2024 年 31 巻 1 号 p. 36-47
Aims: Small dense (sd) low-density lipoprotein (LDL)-cholesterol (C) is the most powerful predictor of cardiovascular (CV) disease among lipid biomarkers and is generated by hypertriglyceridemia and insulin resistance. Metabolic dysfunction-associated fatty liver disease (MAFLD) is a newly proposed liver disease with a high CV risk. We investigated the specific association of sdLDL-C with MAFLD beyond triglycerides (TG) and obesity
Methods: Participants were 839 non-alcoholic drinkers with type 2 diabetes enrolled in a regional diabetes cohort. Fatty liver (FL) and visceral fat area (VFA) was detected by computed tomography scan. sdLDL-C and LDL-TG were measured by our established homogeneous assay. TG rich lipoprotein (TRL) was calculated by subtracting LDL-C plus HDL-C from total-C. Grade of sdLDL-C (≤ 24, 25–34, 35–44, and ≥ 45 mg/dL) was classified according to the Hisayama study.
Results: Compared to non-FL counterparts, FL subjects were younger, predominantly male and smokers; and had higher body mass index (BMI), VFA, hemoglobin A1c, C-peptide, TG, and sdLDL-C, while had similar levels of LDL-C, LDL-TG, and TRL-C. Multivariate logistic analysis revealed that sdLDL-C was the most powerful lipid parameter for identifying FL, independent of TG, HDL-C, BMI, and VFA. The independent association between TG and FL was lost when sdLDL-C was added to the analysis. These results remained the same when lipid-lowering drug users were excluded. After adjustment for confounders, the odds ratio for FL was 2.4–2.7 at sdLDL ≥ 35 mg/dL based on sdLDL ≤ 24 mg/dL.
Conclusions: sdLDL-C levels are specifically elevated in patients with diabetes and MAFLD, independent of TG and VFA, suggesting liver-centered metabolic abnormalities.
See editorial vol. 31: 17-18
Metabolic dysfunction-associated fatty liver disease (MAFLD), formerly named non-alcoholic fatty liver disease (NAFLD), affects about a quarter of the adult population in the world and poses health and economic burden to all countries1). NAFLD is a risk factor for steatohepatitis and liver cancer, as well as cardiovascular (CV) disease2, 3). A number of pathophysiological mechanisms were proposed for an intimate association between MAFLD and CV risk, such as systemic inflammation, insulin resistance, increased oxidative stress, platelet activation, endothelial dysfunction, and abnormal lipid metabolism2). Among them, hypertriglyceridemia seems to be an essential CV risk factor associated with MAFLD because triglycerides (TG) are an indivisible component for liver steatosis.
Atherogenicity of hypertriglyceridemia is mainly explained by increased TG-rich lipoprotein (TRL) remnants4), high prevalence of small dense low-density lipoprotein (sdLDL) particles5), and low high-density lipoprotein (HDL)-cholesterol (C). sdLDL particles are known to promote atherosclerosis more than large buoyant LDL particles5). We have evolved the measurement of sdLDL from a qualitative (gel electrophoresis) method to a quantitative method, a fully automated assay that directly measures sdLDL-C6). This assay system was adopted in well-known cohort studies such as in the community atherosclerosis risk study (ARIC)7), the Multi-ethnic Study of Atherosclerosis (MESA)8), and the Hisayama study9), and all these studies have consistently revealed that sdLDL-C is superior to LDL-C in predicting CV diseases.
Hypertriglyceridemia is both associated with fatty liver (FL) and preponderance of sdLDL particles10). Increased hepatic lipogenesis due to de novo or exogenous fatty acid (FA) fluxes is a common pathophysiological mechanism that promotes Very-low-density lipoprotein (VLDL) production and hepatic steatosis, which leads to develop hypertriglyceridemia and sdLDL generation10, 11). Although several studies have already reported smaller LDL size and increased sdLDL subspecies in individuals with FL or NAFLD12-14), few studies have measured sdLDL-C levels in MAFLD and examined their specificity for the presence of FL. We examined the association between sdLDL-C levels and prevalence of FL in patients with type 2 diabetes, in whom hepatic lipogenesis is generally stimulated11). The aim of the present study is to investigate whether the association between sdLDL-C and MAFLD is independent of TG and whether high sdLDL-C levels could be a specific lipid abnormality in MAFLD.
Subjects participated in the “ViNA” cohort study. The ViNA cohort study began on October 1, 2019, and carried out regular tests, assessment of diabetic complications, and prognostic surveys15). Exclusion criteria were patients with malignancies currently being treated; patients with severe liver, endocrine, and respiratory disorders; and patients undergoing hemodialysis. Thus, patients with active viral hepatitis, cirrhosis, or liver cancer were excluded. To limit the scope of this study to NAFLD, current drinkers, defined as those who drink even small amounts of alcohol, including non-regular drinkers, were excluded. All subjects (n=839) had type 2 diabetes and had been receiving treatment for diabetes at Ebina General Hospital for at least 3 years. Subject characteristics were listed in Table 1. One-fourth of patients were insulin users, and all subjects were treated with the following hypoglycemic agents alone or in combination: a sulfonylurea (27%), metformin (47%), pioglitazone (5%), dipeptidase peptidase-4 inhibitors (64%), sodium-glucose cotransporter (SGLT)-2 inhibitors (26%), or glucagon-like peptide-1 receptor agonists (GLP-1RA) (3%). Subjects with hyperlipidemia were treated with statins (60%), ezetimibe (8%)), fibrates (4%), or omega-3FA (4%) alone or in combination. Because lipid-lowering agents strongly affect plasma lipid concentrations, the correlation between lipids and FL may be distorted by these treatments. Therefore, we excluded all lipid-lowering drug users and analyzed for that subset, as shown in Supplementary Tables 1 and 2. All patients were taught an appropriate diet proposed by the Japan Diabetes Foundation by a dietitian.
| Total subjects | Men | Women | p | Non-Fatty liver | Fatty liver | p | |
|---|---|---|---|---|---|---|---|
| N (male/female) | 839 | 415 | 424 | 554 (255/299) | 285 (160/125) | ||
| Male % | 49% | 100% | 0% | 46% | 56% | ＜0.001 | |
| Fatty Liver % | 34% | 38% | 29% | ＜0.01 | 0% | 100% | |
| Smoker % | 15% | 25% | 5% | ＜0.0001 | 13% | 19% | ＜0.05 |
| Age years | 67 (12) | 66 (12) | 69 (11) | ＜0.0001 | 70 (10) | 62 (12) | ＜0.0001 |
| BMI kg/m2 | 25.3 (4.3) | 25.3 (4.0) | 25.4 (4.5) | ns | 24.3 (3.8) | 27.4 (4.4) | ＜0.0001 |
| VFA cm2 | 152 (78) | 163 (86) | 140 (66) | ＜0.0001 | 128 (66) | 198 (78) | ＜0.0001 |
| SFA cm2 | 167 (82) | 148 (81) | 185 (79) | ＜0.0001 | 148 (71) | 203 (90) | ＜0.0001 |
| SBP mmHg | 132 (15) | 130 (14) | 133 (16) | ＜0.005 | 132 (15) | 132 (15) | ns |
| DBP mmHg | 77 (12) | 78 (12) | 75 (12) | ＜0.0001 | 75 (11) | 80 (12) | ＜0.0001 |
| HbA1c % | 7.4 (0.9) | 7.4 (1.0) | 7.4 (0.8) | ns | 7.3 (0.9) | 7.6 (0.9) | ＜0.001 |
| Glucose mg/dL | 149 (40) | 153 (42) | 145 (38) | ＜0.005 | 149 (43) | 149 (33) | ns |
| Metabolic syndrome % | 61% | 64% | 57% | ＜0.05 | 53% | 75% | ＜0.0001 |
| C-Peptide ng/mL | 1.4 [0.9-2.1] | 1.5 [0.9-2.3] | 1.3 [0.8-1.9] | ＜0.05 | 1.2 [0.8-1.9] | 1.7 [1.3-2.5] | ＜0.0001 |
| hsCRP mg/L | 0.61 [0.27-1.40] | 0.58 [0.26-1.30] | 0.67 [0.29-1.45] | ns | 0.47 [0.23-1.02] | 0.97 [0.51-2.23] | ＜0.0001 |
| eGFR mL/min/1.73m2 | 70 (21) | 69 (21) | 71 (21) | ns | 68 (21) | 74 (21) | ＜0.0001 |
| ALT IU/L | 20 [15-29] | 22 [16-34] | 19 [14-26] | ＜0.0001 | 18 [14-24] | 27 [20-43] | ＜0.0001 |
| γGTP IU/L | 22 [16-35] | 25 [18-39] | 19 [15-30] | ＜0.0001 | 19 [15-29] | 30 [21-52] | ＜0.0001 |
| TG mg/dL | 106 [78-153] | 110 [79-165] | 103 [77-142] | ＜0.05 | 98 [72-133] | 127 [93-181] | ＜0.0001 |
| Total-C mg/dL | 181 (32) | 176 (30) | 187 (33) | ＜0.0001 | 182 (33) | 180 (30) | ns |
| LDL-C mg/dL | 102 (25) | 101 (25) | 103 (25) | ns | 101 (26) | 104 (24) | ns |
| HDL-C mg/dL | 54 (14) | 50 (12) | 58 (14) | ＜0.0001 | 56 (14) | 50 (11) | ＜0.0001 |
| nonHDL-C mg/dL | 127 (29) | 126 (28) | 129 (30) | ns | 126 (29) | 130 (28) | ＜0.05 |
| TRL-C mg/dL | 23 [18-30] | 22 [17-30] | 23 [18-30] | ns | 23 [18-29] | 23 [17-32] | ns |
| sdLDL-C mg/dL | 27 [21-37] | 28 [21-37] | 26 [20-36] | ns | 25 [19-33] | 32 [24-44] | ＜0.0001 |
| LDL-TG mg/dL | 15.4 (5.2) | 14.4 (4.5) | 16.4 (5.8) | ＜0.0001 | 15.1 (4.7) | 16.0 (6.1) | ＜0.05 |
| ApoB mg/dL | 87 (19) | 86 (18) | 88 (19) | ns | 85 (18) | 90 (19) | ＜0.0005 |
| ApoCIII mg/dL | 10.1 [8.2-12.3] | 9.7 [7.8-11.9] | 10.4 [8.7-12.7] | ＜0.0005 | 9.8 [8.1-11.9] | 10.8 [8.6-13.7] | ＜0.0001 |
| ApoE mg/dL | 4.1 (1.2) | 3.8 (1.1) | 4.4 (1.3) | ＜0.0001 | 4.1 (1.2) | 4.2 (1.2) | ns |
| sdLDL-C/LDL-C | 0.27 [0.22-0.34] | 0.27 [0.22-0.36] | 0.26 [0.22-0.33] | ns | 0.25 [0.21-0.31] | 0.32 [0.25-0.40] | ＜0.0001 |
| LDL-C/ApoB | 1.17 (0.14) | 1.18 (0.14) | 1.17 (0.13) | ns | 1.18 (0.13) | 1.15 (0.14) | ＜0.01 |
| LDL-TG/LDL-C | 0.16 (0.06) | 0.15 (0.05) | 0.17 (0.06) | ＜0.0001 | 0.16 (0.05) | 0.16 (0,07) | ns |
| Statins | 60% | 53% | 67% | ＜0.0001 | 58% | 62% | ns |
| Ezetimibe | 8% | 7% | 9% | ns | 7% | 9% | ns |
| Fibrates | 4% | 5% | 4% | ns | 3% | 8% | ＜0.001 |
| Omega3-FA | 5% | 6% | 4% | ns | 4% | 5% | ns |
| Sulfonylurea | 27% | 27% | 27% | ns | 27% | 27% | ns |
| SGLT2 inhibitors | 27% | 26% | 27% | ns | 19% | 41% | ＜0.0001 |
| Metformin | 47% | 45% | 49% | ns | 40% | 61% | ＜0.0001 |
| GLP-1RA | 4% | 2% | 5% | ＜0.05 | 3% | 6% | ＜0.05 |
| DPP-4 inhibitors | 65% | 65% | 64% | ns | 65% | 64% | ns |
| Pioglitazone | 5% | 7% | 4% | ns | 6% | 5% | ns |
| Insulin | 25% | 25% | 26% | ns | 28% | 20% | ＜0.01 |
Fatty liver is diagnosed by CT scan. Metabolic syndrome is defined as visceral fat area (VFA) ≥ 100 cm2 with either high blood pressure (SBP ≥ 130mmHg or DBP ≥ 85mmHg or use of anti-hypertensive drugs) or dyslipidemia (HDL-C ＜40mg/dL or TG ≥ 150mg/dL). Continuous variables are expressed as mean±standard deviation (SD) or median [interquartile range (IQR)]. Differences between groups are examined with either Student’s t-test or Mann Whitney’s test. Differences between categorical variables are examined with chi-square test. ns=not significant, VFA=visceral fat area, SFA=subcutanous fat area, SBP=systolic blood pressure, DBP=diastolic blood pressure, ns=not significant, Apo=apolipoprotein, C=cholesterol, TG=triglycerides
| Non-Fatty liver | Fatty liver | p | |
|---|---|---|---|
| N (male/female) | 198 (113/85) | 84 (57/27) | |
| Male % | 57% | 67% | ns |
| Smoker % | 15% | 16% | ns |
| Age years | 70 (11) | 61 (13) | ＜0.0001 |
| BMI kg/m2 | 23.2 (3.8) | 27.4 (4.4) | ＜0.0001 |
| VFA cm2 | 109 (62) | 209 (72) | ＜0.0001 |
| SFA cm2 | 128 (71) | 197 (83) | ＜0.0001 |
| SBP mmHg | 131 (15) | 132 (17) | ns |
| DBP mmHg | 75 (11) | 81 (13) | ＜0.0001 |
| HbA1c % | 7.3 (0.9) | 7.3 (0.8) | ns |
| Glucose mg/dL | 153 (49) | 145 (33) | ns |
| Metabolic syndrome % | 47% | 78% | ＜0.0001 |
| C-Peptide ng/mL | 1.0 [0.6-1.5] | 1.6 [1.2-2.4] | ＜0.0001 |
| hsCRP mg/L | 0.46 [0.22-1.04] | 1.0 [0.52-2.23] | ＜0.0001 |
| eGFR | 70 (20) | 77 (23) | ＜0.0001 |
| ALT IU | 21 [17-24] | 30 [21-40] | ＜0.0001 |
| GTP IU | 18 [15-26] | 32 [22-49] | ＜0.0001 |
| TG mg/dL | 93 [65-131] | 129 [80-187] | ＜0.0001 |
| Total-C mg/dL | 193 (32) | 184 (28) | ＜0.05 |
| LDL-C mg/dL | 109 (26) | 109 (23) | ns |
| HDL-C mg/dL | 58 (15) | 50 (11) | ＜0.0001 |
| TRL-C mg/dL | 23 [18-30] | 22 [15-33] | ns |
| sdLDL-C mg/dL | 24 [19-32] | 32 [23-46] | ＜0.0001 |
| LDL-TG mg/dL | 15.4 (5.2) | 15.2 (4.2) | ns |
| ApoB mg/dL | 88 (18) | 91 (18) | ns |
| ApoCIII mg/dL | 9.3 [7.7-11.6] | 10.4 [8.3-12.9] | ＜0.0001 |
| ApoE mg/dL | 4.3 (1.2) | 4.2 (1.2) | ns |
| sdLDL-C/LDL-C | 0.22 [0.2-0.3] | 0.29 [0.23-0.39] | ＜0.0001 |
| LDL-C/ApoB | 1.23 (0.11) | 1.20 (0.12) | ＜0.05 |
| LDL-TG/LDL-C | 0.14 (0.04) | 0.14 (0.05) | ns |
| Sulfonylurea | 23% | 22% | ns |
| SGLT2 inhibitors | 13% | 40% | ＜0.0001 |
| Metformin | 36% | 48% | ns |
| GLP-1RA | 2% | 3% | ns |
| DPP-4 inhibitors | 58% | 64% | ns |
| Pioglitazone | 3% | 1% | ns |
| Insulin | 27% | 17% | ns |
Fatty liver is diagnosed by CT scan . Continuous variables were expressed as mean±standard deviation (SD) or median [interquartile range (IQR)]. Differences between groups were examined with either Student’s t-test or Mann Whitney’s test. Differences between categorical variables were examined with the chi-square test.
The p-trend was estimated by Cochran-Armitage trend test for categorical variables or Jonckheere-Terpstra trend test for continuous variables. ns=not significant, VFA=visceral fat area, SFA=subcutanous fat area, SBP=systolic blood pressure, DBP=diastolic blood pressure, ns=not significant, Apo=apolipoprotein, C=cholesterol, TG=triglycerides
| Ln [sdLDL-C] per 1SD | |||
| Odds | 95% CI | p | |
| Age and gender-adjusted | 1.78 | 1.32-2.39 | ＜0.0001 |
| Model 1 | 1.62 | 1.16-2.28 | ＜0.005 |
| Model 1+BMI | 1.71 | 1.19-2.46 | ＜0.005 |
| Model 1+BMI+VFA | 1.70 | 1.09-2.65 | ＜0.02 |
| Model 1+BMI+VFA+Ln [TG] | 1.95 | 1.14-3.34 | ＜0.02 |
| Model 1+BMI+VFA+Ln [TG]+HDL-C | 1.95 | 1.14-3.33 | ＜0.02 |
| Model 1+BMI+VFA+Ln [TG]+HDL-C+apoB | 3.95 | 2.03-7.69 | ＜0.0001 |
| Ln [TG] per 1SD | |||
| Odds | 95% CI | p | |
| Age and gender-adjusted | 1.67 | 1.26-2.21 | ＜0.0005 |
| Age, gender, and Ln [sdLDL-C] | 1.33 | 0.95-1.87 | ns |
| Age, gender, and VFA | 1.22 | 0.83-1.79 | ns |
| Model 1 | 1.25 | 0.86-1.83 | ns |
Model 1: Adjusted for age, gender, DBP, Ln [C-peptide], eGFR, Ln [hsCRP], and use of insulin and SGLT-2 inhibitors
Fatty liver (FL) was diagnosed by the radiologist’s overall judgment including non-detectability of intrahepatic bile ducts and high hepatosplenic contrast on computed tomography (CT) scan images. Severe liver disease was excluded based on the radiologist comments. The diagnosis of MAFLD was based on histological, imaging, and blood biomarker evidence of fat accumulation in the liver (hepatic steatosis), plus one of the following three criteria: overweight/obesity, presence of type 2 diabetes, or evidence of metabolic abnormalities1). Therefore, the present subjects (type 2 diabetes) were diagnosed as MAFLD if FL was detected by CT scan (imaging). In this study, MAFLD is synonymous with NAFLD because alcohol drinkers were excluded.
MeasurementsPlasma samples were taken in the morning after overnight fasting. sdLDL-C and LDL-TG concentrations were measured directly in plasma by the homogeneous method established by our group (Denka Co. Tokyo)6, 16). Non-HDL-C was calculated by subtracting HDL-C from Total-C, and TG rich lipoprotein (TRL) was calculated by subtracting LDL-C plus HDL-C from total-C. The visceral fat area (VFA) and subcutaneous fat area (SFA) were measured by the fat scan program for CT scan (Fujifilm, Tokyo). Apolipoproteins (apo) B, CIII, and E, high sensitive (hs)C-reactive protein (CRP), and C-peptide were measured by commercially available test kits. Grade of sdLDL-C was classified according to the Hisayama study indicating CV risk stratified by sdLDL-C quartile9). sdLDL1 ≤ 24, sdLDL2=25–34, sdLDL3=35–44, and sdLDL4 ≥ 45 mg/dl corresponded to the quartile of sdLDL-C levels (Q1, 2, 3, and 4) in the Hisayama study. Metabolic syndrome in diabetic subjects was defined as VFA ≥ 100 cm2 with either high blood pressure (systolic blood pressure [SBP] ≥ 130 mmHg or diastolic blood pressure [DBP] ≥ 85 mmHg or use of anti-hypertensive drugs) or dyslipidemia (HDL-C ＜40 mg/dL or TG ≥ 50 mg/dL)17).
The study was explained in detail to all subjects who consented to participate, and a written informed consent form was obtained from all participants prior to the study. This study was approved by the Ethics Committee of Ebina General Hospital.
StatisticsAll continuous variables were expressed as mean±standard deviation (SD) or median (interquartile range [IQR]). Differences between groups were examined with either Student’s t-test or Mann Whitney’s test. Significance was evaluated by Wilcoxon signed-rank test for the continuous variables and chi-square test for categorical variables. For non-normally distributed variables, logarithmic transformation (Ln) was performed before analyses of difference or regression. Relationships between the presence of FL and continuous variables were evaluated with univariate logistic analysis and expressed as chi-square values. Multivariate logistic analysis was employed to analyze the independent relationship between sdLDL-C values or sdLDL grade (1, 2, 3, and 4) and the presence of FL after adjustment for other lipids or obesity-related factors. The results were expressed as odds ratios and 95% confidence intervals (CI). The odds ratio was calculated using each explanatory variable divided by SD because the absolute value of each measurement was substantially different. Non-HDL-C and TRL-C were excluded from the multivariate analysis because they were calculated values. The p trend was estimated by Cochran-Armitage trend test for categorical variables or Jonckheere-Terpstra trend test for continuous variables. P-value less than 0.05 was considered statistically significant. Other all analyses were performed using JMP software version 15 (SAS Institute, Cary, NC, USA).
Table 1 lists clinical characteristics and measurements in total subjects, gender difference, and subgroups with or without FL. Men had a higher prevalence of FL, smokers, and metabolic syndrome than women. Men had greater VFA and smaller SFA than women. Liver biomarkers were higher in men than in women. HDL-C was higher in women and TG was higher in men, but there were no significant gender differences in TRL-C, LDL-C, and sdLDL-C. Compared to the non-FL group, the FL group was younger (62 vs. 69 years old), predominately male (56% vs. 45%) and higher prevalence of smokers (18% vs. 12%). The FL group had higher body mass index (BMI), VFA and SFA, and increased ALT, γGTP, and hsCRP levels. SBP was similar but DBP was higher in the FL group. HbA1c, C-peptide, and estimated glomerular filtration rate (eGFR) were higher in the FL group. In terms of plasma lipid parameters, the FL group had higher TG, sdLDL-C, non-HDL-C, and apo B and CIII levels, while LDL-C, LDL-TG, TRL-C, and apoE levels were similar compared to the non-FL group. The sdLDL-C/LDL-C ratio reflecting LDL size was higher, whereas the LDL-C/apoB ratio reflecting LDL buoyancy was lower in the FL group than in non-FL group. The prevalence of metabolic syndrome, defined as the presence of hypertension and/or dyslipidemia based on a large VFA, was significantly higher in the FL group than in the non-FL group (75% vs. 53%). Sixty percent of subjects were treated with statins, resulting in a comparable medication use between the two groups. Fewer subjects used fibrates, with a higher medication use in the FL group (7% vs. 4%). Compared to the non-FL group, more patients in the FL group used metformin, SGLT-2 inhibitors, and GLP-1RA; and fewer used insulin.
Supplementary Table 1 shows the clinical characteristics and measurements of the 282 lipid-lowering drug-free subjects. As with all subjects in Table 1, the FL group (n=84) was younger, had higher BMI, VFA, and SFA; increased ALT, γGTP, C-peptide, and hsCRP levels; and higher DBP and eGFR than the non-FL group (n=198). The FL group had higher TG, sdLDL-C, non-HDL-C, and apo B and CIII levels, while LDL-C, LDL-TG, TRL-C, and apoE levels were similar compared to the non-FL group. The prevalence of metabolic syndrome was significantly higher in the FL group than in the non-FL group (78% vs. 47%).
Table 2 shows the parameters in Table 1 that showed significant differences between the FL and non-FL groups and their association with the presence or absence of FL as determined by logistic analysis. Age, BMI, VFA, and Ln[ALT] were strongly associated with the presence or absence of FL (chi-square value ＞100); Ln[γGTP], Ln[TG], Ln[sdLDL-C], Ln[sdLDL-C/LDL-C], Ln[hsCRP], and Ln[C-peptide] showed moderate association with FL (chi-square value ＞50). For lipid parameters, Ln[sdLDL-C] and Ln[sdLDL-C/LDL-C] showed the strongest association with FL.
| χ2 | p | |
|---|---|---|
| Male | 3.8 | ＜0.05 |
| Age | 104.5 | ＜0.0001 |
| BMI | 105.6 | ＜0.0001 |
| VFA | 147.9 | ＜0.0001 |
| SFA | 78.6 | ＜0.0001 |
| DBP | 46.3 | ＜0.0001 |
| HbA1c | 11.1 | ＜0.001 |
| Ln [C-Peptide] | 51.2 | ＜0.0001 |
| Ln [hsCRP] | 50.3 | ＜0.0001 |
| eGFR | 20.1 | ＜0.0001 |
| Ln [ALT] | 138.4 | ＜0.0001 |
| Ln [γGTP] | 79.8 | ＜0.0001 |
| Ln [TG] | 60.6 | ＜0.0001 |
| HDL-C | 44 | ＜0.0001 |
| NonHDL-C | 4 | ＜0.05 |
| Ln [sdLDL-C] | 63.1 | ＜0.0001 |
| Ln [sdLDL-C/LDL-C] | 66.3 | ＜0.0001 |
| ApoB | 12.5 | ＜0.0005 |
| Ln [ApoCIII] | 18.9 | ＜0.0001 |
| LDL-C/ApoB | 6.8 | ＜0.001 |
Ln=Natural logarithmically transformed. VFA=visceral fat area, DBP=diastolic blood pressure, Apo=apolipoprotein, C=cholesterol, TG=triglycerides
Table 3 shows odds ratios and 95% CIs in the multivariate logistic analysis for the presence or absence of FL with sdLDL-C and TG as explanatory variables. To obtain the odds ratio corresponding to the change for each SD, Ln[sdLDL-C] or Ln[TG] was divided by its respective SD. Model 1 consisted of age, sex, smoking, DBP, HbA1c, Ln[C-Peptide], Ln[hsCRP], eGFR, fibrate, insulin, SGLT-2 inhibitors, and GLP-1 RA usage, which were significantly different between the FL and non-FL groups in Table 1. The odds ratio for Ln[sdLDL-C]/SD was shown to be 1.7 [1.4–2.0] adjusted for age and sex; and 1.5 [1.3–1.9] adjusted for Model 1. This odds ratio did not change when BMI, VFA, HDL-C, and Ln[TG] were added to Model 1. The odds ratio for Ln[TG]/SD was 1.7 [1.4–2.0], similar to [sdLDL-C]/SD. After adjustment for BMI, VFA, and HDL-C in Model 1, the odds ratio decreased to 1.3, but remained significant. However, when Ln[sdLDL-C] was included in the multivariate analysis, Ln [TG] lost its association with FL.
| Ln [sdLDL-C] per 1SD | Ln [TG] per 1SD | |||||
|---|---|---|---|---|---|---|
| Odds | 95% CI | p | Odds | 95% CI | p | |
| Age and gender-adjusted | 1.70 | 1.44 - 2.01 | ＜0.0001 | 1.69 | 1.44 - 1.99 | ＜0.0001 |
| Model 1 | 1.55 | 1.30 - 1.86 | ＜0.0001 | 1.39 | 1.14 - 1.69 | ＜0.002 |
| Model 1+BMI | 1.55 | 1.29 - 1.86 | ＜0.0001 | 1.40 | 1.15 - 1.71 | ＜0.001 |
| Model 1+BMI+VFA | 1.52 | 1.25 - 1.85 | ＜0.0001 | 1.31 | 1.06 - 1.62 | ＜0.02 |
| Model 1+BMI+VFA+HDL-C | 1.50 | 1.23 - 1.83 | ＜0.0001 | 1.27 | 1.01 - 1.59 | ＜0.05 |
| Model 1+BMI+VFA+HDL-C+apoB | 2.13 | 1.55 - 2.93 | ＜0.0001 | 1.22 | 0.95 - 1.56 | ns |
| Model 1+Ln [TG] | 1.49 | 1.20 - 1.86 | ＜0.0005 | - | - | - |
| Model 1+Ln [TG]+BMI+VFA | 1.52 | 1.19 - 1.93 | ＜0.001 | - | - | - |
| Model 1+Ln [TG]+BMI+VFA+HDL-C | 1.54 | 1,21 - 1.97 | ＜0.001 | - | - | - |
| Model 1+Ln [TG]+BMI+VFA+HDL-C+apoB | 2.25 | 1.57 - 3.21 | ＜0.0001 | - | - | - |
| Model 1+Ln [sdLDL-C] | - | - | - | 1.08 | 0.85 - 1.37 | ns |
| Model 1+Ln [sdLDL-C]+BMI+VFA | - | - | - | 1.00 | 0.77 - 1.30 | ns |
| Model 1+Ln [sdLDL-C]+BMI+VFA+HDL-C | - | - | - | 0.95 | 0.71 - 1.25 | ns |
| Model 1+Ln [sdLDL-C]+BMI+VFA+HDL-C+apoB | - | - | - | 0.91 | 0.69 - 1.20 | ns |
Model 1: Adjusted for age, gender, smoking, DBP, HbA1c, Ln [C-peptide], eGFR, Ln [hsCRP], and use of fibrates, insulin, metformin, SGLT2 inhibitors, and GLP-1RA
Ln=Natural logarithmically transformed. sdLDL-C=small dense LDL-cholesterol, TG=triglycerides, VFA=visceral fat area, BMI=body mass index
Supplementary Table 2 shows odds ratios and 95% CIs in the multivariate logistic analysis for the presence or absence of FL with sdLDL-C and TG as explanatory variables in lipid-lowering drug-free subjects. The odds ratio of Ln[sdLDL-C] to FL is about 1.7, and this ratio becomes even larger when Ln [TG], BMI, VFA, and HDL-C are included as variables. On the other hand, Ln [TG] lost its association with FL when adjusted for Ln[sdLDL-C], VFA, and Model 1.
Table 4 shows the clinical characteristics and measurements stratified by sdLDL-C grade. The prevalence of FL was higher with higher sdLDL grades, 21, 35, 46, and 57% for sdLDL1, sdLDL2, sdLDL3, and sdLDL4, respectively. Younger age, BMI, VFA, SFA, DBP, HbA1c, glucose, and C-peptide were associated with higher sdLDL grade. Similarly, the liver markers ALT and GTP were increased with higher sdLDL grade; all lipid parameters, except for HDL-C, were significantly increased with sdLDL grade, while HDL-C was negatively correlated with sdLDL grade. Use of ezetimibe, fibrates, omega-3FA, metformin, and SGLT-2 inhibitors were associated with sdLDL grade.
| sdLDL1 | sdLDL2 | sdLDL3 | sdLDL4 | p -trend | |
|---|---|---|---|---|---|
| sdLDL-C range mg/dL | ≤ 24 | 25-34 | 35-44 | ≥ 45 | |
| sdLDL-C mg/dL | 19 [16-22] | 28 [26-31] | 39 [37-42] | 52 [47-62] | ＜0.0001 |
| Number {%male} | 357 {46%} | 251 {52%} | 125 {47%} | 106 {52%} | ns |
| Age years | 68 (11) | 68 (11) | 66 (12) | 61 (12) | ＜0.0001 |
| Fatty liver % | 21% | 35% | 46% | 57% | ＜0.0001 |
| Smoker % | 14% | 12% | 16% | 19% | ns |
| Metabolic syndrome % | 51% | 61% | 74% | 76% | ＜0.0001 |
| BMI kg/m2 | 24.6 (4.2) | 25.4 (3.8) | 25.6 (4.1) | 27.2 (4.7) | ＜0.0001 |
| VFA cm2 | 135 (78) | 152 (73) | 161 (72) | 191 (75) | ＜0.0001 |
| SFA cm2 | 157 (79) | 165 (76) | 173 (83) | 194 (93) | ＜0.0001 |
| SBP mmHg | 131 (15) | 132 (14) | 133 (15) | 131 (14) | ns |
| DBP mmHg | 74 (10) | 76 (11) | 78 (11) | 79 (14) | ＜0.0005 |
| HbA1c % | 7.2 (0.8) | 7.3 (0.8) | 7.7 (1.0) | 7.5 (1.0) | ＜0.0001 |
| Glucose mg/dL | 145 (39) | 147 (36) | 156 (37) | 157 (49) | ＜0.005 |
| C-Peptide ng/mL | 1.15 [0.76-1.80] | 1.37 [0.92-1.96] | 1.62 [1.07-2.63] | 1.79 [1.14-2.82] | ＜0.0001 |
| hsCRP mg/L | 0.48 [0.24-1.27] | 0.58 [0.27-1.15] | 0.69 [0.34-1.52] | 0.97 [0.54-2.05] | ＜0.0001 |
| eGFR mL/min/1.73m2 | 69 (20) | 70 (20) | 69 (20) | 71 (23) | ns |
| ALT IU/L | 18 [14-25] | 20 [15-28] | 24 [17-38] | 24 [18-46] | ＜0.0001 |
| γGTP IU/L | 19 [14-28] | 22 [16-32] | 28 [19-47] | 34 [19-63] | ＜0.0001 |
| TG mg/dL | 84 [62-107] | 106 [84-150] | 141 [110-189] | 191 [143-288] | ＜0.0001 |
| Total-C mg/dL | 164 (26) | 184 (27) | 193 (28) | 214 (28) | ＜0.0001 |
| LDL-C mg/dL | 87 (19) | 106 (20) | 113 (22) | 126 (24) | ＜0.0001 |
| HDL-C mg/dL | 56 (13) | 54 (14) | 51 (13) | 49 (11) | ＜0.0001 |
| nonHDL-C mg/dL | 108 (20) | 130 (21) | 142 (20) | 165 (25) | ＜0.0001 |
| TRL-C mg/dL | 20 [16-24] | 24 [18-29] | 28 (11) | 38 (17) | ＜0.0001 |
| LDL-TG mg/dL | 13 [10-16] | 14 [12-17] | 15 [13-18] | 17 [15-20] | ＜0.0001 |
| ApoB mg/dL | 73 (12) | 89 (11) | 98 (11) | 113 (17) | ＜0.0001 |
| ApoCIII mg/dL | 8.6 [7.2-10.0] | 10.5 [8.8-12.0] | 11.9 [10.0-14.5] | 14.3 [12.0-17.4] | ＜0.0001 |
| ApoE mg/dL | 3.7 (1.0) | 4.0 (1.1) | 4.4 (1.2) | 5.1 (1.4) | ＜0.0001 |
| Statins | 60% | 61% | 60% | 51% | ns |
| Ezetimibe | 7% | 5% | 14% | 11% | ＜0.05 |
| Fibrates | 3% | 5% | 8% | 6% | ＜0.05 |
| Omega3-FA | 2% | 4% | 8% | 5% | ＜0.05 |
| Sulfonylurea | 27% | 27% | 21% | 31% | ns |
| SGLT2 inhibitors | 23% | 24% | 34% | 38% | ＜0.005 |
| Metformin | 43% | 47% | 51% | 53% | ＜0.05 |
| GLP-1RA | 4% | 2% | 6% | 4% | ns |
| DPP-4 inhibitors | 61% | 67% | 68% | 65% | ns |
| Pioglitazone | 6% | 5,5% | 5% | 4% | ns |
| Insulin | 26% | 23% | 25% | 23% | ns |
Table 5 shows the odds ratios and 95% CIs obtained by multivariate logistic analysis for the presence of FL as an explanatory variable for sdLDL grade. sdLDL2, sdLDL3, and sdLDL4 age-adjusted odds ratios for FL were 2.1, 2.9, and 3.7, respectively, with sdLDL1 as the reference. Model1 consisted of age, DBP, HbA1c, C-Peptide, fibrate, omega-3FA, ezetimibe, metformin, and use of SGLT-2 inhibitors. After adjustment for Model1, the odds ratios for sdLDL2, sdLDL3, and sdLDL4 to FL were 2.0, 2.4, and 2.7, respectively, based on sdLDL1. Similar odds ratio trends were observed when Ln[TG], HDL-C, BMI, and VFA were entered into Model 1 as explanatory variables.
|
sdLDL1 ＜24 mg/dL |
sdLDL2 25-34 mg/dL |
sdLDL3 35-44 mg/dL |
sdLDL4 ＞45 mg/dL |
|||||
|---|---|---|---|---|---|---|---|---|
| Odds | 95% CI | Odds | 95% CI | Odds | 95% CI | Odds | 95% CI | |
| Age -adjusted | 1.00 | (Reference) | 2.01 | 1.38 - 2.93 | 2.93 | 1.86 - 4.60 | 3.73 | 2.30 - 6.04 |
| Model 1 | 1.00 | (Reference) | 1.99 | 1.33 - 2.98 | 2.40 | 1.46 - 3.94 | 2.69 | 1.59 - 4.55 |
| Model 1+Ln[TG] | 1.00 | (Reference) | 1.90 | 1.25 - 2.89 | 2.23 | 1.31 - 3.78 | 2.36 | 1.28 - 4.37 |
| Model 1+Ln[TG]+HDL-C | 1.00 | (Reference) | 1.96 | 1.29 - 3.00 | 2.33 | 1.37 - 3.97 | 2.62 | 1.41 - 4.87 |
| Model 1+Ln[TG]+HDL-C+BMI | 1.00 | (Reference) | 1.91 | 1.25 - 2.92 | 2.39 | 1.40 - 4.08 | 2.47 | 1.31 - 4.63 |
| Model 1+Ln[TG]+HDL-C+BMI+VFA | 1.00 | (Reference) | 1.77 | 1.11 - 2.82 | 2.43 | 1.38 - 4.26 | 2.41 | 1.22 - 4.76 |
Model1: Adjusted for age, Ln[C-peptide], Ln[hsCRP], HbA1c, and DBP, and use of metformin, SGLT-2 inhibitors, fibrates, omega-3FA, and ezitimebe.
sdLDL-C levels were substantially elevated in patients with type 2 diabetes presenting FL. Alcohol drinkers were excluded from the study, so the present subjects with FL correspond to NAFLD. Hosoyamada et al 18), Sugino et al 19), and Kikkawa et al 20) have already reported significantly higher sdLDL-C levels in FL patients using our assay. However, these results cannot be referred to NAFLD due to the heterogeneity of the subjects, including alcohol drinkers. Many studies have shown that patients with metabolic syndrome have more sdLDL particles and higher sdLDL-C levels21-24). Metabolic syndrome is often accompanied with FL25), and therefore, increased sdLDL-C in FL might simply be due to concomitant metabolic syndrome. Indeed, VFA and BMI were strongly correlated with the presence of FL. Therefore, we carefully examined the specific association between sdLDL and FL, independent of metabolic syndrome. The results showed that sdLDL-C was associated with FL independently of BMI and VFA, suggesting the presence of liver-centered metabolic abnormalities, that increase sdLDL-C levels besides visceral fatty obesity.
A new finding of this study revealed that sdLDL-C is more closely related to FL than TG, a component inseparable from FL. This relationship became even clearer when those taking lipid-lowering drugs were excluded. The sdLDL-C value is determined by the number of LDL particles and the percentage of small LDL particles26). The former is regulated by production of apoB-containing lipoprotein particles in the liver and the latter by TG-cholesteryl ester exchange between TRL and LDL particles in the plasma circulation12). The finding that sdLDL and FL are more closely related than TG and FL may suggest that hepatic VLDL production has a stronger effect on sdLDL-C levels than TG-induced remodeling of LDL particles. TG-rich VLDL1 particles are preferentially produced by insulin resistance, leading to the formation of sdLDL particles27, 28). On the other hand, VLDL1 production is accelerated by increased lipogenesis, a major cause of hepatic steatosis3). These liver-centered metabolic abnormalities may explain the close association of sdLDL and FL via VLDL1 production. VLDL1 production is not necessarily correlated with plasma TG concentration, since plasma TG is also determined by concentrations of chylomicrons and VLDL2, and TG clearance with lipoprotein lipase (LPL). In addition, hepatic TG lipase (HTGL) may explain the dissociation of sdLDL and TG associated with FL; HTGL is increased in FL29) and it stimulates the formation of sdLDL particles12); unlike LPL, HTGL activity is poorly correlated with plasma TG concentration30). Campanella, et al reported that TRL-C is independently associated with the severity grade of NAFLD as determined by ultrasound echogram31). In our cross-sectional study, there was no significant difference in TRL-C between patients with and without FL, but our results do not rule out the possibility that TRL-C predicts NAFLD. A prospective and interventional approach, such as that reported by Campanella, is needed to clarify the causal relationship between TRL-C and liver disease outcomes.
The target values for plasma lipids in CVD prevention guidelines in Japan were set based on the Hisayama study9), which also reported the association between sdLDL-C quartile and CV events. The prevalence of coronary artery disease was more than five times higher in the upper quartile of sdLDL-C (Q4 ≥ 45 mg/dL) at baseline than in the lower quartile of sdLDL-C (Q1 ≤ 25 mg/dL), and the cutoff value (35 mg/dL) was two times higher9). Since normal/abnormal sdLDL-C values are officially undetermined and it is useful to imagine the CV risk of FL patients, we adopted the same sdLDL-C grading in this study as in the Hisayama study. The FL prevalence of sdLDL-C ＞35 mg/dL resulted in odds ratios of 2.4–2.7 based on sdLDL-C ≤ 25 mg/dL, even when adjusted for TG and VFA. Thus, sdLDL-C ＞35 mg/dL may be a double risk for FL and CV. DeFilippis32) reported in the MESA study that various sizes of LDL particles measured by nuclear magnetic resonance spectroscopy and NAFLD was associated with a higher number of LDL particles and a smaller LDL particle size. The MESA study also employed our sdLDL-C direct assay and found that sdLDL-C levels were sensitive biomarkers for predicting CVD events. The results of the MESA study support our findings that sdLDL-C is a common biomarker for FL and CV disease.
MAFLD was recently proposed as a new concept connecting metabolic dysregulation factors and FL1). This disease concept is intended to move away from NAFLD to a more aggressive focus on metabolic factors as the cause of liver damage. MAFLD is considered a CV risk2), as the metabolic abnormalities of atherosclerosis are also closely related to hepatic steatosis. The sdLDL-C level is strongly associated with hypertriglyceridemia, hepatic insulin resistance, and visceral obesity, all common in FL patients. Therefore, high sdLDL-C levels are a representative metabolic abnormality of MAFLD and may explain the high CV risk in this liver disease. Whether elevated sdLDL-C levels in MAFLD can predict CVD events more sensitively than other lipid abnormalities remains to be determined in future prospective studies.
sdLDL-C levels are specifically elevated in patients with diabetes and FL, independent of TG, BMI, and VFA, and thus may be a strong CV risk in MAFLD.
We thank Dr. Tsuyoshi Hirashima, Dr. Ema Aoki, and Dr. Natsuko Suzuki of the Diabetes Center, Ebina General Hospital, for conducting the ViNA cohort, and Mrs. Miyuki Tokudome of Ebina General Hospital for conducting the sample analysis of the ViNA cohort.
Tsutomu Hirano receives advisor fees from Denka Co, and lecture fee from Kowa Co. Noriyuki Satoh and Yasuki Ito are employee of Denka Co.
Funding: This study was partially supported by Denka Co., Ltd. The source of funding was not involved in designing, conducting surveys, analyzing, or interpreting the data.
This study was approved by the Ethics Committee of Ebina General Hospital
Informed Consent: The study was detailed to all subjects who consented to participate, and a written informed consent form was obtained from all participants prior to the study.
Approval date: 11-September -2019, no 115, 2019
