"Stress" and coronary heart disease: psychosocial risk factors

There is intense public interest in possible links between "stress" and coronary heart disease (CHD). Until recently, organisations such as the National Heart Foundation of Australia have only been able to make judgements based on limited data in this area.

In 1988 the National Heart Foundation of Australia published a report, "Stress and cardiovascular disease", which concluded that, although acute catastrophic events might trigger acute myocardial infarction or sudden death, there was insufficient existing evidence from prospective studies that any form of "stress" consistently predicted the subsequent development of CHD.1 The report concluded that psychosocial risk factors had effects on conventional risk factors, but no independent effect.

Since then, a considerable number of prospective cohort studies have examined the links between various forms of stress and the development and prognosis of CHD; there has also been a multitude of reviews, both narrative and systematic. However, these reviews have used different methods and at times have come to different conclusions. Because systematic reviews attempt to find, appraise and summarise the findings of all studies in a systematic and transparent way, these reviews should be the more reliable. Unfortunately, the reported systematic reviews have varied in their quality and come to different conclusions. Recently, methods for critically appraising systematic reviews have been developed, and this position statement is based on a review of the systematic reviews using this methodology.2,3

An Expert Working Group considered all the major suggested psychosocial risk factors ("stressors") to identify evidence of independent associations with CHD.

Although the term "stress" is in general use, it is so imprecise that, in agreement with other review groups,4 the Expert Working Group examined separately those variables that are commonly regarded as components of stress. These include:

• depression, anxiety, panic disorder;

• social isolation and lack of quality social support;

• acute and chronic life events;

• psychosocial work characteristics; and

• Type A behaviour, hostility.

The methods used in formulating this position statement are outlined in Box 1.

There was strong and consistent evidence across all the reviews that depression is an independent risk factor for clinical CHD and its prognosis (evidence rating A; Box 2). The association exists for men and women, subjects living in different countries, and various age groups. Furthermore, the CHD risk is directly related to the severity of depression: a 1–2-fold increase in CHD for minor depression and 3–5-fold increase for major depression (evidence rating A; Box 2).4,10,11,12,13 The strength of the association is of similar magnitude to that of standard risk factors such as smoking or hypercholesterolaemia.

There is strong and consistent evidence across all the reviews that social isolation and lack of quality social support are independent risk factors for CHD onset and prognosis: the risks are increased 2–3-fold and 3–5-fold, respectively (evidence rating A; Box 2). The association exists for both men and women, subjects living in different countries, and various age groups. An association was found in studies that examined some aspect of the size and nature of a person's social network and in studies that examined the type of support received (evidence rating A; Box 2).4,11,14,15

Acute life event "stressors" can trigger CHD events, although it is very difficult to study and quantify the magnitude of effects. Acute "stressors" include significant common events such as bereavement (evidence rating B; Box 2),11 as well as catastrophic events such as earthquakes or terrorist attacks (evidence rating A; Box 2).11,15,16 Although the deleterious physiological effects of acute "stressors" as CHD triggers are well documented, the role of chronic "stressors" in CHD onset and prognosis remains unclear.

This topic refers specifically to the characteristics of the work environment as distinct from the life-event "stressors" referred to above. The studies included in one review4 under psychosocial work characteristics were heterogeneous, with a wide variety of factors being examined individually and collectively. When the results for job control, demands and strain were recalculated, there was not a preponderance of positive over negative studies. The Expert Working Group found no consistency between this review4 and the other two reviews of work-related "stressors".11,17

Reasons for the discordance between the reviews of prospective studies in healthy populations were explored by following a set of steps applicable to all types of systematic reviews, including aetiological and prognostic studies, developed from an algorithm devised to interpret discordant meta-analyses of intervention studies.8 Two of the reviews4,11 covered the job-strain model, job control and the effort–reward model, whereas the third review17 covered only the job-strain model. Of the first two reviews, one4 included twice as many studies as the other11 and summarised their findings more fully. Consequently, the Expert Working Group gave more credence to this "negative" review,4 and concluded that there was neither strong nor consistent evidence of a causal association between work-related "stressors" and CHD (evidence rating C; Box 2).

Type A behaviour pattern refers to a number of personality trait characteristics, including rushed, ambitious and competitive behaviour, impatience, hostility, and intolerance.18 Early positive studies have now been displaced by a large number of studies concluding that Type A behaviour pattern has no effect (evidence rating A).4

One review of prospective studies concluded that there was consistent positive evidence of association between hostility and CHD.13 Two other reviews reported an almost equal number of positive and negative prospective studies in healthy populations.11,19 The most recent review concluded that there was no evidence of association.4

When the discordance between these reviews was examined, we found that the review that found no clear association between hostility and CHD4 included 2–6 times as many large studies as the other reviews, and that the other reviews had only 2–4 primary studies in common with the most recent review.4 As well as including several more recent studies, this review included studies with better measures of hostility and more studies of the general population. Its inclusion of studies of Type A behaviour patterns18 did not account for the preponderance of "negative" studies. The Expert Working Group therefore gave greater credence to this better-quality "negative" review and considered that hostility is not a risk factor for CHD (evidence rating C; Box 2).

A review of primary studies where anxiety was the specific exposure4 (rather than anxiety associated with depression) found an equal number of positive and null findings among both the aetiological and the prognostic studies and concluded there was no association with CHD. Other reviews came to the opposite conclusion or were equivocal.11,12

When the reasons for the discordance between the reviews of aetiological studies were explored, it was found that the reviews which had concluded that there was12 or may be11 an association between anxiety and CHD had included fewer of the "negative" primary studies than the review which concluded that there was no clear association.4 This latter review also summarised the primary studies more fully. For those reasons the Expert Working Group gave more credence to that review.4

In addition, when the reasons for the discordance between the two reviews of prognostic studies were explored, it was found that the review which had concluded that there was no clear association between anxiety and the prognosis of CHD had included 18 large primary studies,4 whereas the review which concluded that there was an association11 included only four primary studies, two of which were small and one which included patients with cardiopulmonary disease. The Expert Working Group therefore gave greater credence to the negative review.4

Patients with panic disorder are subject to episodes of recurring, often inexplicable, psychophysiological arousal. The one review of this area found little evidence to link panic disorder with either CHD development or progression.20

The Expert Working Group concluded there was neither strong nor consistent evidence of a causal association between anxiety and panic disorders and CHD (evidence rating C; Box 2).

This position statement is based on a review of systematic reviews and therefore depends on the rigour with which relevant primary studies have been identified, appraised and summarised. The Expert Working Group found that differences in the primary studies included in the reviews was a common explanation for discordance between them.

Systematic reviews rely almost entirely on published studies and are therefore potentially limited by an important source of bias. Reviews of published data, particularly observational studies, may be misleading. There is no mechanism for identifying the results of unpublished studies, or published studies that have data on psychosocial variables and CHD outcomes but do not report it. This highlights the need for improved search methods. The establishment of an international registry for such studies, with advanced lodgement of study protocols, could address this important issue.21

Intervention studies are not addressed in this position statement. The Expert Working Group found few good intervention studies which addressed single "stressors" and which could provide experimental evidence for or against a causal association. The lack of evidence of the efficacy of specific interventions for depression, social isolation or lack of social support in people with coronary risk factors or after coronary events is an important area that needs further research.

Although one or two reviews cited examples of the effect of a combination of psychosocial risk factors on CHD, none considered the issue systematically, as few primary studies published data on this topic. Thus, the Expert Working Group could do no more than note the possibility of the clustering of psychosocial risk factors.