Intended for healthcare professionals


Psychological stress and cardiovascular disease

BMJ 2002; 325 doi: (Published 10 August 2002) Cite this as: BMJ 2002;325:337

This article has a correction. Please see:

Rose questionnaire is not what it seems

  1. Matthew Hotopf, reader in psychological medicine (m.hotopf{at}
  1. Division of Psychological Medicine, Guy's, King's, and St Thomas's School of Medicine, London SE5 8AZ
  2. New York Medical College, Yonkers, NY 10703, USA
  3. Department of Public Health, University of Glasgow, Glasgow G12 8RZ

    EDITOR—Macleod et al's paper on stress and cardiovascular disease tells us two things.1 Firstly, the Rose angina questionnaire rather inconveniently does not just measure angina in the sense understood by cardiologists.2 Instead it measures chest pain as understood by everyone else.

    Most cases of chest pain in the general population are not due to heart disease, and even in middle aged Scottish men the prevalence of coronary heart disease is low, so the positive predictivevalue of the Rose questionnaire will be poor.3 The relation between stress and chest pain that the questionnaire measures is only a “bias” in as much as it does not fit into the view of cardiovascular epidemiologists. The effect is real (and has important clinical implications to cardiologists) in that the Rose questionnaire is a superb measure of anxiety in young people but will mislead those who interpret its results too credulously.3 The effect probably accounts for anomalies such as the higher rates of angina in women despite their lower rates of coronary heart disease.4

    The second thing the paper tells us is that a weak measure of stress is a poor predictor of cardiovascular events many years later. Contrary to the statement in “This week in the BMJ,” the findings of the study do not do much to “cast doubt over the associations between psychosocial measures and disease outcomes.”

    Many studies indicate a higher mortality for people with depression and other psychiatric disorders.5 Though there may be important residual confounders (or explanatory pathways) to explain these effects, they still need explaining.


    Paper doesn't clarify things

    1. Paul J Rosch, president, American Institute of Stress (stress124{at}
    1. Division of Psychological Medicine, Guy's, King's, and St Thomas's School of Medicine, London SE5 8AZ
    2. New York Medical College, Yonkers, NY 10703, USA
    3. Department of Public Health, University of Glasgow, Glasgow G12 8RZ

      EDITOR—The notion that responses to four questions provide an accurate measurement of stress levels is as naive as the counterintuitive conclusion reached by Macleod et al.1 A wealth of literature confirms a causative or aggravating contribution of anxiety, depression, sudden emotional shock, and other stressors to coronary morbidity and mortality, sudden death, and congestive failure.2 Two articles in one issue of Psychosomatic Medicine were devoted to discussing novel pathways by which such stress related effects may be mediated.34

      Having worked in stress research for over 50 years, I would question the value and alleged “wide” use of the four question Reeder stress inventory. The fact that this inventory has not been much used may explain why it is no longer in vogue. In one study (not cited), this assessment was not consistent with validated instruments such as two anxiety inventories.5

      In addition, high scores with responses to these questions may reflect increased neuroticism, which would essentially exclude men with type A behaviour, which is as significant a risk factor for coronary heart disease as cholesterol concentration, hypertension, and cigarette smoking.

      There is little doubt that reporting bias can influence putative associations between emotions and health, but this paper hardly negates numerous studies showing a relation between “stress” andcoronary events and is not likely to shed light on this controversial subject.


      Authors' reply

      1. Carole Hart, research fellow
      1. Division of Psychological Medicine, Guy's, King's, and St Thomas's School of Medicine, London SE5 8AZ
      2. New York Medical College, Yonkers, NY 10703, USA
      3. Department of Public Health, University of Glasgow, Glasgow G12 8RZ

        EDITOR—Hotopf seems to agree that artefactual associations between self reported psychosocial exposures and self reported symptoms of angina are almost inevitable. However, he seems to suggest that this point is self evident and barely worth making. We disagree. It is hardly a naive minority who seem to give credence to these associations. Indeed, the BMJ and the Lancet regularly report such associations as evidence of substantive effects. 12 Our point is that bias may influence all effect estimates derived in this way, as becomes apparent when subjective outcomes are compared with objective ones. The accompanying table presents our results alongside those from the Whitehall II study on effects of job control to illustrate this.3

        Associations between stress and job control and subjective and objective outcomes in West of Scotland collaborative study and Whitehall II study

        View this table:

        We agree that the Rose questionnaire is not a precise diagnostic instrument. It is a widely used epidemiological survey tool. And clearly it has some validity: coronary mortality was almost trebled in subjects who had positive results on the questionnaire.

        We agree that the Reeder inventory can be criticised on several grounds but reiterate that our points about problems interpreting observational evidence in psychosocial epidemiology are independent of the status of the instrument we used to illustrate these issues. Furthermore, by the conventions of psychometric validation this instrument seems as robust as many of its more modern and popular counterparts.4

        Relations between depression (or other psychosocial factors) and coronary mortality are certainly unlikely to be the product of reporting bias. However, they may well be the product of confounding (for example, by socioeconomic factors or disease severity), as suggested by results of the only large experimental study so far to address this issue.5

        Rosch states that a wealth of literature confirms a causal relation between psychological exposures and heart disease and cites one of his own editorials (devoid of references) in support of this statement. Despite Rosch's assertions, however, the American Heart Association does not currently accept “type A behaviour,” or any other psychosocial exposure, as an established coronary risk factor. Unlike smoking, hypertension, and cholesterol concentration, type A behaviour basically stopped being a coronary risk factor when diagnosed heart disease stopped being more common among the middle class. This perhaps explains why contemporary studies prefer to focus on “hostility”—the component of type A behaviour most reliably associated with disadvantage and hence now with heart disease.6

        We suggest that more light than heat could be shed on the important questions that psychosocial epidemiology seeks to explore if researchers in the field were as critical of evidence that ostensibly supports their cherished theories as that which challenges them.


        View Abstract