Detection of Esophageal Varices Using CT and MRI

Dig Dis Sci (2011) 56:2696 2700 DOI 10.1007/s10620-011-1660-8 ORIGINAL ARTICLE Detection of Esophageal Varices Using CT and MRI Michael J. Lipp Arkady Broder David Hudesman Pauline Suwandhi Steven A. Okon Mitchell Horowitz David J. Clain Patricia Friedmann Albert D. Min Received: 9 November 2010 / Accepted: 22 February 2011 / Published online: 6 March 2011 Ó Springer Science+Business Media, LLC 2011 Abstract Background and Aims The development of esophageal varices in cirrhotic patients carries a significant risk of hemorrhage and associated morbidity/mortality. Universal endoscopic screening, however, is invasive and expensive. Conversely, cirrhotic patients often have imaging findings which suggest portal hypertension. The aim of this study was to evaluate the ability of CT and/or MRI to detect esophageal varices compared to EGD. Methods Medical records from 2000 to 2007 were retrospectively reviewed. CT and/or MRI images were included if performed within 90 days of EGD. Two blinded, experienced radiologists were asked to review images for the presence of esophageal varices, as well as other findings associated with portal hypertension. Sensitivity, specificity, PPV and NPV were calculated using EGD findings as the gold standard. Results A total of 195 patients and 142 patients met criteria for CT and MRI, respectively. The sensitivity of CT to detect EGD varices was 58 89%, but increased to 65 100% when specifically looking at large endoscopic varices. Overall specificity was 68 82%, but increased to 97 100% when applying C4 mm varices criteria. CT was M. J. Lipp (&) D. Hudesman D. J. Clain P. Friedmann A. D. Min Division of Digestive Diseases, Beth Israel Medical Center, First Avenue at 16th Street, 17-BH, New York, NY 10003, USA e-mail: lipp50@gmail.com A. Broder P. Suwandhi Department of Medicine, Beth Israel Medical Center, New York, NY, USA S. A. Okon M. Horowitz Department of Radiology, Beth Israel Medical Center, New York, NY, USA superior to MRI in the detection of endoscopic varices; the addition of other portal hypertension stigmata did not improve results. Conclusions The exclusion of large endoscopic varices by CT, using standardized criteria, may obviate the need or frequency of EGD screening in select patient populations. Alternatively, CT findings highly suggestive of esophageal varices in cirrhotic patients may warrant further investigation and/or treatment. Further studies are needed to validate these findings. Keywords Esophageal varices Computerized tomography Magnetic resonance imaging Noninvasive Introduction The progression to cirrhosis is often accompanied by the development of portal hypertension, as an increase in portal pressure leads to the development of porto-systemic collaterals. Esophageal varices are perhaps the most clinically relevant porto-systemic collaterals. Because of their possible rupture and hemorrhage, esophageal varices remain the most common lethal complication of portal hypertension with a reported mortality rate as high as 20 30% at 6 weeks [1, 2]. The presence of gastroesophageal varices correlates with the Child-Turcotte-Pugh class, and ranges from 40 to 85% of patients with the highest rates in Child- Pugh class C cirrhotics [3]. Size is the most important predictor of variceal hemorrhage with large varices conferring the highest yearly risk of first hemorrhage in as many as 15% of patients [4] and a rebleeding risk (within 1 2 years) of 60% in previously untreated patients [5]. The gold standard in diagnosis of esophageal varices remains esophagogastroduodenoscopy (EGD). Variceal

Dig Dis Sci (2011) 56:2696 2700 2697 size is an important predictor of bleeding risk and should be quantified as either small or large, with a recommended cut-off diameter of 5 mm or the older morphological description of 2? or greater [6, 7]. EGD is recommended once the diagnosis of liver cirrhosis is established, with follow-up surveillance intervals ranging from 1 to 3 years, with the most frequent yearly follow-up recommended in patients with decompensated cirrhosis [8]. Invasive endoscopic examination, however, carries with it high costs and multiple procedure related risks, which are often augmented in patients with decompensated cirrhosis [9]. Therefore, there has been substantial interest in developing a reliable non-invasive method of accurately predicting the degree of cirrhosis and portal hypertension size such as the platelet count, spleen size, Fibrotest, Fibroscan, portal vein diameter, transient elastography and esophageal capsule endoscopy [10]. However, to date the predictive value of these markers remains unclear, with low sensitivities that often result in failure to identify large varices [6]. The presences of esophageal varices, or other findings associated with portal hypertension, are often found when dedicated liver CT and/or MRI are performed on cirrhotic patients for cancer screening or diagnostic indication [11, 12]. The clinical significance of these findings, however, is uncertain, in part because of the absence of insufflation during these studies compared to endoscopy. There have been several recent studies proposing CT and MRI imaging modalities as an alternative means of diagnosing esophageal varices [13 15]. The study by Kim et al. developed radiologic criteria, and concluded that a 3-mm screening threshold on CT was optimal for sensitivity, specificity, and accuracy. A study by Perri et al. used a CT cutoff of 5 mm, and concluded that CT could be a cost-effective screening test for esophageal varices. One of the critiques of this study, however, was that using CT as a primary screening modality would result in patients with large varices being misdiagnosed [16]. The aim of our study was to retrospectively evaluate the effectiveness of CT and/or MRI imaging modalities to detect esophageal varices, when compared to gold standard EGD findings in cirrhotic patients who had imaging performed for other indications. Methods Patients who presented to our medical center between the years of 2000 and 2007 were retrospectively evaluated. We included all patients 18 years and older who had undergone EGD, abdominal computed tomography (CT) with IV contrast and/or magnetic resonance imaging (MRI) with IV contrast. Imaging studies were included if performed within 90 days of EGD. All EGDs were performed for variceal surveillance in cirrhotic patients by at least one experienced endoscopist. The retrospective nature of our study created a scenario where several cut-offs were used to identify esophageal varix size. Investigators performing EGD classified varices using either the 5-mm cut-off [6], the older 2? numbering system [7], and the currently recommended small, medium and large grading system [6]. A grading of 2? corresponds to being greater than or equal to of the normal esophageal lumen being obstructed by varices. Patients who underwent esophageal banding prior to the obtained imaging were excluded. Patient data regarding age, gender, and date of examination were recorded. Radiologists were blinded to prior EGD reports, the patient mix, and were given a specific protocol to evaluate the findings on CT or MRI. The radiologists had 10 and 20 years of post-training experience, respectively, and were each subspecialists in abdominal imaging. The CT equipment used at our institution during the study period included the General Electric Lightspeed Plus-4 detector row, Siemens Emotion 16, and the Siemens Sensation 64-32 (64) detector row. Both dedicated multiphase liver protocol and routine single-phase CT studies were included. The MRI equipment used was the General Electric Signa HD 1.5 T MRI unit. MRI images were obtained with multi-phased dynamic enhancement, spoiled gradient, fat saturation, and were T1 weighted. Based on the prior study by Kim et al., which aimed to detect esophageal varices in cirrhotic patients, esophageal varices on imaging were defined as the probable presence of either discrete enhancing nodular lesions abutting the luminal surface of the esophageal wall or protruding into luminal space or ill defined nodular enhancing lesions appeared to be contacting or protruding into the lumen [14]. Based on these criteria, radiologists used a standardized form to evaluate the assigned images for the presence of varices and other stigmata of portal hypertension. Radiologists were initially asked to determine if the assigned imaging provided adequate views for the detection of esophageal varices. Studies lacking adequate views were excluded for each radiologist. Each radiologist determined the presence of esophageal varices and the approximate size by measuring the diameter of the largest observed varix. A quantitative cutoff for large varices seen on CT or MRI was chosen as greater than or equal to 4 mm in diameter based on previous studies [13, 15]. Other stigmata of portal hypertension evaluated included paraesophageal, gastric and perigastric varices, as well as a recanalized peri-umbilical vein and spleno-renal shunt. The primary endpoint of the study was the positive and negative predictive values (PPV and NPV) of CT and MRI in determining the presence of any and/or large esophageal varices. The identity of patients and their medical records were kept confidential, and all study procedures were

2698 Dig Dis Sci (2011) 56:2696 2700 compliant with the standards set forth by the institutional review board (IRB). Statistical Analysis Data were analyzed with the SAS system version 9.12. Descriptive data analyses were conducted. Sensitivity, specificity, positive and negative predictive values of CT with contrast and MRI readings were computed and expressed as percents with 95% confidence intervals. EGD findings were used as the gold standard. Results Two-hundred ninety-nine patients met the inclusion criteria and were included in the data analysis. Baseline characteristics of these patients are described in Table 1. The mean age of the patient population was 55.2, with a male to female distribution 64.9 and 35.1%, respectively. The overall prevalence of esophageal varices on EGD was 41.4% (124/299), with a 20.3% prevalence of large esophageal varices. Table 1 Baseline patient characteristics Baseline characteristics Value (N = 299) Mean age (years) 55.2 ± 10 Mean time EGD to imaging (days) 27.8 ± 24.4 Gender, n (%) Male 194 (64.9%) Female 105 (35.1%) Presence of endoscopic varices, n (% of patients) Large varices 61 (20.3) Small varices 63 (21.1) EGD esophagogastroduodenoscopy Imaging studies were included if views were deemed adequate by the radiologists (Fig. 1). Out of a total of 195 CT images, radiologist 1 included 138 studies (71%), and radiologist 2 included 165 (85%). Of 142 MRI images that met initial study criteria, radiologist 1 included 126 (89%) and radiologist 2 included 93 (66%). A comparison was made of each radiologist s ability to detect the presence of any varices on CT imaging as compared to EGD (Table 2). This revealed a sensitivity of 89 and 58%, respectively, for radiologists 1 and 2. When the 4-mm variceal size criteria was applied to the interpretation of varices seen on CT, a notable result was an increase in the specificity for radiologists 1 and 2, i.e., 97% and 100%, respectively. The 4-mm CT criteria was also associated with an increase in the positive predictive value of any esophageal varices to 93 and 100% for radiologist 1 and 2, respectively. The negative predictive value for anysized esophageal varices was 88% for radiologist 1 and 72% for radiologist 2. Further analysis was performed to evaluate the ability to detect the presence of large esophageal varices noted on EGD (C2?) using CT imaging (Table 3). Notable findings included 100% sensitivity for radiologist 1 when evaluating any for esophageal varices using CT. Radiologist 2 had a lower sensitivity of 65% using similar criteria. These findings were associated with high negative predictive values of 100% for radiologist 1 and 89% for radiologist 2. Thus, a negative CT reading was able to exclude large endoscopic varices in 100% of cases (59/59) for radiologist 1 and 89% of cases for radiologist 2 (95/107). When the 4-mm CT criteria was applied to the interpretation of large EGD varices, the specificity of radiologists 1 and 2 were 90 and 97%, respectively. These findings correlate with modest positive predictive values of 62% for radiologist 1 and 75% for radiologist 2. Data analysis comparing the radiologists ability to detect EGD varices using MRI images was also performed. Overall, the sensitivity and specificity for detecting the presence of any sized varices was poor when compared to Fig. 1 Radiology assessment of adequate CT and MRI images

Dig Dis Sci (2011) 56:2696 2700 2699 Table 2 Sensitivity, specificity, PPV, NPV of CT for any EGD varices Criteria Observer 1 Observer 2 Values 95% CI Values 95% CI Sensitivity (%) Any size 89 (54/61) 79 95 58 (41/71) 46 69 Size [4 mm 44 (27/61) 32 57 23 (16/71) 14 33 Specificity (%) Any size 68 (52/76) 57 78 82 (77/94) 73 89 Size [4 mm 97 (74/76) 92 100 100 (94/94) 97 100 Positive predictive value (%) Any size 69 (54/78) 58 79 71 (41/58) 58 81 Size [4 mm 93 (27/29) 79 99 100 (16/16) 83 100 Negative predictive value (%) Any size 88 (52/59) 78 95 72 (77/107) 63 80 Size [4 mm 69 (74/108) 59 77 63 (94/149) 55 71 CI confidence interval, PPV positive predictive value, NPV negative predictive value, CT computed tomography, EGD esophagogastroduodenoscopy Table 3 Sensitivity, specificity, PPV, NPV of CT for large EGD varices Criteria Observer 1 Observer 2 Values 95% CI Values 95% CI Sensitivity (%) Any size 100 (30/30) 90 100 65 (22/34) 48 79 Size [4 mm 60 (18/30) 42 76 36 (12/34) 21 52 Specificity (%) Any size 55 (59/107) 46 64 73 (95/131) 65 80 Size [4 mm 90 (96/107) 83 95 97 (127/131) 94 100 Positive predictive value (%) Any size 38 (30/78) 28 50 38 (22/58) 26 51 Size [4 mm 62 (18/29) 44 78 75 (12/16) 50 92 Negative predictive value (%) Any size 100 (59/59) 95 100 89 (95/107) 83 95 Size [4 mm 89 (96/108) 82 94 85 (127/149) 80 91 CI confidence interval, PPV positive predictive value, NPV negative predictive value, CT computed tomography, EGD esophagogastroduodenoscopy CT (sensitivities for radiologists 1 and 2 were 29 and 29%; specificities were 35 and 33%, respectively). Similarly, the positive and negative predictive values for each radiologist was low (PPV of 35 and 33% for radiologists 1 and 2; NPV 53 and 24%, respectively). The presence of other portal hypertension stigmata were evaluated for both CT and MRI imaging. This included the paraesophageal, gastric and perigastric varices, as well as a recanalized peri-umbilical vein and spleno-renal shunt. The addition of these findings, however, did not improve the ability to predict endoscopic Fig. 2 Computed tomography (CT) axial image of the lower esophagus showing a nodular, enhancing, intraluminally protruding 6-mm lesion within the esophageal wall (arrows) varices when compared to the imaging of esophageal varices alone (data not shown) (Fig. 2). Discussion Our results show the potential of CT imaging to detect and/ or exclude the presence of esophageal varices in patients with cirrhosis. CT imaging had a high positive predictive value for diagnosing any-sized esophageal varices (93 100%), as well as an impressive negative predictive value in ruling out large/clinically significant esophageal varices (89 100%). Additionally, our findings suggest that CT is a superior imaging modality to MRI in the detection of esophageal varices. Furthermore, the evaluation of portal hypertension stigmata on imaging, such as paraesophageal, gastric and perigastric varices, recanalized peri-umbilical vein and spleno-renal shunt, did not improve the predictive value when compared to the presence of esophageal varices alone. Our study has several limitations. There are no standardized radiology criteria to classify esophageal varices, and the criteria used in our study were based on those used by Kim et al. [15]. Thus, the radiologists in this study did not receive formal training in interpretation of varices. Additionally, there was variability in the interpretation of adequate image quality. To mimic clinical practice, radiologists were not asked to interpret images they felt had inadequate image quality. Unfortunately, since each radiologist did not evaluate every imaging study independently regardless of perceived quality, we were unable to determine intraobserver and interobserver variability. It is also important to note that although EGD findings were used as the gold standard, the fact that EGDs were performed by

2700 Dig Dis Sci (2011) 56:2696 2700 more that one endoscopist is problematic given the data indicating a lack of agreement between endoscopists in describing varices [17, 18]. Thus, weakness of EGD as a gold-standard for diagnosing esophageal varices is a limitation as well. There may have also been heterogeneity in our patient population, with potential physician bias in determining which patients should have a CT or MRI performed. Also, our database did not include potentially useful patient information such as the presence of portal vein thrombosis and differences between each protocol and type of CT machine. It would have been of interest to determine Childs Pugh scores; however, due to the limitations of a retrospective database we were unable to determine those scores. Finally this was a single center, retrospective study, with uncertain generalizability to other radiologists and patient populations without further studies. Despite these limitations, these results are clinically relevant given the need to develop a sensitive and noninvasive method of accurately predicting the presence of esophageal varices. This study differs from previous trials in that these imaging studies were not performed for the purpose of detecting esophageal varices. The primary indication for the majority of CT and/or MRIs was to evaluate for hepatocellular carcinoma. One potential advantage to evaluating images obtained for another purpose is that patients would not be subjected to additional risks of radiation exposure, and that no additional costs would be incurred. Clinicians would also benefit from understanding how a given radiographic interpretation may change the likelihood of esophageal varices in a given patient with cirrhosis. Thus, the exclusion of large varices by CT findings using standardized size criteria may obviate the need for EGD in select patient populations. Alternatively, CT findings highly suggestive of esophageal varices in cirrhotic patients may warrant further investigation and/ or treatment. In our study we used the radiology criteria developed by Kim et al. [14] using a 4-mm cutoff for large varices. Further studies are needed to test whether these results apply to other institutions, and these standardized criteria may be incorporated into radiology training programs. Given the frequency of CT imaging performed in the care of patients with cirrhosis, however, these findings suggest that there is considerable potential to reduce the costs and/ or frequency of EGD screening in this population. 2. Fontana RJ, Sanyal AJ, Ghany MG, et al. Factors that determine the development and progression of gastroesophageal varices in patients with chronic hepatitis C. Gastroenterology. 2010;138: 2321 2331. 3. Pagliaro L, D Amico G, Pasta L. Portal hypertension in cirrhosis: Natural history. In: Bosch J, Groszmann RJ, et al., eds. Portal Hypertension. Pathophysiology and Treatment. Oxford, UK: Blackwell Scientific; 1994:72 92. 4. El-Serag HB, Everhart JE. Improved survival after variceal hemorrhage over an 11-year period in the department of veterans affairs. Am J Gastroenterol. 2000;95:3566 3573. 5. Bosch J, Garcia-Pagan JC. Prevention of variceal rebleeding. Lancet. 2003;361:952 954. 6. Garcia-Tsao G, Sanyal AJ, Grace ND, et al. Practice Guidelines Committee of the American Association for the Study of Liver Diseases the 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. 7. Dagradi AE. Endoscopy of oesophageal varices. Gastroenterol Jpn. 1972;7:103 105. 8. 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. 9. Eisen GM, Baron TH, Dominitz JA, et al. Complications of upper GI endoscopy. Gastrointest Endosc. 2002;55:784 793. 10. de Franchis R. Non-invasive (and minimally invasive) diagnosis of oesophageal varices. J Hepatol. 2008;49:520 527. 11. Andersson KL, Salomon JA, Goldie SJ, et al. Cost effectiveness of alternative surveillance strategies for hepatocellular carcinoma in patients with cirrhosis. Clin Gastroenterol Hepatol. 2008;6: 1418 1424. 12. Perri RE, Chiorean MV, Fidler JL, et al. A prospective evaluation of computerized tomographic (CT) scanning as a screening modality for esophageal varices. Hepatology. 2008;47:1587 1594. 13. Kim SH, Kim YJ, Lee JM, et al. Esophageal varices in patients with cirrhosis: Multidetector CT esophagography comparison with endoscopy. Radiology. 2007;242:759 768. 14. Kim H, Choi D, Gwak GY, et al. Evaluation of esophageal varices on liver computed tomography: Receiver operating characteristic analyses of the performance of radiologists and endoscopists. J Gastroenterol Hepatol. 2009;24:1534 1540. 15. Kim YJ, Raman SS, Yu NC, et al. Esophageal varices in cirrhotic patients: Evaluation with liver CT. Am J Roentgenol. 2007;188: 139 144. 16. Thabut D, Moreau R, Lebrec D. Screening for esophageal varices: Endoscopy, other tools, or endoscopy and other tools? Hepatology. 2008;47:1434 1436. 17. Bendtsen F, Skovgaard LT, Sørensen TI, et al. Agreement among multiple observers on endoscopic diagnosis of esophageal varices before bleeding. Hepatology. 1990;11:341 347. 18. Cales P, Pascal JP. Gastroesophageal endoscopic features in cirrhosis: Comparison of intracenter and intercenter observer variability. Gastroenterology. 1990;99:1189. Conflict of interest None. References 1. Carbonell N, Pauwels A, Serfaty L, et al. Improved survival after variceal bleeding in patients with cirrhosis over the past two decades. Hepatology. 2004;40:652 659.