Assessment of hepatic fat content in using quantitative ultrasound measurement of hepatic/renal ratio and hepatic echo-intensity attenuation rate

(1)

DOI:

Original papers

Assessment of hepatic fat content in using quantitative ultrasound measurement of hepatic/renal ratio and hepatic echo-intensity attenuation rate

Yun-Sheng Wang

¹

, Gui-Ping Zhang

²

, Xiao Yang

²

, Jun Ye

¹

, Yong-Hong Cao

¹

, Rong Zhang

¹

, Shuai Ye

¹

, Shi-Mei Xing

¹

, Er-Lan Shi

¹

, Ji Zhang

³

, Hu Lian

³

, Jin-Xiang Xia

³

, Qiu Zhang

⁴

, Wu Dai

¹

1Department of Endocrinology, the Second People’s Hospital of Hefei, the Affiliated Hefei Hospital of Anhui Medical University, ²Department of Ultrasound, the Second People’s Hospital of Hefei, the Affiliated Hefei Hospital of Anhui Medical University, ³Department of Magnetic Resonance Imaging, the Second People’s Hospital of Hefei, the Affiliated Hefei Hospital of Anhui Medical University, ⁴Department of Endocrinology, the First Affiliated Hospital of Anhui Medical University, Hefei, Anhui, China

Received 24.03.2020 Accepted 22.05.2020 Med Ultrason

2020, Vol. 22, No 4, 393-401 Corresponding author: Wu Dai

Department of Endocrinology, the Second People’s Hospital of Hefei, the Affiliated Hefei Hospital of Anhui Medical University, Hefei, Anhui, 230011, China

Phone: +86 551 62965716 Fax: +86 551 62965716 E-mail: [email protected]

Introduction

Non-alcoholic fatty liver disease (NAFLD) has be- come a major public health issue with a worldwide epi- demic [1]; the prevalence of NAFLD has been increasing in recent years (30% in Europe and America and about 15% in Asia). In China, it has been estimated that the prevalence rate of NAFLD is 6% to 27%, and the incidence rate of NAFLD is about 3.4% to 9.1% every year [2-4]. The close relationship between NAFLD and diabe- Abstract

Aims: This study aims to evaluate and validate a simple quantitative ultrasound (US) method for determining the hepatic fat content (HFC) based on the combination of quantitative US hepatic/renal ratio (US-HRR) and quantitative US hepatic echo-intensity attenuation rate (US-HAR) as compared with [¹H]-magnetic resonance spectroscopy (¹H-MRS). Material and methods: There were a total of 242 subjects recruited in the present study. All subjects were examined for HFC by quantita- tive US and ¹H-MRS methods. The QUS-HRR and QUS-HAR were calculated from ordinary ultrasound images of liver and kidney with a triple modality 3D abdominal phantom using the Image J software. Results: The results found that US-HRR and US-HAR correlated with ¹H-MRS HFC (US-HRR: r=0.946, p<0.001; US-HAR: r=0.936, p<0.001). The equation for HFC prediction by using quantitative US was: HFC (%) = 28.965 × US-HRR + 218.045 × US-HAR - 8.892. Subgroup analysis in study subjects with body mass index (BMI) ≥28 showed that quantitative US HFC was associated with ¹H-MRS HFC (R²=0.953, p<0.001). Receiver operating characteristic (ROC) analysis observed that the cut-off value of fatty liver diagnosis was 6.71% in using the quantitative US model; the sensitivity and specificity for fatty liver diagnosis were 94.15% and 96.30%, respectively. Variability analysis indicated that there was a relative high degree of consistency in the measurement of HFC with different operators or ultrasonic apparatus. Conclusions: Quantitative US measurement could be regarded as a simple, sensitive tool to accurately assess HFC. It provides a valid alternative to ¹H-MRS as an easy, non-invasive option for the precise estimation of HFC in clinical practice.

Keywords: type 2 Diabetes; hepatic fat content; ultrasound hepatic/renal ratio; ultrasound hepatic echo-intensity attenua- tion rate; magnetic resonance spectroscopy

DOI: 10.11152/mu-2522

(2)

tes has been well demonstrated, where NAFLD associated with the increased risk for insulin resistance, and there was also a higher risk of diabetes prevalence in patients with NAFLD [5,6]. In addition, several studies have revealed that the presence of NAFLD was also associated with central obesity, metabolic disorders and cardiovascular diseases [7-9]. In this context, precise evaluation of hepatic liver content (HFC) is not only of great importance for the early diagnosis of NAFLD, but also is useful in predicting the risk for future occurrence of diabetes, metabolic disorders and cardiovascular diseases.

Currently, non-targeted percutaneous liver biopsy with direct histological visualization is the current gold standard to diagnose NAFLD, but its widespread use is limited by cost, sampling error, and procedure-related morbidity and mortality. In the past few years, a number of studies have reported a high sensitivity of the noninvasive technique of [¹H] magnetic resonance spectroscopy (¹H-MRS) in the assessment of HFC but routine use of

1H-MRS is limited by its cost and availability [10-13].

Ultrasound (US) is the most widely available modality for the initial evaluation of hepatic steatosis and diagnosing NAFLD, but this method could be affected by subjective factors, thus causing an inaccurate result. Qualitative US is able to infer the presence and severity according to qualitative sonographic features of the US hepatic/

renal ratio (US-HRR) or the US hepatic echo-intensity attenuation rate (US-HAR) [14,15]. Previous literature have found a significant correlation of US-HRR with his- tologic steatosis and the hepatic/renal sonographic index in patients with chronic liver diseases [16]; in addition, it has also been revealed that US-HRR and US-HAR associated with the degree of liver steatosis and the development of NAFLD [17-19]. However, given the several shortcomings of previous studies, such as the small sample size, ethnic variations and differed study subjects, the development of an easy, simple and accurate quantitative US measurement for detecting HFC is beneficial for the identification of asymptomatic high-risk NAFLD populations and the evaluation of appropriate therapy response or disease progression in NAFLD patients.

In the present study, we established a quantitative US method to provide a more precise estimation on HFC with a triple modality 3D abdominal phantom combined with quantitative US-HRR and quantitative US-HAR as compared with ¹H-MRS.

Material and methods Study subjects

Two hundred and forty-two study subjects [141con- secutively newly diagnosed type 2 diabetes mellitus

(NT2DM) patients, 48 prediabetes mellitus (PDM) subjects and 53 normal controls (NC)] were recruited from the Department of Endocrinology or the medical examination center at the Second People’s Hospital of Hefei, when they first visited the DM clinic. The diagnosis of T2DM or PDM was verified according to the American Diabetes Association diagnostic criteria (2018) [20]. NC subjects, without any liver or metabolic diseases, were enrolled from the medical examination center. Patients with other causes of liver disease (viral hepatitis, auto- immune hepatitis, Wilson’s disease, hemochromatosis, drug-induced hepatitis, or others) were excluded. An- thropometric measurement and routine laboratory results were obtained from hospital medical records.

All subjects completed the following tests when first entering the study: fasting plasma glucose (FPG), total cholesterol (TC), triglyceride (TG), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), very low-density lipoprotein cholesterol (VLDL-C), alanine aminotransferase (ALT), alkaline phosphatase (ALP), aspartate aminotransferase (AST), γ-glutamyltransferase (GGT), lactate dehydrogenase (LDH), total bilirubin (TBIL), indirect bilirubin (IBIL), direct bilirubin (DBIL), creatinine, uric acid (UA), apolipoprotein (Apo)A1 and ApoB.

All the blood biochemical indices were determined by using a Hitachi 7600 autoanalyzer (Hitachi Ltd., To- kyo, Japan) or immunoturbidimetric assay (Roche/Cobas Integra Tina Quant, Roche Diagnostics).

Standard protocol approvals and patient consents This study was approved by the Ethical Committee of the Second People’s Hospital of Hefei (Hefei, Anhui, China). All the study subjects provided informed consent to participate in this study.

The study was conducted according to the principles of the Declaration of Helsinki (1964) and its amend- ments.

HFC detection by ¹H-MRS

1H-MRS was used to detect the HFC of all subjects.

The right lobe of the liver was located when patients were lying in the supine position [21]. Areas under the water peak and fat peak were recorded. Liver fat content was calculated as [liver fat content (%) = area under the fat peak × 100/(area under the fat peak + area under the water peak)]. Liver fat content ≥5.56% was defined as fatty liver [13].

US-HRR and US-HAR analysis

US examinations were implemented to measure the HFC of all study subjects at the same day with the 1H-MRS detection. Ultrasonic images were analyzed by Image J software (Image J2x, National Institutes of Health, USA;

https://imagej.nih.gov/ij/). Average gray-scale intensities

(3)

were determined in the regions of interest (ROIs) of liver (15.11×15.11 mm) and renal cortex (5.14×5.14 mm). The hepatic/renal ratio was calculated according to the equation: hepatic/renal ratio = mean gray-scale intensity of the liver/mean grayscale intensity of the renal cortex (fig 1a). Two liver ROIs samples were selected at a depth of 4-6 cm from the near-field of the same beam. The distance between the two ROIs were measured (fig 1b). The hepatic echo-intensity attenuation rate was calculated according to the equation: hepatic echo-intensity attenuation rate = (lnAn-lnAf)/(∆d × f) [22]; where An and Af represent the mean echo intensity of the near-field and far- field ROIs, respectively; ∆d is the line distance between the two ROIs and f is the ultrasonic transducer frequency.

Adjustment for quantitative US parameters

To correct the US-HRR and US-HAR, we applied a triple modality 3D abdominal phantom to adjust the normal distribution of abdominal organs of human body [23- 25]. The triple modality 3D abdominal phantom stimulat- ed images were analyzed, then the quantitative US-HRR and quantitative US-HAR were calculated according to the equation: quantitative US-HRR = US-HRR of study subjects / HRR of 3D model; quantitative US-HAR = US-HAR of study subjects - HAR of 3D model.

Variability analysis

To evaluate the consistency of quantitative US-HRR and quantitative US-HAR among different operators in using the same ultrasonic device, the repeated measures on quantitative US-HRR and quantitative US-HAR were performed in 100 study subjects by two independent US specialists who had over 10 years of experience (Gui- Ping Zhang and Xiao Yang). In addition, to compare the variability of different ultrasonic devices on the influence of quantitative US-HRR and quantitative US-HAR, an experienced sonographer independently measured the quantitative US-HRR and quantitative US-HAR among 100 study subjects in using two different types of medical ultrasonic devices, respectively (IE33 [Philip, Germany], transducers [C5-1, Philip, Germany] and [GE Logiq P7, USA], transducers [C1-5 D, GE, USA]).

Statistical analysis

Normal distribution data were represented as a mean

± standard deviation; if data was not in the normal distribution, the median and interquartile range was used.

One-way ANOVA or nonparametric test (Kruskal–Wal- lis test) was utilized for intergroup comparisons. Linear regression analyses were used to detect the associations of quantitative US-HRR/US-HAR HFC with ¹H-MRS HFC. Spearman correlation analysis was performed to investigate the correlation of the quantitative US determined HFC with several clinical/laboratory parameters.

Variability analysis was performed by calculating intra-

class correlation coefficients and depicted Bland-Altman plots [26]. Receiver operating characteristic (ROC) analysis was constructed to evaluate the sensitivity and specificity of quantitative US determined HFC for the diagnosis of NAFLD. Data analysis was performed in using the SPSS23.0 statistical software (SPSS Inc., Chicago, IL, USA) and MedCalc, version 11.4.2.0 (Mariakerke, Bel- gium). Two tailed p<0.05 were considered to be statisti- cally significant.

Results

Characteristics of the study population

There were 188 subjects diagnosed as NAFLD (77.68%, 188/242) based on HFC ≥5.56%.Given the differences of ¹H-MRS HFC, we divided all study subjects into five groups (HFC<5% (n=48), HFC:

2.96±1.12%; 5%≤ HFC<10% (n=65), HFC: 7.39±1.33%;

10%≤HFC<15% (n=50), HFC: 13.18±1.42%; 15%≤

HFC<20% (n=55), HFC: 17.03±1.33%; 20%≤ HFC (n=24), HFC: 24.60±2.86%). There were significant differences in body mass index (BMI), waist circumference (WC), waist-to-hip ratio (WHR), systolic blood pressure (SBP), diastolic blood pressure (DBP), FPG, 2hPG, BUN, UA, TBIL, IBIL, ALT, GGT, TG, TC, HDL-C, LDL-C, VLDL-C, ApoA1, ¹H-MRS HFC, quantitative US-HRR and quantitative US-HAR among those five groups (all p<0.05). However, we did not observe any Fig 1. Ultrasonic liver image with HRR and HAR: a) Ultra- sonic liver image with HRR. ROI-1 and ROI-2 stands for the echo gray histograms of the liver and kidney cortex ROIs;

b) Ultrasonic liver image with HAR. ROI-3 and ROI-4 stands for the near-field and far-field echo gray histograms of the liver.

HAR: hepatic echo-intensity attenuation rate; HRR: hepatic/renal ratio; ROI: Region of interest.

(4)

significant difference regarding gender distribution, age, Cr, DBIL, AST, ALP, LDH and ApoBin those five groups (all p>0.05). In addition, we found that the increase of quantitative US-HRR and quantitative US-HAR was consistent with the gradual increase of ¹H-MRSHFC (Table I).

Correlation between quantitative US parameters and ¹H-MRS HFC

Correlation analysis revealed that quantitative US- HRR and quantitative US-HAR were highly correlated with ¹H-MRS HFC, respectively (r=0.946, p<0.001;

r=0.936, p<0.001) (fig 2).

Defining the quantitative US model for HFC estimation

Linear regression analysis was applied to evaluate quantitative US parameters for HFC estimation. ¹H-MRS HFC was set as a dependent variable, and quantitative US-HRR was included in the regression equation as independent variable. The results indicated that quantitative US-HRR was the strong predictor for HFC (corrected R²=0.895, p<0.001). When quantitative US-HAR combined quantitative US-HRR into the regression model, it showed an improved estimation accuracy (Corrected R²=0.935, p<0.001). The equation for HFC prediction by using quantitative US was: HFC (%) = 28.965 × quantitative US-HRR + 218.045 × quantitative US-HAR - 8.892 (Table II).

Correlation analysis between quantitative USHFC and clinical/laboratory parameters

The quantitative US HFC was positively correlated with BMI, WHR, SBP, DBP, FPG, 2hPG, UA, ALT, GGT, TG, TC, LDL-C and VLDL-C and negatively correlated with LDH, HDL-C and ApoA1 (Table III).

Correlation of ¹H-MRS HFC and quantitative US HFC

To avoid the potential effect of subcutaneous fat in obese subjects on the echo attenuation, we performed a correlation analysis of ¹H-MRS HFC and quantitative US

HFC in study subgroup with BMI ≥ 28. The results indicated that quantitative USHFC was still associated with

1H-MRS HFC (R² = 0.953, p< 0.001) (fig 3).

NAFLD diagnosis by quantitative US and ¹H-MRS ROC analysis revealed that the optimum point of fatty liver diagnosis was 6.71% in using the quantitative US model. Based on the ¹H-MRS HFC, all study subjects were divided into the NAFLD group (HFC ≥5.56%) and non-NAFLD group. When ¹H-MRS was set as the gold standard for diagnosing NAFLD by the quantitative US model, the sensitivity and specificity for NAFLD diagnosis were 94.15% and 96.30%, with the area under curves (AUC) of 0.987 (95%CI: 0.963-0.997). Furthermore, a subgroup analysis was also implemented when ¹H-MRS HFC <11.12%, the sensitivity and specificity for quantitative US model were 95.31% and 90.74%, with the AUC of 0.963 (95%CI: 0.912-0.989).

Variability analysis

Intraclass correlation coefficient was calculated to test the consistency of quantitative US HFC in different operators or medical ultrasonic devices; the results implied that there were relatively high degrees of consist- encies between different operators or ultrasonic devices (Table IV, fig 4)

Bland-Altman analysis for evaluation of the quantitative US HFC

To avoid the omission of subjects with light fatty liver, the cut-off value was set as 2-fold of diagnostic standard (HFC ≥ 5.56% defined as fatty liver). First, 118 study subjects with ¹H-MRS HFC <11.12% were included in the Bland Altman analysis. The results found that there was no significant bias for 118 study subjects considered;

six subjects (6/118) showed an overestimated HFC, and one subject (1/118) had an underestimated HFC. Moreo- ver, when the Bland Altman analysis was also performed in 242 study subjects, the results observed that while twelve subjects (12/242) had a higher HFC, only four subjects (4/242) reported a lower HFC.

Fig 2. Correlation analysis between ¹H-MRS HFC and US-HRR (a), and US-HAR (b). US-HAR: ultrasound hepatic echo-intensity attenuation rate; US-HRR: ultrasound hepatic/renal ratio; ¹H-MRS: [¹H]-magnetic resonance spectroscopy; HFC: hepatic fat content

(5)

Table I. Clinical characteristic of different HFC groups Parameters HFC<5% (n=48)

5%≤ HFC<10% (n=65)

10%≤ HFC<15% (n=50)

15%≤ HFC<20% (n=55)20%≤ HFC (n=24)p value Gender (female/male)29/1929/3622/2824/3113/110.360 Age (years)46.3±9.249.6±10.152.3±10.850.9±11.149.3±9.60.074 BMI (kg/m2)23.60±2.7825.17±2.0026.68±2.0126.92±3.0927.25±2.28<0.001 WHR0.87±0.060.91±0.060.97±0.100.98±0.091.01±0.12<0.001 SBP (mmHg)125±13131±16134±15139±17137±14<0.001 DBP (mmHg)77±981±1182±1187±984±11<0.001 FPG (mmol/L)5.26±1.075.94±1.196.96±1.437.21±1.207.82±0.83<0.001 2hPG (mmol/L)6.49 (5.71, 9.49)9.29 (7.40, 15.35)15.70 (10.88, 17.03)16.13 (12.61, 17.68)17.03 (15.73, 17.88)<0.001 BUN (mmol/L)4.52±1.295.13±1.485.37±1.605.25±1.384.67±1.070.016 Cr (umol/L)60.80±13.6059.76±13.1859.38±16.2359.99±14.8754.71±13.060.531 UA (umol/L)292.00±66.00320.82±64.05345.08±70.15331.49±53.51366.13±65.07<0.001 TBIL (umol/L)13.80 (11.80, 16.19)17.00 (11.60, 22.70)16.95 (13.43, 22.78)15.70 (12.70, 20.60)14.05 (11.05, 17.83)0.038 DBIL (umol/L)4.01±1.294.99±2.604.46±1.784.39±2.163.82±1.300.056 IBIL (umol/L)9.90 (8.10, 11.98)12.50 (8.85, 17.65)12.80 (10.28, 15.93)11.20 (9.00, 15.90)10.20 (7.93, 13.15)0.018 ALT (U/L)21.50 (13.25, 33.50)32.00 (28.00, 36.00)33.00 (31.00, 40.00)35.00 (30.00, 38.00)35.00 (32.00, 38.75)<0.001 AST (U/L)19.00 (16.00, 23.75)21.00 (17.00, 26.00)22.00 (16.00, 27.00)21.00 (17.00, 33.00)19.00 (14.50, 29.50)0.256 GGT (U/L)23.00 (17.00, 38.75)33.00 (22.00, 63.00)33.00 (19.75, 48.75)32.00 (26.00, 54.00)37.00 (26.00, 57.50)0.007 ALP (U/L)69.56±22.0876.80±22.0377.12±19.9275.60±20.5976.75±20.090.362 LDH (U/L)176.08±33.36178.75±36.78175.76±34.98171.09±36.12164.54±31.210.469 TG (mmol/L)1.62±0.732.02±0.782.28±0.682.30±0.712.43±0.51<0.001 TC (mmol/L)4.45±0.704.81±0.495.21±0.425.24±0.445.32±0.31<0.001 HDL-C (mmol/L)1.55±0.381.40±0.301.43±0.321.42±0.301.25±0.180.003 LDL-C (mmol/L)2.34 (2.04, 3.01)2.89 (2.21, 3.31)3.16 (2.79, 3.42)3.18 (2.98, 3.42)3.17 (2.95, 3.40)<0.001 VLDL-C (mmol/L)0.25 (0.18, 0.37)0.33 (0.23, 0.54)0.36 (0.26, 0.62)0.34 (0.24, 0.53)0.45 (0.37, 0.71)0.002 ApoB (g/L)0.94 (0.81, 1.04)0.90 (0.72, 1.06)0.92 (0.74, 1.07)0.86 (0.75, 1.02)0.85 (0.66, 0.93)0.493 ApoA1 (g/L)1.25±0.261.14±0.201.16±0.211.13±0.181.08±0.180.009 HFC%2.96±1.127.39±1.3313.18±1.4217.03±1.3324.60±2.86<0.001 US-HRR0.56±0.040.63±0.050.72±0.050.8±0.050.95±0.09<0.001 US-HAR (MHz-1*cm-1)-0.019 (-0.025, -0.016)-0.007 (-0.008, -0.001)0.004 (0.002, 0.007)0.008 (0.006, 0.013)0.025 (0.018, 0.03)<0.001 ApoA1: Apolipoprotein-A1; ApoB: Apolipoprotein-B; ALP: alkaline phosphatase; AST: aspartate aminotransferase; ALT: alanine transaminase; BMI: body mass index; Cr: creatine; DBIL: direct bilirubin; DBP: diastolic blood pressure; FBG: fasting blood glucose; GGT: γ-glutamyltransferase; HOMA-IR: homeostasis model assessment of insulin resistance; HOMA-β: homa beta cell function index; HDL-C: high-density lipoprotein cholesterol; IBIL: indirect bilirubin; LDH: lactate dehydrogenase; LDL-C: low-density lipoprotein cholesterol; SBP: systolic blood pressure; TC: total cholesterol; TG: triglycerides; TBIL: total bilirubin; UA: uric acid; VLDL-C: very low-density lipoprotein cholesterol; WHR: waist-to-hip ratio; HFC: hepatic fat content; US-HRR: ultrasound hepatic/renal ratio; US-HAR: ultrasound hepatic echo-intensity attenuation rate

(6)

Table II. Quantified US parameters predicts HFC model

Model US-HRR US-HAR Constant Corrected R²

β p β p

HFC 1 50.240 < 0.001 - - -23.752 0.895

2 28.965 < 0.001 218.045 < 0.001 -8.892 0.935

Model 1: US -HRR to estimate 1H-MRS HFC. Model 2: US-HRR and US-HAR to estimate 1H-MRS HFC. US: ultrasound; US-HRR:

ultrasound hepatic/renal ratio; US-HAR: ultrasound hepatic echo-intensity attenuation rate Table III. Correlation of quantitative US quantified HFC with

clinical characteristic.

Parameters Quantitative US quantified HFC

r P

Age 0.096 0.137

BMI 0.500 0.000

WHR 0.561 0.000

SBP 0.287 0.000

DBP 0.266 0.000

FPG 0.574 0.000

2hPG 0.553 0.000

BUN 0.107 0.096

Cr -0.046 0.477

UA 0.298 0.000

TBIL 0.010 0.876

DBIL -0.047 0.464

IBIL 0.019 0.770

ALT 0.341 0.000

AST 0.044 0.492

GGT 0.174 0.007

ALP 0.107 0.097

LDH -0.127 0.048

TG 0.311 0.000

TC 0.464 0.000

HDL-C -0.183 0.004

LDL-C 0.394 0.000

VLDL-C 0.230 0.000

ApoB -0.092 0.155

ApoA1 -0.166 0.010

ApoA1: Apolipoprotein-A1; ApoB: Apolipoprotein-B; ALP: alkaline phosphatase; AST: aspartate aminotransferase; ALT: alanine transaminase; BMI: body mass index; Cr: creatine; DBIL: direct bilirubin; DBP: diastolic blood pressure; FBG: fasting blood glucose; GGT: γ-glutamyltransferase; HOMA-IR: homeostasis model assessment of insulin resistance; HOMA-β: homa beta cell function index; HDL-C: high-density lipoprotein cholesterol; IBIL: indirect bilirubin; LDH: lactate dehydrogenase; LDL-C: low-density lipoprotein cholesterol; SBP: systolic blood pressure; TC: total cholesterol; TG: triglycerides; TBIL: total bilirubin; UA: uric acid;

VLDL-C: very low-density lipoprotein cholesterol; WHR: waist- to-hip ratio; HFC: hepatic fat content; US: ultrasound

Fig 4. Bland–Altman analysis for intra-session repeatability for different operators in HAR (a), HRR (b); for different ultrasonic apparatuses in HAR (c), HRR (d). HAR: hepatic echo- intensity attenuation rate; HRR: hepatic/renal ratio

Fig 3. Correlation analysis between ¹H-MRS HFC and US- HRR (a), and US-HAR (b). US-HAR: ultrasound hepatic echo- intensity attenuation rate; US-HRR: ultrasound hepatic/renal ratio; ¹H-MRS: [¹H]-magnetic resonance spectroscopy; HFC:

hepatic fat content

Table IV. Intraclass correlation coefficient of quantitative US parameters by different operators or apparatuses.

Parameters US parameters Intraclass correlation coefficient p 95% CI

Different operators US-HRR 0.970 <0.001 0.956 0.980

US-HAR 0.978 <0.001 0.967 0.985

Different apparatuses US-HRR 0.972 <0.001 0.958 0.981

US-HAR 0.981 <0.001 0.972 0.987

US: ultrasound; US-HRR: ultrasound hepatic/renal ratio; US-HAR: ultrasound hepatic echo-intensity attenuation rate

(7)

Discussion

Over the past decade, emerging evidence has high- lighted the relationship of an excessive hepatic fat accu- mulation with several disease risks for future occurrence of diabetes, metabolic disorders and cardiovascular diseases [7-9]. Therefore, the early detection of HFC is of great importance in providing information about the liver and in the assessment and prevention of cardio-metabolic risks.

The common recognized gold standard for the assessment of fatty liver is liver biopsy, but this technique is invasive and suffers from sampling problems resulting in diagnostic errors; moreover, the result is also strong- ly influenced by the subjectivity and experience of the pathologist [27,28]. As an alternative, non-invasive ap- proaches can be used to quantify liver steatosis, ¹H-MRS has been recognized as a sensitive, accurate and quantitative evaluation of HFC using non-ionizing radiation, as compared to liver biopsy, ¹H-MRS showed a compara- tive sensitivity (100%) and specificity (97%) in the diagnosis of liver steatosis [13]. However, the use of this approach in clinical practical is limited because of the time required for examination, high equipment require- ments and high expense. In addition, both the acquisition and analysis of the collected spectra require expertise and specialized software which further restricts the availability of this technique.

In the present study, we have investigated the association of quantitative US HFC with ¹H-MRS HFC by en- rolling a total of 242 subjects, and have also constructed the equation of the quantitative US model for HFC estimation. In addition, the variability analysis and Bland- Altman analysis were also utilized to assess the consistency of quantitative US HFC in different operators or medical ultrasonic devices. The present study revealed that the quantitative US-HRR and quantitative US-HAR were highly correlated with ¹H-MRS HFC. In addition, when using quantitative US-HRR and quantitative US- HAR to predict the ¹H-MRS HFC; the results of MLR analysis supported a relatively increased accuracy in using a combination of quantitative US-HRR and quantitative US-HAR for the prediction of HFC. These indicated that the use of the quantitative US model could represent a valid alternative to ¹H-MRS imaging. Consider- ing overweight and obese subjects, quantitative US HFC was still associated with ¹H-MRS HFC in the subgroup of BMI ≥28. These results suggested that the quantitative US model has a good diagnostic performance even in overweight and obese patients.

We have also compared the performance of quantitative US and ¹H-MRS in the diagnosis of NAFLD, the ROC

analysis revealing that the quantitative US model exerted a good sensitivity (94.15%) and specificity (96.30%) in the diagnosis of NAFLD (in compared with ¹H-MRS), with the AUC of 0.987 (95% CI: 0.963-0.997). The quantitative US model showed a cut-off point of 6.71% in the diagnosis of NAFLD, which turned out to be higher than that of ¹H-MRS data (HFC ≥5.56%); this discrepancy may be due to the imperfectly linear relationship between quantitative US HFC and ¹H-MRS HFC. Furthermore, when we set the cut-off value as ¹H-MRS HFC <11.12%, the results revealed that the sensitivity and specificity for quantitative US were 95.31% and 90.74%, with the AUC of 0.963 (95%CI: 0.912-0.989). These facts confirmed that quantitative US is capable of providing a precise and reliable diagnostic value for the assessment of NAFLD and fatty liver degeneration.

Currently, given the features of non-invasive, non- ionizing, inexpensive and widely available in US, several studies have been conducted to use US in the de- termination of HFC [29-31]. However, the sensitivity and specificity of using US in the quantification of HFC have differed. A meta-analysis, performed by Bohte et al, implied that the use of conventional US in the assessment of NAFLD showed a sensitivity of 73% to 91% as compared to MRS, and they also observed that the conventional US does not accurately predict the presence of NAFLD when HFC is <10%, with the sensitivity of 62.2% to 82.1% [32]. Bedossa et al showed that the conventional US had the sensitivity of 55% in diagnosis of NAFLD when HFC <20% [33]. It has also been reported that the sensitivities and specificities of using conventional US in assessing liver fat ranges from 60% to 94%

[34-37]. The different parameter settings among ultrasonic devices, post-processing procedures for US images and ultrasonic operators that may contribute to the varied sensitivities and specificities of using conventional US in assessing HFC, restrict the clinical reliability of using US in the quantification of HFC.

In the present study, we have adopted the triple modality 3D abdominal phantom to adjust the potential errors that may be caused by different ultrasonic operators and devices. We found that quantitative US with a triple modality 3D abdominal phantom adjustment in the assessment of HFC represented a good performance as compared with ¹H-MRS HFC. The results of the variability analysis also indicated that the quantitative USHFC between two independent operators or two types of ultrasonic devices had a good repeatability. The Bland–Alt- man analysis did not reveal a significant bias for most of the study subjects in the use of quantitative US as compared with ¹H-MRS, but it showed that quantitative US had an allowable overestimation in the assessment

(8)

of early liver steatosis. These results, along with a previous study implementing automatic measurements on US images [18,19], supported that the assessment of HFC in using the quantitative US model could decrease the variability related to the subjective visual evaluation, the influence of different ultrasonic devices and operators.

This suggested that the application of quantitative US is an acceptable technique for the detection of fatty liver and NAFLD.

There are some shortcomings that should be observed in the present study. First, liver biopsy is still the common recognized standard for diagnosing fatty liver. Our study used ¹H-MRS HFC as the standard comparison group rather than histology, thus this may lead to a potential discrepancy in the assessment of HFC and the diagnosis of NAFLD. Second, our study has shown that quantitative US exerted an excellent performance for the detection of NAFLD in subjects with moderate and high HFC, but it was limited in subjects with a low HFC. Further- more, the present study was a single-hospital based study with a relatively small study sample size. Thus, it could possibly restrict the generalizability of our findings. A further multi-center study with a large scale population is required to confirm our results.

Conclusions

Overall, our study established and confirmed the application of quantitative US model with a triple modality 3D abdominal phantom for a relatively accurate estimation of HFC through the combination of quantitative US-HRR and quantitative US-HAR, suggesting that quantitative US is capable of being a valid alternative to

1H-MRS as a non-invasive, reliable option with low costs in the clinical assessment of liver fat and diagnosis of NAFLD.

Acknowledgment: This study was supported by Anhui Science and Technology Project (grant number 1604a0802099).

Conflict of interest: none Reference

1. Lonardo A, Byrne CD, Caldwell SH, Cortez-Pinto H, Tar- gher G. Global epidemiology of nonalcoholic fatty liver disease: Meta-analytic assessment of prevalence, incidence, and outcomes. Hepatology 2016;64:1388-1389.

2. Fan JG. Epidemiology of alcoholic and nonalcoholic fatty liver disease in China. J Gastroenterol Hepatol 2013;28 Suppl 1:11-7.

3. Wang FS, Fan JG, Zhang Z, Gao B, Wang HY. The global burden of liver disease: the major impact of China. Hepa- tology 2014;60:2099-2108.

4. Fan JG, Zhu J, Li XJ, et al. Prevalence of and risk factors for fatty liver in a general population of Shanghai, China. J Hepatol 2005;43):508-514.

5. Mansour A, Mohajeri-Tehrani MR, Samadi M, et al. Risk factors for non-alcoholic fatty liver disease-associated hepatic fibrosis in type 2 diabetes patients. Acta Diabetol 2019;56:1199-1207.

6. Younossi ZM, Golabi P, de Avila L, et al. The Global Epi- demiology of NAFLD and NASH in Patients with type 2 diabetes: A Systematic Review and Meta-analysis. J Hepa- tol 2019;71:793-801.

7. Vozarova B, Stefan N, Lindsay RS, et al. High alanine aminotransferase is associated with decreased hepatic insulin sensitivity and predicts the development of type 2 diabetes.

Diabetes 2002;51:1889-1895.

8. Adams LA, Waters OR, Knuiman MW, Elliott RR, Olynyk JK. NAFLD as a risk factor for the development of diabetes and the metabolic syndrome: an eleven-year follow-up study. Am J Gastroenterol 2009;104:861-867.

9. Ballestri S, Zona S, Targher G, et al. Nonalcoholic fatty liver disease is associated with an almost twofold increased risk of incident type 2 diabetes and metabolic syndrome.

Evidence from a systematic review and meta-analysis. J Gastroenterol Hepatol 2016;31:936-944.

10. Thomsen C, Becker U, Winkler K, Christoffersen P, Jensen M, Henriksen O. Quantification of liver fat using magnetic resonance spectroscopy. Magn Reson Imaging 1994;12:487-495.

11. McPherson S, Jonsson JR, Cowin GJ, et al. Magnetic resonance imaging and spectroscopy accurately estimate the severity of steatosis provided the stage of fibrosis is considered. J Hepatol 2009;51:389-397.

12. Idilman IS, Keskin O, Celik A, et al. A comparison of liver fat content as determined by magnetic resonance imaging-proton density fat fraction and MRS versus liver histology in non-alcoholic fatty liver disease. Acta Radiol 2016;57:271-278.

13. Szczepaniak LS, Nurenberg P, Leonard D, et al. Magnetic resonance spectroscopy to measure hepatic triglyceride content: prevalence of hepatic steatosis in the general population. Am J Physiol Endocrinol Metab 2005;288:E462-E468.

14. Shiralkar K, Johnson S, Bluth EI, Marshall RH, Dornelles A, Gulotta PM. Improved method for calculating hepatic steatosis using the hepatorenal index. J Ultrasound Med 2015;34:1051-1059.

15. Wang Q, Ma W, Xia J. Nonalcoholic Fatty Liver Is Associ- ated With Further Left Ventricular Abnormalities in Patients With Type 2 Diabetes Mellitus: A 3-Dimensional Speckle- Tracking Study. J Ultrasound Med 2018;37:1899-1911.

16. Webb M, Yeshua H, Zelber-Sagi S, et al. Diagnostic value of a computerized hepatorenal index for sonographic quantification of liver steatosis. AJR Am J Roentgenol 2009;192:909-914.

17. Mancini M, Prinster A, Annuzzi G, et al. Sonographic hepatic-renal ratio as indicator of hepatic steatosis: comparison with (1)H magnetic resonance spectroscopy. Metabo- lism 2009;58:1724-1730.

(9)

18. Kwon HJ, Kim KW, Lee SJ, et al. Value of the ultrasound attenuation index for noninvasive quantitative estimation of hepatic steatosis. J Ultrasound Med 2013;32:229-235.

19. Xia MF, Yan HM, He WY, et al. Standardized ultrasound hepatic/renal ratio and hepatic attenuation rate to quantify liver fat content: an improvement method. Obesity (Silver Spring) 2012;20:444-452.

20. American Diabetes Association. 2. Classification and Di- agnosis of Diabetes: Standards of Medical Care in Diabe- tes-2018. Diabetes Care 2018;41(Suppl 1):S13-S27.

21. Han A, Andre MP, Deiranieh L, et al. Repeatability and Reproducibility of the Ultrasonic Attenuation Coefficient and Backscatter Coefficient Measured in the Right Lobe of the Liver in Adults With Known or Suspected Nonalco- holic Fatty Liver Disease. J Ultrasound Med 2018;37:1913- 1927.

22. Langton CM, Njeh CF. The measurement of broadband ultrasonic attenuation in cancellous bone--a review of the science and technology. IEEE Trans Ultrason Ferroelectr Freq Control 2008;55:1546-1554.

23. Toshikuni N, Shiroeda H, Ozaki K, et al. Advanced ultrasonography technologies to assess the effects of radiofre- quency ablation on hepatocellular carcinoma. Radiol Oncol 2013;47:224-229.

24. Garcia-Vazquez V, Marinetto E, Guerra P, et al. Assessment of intraoperative 3D imaging alternatives for IOERT dose estimation. Z Med Phys 2017;27:218-231.

25. Won HJ, Kim N, Kim GB, Seo JB, Kim H. Validation of a CT-guided intervention robot for biopsy and radiofrequen- cy ablation: experimental study with an abdominal phantom. Diagn Interv Radiol 2017;23:233-237.

26. Cleophas TJ, Droogendijk J, van Ouwerkerk BM. Validat- ing diagnostic tests, correct and incorrect methods, new de- velopments. Curr Clin Pharmacol 2008;3:70-76

27. Ratziu V, Charlotte F, Heurtier A, et al. Sampling variability of liver biopsy in nonalcoholic fatty liver disease. Gastroen- terology 2005;128:1898-1906.

28. Villeneuve JP, Bilodeau M, Lepage R, Cote J, Lefebvre M.

Variability in hepatic iron concentration measurement from needle-biopsy specimens. J Hepatol 1996;25:172-177.

29. Charatcharoenwitthaya P, Lindor KD. Role of radiologic modalities in the management of non-alcoholic steatohepa- titis. Clin Liver Dis 2007;11:37-54.

30. Mishra P, Younossi ZM. Abdominal ultrasound for diagnosis of nonalcoholic fatty liver disease (NAFLD). Am J Gastroenterol 2007;102:2716-2717.

31. Saadeh S, Younossi ZM, Remer EM, et al. The utility of radiological imaging in nonalcoholic fatty liver disease.

Gastroenterology 2002;123:745-750.

32. Bohte AE, Koot BG, van der Baan-Slootweg OH, et al.

US cannot be used to predict the presence or severity of hepatic steatosis in severely obese adolescents. Radiology 2012;262:327-334.

33. Bedossa P, Patel K. Biopsy and Noninvasive Methods to Assess Progression of Nonalcoholic Fatty Liver Disease.

Gastroenterology 2016;150:1811-1822.e4.

34. Mathiesen UL, Franzen LE, Aselius H, et al. Increased liver echogenicity at ultrasound examination reflects degree of steatosis but not of fibrosis in asymptomatic patients with mild/moderate abnormalities of liver transaminases. Dig Liver Dis 2002;34:516-522.

35. Hepburn MJ, Vos JA, Fillman EP, Lawitz EJ. The accuracy of the report of hepatic steatosis on ultrasonography in patients infected with hepatitis C in a clinical setting:

a retrospective observational study. BMC Gastroenterol 2005;5:14.

36. Joseph AE, Saverymuttu SH, al-Sam S, Cook MG, Max- well JD. Comparison of liver histology with ultrasonography in assessing diffuse parenchymal liver disease. Clin Radiol 1991;43:26-31.

37. Sogabe M, Okahisa T, Hibino S, Yamanoi A. Usefulness of differentiating metabolic syndrome into visceral fat type and subcutaneous fat type using ultrasonography in Japa- nese males. J Gastroenterol 2012;47:293-299.