Despite improvement in the management of gastrointestinal hemorrhage related to portal hypertension, mortality after an episode of variceal bleeding is 20% at 6 weeks.1 Therefore, practice guidelines recommend screening for varices in all patients with cirrhosis because nonselective beta-blockers and band ligation reduce the risk of first bleeding by 50% in patients with medium or large varices.2, 3 To date, upper gastrointestinal endoscopy is the gold standard for screening and must be performed every 2 years in patients without varices and yearly in those with small varices.4 However, this technique is uncomfortable, invasive, and costly. Moreover, half of these patients will not develop varices after 10 years and therefore will undergo unnecessary screening. In addition, in a subgroup of patients with compensated stage 15 mainly hepatitis C cirrhosis, endoscopic screening should not be systematically performed because most of these patients are at low risk of developing esophageal varices as indicated by the hepatic venous pressure gradient below 10 mm Hg6 (D. Lebrec and R. Moreau, unpublished results). Thus, there is an urgent need for noninvasive screening methods for varices in patients with cirrhosis.
Although numerous studies have been performed to assess different tools to diagnose varices, several considerations must be kept in mind when evaluating these studies. Because the target patients have compensated cirrhosis, these tools should be validated in this population only. Because the prognostic value of small varices remains unknown, these diagnostic tools should diagnose medium to large varices. The sensitivity of the tool should be a primary factor, because failure to identify large varices may have serious consequences.
Except for portal hypertension, the mechanisms responsible for the development of varices remain unknown but several laboratory and radiological tests have been correlated with the presence of esophageal varices, for example: platelet count, prothrombin time, splenomegaly, portal vein diameter, ascites, or a combination of these markers.7–10 The diagnostic value of platelet count has been extensively studied. However, the optimal threshold varies markedly from study to study, and its diagnostic value in patients with compensated cirrhosis has been recently questioned.11 Prognostic models that included platelet count and other tests have shown promising results and were highly sensitive,7–10 but validation by additional studies is still required10 or have been disappointing.12 Recently, two noninvasive methods, the FibroTest13 and Fibroscan,14, 15 were used to screen esophageal varices. Cut-offs with high sensitivities should be identified for the FibroTest and be confirmed in a more selected population; results for the Fibroscan are controversial. For these reasons, the predictive accuracy of noninvasive methods is still considered to be insufficient in patients with compensated cirrhosis, and none of them have been recommended for use in clinical practice.4
In this issue of HEPATOLOGY, two noninvasive methods were tested for the detection of esophageal varices: The endoscopic capsule16 and abdominal computed tomography (CT).17 The first article describes a large multicenter study which assessed the diagnostic performance of the capsule using endoscopy as the gold standard. Two-thirds of the patients had compensated cirrhosis. Although the agreement between the esophageal capsule and endoscopy was good (86%), the observed difference between the two techniques was above 10%, so the study was considered negative. There was also good agreement (91%) between the two techniques in differentiating between patients without or with small varices and patients with large varices, but the sensitivity was only 78%. The capsule was safe and was preferred by significantly more patients. The second study is a monocentric prospective study assessing the performance of CT in the detection of esophageal varices. Results were interpreted by two radiologists with at least 10 years experience, whereas the endoscopists were not necessarily experienced. Patients were asked to choose the procedure they preferred, and the cost-effectiveness of three strategies—endoscopy alone, CT alone, or CT followed by endoscopy in patients with small varices at CT—was evaluated. When large varices were found upon the performance of endoscopy, CT had 90% sensitivity in detecting varices irrespective of their size, but only 60% sensitivity in detecting large varices. Patients overwhelmingly preferred CT to endoscopy. Agreement between radiologists was higher than between endoscopists for the size of varices. CT alone was the most cost-effective strategy.
These two well-designed studies were performed in a population of patients with cirrhosis who need to undergo screening, and the sensitivity of both techniques for the detection of large varices was clearly reported. Capsule and CT were safe and much better tolerated than endoscopy, and could thus be good screening tools; moreover, screening with CT was more cost-effective than endoscopic screening. Nevertheless, the results of these studies are disappointing. Indeed, sensitivity for detecting large varices is unsatisfactory: one-third of the patients with large varices would have been misdiagnosed by CT, and sensitivity for the capsule was only 78%, which is insufficient. Therefore, replacing endoscopy with one of these two techniques would result in more misdiagnosed patients with large varices, which is not reasonable.
How can we explain these disappointing results? First, the interpretation of the results of CT and the capsule may vary. In the CT study, although both radiologists were experienced, their agreement for the size of varices was only 75%. In the capsule study, results were better in centers with more than 15 patients and worse than in a previous study where results were only analyzed by experienced operators.18 This emphasizes the importance of the learning curve for interpretation of results from this technique. Thus, we can imagine that improvement of the interpretation of results from these techniques will result in better detection of large varices. Second, and interestingly, CT results improved when endoscopists were experienced, raising the difficult question of interobserver variability for detecting and grading varices with endoscopy, that is, of the weakness of the gold standard. This issue is well addressed in the CT study, where the degree of interobserver variability was determined by randomly taking photographs during endoscopies, then submitting them to five other endoscopists, some experienced and some not. As previously reported,19 agreement for small varices was only fair in experienced endoscopists, because there was disagreement between at least two endoscopists in about one-third of the cases. More important, in 26% of cases, these varices were considered large by at least one endoscopist. As expected, the interobserver agreement for small varices was even lower (49%) in inexperienced endoscopists. In the capsule study, one patient who was determined to have large varices with the capsule but not at endoscopy underwent a second endoscopy that confirmed the presence of large varices. It would have been interesting to perform endoscopies in all patients with the same features. Similarly, in our experience, patients who undergo esophageal endoscopy for screening for potential varices have more chances of being diagnosed with varices than those who undergo this examination for other reasons (D. Thabut and D. Lebrec, unpublished results).
How can the detection of varices be improved? At present, endoscopy, which is the existing gold standard, has many drawbacks, but there is no perfect tool to replace it. Thus, other noninvasive strategies must be developed to decrease the number of endoscopies without increasing mortality from variceal bleeding. One solution is to associate several noninvasive methods to increase screening sensitivity in the same way as the FibroTest and Fibroscan for fibrosis.20 Another option is to indicate endoscopy only when capsule or CT show small varices. This approach was evaluated in the CT study and was less cost-effective. However, in that study the proportion of small varices was relatively high. In addition, because hepatitis C cirrhosis is now more often diagnosed in asymptomatic patients, the proportion of patients without varices may grow. Therefore the cost of this strategy should be re-evaluated in the future.
The best noninvasive tool to replace endoscopy has not yet been found. When designing future studies, the weaknesses of the existing gold standard should probably be taken into account. In particular, attention should be paid to improving the interpretation of endoscopy and reducing the variability of results. We recommend that endoscopies be performed by experienced operators and videos be recorded that could be reanalyzed by experts if there is any doubt. We hope that studies with the next generation of the capsule, which is very promising, will take into account these recommendations.
In conclusion, the gold standard has numerous drawbacks and the perfect screening tool has not been found. Physicians and patients must accept this uncertainty and efforts must continue to find indexes and associate noninvasive methods to screen patients for varices without using endoscopy.