This prospective study was approved by the Ethics Committee of Northwestern and Central Switzerland (EKNZ 2017–01451) and conducted between 2018 and 2020 according to the Declaration of Helsinki. Written informed consent was obtained from all study participants before study participation. It was part of the COmPLETE-project (29). A detailed description of the study design is presented elsewhere (29). In summary, the project aimed to analyze age-related differences in physical functioning and to assess the role of physical fitness in this context. For the current study, targeting the interplay between cardiorespiratory fitness and vascular aging, 564 healthy adults between 20 and 91 years of age were recruited based on a randomized and blinded invitation procedure involving rural and urban areas. 144 patients with cardiovascular diseases aged 26 to 91 years stemmed from the vicinity of the cantons Basel-City and Basel-Country. Details about the recruitment procedure can be found elsewhere (29). After post-procedural data cleaning, full datasets of 360 healthy participants and 99 patients with cardiovascular diseases were used for statistical analyses (Figure 1). Empiric and anthropometric data were collected during a medical interview and examination (29).

FMD and L-FMC were measured semi-automatically with an ECG-guided high-resolution B-mode ultrasound system (UNEX-EF-38G, UNEX Corp., Nagoya, Japan) following a standardized procedure according to up-to-date methodological recommendations (30). Excellent reliability and precision of this method have been demonstrated before (31). Details of measurement procedure, post-procedural data cleaning and automatic calculation of FMD and L-FMC have been described elsewhere (20). Importantly, lower L-FMC values indicate better vasoconstrictor responsiveness, whereas higher FMD values indicate better vasodilator responsiveness.

Cardiopulmonary exercise testing was conducted on an electromagnetically braked cycle ergometer (Ergoselect-200; Ergoline, Bitz, Germany). Gas exchange and ventilatory variables were continuously analyzed (breath-by-breath) using a computer-based system (MetaMax-3B; Cortex Biophysik GmbH, Leipzig, Germany). Before and during tests, participants were encouraged to reach their individual level of maximal exertion (i.e., volitional exhaustion, dyspnea, or fatigue). V.O₂peak was defined as the highest 30-second average of V.O₂ at any point during the test. Participants without a V.O₂-plateau who did not fulfill secondary V.O₂peak criteria were excluded from analysis. A complete description of testing procedures, including ramp protocols based on predicted performance can be found elsewhere (32).

Data analysis was performed using “R” version 3.6.1. Means and standard deviations were calculated for general description of the dataset and the level of significance was set at p = 0.05 for all tests. All tests were two-sided. Independent samples t-test was used for sex-specific differences between healthy adults and patients with cardiovascular diseases. Linear regression models were used to predict associations of age with FMD and L-FMC. Sample selection accounted for age-related variations of disease burden at different ages, and models were additionally adjusted for sex, mean arterial pressure, heart rate and pre-cuff-inflation diameter, which are major effectors of FMD and L-FMC. All continuous variables were modeled using restricted cubic splines with four knots placed at appropriate quantiles. To evaluate the evidence of a moderating effect of V.O₂peak on the relationship between age and FMD, a likelihood ratio test was used comparing a model with an interaction between V.O₂peak and age to a model without such an interaction. Model fits were inspected using diagnostic residual plots. The model for FMD exhibited heteroscedastic residuals. Therefore, heteroscedasticity-robust p-values and 95%–confidence intervals are presented for this model. To visualize the relationship of FMD with age for different levels of cardiorespiratory fitness, model-predicted change of FMD was plotted for different centiles of V.O₂peak (3rd, 15th, 50th, 85th, and 97th) (33) while other variables were fixed at their sex-specific mean. Partial R² was presented to quantify the relative importance of the included variables. To aid interpretation, p-values were converted to surprisal values (s-values) when appropriate using the formula s = −log₂⁡(p) (34). The s-value provides an intuitive interpretation of the information conveyed by a certain p-value against the tested hypothesis. For a specific s-value, p conveys roughly the same evidence against the tested null hypothesis as seeing all heads in s independent tosses of a fair coin.

Data of 360 healthy non-smoking adults without cardiovascular disease (48.1% women) and 99 patients with cardiovascular diseases (28.3% women) were analyzed (Figure 1 and Table 1). Physical activity of the subjects was above a population specific average [details provided elsewhere (33)]. Healthy participants had a very low average 10-year risk of cardiovascular disease of 8.2 ± 8.6% (35) and their age- and sex-specific V.O₂peak was similar (36) or higher (37, 38) than in other large cohorts of healthy adults. Cardiovascular diseases among the patients were arterial hypertension (N = 42), coronary artery disease (N = 41), aortic/mitral valve regurgitation (N = 5), atrial fibrillation (N = 4), dilative cardiomyopathy (N = 4), aortic/mitral valve stenosis (N = 2) and pulmonary hypertension (N = 1). 95 participants received antihypertensive medication including β-blockers in 47 cases. Lipid-lowering and anti-diabetic medication was taken by 52 and 8 patients, respectively. 4 patients were active smokers, 17 were ex-smokers. Left ventricular ejection fraction was mildly reduced (40 – 60%) in 11 patients and severely reduced (<40%) in 18 patients. According to NYHA dyspnea classification, 68, 19, 12 and 0 patients were class I, II, III and IV, respectively. Average FMD was lower in patients with cardiovascular diseases than in healthy adults (men: 4.6 ± 3.2% vs. 6.6 ± 4.0%, women: 4.3 ± 3.5% vs. 7.8 ± 4.3%, p < 0.001), whereas there was only little evidence for a difference in L-FMC between groups (men: −0.6 ± 2.8 vs. −0.3 ± 2.0, p = 0.30, women: −0.7 ± 3.1 vs. −1.2 ± 3.0, p = 0.38). Further measurement data of endothelial function (i.e., shear rate, blood flow, vascular wall thickness) are presented in the Supplementary Table 1.

FMD showed a significant negative association with age in the healthy sample without burden of chronic diseases (adjusted R² = 0.27, partial R² = 0.07, p < 0.001; Supplementary Table 2) and in patients with cardiovascular diseases (adjusted R² = 0.29, partial R² = 0.05, p = 002). L-FMC showed no significant association with age in both samples (Supplementary Table 3).

There was little evidence that the relationship of FMD with age was moderated by V.O₂peak (χ²(5) = 5.37, p = 0.372, s = 1.43). Therefore, models containing no interaction between V.O₂peak and age were presented. The model explained 30.3% of the variance of FMD with V.O₂peak accounting for 2.8%. It predicted a 41% decline of FMD in very high-fit healthy men (V.O₂peak ≥97th percentile) from 20 to 90 years of age compared to 48% in very low-fit healthy men (V.O₂peak ≤3rd percentile) and of 36% in very high-fit healthy women compared to 45% in very low-fit healthy women (Figures 2A,B). In patients with cardiovascular diseases, predicted age-related differences of FMD were 43% in very high-fit men compared to 55% in very low-fit men and 41% in very high-fit women compared to 52% in very low-fit women (Figures 3A,B).

Although statistically non-significant, the predicted 7 – 9% reduction of FMD with higher age in association with a constantly low V.O₂peak in the healthy sample should not be ignored for their potential clinical relevance on an individual level. Notably, a 1% higher FMD may lead to a 13% reduction of cardiovascular morbidity (44, 45) and, thus, might reduce the risk of cardiovascular morbidity considerably, regardless of sex, age or other cardiovascular risk factors (44, 45).

In the older patients with cardiovascular diseases, FMD was 11 (women) – 12% (men) lower in the lowest V.O₂peak percentile compared to the highest, which equals an absolute difference of 0.8%. This is similar to previously observed effects of exercise training in heart failure patients increasing FMD by 1.08% (11), which, again seems clearly relevant, in terms of cardiovascular morbidity (44, 45). This is supported by further observations. For example, one study on Japanese individuals with coronary artery disease observed 50% lower rates of cardiovascular events with every 3.1% total improvement of FMD (46). Furthermore, a reduced FMD was previously associated with higher rates of in-stent restenosis (47) and occurrence of postoperative cardiovascular sequelae (48) in patients with coronary artery disease.

Differences in FMD stratified by V.O₂peak percentile became visible at the age of 30 years in women, thus 20 years earlier than in men. This seems counter-intuitive, as previous evidence suggests a 10 year earlier decline of endothelial function in men than in women indicating their higher risk for early cardiovascular morbidity and mortality (49). However, this study’s observations support the assumption, that V.O₂peak-improving exercise during young adulthood might exert higher favorable effects on vasodilator responsiveness in women, than in men (13).

The quality of phenotyping with direct state-of-the art measurement of cardiorespiratory fitness and endothelial function as well as the large and well-phenotyped samples representing the healthy population and a population with high cardiovascular disease burden are major strengths of this study.

Importantly, results from the regression analyses are visualized in Figures 2, 3 but not based on visual inspection of these graphs. Furthermore, they are not descriptive and do not provide clinical cut-offs for FMD with high or low V.O₂peak. For this, longitudinal studies with multiple measures are needed. Due to their cross-sectional nature, the models do not allow for the conclusion, that people with a high V.O₂peak throughout their adult life will show considerably better FMD than those with low cardiorespiratory fitness, especially as the process of aging does not follow a linear course, thus ultimately requesting longitudinal studies to be assessed. Yet, a person will most likely have a better individual FMD, if a high V.O₂peak is maintained into old age. However, the current results give valuable insights into the relationship between vascular aging and cardiorespiratory fitness, supporting previous claims that direct effects of exercise on vascular function and structure might constitute a relevant part of the beneficial effects of exercise in humans independent of classical cardiovascular risk factors (6, 50).