Background A single assessment of psychological distress, which includes depression and anxiety, has been associated with increased mortality in patients with coronary heart disease, but the prognostic importance of persistence of distress symptoms is less certain.
Aim To determine whether intermittent and/or persistent psychological distress is associated with long-term cardiovascular (CV) and total mortality in patients with stable coronary artery disease.
Methods 950 participants in the Long-Term Intervention with Pravastatin in Ischaemic Disease (LIPID) trial completed at least four General Health Questionnaires (GHQ-30) at baseline and after ½, 1, 2 and 4 years. In a landmark analysis from 4 years, Cox proportional hazards models evaluated the risk of CV and total mortality by increasing levels of psychological distress: never distressed, sometimes any severity (GHQ score >5), persistent mild (GHQ score >5 on three or more occasions) and persistent moderate distress (GHQ score >10) on three or more occasions, over a median of 12.1 (IQR 8.6–12.5) years. The models were both unadjusted and adjusted for known baseline risk factors.
Results Persistent moderate or greater psychological stress was reported on three or more assessments by 35 (3.7%) subjects. These patients had a higher risk of both CV death (adjusted HR 3.94, 95% CI 2.05 to 7.56, p<0.001) and all-cause mortality (adjusted HR 2.85, 95% CI 1.74 to 4.66, p<0.001) compared with patients with no distress. In contrast, patients who reported persistent mild distress (n=73, 7.7%) on three or more visits, and those who met criteria for distress on only one or two assessments (n=255, 26.8%), did not have an increased risk of CV or all-cause mortality during follow-up.
Conclusion In patients with stable coronary artery disease, persistent psychological distress of at least moderate severity is associated with a substantial increase in CV and all-cause mortality.
- coronary heart disease
- psychological distress
Statistics from Altmetric.com
Symptoms of depression and anxiety have been associated with an increased risk of major cardiovascular (CV) events in many studies, both prognostic and aetiological.1–3 Informed by this evidence, expert committees have concluded that depression and anxiety are likely to increase the risk of adverse coronary heart disease events.4 5 However, the optimal strategies for screening and targeting those patients who may benefit from interventions to reduce depression and anxiety are uncertain.
In most previous studies from coronary heart disease populations, psychological well-being was assessed only once, often following an acute coronary event.1 4 However, anxiety and depression often improve with time after a major life event, and the severity of symptoms can vary over time for various reasons. Over the longer term the persistence of ‘distress’ symptoms in addition to their severity may be important determinants of the associated mortality risk. A better understanding of how the burden of these symptoms over time relates to mortality may help to identify persons more likely to benefit from intervention.
The aim of this analysis was to describe associations between occasional and more persistent symptoms of distress, or depression and anxiety, assessed over 4 years and CV and all-cause mortality during the next 12 years in patients with stable coronary heart disease who participated in the Long-Term Intervention with Pravastatin in Ischaemic Disease (LIPID) study.6 7
LIPID study participants had a history of acute myocardial infarction or hospitalisation for unstable angina pectoris in the previous 3–36 months, were aged 31–74 years, and had a fasting serum cholesterol level of 4.0–7.0 mmol/L before randomisation. Patients with heart failure were excluded. Participants were randomly assigned to pravastatin (40 mg/day) or placebo, and followed for a median of 6.0 years. A detailed description of the LIPID study and its major findings have been published elsewhere.6 A psychological substudy of LIPID was designed to determine whether lowering serum cholesterol with pravastatin had any influence on psychological well-being. This study was approved by the relevant ethics committee for each centre and all participants provided informed consent.8
One thousand two hundred and twenty-two LIPID study participants from 36 centres in Australia and New Zealand were randomly selected for an invitation to the psychological study by the Australian National Health and Medical Research Council Clinical Trials Centre in Sydney, Australia. Of these 1130 (93%) agreed to participate and completed the baseline questionnaire. These subjects were asked to complete follow-up questionnaires at the 6-month, 1-year, 2-year and 4-year assessments with a response rate of 90%, 90%, 88% and 77%, respectively. There was no difference in any measures of psychological well-being between pravastatin and placebo groups during the first 4 years of follow-up.8
Of the original 1130 patients, 85 died before the 48-month visit and 95 further patients did not complete four or more questionnaires. The remaining 950 patients were included in the current analysis (figure 1).
Assessment of long-term outcomes
After completion of the median of 6.0 years on randomised treatment, all patients were followed for two further years on open label pravastatin (until 1999). From this time, surviving patients were followed up centrally by the National Health and Medical Research Council Clinical Trials Centre until the end of 2007.7 Information on cause-specific mortality and cancer incidence was also obtained through data linkage with national death registries and state-based cancer registries in Australia and New Zealand.7 The outcome measures were CV death and all-cause mortality obtained for all patients after 48 months for further follow-up over a median of 12.1 (IQR 8.6–12.5) years.
CV death was defined as death resulting from an acute myocardial infarction, sudden cardiac death, heart failure, related to CV procedures, CV haemorrhage or death from cerebrovascular accident, pulmonary embolism, complications of peripheral vascular disease or uncontrolled hypertension
Assessment of distress
The 30-question version of the General Health Questionnaire (GHQ)9 was used to assess ‘distress’ related to anxiety, depression or both (see online supplementary appendix). Examples of questions in the GHQ30 include ‘Have you recently felt constantly under strain’; ‘been finding life a struggle all the time’, ‘been getting scared or panicky for no good reason’ with possible responses ‘not at all, no more than usual, rather more than usual, much more than usual’, and ‘been able to feel warmth for those near you’; ‘felt that you are playing a useful part in things’ with possible responses ‘Better than usual, about the same as usual, less well than usual, much less well’. Severity was assessed from Likert scores, which sum the graded responses (0, 1, 2, 3) to each question. To define a ‘case’ using the GHQ, a different scoring system was used. Responses to each question were coded categorically (0, 0, 1, 1) and the total determined. Two levels of distress were pre-specified using cut-levels from previously reported analyses: mild distress, defined as GHQ score >58 9 and moderate to severe distress defined as a GHQ score >10.10 11 Persistent distress was defined as elevated distress on the majority of the five assessments (at least three).
Cox proportional hazards models were fitted for CV death and all-cause mortality using a landmark analysis where time started at the 48 month visit and all deaths prior to that time were excluded. A landmark analysis was chosen to define patients distress across 48 months rather than at one time point and then to explore the association between distress categories and outcomes. Patients had to have completed 4 or 5 of the five possible questionnaires administered at baseline, 6, 12, 24 and 48 months to be included in the analysis. Values for the risk factors measured at randomisation are used in adjusted models. If the patient did not complete a questionnaire at 48 months but did complete the previous four questionnaires, then the landmark starting point was the visit date at 48 months after randomisation.
The Cox models included distress category and the adjusted model in addition included sex and risk factors previously found to be relevant to outcomes in the LIPID study; treatment group, Australian versus New Zealander, age, systolic blood pressure, diastolic blood pressure, hypertension, diabetes, obesity, cholesterol, high-density lipoprotein cholesterol, current smoker, stroke history, history of myocardial infarction (none, one and more than one), history of coronary artery revascularisation, atrial fibrillation, dyspnoea, angina and white blood cell count.12 The model also adjusted for troponin I, B-type natriuretic peptide (BNP), cystatin C and C-reactive protein (CRP), which were independently associated with CV death or myocardial infarction in the LIPID study.10 Covariates were assessed for proportional hazards using the Kolmogorov-type supremum test. Of all the covariates in the two models, only one violated this assumption, which was considered consistent with a type 1 error.
The relationship between the baseline variables and the distress levels were explored by fitting a proportional odds cumulative logistic regression. All analyses were done in SAS V.9.3 or STATA V.12.1. A sensitivity analysis allowing for the competing risk of non-CV death and CV-death was performed using the Fine and Gray model13 in the SAS CIF macro. C-statistics were calculated using the methods described by Harrell et al 14 and Net Reclassification Index (NRI) statistics using the methods described by Pencina et al.15
Of the 950 participants 587 (62%) reported no distress at any visit, 27% mild or greater distress (GHQ >5) on one or two assessments, 8% mild distress on at least three visits or ≥60% of the time (but not GHQ >10 more than 60% of the time), and 35 subjects (3.7%) had moderate distress (GHQ was >10) on three or more of the five assessments. The distribution of the baseline characteristics of these distress groups are displayed in table 1. Subjects with persistent mild or moderate distress were more likely to report angina and dyspnoea, but were not more likely to have a history of myocardial infarction or to have higher levels of B-type natriuretic peptide or troponin I. CRP level was higher in patients with persistent distress. CV risk factors and other biomarkers were not associated with distress.
During follow-up, there were 398 deaths and 199 CV deaths. Moderate or severe distress (GHQ >10) which was present on at least three of five visits was associated with a twofold to fourfold increase in both CV and total mortality during long-term follow-up (figure 2). The strength of these associations were similar after adjusting for other predictors of mortality. The association between distress and CV death was similar when the analysis was repeated allowing for the competing risk of non-CV death and CV death (unadjusted model not including competing risk p=0.05, model including competing risk, p=0.03). Persistent mild distress was not associated with a clear increase in CV or all-cause mortality. Distress of any severity which was present on only one or two of the assessments was not associated with either CV or all-cause mortality (figure 2). Kaplan-Meier plots for all-cause mortality and CV death by level of distress are displayed in figure 3.
In the multivariate model which included all significant clinical risk factors and biomarkers, persistent moderate distress was an independent risk factor for both all-cause mortality and CV death (table 2). For CV death, the C-statistics for the model before and after including the level of psychological distress were 0.70 and 0.71 respectively and the net reclassification index was 2.13. For all-cause death, the C-statistics for the model before and after including the level of psychological distress were 0.69 and 0.70 respectively, and the net reclassification index was 2.99.
In this study, persistent moderate or severe psychological distress assessed over 4 years was associated with a twofold to fourfold increase in risk of CV and all-cause mortality during the next ~12 years. In contrast distress which was mild or not persistent was not clearly associated with increased mortality. The overall burden of distress, indicated by the average GHQ score, increased progressively from low levels in participants who never reported distress, to high levels in those who reported persistent severe distress. These findings suggest that in patients with stable CHD, long-term mortality risk is related to the cumulative burden of psychological distress.
Evidence from previous studies of general populations without known CV disease supports these conclusions. In an individual participant meta-analysis of 10 prospective cohorts with 68 222 subjects from the Health Survey of England there was a dose-related association between psychological distress assessed by the 12-question version of the GHQ and both all-cause and CV mortality during a mean follow-up of 8.2 years.16 Mortality was almost twice as high for the ~7% of participants in the highest category, with more modest increases in risk with lower levels of distress. This stepwise association between the level of distress and mortality is consistent with the importance of the cumulative burden of distress over time. However, a single assessment does not account for variation in the level of distress, and may therefore underestimate risk for some patients and overestimate it for others.
In the INTERHEART study,17 a case–control study of risk factors for myocardial infarction in 52 countries, permanent stress at home, work or related to finances were each associated with a ~2 fold risk of myocardial infarction, while episodes of stress which were not permanent were associated with a lesser increase in risk. Two general population aetiological studies18 19 also support the conclusion that repeated measures of psychological well-being provide a more reliable indication of increased CV risk related to distress. In the Whitehall II study19 of British civil servants, depression and anxiety were evaluated five times over 10 years using the GHQ-30, as in the current study. Cumulative ‘GHQ caseness’ defined as a GHQ score ≥5 on at least two assessments was associated with a 50% increase in major CHD events, but a GHQ score ≥5 on only one occasion was not associated with cardiac events. In the HUNT 2 study,18 2.7% of a Norwegian population met criteria for mixed symptoms of anxiety and depression on two assessments 10 years apart. These participants had a ~60% increase in the risk of subsequent myocardial infarction, but the much larger proportion of individuals who met criteria for anxiety and depression on only one of the two assessments did not have an increased mortality risk. Long term follow-up of the ‘ENRICHD’20 and ‘SADHART’21 clinical trials, which evaluated different interventions to reduce depression after myocardial infarction, have also reported that patients with persistent depression have an increased risk of adverse events.22–24
Most previous prognostic studies which reported associations between depression and/or anxiety and mortality in patients with known coronary heart disease have been based on a single psychological assessment.1 2 4 Many of these studies assessed patients early after an acute coronary event. At this time, the level of anxiety and depression may be more influenced by the underlying cardiac disease, and symptoms may be more likely to improve over time. In a meta-analysis of these studies, associations were on average stronger for studies with shorter compared with longer term follow-up for outcomes.1 In the current analysis, which included multiple additional psychological assessments over 4 years, and a longer duration of follow-up, a smaller group of patients with persistent distress were identified, and these people had a significantly higher mortality risk over the longer term. The overall improvement in prediction of CV and all-cause mortality by including distress in the models, as estimated by C-statistics and the net reclassification index, were consistent with a small impact on prognosis for the majority of patients with CHD.
Patients with greater symptoms of distress were more likely to report cardiac symptoms, but were not more likely to have history of myocardial infarction, and did not have higher plasma levels of BNP or troponin I, the biomarkers most clearly associated with worse coronary heart disease outcomes, including in the LIPID study.25 Other objective measures of disease severity could not be assessed, because left ventricle ejection fraction or other data that quantify severity of coronary artery disease were not collected. It is therefore plausible that patients with persistent distress had greater CV disease severity. CRP at baseline was higher in patients reporting persistent moderate distress, consistent with previous reports of an association between inflammation and treatment resistant depression.10 26 There was no association between psychological distress and conventional CV risk factors, and adjustment for all co-variables had a modest influence on the HRs for mortality. Sustained depression and anxiety have been associated with greater progression of atherosclerotic disease in some studies,27 28 but not others.29 It is possible a greater cumulative burden of psychological distress increases the probability of triggering an acute coronary event, as observed for short-term stresses.30 31
The study has limitations. The study is observational and cannot determine whether the association between persistent psychological distress and mortality was causal. In a previous analysis from the same LIPID study cohort, we reported a modest association between GHQ score >5 assessed at baseline, 5 months to 3 years after an acute myocardial infarction or unstable angina and mortality.32 However, the association was not significant after adjusting for multiple co-variates, consistent with the possibility that ‘reverse causality’, or more severe cardiac disease increasing depression and anxiety, in part explained the observed associations. In the current analysis, the association between persistent distress and mortality was similar before and after adjusting for multiple covariates, but the possibility of residual confounding cannot be excluded.
The GHQ was designed many years ago to assess anxiety, depression and other non-specific psychological symptoms, and is no longer widely used. However, similar results were observed in large general population cohorts which used the GHQ-30 or shorter GHQ-12 to evaluate psychological well-being.16 19 In the Whitehall II study, the multiple assessments of psychological distress by the GHQ-30 between midlife and old age, were associated with the presence of CHD, obesity and smoking,33 as well as the risk of major CHD events.19 Psychological distress detected by the GHQ may include many patients with post-traumatic stress disorder, which has also been associated with increased CV risk.34 However, the current analysis did not evaluate occurrence of traumatic life events. To evaluate the persistence of distress, the study population was restricted to patients alive at 48 months. This may underestimate the impact of persistent distress and decrease the statistical power of the study by excluding patients who died before this time. To our knowledge, no previous studies have undertaken a systematic assessment of patients with stable coronary heart disease which included both multiple psychological assessments over several years and long-term follow-up for mortality.
The study findings are relevant to possible interventions for managing anxiety and depression in patients with stable CHD. Screening and treating anxiety and depression early after myocardial infarction is currently recommended,4 but randomised clinical trials of interventions to reduce depression and anxiety have reported no reduction in major CV21 35 events or had low-statistical power.36 An alternative strategy would be to focus efforts on identifying patients who have persistent distress for multi-modal or stepped care interventions.37 38 Because repeated interactions are likely to be needed these may be more effectively undertaken by the patient’s primary care or usual healthcare providers.39 The observations from the current study are relevant to the design of future research needed to determine whether screening and targeting of interventions to reduce persistent psychological distress in patients with stable coronary heart disease will also reduce major CV events.
What is already known on this subject?
Depression and anxiety have been associated with increased cardiovascular (CV) events in many studies. In most previous studies, psychological well-being was assessed only once, often following an acute coronary event. However, symptoms of distress often improve and their severity can vary over time. Over the longer term, the persistence of distress symptoms may be an important determinant of mortality risk.
What might this study add?
In patients with stable coronary heart disease moderate or greater psychological distress identified by questionnaire on at least three of five assessments over 4 years was associated with a substantial increase in CV and all-cause mortality over the next 12 years.
How might this impact on clinical practice?
Patients with persistent symptoms of distress have the greatest potential to benefit from interventions to improve psychological well-being. The patient’s primary care or usual healthcare providers may be best placed to identify these patients, and consider appropriate interventions. However further research and particularly randomised controlled trials are needed to determine whether interventions which reduce persistent distress also decrease mortality.
Available on request to the corresponding author
Contributors The manuscript was drafted by RS. All coauthors contributed to the design and management of the LIPID trial and psychological substudy. All authors contributed to critical review of the manuscript.
Funding The LIPID trial was originally funded by Bristol-Myers Squibb. The psychological substudy was undertaken independently as an investigator initiated project. The design analysis and drafting of the manuscript were all undertaken independently of the study sponsor.
Competing interests None declared.
Patient consent Obtained.
Ethics approval Ethics Committees for all participating centers.
Provenance and peer review Not commissioned; externally peer reviewed.
Correction notice This paper has been amended since it was published Online First. Owing to a scripting error, some of the publisher names in the references were replaced with ’BMJ Publishing Group'. This only affected the full text version, not the PDF. We have since corrected these errors and the correct publishers have been inserted into the references.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.