The Current Literature
British Journalof Psychiatry (1992), 161,535—541
Stressful Life Events and Graves' Disease TIRRILHARRIS, FRANCIS CREED and TRAOLACH
S. BRUGHA
“¿The role of stressful life events in the onset of Graves' disease (toxic diffuse goitre) is controversial. However, the numerous early clinicalreports that supported such an association were not adequately controlled and specificity of the diagnosis could be questioned. Later studies have not shown a causal relation, but these studies were small, did not have proper controls, or epidemiological methods were inappropriate. To assess possible associations between life events, heredity, social support, and Graves' disease, we have done a population based case-control study in a defined area with about 1 million inhabitants. Over 2 years, 208 (95%) of 219 eligible patients with newly-diagnosed Graves' disease and 372 (80%) of all selected matched controls answered an identical mailed questionnaire about marital status, occupation, drinkingand smoking habits, physical activity, familialoccurrence of thyroid disease, life events, social support, and personality. Compared with controls, patients claimed to have had more negative life events in the 12 months preceding the diagnosis, and negative life-event scores were also significantly higher (odds ratio 6@3, 95% confidence interval 2@7—14@7, for the category with the highest negative score). Individuals who had relatives with thyroid disease (especially first-degree and second-degree relatives) were more likely to have Graves' disease (3@6, 2@2—5@9). Slightly more patients than controls were divorced (1 .8, 1 @O-3@3) and reported a less frequent intake of alcohol (0@4, 0@2-0@8). When results were adjusted for possible confounding factors in multivariate analyses, risk estimates were almost unchanged. These findings indicate that negative life events and hereditary factors may be risk factors for Graves' disease.―
The summary quoted above is from an article by Winsa et al (1991). The present authors were invited to comment upon the study.
ratio 6.3), and even when the year was divided into three4-monthperiods, thisheldequally foreachone.
central Sweden. Over 2 years, 208 (95°lo)of 219
Secondly, those who had relatives with thyroid disease (especially first- and second-degree relatives) were more likely to have Graves' disease. The prevalence of divorce was higher and job satisfaction lower in cases, compared with controls. However, social support and mastery did not differ between the two groups, and the tendency for cases to show more type-A behaviour than controls was not statistically
eligible patients with newly diagnosed Graves' disease
significant.
and 372 (80°lo) of all controls, matched for sex, date of birth, and current country of residence, and selectedfrom the population registerthat coversthe whole of Sweden, answered an identical mailed questionnaire. This covered marital status, occupa tion, drinking and smoking habits, physical activity,
possible confounding factors in multivariate analyses,
Tirril Harris The authors report a population-based case-control study of toxic diffuse goitre in a defined area of
When the results were adjusted
for
the risk estimates were almost unchanged. There are many reasons why the results of this study should be taken seriously. Most notably, it has avoided some of the methodological shortcomings of
using a modified version of Sarason's Life-Event
previous work: the size of the sample (219 patients) is greater than reported in most other studies, and the fact that they were all newly diagnosed, combined
Survey (LES),
with the low refusal rate (only 5°lo)and the care with
familial occurrence
version
of thyroid
social support
of Sarason's
disease, life events using a shortened
Social Support
Questionnaire
(SSQ), mastery (the authors' own 13-item scale with particular attention to work mastery), and nine of
the statements from the Jenkins Activity Survey to measure type-A behaviour. Two results stand out: firstly, compared with controls, patients had experienced more negative life events in the 12 months preceding the diagnosis (odds 535
which a control group was selected, must make the findings all the more convincing. There is therefore no reason to believe that these findings are the result of selection bias. As with most life-event studies, where data collection must involve retrospective reporting of life-stress, the authors have had to confront the possibility that recall bias might confound their findings. Here, two features of their
536
HARRIS ET AL
results are reassuring: if such bias were the only thing from a conservative perspective, therefore not clearly increasingthe numberof negativeeventsreported, a potential cause of the illness. By contrast, it would reasonably be expected to affect the number
independent events, suchas redundanciesdue to an
of positive events in a corresponding manner, and
employer's mismanagement, enforced house moves due to municipal planning policies, and most deaths, would still be considered potential causal agents, because their origins could not be attributed to the agency of the patient. By combining these two strategies-firstly, the use of putative onset date rather than date of diagnosis for those patients who were able to give plausible reports of their symptoms, and, secondly, performing two sets of analyses, one using all events and the other independent events alone—theauthors could have made their account even more convincing. However, it is possible to extract some reassurance from the patterning of the results they do report. The first involves neutral events (not negative but involving change, such as choosing to leave a job or
this was not found. Moreover, if the effect of biased accounts was to produce a fall-off in reporting with increasing time before interview, the temporal patterning mentioned earlier would not have been expected. The fact that the association between patient status and family history of thyroid disease was stronger for second- than for first-degree relatives is less reassuring: while this may be due to chance, there is also a possibility that patients had become alerted to the topic of thyroid illness, and therefore made more effort to track it down throughout their families. This would certainly be in accord with a conclusion about the relative untrustworthiness of
the reporting of affective disorder in family members using a sample of twins, and the superiority of the family study over the family history method
(Kendler et al, 1991). This is one reason for focusing on the findings concerning stress,rather than on those concerning heredity. That focus is also encouraged by the authors noting that their data gave no indication of what may be called a ‘¿vulnerability effect', in that
the effect of negative life events was similar in individuals with and without familial thyroid disease. This implies that the two factors influence the risk
of Graves' disease through different mechanisms. The authors are aware that taking the 12 months before diagnosis, rather than before disease onset, for the study of stressful experience has its drawbacks, in that some of the events may have occurred between
onset and diagnosis and therefore would not be eligible as putative aetiological agents. There are, however, two possible refmements which could have reduced these drawbacks. Firstly, while the establishment of a date of onset can be difficult for
such a condition as Graves' disease,it is quite possible in a number of instances to obtain reports of the start of a cluster of symptoms around a certain time, in a way which approximates to the reports used to date the onset of depression (Brown &
Harris, 1978), gastrointestinal disorder (Craig, 1989), and menorrhagia (Harris, 1989). It is also possible to rate the source of the negative life events according to their independence from the disorder under study: this allows the investigator to repeat the analyses using independent events only. An event which is not ‘¿independent' could be judged to be the result of an
insidiousappearanceof the illness—for example,a car accident when the patient was driving or the break-up of a relationship after an argument—and,
moving house); there is as much reason to expect
an increased rate of these as of negative events if events are the result of thyrotoxicosis. However, since this effect was reported to be absent, the notion that the high rate of negative scores may be
consequent upon, rather than aetiological for, the disorder is less plausible. Secondly, the differences in negative-event scores between cases and controls were less marked in the two months immediately preceding diagnosis than in the six months before that; whereas if the high rate was mainly due to events which were the result of the disease, the two months before diagnosis would be expected to show the largest differences of the whole year. Taken as a whole, therefore, the authors' failure to relate stress to a precisely dated onset has probably not jeopardised their overall conclusion. Although analyses of the type engendered by
the Sarason Life-Event Survey involve the use of additive scores, it can often be illuminating to use an approach to the data in terms of ‘¿experience of at least one ...‘(e.g. ‘¿at least one markedly negative agent' as opposed to three mildly negative events
summing to the same total score). The authors began to adopt this approach when they presented figures for the experience of at least one death of a close
relative or friend, which was significantly more frequent among the cases. But it would have been interesting to pursue this further, not only in these terms (i.e. the basic classification
of the events, such
as promotions, accidents, marital separations, etc.), but also in terms of more global domains such as work, relationships, and finance. Instruments which employ this level of classification of stress also often allow the documentation in each of these domains of ongoing difficulties, which are still sources of
537
LIFE EVENTS AND GRAVES' DISEASE
distress, although they may not themselves have
which predisposes to helplessness and even passivity.
produced
Work on depression has been responsible for the
any real life event (only less substantial
hassles or incidents, such as a spouse never returning home from work until it is almost time for bed). There is some indication that more of the cases than the controls may have been undergoing work difficulties of this kind, but it is not clear whether the problems really lay in their work situations or whether they might not be showing a hypersensitivity to their job security—inother words, whether they might not have different styles of appraisal rather than different environments from those of the controls. Once again, use of a different instrument which contrasted these two aspects—thesubjective report which indicates the respondent's style of appraisal and the actual features of the more objective context—wouldhave been able to throw light on this type of issue. Certainly, the inclusion of difficulties along with events would allow a more realistic assessment of the overall risk attributable to stress to be calculated, using PAR (population attributable risk). In this connection,
two further
results
merit
comment, since they involve factors traditionally eligible to be mediators between stress and disorder, and so have some bearing on the appraisal process.
Firstly the non-significant trend for more type-A behaviour among cases may suggest that they have personalities
which tend towards
a more active,
agitated response style, which may lead them to report as stressful, work situations which the less agitated
personalities
among
the controls
would
experience as less stressful. There has been some scepticism about the ability of the Jenkins Activity Survey to capture the essence of vulnerability to coronary heart disease, as compared with the original
more challenging interview measure of type-A behaviour (Friedman & Rosenman, 1959). However, in terms of Graves' disease, this may be less relevant: thehastyrestlessness of type-A2behaviourtypified by the Jenkins instrument seems intuitively to be just
as important for thyrotoxicosis (if not more so) as the impatient irascibility of type-Al behaviour, which is increasingly seen as the crucialfactor for myocardial infarction (Matthews et a!, 1977; Dembrowski et a!, 1977). In the light of the earlier
comments, it would be interesting to know if those with the higher Jenkins scores were also those reporting more job insecurity. If they were unrelated,
then speculation that the cases had really experienced more contextually severe job insecurity would obtain some support. However, there can be other kinds of hypersensitive appraisals than those deriving from a type-A personality-for example a pessimistic appraisal style
classic aetiological
model of this second type of
appraisal, and has highlighted the role of social support in preventing depression by promoting optimism
and mastery.
In this study,
absence of even a non-significant
then,
the
trend for social
support to be deficient among the patients is all the more suggestive. It could indicate that this more helpless,lessactivated,responsestyle,whichcanbe mitigated by the support of close others, is of little relevance for this particular thyroid illness. This would accord well with what has already been mentioned as the intuitive impression of Graves' disease as a disorder of over-activation. However, an alternative explanation which must be considered is that the particular measure of support used may be failing to capture the essential features reported in other studies, but this is unlikely, given the track
record of the SSQ in work on distress and depressive symptoms (Sarason et a!, 1992). The contrast between disorders of over-activation and those of disengagement was first suggested nearly 20 years ago (Singer, 1974), but has received an undeserved lack of attention. It has, however, recently been discussed in connection with a parti cular instrument for measuring stressful experience the Life Event and Difficulty Schedule (LEDS)—as
part of a general debate on the specificity hypothesis (Harris & Brown, 1989). Here, it is argued that approaching the categorisation of stress at a particular level, which allows the discrimination of the cognitive-affective qualities of each life event, illuminates the specificity of the aetiology of disease
and highlights the difference between disorders of over- and underinvolvement respectively. Examples of cognitive-affective qualities of stressors producing disorders of disengagement
are losses and failures,
while those producing overengagement are dangers, challenges, and intrusions. Positing a psychosomatic continuum of illnesses, it is speculated, on the basis of detailed findings with the LEDS, that myocardial infarction, secondary amenorrhoea, anxiety, and schizophrenia are disorders of over-activation, while
depression, menorrhagiaandbacterialinfectionsare those of disengagement. In the light of this perspective, it would have been useful if the Swedish authors had been able to describe the cognitive-affective qualities of the negative events recorded among the sufferers from Graves' disease: were there a higher percentage of dangers and intrusions, or even frustrations, than
would appear in an ordinary range of the stressors which tend to precede depression,
where losses
predominate (Brown et al, 1987)? A reanalysis of
HARRIS ET AL
538
the data in this way might provide an interesting pointer, but unfortunately might not be possible without the full range of contextual information which would allow an assessment of the degree of, say, intrusiveness, and which requires a semistructured
face-to-face interview such as the LEDS. In this connection,
it is worth drawing attention
to a study using the LEDS with depressed patients which also examined thyroid function. Not only had patients with increased thyroid hormone output and blunting on the test for thyrotrophin-releasing hormone experienced more ongoing difficulties, most of which represented threats to security in relation ships or at work, but patients with severe danger events had a higher free thyroxine index than those with severe losses (Calloway & Dolan, 1989). None of the sample were clinically hyperthyroid, but the authors speculate as to whether there may not be similar social factors in the onset of true thyrotoxicosis. One of the bases on which they are prepared to embark on such speculations about this parallel is the similarity between, on the one hand, their finding that those with blunting due to high thyroid hormone output reported significantly more parental separations (not deaths) before the age of
11, and on the other hand, the report of a high incidence of childhood loss of parents among those with Graves' disease (Weiner, 1977). Unfortunately, the Swedish study has given no data concerning such childhood life events. Calloway & Dolan note a vulnerability effect, whereby those with blunting due to high thyroid hormone output and an increased number of ongoing difficulties are also those with the childhood separations from parents; they comment that “¿such individuals may have become ‘¿sensitised' to insecurity or danger in later life― by these
childhood experiences. Thus, if the findings concerning thyroid function among depressed patients are any guide to a perspective on Graves' disease, they do seem to offer support to the overengagement hypothesis. Finally, three further findings in the Swedish study should not go unnoticed. The first is the weak association between Graves' disease and frequent episodes of both viral and bacterial infections during the 12 months, which is interpreted by the authors as supporting both the hypothesis that viral infections may be a cause of many autoimmune diseases, and that Graves' disease renders people more susceptible to bacterial infection. Again, little can be concluded without more clarity about the true date of onset, but, in principle, there seems no reason why both processes may not be occurring, though in different periods (before versus after onset). There need therefore be no conflict between this finding and the
theory outlined above, whereby bacterial infections tendto beassociated with situationspredisposing to disengagement rather than overengagement. Indeed, once this more elongated perspective of the course of the illness is accepted, it becomes easier to see how disorders which follow situations of overengagement may then enter the causal chain themselves as disengagement situations. Thus, although it is the consequence of over-involvement, a heart attack may itself be a depressogenic
experience,
leading
to
disengagement. Secondly,
it would be interesting
to know more
about the background to the higher rate of divorce among cases than controls—for example, had the marital separation been initiated by the patients or by their spouses (higher ratings of loss being associated with the latter)? And had there been associated violence (or danger)? Finally, the higher consumption
of alcohol by controls than by cases can also be seen asanothersuggestionin the data that the disorder isoneof overengagement, in that thesufferersfrom Graves' disease had not availed themselves
of one
means of combating the process of overactivation. Francis Creed The authors concluded that life stress plays a part in the aetiology of Graves' disease, but unfortunately,
the study used a self-administered questionnaire, which invalidates this conclusion. Establishing that stressful life events cause illness requires several
conditions to be met (Cooke, 1986; Macs eta!, 1987; Day, 1989): a statistical correlation between life events and onset of disorder, a demonstration that events lead to illness rather than the other way round, repeated results across different populations at different times, and, preferably, an understanding of the role of intervening variables together with a dose—responserelationship between stressful events and illness. Compared with this formidable list, the relationship between stressful life events and the onset of Graves' disease looks far from proven. The first problem is defining onset. These authors have used time of diagnosis, since the latency period
from true onset to diagnosis may vary. This, as the authors admit, means that the life events may have followed rather than preceded onset. The second problem is measurement of life events. The authors presumably did not have the resources to employ an interviewer-rated method, such as the LEDS. This was unfortunate, as the arguments
against using a self-administered questionnaire (because of lack of reliability) are powerful and have often been rehearsed (Brown & Harris, 1978; Paykel, 1983; Dohrenwend et a!, 1987).
LIFE EVENTSAND GRAVES' DISEASE The key issue is whether events are accurately recalled. If not, there is a natural tendency for ill people to seek in their recent past an explanation for
the illness(known as ‘¿effort after meaning').The authors of the present paper argue that this is not the case with their study because positive events were recalled with equal frequency by both the experimental and control groups. This argument does
not hold, however, since negative events may be selectively recalled by those trying to seek a reason for illness. The report does include data regarding additional life events, that is, those not included in the Life-Events Survey (Sarason eta!, 1978). The 208 patients with Graves' disease recalled 83 additional events (mean 0.39 each), compared with only 23 additional events recalled by the 372 controls (0.06 each)—thepatients were presumably searching in their recent past for life stress more diligently than the controls.
539
more definite interpretation. Death of a close relative or friend was recorded significantly, but not dramatically, more often in the index group (15°lov. 10%); but deaths of close relatives alone
would need to be considered, to provide a definite clue that this form of stressful event was im portant. There were more divorced people in the index group (13% v. 8°lo), but no clue as to whether the divorce was recent. The lack of any difference between
cases and controls
in social support
suggests
that recent loss of a crucial relationship was not important; a significant excess of recent negative life events is also often reflected in a reduction of social support. It is unfortunate that this study did not include an objective method of measuring life events, since in other respects it was well designed. The inclusion of a multivariate analysis allowed the relative importance
of life stress and other aetiological factors (e.g.
It is important to note that mean negative score,
family history) to be assessed. No attempt was made
notthenumberof events,wasanalysed(+2 for very positive to —¿2 for very negative). In other words, the experience of life events might have been similar, but the subjects with Graves' disease may have
to relate the severity of biochemical abnormality with
perceived the events in a more negative light. This
the negative events score, which would have provided an interesting test of the dose-response relationship that was so neatly demonstrated in another study concerning immune response and psychological stress
is a direct measure of subjective rating of the event
(Cohen et a!, 1991).
by the respondent;
The balance of evidence of previous work is against a positive association between negative life events and
an objective,
but personally
meaningful, measure is required for aetiological research (Brown & Harris,
1978).
Another important point is that mood was not measured. If the period of weeks following the diagnosis of Graves' disease is associated with depression, this alone would explain the greater negative scores for life events. Depression would be a plausible intervening variable between life stress and onset of an autoimmune disease, as there is evidence that depression, as well as or instead of negative life events, is closely related to alteration of
the immune response (Irwin et a!, 1990). Unfortu nately, the authors did not measure depression, even though it may be present in one-third of subjects with hyperthyroidism (Kathol et a!, 1986).
The failure of the self-administered questionnaire to prompt accurate recall is seen in the ‘¿fall-off' of
events with increasing time from the interview. During the 3 months furthest from interview (i.e. 9—12months before onset), the number of people with a negative events score was half that recorded for the 3 months closest to interview (1—4months
before onset). This cannot be attributed to an excess of negative events just before the diagnosis
of
Graves' disease, as an identical reduction was demonstrated for the control group. Unfortunately, details of the events are not given, though these sometimes allow a fuller picture and a
the onset of Graves' disease (Jadresic, 1990). Only one paper approached a satisfactory method (Gray & Hoffenberg, 1985) and used sufficient patients—the result was negative. Evidence from other physical illness is conflicting. In the majority of physical ill nesses studied with the LEDS, there is no clear link with severely threatening life events, though the exceptions are stroke, where a weak but positive effect has been reported (House et a!, 1990) and multiple sclerosis (MS), where the pattern of events before onset of first symptoms was similar to that found in depression (Grant et a!, 1989). The onset of MS is also related to immunological change, but there is also a problem in defining onset with this
disorder. This is an important area of research, which is gradually increasing our knowledge of the multifactorial causation of disease, but only when objective measures are used (Creed, 1992). T. S. Brugha Winsa et a! claim that their study has shown that “¿stress can be an important factor in the cause of [Graves'] disease―. Before considering this claim more critically, it may be worth pondering for a moment that if they had produced a negative fmding, we might have been more surprised. As Thoits (1991)
540
HARRIS ET AL
has pointed out: ‘¿ ‘¿stress theory can be criticised for
review process of a high-quality general psychiatry or psychology journal, the readers of which would not need to be reminded of the considerable advances in the reliability of life-event measurement that have of physical illness and of poor mental health―. followed on the development of investigator-based Accordingly, the study under consideration here can measures such as the LEDS. be criticised for its failure, apriori, to give particular To take just one objection a little further: there is consideration to the possible theoretical link between good evidence that memory is differentially affected its lack of specificity with respect to outcome; the same theoretical factors (stress exposure, lack of coping resources) are used to explain the occurrence
adversity and Graves' disease, and, in order to do this, to consider what aspects of these two factors might be particularly related to one another. Turning first to the aetiological side of the
by mood and, in particular, that negative events or
equation, Thoits (1991) has argued, according to her identity theory of stress research, that identity
could explain the increased reporting in such cases
irrelevant stressors should influence physical health directly by means of general physiological arousal, whereas identity-relevant stressors should influence psychological symptoms directly. For example, the loss of a highly valued social role or attribute might increase the risk of depression. Thus, the finding by Winsa et a! that the total amount of stress does not
play a part in Graves' disease, but that life events that are self-rated as negative in their effect are significantly associated with it, is contrary to the model proposed by Thoits. Their finding would be consistent instead with a psychological disorder, perhaps most saliently with depression. In fairness to these authors, much research on stressful life events within the field of psychological and psychiatric disorder does not incorporate an attempt to link specific types of events in people with specific types of vulnerability to specific forms of psychiatric disorder. Although the idea is clinically appealing and theoretically more substantial, in practice, attempts to do so have encountered considerable methodological difficulties (e.g. Finlay-Jones &
Brown, 1981). Turning next to the illness side of the equation, what particular issues arise? Graves' disease presents an additional difficulty for stress researchers because of the coexistence in the clinical presentation of psychological symptoms. According to Lishman (1978), “¿psychological disturbance in some degree is universal with thyroid overactivity―; both anxiety and less often depressive features have been clearly described. Thus, any attempt to study the relation ship between stress and coping on the one hand and Graves' disease on the other would require that psychiatric comorbidity be taken into account. These authors appear to have overlooked this, possibly to
the detriment of their hope that we would be wiffing to interpret their findings as they do. Winsa et a! accept that “¿recall bias is a serious concern―. Their choice of a self-completed and self-rated measure of adversity would almost certainly not have passed the
memoriesare better recalledin the presenceof a negative affect (Clark, 1983). The presence of such affects in the clinical manifestation
of thyrotoxicosis
of negative events in comparison with healthy, though otherwise appropriately matched controls.
The argument by Winsa eta! that there was probably no recall bias in their own study, because cases and controls claimed to have equal numbers of reliable supporters and satisfaction with these social relation ships, does not answer the objection, since their measure of social support
(Sarason
et a!, 1983) is a
state measure that does not rely on memory to the same degree that retrospective life-event measures do. However, this negative finding may also have some additional value, although it may be of no direct interest to the aims of their study. A frequent objection to the fmding that deficits in social support
are associated with the presence of psychiatric disorders, such as depression, is that patients view their social world negatively (Henderson eta!, 1981). A physical disorder, such as Graves' disease, in which clinically significant psychological disturbance occurs
more than in the general population, would provide one opportunity to test the strength of this objection. Unfortunately, no data on psychiatric comorbidity were gathered by Winsa et a!, although had their study included such additional information, it could
have been of value to those concerned with exploring this issue. One can only conclude, therefore, that any (secondary) psychological morbidity present in the cases reported in this study was unlikely to have significantly biased their self-reports of personal relationships and their supportiveness. Winsa et a! point out that a previous prospective study, in which the control group consisted of patients with non-toxic goitre, suggested that stress was not a cause. Prospective research on the effects of life events is also problematical (Brown eta!, 1986),and perhaps almost impossible if the investigator is concerned with predicting the first episode of a condition that is rare and otherwise unpredictable. However, there is considerable scope for improving the quality of cross-sectional and retrospective research. The present study was commendable for the method of sampling
used, which was epidemiologically based. Future
LIFE EVENTS AND GRAVES'
research in this area should measure coexisting psychological disturbance, take into account any bias due to it, and make use of methods of measuring life events that have already been shown to be capable of dealing with the problems of recall. In the meantime, the safest answer to the question of whether stress causes Graves' disease is: perhaps.
541
DISEASE
Gii@y, J. & HOFFENBERG, R. (1985) Thyrotoxicosis and stress. Quarterly Journal ofMedicine, New Series,54, 153—160. HARRIS, 1. 0. (1989) Disorders
of menstruation.
In Life Events
and Illness (eds 0. W. Brown & 1. 0. Harris). New York: Guilford Press. & BROWN,0. W. (1989) The LEDS findings in the context of other research: An overview.
In Life Events and
illness(eds 0. W. Brown & 1. 0. Harris). New York: Guilford
Press. HENDERSON, S., BYRNE, D. 0. & DUNCAN-JONES, P. (1981) Neurosis
and The Social Environment. Sydney: Academic Press. HOUSE,
References BROWN, 0.
W.
&
HARRIS,
1.
0.
(1978)
Social
Origins
of
Depression.London:Tavistock. Ar@inanws, B., HARRIS,T., et a! (1986) Social support, self-esteem and depression. Psychological Medicine, 16,813-831. BwuLco, A. T. & HARRIS,T. 0. (1987) Life events, vulnerability, and onset of depression. British Journal of Psychiatry, 150, 30—42. CALLOWAY,P. & DOLAN,R. (1989) Endocrine change and clinical
profiles in depression. In Life Events and Illness(eds G. W. Brown & T. 0. Harris). New York: Guilford Press. Ci@sx,
D (1983) Differential
Effects
of Mood
on the Accessibility
of Positive and Negative Information. Unpublished DPhil Thesis, University of Oxford. COHEN,S., TYRELL,D. A. J. & SMITh, A. P. (1991) Psychological
stress and susceptibility to the common cold. New England Journal of Medicine, 325, 606—612. CoolcE, D. J. (1986) Inferring causality in life event research. Stress
Medicine, 2, 141—152.
Cit@io,T. K. J. (1989)Abdominalpain. In LifeEventsand Illness (eds 0. W. Brown & T. 0. Harris). New York: Guilford Press. CREED,F. H. (1992) Life events and disorder. Current Opinion in
Psychiatry, 5, 300—304. DAY, R. (1989) Schizophrenia. In Life Events and illness (eds 0. W.
Brown&1.0. Harris),pp. 113-137.NewYork:GuilfordPress. DEMBROWSKI, 1. M. MACDOUGALL,J. M. & SHIELDS,J. L. (1977) Physiologic reactions to social challenge in persons evidencing
the type A coronary-pronebehavior pattern. Journal of Human Stress, 3, 2—10. DOHRENWEND, B. P.,
LEVAV, I., SHEoUT,
A.,
DENNIS,
M.,
MOGRIDGE,
L.,
et
al
(1990)
Life
events
and difficulties preceding stroke. Journal of Neurology, Neurosurgery and Psychiatry, 53, 1024—1028.
P. E., et a! (1987)
Life
stress and psychopathology: progress on research begun with Barbara Snell Dohrenwend. American Journal of Community Psychology, 15, 677—715. FIr@aAY-JoHEs,R. & BROWN,0. W. (1981) Types of stressful life
event and the onset of anxiety and depressive disorders. PsychologicalMedicine,11, 803—816. FRIEDMAN,M. & ROSENMAN,R. H. (1959) Association
of specific
overtbehaviorpattern withbloodand cardiovascularfindings.
IRWIN,M., PA1-IERSON,T., SMITH,1. L., et a! (1990) Reduction
of immune function in life stress and depression. Biological Psychiatry. 27, 22—30. JADRESIC, D. P. (1990) Psychiatric
aspects of hyperthyroidism.
Journal of Psychosomatic Research, 34, 603—615. KATHOL,R. 0., TURNER,R. & Dati@nuwr, J. (1986) Depression and anxiety
associated
with hyperthyroidism:
therapy. Psychosomatics, KENDLER,
K.
S.,
SILBERG,
response
to antithyroid
27, 501—505. J.
L.,
NEALE,
M.
C.,
et
a! (1991)
The
family history method. Whose psychiatric history is measured? American Journal of Psychiatry, 148, 1501-1504. LISHMAN,
W.
A.
(1978)
Organic
Psychiatry.
Oxford:
Blackwell.
LEVINE,H. M., BASHAM,R. B., et a! (1983) Assessing social support: the social support questionnaire. Journal of
Personalityand SocialPsychology,44, 127—139. [email protected],S., VINGERHOETS, A. & VANHECK,0. (1987) The study of stress and disease: some developments and requirements. Social Science and Medicine, 25, 567—578. MArrHEWS, K. A., GLASS, D. C., ROSENMAN,R. H., et a! (1977)
Competitive drive, pattern A and coronary heart disease: a further analysis of some data from the Western Collaborative Study. Journal of Chronic Diseases, 30, 489—498. PAYKEL,E. 5. (1983) Methodological
aspects of life events research.
Journal of Psychosomatic Research, 27, 341—352. SARASON,I., JOHNSON,J. H. & SIEGEL,J. M. (1978) Assessing the
impact of life changes: Development of the Life Experiences Survey. Journal of Consulting and Clinical Psychology, 46, 932—946. SARASON,B. R. & PIERCE, 0. R. (1992) Three contexts of
social support. In The Meaning and Measurement of Social Support (eds H. 0. F. Veiel & U. Baumann). New York: Hemisphere. Su@nE1t, M. 1. (1974)Presidential address. Engagement-involvement:
a centralphenomenonin psychophysiological research.Psycho somatic Medicine, 36, 1—17. THOIrs, P. A. (1991) On merging
identity
theory
and stress
research.SocialPsychologyQuarterly,54, 1—112.
Journalof theAmericanMedicalAssociation,169, 1280—1296. WEINER,H. (1977) Psychobiology and Human Disease. New York:
GRANr,I., MCDONALD, W. I. PATrERSON, 1., et a! (1989)Multiple sclerosis. In Life Events and illness (eds 0. W. Brown & 1. 0. Harris), pp. 295—311.New York: Guilford Press.
@
Elsevier. WINSA, B., ADAMI, H.-0.,
BERGSTROM, R., eta! (1991) Stressful
life eventsand Graves' disease.Lancet, 338, 1475—1479.
Tirril Harris, Research Officer, Department of Social Policy and Social Science, Royal Holloway and Bedford New College, University of London, 11 Bedford Square, London WCJB 3RA; Francis Creed, MD, FRCP, FRCPsych, Professor, Department of Psychiatry, Rawnsley Building, Manchester Royal Infirmary, Manchester M13 9WL; Traolach S. Brugha MD, MRCPsych, Senior Lecturer, Department of Psychiatry, University of Leicester, Leicester LE2 7LX
British Journalof Psychiatry (1992), 161,535—541
Stressful Life Events and Graves' Disease TIRRILHARRIS, FRANCIS CREED and TRAOLACH
S. BRUGHA
“¿The role of stressful life events in the onset of Graves' disease (toxic diffuse goitre) is controversial. However, the numerous early clinicalreports that supported such an association were not adequately controlled and specificity of the diagnosis could be questioned. Later studies have not shown a causal relation, but these studies were small, did not have proper controls, or epidemiological methods were inappropriate. To assess possible associations between life events, heredity, social support, and Graves' disease, we have done a population based case-control study in a defined area with about 1 million inhabitants. Over 2 years, 208 (95%) of 219 eligible patients with newly-diagnosed Graves' disease and 372 (80%) of all selected matched controls answered an identical mailed questionnaire about marital status, occupation, drinkingand smoking habits, physical activity, familialoccurrence of thyroid disease, life events, social support, and personality. Compared with controls, patients claimed to have had more negative life events in the 12 months preceding the diagnosis, and negative life-event scores were also significantly higher (odds ratio 6@3, 95% confidence interval 2@7—14@7, for the category with the highest negative score). Individuals who had relatives with thyroid disease (especially first-degree and second-degree relatives) were more likely to have Graves' disease (3@6, 2@2—5@9). Slightly more patients than controls were divorced (1 .8, 1 @O-3@3) and reported a less frequent intake of alcohol (0@4, 0@2-0@8). When results were adjusted for possible confounding factors in multivariate analyses, risk estimates were almost unchanged. These findings indicate that negative life events and hereditary factors may be risk factors for Graves' disease.―
The summary quoted above is from an article by Winsa et al (1991). The present authors were invited to comment upon the study.
ratio 6.3), and even when the year was divided into three4-monthperiods, thisheldequally foreachone.
central Sweden. Over 2 years, 208 (95°lo)of 219
Secondly, those who had relatives with thyroid disease (especially first- and second-degree relatives) were more likely to have Graves' disease. The prevalence of divorce was higher and job satisfaction lower in cases, compared with controls. However, social support and mastery did not differ between the two groups, and the tendency for cases to show more type-A behaviour than controls was not statistically
eligible patients with newly diagnosed Graves' disease
significant.
and 372 (80°lo) of all controls, matched for sex, date of birth, and current country of residence, and selectedfrom the population registerthat coversthe whole of Sweden, answered an identical mailed questionnaire. This covered marital status, occupa tion, drinking and smoking habits, physical activity,
possible confounding factors in multivariate analyses,
Tirril Harris The authors report a population-based case-control study of toxic diffuse goitre in a defined area of
When the results were adjusted
for
the risk estimates were almost unchanged. There are many reasons why the results of this study should be taken seriously. Most notably, it has avoided some of the methodological shortcomings of
using a modified version of Sarason's Life-Event
previous work: the size of the sample (219 patients) is greater than reported in most other studies, and the fact that they were all newly diagnosed, combined
Survey (LES),
with the low refusal rate (only 5°lo)and the care with
familial occurrence
version
of thyroid
social support
of Sarason's
disease, life events using a shortened
Social Support
Questionnaire
(SSQ), mastery (the authors' own 13-item scale with particular attention to work mastery), and nine of
the statements from the Jenkins Activity Survey to measure type-A behaviour. Two results stand out: firstly, compared with controls, patients had experienced more negative life events in the 12 months preceding the diagnosis (odds 535
which a control group was selected, must make the findings all the more convincing. There is therefore no reason to believe that these findings are the result of selection bias. As with most life-event studies, where data collection must involve retrospective reporting of life-stress, the authors have had to confront the possibility that recall bias might confound their findings. Here, two features of their
536
HARRIS ET AL
results are reassuring: if such bias were the only thing from a conservative perspective, therefore not clearly increasingthe numberof negativeeventsreported, a potential cause of the illness. By contrast, it would reasonably be expected to affect the number
independent events, suchas redundanciesdue to an
of positive events in a corresponding manner, and
employer's mismanagement, enforced house moves due to municipal planning policies, and most deaths, would still be considered potential causal agents, because their origins could not be attributed to the agency of the patient. By combining these two strategies-firstly, the use of putative onset date rather than date of diagnosis for those patients who were able to give plausible reports of their symptoms, and, secondly, performing two sets of analyses, one using all events and the other independent events alone—theauthors could have made their account even more convincing. However, it is possible to extract some reassurance from the patterning of the results they do report. The first involves neutral events (not negative but involving change, such as choosing to leave a job or
this was not found. Moreover, if the effect of biased accounts was to produce a fall-off in reporting with increasing time before interview, the temporal patterning mentioned earlier would not have been expected. The fact that the association between patient status and family history of thyroid disease was stronger for second- than for first-degree relatives is less reassuring: while this may be due to chance, there is also a possibility that patients had become alerted to the topic of thyroid illness, and therefore made more effort to track it down throughout their families. This would certainly be in accord with a conclusion about the relative untrustworthiness of
the reporting of affective disorder in family members using a sample of twins, and the superiority of the family study over the family history method
(Kendler et al, 1991). This is one reason for focusing on the findings concerning stress,rather than on those concerning heredity. That focus is also encouraged by the authors noting that their data gave no indication of what may be called a ‘¿vulnerability effect', in that
the effect of negative life events was similar in individuals with and without familial thyroid disease. This implies that the two factors influence the risk
of Graves' disease through different mechanisms. The authors are aware that taking the 12 months before diagnosis, rather than before disease onset, for the study of stressful experience has its drawbacks, in that some of the events may have occurred between
onset and diagnosis and therefore would not be eligible as putative aetiological agents. There are, however, two possible refmements which could have reduced these drawbacks. Firstly, while the establishment of a date of onset can be difficult for
such a condition as Graves' disease,it is quite possible in a number of instances to obtain reports of the start of a cluster of symptoms around a certain time, in a way which approximates to the reports used to date the onset of depression (Brown &
Harris, 1978), gastrointestinal disorder (Craig, 1989), and menorrhagia (Harris, 1989). It is also possible to rate the source of the negative life events according to their independence from the disorder under study: this allows the investigator to repeat the analyses using independent events only. An event which is not ‘¿independent' could be judged to be the result of an
insidiousappearanceof the illness—for example,a car accident when the patient was driving or the break-up of a relationship after an argument—and,
moving house); there is as much reason to expect
an increased rate of these as of negative events if events are the result of thyrotoxicosis. However, since this effect was reported to be absent, the notion that the high rate of negative scores may be
consequent upon, rather than aetiological for, the disorder is less plausible. Secondly, the differences in negative-event scores between cases and controls were less marked in the two months immediately preceding diagnosis than in the six months before that; whereas if the high rate was mainly due to events which were the result of the disease, the two months before diagnosis would be expected to show the largest differences of the whole year. Taken as a whole, therefore, the authors' failure to relate stress to a precisely dated onset has probably not jeopardised their overall conclusion. Although analyses of the type engendered by
the Sarason Life-Event Survey involve the use of additive scores, it can often be illuminating to use an approach to the data in terms of ‘¿experience of at least one ...‘(e.g. ‘¿at least one markedly negative agent' as opposed to three mildly negative events
summing to the same total score). The authors began to adopt this approach when they presented figures for the experience of at least one death of a close
relative or friend, which was significantly more frequent among the cases. But it would have been interesting to pursue this further, not only in these terms (i.e. the basic classification
of the events, such
as promotions, accidents, marital separations, etc.), but also in terms of more global domains such as work, relationships, and finance. Instruments which employ this level of classification of stress also often allow the documentation in each of these domains of ongoing difficulties, which are still sources of
537
LIFE EVENTS AND GRAVES' DISEASE
distress, although they may not themselves have
which predisposes to helplessness and even passivity.
produced
Work on depression has been responsible for the
any real life event (only less substantial
hassles or incidents, such as a spouse never returning home from work until it is almost time for bed). There is some indication that more of the cases than the controls may have been undergoing work difficulties of this kind, but it is not clear whether the problems really lay in their work situations or whether they might not be showing a hypersensitivity to their job security—inother words, whether they might not have different styles of appraisal rather than different environments from those of the controls. Once again, use of a different instrument which contrasted these two aspects—thesubjective report which indicates the respondent's style of appraisal and the actual features of the more objective context—wouldhave been able to throw light on this type of issue. Certainly, the inclusion of difficulties along with events would allow a more realistic assessment of the overall risk attributable to stress to be calculated, using PAR (population attributable risk). In this connection,
two further
results
merit
comment, since they involve factors traditionally eligible to be mediators between stress and disorder, and so have some bearing on the appraisal process.
Firstly the non-significant trend for more type-A behaviour among cases may suggest that they have personalities
which tend towards
a more active,
agitated response style, which may lead them to report as stressful, work situations which the less agitated
personalities
among
the controls
would
experience as less stressful. There has been some scepticism about the ability of the Jenkins Activity Survey to capture the essence of vulnerability to coronary heart disease, as compared with the original
more challenging interview measure of type-A behaviour (Friedman & Rosenman, 1959). However, in terms of Graves' disease, this may be less relevant: thehastyrestlessness of type-A2behaviourtypified by the Jenkins instrument seems intuitively to be just
as important for thyrotoxicosis (if not more so) as the impatient irascibility of type-Al behaviour, which is increasingly seen as the crucialfactor for myocardial infarction (Matthews et a!, 1977; Dembrowski et a!, 1977). In the light of the earlier
comments, it would be interesting to know if those with the higher Jenkins scores were also those reporting more job insecurity. If they were unrelated,
then speculation that the cases had really experienced more contextually severe job insecurity would obtain some support. However, there can be other kinds of hypersensitive appraisals than those deriving from a type-A personality-for example a pessimistic appraisal style
classic aetiological
model of this second type of
appraisal, and has highlighted the role of social support in preventing depression by promoting optimism
and mastery.
In this study,
absence of even a non-significant
then,
the
trend for social
support to be deficient among the patients is all the more suggestive. It could indicate that this more helpless,lessactivated,responsestyle,whichcanbe mitigated by the support of close others, is of little relevance for this particular thyroid illness. This would accord well with what has already been mentioned as the intuitive impression of Graves' disease as a disorder of over-activation. However, an alternative explanation which must be considered is that the particular measure of support used may be failing to capture the essential features reported in other studies, but this is unlikely, given the track
record of the SSQ in work on distress and depressive symptoms (Sarason et a!, 1992). The contrast between disorders of over-activation and those of disengagement was first suggested nearly 20 years ago (Singer, 1974), but has received an undeserved lack of attention. It has, however, recently been discussed in connection with a parti cular instrument for measuring stressful experience the Life Event and Difficulty Schedule (LEDS)—as
part of a general debate on the specificity hypothesis (Harris & Brown, 1989). Here, it is argued that approaching the categorisation of stress at a particular level, which allows the discrimination of the cognitive-affective qualities of each life event, illuminates the specificity of the aetiology of disease
and highlights the difference between disorders of over- and underinvolvement respectively. Examples of cognitive-affective qualities of stressors producing disorders of disengagement
are losses and failures,
while those producing overengagement are dangers, challenges, and intrusions. Positing a psychosomatic continuum of illnesses, it is speculated, on the basis of detailed findings with the LEDS, that myocardial infarction, secondary amenorrhoea, anxiety, and schizophrenia are disorders of over-activation, while
depression, menorrhagiaandbacterialinfectionsare those of disengagement. In the light of this perspective, it would have been useful if the Swedish authors had been able to describe the cognitive-affective qualities of the negative events recorded among the sufferers from Graves' disease: were there a higher percentage of dangers and intrusions, or even frustrations, than
would appear in an ordinary range of the stressors which tend to precede depression,
where losses
predominate (Brown et al, 1987)? A reanalysis of
HARRIS ET AL
538
the data in this way might provide an interesting pointer, but unfortunately might not be possible without the full range of contextual information which would allow an assessment of the degree of, say, intrusiveness, and which requires a semistructured
face-to-face interview such as the LEDS. In this connection,
it is worth drawing attention
to a study using the LEDS with depressed patients which also examined thyroid function. Not only had patients with increased thyroid hormone output and blunting on the test for thyrotrophin-releasing hormone experienced more ongoing difficulties, most of which represented threats to security in relation ships or at work, but patients with severe danger events had a higher free thyroxine index than those with severe losses (Calloway & Dolan, 1989). None of the sample were clinically hyperthyroid, but the authors speculate as to whether there may not be similar social factors in the onset of true thyrotoxicosis. One of the bases on which they are prepared to embark on such speculations about this parallel is the similarity between, on the one hand, their finding that those with blunting due to high thyroid hormone output reported significantly more parental separations (not deaths) before the age of
11, and on the other hand, the report of a high incidence of childhood loss of parents among those with Graves' disease (Weiner, 1977). Unfortunately, the Swedish study has given no data concerning such childhood life events. Calloway & Dolan note a vulnerability effect, whereby those with blunting due to high thyroid hormone output and an increased number of ongoing difficulties are also those with the childhood separations from parents; they comment that “¿such individuals may have become ‘¿sensitised' to insecurity or danger in later life― by these
childhood experiences. Thus, if the findings concerning thyroid function among depressed patients are any guide to a perspective on Graves' disease, they do seem to offer support to the overengagement hypothesis. Finally, three further findings in the Swedish study should not go unnoticed. The first is the weak association between Graves' disease and frequent episodes of both viral and bacterial infections during the 12 months, which is interpreted by the authors as supporting both the hypothesis that viral infections may be a cause of many autoimmune diseases, and that Graves' disease renders people more susceptible to bacterial infection. Again, little can be concluded without more clarity about the true date of onset, but, in principle, there seems no reason why both processes may not be occurring, though in different periods (before versus after onset). There need therefore be no conflict between this finding and the
theory outlined above, whereby bacterial infections tendto beassociated with situationspredisposing to disengagement rather than overengagement. Indeed, once this more elongated perspective of the course of the illness is accepted, it becomes easier to see how disorders which follow situations of overengagement may then enter the causal chain themselves as disengagement situations. Thus, although it is the consequence of over-involvement, a heart attack may itself be a depressogenic
experience,
leading
to
disengagement. Secondly,
it would be interesting
to know more
about the background to the higher rate of divorce among cases than controls—for example, had the marital separation been initiated by the patients or by their spouses (higher ratings of loss being associated with the latter)? And had there been associated violence (or danger)? Finally, the higher consumption
of alcohol by controls than by cases can also be seen asanothersuggestionin the data that the disorder isoneof overengagement, in that thesufferersfrom Graves' disease had not availed themselves
of one
means of combating the process of overactivation. Francis Creed The authors concluded that life stress plays a part in the aetiology of Graves' disease, but unfortunately,
the study used a self-administered questionnaire, which invalidates this conclusion. Establishing that stressful life events cause illness requires several
conditions to be met (Cooke, 1986; Macs eta!, 1987; Day, 1989): a statistical correlation between life events and onset of disorder, a demonstration that events lead to illness rather than the other way round, repeated results across different populations at different times, and, preferably, an understanding of the role of intervening variables together with a dose—responserelationship between stressful events and illness. Compared with this formidable list, the relationship between stressful life events and the onset of Graves' disease looks far from proven. The first problem is defining onset. These authors have used time of diagnosis, since the latency period
from true onset to diagnosis may vary. This, as the authors admit, means that the life events may have followed rather than preceded onset. The second problem is measurement of life events. The authors presumably did not have the resources to employ an interviewer-rated method, such as the LEDS. This was unfortunate, as the arguments
against using a self-administered questionnaire (because of lack of reliability) are powerful and have often been rehearsed (Brown & Harris, 1978; Paykel, 1983; Dohrenwend et a!, 1987).
LIFE EVENTSAND GRAVES' DISEASE The key issue is whether events are accurately recalled. If not, there is a natural tendency for ill people to seek in their recent past an explanation for
the illness(known as ‘¿effort after meaning').The authors of the present paper argue that this is not the case with their study because positive events were recalled with equal frequency by both the experimental and control groups. This argument does
not hold, however, since negative events may be selectively recalled by those trying to seek a reason for illness. The report does include data regarding additional life events, that is, those not included in the Life-Events Survey (Sarason eta!, 1978). The 208 patients with Graves' disease recalled 83 additional events (mean 0.39 each), compared with only 23 additional events recalled by the 372 controls (0.06 each)—thepatients were presumably searching in their recent past for life stress more diligently than the controls.
539
more definite interpretation. Death of a close relative or friend was recorded significantly, but not dramatically, more often in the index group (15°lov. 10%); but deaths of close relatives alone
would need to be considered, to provide a definite clue that this form of stressful event was im portant. There were more divorced people in the index group (13% v. 8°lo), but no clue as to whether the divorce was recent. The lack of any difference between
cases and controls
in social support
suggests
that recent loss of a crucial relationship was not important; a significant excess of recent negative life events is also often reflected in a reduction of social support. It is unfortunate that this study did not include an objective method of measuring life events, since in other respects it was well designed. The inclusion of a multivariate analysis allowed the relative importance
of life stress and other aetiological factors (e.g.
It is important to note that mean negative score,
family history) to be assessed. No attempt was made
notthenumberof events,wasanalysed(+2 for very positive to —¿2 for very negative). In other words, the experience of life events might have been similar, but the subjects with Graves' disease may have
to relate the severity of biochemical abnormality with
perceived the events in a more negative light. This
the negative events score, which would have provided an interesting test of the dose-response relationship that was so neatly demonstrated in another study concerning immune response and psychological stress
is a direct measure of subjective rating of the event
(Cohen et a!, 1991).
by the respondent;
The balance of evidence of previous work is against a positive association between negative life events and
an objective,
but personally
meaningful, measure is required for aetiological research (Brown & Harris,
1978).
Another important point is that mood was not measured. If the period of weeks following the diagnosis of Graves' disease is associated with depression, this alone would explain the greater negative scores for life events. Depression would be a plausible intervening variable between life stress and onset of an autoimmune disease, as there is evidence that depression, as well as or instead of negative life events, is closely related to alteration of
the immune response (Irwin et a!, 1990). Unfortu nately, the authors did not measure depression, even though it may be present in one-third of subjects with hyperthyroidism (Kathol et a!, 1986).
The failure of the self-administered questionnaire to prompt accurate recall is seen in the ‘¿fall-off' of
events with increasing time from the interview. During the 3 months furthest from interview (i.e. 9—12months before onset), the number of people with a negative events score was half that recorded for the 3 months closest to interview (1—4months
before onset). This cannot be attributed to an excess of negative events just before the diagnosis
of
Graves' disease, as an identical reduction was demonstrated for the control group. Unfortunately, details of the events are not given, though these sometimes allow a fuller picture and a
the onset of Graves' disease (Jadresic, 1990). Only one paper approached a satisfactory method (Gray & Hoffenberg, 1985) and used sufficient patients—the result was negative. Evidence from other physical illness is conflicting. In the majority of physical ill nesses studied with the LEDS, there is no clear link with severely threatening life events, though the exceptions are stroke, where a weak but positive effect has been reported (House et a!, 1990) and multiple sclerosis (MS), where the pattern of events before onset of first symptoms was similar to that found in depression (Grant et a!, 1989). The onset of MS is also related to immunological change, but there is also a problem in defining onset with this
disorder. This is an important area of research, which is gradually increasing our knowledge of the multifactorial causation of disease, but only when objective measures are used (Creed, 1992). T. S. Brugha Winsa et a! claim that their study has shown that “¿stress can be an important factor in the cause of [Graves'] disease―. Before considering this claim more critically, it may be worth pondering for a moment that if they had produced a negative fmding, we might have been more surprised. As Thoits (1991)
540
HARRIS ET AL
has pointed out: ‘¿ ‘¿stress theory can be criticised for
review process of a high-quality general psychiatry or psychology journal, the readers of which would not need to be reminded of the considerable advances in the reliability of life-event measurement that have of physical illness and of poor mental health―. followed on the development of investigator-based Accordingly, the study under consideration here can measures such as the LEDS. be criticised for its failure, apriori, to give particular To take just one objection a little further: there is consideration to the possible theoretical link between good evidence that memory is differentially affected its lack of specificity with respect to outcome; the same theoretical factors (stress exposure, lack of coping resources) are used to explain the occurrence
adversity and Graves' disease, and, in order to do this, to consider what aspects of these two factors might be particularly related to one another. Turning first to the aetiological side of the
by mood and, in particular, that negative events or
equation, Thoits (1991) has argued, according to her identity theory of stress research, that identity
could explain the increased reporting in such cases
irrelevant stressors should influence physical health directly by means of general physiological arousal, whereas identity-relevant stressors should influence psychological symptoms directly. For example, the loss of a highly valued social role or attribute might increase the risk of depression. Thus, the finding by Winsa et a! that the total amount of stress does not
play a part in Graves' disease, but that life events that are self-rated as negative in their effect are significantly associated with it, is contrary to the model proposed by Thoits. Their finding would be consistent instead with a psychological disorder, perhaps most saliently with depression. In fairness to these authors, much research on stressful life events within the field of psychological and psychiatric disorder does not incorporate an attempt to link specific types of events in people with specific types of vulnerability to specific forms of psychiatric disorder. Although the idea is clinically appealing and theoretically more substantial, in practice, attempts to do so have encountered considerable methodological difficulties (e.g. Finlay-Jones &
Brown, 1981). Turning next to the illness side of the equation, what particular issues arise? Graves' disease presents an additional difficulty for stress researchers because of the coexistence in the clinical presentation of psychological symptoms. According to Lishman (1978), “¿psychological disturbance in some degree is universal with thyroid overactivity―; both anxiety and less often depressive features have been clearly described. Thus, any attempt to study the relation ship between stress and coping on the one hand and Graves' disease on the other would require that psychiatric comorbidity be taken into account. These authors appear to have overlooked this, possibly to
the detriment of their hope that we would be wiffing to interpret their findings as they do. Winsa et a! accept that “¿recall bias is a serious concern―. Their choice of a self-completed and self-rated measure of adversity would almost certainly not have passed the
memoriesare better recalledin the presenceof a negative affect (Clark, 1983). The presence of such affects in the clinical manifestation
of thyrotoxicosis
of negative events in comparison with healthy, though otherwise appropriately matched controls.
The argument by Winsa eta! that there was probably no recall bias in their own study, because cases and controls claimed to have equal numbers of reliable supporters and satisfaction with these social relation ships, does not answer the objection, since their measure of social support
(Sarason
et a!, 1983) is a
state measure that does not rely on memory to the same degree that retrospective life-event measures do. However, this negative finding may also have some additional value, although it may be of no direct interest to the aims of their study. A frequent objection to the fmding that deficits in social support
are associated with the presence of psychiatric disorders, such as depression, is that patients view their social world negatively (Henderson eta!, 1981). A physical disorder, such as Graves' disease, in which clinically significant psychological disturbance occurs
more than in the general population, would provide one opportunity to test the strength of this objection. Unfortunately, no data on psychiatric comorbidity were gathered by Winsa et a!, although had their study included such additional information, it could
have been of value to those concerned with exploring this issue. One can only conclude, therefore, that any (secondary) psychological morbidity present in the cases reported in this study was unlikely to have significantly biased their self-reports of personal relationships and their supportiveness. Winsa et a! point out that a previous prospective study, in which the control group consisted of patients with non-toxic goitre, suggested that stress was not a cause. Prospective research on the effects of life events is also problematical (Brown eta!, 1986),and perhaps almost impossible if the investigator is concerned with predicting the first episode of a condition that is rare and otherwise unpredictable. However, there is considerable scope for improving the quality of cross-sectional and retrospective research. The present study was commendable for the method of sampling
used, which was epidemiologically based. Future
LIFE EVENTS AND GRAVES'
research in this area should measure coexisting psychological disturbance, take into account any bias due to it, and make use of methods of measuring life events that have already been shown to be capable of dealing with the problems of recall. In the meantime, the safest answer to the question of whether stress causes Graves' disease is: perhaps.
541
DISEASE
Gii@y, J. & HOFFENBERG, R. (1985) Thyrotoxicosis and stress. Quarterly Journal ofMedicine, New Series,54, 153—160. HARRIS, 1. 0. (1989) Disorders
of menstruation.
In Life Events
and Illness (eds 0. W. Brown & 1. 0. Harris). New York: Guilford Press. & BROWN,0. W. (1989) The LEDS findings in the context of other research: An overview.
In Life Events and
illness(eds 0. W. Brown & 1. 0. Harris). New York: Guilford
Press. HENDERSON, S., BYRNE, D. 0. & DUNCAN-JONES, P. (1981) Neurosis
and The Social Environment. Sydney: Academic Press. HOUSE,
References BROWN, 0.
W.
&
HARRIS,
1.
0.
(1978)
Social
Origins
of
Depression.London:Tavistock. Ar@inanws, B., HARRIS,T., et a! (1986) Social support, self-esteem and depression. Psychological Medicine, 16,813-831. BwuLco, A. T. & HARRIS,T. 0. (1987) Life events, vulnerability, and onset of depression. British Journal of Psychiatry, 150, 30—42. CALLOWAY,P. & DOLAN,R. (1989) Endocrine change and clinical
profiles in depression. In Life Events and Illness(eds G. W. Brown & T. 0. Harris). New York: Guilford Press. Ci@sx,
D (1983) Differential
Effects
of Mood
on the Accessibility
of Positive and Negative Information. Unpublished DPhil Thesis, University of Oxford. COHEN,S., TYRELL,D. A. J. & SMITh, A. P. (1991) Psychological
stress and susceptibility to the common cold. New England Journal of Medicine, 325, 606—612. CoolcE, D. J. (1986) Inferring causality in life event research. Stress
Medicine, 2, 141—152.
Cit@io,T. K. J. (1989)Abdominalpain. In LifeEventsand Illness (eds 0. W. Brown & T. 0. Harris). New York: Guilford Press. CREED,F. H. (1992) Life events and disorder. Current Opinion in
Psychiatry, 5, 300—304. DAY, R. (1989) Schizophrenia. In Life Events and illness (eds 0. W.
Brown&1.0. Harris),pp. 113-137.NewYork:GuilfordPress. DEMBROWSKI, 1. M. MACDOUGALL,J. M. & SHIELDS,J. L. (1977) Physiologic reactions to social challenge in persons evidencing
the type A coronary-pronebehavior pattern. Journal of Human Stress, 3, 2—10. DOHRENWEND, B. P.,
LEVAV, I., SHEoUT,
A.,
DENNIS,
M.,
MOGRIDGE,
L.,
et
al
(1990)
Life
events
and difficulties preceding stroke. Journal of Neurology, Neurosurgery and Psychiatry, 53, 1024—1028.
P. E., et a! (1987)
Life
stress and psychopathology: progress on research begun with Barbara Snell Dohrenwend. American Journal of Community Psychology, 15, 677—715. FIr@aAY-JoHEs,R. & BROWN,0. W. (1981) Types of stressful life
event and the onset of anxiety and depressive disorders. PsychologicalMedicine,11, 803—816. FRIEDMAN,M. & ROSENMAN,R. H. (1959) Association
of specific
overtbehaviorpattern withbloodand cardiovascularfindings.
IRWIN,M., PA1-IERSON,T., SMITH,1. L., et a! (1990) Reduction
of immune function in life stress and depression. Biological Psychiatry. 27, 22—30. JADRESIC, D. P. (1990) Psychiatric
aspects of hyperthyroidism.
Journal of Psychosomatic Research, 34, 603—615. KATHOL,R. 0., TURNER,R. & Dati@nuwr, J. (1986) Depression and anxiety
associated
with hyperthyroidism:
therapy. Psychosomatics, KENDLER,
K.
S.,
SILBERG,
response
to antithyroid
27, 501—505. J.
L.,
NEALE,
M.
C.,
et
a! (1991)
The
family history method. Whose psychiatric history is measured? American Journal of Psychiatry, 148, 1501-1504. LISHMAN,
W.
A.
(1978)
Organic
Psychiatry.
Oxford:
Blackwell.
LEVINE,H. M., BASHAM,R. B., et a! (1983) Assessing social support: the social support questionnaire. Journal of
Personalityand SocialPsychology,44, 127—139. [email protected],S., VINGERHOETS, A. & VANHECK,0. (1987) The study of stress and disease: some developments and requirements. Social Science and Medicine, 25, 567—578. MArrHEWS, K. A., GLASS, D. C., ROSENMAN,R. H., et a! (1977)
Competitive drive, pattern A and coronary heart disease: a further analysis of some data from the Western Collaborative Study. Journal of Chronic Diseases, 30, 489—498. PAYKEL,E. 5. (1983) Methodological
aspects of life events research.
Journal of Psychosomatic Research, 27, 341—352. SARASON,I., JOHNSON,J. H. & SIEGEL,J. M. (1978) Assessing the
impact of life changes: Development of the Life Experiences Survey. Journal of Consulting and Clinical Psychology, 46, 932—946. SARASON,B. R. & PIERCE, 0. R. (1992) Three contexts of
social support. In The Meaning and Measurement of Social Support (eds H. 0. F. Veiel & U. Baumann). New York: Hemisphere. Su@nE1t, M. 1. (1974)Presidential address. Engagement-involvement:
a centralphenomenonin psychophysiological research.Psycho somatic Medicine, 36, 1—17. THOIrs, P. A. (1991) On merging
identity
theory
and stress
research.SocialPsychologyQuarterly,54, 1—112.
Journalof theAmericanMedicalAssociation,169, 1280—1296. WEINER,H. (1977) Psychobiology and Human Disease. New York:
GRANr,I., MCDONALD, W. I. PATrERSON, 1., et a! (1989)Multiple sclerosis. In Life Events and illness (eds 0. W. Brown & 1. 0. Harris), pp. 295—311.New York: Guilford Press.
@
Elsevier. WINSA, B., ADAMI, H.-0.,
BERGSTROM, R., eta! (1991) Stressful
life eventsand Graves' disease.Lancet, 338, 1475—1479.
Tirril Harris, Research Officer, Department of Social Policy and Social Science, Royal Holloway and Bedford New College, University of London, 11 Bedford Square, London WCJB 3RA; Francis Creed, MD, FRCP, FRCPsych, Professor, Department of Psychiatry, Rawnsley Building, Manchester Royal Infirmary, Manchester M13 9WL; Traolach S. Brugha MD, MRCPsych, Senior Lecturer, Department of Psychiatry, University of Leicester, Leicester LE2 7LX
