Portal Hypertension in Patients with Liver Cirrhosis: Diagnostic Accuracy of Spleen Stiffness 1

This copy is for personal use only. To order printed copies, contact reprints@rsna.org Yoshitaka Takuma, MD Kazuhiro Nouso, MD Youichi Morimoto, MD Junko Tomokuni, BA Akiko Sahara, BA Hiroyuki Takabatake, MD Kazuhiro Matsueda, MD Hiroshi Yamamoto, MD Portal Hypertension in Patients with Liver Cirrhosis: Diagnostic Accuracy of Spleen Stiffness 1 Purpose: Materials and Methods: To evaluate the accuracy of spleen stiffness (SS) and liver stiffness (LS) measured by using acoustic radiation force impulse imaging in the diagnosis of portal hypertension in patients with liver cirrhosis, with the hepatic venous pressure gradient (HVPG) as a reference standard. Institutional review board approval and informed consent were obtained for this prospective single-center study. From February 2012 to August 2013, 60 patients with liver cirrhosis (mean age, 70.8 years; age range, 34 88 years; 34 men, 26 women) with HVPG, LS, and SS measurements and gastrointestinal endoscopy and laboratory data were included if they met the following criteria: no recent episodes of gastrointestinal bleeding, no history of splenectomy, no history of partial splenic embolization, no history of b-blocker therapy, and absence of portal thrombosis. The efficacy of the parameters for the evaluation of portal hypertension was analyzed by using the Spearman rank-order correlation coefficient and receiver operating characteristic (ROC) curve analysis. Original Research n Ultrasonography Results: The correlation coefficient between SS and HVPG (r = 0.876) was significantly better than that between LS and HVPG (r = 0.609, P,.0001). The areas under the ROC curve of SS for the identification of clinically important portal hypertension (HVPG 10 mm Hg), severe portal hypertension (HVPG 12 mm Hg), esophageal varices (EVs), and high-risk EVs were significantly higher (0.943, 0.963, 0.937, and 0.955, respectively) than those of LS, spleen diameter, platelet count, and platelet count to spleen diameter ratio (P,.05 for all). SS could be used to accurately rule out the presence of clinically important portal hypertension, severe portal hypertension, EVs, and high-risk EVs (negative likelihood ratios, 0.051, 0.056, 0.054, and 0.074, respectively). 1 From the Departments of Gastroenterology (Y.T., Y.M., H.T., K.M., H.Y.) and Laboratory Medicine (J.T., A.S.), Kurashiki Central Hospital, Okayama, Japan; and Department of Gastroenterology and Hepatology, Okayama University Graduate School of Medicine, Dentistry, and Pharmaceutical Sciences, Okayama, Japan (K.N.). Received March 26, 2015; revision requested May 11; revision received June 28; accepted July 10; final version accepted August 27. Address correspondence to Y.T., Department of Internal Medicine, Hiroshima City Hospital, 7-33 Motomachi, Naka-ku, Hiroshima 730-8518, Japan (e-mail: takuma@ enjoy.ne.jp). Conclusion: SS is reliable and has better diagnostic performance than LS for identifying portal hypertension in liver cirrhosis. q RSNA, 2015 Online supplemental material is available for this article. q RSNA, 2015 Radiology: Volume 279: Number 2 May 2016 n radiology.rsna.org 609

The development of portal hypertension is a common consequence of chronic liver diseases and leads to the major complications of liver cirrhosis, such as ascites, hepatic encephalopathy, variceal bleeding, and decompensation. Furthermore, decompensation is the most important predictor of prognosis and mortality in patients with liver cirrhosis (1). Thus, early diagnosis of portal hypertension is essential for adequate treatment to reduce the mortality rate of portal hypertension related complications (2). Measurement of the hepatic venous pressure gradient (HVPG) has been accepted as the reference standard for portal hypertension assessment in Advances in Knowledge nn In patients with liver cirrhosis, both spleen stiffness (SS) and liver stiffness (LS) measured by using acoustic radiation force impulse (ARFI) imaging were linearly correlated with hepatic venous pressure gradient (HVPG) (SS: r = 0.876; LS: r = 0.609), and the correlation coefficient between SS and HVPG was significantly higher than that between LS and HVPG (P,.0001). nn SS was the most accurate predictive factor for clinically important portal hypertension (HVPG 10 mm Hg) (area under the receiver operating characteristic curve [AUC], 0.943) and severe portal hypertension (HVPG 12 mm Hg) (AUC, 0.963), and SS had significantly higher AUCs than the other measured parameters of LS, spleen diameter, platelet count, and platelet count to spleen diameter ratio (P,.05 for all). nn SS could be used to accurately rule out the presence of severe and clinically important portal hypertension, esophageal varices (EVs), and high-risk EVs (negative likelihood ratios, 0.051, 0.056, 0.054, and 0.074, respectively). patients with cirrhosis. Patients with clinically important portal hypertension (HVPG 10 mm Hg) are at increased risk of developing varices, while patients with severe portal hypertension (HVPG 12 mm Hg) are at risk for variceal bleeding, with mortality rates ranging from 20% to 35% (3). However, measurement of HVPG is an invasive procedure associated with complications and is costly. Thus, there is a need for accurate and noninvasive methods of assessing the progression of portal hypertension. Liver stiffness (LS) measured by using transient elastography is a rapid and noninvasive method for the diagnosis of portal hypertension in patients with liver cirrhosis (4 6). However, Vizzutti et al (4) confirmed a poor correlation between the HVPG and LS when the HVPG is greater than 12 mm Hg, and LS measurements obtained by using transient elastography have been considered a useful but not a suitable method for identifying esophageal varices (EVs) (7). Colecchia et al (8) have reported that measuring spleen stiffness (SS) by using transient elastography was another feasible method for identifying portal hypertension in patients with hepatitis C virus induced cirrhosis. However, transient elastography has some limitations. Measuring LS with transient elastography is difficult in patients who are obese (although a newer probe allows measurements in obese patients) or who have narrow intercostal spaces and is impossible in patients with ascites (9). Acoustic radiation force impulse (ARFI) imaging can be performed with clear observation of the actual measuring site with B-mode imaging and can be used even in obese patients and in patients with ascites (10 14). In our recent study (11), SS evaluated by using ARFI imaging was shown to be closely correlated with the presence of EVs; Implication for Patient Care nn SS measured by using ARFI imaging may be used in the assessment of portal hypertension associated with liver cirrhosis. however, that study did not have the reference standard of HVPG. The aim of our current study was to evaluate the accuracy of SS and LS measured by using ARFI imaging in the diagnosis of portal hypertension in patients with liver cirrhosis, with HVPG as a reference standard. Materials and Methods Patients This study was performed in accordance with the guidelines of and was approved by our institutional review board. Written informed consent was obtained. Eighty patients with liver cirrhosis underwent assessment of portal pressure and endoscopic findings before starting prophylactic treatment with a b- blocker or endoscopic treatment; these patients were considered for inclusion in the study between February 2012 and August 2013 on being referred to the Kurashiki Central Hospital. Published online before print 10.1148/radiol.2015150690 Content codes: Radiology 2016; 279:609 619 Abbreviations: ALT = alanine aminotransferase ARFI = acoustic radiation force impulse AUC = area under the receiver operating characteristic curve CI = confidence interval EV = esophageal varix HVPG = hepatic venous pressure gradient LS = liver stiffness NLR = negative likelihood ratio NPV = negative predictive value PLR = positive likelihood ratio PPV = positive predictive value PSR = platelet count to spleen diameter ratio SS = spleen stiffness Author contributions: Guarantor of integrity of entire study, Y.T.; study concepts/ study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; agrees to ensure any questions related to the work are appropriately resolved, all authors; clinical studies, Y.T., Y.M., J.T., A.S., H.T., K.M., H.Y.; statistical analysis, Y.T.; and manuscript editing, Y.T., K.N. Conflicts of interest are listed at the end of this article. 610 radiology.rsna.org n Radiology: Volume 279: Number 2 May 2016

Figure 1 Figure 1: Flowchart of study participants. Of these 80 patients, 18 were excluded for the following reasons: episodes of recent (<6 months) gastrointestinal bleeding (n = 8), a history of splenectomy (n = 2), a history of partial splenic embolization (n = 2), a history of b-blocker therapy (n = 4), and the presence of portal thrombosis (n = 2) (Fig 1). A diagnosis of cirrhosis was determined as described previously (12,13) on the basis of the results of histologic examination of liver tissue or combined physical, laboratory, and radiologic findings. ARFI measurements in each patient were obtained by using an ultrasonography (US) system (Acuson S2000; Siemens, Erlangen, Germany) by one of two sonographers (J.T. and A.S., with 10 and 15 years of experience, respectively) who were blinded to the clinical data. After an overnight fast, each patient was placed in the supine position and underwent ARFI imaging with B-mode imaging. A region of interest (a fixed-dimension 1 3 0.5-cm box; maximum evaluable depth, 5.5 cm) in the liver and spleen parenchyma that was free of large blood vessels was selected. LS was measured in the right lobe of the liver, 1 cm below the liver capsule, by using the intercostal approach. SS was measured 1 cm below the spleen capsule by using the intercostal approach. ARFI shear-wave velocity was measured in meters per second. According to previous reports (10,14,15), more than five successful measurements should be performed for each patient. Thus, five valid measurements were performed in the liver and in the spleen of each patient, and median values were calculated. LS or SS measurement failure was defined as when there were zero valid shots, and unreliable measurements were defined as when the ratio of the interquartile range to the median value was greater than 30% or when the success rate was less than 60% (9,16). The maximum spleen bipolar diameter was estimated by using US and was expressed in millimeters. Platelet count to spleen diameter ratio (PSR) (17) was calculated for all patients. Clinical and laboratory parameters were measured in each patient on the day of US, which included ARFI imaging. Within 1 week (mean, 2.3 days; range, 1 6 days) after the measurement of LS and SS, HVPG measurements were performed in patients who had fasted overnight by a hepatologist (Y.T., with 18 years of experience with HVPG measurement) who was blinded to the US data. The right hepatic vein was catheterized percutaneously through the femoral vein, and pressure was recorded in both the wedged position and the free position with a 5- or 6-F balloon-tipped catheter (Terumo, Tokyo, Japan). HVPG was determined by subtracting the free hepatic vein pressure from the wedged hepatic vein pressure, and all measurements were performed in triplicate. Clinically important portal hypertension was defined as an HVPG of 10 mm Hg or greater, and severe portal hypertension was defined as an HVPG of 12 mm Hg or greater, according to the Baveno V criteria (18). After the US examinations, all patients also underwent screening upper gastrointestinal endoscopy, and the presence of EVs was determined by one of six endoscopists (each with more than 7 years of experience), who were blinded to the US data and HVPG. EVs were classified on the basis of the criteria for describing endoscopic findings of esophagogastric varices in Japan (19). The severity of EVs was classified as follows: A score of F1 indicated straight and small-caliber varices; a score of F2, beady varices; and a score of F3, tumor-shaped varices. Presence of a red color indicated a high risk of variceal bleeding. EVs in danger of rupture (high-risk EVs) were defined as F2 or F3 EVs or as F1 EVs with red color signs or Child-Pugh class C disease according to the Baveno V criteria (18). Low-risk EVs were defined as F1 EVs without red color signs or Child-Pugh class C disease. If the patients had high-risk EVs or clinically important portal hypertension, administration of a b-blocker or endoscopic treatment was considered. Complete evaluation for each patient (US and endoscopic measurements) was performed within 3 months (mean, 1.4 months; range, 1 day to 2.90 months). Statistical Analysis After we analyzed the normal or nonnormal distribution of the continuous variables with the Kolmogorov-Smirnov test, we examined differences between continuous and categoric variables using the Student t test (for normally distributed variables), the Mann-Whitney test (for non-normally distributed variables), and the x 2 test. Variables (platelet count, prothrombin time, albumin level, alanine aminotransferase [ALT] level, LS, and SS) found to be associated with the presence of clinically important portal hypertension and severe portal hypertension at univariate analysis (P,.05) were entered into Radiology: Volume 279: Number 2 May 2016 n radiology.rsna.org 611

multivariate backward stepwise logistic regression analysis. The interaction between these variables was also tested. SS and LS were separately entered into each multivariate analysis because of the strong correlation between SS and LS (r = 0.689), which might cause errors in multivariate analysis. To avoid the effect of collinearity, Child-Pugh classification and PSR (17) were not included in the multivariate model because these parameters are composite parameters. Linear correlations between noninvasive parameters of portal hypertension and HVPG were assessed by using the Spearman rankorder correlation coefficient (r). To compare the performance of two correlation coefficients, we performed the Steiger Z test for correlated correlations within a population (20). By convention, values greater than 1.96 are considered significant if a two-tailed test is performed. Receiver operating characteristic curve analysis of the parameters was used to assess their diagnostic value for identifying clinically important portal hypertension, severe portal hypertension, EVs, and high-risk EVs. The method suggested by DeLong et al (21) was used to compare the areas under the receiver operating characteristic curve (AUCs) of various parameters of portal hypertension such as SS, LS, spleen diameter, platelet count, and PSR. The overall performance of the models was evaluated with the Nagelkerke R 2 and the Brier score. A higher R 2 and a lower Brier score indicate better discriminative performance (Brier scores range from 0 [perfect] to 0.25 [worthless]) (22). Calibration between predicted and observed risk was tested by using the Hosmer- Lemeshow test. A nonsignificant P value with the Hosmer-Lemeshow test indicated an acceptable calibration. Diagnostic value was calculated by using sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy, positive likelihood ratio (PLR), and negative likelihood ratio (NLR). Cutoff values were chosen according to the aim of the screening test (to rule out the presence of any condition), choosing the lowest NLR. Data are given as means or as values with 95% confidence intervals (CIs). For all analyses, P,.05 was considered to indicate a statistically significant difference. Data were analyzed by using SPSS 16.0 J for Windows (SPSS, Chicago, Ill), MedCalc (MedCalc Software, Broekstraat, Mariakerke, Belgium), and R software, version 3.0.1. Results Baseline Characteristics of Patients with Liver Cirrhosis Among the 62 patients who satisfied the inclusion criteria, two had an Table 1 Baseline Patient Characteristics inconclusive SS measurement because the spleen was poorly visualized secondary to obesity or gastrointestinal gas. In total, there were 60 patients in the final analysis group. Table 1 shows the clinical, biochemical, endoscopic, and US characteristics of the 60 patients. The mean age was 70.8 years, 34 patients (56.7%) were male, and 36 patients (60.0%) had histologically proven cirrhosis. A total of 34 patients (56.7%) had clinically important portal hypertension, 29 (48.3%) had severe portal hypertension, 24 (40.0%) had EVs, and 16 (26.7%) had high-risk EVs. Variable Datum Range Age (y) 70.8 6 9.9 34 88 Age of male patients/age of female patients (y) 69.7 6 12.0/72.2 6 6.2* 34 84/63 88 Sex (M/F) 34/26 Body mass index (kg/m 2 ) 23.4 6 3.6 13.3 36.0 Histologically proven cirrhosis 36 (60.0)... Cause of cirrhosis Hepatitis B virus 8 (13.3)... Hepatitis C virus 35 (58.3)... Alcohol 6 (10.0)... Other 11 (18.3)... Presence of clinically important portal hypertension 34 (56.7)... Presence of severe portal hypertension 29 (48.3)... Presence of EVs 24 (40.0)... Presence of high-risk EVs 16 (26.7)... Presence of ascites 6 (10.0)... Platelet count (310 4 /ml) 9.6 6 4.0 2.2 22.0 Prothrombin time (%) 78.7 6 12.3 54 108 Albumin level (g/dl) 3.4 6 0.6 1.9 4.6 Total bilirubin level (mg/dl) 0.9 6 0.5 0.2 2.5 AST level (IU/L) 58.7 6 28.8 22 156 ALT level (IU/L) 47.2 6 26.3 11 118 Child-Pugh score 6.2 6 1.3 5 11 Child-Pugh classification A 41 (68.3)... B 18 (30.0)... C 1 (1.7)... HVPG (mm Hg) 11.9 6 5.2 4 25 LS (m/sec) 2.61 6 0.61 1.32 3.95 Spleen stiffness (m/sec) 3.40 6 0.41 2.50 4.44 Spleen length (mm) 112.2 6 25.6 77 190 PSR 610.6 6 435.8 211.5 1837.7 Note. Data are means 6 standard deviations or numbers of patients, with percentages in parentheses. AST = aspartate aminotransferase. * P =.99. 612 radiology.rsna.org n Radiology: Volume 279: Number 2 May 2016

Clinical Parameters for Identifying Clinically Important Portal Hypertension and Severe Portal Hypertension Platelet count, prothrombin time, albumin level, ALT level, Child-Pugh classification, LS, SS, and PSR were associated with the presence of clinically important portal hypertension according to univariate logistic regression Table 2 analysis (P =.009, P =.012, P =.043, P =.005, P =.023, P,.001, P,.001, and P =.008, respectively) (Table 2). SS was selected as an independent parameter associated with the presence of clinically important portal hypertension after adjustment for platelet count in multivariate logistic regression analysis (P,.001) (Table 2). Univariate and Multivariate Analysis of Clinical Parameters for Patients with HVPG Greater than 10 mm Hg Variable Univariate Analysis Multivariate Analysis Odds Ratio P Value Odds Ratio P Value Age 0.986 (0.935, 1.040).596...... Male sex 0.703 (0.249, 1.986).506...... Body mass index 1.151 (0.977, 1.355).093...... Viral cause of cirrhosis 1.721 (0.555, 5.335).347...... Platelet count 0.817 (0.700, 0.955).009 0.797 (0.601, 1.055).113 Prothrombin time 0.938 (0.893, 0.986).012...... Albumin level 0.368 (0.140, 0.970).043...... Total bilirubin level 3.099 (0.838, 11.465).090...... AST level 0.989 (0.971, 1.007).221...... ALT level 0.966 (0.944, 0.989).005 0.957 (0.926, 0.989).008 Child-Pugh class A disease 0.230 (0.065, 0.814).023...... LS 17.4 (3.7, 81.1),.001 27.3 (4.0, 184.2).001 Spleen stiffness 5016.3 (84.2, 298815.8),.001 4641.7 (59.8, 360460.0),.001 Spleen diameter 1.189 (0.954, 1.481).124...... PSR 0.998 (0.997, 0.999).008...... Note. Data in parentheses are 95% CIs. AST = aspartate aminotransferase. Furthermore, LS was also selected as an independent parameter associated with the presence of clinically important portal hypertension after adjustment for ALT level in multivariate logistic regression analysis (P =.001). Platelet count, prothrombin time, albumin, ALT level, Child-Pugh classification, LS, SS, spleen diameter, and PSR were significant parameters associated with the presence of severe portal hypertension in univariate logistic regression analysis (P =.007, P =.003, P =.022, P =.006, P =.002, P,.001, P,.001, P =.010, and P =.001, respectively) (Table E1 [online]). SS was selected as an independent parameter associated with the presence of severe portal hypertension after adjustment for platelet count in multivariate logistic regression analysis (P,.001). Furthermore, LS was also selected as an independent parameter associated with the presence of severe portal hypertension after adjustment for ALT level and spleen diameter in multivariate logistic regression analysis (P =.001). Correlations of SS, LS, and HVPG Both SS and LS were strongly linearly correlated with HVPG (SS: r = 0.876; LS: r = 0.609) (Fig 2). The correlation coefficient between SS and HVPG was significantly higher than that between Figure 2 Figure 2: Scatterplots show correlations between HVPG and, A, SS and, B, LS. Radiology: Volume 279: Number 2 May 2016 n radiology.rsna.org 613

Figure 3 Figure 3: A, B, Scatterplots show correlations between HVPG and, A, SS and, B, LS in patients with an HVPG of 10 mm Hg or greater (n = 34). C, D, Scatterplots show correlations between HVPG and, C, SS and, D, LS in patients with an HVPG of less than 10 mm Hg (n = 26). LS and HVPG (P,.0001). In patients with an HVPG of 10 mm Hg or greater (n = 34), both SS and LS were significantly correlated with HVPG (SS: r = 0.764; LS: r = 0.426) (Fig 3). The correlation coefficient between SS and HVPG was significantly higher than that between LS and HVPG (P =.017). In patients with an HVPG of less than 10 mm Hg (n = 26), there was a significant correlation between HVPG and SS (r = 0.451), while no correlation was observed between HVPG and LS (r = 0.034) (Fig 3). AUCs of SS and Other Noninvasive Parameters for Identifying Clinically Important Portal Hypertension and Severe Portal Hypertension Among these parameters, SS was the most accurate diagnostic factor for clinically important portal hypertension (AUC, 0.943; 95% CI: 0.852, 0.987) and severe portal hypertension (AUC, 0.963; 95% CI: 0.880, 0.995), and both AUCs of SS were significantly higher than those of LS, spleen diameter, platelet count, and PSR (P,.05 for all) (Table 3). Indicating its overall diagnostic performance for clinically important portal hypertension and severe portal hypertension, SS had high Nagelkerke R 2 values (0.710 for clinically important portal hypertension and 0.770 for severe portal hypertension) and low Brier scores (0.100 for clinically important portal hypertension and 0.080 for severe portal hypertension) (Table 3). In addition, the Hosmer-Lemeshow test for calibration between predicted and observed risk was applied to the logistic models for identifying clinically important and 614 radiology.rsna.org n Radiology: Volume 279: Number 2 May 2016

Table 3 Performance of Clinical Parameters for Association with Portal Hypertension Parameter AUC severe portal hypertension, and nonsignificant P values for SS indicated acceptable calibration (Table 3). Thus, SS had the best overall performance for the diagnosis of clinically important and severe portal hypertension. AUCs of SS and Other Noninvasive Parameters for Identifying EVs and High- Risk EVs For identifying EVs and high-risk EVs, SS was the most accurate diagnostic factor for EVs (AUC, 0.937; 95% CI: 0.844, 0.984) and high-risk EVs (AUC, 0.955; 95% CI: 0.867, 0.992) among these parameters, and both AUCs for association with EVs and high-risk EVs of SS were significantly higher than those of LS, spleen diameter, platelet count, and PSR (P,.05 for all for both parameters) (Table E2 [online]). For identifying EVs and high-risk EVs, SS had high Nagelkerke R 2 values (0.684 for EVs and 0.717 for high-risk EVs), low Brier scores (0.099 for EVs and 0.081 for high-risk EVs), and nonsignificant P values in the Hosmer-Lemeshow test (P =.675 for EVs and P =.816 for high-risk EVs) (Table E2 [online]). Diagnostic Ability of SS for Identifying Clinically Important Portal Hypertension, Brier Score Nagelkerke R 2 Value Clinically important portal hypertension Spleen stiffness 0.943 (0.852, 0.987) 0.100 0.710.599 LS 0.833 (0.714, 0.917) 0.163 0.434.839 Spleen diameter 0.598 (0.463, 0.722) 0.236 0.056.572 Platelet count 0.715 (0.584, 0.824) 0.213 0.165.967 PSR 0.712 (0.580, 0.821) 0.215 0.173.168 Severe portal hypertension Spleen stiffness 0.963 (0.880, 0.995) 0.080 0.770.208 LS 0.846 (0.745, 0.948) 0.159 0.445.408 Spleen diameter 0.699 (0.565, 0.832) 0.220 0.165.213 Platelet count 0.744 (0.616, 0.872) 0.207 0.196.881 PSR 0.778 (0.660, 0.895) 0.192 0.290.243 Note. Data in parentheses are 95% CIs. Hosmer-Lemeshow P Value Severe Portal Hypertension, EVs, and High-Risk EVs The SS cutoff values of 3.10, 3.15, 3.36, and 3.51 m/sec were selected to rule out the presence of clinically important portal hypertension, severe portal hypertension, EVs, and high-risk EVs (sensitivity, 97.1%, 96.6%, 95.8%, and 93.8%, respectively; NLR, 0.051, 0.056, 0.054, and 0.074, respectively) (Table 4). Among 24 patients with EVs, 23 (95.8%) had high SS values ( 3.36 m/sec) (Fig 4), while one patient had a low SS value (3.18 m/sec) and had small EVs without the red color sign. No high-risk EV was misdiagnosed with the SS cutoff value at 3.36 m/sec, which was proposed to rule out EVs. Moreover, no EVs were observed when the SS cutoff value was set at 3.10 m/sec, which was proposed to rule out clinically important portal hypertension. Discussion We found that the AUCs of SS for identifying clinically important portal hypertension, severe portal hypertension, EVs, and high-risk EVs were excellent (0.943, 0.963, 0.937, and 0.955, respectively). Similar to results of previous studies (23,24), our results showed that the AUC for identifying EVs with LS measured at ARFI imaging was 0.789, and LS was therefore considered useful, but not excellent. Thus, these results suggest that, in most patients evaluated by using SS, SS could be used to diagnose clinically important portal hypertension and severe portal hypertension associated with EVs and high-risk EVs. Thus, if our results are corroborated by future prospective studies, SS could potentially be used as an indication for screening endoscopy and prophylactic treatments such as b- blocker therapy or endoscopic variceal ligation. Moreover, the SS cutoff value of 3.51 m/sec proposed to rule out high-risk EVs was able to exclude the EVs in danger of rupture with a sensitivity of 93.8% and an NLR of 0.074. Thus, an SS cutoff value of 3.51 m/sec is a good threshold for the indication of prophylactic treatment for EV rupture. Furthermore, no high-risk EVs were misdiagnosed with the SS cutoff value of 3.36 m/sec proposed to rule out EVs, and no EV was observed when the SS cutoff value was less than the 3.10 m/sec proposed to rule out clinically important portal hypertension. In addition, the cutoff values of SS for EVs and high-risk EVs in the present study were higher (3.36 and 3.51 m/sec, respectively) than those in our prior study (11) (3.18 and 3.30 m/sec, respectively), because our present study had different risk factors for inclusion compared with our previous study. EV formation may start when HVPG increases to greater than 10 mm Hg (25), and clinical decompensation in the form of EVs and EV rupture may develop when HVPG increases over a threshold value of 10 12 mm Hg (7,26). Moreover, EV rupture does not occur when the HVPG is less than 12 mm Hg (25). In this study, we clearly demonstrated that the correlation coefficient between SS and HVPG was significantly higher than that between LS and HVPG (r = 0.876 vs r = 0.609, P,.0001). Furthermore, the correlation coefficient between SS and HVPG was significantly higher than that between LS and HVPG (r = 0.764 vs r = 0.426, P =.017) in patients with an HVPG of 10 mm Hg or greater. Radiology: Volume 279: Number 2 May 2016 n radiology.rsna.org 615

Table 4 Cutoff Values and Diagnostic Accuracies of SS for Association with Portal Hypertension and EVs Cutoff Value (m/sec) Sensitivity Specificity PPV NPV PLR NLR Accuracy Parameter 3.10 33/34 (97.1) [84.7, 99.9] 15/26 (57.7) [36.9, 76.6] 33/44 (75.0) [59.7, 86.8] 15/16 (93.7) [69.8, 99.8] 2.294 [1.459, 3.607] 0.051 [0.007, 0.361] 48/60 (80.0) [68.2, 88.2] Clinically important portal hypertension 3.15 28/29 (96.6) [82.2, 99.9] 19/31 (61.3) [42.2, 78.2] 28/40 (70.0) [53.5, 83.4] 19/20 (95.0) [75.1, 99.9] 2.494 [1.593, 3.905] 0.056 [0.008, 0.394] 47/60 (78.3) [66.4, 86.9] Severe portal hypertension EVs 3.36 23/24 (95.8) [78.9, 99.9] 28/36 (77.8) [60.8, 89.9] 23/31 (74.2) [55.4, 88.1] 28/29 (96.6) [82.2, 99.9] 4.312 [2.327, 7.991] 0.054 [0.008, 0.368] 51/60 (85.0) [73.9, 91.9] High-risk EVs 3.51 15/16 (93.8) [69.8 99.8] 37/44 (84.1) [69.9, 93.4] 15/22 (68.2) [45.1, 86.1) 37/38 (97.4) [86.2, 99.9] 5.893 [2.953, 11.761] 0.074 [0.011, 0.498] 52/60 (86.7) [75.8, 93.1] Note. Data in parentheses are percentages, with 95% CIs in square brackets. A few investigators have assessed the correlation coefficient between SS and HVPG. Hirooka et al (27) reported that although both SS and LS values obtained by using real-time tissue elastography were linearly correlated with HVPG, the r value was higher for SS than for LS (r = 0.854 and r = 0.510, respectively). Colecchia et al (8) concluded that both SS and LS values obtained by using transient elastography were strongly correlated with HVPG (r = 0.885 and r = 0.836, respectively). However, these previous studies did not provide the statistical comparison of the strength of the two correlation coefficients (SS and HVPG vs LS and HVPG). Talwalkar et al (28) suggested that SS values greater than 10.5 kpa measured at magnetic resonance elastography in compensated cirrhosis were associated with EVs in all patients. These results indicate that additional assessment of SS may be better than measuring LS for identifying portal hypertension and EVs. In a systematic review (29) of 12 studies of the diagnostic accuracy of transient elastography based LS measurements, the diagnostic odds ratios for detecting the presence of any EV and large EVs were 7.5 (95% CI: 4.5, 12.7) and 8.8 (95% CI: 5.9, 13.2), respectively. Comparable diagnostic odds ratios for SS were 19.3 and 12.6, respectively, and the diagnostic performance of SS was significantly better than that of LS. SS is potentially characterized by a wider range of applications than is LS, because SS directly reflects the hemodynamic changes of extrahepatic factors caused by increases in portal pressure such as hyperdynamic splanchnic circulation and portosystemic collateral vessels (4,8,29). An elevated SS value may be caused not only by congestion of the red pulp and tissue hyperplasia but also by diffuse fibrosis of spleen trabeculae (30,31). Vizzutti et al (4) reported that strong correlation between HVPG and LS was observed in patients with an HVPG of less than 10 mm Hg and in those with an HVPG of less than 12 mm Hg (r = 0.81, P,.0001 and r = 0.91, P,.0001, respectively). In contrast, in 616 radiology.rsna.org n Radiology: Volume 279: Number 2 May 2016

Figure 4 Figure 4: Dot diagrams show findings of SS in the absence and presence of, A, clinically important portal hypertension, B, severe portal hypertension, C, EVs, and, D, high-risk EVs in patients with cirrhosis. Solid line = optimal cutoff value. Dot = a patient. our study, there was poor correlation between HVPG and LS in patients with an HVPG of less than 10 mm Hg (r = 0.034, P =.869). The LS measuring range of Vizzutti and colleagues was very large (approximately 3 60 kpa) compared with that in the present study (1.32 3.95 m/ sec). Differences in patient populations might in part explain this difference. More recently, Elkrief et al (32) reported that the diagnostic performance of LS measured by using shear-wave elastography was significantly better than that of SS for the diagnosis of clinically important portal hypertension. We infer that a difference in the prevalence of decompensated liver cirrhosis between their study and the present study (69.6% vs 31.7%) might be one of the reasons for this discrepancy. An important clinical advantage of the ARFI device is that the success rate of ARFI measurements is higher than that of transient elastography measurements (33,34). In fact, the rates of unsuccessful measurements of LS and SS at ARFI imaging in our study were 0% and 3.2%, respectively. In contrast, rates of unsuccessful measurements of LS and SS at transient elastography have been reported to be 18.9% (9) and 14.6% (23), respectively. Furthermore, in a prior study (11), we showed that LS and SS measurements consistently had excellent reproducibility, with intraobserver and interobserver agreement intraclass correlation coefficients of ARFI measurements of 0.97 and 0.97 for LS and 0.98 and 0.98 for SS, respectively. We found that SS measurement at ARFI imaging had a high diagnostic performance for defining the degree of portal hypertension (HVPG 10 or 12 mm Hg). However, our study was a cross-sectional study. More recently, Colecchia et al (35) reported that SS value obtained by using transient elastography represented an accurate predictor of clinical decompensation, Radiology: Volume 279: Number 2 May 2016 n radiology.rsna.org 617

with an accuracy at least equivalent to that of HVPG, in patients with compensated cirrhosis in a longitudinal cohort study. Therefore, a longitudinal cohort study needs to be performed of the prediction of complications of cirrhosis with ARFI imaging. The limitations of this study included the facts that the sample size of 60 participants was relatively small, the timing of SS measurements was not fixed, and this was a single-center study without external validation. We used the current population to set our thresholds for analysis, which can lead to an overestimation of results. We did this because our present population had different risk factors for inclusion than did the population in our previous study. In conclusion, SS measured by using ARFI imaging provides excellent diagnostic performance for identifying portal hypertension in liver cirrhosis. Our findings are complementary to those of previous studies that suggest a role for SS measurements in the treatment of patients with liver cirrhosis. The measurement of SS could help select suitable patients for procedures such as screening endoscopy or prophylactic treatment for decompensation. Disclosures of Conflicts of Interest: Y.T. disclosed no relevant relationships. K.N. disclosed no relevant relationships. Y.M. disclosed no relevant relationships. J.T. disclosed no relevant relationships. A.S. disclosed no relevant relationships. H.T. disclosed no relevant relationships. K.M. disclosed no relevant relationships. H.Y. disclosed no relevant relationships. References 1. D Amico G, Garcia-Tsao G, Pagliaro L. Natural history and prognostic indicators of survival in cirrhosis: a systematic review of 118 studies. J Hepatol 2006;44(1):217 231. 2. Bosch J, Garcia-Pagán JC, Berzigotti A, Abraldes JG. Measurement of portal pressure and its role in the management of chronic liver disease. Semin Liver Dis 2006; 26(4):348 362. 3. Jensen DM. Endoscopic screening for varices in cirrhosis: findings, implications, and outcomes. Gastroenterology 2002;122(6):1620 1630. 4. Vizzutti F, Arena U, Romanelli RG, et al. Liver stiffness measurement predicts severe portal hypertension in patients with HCV-related cirrhosis. Hepatology 2007;45(5):1290 1297. 5. Bureau C, Metivier S, Peron JM, et al. Transient elastography accurately predicts presence of significant portal hypertension in patients with chronic liver disease. Aliment Pharmacol Ther 2008;27(12):1261 1268. 6. Lemoine M, Katsahian S, Ziol M, et al. Liver stiffness measurement as a predictive tool of clinically significant portal hypertension in patients with compensated hepatitis C virus or alcohol-related cirrhosis. Aliment Pharmacol Ther 2008;28(9):1102 1110. 7. Castera L, Pinzani M, Bosch J. Non invasive evaluation of portal hypertension using transient elastography. J Hepatol 2012;56(3): 696 703. 8. Colecchia A, Montrone L, Scaioli E, et al. Measurement of spleen stiffness to evaluate portal hypertension and the presence of esophageal varices in patients with HCV-related cirrhosis. Gastroenterology 2012;143(3): 646 654. 9. Castéra L, Foucher J, Bernard PH, et al. Pitfalls of liver stiffness measurement: a 5-year prospective study of 13,369 examinations. Hepatology 2010;51(3):828 835. 10. Friedrich-Rust M, Nierhoff J, Lupsor M, et al. Performance of Acoustic Radiation Force Impulse imaging for the staging of liver fibrosis: a pooled meta-analysis. J Viral Hepat 2012;19(2):e212 e219. 11. Takuma Y, Nouso K, Morimoto Y, et al. Measurement of spleen stiffness by acoustic radiation force impulse imaging identifies cirrhotic patients with esophageal varices. Gastroenterology 2013;144(1):92 101.e2. 12. Di Lelio A, Cestari C, Lomazzi A, Beretta L. Cirrhosis: diagnosis with sonographic study of the liver surface. Radiology 1989;172(2): 389 392. 13. Schuppan D, Afdhal NH. Liver cirrhosis. Lancet 2008;371(9615):838 851. 14. Takahashi H, Ono N, Eguchi Y, et al. Evaluation of acoustic radiation force impulse elastography for fibrosis staging of chronic liver disease: a pilot study. Liver Int 2010; 30(4):538 545. 15. Yoneda M, Suzuki K, Kato S, et al. Nonalcoholic fatty liver disease: US-based acoustic radiation force impulse elastography. Radiology 2010;256(2):640 647. 16. Friedrich-Rust M, Wunder K, Kriener S, et al. Liver fibrosis in viral hepatitis: noninvasive assessment with acoustic radiation force impulse imaging versus transient elastography. Radiology 2009;252(2):595 604. 17. Giannini E, Botta F, Borro P, et al. 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 2003;52(8):1200 1205. 18. 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(4):762 768. 19. Idezuki Y. General rules for recording endoscopic findings of esophagogastric varices (1991). Japanese Society for Portal Hypertension. World J Surg 1995;19(3):420 422; discussion 423. 20. Steiger JH. Tests for comparing elements of a correlation matrix. Psychol Bull 1980;87(2): 245 251. 21. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988;44(3):837 845. 22. Gerds TA, Andersen PK, Kattan MW. Calibration plots for risk prediction models in the presence of competing risks. Stat Med 2014;33(18):3191 3203. 23. Stefanescu H, Grigorescu M, Lupsor M, Procopet B, Maniu A, Badea R. Spleen stiffness measurement using Fibroscan for the noninvasive assessment of esophageal varices in liver cirrhosis patients. J Gastroenterol Hepatol 2011;26(1):164 170. 24. Foucher J, Chanteloup E, Vergniol J, et al. Diagnosis of cirrhosis by transient elastography (FibroScan): a prospective study. Gut 2006;55(3):403 408. 25. Garcia-Tsao G, Groszmann RJ, Fisher RL, Conn HO, Atterbury CE, Glickman M. Portal pressure, presence of gastroesophageal varices and variceal bleeding. Hepatology 1985;5(3):419 424. 26. Ripoll C, Groszmann R, Garcia-Tsao G, et al. Hepatic venous pressure gradient predicts clinical decompensation in patients with compensated cirrhosis. Gastroenterology 2007;133(2):481 488. 27. Hirooka M, Ochi H, Koizumi Y, et al. Splenic elasticity measured with real-time tissue elastography is a marker of portal hypertension. Radiology 2011;261(3):960 968. 28. Talwalkar JA, Yin M, Venkatesh S, et al. Feasibility of in vivo MR elastographic splenic stiffness measurements in the as- 618 radiology.rsna.org n Radiology: Volume 279: Number 2 May 2016

sessment of portal hypertension. AJR Am J Roentgenol 2009;193(1):122 127. 29. Singh S, Eaton JE, Murad MH, Tanaka H, Iijima H, Talwalkar JA. Accuracy of spleen stiffness measurement in detection of esophageal varices in patients with chronic liver disease: systematic review and meta-analysis. Clin Gastroenterol Hepatol 2014;12(6):935 45.e4. 30. Manenti A, Botticelli A, Gibertini G, Botticelli L. Experimental congestive splenomegaly: histological observations in the rat. Pathologica 1993;85(1100):721 724. 31. Terayama N, Makimoto KP, Kobayashi S, et al. Pathology of the spleen in primary biliary cirrhosis: an autopsy study. Pathol Int 1994;44(10-11):753 758. 32. Elkrief L, Rautou PE, Ronot M, et al. Prospective comparison of spleen and liver stiffness by using shear-wave and transient elastography for detection of portal hypertension in cirrhosis. Radiology 2015;275(2):589 598. 33. Rifai K, Cornberg J, Mederacke I, et al. Clinical feasibility of liver elastography by acoustic radiation force impulse imaging (ARFI). Dig Liver Dis 2011;43(6):491 497. 34. Rizzo L, Calvaruso V, Cacopardo B, et al. Comparison of transient elastography and acoustic radiation force impulse for noninvasive staging of liver fibrosis in patients with chronic hepatitis C. Am J Gastroenterol 2011;106(12):2112 2120. 35. Colecchia A, Colli A, Casazza G, et al. Spleen stiffness measurement can predict clinical complications in compensated HCVrelated cirrhosis: a prospective study. J Hepatol 2014;60(6):1158 1164. Radiology: Volume 279: Number 2 May 2016 n radiology.rsna.org 619