386

Relation of Aortic Stiffness to Factors Modifyingthe Risk of Atherosclerosis in Healthy People

Markku Kupari, Pauli Hekali, Pekka Keto, Veli-Pekka Poutanen,Matti J. Tikkanen, Carl-Gustaf Standertskjold-Nordenstam

Abstract To identify factors predicting aortic stiffness, westudied the modulus of elasticity of the thoracic aorta inrelation to sex, obesity, blood pressure, physical activity,smoking, ethanol consumption, salt intake, and serum lipidand insulin levels in 55 healthy people born in 1954. Atransverse cine magnetic resonance image of the thoracic aortawas made, and the modulus of elasticity was determined asbrachial artery cuff pulse pressure/aortic strain, where strainwas determined as the ratio of pulsatile aortic luminal areachange to the diastolic luminal area. The average of measure-ments made in the ascending and descending aorta was used asthe elastic modulus of the thoracic aorta. Habitual physicalactivity, smoking, and alcohol use were quantified by 2-monthprospective daily recording and salt intake by 7-day foodrecords. The aortic elastic modulus ranged from 100 to 2091103 dyne/cm2 (median, 390 103 dyne/cm2). In multiple regres-

The stiffness of the thoracic aorta influences aorticconduit function, contributes to blood pressureand left ventricular load, and may also modify

the aortocoronary blood flow.13 People with coronaryartery disease have abnormally rigid aortas,3"6 andnoninvasive measurement of aortic distensibility hasbeen considered for a targeted screening of coronaryatheroma.7 For these reasons it has become timely toknow what factors modify aortic pulsatility and in whatway. It is generally agreed that the stiffness of the aortaincreases with age and in hypertension,'-2-6-10 but knowl-edge of other potential predictors is insufficient andpartly contradictory.6'813 Concerning the relation ofaortic stiffness to blood cholesterol level, some reportsshow a positive association,1112 others a negative one,613

and the rest no association.810 Although part of thesedifferences may be attributed to the modifying effect ofage,11-13 confusing contrasts between the data6-12 stillremain.

The purpose of this study was to examine the predic-tors of aortic stiffness in a sample of the general adultpopulation homogeneous for age and free of significantcardiovascular disease. In principle, aortic stiffness canbe assessed noninvasively either by studying the relationbetween pulsatile changes in blood pressure and aorticluminal size or more indirectly by measuring the pulsewave velocity through the thoracoabdominal aorta.1 In

Received June 18, 1993; revision accepted December 8, 1993.From the Division of Cardiology (First Department of Medi-

cine) and the Department of Radiology, Helsinki (Finland) Uni-versity Central Hospital.

Reprint requests to Dr Markku Kupari, Division of Cardiol-ogy, Helsinki University Central Hospital, FIN-00290 Helsinki,Finland.

sion analyses, log,0 aortic elastic modulus was related directlyto mean blood pressure (standardized coefficient [/3]=.37,/)=.002), serum high-density lipoprotein cholesterol (/3=.36,P=.O12), square root of daily energy expenditure in physicalactivity (/3=.33, P=.005), and log10 serum insulin (/3=.27,/*=.047) and inversely to serum low-density lipoprotein cho-lesterol (0=-.26, P=.O35). A relation to salt intake was alsoobserved, but the regression slope was dependent on meanblood pressure (P=.005 for interaction). These data suggestthat many modifiable constitutional and lifestyle characteris-tics may contribute to the stiffness of the thoracic aorta.(Arterioscler Thromb. 1994;14J86-394.)

Key Words • aortic stiffness • magnetic resonanceimaging • lipids • insulin • physical activity • sodiumintake

this study, we used cine magnetic resonance imaging(MRI) to measure the systolic and diastolic cross-sectional areas of the ascending and descending tho-racic aorta and calculated the modulus of elasticity14 asan index of aortic stiffness. The present report describesthe relations of aortic stiffness to those factors known tomodify the risk of cardiovascular diseases, such as sex,obesity, blood lipid levels, physical activity, smoking,alcohol consumption, sodium intake, and serum insulinlevel.

Methods

Study PopulationThis work was carried out as a substudy to a comprehensive

cardiovascular assessment of a sample of the population livingin Helsinki and born in 1954. Of the target population (3730men, 4250 women), 120 randomly selected individuals wereinvited, and 93 (78%) took part in the assessment. Of them, 55people (31 men) underwent an MRI examination of thethoracic aorta; they constitute the population of this study. All93 members of the primary sample could not be included inthe present work because of the limited MRI resourcesallocated to our use.

The participants were aged 36 to 37 years at the time of ourstudy. The mean height and weight (±SD) of the 55 subjectsstudied by MRI were 172±9 cm and 71 ±13 kg, respectively.Their body mass index averaged 23.9±3.4 kg/m2 (range, 17.3 to34.4 kg/m2). All were free of cardiovascular and metabolicdiseases according to the medical history, physical examina-tion, 12-lead electrocardiogram, a complete cardiac ultra-sound study, and laboratory tests including peripheral bloodcount, fasting blood glucose, serum creatinine, and liver en-zymes. Several subjects had been in follow-up for borderlineblood pressure elevation, but none had been diagnosed ashaving hypertension or had used drugs for it. Fifteen partici-

by guest on October 15, 2017

http://atvb.ahajournals.org/D

ownloaded from

Kupari et al Predictors of Aortic Stiffness 387

pants had a positive family history of coronary heart disease,and 13 had a family history of hypertension.

Study DesignThe study started with a prospective 2-month collection of

lifestyle data. Each subject was given a pocket diary forrecording daily physical activity and consumption of alcoholand tobacco. Another diary was given for a 7-day food recorddesigned to estimate dietary salt intake. The subjects wereasked not to change their lifestyle or diet because of partici-pation in the study. After an average of 65 days of lifestylerecording (range, 46 to 74 days), the participants returned thediaries and underwent determination of aortic stiffness byMRI and measurement of serum lipid and insulin levels. Theprotocol was approved by the institutional ethics committee.

Assessment of Physical Activity, Alcohol Use,Smoking, and Sodium Intake

The subjects recorded each day the type of leisure-timephysical activity they had practiced and the exact time spent init. The energy expended was calculated as suggested by Wilsonet a l" by multiplying the metabolic equivalent value (MET; 1MET=the energy expended by a person at rest, ie, ^ l kcal/kgper hour) of each activity by the time (hours) spent in it. TheMET values of different activities were adopted from previ-ously published tables.15 The average energy expenditure perday of follow-up was used as a physical activity index. Sup-porting its utility, the index (MET • h/d) correlated negativelyin the primary population sample with body mass index(r=-.27, P=.O1O) and resting heart rate ( /=-.23, />=.O39)and positively with daily energy intake by the food consump-tion records (see below) (r=.24, P=.O28). Occupational phys-ical activities were not included in this index but were esti-mated by questioning at the subjects' initial assessment.Depending on the type of daily work, the average occupationalphysical workload was arbitrarily graded as minimal, light,moderate, or heavy.

The types and amounts of all alcoholic beverages and thenumber of cigarettes consumed each day were also enteredinto the diary. The drinks were converted to grams of absoluteethanol, added up, and divided by body weight and by thenumber of follow-up days to obtain an estimate of dailyethanol consumption (grams per kilogram). Supporting thevalidity of the data, the daily ethanol consumption correlateddirectly with serum concentrations of -y-glutamyltransferase(r=.28, />=.0O9) and high-density lipoprotein cholesterol(HDL-C) (sex-adjusted partial r=.26, P=.O19) in our primarypopulation sample. The average daily cigarette consumptionwas calculated in the same way.

The data on salt intake were collected by means of a 7-dayfood record designed and analyzed by an expert nutritionist.The recording days were distributed evenly over the 2-monthfollow-up period, and each day of the week was covered once.The subjects were given written and illustrated instructions toensure correct data entry into the food records. The subjectsreturned the records personally to the nutritionist, whochecked the food consumption entries for adequacy and askedcomplementary questions if necessary. The records wereanalyzed for energy and sodium intake using commercialsoftware (UNILEVER DIETARY ANALYSIS PROGRAM, UnileverInc) and a database on nutrient and mineral composition ofFinnish foods.16 The average sodium intake (milliequivalentsper day) was calculated for the purposes of the present work.Earlier research in our country has shown that sodium intakecalculated from food records can be used to estimate saltintake.17

Laboratory TestsVenous blood was sampled after an overnight fast. Serum

total cholesterol was determined by a commercial kit (Boeh-ringer) and the concentration of triglycerides in an Autoana-

lyzer II (Technicon Instruments). HDL-C was separated fromserum by precipitation of the other lipoproteins with heparin-manganese chloride. The supernatant was further fractionatedinto HDL r C and HDLj-C by precipitation with 0.11% dextransulfate.18 Low-density lipoprotein cholesterol (LDL-C) wascalculated using the Friedewald equation.19 Serum insulin wasmeasured with a commercial radioimmunoassay kit (Pharma-cia). The measurements were made in the same batch of theassay to eliminate any interassay variation.

MRI Study of Aortic DistensibilityThe MRI studies were made using a 1.0-T superconducting

unit (Siemens Magnetom 42 SP), a body coil, and electrocar-diogram triggering. For examination of the pulsatile changes ofthe cross-sectional luminal areas of the ascending and de-scending thoracic aorta, a cine examination was acquired in aplane transecting the aorta transversally at the level of thepulmonary artery bifurcation (Fig 1). A two-dimensionalgradient echo sequence was used with a repetition time of 50milliseconds and an echo time of 12 milliseconds; the flip anglewas 30°, the matrix size 128x256, and the slice thickness 7 mm.Brachial artery systolic, diastolic, and pulse pressures wereaveraged over two measurements made by the cuff method(Korotkoff phases I and V) just before and after the MRIstudy. Mean blood pressure was calculated as diastolic pres-sure plus one third pulse pressure.

The MRI studies were analyzed without knowledge of thesubject under assessment. The smallest diastolic and largestsystolic circumferences of the ascending and descending tho-racic aorta were traced with a mouse-driven cursor in anoff-line image-analysis system (Radgop/wiz, Contextvision,Struers Vision AB). The aortic luminal areas were determinedby multiplying the number of pixels by the pixel size. Aorticstrain was determined as the change of the aortic luminal areafrom its diastolic minimum to its systolic maximum divided bythe diastolic area (Fig 1). The aortic elastic modulus14 wascalculated in dynes per square centimeter as 1332 • pulsepressure/strain, where the coefficient converts millimeters ofmercury to dynes per square centimeter. The mean of valuesof elastic modulus in the ascending and descending aorta wascalculated as an index of the stiffness of the thoracic aorta foreach individual.

For assessment of the repeatability of analyzing the MRIexaminations, eight studies were measured twice in a blindedmanner. The absolute difference of the paired data as apercentage of their average was 2.1 ±2.0% (mean±SD) for thediastolic area of the ascending aorta, 3.0±3.1% for the dia-stolic area of the descending aorta, 21.8±21.0% for thepulsatile area change in the ascending aorta, and 22.5±20.6%for the area change in the descending aorta. The reproducibil-ity of the modulus of elasticity was 25.0± 18.8% in the ascend-ing aorta, 24.6±23.4% in the descending aorta, and 16.2±11.1% for the average value in the thoracic aorta. To quantifythe reproducibility of serial MRI examinations, the aorticimaging was made twice 1 week apart in eight healthy people.The difference between the two studies was 7.9±14.0% for thediastolic area of the ascending aorta, 11.7±13.3% for the areaof the descending aorta, 19.5±22.5% for the pulsatile areachange in the ascending aorta, and 38.8±37.0% for the areachange in the descending aorta. The reproducibility of themodulus of elasticity was 26.3±23.3% in the ascending aorta,34.5±32.1% in the descending aorta, and 25.1 ±20.2% for theaverage value in the thoracic aorta.

StatisticsAll variables were tested for normal distribution by the

Kolmogorov-Smirnov one-sample test. Group means werecompared by the Student's t test or ANOVA. The values ofelastic modulus in the ascending and descending thoracic aortawere compared with the Student's paired t test. MultivariateANOVA was used to assess whether the present study group

by guest on October 15, 2017

http://atvb.ahajournals.org/D

ownloaded from

388 Arteriosclerosis and Thrombosis Vol 14, No 3 March 1994

FIG 1. End-diastolic (left) and end-systolic (right) transverse images of ascending and descending thoracic aorta (AA and DA,respectively) at the level of the pulmonary artery (PA) bifurcation. Diastolic and systolic areas of the ascending aorta measured 5.6 and7.0 cm2, respectively (strain, 0.25). Corresponding areas of the descending aorta were 3.5 and 4.1 cm2 (strain, 0.18).

differed from those members of our population sample whowere not studied by MRI. Bivariate correlation coefficientswere calculated by the Pearson product-moment method(continuous versus continuous variables) or by Spearman'srank method (continuous versus discrete variables). The inde-pendent relations of the aortic elastic modulus to its potentialpredictors were analyzed by stepwise multiple linear regres-sion. Because the relations were curved in bivariate data plots,logarithmic values of elastic modulus were used. The data onethanol use, physical activity, and serum insulin and triglycer-ide levels were grossly skewed toward high values and wereincluded as their logarithms or square roots in the analyses. Totest for interactions between the explanatory variables, we ranadditional analyses by including appropriate product terms inthe model. The associations are reported as multiple regres-sion coefficients (b) (±SEM) between logl0 aortic elasticmodulus and the independent variables. Standardized (unit-independent) regression coefficients (/3) and squared multiplecorrelation coefficients (R1) were also calculated. The otherdata are given as mean±SD unless indicated otherwise.Adjusted values for elastic modulus were calculated using themultiple regression coefficients. Values of P<.05 were consid-ered statistically significant. All calculations were performedon a microcomputer using commercially available statisticalsoftware (SYSTAT version 5.2, Systat Inc).

ResultsCharacteristics of the Subjects

Table 1 summarizes the blood pressure measure-ments, aortic cross-sectional areas, biochemical data,and most lifestyle variables in the study group. Themodulus of elasticity averaged 572±735 If? dyne/cm2

(median, 285) in the ascending aorta and 436±306 103

dyne/cm2 (median, 390) in the descending thoracicaorta. The difference was not statistically significant(P=.89O for log-transformed data), and all further analy-ses were made using the average elastic modulus of thethoracic aorta as an index of aortic stiffness. Thirty-three subjects were nonsmokers, 12 smoked 1 to 10cigarettes per day, and 10 smoked more than 10 ciga-

rettes per day. The occupational physical workload wasconsidered minimal in 25 individuals, light in 15, mod-erate in 13, and heavy in 2. A multivariate ANOVA onbody mass index, systolic and diastolic blood pressures,square root physical activity index, sodium intake,square root ethanol consumption, serum HDL-C, se-rum LDL-C, logio serum triglycerides, and log10 seruminsulin gave an overall F value of 0.97 (P=M) betweenthe study group (n=55) and those members (n=38) ofthe population sample who could not be examined byMRI. No statistically significant group differences werefound in any individual variable.

Univariate and Multivariate Correlates of theElastic Modulus of the Thoracic Aorta

The modulus of elasticity of the thoracic aorta aver-aged 508±414 103 dyne/cm2 in the study group. The datadistribution was skewed toward high values (P=.001),with a median of 390 103 dyne/cm2 and a range of 100 to2091 103 dyne/cm2. The log,0 elastic modulus averaged5.62±0.27 in men and 5.58±0.32 in women (P=.6O),5.60±0.32 in people with a family history of hyperten-sion and 5.60±0.26 in those without (P=.99), and5.53±0.26 in people with a family history of coronaryheart disease and 5.61 ±0.29 in those without (P=.35).The logio elastic modulus was not related to smoking,either: the average was 5.62±0.30 in nonsmokers,5.58±0.24 in people smoking 1 to 10 cigarettes per day,and 5.59±0.25 in those smoking more than 10 cigarettesper day (P=J1). The other univariate associations withlog,0 aortic elastic modulus are summarized in Table 2.

Stepwise multiple linear regression analyses weremade with the following explanatory variables: sex(male=l, female=2), body mass index, mean bloodpressure, smoking (nonsmoker=0,1 to 10 cigarettes perday=l, >10 cigarettes per day=2), occupational physi-cal workload (minimal = 0, light = l, moderate = 2,heavy=3), square root physical activity index, square

by guest on October 15, 2017

http://atvb.ahajournals.org/D

ownloaded from

Kupari et al Predictors of Aortic Stiffness 389

TABLE 1. Characteristics of the Study Population (N=55)

Characteristic

Systolic blood pressure, mm Hg

Diastolic blood pressure, mm Hg

Mean blood pressure, mm Hg

Pulse pressure, mm Hg

Luminal area of ascending thoracic aorta, cm2

Diastolic

Systolic

Luminal area of descending thoracic aorta, cm2

Diastolic

Systolic

Serum LDL-C, mmol/L

Serum HDL-C, mmol/L

Serum HDLj-C, mmol/L

Serum HDL3-C, mmol/L

LDL-HDL cholesterol ratio

Serum triglycerides, mmol/L

Serum insulin, mU/L

Ethanol consumption, (g/kg)/d

Sodium intake, mEq/d

Physical activity index, MET • h/d

Mean±SD

117±11

79±16

92±7

38±8

6.2±1.2

7.2±1.2

3.3±0.7

3.8+0.9

2.81 ±0.86

1.72±0.41

0.66±0.28

1.06±0.22

1.78±0.78

1.15±0.79

6.5±4.8

0.31 ±0.28

160 + 52

2.6±2.4

Median

118

78

90

38

6.3

7.3

3.3

3.7

2.79

1.69

0.62

1.04

1.70

0 82

4.8

0.18

148

2.1

Range

100-148

68-95

80-106

24-60

3.6-9.2

4.8-9.9

2.3-6.1

2.7-7.4

1.24-4.79

1.06-3.15

0 02-1.22

0.72-2.03

0 47-3.76

0.42-4.20*

1.5-30.4*

0-1.04*

47-291

0-13.8*

LDL-C indicates low-density lipoprotein cholesterol; HDL-C, high-density lipoproteln cholesterol; and MET,metabolic equivalent.

*Data distribution statistically significantly skewed (P<.01) according to Kolmogorov-Smirnov test.

root daily ethanol consumption, daily sodium intake,serum LDL-C and HDL-C, and log10 values of serumtriglyceride and insulin levels. The analyses showed thatLDL-C, HDL-C, mean blood pressure, daily physicalactivity, and serum insulin levels were statistically sig-nificant independent correlates of log,0 aortic elasticmodulus. Their effects are summarized in Table 3,which also shows that the LDL-HDL cholesterol ratiocould be substituted for LDL-C and HDL-C withoutany loss in the explanatory power of the equation. WhenHDL-C was replaced by its subfractions (one at a time)in model 1 of Table 3, aortic elastic modulus was relatedto HDL r C (j3=.42, P=.004) but not to HDL3-C (/3=.13,P=.3O). Fig 2 illustrates further the relations of log,0aortic elastic modulus to the LDL-HDL ratio, meanblood pressure, physical activity, and serum insulin. Todisplay independent associations, the data on log]0

aortic elastic modulus were adjusted for the effects ofthe other predictors using the coefficients shown inTable 3.

Tables 2 and 3 show that the univariate and multi-variate test results agreed relatively well, apart from thelack of univariate association between aortic elasticmodulus and serum insulin levels. Statistically this wasdue to an inverse relation of log10 serum insulin withHDL-C (r=-.43, P=.000). When the effect of HDL-Con log10 aortic elastic modulus was adjusted for, thepartial correlation coefficient between log10 aortic elas-tic modulus and log,o serum insulin rose from .06(P=.70) to .29 (P=Ml). It is also noteworthy that log10

aortic elastic modulus showed a borderline correlation

with both the physical activity index and occupationalphysical workload in univariate analysis, but only theformer association was statistically significant in multi-ple regression analysis. There was no association be-tween aortic size (diastolic area of the ascending aorta)and either leisure-time physical activity (r=.003, P=.99)or the grade of occupational physical workload (r=.O8,P=.57).

Additional analyses were run to identify any interac-tional effects between the explanatory variables on theaortic elastic modulus. A statistically significant interac-tion between blood pressure and sodium intake wasuncovered. Table 4 shows the results of adding sodiumintake and its product term with mean blood pressure tothe regression equation of model 2 in Table 3. Theexplanatory power of the model rose from 50% to 59%.Fig 3 is a three-dimensional illustration of the effects ofmean blood pressure and sodium intake on the elasticmodulus of the thoracic aorta. No other interactionswere observed.

DiscussionOur study shows that the stiffness of the thoracic

aorta is related to blood pressure, leisure-time physicalactivity, and serum lipid and insulin levels in peoplehomogeneous for age and free of detectable cardiovas-cular disease. Dietary salt may also be of significance,but this association is less clear and appears to dependon blood pressure. Totally 50% to 60% of the varianceof aortic elastic modulus in our study group could beexplained by the above factors. Part of the remaining

by guest on October 15, 2017

http://atvb.ahajournals.org/D

ownloaded from

390 Arteriosclerosis and Thrombosis Vol 14, No 3 March 1994

TABLE 2. Blvarlate Correlation Coefficients BetweenLog10 Aortic Elastic Modulus and SelectedSubject Characteristics

Characteristic r

Body mass index, kg/m2

Systolic blood pressure, mm Hg

Diastolic blood pressure, mm Hg

Mean blood pressure, mm Hg

Square root physical activity index,

Occupational physical workload*

Square root daily ethanol consumption,

Serum HDL-C, mmol/L

Serum HDLrC, mmol/L

Serum HDL3-C, mmol/L

Serum LDL-C, mmol/L

LOL-HDL cholesterol ratio

Log10 serum triglycerides, mmol/L

Log)c) serum Insulin, mU/L

r indicates bivariate correlation coefficient; P, statistical signif-icance of the correlation; MET, metabolic equivalent; HDL-C,high-density lipoprotein cholesterol; and LDL-C, low-densitylipoprotein cholesterol.

*Values given to occupational physical workload: minimal=0,light=1, moderated, heavy=3.

variation probably reflects the random error of theelastic modulus measurements, but certainly there canbe factors (eg, genetic or neuroendocrine) we did notassess that also contribute.

Methodological ConsiderationsThe subjects of this study were derived from a ran-

dom sample of the population born in 1954 and living inHelsinki. The primary sample was representative of thetarget population because the participation rate wasclose to 80% (93 of 120 people).20 Although the sub-groups studied (n=55) and not studied (n = 38) by MRIwere not statistically significantly different regarding the

.02

.47

.32

.46

.26

.28

.14

.39

.35

.28

- .34

-.39

- .19

.06

.887

.001

.024

.002

.068

.052

.321

.005

.012

.046

.015

.005

.187

.703

characteristics of interest in this work, there was a malepredominance in the studied subgroup compared withthe target population. The age homogeneity of thesubjects, which was intended to help avoid the con-founding effect of age in the analyses, is another limi-tation to the applicability of our data.

We used cine MRI of the thoracic aorta with nonin-vasive brachial artery blood pressure measurements tostudy aortic stiffness. Mohiaddin et al21 first reportedthe utility of MRI, and we have used this techniquepreviously to study aortic distensibility in Marfan'ssyndrome.22 The applicability of brachial cuff bloodpressures to the assessment of aortic stiffness has beenverified earlier.2-23 We used the pulsatile change in theaortic luminal area instead of change in the diameter tocalculate aortic strain, and therefore our data representthe volume elastic modulus per unit length of theaorta.14 It is worth noting that although the pressure-volume and pressure-diameter relations reflect arterialstiffness in the same way,14 the values of elastic modulusby the area change are one half of the values calculatedby the diameter change.24

The reproducibility of the aortic luminal areas in ourexperience (2.1% to 7.9%) compares well with the 6%reproducibility of area measurements given by Mohiad-din et al.21 The data on pulsatile area change and elasticmodulus were less reproducible, however, because thetrue area change is small (on average, 15% of thediastolic area) compared with the random measurementerror. The less than ideal reproducibility of the mea-surements should not detract from the validity of ourmain findings, however, because the random nature ofthe variation acts to weaken the possibility of identify-ing true associations rather than to create spuriousones. Furthermore, this limitation is not unique to MRI;Isnard et al2 reported that the reproducibility of twoultrasound measurements was 2.4% to 3.6% for aorticdiameters but up to 23% for the pulsatile diameterchange. Because the values of the elastic modulus in theascending and descending segments of the thoracicaorta were not significantly different, we could use theirmean, which had better reproducibility (see "Meth-

TABLE 3. Multiple Regression Models Showing Independent Linear Correlates of Log10 Aortic Elastic Modulus in theStudy Population

Variable*

Mean blood pressure, mm Hg

LDL-C, mmol/L

HDL-C, mmol/L

LDL-HDL cholesterol ratio

Square root physical activity index, VMET • h/d

Log10 serum insulin, mU/L

Constant

R*

b±SEM

.016±.005

-.089±.041

.253±.O97

.140+.047

.314 + .154

3.52

.49

Model 1

P.37

-.26

.36

.33

.27

P

.002

.035

.012

.005

.047

b±SEM

.019±.OO5

-.196±.O43

.150+.046

.289±.141

3.81

.50

Model 2

P.44

- .52

.36

.24

P

.000

.000

.002

.046

b indicates multiple linear regression coefficient; p, standardized regression coefficient; LDL-C, low-density lipoprotein cholesterol;HDL-C, high-density lipoprotein cholesterol; MET, metabolic equivalent; and ft2, square of the multiple correlation coefficient An ellipsisindicates that LDL-HDL cholesterol ratio was included interchangeably with LDL-C and HDL-C in the analysis.

*Sex, body mass index, serum triglyceride levels, smoking, ethanol consumption, occupational physical workload, and sodium intakewere also included In the analyses but had no statistically significant independent influence on aortic elastic modulus.

by guest on October 15, 2017

http://atvb.ahajournals.org/D

ownloaded from

Kupari et al Predictors of Aortic Stiffness 391

10000

2 looo

a.a

u

100

10

10000

log10(y)=-0.20«x+5.96r=-0.59. p=0.000

10000 r

a

•a2 IOOO

100

o

10

log(y)=0.02'x+3.89r=0.53. p=0.000

1 2 3

LDL/HDL-rmtio

75 85 95 105

Mean blood prenure (mmHg)

115

tr-Su

2 1000m99

100

o

10

• •

: . •

log10(y)=0.15'xu+5.39r-0.45. p=0.001

10000

a|•o

2 iooo

3

"8aS 100

1 4 9

Phyiicil activity Index (MET/day)

16 10

log,,(y)=0.29'x+5.40r=0.32. p=0.046

0.0 05 1.0

Log,, icrum insulin in mU/1

1.5

FIG 2. Scatterplots show independent relations of the modulus of elasticity of the thoracic aorta to the ratio of low-density tohigh-density lipoprotein cholesteral (LDL/HDL) (top left), mean blood pressure (top right), physical activity Index (bottom left), and seruminsulin (bottom right). In each panel, the data on elastic modulus have been adjusted to the means of the other Independent variablesin the study group using the multiple regresston coefficients of model 2 in Table 3. Note that the scale of elastic modulus (y axis in eachplot) Is logarithmic and the scale of physical activity index (x axis in bottom left panel) Is square-root transformed. MET indicatesmetabolic equivalent.

ods") in identifying the predictors of aortic stiffness inour study population.

Potential Predictors of Aortic Elastic ModulusThe effects of age and blood pressure on aortic

stiffness have been established previously.1-3610-21-23 Theage effect was deliberately eliminated in our study, butthe influence of blood pressure was clearly seen eventhough nearly all subjects were normotensive at the timeof the MRI study. The association of aortic elasticmodulus with the prevailing mean blood pressure isunderstandable because the passive elastic properties ofany arterial wall (including aortic) dictate that its dis-tensibility decreases as the distending pressure in-creases.1 Whether there are sex differences in aorticstiffness has been a matter of disagreement. Some

earlier studies showed that the pulse wave velocitythrough the descending aorta was higher in men than inwomen,25-26 suggesting stiffer aortas in men, but otherreports have not found sex differences in either thepulse wave velocity8-9-27 or the indexes of aortic stiffnessbased on direct imaging of aortic pulsations.6-28 In ourstudy, the aortic elastic modulus was independent ofsex, and there also were no sex interactions, indicatingthat the regressions given in Tables 3 and 4 were notstatistically significantly different between men andwomen.

Aortic stiffness was unrelated to body mass index,smoking, and ethanol consumption in our study popula-tion. No earlier data on aortic stiffness in relation to thesefactors exist. Smoking and obesity may reduce the disten-sibility of the peripheral arteries,29-30 but the composition

by guest on October 15, 2017

http://atvb.ahajournals.org/D

ownloaded from

392 Arteriosclerosis and Thrombosis Vol 14, No 3 March 1994

TABLE 4. Independent Linear Correlates of Log,0 Aortic Elastic Modulus in a Model IncludingSodium Intake and Its Interaction With Mean Blood Pressure

Variable

Mean blood pressure, mm Hg

LDL-HDL cholesterol ratio

Square root physical activity index, VMET • h/d

Logio serum insulin, mU/L

Sodium intake, mEq/d

Mean blood pressure • sodium intake, mm Hg • mEq/d

Constant

R2

b±SEM

.072 ±.018

-.204±.044

.179±.O49

.382±.137

.027+.009

-.0003±.0001

-1.12

.59

P

.000

.000

.001

.008

.006

.005

b indicates multiple linear regression coefficient; LDL, low-density lipoprotein; HDL, high-density lipopro-tein; MET, metabolic equivalent; and R2, square of the multiple correlation coefficient. The data are multiplelinear regression coefficients constituting an equation to predict log,0 aortic elastic modulus from theexplanatory variables

of the arterial walls is also different, with the relativeamounts of elastin, collagen, and smooth muscle changingmarkedly from the aorta to the periphery.31

The higher aortic stiffness in physically more activepeople is surprising and disagrees at first sight with thedata of Mohiaddin et al21 of increased aortic volumecompliance in active athletes. However, unlike us, Mo-hiaddin et al did not adjust the pulsatile aortic areachange to the diastolic size of the aorta (which may havebeen increased in the athletes), and therefore their dataare not directly comparable with ours. Furthermore, itcannot be excluded that athletes have a relatively highaortic compliance for constitutional reasons rather thanas a result of physical training. There is no factualexplanation of how the thoracic aorta could becomemore stiff with physical activity, but possibly the regu-larly recurring increases of blood pressure with exercise

FK3 3. Computer-generated three-dimensional model showsthe association of log10 aortic elastic modulus with mean bloodpressure and sodium intake. The model represents the equationz= -0.95+0.072 • x+0.027 • y-0.000295 • x • y, where x is meanblood pressure, y is sodium intake, and z is log10 aortic elasticmodulus adjusted for the effects of the ratio of low-density tohigh-density lipoprotein cholesterol, physical activity, and seruminsulin level. The model is based on the regression coefficientsshown in Table 4.

could be of importance. We emphasize, however, thatthe associations found in this study, including the pre-dictive value of physical activity, are cross-sectional:they may suggest causal relations but are no proofthereof. The lack of precise estimates of physical fitness,such as maximal oxygen uptake, is also a limitation ofthese analyses.

The association of aortic elastic modulus with bloodlipid levels was the most conspicuous finding in our study:The higher the LDL-C and the lower the HDL-C, the lessthe rigidity of the thoracic aorta. This accords well withthe findings of Dart et al6 showing an inverse associationof aortic stiffness with serum cholesterol (total) but dis-agrees with the very recent data of Hopkins et al12

suggesting, by contrast, a positive association of aorticstiffness with serum cholesterol (LDL-C and the LDL-C/HDL-C ratio) in healthy adults. However, when the effectof age was taken into account in the study by Hopkins etal, LDL-C lost its significance, and the positive relation ofaortic stiffness to LDL-C/HDL-C was evident in menonly.12 Other studies in Asian89 or Western10 populationshave found no association between aortic stiffness andblood cholesterol. In familial hypercholesterolemia, aorticstiffness has been reported to be reduced in youth13 butincreased in adults." As suggested previously by Lehmannet al,11 these seemingly contradictory data can be recon-ciled in part by the known behavior of aortic stiffness inexperimental atherosclerosis. Newman et al32 and Farraret al33 have shown that in the early stages of diet-inducedatheromatosis in animals, when cholesterol accumulationand foam cells predominate in the intimal lesions, theaorta becomes more distensible than normal, only tostiffen later when the lesions acquire an increased amountof collagen and calcification. Our study group was rela-tively young and free of signs of cardiovascular disease. Itis therefore possible that our subjects with higher LDL-Cand lower HDL-C levels had less stiff aortas because theyhad early nonsclerotic atheromatous changes in the aorticwall. The relation of aortic elastic modulus to HDL2-C butnot to HDL3-C supports this idea; the former subtractionis known to associate better with the risk of atherosclero-sis. If measurement of aortic stiffness is used to screenpeople with hypercholesterolemia, as has been consid-ered,7 the complexity of these relations must be kept inmind. Although a stiff aorta increases the likelihood of

by guest on October 15, 2017

http://atvb.ahajournals.org/D

ownloaded from

Kupari et al Predictors of Aortic Stiffness 393

concomitant atherosclerotic aortic and coronary involve-ment, reduced stiffness is no guarantee against the devel-opment of cardiovascular disease.

One of the novel findings of our study was the positiveassociation of the modulus of aortic elasticity to seruminsulin level. Neutel et al34 have shown previously thatperipheral arterial compliance is inversely related to se-rum insulin in hypertensive men. On the other hand,Hopkins et al12 found no association of aortic distensibilitywith serum insulin in healthy people. Our data and thefindings of Neutel et al34 accord with epidemiologic obser-vations incriminating insulin as a risk factor of atheroscle-rosis.35 The way insulin links up with arterial changes is notknown in detail, but the possible mechanisms includeeffects that promote connective tissue synthesis and pro-liferation of smooth muscle in the vessel wall. Receptorsfor insulin and insulin-like growth factor have been iden-tified in the aortic wall.36 Nonetheless, Hopkins et al12

found no age-independent relation between aortic disten-sibility and serum insulin-like growth factor-I in healthypeople. In diabetes, the overall rigidity of the thoracoab-dominal aorta is increased, at least in adults with non-insulin-dependent disease,37 but whether the ascending ordescending thoracic aorta or both are stiffened has notbeen studied.

The studies of Avolio et al9-27 have suggested thatpeople with a low sodium intake may have less stiffaortas than those consuming salt in higher amounts.,Inour study, the association of aortic stiffness with saltintake was more complex because the regression wasmodified by the level of blood pressure. As Fig 3 shows,the association of aortic elastic modulus with habitualsodium intake was strongly positive in people with lowmean blood pressure but faded away in people withblood pressure that was higher but still within thenormotensive range. Because of the relatively smallstudy group, the interaction must be viewed with cau-tion. We interpret the data only as suggesting thatdietary salt intake is one potential predictor of aorticstiffness and that its effect may depend on prevailingblood pressure.

ImplicationsOur data suggest that blood pressure, physical activ-

ity, serum lipid and insulin levels, and possibly sodiumintake are potential predictors of the stiffness of thethoracic aorta in the general population. Although theseassociations remain to be shown as causal, theystrengthen the idea of aortic distensibility being suscep-tible to therapeutic or lifestyle interventions. Althoughsome loss of distensibility may be predestined to occurwith aging, undue stiffening may be preventable orreversible. As a key determinant of systolic blood pres-sure and left ventricular load, the stiffness of thethoracic aorta may become a new therapeutic target inthe future.

AcknowledgmentsThis study was supported by a grant from the Foundation for

Alcohol Studies, Helsinki, Finland. We are grateful to MarkkuVentila, MSc, Marjaana Lahti-Koski, MSc, Tuovi Pekuri, RN,and Soili Sirenius, RN, for advice and skillful assistance duringthe study.

References1. Simon AC, O'Rourke MF, Levenson J. Arterial distensibility and

its effect on wave reflection and cardiac loading in cardiovasculardisease. Coronary Artery Dis. 1991;2:1111-1120.

2. Isnard RN, Pannier BM, Laurent S, London GM, Diebold B, SafarME. Pulsatile diameter and elastic modulus of the aortic arch inessential hypertension: a noninvasive study. J Am Coil Cardiol.1989;13:399-405.

3. Bogren HG, Mohiaddin RH, Klipstein RK, Firmin DN, UnderwoodRS, Rees SR, Longmore DB. The function of the aorta in ischemicheart disease: a magnetic resonance and angiographic study of aorticcompliance and blood flow patterns. Am Heart J 1989;118:234-247.

4. Stefanadis C, Wooley CF, Bush CA, Kolibash AJ, Boudoulas H.Aortic distensibility abnormalities in coronary artery disease. AmJ Cardiol 1987;59:1300-1304.

5. Hirai T, Sasayama S, Kawasaki T, Yagi S. Stiffness of systemicarteries in patients with myocardial infarction: a noninvasivemethod to predict severity of coronary atherosclerosis. Circulation.1989;80:78-86.

6. Dart AM, Lacombe F, Yeoh JK, Cameron JD, Jennings GL,Laufer E, Esmore DS. Aortic distensibility in patients with isolatedhypercholesterolemia, coronary artery disease, or cardiactransplant. Lancet. 1991,338:270-273.

7. Aortic distensibility and screening for coronary atheroma. Lancet.1991;338:288. Editorial.

8. Avolio AP, Chen S, Wang R, Zhang C, Li M, O'Rourke MF.Effects of aging on changing arterial compliance and left ven-tricular load in a northern urban Chinese community. Circulation.1983;68:50-58.

9. Avolio AP, Fa-Quan D, Wei-Qiang L, Yao-Fei L, Zhen-Dong H,Lian-Fen X, O'Rourke MF. Effects of aging on arterial distensi-bility in populations with high and low prevalence of hypertension:comparison between urban and rural communities in China. Cir-culation 1985;71:202-210.

10. Nichols WW, O'Rourke MF. Aging, high blood pressure anddisease in humans. In: Nichols WW, O'Rourke MO, eds.McDonald's Blood Flow in Arteries Theoretical, Experimental andClinical Principles. London, England: Edward Arnold; 1990:398-420.

11. Lehmann ED, Watts GF, Gosling RG. Aortic distensibility andhypercholesterolemia. Lancet. 1992;340:l 171-1172. Letter.

12. Hopkins KD, Lehmann ED, Gosling RG, Parker JR, Sonksen PH.Biochemical correlates of aortic distensibility m vivo in normalsubjects. Clin Sci. 1993;84593-597.

13. Lehmann ED, Watts GF, Fatemi-Langroud B, Gosling RG. Aorticcompliance in young patients with heterozygous familial hypercho-lesterolemia. Clm Sci. 1992,83:717-721.

14. Peterson LH, Jensen RE, Parnell J. Mechanical properties ofarteries in vivo. Circ Res. 1960,8:622-639.

15. Wilson PWF, Paffenbarger RS, Morris JN, Havlik RJ. Assessmentmethods for physical activity and physical fitness in populationstudies: report of a NHLBI workshop. Am Heart J. 1986;111:1177-1192.

16. Rastas M, Seppanen R, Knuts L-R, Karvetti R-L, Varo P. NutrientComposition of Foods. 2nd ed. H5meenlinna, Finland: Karisto;1989.1-457.

17. Pietinen P. Estimating sodium intake from food consumption data.Ann Nutr Metab. 1982;26:90-99.

18. Gidez LJ, Miller GJ, Burstein M, Eder HA. Analyses of plasmahigh-density lipoprotein subclasses by a precipitation procedure:correlations with preparative and analytical ultracentrifugation. In:LJppel K, ed. Report of the High Density Lipoprotein MethodologyWorkshop. 1979:328. NIH publication No 79-1661.

19. Friedewald WT, Levy Rl, Fredrickson DS. Estimation of the con-centration of low density lipoprotein in plasma without use of thepreparative ultracentrifuge. Clm Chem. 1972; 18:499-502.

20. Evans SJW. Good surveys guide. Br Med J. 1991^302:302-303.21. Mohiaddin RH, Underwood SR, Bogren HG, Firmin DN,

Klipstein RH, Rees RSO, Longmore DB. Regional aortic com-pliance studied by magnetic resonance imaging: the effects of age,training, and coronary artery disease. Br Heart J. 1989;62:90-96.

22. Savolainen A, Keto P, Hekali P, Nisula L, Kaitila I, Viitasalo M,Poutanen V-P, Standertskj6ld-Nordenstam C-G, Kupari M. Aorticdistensibility in children with the Marfan syndrome. Am J Cardiol1992; 70:691-693.

23. Stefanadis C, Stratos C, Boudoulas H, Kourouklis C, Toutouzas P.Distensibility of the ascending aorta: comparison of invasive andnoninvasive techniques in healthy men and in men with coronaryartery disease. Eur Heart J. 1990;! 1.-990-996.

by guest on October 15, 2017

http://atvb.ahajournals.org/D

ownloaded from

394 Arteriosclerosis and Thrombosis Vol 14, No 3 March 1994

24. Hickler RB. Aortic and large artery stiffness: current methodologyand clinical correlations. Clin Cardiol. 1990;13:317-322.

25. Schimmler W. Untersuchungen zu Elastizitatsproblemen derAorta. Arch Kreislaufforschung. 1965;47:189-205.

26. Laogun AA, Gosling RG. In vivo arterial compliance in man. ClinPhys PhysiolMeas. 1982;3:201-212.

27. Avolio AP, Clyde KM, Beard TC, Cooke HM, Ho KKL, O'RourkeMF. Improved arterial distensibility in normotensive subjects on alow-salt diet. Arteriosclerosis. 1986;6:166-169.

28. Lacombe F, Dart A, Dewar E, Jennings G, Cameron J, Laufer E.Arterial elastic properties in man: a comparison of echo-Dopplerindices of aortic stiffness. Eur Heart J. 1992;13:1040-1045.

29. Toto-Moukou JJ, Achimastos A, Asmar RG, Hugues CJ, SafarME. Pulse wave velocity in patients with obesity and hypertension.Am Heart J. 1986;112:136-140.

30. Levenson J, Simon AC, Cambien FA, Beretti C. Cigarette smokingand hypertension: factors independently associated with blood hyper-viscosity and arterial rigidity. Arteriosclerosis. 1987;7:572-577.

31. Nichols WW, O'Rourke MF. Properties of the arterial wall. In:Nichols WW, O'Rourke MF, eds. McDonald's Blood Flow in

Arteries. Theoretical, Experimental and Clinical Principles. London,England: Edward Arnold; 1990:77-124.

32. Newman DL, Gosling RG, Bowden NLR. Changes in aortic dis-tensibility and area ratio with the development of atherosclerosis.Atherosclerosis. 1971;14:231-240.

33. Farrar DJ, Bond G, Riley WA, Sawyer JK. Anatomic correlates ofaortic pulse wave velocity and carotid artery elasticity during ath-erosclerosis progression and regression in monkeys. Circulation.1991;83:1754-1763.

34. Neutel JM, Smith DHG, Graettinger WF, Weber MA. Dependencyof arterial compliance on circulating neuroendocrine and metabolicfactors in normal subjects. Am J Cardiol. 1992;69:1340-1344.

35. Pyorala K. Relationship of glucose tolerance and plasma insulin tothe incidence of coronary heart disease: results from two popu-lation studies in Finland. Diabetes Care. 1979;2:131-141.

36. King GL, Goodman AD, Buzney S, Moses A, Kahn CR. Receptorsand growth-promoting effects of insulin and insulin-like growthfactor on cells from bovine retinal capillaries and aorta. J ClinInvest. 1985;75:1028-1036.

37. Lehmann ED, Gosling RG, Sonksen PH. Arterial wall compliancein diabetes. Diabetic Med. 1992;9:114-119.

by guest on October 15, 2017

http://atvb.ahajournals.org/D

ownloaded from

M Kupari, P Hekali, P Keto, V P Poutanen, M J Tikkanen and C G Standerstkjöld-NordenstamRelation of aortic stiffness to factors modifying the risk of atherosclerosis in healthy people.

Print ISSN: 1079-5642. Online ISSN: 1524-4636 Copyright © 1994 American Heart Association, Inc. All rights reserved.

Avenue, Dallas, TX 75231is published by the American Heart Association, 7272 GreenvilleArteriosclerosis, Thrombosis, and Vascular Biology

doi: 10.1161/01.ATV.14.3.3861994;14:386-394Arterioscler Thromb Vasc Biol. 

http://atvb.ahajournals.org/content/14/3/386World Wide Web at:

The online version of this article, along with updated information and services, is located on the

  http://atvb.ahajournals.org//subscriptions/

at: is onlineArteriosclerosis, Thrombosis, and Vascular Biology Information about subscribing to Subscriptions:

  http://www.lww.com/reprints

Information about reprints can be found online at: Reprints: 

document.Permissions and Rights Question and AnswerFurther information about this process is available in theis being requested is located, click Request Permissions in the middle column of the Web page under Services.Clearance Center, not the Editorial Office. Once the online version of the published article for which permission

can be obtained via RightsLink, a service of the CopyrightArteriosclerosis, Thrombosis, and Vascular Biology Requests for permissions to reproduce figures, tables, or portions of articles originally published inPermissions:

by guest on October 15, 2017

http://atvb.ahajournals.org/D

ownloaded from