Acute stress promotes transient elevation of blood pressure, but there is no consistent evidence that this effect results in hypertension. In this systematic review of cohort and case–control studies that investigated the association between psychosocial stress and hypertension, we conducted a complete search up to February 2007 in MEDLINE, EMBASE, PSYCINFO and LILACS, through a search strategy that included eight terms to describe the exposure, six related to the design of the studies and one term for outcome. The quality was assessed by the Newcastle–Ottawa Quality Assessment Scale. The selection was done in duplicate by two teams of independent reviewers. Among 82 studies selected in the second phase, only 14 (10 cohort studies and 4 case–control studies), totalling 52 049 individuals, fulfilled the selection criteria. The average quality of the studies was 6.6±1.3 in a 9-point scale. Acute life events were associated with hypertension in one and were not associated in two studies. Five out of seven studies found a significant and positive association between measures of chronic stress and hypertension, with risk ratios ranging from 0.8 to 11.1. Three out of five studies reported high and significant risks of affective response to stress for hypertension, one a significant risk close to a unit and one reported absence of risk. Acute stress is probably not a risk factor for hypertension. Chronic stress and particularly the non-adaptive response to stress are more likely causes of sustained elevation of blood pressure. Studies with better quality are warranted.
The interaction between a genetic background and environmental and behavioural exposures, such as the excess of salt, fat and alcohol consumption, accounts for most but not all cases of hypertension. Stress has long been listed as a potential and important cause of hypertension1 among other potential risk factors such as low potassium consumption, low physical activity and sleep abnormalities.
Acute stress can induce transient elevation of blood pressure, but it is still unclear whether this effect results in sustained elevation of blood pressure and hypertension.2 The results of the INTERHEART study, which identified an independent risk of stress for myocardial infarction,3 have renewed the interest on stress as a risk factor for cardiovascular diseases.
Stress has been defined as a process in which environmental demands exceed the adaptive capacity of an organism, resulting in psychological and biological changes that may place persons at risk for disease.4 Both stressors and the adaptive response of individuals (the transaction between stressor and response) have been defined in different ways in experimental and epidemiological studies. In addition, epidemiological studies have different designs, follow-up times, sample criteria and variable definitions of hypertension. To our knowledge, there is no systematic review or meta-analysis of cohort and case–control studies focusing on stress as a risk factor for hypertension, probably because of the large variability of exposure and outcomes used in the original studies. There is only one systematic review of cohort studies showing that anxiety, depression, and anger control and social support were associated with the incidence of hypertension.5 The variability between studies precluded the undertaking of a meta-analysis. Therefore, we summarized the evidence about the association between stress and hypertension in observational studies through a systematic review, combining studies by design and type of stressors and adaptive response.
Criteria for study inclusion
We restricted our analysis to cohort and case–control studies. In view of the paucity of longitudinal studies, case–control studies were required to increase the number of useful studies. Despite their higher propensity for bias, case–control studies assume that exposition occurred before the event. Cross-sectional studies were not included, as they are still more heterogeneous in terms of measures of stress and hypertension than the cohort and case–control studies. Further, causality cannot be inferred and there is currently no established method for the systematic review of such studies. The outcome of the studies should be the diagnosis of hypertension as a binary variable. Cases should be participants with hypertension and control individuals without such diagnosis. Participants were from both genders and at least 18 years of age. Cohort studies with participants with hypertension at the baseline evaluation were excluded.
Studies that measured imposed physical–mental ‘loads’ or ensuing pathophysiological states6 were included in the process of study selection. Experimental studies, demographic studies and studies of personality and coping (set of strategies used by individuals to adapt themselves to adverse or stressful events) were not included. Studies that included participants with psychiatric diagnoses, such as anxiety and depression, were also excluded.
We conducted a complete search up to February 2007 in MEDLINE, EMBASE, PSYCINFO and LILACS, through search strategy and terms (full-text and keywords) presented in the panel. The search terms were selected from the MeSH Database, selecting 5 of 95 terms related to stress, and also terms suggested by MeSH (‘see also’). Additionally, we searched the references of the articles selected in the databases. We did not apply any restriction in regard to the language of the publication and did not use any strategy of search for studies that were not published.
Extraction of data
In phase I, the titles and abstracts resultant of the search strategy was evaluated to select studies with potential to fulfil the inclusion criteria. Most studies were excluded at this step, as they were not clearly a cohort or case–control studies or they provided other evidence of inadequacy with regard to the selection criteria. In phase II, the selected articles were evaluated in their full-text format to decide whether they fulfilled the criteria of eligibility.
The following information was extracted from the studies that fulfilled the eligibility criteria: first author, country and year of publication, age composition, sample size, criteria of definition of exposure and outcome, duration of follow-up (cohort), number of events according to the factor in study, variables controlled in the multivariate analysis and effect measures (relative risk, odds ratio (OR) or hazards ratio). In view of the variable definition of exposure both in terms of stressors and adaptive response, we aggregated the studies by design, acute and chronic nature of exposure to stressors, and by studies that evaluated the affective response. Studies of affective response were defined on the basis of use of any questionnaire oriented to the perception of stress. The studies used quite variable instruments, but mostly rating scales. Hypertension was defined by the authors of the studies, and when more than one criterion was used we chose blood pressure 140/90 mm Hg or use of blood pressure drugs.
Appraisal of the quality of studies
We used the Newcastle–Ottawa Scale to assess the quality of studies.7 A score of quality was calculated on the basis of three major components: selection of the groups of study (0–4 points), quality of the adjustment for confounding (0–2 points) and ascertainment of the exposure or outcome of interest in the case–control or cohorts, respectively (0–3 points). The maximum score would be 9 points, representing the highest methodological quality (Table 1). Adjustment for baseline blood pressure values, duration of follow-up of at least 5 years and losses that did not exceed 10% for cohort studies, and adjustment for age and non-response rates below 10% in case–control studies were criteria of higher quality. In addition, we employed funnel plots to examine the possibility of publication bias separately for the different definitions of stressors and adaptive responses.
Extraction of data and appraisal of the quality of studies were carried out in duplicate by two teams of independent reviewers (GW and FPF, FTC and AMA), who were previously trained and certified. The first author resolved the divergences between the reviewers.
The initial search in abstracts identified 593 articles that were potentially eligible. The large number of studies selected in the first phase was explained by the ease of primary selection, as stress and hypertension are very common terms. Eighty-two studies fulfilled the criteria of eligibility for the first phase and were fully evaluated. Seventy-three studies were excluded in the second phase, mostly because they did not have exposures or outcomes defined as required for this review (Figure 1). The list of these articles may be requested from the correspondent author. On the basis of the list of references of the nine studies, it was possible to identify five others that fulfilled the criteria of eligibility and were therefore included in the analysis.
Description of the studies
Ten studies were cohorts and four case–controls, totalling 52 049 individuals. The study of Radi et al.8 reported two different forms of exposures and was analysed separately. Details of studies are presented in Table 2.
The largest cohort, with 30 330 black women and 2316 cases of hypertension, included more participants than the sum of the remaining studies.21 Cohort studies totalled 50 471 individuals and identified 7800 cases of incident hypertension (15.5%). Nine different definitions of outcome were employed and only three8, 10, 16 used the most contemporaneous definition of hypertension (blood pressure 140/90 mm Hg or use of blood pressure-lowering agents). All studies were adjusted for age and for a variable set of confounders, including body mass index in nine and previous blood pressure levels and education in five. The mean age of participants at entry also varied, but most studies were carried out with young or middle-age adults. The study by Levenstein et al.15 included only adults, and the study by Sparrow et al.19 did not present a measure of central tendency. The study by Markovitz et al.18 supplied information on the racial distribution of the sample, which included 45.2% of black individuals, and the study by Cozier et al. was restricted to black women. The duration of follow-up of cohorts varied from 2.5 to 20 years.
Methodological quality of the studies
The quality of the studies varied from 3 to 8 points, with a mean of 6.6±1.3 and a median of 7 (Table 1). The studies with higher scores10, 16 did not get a score of 9 because their samples were not representative of communities. Nakanishi et al.16 derived their cohort from the workers of only one company, and Perez et al.10 informed that their cases had originally come from the health services of Yarumal, Colombia, but did not inform whether they were representative of the individuals who needed medical care. The study with the lower score was exactly the largest one.21 The authors derived their sample from subscribers of a magazine; exposures and outcome were self-reported, and the adjustment was made only for reported age and body mass index, besides having a short time of follow-up and high proportion of losses (26%). The funnel plot for acute and chronic studies was unequivocally symmetric, suggesting that publication bias was unlikely (Figure 2a). For studies of affective response, the funnel was not so evident, but there were fewer studies (Figure 2b).
Association between different definitions of stress and hypertension
The studies were grouped by design (cohort or case–control study) and definition of stress (acute, chronic or affective response to stress). All studies of chronic stress were occupational. The case–control study by Radi et al.8 included life events and occupational stress. When more than one risk ratio was presented, the higher and statistically significant estimate was chosen. The risk ratios for hypertension and the corresponding confidence intervals, when available, are presented in Figure 3.
The three studies, two cohorts and one case–control, which assessed the association between acute stress and hypertension reported null, inverse or positive association. Perini et al.12 evaluated a list of 14 events as a continuous variable. Radi et al.8 defined exposure by the report of any among a list of 14 events. Dorn et al.13 measured exposure to a disaster (being the parent of an adolescent involved in a mass burn incident) and found a direct association. After this episode, the victims’ parents were forced to deal with a great number of stressful experiences in their function as caregivers, such as physical disabilities and emotional scars.
Four studies8, 9, 18, 17 used the Job Content Questionnaire,22 an instrument to measure occupational stress, focusing on the forms of work organization. Fauvel et al.17 reported an inverse association and the remaining studies identified a statistically significant association between stress and hypertension. Three studies used singular items of exposure. Kahn et al.14 investigated a list of 90 exposures, 30 being items of current chronic stressors. Only two of these, both occupational items, were associated with hypertension. Levenstein et al.15 tested only two indicators of stress on the job, describing a risk for ‘worried about keeping the job’ and null association for ‘average or not good at doing the job’. Nakanishi et al.16 reported an inverse relation between working hours and hypertension. Four studies exclusively studied employees. Two of them8, 9 had selected workers of diverse companies and the other two, those of a single company.16, 17
Three out of the five studies included10, 11, 20 demonstrated a statistically significant association between stress, measured by different scales, and hypertension. Everson et al.,20 El-Shafei et al.11 and Perez et al.10 identified risk ratios for hypertension between 1.6 and 33.7 for hopelessness, severe stress and ‘tension-anxiety’. Sparrow et al.19 derived five measures from stress from the Medical Cornell Index and did not identify an association between stress and hypertension. In the same way, Cozier et al.21 adapted one instrument of perceived racism, composed for eight items, and found a few risk ratios for hypertension but very close to null.
This qualitative systematic review tried to include the aggregate of all observational evidences with a cohort and case–control design that had evaluated the association between stress and hypertension. The number of studies selected in the first phase of the searching strategy was impressive, but just a few studies fulfilled the criteria of selection. Even for this small number of studies, the quality and number of individuals investigated were not satisfactory; for instance, the largest study21 had the lowest score of quality. Moreover, stressors and adaptive responses were quite variable in the studies. It was therefore impossible to summarize the evidence for the association between a unique and well-defined characterization of exposure to stressors and the subsequent adaptive response and the incidence of hypertension. We then aggregated the studies by design and main exposures, looking at the trends for risk in each situation. Overall, only when the studies explored the affective response to stressors was there a consistent risk for hypertension.
The association between acute stress and hypertension was evaluated in only three studies,8, 12, 13 which had inconsistent results. The case–control study showed an inverse association in men and women,8 but the probability of a recall bias cannot be ruled out. One of the cohort studies showed a null association12 and the other a positive association.13 From the studies of chronic stress, five reported at least one significant risk for hypertension among several measures of stress,8, 9, 14, 15, 18 and two were negative.16, 17 The positive studies had a mean average score of quality of 6.6, in comparison with a score of 7.5 for the two studies with negative findings. The distinction of acute and chronic stress was in some way arbitrary, as some acute events have long-lasting repercussion. An example is the study by Dorn et al.,13 who studied parents of adolescent fire victims. This acute stressful event definitely has a long-lasting duration. On the other hand, studies classified primarily by the presence of chronic stressors had repetitive and recurrent exposition to stressors, particularly on the job.
From the studies that used the Job Content Questionnaire,8, 9, 18, 17 only Fauvel et al.17 did not find a positive association. In this study, the cumulative incidence of hypertension seems to be even higher among those not exposed to stress. A limitation of all studies, but not of the study of Markovitz et al.,18 is that their authors assumed that exposure had not varied along the follow-up. Markovitz et al.12 applied the questionnaire at the start and end of the follow-up and found an OR of 2.06 (confidence interval (CI) 95% 1.01–4.26) for the increase in the ratio job demand/decision latitude. Another potential limitation of the cohort studies was the losses in follow-up, which varied from 0 to 68%. Studies that presented original analysis had fewer losses,12, 13, 14, 16, 17 which were higher in re-analysis, mostly because of incomplete data.8, 15, 18, 19, 20 There was no trend for an association between the proportion of losses and the estimates of risk. Moreover, there is no evident reason to suppose that the losses had any systematic direction.
Studies of affective response seem to be more appropriate to investigate the association between stress events and hypertension. They integrate both the stressor events, acute or chronic, the response to them and characteristics of the individuals, social support and other known and unknown mechanisms that could link stress to hypertension.24 A large number of instruments to measure the affective response have been presented in the literature. From the five studies that addressed this exposure and fulfilled the criteria of inclusion in this systematic review, one19 did not find an association between the affective response to stress and hypertension, and three10, 11, 20 reported positive association and a trend for a dose–response effect. The ORs were 1.27 (95% CI 0.79–2.07),16 2.85 (95% CI 1.59–5.11)14 and 8.70 (95% CI 2.61–28.86)20 for moderate stress and OR 3.22 (95% CI 1.56–6.67),16 5.02 (95% CI 2.25–11.19)14 and 10.10 (95% CI 3.02–33.72)20 for high stress. The study by Cozier et al.21 had a low score of quality and explored the association of a large number of exposures with self-reported hypertension.
These findings are in accordance with the results of laboratory experiments. It has been suggested that individuals who presented a more intense haemodynamic response to stressful tasks have a familial predisposition to hypertension.25 Impaired cardiovascular recovery following stress seems to predict the risk of developing hypertension.26 High cardiovascular reactivity to stress would depend on an interaction between genetic factors (familial history) and environment (frequency of exposure to stressors).27 The response to exposures could be influenced by former experiences.28 These sympathetic-induced haemodynamic responses to stressful events29 could lead to endothelial dysfunction and proliferation of smooth muscular cells,30 perpetuating the hypertensive state. The investigation of the link between stress events, affective response and hypertension would require further studies in terms of frequency and magnitude of exposure and duration of the cardiovascular recovery after stress. Beyond the direct physiopathological consequences of stress, it could be associated with unhealthy habits such as dietary habits, excessive alcoholic beverages consumption and low levels of physical activity.29 Clinical trials demonstrating that measures to control stress are effective in lowering blood pressure would be helpful in confirming the link between stress and hypertension. One protocol for a systematic review of five clinical trials was registered with the Cochrane Library, focusing on the effect of relaxation in the treatment of patients with essential hypertension.31
Anxiety and depression could be part of the maladaptive response to stressors, and therefore would be in the chain of causation of hypertension and other illnesses. These conditions are psychiatric conditions that require formal criteria for diagnosis, and could occur independently of identifiable triggering factors, such as environmental stressors. A quantitative review identified that psychological factors could be associate with hypertension,5 but the isolate contribution of each component has been questioned.32
To our knowledge, this is the first systematic review addressing the association between stress and hypertension in cohort and case–control studies. We had planned to undertake a meta-analysis, but the heterogeneity between the studies, with regard to design, definition of exposure and outcome and control for confounding, excluded this possibility. The search for studies was conducted in four different databases and references of the studies. The use of methodological filters of observational studies increased the specificity of title and abstract resultant of the search strategy. The use of two teams of independent reviewers probably enhanced the quality of study selection and evaluation. The main limitation of our systematic review was the absence of an attempt to identify unpublished investigations, which could have biased our findings as a consequence of a publication bias. The small number of studies that fulfilled our eligibility criteria, however, suggests that it is unlikely that a properly conducted study did not report its findings.
In conclusion, the few cohort and case–control studies addressing the association between stress and the incidence of hypertension used variable definitions of stressors and the adaptive response to them. In addition, some of these studies had low methodological quality. With the aggregation of studies by exposure and response, it is possible to conclude that acute stressful events have no consistent association with incident hypertension. Chronic stress and particularly the non-adaptive response to stress are more likely causes of sustained elevation of blood pressure.
This study was funded by grants from CNPq, CAPES and FAPERGS.
About this article
Nature Reviews Cardiology (2017)