Illness and Personality
Illness Experience and Health Care
’The virtues of science are skepticism and independence of thought.’
In this chapter, we examine claimed associations between individual differences in personality and other psychological characteristics on susceptibility to illness. This is a field of high expectations and false hopes. For this reason, we adopt a sceptical attitude towards some of the worst examples of hyped science, wishful thinking and blatant exaggeration. We discuss the scientific problems involved in investigating and explaining links between personality and physical illness. We assess contemporary research on personality and illness with particular reference to coronary heart disease and cancer.
This chapter critically reviews the literature on personality, health and illness. As we have seen in other chapters, the ’biopsychosocial model’ (BPSM) has been proposed as an alternative to biomedicine. The banner status of the BPSM has opened the floodgates to the idea that psychological processes can directly or indirectly influence well-being and the course of illness. However, there have been many difficulties establishing any solid findings. ’Exciting’ new findings on the ’power of the mind’ to cure cancer, prevent heart disease and to live a long and happy life are frequently reported in the media but rarely are these substantiated in dispassionate reviews. What we can learn from this field is a valuable object-lesson in sceptical science. An evidence-based and critical appreciation of current knowledge serves as a shield against the bad science and flawed studies put out by journals and media who put hype before truth. A recent example of distorted reviewing by a well-known health psychology journal is described in Box 18.1.
Box 18.1 Misleading Review on Slowing Cancer Progression and Increasing Survival Time Using Psychological Intervention
Some researchers have extravagantly claimed that their interventions extended their patients’ survival but, apart from this one highly publicized report, the evidence is non-existent. Spiegel et al. (1989) reported a positive effect of psychosocial treatment on the survival of patients with metastatic breast cancer, claiming an increased survival time of about 18 months. Although the findings have never been replicated, it raised many false hopes among cancer sufferers, a false hope that became reinforced by practitioners offering hyped-up and glamorous-sounding but totally ineffective therapies. Edwards et al. (2004) and Chow et al. (2004) both reported systematic reviews of psychosocial interventions in which effects on survival proved statistically non-significant.
This review by Spiegel (2014) in the British Journal of Health Psychology repeats the misleading claim of Spiegel et al. (1989) that cancer progression can be slowed by psychological intervention. The review provides an instructive example of how the national psychological society representing the science and practice of psychology in the UK, the British Psychological Society, lends credence to a distorted view of scientific evidence. The journal has allowed privileged access to the author by inviting his paper. The publication repeats a misleading claim originally made by the same author 25 years earlier concerning alleged survival benefits of psychotherapy in a manner that could be detrimental to patients and their families. The review is biased in numerous ways: only supportive studies are cited; no reference is made to their statistical and methodological flaws; and finally, important studies that produced negative findings are not cited but are ignored, even when they are of high quality. The evidence cited by Spiegel (2014) is listed in the following table.
All eight studies tabled by Spiegel (2014) as evidence of benefit contain major methodological and/or statistical flaws (Coyne et al., 2007, 2009). One can conclude that the normal editorial and peer-review process lapsed in the acceptance of this paper for publication. The truth is that there is no evidence from properly controlled trials that psychological intervention prolongs survival in cancer. The British Journal of Psychiatry published an editorial two years later in which Spiegel’s (2014) review was cited as evidence that psychotherapeutic interventions are effective in surviving cancer (Bhui, 2016). Both of these national journals have neglected the obvious flaws pointed out by James Coyne and done a disservice to science and to patients by publishing a distorted review of the evidence on the effect of psychotherapy on cancer outcomes.
Source: Coyne et al. (2007, 2009)
Biopsychosocial Model as a ’Blueprint’
Biomedicine is based on the belief that physical illness has physical causes and requires physical treatment. While there is increasing acceptance of the importance of health behaviours, emotions and social issues among the causes of illness, the assumption remains that the actual mechanisms are physical, including such things as viruses, bacteria, carcinogens and physiological abnormalities. Partly as a result of DNA research and the influence of the pharmaceutical industry, this physicalist approach is increasingly being applied to psychological disorders, and conditions ranging from schizophrenia to mild cases of depression are commonly attributed to biochemical imbalances, often thought to be genetic.
Engel (1977: 132) famously proposed a biopsychosocial model (BPSM) that purported to provide ’a blueprint for research, a framework for teaching, and a design for action in the real world of health care’. Engel claimed that BPSM is a scientific model. However, Engel’s so-called BPSM is really a framework (see Chapter 1) within which true models and theories can be created. Engel’s BPSM does not meet the necessary criteria for a scientific model, which would be the capability to explain and predict observed phenomena. The BPSM offers no such capability because it is only a ’framework for teaching’, an idea about how to approach the business of health care. BPSM offers an open invitation to health care practitioners to consider psychological and social experience as complementary to the biological condition of the physical body. Although the BPSM does not enable prediction or explanation, it has provided a rationale, a banner for the role of emotion, thought and behaviour in the study of health and illness.
A popular belief is that people with certain dispositions may be particularly susceptible to heart disease or cancer and that psychological disturbances may be a cause of much physical illness. People talk of ’mental toughness’, being ’strong willed’, ’centred’, ’resilient’ or ’positive’. Alternative treatments and complementary therapies for physical diseases often have psychological components, including stress management programmes, relaxation, breathing exercises and meditation. It is claimed by some proponents of such treatments that they are capable of correcting ’psychic’ or ’energy’ imbalances that have triggered the symptoms in the first place. Others claim to be able to train people to become more ’positive’ and to improve resilience to the impact of serious illnesses such as cancer. These claims rest on a bedrock of folklore, myth and ’old wives’ tales’ that persist to the present day.
In England in the late sixteenth and seventeenth centuries it was widely believed that the happy man would not get plague. In 1871 the physician who treated Alexander Dumas for cancer wrote that among the principal causes of cancer were ’deep and sedentary study and pursuits, and feverish and anxious agitation of public life, the cares of ambition, frequent paroxysms of rage, violent grief’ (Sontag, 2002: 52). At about the same time in England, one doctor advised patients that they could avoid cancer by being careful to bear the ills of life with equanimity; above all things, not to ’give way’ to any grief. At this time also, tuberculosis (TB) was often thought to come from too much passion, afflicting the reckless and sensual, or else to be a disease brought on by unrequited love. In fact, TB was often called consumption, hence the appearance in the English language of metaphors such as consuming passion.
Before jumping to judgement on these fanciful claims, we must take stock of some of the far-fetched claims put out today under the umbrella of ’Positive Psychology’. Many beliefs that psychologists study are entrenched in popular narratives about mind, body and spirit. These cultural heirlooms do not simply lie down and die when a few negative findings are published. False beliefs and fanciful tales are highly resilient. The inkblot that is the biopsychosocial model facilitates their survival.
Explaining Links between Personality and Physical Illness
It is generally agreed that the psychosomatic approach in psychoanalysis and the proponents of psychosomatic medicine failed to produce convincing evidence of any causal connections between psychological characteristics and physical illness, or to demonstrate that their therapeutic interventions were effective (Holroyd and Coyne, 1987). The period since the 1960s has seen the growth of a large empirical literature on the statistical relationship between personality, as assessed by a wide variety of standardized tests, and physical illness. These studies often derive their hypotheses from the earlier speculations of the psychosomatic approach but seek to rectify its defects by carefully analysing statistical evidence. There is, however, one major defect that cannot be overcome because it is intrinsic to this type of research. The evidence is obtained from correlational investigations rather than true experiments and, as a consequence, findings are open to a wide range of interpretations. Obviously, no ethical investigator can assign people with different personalities at random to experimental participants and then study their subsequent proneness to illness. All that the investigator can normally do is to administer personality tests to the participants and obtain measures of their illness status. If the personality and illness scores are obtained at the same time, then this is a cross-sectional or correlational study. In a prospective study, personality scores are taken at time 1 (T1) and health outcomes are measured at time 2 (T2). The prospective study is a statistically more powerful methodology because it allows for the possibility of analysing causal relationships between antecedent (T1) measures and outcome measures at T2. Owing to the more complex and resource-intensive nature of prospective studies, cross-sectional studies are far more common.
There are several possible mechanisms that could mediate associations between personality and illness, including accumulated stress, disruption of the immune and endocrine systems, and increased chronic inflammation. Personality may influence cancer risk indirectly via poor health behaviours, such as smoking and not participating in screenings.
It is not possible to infer causation from such correlation and, in the case of personality—illness correlations, it is possible to illustrate this by considering a range of important problems of interpretation. The issues are perennial problems in the literature and are essentially logical in nature. We summarize these in Table 18.1 along with safeguards that readers of research papers need to apply when interpreting study findings.
Direction of Causality
Psychological characteristics may be a cause of physical illness, but they can also be a consequence of it. This is a particular problem for cross-sectional studies that simultaneously assess personality traits and illness. For example, if patients with a history of coronary illness have higher scores on anxiety and depression than healthy controls, should we conclude that anxiety and depression are risk factors for coronary illness or that a history of coronary problems can cause people to become more anxious and depressed? Obviously, the existence of any causal relationship cannot be inferred from this type of study. The flaw may seem trivially obvious, but it is surprising how many cross-sectional studies are to be found in the research literature on associations between personality and illness.
The problem cannot always be resolved by conducting prospective studies to investigate the extent to which the current personalities of healthy participants are predictive of future illness. The reason for this is that many major illnesses take a long time to develop and it is frequently the case that patients have experienced unusual and disturbing symptoms for some time before a diagnosis is given. It is therefore possible that psychological characteristics that appear to be a cause of subsequent illness are in fact a consequence of symptoms of developing illness occurring prior to diagnosis. Prospective studies are clearly far superior to cross-sectional studies, but they do not necessarily eliminate the problem of how to determine the direction of causality.
A correlation may be found between two variables A and B when there is neither a direct effect of A on B, nor of B on A, but because a third background variable C has an effect on both. Galen’s explanation for the association he believed to exist between melancholy and breast cancer is a good illustration. Here an excess of black bile is the background variable that he hypothesized to be a cause of both melancholy and breast cancer. As a further illustration, suppose that some people have a history of childhood illness that leaves them constitutionally weak and prone to further illness. Suppose also that this history of childhood illness has had a deleterious effect on their personality development, perhaps by limiting their opportunities for social development. The inclusion of a number of such individuals in a sample along with constitutionally stronger individuals who have had a healthy childhood could produce a non-causal correlation in the sample between adult personality and adult proneness to illness, because both are influenced by childhood illness. Further examples of background variables, notably genetic predispositions, are discussed by Holroyd and Coyne (1987) and Suls and Rittenhouse (1990).
Self-Reporting of Illness
Stone and Costa (1990) and Cohen and Williamson (1991) pointed out that much health psychology research relies on self-reported illness rather than biologically verified disease. This does not only apply to minor ailments such as colds and flu. The diagnosis of angina pectoris, for example, is frequently based solely on patients’ reports of chest pain. Stone and Costa argue that this reliance on self-reported illness is particularly unsatisfactory when considering research into links between personality and illness. They point out that there is extensive evidence that psychological distress is associated with somatic complaints but not with organic disease. Since many personality test scores may be interpreted, at least to some extent, as measures of distress, it follows that correlations between test scores and self-reported illness may provide a false indication of a link between personality and disease when all that they really show is that neurotic individuals are the ones most likely to complain of being ill. Friedman and Kern (2014) consider these issues in some detail, and recommend that personality—illness research should focus primarily on mortality and longevity as outcome measures. However, it should not be assumed that any discrepancy between self-reported illness and biologically verified illness is necessarily an indication of neuroticism. Adler and Matthews (1994) noted that there is evidence that perceived health predicts mortality independently of biological risk factors, leading them to conclude that self-reported health provides useful information over and above direct biological indications. The association between perceived health and mortality has now been confirmed by many studies, and Jylhä (2009) provides a detailed analysis of possible reasons for it.
Dimensions of Personality
Personality testing is an inexact science with little agreement as to what the basic dimensions of personality really are, or even whether the question is worth asking. Three influential theories have been those of Eysenck, who argued initially that there were only two dimensions, extraversion and neuroticism, and subsequently added a third, psychoticism; Cattell, who believed he had identified 18; and McCrae and Costa (2003) who settled for five. These so-called ’Big Five’ personality traits have generated a considerable amount of recent research. They consist of extraversion/introversion, agreeableness/antagonism, conscientiousness, neuroticism/emotional stability, and openness to experience.
A problem that arises in research using personality tests is that similar items can often be found in tests that are supposed to be measuring different traits so that, not surprisingly, scores for the same group of individuals given both tests may be highly intercorrelated. Consider some of the measures that are frequently used in research into the links between personality and illness. The individual who scores high on anxiety is also likely to score high on depression, neuroticism and pessimistic explanatory style, and correspondingly low on self-esteem, self-efficacy, hardiness and sense of coherence. The common element that may run through all of these measures is probably best labelled, following Stone and Costa (1990), as distress proneness or negative affectivity. Suls and Bunde (2005) discuss this issue in a detailed review of research linking anger/hostility, anxiety and depression with cardiovascular disease. They note that strong correlations exist between all three traits that make it difficult to establish which of the associations with cardiovascular disease is of primary importance, or whether a more general trait of negative affectivity is the key variable.
Physiological Mechanisms versus Health Behaviour
Psychological characteristics may be linked to illness, either by way of physiological variables with which they are associated or, more indirectly, by way of their relationship to health behaviour. Health psychologists who conduct research into the relationship between personality and illness are also primarily interested in physiological pathways. However, it is generally acknowledged that more prosaic explanations for correlations between personality and illness may be derived from the fact that personality differences are often associated with differences in health behaviour (Suls and Rittenhouse, 1990; Miller et al., 1996; Stone and McCrae, 2007). Characteristics such as anxiety, depression, neuroticism and hostility have been variously shown to be associated with levels of smoking and alcohol consumption, diet and exercise, sleep disturbance, likelihood of seeking medical advice in the early stages of a disease and the likelihood of adhering to recommendations subsequently. Any of these variables, or some combination of them, could be invoked to account for an empirical correlation between personality and illness.
Data Mining, p-Hacking and Data Selection
Data mining (also known as data dredging, fishing or snooping, and p-hacking) is the seeking of patterns in data that can be presented as statistically significant without first devising a specific hypothesis as to the underlying causality. It is a process of falsification that can be completely hidden from external observers and that is guaranteed to produce publishable results. Results obtained using these methods will almost always be unrepeatable.
Minimally Significant Effect
Any study with a finding based on a single p level of .05 must be interpreted cautiously. There are many different reasons why a single p < .05 finding may find its way into print. The only way to support a single p < .05 finding is replication. Without that, it is prudent to ignore the finding.
Fraud involves the fabrication of data to present a positive confirmation of a hypothesis or desired outcome. Fortunately, known cases of fraud are relatively rare, but they do exist.
The Importance of Intervention Studies
The following brief survey of statistical studies of the associations between personality and physical illness indicates a number of promising findings. However, for the reasons that have been presented, all of these findings are subject to a range of interpretations so that true causal links have not yet been definitively established.
Friedman and Kern (2014) concluded that, in order to make a convincing case for causality, it is necessary not only to demonstrate the existence of an association, but also to show that interventions designed to reduce a ’toxic’ aspect of personality, or enhance a beneficial one, have the effect of reducing disease risk and increasing longevity. Interventions may obviously be of value in reducing distress in individuals suffering from serious illness, but the question that also needs to be addressed is: Can they produce an improved prognosis for the disease in question? Sadly, this has not yet been convincingly shown in any study. We now review the evidence in order to indicate the areas that look most promising for future research using interventions.
The Type A Personality, Hostility and Coronary Heart Disease
Type A and B Personalities
Speculation about an association between the Type A and Type B personalities and coronary heart disease (CHD) has a history that dates back more than 50 years (Riska, 2000). The distinction between the two personalities was introduced in the mid-1950s by cardiologists Meyer Friedman and Ray Rosenman, although, as already noted, their ideas can be traced back further to the work of Alexander in psychosomatic medicine. The Type A personality, thought to be at greater risk of CHD, is described as highly competitive and achievement oriented, not prepared to suffer fools gladly, always in a hurry and unable to bear delays and queues, hostile and aggressive, inclined to read, eat and drive very fast, and constantly thinking what to do next, even when supposedly listening to someone else. In contrast to this, the Type B personality is relaxed, laid back, lethargic, even-tempered, amiable and philosophical about life, relatively slow in speech and action, and generally has enough time for everyone and everything. The Type A personality has much in common with Galen’s choleric temperament, the Type B with the phlegmatic (see Table 1.1). It is well known that men are at greater risk of CHD than women, and Riska (2000) made an interesting argument for the view that the concept of the Type A personality was an attempt to ’medicalize’ and ’pathologize’ traditional concepts of masculinity.
The key pioneering study of Type A personality and CHD was the Western Collaborative Group Study (WCGS), in which over 3,000 Californian men, aged from 39 to 59 at entry, were followed up initially over a period of 8.5 years, later extending over 22 years. When results were reported at the 8.5-year follow-up, it appeared that Type As were twice as likely compared with Type Bs to suffer from subsequent CHD. Of the sample, 7% developed some signs of CHD and two-thirds of these were Type As. This increased risk was apparent even when other risk factors assessed at entry, such as blood pressure and cigarette smoking, were statistically controlled for. Similar results were subsequently published from another large-scale study conducted in Framingham, Massachusetts, this time with both men and women in the sample, and by the early 1980s it was confidently asserted that Type A characteristics were as much a risk factor for heart disease as high blood pressure, high cholesterol levels and smoking.
However, later research failed to support these early findings. When Ragland and Brand (1988) conducted a 22-year follow-up of the WCGS, using CHD mortality as the crucially important measure, they failed to find any consistent evidence of an association. Much further research continued to be published up to the late 1980s, yielding few positive findings. Reviewing this evidence, Myrtek (2001) suggests that the modest number of positive findings that did exist were the result of over-reliance on angina as the measure of CHD. As we have already pointed out, this is an unreliable measure because it is frequently based solely on self-reported chest pain. Considering studies that adopted hard criteria, including mortality, Myrtek concludes that we can be confident that the Type A personality is not a risk factor for CHD.
It can take a long time for a popular belief to fade away when there is a lack of evidence to support it. The extensive coverage still given to the Type A—CHD hypothesis by textbook writers is a good illustration of this. Researchers may be a little quicker to react, as is indicated by the decline in publications in this field from the early 1990s. In fact, it was largely replaced by an alternative hypothesis that was itself generated by the analysis of Type A—CHD research. This hypothesis is that hostility is the key dimension of personality that is associated with CHD.
The Type A personality, as described briefly in the last section, contains a number of components that are not necessarily closely correlated. Measures of Type A and B personalities often included sub-components that could be separately analysed for their association with subsequent CHD. When this was done it emerged that there was only one component that did seem to have some predictive power, and this component was anger or hostility. Research into links between anger/hostility and CHD became as popular in the 1990s as Type A research had been in previous years, but unfortunately with a very similar conclusion. By the end of the decade a number of reviews, including that of Myrtek (2001), found the studies to be of very mixed quality, with inconsistent results. There was some evidence of a statistically significant but very weak relationship for prospective studies of initially healthy individuals, but not for studies that have followed up patients already diagnosed with CHD.
In a curious mirroring of the breakdown of the Type A personality into sub-components, which led to the hostility—CHD research, the hostility researchers themselves reacted to disappointing findings by breaking hostility down into separate components. These included cynicism, mistrust, verbal and physical aggressiveness, and overt and experienced aggressiveness. It was proposed that more attention should be given to these sub-components in order to discover which are the most hazardous for health. However, when reviewing this area of research, Suls and Bunde (2005) noted that there is considerable overlap between measures of these sub-components of anger/hostility, with similar items being included in ostensibly different measures. Suls and Bunde also confirm Myrtek’s earlier conclusion that evidence of an association between hostility, however measured, and subsequent CHD suggests a weak relationship, possibly no more than a side effect of the correlation of hostility measures with anxiety and depression, characteristics that will be considered later in this chapter because they appear to have a more substantial association with CHD.
Support of the Tobacco Industry for Personality—Illness Research
In view of the disappointing results achieved by research into the Type A personality, hostility and CHD, it may well be asked why it obtained so much publicity over more than 40 years. The reason may be connected with the high level of support this research has received from the US tobacco industry. Petticrew et al. (2012) have established this by analysing material lodged at the Legacy Tobacco Documents Library, a vast collection of documents that the companies were obliged to make public following litigation in 1998. These documents show that, for over 40 years from the 1950s, the industry heavily funded research into links between personality and both CHD and cancer, hoping to demonstrate that these personality variables were associated with cigarette smoking, thereby undermining claims about causal links between smoking and disease. Thus, for example, if it could be shown that Type A personalities were both more likely to smoke than Type Bs, and more likely to develop CHD, then it could be argued that smoking might be just an innocent background variable. Further to this, the Philip Morris company funded Meyer Friedman, the originator of Type A research, for the Meyer Friedman Institute, conducting research aiming to show that Type A personalities could be changed by interventions, thereby presumably reducing proneness to CHD even if they continued to smoke. Petticrew et al. also show that, while most Type A—CHD studies were not funded by the tobacco industry, most of the ones that found positive results were tobacco-funded. As has been pointed out in many areas of science, positive findings invariably get a great deal more publicity than negative findings and rebuttals.
Another researcher who found an association with the tobacco industry to be a useful source of funding was the late H.J. Eysenck. Eysenck was a student of the notorious Cyril Burt, who is known to have fabricated researchers and data to establish his false eugenic theory of intelligence. Pringle (1996) reported that Eysenck received nearly £800,000 to support his research on personality and cancer, with results that were a spectacular exception to the general run of negative findings in this field, findings that are discussed later in this chapter. Eysenck (1988) claimed that personality variables are much more strongly related to death from cancer than even cigarette smoking. This is a staggering claim. In two subsequent papers, with R. Grossarth-Maticek, the two researchers appeared to have identified personality types that increase the risk of cancer by about 120 times and heart disease by about 25 times (Eysenck and Grossarth-Maticek, 1991; Grossarth-Maticek and Eysenck, 1991). They also claimed to have tested a new method of psychological treatment that could reduce the death rate for disease-prone personalities over the next 13 years from 80% to 32%.
These extraordinary claims were not received favourably by others working in this field. In a comprehensive and highly respected review of the subject, Fox (1988) dismissed earlier reports by Eysenck and Grossarth-Maticek as ’simply unbelievable’, and the 1991 papers, which were the first moderately detailed accounts of their research, were subjected to devastating critiques by Pelosi and Appleby (1992, 1993) and by Amelang and Schmidt-Rathjens (1996). The ’cancer prone personality’ was not clearly described and seems to have been an odd amalgam of emotional distance and excessive dependence. After pointing out a large number of errors, omissions, obscurities and implausible aspects of the data, in a manner reminiscent of Leon Kamin’s now-legendary analysis of Cyril Burt’s twin IQ data (Kamin, 1977), Pelosi and Appleby comment:
It is unfortunate that Eysenck and Grossarth-Maticek omit the most basic information that might explain why their findings are so different from all the others in this field. The methods are either not given or are described so generally that they remain obscure on even the most important points. … Also essential details are missing from the results, and the analyses used are often inappropriate. (Pelosi and Appleby, 1992: 1297).
They never used the word ’fraud’. They didn’t need to.
Depression, Stress and Coronary Heart Disease
There have been a number of reviews that have concluded, on the basis of prospective studies, that there are substantial associations between both anxiety and depression and subsequent CHD (Hemingway and Marmot, 1999; Krantz and McCeney, 2002; Wulsin and Singal, 2003; Lett et al., 2004; Suls and Bunde, 2005). These associations have been found in studies of patients with clinically diagnosed distress and in general population studies. Anxiety seems to predict sudden cardiac death rather more than other types of CHD, and phobic, panic-like anxiety is a particularly strong predictor: Haines et al. (1987) found that sufferers were three times more at risk of sudden cardiac death over the next seven years compared with non-sufferers. Very similar results were subsequently found by Kawachi et al. (1994) in a two-year follow-up of 33,999 initially healthy US male health professionals.
Depression is predictive of a wider range of CHD than anxiety. In a UK study with 19,649 participants who were initially free of clinical manifestations of heart disease, Surtees et al. (2008a) found, with an average follow-up period of 8.5 years, that those assessed as suffering from a major depressive disorder were 2.7 times more likely to die from ischaemic heart disease over the follow-up period than those who did not, independently of age, sex, smoking, systolic blood pressure, cholesterol, physical activity, body mass index, diabetes, social class, heavy alcohol use and antidepressant medication use. In a very large prospective study of 96,376 post-menopausal women, Wassertheil-Smoller et al. (2004) report that depressive symptoms were substantially associated with death from cardiovascular disease after adjusting for age, race, education, income, diabetes, hypertension, smoking, cholesterol level, body mass index and physical activity.
These findings for anxiety and depression are impressive, but they should be considered alongside our earlier discussion of problems in interpreting personality—illness correlations. Hemingway and Marmot (1999) point out that these problems are particularly acute in this area. Anxiety and depression are certainly consequences of CHD as well as possible causes of it. Furthermore, symptoms of incipient CHD, such as breathlessness and chest pains, may occur for years prior to diagnosis, and lead in turn to experienced anxiety and depression. In this way, prospective studies could give the impression that anxiety and depression are causes of CHD, when in fact the direction of causality is the other way around.
Stress is often cited as a causal factor in coronary heart disease. Possible neurobiological pathways are illustrated in Figure 18.1. The neurobiological model of stress, unlike the BPSM, is a genuine scientific model. It enables predictions to be tested, shows relationships between processes that can be observed, and combines biological and psychosocial processes in a single system.
According to the neurobiological model: the stress response involves the central activation of brain systems responsible for the analysis of the environment. This response is important and appears not to be deleterious, as it promotes a physiological balance in response to classic and normal environmental stressors. However, in the case of chronic and mainly psychosocial stressors, the allostatic system may be overwhelmed with hyperactivation of the hypothalamic—pituitary—adrenal axis and the autonomic nervous system with dysregulation of blood pressure and cortisol levels. In addition, an immuno-inflammatory response occurs with the production of inflammatory cytokines. If this phenomenon lasts for a long time because of chronic adversity (work or social stress, for instance), the pathophysiological effects can lead to metabolic disturbances (glucose and lipid dysregulation), metabolic syndrome and cardiovascular disease. Then, psychological factors like perceived stress, coping style, personality traits, or social support might modulate the stress response. (Chauvet-Gelinier and Bonin, 2017)
As Miller et al. (2009) pointed out, one practically useful way of resolving some of the methodological problems in interpreting findings in this area would be a robust demonstration that interventions designed to reduce stress in patients suffering from CHD could produce an improved prognosis. Unfortunately, the two most substantial trials that have been conducted so far, each targeting depression and with more than 2,000 participants, showed no effect of interventions. Berkman et al. (2003) evaluated a programme that included cognitive behavioural therapy supplemented with antidepressants for the more severely depressed patients. Van Melle et al. (2007) evaluated an intervention consisting simply of treatment with antidepressants. Neither study found any evidence of an improvement in event-free survival for the intervention groups compared with controls.
Figure 18.1 Neurobiological action of stress
Source: Reproduced from Chauvet-Gelinier and Bonin (2017)
Sin et al. (2016) prospectively examined relationships of depressive symptoms with behavioral and lifestyle factors among 667 patients with stable coronary heart disease. The lifestyle behaviours consisted of physical activity, medication adherence, body mass index, waist to hip ratio, sleep quality and smoking status. These were assessed at baseline and five years later. Sin et al. found that greater depressive symptoms at baseline predicted poorer lifestyle behaviours five years later (less physical activity, lower medication adherence, higher body mass index, higher waist to hip ratio, worse sleep quality and smoking). Baseline lifestyle behaviours predicted five-year change in depressive symptoms.
Some studies have indicated that the relationship between emotional distress and illness is not restricted to heart disease. Mykletun et al. (2007) examined data from a population-based health study of 61,349 participants. With a mean follow-up of 4.4 years, they found that depression was equally associated with all disease-related causes of death, not just CHD. They did not find any evidence of an association with anxiety. On the other hand, Grossardt et al. (2009), in a follow-up of 7,080 participants originally tested between 1962 and 1965, found that pessimistic, anxious and depressive personality traits were each predictive of all-cause mortality. Further research is obviously needed to clarify reasons for the differences between the findings of the two studies. However, they both suggested that the association between psychological distress and disease-related death extends beyond CHD.
Depression, Hope and Cancer
The widely held belief that depression is also an important factor in the onset and subsequent development of cancer has received little support from research. Adler and Matthews (1994) reviewed three large-scale prospective studies of the relationship between depression and both the incidence of and mortality from cancer. In these studies, initially healthy samples of up to 9,000 were followed up over periods ranging from 10 to 20 years and no associations were found between depression and either cancer onset or mortality. Since then two large-scale studies have produced conflicting results. Penninx et al. (1998) carried out a prospective study of 1,708 men and 3,117 women aged 71 and over. They found a significantly increased incidence of cancer for those who were diagnosed as suffering from chronic depression, indicated by repeated assessments of symptoms over six years. On the other hand, Whooley and Browner (1998) undertook a prospective study over six years of 7,519 women aged 67 or over and analysed the relationship between depression and subsequent mortality from (a) cancer, (b) cardiovascular disease and (c) all other diseases. They found no relationship between depression and cancer, but a strong relationship with both cardiovascular disease and all other diseases.
Negative findings have also been reported from follow-up studies of patients who have been treated for cancer. For example, Barraclough et al. (1992) followed up 204 patients who had received surgery for breast cancer over 42 months after surgery. They used a very detailed interview schedule, which included the assessment of prolonged major depression before surgery and during the follow-up period. They found no relationship at all between depression and relapse. Relapse was also unrelated to stress, including bereavement, long-term social difficulties and lack of a confiding relationship.
Coyne et al. (2007) noted the persistence of the belief that psychotherapy promotes survival in people with cancer in the face of contradictory findings. They provide a systematic critical review of the relevant literature and conclude:
No randomized clinical trial designed with survival as a primary endpoint and in which psychotherapy was not confounded with medical care has yielded a positive effect. Among the implications of the review is that an adequately powered study examining effects of psychotherapy on survival after a diagnosis of cancer would require resources that are not justified by the strength of the available evidence. (Coyne et al., 2007: 367)
Psychological resources such as hope, mindfulness and spirituality, have been identified as potential resources for patients to cope with the course of cancer and its treatment by alleviating emotional distress and improving quality of life (Aspinwall and MacNamara, 2005; Coughlin, 2008; Hou et al., 2010; Lim et al., 2016). These resources are said to be amenable to individual control and have been suggested as targets for psychosocial interventions (Pitceathly et al., 2013). Mahendran et al. (2017) followed a sample of newly diagnosed Asian cancer patients over a one-year period to identify early opportunities to improve patient outcomes. They aimed to investigate the relative predictive value of hope, mindfulness, spirituality and life satisfaction on cancer mortality. They hypothesized that metastatic and advanced stage cancer predict higher cancer mortality, and greater availability of psychological resources predict lower cancer mortality. Mindfulness, spirituality and life satisfaction were all non-significant as predictors of cancer mortality. They found that metastatic cancer, advanced stage cancer and hope predicted cancer mortality after adjusting for socio-demographic and medical variables. However, this result must be interpreted with caution as it was significant at only the .05 level, and could have been a Type I error. Also, cancer mortality was evaluated over a single year of follow-up with newly diagnosed patients, and the association between psychological resources and mortality over five or ten-year follow-up remains unknown. Finally, Mahendran et al.’s (2017) study did not include potential moderating variables such as health behaviours or social support.
Schofield et al. (2016) investigated associations between resources of hope, optimism, anxiety, depression, health utility and survival in patients starting first-line chemotherapy for metastatic colorectal cancer. This cross-sectional study included 429 patients with metastatic colorectal cancer in a randomized controlled trial of chemotherapy with a median follow-up of 31 months. They completed questionnaires assessing hopefulness, optimism, anxiety and depression, and health utility. Univariable analyses showed that overall survival was associated negatively with depression and positively with health utility and hopefulness. Schofield et al. (2016) concluded that depression and health utility, but not optimism, hope or anxiety, were associated with survival after controlling for known prognostic factors in patients with advanced colorectal cancer. This was not a randomized clinical trial designed with survival as a primary endpoint.
It would be a sensible strategy to switch the effort and research funds expended on cross-sectional studies of cancer survival and psychological resources to high-powered prospective studies using randomized controlled trials.
Conscientiousness and Longevity
Conscientiousness is one of the five factors of personality proposed by McCrae and Costa in their influential theory that was originally developed in 1985. A considerable amount of research has shown that this characteristic is associated with longevity. Many studies, including some long-term prospective ones, have find positive evidence for an association between conscientiousness and longevity.
In a meta-analysis that pooled the results of 20 independent samples, Kern and Friedman (2008) found a modest but significant correlation of 0.11 between conscientiousness and longevity. In a subsequent review, Friedman and Kern (2014) drew attention to more research that confirms the existence of this relationship, which also extends from childhood through adult life. They suggest a number of pathways mediating the association. An important one is health behaviour. For example, Bogg and Roberts (2004) pooled the results from 194 studies that incorporated measures of conscientiousness-related traits and assessments of any of the leading behavioural contributors to mortality (tobacco use, diet and activity patterns, excessive alcohol use, violence, risky sexual behaviour, risky driving, suicide and drug use). They found that conscientiousness-related traits were negatively related to all risky health-related behaviours and positively related to all beneficial health-related behaviours.
Other factors may also be at work in mediating the relationship between conscientiousness and health. While it seems clear that health behaviour is a major one, social environmental correlates of conscientiousness, including social status, may also be important. In a 17-year follow-up of a UK sample of 6,800 participants, aged from 34 to 55 at recruitment, Hagger-Johnson et al. (2012) found that health behaviours and socio-econonmic status both played a significant part, but that the relationship was not fully accounted for by these factors. Other possibilities are considered by Friedman and Kern (2014) and by Bogg and Roberts (2013). For example, conscientiousness may ameliorate the effects of stress on health.
A major 40-year prospective study in Hawaii investigated a lifespan health behaviour mechanism relating childhood conscientiousness to adult clinical health (Hampson et al., 2015). Children in entire elementary school classrooms on two Hawaiian islands were assessed between 1959 and 1967 on their personality traits by their teachers towards the end of one school year. Then in 1998, efforts were made to find these same people, who by then were middle-aged adults. There were 2,418 in the original child cohort, 79 of whom were already deceased and 19 only had first names recorded, leaving 2,320 to locate. Of these, an amazing 1,938 (84%) were found. From those successfully identified, 36 refused further contact and one was illiterate, reducing the sample to 1,901. Of these, 1,387 (73%) were recruited and completed at least one questionnaire. To be included in the study, participants had to have participated in the medical and psychological examination at 51 years, and to have completed the first questionnaire. These requirements limited the sample to 372 men and 387 women (n = 759).
The investigators studied the associations between the Big Five personality traits at mean age 10, adult Big Five personality traits, and adult clinically assessed ’dysregulation’ at mean age 51. Dysregulation was defined as a summary of dysregulated blood glucose, blood pressure and lipids. A retrospective, cumulative measure of lifespan health-damaging behaviour (lifetime smoking, physical inactivity, and body mass index from age 20) were assessed in the Hawaii Personality and Health Cohort (n = 759). Structural equation modelling was used to test the conceptual model with direct and indirect paths from childhood Conscientiousness to adult Conscientiousness, lifespan health-damaging behaviours, educational attainment, adult cognitive ability and adult clinical health.
For both men and women, childhood Conscientiousness influenced health-damaging behaviours through educational attainment, and lifespan health-damaging behaviours predicted dysregulation. Although childhood Conscientiousness predicted adult Conscientiousness, the latter did not predict any other variables in the model. The pathways of influence differed between the two genders. For men, childhood Conscientiousness predicted dysregulation through educational attainment and health-damaging behaviours. For women, childhood Conscientiousness predicted dysregulation through educational attainment and adult cognitive ability.
According to Hampson et al. (2015), childhood Conscientiousness appears to influence health assessed more than 40 years later through complex processes including educational attainment, cognitive ability and the cumulative effects of health behaviours, but not adult Conscientiousness. However, it must be noted that the effects are relatively small. The models account for only 7.4% of the variance in dysregulation for women and 16.5% of the variance in dysregulation for men.
A question that naturally arises at this point is: Can interventions be developed to increase conscientiousness, thereby potentially leading to improved health? Here there are grounds for scepticism. Given that health behaviours play an important role in mediating the association, it is likely to be more productive to focus on health promotion campaigns and interventions aimed directly at specific behaviours, such as dysregualtion, smoking, drinking, diet and exercise, rather than programmes aimed at conscientiousness as a general characteristic.
Positive Psychology: A Cautionary Tale
In contrast to the focus on the health consequences of negative characteristics, such as anxiety and depression, a considerable amount of recent research within ’Positive Psychology’ has been concerned with the potential health benefits of positive characteristics, such as optimism, life satisfaction and self-esteem. There has been a massive amount of hype about positive psychology, but is there any solid evidence to support these conjectures?
One example is the work of Surtees et al. (2003) on sense of coherence and health. Originally proposed as a measurable characteristic by Antonovsky (1979), sense of coherence is described as the ability to perceive one’s world as meaningful and manageable. Surtees et al. carried out a prospective study over six years of the relationship between sense of coherence and mortality from all causes for a very large UK sample of 20,579 participants aged 41—80 years. They found that a strong sense of coherence was associated with a 30% reduction in mortality from all causes and also, more specifically, for cardiovascular disease. The explanation for this probably lies in the fact that sense of coherence is inversely correlated with anxiety and depression. In a study conducted in Finland of 4,642 men and women aged from 25 to 74, Konttinen et al. (2008) found inverse correlations of 0.62 between sense of coherence and depression for both men and women; the inverse correlations between sense of coherence and anxiety were 0.57 for men and 0.54 for women. It follows that the findings of Surtees et al. could be basically the corollary of the positive correlations already discussed between anxiety and depression and both cardiovascular disease and all-cause mortality.
Boehm and Kubzansky (2012) provide a broad critical review of research on the association between psychological well-being and cardiovascular health. They note the existence of many findings that confirm this association, while pointing out the potential significance of the correlations between psychological well-being and the health behaviours that are already known to be associated with cardiovascular health. It has not been demonstrated that changes to an individual’s psychological well-being, whether the result of interventions or for other reasons, have a direct effect on cardiovascular health, independently of known risk factors such as smoking, diet and exercise.
In a combative paper in The Lancet, Sloan (2011) notes that beliefs in the health benefits of positive thinking have been widespread in the USA since the middle of the nineteenth century and are especially prevalent today. They may seem harmless enough even if there is a lack of convincing evidence in their favour, but, as Sloan points out, they are associated with the view that individuals are largely responsible for their own health, which is convenient for governments wishing to minimize spending on public health services. They may also be doubly pernicious if they are false.
Coyne and Tennen (2010) examined four widely accepted claims in the positive psychology literature regarding beneficial adaptational outcomes among individuals living with cancer:
1. The alleged role of positive factors such as a ’fighting spirit’ in extending the life of persons with cancer.
2. The alleged effects of interventions cultivating positive psychological states on immune functioning and cancer progression and mortality.
3. The alleged value benefit finding.
4. The alleged post-traumatic growth following serious illness such as cancer and other highly threatening experiences.
Coyne and Tennen’s analysis suggested that these four types of claim are routinely made in the positive psychology literature but none is supported by the available evidence. In particular, they found that claims about the adaptational value of benefit-finding and post-traumatic growth among cancer patients and the implausibility of claims that interventions that enhance benefit-finding improve the prognosis of cancer patients by strengthening the immune system lacked coherence. Coyne and Tennen (2010: 16) concluded by urging positive psychologists to ’rededicate themselves to a positive psychology based on scientific evidence rather than wishful thinking’.
More Null Results
The role of personality in cancer risk has been controversial, and, as we have seen, the evidence remains inconclusive. Jokela et al. (2014) pooled data from six prospective cohort studies (British Household Panel Survey; Health and Retirement Study; Household, Income and Labour Dynamics in Australia; Midlife in the United Survey; Wisconsin Longitudinal Study Graduate; and Sibling samples) in a meta-analysis to examine whether personality traits of the Five Factor Model (extraversion, neuroticism, agreeableness, conscientiousness and openness to experience) were associated with the incidence of cancer and cancer mortality in 42,843 cancer-free men and women at baseline (mean age 52.2 years, 55.6% women). For an average follow-up of 5.4 years, there were 2,156 incident cancer cases. In a meta-analysis adjusted for age, sex and race/ethnicity, none of the Big Five personality traits was associated with the incidence of all cancers or any of the six site-specific cancers included in the analysis (lung, colon, breast, prostate, skin and leukaemia/lymphoma). In the three cohorts with cause-specific mortality data (421 cancer deaths among 21,835 participants), none of the personality traits was associated with cancer mortality. These data suggest that personality is not associated with increased risk of incidents of cancer or cancer-related mortality. It is fanciful, irresponsible and actually unethical to suggest that interventions based on the belief that cancer can be beaten by ’mind power’ have any effect on cancer mortality. Positive psychologists please take note! If only this null effect were more widely appreciated, sufferers would be spared from wasting their money on expensive private treatments based on unsubstantiated claims and also, perhaps more seriously, be spared from blaming themselves for having become ill in the first place, and for failing to get better. As a service to patients, national professional psychological societies such as the British Psychological Society and the American Psychological Association should promote a balanced view of the evidence rather than promote ’quack’ psychological interventions for physical illnesses such as cancer.
1. Investigations are needed to distinguish between personality variables that are associated with biologically verified illness and those that are associated only with self-reported illness.
2. There is a need for studies to establish which dimensions of personality are directly associated with health-relevant physiological variables, and to distinguish them from those that are primarily associated with health behaviours.
3. Outcome studies are needed to assess the effectiveness of interventions designed to modify psychological characteristics that are suspected of being hazardous to health.
4. In the case of anxiety and depression, interventions would obviously be worthwhile if they relieve these conditions, whether or not they influence health outcomes.
1. From the ancient Greeks to modern times, medical practitioners have usually believed that there is a physical basis to all illness, including psychological disorders. In contrast to this, there has also been a holistic tradition that has placed an emphasis on the role of psychological factors.
2. Health psychologists have found it very difficult to determine whether personality is associated with susceptibility to physical disease directly through physiological mechanisms or indirectly by way of health behaviour, or whether the data are best explained by statistical artefacts and flaws in the design of the studies from which they are obtained.
3. We have identified eight different ways in which a positive association between a personality trait and an illness measure can be obtained in a misleading manner.
4. Early indications that the Type A personality is a risk factor for CHD were not confirmed by later studies. Attention then shifted to hostility, but this variable now seems only to be a weak predictor, if a predictor at all.
5. Anxiety, especially phobic, panic-like anxiety, and depression are both associated with an increased risk of CHD, although a number of different interpretations of these associations are possible.
6. There is no clear-cut evidence to support the view that personality variables are associated with risk of cancer or of relapse following treatment. Interventions based on treatment for depression have not been shown to improve the survival chances of cancer sufferers.
7. There is substantial evidence that ’positive’ psychological characteristics, including sense of coherence, are associated with reduced risk of CHD and all-cause mortality. This may be the inverse of the parallel findings for anxiety and depression, which have strong negative correlations with positive characteristics.
8. Conscientiousness has been shown in many studies to be predictive of longevity. A major reason seems to be that this characteristic is negatively related to many risky health-related behaviours and positively related to beneficial health-related behaviours.
9. The study by Jokela et al. (2014), which pooled data from six prospective studies, suggests that personality is not associated with increased risk of incidents of cancer or cancer-related mortality. Interventions based on the belief that cancer can be beaten by ’mind power’ have no effect whatsoever on cancer mortality.
10. As a service to patients, national professional psychological societies such as the British Psychological Society and the American Psychological Association need to promote a balanced view of the scientific evidence and avoid occasionally giving the impression of promoting ’quack’ psychological interventions for physical illnesses such as cancer.