SUPPLEMENTARY DATA
Supplement to: Würtz P, Raiko JR, Magnussen CG, et al. High-throughput quantification of
circulating metabolites improves prediction of subclinical atherosclerosis.
Supplementary methods
Conventional risk factors; NMR spectroscopy; Metabolite quantification
Coefficients of variation and tracking of tyrosine, glutamine, and docosahexaenoic acid
Cross-sectional associations of tyrosine and glutamine with carotid IMT
Associations of tyrosine and glutamine with coronary artery disease
Figure S1: Inter-correlations of lipid measures and assayed metabolites
Table S1: Baseline metabolite levels
Table S2: Odds ratios for 6-year incident high IMT and/or plaque for all assayed small
molecules and serum extract metabolites
Table S3: Characteristics of the additional populations studied for cross-sectional validation
of tyrosine and glutamine with carotid IMT and presence of coronary artery disease
Table S4: Predictive performance of metabolite quantification in addition to conventional
lipids.
Table S5: Odds ratios and predictive performance of alternative outcomes representing
increased subclinical atherosclerosis
Table S6: Odds ratios and evaluation of prediction models for progression and regression of
carotid IMT
Predictive performance using alternative cutpoints to denote high IMT
Figure S2: Area under the ROC curve and net reclassification index for alternative percentiles
used as cut-point to denote high IMT
Table S7: Predictive performance with random division of the study population for split risk
algorithm calculation and comparison of risk discrimination
References
1
Supplemental methods
Conventional risk factors
Height, weight, and waist circumference were measured and body mass index calculated as
weight/height2 in units of kg/m2. Blood pressure was measured at sitting position with a
random zero sphygmomanometer and the mean of three measurements used. Cigarette
smoking was assessed with a questionnaire and smoking on a daily basis (yes/no) was used
to define smoking status. Family history of cardiovascular disease was defined as positive if
either parent had cardiovascular disease, diagnosed coronary event or cerebroal vascular
disease before the age of 55.
Venous blood samples were drawn after an overnight fast and stored at –70°C.
Standard enzymatic methods were used for measurement of serum total cholesterol,
triglycerides and HDL-cholesterol. LDL-cholesterol was calculated using the Friedewald
formula1 and further directly measured for 746 individuals. Apolipoprotein A-1 and
apolipoprotein B were analysed immunoturbidimetrically (Orion Diagnostica, Espoo,
Finland). High-sensitivity serum C-reactive protein was determined turbidimetrically (CRPUL, Wako, USA) on an automated analyser (AU400, Olympus, Japan). Serum glucose was
determined enzymatically (Glucose Reagent, Olympus, Ireland). Details of the study
population and ultrasound assessment have been described previously.2
NMR spectroscopy
Metabolites were quantified from a high-throughput platform with NMR data measured
using Bruker AVANCE III spectrometer operating at 500 MHz. From each sample, three NMR
spectra were recorded; two spectra were measured from native serum and one from serum
lipid extracts.3 Measurements of native serum samples (300 µl) and serum lipid extracts
were conducted at 37°C and at 22°C, respectively.
The spectrum used for lipoprotein quantification was acquired with a standard pulse
sequence with water peak suppression. The small molecule data were acquired with a pulse
sequence that suppresses most of the broad macromolecule and lipoprotein lipid signals.
Lipids were extracted from the serum samples using a standard protocol where lipoprotein
particles are broken down with methanol and dichloromethane. A standard pulse sequence
was used to acquire the data. Details of the NMR spectroscopy and metabolite
quantification have been described previously.4,5
2
Metabolite quantification
The 14 lipoprotein subclasses were calibrated via high-performance liquid chromatography6
and are as follows: extremely-large VLDL (with particle diameters from approximately 75 nm
upwards and with a possible contribution of chylomicron particles), five VLDL subclasses
(average particle diameters of 64.0 nm, 53.6 nm, 44.5 nm, 36.8 nm, and 31.3 nm), IDL (28.6
nm), three LDL subclasses (25.5 nm, 23.0 nm, and 18.7 nm), and four HDL subclasses (14.3
nm, 12.1 nm, 10.9 nm, and 8.7 nm).
The total NMR-visible lipoprotein lipid amount was converted into information on
the total lipid concentration in various lipoprotein subclasses by a regression model
approach to isolate the different lipoprotein categories and calibrated to units of mmol/l. 7 In
total, 17 spectra (0.9%) failed the automated lipoprotein quantification protocol, including 7
that were outside the boundaries of the Friedewald LDL-C equation.1 Regression models
were used for quantification of the small molecules in absolute concentrations. Serum
extract lipid constituents were calibrated to units of mmol/l using the total cholesterol
concentration in native serum. No internal standards were used for quantification.
The lipoprotein subclasses were quantified in terms of total lipid concentrations.
Therefore the sums of the subclass measures do not directly correspond to the cholesterol
concentrations of a given major lipid class. For instance, IDL is the total lipid concentration
of lipoproteins with that particular lipoprotein size (average 28.6 nm) whereas IDL-C is the
cholesterol content of that lipoprotein fraction. In contrast to the Friedewald estimated LDLC, the LDL-CNMR measure does not contain the IDL-C fraction and therefore concentrations
of LDL-CNMR in Table S1 may appear low.8
In addition to lipoprotein subclass assessment in terms of total lipid concentrations
[mmol/l], NMR spectroscopy allows for quantification in units of particle number.9 This can
be achieved by approximating the lipoprotein particle diameter and core lipid volume and
mass. The total lipoprotein lipid and particle number concentrations are different
approximations based on the total NMR-visible lipid amount.10 Essentially identical results
were obtained using lipoprotein particle numbers instead of total lipoprotein lipid
concentrations in terms of odds ratios, and derivation and evaluation of the prediction
models.
3
Coefficients of variation and 6-year tracking of tyrosine, glutamine, and docosahexaenoic
acid
To assess the biological stability of the novel biomarkers highlighted in this study we
determined the coefficients of variation and the 6-year tracking of tyrosine, glutamine, and
docosahexaenoic acid. Coefficients of variation were calculated from triplicates of
individually prepared serum samples from six individuals and measured on five consecutive
days. Tracking was estimated from Spearman’s correlation coefficients of the circulating
metabolite concentrations measured in 2001 and 2007. These analyses suggest that the
biomarkers were quantified with a high degree of accuracy and that fasting levels of these
biomarkers are biologically stable over long periods of time.
Coefficient of
6-year tracking
variation [%]
coefficients*
Tyrosine
4.5
0.42
Glutamine
2.6
0.42
Docosahexaenoic acid
3.7
0.52
* P<0.0005 for all.
4
Cross-sectional associations of tyrosine and glutamine with carotid IMT
Circulating levels of tyrosine and glutamine are potential novel biomarkers for the
development of subclinical atherosclerosis. In addition to associations with incidence and
prevalence of the combined outcome of high IMT and/or carotid plaque, both amino acids
were associated with 6-year prevalence of these outcomes when analysed separately
(Supplementary Table 4). Nevertheless, further confirmation of these biomarkers with
measures of subclinical atherosclerosis is warranted. In the lack of additional prospective
data on carotid IMT in young adults, we tested cross-sectional associations of tyrosine and
glutamine with carotid IMT in an older, independent population. To this end amino acid
quantification was conducted in a subset of the Health 2000 cohort with cross-sectional
data on carotid IMT. The Health 2000 study is a Finnish health survey carried out in 20002001.11 The overall study cohort (8028 persons) was representative of the Finnish
population of age 30 years and above. To study cardiovascular disease risk factors more
thoroughly, a supplemental study was carried out which included carotid ultrasound
examination. Subjects in this substudy were 45 years and older. The study group was
formed of 1044 subjects without usage of lipid medication and with carotid IMT and amino
acid data available. Characteristics of the Health 2000 substudy population are shown in
Supplementary Table 3. Ultrasound assessment was performed according to a standardized
protocol using a 7.5 MHz linear array transducer as described previously.12 Mean (SD)
carotid IMT was 0.80 (0.16) mm. Amino acids were quantified using the same experimental
setup as for the Cardiovascular Risk in Young Finns Study.
Associations of the two amino acids were further evaluated cross-sectionally in the
Cardiovascular Risk in Young Finns Study in the subset of individuals only attending the
follow-up studies in either 2001 or 2007, and therefore excluded from prospective analyses.
This study group of 830 individuals had a mean (SD) carotid IMT of 0.61 (0.09) mm.
Characteristics of the additional study population from the Cardiovascular Risk in Young
Finns Study are shown in Supplementary Table 3. Tyrosine has recently been associated
with insulin resistance and the risk for development of type 2 diabetes13, however, including
HOMA-insulin resistance as covariate in the models gave essentially unaltered results.
Tyrosine, glutamine, triglycerides, glucose, HOMA-insulin resistance as well as carotid IMT
were loge-transformed prior to linear regression analyses.
5
Associations of tyrosine and glutamine with coronary artery disease
A prior mass spectrometry-based metabolic profiling study has shown glutamine/glutamate
as well as a principal component factor including tyrosine to be associated with the
presence of coronary artery disease.14 In order to further study the role of tyrosine and
glutamine as biomarkers for clinical manifestations of atherosclerosis we analysed
associations of these two amino acids with the presence and severity of coronary artery
disease in an independent population. The Angiography and Genes Study (ANGES) consisted
of 1000 patients referred to coronary angiography because of chest pain and clinically
suspected CAD.15,16 The study protocol was approved by the local Ethics Committee of the
Tampere University Hospital. Coronary angiography was performed using the standard
Judkins’ technique. Transluminal narrowing of at least 50% of any major coronary artery
(left anterior descending, left circumflex or right coronary arteries) was employed for the
diagnosis of coronary artery disease. Individuals with stenosis <50% in every major
epicardial coronary artery were used as controls. The number of arteries with >50% stenosis
was used to signify the severity of coronary artery disease. Angiography-based diagnosis of
coronary artery disease and amino acid profiling was available for 967 individuals.
Characteristics of the Angiography and Genes Study population are shown in Supplementary
Table 3.
6
Figure S1. Inter-correlations of lipid measures and assayed metabolites. The colour coding
indicates Pearson’s correlation coefficients of the conventional lipids and NMR-based
lipoprotein lipid and metabolite measures analyzed in the current study (n=1595).
7
Table S1. Baseline metabolite levels
IMT<90th percentile
IMT≥90th percentile or
(n=1445)
plaque (n=150)
Total cholesterol
4.9 (0.95)
5.4 (1.1)
0.0002
IDL-cholesterol
0.73 (0.18)
0.82 (0.21)
0.0003
LDL-cholesterol
1.9 (0.54)
2.3 (0.64)
<0.0001
HDL-cholesterol
1.6 (0.39)
1.5 (0.32)
0.31
Total triglycerides
1.1 [0.9-1.5]
1.4 [1.0-1.8]
0.0006
VLDL-triglycerides
0.67 [0.46-1.0]
0.89 [0.58-1.3]
0.006
0.12 [0.10-0.15]
0.10
P
Major lipoprotein fractions [mmol/l]
IDL-triglycerides
0.11 [0.091-0.14]
Lipoprotein subclasses (total lipid concentration) [mmol/l]
Extremely large VLDL
0.007 [0.0005-0.020]
0.012 [0.004-0.034]
0.06
Very large VLDL
0.029 [0.008-0.071]
0.048 [0.017-0.11]
0.10
Large VLDL
0.16 [0.090-0.32]
0.25 [0.13-0.47]
0.04
Medium VLDL
0.45 [0.31-0.68]
0.61 [0.39-0.89]
0.002
Small VLDL
0.56 [0.44-0.73]
0.70 [0.54-0.89]
0.0001
Very small VLDL
0.48 [0.40-0.58]
0.52 [0.45-0.65]
0.15
IDL
1.2 (0.29)
1.3 (0.33)
0.002
Large LDL
1.5 (0.36)
1.7 (0.43)
<0.0001
Medium LDL
0.86 (0.23)
1.0 (0.28)
<0.0001
Small LDL
0.54 (0.15)
0.64 (0.19)
<0.0001
Very large HDL
0.37 (0.22)
0.30 (0.17)
0.37
Large HDL
0.81 (0.40)
0.62 (0.32)
0.06
Medium HDL
1.0 (0.24)
0.98 (0.22)
0.22
1.3 (0.16)
0.27
Small HDL
1.2 (0.15)
Small molecules [standardized concentration units]
Glutamine
16 (2.6)
17 (2.4)
0.02
Histidine
2.2 (0.36)
2.3 (0.35)
0.04
Tyrosine
1.7 [1.5-2.0]
1.9 [1.7-2.1]
0.004
Alanine
12 [11-14]
13 [11-14]
0.11
Isoleucine
1.7 [1.4-2.1]
1.9 [1.6-2.2]
0.21
Leucine
2.5 [2.1-2.9]
2.7 [2.3-3.1]
0.03
Phenylalanine
2.3 (0.37)
2.4 (0.40)
0.36
Valine
6.4 [5.6-7.5]
7.0 [6.1-7.8]
0.56
Lactate
29 [24-35]
30 [25-35]
0.54
8
Pyruvate
2.5 [2.0-3.0]
2.6 [2.1-3.3]
0.33
Glycerol
2.7 [2.0-3.4]
2.5 [1.9-3.4]
0.76
Citrate
3.2 (0.74)
3.1 (0.73)
0.34
α-1 acid glycoprotein
39 [36-44]
40 [37-46]
0.15
3-hydroxybutyrate
2.7 [2.2-3.7]
2.7 [2.3-3.4]
0.002
Acetate
1.3 [1.1-1.6]
1.3 [1.1-1.6]
0.49
Acetoacetate
1.2 [0.91-1.8]
1.3 [0.95-1.7]
0.13
Creatinine
2.0 [1.8-2.3]
2.0 [1.8-2.3]
0.10
2.4 (1.1)
2.3 (1.1)
0.26
3.6 (0.69)
3.9 (0.82)
0.0001
ω-6 fatty acids [mmol/l]
3.5 [3.1-3.9]
3.8 [3.3-4.2]
0.007
ω-3/ω-6
0.11 (0.033)
0.10 (0.029)
0.05
Linoleic acid [mmol/l]
3.0 [2.7-3.4]
3.3 [2.9-3.7]
0.007
0.16 [0.12-0.20]
0.15 [0.11-0.19]
0.47
Total fatty acids [mmol/l]
10 [8.9-12]
11 [9.3-13]
0.009
Free cholesterol [mmol/l]
1.4 (0.28)
1.5 (0.32)
0.002
0.86 (0.20)
0.87 (0.19)
0.2
Phosphatidylcholine [mmol/l]
2.0 [1.8-2.3]
2.0 [1.8-2.3]
0.22
Sphingomyelins [mmol/l]
0.29 (0.061)
0.30 (0.065)
0.08
ω-3 fatty acids [mmol/l]
0.35 [0.29-0.44]
0.38 [0.30-0.45]
0.48
6.3 [5.4-7.7]
7.1 [5.8-8.3]
0.02
0.057 (0.017)
0.054 (0.014)
0.05
3.0 [2.5-3.8]
3.4 [2.8-4.1]
0.06
2.0 [1.7-2.3]
2.1 [1.8-2.5]
0.17
9.7 (0.18)
9.7 (0.17)
0.39
7.7 (0.60)
7.8 (0.60)
0.20
Urea
Serum extract metabolites
Esterified cholesterol
[mmol/l]
Docosahexaenoic acid
[mmol/l]
Total phosphoglycerides
[mmol/l]
ω-7, ω-9, and saturated fatty
acids [mmol/l]
ω-3/ω-7, ω-9, and saturated
fatty acids
Monounsaturated fatty acids
[mmol/l]
Other polyunsaturated fatty
acids than linoleic acid
[mmol/l]
Average methylene groups in
fatty acid chain
Average methylene groups
per double bond
9
Average double bonds per
fatty acid chain
1.3 (0.079)
1.3 (0.077)
0.15
0.53 (0.031)
0.53 (0.028)
0.11
0.67 (0.076)
0.66 (0.072)
0.12
18 (0.18)
18 (0.17)
0.27
Ratio of bisallylic groups to
double bonds
Ratio of bisallylic groups to
total fatty acids
Average fatty acid chain
length
Baseline metabolite levels of the study population. Values are mean (SD) for normally
distributed variables and median [25th - 75th percentile] for variables with skewed
distributions. P-values for comparison with t-tests and Kolmogorov-Smirnov tests were
adjusted for sex, age, and body mass index.
10
Table S2: Odds ratios for 6-year incident high IMT and/or plaque for all assayed small
molecules and serum extract metabolites
OR
95% CI
P
Glutamine
1.38
(1.13-1.68)
0.001
Histidine
1.23
(1.02-1.47)
0.03
Tyrosine
1.33
(1.10-1.60)
0.003
Alanine
1.18
(0.98-1.42)
0.09
Isoleucine
1.02
(0.80-1.30)
0.85
Leucine
1.07
(0.88-1.30)
0.50
Phenylalanine
1.02
(0.84-1.24)
0.86
Valine
1.03
(0.84-1.27)
0.77
Lactate
1.11
(0.95-1.30)
0.19
Pyruvate
1.11
(0.94-1.33)
0.22
Glycerol
1.02
(0.83-1.25)
0.86
Citrate
0.92
(0.76-1.12)
0.39
α-1 acid glycoprotein
1.05
(0.83-1.32)
0.68
3-hydroxybutyrate
1.09
(0.90-1.31)
0.38
Acetate
0.96
(0.79-1.16)
0.66
Acetoacetate
1.10
(0.93-1.31)
0.28
Creatinine
0.92
(0.75-1.14)
0.44
Urea
0.95
(0.79-1.13)
0.55
Esterified cholesterol
1.38
(1.03-1.85)
0.03
ω-6 fatty acids
1.29
(1.01-1.65)
0.04
ω-3/ω-6
0.81
(0.67-0.98)
0.03
Linoleic acid
1.32
(1.05-1.65)
0.02
Docosahexaenoic acid
0.74
(0.60-0.92)
0.007
Total fatty acids
1.11
(0.80-1.54)
0.55
Free cholesterol
1.12
(0.85-1.49)
0.42
Total phosphoglycerides
1.00
(0.74-1.34)
0.98
Phosphatidylcholine
0.95
(0.74-1.21)
0.68
Sphingomyelins
1.03
(0.82-1.30)
0.80
ω-3 fatty acids
0.89
(0.73-1.08)
0.24
ω-7, ω-9 and saturated fatty acids
1.01
(0.73-1.40)
0.96
ω-3/ω-7, ω-9, and sat. fatty acids
0.85
(0.70-1.03)
0.10
Small molecules
Serum extract metabolites
11
Monounsaturated fatty acids
1.01
(0.73-1.40)
0.95
0.86
(0.68-1.09)
0.20
1.03
(0.85-1.25)
0.76
1.11
(0.89-1.40)
0.35
0.87
(0.69-1.10)
0.23
0.87
(0.71-1.08)
0.21
fatty acids
0.87
(0.69-1.09)
0.23
Average fatty acid chain length
0.87
(0.71-1.08)
0.20
Other polyunsaturated fatty acids
than linoleic acid
Average methylene groups in fatty
acid chain
Average methylene groups per
double bond
Average double bonds per fatty acid
chain
Ratio of bisallylic groups to double
bonds
Ratio of bisallylic groups to total
Odds ratios (OR) and 95% confidence intervals (CI) for incident carotid IMT≥90 th percentile
or plaque at follow-up (2007) according to metabolite measures at baseline (2001). Odds
ratios are adjusted for sex, baseline age, body mass index, systolic blood pressure, glucose,
total-C, HDL-C, and triglycerides. Values are expressed per 1-SD increase in the predictor
variable. Metabolites shown in italic were included in the selection procedure for derivation
of prediction models.
12
Table S3. Characteristics of the additional populations studied for cross-sectional validation
of tyrosine and glutamine with carotid IMT and presence of coronary artery disease
Subset of the
Age [yr]
Angiography and
Cardiovascular Risk
Health 2000 subset
Genes Study
Genes Study
in Young Finns
(n=1044)
Non-CAD patients
Presence of CAD
(n=374)
(n=593)
Study† (n=830)
Male sex [%]
Angiography and
49 (45-52)
41 (38-44)
45 (40-51)
76 (72-79)
33.5 (5.8)
57.7 (7.9)
59.2 (10.0)
64.5 (9.49)
25.7 (4.8)
27.1 (4.6)
28.0 (4.6)
28.1 (4.1)
120 (14)
137 (22)
142 (20)
145 (23.2)
27 (24-30)
23 (20-25)
14 (11-18)
14 (11-17)
5.1 (1.0)
5.7 (0.9)
4.6 (1.0)
4.4 (1.0)
1.3 (0.3)
1.6 (0.4)
1.3 (0.4)
1.1 (0.3)
1.4 (0.9)
1.3 (0.9)
1.5 (1.1)
1.6 (1.2)
Body mass index
[kg/m2]
Systolic blood pressure
[mm Hg]
Current smoker [%]
Total-C [mmol/l]
HDL-C [mmol/l]
Triglycerides [mmol/l]
†: Individuals not included in the prospective analyses since they only attended one of the field-study surveys.
13
Table S4: Predictive performance of metabolite quantification in addition to conventional
lipids.
The predictive performance of the NMR-based lipoprotein measures and metabolite
biomarkers beyond the Framingham Risk Score including conventional lipids was assessed. A
prediction model was derived in a similar manner to the Extended model presented in Table
4, however with the conventional lipids total-C and HDL-C were forced here into the model.
In this analysis, the conventional lipids were accompanied by NMR-based LDL-C, IDL, small
LDL and docosahexaenoic acid as well as tyrosine also remained in the model. Similar
predictive performance was obtained as for the Extended model in Table 4, thus confirming
the augmented predictive value of metabolite quantification. On the other hand, the slightly
lower AUC obtained when including total-C and HDL-C in the prediction model in
comparison to the more parsimonious Extended model suggest no additional benefit of
standard lipid testing in combination with the employed high-throughput metabolite
quantification experimentation.
Model
AUC
PAUC
NRI [%]
PNRI
IDI [%]
-
-
-
-
0.06
16.4
0.002
3.2
PIDI
(95% CI)
Reference model: age, sex, systolic
0.737
BP, smoking status, glucose,
(0.699-0.775)
total-C, HDL-C
Age, sex, systolic BP, smoking
status, glucose, total-C, HDL-C +
0.761
LDL-CNMR, IDLNMR, small LDLNMR,
(0.722-0.799)
docosahexaenoic acid, tyrosine
P-values are in comparison to the Reference model.
-
<0.0001
14
Table S5: Odds ratios and predictive performance of alternative outcomes representing
increased subclinical atherosclerosis
Outcome
(population size/
Prevalence of plaque
Prevalence of carotid IMT
Prevalence of carotid IMT
(n=1692/50)
≥ 90th percentile
≥ 90th percentile and/or
(n=1692/169)
plaque
number of events)
(n=1692/205)
Metabolite measure
OR [95% CI] P-value
OR [95% CI] P-value
OR [95% CI] P-value
1.84 [1.42-2.38] P<0.0001
1.09 [0.92-1.29] P=0.31
1.24 [1.06-1.44] P=0.006
1.82 [1.41-2.35] P<0.0001
1.14 [0.97-1.35] P=0.12
1.29 [1.10-1.50] P=0.001
1.61 [0.81-3.18] P=0.17
1.12 [0.80-1.59] P=0.50
1.13 [0.81-1.56] P=0.48
0.88 [0.63-1.24] P=0.47
1.20 [0.97-1.50] P=0.10
1.39 [1.11-1.74] P=0.004
1.94 [1.46-2.58] P<0.0001
1.05 [0.77-1.43] P=0.76
0.83 [0.68-1.01] P=0.06
1.04 [090-1.21] P=0.58
1.18 [1.01-1.38] P=0.04
1.14 [0.95-1.36] P=0.15
0.79 [0.65-0.96] P=0.02
0.84 [0.70-1.01] P=0.06
1.08 [0.94-1.25] P=0.25
1.24 [1.07-1.44] P=0.005
1.30 [1.10-1.53] P=0.002
0.85 [0.72-1.02] P=0.08
1.52 [1.18-1.95] P=0.001
⠀ Apolipoprotein A-1
Major lipoprotein fraction (NMR)
Total cholesterol
1.87 [1.46-2.40] P<0.0001
IDL-cholesterol
1.81 [1.41-2.33] P<0.0001
LDL-cholesterol
2.01 [1.54-2.61] P<0.0001
HDL-cholesterol
0.89 [0.62-1.28] P=0.53
Total triglycerides
1.54 [1.18-2.01] P=0.002
VLDL-triglycerides
1.43 [1.09-1.87] P=0.009
IDL-triglycerides
1.69 [1.31-2.18] P<0.0001
Lipoprotein subclasses (NMR)
Extremely large VLDL
1.07 [0.82-1.38] P=0.64
Very large VLDL
1.16 [0.90-1.50] P=0.26
Large VLDL
1.25 [0.97-1.62] P=0.09
Medium VLDL
1.46 [1.12-1.90] P=0.005
Small VLDL
1.99 [1.52-2.62] P<0.0001
Very small VLDL
1.70 [1.33-2.17] P<0.0001
IDL
1.76 [1.37-2.26] P<0.0001
Large LDL
1.94 [1.50-2.50] P<0.0001
Medium LDL
2.03 [1.56-2.65] P<0.0001
Small LDL
1.91 [1.46-2.49] P<0.0001
Very large HDL
0.82 [0.57-1.19] P=0.30
Large HDL
0.67 [0.44-1.02] P=0.06
Medium HDL
0.86 [0.61-1.21] P=0.37
Small HDL
1.37 [1.05-1.79] P=0.02
Small molecules (NMR)†
Glutamine
1.47 [1.06-2.03] P=0.02
Histidine
1.10 [0.807-1.49] P=0.55
Tyrosine
1.45 [1.06-1.98] P=0.02
Serum extract metabolites (NMR)†
Esterified cholesterol
1.78 [1.20-2.64] P=0.004
1.23 [1.03-1.46] P=0.019
1.24 [1.06-1.44] P=0.006
1.16 [0.98-1.37] P=0.08
1.19 [1.01-1.40] P=0.04
1.29 [1.09-1.52] P=0.003
0.87 [0.70-1.08] P=0.20
0.98 [0.82-1.16] P=0.78
0.98 [0.82-1.15] P=0.77
0.98 [0.83-1.17] P=0.84
1.31 [1.13-1.53] P=0.0004
1.31 [1.12-1.52] P=0.0005
1.45 [1.24-1.70] P<0.0001
0.88 [0.72-1.07] P=0.19
1.09 [0.94-1.28] P=0.25
1.08 [0.93-1.27] P=0.31
1.08 [0.92-1.26] P=0.34
0.98 [0.84-1.14] P=0.77
0.99 [0.84-1.16] P=0.88
0.98 [0.83-1.15] P=0.81
0.99 [0.84-1.17] P=0.92
1.09 [0.91-1.30] P=0.36
1.02 [0.86-1.20] P=0.86
1.14 [0.97-1.35] P=0.12
1.24 [1.05-1.47] P=0.01
1.27 [1.08-1.51] P=0.005
1.23 [1.04-1.46] P=0.02
0.96 [080-1.19] P=0.72
0.83 [0.66-1.05] P=0.12
0.78 [0.63-0.94] P=0.01
0.93 [0.78-1.12] P=0.46
1.01 [0.87-1.17] P=0.90
1.05 [0.91-1.21] P=0.53
1.06 [0.92-1.23] P=0.41
1.11 [0.95-1.29] P=0.19
1.27 [1.08-1.49] P=0.004
1.12 [0.96-1.30] P=0.15
1.25 [1.07-1.45] P=0.004
1.39 [1.19-1.62] P<0.0001
1.45 [1.24-1.69] P<0.0001
1.39 [1.19-1.63] P<0.0001
0.94 [0.77-1.14] P=0.51
0.79 [0.64-0.98] P=0.03
0.78 [0.65-0.94] P=0.009
1.04 [0.88-1.22] P=0.67
1.26 [1.04-1.53] P=0.02
1.18 [0.99-1.41] P=0.06
1.36 [1.14-1.62] P=0.0008
1.30 [1.09-1.55] P=0.004
1.17 [1.00-1.38] P=0.05
1.37 [1.16-1.62] P=0.0002
1.26 [0.93-1.70] P=0.13
1.42 [1.09-1.85] P=0.009
Lipoprotein and lipid measures (laboratory)
Total cholesterol
LDL-cholesterol
⠀ (Friedewald)
LDL-cholesterol
⠀ (direct measure)
HDL-cholesterol
Total triglycerides
Total-C/HDL-C
Apolipoprotein B
Apolipoprotein A-1
Apolipoprotein B/
15
ω-6 fatty acids
1.75 [1.24-2.49] P=0.002
1.17 [0.91-1.49] P=0.22
1.28 [1.02-1.60] P=0.030
ω-3/ω-6 fatty acids
1.12 [0.88-1.44] P=0.36
0.88 [0.74-1.04] P=0.14
0.914 [0.78-1.07] P=0.25
Linoleic acid
1.65 [1.20-2.28] P=0.002
1.19 [0.95-1.50] P=0.12
1.29 [1.05-1.58] P=0.02
Docosahexaenoic acid
1.01 [0.758-1.35] P=0.93
0.83 [0.68-1.01] P=0.06
0.83 [0.70-1.00] P=0.05
Comparison of prediction models based on discrimination and reclassification indices
Reference model: age, sex,
systolic BP, smoking,
glucose, total-C, HDL-C
Extended model: nonlaboratory risk factors*,
glucose, LDL-CNMR,
medium HDL, docosahexaenoic acid and tyrosine
AUC [95% CI]
P-value
NRI; P-value
IDI; P-value
AUC [95% CI]
P-value
NRI; P-value
IDI; P-value
AUC [95% CI]
P-value
NRI; P-value
IDI; P-value
0.757 [0.695-0.818]
0.749 [0.713-0.786]
0.754 [0.721-0.787]
0.751 [0.688-0.814]
P=0.61
0.8%; P=0.75
0.6%; P=0.19
0.776 [0.740-0.811]
P=0.04
14.5%; P=0.003
2.8%; P<0.0001
0.780 [0.748-0.812]
P=0.001
14.5%; P=0.0005
3.3%; P<0.0001
Odds ratios were adjusted for sex, baseline age, body mass index, systolic blood pressure, smoking status and
glucose. †Small molecules and serum extract metabolite associations were further adjusted for total-C, HDL-C,
and triglycerides. Odds ratios are expressed for 1-SD increase in the metabolites.
16
Table S6: Odds ratios and predictive performance of progression and regression of IMT
Outcome
Association measure
Progression of carotid IMT
ΔIMT ≥ 80th percentile
(11.3µm) n=1681/327
Regression of carotid IMT
ΔIMT < 0 µm
n=1681/496
OR [95% CI] P-value
OR [95% CI] P-value
Lipoprotein and lipid measures (laboratory)
Total cholesterol
1.04 [0.91-1.18] P=0.59
1.09 [0.96-1.25] P=0.20
LDL-cholesterol (Friedewald)
LDL-cholesterol (direct)
1.13 [0.91-1.40] P=0.26
HDL-cholesterol
0.79 [0.68-0.91] P=0.002
Total triglycerides
1.07 [0.93-1.22] P=0.32
Total-C/HDL-C
1.20 [1.05-1.37] P=0.008
Apolipoprotein B
1.17 [1.02-1.35] P=0.03
Apolipoprotein A-1
0.90 [0.79-1.04] P=0.15
Apolipoprotein B/
1.20 [1.04-1.38] P=0.01
Apolipoprotein A-1
Major lipoprotein fraction (NMR)
Total cholesterol
1.07 [0.94-1.21] P=0.32
IDL-cholesterol
1.09 [0.96-1.24] P=0.17
LDL-cholesterol
1.15 [1.01-1.31] P=0.034
HDL-cholesterol
0.79 [0.68-0.93] P=0.004
Total triglycerides
1.08 [0.94-1.24] P=0.29
VLDL-triglycerides
1.07 [0.93-1.22] P=0.36
IDL-triglycerides
1.05 [0.92-1.19] P=0.51
Lipoprotein subclasses (NMR)
Extremely large VLDL
1.00 [0.88-1.14] P=0.96
Very large VLDL
1.04 [0.92-1.19] P=0.52
Large VLDL
1.05 [0.92-1.20] P=0.51
Medium VLDL
1.07 [0.94-1.23] P=0.30
Small VLDL
1.24 [1.07-1.42] P=0.003
Very small VLDL
1.07 [0.94-1.22] P=0.33
IDL
1.07 [0.94-1.22] P=0.33
Large LDL
1.13 [0.99-1.28] P=0.073
Medium LDL
1.15 [1.01-1.31] P=0.04
Small LDL
1.12 [0.98-1.28] P=0.09
Very large HDL
0.84 [0.72-0.98] P=0.03
Large HDL
0.76 [0.64-0.90] P=0.002
Medium HDL
0.83 [0.72-0.95] P=0.009
Small HDL
0.98 [0.86-1.11] P=0.72
Small molecules (NMR)†
Glutamine
1.12 [0.97-1.29] P=0.14
Histidine
1.12 [0.98-1.27] P=0.09
Tyrosine
1.11 [0.97-1.28] P=0.14
Serum extract metabolites (NMR)†
Esterified cholesterol
1.17 [0.94-1.46] P=0.16
ω-6 fatty acids
1.12 [0.94-1.35] P=0.21
ω-3/ω-6 fatty acids
0.87 [0.75-1.00] P=0.05
Linoleic acid
1.14 [0.96-1.36] P=0.13
Docosahexaenoic acid
0.87 [0.75-1.01] P=0.07
0.86 [0.76-0.98] P=0.02
0.88 [0.77-0.99] P=0.04
0.99 [0.82-1.19] P=0.91
0.99 [0.87-1.13] P=0.88
0.90 [0.79-1.03] P=0.12
0.87 [0.75-1.00] P=0.05
0.83 [0.72-0.95] P=0.006
0.98 [0.86-1.11] P=0.73
0.86 [0.74-0.99] P=0.03
0.83 [0.73-0.94] P=0.002
0.84 [0.74-0.95] P=0.005
0.81 [0.71-0.92] P=0.001
0.97 [0.85-1.10] P=0.63
0.91 [0.80-1.05] P=0.19
0.94 [0.82-1.07] P=0.33
0.94 [0.83-1.06] P=0.29
0.89 [0.78-1.02] P=0.10
0.89 [0.78-1.02] P=0.09
0.93 [0.81-1.06] P=0.25
0.94 [0.82-1.07] P=0.35
0.86 [0.75-0.99] P=0.03
0.89 [0.78-1.00] P=0.06
0.86 [0.76-0.97] P=0.01
0.83 [0.73-0.94] P=0.003
0.81 [0.71-0.92] P=0.001
0.79 [0.69-0.90] P=0.0004
0.92 [0.81-1.05] P=0.22
1.02 [0.89-1.17] P=0.75
1.02 [0.91-1.15] P=0.74
0.97 [0.87-1.09] P=0.65
0.87 [0.764-0.99] P=0.04
0.86 [0.76-0.97] P=0.01
0.89 [0.78-1.01] P=0.08
0.80 [0.64-1.00] P=0.05
0.90 [0.76-1.07] P=0.25
1.02 [0.91-1.16] P=0.70
0.90 [0.76-1.07] P=0.22
1.01 [0.88-1.16] P=0.87
Comparison of prediction models based on discrimination and reclassification indices
Model
AUC [95% CI] P-value
NRI; P-value
AUC [95% CI] P-value
NRI; P-value
17
Reference model: age, sex,
systolic BP, smoking, glucose,
total-C, HDL-C
Extended model: non-lab risk
factors*, glucose, LDL-CNMR,
medium HDL, docosahexaenoic acid and tyrosine
IDI; P-value
IDI; P-value
0.710 [0.680-0.740]
0.745 [0.720-0.770]
0.710 [0.680-0.741] P=0.85
1.7%; P=0.52
0.6%; P=0.07
0.748 [0.723-0.773] P=0.37
4.5%; P=0.004
0.7%; P=0.003
Odds ratios and 95% confidence intervals adjusted for sex, baseline age, body mass index, systolic blood
pressure, smoking status, glucose, and baseline carotid IMT. †Small molecules and serum extract metabolite
associations were further adjusted for total-C, HDL-C and triglycerides. Odds ratios are expressed for 1-SD
increase in the metabolites.
18
Predictive performance using alternative cutpoints to denote high IMT
Because there is no clinical consensus on what signifies high IMT, we examined predictive
performance of the models using alternate cut-points (≥80th-96th percentile) to define high
IMT. The results are illustrated in Figure S2. Improvements in AUC reached statistical
significance in the range of IMT≥82th-94th percentile. NRI was generally above 10% with
respect to the reference model.
Figure S2. Area under the ROC curve (AUC) and net reclassification index (NRI) for
different percentiles used as cut-point to denote high IMT.
Error bars for AUC denote standard error. * and ** denote P<0.05 and P<0.005,
respectively, for comparison of AUC and NRI of the extended model with respect to the
reference model.
19
Predictive performance with random division of the study population for split risk
algorithm calculation and comparison of risk discrimination
In order to mimic validation of the risk assessment in an independent cohort we derived the
risk algorithm weighting for the two prediction models in one random half of the study
population (n=798) and compared the predictive performance in the other half of the study
population. Median risk prediction metrics for 100 random divisions of the study population
into training and validation parts are shown in Table S6. The same variables were included in
the models for all comparisons. Essentially similar predictive performance was achieved in
terms of absolute increase in AUC and NRI as the results for the whole study population
presented in Table 4. However, due to the smaller size of the validation set (n=797) the
comparison of AUC to the reference model was not statistically significant (PAUC=0.09).
Table S7. Predictive performance with random division of the study population for split
risk algorithm calculation and comparison of risk discrimination
Model
AUC
PAUC
NRI
PNRI
[%]
Reference model: age, sex, systolic BP,
smoking status, glucose, total-C, HDL-C
IDI
PIDI
[%]
0.733
(0.678-0.787)
-
-
-
-
-
0.757
(0.704-0.814)
0.09
15.4
0.04
3.04
0.003
Extended model: non-laboratory risk factors*,
glucose, LDL-CNMR, medium HDL,
docosahexaenoic acid and tyrosine
P-values are for comparison of the extended model with respect to the reference model.
AUC = area under receiver-operating characteristic curve;
NRI = net reclassification index; IDI = integrated discrimination improvement.
* Non-laboratory risk factors: age, sex, systolic BP, smoking status.
20
Table S7.
21
References:
1.
Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concentration of low-
density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge.
Clin Chem. 1972;18:499-502.
2.
Raitakari OT, Juonala M, Rönnemaa T, Keltikangas-Järvinen L, Räsänen L, Pietikäinen
M, Hutri-Kähönen N, Taittonen L, Jokinen E, Marniemi J, Jula A, Telama R, Kähönen M,
Lehtimäki T, Åkerblom HK, Viikari JS. Cohort profile: the cardiovascular risk in Young Finns
Study. Int J Epidemiol. 2008;37:1220-1226.
3.
Tukiainen T, Tynkkynen T, Mäkinen VP, Jylänki P, Kangas A, Hokkanen J, Vehtari A,
Gröhn O, Hallikainen M, Soininen H, Kivipelto M, Groop PH, Kaski K, Laatikainen R, Soininen
P, Pirttila T, Ala-Korpela M. A multi-metabolite analysis of serum by 1H NMR spectroscopy:
early systemic signs of Alzheimer's disease. Biochem Biophys Res Commun. 2008;375:356361.
4.
Soininen P, Kangas AJ, Würtz P, Tukiainen T, Tynkkynen T, Laatikainen R, Järvelin MR,
Kähönen M, Lehtimäki T, Viikari J, Raitakari OT, Savolainen MJ, Ala-Korpela M. Highthroughput serum NMR metabonomics for cost-effective holistic studies on systemic
metabolism. Analyst. 2009;134:1781-1785.
5.
Inouye M, Kettunen J, Soininen P, Silander K, Ripatti S, Kumpula LS, Hämäläinen E,
Jousilahti P, Kangas AJ, Männistö S, Savolainen MJ, Jula A, Leiviskä J, Palotie A, Salomaa V,
Perola M, Ala-Korpela M, Peltonen L. Metabonomic, transcriptomic, and genomic variation
of a population cohort. Mol Syst Biol. 2010; 6:441.
6.
Okazaki M, Usui S, Ishigami M, Sakai N, Nakamura T, Matsuzawa Y, Yamashita S.
Identification of unique lipoprotein subclasses for visceral obesity by component analysis of
cholesterol profile in high-performance liquid chromatography. Arterioscler Thromb Vasc
Biol. 2005;25:578-584.
7.
Vehtari A, Mäkinen VP, Soininen P, Ingman P, Mäkelä SM, Savolainen MJ,
Hannuksela ML, Kaski K, Ala-Korpela M. A novel Bayesian approach to quantify clinical
variables and to determine their spectroscopic counterparts in 1H NMR metabonomic data.
BMC Bioinformatics. 2007;8 Suppl 2:S8.
8.
Niemi J, Mäkinen VP, Heikkonen J, Tenkanen L, Hiltunen Y, Hannuksela ML,
22
Jauhiainen M, Forsblom C, Taskinen MR, Kesäniemi YA, Savolainen MJ, Kaski K, Groop PH,
Kovanen PT, Ala-Korpela M. Estimation of VLDL, IDL, LDL, HDL(2), apoA-I, and apoB from the
Friedewald inputs-apoB and IDL, but not LDL, are associated with mortality in type 1
diabetes. Ann Med. 2009;41:451-461.
9.
Jeyarajah EJ, Cromwell WC, Otvos JD. Lipoprotein particle analysis by nuclear
magnetic resonance spectroscopy. Clin Lab Med. 2006;26:847-870.
10.
Ala-Korpela M, Soininen P, Savolainen MJ. Letter by Ala-Korpela et al regarding
article, "Lipoprotein particle profiles by nuclear magnetic resonance compared with
standard lipids and apolipoproteins in predicting incident cardiovascular disease in women".
Circulation. 2009;120:e149; author reply e150.
11.
Sipilä K, Moilanen L, Nieminen T, Reunanen A, Jula A, Salomaa V, Kaaja R, Kukkonen-
Harjula K, Lehtimäki T, Kesäniemi YA, Koivistoinen T, Nieminen MS, Tuomilehto J, Kähönen
M. Metabolic syndrome and carotid intima media thickness in the Health 2000 Survey.
Atherosclerosis. 2009;204:276-281.
12.
Samani NJ, Raitakari OT, Sipilä K, Tobin MD, Schunkert H, Juonala M, Braund PS,
Erdmann J, Viikari J, Moilanen L, Taittonen L, Jula A, Jokinen E, Laitinen T, Hutri-Kähönen N,
Nieminen MS, Kesäniemi YA, Hall AS, Hulkkonen J, Kähönen M, Lehtimäki T. Coronary artery
disease-associated locus on chromosome 9p21 and early markers of atherosclerosis.
Arterioscler Thromb Vasc Biol. 2008;28:1679-1683.
13.
Wang TJ, Larson MG, Vasan RS, Cheng S, Rhee EP, McCabe E, Lewis GD, Fox CS,
Jacques PF, Fernandez C, O'Donnell CJ, Carr SA, Mootha VK, Florez JC, Souza A, Melander O,
Clish CB, Gerszten RE. Metabolite profiles and the risk of developing diabetes. Nat Med
2011;17:448-53.
14.
Shah SH, Bain JR, Muehlbauer MJ, Stevens RD, Crosslin DR, Haynes C, Dungan J,
Newby LK, Hauser ER, Ginsburg GS, Newgard CB, Kraus WE. Association of a peripheral
blood metabolic profile with coronary artery disease and risk of subsequent cardiovascular
events. Circ Cardiovasc Genet 2010;3:207-214.
15.
Mennander A, Kuukasjärvi P, Laurikka J, Nikus K, Karhunen PJ, Tarkka M, Lehtimäki T.
Diagnostic performance of plasma high sensitive C-reactive protein in detecting three-vessel
coronary artery disease: modification by apolipoprotein E genotype. Scand J Clin Lab Invest
23
2008;68:714-719.
16.
Raitoharju E, Seppälä I, Levula M, Kuukasjärvi P, Laurikka J, Nikus K, Huovila AP,
Oksala N, Klopp N, Illig T, Laaksonen R, Karhunen PJ, Viik J, Lehtinen R, Pelto-Huikko M,
Tarkka M, Kähönen M, Lehtimäki T. Common variation in the ADAM8 gene affects serum
sADAM8 concentrations and the risk of myocardial infarction in two independent cohorts.
Atherosclerosis 2011;218:127-133.
17.
Hosmer DW, Lemeshow S. Applied Logistic Regression. 2nd ed. New York, NY: John
Wiley & Sons Inc; 2000.
24