Performance of Liver Stiffness Measurement from FibroScan was Affected by Glucose Metabolism in Patients with Nonalcoholic Fatty Liver Disease

Background: The performance of liver stiffness measurement (LSM) obtained using FibroScan can be affected by several factors and cut-off values were different for brosis caused by various etiologies. The aim of this study was to evaluate the diagnostic performance of LSM in nonalcoholic fatty liver disease (NAFLD) patients with abnormal glucose metabolism and investigate whether LSM value would be affected by metabolic indicators. Methods: The study involved 91 NAFLD patients with abnormal glucose metabolism who underwent liver biopsy. The receiver operator characteristic (ROC) curves were used to evaluate the diagnostic accuracy, with the biopsy results taken as the gold standard. Multivariate linear regression and subgroup analysis were used to determine the correlated indicators. Results: The areas under ROC curves (AUROCs) of LSM values in diagnosing brosis stage ≥ 1, 2, 3 and 4 were 0.793 (95% condence interval [CI]: 0.695-0.871), 0.764 (95% CI: 0.663-0.846), 0.837 (95% CI: 0.744-0.906) and 0.902 (95% CI: 0.822-0.955), with cut-off values of 6.3, 7.6, 8.3 and 13.8 kPa, respectively. Multivariate linear regression demonstrated that HbA1c (β=0.200, p=0.038) and AST (β=0.200, p=0.044) were independently associated with LSM value after adjustment for brosis stage from liver biopsy. Subgroup analysis revealed that LSM values were slightly higher in patients with HbA1c ≥ 7% than those HbA1c <7%, and in patients with BMI ≥ 30kg/m 2 than those BMI <30kg/m 2 . Conclusions: FibroScan was valuable for the evaluation of liver brosis in NAFLD patients with abnormal glucose metabolism. It was recommended for evaluating severe brosis, especially for excluding advanced brosis. Glucose metabolism state may affect LSM value.


Introduction
Nonalcoholic fatty liver disease (NAFLD) has become the predominant cause of chronic liver injury worldwide, which refers to the presence of ≥ 5% hepatic steatosis (HS) without other competing etiologies, including chronic viral hepatitis, excessive alcohol consumption, use of steatogenic medication or hereditary disorders [1]. Its spectrum ranges from nonalcoholic fatty liver (NAFL), nonalcoholic steatohepatitis (NASH) to cirrhosis or even carcinoma, and is associated with the features of metabolic syndrome such as hypertension, insulin resistance, diabetes mellitus (DM) and dyslipidemia, increases risks of cardiovascular disease, and accelerates the progression of underlying disease, leading to severe consequences [2][3][4]. NALFD patients with DM is prone to develop NASH, liver brosis and cirrhosis, and even liver cancer. The overall prevalence of NAFLD among patients with type 2 diabetes mellitus (T2DM) is 55%, more than 2-fold higher than in the general population, with a very high rate of NASH [5][6][7]. In the course of NAFLD, liver brosis, an important predictor of adverse prognosis, is the most relevant target for early diagnosis and treatment, as it has been associated with further deterioration of cirrhosis and increased overall mortality [8,9]. In our recent study, brosis occurred in up to 50% of patients with both NAFLD and T2DM [10]. Considering the huge number of patients with T2DM, the burden of the management of NAFLD seems to be enormous. Although the effective therapies for NAFLD has not been established, early intervention can signi cantly improve the poor prognosis of NAFLD. Therefore, the accurate and early detection of NAFLD in patients with abnormal glucose metabolism, especially the staging of liver brosis, is crucial [8,11,12].
Liver biopsy, an invasive procedure, has been recommended as the gold standard for diagnosis and classi cation of NAFLD, but the limitations of possible bleeding risks and sampling errors make it unsuitable for screening and frequent monitoring [13][14][15]. Therefore, non-invasive alternatives to liver biopsy have been investigated, such as serum biomarkers, clinical scoring systems and imaging tests including ultrasonography, FibroScan, magnetic resonance elastography (MRE), proton magnetic resonance spectroscopy ( 1 H-MRS) and magnetic resonance imaging-derived proton density fat fraction (MRI-PDFF) etc. [16,17] Among them, FibroScan (EchoSens, Paris, France) is recommended by the 2018 NAFLD guidance because of its convenience, clinically accessibility, low cost and simultaneous measurement of brosis and steatosis 16 . Liver stiffness measurement (LSM) obtained using FibroScan is one parameter for the diagnosis and quanti cation of liver brosis by measuring mechanical or ultrasound shear wave propagation through the hepatic parenchyma [18,19].
There exist several studies assessing diagnostic performance of FibroScan, most of which targeted at patients with chronic hepatitis B [20,21], while others focused on NAFLD [18,22,23]. Transient elastography expert consensus pointed out that there are differences in the cut-off values in patients with liver brosis caused by various etiologies, consists of Hepatitis B, Hepatitis C and NAFLD etc. Factors such as liver in ammation activity manifested by alanine aminotransferase (ALT) or increased bilirubin levels, excessive alcohol intake and eating may lead to increase of LSM values [24]. In view of the promoting effect of abnormal glucose metabolism on liver disease, NAFLD patients with abnormal glucose metabolism may have speci c cut-off values differentiated from general NAFLD patients. However, up to now, there are no studies focusing on these population and lack of cut-off values for them.
The aim of this study was to evaluate the diagnostic performance of FibroScan and obtain cut-off values in NAFLD patients with abnormal glucose metabolism, and investigate whether metabolic indicators would affect the measurement of FibroScan, in order to provide a clinical advice for the application of FibroScan in the diagnosis and evaluation of NAFLD patients.

Design and subjects
This cross-sectional study included 91 NAFLD patients evaluated in Zhongshan Hospital, Fudan University between July 2015 and December 2019, 89 of whom underwent liver biopsy while 2 others were identi ed cirrhosis using ultrasound. The time interval between FibroScan examination and biopsy was < 2 weeks. Patients with a history of excessive alcoholic consumption (> 20 g for men or > 10 g for women/day), chronic viral hepatitis, drug use and any other causes associated with liver injury were excluded. The study was approved by the Human Research Ethics Committee of Zhongshan Hospital Clinic and the whole process was conducted to the ethical guidelines of the Declaration of Helsinki.
Written informed consent was also obtained from all patients.

Liver biopsy and histopathologic evaluations
Liver biopsy samples were obtained from right liver lobe of NAFLD patients using 16-gauge needle under ultrasound guidance. 2 specimens were obtained from each person to ensure su cient sample size for analysis and to reduce error. All biopsy specimens were evaluated by two experienced pathologists blinded to the clinical and biological data. Histopathological ndings were reported according to Non-Alcoholic Steatohepatitis Clinical Research Network Scoring System [25]. Liver steatosis was categorized as grade 1, grade 2 and grade 3, while brosis was staged from stage 0 to stage 4 in accordance with the scoring system above. The NAFLD activity score (NAS) is the sum of scores for hepatic steatosis (grade 0-3), hepatocyte ballooning (grade 0-2) and lobular in ammation (grade 0-3). Another SAF scoring system was also adopted, consists of steatosis (grade 0-3), activity (grade 0-4) and brosis (stage 0-4). The classi cation criteria are described in the Table 1. All data are expressed as mean ± SD, medians (interquartile range), or n (%), as appropriate. BMI, body mass index; LSM, liver stiffness measurement.

LSM measurement
LSM measurement was performed with the FibroScan using M probe. Details of measurement are described in several previous studies, performed with the same machine by the same experienced operator blinded to other non-invasive methods and biopsy results [23,26]. The whole examination duration was less than 5 minutes. 10 valid measurements were obtained from each patient, and then the success rate, the ratio of the successful measurement times over the total times, was calculated. The result was considered as reliable only when the success rate was ≥ 60% and the interquartile range (IQR)/median was ≤ 30%. Median value was kept as representative result. Liver stiffness measurement results are expressed with kilopascal (kPa).

Basic characteristics collection
Medical history of each patient was collected when admitted, including general physical characteristics, history of chronic diseases etc. Hypertension was diagnosed according to criteria [27]. Fasting blood samples were collected locally and then shipped to clinical laboratory of Zhongshan Hospital for assessment of the blood glucose, lipid pro es and other blood biochemical parameters.

Statistical analysis
Continuous variables with normal distribution were summarized as mean ± SD while those without were described with median (interquartile range). Categorical variables were summarized as frequencies and percentages. The software SPSS (version 23.0) was used for data analysis. The graphs were performed with Graphpad (version 8.4) and Medcalc (version 19.1). The receiver operator characteristic (ROC) curves were performed. The diagnostic accuracy was evaluated by the areas under ROC curves (AUROCs) and the optimal cut-off values were de ned by Youden's index, with sensitivity, speci city, positive predictive value (PPV) and negative predictive value (PPV) calculated. The unpaired t test and Kruskal-Wallis test with Dunn's multiple correction were used for univariate comparisons between groups. The independent correlation was analyzed using multivariate linear regression analysis. The Spearman's rank correlation coe cient was used to assess the correlation between the liver histopathologic degree and FibroScan results in patients underwent biopsy. P value < 0.05 was considered statistically signi cant.

Patient characteristics
In this study, a total of 91 patients with NAFLD were involved. All patients received blood biochemical examination. LSM and CAP values obtained using FibroScan were attempted in all patients. The physical, clinical, serological, and histologic characteristics were detailed in Table 1.

Evaluation of diagnostic accuracy of FibroScan on liver brosis in patients with NAFLD
The LSM was used to assess the stage of liver brosis measured by FibroScan in patients with NAFLD.  Table 2. It seemed that the diagnostic accuracy of LSM improved as the histologic severity of hepatic brosis increased. The diagnostic AUROC at stage 4 even reached up to 0.902, indicating FibroScan is an ideal machine to evaluate the severity of liver brosis, especially in more severe degrees. Besides, the cut-off value for each stage with sensitivity ≥ 90% or speci city ≥ 90% were added in Additional le 1- Table 1.

In uence of metabolic indicators on FibroScan in detecting liver brosis
We next investigated whether metabolic state would affect FibroScan measurements. The relationship between metabolic indicators including age, BMI, hypertension, fasting glucose, 2 h glucose, hemoglobin A1c (HbA1c), total cholesterol (TC), triglycerides (TG), LDL cholesterol, HDL cholesterol, alanine aminotransferase (ALT) and aspartate aminotransferase (AST) and LSM value was investigated using multivariate linear regression analysis after adjustment of liver brosis, as shown in Table 3. The results demonstrated that HbA1c (β = 0.200, p = 0.038) and AST (β = 0.200, p = 0.044) were independently correlated with LSM value after adjustment for the liver brosis stage. BMI (β = 0.164, p = 0.092) also entered the regression model, while no statistical signi cance was obtained. In order to further investigate the in uence of metabolic indicators on FibroScan measurements, the subgroup analysis of LSM value was conducted grouped by BMI < 30 kg/m 2 or BMI ≥ 30 kg/m 2 and HbA1c < 7% or HbA1c ≥ 7%. The brosis stage was divided into stage 0-1 and stage 2-4 (signi cant brosis). The violin plots in Fig. 2A-B demonstrated a trend that LSM values were slightly higher for patients with BMI ≥ 30 kg/m 2 than those for patients with BMI < 30 kg/m 2 , and statistically signi cant was detected in patients with signi cant brosis. Similar increase trend was also observed in patients with HbA1c ≥ 7% than compared with those HbA1c < 7%, and statistically signi cant was detected in patients without signi cant brosis. In view of the results above, we then conducted the subgroup comparison of cut-off value. When the sensitivity or speci city were consistent between subgroups, the cut-off value of patients with BMI ≥ 30 kg/m 2 or HbA1c ≥ 7% was higher than those without them, as shown in Table 4. HbA1c and AST were independently associated with LSM value after adjustment for the liver brosis. 95% CI, 95% con dence interval; Se, sensitivity; Sp, speci city; PPV, positive predictive value; NPV, negative predictive value. †Cut-off value was determined when the sensitivity or speci city are consistent between subgroups.

Correlation analysis of LSM value with other histologic parameters
We also investigated the relationship between LSM value of FibroScan and other pathologic degrees, which was performed by Spearman correlation analysis, as shown in Additional le 1- Table 2. LSM value was correlated with the grade of ballooning (r = 0.258, p = 0.007) and in ammation (r = 0.241, p = 0.013).
In addition, the distribution of LSM value classi ed in accordance with other histological parameters was performed in Additional le 2- Fig. 1. There was a trend of stepwise increase as the liver ballooning aggravated, indicated that LSM value can be used to roughly assess the severity of ballooning, but no signi cant trend in LSM value with the increase of in ammation or steatosis degree.

Discussion
To the best of our knowledge, this is the rst study that evaluated the diagnostic accuracy of FibroScan and obtained the cut-off value for NAFLD patients accompanied by abnormal glucose metabolism. Furthermore, we investigated the in uence of metabolic indicators on FibroScan measurement. This cross-sectional study demonstrated that LSM values obtained using FibroScan achieved good diagnostic performance of liver brosis, especially in the more severe histologic stage, as the AUROC with a diagnosis of advanced brosis (stage ≥ 3) reached up to more than 80% and even 90% for cirrhosis (stage ≥ 4). It is worth noting that the NPV of stage ≥ 3 was more than 90%, indicated that FibroScan has high e cacy for excluding advanced brosis and cirrhosis. In addition, our study suggested that HbA1c was independently associated with LSM value. Glucose metabolism state may affect LSM value measurement, and its value was elevated in patients with poor glycemic control, which further emphasized the signi cance of this study.
Several studies have previously investigated the accuracy of FibroScan in patients with NAFLD, but our study targeted at people with NAFLD accompanied by abnormal glucose metabolism [18,22,23]. The LSM cut-off values in our study were within the interval of most previous studies, but the value 8.3 kPa for diagnosing advanced brosis was lower than 9.9 kPa recommended by the 2018 NAFLD guidance, which may be due to the differences of ethnicity and metabolic state [16]. All subjects in this study are Chinese, with lower BMI and thinner subcutaneous fat thickness than American, and all of them were accompanied with abnormal glucose metabolism. Consistent with previous studies, our results further con rmed that FibroScan is more suitable for assessing severe brosis, especially for ruling out advanced brosis, which can reduce the demand for liver biopsy to some extent.
It has been reported that several factors may affect the FibroScan measurement, such as ALT, increased bilirubin levels, excessive alcohol intake and eating. Different cut-off values were suggested in patients with chronic liver disease caused by different etiologies [24]. However, the effect of metabolic indicators on FibroScan has not been noticed so far. Based on previous studies, we hypothesized that metabolic indicators may affect LSM value. As we expected, after adjusting liver brosis stage, the HbA1c and AST were independently associated with LSM value by multivariate linear regression analysis. The effect of glucose metabolism on LSM measurement may be due to the fact that T2DM patients have more severe liver in ammation, and LSM value is closely associated with in ammatory injury [28]. Similarly, AST level was a typical marker of liver in ammation activity. BMI and obesity were reported as independent risk factors of unreliable measurements as well as measuring failure [29]. A prospective multicenter study pointed out that skin capsular distance ≥ 25 mm lead to overestimation of LSM value [30]. In our study, BMI also affect the LSM measurement but no statistical signi cance was obtained because of the small sample size. Its effect may be attributed to in uence on both subcutaneous fat thickness and metabolic state. In addition, there was a trend that the LSM values of patients with BMI ≥ 30 kg/m 2 or HbA1c ≥ 7% was slightly higher than patients without them. The cut-off values of these two subgroups also revealed increase trends in the case of consistent sensitivity or speci city. As the M probe is still the main tool for LSM measurement, the current optimal cut-off values are unsuitable for patients with abnormal glucose metabolism. Different cut-off values need to be considered for different BMI or different metabolic state in clinical application. This is also the reason why we gave speci c cut-off values for this population in our study. The results need to be further con rmed by expanding subject number in the future.
Our study had several limitations. Firstly, the data featured skewed distribution. The majority of patients involved have moderate NASH. There was a relative lack of data for patients with NAFL and severe brosis, as liver biopsy is not recommended by the guideline for these parts of the population. The other limitation was that subgroup analysis revealed a general trend, but due to the small sample size, statistically signi cance was not detected in some subgroup comparisons. Similar analysis needs to be conducted by expanding the number of subjects in the future. Besides, more detailed subgrouping and the cut-off value comparison for each histologic degree can be carried out.
In conclusion, FibroScan is con rmed to be a relatively accurate diagnostic approach for evaluating liver brosis among NAFLD patients with abnormal glucose metabolism. It is valuable for evaluating severe brosis, especially for excluding advanced brosis. Glucose metabolism state should be considered in clinical application of LSM measurement.