Inflammation Markers are Associated with Cardiovascular Diseases Risk in Adolescents: The Young Hearts Project 2000

Abstract

Purpose

The traditional approach for identifying subjects at risk from cardiovascular diseases (CVD) is to determine the extent of clustering of biological risk factors adjusted for lifestyle. Recently, markers of endothelial dysfunction and low grade inflammation, including high sensitivity C-reactive protein (hsCRP), soluble intercellular adhesion molecules (sICAM), and soluble vascular adhesion molecules (sVCAM), have been included in the detection for high risk individuals. However, the relationship of these novel biomarkers with CVD risk in adolescents remains unclear. The purpose of this study, therefore, was to establish the association of hsCRP, sICAM, and sVCAM with CVD risk in an adolescent population.

Methods

Data from the Young Hearts 2000 cross-sectional cohort study, carried out in 1999–2001, were used. From a total of 2,017 male and female participants, 95 obese subjects were identified and matched according to age, sex, and cigarette smoking, with 95 overweight and 95 normal-weight adolescents. Clustered CVD risk was computed using a sum of Z-scores of biological risk factors. The relationship was described using multiple linear regression analyses.

Results

hsCRP, sICAM, and sVCAM showed significant associations with CVD risk. hsCRP and sICAM had a positive relation with CVD risk, whereas sVCAM showed an inverse relationship. In this study, lifestyle factors showed no relation with CVD risk.

Conclusion

The results fit the hypothesized role of low grade inflammation and endothelial dysfunction in CVD risk in asymptomatic adolescents. The inverse relationship of VCAM, however, is hard to explain and indicates the complex mechanisms underlying CVD. Further research is needed to draw firm conclusions on the biomarkers used.

See Editorial p. 319

Cardiovascular diseases (CVD) remain prominent among the major chronic diseases, causing widespread disability and a shortened life span. To prevent CVD, pre-emptive interventions aimed at lifestyle habits from an early age are required, before associated pathologies become established [1–3]. The classical approach to detect subjects at risk is to determine the presence of biological risk factors such as blood pressure and cholesterol level, and lifestyle risk factors such as physical activity and daily caloric intake [4–6]. These classical risk factors tend to cluster and work synergistically, so that the clustered risk exceeds that of the sum of risk factors in isolation. Therefore, CVD risk clusters are especially useful in the detection of high risk individuals [7–9]. Recently, CVD risk has been refined by including various markers indicative of low grade inflammation and endothelial dysfunction [10–12]. The practical usefulness of these markers derives from their appearance in the early stages of a causative pathway of obesity-related vascular damage [11,13].

So far, the systemic inflammation marker, high sensitivity C-reactive protein (hsCRP), has been shown to be upregulated in pre-clinical as well as advanced stages of vascular damage [14,15]. Markers specific for endothelial dysfunction are soluble vascular adhesion molecules (sVCAM) and intercellular (sICAM) adhesion molecules [3,13,16,17]. Although the physiological mechanism of sICAM and sVCAM in endothelial damage remains obscure, it is associated with upregulation of sCAMs in different stages of vascular damage [3,16,17]. However, although the single markers have been found to be related to several CVD risk factors in adult populations, the association with CVD risk in adolescents is unknown. The aim of this study, therefore, was to examine the relationship of hsCRP, sICAM, and sVCAM with CVD risk in adolescents, independent of the widely accepted biological and lifestyle risk factors. Information on CVD risk in adolescents is important as it may direct preventive strategies. Additionally, it may provide a valuable insight in the etiology of inflammation, premature atherosclerosis, and childhood adiposity.

Study population and methods

Study population

The Young Hearts Project 2000 is a cross-sectional cohort study which took place in 1999–2001 following a previous study, Young Hearts I, described in detail elsewhere [18]. The Young Hearts Project 2000 had a total of 2,017 participants from Northern Ireland, consisting of 12- and 15-year-olds from 36 nationally representative schools; the response rate was 65.4%. Eligible subjects were selected using a computer-generated random number list, divided into groups according to age and gender, resulting in approximately 500 participants in each of the four age/sex groups. Age was defined as years of age ± 6 months (e.g., a 12-year-old child ranges from 11.5 to 12.5 years). Inflammation markers were only measured in a subsample of the total population. In this subsample, a distinction was made between normal, overweight, and obese subjects (according to the classification by Cole et al. [19]). The prevalence of overweight and obesity in the complete cohort was 16.2% and 14.7%, respectively. In the present study sample, all obese adolescents (n = 95) were selected and matched according to age, gender, and smoking status, with overweight and normal weight children. Because of problems in obtaining a blood sample from three of the obese adolescents, the final number of participants in each of the three groups was reduced to 92. The total number of participants in this study was therefore = 276.

Ethical approval was obtained from the Research Ethics Committee of Queens University Belfast. Written informed consent was signed by the participants and the participants' parent or guardian.

Surveys and measurements

For each child, the examination included a medical examination, cardiorespiratory fitness test, dietary examination, and a parental and physical activity questionnaire. The medical examination included weight (nearest 100 g, Seca 770 electronic weighing scale) and height (nearest mm, Holtain stadiometer) using standard techniques, with participants in light clothing and without shoes. Pubertal status was visually assessed using Tanner pubic hair development stages [20]. Body mass index was calculated as weight (kg)/height² (m).

Peripheral venous blood (∼16 mL) was obtained by venepuncture in fasting state and collected into serum-z-clot activator EDTA-sodium or lithium heparin-coated tubes. Samples were transported in a chilled insulated container, and serum/plasma was removed after centrifugation at 100 × g for 10 minutes at 4°C within 4 hours from venepuncture and stored at −70°C.

Clustered CVD risk

Because CVD risk is seldom clinically defined at this age, a clustered score for biological risk factor can be used to specify subjects at risk [7–9].Clustered CVD risk in the present study was computed using scores for mean arterial pressure (MAP), LDL:HDL cholesterol ratio, cardiorespiratory fitness, skinfolds, and triglyceride levels. Blood pressure was measured twice from the right arm using a Hawksley random zero sphygmomanometer, with subjects sitting down quietly beforehand for at least 5 minutes. Systolic blood pressure (SB) was recorded as the mean of the two values for Korotkoff phase 1, whereas the diastolic blood pressure (DB) was based on the mean of the two values for Korotkoff phase V (15-year-old) or phase IV (12-year-old). Mean arterial pressure was defined as [2DB + SB]/3). Total plasma-cholesterol, HDL-cholesterol, and plasma triglyceride were determined (Boeringer, Mannheim, Germany; Cobas Fara automated analyzer) according to the World Health Organization (WHO) standards of quality control. Four skinfold thickness measurements (triceps, biceps, subscapula, and suprailiac; mm) were taken in duplicate on the left side of the body according to Durnin and Rahaman [21]. Cardiorespiratory fitness was determined with the 20-meter endurance shuttle test. Subjects ran a distance of 20 m at a fixed speed, and the pace was increased by .5 km/hr every minute. The score was recorded as the number of successfully completed laps at voluntary exhaustion [22].

Lifestyle

Lifestyle factors were physical activity, caloric intake, and smoking status. Physical activity was estimated by a self-administered recall questionnaire under the supervision of an exercise physiologist. The questionnaire assessed frequency, intensity, and duration of habitual activity and allocated scores out of 100 [18,23–25]. Dietary information was recorded by a nutritionist using a 7-day recall diet history interview [26]. Each nutritionist consulted with a similar number of participants from each group, and inter-observer reproducibility was confirmed before sampling. The mean daily caloric intake (kcals) was calculated from the 7-day recall diet history interview using the nutrition analysis software programme WISP (Tinuviel Software, Warrington, UK), and recorded for each subject. Smoking status was determined using a self-administered questionnaire on smoking and daily number of cigarettes smoked.

Inflammation markers

Plasma levels of sICAM and sVCAM were determined using commercially available monoclonal ELISA kits (Diaclone, Immunodiagnostic Systemd, Ltd) according to the recommendations of the manufacturer. Mean inter-assay coefficients of variation for sICAM and sVCAM ELISAs were 4.7% and 4.2%, respectively. Plasma hsCRP was quantified by a high sensitivity latex-enhanced immunoturbidimetric assay (Wako Chemicals GmbH) using a Cobas Fara automated analyzer (Roche Diagnostics, UK). The intra-assay and inter-assay coefficients of variation were 5.0%. The operator was blinded to sample classification for all laboratory procedures.

Statistical analyses

Clustered CVD risk scores were computed using a sum of age- and gender-specific z-scores. Triglyceride, cholesterol ratio, and cardiorespiratory fitness measures were log transformed to meet the assumption of normal distribution. The existence of clustering between the biological risk factors was determined with pair-wise Pearson correlation coefficients.

To determine whether inflammation markers relate to clustered CVD risk independent of established lifestyle risk factors, linear regression analyses were used. Regression analyses required that each participant had a complete dataset for all variables. Missing data were not imputed because of the small numbers of missing data. The total number of participants included in the regression analyses was n = 251. First, univariate analyses were performed to analyze the relationship between single determinants and clustered CVD risk. Second, a multiple regression analysis with a forward selection procedure was performed in which both lifestyle risk factors and inflammation markers were included as possible determinants. A threshold significance value of 10% (p = .10) was assigned for the inclusion of variables in the final prediction model.

R² were used to give an indication of the quality of the final prediction model. All statistical analyses were performed in Statistical Package for the Social Sciences (SPSS Inc, Chicago, IL), version 16.0.

Results

Table 1 shows descriptive information for all variables measured within the present study.

Table 1.

Descriptive information on CVD risk, lifestyle factors and inflammation markers in the Young Hearts 2000

	Boys		Girls
	12 years	15 years	12 years	15 years
	n = 80	n = 46	n = 82	N = 68
BMI	24.2 (5.2)		25.0 (5.1)
Tanner stage
<3	98.7	26.1	91.4	4.4
>3	1.3	73.9	8.6	95.6
Sum4Skinfolds (mm)	64.4 (32.0)		71.8 (28.3)
MAP (2DB + SB)/3 (mm Hg)	97.2 (11.8)		96.6 (11.8)
Systolic blood pressure (mm Hg)	113 (14.2)		112 (14.0)
Diastolic blood pressure (mm Hg)	65.5 (9.0)		66.4 (9.8)
LDL/HDL cholesterol ratio	.29 (.23–.34)		.31 (.26–.37)
LDL cholesterol (mg/dL)	3.0 (.75) (= .08 mmol/L)		2.9 (.67) (= .08 mmol/L)
HDL cholesterol (mg/dL)	1.2 (.34) (= .03 mmol/L)		1.3 (.29) (= .03 mmol/L)
Triglyceride (mmol/L)	.85 (.63–1.10)		.75 (.52–1.10)
Cardiorespiratory fitness (20-MST: number of laps)	54 (38–77)		34 (26–50)
Total Z score (age and sex specific)	0.0 (3.5)		0.0 (3.4)
Physical activity (Baecke score)	30.5 (15.9)		21.8 (12.4)
Energy intake (kcal)	2797 (886)		2489 (942)
Smoking	5%		18%
sICAM (mg/L)	836 (191)		746 (193)
sVCAM (mg/L)	1070 (227)		995 (236)
hsCRP (mg/L)	.87 (.99)		.77 (.95)

BMI = body mass index; MAP = mean arterial pressure; 20-MST = 20-meter endurance shuttle test.

Data are mean ± standard deviation (SD) or median (interquartile range) for continuous variables or percent for dichotomous/categorical variables.

Table 2 shows the pair-wise Pearson correlation matrix for the biological risk factors. On the basis of the partial correlations, clustering of biological risk factors seems to exist, that is, all Pearson correlations between the biological risk factors were positive and significant. Table 3 shows the pair-wise Pearson correlation matrix for inflammation markers with univariate and clustered CVD risk variables.

Table 2.

Pearson correlation matrix of CVD risk variables

	Blood pressure	Triglycerides	Cholesterol ratio	Low Fitness
Triglycerides	.27∗
Cholesterol ratio	.37∗	.53∗
Low cardio-respiratory Fitness	.17∗	.15∗	.18∗
Sum 4 Skinfolds	.54∗	.41∗	.43∗	.49∗

Correlations∗ (r) are significant (p < .05); total n = 251

Table 3.

Pearson correlation matrix of inflammation markers with univariate and clustered CVD risk variables

	ICAM	VCAM	hsCRP
Cholesterol ratio	−.23∗	.09	−.21∗
Triglycerides	.34∗	.03	.21∗
Low cardio repiratory Fitness	.21∗	−10	.30∗
Blood pressure	.05	−.12	.05
Sum 4 Skinfolds	.24∗	−.23∗	.38∗
Total Z score for CVD risk	.34∗	−.16∗	.39∗

Correlations∗ (r) are significant (p < .05); total n = 251

Table 4 shows the results of the crude and multiple regression analyses. No significant associations were found for the lifestyle variables. Inflammation markers, however, did show significant associations with clustered CVD risk. The best prediction model included hsCRP, sICAM, and sVCAM and showed an explained variance of 26 %.

Table 4.

Regression models

	Univariatea			Multipleb
	Bèta (standardized)	95% CI	p-value	Bèta (standardized)	95% CI	p-value
Physical activity	−.081	−.043 to .202	.201
Smoking status	−.037	−.158. to .086	.599
Caloric intake per 100 kCals	−.002	−.124 to.121	.981
sICAM	.340	.23 to .44	<.001	.32	21 to .44	<.001
hsCRP	.389	.28 to .49	<.001	.29	.18 to .40	<.001
sVCAM	−.161	−.28 to −.04	.010	−.24	−.36 to −.16	<.001

^aInterpretation of standardized regression coefficient (Bèta) of for example physical activity (B = −.081): scoring 1SD higher in the physical activity score results in a change of - .081 on the CVD risk score.

^bThe multiple regression model is the model that includes all significant predictors for CVD risk.

Discussion

To our knowledge, this is the first study to associate markers of inflammation and endothelial dysfunction with clustered CVD risk in adolescents. The main finding of this study was that these markers were strong predictors for CVD risk in adolescents, unlike lifestyle factors which were not predictive.

We studied three markers of low grade inflammation and endothelial dysfunction. However, not all these markers showed similar patterns of association with CVD risk. The markers sICAM and hsCRP were positively related to CVD risk, whereas an inverse relation for sVCAM was found. This unexpected inverse relation can possibly be explained in two ways. First, membrane-bound ICAM and VCAM are involved in leukocyte adhesion and migration to the vessel wall as seen in atherosclerosis [3,16,27]. These markers, however, cannot be easily measured without invasive techniques. Therefore, soluble adhesion molecules (sCAM) are used as a surrogate for membrane bound levels, based on the assumption that soluble levels are indicative of the processes at the endothelium. Although the soluble markers have previously been found to be elevated in children at risk, there is no guarantee that the elevated levels solely indicate vascular damage. Elevated levels could also indicate other events, for instance renal dysfunction [3]. Second, there is a difference in origin of sICAM and sVCAM. Although sVCAM is mainly a product of endothelial cells, sICAM is produced by many other cell types throughout the body [16,17]. Current understanding of the roles and effects of sICAM and sVCAM remains incomplete. Some studies have found that the contribution of sICAM and sVCAM in atherosclerosis is similar, whereas others have found high expression of sICAM in healthy subjects associated with CVD risk and high levels of sVCAM in ongoing stages of atherosclerosis and CVD outcomes [3,17,28–30]. Such associations of VCAM need not exist in a healthy adolescent population as currently used. The inverse relation of sVCAM as found in the current study could, therefore, indicate lower incidence of ongoing stages of vascular damage in this population.

Limitations of the study design merit consideration. Although the study provides a representative population, the cross-sectional design of the study cannot provide information on change over time. Nevertheless, a clustered risk score is an accepted method of prediction of CVD risk [4]. All variables used in the clustered risk were taken into account using the same weighting. Published data do not indicate that one variable should be weighted more heavily than another [4,5]. Although lifestyle variables did not contribute to CVD risk prediction in this study, we should be careful with the interpretation of this result because lifestyle variables were measured by self-report. Finally, the rather large percentage of obese and overweight subjects compared with a normal population did not influence the results, that is, the final regression model remained more or less the same even when analyses were performed separately for the body mass index groups (data not shown).

Conclusion

Markers of inflammation and endothelial dysfunction refine classical CVD risk prediction. Results not only confirm a role for inflammation and endothelial dysfunction but also prompt further research on the mechanistic level. At this point, conclusions to redirect the prevention of CVD are preliminary and thus efforts should remain concentrated on classical risk prediction and preventive strategies.