Introduction
Various negative psychological traits have previously been associated with cardiovascular disease (CVD) incidence, CVD mortality, and all-cause mortality. Early studies reported positive associations between type A behavior and CVD incidence and mortality (e.g.,
1,
2), and studies since have reported positive associations between hostility and CVD [
3‐
6], hostility and all-cause mortality [
6], and anger expression and CVD [
3,
6,
7]. Depression has been repeatedly positively associated with CVD and all-cause mortality (e.g.,
4,
8‐
12), and studies have demonstrated various associations between CVD and a variety of other psychological traits (
4,
6,
13 and see
14 for a review).
Explanations for associations between psychological traits, CVD, and mortality are largely based in physiology. Negative psychological traits and CVD are most commonly linked via increases in autonomic nervous system (ANS) responding to stress or ‘cardiac reactivity’ (increases in heart rate, systolic, and diastolic blood pressure in response to stress), and via increased activity of the hypothalamic-pituitary-adrenal (HPA) axis, with subsequent impacts on immune and inflammatory systems [
4,
9,
10,
15‐
17], although other pathways may also play a role [
4,
17]. In relation to increased ANS responding, Pizzi et al., 2008 [
4], found differences in heart rate variability between depressed and non-depressed individuals, type A behaviors, hostility, anger, and aggression have all been associated with increased cardiac reactivity [
14,
18], and individuals defined as hostile or angry compared to others have demonstrated increased heart rate and increased blood pressure in response to stressful tasks [
19,
20]. In relation to the HPA axis, Pope & Smith, 1991 [
21], reported increased cortisol levels in hostile compared to low-hostile men during everyday activities, Suarez et al., 1998 [
22], showed increased cortisol in hostile men under stress compared to low-hostile men, and Steptoe et al., 2000 [
23], found increased free cortisol in angry individuals compared to non-angry individuals when under high job strain. In relation to immune and inflammatory activity, Pizzi et al., 2008 [
17], found differences in the levels of various inflammatory markers between depressed and non-depressed individuals, Howren et al., 2009 [
16], found positive associations between depression and various inflammatory markers in a meta-analysis, and Stewart et al., 2008 [
24], found associations between depression, hostility, and inflammatory markers, in combination. Empana et al., 2005 [
11], also found positive associations between depressive symptoms and inflammatory markers after adjustment for classic CVD risk factors. Cardiac reactivity, increased cortisol and increased inflammatory responses have also been demonstrated in association with CVD and mortality [
4,
11,
17,
25,
26].
Behavioral explanations, however, are also possible. Various negative psychological traits and CVD are associated with poor health behaviors, such as smoking, poor diet, and low physical activity [
3,
6,
27], and it may be these poor health behaviors that account for the relationships between negative psychological traits and CVD incidence and mortality. DiMatteo et al., 2000 [
28], demonstrate associations between depression and anxiety and poor treatment compliance. Scherwitz et al., 1992 [
29] report increased smoking, marijuana use, alcohol, and energy intake in hostile compared to low-hostile individuals. Shin et al., 2005 [
30], find a strong association between anger and poor sleep, Falk Dahl & Dahl, 2010 [
13], report associations between anxiety, alcohol problems, and low levels of exercise, and Siegler et al., 1992 [
31], demonstrate positive associations between hostility and smoking, BMI, lipid levels, and caffeine intake 21–23 years later. In relation to disease and mortality, previous analyses of the cohort under consideration have shown associations between smoking and CVD events and all-cause mortality, after controlling for various demographic variables [
32], associations between alcohol intake patterns and CVD, after controlling for demographic and classic CVD risk factors [
33], and a contribution from smoking, fruit and vegetable intakes, physical activity, and alcohol intakes to the observed socioeconomic gradient in CVD events and all-cause mortality after controlling for demographic and classic CVD risk factors [
34].
Associations between psychological traits and disease outcome, however, are complicated by possible overlap between psychological variables, and possible interactions between psychological variables and behavior [
3,
6,
7,
19].
Particular emphasis has also previously been given to the conciliatory role of the more positive psychological variable of social support. Social support is well-recognized to have positive health benefits [
35,
36]. In relation specifically to CVD, emotionally supportive social support has been found to protect against CVD [
4], and low levels of emotional support have been associated with negative health outcomes [
4,
37]. Rosengren and colleagues [
37], for example, reported a protective effect of social support for new CVD events in middle-aged men, and Krumholz and colleagues [
38], showed emotional support as a strong predictor of cardiac events in hospital patients when accounting for disease severity on hospital admission and instrumental help for assisting living. Social support has also been found to result in reduced ANS reactivity, reduced cortisol reactivity in response to stressors, increased inflammatory and immune responding [
36,
39‐
42], and has been linked to CVD and mortality via these mechanisms. Effects of stress on cardiac function have been found to be lower in those with social support compared to those without [
35], and social isolation in combination with high stress has been found to predict mortality in post Myocardial Infarction (MI) patients [
43].
Interactions with negative psychological traits, however, may limit the impact of social support in certain individuals [
20,
44]. Of relevance to the preceding discussions, depression, hostility, and anger particularly, may undermine the possible beneficial influence of social support [
44]. Depression is often characterized by social withdrawal and a negative interpretation of the actions of others [
28]. Hostility and anger are defined by mistrust and cynicism of others, a negative interpretation of others, and often also a negative attitude or reaction to others [
6,
20]. Social withdrawal, negative interpretations of others, including authority, and negative attitudes and behaviors towards others will limit the possibility of and possible benefit attained from social support by depressed, hostile, or angry individuals [
6,
28,
39]. Holt-Lunstad and colleagues [
20], for example, demonstrate decreased ratings of the friendliness of friends from hostile compared to non-hostile individuals despite no differences in observed friend behaviors. Lepore [
39] also demonstrated that hostile individuals do not benefit from the support of strangers, while non-hostile individuals do benefit.
Associations between psychological traits and CVD have been found not only in relation to CVD incidence but also for CVD prognosis and progression [
3,
4,
9]. Impact on the relationships between negative psychological traits and CVD by behavioral variables could provide valuable suggestions for intervention, both for the prevention of CVD and for its treatment [
3,
6]. This analysis aimed to investigate the impact of behavioral variables on the associations between negative psychological traits and CVD incidence, CVD mortality, and mortality from all-causes. Four lifestyle behaviors were considered—smoking, alcohol intake, fruit and vegetable consumption, and physical activity. Analyses were conducted both with and without prior accounting for other psychological variables, including a measure of social support.
Discussion
These findings firstly demonstrate relationships between both depression and hostility and CVD mortality and all-cause mortality, when adjusting for demographic and biological risk factors. These relationships have previously been demonstrated elsewhere (e.g.,
3,
9,
12).
Secondly, these relationships were attenuated by the inclusion of lifestyle behaviors in the predictive models. These findings demonstrate the importance of behavioral variables in these relationships. Behaviors have previously been hypothesized as the mechanism through which various characteristics and traits such as depression and hostility affect disease and mortality. Whooley and colleagues [
51], for example, find no association between depression and CVD events after controlling for alcohol use, smoking, physical activity, and medication non-adherence. Chida & Steptoe [
3] find no association between hostility and anger and CVD incidence after controlling for smoking, physical activity, and BMI as well as socio-economic status. Chida & Hamer [
15] also found a reduction in associations with cardiac reactivity after controlling for behavioral variables. A role for behavior may have considerable implications for treatment and secondary prevention. Behaviors may offer an alternative route for intervention than that offered through the treatment of psychological variables. Repeated research shows beneficial impacts of behavioral interventions (e.g.,
52).
Effects sizes are small, but these effect sizes represent only the direct effects of the considered lifestyle behaviors on the relationships between depression, hostility, and all outcome variables. Additional effects are also likely, as a result of impacts of the lifestyle behaviors on demographic and biological risk factors and on the psychological variables themselves. Fruit and vegetable intake and physical activity, for example, are known to impact on many of the biological risk factors for CVD, including some of those controlled in analyses here—blood pressure, cholesterol levels, BMI, and diabetes (e.g.,
53‐
57). Smoking and alcohol use can also have impacts on these biological risk factors (e.g.,
58,
59). The lifestyle behaviors investigated may also impact on depression, hostility, and social support (e.g.,
60‐
62). The distinction between behavioral, biological, and psychological risk factors is often unclear, and relationships between all three are likely to be much more complex than is suggested by our analyses. The lifestyle factors investigated furthermore may only represent a subset of those behaviors that impact on the relationships between psychological health and disease/mortality (e.g.,
28‐
31).
Of the behaviors investigated, smoking contributed most significantly to mortality, although evidence for a role from fruit and vegetable intake and high alcohol consumption was also found. The important contribution of smoking to CVD disease and mortality is well-recognized (e.g.,
32). Associations between fruit and vegetable intake and CVD disease, mortality, and all-cause mortality are also well-known (e.g.,
63,
64), and associations between alcohol intake and all-cause mortality are well-recognized (e.g., [
65]). Of interest in our data, physical activity was not important for CVD incidence or mortality, but this result may be specific to the measurement of physical activity used, and our specific sample characteristics—middle-aged men.
The lifestyle behaviors were found to be important, furthermore, both when other psychological variables were also accounted for and when not. Comparisons between models 1 and 3 suggest that the additional psychological variables do impact on the primary relationships with all outcomes, although effect sizes are smaller than those for the lifestyle behaviors. These small effects may have resulted from the measures used but are likely to also demonstrate the high inter-relation between psychological variables [
43]. The similar patterns in findings for depression and hostility also suggest a close relationship between these variables. Associations between psychological variables have previously been reported [
3,
6,
7,
19,
24].
Interestingly, none of the relationships were attenuated entirely by the inclusion of the lifestyle behaviors in analyses, and relationships between both depression and hostility and all-cause mortality remained significant. Independent associations between depression, hostility, and all-cause mortality have previously also been found elsewhere. Miller and colleagues [
6], for example, also report a significant independent association between hostility and CVD incidence and mortality after controlling for behavioral variables. These authors, however, also suggest that accounting for behaviors in studies may not always be adequate [
6]. Behaviors in addition to those frequently measured may have additional impact. In relation to depression and hostility, medication compliance [
28], and quality as opposed to quantity of social support [
44] are obvious suggestions, but other characteristics and behaviors, such as those related to childhood experiences or exposures, may also have impacts in these relationships [
3]. These early experiences are, however, very difficult to control for.
The strengths of these analyses clearly lie in the prospective nature of the data on which the analyses are conducted, the large sample size and the long (10 year) follow-up period involved. Limitations lie in the measures used for the assessment of the psychological variables, the limited behaviors that were measured, the self-report measures used to study these behaviors, and the possibility that these behaviors may have changed over the course of the study period—assessments of behavior were only made at the start of the study. Psychological variables were not assessed using complete validated measures [
66], but were instead assessed using a composite questionnaire. We have no data to compare our questionnaire scores to the scores of validated questionnaires, but similar levels of depressive symptoms have been reported using the CES-D scale in older (13.2 %) and elderly (21.5 %) general populations from Europe [
67], and repeat analyses using the questions from the Welsh Pure Depression Scale and the questions from the Cook-Medley Hostility scale in place of our composite measures of depression and hostility reveal the same patterns as those presented (data not shown). It is possible that our findings are a result at least in part of our use of a composite measure, but given the comparability between our findings and those of others, we think this is unlikely. The limitations of self-report for our behavioral measures and our use of limited behaviors are acknowledged, but health behaviors are known to typically cluster highly, thus assessment of further behaviors may have little impact on our findings. The assessment of behavior only at the start of the study however, not only limits our abilities to monitor changes over time but also limits our abilities to study these using more formal mediation analyses due to the temporal precedence requirement for mediation models [
68]. While novel procedures are currently under development (e.g., see
69), formal mediation analyses for use in survival (time-to-event) data are currently not well established [
70]. The analysis was also restricted to middle aged (50–59 year old) men, while sex and age differences in CVD, mortality, heath behaviors, negative psychological traits, and social support are well-known [
3,
9,
17,
23,
24,
29,
41].
In conclusion, this analysis demonstrates positive associations between depression and hostility and mortality from CVD and all causes. These associations, however, were reduced when accounting for lifestyle behaviors. These findings demonstrate the importance of health behaviors in the relationships between negative psychological traits and mortality. These findings may suggest possibilities for treatment and secondary prevention.
Acknowledgments
The PRIME Study is organized under an agreement between INSERM and the Merck, Sharpe and Dohme-Chibret Laboratory, with the following participating Laboratories: The Strasbourg MONICA Project, Department of Epidemiology and Public Health—EA1801, Université Louis Pasteur, Faculté de Médecine, Strasbourg, France (D. Arveiler, B. Haas); The Toulouse MONICA Project, INSERM, U558; Department of Epidemiology, Université Paul Sabatier–Toulouse Purpan, Toulouse, France (J. Ferrières, JB. Ruidavets); The Lille MONICA Project, INSERM, U744, Lille; Institut Pasteur de Lille, Lille; Université de Lille 2, Lille, France (P. Amouyel, M. Montaye); The Department of Epidemiology and Public Health, Queen’s University, Belfast, Northern Ireland (A. Evans, J. Yarnell, F. Kee); The Department of Atherosclerosis, INSERM, U545, Lille; Institut Pasteur de Lille, Lille; Université de Lille 2, Lille, France (G. Luc, JM. Bard); The Laboratory of Haematology, INSERM, U626, Marseille, Hôpital La Timone, Marseille, France (I. Juhan-Vague, Pierre Morange); The Laboratory of Endocrinology, INSERM U563, Toulouse, France (B. Perret); The Vitamin Research Unit, The University of Bern, Bern, Switzerland (F. Gey); Nutrition and Metabolism Group, Centre for Clinical and Population Sciences, Queen’s University Belfast, Northern Ireland (Jayne Woodside, Ian Young); The DNA Bank, INSERM U525, Paris, France (F. Cambien); The Coordinating Center, INSERM, U780, Villejuif; France (P. Ducimetière, A. Bingham).
We thank the following organizations which allowed the recruitment of the PRIME subjects: the health screening centers organized by the Social Security of Lille (Institut Pasteur), Strasbourg, Toulouse and Tourcoing; Occupational Medicine Services of Haute-Garonne, of the Urban Community of Strasbourg; the Association Interentreprises des Services Médicaux du Travail de Lille et environs; the Comité pour le Développement de la Médecine du Travail; the Mutuelle Générale des PTT du Bas-Rhin; the Laboratoire d’Analyses de l’Institut de Chimie Biologique de la Faculté de Médecine de Strasbourg; the Department of Health (NI) and the Northern Ireland Chest Heart and Stroke Association.
We also thank the members of the event validation Committees: Pr L. Guize, Dr C. Morrison, Dr M-T. Guillanneuf, Pr M. Giroud; and the Alliance Partnership Programme for its financial support.