Is the Direct Association Between Depression and CHD Plausible?
Depression has various direct and indirect biological effects which may explain its link with cardiovascular disease. Current evidence suggest that depression may act through an autonomically-mediated increase in myocardial workload (and thus ischemia), or in ventricular irritability; diminished beat-to-beat heart rate variability; or increased atherogenesis, through relatively direct effects on platelet function or hypercortisolemia-mediated effects on other cardiovascular risk factors such as blood pressure or lipid levels [9, 10]. Vascular endothelial dysfunction of the coronary arteries has also been observed in a number of cohorts of depressed patients .
Depression-related dysregulation of the sympathoadrenal system and HPA axis has been observed to cause an increase in circulating catecholamines and serum cortisol resulting in surges in heart rate and blood pressure, which increases the risk of atherosclerotic plaque rupture and acute coronary thrombosis . High levels of catecholamines also increase the irritability of the heart muscle which can lead to ventricular arrhythmias and modify the function of circulating platelets [9, 11]. Increased cortisol levels lead to inflammation, excessive clotting, metabolic syndrome, induction of hypercholesterolemia and hypertryglyceridemia, and injury of vascular endothelial cells [2, 9]. Effects on serotonin metabolism have also been shown to lead to additional alteration of cardiac function, increased platelet activity, and vasoconstriction . Changes in inflammatory immune response, such as raised inflammatory markers, may also predict the development of cardiovascular disease [2, 11].
A number of behavioral pathways have been identified in the link between mental disorders and cardiovascular disease. Patients who suffer from a mental disorder are less likely to adopt a healthy lifestyle or to reduce cardiovascular risk factors. Physical activity and diet control are less frequently undertaken and smoking and alcohol consumption rates are often higher [2, 11–13]. In addition, obesity has been strongly associated with both cardiovascular disease and mental disorders such as depression, bipolar disorder and panic disorder . Adherence to treatment is often less in depressed patients compared to non-depressed patients . Depressed patients are known to access social services less frequently and insufficient social support has been strongly linked to the development of cardiovascular disease .
Establishing the temporal relationship of depression antecedent to CHD is a challenge. Both depression and early stages of cardiovascular disease can go undiagnosed for a substantial period of time. In addition, the relationship between depression and CHD appears to be bi-directional. Increased prevalence of depression post-cardiac event and its negative prognostic effect on cardiac patients has been clearly demonstrated [14, 15]. Proof of a temporal relationship can be provided by only including prospective studies in which patients with previous CHD are excluded from the study, or at the very least excluded from the analysis. However, Rugulies and colleagues  tested this hypothesis by pooling effect sizes for all relevant studies and then for studies which control for sub-threshold CHD by intensive medical examinations at baseline. They report a small reduction in the effect size (from 1.77 to 1.51) but it remains statistically significant.
It can further be argued that if sub-threshold CHD were present at baseline, relative risk (RR) of outcome would be higher within a short period of time and tail off with longer follow-up. The literature does not support this. Data from the John Hopkins Precursors Study showed that risk of myocardial infarction (MI) after 40 years of follow-up did not differ from risk after only 10 years (RR of 2.12 and 2.1, respectively) . Implications of these findings are that possible sub-threshold CHD at baseline does not significantly affect the risk of incident CHD in depressed samples.
Evidence for Association between Depression and CHD
Several reviews of the association between depression and incident CHD have been published. Specifically four meta-analyses have been published, reporting pooled relative risk for depression and incident CHD between 1.57 and 2.69 [3, 5–7]. A further four qualitative reviews were identified, all reporting a high degree of evidence for the role of depression in the development of CHD [13–15, 17]. A number of authors conclude that the risk of depression on incident CHD may be equal to that of smoking [6, 7, 17].
Three other qualitative reviews [9, 11, 12] document increased cardiovascular morbidity and mortality in patients with depressive symptoms or major depression. A strong suggestion of a dose-response relationship between depression and CHD was identified [5, 7, 11, 14]. Depression meeting diagnostic criteria was associated with a higher risk of CHD compared to depressive symptoms [5, 7]. Existence of a dose-response relationship between exposure and outcome supports likelihood of causation .
Association is stronger with MI than more broadly defined CHD
A comprehensive meta-analysis based on 28 studies reported a pooled relative risk of 1.60 (95%CI 1.34-1.92) for MI compared to 1.48 (95%CI 1.29-1.69) and 1.63 (1.26-2.12) for CHD and CVD, respectively . It was only when the outcome was narrowed to MI alone that heterogeneity between studies was eliminated. This suggests that classification of outcome (MI), as opposed to other factors was an important source of heterogeneity between the studies. An important limitation of this review is that not all included studies controlled for CHD at baseline. Yet the implication of this finding is that the evidence for depression as an independent risk factor is more consistent for incident MI than for CHD.
The literature on depression and cardiovascular disease displays large heterogeneity. Lack of consistency in methodology makes the task of comparing studies and pooling data for meta-analysis difficult and needs to be carefully considered in any systematic review. For example, it has been shown that the prognostic impact of initial depression increases over time, with higher odds ratios for mortality post diagnosis of depression in CHD in studies with longer follow-up periods .
Defining and Measuring Depression
For the purposes of the GBD, depressive disorders are required to meet the Diagnostic and Statistical Manual of Mental Disorders (DSM)  diagnostic criteria for major depressive disorder (F32-F33) or dysthymic disorder (F34.1), or the WHO International Classification of Diseases (ICD) criteria for depressive disorders including dysthymia (F32.0-F32.2, F32.8, F33.0-F33.2, F33.8, F33.9, F34.1).
However, identification and measurement of depression is complex and the refinement of adequate, quantitative measures is ongoing. Measures of depression used in published studies range from self-report of mood to strict adherence to diagnostic criteria, with a myriad of other surveys and questionnaires somewhere between and varying on the spectra of reliability and validity [11, 21].
There exists variation in the cut-off values of 'diagnostic' scoring within studies which use the same measurement instrument [22–25]. The ability to compare findings from various studies employing differing measurement methods is therefore compromised. Case identification according to DSM or ICD criteria is the ideal. Where a study falls short of this, for example in large cohort studies, a diagnosis by a clinician or the use of a measurement tool which has been shown to acceptably represent DSM or ICD criteria is considered sufficient for the case ascertainment of depression. Additional complication comes about by variations in the period of assessment (for example, depression experienced during the past-week versus past 12-months).
Defining and Measuring Coronary Heart Disease
CHD is considered to encompass ischemic heart disease, myocardial infarction and angina, included cases of ICD-9 codes 410-414 and 429, or ICD-10 codes 120-125. The WHO criteria for myocardial infarction (MI) are also accepted. While the term Coronary Heart Disease was recommended by Tunstall-Pedoe (2001) , many studies dealing with the relationship between depression and disease of the coronary vessels use alternative nomenclature, such as Coronary Artery Disease (CAD) or Ischemic Heart Disease (IHD). All essentially refer to the disease process of atherosclerosis of the coronary arteries leading to myocardial infarction and angina .
MI provides a very well defined clinical marker, both in terms of a clearly definable 'hard' endpoint for cardiovascular disease , as well as a marker in time, allowing temporal correlation between MI and depressed mood.
Although case-definitions for CHD are clear in theory, signs of CHD can be very difficult to detect and subclinical in nature. Hence ruling out CHD at baseline and proving an unbiased etiological relationship is challenging . This is particularly true for angina and it is for this reason that angina was not accepted as a proxy for CHD.
Is There Residual Uncontrolled Confounding?
Although the obvious confounding variables should be controlled for as a requirement for inclusion in a quantitative review, residual uncontrolled confounding cannot be ruled out due to the complex nature of the depression and CHD relationship. For example, a depressive state may be associated with other CVD risks such as smoking and other chronic medical conditions.
Reviews of the literature reporting associations between psychosocial factors and CHD found strength of evidence for depression as a causative factor to be strong compared to anxiety where the effect was less consistent [2, 15]. A role for anxiety in the development of CHD is biologically plausible. Anxiety, as with depression, affects the autonomic regulation of the heart via the hypothalamic-pituitary-adrenocorticol (HPA) axis [10, 29]. Chronic over-activation of these systems due to anxiety would help to explain the increased frequency of cardiovascular risk factors in patients with anxiety disorders . Reduced heart rate variability in those with anxiety disorders could also involve impaired vagal control . In the hypertensive state a chronic anxiety disorder may trigger increased levels of epinephrine and norephinephrine which could also increase the heart rate, total peripheral resistance and even increase the circulating fatty acids . There is significant evidence which links anxiety disorders to sudden cardiac death [30, 31]. However, anxiety has not been associated with myocardial infarction in these studies [12, 31].
While some reviews have found an association between anxiety and CHD , they included fewer of the primary studies with negative findings than the review by Kuper et al [15, 33]. We have concluded there is neither strong nor consistent evidence for a causal association between anxiety disorders and CHD .
However, depression and anxiety are strongly co-morbid [34, 35]. Hence we need to consider anxiety as a confounder in the relationship between depression and CHD. Given the conclusion above that anxiety disorders (alone) do not have a proven etiological relationship with CHD we examined the evidence for the relationship between major depressive disorder and CHD controlling for other mental disorders. In one study the odds ratio (OR) for incident MI was only slightly reduced once phobia, panic disorder and alcohol use/dependence were controlled for (4.16 95%CI 1.49-11.62 compared to 4.54 95%CI 1.65-12.44) . We therefore concluded that there is insufficient evidence for co-morbid anxiety substantially mediating the relationship between depression and CHD.
A number of studies have reported an association between use of anti-depressant medication and fatal and nonfatal CHD [37, 38]. This raises the issue as to whether anti-depressant medication, rather than depression, is a risk factor in development of CHD. A limitation of many earlier papers considering the role of antidepressants and CHD [37, 38] is that few controlled for the presence of depression or other mental disorders. More recent studies examining depression and CHD have controlled for the use of medications. Retrospective analysis of a primary care sample with IHD found that use, dose and duration of tricyclic anti-depressant consumption were not statistically significant in the development of ischemic heart disease when diagnosis and duration of depression were included in the regression model . In a prospective follow-up of the Baltimore Epidemiologic Catchment Area reported use of psychotropic medication had a marginal effect on the RR of incident MI in a depressed sample (RR 4.16 compared to 4.14) .