Author + information
- Received July 26, 1999
- Revision received April 21, 2000
- Accepted June 28, 2000
- Published online November 15, 2000.
- Peter Angerer, MD∗,* (, )
- Uwe Siebert, MD, MPH†,
- Wolfgang Kothny, MD∗,
- Dagmar Mühlbauer, MD∗,
- Harald Mudra, MD∗ and
- Clemens von Schacky, MD∗
- ↵*Reprint requests and correspondence:
Dr. Peter Angerer, Medizinische Klinik, Klinikum Innenstadt der Universität München, Ziemssenstrasse 1, D-80336, Munich, Germany.
This prospective cohort study of patients with coronary artery disease (CAD) sought to determine the impact of social support, anger expression and cynical hostility on progression of coronary atherosclerosis as shown by angiography.
Low social support, high levels of expressed anger and cynical hostility are correlated to increased CAD morbidity and mortality. However, the impact of these factors, alone or together, on progression of human coronary atherosclerosis is unknown.
Of 223 patients with CAD documented by standardized angiography at baseline, 162 had a second angiogram after two years. An expert panel who had no knowledge of the patients’ characteristics evaluated the films pairwise to determine disease progression. At baseline, all patients were asked to answer three self-report questionnaires: questions concerning emotional social support, the State-Trait-Anger-Expression Inventory (STAXI) and the Cook-Medley cynical hostility scale. Each patient’s clinical and laboratory status was followed.
Questionnaires and angiographic follow-up data were available for 150 patients. Bivariate analysis of the psychological variables showed a higher risk of progression only for patients who scored high on STAXI anger-out or low on social support. In the multivariate analysis, when adjusting for confounding variables and examining the interaction between psychological variables, only patients with both high anger-out and low social support were at highly increased risk for progression (odds ratio 30, confidence interval [CI] 5.5 to 165.1; RR 3.19).
Patients with CAD and low emotional social support who express anger outwardly are at a highly increased risk of disease progression, independent of medication or other risk factors.
Both low social support and high hostility are thought to increase a person’s risk of developing coronary artery disease (CAD) and to die from it (1). Direct and indirect pathophysiologic mechanisms may be involved in this relation. Both psychological factors have been shown to hyperactivate neuroendocrine systems, especially the sympathetic nervous system, and to activate platelets (1). They also promote adverse health behaviors (1). It is largely unknown, though, whether these two conditions directly influence atherogenesis in human coronary arteries.
Emotional support is probably the component of social support with the strongest influence on CAD (2,3). Hostility in a wider sense, as it is used here, also has several components (4). The potential for hostility (i.e., aggressive verbal or physical responses when angry) was found to be consistently related to CAD (4–6). The potential for hostility predicts restenosis (7), and anger predicts recurrent events, both after percutaneous transluminal coronary angioplasty (PTCA) (8). Suppressed anger, commonly believed to be harmful, did not predict CAD or all-cause mortality (4). So-called cynical hostility, as defined by Cook and Medley, is associated with CAD to a lesser extent; it can be described as an attitude of resentment, distrust, cynicism and suspicion (4,9). Some studies suggested that high hostility and lack of social support may act synergistically on CAD pathogenesis and expression (10–12).
In the present two-year prospective angiographic study, we sought to examine our hypothesis that each of the following psychological characteristics promotes progression of coronary atherosclerosis in patients with CAD: low emotional social support, as assessed by the self-report questionnaires used by Williams et al. (2); experience and outwardly directed expression of anger, as assessed by the State-Trait-Anger-Expression Inventory (STAXI) (13); and cynical hostility, as assessed by the Cook-Medley hostility scale (9). Furthermore, we hypothesized a positive interaction between the psychological characteristics.
Patients were participants in the Study on prevention of Coronary atherosclerosis by Intervention with Marine Omega-3 fatty acids (SCIMO) trial, a prospective, randomized, placebo-controlled, double-blind, single-center study. SCIMO assessed the effect of treatment with omega-3 fatty acids for two years on coronary atherosclerosis, as demonstrated angiographically. A detailed description of subjects and methods has been published (14).
All patients (n = 2,284) referred to our institution between September 1992 and May 1994 for diagnostic coronary angiography were screened for eligibility. Patients whose angiogram demonstrated >20% stenosis in at least one vessel were eligible, unless they met one of the predefined exclusion criteria (e.g., age >75 years, revascularization in more than one vessel [PTCA or coronary artery bypass graft operation] planned or already performed in the preceding six months, hemodynamically relevant left main coronary artery stenosis or proximal stenosis in all three main vessels, left ventricular ejection fraction <35%, history of cardiac transplantation, diabetes mellitus, psychiatric disorders) (14). A total of 396 patients were eligible; 338 could be approached; and 223 agreed to participate and gave informed, written consent that included follow-up coronary angiography.
The study was approved by the Ethics Committee of the Faculty of Medicine of the Ludwig-Maximilians-University of Munich, and was conducted according to the European Good Clinical Practice guidelines.
History, clinical status, laboratory work-up and electrocardiogram (ECG) were obtained at the beginning and after 1, 6, 12, 18 and 24 months. After 24 months, follow-up angiography was performed according to a highly standardized protocol, as described previously (14,15).
Assessment of coronary angiograms
Three expert cardiologists who had no knowledge of all possible aspects of the study, such as treatment, temporal order of the films and patient characteristics, including psychological status, evaluated all pairs of films, working simultaneously. The severity of all stenoses (percent lumen narrowing) and their relative clinical significance (e.g., proximal or distal location) was determined on each film. If one of the three coronary vessels had had PTCA six months before or any time during the study period, it was excluded from analysis (16). On the basis of this evaluation, the global difference between the baseline and follow-up angiogram was graded semiquantitatively on a scale from −3 to +3, as used in the Cholesterol Lowering Atherosclerosis Study (CLAS) (15). A score of 0 indicates no difference; 1 indicates a definitely discernible but small difference; 2 indicates an intermediate difference; and 3 indicates an extreme difference. Left ventricular ejection fraction was assessed on the baseline ventriculogram.
The extent of coronary atherosclerosis at the beginning was defined as the number of stenoses with ≥50% lumen narrowing; the clinical severity of the disease was defined as the number of main vessels with ≥70% lumen narrowing. When comparing the follow-up with the baseline angiogram, a score of +1, +2 or +3 (small, intermediate or extreme progression, respectively) was defined as progression; a score of 0, −1, −2 or −3 (no change, small, intermediate or extreme regression, respectively) was defined as no progression.
Clinical events were not considered as representing progression of atherosclerosis because other biologic mechanisms, such as plaque rupture in acute coronary syndromes, or complex decisions, such as those for PTCA, were involved. In addition, most PTCAs were either scheduled as a consequence of the baseline angiogram or were restenoses, resulting from a different pathomechanism (17).
Assessment of psychological characteristics
At the baseline visit, all patients received a paper-and-pencil version of three questionnaires for self-evaluation, were instructed in a standardized way and were reassured that their answers would be evaluated anonymously. Emotional social support was assessed by four questions used in previous research (2) to measure the functional aspect of emotional social support (e.g., a person’s perception of his or her relationships as supportive). This scale has not been psychometrically validated, but the distribution of answers obtained by Williams et al. (2) was very similar to ours. A score ranging from 0 to 4 was created by summing the answers (after inversion where appropriate). The scale was condensed by collapsing the scores 2 to 4, because of small numbers in these categories. Thus, a score of 0 indicates a high level of social support; 1 = medium level; and 2 to 4 = low level.
The German version of Spielberger’s STAXI (13,18) was used to measure angry temperament and angry reaction (trait anger), as well as forms of anger expression: “anger-out” denotes an outward direction of anger; “anger-in” denotes a suppression or nonexpression of angry emotions; and “anger-control” denotes use of strategies to control angry emotions. Cynical hostility was measured by the answers to 50 questions of the German version of the Minnesota Multiphasic Personality Inventory (MMPI), as defined by Cook and Medley (9). This is referred to as the full scale. Barefoot et al. (19) defined three subsets that altogether predicted mortality better than the full scale: cynicism, hostile affect and aggressive responding. The sum of these subsets (19) is referred to as combined subsets.
Questionnaires were considered valid if not more than one question concerning social support, and each of the anger (STAXI) scales (13,18), and not more than five questions concerning cynical hostility were left unanswered (9).
All analyses were performed using the SPSS statistical package, version 6.1.1. After the clinical data file was closed, the sealed envelopes were opened and questionnaires were evaluated by a person who had no knowledge of the patients’ characteristics, particularly the results of the expert panel.
Social support was treated as ordinal. Cynical hostility and STAXI scores were primarily used with no transformation and treated as quantitative variables.
First, bivariate analysis was performed 1) to examine the cross-sectional association of all psychological variables with the baseline characteristics relevant to progression of coronary atherosclerosis and 2) to examine the longitudinal association of all relevant baseline variables with the outcome (i.e., CAD progression). The statistical tests appropriate to the nature of the data were used (i.e., chi-square test, Student t test, Mann-Whitney U test). A p value ≤0.05 was considered significant.
Second, a multiple logistic regression model with CAD progression as the dependent variable was built in steps to identify and adjust for potential confounding variables and to examine the interaction between psychological variables. Step 1: Psychological variables (i.e., all scales and subscales) were selected as independent variables if they were associated with outcome (CAD progression) in the bivariate screening procedure at level p ≤ 0.20 (20). Those selected were forced into the model for further steps. Step 2: To assess confounding, the change-in-estimate criterion (21) was used, defined as ≥10% change in the odds ratio (OR) of at least one psychological variable forced into the model, as described earlier. This criterion was chosen to achieve maximal validity of the relation between psychological testing and outcome (21,22).
Thirteen potential confounding variables were preselected from all factors known to promote progression of coronary atherosclerosis: age, gender, positive family history of premature CAD, marital status, number of stenoses ≥50% at baseline, treatment with lipid-lowering drugs and study medication (omega-3 fatty acids), mean diastolic and systolic blood pressure, mean high (HDL) and low density lipoprotein (LDL) cholesterol, mean plasma glucose and mean number of smoked cigarettes, all during follow-up. Potential confounding variables were entered into the model by a forward selection strategy and removed if they did not meet the change-in-estimate criterion. A p value <0.05 was considered to be significant for the association of each psychological variable with coronary disease progression after adjustment for confounding. Step 3: The interaction between psychological variables was examined in the fully adjusted model. Using the likelihood-ratio test, a p value <0.05 was defined as significant.
Comparison of patients with and without complete data
Of 223 patients included in the study, 179 returned a completed questionnaire. Of these 168 cynical hostility questionnaires, 166 STAXI trait-anger, 164 anger-out, 164 anger-in and 166 anger-control scales, and 179 social support ratings were considered valid. A total of 162 patients participated in the study according to the protocol, including repeat coronary angiography at 24 months. For 150 of them, at least one complete psychological assessment (cynical hostility, STAXI or social support) could be related to coronary progression. Patients with complete data did not differ significantly, in terms of any relevant social, medical and (as far as information was available) psychological baseline characteristics, from patients who returned completed questionnaires but refused follow-up angiography (n = 29), those refusing the questionnaires but returning for angiography (n = 12) and those who refused both (n = 32) (data not shown).
Cross-sectional analysis at baseline
Bivariate analysis examining the cross-sectional associations between psychological variables and other baseline characteristics identified age as the only coronary risk factor that was inversely correlated with trait-anger (r = −0.16, p = 0.041) and anger-in (r = −0.22, p = 0.005).
Among the angiographic findings, cynical hostility (full scale) and cynical hostility (combined subsets) were significantly associated with the presence of at least one hemodynamically significant stenosis (>70%) (OR 1.06, 95% CI 1.00 to 1.24 and OR 1.11, 95% CI 1.01 to 1.21 per increment on the scale, respectively). Lack of social support correlated with cynical hostility (r = 0.27, p = 0.001) and anger-in (r = 0.25, p = 0.003), but not with anger-out.
Bivariate association of baseline data with CAD progression within two years
A total of 51 patients showed progression (45 with small and 6 with intermediate progression); 149 showed no progression (85 with no change, 12 with small and 2 with intermediate regression). Of all baseline variables, only the number of coronary segments with ≥50% stenosis, marital status and, with marginal significance, mean plasma glucose were able to distinguish between patients who would progress and those who would not (Table 1). Patients living with a partner (marital status) at the beginning of the investigation were more likely to experience progression on bivariate analysis (Table 1).
The bivariate association of psychological variables at baseline with CAD progression is presented in Table 2. Because the magnitude of the OR is dependent on the number of items in each scale, additionally, the original numeric scales were transformed into tertiles to facilitate comparison of ORs. The p values and further variable selection are based on the original scales. Only low social support and high anger-out scores were related to progression.
In a secondary analysis, the semiquantitative score was used as the coronary outcome. As for the dichotomous coronary outcome, among all psychological variables, only anger-out and social support were associated with the score (r = 0.16, p = 0.057 and r = 0.16, p = 0.050, respectively), and no other variable met the criterion for selection into the multivariate model.
Two patients had a myocardial infarction, and the angiogram of both was classified as showing progression. One patient had single-vessel bypass surgery and, after excluding this vessel from the analysis, was classified as showing no progression. Twenty-eight patients had PTCA. After exclusion of all vessels in which a segment had been dilated within the study, 21 patients were classified as showing no progression and seven as showing progression. On univariate analysis, the ORs for progression per increment of the psychological scales of patients without myocardial infarction, bypass surgery or PTCA during the study were very similar to those reported in Table 2: only anger-out (OR 1.12, 95% CI 1.03 to 1.23) and low and medium social support (OR 1.12, 95% CI 0.45 to 2.82 and OR 4.33, 95% CI 1.56 to 12.00, respectively) were significantly related to outcome.
Variable selection in the multivariate model
As a result of the bivariate screening, only social support and anger-out met the p < 0.2 selection criterion (Table 2) and were forced into a multivariate logistic regression model as psychological variables of interest. Of the 13 potential confounding variables, only age, marital status, number of stenoses ≥50%, high density lipoprotein (HDL) cholesterol and intake of lipid-lowering drugs met the change-in-estimate criterion and were controlled for in the final model (Tables 3 and 4). ⇓The remaining eight variables were removed from the model, regardless of their relation to progression in the bivariate analysis (Table 1).
Multivariate association of baseline data with CAD progression within two years
In the logistic regression model, only social support was significantly related to progression of coronary atherosclerosis after 24 months. This relation was independent of CAD extent, as it remained significant when examined separately in patients with a smaller number (0 to 1) of segments with ≥50% stenosis and in patients with a higher number (≥2) at baseline.
Using the fully adjusted model, a second multiple model was built by adding the interaction term social support combined with STAXI anger-out (data not shown). This interaction was significantly related to progression (p = 0.035).
To describe this effect for distinct groups, the original social support scale (low, medium and high support) was combined with two groups obtained by the median split of the STAXI anger-out scale (≤12/≥13 score points) (Table 4). The scale was not split to a higher degree to avoid cell-expected numbers becoming too small. As compared with patients in the high support/low anger reference group, patients in the low support/high anger group had an OR of 30 for coronary progression (95% CI 5.5 to 165.1).
Because the OR overestimates the RR (which cannot be obtained directly by logistic regression analysis) if the so-called rare disease assumption is not met, as in this case, the RR was calculated for a median covariate setting (Table 4) (i.e., by calculating the probability of progression for patients with median values in all covariates (age 58 years, married, two coronary segments with ≥50% stenosis, HDL cholesterol 1.3 mmol/liter, lipid-lowering drugs during six months of follow-up). For patients in the low support/high anger group, this translates to a RR of 3.19.
The potential influence of the 29 patients who dropped out during the study but completed questionnaires at baseline on the associations presented in Table 4 was examined by making two extreme assumptions: 1) all patients with high support and low anger (n = 7, reference group), but no patients with low support and high anger (n = 6, high risk group), had progression; and 2) no patients with high support and low anger, but all patients with low support and high anger had progression. Further, both assumptions were based on that one-third of the patients in the remaining four medium-risk groups had progression. For assumption 1, the OR of the high risk group diminished to 4.96 (CI 1.49 to 16.51); for assumption 2, the OR increased to 62.05 (CI 11.89 to 323.69).
Effect modification by treatment with omega-3 fatty acids with respect to social support and anger-out was examined using interaction terms in the logistic regression model. The effect sizes of the estimated risks associated with the psychological measures were not different (p = 0.848) between the active treatment and placebo groups.
This investigation examined, for the first time to our knowledge, the influence of psychological factors on the progression of coronary atherosclerosis in a prospective angiographic study. We demonstrated that low perceived social support and, to a lesser extent, high outwardly expressed anger, both reported at baseline, were associated with accelerated progression within two years of follow-up in a bivariate analysis. Only when both psychological factors were present at a high level and acted in concert was this associated with a markedly elevated risk (OR 30.0, RR 3.19) in a multivariate analysis controlling for other known risk factors. For the psychological characteristics—cynical hostility, angry temperament, suppression of anger and control of anger—there was no significant association with progression.
Interaction of anger and social support
High anger and low social support influenced progression of CAD by acting synergistically (Table 4), as we hypothesized based on previous observations (10,11). Owing to the prespecified variable selection process, the interaction of these two but not of other psychological measures was tested. Dealing with this issue, there is only one prospective angiographic study conducted with cynomolgus monkeys, which demonstrated that destruction of social bonds and thereby induction of aggression accelerates atherogenesis in coronary arteries (23). Thus, our study adds weight to the idea that not one but two personality features, acting jointly, deleteriously affect the course of CAD, thereby defining the high risk group more precisely.
Emotional versus structural social support
We found the association of the emotional quality, as compared with the structural quality (i.e., being married), of social support with CAD progression was stronger in bivariate and multivariate analyses. A recent meta-analysis concluded that the quality of support, in particular, emotional support, showed the largest effect on CAD (3). In contrast, the quantity of support (e.g., social network size, marital status, number of friends or relatives) was observed to predominantly influence CAD mortality, possibly reflecting the availability of quick medical assistance (2,24). For both qualities, the association with CAD may be different in women (25). However, in our study, the effect was not modified by gender.
In this study, cross-sectional analysis at baseline showed an association between cynical hostility and the presence of significant (≥70%) coronary stenosis, but no association with progression in the two following years. Two explanations are plausible: our patients scored high on the cynical hostility scale (mean 24.1, median 24), as compared with those patients in studies that found an association (cutpoint 10 , mean 16.1  and mean 14.65 ). Thus, selection of a highly hostile study group may have made the difference between high and low cynical hostility undetectable in our study (so-called disease-based spectrum bias) (4,28). Second, it was recently demonstrated that increased risk for all-cause mortality, cardiovascular mortality and nonfatal myocardial infarction of highly cynical hostile men is primarily mediated through coronary risk factors (29). Intense treatment of risk factors in our patients may have leveled off any difference in risk factor profile between more and less cynically hostile patients.
Definition of coronary outcome
A global change on the coronary angiograms was assessed visually by an expert panel using a semiquantitative scale (30). Studies directly comparing evaluation by the expert panel and by quantitative coronary angiography showed good to excellent agreement for both methods if the change of natural coronary atherosclerosis (not restenosis after PTCA) was assessed (31,32). The expert panel appears to be even more superior if the coronary angiographic change visualized over the entire coronary artery tree is the outcome of interest (33). With regard to the published data, we had hypothesized that the psychological factors studied would promote progression, and we therefore chose a dichotomous primary coronary outcome. In a secondary analysis, similar associations were found for psychological variables with a semiquantitative coronary outcome, suggesting that regression, the type of change that is defined as “no progression” in the dichotomous outcome, is not differently influenced by the psychological measures.
Lipids and coronary outcome
Low density lipoprotein cholesterol levels were not associated with coronary outcome. This is in accordance with several lipid-lowering trials that showed only weak and inconsistent correlations between baseline LDL cholesterol, in-trial LDL cholesterol or absolute reduction of LDL cholesterol levels and angiographic progression of atherosclerosis (34). The lipid levels were not exceedingly high in this study, whereas the correlations are more likely to be found with high baseline levels (34).
Our predefined hypothesis was tested in an investigation designed primarily to study the effect of omega-3 fatty acids on progression of coronary atherosclerosis. This shortcoming has its ethical reasons and characterizes all clinical studies on psychosocial factors that use invasively measured end points. Our data are observational by nature because the psychosocial conditions studied cannot be randomized. Only an intervention trial could achieve a higher level of evidence. Some of the potential confounding factors were measured during the study because a change in the angiograms occurred during this time. Previous studies that found any associations between LDL or HDL cholesterol and coronary progression found them to be strongest for lipid levels during the trial (35). Variables measured during follow-up could be interpreted as both confounding variables and intermediate steps in the causal pathway between psychological measures and progression. In the latter case, we would have underestimated the effect. Yet, previous studies have found expressed anger and social support to be associated with CAD independent of standard risk factors (4,36).
Approximately one-third of the participants had either no second angiogram or no questionnaire, or neither. Our data indicate that these participants did not differ in any respect from the remaining participants. Most people in this group ended participation within the first month after randomization. Thus, it seems very unlikely that outcome (progression) triggered the selection and led to selection bias. Yet, scores for lack of social support and for outwardly expressed anger were (nonsignificantly) higher in patients who refused a second angiogram. Sensitivity analysis showed that even if all patients in the low risk group but none in the high risk group (referring to the results in Table 4) had progression, the association would be weaker but still significant, indicating robustness of the results against differential selection bias. Still, we have no psychological questionnaires for 44 of 223 patients, thus somewhat limiting the generalizability. However, the magnitude of the OR and RR for progression in the low support/high anger group, as well as the support from epidemiologic and experimental studies, corroborates our results.
We identified patients with CAD, characterized by a combination of self-reported high outwardly expressed anger and low social support, who are at a highly increased risk (OR 30, 95% CI 5.5 to 165.1; RR 3.19) for CAD progression. This finding contributes to the understanding of the pathogenesis of CAD. Defining a specific high risk group is a prerequisite for developing a targeted prevention strategy.
☆ This study was supported in large part by the Bundesministerium für Forschung und Technologie, Bonn; through Gesellschaft für Strahlenforschung (GSF, 07ERG03), Neuherberg and Deutsche Forschungsanstalt für Luft und Raumfahrt (DLR, 01 EA 9501/7), Köln; and by the Wilhelm Sander-Stiftung (93.032), Munich, Germany. It was also supported by the Deutsche Forschungsgemeinschaft (Scha 398), Bonn, Germany. Pronova, A.S., Lysaker, Norway, provided the study medication and funds for monitoring.
- coronary artery disease
- confidence interval
- electrocardiogram or electrocardiographic
- high density lipoprotein
- low density lipoprotein
- odds ratio
- percutaneous transluminal coronary angioplasty
- risk ratio
- Study on prevention of Coronary atherosclerosis by Intervention with Marine Omega-3 fatty acids
- State-Trait-Anger-Expression Inventory
- Received July 26, 1999.
- Revision received April 21, 2000.
- Accepted June 28, 2000.
- American College of Cardiology
- Rozanski A,
- Blumenthal J.A,
- Kaplan J
- Greenwood D.C,
- Muir K.R,
- Packham C.J,
- Madeley R.J
- Kawachi I,
- Sparrow D,
- Spiro A,
- Vokonas P,
- Weiss S.T
- Thomas S.A,
- Friedmann E,
- Wimbush F,
- Schron E
- Blumenthal J.A,
- Burg M.M,
- Barefoot J,
- Williams R.B,
- Haney T,
- Zimet G
- Orth-Gomer K,
- Unden A.L
- Spielberger C.D
- Mudra H,
- Klauss V,
- Blasini R,
- et al.
- Schwenkmezger P,
- Hodapp V,
- Spielberger C.D
- Barefoot J.C,
- Dodge K.A,
- Peterson B.L,
- Dahlstrom W.G,
- Williams R.J
- Maldonado G,
- Greenland S
- Miettinen O.S,
- Cook E.F
- Kaplan J.R,
- Manuck S.B,
- Clarkson T.B,
- Lusso F.M,
- Taub D.M,
- Miller E.W
- Seeman T.E,
- Syme S.L
- Williams R.J,
- Haney T.L,
- Lee K.L,
- Kong Y.H,
- Blumenthal J.A,
- Whalen R.E
- Shekelle R.B,
- Gale M,
- Ostfeld A.M,
- Paul O
- Everson S.A,
- Kauhanen J,
- Kaplan G.A,
- et al.
- Mack W.J,
- Selzer R.H,
- Pogoda J.M,
- et al.
- Brown B.G,
- Hillger L.A,
- Lewis C,
- et al.
- Orth Gomer K,
- Horsten M,
- Wamala S.P,
- et al.