Hepatic Steatosis and Fibrosis in Type 2 Diabetes: A Risk-Based Approach to Targeted Screening

1Liver and Pancreatobiliary Diseases Research Center, Digestive Disease Research Institute, Shariati Hospital, Tehran University of Medical Sciences, Tehran, Iran 2Endocrine Research Center, Institute of Endocrinology and Metabolism, Iran University of Medical Sciences, Tehran, Iran 3Research Center for Prevention of Cardiovascular Disease, Institute of Endocrinology and Metabolism, Iran University of Medical Sciences, Tehran, Iran


Introduction
Non-alcoholic fatty liver disease (NAFLD) is a common chronic liver disease, encompassing a range of histopathologic abnormalities from simple steatosis to advanced fibrosis and cirrhosis. 1 The reported prevalence of NAFLD in the general population varies from 19% to 46% based on the study population, age, and the tools used for diagnosis. 2 Based on a meta-analysis published in 2016, NAFLD is estimated to affect approximately one third of the Iranian population (33.9%, 95% CI: 26.4%-41.5%). 2 It is well known that NAFLD is associated with obesity, type 2 diabetes (T2DM), and insulin resistance. 3 On the other hand, diabetes is an independent risk factor of advanced fibrosis in patients with NAFLD. 4 In addition, the incidence of cirrhosis is higher in patients with diabetes. 5 With the steady increase in the prevalence of T2DM, the burden of NAFLD is accordingly expected to increase. Therefore, early diagnosis and management are required to prevent progressive liver disease.
Routine screening of steatosis and liver fibrosis in patients with diabetes is not generally recommended due to the lack of precise and non-invasive methods for diagnosis of liver fibrosis, as well as limited therapies available. 6 While liver biopsy is considered to be the gold standard method of liver disease diagnosis, it is not routinely performed for NAFLD detection due to its invasive nature. Currently, ultrasound is the most commonly used tool for detecting fatty liver. However, abdominal ultrasound is not susceptible to mild steatosis. 7 Computed tomography (CT) scan also has a low sensitivity in detecting earlier stages of fatty liver, while at the same time exposing patients to radiation. Although magnetic resonance imaging (MRI) is considered to be the most accurate quantitative method to measure liver fat, its high cost is an important barrier for routine use in clinical practice.
Fibroscan is a non-invasive tool that measures hepatic elasticity using ultrasound transient elastography (TE). [8][9][10] With a sensitivity of 94.80% and specificity of 77.08%, TE can be used as an acceptable noninvasive screening tool in large populations. 11 Given that liver stiffness is correlated to liver fibrosis, TE can be used to detect liver cirrhosis with high accuracy. 12 The fibroscan device has different sized probes for various body sizes. To overcome body weight as a potential barrier in obese individuals, an XL probe as well as a controlled attenuation parameter (CAP) can be implemented on the device to measure the level of steatosis. 13 Existing evidence indicates that the prevalence, severity and outcomes of NAFLD could be influenced by some underlying factors including ethnicity. 14 To the best of our knowledge, there is currently no study on liver steatosis and fibrosis using fibroscan in individuals with T2DM compared to those without diabetes matched for age, sex, and body mass index (BMI) in the Iranian population. Considering NAFLD as a multisystem disease affecting extra-hepatic organs and its association with hard endpoints such as cardiovascular disease, we aimed to explore the frequency and predictors of NAFLD in individuals with T2DM and to identify high risk subgroups.

Materials and Methods
This was a cross-sectional study including 100 individuals with T2DM and 100 without diabetes matched for age, sex, and BMI by category/mean matching method.
The study included patients 30 to 65 years of age who had diabetes for at least 3 years. The following exclusion criteria were applied: history of liver disease or symptoms of active liver disease, evidence of other chronic liver diseases such as HBS Ag + , Anti-HCV Ab + , history of alpha-1 anti-trypsin deficiency, autoimmune hepatitis, liver damage caused by drugs, primary biliary cirrhosis, primary sclerosing cholangitis, biliary obstruction in previous radiological studies, advanced complications of diabetes (CKD ≥ Stage III, CHF, Advanced CVD), excessive alcohol consumption (>20 g/d), pregnancy, and intake of medications that can lead to NAFLD, namely, corticosteroids, tamoxifen, amiodarone and methotrexate.
Sample Size Calculation Sample size was determined using the G*Power software. Considering the prevalence of liver fibrosis in individuals with diabetes (40.3%) and non-diabetic people (17%) reported in previous studies, 15 a sample of 160 subjects (80 in each groups) could provide sufficient power (80%) at the 5% significance level. However, a total of 200 subjects were recruited to increase the study power. Furthermore, given previous studies suggesting 10 subjects per variable as an appropriate sample size for linear and logistic regression models, 16,17 the determined sample size would be enough to ensure the evaluation of main predictors of liver steatosis and fibrosis in the study population (19 candidate variables in overall participants and 9 candidate variables in each study group, as described in the Statistical Analysis section).
Clinical Assessment Demographic and anthropometric measurements were obtained by a trained investigator. BMI was calculated by dividing body weight (in kilograms) by the square of the height (in meters). A fasting blood sample was taken for the measurement of metabolic variables including fasting blood sugar (FBS), glycosylated hemoglobin (HbA1c), triglyceride (TG), total cholesterol (Chol), high density lipoprotein (HDL), and low-density lipoprotein (LDL), aspartate aminotransferase (AST), alanine aminotransferase (ALT), gamma-Glutamyl transpeptidase (GGT), alkaline phosphatase (ALP), insulin level, and viral markers (HBS Ag and Anti-HCV). Insulin resistance was assessed by the homeostatic model assessment -insulin resistance (HOMA-IR) based on the following formula: [fasting glucose (mg/dL) × insulin (μU/mL)] divided by 405. 18 Blood glucose levels were determined by the enzymatic colorimetric method. Enzymatic assays were also used to determine the levels of serum triglyceride, total cholesterol and HDL-cholesterol. All analyses were performed using Pars Azmun diagnostic kits (Pars Azmun Co., Tehran, Iran) with a between-and within-run coefficient of variation <6.2 %. LDL-cholesterol was calculated using a modified version of the Friedewald equation. 19 Serum insulin (ECLIA method), ALT (IFCC method), AST (IFCC without pyridoxal phosphate activation), GGT (acc.Ver.1 Standardized against IFCC) and AlP (acc. IFCC Gen2 ALP2L) were determined using Roche Diagnostics kits (Roche Cobas 6000 analyzer) with coefficient of variations <3.2%.
Dietary intake was assessed using a food frequency questionnaire (FFQ), which was completed for all participants during a face-to-face interview. To assess physical activity, the short form of the international physical activity questionnaire (IPAQ-SF) was used categorizing physical activity levels as low, moderate, and high. 20 All TE was performed using FibroScan ® (Echosens, Paris, France) by a trained specialist physician who was blinded to the subjects. Participants were advised to be fasted for at least 3 hours prior to their TE test. For each participant, at least 10 valid measurements were obtained, with an interquartile range to median ratio <0.3. The final score was reported as the median of individual measurements. Based on the measurements, a CAP score (amount of steatosis in decibels per meter (dB/m) of liver tissue) was designated to each individual, from which NAFLD diagnosis and the grade of steatosis were determined. Steatosis can range from 100 to 400 dB/m; individuals with a score greater than 238 dB/m were diagnosed with NAFLD, with the following CAP score ranges, indicating the grade of steatosis: grade 1 (S1): 238 dB/m ≤ a CAP score ≤259 dB/m, grade 2 (S2): 260 dB/m ≤ CAP score ≤ 291 dB/m, and grade 3 (S3): 292 dB/m ≤ CAP score. 21,22 Hepatic fibrosis and its stages were assessed by the METAVIR scoring system. The five stages of hepatic fibrosis include: F0: no fibrosis: mild fibrosis (F1 < 7.1 KPa), significant fibrosis (7.1 ≤ F2 < 9.5 kPa), severe fibrosis (9.5 ≤ F3 < 12.5 kPa), cirrhosis (F4 ≥ 12.5 kPa). A fibrosis score equal to or greater than F2 was considered as significant fibrosis. 23,24 Statistical Analysis Statistical analyses were carried out using IBM SPSS Statistics for Windows (version 20.0 IBM Corp. Released 2011, Armonk, NY). Continuous data were assessed for normality using histograms as well as the skewness and kurtosis statistics, in which the values outside the range of -1 to +1 were considered as being non-normal. Demographic and clinical characteristics of participants in the study groups are presented as proportions, means ± standard deviation, or medians with interquartile range (IQR) for skewed data. Between-group comparisons were performed using independent sample t test and non-parametric Mann-Whitney U test for normally distributed and nonnormal data, respectively. Categorical data were compared using χ 2 test if at least 80% of the cells had an expected count greater than 5, or Fisher exact test if this assumption was not met. A Spearman's partial correlation analysis was undertaken to examine the correlations of the fibrosis score with various clinical parameters, after controlling for age and gender. The significant predictors of hepatic fibrosis and steatosis by diabetes status were explored using binary logistic regression analyses and multivariable linear regression analyses with backward procedure, considering an alpha level of 0.157 for backward selection. In linear regression analyses, the candidate predictors for inclusion in the multivariable models in overall participants have been general covariates of age, gender, smoking status, diabetes status and statin medication as well as those with P value < 0.1 in the univariate correlation analysis with CAP score, including: BMI, waist-hip ratio (WHR), systolic blood pressure (SBP), fasting blood glucose (FBG), HOMA, TG, HDL, ALT, AST, GGT, A1C. For performing multivariable analyses in each study group, the candidate predictors consisted of gender and statin medication as well as those with P value < 0.1 in the univariate correlation analysis with CAP score in the study groups, including the BMI, WC/WHR, HOMA, TG, ALT, AST, GGT (a total of 9 variables). Similarly, the candidate variables for inclusion in logistic regression analysis in overall participants were general covariates of age, gender, smoking status and diabetes status as well as those with P value < 0.1 in the univariate correlation analysis with fibrosis score, including: BMI, WC, WHR, SBP, BPD, FBG, HOMA, TG, HDL, LDL, ALT, AST, GGT, A1C and CAP score. However, due to low effective sample size, we had limitations in performing logistic regression analyses in the study groups separately. Thus, only the results of the analysis in the overall participants have been presented. All tests were 2-tailed, and P ≤ 0.05 was considered statistically significant.

Results
Of 200 participants enrolled initially, two participants without diabetes were excluded later because of positive HBsAg and anti-HCV test results. Accordingly, data from 100 individuals with T2DM, and 98 individuals without diabetes matched for age, sex, and BMI were used in the final analyses. Table 1 presents the demographic, clinical and biochemical characteristics of the participants by diabetes status. As expected, there was no significant difference between the two study groups in terms of age, sex, and BMI; however, those with diabetes had significantly higher mean SBP, diastolic blood pressure (DBP), fasting blood glucose, HbA1C, HOMA-IR, AST, GGT and total energy intake compared to their counterparts without diabetes (all P values < 0.05).
Results of the fibroscan test (Table 2) indicated that, despite comparable rates of NAFLD (S1-S3) in both study groups (T2DMs: 87% vs. non-T2DM: 78.6%, P = 0.11), individuals with diabetes had a significantly higher mean CAP score (294.90 ± 53.12 vs. 269.78 ± 45.05 dB/m; P < 0.001), resulting in a significantly higher rate of severe steatosis (S3) compared to those without diabetes (50% vs. 31.6%, P = 0.009). Similarly, our results showed a significantly higher mean fibrosis score in individuals with diabetes compared to that in individuals without diabetes [5.50 kPa (4.50-6.10) vs. 4.50 kPa (4.00-5.50), P < 0.001]. Given that fibrosis levels greater than F2 are clinically more significant, affecting patient outcomes, individuals in this study were categorized as those with no or minimal fibrosis (F0-F1), and those with advanced fibrosis (F2-F4). Advanced fibrosis was accordingly diagnosed in 13% of people with diabetes, while its frequency was 4% in individuals without diabetes (P = 0.02).
We also found that patients with diabetes and advanced fibrosis had significantly higher BMI, WC, WHR, ALT, AST, and CAP score compared to those with no or minimal fibrosis (P < 0.05); however, duration of diabetes did not affect fibrosis level in this study (Table 3). On the other hand, people without diabetes with and without advanced stages of liver fibrosis only differed significantly in BMI (Table 4). Table 5 demonstrates the correlation of fibrosis scores with demographic and clinical parameters by diabetes status. BMI, WC, WHR, fasting insulin, GGT and CAP score had positive correlations with fibrosis scores in both groups. Also, HOMA-IR was correlated with hepatic fibrosis in patients with diabetes. However, the correlation did not remain statistically significant when they were adjusted for age, sex, BMI, and WC.
Using multivariable linear regression analyses for the entire group (Table 6), diabetes status, BMI, WHR and ALT were significantly associated with higher CAP scores (model adjusted R 2 : 0.47), with BMI having the greatest influence (β = 0.49, P < 0.001). A subsequent separate analysis for the two study groups highlighted BMI, WHR, AST and female gender as significant determinants of CAP scores in patients with diabetes (model adjusted R 2 : 0.50); and WC, serum TG level, and statin medication in people without diabetes (model adjusted R 2 : 0.43). Moreover, we found that in people with diabetes, BMI and WHR could each explain 43% and 18% of variations in CAP score, respectively.

Discussion
This study shows that people with diabetes are more likely to display more severe forms of steatosis and fibrosis compared to people without diabetes with the same age, sex and BMI. CAP score, HOMA-IR, higher BMI, higher WC, WHR, and AST were contributing factors to more severe NAFLD.
Parallel to the increased incidence of diabetes, NAFLD, as the most common cause of chronic liver diseases, has also increased. 25,26 Recent studies showed NAFLD as a multisystem disease involving liver, cardiovascular, and Continuous variables are expressed as mean ± SD or as median (IQR) for skewed data. Categorical variables are presented as n (% within group). Betweengroup comparisons were performed using independent sample t-test and non-parametric Mann-Whitney U test for normally distributed and skewed data, respectively. Categorical data were compared using χ2 test. a Determined in 70 diabetic participants who did not take insulin medication; b Evaluated in 99 diabetics and 98 non-diabetic participants.    Continuous variables are expressed as mean ± SD or as median (IQR) for skewed data unless otherwise noted. Categorical variables are presented as n (% within liver fibrosis stage). Between-group comparisons were performed using independent sample t-test and non-parametric Mann-Whitney U test for normally distributed and skewed data, respectively. Categorical data were compared using χ 2 test. cerebrovascular systems. 25,27 Diabetes and NAFLD increase the risk of overall mortality by 2.2 folds 28 and in people with diabetes, NAFLD increases the risk of cardiovascular events by 1.87 folds. 29 In addition, diabetes increases the risk of hepatocellular carcinoma as well as hepatocellular carcinoma mortality by 2-3 folds. [30][31][32] Therefore, timely detection and treatment of liver steatosis and fibrosis is crucial to improve long-term prognosis of individuals with diabetes. However, up to now, routine screening has not been recommended for NAFLD due to cost, limited treatment options, and current ineffectiveness of diagnostic modalities. 33 Furthermore, there is controversy regarding whether a screening program should be implemented for the total population or only high risk groups, namely people with diabetes. 34 Although liver biopsy is the gold standard to assess NAFLD, it is an invasive method with a fatality rate of 0.05%. 35 Moreover, it is not feasible in a large population. Therefore, only a limited number of studies have used liver biopsy to assess NAFLD in patients with diabetes. 7,36,37 Some studies used magnetic resonance spectroscopy (MRS) to evaluate NAFLD, 38,39 though it is expensive with limited accessibility. Considering MRS > 5.5% as indicator of NAFLD, 60.3% of patients with diabetes were diagnosed with NAFLD in one study. 39 In another study on 100 patients with diabetes conducted by Doycheva et al, the prevalence of advanced fibrosis (magnetic resonance elastography [MRE] > 3.6 kPa) was 7.1% using MRE. 28 In a study that was conducted by Park et al, 40 MRE was compared with TE with XL probe in detection of fibrosis and steatosis in patients with biopsy-proven NAFLD. They found MRI to have higher diagnostic accuracy for fibrosis and steatosis in NAFLD patients, [41][42][43] while TE had higher negative predictive value for diagnosing significant fibrosis (stages 2-4), severe fibrosis (stages 3-4), as well as cirrhosis. 44,45 Furthermore, in another study conducted by Eddowes et al, it was demonstrated that TE has high applicability as well as a low failure rate in patients with potential NAFLD with area under the receiver operating characteristic curve values ranging from 0.7 to 0.89, showing that histological components or choice of probe do not influence the Continuous variables are expressed as mean ± SD or as median (IQR) for skewed data unless otherwise noted. Categorical variables are presented as n (% within liver fibrosis stage). Between-group comparisons were performed using independent sample t-test and Mann-Whitney U test for normally distributed and skewed data, respectively. Categorical data were compared using Fisher exact test. measurements. 46 Goh et al 15 reported different clinical spectra of NAFLD in people with diabetes compared to people without diabetes using liver biopsy as the gold standard. The mean BMI, AST, and ALT in people with diabetes in that study were higher compared to our population with diabetes. Consistent with our findings, duration of diabetes was not different between people with and without significant fibrosis. However, advanced fibrosis was reported to be present in 40.3% of people with diabetes compared to 17% in those without diabetes. In our study, significant fibrosis was found in 13% of people with diabetes, and 4% of those without diabetes. This may be due to different stages of chronic inflammation, 47 oxidative stress, 48 as well as the up-regulation of hepatotoxic inflammatory cytokines in diabetes. 49 Other mechanisms might be highlevel of glucose and insulin concentrations which can stimulate connective tissue growth factor expression -a determining factor in progressive hepatic fibrosis. 49 In a study conducted by Nakahara et al 50 in a large patient cohort (N = 1365) with biopsy-proven NAFLD, they found a parallel rise in the prevalence of diabetes and degree of fibrosis, indicating diabetes as a significant risk factor for advanced fibrosis.
Although the prevalence of advanced fibrosis may be lower among people without diabetes, due to larger population of people without diabetes, the absolute number of severe NAFLD in individuals without diabetes may not be lower than people with diabetes.
In another study on 939 patients with T2DM in Edinburgh, the prevalence of NAFLD was 42.6% using ultrasound. 6 The disparity in the prevalence of NAFLD between different studies may be due to the variations in sensitivity of diagnostic modalities as well as different methods of sample selection, dietary, lifestyle habits of participants, and different patient characteristics. 51 The results of our study showed that the female gender was a risk factor for NAFLD in patients with diabetes. While in another cross-sectional study on 2839 patients with diabetes, the prevalence of NAFLD was reported to be 71.1% in men and 68% in women, though the age range was not different to ours. But the duration of diabetes in that study was higher in people with NAFLD compared to ours. 29 Consistent with our findings, in a study conducted by Summart et al 52 on 34 709 participants (27 073 females and 7636 males), they found a higher NAFLD prevalence in women, which increased with age, with the largest difference among the population of 56-60 years old as well as patients with diabetes. The age range can be justified by natural physiologic changes in women during the postmenopausal period including insulin resistance, central obesity, adipose distribution,  and sex hormones. 53 Eliminating the effect of estrogen as a powerful antioxidant may lead to fatty acid oxidant decrease, while increasing lipogenesis within the liver resulting in visceral fat accumulation and NAFLD. 54 In addition, estrogen is a protective factor against diabetes 55 ; therefore, the risk of NAFLD increases in the postmenopausal period. Moreover, insulin resistance in multiple organs such as the liver, muscle and adipose tissue is accompanied by higher gluconeogenesis in the liver, insulin clearance reduction, and impaired glucose uptake by muscles, all of which lead to elevated plasma glucose concentrations. 56 Insulin resistance in the adipose tissue can also result in higher levels of free fatty acids and inflammatory cytokines. 14 Therefore, transaminase levels increase in patients with NAFLD; however, it may not occur in all patients with diabetes. 57 Moreover, Our study demonstrated that BMI, WC, WHR, HOMA-IR, and CAP score were associated with advanced liver fibrosis, while in a study conducted by Zhao et al, 58 older age, high BMI, smoking, and low PLT levels were reported as independent risk factors for liver fibrosis in patients with diabetes.
Consistent with the results of the study conducted by Zhao et al, our study revealed that high blood pressure as well as GGT and HOMA-IR had a significant association with hepatic steatosis in patients with diabetes. In addition, HOMA-IR has been reported as an independent risk factor for NAFLD and liver fibrosis both in people with diabetes 59 and without diabetes. 4 Therefore, it is reasonable to recommend hepatic steatosis screening in patients with diabetes, especially those who are overweight with high GGT and HOMA-IR as well as high blood pressure.
GGT also had a significant association with hepatic steatosis in individuals without diabetes. GGT, as one of the key enzymes responsible for glutathione metabolism, has been shown to be closely related with cardiovascular disease as well as central obesity. 60 Our study also demonstrated higher GGT in grade 3 steatosis in patients with diabetes compared to individuals without diabetes with grade 3 steatosis, suggesting GGT as a biomarker for hepatic steatosis in patients with diabetes. 58 Strength and Limitation To date, a limited number of studies have explored liver steatosis and fibrosis in people with T2DM compared to individuals without diabetes matched for age, sex, and BMI. The strengths of our study include recruitment of subjects from diabetes clinics in routine clinical practice and age, sex and BMI matched individuals without diabetes, pragmatic use of a combination of readily available biomarkers, and use of transient elastography for liver assessments. In addition, diet and physical activity were assessed. On the other hand, our study had some limitations. The platelet count, as one possible risk factor for liver fibrosis, was not measured. In addition, the sample size was relatively small and patients were enrolled in a single center. A multicenter controlled randomized study is warranted to validate the findings of our study.
In conclusion, individuals with diabetes are more likely to have severe fibrosis. Obesity, central obesity, female gender, elevated liver enzymes, and higher degree of insulin resistance are risk factors predicting severity of NAFLD.
The findings highlight the potential importance of early screening for NAFLD in high risk subgroups of patients with diabetes as well as overweight people with high GGT, high HOMA-IR and high blood pressure to initiate timely behavioral and/or medical intervention.