PDF
Citation
Print
INTRODUCTION
Steatotic liver disease (SLD) is the leading cause of chronic liver disease. In Korea, the incidence/prevalence of metabolic dysfunction-associated steatotic liver disease (MASLD) in 2021 were 13.95/34.23 per 1,000 person and increasing trend was reported [1]. It representing a substantial clinical burden and a public health concern. Most of SLD is considered to reflect the hepatic component of the metabolic syndrome, since it has strong association with insulin resistance, obesity, and dyslipidemia. Recently, a novel definition of MASLD replacing non-alcoholic fatty liver disease (NAFLD) was proposed in a conception that SLD and metabolic syndromes share a common pathway of pathogenesis. MASLD encompass hepatic steatosis status with cardiometabolic criteria, and it can identify more patients at risk for cirrhosis and liver cancer [2,3].
Thyroid hormone has important role in lipid metabolism and energy homeostasis. After thyroxine is converted to the active hormone, triiodothyronine (T3), by deiodinase, it enters target cells such as hypothalamus, adipose tissue, muscle, and hepatocyte via thyroid hormone receptor (THR). In hepatocyte, T3 affects regulation of hepatic fat accumulation through multiple pathways, including stimulation of free fatty acid delivery to the liver for re-esterification to triglyceride (TG), or increasing fatty acid β-oxidation [4,5].
The relationship between thyroid hormone marker, especially T3, and metabolic disease factor such as high body mass index (BMI), insulin resistance, diabetes also has been suggested in recent decades [6-8]. Considering SLD/MASLD is recognized as a hepatic presentation of metabolic syndrome, it is reasonable that T3 is associated with SLD. The large cohort study (The Lifelines Cohort Study) demonstrated that NAFLD was independently associated with a high‐normal free T3 (FT3) level [9]. Also, a study in the middle-aged and elderly euthyroid subjects showed that high-normal FT3 are independently associated with a higher risk of NAFLD [10]. However, the cause-consequence effect may not be established due to cross-sectional design or lack of survival analysis with adjustment. Furthermore, the association between MASLD and T3 has not been researched in longitudinal design though a novel concept of MASLD replacing NAFLD in clinical practice [2,3].
Therefore, in this long-term follow-up longitudinal data, we aimed to explore whether the thyroid hormone maker, especially total T3 (TT3) is associated with either the development of SLD or the resolution of SLD. In addition, we speculated the relationship between TT3 and modified MASLD presenting the emerging concept of MASLD.
METHODS
Study design and study population
In this retrospective longitudinal study, we assessed 2,372 participants who completed annual or biennial examinations between 2007 and 2014 at Gyeongsang National University Hospital (GNUH) healthcare center in South Korea for eligibility. Among them, the participants who aged ≥18 years old and underwent medical examinations with serial thyroid hormone measurements ≥4 times were included (n=2,246). The participants with abnormal initial results of the thyroid function test (n=294), missing data (n=272), and a history of thyroid disease or recent medication affecting thyroid function (n=15) were excluded. We also excluded the participants with liver disease such as hepatitis B virus infection (n=113), hepatitis C virus infection (n=14), liver cirrhosis (n=3), and hepatocellular carcinoma (n=1), and pregnant participant (n=0), which could alter TT3 level significantly [11]. Finally, 1,534 participants were included (baseline cohort). The baseline cohort was used for investigating the association between prevalent SLD or modified MASLD and thyroid hormone markers. Among them, the participants with initial SLD were used for SLD resolution cohort (n=318). After excluding the participants who already have SLD disease at baseline, we generated SLD risk cohort (n=1,216) to figure out whether the thyroid hormone marker related to prevalence of SLD was also associated with incident SLD. To investigate the relationship between the thyroid hormone markers and modified MASLD, we used 269 participants with modified MASLD at baseline (MASLD resolution cohort), and 1,265 participants without initial modified MASLD (MASLD risk cohort) (Fig. 1). The Institutional Review Board (IRB) of GNUH approved this study (No. 2022-02-003) and the requirement for informed consent was waived by the IRB.
Data collection
Demographic and anthropometric characteristics were obtained. BMI was calculated as the body weight (kg) divided by the height squared (m2) that were measured by trained nurses. All participants underwent an overnight fasting and the following laboratory data were measured by blood samples: the levels of thyroid stimulating hormone (TSH), free thyroxine (FT4), TT3, TG, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), aspartate aminotransferase (AST), alanine transaminase (ALT), and hemoglobin A1c (HbA1c). The thyroid hormone including TSH, FT4, TT3 levels were determined by an electro-chemiluminescent immunoassay (Roche Diagnostics Ltd., Mannheim, Germany) (reference ranges of 0.27–4.2 mU/L for TSH, 0.93–1.70 ng/dL for FT4, and 80–200 ng/dL for TT3). Well-trained radiologist performed liver ultrasonography (USG) using an ultrasound system (APLIO I70, Canon, Tokyo, Japan).
Definitions of study outcome and subgroups
SLD was defined as USG diagnosed SLD according to the Asia-Pacific Guidelines—the presence of at least two of the followings: (1) diffusely increased echogenicity liver with liver echogenicity greater than kidney or spleen; (2) vascular blurring; and (3) deep attenuation of ultrasound signal [12]. For sensitivity analysis, we used modified MASLD which defined the participants with SLD in addition to one of the following four criteria: (1) BMI ≥23 kg/m2; (2) presence of type 2 diabetes mellitus or prediabetes; (3) TG ≥150 mg/dL; and (4) HDL-C ≤40 mg/dL (male)/≤50 mg/dL (female), and prediabetes. The definition was adopted from diagnostic criteria of MASLD excluding hypertension (HTN), waist circumference (WC), and alcohol consumption [2,3].
To investigate the association between prevalence of SLD and thyroid hormone makers, TSH, FT4, and TT3 at baseline were used. When we explore the association between development or resolution of SLD and thyroid hormone makers, mean of thyroid hormone makers (mean TT3, mean FT4, mean TT3) which defined as sum of all measured thyroid hormone levels divided by the number of measurements were adopted for quantitative evaluation of thyroid hormone exposure instead of one spot value of baseline. After determining that mean TT3 as a continuous variable is associated with SLD, we further defined mean TT3 into categorical variable: (1) high mean TT3 group—the participants who recorded mean TT3 higher than median mean TT3 (109.45 ng/dL); (2) low mean TT3 group—the participants who recorded mean TT3 same or lower than median mean TT3 (109.45 ng/dL).
BMI was classified into three subgroups (<23 [normal], 23– 25 [overweight], and ≥25 kg/m2 [obese]) using Asia-Pacific BMI classification [13]. Bodyweight gain group was defined as bodyweight increment over 5% and bodyweight loss group was defined as bodyweight decrement over 5% from the baseline during follow-up. Stable bodyweight group included participants whose bodyweight changed less than 5% from the baseline [14,15].
Statistical analysis
All continuous variables are presented as the median and interquartile range (IQR), and all categorical variables are presented as the number and percentage. To investigate the effect of thyroid hormone marker on prevalence of SLD and modified MASLD, binary logistic regression models were generated in baseline cohort. The odds ratio (OR) with 95% confidence interval (CI) were reported. To investigate the effect of thyroid hormone marker on incidence of SLD or modified MASLD, Kaplan–Meier curves by the log-rank test and Cox proportional hazards regression models were generated in SLD risk cohort and modified MASLD risk cohort, respectively. To figure out the relationship between the thyroid hormone markers and resolution of SLD or modified MASLD, Kaplan–Meier curves by the log-rank test and Cox proportional hazards regression models were also used in SLD resolution cohort and modified MASLD resolution cohort, respectively. Age, gender, BMI, HbA1c, TG, HDL-C, LDL-C, AST, and ALT were adjusted in multivariable analysis. The hazard ratio (HR) with 95% CI were reported. IBM SPSS Statistics for Windows version 22.0 (IBM Co., Armonk, NY, USA) was used for the statistical analysis and P values of <0.05 with two-sided were considered statistically significant.
RESULTS
Baseline characteristics
The baseline characteristics of the baseline cohort, SLD risk cohort, and SLD resolution cohort are presented in Table 1. Of the 1,534 participants in baseline cohort, women comprised 34.9%, median age was 43.0 years (IQR, 37.0 to 50.0), and median BMI was 23.7 kg/m2 (IQR, 21.8 to 25.8). The median serum TSH level, FT4 level, and TT3 levels were 1.71 mU/L (IQR, 1.19 to 2.38), 1.32 ng/dL (IQR, 1.21 to 1.43), and 108.60 ng/dL (IQR, 97.71 to 119.00), respectively. In SLD risk cohort (n=1,216), median age at baseline was 43.0 years (IQR, 37.0 to 50.0), and there were 432 women (35.5%). The median BMI was 23.6 kg/m2 (IQR, 21.6 to 25.7), and the median serum TSH level, FT4 level, and TT3 levels were 1.74 mU/L (IQR, 1.18 to 2.39), 1.32 ng/dL (IQR, 1.21 to 1.43), and 107.80 ng/dL (IQR, 97.23 to 118.10), respectively. In SLD resolution cohort (n=318), median age at baseline was 45.0 years (IQR, 38.0 to 50.0), and 32.7% were women. The median BMI was 24.2 kg/m2 (IQR, 22.3 to 26.2), and the median serum TSH level, FT4 level, and TT3 levels were 1.66 mU/L (IQR, 1.20 to 2.34), 1.30 ng/dL (IQR, 1.21 to 1.44), and 111.45 ng/dL (IQR, 99.55 to 122.17), respectively.
Association between baseline TT3 and prevalence of SLD
In baseline cohort, 20.7% (318/1,534) of the participants had SLD. Binary logistic regression analysis showed that high TT3 level was independently associated with SLD (adjusted OR, 1.01; 95% CI, 1.00 to 1.02; P=0.002) even after adjusting other factors. However, TSH level (adjusted OR, 1.04; 95% CI, 0.90 to 1.22; P=0.542) or FT4 level (adjusted OR, 1.18; 95% CI, 0.52 to 2.64; P=0.683) had no association with prevalent SLD. High BMI (adjusted OR, 1.05; 95% CI, 1.00 to 1.10; P=0.039) and high HbA1c (adjusted OR, 1.94; 95% CI, 1.66 to 2.27; P<0.001) were also independent factors associated with the prevalence of SLD (Table 2).
Association between mean TT3 and the development of SLD
In SLD risk cohort, 66.3% (807/1,216) of the participants developed SLD with median time to onset as 24.0 months during the follow-up period of a median of 63.0 months (IQR, 51.2 to 71.0). Among the three thyroid hormone makers, mean TT3 was positively associated with occurrence of SLD in the multivariable analysis (adjusted HR, 1.01; 95% CI, 1.00 to 1.02; P=0.002; model 4) (Table 3). There was no significant difference in baseline characteristics (age, sex, BMI, AST/ALT level, HbA1c level, lipid profile, TSH level, and FT4 level) except TT3 level between high and low mean TT3 groups (Supplemental Table S1). Participants with high mean TT3 had a significantly higher risk to develop SLD than those of low mean TT3 (adjusted HR, 1.20; 95% CI, 1.05 to 1.38; P=0.008; model 4). The statistical significance was absent for mean TSH (adjusted HR, 0.94; 95% CI, 0.87 to 1.01; P=0.132; model 4) and for mean FT4 (adjusted HR, 1.01; 95% CI, 0.45 to 2.26; P=0.976; model 4) (Table 3). Other factors associated with SLD occurrence were old age (adjusted HR, 1.00; 95% CI, 1.00 to 1.01; P=0.034), high HbA1c (adjusted HR, 1.12; 95% CI, 1.03 to 1.23; P=0.009) and female gender (adjusted HR, 0.81; 95% CI, 0.69 to 0.95; P=0.010) (Supplemental Table S2).
Association between mean TT3 and the resolution of SLD
In SLD resolution cohort, SLD resolved in 79.5% (253/318) of the participants with median time to onset as 14.0 months during the follow-up period of a median of 60.0 months (IQR, 48.0 to 70.0). Among the three thyroid hormone makers, mean TT3 was inversely associated with resolution of SLD in the multivariable analysis (adjusted HR, 0.97; 95% CI, 0.96 to 0.99; P=0.005; model 4). Participants with high mean TT3 had a significantly lower chance to resolve SLD than those of low mean TT3 (adjusted HR, 0.66; 95% CI, 0.51 to 0.85; P=0.001; model 4) (Table 4, Supplemental Table S2). The statistical significance was absent for mean TSH (adjusted HR, 1.07; 95% CI, 0.90 to 1.28; P=0.390; model 4) and for mean FT4 (adjusted HR, 0.59; 95% CI, 0.14 to 2.35; P=0.457; model 4) (Table 4).
Sensitivity analysis using the modified MASLD
The baseline characteristics of the baseline cohort, modified MASLD risk cohort, and modified MASLD resolution cohort are shown in Supplemental Table S3. In sensitivity analysis using the modified MASLD instead of simple SLD, similar associations between TT3 and modified MASLD were observed. In baseline cohort, 269 (17.5%) participants have prevalent modified MASLD. High TT3 level was also positively associated with modified MASLD (adjusted OR, 1.01; 95% CI, 1.00 to 1.02; P=0.001, not shown in table) in binary logistic regression analysis with adjusting other factors. In modified MASLD risk cohort, 631 out of 1,265 participants (49.9%) developed modified MASLD during the follow-up period of a median of 63.0 months (IQR, 52.0 to 71.0). In the multivariable analysis, the statistical significance was sustained for modified MASLD (adjusted HR, 1.01; 95% CI, 1.00 to 1.02; P=0.003 for mean TT3; model 4) (adjusted HR, 1.29; 95% CI, 1.10 to 1.51; P=0.001 for high mean TT3; model 4). In modified MASLD resolution cohort, modified MASLD resolved in 210 out of 269 participants (78.1%). Multivariate Cox hazard models showed that mean TT3 is negatively associated with resolution of modified MASLD (adjusted HR, 0.96; 95% CI, 0.95 to 0.98; P<0.001; model 4). Compared with participants with low mean TT3, those in high mean TT3 were significantly inversely associated with the resolution of modified MASLD (adjusted HR, 0.71; 95% CI, 0.54 to 0.94; P=0.018; model 4) (Table 5, Supplemental Table S4).
Subgroup analysis according to baseline BMI
After dividing the participants in the SLD risk cohort into three subgroups according to baseline BMI (<23, 23–25, and ≥25 kg/m2), Kaplan–Meier curves for SLD risk according to the mean TT3 status in normal (log-rank P=0.966) (Fig. 2A), overweight (log-rank P=0.142) (Fig. 2B), and obese participants (log-rank P=0.001) (Fig. 2C) were generated, respectively. In the multivariate analysis, the adjusted HRs of high mean TT3 on incident SLD risk were 1.00 (95% CI, 0.80 to 1.25; P=0.970), 1.21 (95% CI, 0.91 to 1.60; P=0.176), and 1.51 (95% CI, 1.15 to 1.84; P=0.001), respectively. The effect of high mean TT3 level on incident SLD risk was evident in obese subgroups (Fig. 2, Supplemental Table S5). In this obese group, there was no significant difference in BMI between the low mean TT3 and high mean TT3 groups (26.4 [IQR, 25.7 to 27.7] vs. 26.8 [IQR, 25.7 to 28.0], P=0.302, not shown in table).
Subgroup analysis according to bodyweight change
When the participants in the SLD risk cohort divided into three subgroups according to bodyweight change during follow-up (bodyweight loss group, stable bodyweight group, bodyweight gain group), Kaplan–Meier curves for incident SLD according to the mean TT3 status showed that the participants with high mean TT3 were more likely to develop SLD than those with low mean TT3 in bodyweight loss group (log-rank P=0.018) (Fig. 3A) and stable bodyweight group (log-rank P=0.001) (Fig. 3B), whereas not in bodyweight gain group (log-rank P=0.073) (Fig. 3C). Similar results were shown in multivariable analysis as follows according to bodyweight change: bodyweight loss group- adjusted HR 1.39 (95% CI, 1.01 to 1.92; P=0.039), stable bodyweight group- adjusted HR 1.38 (95% CI, 1.14 to 1.67; P=0.001), and bodyweight gain group- adjusted HR 0.77 (95% CI, 0.58 to 1.03; P=0.080) (Supplemental Table S6).
DISCUSSION
In this large and long-term followed longitudinal cohort study, we firstly demonstrated that TT3 was associated with development and resolution of SLD. We also showed that TT3 was also related to development and resolution of modified MASLD. The association of high mean TT3 with development of SLD was most prominent in obese participants (BMI of ≥25 kg/m2), and the association was sustained in bodyweight stable or bodyweight loss group.
Several previous cross-sectional studies in euthyroid subjects showed a positive correlation between FT3 and prevalent NAFLD [9,10,16]. In line with the study, we also found that higher TT3 level rather than TSH or FT4 was associated with increased prevalence of SLD. However, studying relationship between TT3 and development of SLD was an unmet need thus far—this is the first longitudinal study using survival analysis to suggest that TT3 is associated with development of SLD. The participants with high TT3 level during follow-up were more likely to develop SLD comparing those with low TT3 level during follow-up (HR, 1.20). A previous study with longitudinal design showed that increased FT3 was indicated to be independently associated with increased hepatic steatosis [16]. However, the study did not apply survival analysis and average follow-up duration was short (2.2 years). About 2 years of follow-up duration might be insufficient considering that median onset duration for the SLD occurrence was over 2 years in current study. Comparing the previous study, we traced the participants over 5 years (median 61.0 months). Notably, we also showed that TT3 level was also associated with the resolution of SLD when maintained in the lower normal level. It might assume an even stronger association between TT3 level and SLD. There are a few studies reporting that T3 level is not associated with the SLD [17,18], and it could be explained by the fact that the difference of baseline demographics such as BMI or ethnicity.
A novel concept of MASLD is now replacing the NAFLD [2,3]. For sensitivity analysis, we assessed modified MASLD as an outcome replacing simple steatosis, and the result was similar to the main analysis. The results indicate that the previous association with the T3 and NAFLD can be postulated in the context of the MASLD. Also, the fact that the sensitivity analysis including modified MASLD was concordant with the main analysis supports the validity of our findings. Although lack of baseline data about HTN, WC, and alcohol consumption limited the definite diagnosis of MASLD, in our knowledge, this is the first study which evaluated the correlation between TT3 and development & resolution of SLD using the novel concept of MASLD.
The possible pathophysiology linking TT3 and SLD/MASLD are as follows. At first, T3 levels potentially influence the SLD/MASLD related with metabolic disease. Much data had shown the positive association of T3 levels with obesity, insulin resistance, or metabolic syndrome: euthyroid obese individuals exhibit higher FT3 levels [19], while FT3 levels decrease following weight loss [20]. A large study from the Netherlands provided that highest FT3 quartiles were related with over 50% increased risk of metabolic syndrome [21]. SLD/MASLD is considered as a liver manifestation of metabolic disease, and metabolic disease such as obesity or diabetes influences both SLD and T3 levels. In this context, the relationship between SLD/MASLD development/resolution and TT3 levels in current study is reasonable. Second, high TT3 could be a result of the compensatory mechanism. Obesity is proven cause of SLD, and obese patients tend to have an increased deiodinase activity which conversion of FT4 to FT3 [6,19] as a compensatory mechanism to prevent further fat accumulation. In line with the concept, the impact of high mean TT3 on development of SLD is prominent in obese participants in subgroup analysis of this study. To completely explain, nevertheless, it is insufficient because the impact of high mean TT3 on development of SLD remained despite steady or even declining body weight group when we performed a subgroup analysis according to bodyweight change during follow-up. Third, it might be the manifestation of resistance of hepatic T3 uptake caused by repeated fat accumulation. Excessive fat in hepatocytes modified level of deiodinase via interleukin 6, tumor necrosis factor α, and reduction of antioxidants, which is responsible for inactivating T3 [22,23]. In addition, serum T3 signals gene express through THR-β isoform [24-26], and obese patients showed decreased the THR in peripheral cells [27]. Recently, liver-directed selective THR-β agonist (resmetirom), which could improve the resistance of hepatic T3 uptake, showed superiority over placebo in nonalcoholic steatohepatitis resolution (25.9% [80 mg] vs. 29.9% [100 mg] vs. 9.7% [placebo]) or improvement in fibrosis (24.2% [80 mg] vs. 25.9% [100 mg] vs. 14.2% [placebo]) [24,28,29]. The hopeful result of the trial and evidence of hepatic T3 uptake resistance in current study offer promising drug target for SLD.
The current study has several strengths. First, we comprehensively demonstrated the association between TT3 and SLD through prevalence, incidence, or resolution. To our knowledge, this is the first study evaluating the relationship between TT3 and development and resolution of SLD using survival analysis in longitudinal data. Second, the sensitivity analysis adopting concept of MASLD not only confirms the reliability of our study but also shed light on future research of relationship between MASLD and thyroid hormone. Third, we proposed the possible mechanism (the concept of hepatic T3 resistance) that TT3 is positively associated with incidence of SLD through subgroup analysis and review of other articles. However, the present study also has limitations. First, we used USG to define SLD, which may not be objective. Despite this shortage, USG is the most widely accepted and recommended as first diagnostic modality for hepatic steatosis, and well-trained radiologist performed USG in current study. Also, there was only two pathological obese patients (BMI ≥35 kg/m2) who have possibility of prominent low sensitivity/specificity of USG to diagnosis SLD [30] in current study. The further study using more objective and quantifiable modality such as computed tomography or magnetic resonance imaging [31,32] is needed. Second, we could not use accurate definition of MASLD due to lack of HTN or WC data. Moreover, the data have no alcohol consumption information, so modified MASLD in current study might include alcohol-related liver disease (ALD) or MASLD and increased alcohol intake (MetALD) [3]. Therefore, a future study with accurate definition of MASLD is warranted. Third, while this study indicates an association between TT3 levels and the development/resolution of SLD in longitudinal design, it could be a phenomenon rather than establishing a causal relationship. We should keep in mind that SLD could elevate TT3 levels by increasing thyroxine-binding globulin (TBG) levels. At last, this was a retrospective study involving the participants of a single ethnicity. Therefore, the results should be verified in additional large, and prospective studies.
In conclusion, TT3 level in normal range was associated with development and resolution of either SLD or modified MASLD in Korean euthyroid adults. It could be possible evidence of hepatic T3 uptake resistance that related with future drug target for SLD.
XML Download