Low Serum Total Testosterone Is Associated with Non-Alcoholic Fatty Liver Disease in Men but Not in Women with Type 2 Diabetes Mellitus

Materials and Methods There were 1155 patients with T2DM included in the analysis. Serum levels of total testosterone and the precursors of androgens, including androstenedione, DHEA, and DHEAS, were quantified using liquid chromatography-tandem mass spectrometry assays. Results The risk of NAFLD decreased as total testosterone concentration increased in men with T2DM. After adjusting for age, current smoking, current drinking, body mass index, duration of T2DM, diastolic blood pressure, total cholesterol, triglycerides, low-density lipoprotein/high-density lipoprotein cholesterol ratio, uric acid, C-reactive protein, and sex hormones in model 4, the adjusted odds ratio (OR) and 95% confidence interval (CI) of NAFLD for tertile3 vs tertile1 was 0.37 (0.17–0.77; P = 0.024 for trend). When taken as a continuous variable, this association was still robust in model 4 (OR, 0.58; 95% CI, 0.42–0.80; P < 0.05). No significant associations were found between increasing levels of the precursors of androgens and the odds of NAFLD in men with T2DM (all P > 0.05). Moreover, women showed no significant associations of total testosterone, androstenedione, DHEA, and DHEAS, with the odds of NAFLD (all P > 0.05). Conclusions Serum total testosterone was independently associated with the risk of NAFLD among men with T2DM. This study highlights the potential role of testosterone as a risk factor for NAFLD in patients with T2DM.


Introduction
Non-alcoholic fatty liver disease (NAFLD) is a metabolic liver disease, which has become a major health issue, affecting 25% of the adult population worldwide [1]. NAFLD ranges from simple fatty liver to non-alcoholic steatohepatitis (NASH) and may progress to liver cirrhosis and cancer [2]. Due to the increasingly unhealthy lifestyles in recent years, China has the highest prevalence and incidence of NAFLD in Asia [3]. Moreover, NAFLD with metabolic syndrome increases the risk of all-cause, liver-specific, and cardiovascular death [4].
Previous studies have reported that men and women differ in their susceptibility to metabolic diseases [5]. Accumulating evidence has demonstrated that sex steroid hormones play a significant role in the progression of cardiovascular and metabolic diseases. e sex and menopausal status differences in the NAFLD incidence [6][7][8] suggest the potential role of sex hormones as mediators in the progression of this disease. e prevalence of NAFLD is higher among patients with metabolic diseases than in the general population [9]; it is predicted that 75% of patients with type 2 diabetes mellitus (T2DM) have some form of NAFLD [10]. In these patients, it is necessary to identify the factors that are related to a high risk of NAFLD. Low testosterone is independently related to hepatic steatosis and NASH in men [11][12][13]. However, in patients with T2DM, few research studies have reported the association of testosterone and the precursors of androgens with risk of NAFLD, including androstenedione, dehydroepiandrosterone (DHEA), and dehydroepiandrosterone sulfate (DHEAS).
Accordingly, the aim of this cross-sectional study was to explore the associations of testosterone, androstenedione, DHEA, and DHEAS with the risk of NAFLD in patients with T2DM.

Study Subjects.
e medical records of 1,416 consecutive patients with T2DM were reviewed from October 12, 2020, to December 31, 2021. All patients were hospitalized, and their sex steroid hormone levels were measured at the Department of Endocrinology and Metabolism, Tianjin Medical University General Hospital. If multiple inpatient records were available for the same individual, only one record was selected. e exclusion criteria were as follows: (1) lack of abdominal ultrasonography examination; (2) ethanol consumption of ≥210 g/week for men and ≥140 g/ week for women; and (3) viral or autoimmune hepatitis, drug-induced liver disease, Wilson's disease, Cushing's syndrome, and hypothyroidism. Figure 1 illustrates the study population identification process. Based on the exclusion criteria, a total of 1155 patients with T2DM were included in the final analysis.
is study was approved by the institutional review board of Tianjin Medical University General Hospital. e requirement for informed consent was waived (approval number: IRB2020-YX-027-01) because the patient information was extracted from electronic medical records at the Department of Endocrinology and Metabolism, and the patients' identities were kept anonymous, except for the date of birth.

Information Collection and Measurements of Androgens.
Data including age, sex, lifestyle characteristics, insurance, medical history of diabetes, hypertension and dyslipidemia, height, weight, systolic blood pressure, diastolic blood pressure, glycosylated hemoglobin (HbA1c), total cholesterol (TC), triglycerides (TG), high-density lipoprotein (HDL), low-density lipoprotein (LDL), uric acid (UA), C-reactive protein (CRP), aspartate aminotransferase (AST), and alanine aminotransferase (ALT) were obtained from the medical records. Body mass index (BMI) was calculated as the patients' weight in kilograms divided by the square of the height in meters. Alcohol drinking was assessed by selfreport. Excess alcohol consumption was defined as drinking more than 210 g/week for men and 140 g/week for women, respectively.
Total testosterone (TT) and precursors of androgens, including androstenedione, DHEA, and DHEAS, were quantified using liquid chromatography-tandem mass spectrometry (LC-MS/MS) assays in the Laboratory of Endocrinology and Metabolism at Tianjin Medical University General Hospital. Patients' venous blood samples were drawn in the morning after hospital admission following at least 10 hours of fasting. An extraction liquid for steroids (Hangzhou Calibra Diagnostics Co., LTD., Zhejiang, China) was used for sample pretreatment. Androgens were measured using a Jasper ™ HPLC system coupled to an AB SCIEX Triple Quad ™ 4500MD mass spectrometer with a heated nebulizer ionization source in positive ion mode. Quantitative data analysis was performed using Multi-Quant ™ MD 3.0.2 software. Calibration curves of each set of samples were generated by plotting the area ratios of the analyte to the internal standard peak of the six calibration standards; two quality control samples were contained in each set of samples. e plot used linear regression with 1/x 2 weighting, and the correlation coefficients were all greater than 0.99.

Definitions.
Diabetes was defined as a fasting blood glucose level ≥7.0 mmol/L, 2-hour blood glucose level ≥11.1 mmol/L, HbA1c level ≥6.5%, previous medical history of diagnosed diabetes, or use of antidiabetic drugs [14]. Dyslipidemia was defined as TC ≥ 6.2 mmol/L, TG ≥ 2.3 mmol/L, LDL cholesterol ≥4.1 mmol/L, HDL cholesterol <1.0 mmol/L, or use of lipid lowering drugs [15]. Hypertension was defined as a systolic blood pressure ≥140 mmHg, diastolic blood pressure ≥90 mmHg, previous medical history of diagnosed hypertension, or use of antihypertensive drugs [16]. e diagnosis of NAFLD was based on the abdominal ultrasound images. Experienced sonographers performed ultrasound at Tianjin Medical University General Hospital following a uniform standard protocol who were blinded to the patients' clinical information. NAFLD was diagnosed on the basis of high-level echoes from the hepatic parenchyma, differences in hepatorenal echo, echo penetration deep into the liver, and vascular blurring on ultrasonography [17].

Statistical
Analyses. Normally distributed continuous variables were shown as mean ± standard deviation (SD) and compared between the groups using Student's t-test. Nonnormally distributed continuous variables were expressed as medians with interquartile ranges (IQRs) and compared between the groups using the Mann-Whitney U test. Categorical variables were presented as numbers with frequencies and were compared between the groups using the chi-squared test. Binary logistic regression analyses were used to assess the contribution of the androgens to NAFLD after adjusting for confounding factors in four models: model 1 was adjusted for age and post-menopausal status (only for women); model 2 was adjusted for model 1 plus current smoking and current drinking; model 3 was adjusted for model 2 plus BMI, duration of T2DM, diastolic blood pressure, TC, TG, LDL/HDL cholesterol ratio, UA, and CRP; and model 4 was mutually adjusted for sex hormones in the basis of model 3. Factors that showed statistical significance in the univariate analysis, including age, post-menopausal status, BMI, duration of T2DM, diastolic blood pressure, TC, TG, LDL/HDL cholesterol ratio, UA, and CRP, or were considered clinically relevant, such as smoking and drinking, were included in the binary logistic regression analyses. e associations were shown as odds ratios (ORs) and 95% confidence intervals (CIs); P < 0.05 (two-tailed test) was considered statistically significant. SPSS for Windows (version 25.0; SPSS, Chicago, IL, USA) was used for all analyses.

Characteristics of the Study Population with and without NAFLD.
In the study population of 600 men, NAFLD was observed in 383 (63.8%). BMI, diastolic blood pressure, TC, TG, LDL-C, LDL/HDL cholesterol ratio, UA, AST, and ALT were higher in men with NAFLD than in those without NAFLD (all P < 0.05). Men diagnosed with NAFLD also had a higher prevalence of dyslipidemia (P � 0.002). However, the mean age, duration of T2DM, and HDL-C were lower in men with NAFLD than in those without NAFLD (all P < 0.001). As for TT and precursors of androgens, men diagnosed with NAFLD had higher DHEA and DHEAS levels but lower TT levels (all P < 0.001) ( Table 1).
A total of 375 (67.6%) of the study population of 555 women were diagnosed with NAFLD. Similar to men, women with NAFLD had a higher BMI, systolic blood pressure, diastolic blood pressure, TC, TG, LDL-C, LDL/ HDL cholesterol ratio, UA, CRP, AST, and ALT but lower mean age, duration of T2DM, and HDL-C than those without NAFLD (all P < 0.05). Women diagnosed with NAFLD had higher dyslipidemia but lower post-menopausal status rates (all P < 0.05). e level of DHEAS was higher in women with NAFLD (P � 0.016, Table 2). Figure 2shows the prevalence of NAFLD in men and women according to the tertiles of serum TT, androstenedione, DHEA, and DHEAS. e percentage of men with NAFLD significantly decreased with increasing tertiles of serum TT, with 73.0% in ter-tile1, 68.0% in tertile2, and 50.5% in tertile3 (P < 0.001). Furthermore, the prevalence of NAFLD in men gradually increased in line with increasing tertiles of DHEA and DHEAS, with 56.5% and 52.0% in tertile1, 62.0% and 66.5% in tertile2, and both 73.0% in tertile3, respectively (all P < 0.05). In women, we found that the percentage of NAFLD increased with increasing tertiles of DHEAS, with 65.9% in tertile1, 60.0% in tertile2, and 76.8% in tertile3 (P � 0.026).

Associations of TT, Androstenedione, DHEA, and DHEAS
with the Risk of NAFLD. Table 3 presents the associations of TT, androstenedione, DHEA, and DHEAS with risk of NAFLD in men after adjusting for confounding variables in the multivariate logistic regression analysis. When only adjusted for age in model 1, the risk of NAFLD decreased with an increasing TTconcentration. Even after adjusting for Data from 1292 unique inpatients were collected 83 records excluded due to lack of abdominal ultrasound 11 records excluded due to age less than 18  Analysis data set of 1155 patients with type 2 diabetes mellitus 44 records excluded due to heavy drinking 7 records excluded due to hepatitis B 1 records excluded due to hepatitis C 1 records excluded due to autoimmune hepatitis 1 records excluded due to drug-induced liver disease 7 records excluded dueto cushing syndrome 65 records excluded due to hypothyroidism 7 records excluded due to liver cirrhosis 4 records excluded due to hepatic abscess confounding variables in models 2, 3, and 4, this association remained robust. e fully adjusted OR (95% CI) for tertile3 vs tertile1 was 0.37 (0.17-0.77; P � 0.024 for trend) in model 4. Similarly, when taken as a continuous variable, each SD increment in the TT level was significantly related to a 42% decrease in the odds of developing NAFLD in model 4 (OR, 0.58; 95% CI, 0.42-0.80; P < 0.05). No significant relationships were shown between increasing levels of androstenedione, DHEA, DHEAS, and the risk of NAFLD (all P > 0.05). As shown in Table 4, women with T2DM showed no significant associations of increasing levels of TT, androstenedione, DHEA, and DHEAS, with the odds of NAFLD (all P > 0.05).

Discussion
In this cross-sectional study, we found that serum TT was significantly associated with NAFLD prevalence in men with T2DM.
is finding remained significant even after adjusting for potential confounding factors, including age, current smoking, current drinking, BMI, duration of T2DM, diastolic blood pressure, TC, TG, LDL/HDL cholesterol ratio, UA, and CRP. In contrast, no statistically significant associations were found between TT levels and the risk of NAFLD in women with T2DM. Additionally, the participants showed no statistically significant relationships between increasing levels of androstenedione, DHEA, DHEAS, and NAFLD prevalence in men and women with T2DM.
Previous studies have shown that testosterone is involved in the incidence and progression of cardiovascular and metabolic diseases [18,19]. NAFLD, which is causally related to metabolic syndrome [20], shares many potential risk factors for metabolic diseases. e inverse association between testosterone levels and the risk of NAFLD has been found in many studies among men. Nationally representative cross-sectional analyses in the United States (US) showed that the TT concentration was inversely related to the risk of NAFLD in men, independent of the age, obesity, and lifestyle [13,21]. In addition, a cohort study by the multicenter NASH Clinical Research Network found that low testosterone levels were related to the development of NASH and fibrosis severity among men with NAFLD [12]. Moreover, in men with T2DM, an inverse association was found between serum TT and markers of liver fibrosis, including the FIB-4 index and 7S domain of type IV collagen [22]. Consistent with these studies, we concluded that, in men with T2DM, serum TT was significantly related to the  International Journal of Endocrinology risk of NAFLD after adjusting for potential confounding factors, including age, BMI, lifestyle, lipids, and uric acid. erefore, an examination for NAFLD is essential for male patients with a low level of TT in clinical practice.
Mechanistically, the link between the liver and sex hormones is complex and bidirectional. One previous study showed that testosterone activated the androgen receptor/ β-catenin interaction and subsequently translocated this complex to the nucleus, thus downregulating adipogenic transcription factors and inhibiting adipogenic differentiation [23]. In men, testosterone has been shown to inhibit lipoprotein lipase activity in the adipose tissue [24], which provides fat cells with free fatty acids for intracellular esterification by hydrolyzing circulating triglyceride-rich lipoproteins. For this reason, testosterone deficiency may contribute to visceral obesity and subsequent insulin resistance, which was found to upregulate glucose-6-phosphatase levels and increase hepatic gluconeogenesis [25]. In addition, the lack of hepatic androgen receptors decreased fatty acid beta-oxidation and increased de novo lipid synthesis, resulting in hepatic steatosis and insulin resistance [26]. On the other hand, high plasma levels of IL1β, which is a type of pro-inflammatory cytokine mainly produced by macrophages, may lead to decreased hepatic expression of sex hormone-binding globulin (SHBG) in patients with obesity [27]. us, fat deposition in the liver may affect the reproductive axis by regulating sex hormone metabolism through SHBG [28].
Significant benefits of testosterone treatment have been observed in animal and human research. In castrated male rats, testosterone replacement decreased micro and macrovesicular fat and apoptosis in hepatocytes, thus improving the NAFLD [29]. Similarly, long-term testosterone therapy produced a significant decline in the body weight, waist circumference, BMI, and lipid profile in men with hypogonadism [30][31][32]. In addition, testosterone therapy decreased the fatty liver index, gamma-glutamyl-transferase, and triglycerides, produced metabolically healthier preadipocytes, and improved mitochondrial functioning and lipid handling in hypogonadal men [33,34]. Further prospective studies with larger sample sizes are needed to better understand the value of testosterone replacement for patients with metabolic diseases.
In women, the relationship between testosterone levels and the odds of NAFLD is controversial. A US study using data from 2011 to 2012 National Health and Nutrition Examination Survey reported that low TT levels were related to suspected NAFLD, diagnosed when serum ALT was greater than 19 IU/L, among post-menopausal women [21]. However, in the prospective Coronary Artery Risk Development in Young Adults study, high free testosterone was found to be related to NAFLD prevalence among middleaged women, independent of traditional risk factors [35]. In middle-aged and elderly Chinese women living in the community, the calculated serum free testosterone and bioavailable testosterone levels were found to be positively associated with the risk of NAFLD, but TT was not significantly related to NAFLD [36]. In our study, we did not find a statistically significant relationship between TT and the odds of NAFLD among women with T2DM. In general, there are three different forms of testosterone in the circulation: free testosterone, testosterone bound to albumin, and testosterone bound to SHBG. Bioavailable testosterone forms, including free testosterone and albumin-bound testosterone, are readily available to cells. ese different forms of testosterone could partly explain the different conclusions of the above studies, although further research is needed to clarify this issue.  International Journal of Endocrinology As is well known, androstenedione, DHEA, and DHEAS, as precursors of androgens, are converted into testosterone and dihydrotestosterone in target organs and bind to androgen receptors to produce the effects of androgens. DHEA has been shown to increase insulin sensitivity [37] as well as predict all-cause and cardiovascular disease death [38]. e level of DHEA was lower in obese children aged 8.5 to 18.0 years with NAFLD than in those without liver disease [39]. An animal study also proved that DHEA can reduce tumor necrosis factor-α-induced hepatocyte apoptosis and, therefore, decrease liver injury in mice [40]. However, in the current study, no significant association was found between DHEA and the risk of NAFLD in either men or women with T2DM. ese differences could possibly be explained by disparities in the study designs and populations.
DHEAS is the most abundant androgen precursor in circulation. In a retrospective study of women with polycystic ovary syndrome, the level of DHEAS was similar in patients with and without NAFLD when using ultrasound to diagnose this disease [41]. Nevertheless, other studies have reported that a low level of circulating DHEAS is histologically a crucial determinant of the severity of fibrosis in NAFLD [42,43]. In the present cross-sectional study, we found no statistically significant relationship between DHEAS and the odds of NAFLD among individuals with T2DM.
e differences in the criteria used to diagnose NAFLD as well as the specific populations included could be partly responsible for the conflicting results.
Androstenedione, which serves as a precursor in the biosynthesis of testosterone, is produced by the gonads and adrenal glands of both sexes. Few research studies have examined the relationship between androstenedione and the risk of NAFLD. Consistent with the results in children [44], this association was also not found to be statistically significant in patients with T2DM in this study.
is study has several limitations. First, we could not establish a causal relationship between androgens and NAFLD due to the cross-sectional nature of our study. Second, since the level of SHBG was not measured in our patients, we could not evaluate the role of free and bioavailable testosterone in the risk of developing NAFLD in patients with T2DM. ird, exercise and eating habits were not included in this study, which may be confounding factors in the development of NAFLD. Finally, the lack of histological evidence of NAFLD is another potential weakness of our study. Nevertheless, abdominal ultrasound is the preferred test for NAFLD in epidemiological studies because of the invasive nature of liver biopsy.
In conclusion, we found that serum TT was independently associated with the risk of NAFLD among men with T2DM in this cross-sectional study. In contrast, no statistically significant associations were found between precursors of androgens, including androstenedione, DHEA, and DHEAS, and the risk of NAFLD in men and women with T2DM. is study highlights the potential role of testosterone as a risk factor for NAFLD in patients with T2DM. Further longitudinal studies are needed to establish causal associations between androgens and the risk of NAFLD in the future.
Data Availability e dataset generated and analyzed during the current study is available from the corresponding author upon reasonable request.

Disclosure
Xinxin Zhang, Jinfeng Xiao, and Qi Liu are co-first authors.

Conflicts of Interest
e authors declare that they have no conflicts of interest.

Authors' Contributions
Xinxin Zhang, Jinfeng Xiao, and Qi Liu contributed equally to this study.