| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From the Departments of Psychology (L.S.P.) and Psychiatry and Behavioral Sciences (A.A.S., J.E.S.), State University of New York at Stony Brook, Stony Brook, NY.
Address reprint requests to: Laura S. Porter, PhD, Department of Psychology, CB 3270, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599-3270.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONPRESENT STUDYMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
METHODS: One hundred college students completed state and trait versions of the Spielberger (1) anger expression scale. State measures were completed in response to specific anger-provoking situations as they occurred over a 7-day period. Ambulatory blood pressure was recorded on one of these days.
RESULTS: Moderate correlations were observed between trait and state anger expression. Significant associations were found between a number of situational variables and state anger expression scales. Neither trait nor state anger expressions scales were related to blood pressure levels.
CONCLUSIONS: These results indicate that trait and state measures of anger expression are not equivalent and that situational factors play an important role in anger expression. Situational variability may be an important factor in determining the health consequences of anger expression.
Key Words: anger expression • ambulatory blood pressure • behavioral cardiology • situational variability
Abbreviations: ABP = ambulatory blood pressure; DBP = diastolic bloodpressure; SBP = systolic blood pressure; VAS = visualanalogue scale.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONPRESENT STUDYMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Numerous studies have tested this hypothesis, including prospective and cross-sectional studies of the relationship of anger expression to resting blood pressure (5–13), lipid concentration (14), the severity of angiographically documented coronary stenosis (15, 16), and coronary heart disease and all-cause mortality (17–21).
In general, the results can be characterized as inconsistent. Although a number of studies have found evidence suggesting that suppressed anger is indeed associated with poor health outcomes (eg, 5–8, 12, 15, 18), other studies have found the opposite, that the outward expression of anger leads to illnesses such as hypertension and coronary heart disease (eg, 11, 16, 19, 21). Still other studies have found no association between anger expression and health outcomes (eg, 13, 17, 20). In a meta-analysis of the literature pertaining to resting blood pressure, Suls et al. (4) concluded that high scores on the Spielberger "anger-in" scale (a trait measure of suppression) were associated with higher blood pressure levels, although only in studies assessing blood pressure with a single measurement, a method that tends to be unreliable (22). The "anger-out" scale (a trait measure of outward expression) showed an inverse relationship with SBP but not DBP levels.
One reason for these discrepant findings may be the conceptualization of anger expression as a stable personality trait rather than as a state that fluctuates according to situational characteristics or demands. The vast majority of research on anger expression has been conducted with the assumption that anger expression is a trait, that is, a stable personality disposition that transcends the influence of situational context or time. This assumption is held despite the fact that, in one of the earliest studies in this area, Funkenstein et al. (23) noted that anger expression does indeed vary according to situational factors.
Although the issue of situational variability has been acknowledged (7, 14), it has not been explored to any extent. This oversight is particularly salient in light of the increasing evidence for situational variability in other constructs that were traditionally considered traits. In the coping literature, for example, there has been increasing recognition that certain types of situations are reliably associated with particular coping strategies and that individuals vary their coping responses on the basis of the content of the problem with which they are faced (eg, Ref. 24). Anger expression is also likely to be associated with situational factors, such as the person who instigates the anger and the location in which the incident takes place. For example, the same individual who yells and slams doors during an argument with a family member at home may inwardly seethe at an unreasonable boss at work without saying anything. This research area would benefit from an examination of situational variability in anger expression to determine whether, and how, anger expression varies according to situational characteristics. It may be that frequent exposure to situational factors that are related to a particular mode of anger expression across subjects (ie, a "coronary-prone situation") is more predictive of adverse health outcomes than individual difference variables or a "coronary-prone personality" (17).
|
PRESENT STUDY |
|---|
|
TOPABSTRACTINTRODUCTIONPRESENT STUDYMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Second, we predicted there would be reliable associations between certain situational variables and state anger expression scales. For instance, situations that occur in public as opposed to private locations and in which the target is a superior rather than a peer or subordinate should be associated with more anger suppression and less expression. Who is present during the incident may also influence anger expression: The presence of the target may encourage anger expression, whereas the presence of others may inhibit expression. In addition, the importance of cognitive/psychological appraisal variables has been recognized in the coping literature (eg, 24, 27–29), and we suspected that variables such as anger intensity and perceived control would also be associated with anger expression.
Third, we examined associations between trait and state anger expression scales and ABP. We hypothesized that, should there be reliable associations between anger expression and blood pressure, they were more likely to be found with state measures of anger expression scales than with trait measures.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONPRESENT STUDYMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Measures
Demographic and health questionnaire.
Participants provided information about a number of health and demographic variables that have been shown to be associated with blood pressure levels, including age, sex, height, weight, and race. They also indicated whether they smoked and whether either of their parents was hypertensive.
Trait assessment of anger expression.
Participants completed self-report questionnaires concerning their general tendencies to experience and express anger, including the Spielberger anger expression scale (1). The anger expression scale is a 20-item measure that instructs participants to indicate how often they behave in a particular manner when they are feeling angry. Responses are made on a four-point scale from "almost never" to "almost always." The items reflect two relatively independent anger expression dimensions labeled anger-in and anger-out. Examples of items loading on the anger-in scale are "keep things in" and "angrier than willing to admit." Anger-out items include "say nasty things" and "argue with others." The anger expression scale was used because of its relatively wide use as a measure of anger expression as well as its demonstrated psychometric properties (1).
State assessments.
To assess state anger expression, participants were given booklets containing brief questionnaires that they were instructed to complete in the event of an anger-provoking situation. Seven days was chosen as an adequate time period over which to collect data on the basis of a similar study by Averill (31), in which participants prospectively reported a mean of 7.3 anger incidents over the course of a week. The questionnaire consisted of three sections: descriptive questions, anger expression items, and cognitive/psychological appraisal items. The descriptive and appraisal items were adapted from those used by Averill (31).
Participants were initially asked to provide a brief description of the incident that made them angry. They were then asked a series of questions about the nature of the incident: the intensity of their anger (measured with a 10-point VAS anchored at "very mild" and "very intense; as angry as most people ever become"); the target of the participant’s anger ("a person/an inanimate object/a group of people or an organization/society or ‘things in general’/other"); if the target was a person, the participant’s relationship with the target ("someone you know well and like/someone you know well and dislike/an acquaintance/a stranger") and the power relationship between the participant and the target ("someone with authority over you/someone over whom you have authority/an equal or peer"); the location of the incident ("public/private"); the presence of the target ("yes/no"); and the presence of people other than the target ("yes/no").
The second section contained a modified version of the anger expression scale in which participants were instructed to indicate whether they had engaged in certain behaviors during the specific incident being reported using a dichotomous (yes/no) scale. The dichotomous scale was chosen over the four-point scale to reduce the subject burden for the subject for the situation-specific reports. The content and number of items were identical to those of the original anger expression scale.
The final section of the questionnaire consisted of questions about the outcome of the situation, including whether the situation was resolved; if the situation was resolved, whether the participant was satisfied with the resolution; the participant’s perceived degree of control over the resolution (measured on a 10-point VAS anchored at "none" and "a lot"); and whether the participant felt that, everything considered, the incident was "beneficial/productive" or "harmful/nonproductive" (outcome).
Ambulatory blood pressure.
Blood pressure was measured using the SpaceLabs model 90207 ABP monitoring device. Evidence has been obtained for the validity of measurements collected with the SpaceLabs 90207 device (32, 33). At each blood pressure reading, participants completed a short questionnaire assessing their location, posture, and activities, variables that have been shown to affect ABP readings (34, 35).
Participants wore the monitoring device for 1 waking day (approximately 24 hours, removing the monitor while sleeping). All blood pressure monitoring took place on weekdays. The accuracy of the monitor was assured by checking that monitor readings were within 5 mm Hg of those obtained with a stethoscope and mercury machine in consecutive measurements from the same arm. Participants were fitted with the ambulatory device on their nondominant arm and were given verbal and written instructions about its operation and care. They were instructed to proceed with their usual daily activities and to keep their arm still whenever the cuff inflated. They were also instructed on how to terminate a reading in the event that the cuff became painfully tight or inflated when not on their arm. Several readings were then taken to test the ambulatory recording procedure. The monitoring device was programmed to automatically take blood pressure readings every 45 minutes over the course of the day.
Procedure
Participants completed the self-report questionnaires at the start of the study. For the following 7 days, they were asked to report on anger-provoking situations as they occurred during the course of their everyday lives. Reports were event-driven; that is, participants were given instructions to complete a questionnaire whenever they became angry. Participants were instructed to report both minor and intense experiences of anger, as opposed to annoyance (31), and examples of possible anger-arousing events were given (being cut off by another driver, receiving an unfair grade on an exam, or the landlord not shoveling a snowy walkway). This type of event-driven recording was chosen over retrospective reports to reduce biases associated with retrospective recall, which include active distortion as well as simple forgetting (36). Event-driven recording has been used successfully in other studies (eg, monitoring of smoking urges and lapses; 37, 38).
The disadvantage of the event-driven methodology is its susceptibility to reporting biases. Individuals may use different criteria concerning, for example, the intensity of anger that stimulates them to make a report. There may also be differences in subjects’ willingness to make reports given the additional subject burden this entails. This latter problem can be minimized by keeping questionnaires short and simple to complete, thereby reducing subject burden. An alternative method of collecting momentary data across a number of time points involves prompting subjects at random times to report on their current situation. Although this approach eliminates some of the problems associated with event-driven reports, the latter was chosen for this study because it is better able to capture relatively low-frequency events, such as the occurrence of anger (see Ref. 39 for a discussion of time-sampling procedures).
Data Analysis
The three hypotheses of this study required different analytic techniques depending on the nature of the data involved. To examine the first hypothesis, the degree of correspondence between multiple state anger expression scales and the trait scales, a nested analysis of variance procedure was used. This method provides a slightly more accurate estimate of the correlation between the trait and state measures than a simple correlation because it is implicitly based on each participant’s true mean on the state scales rather than on the observed sample mean (26). The second hypothesis, regarding the association between anger expression and situational variables, was addressed through the use of multilevel regression models, which explicitly take into account the two levels of sampling that occur in the data, within-person and between-person. This allowed us to address the question of whether anger expression was related to situation variables on an incident-by-incident basis while controlling for individual differences. For the third hypothesis, regarding the association between anger expression and blood pressure, the state anger expression scales were aggregated across each person’s reports and then a general linear regression model was used. The blood pressures were also aggregated for each person.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONPRESENT STUDYMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
|
State assessments.
Anger booklets were obtained from 97 participants (39 men and 58 women). On average, participants reported 4.1 (SD = 3.1) incidents of anger over the 7-day study period, with a range of 0 to 17. Scores on the Spielberger anger-in and anger-out scales were computed for each incident and then averaged across incidents for each participant. Because the state items were dichotomously scored, the possible range for each scale was 0 to 10. The mean situation-specific anger-in score was 2.9 (SD = 1.7), and the mean anger-out score was 2.5 (SD = 1.3). There were no significant gender differences on aggregated anger-in or anger-out scales.
ABP readings.
Ninety-eight participants completed ABP monitoring. One participant did not wear the blood pressure monitor, and another was unable to complete monitoring because of pain associated with the inflation of the blood pressure cuff. On average, 12.3 blood pressure readings were obtained per participant (SD = 4.1), with a range from 2 to 22.1Average SBP was 125.9 (SD = 9.1) for men and 116.6 (SD = 9.5) for women. Average DBP was 73.8 (SD = 6.9) for men and 72.4 (SD = 6.1) for women. Participants’ average ambulatory SBP and DBP readings were used as the dependent variables in the analyses that follow.2
To ensure that the days on which blood pressure was collected were generally representative of the entire week, we compared days with blood pressure monitoring to other study days in terms of the number of anger incidents reported and average anger intensity. Participants reported an average of 0.66 (SD = 0.53) incidents per day on days that they were not wearing the blood pressure monitor, with a mean anger intensity of 5.0 (SD = 1.8). They reported an average of 0.87 (SD = 1.4) incidents on the day they wore the blood pressure monitor, with a mean anger intensity of 4.8 (SD = 1.8). Paired t tests indicated that the difference in frequency was nonsignificant (t(94) = 1.58, p = .12), although it was somewhat suggestive of increased reporting on days on which blood pressure was monitored, and that the difference in anger intensity was not significant (t(33) = 0.45, p = .66). (Note: The difference in mean anger intensity was based on data from the 34 participants who reported incidents both with and without the blood pressure monitor.)
To further explore any possible influence of the blood pressure monitor on anger expression, we also compared participants’ mean anger expression in response to incidents that took place while the blood pressure monitor was being worn with those that took place when the monitor was not worn. Paired t tests indicated that there were no differences in anger-in (t(32) = 0.22, p = .82) or anger-out (t(32) = 0.39, p = .70).
Correspondence Between Trait and State Measures of Anger Expression
It was hypothesized that trait measures of anger expression would be moderately associated with state measures but that there would be enough discrepancy to indicate that the two methods of assessment are not interchangeable. To examine the degree of correspondence between trait and state anger expression scales, a nested analysis of variance was performed. In this method, individual incident anger-in and anger-out scores are associated with trait scores in between-subject correlations. The variables entered into the nested ANOVA are the anger-in and anger-out scores for each incident and the corresponding trait measures.
As predicted, the correlations between trait and state anger expression were in the moderate range: 0.37 for total anger expression, 0.45 for anger-in and 0.48 for anger-out. On average, trait measures of anger expression accounted for 19% of the variance in state measures.
Predicting State Anger Expression Scales From Situational Characteristics and Appraisals
The second hypothesis was that, across participants, there would be reliable associations between situational variables and anger expression scales. A multilevel random effects model, which treats both persons and reports as random sources of variance, was used for these analyses (40). This strategy is preferable to ordinary regression analyses because the latter ignores the fact that reports are nested within persons, which, in turn, can bias inferential tests and limit the generalizability of the results (41, 42). PROC MIXED in SAS (43) was used to construct linear repeated-measures models that explicitly take into account the two levels of sampling by specifying a compound symmetry error structure.
The dependent variables in these analyses were individual incident scores on anger-in and anger-out. There were two classes of situational predictor variables: (1) objective situational characteristics, which include the target of the participant’s anger, the participant’s relationship to the target, the power relationship between the participant and the target, the location in which the incident took place, and the presence of the target and presence of others; and (2) mood and appraisal variables, including whether the participant was still angry at the time of the report, anger intensity, whether the situation was resolved, the amount of control the person had over resolving the situation, and whether the outcome of the situation was beneficial/productive or harmful/unproductive.3
Analyses involving objective situational characteristics were performed twice. The first analysis included all incidents reported. Predictor variables for this analysis were target, location, presence of the target, and presence of others. The second analysis included only incidents in which the target of the participant’s anger was a specific person. In addition to the above predictor variables, this second analysis included variables assessing the participant’s relationship to the target and the power relationship between the participant and the target, variables that are relevant only when the target is a specific person.
In the analysis of all incidents, target, location, and the presence of the target all showed significant associations with anger expression scales (see Table 2). Target was a significant predictor of anger-in (p < .01): Compared with incidents in which anger was directed toward another person, when anger was directed toward an inanimate object, scores on anger-in were lower. Location was a significant predictor of anger-out (p < .0001), with higher scores on anger-out occurring in private as compared with public locations. The presence of the target was also associated with anger-out (p < .001). When the target was present, participants were more likely to use anger-out. The presence of people other than the target was not significantly related to either anger-in or anger-out.
|
|
|
There were two steps to each set of analyses (one set for trait measures and one set for state measures). First, anger-in and anger-out scores were entered in separate analyses, after the covariates, to examine the effect of each anger expression scale by itself. Second, the two scale scores were entered as simultaneous predictors to determine their combined effect. There were no significant findings with regard to the state or trait anger expression scales (for all, p > .05). Results for the separate and simultaneous analyses were virtually identical; those for the separate analyses are presented in Table 5. n = 92.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONPRESENT STUDYMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
Furthermore, both objective situational characteristics and mood and appraisal variables showed significant associations with anger expression scales. Situational characteristics that predicted anger expression included the target of the participant’s anger (ie, a person vs. an inanimate object), the relationship between the participant and target (ie, a close friend vs. a stranger), the authority of the target (subordinate vs. peer), location (public vs. private), and the physical presence of the target. Mood and appraisal variables, including whether the participant was still angry at the time of the report, anger intensity, whether the situation was resolved, and the participant’s perception of control over the resolution, were also associated with anger expression scales. The results are consistent with a number of conceptually obvious associations (eg, that individuals are less likely to express their anger outwardly in public places). They also revealed some less obvious but potentially important associations regarding the influence of the relationship between an individual and the target of his or her anger. For instance, individuals are most likely to express their anger outwardly when their anger is directed toward someone whom they know well and like, and they are least likely to suppress their anger when it is directed toward a subordinate. Taken together, the large number of significant findings in this study suggests that future research should consider assessing anger expression from a state perspective to address issues of situational variability.
Neither the state nor the trait anger expression scales in this study were related to ABP levels. Given the inconsistency of the results of prior studies, the lack of findings in regard to the trait measure is not particularly surprising. Although the meta-analysis by Suls et al. (4) indicated that the Spielberger anger-in and anger-out scales were significantly associated with resting blood pressure, this applied only to studies that used single blood pressure readings; their findings did not extend to those studies measuring blood pressure over multiple time points as in this study. The findings of the meta-analysis regarding anger-in were also weakened because of heterogeneity between study outcomes. Given the effect sizes reported in the meta-analysis, the present study had more than adequate power (96%) to detect the relationships reported by Suls et al. (4) between anger-in and SBP and DBP, although not enough (ie, <80%) to detect the weaker associations between anger-out and SBP.
This is the first study to assess the relationship between blood pressure and aggregated state measures of anger expression. There are a number of possible explanations for the lack of significant findings. One concerns the limited number of state assessments obtained per participant. On average, participants completed only four assessments each. This may not be a sufficient number to adequately represent participants’ typical experiences or ways of responding. There are also likely to be differences in the reliability and validity of data collected for each participant based on the number of reports they provided and the situational variability within those reports.
A second possibility is that the use of college students as participants in this study precluded the detection of associations that may be evident in other populations. Previous studies have predominantly used community samples, which may result in greater variations in blood pressure levels. Older adults may also be more likely to have developed habitual ways of expressing anger that are less subject to situational variability. Thus, the results of this study may not be generalizable to other populations. The association between state anger expression and blood pressure levels, as well as the questions pertaining to situational variability in anger expression, warrant replication in a sample of older adults.
Although the average frequency of anger incidents was lower than what was anticipated on the basis of Averill’s (31) finding of an average of 7 incidents reported prospectively over a week, it was in the range of "several times a day to several times a week" (see Ref. 44, p. 1146) established by previous research. Unfortunately, however, the low frequency of incidents reported in this study may have had the effect of weakening the correspondence between trait and state measures. As noted previously, we cannot assume that the situations reported were representative of participants’ typical experiences. We also do not know whether they are the types of experiences that participants were thinking about when completing the trait assessment. Future studies may need to include questions aimed at increasing the comparability of the state and trait measures, for instance, asking participants how typical each anger episode was.
Another potential problem with the methodology in this study is that, although participants were instructed to complete a questionnaire at the time they became angry, we cannot be certain that they followed these instructions or completed the diary all at once before returning it at the end of the week. If a large amount of the data had been reported retrospectively, however, there probably would have been higher correlations between state and trait anger expression scales, because longer recall periods of specific events tend to result in more trait-like reporting (45, 46). The best way to address this type of problem is through the use of more sophisticated data collection techniques, such as palm-top computers (eg, see Refs. 26, 37, and 38), although these can be prohibitively expensive to use.
Despite these limitations, this study also has a number of notable strengths. First, this is the first study to obtain multiple, detailed assessments of state anger expression in a naturalistic design. Previous studies have relied on trait measures of anger expression or have used hypothetical situations, assuming that people are able to report accurately how they typically respond when they are angry and that these responses can be generalized across situations and time. The assessment of anger responses in a naturalistic design is advantageous in that it minimizes biases associated with retrospective recall and increases ecological validity. In addition, repeated state assessments allow examination of situational variability in anger expression.
A second strength of this study is the use of multiple ABP readings. The majority of previous studies have relied on office or clinic measurements of blood pressure despite the fact that these types of measures do not accurately reflect 24-hour ambulatory readings and that ambulatory readings have been found to be more predictive of the time course and complications of hypertension than clinic measures (47–49). Studies by Shapiro et al. (13, 51) and Jamner et al. (50) are noteworthy exceptions, although they predicted situation-specific blood pressure readings from trait anger expression measures, whereas this study predicted average blood pressure from aggregated situation-specific anger expression assessments.
In conclusion, the results of this study point to the need to reconsider the way in which the health consequences of anger expression are studied. Understanding the effect of anger expression on physiological processes may require a more comprehensive model than the traditional perspective of anger expression as a stable personality disposition. This is the first study to compare trait and state assessments of anger expression as well as to specifically examine how anger expression varies according to situational characteristics. The results are consistent with the hypotheses that trait and state measures of anger expression are not equivalent and that situational factors play an important role in anger expression.
In general, research on anger expression would benefit from an increased focus on methodological issues. There is little evidence regarding the reliability and validity of currently used measures of anger expression. In addition, new assessment tools, designed for the purposes of examining situational variability, need to be developed. Although these recommendations give rise to a number of difficult conceptual and methodological issues, increasing the complexity of research designs in this area is likely to be the only way to achieve any definitive answers concerning the relationship between anger expression and blood pressure or other health outcomes.
|
NOTES |
|---|
|
TOPABSTRACTINTRODUCTIONPRESENT STUDYMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
An adjusted mean blood pressure was also calculated for each participant using information concerning participants’ location, posture, and activities during the blood pressure readings. This was accomplished by performing a general linear regression analysis predicting individual SBP and DBP readings from location, posture, and activities, and producing a least squares mean for each participant based on this regression equation. Analyses performed using adjusted blood pressure means as dependent variables did not differ from those using unadjusted blood pressure means.
The item pertaining to satisfaction with the resolution was excluded from analysis due to the large proportion of missing data for this variable. Because participants were instructed to answer this question only if the situation was resolved, inclusion of this variable would have resulted in the exclusion of the 44% of the situations that had not been resolved.
Received for publication September 16, 1998.
Revision received April 19, 1999.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONPRESENT STUDYMETHODSRESULTSDISCUSSIONNOTESREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  T. M. Starner and R. M. Peters Anger Expression and Blood Pressure in Adolescents The Journal of School Nursing, December 1, 2004; 20(6): 335 - 342. [Abstract] [PDF]
|
|
|
|
 |
|
 V. Chikani, D. Reding, P. Gunderson, and C. A. McCarty Wisconsin Rural Women's Health Study Psychological Factors and Blood Cholesterol Level: Difference between Normal and Overweight Rural Women Clin. Med. Res., February 1, 2004; 2(1): 47 - 53. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. M. Eng, G. Fitzmaurice, L. D. Kubzansky, E. B. Rimm, and I. Kawachi Anger Expression and Risk of Stroke and Coronary Heart Disease Among Male Health Professionals Psychosom Med, January 1, 2003; 65(1): 100 - 110. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. E. Williams, F. J. Nieto, C. P. Sanford, D. J. Couper, H. A. Tyroler, and A. Algra The Association Between Trait Anger and Incident Stroke Risk: The Atherosclerosis Risk in Communities (ARIC) Study * Editorial Comment: The Atherosclerosis Risk in Communities (ARIC) Study Stroke, January 1, 2002; 33(1): 13 - 20. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
