doi: 10.2169/internalmedicine.

4954-20
Intern Med Advance Publication
http://internmed.jp

【 ORIGINAL ARTICLE 】

The Association between the Triglyceride to High-density

Lipoprotein Cholesterol Ratio and Low-density Lipoprotein
Subclasses

Kengo Moriyama

Abstract:
Objectives The triglyceride (TG)/high-density lipoprotein cholesterol (HDL-C) ratio is related to insulin re-
sistance (IR). However, information about whether or not the TG/HDL-C ratio is associated with low-density
lipoprotein (LDL) subclasses in the Japanese population is limited.
Methods In total, 1068 Japanese subjects who underwent an annual health examination and who were not
taking medications were recruited. The association between the TG/HDL-C ratio and LDL subclasses was in-
vestigated using correlation, multiple regression, and receiver operating characteristic analyses.
Results A correlation analysis revealed that both malondialdehyde-modified low-density lipoprotein (MDA-
LDL) and small dense low-density lipoprotein cholesterol (sdLDL-C) were positively associated with the TG/
HDL-C ratio. Furthermore, a multiple linear regression analysis revealed that the TG/HDL-C ratio was posi-
tively associated with MDA-LDL and sdLDL-C in both men and women. The multiple logistic regression
analysis also revealed that the TG/HDL-C ratio was positively associated with the upper tertile of MDA-LDL
and sdLDL-C in men and women. The LDL-C levels increased with the increasing TG/HDL-C ratio. The
MDA-LDL and sdLDL-C are known to be positively associated with LDL-C. However, within the same
LDL-C range, both MDA-LDL and sdLDL-C levels increased with the TG/HDL-C ratio, except for MDA-
LDL levels in the LDL-C <112 mg/dL group in women. These results further supported the notion that the
TG/HDL-C ratio was positively associated with the MDA-LDL and sdLDL-C levels, especially in the higher
LDL-C range, in both men and women. The optimal cut-off points of the TG/HDL-C ratio for the upper ter-
tile of MDA-LDL and sdLDL-C were 1.85 and 2.03 in men and 0.88 and 1.30 in women, respectively.
Conclusion The TG/HDL-C ratio is positively associated with MDA-LDL and sdLDL-C in Japanese sub-
jects. The relationship was particularly notable in subjects with high LDL-C levels.

Key words: triglyceride to high-density lipoprotein cholesterol ratio, malondialdehyde-modified low-density

lipoprotein, small dense low-density lipoprotein cholesterol

(Intern Med Advance Publication)

(DOI: 10.2169/internalmedicine.4954-20)

significantly correlated with insulin resistance (IR) (1-3) and

Introduction may be a useful predictor for the development of diabe-
tes (4), coronary heart disease (CHD), and cardiovascular
A high level of low-density lipoprotein cholesterol (LDL- mortality (5). Furthermore, it is a better screening index for
C) plays a key role in the development and progression of IR than the homeostasis model assessment of insulin resis-
atherosclerosis and cardiovascular disease (CVD). Recently, tance (HOMA-IR) (6). However, few studies have investi-
the use of the triglyceride (TG)/high-density lipoprotein cho- gated this ratio as an indicator of LDL subclasses, such as
lesterol (HDL-C) ratio as a compliment to LDL-C for pre- malondialdehyde-modified low-density lipoprotein (MDA-
dicting CVD has been proposed. The TG/HDL-C ratio is LDL) and small dense low-density lipoprotein cholesterol

Department of Clinical Health Science, Tokai University School of Medicine, Japan

Received: March 26, 2020; Accepted: May 12, 2020; Advance Publication by J-STAGE: July 14, 2020
Correspondence to Dr. Kengo Moriyama, [email protected]

1
 Intern Med Advance Publication DOI: 10.2169/internalmedicine.4954-20

(sdLDL-C).
Measurements
LDL comprises heterogeneous subclasses of particles with
varying density, size, electrophoretic mobility, relative lipid- The waist circumference (WC) was measured at the level
protein proportions, and binding affinity (7). According to of the umbilicus during slight expiration while the partici-
size and density, LDL particles are fractionated into large, pant was in standing position. Blood pressure (BP) was
buoyant LDL (lbLDL; diameter !25.5 nm) and small, dense measured in the upper right arm using an automatic BP
LDL (sdLDL; diameter <25.5 nm) (8). Recently, a direct monitor device (TM-2655P; A&D, Tokyo, Japan) while the
method for the quantification of sdLDL-C using an auto- participant was seated. Blood samples were collected in
analyzer was established (9). sdLDL is directly correlated heparin-coated tubes early in the morning after an overnight
with serum TG and inversely with serum HDL-C (8). fast. The fasting plasma glucose (FPG) level was measured
High sdLDL-C levels are currently considered a risk fac- with the L-type Glu 2 kit using the hexokinase/glucose-6-
tor for CVD. Compared with large LDL, sdLDL-C exhibits phosphate dehydrogenase method (Wako Pure Chemicals,
a lower affinity to LDL receptor and longer half-life in the Osaka, Japan). The fasting immunoreactive insulin (FIRI)
plasma, binds to arterial proteoglycans more tightly, pene- levels were measured using a fluorescence enzyme immu-
trates the arterial subendothelial space more easily, and is noassay (ST AIA-PACK IRI; Toso, Tokyo, Japan). The
more susceptible to chemical modification, including oxida- HOMA-IR was calculated as follows: FPG (mg/dL) × FIRI
tion (10). An increase in reactive oxygen species, which is (μU/mL)/405 (17). The serum high-sensitivity C-reactive
often accompanied by various cardiovascular risk factors, protein (hsCRP) levels were measured using latex agglutina-
such as dyslipidemia, hypertension, and diabetes mellitus, tion turbidimetry. The LDL-C levels were calculated using
leads to the formation of oxidized LDL (11). the Friedewald formula (18). The HDL-C and TG levels
Malondialdehyde-LDL (MDA-LDL) is a major form of oxi- were measured using visible spectrophotometry (Determiner
dized LDL that plays a key role in the progression of athe- L HDL-C and Determiner L TG II, respectively; Kyowa
rosclerosis (10). Increased serum MDA-LDL levels were Medex, Tokyo, Japan). The MDA-LDL levels were meas-
found to be associated with coronary artery disease ured with an enzyme-linked immunosorbent assay using
(CAD) (12-14) and coronary artery calcification (15). Fur- monoclonal antibodies specific for MDA-LDL (clone ML25)
thermore, serum MDA-LDL levels are positively correlated and to apolipoprotein B (clone AB16) (Sekisui Medical, To-
with the carotid intima-media thickness (13, 16). Therefore, kyo, Japan). The sdLDL-C levels were measured using a ho-
it might be important to measure not only the LDL-C levels mogeneous method (sdLDL-Ex; Denka Seiken, Tokyo, Ja-
but also the MDA-LDL and sdLDL-C levels. However, pan). The uric acid (UA) levels were measured with an L-
measurements of LDL subclasses are not commonly per- Type UA M kit using the uricase-N-(3-sulfopropyl)-3-
formed and can be expensive in clinical settings. Given that methoxy-5-methylaniline method (Wako Pure Chemicals).
both the TG/HDL-C ratio and LDL subclasses are associated Verbal consent for the use of anonymized health records
with IR, atherosclerosis, and CVD, there is a possibility of a was obtained from all study participants. The study protocol
relationship between the serum TG/HDL-C ratio and LDL was approved by the Ethics Committee of the Tokai Univer-
subclasses. However, such an association has not been as- sity School of Medicine (protocol number: 14R-109).
sessed thus far.
Statistical analyses
The present study therefore investigated whether or not
the TG/HDL-C ratio is associated with the MDA-LDL or Data were expressed as the mean ± standard deviation or
sdLDL-C level in Japanese adults. median (interquartile range). Normality was examined using
the Kolmogorov-Smirnov test. Bonferroni’s multiple com-
Subjects and Methods parison test was used to compare mean values across three
or more groups. Student’s t-test was used to compare mean
values between two groups. The associations between study
Subjects
variables were investigated using Pearson’s correlation coef-
A total of 1329 subjects who underwent an annual health ficient. The MDA-LDL and sdLDL-C levels were compared
examination at the Health Evaluation and Promotion Center after stratifying subjects according to the sex, TG/HDL-C,
of Tokai University Hachioji Hospital between April 2011 and LDL-C levels. Since the reference ranges for MDA-
and March 2015 were included in this cross-sectional study. LDL and sdLDL-C were uncertain, upper tertile values
After excluding 261 subjects who were taking medications (157.2 [U/L] and 44.4 [μg/dL] for men, and 130.0 [U/L]
for hypertension, diabetes mellitus, dyslipidemia, hyperurice- and 35.5 [μg/dL] for women) were determined.
mia, or chronic renal disease as well as those with a history A multiple linear regression analysis was performed to
of stroke, coronary artery disease, or chronic renal failure, identify the significant determinants of MDA-LDL or
1068 subjects were included in the final analysis. Medical sdLDL-C. The age, body mass index (BMI), WC, systolic
histories were obtained using self-administered question- and diastolic BP, FPG, FIRI, TG/HDL-C ratio, and hsCRP
naires and via interviews conducted by nurses. were used as independent variables in the multiple linear re-
gression analysis of MDA-LDL or sdLDL-C. We then per-

2
 Intern Med Advance Publication DOI: 10.2169/internalmedicine.4954-20

Table 1. Characteristics of Study Subjects.

Men (n=651) Women (n=417)

Age 55.5±12.1 57.6±11.7*
BMI (kg/m2) 23.8±3.2 21.7±2.9**
Waist circumference (cm) 84.1±8.5 78.1±8.5**
Systolic BP (mmHg) 122.0±17.0 117.8±19.0**
Diastolic BP (mmHg) 78.5±12.8 71.6±12.3**
FPG (mg/dL) 102.9±18.9 98.2±18.6**
FIRI (μIU/mL) 6.24±4.64 5.08±2.99**
HOMA-IR 1.64±1.57 1.27±0.94**
TG (mg/dL) 103.0 [73.0,147.0] 73.0 [55.0,103.0]**
HDL-C (mg/dL) 59.4±14.5 76.4±16.8**
TG to HDL-C ratio 1.74 [1.13,2.89] 0.95[0.67,1.53]**
LDL-C (mg/dL) 121.6±30.7 127.2±33.3*
non-HDL-C (mg/dL) 146.3±34.1 143.9±35.9
UA (mg/dL) 6.3±1.2 4.6±0.94**
MDA-LDL (U/L) 142.4±43.0 121.0±41.0**
sdLDL-C (mg/dL) 40.0±16.8 32.9±13.2**
hsCRP (mg/dL) 0.04 [0.02,0.09] 0.03 [0.02,0.07]
Variables are given as means±standard deviations or median [inter-quartile range].
BMI: body mass index, BP: blood pressure, FPG: fasting plasma glucose, FIRI:
fasting immunoreactive insulin, HOMA-IR: homeostasis model assessment-insu-
lin resistance, TG: triglyceride, HDL-C: high-density lipoprotein cholesterol,
LDL-C: low-density lipoprotein cholesterol, non-HDL-C: non high-density lipo-
protein cholesterol, UA: uric acid, MDA-LDL: malondialdehyde-modified low-
density lipoprotein, sdLDL-C: small-dense low-density lipoprotein cholesterol,
hsCRP: high-sensitivity C-reactive protein
**p<0.01, *p<0.05 by paired t- test.

formed a multiple logistic regression analysis to calculate respectively. Most markers were accentuated in men, except
the odds ratios (ORs) for the upper tertile of MDA-LDL or for the age, HDL-C level, and LDL-C level.
sdLDL-C using the same variables utilized in the multiple The association between the logarithmic transformed the
linear regression analysis. A stepwise procedure was used to TG/HDL-C ratio [ln(TG/HDL-C)] and the MDA-LDL or
select variables for multiple regression analyses. A receiver sdLDL-C level was investigated using a correlation analysis.
operating characteristic (ROC) curve was prepared in order The Pearson’s correlation coefficients of ln(TG/HDL-C) for
to evaluate the discriminatory ability for the variables, and the MDA-LDL level in men and women were 0.388 and
the area under the curve (AUC) with its 95% confidence in- 0.333, respectively (Fig. 1a and b). The Pearson’s correla-
terval (CI) was calculated. In order to determine the optimal tion coefficients of ln(TG/HDL-C) for the sdLDL-C level in
cut-off point of TG/HDL-C, the square root of [(1 - sensitiv- men and women were 0.641 and 0.522, respectively
ity)2 + (1 - specificity)2] was calculated, which was the point [Fig. 1(c) and (d)]. These results indicated that both MDA-
on the ROC curve with the shortest distance from the upper LDL and sdLDL-C levels were positively associated with
left corner. the TG/HDL-C ratio.
All statistical analyses were performed using SAS studio The determinants of MDA-LDL or sdLDL-C were identi-
software program, version 3.4 (SAS Institute, Cary, NC, fied by a multiple linear regression analysis (Table 2).
USA). All p-values were two-tailed, and a p-value of <0.05 Among the variables included in this study (age, BMI, WC,
was considered statistically significant. systolic and diastolic BP, FPG, FIRI, and TG/HDL-C ratio),
the BMI and the TG/HDL-C ratio for men (Table 2a) and
Results the age, WC, FPG and the TG/HDL-C ratio for women (Ta-
ble 2b) were selected for MDA-LDL using a stepwise pro-
All variables evaluated in this study are shown in Table 1, cedure. The analysis revealed that the BMI and the TG/
with data stratified according to sex. Of the 1068 subjects HDL-C ratio were positively associated with the MDA-LDL
included, 417 (39.0%) were women. The mean age, median level in men (Table 2a), while the age, WC, FPG level, and
TG level, mean HDL-C level, and median the TG/HDL-C TG/HDL-C ratio were positively associated with the MDA-
ratio of the men were 55.5 years, 103.0 mg/dL, 59.4 mg/dL, LDL level in women (Table 2b).
and 1.74, respectively. The mean age, median TG level, For sdLDL-C, the BMI and TG/HDL-C ratio in men (Ta-
mean HDL-C level, and median the TG/HDL-C ratio of the ble 2a) and the age, WC, diastolic BP, FPG level, and TG/
women were 57.6 years, 73.0 mg/dL, 76.4 mg/dL, and 0.95, HDL-C ratio in women (Table 2b) were selected using a

3
 Intern Med Advance Publication DOI: 10.2169/internalmedicine.4954-20

(a) Men (b) Women

r = 0.388 [0.321,0.451], p<0.001 r = 0.333 [0.244,0.415], p<0.001
350 300

MDA-LDL (U/L)
300

MDA-LDL (U/L)
250
250
200
200
150
150
100 100
50 50
0 0
-1.5 -1.0 -0.5 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 -1.5 -1.0 -0.5 0 0.5 1.0 1.5 2.0
Ln(TG/HDL-C) Ln(TG/HDL-C)

(c) Men (d) Women

r = 0.641 [0.593,0.684], p<0.001 r = 0.522 [0.448,0.588], p<0.001
100
100

sd-LDL-C (Pg/dL)
sd-LDL-C (Pg/dL)

80
80
60
60
40 40

20 20
0 0
-1.5 -1.0 -0.5 0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 -1.5 -1.0 -0.5 0 0.5 1.0 1.5 2.0
Ln(TG/HDL-C) Ln(TG/HDL-C)

Figure 1. Scatter plots and regression lines for the comparisons of MDA-LDL and sdLDL-C levels
(a, c: men and b, d: women) and ln (TG/HDL-C). Pearson’s correlation coefficient with 95% confi-
dence intervals is indicated on the graph. MDA-LDL: malondialdehyde-modified low-density lipo-
protein, sdLDL-C: small-dense low-density lipoprotein cholesterol, ln (TG/HDL-C): logarithmic
transformed triglyceride to high-density lipoprotein cholesterol ratio

Table 2. Multiple Linear Regression Analysis for the MDA-LDL and SdLDL-C.

(a) Men
MDA-LDL sdLDL-C
RC SRC t p RC SRC t p
BMI 2.43897 0.17940 4.58 <0.0001 0.89203 0.16791 4.96 <0.0001
TG/HDL-C 4.99563 0.23936 6.11 <0.0001 4.13176 0.50659 14.97 <0.0001
Variable selection was made by a stepwise procedure. MDA-LDL: malondialdehyde-modified low-density lipoprotein, sdLDL-
C: small-dense low-density lipoprotein cholesterol, RC: regression coefficient, SRC: standardized regression coefficient, BMI:
body mass index, TG: triglyceride, HDL-C: high-density lipoprotein cholesterol

(b) Women
MDA-LDL sdLDL-C
RC SRC t p RC SRC t p
Age 0.72496 0.206.3 4.55 <0.0001 0.12239 0.10828 2.69 0.0074
WC 0.83574 0.17278 3.62 0.0003 0.18355 0.11814 2.72 0.0067
Diastolic BP ---- ---- ---- ---- 0.12097 0.11295 2.79 0.0056
FPG 0.32901 0.14922 3.26 0.0012 0.10325 0.14578 3.60 0.0004
TG/HDL-C 10.00933 0.20149 4.35 <0.0001 7.36252 0.46141 11.27 <0.0001
Variable selection was made by a stepwise procedure. MDA-LDL: malondialdehyde-modified low-density lipoprotein, sdLDL-
C: small-dense low-density lipoprotein cholesterol, RC: regression coefficient, SRC: standardized regression coefficient, WC:
waist circumference, BP: blood pressure, FPG: fasting plasma glucose, TG: triglyceride, HDL-C: high-density lipoprotein choles-
terol

4
 Intern Med Advance Publication DOI: 10.2169/internalmedicine.4954-20

Table 3. Multiple Logistic Regression Analysis for the Upper Tertile of MDA-LDL and SdLDL-C.

(a) Men
Upper tertile of MDA-LDL Upper tertile of sdLDL-C
RC SE OR 95% CI p RC SE OR 95% CI p
Age 0.0442 0.0109 1.045 1.023-1.068 <0.0001 0.0248 0.0109 1.025 1.004-1.047 0.0225
WC 0.0580 0.0147 1.060 1.030-1.091 <0.0001 0.0455 0.0152 1.047 1.016-1.078 0.0028
Diastolic BP ---- ---- ---- ---- ---- 0.0229 0.0100 1.023 1.003-1.043 0.0221
TG/HDL-C 0.6275 0.1569 1.873 1.377-2.547 <0.0001 0.9607 0.1701 2.613 1.873-3.647 <0.0001
Variable selection was made by a stepwise procedure. MDA-LDL: malondialdehyde-modified low-density lipoprotein, sdLDL-C: small-dense low-
density lipoprotein cholesterol, RC: regression coefficient, SE: standard error, OR: odds ratio, CI: confidence interval, WC: waist circumference, BP:
blood pressure, TG: triglyceride, HDL-C: high-density lipoprotein cholesterol

(b) Women
Upper tertile of MDA-LDL Upper tertile of sdLDL-C
RC SE OR 95% CI p RC SE OR 95% CI p
Age -0.0254 0.0078 0.975 0.960-0.990 0.0011 ---- ---- ---- ---- ----
Systolic BP 0.0157 0.0054 1.016 1.005-1.027 0.0034 ---- ---- ---- ---- ----
BMI ---- ---- ---- ---- ---- 0.0867 0.0350 1.091 1.018-1.168 0.0133
TG/HDL-C 0.2687 0.0500 1.308 1.186-1.443 <0.0001 0.9297 0.0908 2.527 2.115-3.020 <0.0001
Variable selection was made by a stepwise procedure. MDA-LDL: malondialdehyde-modified low-density lipoprotein, sdLDL-C: small-dense low-
density lipoprotein cholesterol, RC: regression coefficient, SE: standard error, OR: odds ratio, CI: confidence interval, BP: blood pressure, BMI: body
mass index, TG: triglyceride, HDL-C: high-density lipoprotein cholesterol

stepwise procedure. The analysis revealed that the BMI and gression analysis revealed that the TG/HDL-C ratio was
TG/HDL-C ratio were positively associated with the sdLDL- positively associated with the upper tertile of MDA-LDL
C level in men (Table 2a), while the age, WC, diastolic BP, and sdLDL-C levels in both men and women.
FPG and the TG/HDL-C ratio were positively associated We previously reported that both the MDA-LDL and
with the sdLDL-C level in women (Table 2b). Collectively, sdLDL-C levels were positively correlated with the LDL-C
the multiple linear regression analysis revealed that the TG/ levels19). Indeed, the MDA-LDL and sdLDL-C levels had a
HDL-C ratio was positively correlated with the MDA-LDL moderate to strong association with the LDL-C levels in
and sdLDL-C levels in both men and women. men and women in the present study (Supplemental Fig.). In
The determinants for the upper tertile of MDA-LDL level addition, the LDL-C levels increased with the TG/HDL-C
were analyzed using a multiple logistic regression analysis ratio in both men and women based on a comparison of the
(Table 3). When we analyzed the same variables used in the LDL-C levels after subjects were stratified according to sex
multiple linear regression analysis (age, BMI, WC, systolic and the TG/HDL-C ratio (Fig. 2). To rule out the possibility
and diastolic BP, FPG, FIRI, and the TG/HDL-C ratio), age, that the MDA-LDL and sdLDL-C levels increased due to
systolic BP, and the TG/HDL-C ratio for men (Table 3a) elevated LDL-C levels, we compared the MDA-LDL and
and age, WC, and the TG/HDL-C ratio for women (Ta- sdLDL-C levels after subjects were stratified according to
ble 3b) were selected using a stepwise procedure. The re- the combination of LDL-C and the TG/HDL-C ratio
sults of the analysis revealed that systolic BP and the TG/ (Fig. 3). Within the same LDL-C range, both the MDA-LDL
HDL-C ratio for men (Table 3a) and age, WC, and the TG/ and sdLDL-C levels increased with the TG/HDL-C ratio, ex-
HDL-C ratio for women (Table 3b) were positively associ- cept for the MDA-LDL levels in the LDL-C <112 mg/dL
ated, whereas age in men was negatively associated with the group of women. These results further supported the notion
upper tertile of MDA-LDL. that the MDA-LDL and sdLDL-C levels were positively as-
The determinants for the upper tertile of sdLDL-C level sociated with the TG/HDL-C ratio, especially in the high-
were analyzed using a multiple logistic regression analysis LDL-C range, in both men and women.
(Table 3). When we analyzed the same variables used in the Fig. 4 shows the ROC curve for evaluating the discrimi-
multiple linear regression analysis, the BMI and TG/HDL-C natory ability for 3rd tertile values of MDA-LDL and
ratio for men (Table 3a) and the age, WC, diastolic BP, and sdLDL-C. The AUCs (95% CIs) were 0.719 (0.678, 0.759)
TG/HDL-C ratio for women (Table 3b) were selected using for MDA-LDL and 0.856 (0.826, 0.887) for sdLDL-C in
a stepwise procedure. The results of the analysis revealed men and 0.680 (0.627, 0.783) for MDA-LDL and 0.707
that the BMI and TG/HDL-C ratio for men (Table 3a) and (0.654, 0.761) for sdLDL-C in women.
the age, WC, diastolic BP, and TG/HDL-C ratio for women The optimal cut-off points of the TG/HDL-C ratio yield-
(Table 3b) were positively associated with the upper tertile ing the minimum value of the square root of [(1 - sensitiv-
of the sdLDL-C level. Collectively, this multiple logistic re- ity)2 + (1 - specificity)2] for the upper tertile of MDA-LDL

5
 Intern Med Advance Publication DOI: 10.2169/internalmedicine.4954-20

(a) Men (b) Women

** **
140 ** 160 **
120 140

LDL-C (mg/dL)
LDL-C (mg/dL)
100 120
80 100
80
60
60
40
40
20 20
(214) (220) (217) (138) (138) (141)
0 0
< 1.28 1.28 – • 2.40 < 0.73 0.73 – • 1.29
< 2.40 < 1.29
TG/HDL-C TG/HDL-C
Figure 2. A bar graph of the mean LDL-C values with 95% confidence intervals after stratifying
the subjects according to sex and TG/HDL-C values. LDL-C: low-density lipoprotein cholesterol, TG/
HDL-C: triglyceride to high-density lipoprotein cholesterol ratio. **p<0.01 according to Bonferroni’s
multiple comparison test

(a) Men (b) Women

**
** 180
200 ** 160 **
** * **
MDA-LDL (U/L)

MDA-LDL (U/L)

** TG/HDL- 140 TG/HDL-

150 120
C
< 1.28 C
< 0.73
1.28 - < 2.40 100 0.73 - < 1.29
100 80
• 2.40 • 1.29
60
50 40
20
0 (89) (58) (69) (78) (81) (58) (47) (81) (90) (n) 0 (65) (44) (29) (46) (50) (43) (27) (43) (69) (n)
< 108 108- < 134 • 134 < 112 112- < 143 • 143
LDL-C (mg/dL) LDL-C (mg/dL)

(c) Men (d) Women

** **
** 60 **
70
sd-LDL-C (Pg/dL)

** **
sd-LDL-C (Pg/dL)

60 ** ** ** ** 50 **
50 TG/HDL- ** * TG/HDL-
40
C
< 1.28 ** C
< 0.73
40 0.73 - < 1.29
1.28 - < 2.40 30
30 • 1.29
• 2.40 20
20
10 10
0 (89) (58) (69) (78) (81) (58) (47) (81) (90) (n)
0 (65) (44) (29) (46) (50) (43) (27) (43) (69) (n)

< 112 112- < 143 • 143

< 108 108- < 134 • 134
LDL-C (mg/dL)
LDL-C (mg/dL)

Figure 3. A bar graph of the mean MDA-LDL and sdLDL-C levels (a, c: men and b, d: women)
with 95% confidence intervals after stratifying the subjects according to sex, the TG/HDL-C ratio,
and LDL-C values. MDA-LDL: malondialdehyde-modified low-density lipoprotein, sdLDL-C: small-
dense low-density lipoprotein cholesterol, LDL-C: low-density lipoprotein cholesterol, TG/HDL-C:
triglyceride to high-density lipoprotein cholesterol ratio. *p<0.05, **p<0.01 according to Bonferroni’s
multiple comparison test.

and sdLDL-C were 1.85 and 2.03 in men and 0.88 and 1.30 specificity for sdLDL-C of 0.778 and 0.803 in men, and
in women, respectively. The optimal cut-off points were also 0.786 and 0.553 in women, respectively.
the points that maximized the product of the sensitivity and Since the LDL size is negatively regulated by the serum
specificity, with a sensitivity and specificity for MDA-LDL TG levels and is significantly reduced in hypertriglyc-
of 0.644 and 0.703 in men and 0.543 and 0.748 in women, eridemic subjects, we compared the discriminatory ability
respectively. The optimal cut-off points with sensitivity and for the variables of the TG/HDL-C ratio and TG (Supple-

6
 Intern Med Advance Publication DOI: 10.2169/internalmedicine.4954-20

(a) Men (b) Women

TG/HDL-C 1.85 (0.644, 0.703) TG/HDL-C 0.88 (0.543, 0.748)
1.0 1.0

Sensitivity
Sensitivity
0.8 0.8
0.6 0.6
0.4 0.4
AUC=0.719 AUC=0.680
0.2 [0.678,0.759] 0.2 [0.627,0.733]
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
(1-Specificity) (1-Specificity)
MDA-LDL•157.2 (U/L) MDA-LDL•130.0 (U/L)

TG/HDL-C 2.03 (0.778, 0.803) TG/HDL-C 1.30 (0.786, 0.553)

1.0 1.0
Sensitivity

0.8 0.8

Sensitivity
0.6 0.6
0.4 AUC=0.856 0.4 AUC=0.707
0.2 [0.826,0.887] 0.2 [0.654,0.761]
0.0 0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
(1-Specificity) (1-Specificity)
SdLDL-&•44.4 (Pg/dL) SdLDL-&•35.5 (Pg/dL)

Figure 4. The ROC curves of the TG/HDL-C ratio for predicting the upper tertile of the MDA-LDL
and sdLDL-C levels. The AUC with its 95% CI and optimal cut-off point (sensitivity, specificity) for
the TG/HDL-C ratio are shown in the graph. ROC: receiver operator characteristic, TG/HDL-C:
triglyceride to high-density lipoprotein cholesterol ratio, MDA-LDL: malondialdehyde-modified low-
density lipoprotein, sdLDL-C: small-dense low-density lipoprotein cholesterol, AUC: area under the
curve, CI: confidence interval

mental Table). Based on the results of ROC analyses, the In this study, we analyzed the association between the
discriminatory abilities of the TG/HDL-C ratio and TG for TG/HDL-C ratio and MDA-LDL or sdLDL-C level using
MDA-LDL and sdLDL-C were higher in men than in correlation and multiple regression analyses. Since the TG/
women. In addition, both the TG/HDL-C ratio and TG HDL-C ratio varies between men and women based on a
showed a higher discriminatory ability for sdLDL-C than previous report (20), we analyzed the data after subjects had
MDA-LDL, especially in men. The cut-off points of the TG/ been stratified according to sex. The analysis of Pearson’s
HDL-C ratio and TG for high MDA-LDL and sdLDL-C correlation coefficient for MDA-LDL and sdLDL-C levels
were both higher in men than in women, and those cut-off revealed that the correlation coefficients were higher in men
points for sdLDL-C were higher than for MDA-LDL in both than in women. The multiple linear regression analysis indi-
men and women. Taken together, these findings suggest that cated that the standardized regression coefficients for MDA-
the TG/HDL-C ratio was a useful predictor for not only LDL and sdLDL-C were higher in men than in women.
high sdLDL-C but also MDA-LDL compared to TG in both Based on the ORs in the multiple logistic regression analy-
men and women. sis, the TG/HDL-C ratio was a stronger determinant for the
upper tertile of both MDA-LDL and sdLDL-C levels in
Discussion women than in men. However, gender differences in the re-
lationship between the TG/HDL-C ratio and the MDA-LDL
We showed that the TG/HDL-C ratio is associated with or sdLDL-C/upper tertile of MDA-LDL or sdLDL-C level in
both MDA-LDL and sdLDL-C levels in Japanese subjects. these analyses were not notable. The exact reason for the
These results suggest that the TG/HDL-C ratio may be use- gender differences in this study was unclear; however, it is
ful for assessing the risk of CVD. To evaluate atherogenic speculated that higher HDL-C and LDL-C levels in women,
LDL, assessments may focus on not only LDL-C levels but which are particularly increased after menopause, than in
also the TG/HDL-C ratio. men might have been involved.

7
 Intern Med Advance Publication DOI: 10.2169/internalmedicine.4954-20

Why the TG/HDL-C ratio more strongly influenced the viduals; therefore, the effect of ethnicity on the relationship
sdLDL-C than the MDA-LDL levels was unclear based on between the TG/HDL-C levels and LDL subclasses was not
the Pearson’s correlation coefficient, standardized regression assessed. Finally, our dataset was small; therefore, our find-
coefficient, and OR. The LDL size was negatively regulated ings might not be generalizable to all Japanese individuals.
by the serum TG levels and was significantly reduced in hy- In conclusion, the TG/HDL-C ratio is associated with
pertriglyceridemic subjects. The formation of sdLDL is LDL subclasses in healthy Japanese subjects. The quantifica-
closely associated with IR and hypertriglyceridemia (21). It tion of the MDA-LDL and sdLDL-C levels is not a routine
is well recognized that the TG and HDL-C levels are in- test; given our data, the routine use of triage tests for CVD
versely related, so high TG and low HDL-C levels lead to a based on an increased TG/HDL-C ratio might help identify
high TG/HDL-C ratio and high sdLDL-C levels. Although subjects with metabolic abnormalities. Since both TG and
sdLDL is prone to oxidization, oxidized LDL results from HDL-C levels are routinely evaluated, the TG/HDL-C ratio
the exposure of LDL to several oxidizing agents, enzymes, might be a useful indicator for LDL subclasses, particularly
and products of myeloperoxidase (22), suggesting that not when the TG/HDL-C ratio and LDL-C levels are both con-
all MDA-LDL came from sdLDL. This may have contrib- sidered for the evaluation.
uted to the TG/HDL-C ratio differently affecting the deter-
mination of the MDA-LDL and sdLDL-C levels. The author states that he has no Conflict of Interest (COI).
The LDL-C and MDA-LDL or sdLDL-C levels had a
moderate to strong association in both men and women Acknowledgement
(Supplemental Fig). The mean LDL-C level was signifi- The author would like to thank the staff at the Health Evalu-
cantly higher in women than in men, while the mean MDA- ation and Promotion Center, Tokai University Hachioji Hospital,
LDL and sdLDL-C levels were significantly higher in men for providing assistance in data collection.
than in women (Table 1). Furthermore, some subjects with
References
LDL-C levels in the lowest and middle tertiles had relatively
elevated MDA-LDL and sdLDL-C levels among both men 1. McLaughlin T, Abbasi F, Cheal K, Chu J, Lamendola C, Reaven
and women (Fig. 3). Thus, the estimation of the atherogenic G. Use of metabolic markers to identify overweight individuals
MDA-LDL or sdLDL-C levels based on the LDL-C level who are insulin resistant. Ann Intern Med 139: 802-809, 2003.
2. McLaughlin T, Reaven G, Abbasi F, et al. Is there a simple way
alone is insufficient. Accordingly, the use of MDA-LDL or to identify insulin-resistant individuals at increased risk of cardio-
sdLDL-C levels combined with the LDL-C level and the vascular disease?. Am J Cardiol 96: 399-404, 2005.
TG/HDL-C ratio might provide a better estimation than us- 3. Kim JS, Kang HT, Shim JY, Lee HR. The association between the
ing the LDL-C level alone. triglyceride to high-density lipoprotein cholesterol ratio with insu-
Since the quantitation of MDA-LDL and sdLDL-C levels lin resistance (HOMA-IR) in the general Korean population: based
on the National Health and Nutrition Examination Survey in 2007-
is not a routine test, its use in the estimation of CVD risk is 2009. Diabetes Res Clin Pract 97: 132-138, 2012.
not common in healthy subjects. In contrast, the TG, HDL- 4. de León AC, Coello SD, González DA, et al. Impaired fasting glu-
C, and LDL-C levels are routinely measured, suggesting that cose, ancestry and waist-to-height ratio: main predictors of inci-
the use of both the TG/HDL-C ratio and LDL-C level to dent diagnosed diabetes in the Canary Islands. Diabet Med 29:
identify individuals who are at an increased risk for CVD 399-403, 2012.
5. Hadaegh F, Khalili D, Ghasemi A, Tohidi M, Sheikholeslami F,
may be more practical than using the TG/HDL-C ratio or Azizi F. Triglyceride/HDL-cholesterol ratio is an independent pre-
LDL-C per se. Furthermore, we recently reported that TG/ dictor for coronary heart disease in a population of Iranian men.
HDL-C ratio is associated with IR, components of metabolic Nutr Metab Cardiovasc Dis 19: 401-408, 2009.
syndrome (MetS), and lifestyle habits (23). The TG/HDL-C 6. Liang J, Fu J, Jiang Y, Dong G, Wang X, Wu W. TriGlycerides
ratio might therefore be a useful indicator for evaluating and high-density lipoprotein cholesterol ratio compared with ho-
meostasis model assessment insulin resistance indexes in screening
CVD and also suggest ways to prevent and improve athero-
for metabolic syndrome in the Chinese obese children: a cross sec-
sclerotic disorders. tion study. BMC Pediatr 15: 138, 2015.
The discrimination abilities of the TG/HDL-C ratio for 7. Carmena R, Duriez P, Fruchart JC. Atherogenic lipoprotein parti-
MDA-LDL and sdLDL-C were as good as those of TG in cles in atherosclerosis. Circulation 109: 2-7, 2004.
both men and women (Supplemental Table). However, the 8. Kathiresan S, Otvos JD, Sullivan LM, et al. Increased small low-
density lipoprotein particle number: a prominent feature of the
cut-off points of TG for MDA-LDL and sdLDL-C were
metabolic syndrome in the Framingham Heart Study. Circulation
both <150 mg/dL. It is well recognized that the diagnostic 113: 20-29, 2006.
criteria of MetS and hypertriglyceridemia include a TG level 9. Ito Y, Fujimura M, Ohta M, Hirano T. Development of a homoge-
! 150 mg/dL. For this reason, TG cut-off points of <150 mg/ neous assay for measurement of small dense LDL cholesterol.
dL can be ignored and may not be practical in clinical set- Clin Chem 57: 57-65, 2011.
10. Orekhov AN, Bobryshev YV, Sobenin IA, Melnichenko AA,
tings.
Chistiakov DA. Modified low density lipoprotein and lipoprotein-
The major limitation of this study is its cross-sectional containing circulating immune complexes as diagnostic and prog-
design, as it inhibited the assessment of the causal relation- nostic biomarkers of atherosclerosis and type 1 diabetes macrovas-
ship between the TG/HDL-C ratio and LDL subclasses. All cular disease. Int J Mol Sci 15: 12807-12841, 2014.
participants in this study were middle-aged Japanese indi- 11. Navab M, Ananthramaiah GM, Reddy ST, et al. The oxidation hy-

8
 Intern Med Advance Publication DOI: 10.2169/internalmedicine.4954-20

pothesis of atherogenesis: the role of oxidized phospholipids and 18. Friedewald WT, Levy RI, Fredrickson DS. Estimation of the con-
HDL. J Lipid Res 45: 993-1007, 2004. centration of low-density lipoprotein cholesterol in plasma, with-
12. Holvoet P, Mertens A, Verhamme P, et al. Circulating oxidized out use of the preparative ultracentrifuge. Clin Chem 18: 499-502,
LDL is a useful marker for identifying patients with coronary ar- 1972.
tery disease. Arterioscler Thromb Vasc Biol 21: 844-848, 2001. 19. Moriyama K, Takahashi E. Evaluation of malondialdehyde low-
13. Tanaga K, Bujo H, Inoue M, et al. Increased circulating density lipoprotein stratified by low-density lipoprotein cholesterol.
malondialdehyde-modified LDL levels in patients with coronary Clin Lab 63: 1179-1186, 2017.
artery diseases and their association with peak sizes of LDL parti- 20. Salazar MR, Carbajal HA, Espeche WG, et al. Relation among the
cles. Arterioscler Thromb Vasc Biol 22: 662-666, 2002. plasma triglyceride/high-density lipoprotein cholesterol concentra-
14. Miyazaki T, Shimada K, Sato O, et al. Circulating tion ratio, insulin resistance, and associated cardio-metabolic risk
malondialdehyde-modified LDL and atherogenic lipoprotein pro- factors in men and women. Am J Cardiol 109: 1749-1753, 2012.
files measured by nuclear magnetic resonance spectroscopy in pa- 21. Berneis KK, Krauss RM. Metabolic origins and clinical signifi-
tients with coronary artery disease. Atherosclerosis 179: 139-145, cance of LDL heterogeneity. J Lipid Res 43: 1363-1379, 2002.
2005. 22. Mello AP, da Silva IT, Abdalla DS, Damasceno NR. Electronega-
15. Asamiya Y, Yajima A, Tsuruta Y, Otsubo S, Nitta K. Oxidised tive low-density lipoprotein: origin and impact on health and dis-
LDL/LDL-cholesterol ratio and coronary artery calcification in ease. Atherosclerosis 215: 257-265, 2011.
haemodialysis patients. Nutr Metab Cardiovasc Dis 23: 619-627, 23. Moriyama K. Associations Between the Triglyceride to High-
2013. Density Lipoprotein Cholesterol Ratio and Metabolic Syndrome,
16. Hayashi Y, Okumura K, Matsui H, et al. Impact of low-density Insulin Resistance, and Lifestyle Habits in Healthy Japanese. Me-
lipoprotein particle size on carotid intima-media thickness in pa- tab Syndr Relat Disord in press .
tients with type 2 diabetes mellitus. Metabolism 56: 608-613,
2007. The Internal Medicine is an Open Access journal distributed under the Creative
17. Matthews DR, Hosker JP, Rudenski AS, et al. Homeostasis model Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. To
assessment: insulin resistance and beta-cell function from fasting view the details of this license, please visit (https://creativecommons.org/licenses/
plasma glucose and insulin concentrations in man. Diabetologia by-nc-nd/4.0/).

28: 412-419, 1985.

Ⓒ The Japanese Society of Internal Medicine

Intern Med Advance Publication