Of the 25 895 respondents to the baseline survey in 1998, 234 had moved abroad and could not be included in the follow-up. Data on CHD and CBVD were linked to survey responses from national health registers on the basis of a written consent from 24 128 (93%) participants. A total of 23 282 participants with complete data on covariates were included in the analyses. A total of 203 incident CHD events (fatal or non-fatal ischaemic heart disease events) and 129 incident CBVD events (fatal and non-fatal stroke events: 27 subarachnoid haemorrhages, 23 intracerebral haemorrhages, 55 cerebral infarction, 3 other non-traumatic intracranial haemorrhage, 19 other CBVDs, 2 sequelae of cerebral infarction) were documented during the follow-up.
presents the differences in CHD, CBVD and depression as a function of baseline sample characteristics. CHD and CBVD events were higher in men, older participants, those with lower educational level, current smokers (not for CBVD), high-alcohol consumers, obese participants and those with a sedentary life style and with hypertension or diabetes (P ≤ 0.026). Depression among participants at baseline (BDI score ≥10 or filling antidepressant prescriptions) was more likely among women, older people, those with a lower education, current smokers, high alcohol consumers, obese people, those more likely to have a sedentary life style and those with hypertension or diabetes (P ≤ 0.01).
Number of incident CHD, CBVD and depression as a function of covariates
presents the associations between depression and subsequent CHD and CBVD events using Cox regression analysis. In model 1, when adjusted for socio-demographic variables, the HR for CHD was 1.66 [95% confidence interval (95% CI) 1.24–2.24] for participants with mild to severe depressive symptoms and 2.04 (95% CI 1.27–3.27) for participants who filled antidepressant prescriptions when compared with those without depression markers. The risk of CHD for 1-unit increase in the score on the BDI (continuous variable) was 1.04 (95% CI 1.02–1.06). For CBVD, the HR adjusted for socio-demographic factors was 1.01 (95% CI 0.67–1.53) for participants with mild to severe depressive symptoms, 1.77 (95% CI 0.95–3.29) for those who filled antidepressant prescriptions and 1.01 (95% CI 0.99–1.04) for 1-unit increase on the BDI (continuous variable). In models 2 and 3, adjusted for behavioural and biological risk factors respectively, these associations were attenuated but the association between depression and CHD was robust to these adjustments. In model 4, including simultaneously adjusted for all aforementioned variables, these associations were further reduced and only the association with CHD remained evident. The corresponding fully adjusted HRs were 1.47 (1.08–1.99) and 1.72 (1.06–2.77) for CHD. For CBVD, these HRs were 0.87 (0.57–1.32) and 1.52 (0.81–2.84).
Depression as a predictor of CHD and CBVD
To test the robustness of our findings, we repeated the analyses excluding CHD and CBVD events that occurred in the first 2 years of follow-up. These analyses provided a similar pattern of associations as those presented in . For CHD, the number of events was reduced by 31% (n = 145), but the unadjusted HR for participants with mild to severe depression (BDI score ≥10) and for those who filled prescriptions for antidepressant drugs remained at 2.08 (P < 0.001) and 2.15 (P < 0.006) when compared with those who were not depressed. The corresponding fully adjusted HRs were 1.94 (P < 0.001) and 2.00 (P = 0.015). For CBVD, the number of events was reduced by 26% (n = 97) and the corresponding unadjusted HRs were 1.03 (P = 0.898) and 1.94 (P = 0.060). The fully adjusted HRs were 0.89 (P = 0.626) and 1.66 (P = 0.155).
We also repeated the analyses excluding definite angina (ICD-10 code I20) events and considering only fatal and non-fatal myocardial infarction (n = 142). The HRs from model 1 for participants with mild to severe depression (BDI score ≥10) and for those who filled prescriptions for antidepressant were 1.95 (P < 0.001) and 2.56 (P < 0.001), respectively, when compared with those who were not depressed. The corresponding fully adjusted HRs were 1.62 (P < 0.008) and 2.08 (P = 0.007). These results are highly consistent with those obtained including definite angina, leading us to conclude that the results reported here are not driven by ‘soft’ endpoints.
In addition to these analyses, we examined the association between the severity of depressive symptoms and the risk of CHD and CBVD by using the standard cut-offs as follows:23
scores of 0–9 indicated no depression, 10–18 indicated mild depression, 19–29 indicated moderate depression and 30–63 indicated severe depression. For CHD, the unadjusted HRs were 1.57 (P
= 0.008) for participants with mild, 1.81 (P
= 0.029) for those with moderate and 2.80 (P
= 0.042) for those with severe depressive symptoms. The corresponding fully adjusted HRs were 1.45 (P
= 0.0325), 1.58 (P
= 0.097) and 2.15 (P
= 0.784). For CBVD, the corresponding unadjusted HRs were 0.93 (P
= 0.7553), 0.99 (P
= 0.684) and 2.68 (P
= 0.094) and fully adjusted HRs were 0.82 (P
= 0.435), 0.79 (P
= 0.577) and 1.97 (P