Application of the platelet count/spleen diameter ratio to rule out the presence of oesophageal varices in patients with cirrhosis: A validation study based on follow-up

Results. During the follow-up, 27 patients (40%) developed oesophageal varices. Patients with higher baseline platelet count/spleen diameter ratios ( p < 0.0001) as well as a ratio above 909 were less likely to develop oesophageal varices ( p < 0.0005). At follow-up, a platelet count/spleen diameter ratio ≤909 had 100% negative predictive value and 84% efficiency in identifying the presence of oesophageal varices. Conclusions. The use of the platelet count/spleen diameter ratio proved to be an effective means for ruling out the presence of oesophageal varices even in the longitudinal follow-up of patients. Abbreviations OV oesophageal varices ROC receiver operating characteristics S.D. standard deviation 95% CI 95% confidence interval Keywords Cirrhosis Diagnosis Oesophageal varices Platelets Spleen Surveillance 1 Introduction The development of oesophageal varices (OV) is one of the most important complications of liver cirrhosis. Identifying the presence of OV is a fundamental part of the diagnostic work-up of cirrhotic patients [1] . Upper digestive endoscopy is the only means to diagnose and grade OV. Although patients with large OV are at a higher risk of bleeding, the presence of any OV pinpoints cirrhotic patients who require particular attention. In fact, presence of any OV determines modifications in the patients’ follow-up and is an important prognostic marker in the course of the disease [1,2] . Furthermore, recent studies have shown that identifying patients with small OV is of utmost importance since beta-blocker therapy seems to prevent increases in the size of OV and helps decrease the variceal bleeding rate [3] . Due to the current and future relevance of chronic liver disease in the general population and to the impact of OV in the course of the disease, the Baveno Consensus Conference proposed some simple guidelines to rationalise endoscopic screening for OV that were followed by both the American College of Gastroenterology and the American Association for the Study of Liver Disease [4–6] . However, these guidelines are derived from suggestions made by experts rather than being evidence-based [7] . This limitation, together with the considerable workload that OV screening and surveillance represent for endoscopy units, has recently led many researchers to try to identify some parameters that could noninvasively diagnose OV [8–15] . In most of these studies, the parameter that was most commonly associated with the presence of OV was thrombocytopenia, though it lacked adequate sensitivity to actually be proposed in clinical practice. Furthermore, platelet count did not achieve high specificity for the diagnosis of OV likely due to the multi-factorial aetiology of thrombocytopenia in patients with liver cirrhosis [16,17] . In fact, the presence of OV is merely due to portal hypertension, while thrombocytopenia may be the result of various factors besides splenic pooling of the platelets [18] . Taking into consideration the aforementioned factors, we recently tried to exploit the meaning of thrombocytopenia by normalising the platelet count to spleen diameter as measured by means of abdominal ultrasound. Our aim was to identify the decrease in platelet count most likely caused by hypersplenism. In the first set of cirrhotic patients, we found that a platelet count/spleen diameter ratio that had been identified by means of receiver operating characteristics (ROC) curve (i.e. 909) was able to rule out the presence of OV in cirrhotic patients with 100% prevalence-adjusted negative predictive value, and we were able to reproduce the same diagnostic yield of this cut-off in the subsequent series of cirrhotic patients [19] . Other researchers further evaluated the usefulness of the platelet count/spleen diameter ratio in a different set of patients and confirmed our results [20] . On the basis of these results, our diagnostic tool was validated in various settings, thus fulfilling the initial criteria for generalisability of scientific results [21] . In this study, our aim was to perform longitudinal validation of the platelet count/spleen diameter ratio for the noninvasive diagnosis of the absence of OV. We deemed it of interest to evaluate whether patients with higher platelet count/spleen diameter ratio or a ratio above 909 carried a lower risk of developing OV, and to assess whether the platelet count/spleen diameter ratio 909 cut-off maintained its diagnostic efficiency. In order to carry this out, we followed-up the patients without OV in our previous study who underwent OV surveillance at scheduled interval. 2 Patients and methods The initial cohort of this study included 106 patients without OV who participated in our previous study (61 Child–Pugh's class A, 30 class B, 15 class C) [19] . After initial endoscopy, we suggested that patients should repeat surveillance endoscopy every other year (Child–Pugh's class A patients), or every year (Child–Pugh's class B and C patients), according to the current clinical practice followed at our institution. Furthermore, we also suggested that patients should have complete clinical and biochemical evaluation, and ultrasonography of the upper abdomen each time they underwent endoscopy. Sixty-one of the 106 patients (58% of the initial cohort) regularly attended our outpatient clinic and did the suggested work-up at the scheduled intervals. On the other hand, although regularly attending our outpatient clinic, 14 patients (13% of the initial cohort) did not follow the suggested agenda and underwent evaluation at irregular intervals. Six out of 75 patients who underwent either regular or irregular follow-up died (median age, 80 ± 8 years; causes of death were stroke in two patients, bladder carcinoma in one patient, car accident in one patient and liver failure in two patients) and one underwent orthotopic liver transplantation (age, 64 years) before undergoing their first follow-up endoscopy. Thirty-one patients (29% of the initial cohort) did not show-up at the scheduled interval following the initial endoscopy and were contacted on telephone. Eleven of them were being treated at other hospitals, nine patients could not be located, seven patients refused to undergo examination and four patients had died. Therefore, 68 patients (male/female, 43/25; aetiology of cirrhosis, viral/non-viral, 53/15) made up the final cohort of this study ( Fig. 1 and Table 1 ). Gender (male/female, 31/7 versus 43/25, χ 2 = 0.141, p = 0.153) and baseline age (63 years versus 62 years, p = 0.623), Child–Pugh's class distribution (A/B/C, 21/9/8 versus 40/21/7, χ 2 = 0.152, p = 0.287), aetiology of liver disease (viral/non-viral, 29/9 versus 53/15, χ 2 = 0.005, p = 0.962) and platelet count/spleen diameter ratio (1324 ± 686 versus 1375 ± 786, p = 0.624) of the patients who, for any reason, did not participate in the follow-up program were no different from the ones of patients who were followed up at our outpatient clinic. The study started after patients included in our previous study performed screening endoscopy (January 1998–July 2002), and ended when all the eligible patients had completed at least one follow-up examination (August 2004). Patients who had not developed OV as of August, 2004, were censored at their most recent evaluation. Data collected at first diagnosis of OV or last follow-up examination in patients who did not develop OV were used for the purpose of the study. Patients’ work-up included upper digestive endoscopy, complete biochemical and clinical examination, and ultrasonography of the upper abdomen. All endoscopies were performed in a single endoscopy unit by means of a video-endoscope, and the same criteria were used to classify all the cases of OV. Briefly, OV were classified using the small and large classification, where small OV were defined as varices that flatten with insufflation or minimally protrude into the oesophageal lumen, while large OV were defined as varices that protrude into the oesophageal lumen and touch each other (presence of confluence), or that fill at least 50% of the oesophageal lumen. Patients were subdivided into two groups according to the presence/absence of OV. Spleen ultrasonography was carried out using an AU 570 Asynchronous (Esaote Biomedica, Genova, Italia) by experienced operators. Maximum spleen bipolar diameter was estimated by means of ultrasound scan and was expressed in millimetres. The coefficient of variation for repeated measurements of this parameter was less than 2%. Both endoscopy and ultrasonography operators were blinded to biochemical and instrumental results. We calculated the platelet count/spleen diameter ratio of all patients. Platelet count, spleen diameter and the platelet count/spleen diameter ratio in patients with or without OV were compared by means of the Mann–Whitney U -test. χ 2 -Test was used to compare proportions. The ROC curves were applied in order to calculate the accuracy of the platelet count, spleen diameter and platelet count/spleen diameter ratio for the diagnosis of OV at follow-up. The platelet count/spleen diameter ratio cut-off that we had identified in our previous study (i.e. 909) was applied to this cohort of patients, and its sensitivity, specificity and prevalence-adjusted negative and positive predictive values for identifying the presence of OV were calculated. Efficiency (true positives + true negatives/ n ) of the platelet count/spleen diameter ratio 909 cut-off for ruling out the presence of OV was also calculated. Data are shown as the mean and standard deviation (S.D.), and percentages are shown as absolute value and 95% confidence interval (95% CI). A p -value <0.05 was considered statistically significant for all analyses. 3 Results Mean time for the diagnosis of any OV, or for last evaluation was 880 days (S.D., 396). During this period, 27 patients (40%) developed OV. First diagnosis showed that they were small in 23 patients (85%) and large in 4 patients (15%). Patients who did not follow the suggested examination schedule were more frequently diagnosed with large OV at the first checkup as compared to patients who followed the suggested agenda (1/57 patients versus 3/11 patients, p = 0.01, χ 2 -test). None of the patients had OV bleeding prior to the detection of OV. 3.1 Prognostic study There were no significant differences between patients who did or did not develop OV with regards to gender (male/female, 16/11 versus 27/14), age (65 ± 7 years versus 65 ± 11 years) and aetiology of cirrhosis (viral/non-viral, 19/8 versus 34/7). When OV was diagnosed, Child–Pugh's class was found to have worsened in 15 patients as compared to the baseline. Table 2 shows the changes in Child–Pugh's class that were observed during the follow-up, as well as the number of patients who had OV, all listed according to variations in Child–Pugh's class. During the first evaluation, OV was found in all the patients who were in Child–Pugh's class C at baseline. Patients who were free from OV at follow-up had higher baseline platelet count/spleen diameter ratio (1690 ± 800 versus 896 ± 461, p < 0.0001). We applied the platelet count/spleen diameter ratio cut-off that had been identified in our previous study (i.e. 909) to this study population. Patients with baseline platelet/count spleen diameter ratios above 909 were less likely to develop OV as compared to patients with baseline ratios below 909 (10/46 versus 17/22, p < 0.0005, χ 2 -test). 3.2 Diagnostic study Platelet count, spleen diameter and platelet count/spleen diameter ratio at follow-up were significantly different between patients who had and did not have OV ( Table 3 ). The accuracy (area under the ROC curve) of the follow-up platelet count spleen/diameter ratio in predicting the presence of OV was 93.5% ( c -index = 0.935; 95% CI, 0.848–0.980; Fig. 2 ), while the accuracy of the follow-up platelet count and spleen diameter alone was 92.1% (95% CI, 82.9–97.2%) and 75.2% (95% CI, 63.2–84.9%), respectively. Sensitivity and specificity for identifying patients with OV of a platelet count/spleen diameter ratio ≤909 were 100% (95% CI, 87–100) and 73% (95% CI, 57–82), respectively. Furthermore, a platelet count/spleen diameter ratio ≤909 had 100% prevalence-adjusted negative predictive value (95% CI, 88–100) and 71% prevalence-adjusted positive predictive value (95% CI, 54–85) in predicting the presence of OV ( Fig. 3 ). Finally, the efficiency (true positives + true negatives/ n ) of the platelet count/spleen diameter ratio 909 cut-off for ruling out the presence of OV in this series was 84%. 4 Discussion The noninvasive assessment of the presence/absence of OV implies a series of medical and social issues [22] . In fact, from a medical point of view, such a noninvasive approach has to fulfil some well-defined characteristics that allow us to identify patients without OV and enable us to concentrate on those who are most likely to have OV. In particular, the accuracy of such a noninvasive approach must represent the best compromise between the risk of missing an important diagnosis, such as OV, and the lack of clinical usefulness because of the low specificity of the screening tool. Therefore, in this setting, the ideal screening tool should have a sensitivity approaching 100% so as to keep the highest safety profile, and high specificity in order to avoid performing unnecessary endoscopies on patients without OV, thus preserving the cost–benefit profile. A good cost–benefit profile would allow for diffusion of the screening tool, and consequently, better use of available resources and an improvement in the management of endoscopy centres. In this study, we moved along the validation pathway of our newly devised criterion for the noninvasive assessment of OV in cirrhotic patients. We re-evaluated the patients who were OV-free in our previous study [19] . This study was carried out in a clinical practice setting, in that some of the patients were regularly followed up at our outpatient clinic, while others were not, and yet others were lost to follow-up [23] . Although 29% of the patients could not be followed up and some died before undergoing the first surveillance endoscopy, their main clinical characteristics (age, gender, aetiology and severity of liver disease, and baseline platelet count/spleen diameter ratio) were not significantly different from those of the study cohort, and therefore, it seems unlikely that the dropouts may have influenced the results of the study. Finally, as often happens in the clinical practice, some patients were compliant with the suggested examination schedule while others were not. In this regard, we observed that patients who did not follow the suggested surveillance agenda were more likely to have large OV at first diagnosis. This finding further underscores the importance of adherence to surveillance programs. Overall, we observed that the rate of appearance of OV was in keeping with previously published series [1,24–26] , although it must be stressed that a formal comparison can be difficult to perform since the rate of OV varies according to the prevalence of severe disease in the studied population and on the timing of the follow-up. First of all, we found that patients who were OV-free at follow-up had higher baseline platelet count/spleen diameter ratios as compared to patients who developed OV. Furthermore, patients who were considered ‘false positive’ in our previous study (i.e. patients with a baseline platelet count/spleen diameter ratio ≤909 but without OV) were more likely to have OV at follow-up as compared to patients with higher platelet count/spleen diameter ratios. We feel that this is an important detail since other studies were not able to identify parameters associated with an increased risk of developing OV [26] . Furthermore, it emphasises the need for strict surveillance in patients with low platelet count/spleen diameter ratios, but without OV. Conversely, patients with higher platelet count/spleen diameter ratios or with a ratio above 909 were less likely to develop OV. Another aim of the study was to assess the overall diagnostic accuracy of the platelet count/spleen diameter ratio, and whether the previously identified cut-off of 909 could continue to maintain good specificity in ruling out the presence of OV even during the follow-up of patients. At OV diagnosis or last examination, there was a significant difference in platelet count and spleen diameter between patients with or without OV, even though the platelet count/spleen diameter ratio had the highest accuracy in ruling out the presence of OV. A platelet count/spleen diameter ratio of 909 had 100% sensitivity and prevalence-adjusted negative predictive value for ruling out the presence of OV. This assured us that all the patients with a higher platelet count/spleen diameter ratio value who hypothetically would not have undergone surveillance endoscopy at follow-up were OV-free, thus keeping a very good safety profile. Moreover, the 909 cut-off had 71% prevalence-adjusted positive predictive value, therefore, allowing us to perform endoscopy with a satisfactory cost–benefit profile. Finally, the efficiency of the platelet count/spleen diameter ratio 909 cut-off for ruling out the presence of OV was 84%. A surrogate diagnostic tool must be biologically plausible, besides being statistically accurate and clinically sound. In this setting, the high diagnostic accuracy of the platelet count/spleen diameter ratio seems to reflect its pathophysiological background. In fact, studies that evaluated noninvasive diagnosis of OV found that splenomegaly seemed to be highly sensitive but less specific, while thrombocytopenia showed high specificity but lower sensitivity. In the study by Chalasani et al., platelet count and splenomegaly independently predicted the presence of OV, and the two variables did not interact [8] , while in the study by Madhotra et al., 32% of the patients whose platelet counts were below 68,000 did not have splenomegaly [14] . These results can be accounted for by a number of explanations. On the one hand, thrombocytopenia has an important clinical meaning in cirrhotic patients and can be ascribed, besides splenic pooling of the platelets due to portal hypertension, to immune-mediated mechanisms and decreased thrombopoietin synthesis [16,17,27,28] . On the other hand, the use of the presence/absence of splenomegaly as a dichotomous variable did not allow modulation of the results since any increase in spleen dimension above the upper normal value was categorised as positive. Moreover, this variable was assessed in various fashions, and, for example, palpatory splenomegaly is less reliable than CT or ultrasound evaluation [29] . In our studies, we used the spleen size as measured by ultrasound, since it has a very low intra- and inter-observer variability, which is lower than Doppler assessment of hepatic and portal blood vessels as well as of portal vein diameter [30,31] . Integrating the two parameters into a single one eventually exploited the pathophysiological meaning of both, and allowed for the use of a continuous scale, thus improving the modulation of results. The fact that the platelet count/spleen diameter ratio was used to assess the presence/absence of any OV rather than large OV could be considered a drawback of the study. In fact, prophylactic treatment with beta-blockers is currently recommended only for patients with large OV [4–6] . However, recent studies put forward the hypothesis that ‘universal’ beta-blocker therapy without endoscopic screening for OV could be a cost-effective option in cirrhotic patients [32,33] , and other studies have shown that beta-blocker therapy is able to reduce the OV growth from small to large and decrease the risk of variceal bleeding [3] . Although this study was not planned to evaluate a cost-effective analysis, by putting together these results and adding our own, we can hypothesise that the use of the platelet count/spleen diameter ratio would be useful for ruling out patients who do not have OV, further increasing the cost–benefit of ‘universal’ beta-blocker prophylaxis. On the basis of the above-mentioned points, a clinical algorithm that includes platelet count/spleen diameter ratio could be incorporated into the diagnostic work-up of cirrhotic patients. If this was implemented, patients whose platelet count/spleen diameter ratio was above the 909 cut-off would not undergo beta-blocker therapy since they should not have OV. On the other hand, 71% of the patients with a platelet count/spleen diameter ratio below 909 would benefit from beta-blocker therapy, and would further increase the cost-efficacy profile as compared to universal empiric treatment. Aetiology of liver disease in this study was clearly similar to the aetiology of the patients included in our previous study, meaning a prevalence of patients with cirrhosis of viral origin. Although separate analysis of the subgroup of patients with alcoholic cirrhosis alone demonstrated that the platelet count/spleen diameter ratio maintains an excellent diagnostic accuracy [34] , and this finding was confirmed in independent series [20] , the low number of patients with cholestatic liver disease included in our studies does not allow generalisation of results to this subgroup of patients. Nevertheless, a recent study showed that, as is the case for patients with post-necrotic cirrhosis, thrombocytopenia and advanced disease are the parameters associated with the presence of OV in patients with primary sclerosing cholangitis [15] . Therefore, it would be interesting to evaluate the diagnostic yield of the platelet count/spleen diameter ratio in this subgroup of patients as well. In conclusion, this study shows that the use of the platelet count/spleen diameter ratio has a prognostic value and maintains its diagnostic accuracy for ruling out the presence of OV, even in longitudinal follow-up of patients. Further validation of this parameter in independent cohorts of patients is needed before proposing this tool as a screening method in clinical practice. Conflict of interest statement None declared. References [1] D.M. Jensen Endoscopic screening for varices in cirrhosis: findings, implications, and outcomes Gastroenterology 122 2002 1620 1630 [2] G. Fattovich M. Pantalena I. Zagni G. Realdi S.W. Schalm E. Christensen European concerted action on viral hepatitis (EUROHEP). Effect of hepatitis B and C virus infections on the natural history of compensated cirrhosis: a cohort study of 297 patients Am J Gastroenterol 97 2002 2886 2895 [3] C. Merkel R. Marin P. Angeli P. Zanella M. Felder E. Bernardinello A placebo-controlled clinical trial of nadolol in the prophylaxis of growth of small esophageal varices in cirrhosis Gastroenterology 127 2004 476 484 [4] R. de Franchis J.P. Pascal E. Ancona A.K. Burroughs M. Henderson W. Fleig Definitions, methodology and therapeutic strategies in portal hypertension. A Consensus Development Workshop, Baveno, Lake Maggiore, Italy, April 5 and 6, 1990 J Hepatol 15 1992 256 261 [5] N.D. Grace Diagnosis and treatment of gastrointestinal bleeding secondary to portal hypertension. American College of Gastroenterology Practice Parameters Committee Am J Gastroenterol 92 1997 1081 1091 [6] N.D. Grace R.J. Groszmann G. Garcia-Tsao A.K. Burroughs L. Pagliaro R.W. Makuch Portal hypertension and variceal bleeding: an AASLD single topic symposium Hepatology 28 1998 868 880 [7] J.A. Talwalkar P.S. Kamath Screening for esophageal varices among patients with cirrhosis of the liver Am J Gastroenterol 96 2001 3039 3040 [8] N. Chalasani T.F. Imperiale A. Ismail G. Sood M. Carey C.M. Wilcox Predictors of large esophageal varices in patients with cirrhosis Am J Gastroenterol 94 1999 3285 3291 [9] A. Zaman R. Hapke K. Flora H.R. Rosen K. Benner Factors predicting the presence of esophageal varices or gastric varices in patients with advanced liver disease Am J Gastroenterol 94 1999 3292 3296 [10] C. Pilette F. Oberti C. Aubé M.C. Rousselet P. Bedossa Y. Gallois Non-invasive diagnosis of esophageal varices in chronic liver disease J Hepatol 31 1999 867 873 [11] F.H. Ng S.Y. Wong C.K. Loo K.M. Lam C.W. Lai C.S. Cheng Prediction of oesophagogastric varices in patients with cirrhosis J Gastroenterol Hepatol 14 1999 785 790 [12] F. Schepis C. Cammà D. Niceforo A. Magnano S. Pallio M. Cinquegrani Which patients should undergo endoscopic screening for esophageal varices detection? Hepatology 33 2001 333 338 [13] A. Zaman T. Becker J. Lapidus K. Benner Risk factors for the presence of varices in cirrhotic patients without a history of variceal hemorrhage Arch Intern Med 161 2001 2564 2570 [14] R. Madhotra H.E. Mulcahy I. Willner A. Reuben Prediction of esophageal varices in patients with cirrhosis J Clin Gastroenterol 34 2002 81 85 [15] C.O. Zein K.D. Lindor P. Angulo Prevalence and predictors of esophageal varices in patients with primary sclerosing cholangitis Hepatology 39 2004 204 210 [16] M. Peck-Radosavljevic Thrombocytopenia in liver disease Can J Gastroenterol 14 Suppl. D 2000 60D 66D [17] E. Giannini F. Botta P. Borro F. Malfatti A. Fumagalli E. Testa Relationship between thrombopoietin serum levels and liver function in patients with chronic liver disease related to hepatitis C virus infection Am J Gastroenterol 98 2003 2516 2520 [18] P.A. McCormick The spleen, hypersplenism, and other relationships between the liver and spleen J. Bircher J.-P. Benhamou N. McIntyre M. Rizzetto J. Rodés Oxford textbook of clinical hepatology 1999 Oxford University Press Oxford 787 795 [19] E. Giannini F. Botta P. Borro D. Risso P. Romagnoli A. Fasoli Platelet count spleen/diameter ratio: proposal and validation of a non-invasive parameter to predict the presence of oesophageal varices in patients with liver cirrhosis Gut 52 2003 1200 1205 [20] T.A. Zimbwa C. Blanshard A. Subramaniam Platelet count/spleen diameter ratio as a predictor of oesophageal varices in alcoholic cirrhosis Gut 53 2004 1055 [21] A.C. Justice K.E. Covinsky J.A. Berlin Assessing the generalizability of prognostic information Ann Intern Med 130 1999 515 524 [22] R. de Franchis A. Dell’Era F. Iannuzzi Diagnosis and treatment of portal hypertension Dig Liver Dis 36 2004 787 798 [23] A.K. Ghosh Refusal of endoscopy in cirrhotic patients with Child–Pugh class A without prior variceal bleeding: an internist dilemma Arch Intern Med 162 2002 947 948 [24] P. Cales H. Desmorat J.P. Vinel J.P. Caucanas A. Ravaud P. Gerin Incidence of large oesophageal varices in patients with cirrhosis: application to prophylaxis of first bleeding Gut 31 1990 1298 1302 [25] P. Gentilini G. Laffi G. La Villa R.G. Romanelli G. Buzzelli V. Casini-Raggi Long course and prognostic factors of virus-induced cirrhosis of the liver Am J Gastroenterol 92 1997 66 72 [26] M. Merli G. Nicolini S. Angeloni V. Rinaldi A. De Santis C. Merkel Incidence and natural history of small esophageal varices in cirrhotic patients J Hepatol 38 2003 266 272 [27] P.J. Pockros A. Duchini R. McMillan L.M. Nyberg J. McHutchison E. Viernes Immune thrombocytopenic purpura in patients with chronic hepatitis C virus infection Am J Gastroenterol 97 2002 2040 2045 [28] S. Liangpunsakul B.J. Ulmer N. Chalasani Predictors and implications of severe hypersplenism in patients with cirrhosis Am J Med Sci 326 2003 111 116 [29] S.G. Tamayo L.S. Rickman W.C. Mathews S.C. Fullerton A.E. Bartok J.T. Warner Examiner dependence on physical diagnostic tests for the detection of splenomegaly: a prospective study with multiple observers J Gen Intern Med 8 1993 69 75 [30] J. O’Donohue C. Ng S. Catnach P. Farrant R. Williams Diagnostic value of Doppler assessment of the hepatic and portal vessels and ultrasound of the spleen in liver disease Eur J Gastroenterol Hepatol 16 2004 147 155 [31] B. Winkfield C. Aubé P. Burtin P. Cales Inter-observer and intra-observer variability in hepatology Eur J Gastroenterol Hepatol 15 2003 959 966 [32] M.R. Arguedas G.R. Heudebert M.A. Eloubeidi G.A. Abrams M.B. Fallon Cost-effectiveness of screening, surveillance, and primary prophylaxis strategies for esophageal varices Am J Gastroenterol 97 2002 2441 2452 [33] B.M. Spiegel L. Targownik G.S. Dulai H.A. Karsan I.M. Gralnek Endoscopy screening for esophageal varices in cirrhosis: is it ever cost-effective? Hepatology 37 2003 366 377 [34] E. Giannini F. Botta R. Testa Prediction of oesophageal varices with platelet count/spleen diameter ratio or platelets alone Gut 53 2004 914 915