The hypothalamic–pituitary–adrenal axis activity as a predictor of cardiovascular disease, type 2 diabetes and stroke

Authors


Dr Roland Rosmond, Department of Heart and Lung Diseases, Sahlgrenska University Hospital, S-413 45 Göteborg, Sweden (fax: +46 31 82 65 40).

Abstract

Abstract. Rosmond R, Björntorp P (University of Göteborg, Sahlgrenska University Hospital, Göteborg, Sweden). The hypothalamic–pituitary–adrenal axis activity as a predictor of cardiovascular disease, type 2 diabetes and stroke. J Intern Med 2000; 247: 188–197.

Objectives. The hypothalamic–pituitary–adrenal (HPA) axis, the mediator of cortisol, plays a central role in the homeostatic processes. In this study, we addressed the potential impact of HPA axis activity on established anthropometric, metabolic and haemodynamic risk factors for cardiovascular disease (CVD), type 2 diabetes mellitus and stroke.

Design. A cross-sectional study.

Subjects. A subgroup of 284 men from a population sample of 1040 at the age of 51 years.

Main outcome measures. Anthropometric measurements included body mass index (BMI, kg m–2), waist/hip circumference ratio (WHR) and abdominal sagittal diameter (D). Overnight fasting values of blood glucose, serum insulin, triglycerides, total, low (LDL) and high density (HDL) lipoprotein cholesterol, as well as resting heart rate and blood pressure, were also determined. By using repeated diurnal salivary cortisol measurements during everyday conditions, methods were developed to characterize the status of the HPA axis, and set in relation to the anthropometric, metabolic and haemodynamic measurements.

Results. In bivariate analyses, risk factors intercorrelated in clusters of anthropometric (BMI, WHR, D), metabolic (insulin, glucose and their ratio, triglycerides, cholesterol [total and LDL], HDL cholesterol [negative]) and haemodynamic (systolic and diastolic blood pressure and heart rate) measurements. This was also the case in the two-dimensional scaling analysis, where, however, HDL separated out. A normal HPA axis status, characterized by high variability and morning cortisol values, as well as a clear response to a standardized lunch and dexamethasone suppression test, was then introduced by a statistical weighting procedure. This did not essentially change the results of either the bivariate correlation matrix or the two-dimensional scaling analysis. A similar introduction of a pathological HPA axis, characterized by low variability and morning cortisol values, a poor lunch-induced cortisol response and a blunted dexamethasone suppression of cortisol, changed the results markedly. Now strong and consistent correlations were found not only within but also between different clusters of risk factors, which also congregated into one distinct cluster, again except for HDL cholesterol.

Conclusions. These results disclose the prospect of an overriding function of a pathological HPA axis on other, established risk factors for CVD, type 2 diabetes and stroke. Its close association to HPA axis dysfunction may explain the previously reported powerful risk indication of abdominal obesity for the diseases mentioned. The HPA axis abnormality has been reported to be a characteristic consequence of frequently repeated or chronic environmental stress challenges.

Introduction

Risk factors are characteristics which, based on epidemiological evidence, are associated with an increased probability of a specific disease, although they are not necessarily causal factors. A number of risk factors have been identified for cardiovascular disease (CVD), type 2 (non-insulin-dependent) diabetes mellitus and stroke. For several of these risk factors, reasonable underlying pathogenic mechanisms have been shown or suggested [1].

The endocrine system, together with the nervous system, provides most of the extracellular control of specialized tissues to function as integrated organs. The ideas that frequent or persistent challenges of the hypothalamic–pituitary–adrenal (HPA) axis may constitute a base for pathophysiological consequences in the periphery of the body stems from the central role played by the HPA axis in the homeostatic processes. Although biologically plausible, this complex hypothesis has been difficult to study in humans [2], presumably because of methodological problems.

The HPA axis regulates the adrenal secretion of hormones, one of which is cortisol . Cortisol has effects at nearly all levels in the human body [3], including an important role in lipid and glucose metabolism. Elevated cortisol levels, if prolonged, lead to a redistribution of body fat characterized by truncal obesity, including hypertension and type 2 diabetes, as is clearly seen in Cushing’s syndrome [4]. In contrast, cortisol deficiency, as in Addison’s disease, leads to impaired fat mobilization and utilization, weight loss, hypotension and impaired gluconeogenesis with hypoglycaemia [5].

It may therefore be speculated that an abnormal cortisol secretion as a result of frequent or persistent challenges of the HPA axis might constitute a powerful risk indicator for serious disease [2]. The scarcity of data on the HPA axis status in relation to disease is probably due to the difficulties involved in measuring a chronic adaptation of this axis in humans. Since the HPA axis responds sensitively to external stimulation such as invasive procedures as well as to the unaccustomed milieu of a laboratory or hospital setting, the assessment of cortisol has to be performed during everyday life and without blood sampling in order to yield reliable and useful results. Furthermore, the dexamethasone suppression test, originally described by Liddle [6], is usually required to detect an abnormal feedback regulation of cortisol secretion by the glucocorticoid receptors in the hippocampus region [7]. This also has to be estimated to obtain a complete picture of the different characteristics of the HPA axis activity and regulation.

The assessment of cortisol in saliva provides several advantages over blood cortisol measurements, as the collection procedure is non-invasive and stress-free, making it ideal for use in psychoneuroendocrinological research [8–10]. Since salivary cortisol sampling is laboratory-independent, it can be applied under a variety of field settings. Cortisol in saliva represents the unbound (‘free’) hormone fraction and reflects accurately the free fraction of cortisol in plasma [10].

By a series of saliva samplings (n = 7) throughout the day, in which cortisol levels were measured, we estimated the resilience and plasticity of the HPA axis. A standardized lunch was provided as a physiological stimulus. As a measurement of the feedback inhibition, suppression of cortisol by overnight low-dose dexamethasone (0.5 mg × 1) was similarly registered.

The aim of the present study was to evaluate the potential impact of the HPA axis activity on established anthropometric, metabolic and haemodynamic risk factors for CVD, type 2 diabetes and stroke.

Materials and methods

Subjects

We identified all men born during the first 6 months of 1944, and living in Göteborg, Sweden (n = 1302) in 1992, from the National Population Register in Göteborg, and mailed a questionnaire to be answered and returned [11–13]. A total of 1040 (80%) responded. Based on self-measured waist/hip circumference ratio (WHR), the following three subgroups, each of 150 men, were selected for further studies: the lowest (≤ 0.885) and the highest values (≥ 1.01), as well as men around the arithmetic mean (0.94–0.96). These men were examined in 1995 at the age of 51 years, and 284 (63%) volunteered to participate. Amongst the men, 44 (15.5%) reported that they had been considered by a physician to be hypertensive. Eight were treated with angiotensin-converting enzyme (ACE) inhibitors and 25 with β-adrenergic antagonists or calcium-channel blockers. Six men had type 2 diabetes. Further details have been reported previously [1415]. None was excluded.

The study was approved by the Ethical Committee of the Medical Faculty of the University of Göteborg, and by the Swedish Data Inspection Board.

Examinations

The same team of research nurses and technicians performed the examinations in the morning after the subjects had fasted overnight.

Haemodynamic measurements. Two blood pressure (BP) measurements were taken on the right arm with the participants sitting, using a random-zero mercury sphygmomanometer [16], with the auscultation site at heart level, a peak inflation level of 30 mmHg above radial pulse disappearance, a cuff deflation rate of 2–3 mmHg s–1 and the pressure value recorded to the nearest even digit. The heart rate was recorded simultaneously, and the individual mean systolic and diastolic blood pressure was calculated as the mean of the two measurements.

Anthropometric measurements. Body weight was measured to the nearest 0.1 kg, and height was measured to the nearest 0.01 m. Body mass index (BMI) was calculated as weight divided by the height squared (kg m–2) [17]. The waist or abdominal circumference was measured midway between the lower costal and the iliac crest, and the hip or gluteal circumference was measured at the level of the great trochanters [18]; the WHR was calculated. The abdominal sagittal diameter (D [in cm]), an anthropometric approximation of visceral fat mass [19], was determined as the distance between the examination table and the highest point of the abdomen in the recumbent position.

Laboratory measurements. Quantitative measurements of insulin in serum were achieved by radioimmunoassay (RIA) (Pharmacia Insulin RIA 100, Kabi Pharmacia Diagnostics, Uppsala, Sweden), and glucose in whole blood was determined using the automated glucose analyser ESAT 6660 from Eppendorf [20]. Triglycerides, total and HDL cholesterol were measured with an enzymatic procedure in a Boehringer Mannheim Cobas Fara II (Boehringer Mannheim, Germany). LDL cholesterol was determined using the Friedewald formula [21]. The analytical technique in assaying salivary cortisol levels was RIA (Orion Diagnostica, Turku, Finland). This method has negligible cross-reactivity with other endogenous glucocorticoids and cortisol metabolites as well as with exogenous glucocorticoids such as dexamethasone (0.1%).

Cortisol measurements. The assessment of cortisol was performed using a sampling device called Salivette (Sarstedt Inc., Rommelsdorf, Germany), where salivary cortisol concentrations were determined. The Salivette consists of a small cotton swab inside a centrifugation tube [10]. The participants were asked to chew on the cotton swab for 45–60 s and store the Salivettes at + 8°C, returning them to the laboratory when all samples were completed. The salivary samples can be stored at room temperature for up to 30 days without influencing the concentration of cortisol significantly [10].

On a random working day, the participants delivered repeated salivary cortisol samples. Samples were obtained in the morning (08.00–09.00 h), then at 11.45 h, and 30, 45 and 60 min after a standardized lunch at 12.00 h, at 17.00 h, and finally just before bedtime. Within these periods, relatively small changes in unstimulated cortisol values occur and therefore a satisfactory estimation of the circadian rhythm can be obtained [9]. The lunch was provided by the laboratory and contained 266–277 kcal (protein, 18.2–21.0 g; carbohydrate, 33.6–36.4 g; fat, 5.6–6.5 g). Careful oral and written instructions were provided to avoid misunderstanding. The feasibility of the procedure and the accuracy of salivary cortisol in relation to plasma (r > 0.90) were ascertained in 40 men before the study, but these men were not included in the results.

The day after these measurements a dexamethasone suppression test was performed. Preliminary examinations [22] showed that using a low dose (0.5 mg × 1) of dexamethasone reveals mild abnormalities in the ability to control the HPA axis by feedback inhibition that cannot be encountered with the conventionally used dose of 1 mg [23]. Consequently, the participants were given one tablet of dexamethasone (Decadron, MSD, Sollentuna, Sweden) of 0.5 mg and two Salivettes. A saliva sample was collected in the morning before breakfast (08.00–09.00 h). At 22.00 h, the dexamethasone tablet was taken, and the following morning the salivary sampling was repeated. The decrease in salivary cortisol level after dexamethasone administration was calculated as the arithmetic mean of the non-inhibited morning cortisol levels (including that of the day curve measurements) minus the cortisol level after dexamethasone intake.

Statistical analysis

All data analyses were performed with SPSS for Windows, release 7.5 (SPSS Inc., Chicago, IL, USA). The two-tailed test was used throughout, and a P < 0.05 was considered as statistically significant. The P-values were adjusted for multiple (simultaneous) tests by utilizing Hommel’s modifications of the Bonferroni procedure [24]. Standard methods were used to calculate descriptive statistics. The calculation of cortisol variability has been described in detail previously [15]. Briefly, the cortisol variance (νi) was calculated for each individual based on the repeated measurements of salivary cortisol (n = 7). High (νi) and low (ωi = 1/νi) cortisol variabilities were then utilized in the correlation analyses as weighting variables. The correlation matrices are given by Pearson’s product–moment correlations.

To detect meaningful underlying pattern or structure in the observed correlation matrices, we performed multidimensional scaling (MDS) [25]. The object of MDS is to find the structure in a set of similarity (or dissimilarity) measures between variables (or cases). Pairs of variables with the highest correlations are plotted closest together; those with the lowest correlations are further apart. We used a two-dimensional solution since it has the virtue of simplicity. To estimate the distortion of the data caused by representation in two dimensions, we used the goodness-of-fit measure known as Kruskal’s stress [25]. The better the fit, the smaller is Kruskal’s stress, and less than 10% is generally considered as acceptable [25].

Non-response analyses were performed using tests appropriate to the scale of measurement of each variable [11–13]. The statistical significance was relaxed to P ≤ 0.10 in order to increase the sensitivity to detect potential selection bias.

Results

There was no difference in measures of health status between non-responders and responders, evaluated from previously obtained information [11–13], concerning hypertension, diabetes mellitus, myocardial infarction, stroke and angina pectoris. Moreover, non-responders and responders showed similar patterns of association with socioeconomic status, as measured, for instance, by educational level and occupation.

Table 1 presents the basic (descriptive) data of the study population, and summarizes data collected on a single variable to describe the larger, unobserved population. The geometric mean of BMI was 26.0 kg m–2 (SD = 4.1), and the means of WHR and abdominal sagittal diameter were 0.94 (SD = 0.07) and 22.4 cm (SD = 3.8), respectively.

Table 1.  Baseline characteristics of the total study population. Values are given as the geometric mean and standard deviation (SD) with 90% central range
 Geometric
median (SD)
90% central
range
Systolic blood pressure (mmHg)128.9 (18.0)104.2–162.0
Diastolic blood pressure (mmHg)83.2 (10.8)68.0–104.0
Heart rate (beats min–1) 68.3 (10.6)52.0–88.0
Body mass index (kg m–2) 26.0 (4.1)20.5–34.1
Waist/hip circumference ratio0.94 (0.07)0.83–1.04
Abdominal sagittal diameter (cm)22.4 (3.8)17.3–30.7
Fasting insulin (mU L–1) 10.5 (10.9)5.0–31.6
Fasting glucose (mmol L–1) 4.5 (1.0)3.7–6.0
Insulin/glucose ratio2.3 (2.7)1.1–5.9
Triglycerides (mmol L–1) 1.6 (1.1)0.8–3.9
Cholesterol, total (mmol L–1) 6.1 (1.1)4.4–8.0
Cholesterol, LDL (mmol L–1) 4.0 (1.0)2.5–5.9
Cholesterol, HDL (nmol L–1) 1.2 (0.3)0.8–1.9

Two types of cortisol secretory pattern were identified [15]. The first is characterized by a high morning cortisol value (19.8 nmol L–1), a normal response to dexamethasone (r = 0.70, P < 0.001) along with a brisk cortisol response to lunch (P < 0.001), and at 17.00 h (5.1 nmol L–1) less than 74.2% of the morning value. The average diurnal cortisol secretion was 12.6 nmol L–1. The other secretory pattern of cortisol identified is characterized by a low morning cortisol value (10.1 nmol L–1), a blunted suppression of cortisol by overnight low-dose dexamethasone (r = – 0.70, P < 0.001) as well as a poor lunch-induced cortisol response (P = 0.103), and at 17.00 h (4.9 nmol L–1) less than 51.5% of the morning value. The average diurnal cortisol secretion was 5.9 nmol L–1.

The correlation matrix in Table 2 shows the results of Pearson’s correlation analysis between the measured variables, except for the cortisol measurements. A multitude of significant correlations were found, and were those expected within clusters of anthropometric (BMI, WHR, D) metabolic (insulin, glucose, triglycerides and HDL cholesterol [negative], as well as total and LDL cholesterol) and haemodynamic variables (blood pressures and heart rate). BMI, WHR and D correlated with all other variables except total and LDL cholesterol.

Table 2.  The correlation matrix between risk factors. Marked (*) correlations are significant at P < 0.05
 SBPDBPHRBMIWHRDInsulinGlucoseIRTGCHLDL
  1. BMI, body mass index; CH, cholesterol (total); D, abdominal sagittal diameter; DBP, diastolic blood pressure; HDL, cholesterol (HDL); HR, heart rate; IR, insulin/glucose ratio; LDL, cholesterol (LDL); SBP, systolic blood pressure; TG, triglycerides; WHR, waist/hip circumference ratio.

DBP0.79*           
HR0.180.20*          
BMI0.36*0.35*0.12         
WHR0.32*0.31*0.130.60*        
D0.41*0.39*0.19*0.85*0.61*       
Insulin0.27*0.21*0.25*0.44*0.35*0.43*      
Glucose0.30*0.150.21*0.28*0.22*0.38*0.22*     
IR0.180.140.20*0.32*0.26*0.29*0.94*– 0.06    
TG0.27*0.27*0.090.37*0.33*0.44*0.35*0.39*0.22*   
CH0.170.24*0.090.04– 0.040.040.06– 0.020.060.32*  
LDL0.120.21*0.090.03– 0.060.010.04– 0.100.070.080.90* 
HDL–0.11–0.17–0.07–0.33*–0.25*–0.37*–0.31*–0.23*–0.24*–0.51*0.02–0.11

These results are graphically displayed as a two-dimensional plot in Fig. 1 (Kruskal’s stress = 9.9%). As seen in the figure, the same clusters as above tend to line up along a limited range of values of dimension 1. The anthropometric cluster (BMI, WHR, D) is collected at –0.82 to –0.66; the metabolic cluster of insulin, glucose, insulin/glucose ratio (IR) and triglycerides at –0.73 to –0.47; the cholesterol cluster around 0.83; and the haemodynamic cluster at –0.07–0.27; whilst HDL cholesterol is isolated at a value of 2.59.

Figure 1.

The correlation between risk factors displayed as a two-dimensional plot derived by multidimensional scaling. BMI, body mass index; CH, cholesterol (total); D, abdominal sagittal diameter; DBP, diastolic blood pressure; HDL, cholesterol (HDL); HR, heart rate; IR, insulin/glucose ratio; LDL, cholesterol (LDL); SBP, systolic blood pressure; TG, triglycerides; WHR, waist/hip circumference ratio.

Table 3 shows the same correlation matrix weighted for a high variability of the diurnal cortisol. As compared with the unweighted figures in Table 2, essentially the same results were obtained.

Table 3.  The correlation matrix between risk factors weighted for a high variability of diurnal cortisol. Marked (*) correlations are significant at P < 0.05
 SBPDBPHRBMIWHRDInsulinGlucoseIRTGCHLDL
  1. BMI, body mass index; CH, cholesterol (total); D, abdominal sagittal diameter; DBP, diastolic blood pressure; HDL, cholesterol (HDL); HR, heart rate; IR, insulin/glucose ratio; LDL, cholesterol (LDL); SBP, systolic blood pressure; TG, triglycerides; WHR, waist/hip circumference ratio.

DBP0.81*           
HR0.22*0.28*          
BMI0.43*0.27*0.19*         
WHR0.44*0.34*0.19*0.64*        
D0.47*0.30*0.170.84*0.63*       
Insulin0.37*0.24*0.29*0.53*0.44*0.39*      
Glucose0.33*0.25*0.120.40*0.28*0.44*0.04     
IR0.30*0.150.24*0.45*0.38*0.31*0.97*–0.12    
TG0.31*0.24*0.110.41*0.37*0.38*0.34*0.20*0.27*   
CH0.20*0.27*–0.060.070.010.09–0.100.31*–0.130.21*  
LDL0.22*0.30*–0.050.110.040.14–0.090.34*–0.120.100.96* 
HDL–0.32*–0.30*–0.11–0.48*–0.44*–0.49*–0.34*–0.26*–0.27*–0.55*–0.02–0.19*

When plotted two-dimensionally ( Fig. 2; Kruskal’s stress = 8.7%), the anthropometric cluster is collected around –0.63 to –0.53 in dimension 1; the metabolic cluster of insulin, glucose, IR ratio and triglycerides around –0.89 to –0.03; the cholesterol cluster around 0.59–0.73; and the haemodynamic cluster around –0.11–0.28; whilst HDL is again found alone at about 2.26.

Figure 2.

The correlation between risk factors weighted for a high variability of diurnal cortisol displayed as a two-dimensional plot derived by multidimensional scaling. BMI, body mass index; CH, cholesterol (total); D, abdominal sagittal diameter; DBP, diastolic blood pressure; HDL, cholesterol (HDL); HR, heart rate; IR, insulin/glucose ratio; LDL, cholesterol (LDL); SBP, systolic blood pressure; TG, triglycerides; WHR, waist/hip circumference ratio.

The results of weighting for a low variability in the diurnal cortisol are shown in Table 4. In this case, all correlations became significant, with the exception of the glucose–IR and glucose–LDL relationships. The HDL correlations are consistently negative.

Table 4.  The correlation matrix between risk factors weighted for a low variability of diurnal cortisol. Marked (*) correlations are significant at P < 0.05
 SBPDBPHRBMIWHRDInsulinGlucoseIRTGCHLDL
  1. BMI, body mass index; CH, cholesterol (total); D, abdominal sagittal diameter; DBP, diastolic blood pressure; HDL, cholesterol (HDL); HR, heart rate; IR, insulin/glucose ratio; LDL, cholesterol (LDL); SBP, systolic blood pressure; TG, triglycerides; WHR, waist/hip circumference ratio.

DBP0.88*           
HR0.48*0.45*          
BMI0.72*0.67*0.44*         
WHR0.65*0.62*0.35*0.73*        
D0.75*0.68*0.48*0.92*0.75*       
Insulin0.46*0.40*0.49*0.42*0.33*0.46*      
Glucose0.77*0.55*0.43*0.55*0.57*0.64*0.35*     
IR0.22*0.23*0.38*0.24*0.15*0.25*0.94*0.04    
TG0.44*0.38*0.30*0.50*0.48*0.57*0.32*0.50*0.18*   
CH0.29*0.38*0.20*0.24*0.28*0.27*0.25*0.13*0.25*0.25*  
LDL0.18*0.29*0.16*0.16*0.14*0.16*0.25*0.020.29*0.06*0.95* 
HDL–0.16*–0.16*–0.21*–0.31*–0.12*–0.33*–0.38*–0.24*–0.31*–0.58*–0.07*–0.16*

When these results are plotted two-dimensionally ( Fig. 3; Kruskal’s stress = 3.3%), it is seen that they are all assembled around –0.31 to –0.25 in dimension 1, and within a narrow range in the second dimension (–0.030–0.044). The HDL cholesterol is further out at 3.46 in dimension 1.

Figure 3.

The correlation between risk factors weighted for a low variability of diurnal cortisol displayed as a two-dimensional plot derived by multidimensional scaling. BMI, body mass index; CH, cholesterol (total); D, abdominal sagittal diameter; DBP, diastolic blood pressure; HDL, cholesterol (HDL); HR, heart rate; IR, insulin/glucose ratio; LDL, cholesterol (LDL); SBP, systolic blood pressure; TG, triglycerides; WHR, waist/hip circumference ratio.

Discussion

This study was performed with the aim of highlighting the importance of the HPA axis in human health. This was accomplished by introducing the function of the HPA axis in terms of cortisol variability amongst the established risk factors for CVD, type 2 diabetes and stroke. A normal diurnal cortisol variation is a pattern in which cortisol levels are high and varying in the morning, and from 16.00 h to midnight less than 75% of the morning values [26]. Patients with Cushing’s syndrome or Addison’s disease do not exhibit this diurnal variation [45].

The assessment of HPA axis function in the present study included measurements of cortisol levels under basal conditions and functions during provocation by food intake or suppression by dexamethasone. By repeated sampling (n = 7) of salivary cortisol over a random working day, two types of diurnal cortisol secretory patterns were singled out, representing extremes in terms of the resilience and plasticity of the HPA axis. These measurements also included the HPA axis response to the physiological challenge of food intake. The first is characterized by a high morning cortisol peak, a normal circadian rhythm (high variability) and feedback regulation (dexamethasone) along with a brisk cortisol response to lunch. The other is characterized by the absence of a morning cortisol peak and circadian rhythm (low variability), a blunted suppression of cortisol by dexamethasone and a poor lunch-induced cortisol response. The dexamethasone suppression test correlated strongly with these patterns of cortisol variation. A normal HPA axis is efficiently suppressed by exogenous glucocorticoids such as dexamethasone, whereas a pathological HPA axis exhibits the opposite [27].

In short, this means that the correlations between the established risk factors were performed taking a normal (high cortisol variability) and a pathological (low cortisol variability) HPA axis function into account.

Before interpreting the results, several potential biases should be cautiously examined. Although conditions characterized by an abnormal cortisol secretion may distort the results, the prevalence of Cushing’s syndrome or Addison’s disease in the population studied is extremely low, and subjects with these diseases would be revealed by failure to suppress endogenous cortisol secretion normally when dexamethasone was administered. Furthermore, anorexia is very rare at this age, especially in men. The alcohol history to hand did not reveal any indications of alcoholism. Whilst there is potential for selection bias due to non-response, the non-responders did not differ substantially from the responders in the prevalence of the characteristics examined.

In the correlation analysis between established risk factors, without considering HPA axis function, expected correlations were observed. The risk factors tended to be assembled in clusters, found also in the two-dimensional scaling analysis. This was not appreciably changed when a normal HPA axis function was taken into consideration. When, however, a pathological HPA axis function was included, an entirely different picture emerged. Now essentially all risk factors showed intercorrelations and all lined up within a narrow range in the two-dimensional scaling analysis. This demonstrates that a poorly regulated HPA axis is tightly associated with the other risk factors measured.

Although epidemiological evidence per se is insufficient to establish causality, these findings provide circumstantial evidence that the functional status of the HPA axis is involved in the expression of established risk factors for CVD, type 2 diabetes and stroke. The following arguments support this view .

The HPA abnormalities described above are most likely associated with abdominal obesity and insulin resistance, as seen in Cushing’s syndrome, via mechanisms which have been reviewed recently [28]. Furthermore, potential causal associations between insulin resistance and hypertriglyceridaemia, glucose homeostasis and blood pressure have been put forward [2930]. There are thus reasonable possibilities that a malfunction of the HPA axis of the kind described here might actually be causally connected to several of the risk factors analysed.

During the last decades, increasing attention has revealed that upper body fat distribution is a strong risk predictor of CVD, type 2 diabetes and stroke in both men and women [31–34]. Since there is considerable evidence that cortisol may direct excess body fat to visceral depots [28], abdominal obesity may be explained by HPA axis perturbations [28]. In fact, abdominal obesity may be an indicator of a malfunction of the HPA axis, which might be the primary pathogenetic factor.

In the two-dimensional scaling analyses, HDL was isolated from the other risk factors in both the metabolic cluster (insulin, glucose and triglycerides) and the cholesterol cluster (total and LDL cholesterol). This might be due to the consistent negative relationship to other risk factors, and HDL levels are known to be inversely correlated to the risk of developing CVD [35]. This might also be due to different regulatory mechanisms for HDL than for other lipoproteins [36].

In summary, the results of this study indicate that established risk factors for CVD, type 2 diabetes and stroke form anthropometric, metabolic and haemodynamic clusters in correlation analyses, which is an expected finding that has been reported repeatedly previously. When the function of a normal HPA axis was considered, as described above, this clustering did not change. However, when a pathological HPA axis in terms of a low morning cortisol value and diurnal variability as well as poor dexamethasone suppressibility was considered, another picture emerged. In this case, most of the risk factors intercorrelated and seemed, with the exception of a low HDL, to form one tightly assembled cluster. This might mean that a malfunctioning HPA axis regulation is in fact an overriding factor for most other, established risk factors.

As mentioned in the introduction, the HPA axis responds sensitively to external stimulation [2727]. A common, powerful group of activators are those included under the concept of stress. As the homeostasis is constantly threatened by internal or external adverse factors – stressors – stress is usually defined as a state of threatened homeostasis [27]. There are physical stressors such as cold, trauma, fever and infection; and psychological stressors such as social subordination, anxiety and depression [7]. We have recently shown that stress-related cortisol secretion based on low cortisol variability is strongly related to anthropometric, metabolic and haemodynamic risk factors, as well as to endocrine abnormalities other than those of the HPA axis [15]. Moreover, we have identified a subgroup of elevated BMI, WHR and D, where a blunted dexamethasone response is found, associated with traits of anxiety and depression [14]. The findings described in this report might therefore give a clue to how environmental stress affects human health. We have previously observed associations with psychological and socioeconomic handicaps in the men examined which would be expected to provide such an environment [1112].

Acknowledgements

The authors thank (in alphabetic order) Christina Jonsteg, George Lappas, Katarina Romanus, Erik Tengdahl and Inga-Lill Åhs for excellent technical assistance. The study was supported by grants from the Swedish Medical Research Council (K97-19X-00251-35A).

Received 19 November 1998; accepted 22 July 1999.

Ancillary