Cross-sectional associations between physical activity and selected coronary heart disease risk factors in young adults. The Cardiovascular Risk in Young Finns Study

Abstract

Objective. Physical activity (PA) may reduce the risk of coronary heart disease (CHD) by inducing beneficial changes in several risk factors. We studied the associations between PA and a range of risk markers of CHD in young adults.

Methods and results. We measured serum lipoproteins, oxidized LDL, adipokines, inflammatory markers, metabolic markers, and arginine metabolites in 2,268 individuals (age 24–39 y). Participants were asked frequency, duration, and intensity of PA in leisure time. In addition, commuting to work was assessed. In both sexes, PA was inversely associated with waist circumference (all P < 0.0001). After controlling for sex, age, and waist circumference, PA was directly associated with HDL-cholesterol and apolipoprotein A1, and inversely with heart rate, smoking, oxidized LDL, apolipoprotein B, insulin, glucose, C-reactive protein, leptin, L-arginine, and phospholipase A2 activity (all P < 0.05).

Conclusion. These population-based data are consistent with the idea that the beneficial effects of PA on CHD risk are mediated by favorable influences on several risk factors, as judged by independent relations to markers of lipoprotein metabolism, glucose metabolism, and inflammation. These associations reflect beneficial effects on cardiovascular health in both sexes and may offer mechanistic insights for the inverse association between PA and CHD.

Key words::

• CHD
• physical activity
• risk factors

Abbreviations
ADMA	=	asymmetric dimethyl arginine
apoA1	=	apolipoprotein A1
apoB	=	apolipoprotein B
BMI	=	body mass index
BP	=	blood pressure
CHD	=	coronary heart disease
CRP	=	C-reactive protein
HOMA-IR	=	homeostasis model assessment of insulin resistance
MET	=	metabolic equivalent index
oxLDL	=	oxidized LDL
PA	=	physical activity
SAA	=	serum amyloid A
SDMA	=	symmetric dimethyl arginine
sPLA2	=	phospholipase A2 activity

Key messages

• This population-based study shows that the beneficial effects of physical activity on cardiovascular risk are mediated by favorable influences on several risk factors, as shown by independent relations to markers of lipoprotein metabolism, glucose metabolism, and inflammation.

• These associations offer mechanistic insights for the inverse association between physical activity and coronary heart disease.

Introduction

Physical activity (PA) is inversely related with coronary heart disease (CHD). The protective mechanism may include multiple influences on cardiovascular risk factors. High levels of PA are generally associated with lower body fat mass (1). Beneficial influences on lipid profile have been reported to be partly mediated by fat mass, but an independent direct relation between PA and HDL-cholesterol is well established (2,3). Furthermore, higher levels of PA are related with increased insulin sensitivity (4)—an effect entirely or partly mediated by changes in fat mass (5). Current evidence also support that PA may reduce cardiovascular risk by reducing levels of inflammatory markers: increased PA leads to lower circulating levels of C-reactive protein (6) and leptin (7), and higher levels of adiponectin (8). There is a lack of population-based data on the relations of PA and new cardiovascular risk factors, such as phospholipase A2 activity, asymmetric dimethyl arginine (ADMA) and amyloid-A that may also increase cardiovascular risk by contributing to the vascular inflammation.

The aim of this analysis was to study the associations between PA and selected CHD risk factors, including new biomarkers with a limited amount of population-based data.

Material and methods

Subjects

The Cardiovascular Risk in Young Finns Study is a follow-up study of atherosclerosis risk factors. The first cross-sectional survey was conducted in 1980. The original sample size was 4,320 children and adolescents aged 3, 6, 9, 12, 15, and 18 years. There were 3,596 participants (83.2% of those invited) in 1980. The details of the study design have been published (9).

This study is based on the 21-year follow-up data which were collected in 2001. A total of 3,456 subjects were invited; 2,620 subjects (aged 24–39 y) from the original cohort, i.e. 75.8%, returned the questionnaire, and 2,268 (54% women) participated in the clinical study (65.6%). Not all of the subjects who returned the questionnaire took part in the clinical study, and therefore the number of subjects varies between 1,951 and 2,268 (maximum number of subjects was used in all analyses). Clinical studies and PA data collections were performed at the same time period during the winter months to minimize seasonal variation. However, the participants were asked to estimate their usual amount of leisure time PA, and commuting to work was assessed by taking into account both winter- and summer-times. We found no significant variation in PA across the months of data collection period (between October and January) (data not shown). Participants gave written informed consent, and the study was approved by local ethics committees.

Clinical examination

Weight was measured, in light clothing without shoes, using digital scales with an accuracy of 0.1 kg, and height was measured by a wall-mounted statiometer with 0.1 cm accuracy. Body mass index (BMI) was calculated with the formula:

BMI = weight (kg)/[height (m)]².

Waist circumference was measured midway between iliac crest and the lowest rib at the mid-axillary line using an unstretchable plastic-covered cloth measuring tape with an accuracy of 0.1 cm. Blood pressure (BP) was measured with a random zero sphygmomanometer. Korotkoff's fifth sound was used as a sign of diastolic blood pressure (DBP) and first sound as the sign of systolic blood pressure (SBP). Readings to the nearest even number of millimeters of mercury were performed at least three times on each subject. The average of these measurements was used in the analyses. Heart rate was measured simultaneously. Alcohol consumption, estimated as number of standard drinks per week during the previous week, and smoking habits were assessed by a self-administered questionnaire. Those smoking daily were considered as smokers. We also stratified smoking status into non-smokers (never smoked), less frequent and infrequent smokers, and daily smokers, with essentially similar results (data not shown). Socio-economic status was assessed by the level of education (total school years).

Physical activity

PA was assessed by a questionnaire. Participants were asked the frequency of participation in PA and its intensity during leisure time. In addition, commuting to work was assessed. Intensity was evaluated by asking how one usually performs physical exercise: 1) usually not becoming out of breath or sweating, 2) becoming out of breath and sweating slightly, (3) becoming out of breath and sweating considerably. We found direct correlations with the intensity category and the frequencies of engaging in high-intensity sports, such as jogging (r = 0.23; P < 0.001), ball games (r = 0.18; P < 0.001), racket games (r = 0.15; P < 0.001), and ice-hockey (r = 0.15; P < 0.001). Frequency was assessed by a question: How often one performs physical exercise during spare time to become out of breath or sweat: 1) not at all, 2) once a month, 3) once a week, 4) 2–3 times a week, 5) 4–6 times a week, 6) daily. Duration was assessed by inquiring the average duration of a single instance of physical exercise: 1) under 20 minutes, 2) 20–40 minutes, 3) 40–60 minutes, 4) over 60 minutes. When estimating the PA during commuting to the work-place, the length of the journey and whether it was traveled by foot or by bicycle was considered. Subjects could also choose public transportation or own car. This was asked separately for winter- and summer-times. A metabolic equivalent (MET) index for PA (later ‘MET-index’) was calculated from the product of intensity × frequency × duration and commuting PA. When estimating the PA during commuting to the work-place, the length of the journey and whether it was traveled by foot or by bicycle was considered. The coefficients for the variables were estimated from the existing tables (10). One MET is the consumption of 1 kcal by a person per weight kilogram per hour at rest. The MET-index ranged between 0 and 120 METh/wk.

To validate the MET-index questionnaire, we performed an experimental study, where we included 45 young adults (age range 23–55 y; 48% women). The participants filled in the questionnaire, and their PA was measured with accelerometers and pedometers for a period of 1 week. The MET-index and its main components, i.e. intensity, frequency, and duration, correlated significantly with the volume of movement assessed with accelerometers (r values 0.26–0.40) and the number of steps measured with pedometers (r values 0.30–0.39). The correlation coefficients derived in this validation study were of same magnitudes that have been demonstrated in other similar studies (11).

In addition, we collected step data using validated pedometers in 1,934 individuals from the whole study population. Participants wore an Omron Walking Style One (HJ-152R-E, Omron Healthcare Europe, Hoofddorp, The Netherlands) step counter for a period of 1 week. Similarly, as in the small validation study done in an independent population, significant correlations were seen between the number of steps and MET-index, as well as the number of steps and individual components of the MET-index (Table I). The correlation between the number of steps measured with pedometers and the amount of movement measured with ActiGraph accelerometers (GT1M, ActiGraph, FL, USA) was r = 0.966 (P < 0.001; n = 45).

Table I. Spearman's rank order correlations between MET-index, its individual components, and the pedometer measurements of total steps and aerobic steps (aerobic steps are those taken during activities that last for at least 10 minutes without interruption).

	Total steps		Aerobic steps
	r value	P value	r value	P value
Frequency of intensive participation	0.297	<0.0001	0.429	<0.0001
Intensity level	0.054	0.02	0.100	<0.0001
Duration of participation	0.188	<0.0001	0.217	<0.0001
Commuting at summer	0.238	<0.0001	0.297	<0.0001
Commuting at winter	0.240	<0.0001	0.274	<0.0001
MET-index (METh/week)	0.323	<0.0001	0.423	<0.0001

Frequency of intensive participation (6 categories, please see Methods for details). Intensity level (3 categories, please see Methods for details). Duration of participation (4 categories, please see Methods for details). Commuting at summer (4 categories, please see Methods for details). Commuting at winter (4 categories, please see Methods for details).

MET = metabolic equivalent.

To address the reliability of the MET-index, we calculated Cronbach's alpha (r = 0.65); values greater than 0.6 are commonly considered to indicate good internal consistency (12).

Biochemistry

Venous blood samples were drawn after an overnight fast. All lipid and lipoprotein determinations were performed using standard methods, as described previously (13,14). Serum LDL-cholesterol was calculated using the Friedewald formula (15) in subjects with triglycerides below 4 mmol/L. The oxidized LDL (oxLDL) concentrations were determined using a Mercodia test (16). High-sensitivity C-reactive protein (CRP) concentrations were analyzed by latex turbidometric immunoassay, glucose concentrations enzymatically, and homocysteine concentrations with microparticle enzyme immunoassay kit. Details of the methods have been presented elsewhere (14,17). Serum insulin was measured by microparticle enzyme immunoassay kit (Abbott Laboratories, Diagnostic Division, Dainabot, Ireland). The homeostasis model assessment of insulin resistance (HOMA-IR) method was used to estimate insulin resistance as described [(fasting insulin mU/mL × fasting glucose mmol/L)/22.5] (18). Serum leptin and adiponectin concentrations were determined in duplicate using a commercially available double-antibody radioimmunoassay kit (Human Leptin RIA Kit and Human Adiponectin RIA Kit; Linco Research, St Charles, MO, USA). Serum L-arginine, ADMA, and symmetric dimethyl arginine (SDMA) levels were determined by high-performance liquid chromatography method with precolumn o-phthaldialdehyde derivatization. The precision (CV%) for a plasma pool (n = 77) for arginine, ADMA, and SDMA within series was 7.5%, 5.7%, and 6.5% and between series 12.9%, 10.6%, and 12.1%, respectively. Serum creatinine concentration was measured by photometric Jaffe assay (Olympus Diagnostica GmbH). Serum phospholipase A2 activity (sPLA2) concentration was measured with a sandwich-type enzyme-linked immunosorbent assay and serum sPLA2 activity by a selective fluorometric assay http://atvb.ahajournals.org/cgi/content/full/27/5/1177 - R18-139532http://atvb.ahajournals.org/cgi/content/full/27/5/1177 - R19-139532 by using fluorescent substrate 1-hexadecanoyl-2-(1-pyrenedecanoyl)-sn-glycero-3 phosphomethanol, sodium salt (Interchim, Montluçon, France) (19). The minimum detectable activity was 0.10 nmol/min/mL, and the intra- and inter-assay CV was <10%. Serum amyloid A (SAA) concentrations were measured with an ELISA kit with a detection limit of <0.004 mg/L (Human SAA, Biosource International, Camarillo, CA, USA).

Statistical methods

We tested whether there was a significant difference in risk marker levels between sexes by using a t test, non-parametric median test, generalized linear model, or chi-square test, as appropriately (Table III). The subjects were divided into age- and sex-specific MET-index tertiles. The tertile cut-points by age-group and sex are shown in Table II. The associations between tertiles of PA and risk markers were studied by calculating Spearman's correlation coefficients. Adjustment for waist circumference, smoking, and years of education was done by calculating partial correlation coefficients. Regression modeling was used to assess whether a difference in PA existed between men and women. To examine whether the associations between PA and risk markers were similar in men and women, we tested for statistical interactions. The regression models included each risk variable as the outcome variable and MET-index, sex, and MET-index × sex interaction term as independent variables. If the MET-index × sex interaction term was significant (P < 0.05), the association between a risk marker and PA tertiles was calculated separately for men and women.

Table II. MET-index (METh/week) tertile cut-points in each age-group in men and women.

	33.3 percentile	66.7 percentile
Women Age
24	12	28
27	6	20
30	9	21
33	5	20
36	5	20
39	5	20
Men Age
24	10	33
27	4	22
30	5	29
33	3	20
36	3	20
39	3	20

MET = metabolic equivalent.

Table III. Base-line characteristics. Values are mean and standard deviation unless stated otherwise.

	Women		Men
Variable	Mean	STD	Mean	STD
MET-index (median, range)	14.0	0–111	12.0	0–120
Leisure time MET-index (median, range)	8.3	0–110	8.3	0–93
Age (y)	31.6	5.0	31.5	5.0
Waist circumference (mm)	794	114	899	108
BMI (kg/m^2)	24.3	4.4	25.8	3.9
SBP (mmHg)	113	12	122	12
DBP (mmHg)	69	10	73	11
Heart rate/min	68	9	65	9
Daily smokers (%)	18		28
Drinks per week	3.4	4.6	8.8	10.4
Education (school years)	14.9	3.0	14.2	3.1
Total cholesterol (mmol/L)	5.1	0.9	5.3	1.0
Triglycerides (mmol/L)	1.2	0.7	1.5	1.0
LDL-c (mmol/L)	3.2	0.8	3.4	0.9
oxLDL (U/L)	79.4	23.4	88.6	25.9
HDL-c (mmol/L)	1.4	0.3	1.2	0.3
LDL/HDL-ratio	2.4	0.8	3.1	1.2
ApoA1 (g/L)	1.6	0.3	1.4	0.2
ApoB (g/L)	1.0	0.2	1.1	0.3
ApoB/ApoA1 ratio	0.7	0.2	0.8	0.2
Glucose (mmol/L)	4.9	0.7	5.2	1.0
Insulin (mU/L)	7.9	6.0	7.7	5.8
HOMA-IR	1.8	1.9	1.9	2.1
CRP (mg/L)	2.3	4.4	1.5	3.3
Serum amyloid A (μg/L)	25.6	88.8	20.8	77.7
Homocysteine (μmol/L)	9.0	3.3	10.8	4.0
Leptin (ng/mL)	15.9	10.2	5.5	4.3
Adiponectin (μg/mL)	11.0	4.5	7.4	3.3
ADMA (μmol/L)	0.6	0.2	0.6	0.1
SDMA (μmol/L)	0.4	0.1	0.4	0.1
L-arginine (μmol/L)	116.5	38.4	111.1	37.8
sPLA2 type IIA (ng/mL)	4.7	3.2	3.0	2.3
sPLA2 activity (nmol/mL/min)	1.8	0.6	1.4	0.5
Creatinine (μmol/L)	65.7	9.0	80.0	9.9

All differences between sexes P < 0.05, except for leisure time MET-index, age, insulin, HOMA-IR, and ADMA.

MET = metabolic equivalent; BMI = body mass index; SBP = systolic blood pressure; DBP = diastolic blood pressure; LDL-c = low-density lipoprotein cholesterol; oxLDL = oxidized low-density lipoprotein; HDL-c = high-density lipoprotein cholesterol; ApoA1 = apolipoprotein A1; ApoB = apolipoprotein B; HOMA-IR = homeostasis model assessment of insulin resistance; CRP = C-reactive protein; ADMA = asymmetric dimethyl arginine; SDMA = symmetric dimethyl arginine; sPLA2 = phospholipase A.

Regression modeling was used to assess whether there was an interaction with PA and age. We tested the interaction between age and PA for those variables that were independently associated with PA. The regression models included each risk variable as the outcome variable and MET-index, age, and MET-index × age interaction term as independent variables.

All statistical analyses were done with the Statistical Analysis System software version 9.2, and statistical significance was inferred at a 2-tailed probability value < 0.05.

Results

The characteristics of study subjects are shown in Table III. Women were more active than men (MET-index median 14 versus 12; P = 0.045). This difference was due to the fact that women were more active commuters. There was no significant difference in the MET-index that was calculated based only on leisure time physical activities (Table III). During summer-time 18.2% of women reported that they walked and 25.1% bicycled to work. Only 7.3% of men walked and 17.3% bicycled to work. During winter-time 25.0% of women walked and 4.2% bicycled to work. In men 9.4% walked and 7.4% bicycled to work during winter-time. All risk markers except leisure time MET-index, age, insulin, HOMA-IR, and ADMA levels differed significantly by sex (Table III). The MET-index correlated inversely with age (r = 20.10; P < 0.0001).

The representativeness of the cohort

The characteristics at the original base-line study (1980) were compared between the participants of the present study (64%) and non-participants (36%). No statistically significant difference was seen in the base-line PA level. The proportion of physically active subjects (participating 4–6 times per week in physical exercise) was 84% versus 82% (P = 0.26) among participants and non-participants, respectively. In 2001, there were 1,827 individuals with complete data and 441 subjects that had non-complete data (i.e. missing at least one measurement). The distribution PA was similar in groups with complete and non-complete data (13 versus 12 METh/week; P = 0.47, respectively). The distributions of sex (P = 0.21), waist circumference (P = 0.35), and BMI (P = 0.60) were also similar between these two groups.

Relations to risk markers

Associations between PA and risk markers are shown in Tables IV, V, VI, and VII. Age- and sex-adjusted MET-index tertiles were inversely associated with waist circumference, BMI, heart rate, daily smoking, serum triglycerides, serum LDL-cholesterol, serum oxLDL, LDL/HDL-cholesterol ratio, total cholesterol, apoB, apoB/apoA1 ratio, insulin, glucose, HOMA-IR, CRP, SAA, leptin, L-arginine, and sPLA2 activity, and directly with HDL-cholesterol, apoA1, creatinine, adiponectin, and years of education.

Table IV. Associations between physical activity level (I–III) and obesity indices, blood pressure, heart rate, smoking, alcohol use, and education.

	I^a		II^b		III^c
Age- and sex-specific MET-index tertiles	Mean	(95% CI)	Mean	(95% CI)	Mean	(95% CI)	P value	P value adjusted for waist	P value adjusted for waist and smoking	P value sex interaction term
Age (y)	31.8	(31.4–32.1)	31.3	(30.9–31.7)	31.6	(31.2–31.9)
Waist circumference (mm)	855	(845–865)	848	(838–858)	825	(816–833)	<0.0001			0.39
BMI (kg/m^2)	25.3	(25.0–25.6)	25.0	(24.7–25.3)	24.6	(24.3–24.8)	0.0048			0.10
SBP (mmHg)	116	(115–117)	117	(116–118)	117	(116–118)	0.19			0.73
SBP (mmHg)^d	116	(115–117)	117	(116–118)	118	(117–119)
DBP (mmHg)	70	(70–71)	72	(71–72)	71	(70–72)	0.64			0.28
DBP (mmHg)^d	70	(70–71)	71	(71–72)	71	(70–72)
Heart rate/min	68	(68–69)	67	(66–68)	65	(64–65)	<0.0001	<0.0001	<0.0001	0.0181
Heart rate/min^d	68	(68–69)	67	(66–67)	64	(64–65)
Smokers (%)	30	(27–33)	23	(20–26)	14	(12–17)	<0.0001	<0.0001		0.17
Drinks per week	5.7	(5.1–6.3)	6.1	(5.5–6.8)	5.8	(5.3–6.4)	0.12			0.52
Drinks per week^d	5.7	(5.0–6.3)	6.2	(5.6–6.8)	6.4	(5.7–7.0)
Years of education	14.1	(13.9–14.3)	14.7	(14.5–14.9)	15.0	(14.8–15.2)	<0.0001	<0.0001	0.0001	0.17
Years of education^d	14.4	(14.2–14.6)	14.7	(14.5–15.0)	14.9	(14.6–15.1)

MET = metabolic equivalent; I = inactive; II = moderately active; III = active; 95% CI = 95% confidence interval for the mean; BMI = body mass index; SBP = systolic blood pressure; DBP = diastolic blood pressure.

^aNumber of participants varied between 697 and 808.

^bNumber of participants varied between 618 and 732.

^cNumber of participants varied between 636 and 728.

^dWaist and smoking adjusted means and 95% confidence intervals for the mean.

Table V. Associations between physical activity level (I–III) and lipids, apolipoproteins, and oxidized LDL.

	I^a		II^b		III^c
Age- and sex-specific MET-index tertiles	Mean	(95% CI)	Mean	(95% CI)	Mean	(95% CI)	P value	P value adjusted for waist	P value adjusted for waist and smoking	P value sex interaction term
Total cholesterol (mmol/L)	5.24	(5.17–5.32)	5.16	(5.09–5.24)	5.10	(5.02–5.17)	0.0099	0.2150		0.05
Total cholesterol (mmol/L)^d	5.19	(5.12–5.26)	5.14	(5.07–5.21)	5.13	(5.05–5.20)
Triglycerides (mmol/L)	1.4	(1.3–1.5)	1.4	(1.3–1.4)	1.2	(1.2–1.3)	0.0010	0.2191		0.16
Triglycerides (mmol/L)^d	1.4	(1.3–1.4)	1.4	(1.3–1.4)	1.3	(1.2–1.3)
LDL-c (mmol/L)	3.36	(3.29–3.42)	3.27	(3.20–3.33)	3.22	(3.16–3.29)	0.0015	0.0501		0.14
LDL-c (mmol/L)^d	3.32	(3.26–3.38)	3.25	(3.19–3.32)	3.25	(3.19–3.31)
oxLDL (U/L)	86.1	(84.3–88.0)	83.7	(81.7–85.7)	80.9	(79.0–82.8)	<0.0001	0.0121	0.0151	0.47
oxLDL (U/L)^d	84.7	(82.9–86.5)	83.1	(81.3–85.0)	82.1	(80.3–84.0)
HDL-c (mmol/L)	1.27	(1.25–1.29)	1.27	(1.25–1.30)	1.33	(1.31–1.35)	0.0001	0.0052	0.0354	0.59
HDL-c (mmol/L)^d	1.27	(1.25–1.30)	1.27	(1.25–1.30)	1.30	(1.28–1.33)
LDL-c/HDL-c	2.8	(2.7–2.9)	2.7	(2.6–2.8)	2.6	(2.5–2.7)	<0.0001	0.0003	0.0021	0.31
LDL-c/HDL-c^d	2.8	(2.7–2.8)	2.7	(2.6–2.8)	2.7	(2.6–2.7)
ApoA1 (g/L)	1.48	(1.46–1.50)	1.49	(1.47–1.51)	1.52	(1.50–1.54)	0.0037	0.0202	0.0683	0.20
ApoA1 (g/L)^d	1.48	(1.46–1.50)	1.48	(1.47–1.50)	1.50	(1.48–1.52)
ApoB (g/L)	1.09	(1.07–1.11)	1.07	(1.05–1.09)	1.03	(1.01–1.05)	<0.0001	0.0195	0.0396	0.10
ApoB (g/L)^d	1.07	(1.05–1.09)	1.06	(1.04–1.08)	1.05	(1.03–1.07)
ApoB/ApoA1	0.76	(0.74–0.77)	0.74	(0.72–0.75)	0.70	(0.68–0.72)	<0.0001	0.0008	0.0070	0.48
ApoB/ApoA1^d	0.74	(0.73–0.76)	0.73	(0.72–0.75)	0.72	(0.71–0.74)

MET = metabolic equivalent; I = inactive; II = moderately active; III = active; 95% CI = 95% confidence interval for the mean; LDL-c = low-density lipoprotein cholesterol; oxLDL = oxidized low-density lipoprotein; HDL-c = high-density lipoprotein cholesterol; ApoA1 = apolipoprotein A1; ApoB = apolipoprotein B.

^aNumber of participants varied between 697 and 808.

^bNumber of participants varied between 618 and 732.

^cNumber of participants varied between 636 and 728.

^dWaist and smoking adjusted means and 95% confidence intervals for the mean.

Table VI. Associations between physical activity level (I–III) and glucose metabolism markers and inflammatory markers.

	I^a		II^b		III^c
Age- and sex-specific MET-index tertiles	Mean	(95% CI)	Mean	(95% CI)	Mean	(95% CI)	P value	P value adjusted for waist	P value adjusted for waist and smoking	P value sex interaction term
Glucose (mmol/L)	5.1	(5.0–5.2)	5.1	(5.0–5.1)	5.0	(4.9–5.0)	0.0032	0.0433	0.0835	0.33
Glucose (mmol/L)^d	5.1	(5.1–5.2)	5.1	(5.0–5.1)	5.0	(5.0–5.1)
Insulin (mU/L)	8.6	(8.0–9.1)	8.0	(7.6–8.4)	6.8	(6.5–7.1)	<0.0001	<0.0001	<0.0001	0.09
Insulin (mU/L)^d	8.4	(8.0–8.8)	7.9	(7.5–8.3)	7.1	(6.7–7.5)
HOMA-IR	2.1	(1.9–2.3)	1.9	(1.7–2.0)	1.5	(1.4–1.6)	<0.0001	<0.0001	<0.0001	0.08
HOMA-IR^d	2.0	(1.9–2.2)	1.8	(1.7–2.0)	1.6	(1.5–1.8)
CRP (mg/L)	2.1	(1.8–2.4)	2.0	(1.6–2.3)	1.6	(1.4–1.9)	<0.0001	0.0095	0.0144	0.21
CRP (mg/L)^d	2.0	(1.7–2.3)	1.9	(1.6–2.2)	1.7	(1.4–2.0)
Serum amyloid A (μg/L)	25.4	(19.0–31.9)	25.9	(17.2–34.7)	18.6	(16.0–21.1)	0.0270	0.1658		0.89
Serum amyloid A (μg/L)^d	25.0	(18.6–31.3)	26.0	(19.4–32.7)	18.9	(12.3–25.6)
Homocysteine (μmol/L)	9.9	(9.6–10.2)	9.9	(9.5–10.2)	9.7	(9.5–10.0)	0.4492			0.61
Homocysteine (μmol/L)^d	9.9	(9.6–10.2)	9.9	(9.6–10.2)	9.8	(9.5–10.1)
sPLA2 type IIA (ng/mL)	3.99	(3.77–4.22)	3.88	(3.64–4.12)	3.97	(3.75–4.18)	0.8765			0.94
sPLA2 type IIA (ng/mL)^d	3.96	(3.74–4.18)	3.87	(3.64–4.10)	3.95	(3.72–4.18)
sPLA2 activity (nmol/mL/min)	1.65	(1.61–1.70)	1.61	(1.56–1.66)	1.60	(1.55–1.64)	0.0410	0.0223	0.0683	0.98
sPLA2 activity (nmol/mL/min)^d	1.64	(1.60–1.69)	1.61	(1.57–1.66)	1.59	(1.55–1.64)

MET = metabolic equivalent; I = inactive; II = moderately active; III = active; 95% CI = 95% confidence interval for the mean; HOMA-IR = homeostasis model assessment of insulin resistance; CRP = C-reactive protein; sPLA2 = phospholipase A2.

^aNumber of participants varied between 697 and 808.

^bNumber of participants varied between 618 and 732.

^cNumber of participants varied between 636 and 728.

^dWaist and smoking adjusted means and 95% confidence intervals for the mean.

Table VII. Associations between physical activity level (I–III) and adipokines and arginine metabolism markers.

	I^a		II^b		III^c
Age- and sex-specific MET-index tertiles	Mean	(95% CI)	Mean	(95% CI)	Mean	(95% CI)	P value	P value adjusted for waist	P value adjusted for waist and smoking	P value sex interaction term
Leptin (ng/mL)	12.3	(11.6–13.1)	11.0	(10.2–11.7)	9.9	(9.2–10.6)	<0.0001	<0.0001	<0.0001	0.35
Leptin (ng/mL)^d	12.0	(11.3–12.7)	10.6	(9.9–11.3)	10.0	(9.3–10.7)
Adiponectin (μg/mL)	9.1	(8.8–9.4)	9.1	(8.8–9.4)	9.8	(9.4–10.1)	0.0051	0.3685		0.0050
Adiponectin (μg/mL)^d	9.3	(9.0–9.6)	9.1	(8.8–9.5)	9.4	(9.1–9.7)
ADMA (μmol/L)	0.60	(0.59–0.61)	0.59	(0.57–0.60)	0.59	(0.57–0.60)	0.1564			0.10
ADMA (μmol/L)^d	0.60	(0.59–0.61)	0.59	(0.58–0.60)	0.59	(0.57–0.60)
SDMA (μmol/L)	0.39	(0.38–0.40)	0.38	(0.38–0.39)	0.40	(0.39–0.41)	0.4710			0.0075
SDMA (μmol/L)^d	0.39	(0.39–0.40)	0.39	(0.38–0.39)	0.40	(0.39–0.40)
L-arginine (μmol/L)	116.6	(113.8–119.4)	113.2	(110.2–116.2)	111.9	(108.9–114.9)	0.0130	0.0127	0.0397	0.66
L-arginine (μmol/L)^d	116.0	(113.1–118.8)	113.0	(110.0–116.0)	112.3	(109.3–115.3)
Creatinine (μmol/L)	71.0	(70.2–71.8)	72.1	(71.3–73.0)	74.0	(73.0–75.0)	<0.0001	<0.0001	<0.0001	0.77
Creatinine (μmol/L)^d	71.1	(70.3–72.0)	72.2	(71.4–73.1)	74.5	(73.6–75.4)

MET = metabolic equivalent; I = inactive; II = moderately active; III = active; 95% CI = 95% confidence interval for the mean; ADMA = asymmetric dimethyl arginine; SDMA = symmetric dimethyl arginine.

^aNumber of participants varied between 697 and 808.

^bNumber of participants varied between 618 and 732.

^cNumber of participants varied between 636 and 728.

^dWaist and smoking adjusted means and 95% confidence intervals for the mean.

After adjustment for waist circumference, the MET-index remained inversely associated with heart rate, daily smoking, serum oxLDL, LDL/HDL-cholesterol ratio, apoB, apoB/apoA1 ratio, insulin, glucose, HOMA-IR, CRP, leptin, L-arginine, and sPLA2 activity, and directly with HDL-cholesterol, apoA1, and creatinine. These associations remained significant also after adjustment for years of education and if a categorized variable obesity (non-obese BMI less than 30 kg/m² versus obese BMI at or greater than 30 kg/m²) was used as covariate instead of waist (the only exception was L-arginine: its association with PA was diluted to border-line significant after adjustment for obesity; P = 0.07). After further adjustment for daily smoking, the associations with apoA1, glucose, and sPLA2 activity became non-significant.

Interaction analysis suggested significant sex differences in the relations of PA with heart rate (stronger inverse association in men), adiponectin (direct relation in women only), and SDMA (direct relation in men only) (Table VIII).

Table VIII. Sex-stratified analysis for physical activity and risk markers in which the MET-index × sex interaction term was significant (P < 0.05).

	I		II		III
MET-index tertiles	Mean	(95% CI)	Mean	(95% CI)	Mean	(95% CI)	P value for trend	P value adjusted for age	P value adjusted for age and waist	P value adjusted for age, waist, and smoking
Women
Heart rate/min	69	(68–70)	68	(67–69)	67	(66–68)	0.0011	0.0004	0.0024	0.0009
Adiponectin (μg/mL)	10.5	(10.1–11.0)	10.7	(10.2–11.1)	11.8	(11.3–12.3)	<0.0001	0.0002	0.0184	0.0490
SDMA (μmol/L)	0.38	(0.37–0.39)	0.38	(0.37–0.39)	0.38	(0.37–0.39)	0.47
Men
Heart rate/min	67	(66–68)	66	(65–67)	61	(60–62)	<0.0001	<0.0001	<0.0001	<0.0001
Adiponectin (μg/mL)	7.2	(6.9–7.6)	7.5	(7.1–7.9)	7.3	(7.0–7.7)	0.45
SDMA (μmol/L)	0.40	(0.39–0.41)	0.39	(0.38–0.40)	0.42	(0.41–0.44)	0.0098	0.0089	0.0435	0.0386

MET = metabolic equivalent; I = inactive; II = moderately active; III = active; 95% CI = 95% confidence interval for the mean; SDMA = symmetric dimethyl arginine.

There was an inverse association between PA and oxLDL, and it seemed stronger in younger age-groups. The sex, waist, and daily smoking adjusted correlations were r = 20.12, P = 0.005 in 24–27-years-olds; r = 20.10, P = 0.009 in 30–33-year-olds; and r = 0.03, P = 0.43 in 36–39-year-olds.

The MET-index was not significantly associated with BP, alcohol consumption, serum homocysteine, ADMA, or sPLA2 type IIA.

Discussion

We found that PA was associated with several cardiovascular risk markers. In general, the observed associations were similarly seen in both sexes and across all ages. Interaction analysis suggested significant sex differences only in the relations with heart rate (stronger in men), adiponectin (women only), and SDMA (men only). Although the direct relation between HDL-cholesterol and PA has been established in previous studies in men, some studies have failed to show this association in women (20).

Lipids and lipoproteins

We found that PA was strongly and directly associated with HDL-cholesterol and apoA1 (HDL's major apolipoprotein) and observed no evidence of sex-specific effects on these two markers. PA lowers LDL-cholesterol at least in men (21), but the effect is probably mediated through changes in fat mass (22). In line with this, we found that the inverse relation between LDL-cholesterol and PA became non-significant after controlling for waist circumference. Interestingly, however, the inverse association between apoB (LDL's apolipoprotein) and PA remained significant after the adjustment. There was also an inverse association between oxLDL and PA. It has been previously shown that acute PA increases (23) and weight loss decreases serum oxLDL levels (16). However, in our study the inverse relation remained after controlling for waist circumference. PA may influence lipoprotein metabolism by several mechanisms, including increased skeletal muscle lipoprotein lipase activity (24), decreased cholesterol ester transferase protein concentration/activity, increased lecithin cholesterol acyltransferase activity (25), and decreased hepatic triglyceride lipase activity (26). Interaction analysis suggested significant age difference in the relations of PA and oxLDL. The inverse association with PA seemed stronger in younger age-groups. We have no plausible explanation for this observation, but it may suggest a declining protective influence of PA on LDL oxidation with aging.

Glucose and insulin

PA lowers insulin resistance (27), and much of the effect may be mediated through changes in fat mass, but some effects may be independent (28). We found that PA was inversely associated with insulin, glucose, and HOMA-IR. These associations were independent of fat mass and similar in men and women. Possible mechanisms include increased muscle GLUT4 content, increased glucolytic flux capacity, and improved insulin action in skeletal muscle (29).

Arginine metabolites

L-arginine is a semi-essential amino acid. Part of L-arginine is degraded to creatine (30). Circulating creatine is taken up by skeletal muscle and nerves, where it is phosphorylated and undergoes non-enzymic dehydration to yield creatinine (30). Thus one possible mechanism explaining the observed inverse relation between PA and L-arginine is that the muscles of physically active subjects consume more creatine, which diminishes the amount of available serum L-arginine. In line with this, we found also a direct association between PA and serum creatinine. ADMA and SDMA are methylated metabolites of L-arginine (31). ADMA is an endogenous analog of L-arginine that may interfere with nitric oxide metabolism by acting as a competitive inhibitor of nitric oxide synthesis (32). SDMA is an isomer of ADMA that is not directly capable of inhibiting nitric oxide synthase (33), but it may indirectly limit nitric oxide generation by reducing intracellular availability of L-arginine (34). Because of these characteristics ADMA and SDMA are possible risk factors for endothelial dysfunction. We found that PA was directly associated with SDMA in men. SDMA concentrations correlate with the creatinine clearance (35), and creatinine is a break-down product of creatine phosphate in muscle (36). Therefore, our finding might be explained by men's larger muscle mass.

Adipokines

In line with previous studies (37) we found that PA was inversely associated with serum leptin levels regardless of the amount of body fat. In rats, exercise training reduces leptin expression (38). In humans, the mechanism by which PA influences leptin levels is unknown. We found that PA was independently associated with adiponectin in women, but not in men. A direct association between PA and serum adiponectin has been previously reported (39).

Inflammatory markers

We found that the CRP levels were inversely associated with PA. Part of the association may be mediated by changes in fat mass and decreased adipocyte production of interleukin-6 (40). However, this association remained significant also after controlling for waist circumference, thus suggesting an effect of PA on CRP level that is independent of visceral adiposity. Phospholipase A2 enzymes hydrolyze phospholipids to generate lysophospholipids and fatty acids, leading to the activation of various immunoinflammatory processes related to the pathogenesis of atherosclerosis (41). Several studies have shown increased levels of sPLA2 type IIA or increased sPLA2 activity in patients with cardiovascular disease (42). There is a lack of information regarding the effects of PA on phospholipase enzymes. One available intervention study in 77 patients with type 2 diabetes found no effect of 24 weeks of exercise training (aerobic exercise for 2 hours daily) on sPLA2 activity (43). In our study, however, PA was inversely associated with sPLA2 activity. This association was similar in both sexes and independent of obesity. Thus, these data suggest a direct effect of PA on sPLA2 activity.

Smoking

In line with previous observations (44) PA was strongly inversely associated with smoking in both sexes. We have earlier shown in this cohort (45) that young subjects leading a sedentary life started smoking more often compared to physically more active youths. In contrast, those remaining physically active hardly ever started smoking during transition from adolescence to young adulthood (45). The results of the present analysis indicate that the inverse relationship between PA and smoking persists into adulthood. There are several mechanistic behavioral explanations for this inverse relation, including effects on mood and experiences of reward (46). When daily smoking was taken into account as a covariate, we found that the relations of PA on glucose, PLA2 activity, and ApoA1 were marginally diluted, suggesting confounding. It has been previously shown that smoking is inversely associated with ApoA1 (47). However, in our study the direct relation with HDL-cholesterol remained highly significant after controlling for smoking. In addition, we found that PA was directly associated with years of education. However, taking into account socio-economic status did not dilute the relations between PA and risk factors.

Hemodynamics

We found that PA was inversely associated with heart rate in both men and women. Lower resting heart rates have been reported in subjects who participate in sports (48) and are partially the result of increased parasympathetic tone (49). The mechanisms may include improved cardiovagal baroreflex sensitivity and regulation of the autonomic outflow induced by PA (50). In older persons, hypertensive adults, and obese individuals, PA is frequently associated with lower blood pressure and reduced risk for the development of hypertension (51). However, in younger subjects, no association between PA and blood pressure has been generally observed among normotensive individuals (52). In line with this, we were unable to detect a significant inverse association between PA and blood pressure in our relatively young and normotensive population with an average systolic blood pressure of 113 mmHg in women and 122 mmHg in men. The failure to detect such a relation may be due to inaccuracies in our MET-index, or because no such association exists in our population. We suspect the latter, as our MET-index was sensitive enough to capture the anticipated and strong relations of PA to waist circumference and insulin.

Limitations

We constructed the MET-index with a questionnaire that included questions on the frequency and intensity of leisure time PA and commuting to work. There were potential inaccuracies in the individual components. For example, the intensity of PA was estimated with a question that had only three categories. Despite this, the intensity question correlated directly with high-intensity sport activities, providing evidence of its construct validity. In the frequency question, the two categories at the high end had wide ranges: 2–3 or 4–6 times a week. Therefore, when constructing the MET-index we allocated the same coefficient for these choices so that this inaccuracy would not cause artificial differences in the index. However, similar results were seen if different coefficients were allocated for these two choices, suggesting that not much information is lost by combining these two categories. The highest category for the maximum duration of the exercise was over 60 minutes, thus we were unable to rank subjects above this cut-point. Despite these potential limitations, we observed reasonable correlations between the MET-index and objective measurements of PA assessed by pedometers and accelerometers. Nevertheless, the results should be interpreted with caution as the potential inaccuracies in the PA data may lead to under-estimations of the relations between MET-index and risk markers.

In our study, the relations between PA and known correlates of PA, such as waist circumference, HDL-cholesterol, insulin, and blood pressure, are similar to those observed in other population-based studies in young adults that have used questionnaires to estimate PA (53–55). Due to differences in the methods of measuring PA and the way the results have been reported it is, nevertheless, difficult to compare the magnitude of these relations between populations.

The study design was cross-sectional. Longitudinal studies and intervention studies are needed to confirm the observed associations. Finally, our population consists of white European subjects, and the results may not be generalized to other ethnic groups.

Summary

These population-based data in young adults demonstrate that increased levels of PA are associated with a wide range of metabolic variables, including markers of lipid metabolism, glucose/insulin metabolism, and inflammation. These associations reflect beneficial effects on cardiovascular health in both sexes and may offer mechanistic insights for the inverse association between PA and CHD.

Acknowledgements

Irina Lisinen and Ville Aalto are acknowledged for skillful data management and analysis. This study was financially supported by the Academy of Finland (grants no. 77841, 210283, 121584, and 34316), the Social Insurance Institution of Finland, the Turku University Foundation, the Juho Vainio Foundation, Finnish Foundation for Cardiovascular Research, Research funds from the Tampere and Turku University Hospitals, the Research Foundation of Orion Corporation, and the Finnish Cultural Foundation.

Declaration of interest: The authors state no conflict of interest and have received no payment in preparation of this manuscript.