| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From the Office of Analysis (B.S.J.), Epidemiology and Health Promotion, National Center for Health Statistics, US Centers for Disease Control and Prevention, Hyattsville, MD; and Epidemiology Program Office (J.F.L.), US Centers for Disease Control and Prevention, Atlanta, GA.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
METHODS: A population-based cohort of 3310 initially normotensive and chronic disease–free persons in the NHANES I Epidemiologic Follow-up Study was tracked through four follow-up waves (maximum, 22 years). The association between hypertension and baseline negative affect was analyzed using Cox proportional hazards regression, adjusting for baseline age, sex, race, education, smoking, alcohol use, diastolic and systolic blood pressure, body mass index, and change in body mass index as a time-dependent covariate. Negative affect was based on combined symptoms of depression and anxiety. Hypertension end points included 1) self-reported, 2) treated (prescription of antihypertensive medications), and 3) incident (blood pressure 
160/95 mm Hg or treated) hypertension. Blood pressure measurements were obtained only at baseline and the first follow-up examination (maximum, 13 years).
RESULTS: Increased negative affect was associated with elevated risk for self-reported, treated, and incident hypertension at first follow-up. Through four waves of follow-up, high negative affect was associated with treated hypertension in baseline risk–adjusted models for white women (relative risk [RR] = 1.73, 95% confidence interval [95% CI] = 1.30–2.30), black women (RR = 3.12, 95% CI = 1.24–7.88), and all men (RR = 1.56, 95% CI = 1.08–2.25). Time-dependent covariate models produced similar RRs.
CONCLUSIONS: Negative affect is predictive of development of hypertension. For treated hypertension, white women and all men with increased negative affect had similarly elevated RRs, whereas black women with increased negative affect had substantially higher RRs.
Key Words: negative affect • anxiety • depression • hypertension • longitudinal • incidence
Abbreviations: CI = confidence interval; BMI = body mass index; DBP = diastolic blood pressure; GWB-A = General Well-Being Schedule, Relaxed vs. Anxious scale; GWB-D = General Well-Being Schedule, Cheerful vs. Depressed scale; NHANES I = first National Health and Nutrition Examination Survey; NHEFS = NHANES I Epidemiologic Follow-up Study; RR = relative risk; SBP = systolic blood pressure.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
In comparison to the number of cross-sectional studies, there are fewer prospective studies of negative affect on the subsequent development of hypertension. These studies, summarized elsewhere (1, 7, 20–25), show inconsistent results due, in part, to lack of adequate sample size, use of unstandardized psychological measures, and/or insufficient follow-up interval. Among these, even fewer (1, 20, 22, 23) studied this relationship among women and generally included only white women. Goldberg et al. (20) did not show an effect of anxiety or depression on hypertension in white women, whereas in an earlier study, Markovitz et al. (22) showed an effect in white and black women. However, the Markovitz et al. study included only 33 black women out of a total of 468 women, and both the Markovitz et al. and Goldberg et al. studies had a relatively short follow-up interval. A later study, also by Markovitz et al. (23), of white men and white women that had a longer follow-up interval found an effect in middle-aged white men only. Thus, results of studies of the effect in white women are inconsistent, and there is virtually no information available for black women. A study by Jonas et al. (1), which had a large sample size and long follow-up interval and used standardized measures of anxiety and depression, found an effect in both whites and blacks but did not further stratify by gender because of insufficient accumulation of hypertension cases. Therefore, it is important to further study the putative association of negative affect, as measured by anxiety and/or depression symptomatology, with subsequent hypertension development stratified by both race and sex.
A number of studies have demonstrated an increased prevalence of negative affect in women as compared with men (26) and that black women have higher rates of negative affect than do white women (27). Cornoni-Huntley et al. (28) found higher incidence rates of hypertension in black women as compared with white men, whereas the incidence rates for white women were roughly comparable to those of white men. Klungel et al. (29) found that women with hypertension were 33% more likely to be treated pharmacologically for hypertension than were hypertensive men.
We examined the role of negative affect, as measured by standardized assessments of anxiety and/or depression symptomatology, in the subsequent development of hypertension in the NHEFS. This analysis is larger than most previous investigations, is based on a national probability sample of the United States, has a follow-up period of up to 22 years, and, as a result, has sufficient accumulated cases of hypertension to examine the impact of race for women. The analyses could not be stratified by race for men because of insufficient accumulation of hypertension cases for black men. We hypothesized that after adjusting for possible confounders, negative affect symptomatology in white women, black women, and all men would be associated with an increased risk of hypertension.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
NHEFS is a longitudinal study of the 14,407 participants in NHANES I who were 25 to 74 years of age at the time of their survey examination from 1971 to 1975. Of those 14,407 participants, 6913 received a more detailed series of questions that included measures of anxiety and depression. Personal interviews and results of physical and laboratory examinations from NHANES I provided the baseline data for NHEFS. Four waves of follow-up were conducted during 1982–1984, 1986, 1987, and 1992 (33–35). Information was collected from interviews, healthcare facility records, and death certificates for decedents.
This analysis included white and black persons in NHANES I who were 25 to 64 years old at baseline and who received the more detailed series of questions. Of the 5606 white and black persons eligible for study through the four follow-up waves, we excluded 1433 persons with hypertension at baseline, defined as SBP 160 mm Hg, DBP 95 mm Hg, or past or current prescription of antihypertensive medications. Jonas et al. (1), using NHANES I data, found lower self-reported psychological well-being in normotensive subjects who had ever been told by a physician that they had hypertension or high blood pressure. Therefore, we also excluded these 362 respondents from further analyses. An additional 198 persons were excluded because of a baseline diagnosis of diabetes, stroke, or coronary heart disease. In addition, 303 persons were excluded because of incomplete data on other covariates. After all exclusions, data for 3310 persons were available for analysis.
Psychological Measures
We examined negative affect by combining two four-item scales from the General Well-Being Schedule (36) to measure symptoms of depression and anxiety at baseline. The Relaxed vs. Anxious scale (GWB-A) assesses nervousness, stress, strain, pressure, anxiety, worry, and tension. The Cheerful vs. Depressed scale (GWB-D) assesses unhappiness, sadness, discouragement, hopelessness, and lack of cheerfulness. Scores for each scale range from 0 to 25, with lower scores indicating greater anxiety or depression and higher scores indicating greater relaxation or cheerfulness.
The GWB-A and GWB-D measure self-reported symptoms of depression and anxiety and are not synonymous with clinical diagnoses of depression and anxiety. However, the GWB-A and GWB-D are highly correlated with clinically trained interviewers’ ratings of anxiety and depression, respectively. The GWB-A is also highly correlated with other instruments designed to measure anxiety, including the Minnesota Multiphasic Personality Inventory and the Psychiatric Symptoms Scale (36), and the GWB-D is highly correlated with other instruments designed to diagnose clinical depression, including the Zung Depression Scale and the Psychiatric Symptoms Scale (36). In the detailed sample of NHANES I, the computed internal consistencies of the GWB-A and GWB-D scales were 0.78 and 0.77 (37). Furthermore, Costa et al. (38) demonstrated that level of negative affect is relatively stable over time in a 9-year longitudinal study of 4942 adults (initial age, 25–74 years) using data from NHEFS. Given the size and representativeness of the sample, this is strong evidence of the stability of mean levels of negative affect in adulthood.
Negative affect was used to investigate the association between hypertension because of the high correlation (r = 0.75) between the GWB-A and GWB-D scales. In these analyses, the GWB-A and GWB-D scales were trichotomized as follows: 1) high symptomatology, scores of 0 to 12; 2) intermediate symptomatology, scores of 13 to 18; and 3) low symptomatology, scores of 19 to 25. The high symptomatology group has been shown to give a good approximation of high depression symptomatology (39). The intermediate and low symptomatology groups were chosen so that approximately 50% of the study sample was in the low anxiety symptomatology group. High negative affect was defined as high symptomatology on either the GWB-A, GWB-D, or both. Intermediate negative affect was defined as intermediate symptomatology on either the GWB-A, GWB-D, or both. Low negative affect was therefore defined as low symptomatology on both the GWB-A and GWB-D.
Hypertension Assessment
Baseline.
At the beginning of the baseline physical examination, the physician measured blood pressure with the examinee seated using a mercury sphygmomanometer or an aneroid sphygmomanometer. For DBP, the level was recorded at the level of complete cessation of Korotkoff’s sounds or, if there was no cessation, at the point of muffling. Measurements were recorded to the nearest 2 mm on the scale. Blood pressure was measured in the right arm with the bell of the stethoscope used for auscultation. Blood pressure measurements using the aneroid instruments did not differ significantly from those obtained with mercury instruments (40). Respondents were asked if they had ever been told by a physician that they had hypertension or high blood pressure.
Follow-up.
At each of the four follow-up interviews, respondents were asked if they had ever been told by a physician that they had hypertension or high blood pressure (reported hypertension). Respondents were then asked if a physician had ever prescribed medication for hypertension or high blood pressure (treated hypertension). Blood pressure measurements were performed only at the 1982–1984 follow-up, which limited identification of all incident cases (incident hypertension) to this follow-up wave. Thus, to compare these three definitions, reported, treated, and incident hypertension end points were constructed at the 1982–1984 follow-up wave. These analyses were based on all persons because there were insufficient cases to stratify by race and sex.
Persons who reported at any of the four follow-up interviews that a physician had prescribed antihypertensive medications form an additional group with the end point of treated hypertension; for this end point, incident cases were accumulated through the 1992 follow-up. The analyses were stratified by race and sex into three groups: 1) white women, 2) black women, and 3) white and black men combined (due to insufficient accumulation of hypertension cases for black men).
Baseline Variables
Smoking status was obtained at baseline for the subsample that received the detailed examination. For persons whose smoking status was not obtained, status at baseline either was derived from questions on the 1982–84 follow-up interview on lifetime smoking history or was imputed (41, 42). The validity of this approach has been documented (41, 42). The baseline medical history questionnaire and supplement included questions about selected conditions diagnosed by a physician (30, 32), including diabetes, stroke, or coronary heart disease. Alcohol use at baseline was obtained by asking the respondent "How often do you drink?" Other baseline variables were measured as described elsewhere (30, 32, 33).
Statistical Methods
Incidence rates of treated hypertension through 1992 were calculated as the number of incident cases per 1000 person-years at risk. Estimates of the risk of hypertension for persons with high or intermediate negative affect symptomatology relative to those with low symptomatology were derived from Cox proportional hazards models computed using the PHREG procedure of SAS software (43). All proportional hazards models included age at baseline in years, educational level (less than high school graduate, high school graduate or more), cigarette smoking (never or former smoker, current smoker), alcohol use (every day and just about every day, less frequent usage), baseline DBP and SBP (mm Hg), BMI (kg/m2), and change in BMI.
Change in BMI was based on measured weight in the 1982–1984 follow-up examination. Change in BMI was based on self-reported weight in the 1986, 1987, and 1992 follow-up interviews. Thus, the time-dependent covariate models that compare the three hypertension end points in 1982–1984 use change in BMI, which is based on measured weight change from baseline to this first follow-up. However, the time-dependent covariate models examining treated hypertension through the 1992 follow-up are based on measured weight at baseline compared with reported weight for the last three follow-up interviews. Weight data from the last available follow-up interview were used for persons who did not have treated hypertension. For persons with treated hypertension, weight data at the interview corresponding to their first report of prescription medications were used.
The mean length of follow-up for the treated hypertension group through 1992 was 17 years (range, 7–22 years). The incident date was estimated as the date of the follow-up interview at which antihypertensive medication prescription was first reported. Persons who did not report antihypertensive medication prescription were censored at the date of their last interview. The interview date was used as a conservative estimate of the incident date because no information was available about when medications were first prescribed. For decedents whose proxy interview indicated the prescription of antihypertensive medication, the date of death was used as the incident date.
To assess the effects of the complex survey design on the results, the Cox models were also analyzed using the "proc survival" procedure in the SUDAAN statistical package (44, 45). SUDAAN uses a Taylor series approximation method to compute variances. The NHANES I sample weights for detailed sample locations 1–100 and revised strata and primary sampling unit codes were used (46). The estimates derived from the unweighted analyses were generally consistent with those from analyses incorporating the complex survey design. Therefore, the estimates from the unweighted Cox regression models are presented.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
The association between negative affect and hypertension was highlighted by a consistent set of significant RRs in the various statistical models. The overall models, which compared levels of negative affect, different hypertension end points, and baseline vs. time-dependent covariates, also demonstrated general consistency in RR effect sizes. The RR effect size for high negative affect was the largest of all categorical variables entered in the models. The continuous variables, which included DBP, SBP, BMI, and change in BMI, were all significant predictors of subsequent hypertension development. To put negative affect in context, the RR effect size for high negative affect was roughly comparable to a 10-point difference in baseline DBP or to a 10-point change in BMI over time.
Negative affect was based on combined symptoms of anxiety and depression. Baseline anxiety and depression, as measured by the GWB-A and GWB-D scales, were highly correlated. It was not possible to assess the independent effects of anxiety and depression on the development of hypertension in this analysis because the number of respondents with high anxiety and low depression or the reverse was too small. Several authors have noted significant neurobiological (47), psychological (48, 49), and psychometric (50, 51) overlap among anxiety and depression. Thus, we conclude that negative affect, as measured by combined symptoms of anxiety or depression, may increase the risk of hypertension. More research is needed to determine whether there are independent effects of anxiety and depression on the development of hypertension.
The pathways through which negative affect symptoms could increase the risk of subsequent hypertension remain to be determined. Musselman et al. (52) suggest several possible mechanisms. For example, the increased plasma norepinephrine seen in both depression and hypertension may be causal. Also, increased arousal and mobilization of energy stores seen in anxiety and depression may impact the development of hypertension by making the heart more reactive. Alterations of serotonin-mediated platelet activation described in affective disorders, most notably depression, may be independently associated with hypertension. The present study did not have data to examine any of these hypotheses. Additional research is needed to examine these and other potential pathways.
An indirect pathway through which negative affect may operate is that elevated levels may increase risk behaviors, which may, in turn, increase the risk of hypertension. In these analyses, adjustment for a variety of baseline risk factors did not eliminate the increased risk associated with negative affect symptomatology. Black women also have higher prevalence of excess weight than white women (53). However, baseline BMI as well as changes in BMI over time were included as potential confounders in this study. Although they are both significant predictors of subsequent hypertension, they also did not eliminate the increased risk associated with negative affect symptomatology. There may also be other indirect mechanisms that explain the pattern of higher risk ratios among black women in comparison with white women or men. Krieger and Sidney (54) found that black women had higher blood pressure if they experienced discrimination in more than two of seven specified situations. High blood pressure is also more common in people of lower socioeconomic status, which is disproportionately higher in black women (55). Marmot et al. (56) reported that men and women of lower employment grade had higher rates of coronary heart disease, which they postulate is due in part to their concomitant lack of autonomy. It is therefore possible that the levels of negative affect observed in our study for black women may coexist with discrimination and/or resultant lack of autonomy associated with lower socioeconomic status. These external stressors may act synergistically, directly causing autonomic hyperresponsiveness, leading to higher risk ratios for hypertension.
Treated hypertension for analyses through 1992 was chosen by necessity because blood pressure measurements were conducted only at the 1982–1984 follow-up. However, use of this end point may have several advantages. Because high negative affect persons may be more likely to have blood pressure reactivity than other persons (6–8, 18), the use of treated hypertension as an end point may reduce the potential bias of false-positive diagnoses of hypertension based on blood pressure readings obtained at a single examination. Treated hypertension may also represent a more conservative choice with respect to reported hypertension. Of the 477 persons with reported hypertension at the 1982–1984 follow-up, 359 (75.3%) were treated hypertension cases. Nevertheless, persons with high negative affect may more often ask for and be prescribed antihypertensive medications. Cases of treated hypertension for high, intermediate, and low levels of negative affect constituted 70.9%, 80.9%, and 71.9% of cases of reported hypertension, respectively. Therefore, when reported and treated hypertension were compared with respect to levels of negative affect, no consistent pattern was found.
Conditions such as hypertension are generally underreported in household interview surveys. One could speculate that respondents with elevated levels of negative affect are less likely to underreport conditions than they actually have or are more likely to overreport conditions that they do not have. Specific to this analysis, they may overreport that a physician told them they had high blood pressure or hypertension or that they were prescribed antihypertensive medications. Although such ascertainment bias cannot be directly assessed in this data, we did several analyses to examine this issue indirectly. NHEFS contained several medical conditions, namely, kidney disorder or kidney stone, urinary tract disorder, and cataracts, that have no theoretical association with negative affect. A generalized ascertainment bias would be expected to operate even on neutral medical conditions such as these. Thus, we would expect to find increased RRs of these conditions to be associated with high levels of negative affect. Models were constructed for these three conditions, and no consistent pattern was detected. On the basis of these indirect analyses, we concluded that ascertainment bias was not a likely explanation for the associations observed.
Even though we found no consistent evidence that increased negative affect was increasing the likelihood of receiving hypertension treatment or that a generalized ascertainment bias was operating, the lack of direct control over the issue of ascertainment bias in this report must be regarded as a limitation of these analyses. Another potential limitation concerns the use of measured weight at the 1982–1984 follow-up vs. self-reported weight in successive follow-ups through 1992 to construct change in BMI scores. Rowland (57) found that even though measured and self-reported weights were reported on average with small errors, the unreliability in self-reporting increased directly with the magnitude of the person’s overweight status. The results of our time-dependent covariate models both at the 1982–1984 follow-up and through 1992 indicate that change in BMI has little impact on the RRs of negative affect associated with hypertension. Because change in BMI is based on measured weight at both baseline and the 1982–1984 follow-up, we consider these results reasonable at this follow-up. However, the use of self-reported weight at subsequent follow-up interviews limits our ability to generalize the results of these time-dependent covariate models through 1992. Nevertheless, because effect sizes of baseline covariate models compared with baseline and time-dependent covariate models at the 1982–1984 follow-up are virtually the same, we must conclude that change in BMI has limited impact as a pathway through which negative affect leads to the development of hypertension.
In summary, the present study provides evidence of an association between negative affect symptomatology and subsequent onset of hypertension among white women, black women, and all men aged 25 to 64 years. Among persons with treated hypertension tracked through 1992, black women with high or intermediate negative affect were at substantially higher risk. Elevated negative affect symptomatology was found in all subgroups of the selected risk factors studied. Thus, negative affect is a risk factor for all persons, not just those who are at elevated risk because of their status on other, more established risk factors. Both intermediate and high levels of negative affect symptomology generally revealed a risk gradient for hypertension. Dimsdale (2) states that subsyndromal symptomatology can have significant health consequences. The high prevalence of intermediate negative affect coupled with consistently elevated RRs suggests that this group may have substantial hypertension risk attributable to subsyndromal symptomatology. It remains unclear how the association between negative affect and hypertension can be explained and by which direct and/or indirect pathways. However, regardless of which pathways can be confirmed, because of the high prevalence of both conditions, the relationship between negative affect and hypertension is of considerable public health importance. Additional studies are needed to elucidate the pathways for the effects of negative affect on hypertension incidence among women and men in general and black women in particular.
Received for publication October 27, 1998.
Revision received August 10, 1999.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  B. Hildrum, A. Mykletun, J. Holmen, and A. A. Dahl Effect of anxiety and depression on blood pressure: 11-year longitudinal population study The British Journal of Psychiatry, August 1, 2008; 193(2): 108 - 113. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. H. Miller and C. L. Raison Immune System Contributions to the Pathophysiology of Depression Focus, January 1, 2008; 6(1): 36 - 45. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Pieper, J. F. Brosschot, R. van der Leeden, and J. F. Thayer Cardiac Effects of Momentary Assessed Worry Episodes and Stressful Events Psychosom Med, November 1, 2007; 69(9): 901 - 909. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. C. Thurston and L. D. Kubzansky Multiple Sources of Psychosocial Disadvantage and Risk of Coronary Heart Disease Psychosom Med, October 1, 2007; 69(8): 748 - 755. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Dorn, C. J. Yzermans, H. Guijt, and J. van der Zee Disaster-related Stress as a Prospective Risk Factor for Hypertension in Parents of Adolescent Fire Victims Am. J. Epidemiol., February 15, 2007; 165(4): 410 - 417. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. V. Ostir, I. M. Berges, K. S. Markides, and K. J. Ottenbacher Hypertension in older adults and the role of positive emotions. Psychosom Med, September 1, 2006; 68(5): 727 - 733. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. L. Yan, K. Liu, K. A. Matthews, M. L. Daviglus, T. F. Ferguson, and C. I. Kiefe Psychosocial Factors and Risk of Hypertension: The Coronary Artery Risk Development in Young Adults (CARDIA) Study JAMA, October 22, 2003; 290(16): 2138 - 2148. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T. Rutledge and B. E. Hogan A Quantitative Review of Prospective Evidence Linking Psychological Factors With Hypertension Development Psychosom Med, September 1, 2002; 64(5): 758 - 766. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. C. Light Hypertension and the Reactivity Hypothesis: The Next Generation Psychosom Med, September 1, 2001; 63(5): 744 - 746. [Full Text] [PDF]
|
|
|
|
 |
|
 S. Levenstein, M. W. Smith, and G. A. Kaplan Psychosocial Predictors of Hypertension in Men and Women Arch Intern Med, May 28, 2001; 161(10): 1341 - 1346. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
