.1 (85.7–116.47) 59 (72)
P
(72.41–77.79) (1.41–1.81) (3.11–3.34) (123.7–144.4) (29)
0.002 0.016 0.008 0.000 0.011 0.000
(55) (38) (6) (51) (65)
0.000 0.000
(79.37–91.16) (0.79–1.00) (120.5–162.00) (47) (52)
0.047 0.036 0.001 0.003 0.000
153.9 (119.47–188.37) 54 (57)
0.000 0.024
LV + , presence of medium-sized or large esophageal varices; LV − , absence of medium-sized or large esophageal varices; PHT-US, presence of ultrasound signs of portal hypertension.
Table 3 Predictive capacity of the presence of medium-sized/large esophageal varices of the variables presenting significant differences in the multivariate analysis
Test All patients (n = 345) PHT-US 87.9 Platelets (cutoff 77.2 9 120 × 10 cells/l) Plasma albumin (cutoff 44.8 3.14 mg/dl) 96.5 Most favorable groupa Child A patients (n = 172) PHT-US 83.3 Plasma albumin (cutoff 54.8 3.6 mg/dl) 92.9 Most favorable groupb Child B + C patients (n = 181) 76.5 Platelets (cutoff 117 × 109 cells/l)
NPV (%)
LR +
LR −
Patients with negative tests [n (%)]
(80.1–93.2) 34.9 (28.8–41.5) 40.64 (34.5–497.0) (68.4–88.5) 44.5 (38.4–51.7) 41.3 (34.5–48.2)
85.1 (76.3–91.6) 79.7 (71.7–86.3)
1.4 (1.2–1.5) 1.4 (1.2–1.6)
0.4 (0.2–0.6) 0.5 (0.4–0.8)
95 (27) 128 (37)
(35.6–54.3)
69 (62.6–74.9)
42.28 (33.4–51.5)
71.2 (64.7–77.0)
1.5 (1.1–1.9)
0.8 (0.7–1.0)
222 (64)
(91.3–99.2)
18.8 (13.9–24.5)
37.4 (31.9–43.2)
91.5 (79.6–97.6)
1.2 (1.1–1.3)
0.2 (0.1–0.5)
47 (14)
(68.6–93.0) 50.8 (41.7–59.8) (38.7–70.2) 68.3 (54.4–76.3)
36.1 (26.6–46.5) 36.5 (24.5–49.6)
90.1 (80.0–95.9) 81.9 (73.9–88.7)
1.7 (1.35–2.1) 0.3 (0.2–0.7) 1.7 (1.2–2.5) 0.7 (0.5–0.9)
72 (43) 105 (63)
(80.5–98.4) 32.5 (24.5–41.5)
31.5 (23.4–40.4)
93.2 (81.3–98.49) 1.4 (1.2–1.6)
(64.6–85.9)
48.2 (38.4–58.0)
S (%)
Sp (%)
42.9 (32.9–53.3)
PPV (%)
72.41 (59.1–83.3)
1.3 (1.1–1.7)
0.2 (0.1–0.7)
62 (37)
0.6 (0.3–0.9)
58 (33)
LR − , negative likelihood ratio; LR + , positive likelihood ratio; NPV, negative predictive value; PHT-US, presence of ultrasound signs of portal hypertension; PPV, positive predictive value; S, sensitivity; Sp, specificity. a Absence of US-portal hypertension, albumin, and platelets above the mean. b Absence of US-portal hypertension and albumin above the mean.
albumin concentration. At levels above 3.6 mg/dl, plasma albumin had a negative predictive value for the presence of LV of 81.9%, which could reduce the number of screening endoscopies by 63% (OR: 2.6; 95% CI: 1.27–5.31; P < 0.0087). The absence of signs of portal hypertension together with a plasma albumin level above 3.6 mg/dl had a negative predictive value of the presence of LV of 93% (OR: 6.56; 95% CI: 2.42–17.80; P < 0.0002), which would exclude 37% of patients from the endoscopic screening program, but would misclassify 71 patients. In this subgroup, platelets did not have independent predictive value.
Differences in the usefulness of abdominal ultrasound in decompensated patients
In decompensated patients, the only variable with an independent predictive value of the presence of LV was platelet count: values above 117 × 109 cells/l (OR: 2.44; 95% CI: 1.22–4.85; P < 0.0093) had a negative predictive value of 72%, which would allow 35% of decompensated cirrhotic patients to be excluded from the endoscopic screening program. In this patient subgroup, neither abdominal US nor any other variable could predict the presence or absence of LV.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Accuracy of ultrasound in varices Sort et al. 1339
Discussion Our study shows that abdominal US has a high sensitivity and a high negative predictive value for the detection of LV in patients with liver cirrhosis. These results can be improved by combining this radiological examination (which is available at almost all medical centers) with basic analytical parameters. In patients with cirrhosis as a whole, the use of US can reduce the need for screening endoscopies by 30%. The reduction is achieved mainly in patients with compensated cirrhosis, in whom 43% of the examinations may be avoided; in decompensated cirrhosis, it does not seem to yield any benefit. Significantly, this approach does not increase costs because abdominal US has to be performed every 6 months in patients with liver cirrhosis as screening for the appearance of hepatocellular carcinoma [19], and the US results are as reliable as those obtained with other more innovative diagnostic methods such as TE. Since the development of effective drugs for the prevention of upper digestive tract bleeding because of portal hypertension, it has been proposed that patients with liver cirrhosis should undergo periodical endoscopic screening to detect the cases most likely to present LV. In fact, in 1996, the American Association for the Study of Liver Diseases (AASLD) recommended endoscopic screening in all patients with decompensated cirrhosis or with compensated cirrhosis and a portal diameter greater than 13 mm or thrombocytopenia (< 140 × 109 cells/l) [20]. The results of our study clearly support these criteria. In addition, cirrhotic patients should also be screened for varices at diagnosis of the disease according to the Baveno V consensus workshop; although not specifically designed to confirm this recommendation, the very high negative predictive value of US obtained in our study may not support this approach [6]. The AASLD’s recommendation led to a series of investigations to determine whether the requirement for endoscopic screening could be restricted even further in these patients. Some of these studies have used US data, but none have specifically studied the usefulness of abdominal US. We reviewed three studies designed to identify cirrhotic patients who would not require UGE for screening [8,9,21]. All three were carried out in patients
Table 4
with compensated cirrhosis or who had never presented bleeding. They found that between 20 and 45% of patients had LV, in agreement with the rate of 35% observed in our study. These studies conclude that platelet concentration is an independent predictive variable of the presence of LV, but differ with respect to the importance of other factors such as splenomegaly [8], the presence of spiders [9], or liver function parameters [9,21]. Although none of these studies was designed to assess the diagnostic accuracy of US, the study by Schepis et al. [21] found that a portal diameter above 13 mm was associated independently with the presence of LV. These results are in agreement with ours; however, our study assessed the dichotomous variable presence/ absence of portal hypertension on US and did not evaluate separately each of the US findings that indicate the presence of portal hypertension (portal diameter, splenomegaly, ascites, and collateral circulation). The study of Schepis and colleagues also included only compensated patients. The results of these three studies are in agreement with ours in terms of the possibility of reducing the number of screening endoscopies performed in patients with liver cirrhosis by between 30 and 40% (Table 4). However, it has been suggested that these values could be improved even further by the use of more modern diagnostic methods such as TE, a noninvasive technique based on the measurement of liver stiffness as a marker of cirrhosis progression. Although the results obtained with TE are very promising, in its clinical guide for 2007 [22], the AASLD did not consider that the procedure was sufficiently consolidated to modify the indications of endoscopic screening of LV currently in place. Several studies have confirmed the high diagnostic accuracy of TE for screening for the existence of LV [10–13]; its predictive value ranges between 93% [10] and 100% [13] in different patient groups, and it was found to have a higher diagnostic value than splenomegaly, platelet concentration, or other analytical procedures [10]. However, these studies analyzed highly heterogeneous groups of patients (including compensated and decompensated patients, with or without previous upper digestive tract bleeding, patients coinfected with HIV); the cutoff points used do
Main results of the studies in the literature
References This study Chalasani et al. [8] Pilette et al. [9] Schepis et al. [21] Kazemi et al. [10] Bureau et al. [11] Castéra et al. [12] Montes Ramirez et al. [13]
Quoted
n
LV (%)
Predictive variables
Unnecessary UGE (%)
– 7 8
453 346 143
Type of patients Child A/B/C No UGIB No UGIB
35 20 45
27 28 –
15 9 10 11 12
143 165 150 70 85
Compensated Compensated Child A/B/C Compensated HIV +
20 28 48 19 20
US, platelets, albumin Splenomegaly, platelets Platelets, prothrombin time, spiders Platelets NPV 93% – NPV 84% NPV 100%
43 30 26 15–25
LV, large varices; No UGIB, no upper gastrointestinal bleeding; NPV, negative predictive value; UGE, upper gastrointestinal endoscopy; US, ultrasound.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
1340
European Journal of Gastroenterology & Hepatology 2014, Vol 26 No 12
not coincide, varying from 19 kPa [10] to 30.5 kPa [12], and in fact, the percentage of endoscopies avoided is no higher than that reported in our study. Our study shows that abdominal US has a high negative predictive value in screening for LV in patients with cirrhosis. Although other studies had already shown this capacity, the use of the dichotomous variable presence/ absence of signs of portal hypertension may have made the variable more robust and may have increased its negative predictive value, which reached 93.2% in the most favorable group. Our results also show that there is little difference between the use of abdominal US and TE, and that, as abdominal US is performed routinely every 6 months in patients with liver cirrhosis, the economic cost involved will be lower. Finally, almost all the studies indicate that ∼ 30% of screening endoscopies could be avoided if a method was used to detect patients at a low risk of having LV. In our study, this proportion increased to 43% in the subgroup of patients with compensated liver cirrhosis, although in patients with decompensated liver cirrhosis, US was not effective. The aim of all these studies is to reduce the number of endoscopic examinations. The impact of the application of combined criteria (US and analytical data) on the reduction in the total number of endoscopies performed in our study is limited (a reduction of 14%), but if we apply US criteria alone, the number of explorations avoided increases to 27%. The rate of misclassification would increase from 8.5 to 14.9%, but if we bear in mind the estimates that each year at most 15% of patients with LV will present digestive tract bleeding, the actual clinical impact of the increase in patients classified incorrectly is very low. Our study presents a number of limitations. First, the definition of cirrhosis is based in most cases on clinical and US data, although it is well known that some patients with liver cirrhosis may present normal US and analytical tests. Therefore, the number of patients with Child stage A cirrhosis may be undervalued. However, other studies also show similar (or even lower) proportions of patients with preserved liver function [7,10], and so, our study population appears to be comparable. Second, the study evaluates the presence of portal hypertension according to US findings, but, as is well known, some patients with portal hypertension measured by liver hemodynamic studies do not present US alterations [17]. There is no doubt that a hemodynamic study would make the results more robust, but not all medical centers have the necessary equipment. Finally, the retrospective nature of the study gives rise to a range of limitations: (a) the evaluations performed by the different endoscopists of the size of the varices were not compared; inter-rater variability in the classification of varices is known to be large, and this may have influenced the results. (b) The effect of an US-based screening program on the natural
evolution of the disease cannot be assessed, nor its effect on the costs deriving from this screening. (c) We cannot directly compare US and TE; at present, we can only speculate about the results and costs of the two approaches. Conclusion
The results of our study indicate that (a) the absence of US signs of portal hypertension is a highly reliable indicator of a low risk of presenting LV. (b) The expected reduction in the number of screening endoscopies when US criteria are used is ∼ 30% in patients with cirrhosis as a whole and over 40% in patients with compensated disease. (c) The negative predictive value of US in patients with compensated cirrhosis increases when it is combined with analytical data, but this combination markedly reduces the number of endoscopies avoided: with the current pressure on health services to reduce costs, it may be acceptable to use US as the sole screening technique as the expected number of episodes of bleeding is very low. (d) As the results obtained with US and TE are similar, comparative studies of their cost effectiveness are needed that also include a comparison with endoscopy as the gold standard in the screening of LV. (e) In compensated cirrhosis, the timing of endoscopic screenings should be flexible and should be guided by the results of the analytical tests and abdominal US.
Acknowledgements The authors thank Dr Rafael Fernández and Dr Jesús Montesinos for their critical review of the manuscript. They also thank Anna Arnau for her statistical support. Conflicts of interest
There are no conflicts of interest.
References 1
2
3
4
5
6
7
D’Amico G. Natural history of cirrhosis and prognosis. In: Arroyo V, Abraldes JG, Gines P, Sánchez-Tapias J, Forns X, Bataller R, et al., editors. Treatment of liver diseases. Barcelona: Grupo Ars XXI de comunicacion SL; 2009. pp. 143–152. Pagliaro L, D’Amico G, Pasta L, Politi F, Vizzini G, Traina M, et al. Portal hypertension in cirrhosis: natural history. In: Bosch J, Goszmann RJ, editors. Portal hypertension. Pathophysiology and treatment. Oxford, UK: Blackwell Scientific; 1994. pp. 72–92. North Italian Endoscopic Club for the Study and Treatment of Esophageal Varices. Prediction of the first variceal hemorrhage in patients with cirrhosis of the liver and esophageal varices. A prospective multicenter study. N Engl J Med 1988; 319:983–989. De Franchis R. Updating consensus in portal hypertension: report of the Baveno III consensus workshop on definitions, methodology and therapeutic strategies in portal hypertension. J Hepatol 2000; 33:846–852. D’Amico G, Garcia-Tsao G, Cales P, Escorsell A, Nevens F, Cestari R, et al. Diagnosis of portal hypertension: how and when. In: De Franchis R, editor. Portal hypertension III. Proceedings of the third Baveno international consensus workshop on definitions, methodology and therapeutic strategies. Oxford, UK: Blackwell Science; 2001. pp. 36–64. De Franchis R. Baveno V Faculty. Revising consensus in portal hypertension: report of the Baveno V consensus workshop on methodology of diagnosis and therapy in portal hypertension. J Hepatol 2010; 53:762–768. Calès P, Desmorat H, Vinel JP, Caucanas JP, Ravaud A, Gerin P, et al. Incidence of large esophageal varices in patients with cirrhosis. Application to prophylaxis of first bleeding. Gut 1990; 31:1298–1302.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Accuracy of ultrasound in varices Sort et al. 1341
8 Chalasani N, Imperiale T, Ismail A, Sood G, Carey M, Wilcox M, et al. Predictors of large esophageal varices in patients with cirrhosis. Am J Gastroenterol 1999; 94:3285–3291. 9 Pilette C, Oberti F, Aube C, Rousselet MC, Bedossa P, Gallois Y, et al. Noninvasive diagnosis of esophageal varices in chronic liver diseases. J Hepatol 1999; 31:867–873. 10 Kazemi F, Kettaneh A, N’kontchou G, Pinto E, Ganne-Carrie N, Trinchet JC, Beaugrand M. Liver stiffness measurement selects patients with cirrhosis at risk of bearing large esophageal varices. J Hepatol 2006; 45:230–235. 11 Bureau C, Metivier S, Peron JM, Selves J, Robic MA, Gourrauds PA, et al. Transient elastography accurately predicts presence of significant portal hypertension in patients with chronic liver disease. Aliment Pharmacol Ther 2008; 27:1261–1268. 12 Castéra L, Le Bail B, Roudot-Thoraval F, Bernard PH, Fourcher J, Merrouche W, et al. Early detection in routine clinical practice of cirrhosis and esophageal varices in chronic hepatitis C: comparison of transient elastography (fibroscan) with standard laboratory tests and non-invasive scores. J Hepatol 2009; 50:59–68. 13 Montes Ramirez M, Pascual-Pareja J, Sánchez-Conde M, Bernardino JI, Zamora Vargas FX, Miralles P, et al. Transient elastography to rule out esophageal varices and portal hypertensive gastropathy in HIV-infected individuals with liver cirrhosis. AIDS 2012; 26:1807–1812. 14 Cottone M, D’Amico G, Maringhini A, Amuso M, Sciarrino E, Traina M, et al. Predictive value of ultrasonography in the screening of non-ascitic cirrhotic patients with large varices. J Ultrasound Med 1986; 5:189–192. 15 Rector WG Jr, Campra J, Ralls PW, Charms M. Utility and limitations of splanchnic venous ultrasonography in diagnosis of portal hypertension. J Clin Ultrasound 1986; 14:689–696.
16
17
18
19
20
21
22
Macías-Rodríguez MA, Rendón-Unceta P, Martínez-Sierra MC, TeyssiereBlas I, Diaz-García F, Martín-Herrera L. Prognostic usefulness of ultrasonographic signs of portal hypertension in patients with Child–Pugh A liver cirrhosis. Am J Gastroenterol 1999; 94:3595–3600. Vilgrain V, Lebrec D, Menu Y, Scherrer A, Nahum H. Comparison between ultrasonographic signs and the degree of portal hypertension in patients with cirrhosis. Gastrointest Radiol 1990; 15:218–222. Schrier RW, Arroyo V, Bernardi M, Epstein M, Henriksen JGH, Rodés J. Peripheral arteriolar vasodilation hypothesis: a proposal for the initiation of renal sodium and water retention in cirrhosis. Hepatology 1988; 8:1151–1157. European Association For The Study Of The Liver; European Organisation For Research And Treatment Of Cancer. EASL-EORTC: clinical practice guidelines; management of hepatocellular carcinoma. J Hepatol 2012; 56:908–943. Grace ND, Groszmann RJ, Garcia-Tsao G, Burroughs AK, Pagliaro L, Makuch RW, et al. Portal hypertension and variceal bleeding: an AASLD single topic symposium. Hepatology 1998; 28:868–880. Schepis F, Cammà C, Niceforo D, Magnano A, Pallio S, Cinquegrani M, et al. Which patients with cirrhosis should undergo endoscopic screening for esophageal varices detection? Hepatology 2001; 33:333–338. Garcia-Tsao G, Sanyal A, Grace N, Carey W. Practice Guidelines Committee of the American Association for the Study of Liver Diseases; Practice Parameters Committee of the American College of Gastroenterology. Prevention and management of gastroesophageal varices and variceal hemorrhage in cirrhosis. Hepatology 2007; 46:922–938.
Copyright © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
