Cardio- and cerebrovascular disease (CVD) remains the leading cause of death in Europe with atherosclerosis as the main underlying condition (1, 2). Observations from autopsy studies demonstrate a significant association between cardiovascular risk factors (cvRF) and atherosclerotic lesions even in early life (3). Based on this finding, it is well established that cvRF are predictive of future development of CVD (4). Great importance must therefore be given to the prevention of cvRF from a young age on.

In several studies, physical activity (PA) has been demonstrated to be one of the essential preventive interventions for atherosclerosis (5, 6). To investigate the effect of PA on atherosclerotic vascular changes, various studies have been conducted in adult study cohorts (7, 8). Regarding children and adolescents, however, available evidence is still scarce. Carotid intima media thickness (cIMT) and carotid-femoral pulse-wave velocity (cfPWV) offer an opportunity to measure early atherosclerosis—or more appropriately termed early vascular ageing (EVA)—non-invasively in vivo (9–15). These markers have been associated with various cvRF at young age (16, 17).

Contrary to what might be expected, we and others described a positive association between PA and cIMT (17–21) which could be interpreted that high PA could have unfavorable effects on vascular ageing in children and adolescents. To further investigate this primary unintuitive outcome, the aim of this work is to perform a depth secondary analysis of the large and homogenous EVA-Tirol study population of two assessments of PA [self-reported minutes of sports per day and the validated Baecke Score (BS) (22)] with two independent measures of EVA (cIMT and cfPWV) with additional parameters of PA and EVA.

Data that support the findings of this study are available from the corresponding authors upon reasonable request after signing the appropriate data transfer agreement.

The analysis was performed with 1,000 participants from the intervention group (all participants who completed the follow-up examination) and 529 participants from the control group. After excluding participants with an incomplete data set, 1,001 (65.5%) individuals could be included in the analyses below (see flow chart in Figure 1). There were no significant differences between participants in- and excluded in the present evaluation (data not shown). 558 (55.7%) of the participants were female. The mean age of the study population at the time of the examinations was 17.8 years (standard deviation 0.9, range 16–23 years). The mean cIMT was 417.8 (± 49.0) µm [female: 407.6 (±44.5) µm, male: 430.7 (±51.5) µm; p < 0.001]. The mean value of cfPWV was 6.1 (±0.8) m/s [females: 6.0 (±0.7) m/s, males: 6.3 (±0.9) m/s; p < 0.001].

The median daily PA was 40.0 (25.0–60.0) min, with a significant difference (p < 0.001) between female: 30.0 (20.0–60.0) min and male: 51.4 (30.0–70.0) min study participants. Taking the Baecke score as the second parameter for PA [median of the entire study population: 7.9 (7.1–8.6)], this gender-specific difference [females: 7.8 (7.0–8.5), males: 8.0 (7.3–9.0)] was confirmed (p < 0.001). Further details of the participants can be found in Table 1.

cIMT was positively correlated with PA measured as minutes sports per day (r = 0.152, p < 0.001) or by the BS (r = 0.130, p < 0.001) indicating that a higher PA was associated with an increased vessel wall thickness. This association remained robust in the multivariable regression analysis adjusting for sex, age, BMI, systolic blood pressure, ALT, HDL-C, PY and fasting serum glucose (Table 2).

cfPWV was positively correlated with PA measured in minutes sports per day (r = 0.065, p = 0.040). This association was rendered insignificant after adjustment for sex, age and cvRF (p = 0.554). When using the BS as measurement for PA no correlation (univariable) was found with cfPWV, yet in a multivariable linear regression analysis adjusting for the factors above, a weak inverse association was found with an unstandardized regression coefficient (95% confidence interval) of −0.045 (−0.084 to −0.006) (p = 0.023) (Table 2).

We have previously reported a positive association between PA measured in minutes sports per day and cIMT in this cohort (17). The novelty of our present analysis is to reproduce this finding with the BS as an alternative parameter for PA providing a more complete representation of activity behavior and explore the effect of PA on another parameter of EVA.

It is counter-intuitive that a higher PA is associated with a higher cIMT (i.e., vessel wall thickness), an established surrogate marker for atherosclerosis and future risk of CVD (30, 31), especially as PA has been proven to have a protective effect against the development of CVD (32).

Previous studies have seen similar effects (18–21), but other studies also have found an inverse association of PA on cIMT (33–35). Yet the latter studies have examined considerably younger children (33) or have demonstrated a modest reduction in cIMT after a PA intervention program in obese adolescents and not a cross-sectional association (34, 35).

Current data on the pathophysiology of the relationship between PA and cIMT are very limited. One possible explanation is that increased blood flow during PA leads to an elevated shear stress on the arterial vessel wall, and thereby to an release of prostacyclin and an enhanced synthesis of nitric oxide by the endothelial nitric oxide synthase (36–38). The resulting dilation of the carotid artery leads to a stimulation of endothelial progenitor cells with subsequent arteriogenesis and endothelial repair mechanisms (39, 40).

An alternative explanation would be that moderate/vigorous PA leads to adaptive processes in the Tunica media and the vascular smooth muscle cells, collagen and elastin localized there. Blood pressure peaks during PA result in a mechanical stimulus that leads to a change in the proportion of smooth muscle cells in the intimal layer and to a relative increase in the ratio of collagen to elastin in the media layer to maintain “tensional homeostasis” (41). An over-all increase in cIMT would therefore not be due to an atherosclerotic change in the vessel wall, but rather a physiological adaptive process (21).

When analyzing another parameter for EVA—the cfPWV, reflecting primarily the stiffness of the aorta (with higher values indicating an accelerated EVA) (15, 42)—we have found no independent association between PA measured in minutes per day and a small inverse (i.e., favourable), independent effect of PA assessed by the BS. Therefore, we cannot confirm that PA leads to an increased EVA in the aorta but might already have some protective properties. Other measures of arterial elasticity or stiffness used in published cohorts were not or positively influenced by PA (19, 43–45).

As autopsy studies have demonstrated that earliest morphological changes of arteriosclerosis occur in the vessel wall of the abdominal aorta, even before those in the carotid artery (46). This observation is supported by the study results of Weberruß et al, Pahkala et al. and Königsstein et al, who found in their analyses of adolescents the first signs of EVA earlier in the abdominal aorta (measured as aortic IMT) than in the carotid arteries (measured as cIMT) (43, 47, 48). It is counter-intuitive that PA leads to a premature atherosclerotic vessel wall thickening in the common carotid arteries and to the opposite effect in the aorta (reduced aortic stiffness). Considering these findings, our results support the hypothesis of Baumgartner et al. that the increased IMT in the carotid arteries of physical active adolescents are not due to atherosclerotic wall changes but rather reflect physiologic adaptive processes in the vessel wall (20).

To the best of our knowledge this is the first study to specifically address the issue of a positive association between PA and cIMT in a community-based cohort of adolescents. The strengths of our studies include the large and homogeneous study cohort of young participants and the high quality of the assessment of cvRF and health-specific lifestyle parameters through structured, physician-guided examinations and interviews. The size of our study cohort and the structured assessment of cvRF assessed, enabled a comprehensive consideration of relevant confounders.

There are also some limitations. Even though we have used well-established assessment instruments for PA we cannot exclude a recall bias, and we do not have objective measurements e.g., by accelerometers, pedometers or heart rate monitors. In addition, we had to exclude approximately one third of our population due to missing variables (see flow chart in Figure 1). We have also recruited a sole Middle-European cohort, and our findings may not be generalizable to other communities and populations. Finally, we recognize that the interpretation of our data is—in part—speculative. However, like others before us (19, 21), we must point out that the described topic of the effect of PA on EVA is still under-researched, especially in physiological terms. Valuable insights could be gained from large prospective cohorts monitoring the development of EVA as well as from basic research focusing on vascular remodeling under high PA in healthy blood vessels.

PA is associated with an increased cIMT but showed a trend to a reduced aortic stiffness as measured by cfPWV. Therefore, based on our results and the currently available evidence, we conclude that it is unlikely that a high PA leads to an accelerated EVA. It is very tempting to speculate that PA leads to a physiologic adaptive processes in the cIMT that do not resemble early atherosclerosis, yet in the long run, protects from vascular ageing, atherosclerosis and cardiovascular events. Further studies are needed.