Objectives: Our objective was to assess the accuracy of computed tomography volumetry in estimating liver volume in steatotic patients.
Materials and Methods: We divided 641 liver donors (mean age 27 years, 71% male) into 4 groups according to the extent of steatosis on predonation biopsy: with < 5% comprising group 1, 5% to 9% comprising group 2, 10% to 19% comprising group 3, and ≥ 20% comprising group 4. Graft mass estimation error (%) was calculated as follows: [(computed tomography-measured volume minus graft weight)/graft weight] × 100. We obtained estimation errors, intraclass correlation coefficients, and Pearson correlation coefficients across the groups.
Results: Baseline alanine aminotransferase and gamma-glutamyltranspeptidase values were significantly correlated with extent of steatosis. Mean graft weight and computed tomography-measured volume were 725.9 g and 741.2 mL. Mean estimation errors were comparable (1.5% for group 1, 2.7% for group 2, 3.0% for group 3, and 3.9% for group 4; P = .77). In multivariate linear regression, estimation error was inversely correlated with high-density lipoprotein cholesterol (P = .008). Overall, there was an excellent agreement between measured volume and actual weight, with intraclass correlation coefficients over 0.85 and Pearson correlation coefficients over 0.70 in all groups (P < .001).
Conclusions: Preoperative computed tomography volumetry is an accurate tool for estimating volume regardless of the extent of steatosis.
Key words : Accuracy, Agreement, Fatty liver, Graft mass, Prediction
Liver surgery, including transplant and resection, is increasing worldwide because of advances in surgical techniques, increases in the number of liver donors for living-donor liver transplant, the expansion of transplant criteria, and early detection of primary hepatobiliary malignancies.1,2 Many pretreatment demographic and clinical risk factors for severe postoperative complications have been identified in patients undergoing these surgical procedures: old age, male sex, coagulopathy, hepatic steatosis, liver volume, and so on.3-5 Of these, liver volume is one of the most important factors because it not only influences the recovery of patients after donation or resection but also has implications for graft function in the recipients.6 Therefore, accurate preoperative estimation of liver volume is crucial to the achievement of successful outcomes.
Many reports have demonstrated that liver volumes measured by computed tomography (CT) volumetry are correlated with intraoperative measurements of liver weight, although the 2 are not identical.7-11 In addition, CT volume measurements have been shown to be well-correlated with the values calculated by various numerical formulas for liver weight/volume combining parameters, such as body weight, height, and body surface area.7,8,11 Thus, CT volumetry is widely used for estimating liver volume.12
Because human fat has a density of 0.9 g/cm³ and contains less functional mass (g) than normal liver per unit volume (cm³), there is a possibility of error in estimating liver mass by CT volumetry in patients with hepatic steatosis, especially in severe cases. This idea has been supported by the finding of Chen and associates that liver volume in patients with steatosis decreased after body weight reduction.13 Moreover, Marcos also reported that functional liver mass decreased by 1% with each 1% reduction in fat.12
To test this idea, using data collected from a large cohort of living liver donors with various extents of fatty changes, we measured the accuracy of CT volumetry for estimating actual graft mass in patients with different severities of hepatic steatosis and in normal controls.
Materials and Methods
Healthy subjects who donated right hepatic lobes at our center from November 2004 to April 2012 were identified. Of these, 641 subjects with complete preoperative data, including preoperative liver biopsy and CT volumetry according to the routine protocols of our center,14 were included in this study. The time between CT imaging and liver donation was less than 3 months, with mean (standard deviation [SD]) of 8.8 (16.3) days for all subjects. Body weight changes from the time of preoperative work-up to the time of hepatectomy were negligible, with mean (SD) change of -0.44% (2.89%). Enrolled donors were stratified according to the degree of steatosis detected by preoperative liver biopsy, with < 5% comprising group 1, 5% to 9% comprising group 2, 10% to 19% comprising group 3, and ≥ 20% comprising group 4. The extent of hepatic steatosis was defined by adding the values for macrosteatosis and microsteatosis. There was no other clinically significant damage to the livers, such as inflammation, fibrosis, ballooning, or cirrhosis in any of the donors. This study was approved by the Institutional Review Board of our center, and the requirement to obtain informed consents was waived.
Clinical and laboratory assessment
At the time of preoperative evaluation, we collected demographic and clinical variables including age, sex, body mass index, comorbidities such as hypertension and type 2 diabetes mellitus, liver enzymes including aspartate aminotransferase, alanine aminotransferase, alkaline phosphatase, and gamma-glutamyltranspeptidase, total bilirubin, prothrombin time-international normalized ratio, and lipid profile (total cholesterol, triglyceride, high-density lipoprotein [HDL] cholesterol, and calculated low-density lipoprotein [LDL] cholesterol).
Evaluation of hepatic steatosis
All donors provided written informed consent, and, as a part of predonation evaluation, sonography-guided percutaneous biopsy of the right liver lobe was performed using an intercostal approach with 18-gauge needles (Stericut with coaxial; TSK Laboratory, Tochigi, Japan). Two or more biopsy specimens each approximately 1.5 cm in length were obtained from every subject. Sections were stained with routine hematoxylin and eosin and independently reviewed by an experienced pathologist (EY).
Calculation of preoperative liver volume from computed tomography volumetry
Surgeons with over 5 years of experience in living-donor hepatectomy measured preoperative liver volume by CT volumetry. They manually traced the contour of each hemiliver on a Picture Archiving and Communication System (Petavision II, Hyundai Electronics, Seoul, Korea) with an electronic cursor. Large vessels were subtracted from each image by a vessel segmentation process. The sectional area of each slice on CT images was automatically summed and processed by the arithmetic program embedded in the Petavision II, which generated the volume of each hemiliver. Representative CT images of a subject with 6% hepatic steatosis are presented in Figure 1.
Intraoperative measurement of graft mass
The right hepatic lobes of donors were procured by 3 surgeons (GWS, SH, and TYH), each with more than 5 years of experience in living-donor hepatectomy. After the division of vascular structures, each graft was squeezed gently and shaken to extract the remaining blood from large blood vessels. After a few seconds to permit natural drainage of blood by gravity, the blood-free weight of the graft was measured on an electronic laboratory scale (FD 110; Excell Precision, Jiangsu, China) with a maximum capacity of 7500 g and a standard error of 0.5 g. The scale was calibrated once per week for quality assurance. No cold electrolyte solution or vascular clamp, except for 1 or more small hemoclip (< 1 g in total), was applied to the graft at the time of weighing.
The primary endpoints were the graft mass estimation errors across the 4 groups. Graft mass estimation error (%) was calculated from the following numerical formula: [(liver volume determined by preoperative CT volumetry (mL) minus intraoperative actual graft mass) / intraoperative actual graft mass] × 100, with 1 cm³ of liver graft considered to weigh 1 g.7,15
The chi-square test was used to compare baseline categorical variables, and the Kruskal-Wallis test was used to compare continuous variables across the 4 groups. The calculated graft mass estimation errors (%) were compared across the groups by the Kruskal-Wallis test. The agreement between actual graft mass and CT volumetry-measured liver volume in each group was assessed by the intraclass correlation coefficient using a 1-way random-effects model. Intraclass correlation coefficients were classified as follows:16 poor to fair (below 0.40), moderate (0.41-0.60), excellent (0.61-0.80), and almost perfect (0.81-1.00). The correlation between actual weight and radiologic volume was examined by Pearson test. Univariate linear regression analysis was performed to evaluate the association between graft mass estimation error and predonation variables. All variables, regardless of their probability values, were included in a multivariate analysis (stepwise backward elimination method), and values of P < .05 with 2-tailed probability were considered significant. Data manipulation and statistical analyses were performed with SPSS software version 20 (SPSS: An IBM Company, IBM Corporation, Chicago, IL, USA).
The demographic and clinical characteristics of the enrolled subjects are presented in Table 1. Hepatic steatosis ranged from 0 to 40% (median, 3%). When subjects were stratified according to degree of hepatic steatosis, 341 (53.2%) comprised group 1, 168 (26.2%) comprised group 2, 86 (13.4%) comprised group 3, and 46 (7.2%) comprised group 4.
The mean age (SD) of the enrolled subjects was 27.2 (7.7) years, and 70.8% were male donors. The mean values of alanine aminotransferase and gamma-glutamyltranspeptidase were correlated with the extent of hepatic steatosis, but all were within the reference ranges. The subjects in group 3 had the highest body mass index. No subjects had type 2 diabetes mellitus, and only 1 had hypertension (data not shown). Other variables including total cholesterol, calculated LDL cholesterol, HDL cholesterol, aspartate aminotransferase, alkaline phosphatase, total bilirubin, and prothrombin time-international normalized ratio were comparable across the 4 groups.
Accuracy of computed tomography volumetry according to degree of hepatic steatosis
The mean volume of the right hepatic lobes as determined by preoperative CT volumetry was 741.2 ± 147.7 mL (range, 409.0-1305.0 mL). Intraoperative graft weight ranged from 360.0 g to 1100.0 g, with a mean (SD) value of 725.9 (120.5) g.
The mean liver volume of the right hemiliver measured by CT volumetry was 704.0 mL (interquartile range [IQR], 613-791 mL) in group 1, 758.0 mL (IQR, 667.5-867.3 mL) in group 2, 763.5 mL (IQR, 679.5-877.8 mL) in group 3, and 790.0 mL (IQR, 664.3-887.8 mL) in group 4. The median graft weights measured after procurement of the right hepatic lobes were 700.0 g (IQR, 620.0-770.0 g), 740.0 g (IQR, 670.0-810.0 g), 760.0 g (IQR, 677.5-841.8 g), and 759.0 g (IQR, 690.0-822.5 g) in the corresponding groups (Figure 2).
The calculated mean (SD) errors in estimation of graft mass did not differ significantly across the 4 groups: 1.5% (10.0%) for group 1, 2.7% (13.5%) for group 2, 3.0% (12.5%) for group 3, and 3.9% (17.1%) for group 4 (Figure 3; P = .77).
Agreement as a function of degree of steatosis
The intraclass correlation coefficient analysis showed that the agreement between actual graft mass and liver volume approximated by CT volumetry was almost perfect, with intraclass correlation coefficients of 0.89 (95% confidence interval [CI], 0.87-0.90) for the entire cohort (P < .001), and 0.92 (95% CI, 0.90-0.93) for group 1, 0.86 (95% CI, 0.81-0.89) for group 2, 0.84 (95% CI, 0.76-0.90) for group 3, and 0.84 (95% CI, 0.71-0.91) for group 4 (P < .001 for each case; Table 2). Liver volume measured by CT volumetry was found to be a linear function of actual graft mass, with Pearson correlation coefficients of 0.86 for group 1, 0.77 for group 2, 0.75 for group 3, and 0.73 for group 4, as well as 0.82 for all subjects (P < .001 for each case; Table 2 and Figure 4).
Factors related to errors in estimating graft mass
Univariate analysis revealed that graft mass estimation error was an increasing function of body mass index and a decreasing function of total and HDL cholesterol (Table 3). Only HDL cholesterol remained significant in a multivariate analysis by the stepwise method (R² = 0.011). Low-density lipoprotein cholesterol level was excluded from the multivariate analysis because of its strong correlation with total cholesterol, implying potential collinearity.
In previous reports, CT-based volume measurement has been shown to be an accurate way of estimating actual graft mass and therefore recommended for use in preoperative planning.15,17 However, the effects of replacement of normal liver parenchyma by hepatic steatosis have yet to be evaluated as a potential confounder influencing the accuracy of CT volumetry. Our findings now show that the presence and degree of hepatic steatosis does not in practice affect the accuracy with which CT volume measurements estimate graft weight. This outcome is similar to that of the study of Siriwardana and associates, using 5 popular formulas for estimating graft mass from CT volume; measurements of Chinese liver donors revealed a pattern of estimation errors that was dependent on liver volume and anthropometric values, rather than the extent of steatosis.18
We also found that lower HDL cholesterol level was associated with higher graft mass estimation error. Metabolic syndrome is known to be related to nonalcoholic fatty liver disease,19 and low HDL cholesterol is a vital aspect of metabolic syndrome. Of note is also the finding that more severe hepatic steatosis tended to be associated with a greater error (1.5%, 2.7%, 3.0%, and 3.9% for groups 1, 2, 3, and 4). In addition, although the estimates were generally dependable, their reliability seemed to decline somewhat with the extent of fatty change (intraclass correlation coefficients of 0.92 and 0.86 for groups 1 and 2 and 0.84 for groups 3 and 4; Pearson correlation coefficients of 0.86, 0.77, 0.75, and 0.73 for group 1, 2, 3, and 4). Therefore, it remains possible that the greater the fatty content, the greater the overestimate of hepatic mass by CT. However, this effect, if it exists, is very small.
Hepatic volume is of particular importance in relation to postoperative liver failure following large resection and size-related graft dysfunction after transplant. A sufficient mass of remnant liver has to be guaranteed in subjects who undergo hepatectomy for any reason, including donation or tumor resection, and this requirement is supported by a report that small remnant livers are the main cause of hospital deaths, even in noncirrhotic patients.20 At the same time, liver recipients must receive livers that are able to meet their metabolic demands.21 In small-for-size grafts, the small vascular network may limit tissue oxygenation and cause acute rejection, as well as pulmonary and renal failure,22 whereas large-for-size grafts may result in delayed recovery of liver function and even massive hepatic necrosis or primary graft nonfunction.23
Hepatic steatosis is an established risk factor for primary nonfunctioning of liver transplants and thus a greater problem for recipients than for donors.4 Furthermore, higher rates of blood loss, transfusion, intensive care unit stay, and postoperative complications, including mortality, have been reported in steatotic patients.4,5 Therefore, estimation of liver volume is especially important in these patient populations, and our results show that CT-based volume measurement is accurate and reliable regardless of the severity of the hepatic steatosis.
The traditional method for predicting liver volumes in patients undergoing hepatic resection is to use a numerical formula. In 1995, a Japanese group first deduced a simple formula for estimating liver volume from body surface area.24 From then on, several ethnicity-specific formulas based on anthropometric measurements and/or demographic data have been established in Korea,25 China,26 Germany,27 the United States,28 and Taiwan.29 The Korean and German formulas have been evaluated by comparing calculated values with actual postmortem liver mass or volume measurements,25,27 whereas the liver volumes obtained by using the Chinese, US, and Taiwanese formulas viewed CT measurements as reference standards.26,28,29
The principle of CT volumetry is quite simple:26 first, it measures the area of each cross-sectional image, then it multiplies the area by the slice interval to obtain the slice volume, and finally it sums the individual slice volumes to obtain the total volume of liver.30 Most previous studies have found that this method is sufficiently accurate for most practical applications, although 1 study obtained a discrepancy of as high as 34% with intraoperative measures of liver volume. Among the suggested causes of errors in estimating liver volume by volumetric CT are partial volume effects, examination technique, hepatic physical density, exact contour and segment recognition, intraoperative drainage of liquid from the liver, and hepatic volume variation.7,9,10,30 It is likely that blood volume contributes a significant portion of the error because this volume is included in CT volumetry, whereas graft weight is generally measured in a blood-free manner.31 In the present study, however, we used a blood-free CT technique for all subjects to minimize the effects of blood volume and maximize those of hepatic steatosis.
Because it is generally assumed that the density of healthy liver tissue is 1 g/cm³, volumetric values may in most cases be directly converted to weights.9 However, because human fat has a density of 0.9 g/cm³ and contains less functional mass (g) than normal liver tissue of the same volume (cm³), we assumed that hepatic steatosis may be 1 cause of the discrepancy between actual graft mass and liver volume measured by CT volumetry. The previous work of Chen and associates demonstrating that the volume and fatty content of liver decreased after body weight reduction13 and the subsequent proposal by Marcos that functional liver mass can be assumed to decrease by 1% for each 1% of hepatic steatosis32 were key findings encouraging us to carry out the present study. In our large cohort, we found that CT volumetry was accurate and reliable regardless of the extent of hepatic steatosis and that low HDL cholesterol level, rather than hepatic steatosis, was associated with CT error. However, when we take into consideration that low HDL cholesterol is a component of metabolic syndrome and that our data indicated a slight trend toward increasing error of volumetric CT with increasing severity of steatosis, there remains a possibility of CT-derived overestimation in steatotic subjects, although at a negligible level.
An important limitation of our study is that only graft mass was measured during the surgical procedure with no corresponding volume measurement. Thus, the conversion factor based on the average density of healthy, nonsteatotic liver cannot be evaluated in our cohort; instead, we used the generally accepted value of unity for healthy liver tissue. Further investigations of more refined design are needed to assess the effects of hepatic steatosis on the accuracy of CT volumetry.
In summary, liver donors with different extents of hepatic steatosis yield similar estimates of CT volumetric errors with respect to the actual weights of the right lobe grafts. Thus CT volumetric measurement is an accurate and reliable tool for preoperative estimation of hepatic weight in surgical candidates for liver donation or resection, regardless of steatosis grade.
Volume : 15
Issue : 2
Pages : 196 - 202
DOI : 10.6002/ect.2016.0065
From the 1Department of Gastroenterology Asan Liver Center, Asan Medical Center, University of Ulsan College of Medicine, the 2Department of Gastroenterology, CHA Bundang Medical Center, CHA University, the 3Department of Surgery, Division of Hepatobiliary Surgery, Asan Liver Center, Asan Medical Center, University of Ulsan College of Medicine, and the 4Department of Biostatistics, Asan Medical Center, University of Ulsan College of Medicine, Republic of Korea
Acknowledgements: The authors declare that they have no sources of funding for this study, and they have no conflicts of interest to disclose.
Corresponding author: Ju Hyun Shim or Gi-Won Song, Asan Medical Center, University of Ulsan College of Medicine, 86 Asanbyeongwon-gil, Songpa-gu, Seoul, 138-736, Republic of Korea
Phone: +82 2 3010 3190; +82 2 3010 5796
E-mail: firstname.lastname@example.org; email@example.com
Figure 1. Representative Computed Tomography Image From a 30-Year-Old Male Living Donor With 6% Hepatic Steatosis
Table 1. Baseline Characteristics of the Patients
Figure 2. Box Plots Showing Results for Liver Volumes and Graft Weights Across the 4 Groups*
Figure 3. Graft Mass Estimation Errors (%) According to Extent of Hepatic Steatosis Across the 4 Groups*
Table 2. Intraclass Correlation Coefficients and Pearson Correlation Coefficients Between Graft Mass and Liver Volume Measured by Computed Tomography-Volumetry in All Subjects and Across the 4 Groups
Figure 4. Scatter Plots Showing the Correlation Between Liver Volumes and Graft Weights
Table 3. Univariate and Multivariate Linear Regression Analysis of the Association Between Predonation Variables and Graft Mass Estimation Error (%)