Prediction of posthepatectomy liver failure based on liver stiffness measurement in patients with hepatocellular carcinoma

ARTICLE IN PRESS

Prediction of posthepatectomy liver failure based on liver stiffness measurement in patients with hepatocellular carcinoma Takahiro Nishio, MD,a Kojiro Taura, MD, PhD,a Yukinori Koyama, MD, PhD,a Kazutaka Tanabe, MD,a Gen Yamamoto, MD,a Yukihiro Okuda, MD,a Yoshinobu Ikeno, MD,a Satoru Seo, MD, PhD,a Kentaro Yasuchika, MD, PhD,a Etsuro Hatano, MD, PhD,a Hideaki Okajima, MD, PhD,a Toshimi Kaido, MD, PhD,a Shiro Tanaka, PhD,b and Shinji Uemoto, MD, PhD,a Kyoto, Japan

Background. Posthepatectomy liver failure (PHLF) is a potentially fatal complication, and the accurate prediction of PHLF is essential. Liver stiffness measurement (LSM) has been accepted widely as a noninvasive assessment for liver fibrosis. We aimed to evaluate the usefulness of LSM in predicting PHLF. Methods. One hundred seventy-seven patients with hepatocellular carcinoma who underwent liver resection between August 2011 and October 2014 were analyzed prospectively. LSM was performed by Virtual Touch Tissue Quantification based on acoustic radiation force impulse imaging, and its value was expressed as the shear wave velocity (Vs) [m/s]. The remnant liver volume rate (Rem) was calculated by computed tomography volumetry. PHLF was diagnosed on the basis of the definition from the International Study Group of Liver Surgery. Results. PHLF occurred in 38 patients (21.5%): grade A, 17 patients (9.6%); grade B, 15 patients (8.5%); and grade C, 6 patients (3.4%). The area under the receiver operating characteristic curve of the Vs for predicting PHLF was 0.67 for grade $A, 0.78 for grade $B, and 0.74 for grade C, which was greater than any other preoperative factor for each grade. Multivariate stepwise selection identified 2 significant factors associated with PHLF grade $B: Vs (odds ratio, 2.66; 95% confidence interval, 1.69–4.41, P < .01) and Rem (odds ratio, 0.47; 95% confidence interval, 0.27–0.79, P < .01). The logistic model that included the Vs and Rem resulted in an area under the receiver operating characteristic curve of 0.80 for predicting PHLF grade $B. Conclusion. LSM was useful for the prediction of PHLF and the estimation of the safe Rem range. (Surgery 2015;j:j-j.) From the Department of Surgery,a Graduate School of Medicine; and Department of Pharmacoepidemiology,b Graduate School of Medicine and Public Health, Kyoto University, Kyoto, Japan

LIVER RESECTION HAS BEEN AN EFFECTIVE THERAPEUTIC OPTION IN

PATIENTS 1,2

WITH

HEPATOCELLULAR

CARCINOMA

(HCC). Although advances in diagnosis, operative techniques, and perioperative management This study was supported by the Japan Society for the Promotion of Science (JSPS) Grants-in-Aid for Scientific Research (KAKENHI), Grant Number 24659605. Accepted for publication June 8, 2015. Reprint requests: Kojiro Taura, MD, PhD, Department of Surgery, Graduate School of Medicine, Kyoto University, 54 Kawahara-cho, Shogoin, Sakyo-ku, Kyoto, Japan. E-mail: [email protected]. 0039-6060/$ - see front matter Ó 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.surg.2015.06.024

have contributed to the improved safety and outcomes of liver resection,3,4 excessive liver resections or resections in patients with diseased livers result in insufficient functional volumes of the liver remnant, which increase the risk of posthepatectomy liver failure (PHLF). PHLF is one of the major causes of hepatectomy-related mortality.4-7 The incidence of PHLF has been reported to vary between 1.2 and 32%.5,8-11 The diversity of the reported incidence of PHLF may result partially from the lack of a universally accepted PHLF definition, which impairs the valid comparison of different studies across multiple institutions. To standardize the definition of PHLF, the International Study Group of Liver Surgery (ISGLS) proposed a definition for PHLF in 2010,8 which has SURGERY 1

ARTICLE IN PRESS 2 Nishio et al

been accepted widely. The accurate prediction of PHLF development based on preoperative liver function assessments is required to ensure operative safety and to establish the appropriate indications for liver resection. The prognosis of chronic liver disease is highly dependent on the extent of liver fibrosis, which is a common consequence of advanced liver injury. Precise evaluation of liver fibrosis is a key step to make treatment decisions in patients with chronic liver disease. Liver biopsy has been one of the standard procedures to assess live fibrosis,12 but it has several limitations, such as complications, sampling error, intra- and interobserver variability, and expense.13 It may be unsuitable for a routine use, especially in the setting of preoperative assessment. In the past decade, noninvasive markers of liver fibrosis have been developed as an alternative to liver biopsy, such as hyaluronic acid, aspartate aminotransferase to platelet ratio index (APRI), and FIB4 index.14-16 Liver stiffness measurement is a recently developed and noninvasive method for assessing liver fibrosis17 that has high accuracy and reproducibility in predicting liver fibrosis in patients with chronic liver disease.18-20 To date, most studies have evaluated the performance of transient elastography with the use of Fibroscan (Echosens, Paris, France).20 Recently, Virtual Touchtissue quantification (VTTQ) based on acoustic radio force impulse (ARFI) imaging technology can also be applied as a device to measure liver stiffness.21,22 The accuracy of estimating the degree of liver fibrosis using ARFI was reported to be similar to that of Fibroscan.22-24 Measurement by Fibroscan, which is based on the M-mode and Amode imaging of ultrasonography, is not a realtime procedure. Its potential limitation is that the measurement is difficult in obese patients with narrow intercostal spaces or patients with ascites, and it is affected by the operator’s experience.25 In contrast, ARFI technology, which is a simple real-time procedure based on B-mode imaging, makes it possible to observe the region of interest (ROI), adapt measurement depth according to the skin to liver capsule distance, and measure slim and obese patients as well as patients with ascites.26 This may be particularly advantageous in the preoperative examination setting for patients with liver tumors because the examiners are able to set the ROI in the nontumor area. The relationship between the liver stiffness value and the various complications of cirrhosis such as esophageal varices, portal hypertension,

Surgery j 2015

ascites, development of HCC, and survival has been reported in several studies.27-29 There recently have been several research studies that have investigated the impact of liver stiffness on morbidity and mortality after liver resection30-32; however, few focused on PHLF development according to the universally standardized definition or accounted for the remnant liver volume. Our aim was to assess the usefulness of liver stiffness measured by ARFI for the prediction of PHLF based on the ISGLS definition in patients with HCC. Furthermore, we aimed to establish a novel index that quantitatively assesses the risk of PHLF and enables an estimation of the safety range of the future remnant liver volume. METHODS Patients. We prospectively collected and analyzed data on 177 patients with HCC who underwent liver resection at Kyoto University Hospital, Kyoto, between August 2011 and October 2014. The exclusion criteria were patients undergoing combined resection of an extrahepatic duct and patients undergoing preoperative portal vein embolization. This study was performed in accordance with the ethical guidelines for epidemiological research in Japan and was approved by the Ethics Committee of the Kyoto University Graduate School and Faculty of Medicine (approval code: E1258). Written informed consent was obtained from all study patients. This study was registered with the University Hospital Medical Information Network (unique trial number: UMIN000007172). Data collection. All patients received blood examinations, including a routine liver function test (platelet count, international normalized ratio of prothrombin time [PT-INR], total bilirubin, alanine transaminase, albumin, and ammonia), the measurement of fibrosis markers (hyaluronic acid and type 4 collagen), an indocyanine green retention test, liver stiffness measurement, and computed tomography for preoperative assessment. The total liver volume and the estimated liver resection volume were measured using computed tomography volumetry with a volume analyzer system (Synapse Vincent, Fujifilm, Japan).33,34 The remnant liver volume rate (Rem) was defined as follows: (total liver volume [mL]
resection volume [mL])/total liver volume [mL]. The resection volume was calculated using computed tomography volumetry for major anatomical resections (sectionectomy or more; n = 109), whereas it was substituted by the weight of the resected specimen for minor non-anatomical resections (less than

ARTICLE IN PRESS Surgery Volume j, Number j

sectionectomy; n = 68). The plasma disappearance rate of indocyanine green (KICG) was calculated, and the KICG of the remnant liver (remKICG) was estimated as KICG 3 Rem.35 The APRI and FIB-4 index were calculated as follows: APRI = aspartate transaminase/(upper limit of normal) 3 100/ platelet count [109/L]14; FIB-4 index = (age [in years] 3 aspartate aminotransferase [U/L])/ (platelet count [109/L] 3 alanine aminotransferase [U/L]1/2).15 The pathologic liver fibrosis stage of the resected liver tissue was evaluated according to the Metavir score.36 Measurement of liver stiffness. Liver stiffness was measured using an ACUSON S2000 (Mochida Siemens Medical Systems, Tokyo, Japan) equipped with ARFI-based VTTQ system. In the VTTQ system, the value of tissue stiffness in a userplaced ROI is expressed as the shear wave velocity (Vs) in meters per second (m/s). The liver stiffness measurement was performed in a fasting state within a few days before the surgery. The examined patients were laid in the supine position. The ROI (fixed-dimension 1 3 0.5-cm box) was chosen in the right lobe of the liver in accordance with the previous report,37 from the intercostal space at a depth 4–6 cm from the surface and free of large vascular structures and tumors. Vs values were measured 10 times in every patient and the mean value in m/s was calculated. Valid ARFI measurements were obtained in all 177 patients. The operators were blinded to the clinical data. Operative procedures. The indications for liver resection included a Child-Pugh status of A or B and a future Rem of more than 25% of the whole liver. The liver resection techniques and perioperative care were standard. Open liver resection generally was performed through a bilateral subcostal incision with a vertical midline extension. Intraoperative ultrasonography was performed to detect any major vascular invasion and to mark the transection plane. Transection of the liver parenchyma was performed using a Cavitron ultrasonic surgical aspirator (CUSA; Valleylab, Boulder, NY) and bipolar cautery with a saline irrigation system.38 For laparoscopic liver resections, a combination of CUSA, SonoSurg (Olympus, Tokyo, Japan), BiClamp (ERBE Elektromedizin, T€ ubingen, Germany) tools was used for liver transection. The Pringle maneuver was applied as indicated. Blood tests were routinely performed on postoperative days 1, 2, 3, 5, and 7 and were added according to the patient condition. Definition of operative outcomes. The primary outcome of this study was PHLF development, which was diagnosed based on the ISGLS

Nishio et al 3

definition.8 The ISGLS defines PHLF as an increased PT-INR and concomitant hyperbilirubinemia on or after postoperative day 5. The severity is categorized into three grades, as follows: Grade A, PHLF resulting in abnormal laboratory parameters but requiring no change in the clinical management of the patient; Grade B, PHLF that results in a deviation from the regular clinical management but is manageable without invasive treatment; and Grade C, PHLF that results in a deviation from the regular clinical management and requires invasive treatment. Statistical analysis. Categorical variables were described as frequencies and percentages and were compared using the v2 or Fisher exact test as appropriate. Continuous variables were described as the mean ± standard deviation and were compared using the Mann-Whitney U test. A PHLF prediction model focusing on the outcome of grade $B was built based on logistic regression analysis, and the adjusted odds ratio (OR) per standard deviation change and the 95% confidence interval (CI) were calculated. The predictors were selected through a stepwise procedure using the minimum AICc method among the preoperative liver function indicators (platelet count, PT-INR, total bilirubin, alanine transaminase, albumin, ammonia, KICG, Vs, hyaluronic acid, type 4 collagen) and Rem. The interaction between the variables was tested by Spearman’s rank correlation coefficient (r). The variance inflation factor (VIF) was used to check multicollinearity. We also performed internal validation of the model in terms of discriminatory power and calibration. A receiver operating characteristic analysis was performed to evaluate the discriminatory power of predictors, and the area under the receiver operating characteristic curve (AUROC) and the 95% CI were calculated. Comparison between AUROCs was made using DeLong test. Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were calculated. The Hosmer-Lemeshow goodness-of-fit test was used to calibrate the PHLF prediction model. JMP pro 11 software (SAS institute Inc. Cary, NC) was used in all statistical analyses. The significance level was set at 0.05, and the P values were 2-sided. RESULTS Patient characteristics according to development of PHLF. The characteristics of the studied population are shown in Table I. The underlying liver diseases were: chronic hepatitis B (positive for hepatitis B surface antigen; n = 33, 18.6%); chronic hepatitis C (positive for hepatitis C

ARTICLE IN PRESS 4 Nishio et al

Surgery j 2015

Table I. Study population and patient characteristics according to the development of PHLF Characteristic General background Sex Male Female Age, y Body mass index, kg/m2 Underlying liver disease HB HC HB and HC NBNC Routine blood examination Platelet count, 3109/L PT-INR Total bilirubin, mg/dL Alanine transaminase, IU/L Albumin, g/dL Ammonia, mg/dL Indocyanine green test KICG Liver stiffness Vs, m/s Fibrosis marker Hyaluronic acid, ng/mL Type 4 collagen, ng/mL APRI FIB4 index Compound score Child-Pugh score MELD score Surgical factor Rem Operative time, min Operative blood loss, g Liver fibrosis Metavir score F4 F0 to 3 Operative outcome PHLF, + Grade A Grade B Grade C

P value

Total (N = 177)

PHLF (+) (n = 38)

PHLF (
) (n = 139)

140 (79.1%) 37 (20.9%) 68 ± 10 23.0 ± 3.8

32 (84.2%) 6 (15.8%) 69 ± 9 22.9 ± 3.4

108 (77.7%) 31 (22.3%) 68 ± 10 23.1 ± 3.9

.37

33 66 3 75

3 17 1 17

30 49 2 58

.20

(18.6%) (37.3%) (1.7%) (42.4%)

152 1.08 0.90 40 3.8 45

± ± ± ± ± ±

58 0.09 0.39 42 0.4 18

(7.8%) (44.8%) (2.6%) (44.8%)

132 1.09 1.04 44 3.7 46

± ± ± ± ± ±

52 0.08 0.43 30 0.5 18

(21.6%) (35.3%) (1.4%) (41.7%)

157 1.07 0.86 39 3.8 45

± ± ± ± ± ±

.65 .98

58 0.09 0.36 44 0.4 18

.01 .07 <.01 .08 .37 .62

0.14 ± 0.04

0.13 ± 0.03

0.14 ± 0.04

.17

1.86 ± 0.78

2.16 ± 0.87

1.77 ± 0.74

<.01

171 5.5 1.2 4.0

183 6.0 1.4 4.7

167 5.4 1.2 3.8

554 2.6 2.4 3.4

.01 .61 <.01 <.01

± ± ± ±

495 2.7 2.2 3.3

± ± ± ±

194 3.2 1.2 2.9

± ± ± ±

5.5 ± 0.7 7.8 ± 1.5

5.6 ± 0.8 7.8 ± 1.4

5.5 ± 0.7 7.7 ± 1.5

.17 0.60

0.79 ± 0.19 412 ± 163 1,039 ± 1,382

0.71 ± 0.19 461 ± 162 1,806 ± 1,902

0.81 ± 0.18 399 ± 161 829 ± 1,122

<.01 .03 <.01

41 (24.1%) 129 (75.9%)

14 (36.8%) 24 (63.2%)

27 (20.5%) 105 (79.5%)

.04

38 17 15 6

(21.5%) (9.6%) (8.5%) (3.4%)

Data represent the mean ± standard deviation or the number of patients. Significant P values are represented in bold. The Metavir score was not determined in 7 cases because of their small parenchymal regions. APRI, Aspartate aminotransferase to platelet ratio index; HB, hepatitis B; HC, hepatitis C; KICG, plasma disappearance rate of indocyanine green; MELD, model for end-stage liver disease; NBNC, non-hepatitis B and non-hepatitis C; PHLF, posthepatectomy liver failure; PT-INR, international normalized ratio of prothrombin time; Rem, remnant liver volume rate; Vs, shear wave velocity.

antibody, n = 66, 37.3%); both hepatitis B and C (n = 3, 1.7%); and non-hepatitis B and nonhepatitis C (non-B non-C, negative for hepatitis B surface antigen and hepatitis C antibody; n = 75, 42.4%). Among the non-B non-C HCC patients, 29 patients had alcoholic liver disease, 23 patients

had non alcoholic fatty liver disease, 1 patient had autoimmune hepatitis, 1 patient had primary biliary cirrhosis, 1 patient had Budd-Chiari syndrome, and the other 20 patients were cryptogenic. PHLF occurred in 38 patients (21.5%): 17 patients (9.6%) were categorized as Grade A, 15

ARTICLE IN PRESS Nishio et al 5

Surgery Volume j, Number j

Table II. Receiver operating characteristic analysis of fibrosis markers for predicting the liver fibrosis stage Stage $F2

Vs Hyaluronic acid Type 4 collagen APRI FIB4 index

Stage $F3

Stage F4

AUROC

95% CI

AUROC

95% CI

AUROC

95% CI

0.74 0.66 0.70 0.72 0.70

(0.66–0.81) (0.57–0.74) (0.60–0.78) (0.64–0.79) (0.61–0.77)

0.79 0.76 0.80 0.78 0.76

(0.71–0.85) (0.67–0.83) (0.72–0.86) (0.70–0.84) (0.67–0.82)

0.84* 0.77 0.81 0.75 0.74

(0.75–0.90) (0.67–0.85) (0.72–0.88) (0.65–0.83) (0.64–0.81)

DeLong test: *P < .05 vs FIB4 index. APRI, Aspartate aminotransferase to platelet ratio index; AUROC, area under the receiver operating characteristic curve; CI, confidence interval; Vs, shear wave velocity.

patients (8.5%) as Grade B, and 6 patients (3.4%) as Grade C. The underlying liver diseases according to the prevalence of hepatitis B or C virus infection did not affect the development of PHLF. A lower platelet count and greater total bilirubin were associated with patients with PHLF. Among the noninvasive indicators of fibrosis, the Vs, hyaluronic acid, APRI, and the FIB4 index were greater in patients with PHLF, which was consistent with the greater proportion of pathologic cirrhosis diagnosed by the resected liver tissue pathology. Regarding the surgical factors, the Rem was smaller, the surgical time was longer, and blood loss was greater in patients with PHLF. Prediction of pathologic liver fibrosis. The performance of the Vs and other fibrosis markers in predicting the liver fibrosis stage was analyzed (Table II, and the cut off values are shown in Supplementary Table I). The Vs had high an AUROC for the prediction of either stage of fibrosis and, in particular, had the greatest AUROC (0.84; 95% CI, 0.75–0.90) for the prediction of F4. The mean Vs values increased according to the progression of liver fibrosis: 1.36 ± 0.25 m/s in F0 (n = 22), 1.44 ± 0.40 m/s in F1 (n = 28), 1.60 ± 0.50 m/s in F2 (n = 49), 1.90 ± 0.73 m/s in F3 (n = 30), and 2.66 ± 0.88 m/s in F4 (n = 41). Prediction of PHLF by preoperative factors. We analyzed the performance of liver function indicators for the prediction of PHLF (Table III, and the cut off values are shown in Supplementary Table II). The discriminatory power for predicting every grade of PHLF was compared by receiver operating characteristic analysis among 12 preoperative variables: (1) liver function indicators included in a routine blood test: platelet count, PT-INR, total bilirubin, alanine transaminase, albumin, and ammonia; (2) indocyanine green retention test: KICG; (3) liver stiffness: Vs; and (4) liver fibrosis markers: hyaluronic acid, type 4 collagen, the APRI, and

the FIB4 index. The Vs had greater AUROC values for the prediction of every grade of PHLF than any other preoperative factors including other fibrosis indicators. The Vs was particularly useful in predicting grade $B with an AUROC of 0.78 (95% CI 0.68–0.85), which was greater than that of PT-INR, total bilirubin, ammonia, type 4 collagen, and KICG. The Vs cutoff value of 1.61 m/s for the prediction of PHLF grade $B had a sensitivity of 0.90, specificity of 0.58, PPV of 0.22, and NPV of 0.98. PHLF prediction model incorporating Rem. We performed multivariate analyses to assess the impact of the Rem on PHLF under the condition of given preoperative factors. We particularly focused on a PHLF grade $B, which deviates from the regular clinical course and requires medical intervention. The interactions of 10 preoperative indicators (APRI and FIB4 index, which are commonly derived from the platelet count, and transaminase were excluded from the 12 variables mentioned previously) and the Rem were assessed by Spearman’s rank correlation coefficient (Supplementary Table III). The Vs correlated with the platelet count, hyaluronic acid, and type 4 collagen to a certain extent, with a coefficient of 0.40# jrj #0.51, but not strongly. None of the coefficient values reached jrj = 0.60. The VIF of each variable was less than 5 (the range of VIF: 1.20 to 3.03; platelet count: 1.63, PT-INR: 1.20, total bilirubin: 1.26, alanine transaminase: 1.14, albumin: 1.53, ammonia: 1.15, KICG: 1.59, Vs: 1.59, hyaluronic acid: 2.01, type 4 collagen 3.03, Rem 1.22), indicating permissive multicollinearity. These 10 preoperative factors and the Rem were included in a stepwise procedure (minimum AICc method) to select appropriate variables for the predictive model. The univariate logistic regression analysis of each variable for the prediction of PHLF grade $B and the result of multivariate

ARTICLE IN PRESS 6 Nishio et al

Surgery j 2015

Table III. Receiver operating characteristic analysis of preoperative factors for predicting the PHLF grade Grade $A AUROC Routine blood examination Platelet count PT-INR Total bilirubin Alanine transaminase Albumin Ammonia Indocyanine green test KICG Liver stiffness Vs Fibrosis marker Hyaluronic acid Type 4 collagen APRI FIB4 index

95% CI

Grade $B AUROC

Grade C 95% CI

AUROC

95% CI

0.63 0.59 0.65 0.59 0.55 0.53

(0.53–0.73) (0.49–0.70) (0.55–0.74) (0.49–0.69) (0.44–0.66) (0.42–0.63)

0.68 0.59 0.63 0.70 0.67 0.60

(0.53–0.80) (0.44–0.72) (0.51–0.74) (0.56–0.81) (0.53–0.79) (0.48–0.71)

0.60 0.59* 0.45 0.64 0.68 0.63

(0.29–0.85) (0.28–0.84) (0.29–0.62) (0.41–0.82) (0.37–0.88) (0.42–0.80)

0.57

(0.47–0.67)

0.64

(0.51–0.75)

0.68*

(0.42–0.86)

0.67y,z,x

(0.58–0.75)

0.78y,z,k,{,#

(0.68–0.85)

0.74x,{

(0.58–0.86)

0.64 0.53 0.65x# 0.64

(0.53–0.73) (0.41–0.65) (0.56–0.74) (0.54–0.73)

0.71 0.62 0.73 0.69

(0.58–0.81) (0.43–0.78) (0.63–0.81) (0.58–0.78)

0.35 0.69x 0.68 0.68*

(0.18–0.58) (0.17–0.96) (0.50–0.82) (0.52–0.80)

DeLong test: *P < .05 vs hyaluronic acid. yP < .05 vs albumin. zP < .05 vs PT-INR. xP < .05 vs total bilirubin. kP < .05 vs ammonia. {P < .05 vs KICG. #P < .05 vs type 4 collagen. APRI, Aspartate aminotransferase to platelet ratio index; AUROC, area under receiver operating characteristic curve; CI, confidence interval; KICG, plasma disappearance rate of indocyanine green; PHLF, posthepatectomy liver failure; PT-INR, international normalized ratio of prothrombin time; Vs, shear wave velocity.

stepwise selection are shown in Table IV. Two variables, the Vs (OR 2.66; 95% CI 1.69–4.41) and the Rem (OR, 0.47; 95% CI 0.27–0.79), remained significant factors associated with PHLF grade $B and were maintained in a binary logistic regression model (events per variable: 21/2) as the Vs-Rem index. The equation of the Vs-Rem index (L) is: L ¼ 1:25 3 Vs ½m=s
3:97 3 Rem The probability of PHLF grade $B (PB) is then calculated as follows: PB ¼ 1 = f1 þ expð1:65
LÞg In addition, each combination of a preoperative factor and the Rem was simultaneously included in a logistic regression analysis, and the predictive accuracy of the resulting model designated as preoperative factor-Rem index was compared with the Vs-Rem index (Supplementary Table IV). The Vs-Rem index cutoff value of
0.80 had a sensitivity of 0.95, specificity of 0.60, PPV of 0.24, and NPV of 0.99. The AUROC of the Vs-Rem index for predicting PHLF grade $ B was 0.80

(95% CI 0.70–0.87), which was a better value than that of any other index or that of the single parameter Vs itself (Fig 1). In addition, Vs-Rem index had a greater AUROC than remKICG (AUROC, 0.73; 95% CI 0.59–0.84), a conventional index that also incorporates the impact of the Rem.35 We performed a calibration to validate this model (Fig 2). The total 177 patients were evenly divided into 5 groups according to the rank of the Vs-Rem index (L) value: risk category I (n = 36), II (n = 36), III (n = 35), IV (n = 35), and V (n = 35). The mean L value in each group was
2.17 ± 0.25,
1.59 ± 0.20,
0.91 ± 0.22,
0.14 ± 0.18, and 0.86 ± 0.60, respectively. The mean probability of PHLF grade $B (PB), calculated from the L value (0.022 ± 0.005 in group I, 0.038 ± 0.008 in II, 0.073 ± 0.015 in III, 0.144 ± 0.022 in IV, and 0.321 ± 0.133 in V), corresponded to the observed incidence of PHLF grade $B (0/36 [0.000], 1/36 [0.028], 4/35 [0.114], 6/35 [0.171], and 10/35 [0.286], respectively). The Hosmer-Lemeshow goodness-of-fit test proved that this model was well calibrated (P = .33).

ARTICLE IN PRESS Nishio et al 7

Surgery Volume j, Number j Table IV. Multivariate analysis by stepwise selection for predicting PHLF grade $B Platelet count, 3109/L PT-INR Total bilirubin, mg/dL Alanine transaminase, IU/L Albumin, g/dL Ammonia, mg/dL KICG Vs, m/s Hyaluronic acid, ng/mL Type 4 collagen, ng/mL Rem

OR

95% CI

P value

OR

0.52 1.22 1.36 1.28 0.53 1.21 0.54 2.15 1.08 1.55 0.66

0.29–0.88 0.80–1.82 0.89–2.01 0.89–1.83 0.34–0.80 0.78–1.81 0.31–0.91 1.43–3.28 0.58–1.53 1.06–2.40 0.43–1.03

.02 .30 .13 .14 <.01 .36 .02 <.01 .68 .03 .06

— — — — — — — 2.66 — — 0.47

95% CI

P value

1.69–4.41

<.01

0.27–0.79

<.01

The adjusted ORs per standard deviation change are shown. Significant P values are represented in bold. CI, Confidence interval; KICG, plasma disappearance rate of indocyanine green; OR, odds ratio; PHLF, posthepatectomy liver failure; PT-INR, international normalized ratio of prothrombin time; Rem, remnant liver volume rate; Vs, shear wave velocity.

Fig 1. Receiver operating characteristic analysis of the PHLF risk index for predicting PHLF grade $B. The receiver operating characteristic curves of Vs-Rem index and Vs for the prediction of grade $B are shown (DeLong test: P = .62).

Figure 3 shows the relationship between the Vs and Rem, and PB. Each line indicates the linear function of the Vs and Rem assigned by various L values (L =
1.29, 0.54, 0.27, 0.81, 1.25, and 1.65, in ascending order of risk) based on the estimated probability of PHLF grade $B (PB = .05, .10, .20, .30, .40, and .50, respectively). The graph enables estimation of safety range of Rem when Vs and permissible probability of PHLF grade $B are given. DISCUSSION Despite the advances in operative techniques and perioperative management, PHLF remains a serious complication and a major cause of

mortality after liver resection. The ability to precisely predict PHLF based on preoperative liver function is essential for establishing the appropriate indications for liver resection. Because liver fibrosis is a common consequence of liver injury and it has a strong impact on the prognosis of patients with chronic liver disease, we focused on the utility of the preoperative assessment of liver fibrosis for the prediction of PHLF. In this study, we adopted ARFI-based VTTQ as a method for liver stiffness measurement and evaluated its efficacy in predicting PHLF by comparing with the commonly used preoperative liver function indicators as well as other noninvasive fibrosis markers. Previous studies have reported the usefulness of liver stiffness measurement using Fibroscan as a preoperative assessment tool for liver resection30,31; however, in those previous studies, the definition for PHLF was not standardized, which inhibits the validation of their results among different institutions. In addition, these studies disregarded the Rem, a significant factor influencing the postoperative outcome. We adopted the universally accepted definition for PHLF proposed by ISGLS to enable valid comparison and dissemination of our results. Moreover, we aimed to establish an index that enables quantitative assessment of the risk of PHLF according to the Rem in the operative planning for each patient. We confirmed that the Vs, the liver stiffness value measured by ARFI-based VTTQ system, significantly correlated with the stage of liver fibrosis. The Vs was particularly excellent for the prediction of advanced stage, similarly to previous reports.22,23 On the basis these results, we analyzed the accuracy of the liver stiffness value for the

ARTICLE IN PRESS 8 Nishio et al

Fig 2. Calibration plot of the Vs-Rem index model. Each dot represents the mean probability of PHLF grade $B (PB) calculated from the L value (0.022 ± 0.005 in group I, 0.038 ± 0.008 in II, 0.073 ± 0.015 in III, 0.144 ± 0.022 in IV, and 0.321 ± 0.133 in V) on the x-axis and the observed incidence of PHLF grade $B (0/36 [0.000], 1/36 [0.028], 4/35 [0.114], 6/35 [0.171], and 10/35 [0.286], respectively) on the y-axis.

Fig 3. Relationship between the Vs, Rem, and the probability of PHLF grade $B. Each line indicates the linear function L = 1.25 3 Vs [m/s]
3.97 3 Rem, in which various Vs-Rem index (L) values (L =
1.29, 0.54, 0.27, 0.81, 1.25, and 1.65, in ascending order of risk) are substituted based on the probability of PHLF grade $B (PB) (PB = .05, .10, .20, .30, .40, and .50, respectively).

prediction of PHLF compared with commonly used preoperative indicators and other fibrosis markers. The univariate receiver operating characteristic analysis revealed that the Vs was a useful predictor for PHLF, as demonstrated by its greatest AUROC value for predicting either grade category compared with the other preoperative factors. The

Surgery j 2015

Vs particularly had a greater accuracy for the prediction of PHLF grade $B, which requires some therapeutic management and hence is a great concern in actual clinical settings; suggesting the liver stiffness measured by ARFI was a significant factor for the preoperative risk assessment. We further performed a multivariate analysis and built a model for predicting PHLF grade $B, so as to assess the impact of the Rem as well as to enable the quantitative risk assessment. The Vs and Rem proved to be significant factors, and the obtained equation of this model, the Vs-Rem index, had good diagnostic accuracy for PHLF grade $B, with an AUROC of 0.80. The probability of PHLF grade$B can be calculated from the index value, which helps to estimate the safe Rem range by the preoperative Vs value. To define the specific cutoff value of Vs was possible; however, it might be difficult to simply apply the cutoff to the surgical indication for patients with various clinical backgrounds. We would rather emphasize the usefulness of the quantitative risk assessment, which enables comparison of the calculated risk probability with the benefit of liver resection in individual case. We acknowledge this study has several potential limitations. The number of the outcome of severe PHLF was relatively small, and the further acquisition of cases and the external validation should be future tasks. We narrowed down the predictive variables to 2 significant factors, Vs and Rem, so as to maintain the validity of the predictive model as well as to make it simple to apply the model to future clinical use. ARFI technology is not yet widespread despite its simple and convenient method, which makes it difficult to implement our present results into clinical use in other institutions or under multicenter validation. We hope that the present study contributes to the diffusion and wide acceptance of this technology and that many institutions apply and validate our results. In conclusion, the present study demonstrated the usefulness of liver stiffness measurement using ARFI as a preoperative assessment of liver function. We established an index that quantitatively evaluated the risk of liver resection with the combination of the Vs and Rem that should be useful for surgeons in making therapeutic decisions in patients with HCC. The authors are grateful to Dr Nitta and Dr Mori, who served as clinical investigators and advisors. Study conception and design: Taura, Koyama, Hatano, and Uemoto; acquisition of data: Nishio, Taura, Koyama,

ARTICLE IN PRESS Surgery Volume j, Number j

Tanabe, Yamamoto, Okuda, Ikeno, Seo, Yasuchika, Hatano, Okajima, and Kaido; analysis and interpretation of data: Nishio, Taura, Koyama, Tanabe, Yamamoto, Okuda, Ikeno, and Tanaka; and drafting of manuscript: Nishio, Taura, and Tanaka. All authors participated in the critical revision of the article for important intellectual content, and all authors gave their final approval of the article. SUPPLEMENTARY DATA Supplementary data related to this article can be found online at http://dx.doi.org/10.1016/j.surg.2015.06.024.

REFERENCES 1. de Lope CR, Tremosini S, Forner A, Reig M, Bruix J. Management of HCC. J Hepatol 2012;56(Suppl 1):S75-87. 2. Villanueva A, Hernandez-Gea V, Llovet JM. Medical therapies for hepatocellular carcinoma: a critical view of the evidence. Nat Rev Gastroenterol Hepatol 2013;10:34-42. 3. Ishizawa T, Hasegawa K, Aoki T, Takahashi M, Inoue Y, Sano K, et al. Neither multiple tumors nor portal hypertension are surgical contraindications for hepatocellular carcinoma. Gastroenterology 2008;134:1908-16. 4. Poon RT, Fan ST, Lo CM, Liu CL, Lam CM, Yuen WK, et al. Improving perioperative outcome expands the role of hepatectomy in management of benign and malignant hepatobiliary diseases: analysis of 1222 consecutive patients from a prospective database. Ann Surg 2004;240:698-708. 5. Farges O, Malassagne B, Flejou JF, Balzan S, Sauvanet A, Belghiti J. Risk of major liver resection in patients with underlying chronic liver disease: a reappraisal. Ann Surg 1999; 229:210-5. 6. Schroeder RA, Marroquin CE, Bute BP, Khuri S, Henderson WG, Kuo PC. Predictive indices of morbidity and mortality after liver resection. Ann Surg 2006;243:373-9. 7. Paugam-Burtz C, Janny S, Delefosse D, Dahmani S, Dondero F, Mantz J, et al. Prospective validation of the ‘‘fifty-fifty’’ criteria as an early and accurate predictor of death after liver resection in intensive care unit patients. Ann Surg 2009;249:124-8. 8. Rahbari NN, Garden OJ, Padbury R, Brooke-Smith M, Crawford M, Adam R, et al. Posthepatectomy liver failure: a definition and grading by the International Study Group of Liver Surgery (ISGLS). Surgery 2011;149:713-24. 9. McCormack L, Petrowsky H, Jochum W, Furrer K, Clavien PA. Hepatic steatosis is a risk factor for postoperative complications after major hepatectomy: a matched case-control study. Ann Surg 2007;245:923-30. 10. Mullen JT, Ribero D, Reddy SK, Donadon M, Zorzi D, Gautam S, et al. Hepatic insufficiency and mortality in 1,059 noncirrhotic patients undergoing major hepatectomy. J Am Coll Surg 2007;204:854-62. 11. Kawano Y, Sasaki A, Kai S, Endo Y, Iwaki K, Uchida H, et al. Short- and long-term outcomes after hepatic resection for hepatocellular carcinoma with concomitant esophageal varices in patients with cirrhosis. Ann Surg Oncol 2008;15: 1670-6. 12. Saadeh S, Cammell G, Carey WD, Younossi Z, Barnes D, Easley K. The role of liver biopsy in chronic hepatitis C. Hepatology 2001;33:196-200. 13. Bedossa P, Dargere D, Paradis V. Sampling variability of liver fibrosis in chronic hepatitis C. Hepatology 2003;38:1449-57.

Nishio et al 9

14. Wai CT, Greenson JK, Fontana RJ, Kalbfleisch JD, Marrero JA, Conjeevaram HS, et al. A simple noninvasive index can predict both significant fibrosis and cirrhosis in patients with chronic hepatitis C. Hepatology 2003;38:518-26. 15. Sterling RK, Lissen E, Clumeck N, Sola R, Correa MC, Montaner J, et al. Development of a simple noninvasive index to predict significant fibrosis in patients with HIV/HCV coinfection. Hepatology 2006;43:1317-25. 16. Nunes D, Fleming C, Offner G, Craven D, Fix O, Heeren T, et al. Noninvasive markers of liver fibrosis are highly predictive of liver-related death in a cohort of HCV-infected individuals with and without HIV infection. Am J Gastroenterol 2010;105:1346-53. 17. Ziol M, Handra-Luca A, Kettaneh A, Christidis C, Mal F, Kazemi F, et al. Noninvasive assessment of liver fibrosis by measurement of stiffness in patients with chronic hepatitis C. Hepatology 2005;41:48-54. 18. Castera L, Vergniol J, Foucher J, Le Bail B, Chanteloup E, Haaser M, et al. Prospective comparison of transient elastography, Fibrotest, APRI, and liver biopsy for the assessment of fibrosis in chronic hepatitis C. Gastroenterology 2005;128:343-50. 19. Fraquelli M, Rigamonti C, Casazza G, Conte D, Donato MF, Ronchi G, et al. Reproducibility of transient elastography in the evaluation of liver fibrosis in patients with chronic liver disease. Gut 2007;56:968-73. 20. Friedrich-Rust M, Ong MF, Martens S, Sarrazin C, Bojunga J, Zeuzem S, et al. Performance of transient elastography for the staging of liver fibrosis: a meta-analysis. Gastroenterology 2008;134:960-74. 21. Fahey BJ, Nightingale KR, Nelson RC, Palmeri ML, Trahey GE. Acoustic radiation force impulse imaging of the abdomen: demonstration of feasibility and utility. Ultrasound Med Biol 2005;31:1185-98. 22. Friedrich-Rust M, Wunder K, Kriener S, Sotoudeh F, Richter S, Bojunga J, et al. Liver fibrosis in viral hepatitis: noninvasive assessment with acoustic radiation force impulse imaging versus transient elastography. Radiology 2009;252:595-604. 23. Ebinuma H, Saito H, Komuta M, Ojiro K, Wakabayashi K, Usui S, et al. Evaluation of liver fibrosis by transient elastography using acoustic radiation force impulse: comparison with Fibroscan(Ò). J Gastroenterol 2011;46:1238-48. 24. Sporea I, Bota S, Peck-Radosavljevic M, Sirli R, Tanaka H, Iijima H, et al. Acoustic Radiation Force Impulse elastography for fibrosis evaluation in patients with chronic hepatitis C: an international multicenter study. Eur J Radiol 2012;81: 4112-8. 25. Castera L, Foucher J, Bernard PH, Carvalho F, Allaix D, Merrouche W, et al. Pitfalls of liver stiffness measurement: a 5-year prospective study of 13,369 examinations. Hepatology 2010;51:828-35. 26. Friedrich-Rust M, Nierhoff J, Lupsor M, Sporea I, Fierbinteanu-Braticevici C, Strobel D, et al. Performance of Acoustic Radiation Force Impulse imaging for the staging of liver fibrosis: a pooled meta-analysis. J Viral Hepat 2012;19:e212-9. 27. Foucher J, Chanteloup E, Vergniol J, Castera L, Le Bail B, Adhoute X, et al. Diagnosis of cirrhosis by transient elastography (FibroScan): a prospective study. Gut 2006;55:403-8. 28. Masuzaki R, Tateishi R, Yoshida H, Goto E, Sato T, Ohki T, et al. Prospective risk assessment for hepatocellular carcinoma development in patients with chronic hepatitis C by transient elastography. Hepatology 2009;49:1954-61.

ARTICLE IN PRESS 10 Nishio et al

29. Vergniol J, Foucher J, Terrebonne E, Bernard PH, le Bail B, Merrouche W, et al. Noninvasive tests for fibrosis and liver stiffness predict 5-year outcomes of patients with chronic hepatitis C. Gastroenterology 2011;140:1970-9. 30. Wong JS, Wong GL, Chan AW, Wong VW, Cheung YS, Chong CN, et al. Liver stiffness measurement by transient elastography as a predictor on posthepatectomy outcomes. Ann Surg 2013;257:922-8. 31. Cescon M, Colecchia A, Cucchetti A, Peri E, Montrone L, Ercolani G, et al. Value of transient elastography measured with FibroScan in predicting the outcome of hepatic resection for hepatocellular carcinoma. Ann Surg 2012;256:706-12. 32. Harada N, Shirabe K, Ijichi H, Matono R, Uchiyama H, Yoshizumi T, et al. Acoustic radiation force impulse imaging predicts postoperative ascites resulting from curative hepatic resection for hepatocellular carcinoma. Surgery 2012; 151:837-43. 33. Ohshima S. Volume analyzer SYNAPSE VINCENT for liver analysis. J Hepatobiliary Pancreat Sci 2014;21:235-8. 34. Kubota K, Makuuchi M, Kusaka K, Kobayashi T, Miki K, Hasegawa K, et al. Measurement of liver volume and hepatic

Surgery j 2015

35.

36.

37.

38.

functional reserve as a guide to decision-making in resectional surgery for hepatic tumors. Hepatology 1997;26: 1176-81. Nagino M, Kamiya J, Nishio H, Ebata T, Arai T, Nimura Y. Two hundred forty consecutive portal vein embolizations before extended hepatectomy for biliary cancer: surgical outcome and long-term follow-up. Ann Surg 2006;243: 364-72. Intraobserver and interobserver variations in liver biopsy interpretation in patients with chronic hepatitis C. The French METAVIR Cooperative Study Group. Hepatology 1994;20(1 Pt 1):15-20. Toshima T, Shirabe K, Takeishi K, Motomura T, Mano Y, Uchiyama H, et al. New method for assessing liver fibrosis based on acoustic radiation force impulse: a special reference to the difference between right and left liver. J Gastroenterol 2011;46:705-11. Yamamoto Y, Ikai I, Kume M, Sakai Y, Yamauchi A, Shinohara H, et al. New simple technique for hepatic parenchymal resection using a Cavitron Ultrasonic Surgical Aspirator and bipolar cautery equipped with a channel for water dripping. World J Surg 1999;23:1032-7.