Among participants with no detectable CAC at baseline (n=215, 44.3% of sample), 50 of them developed new incident CAC during the 3 year follow up (see Figure 1). In the overall sample there was a significant increase in log Agatston score between baseline and follow up (2.25 vs 2.65, p<0.001), and 38.2% had progression of CAC as defined by an increase in >10 Agatston units. Participants with CAC progression >10 units were older, more likely to be male, smokers, statin users, have higher blood pressure, lower HDL cholesterol, and higher fibrinogen levels compared to participants without CAC progression (Table 1).

There was considerable variation in the cortisol stress response, with approximately 40% of the sample responding to the stress tasks with an increase in cortisol of at least 1 mmol/l [9]. In analysis adjusted for age and sex, the cortisol stress response (per SD increase) was associated with CAC progression (odds ratio=1.26, 95% CI, 1.02–1.52). In further models the association remained unaltered (odds ratio=1.27, 95% CI, 1.02–1.60) after additional adjustments for follow up time, pre-stress cortisol level, employment grade, use of statins, resting blood pressure, fibrinogen, HDL and LDL cholesterol, body mass index, and smoking. Other independent predictors of CAC progression included age, male sex, smoking, systolic blood pressure, and fibrinogen (Table 2).

We also performed linear analyses to examine the associations between cortisol and relative CAC change. In these analyses, CAC change was associated with age, male sex, systolic BP, fibrinogen, and cortisol reactivity (Table 3). In further multivariate models, including adjustments for age, sex, pre-stress cortisol, employment grade, smoking, resting systolic BP, fibrinogen, lipids, body mass index, and use of statins, the association of cortisol reactivity was marginally attenuated but remained independently associated with relative change in CAC (B=0.011, 95% CI, −0.001–0.023, p=0.066). In analyses stratified by the presence of CAC at baseline, the association between cortisol response and CAC change was only evident in participants without detectable CAC at baseline (fully adjusted B=0.017, 95% CI, 0.006–0.027, p=0.002).

To our knowledge, this is the first study to show a prospective association between cortisol stress reactivity and progression of sub-clinical atherosclerosis. Other studies have examined associations between sympathetic nervous activity and various CVD risk factors. For example, in a small prospective study conducted on Norwegian military personal, norepinephrine responses to mental stress and cold pressor at the baseline examination was associated with insulin resistance and blood pressure at the 18 year follow up assessment [16], [17]. Several population studies have demonstrated associations between diurnal cortisol patterns and CVD; Dekker et al [18] observed an association between total cortisol exposure while awake and higher carotid plaque scores in a sample of older adults, whilst another study showed a greater presence of CAC in younger participants with a flatter diurnal cortisol decline [19]. Also, a flatter slope in cortisol levels across the day was associated with an increased risk of CVD mortality in British civil servants [20], and 24 hr urinary cortisol was associated with CVD death in the InCHIANTI prospective cohort study of older participants [21]. Several studies have also linked raised cortisol levels with metabolic risk factors, including fasting glucose, lipids, and obesity [22], [23]. The findings from clinical patient groups are less clear. Low serum cortisol levels were shown to predict death following myocardial infarction [24] and reduced cortisol stress responses in patients with stable CAD have also been observed [25]. In contrast, elevated fasting cortisol was associated with risk of future cardiac events in patients with chronic heart failure [26], [27]. Interestingly, in our study CAC progression was unrelated to resting cortisol levels or total cortisol production (area under the curve) over the psychophysiological testing period. The equivocal nature of some of these findings might be related to the strong diurnal cortisol pattern that can heavily influence the results. Therefore, single measures of cortisol might not be appropriate to capture the dynamic nature of HPA activity. In this regard, psychophysiological testing is advantageous since extrinsic factors can be tightly controlled.

Previous work has demonstrated that heightened blood pressure responses to laboratory induced stressors is associated with CVD risk, such as progression of IMT and hypertension [3], [4]. In a young, healthy sample of women, aged 20 to 35 years at baseline, each 10 mm Hg change in systolic blood pressure during a video game stressor was associated with a 24% increased odds of having CAC after 13 years follow-up, although there was no association with blood pressure reactivity during a star tracing task [6]. Thus, our null findings on blood pressure responses and CAC are inconsistent with some previous work in this area. Nevertheless, our sample was considerably older than in many previous studies that might partly account for the findings. In addition, previous blood pressure reactivity studies have only collected CAC measures at one point in time and were thus unable to examine CAC progression. Taken together, the different effects of blood pressure and cortisol reactivity on CAC progression shown here highlight the importance of examining both cardiovascular and neuroendocrine indices of stress reactivity in psychophysiological studies.

Relatively few studies have examined risk factors for the progression of CAC. In one of the largest to date, 5756 participants from the Multi-Ethnic Study of Atherosclerosis (MESA) were followed up over 2 years, and results showed that most traditional CVD risk factors were associated with both the risk of developing new incident CAC and increases in existing calcification [15]. However, low and high density lipoprotein cholesterol was only predictive of new incident CAC in MESA. These findings might partly reflect differences in the definition of CAC progression. For example, in the present study LDL cholesterol was the only risk factor associated with new incident CAC (data not shown), although cortisol, smoking, blood pressure and fibrinogen were associated CAC progression when defined as an increase of >10 Agatston units. This might also reflect differences in the mechanisms involved at various stages of the atherosclerotic process.

The mechanisms by which HPA activity directly influences atherosclerosis remain poorly understood, although there is some evidence that increased circulating cortisol levels may promote perivascular inflammation [28], and treatment with glucocorticoids has been shown to enhance calcification within arteriosclerotic lesions [29]. A previous study in healthy participants demonstrated that mental stress-induced endothelial dysfunction and baroreflex impairment was prevented by blocking cortisol production with metyrapone [30]. Thus, heightened cortisol responses may to some extent drive changes in hemodynamic function. Others have reported reduced cortisol stress responses in patients with stable CAD, and suggested that cortisol might act as a powerful anti-inflammatory agent in preventing atherosclerotic processes [8], [25]. In the present study, however, we did not observe any associations between cortisol reactivity and markers of inflammation as indexed by C-reactive protein [9]. We cannot, however, rule out the role of unmeasured confounding risk factors or genetic influences that might account for cortisol response patterns [31] and CHD risk [32]. For example, recent evidence suggests that a common glucocorticoid receptor polymorphism is related to higher pro-inflammatory activity and greater risk of CHD [32].

The present study has a number of strengths and limitations. The major strength is the prospective design of the study that allows greater confidence in interpreting the directionality of the observed relationships. The findings add to the evidence that stress-related processes are associated not only with the occurrence of clinical CVD, but also with progression of underlying coronary disease development. It should be noted that there was a large amount of variability in individual responses to the stressors, and only 40% of participants in this study were defined as cortisol responders, which is consistent with our previous findings from another sample tested with the same behavioural tasks [33]. Cortisol responses to stress tend to be greater when participants are confronted by social-evaluative challenges, rather than psychomotor and problem-solving tasks of the type used here. Cortisol stress responses were measured on a single occasion, and there may be adaptation on repeated testing, although we have previously demonstrated strong reproducibility of these responses over two repeated stress sessions [34].