Article Text

Download PDFPDF

Research paper
Is psycho-physical stress a risk factor for stroke? A case-control study
  1. Jose Antonio Egido1,
  2. Olga Castillo1,
  3. Beatriz Roig1,
  4. Isabel Sanz1,
  5. Maria Rosa Herrero1,
  6. Maria Teresa Garay1,
  7. Ana María Garcia1,
  8. Manuel Fuentes2,
  9. Cristina Fernandez2
  1. 1Stroke Unit, Department of Neurology, Hospital Clinico Universitario San Carlos, Universidad Complutense de Madrid, Instituto de Investigación Sanitaria del Hospital Clínico San Carlos (IdISSC), Madrid, Spain
  2. 2Unidad de Gestión Clínica Servicio de Medicina Preventiva, Unidad de Metodología de Investigación y Epidemiología Clínica, Hospital Clínico San Carlos, Escuela de Enfermería, Universidad Complutense de Madrid, Universidad Camilo Jose Cela. Instituto de Investigación Sanitaria del Hospital Clínico San Carlos (IdISSC), Madrid, Spain
  1. Correspondence to Dr Jose Antonio Egido, Stroke Unit coordinator, Department of Neurology, Hospital Clinico Universitario San Carlos, Avda/ Martin Lagos, s/n. 28040 Madrid, Spain; jegidoh{at}


Background Chronic stress is associated with cardiovascular diseases, but the link with stroke has not been well established. Stress is influenced by life-style habits, personality type and anxiety levels. We sought to evaluate psycho-physical stress as a risk factor for stroke, while assessing gender influences.

Methods Case-control study. Cases: patients (n=150) aged 18–65, admitted consecutively to our Stroke Unit with the diagnosis of incident stroke. Controls: (n=300) neighbours (paired with case ±5 years) recruited from the census registry. Study variables: socio-demographic characteristics, vascular risk factors, psychophysical scales of H&R (Holmes & Rahe questionnaire of life events), ERCTA (Recall Scale of Type A Behaviour), SF12 (QoL scale), GHQ28 (General Health Questionnaire). Statistical analyses included conditional multiple logistic regression models.

Results Mean age was 53.8 years (SD: 9.3). Compared with controls, and following adjustment for confounding variables, significant associations between stroke and stress were: H&R values >150 OR=3.84 (95% CI 1.91 to 7.70, p<0.001); ERCTA (values >24) OR=2.23 (95% CI 1.19 to 4.18, p=0.012); mental SF12 (values >50) OR=0.73 (95% CI 0.39 to 1.37, p=0.330); psychological SF12 (values >50) OR=0.66 (95% CI 0.33 to 1.30, p=0.229), male gender OR=9.33 (95% CI 4.53 to 19.22, p<0.001), high consumption of energy-providing beverages OR=2.63 (95% CI 1.30 to 5.31, p=0.007), current smoker OR=2.08 (95% CI 1.01 to 4.27, p=0.046), ex-smoker OR=2.35 (95% CI 1.07 to 5.12, p=0.032), cardiac arrhythmia OR=3.18 (95% CI 1.19 to 8.51, p=0.022) and Epworth scale (≥9) OR=2.83 (95% CI 1.03 to 7.78, p=0.044).

Conclusions Compared with healthy age-matched individuals, stressful habits and type A behaviour are associated with high risk of stroke. This association is not modified by gender.

  • Stress
  • stroke
  • cerebrovascular disease
  • neuroepidemiology
  • quality of life
  • subarachnoid haemorrhage, ultrasound

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Stroke is one of the principal causes of morbido-mortality worldwide. It is a disease involving considerable disability, with high social and economic repercussions.1 Control of the vascular risk factors plays a fundamental role in the fight against this disease.2–4

Chronic psycho-physical stress is considered the psychological or physical response to stressors persisting >6 months. The patient can be considered as suffering from an adjustment disorder if clinically significant emotional or behavioural symptoms result.5 ,6

Psycho-physical stress can produce neuro-vegetative effects that predispose to psychosomatic diseases.7 ,8 Several studies highlight stress as an independent risk factor in cardiovascular diseases9–13 but there is a dearth of in-depth studies evaluating the psycho-physical bases of stress and stroke.14–16

Psychophysical stress derives from situations that modulate factors such as personality type, quality-of-life (QoL), levels of anxiety or depression and other environmental factors including social- and work-status, family responsibilities, and cultural level.17 These factors have not been fully assessed, to date.

The objective of the present study was to evaluate, in a multimodal standardised approach, the association of psychophysical stress on stroke in a population <65 years of age living within the catchment area of a tertiary level university hospital in the Community of Madrid. Gender differences in risk profiles and stress were explored.

Materials and methods

Study design

The study was a case-control design (ratio of 1:2) paired for age (±5 years), conducted between January 2008 and December 2010 in a working-age population in Madrid.

Case recruitment

Cases were patients consecutively admitted to the Stroke Unit of the Hospital Clínico San Carlos with a primary diagnosis of incident stroke, and aged between 18 and 65 years. Ischaemic aetiologies were classified by the Stroke Unit's neurologists according to modified Trial of Org 10 172 in Acute Stroke Treatment (TOAST) criteria following a detailed evaluation that included an ECG, echocardiography, carotid duplex and neuroimaging with CT and/or MRI. The patients were assessed clinically for competence to answer the study questionnaire. If unable, the help of a family member (or carer) was sought. A pilot study indicated a low concordance between the responses from family member (or carer) and those provided by the patient.18 Hence, any such patient incapacity was an exclusion criterion.

Control recruitment

Control individuals were randomly selected from the patient's local district using census information. We did not pair for gender because one objective of the study was to assess the influence of gender. Control individuals who reported a personal history of stroke were excluded from the study. If a selected control did not respond, refused to participate or was ineligible because of prior stroke, a second or a third age-matched control living in the neighbourhood was invited.

The control individuals were identified by an external clinical research organisation using a random door-to-door search for individuals fulfilling the inclusion criteria. The clinical research organisation was blinded with respect to gender and clinical data. A ‘manual of operations’ was used to standardise data collection.

Study variables

Variables related with psychophysical stress were based on the combined quantitative scores of the 4 scales used:

  • Questionnaire of life events of H&R consisting of 40 items corresponding to stressful experiences lived over the previous year. A score >300 constitutes the high-risk group with an 80% probability of suffering an illness in the near future. An intermediate-risk of 50% is a score between 150 and 300, the low-risk of 30% corresponds to a score <150.19

  • General Health Questionnaire validated in Spain evaluates psychosocial problems in the previous 4 weeks. Consisting of 28 items with four sub-scales, the questionnaire evaluates somatic symptoms of psychological origins such as distress, anxiety, social dysfunction and depression. Values >8 on this scale are considered pathological.20

  • QoL scale (SF-12) provides a profile of general health status with reference to the previous month. Consisting of 12 items derived from eight dimensions of the SF36 questionnaire, it assesses physical function, social function, physical role, emotional role, mental health, vitality, body pain, general health. These dimensions are summed as mental and physical components. A low QoL is defined as a value lower than the median scores of the control group of individuals.21

  • The type A behaviour pattern of the ERCTA (Recall Scale of Type A Behaviour) which has been validated in Spanish, assesses the presence of type A behaviour pattern. Consisting of eight items, the total score varies between 8 and 35, with a score of 24 tending towards a pattern indicative of type A behaviour.22

The clinical variables recorded were diabetes (DM), hypertension (HT), hypercholesterolaemia, history of alterations in cardiac rhythm and atrial fibrillation. A subject was considered diabetic if clinically diagnosed as DM or currently receiving treatment for DM. HT was considered if clinically diagnosed or currently receiving medication for high blood pressure. Hypercholesterolaemia was considered when clinically diagnosed with hypercholesterolaemia or currently receiving hypolipidaemic medication. The history of angina/infarction and history of alterations in cardiac rhythm such as atrial fibrillation was considered when there was evidence of a clinical diagnosis for each of these pathologies. Alcohol consumption was classified in four groups (as a function of the quartiles of consumption of the control group): 0 g alcohol/day; from 0.1 to 3.5 g/day; from 3.5 to 14.4 g/day; and >14.4 g/day.23 Tobacco consumption was established according to the WHO as: non-smoker, ex-smoker and current smoker. This last group includes habitual smokers and occasional smokers.24 Obstructive sleep apnoea syndrome was evaluated on the Epworth scale which measures the diurnal somnolence that reflects the poor quality of night-time sleep. Scores up to 13 indicate light somnolence, up to 19 indicate moderate and >24 indicate severe somnolence.25 The socio-demographic variables included age, gender, civil status, education level, number of children, tobacco consumption, beverages or relaxation-inducing substances (coffee, tea, energy drinks containing cola-caffeine and taurine derivatives, sleeping tablets, anti-depressants, recreational drugs). The intake frequencies were: never, <1/day, between 1 and 2/day, >2/day.

The questionnaires were self-administered. Socio-demographic data and clinical history were recorded by the interviewer using the same standardised format in cases and controls. The interviews were conducted in the hospital during the first week following the stroke. Control individuals were assessed in a face-to-face interview.

Quality control and ethical issues

The agency identifying the control individuals and members of the investigation team underwent quality control testing to guarantee the validity of the information.

The study was approved by the ethics committee of the study centre. Written informed consent was obtained from all the participants prior to inclusion into the study. Data were codified so as to maintain patient and control individuals' anonymity.

Statistical analyses

Sample size was calculated for a confidence level α error of 5%, a power of 80%, an estimated proportion of exposure of controls of 32.4%,26 an OR of 1.8, and with a case-control ratio of 1:2. The sample size calculation indicated 150 cases and 300 controls.

Qualitative variables are expressed with their frequency distributions and quantitative variables as means and SD.

Univariate conditional logistic regression analyses were applied to assess relationships between categorical variables and stroke. The magnitude of association was evaluated using the OR and 95%CI. Interactions between gender and each of the scales used in the study were evaluated while introducing an interaction term into the model. An adjusted logistic conditional regression model was applied to each measurement scale of psychophysical stress.

Adjustment was with those variables which, in the univariate analyses, showed a level of statistical significance of p<0.05, and/or were considered clinically relevant. The overall model was re-applied with only those confounding factors being retained that, following their elimination from the model, produced a change in the OR of the variable of principal exposure (psychosocial scale) of ≤10%. Finally, an adjusted model was applied introducing the scales used in the study that were significant in the univariate analyses, and with adjustment for the confounding factors identified in the earlier models. Significance was set at p=0.05. Each model was applied only to those subjects with complete data on each of the dependent variables, and scores relating to all items in the questionnaire (N=371).

To assess multiple co-linearity, the κ concordance index was calculated for the different candidate confounding factors introduced into the model, and for the different psycho-physical stress scales. The degree of concordance between the independent variables in the model were low (minimum −0.070 and maximum 0.302). The same was observed in evaluating the concordance between the scales of psychophysical stress (minimum −0.340 and maximum 0.340). Tolerance and the incremental factor of variance were calculated for each one of the independent variables in each of the models constructed. The minimum value obtained for tolerance was 0.6563 and the maximum for factor of variance was 1.52. The coefficients between each model (-2 log likelihood) and the residual degree of freedom (N- number of independent variables introduced) were calculated, and none were statistically significant. The data were processed with the STATA V.9.0 (STATA Corp LP, Texas, USA) statistical package.


Consecutive case recruitment continued until the number of cases meeting the study entry requirements (n=150) was achieved. Of the 223 cases evaluated in this process, 73 were excluded because of: comprehension deficit (n=22, 30.1%), previous stroke (n=19, 26%), refusal to participate (n=11, 15.1%), not domiciled in Madrid (n=10, 13.7%), cognitive deterioration (n=5, 6.8%), psychiatric problems (n=3, 4.1%), death (n=3, 4.1%). The mean age of the patients was 53.8 years (±9.3) and that of the control group was 53.6 years (±9.6).

The socio-demographic and clinical characteristics of the group of patients included in the study were compared with those who were excluded. Statistically significant differences were found only in the distribution of the civil status variable (singles included =36 (24%) vs 26 (40%) in the excluded group; p=0.022) and the severity of stroke event (19 (24%) TACI of those included vs 27 (40.9%) of the 73 who had been excluded (p<0.001)).

Table 1 summarises the classification of the cases based on the Oxfordshire Classification, the aetiology of the stroke (TOAST criteria), and the severity scores measured by the Canadian scale and on the modified Rankin scale.

Table 1

Clinical characteristics of the cases

Table 2 summarises the socio-demographic characteristics of the study sample with respect to life-style habits, alcohol and tobacco consumption, risk factors for stroke, and the overall scores on the 4 scales used, together with their corresponding cut-off points.

Table 2

Sociodemographic and lifestyle characteristics of the cases and controls

In the univariate analysis, the demographic and life-style characteristics that were associated with stroke were: civil status, currently employed, consumption of coffee, tea and energy drinks, tobacco, alcohol, diagnosis of HT, hypercholesterolaemia, angina and a value on the Epworth scale >9.

From our database we selected pairs of cases and controls who were recorded as being employed at the time of the study. The re-analysis showed that the significance was maintained between gender and the diagnosis of stroke (OR=5.4; 95% CI 2.5 to 11.7; p<0.001).

The stress evaluation scores showed stronger differences between cases and controls on the H&R and the ERCTA scales. Using clinical criteria, 12 of the 19 patients (63.2%) with TACI had scores ≥150 and 7 (36.8%) had <150 (p= 0.032) on the H&R scale.

Figure 1 depicts the results, stratified by gender, of the relationship of each of the scales with stroke. No statistically significant gender modification was observed in any of the evaluation scales.

Figure 1

Relationship of the psycho-physical stress and quality-of-life scales with ictus between males and females. GHQ, general health questionnaire.

Multivariate analyses

An adjusted explanatory conditional logistic regression model was applied for each of the scales used in measuring psychosocial stress. The following adjustment variables were introduced into the model: gender, civil status, current work, frequency of consumption of energy drinks (> twice a day), tobacco consumption, consumption of alcohol (>14.4 g/day), DM, HT, hypercholesterolaemia, clinically diagnosed angina and/or myocardial infarction, previously diagnosed alterations in cardiac rhythm, and Epworth scale (≥9). Figure 2 summarises the results of the multivariate analyses for each of the scales used. A score of ≥150 in the H&R scale and a score of >24 in the ERCTA scales is significantly related to the presence of stroke. The patients with scores >50 in the psychophysical SF12 scale had a significantly lower frequency of stroke. Following adjustment for potential confounding factors, the score on the GHQ28 scale was not related to stroke.

Figure 2

Multiple conditional regression models: stress scales effect on stroke. Logarithm scale: OR 95% CI. N=371. *adjusted for: gender, energy drinks intake, Epworth scale score. **adjusted for: gender, energy drinks intake, smoke category, hypertension, arrhythmia, Epworth scale score.***adjusted for: gender, energy drinks intake, smoke category, arrhythmia, Epworth scale score. GHQ, general health questionnaire.

Finally, the conditional logistic regression model was adjusted by introducing the psychosocial variables of stress (except the GHQ28) adjusted for the potential confounding factors identified in the previous analysis. The factors that were independently related to stroke were: a score ≥150 on the H&R scale, a score of >24 on the ERCTA scale, masculine gender, consumption of energy drinks more than twice a day, type of smoker, alterations in cardiac rhythm, and a score ≥9 on the Epworth scale (table 3).

Table 3

Multivariate analysis: Relationship of each assessment scales and lifestyle habits with stroke (N=371)

Behavioural factors and stress

The relationships between the consumption of tobacco, alcohol and physical activity versus ERCTA and the H&R scales with respect to cases and controls were assessed. We did not observe, in the univariate analysis, any of these factors being significant in relation to the levels of stress in cases evaluated by these scales. Further, modification of the effect of reduction in the effect of these three factors on stress (between cases and controls) following the introduction of the interaction term into the models of logistic regression indicated no statistically significant differences.


Study results and literature perspective

The grade of stress under which an individual lives is influenced by several socio-cultural factors.27–30 In the present study various dimensions of validated scales were applied, and the results are coherent in terms of relationship of stroke with stress. Other studies have evaluated stress with a single-item questionnaire.31 The INTERSTROKE study,32 which is the largest case-control study dealing with known and emerging risks factor across different countries and races, showed an association between psychosocial stress (and depression) with stroke. However, it only applied an isolated combined measure of general stress in the home and in the workplace over the previous year. Conversely, we used a multimodal approach because psycho-physical stress derives from different situations that modulate such factors as personality type, QoL, anxiety levels, and other environmental factors. Further, in the INTERSTROKE study, for those patients unable to communicate sufficiently to complete the study questionnaire, proxy respondents were used. As we had found (see Methods, above) and as has been previously reported,18 this approach is not valid in assessing responses to psychological questions.

Individuals having lived under stressful conditions in the previous year (H&R scale) were, following adjustment, 3.8-fold more likely to suffer a stroke compared to controls. Other authors have also proposed that psychosocial stress derived from stressful life events increase the risk of cerebral infarct in hypertensive individuals.28

Patterns of behaviour can reflect the capacity to adapt to a stressful life. We found that individuals with high levels of competitiveness and aggression (ERCTA scale >24) are, following adjustment, 2.2-fold more likely to suffer a stroke compared with controls.

For the GHQ28 psychosocial scale, those persons who presented signs of depression had 22% higher likelihood of having a stroke (albeit this did not reach statistical significance). The Caerphilly study15 concluded that middle-aged men with symptoms of psychological distress have a threefold higher likelihood of dying from a stroke. The level of distress and the depression symptoms associated with stress have somatic repercussions such as HT30 and are also associated with poor life-style choices such as low physical activity, tobacco habit, alcoholism and poor dietary habits.33 However, we did not observe that any of the behavioural factors were related to stress in our cases and controls. Over the past year, the prevalence of mental disturbance in Spain has been estimated as 8.48% of the adult population.34 The lack of statistical significance with respect to the GHQ28 scale may be related to the high grade of stress in our control population, which was greater than expected and would require a larger sample size for a more detailed investigation.

In the multivariate analysis of values derived from each scale, in relation to generic aspects of QoL, the individuals who had a better QoL (SF12 >50) had, following adjustment, half the likelihood of suffering a stroke compared to controls. As such, we observed a protector effect of QoL reflected in this scale. Poor socio-cultural conditions of the population, such as unfavourable housing conditions together with a low socio-cultural level, can generate high stress levels which play a significant role in the aetiology of cardiovascular diseases.27 ,35 Conversely, stress has been shown to be associated with material advantage, and which showed a spurious protector effect.36 In our study sample, following adjustment for the stress scales, the statistical significance was lost in both these components of QoL assessment. This effect is produced, principally, by the H&E scale. In our study, in order to minimise the effect of environmental factors, a study sample selection criterion was that the control individuals lived in the same census district as the cases. As such, factors such as education level, social conditions including civil status, number of children, family load, and influence of environment are very similar.

With respect to work status, the number of individuals being actively employed was much greater in cases than in controls, but this variable lost its statistical significance in multivariate analysis. We recorded active employment but, unlike a previous report,37 we did not evaluate the quality nor the grade of stress in the specific employment environment.

Study limitations

To evaluate the bias implied in non-response, we compared the characteristics of the individuals included in the study with those who had been excluded. We observed that there were no statistically significant differences in demographic characteristics between the two groups, except for civil status and extent of the stroke. To achieve the highest level of response possible, all subjects in the study were individually contacted. However, we were not able to exclude completely all those with specific psychological attitudes that could influence non-response.

The exclusion of those patients who, due to their severity of disease or aphasia, were not able to respond on their own to the questionnaire implies a bias in selection. Approximately 40% of those not included presented with extensive stroke, of whom, half had severe aphasia. Hence, these data cannot be extrapolated to this type of patient. However, other authors have studied this group of patients and observed an important association between stress and fatal stroke.14 In our study, half of the cases presented a severe disability, and in 14% of patients with clinical infarct we observed a higher grade of stress (as measured on the H&R scale). Hence, exclusion of more severe patients would tend to minimise the effect of stress.

Some of the risk factors associated with stroke were self-reported and this can produce a differential classification bias. The association of stress with stroke could be influenced by the stress due to the stroke itself, and this implies a bias in the differential classification, but only in the stroke group. However, the scales assess the patient with respect to the period before the stroke, depending on the design and validation of each scale. Conversely, the scales with significant results are the consequence of objective outcomes such as the H&R scale of stressful life19 and of the behaviour type pattern22 and, as such, are less influence by a possible bias in the patient's recall of events and conditions.

It is reasonable to expect that stroke diagnosis and its treatment can change the patients' psycho-physical status and life-style habits. The resulting bias can induce an artificial association in a cross-sectional setting. For example, stroke patients are more likely to quit smoking, which results in selection bias and an underestimation of smoking effect among current smokers together with an overestimation of effect among ex-smokers. In our study we had no means of excluding this possible bias.

Dealing with a hospitalised patient population carries its own limitation in relation to the external validity of the results. The stroke patient sample in our study had similar characteristics as any other hospitalised population with this diagnosis. These included distribution, type of stroke, risk factors and severity, a high proportion of ischaemic stroke of principally cardioembolic and indeterminate aetiology, and a small proportion of hemorrhagic stroke caused by HT.

We selected a limited age range in order to reduce the other confounding vascular disease risk factors and to assess the effect of stress due to employment. As such, our results cannot be extrapolated to other age groups.

Gender, stress and stroke

There were more women in the control population sample and reflect the data provided by Spanish National Institute of Statistics. Gender differences in employment rates were only around 10%.38 Hence, based on the results obtained in our population, gender does not appear to affect the relationship between stress and stroke.


Psycho-physical stress factors related to stressful life-style and type A personality are associated with stroke, independently of other risk factors and unhealthy life-style. We did not observe gender having a significant effect on these findings of psycho-physical stress and stroke. Addressing the influence of psycho-physical factors on stroke could constitute an additional therapeutic line in the primary prevention of stroke in the at-risk population and, as such, warrants further investigation.


View Abstract


  • Funding This study was funded, in part, by a grant from the Health Research Foundation [Fondo Investigacion Sanitaria; FIS PI7/0124] within the European Regional Development Fund [Fondo Europeo de Desarrollo Regional; FEDER].

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval Ethics approval was provided by ethics committee Hospital Clínico San Carlos.

  • Provenance and peer review Not commissioned; externally peer reviewed.