vitaliano et al 2002

A Path Model of Chronic Stress, the Metabolic Syndrome, and Coronary HeartDiseasePETER P. VITALIANO, PHD, JAMES M. SCANLAN, PHD, JIANPING ZHANG, MS, MARGARET V. SAVAGE, PHD,

IRL B. HIRSCH, MD, AND ILENE C. SIEGLER, PHD, MPH

Objective: We tested a theoretical stress model cross-sectionally and prospectively that examined whether rela-

tionships of chronic stress, psychophysiology, and coronary heart disease (CHD) varied in older adult men (N � 47),

older adult women not using hormone replacement therapy (HRT) (N � 64), and older adult women using HRT (N� 41). Method: Structural equations examined relationships of CHD with 1) chronic stress (caring for a spouse with

Alzheimer’s disease and patient functioning), 2) vulnerability (anger and hostility), 3) social resources (supports),

4) psychological distress (burden, sleep problems, and low uplifts), 5) poor health habits (high-caloric, high-fat diet

and limited exercise), and 6) the metabolic syndrome (MS) (blood pressure, obesity, insulin, glucose, and lipids).

Results: Caregiver men had a greater prevalence of CHD (13/24) than did noncaregiver men (6/23) (p � .05) 27 to

30 months after study entry. This was influenced by pathways from caregiving to distress, distress to the MS, and

the MS to CHD. In men, poor health habits predicted the MS 15 to 18 months later, and the MS predicted new CHD

cases over 27 to 30 months. In women, no “caregiving-CHD” relationship occurred; however, 15 to 18 months after

study entry women not using HRT showed “distress-MS” and “MS-CHD” relationships. In women using HRT,

associations did not occur among distress, the MS, and CHD, but poor health habits and the MS were related.

Conclusions: In older men, pathways occurred from chronic stress to distress to the metabolic syndrome, which in

turn predicted CHD. Older women not using HRT showed fewer pathways than men; however, over time, distress,

the MS, and CHD were related. No psychophysiological pathways occurred in older women using HRT. Key words:stress, coronary disease, gender, metabolic syndrome, hormone replacement, path analysis.

AD � Alzheimer’s disease; BP � blood pressure; BMI

� body mass index; CHD � coronary heart disease;

CVD � cardiovascular disease; DBP � diastolic blood

pressure; HDLC � high-density lipoprotein cholester-

ol; HRT � hormone replacement therapy; LV � latent

variable; LVPLS � latent variable partial least square;

MLE � maximum likelihood estimation; MS � meta-

bolic syndrome; MV � manifest variable; PLS � par-

tial least squares; RMS � root mean square; SBP �systolic blood pressure; SES � socioeconomic status;

TG � triglycerides.

INTRODUCTION

There is extensive em

1

pirical support for the hy-

pothesis that chronic stress is associated with cardio-

vascular disease. Greenwood et al. (1) reviewed 14

prospective studies of humans, and almost all of the

studies found relationships of stress with CVD or cor-

onary heart disease. Pickering et al. (2) found that in

persons with high job strain, blood pressure increased

over time, both at home and at work. Unfortunately,

few of these studies focused on men over 65 years of

age and still fewer included older women. Because

older adults are less resilient to stress and illness than

are younger adults and chronic stress may be more

related to CHD in older than in younger adults, this is

an important area of study. Indeed, in the absence of

good health habits, even healthy older adults exhibit

greater risks for CVD. These include greater BP (3),

body fat (3), and insulin and glucose (4). Hence, 72%

of persons over age 65 have CVD, and 25% have CHD

(5). Importantly, CHD prevalence in women increases

dramatically after age 50, suggesting that without hor-

mone replacement therapy, postmenopausal women

may lose much of their CHD “protection” (6).

Our literature review of research on chronic stress and

CHD noted the absence of studies that used theoretical

models to guide their research and hypotheses. More-

over, few studies used prospective designs with natu-

rally occurring (and ecologically valid) chronic stressors

in older men and women. Still fewer studies examined

psychosocial, behavioral, and physiological measures to

represent predisposing, mediating, and outcome vari-

ables within the same investigation. Here, we attempted

to narrow this gap. However, to meet the above require-

ments as well as have a large sample would have been

prohibitively expensive. Instead, we used a moderately

sized sample and a theoretical model of distress to cross-

sectionally and prospectively examine interrelationships

of a natural chronic stressor and psychosocial, physio-

logical, and biomedical variables. We did this in older

adult men, women not using HRT, and women using

HRT. Our model posits that chronic stress, personal vul-

nerabilities, and personal and social resources lead to

psychological distress and poor health habits. Distress

Departments of Psychiatry and Behavioral Sciences and Medicine,

University of Washington (P.P.V., J.M.S., J.Z., M.V.S.), Seattle,

Washington, and Department of Psychiatry and Behavioral Sciences,

Duke University (I.C.S.), Durham, North Carolina.

Address reprint requests to: Peter P. Vitaliano, PhD, University of

Washington, Department of Psychiatry and Behavioral Sciences, Box

356560, Seattle, WA 98195-6560. Email: [email protected]

Received for publication July 27, 2000; revision received June 22, 2001.

418 Psychosomatic Medicine 64:418–435 (2002)

0033-3174/02/6403-0418

Copyright © 2002 by the American Psychosomatic Society

and poor health habits, in turn, lead to physiological

disregulation (Figure 1, top) (7). Later, we will discuss

how the model constructs and pathways from chronic

stress, personal vulnerabilities, and personal and social

resources may be used to predict psychobehavioral re-

sponses (psychological distress and poor health habits).

Fig. 1. Top, theoretical path model for chronic stress and CHD. Bottom, path model for chronic stress and CHD with manifest variables.

STRESS, METABOLIC SYNDROME, CORONARY DISEASE

419Psychosomatic Medicine 64:418–435 (2002)

We will also review how these responses can lead to

physiological disregulation, which may be followed by

CHD.

Pathways From Chronic Stress, Vulnerabilities,

and Resources to Psychological Distress and Poor

Health Habits

The model proposed here is an attempt at a parsimo-

nious and analytic depiction of most aspects of the cog-

nitive-phenomenologic model of stress (CPMS) (8). Here,

insulin levels exist in persons at risk for CHD (50, 47)

and in persons with atherosclerosis with and without

ischemia (51, 52). Hyperinsulinemia also predicts

CHD 5 to 10 years later, independent of diabetes and

other risk factors (47, 50); however, the prediction

results from insulin resistance and not directly from

hyperinsulinemia. The synergistic effects of insulin

resistance, obesity, hypertension, and hyperlipidemia

form a construct called the “metabolic syndrome” (53),

which is a predictor of CHD (47).

Psychosocial Correlates and Predictors of the MS andCHD Distress combined with overeating and inactivity

can lead to greater levels of obesity, insulin, glucose, BP,

and plasma lipids (53). Psychosocial correlates and pre-

dictors of CHD include greater chronic stress, anger, and

hostility (54, 55), higher levels of psychological distress

(55, 56), lower social supports (57), lower SES (58), and

poorer health habits (59). Finally, the lack of positive

experiences is associated with a greater waist-hip ratio

(60) and worse glycemic control (61).

Research Questions

Given the proposed model and its supporting litera-

ture, we examined whether older persons under chronic

stress would manifest greater CHD prevalence than older

persons not under such stress and whether the model

pathways would relate to CHD prevalence. We also ex-

amined several associated questions about the pathways

in Figure 1, top, and their potential roles as mediators of

relationships of chronic stress with CHD:

1. Is variability in distress influenced by chronic

stress, vulnerabilities, and resources?

2. Is variability in poor health habits influenced by

chronic stress and distress?

3. Is variability in the MS influenced and/or pre-

dicted by distress and poor health habits?

4. Is CHD prevalence influenced and/or predicted

by elevated MS values?

5. Do intermediate pathways mediate relationships

between chronic stress and CHD?

6. Given the potential compounding of chronic

stress across time, do the above relationships become

stronger over time?

7. Are the above relationships modified by gender

and HRT use?

To examine these questions, we defined “persons

under chronic stress” as persons caring for a spouse

with Alzheimer’s disease. Spouse caregivers experi-

ence numerous stressors (physical, emotional, and fi-

nancial) (62), hassles (62), anger (63), and depressed

mood (63) relative to matched control subjects, and

such variables are associated with hostility (64). The

demands of caregiving coupled with the biological

vulnerabilities of aging put spouse caregivers at in-

creased risk for CVD and CHD (65) and make them an

appropriate group for examining the above questions.

We excluded persons with diabetes. Diabetes is asso-

ciated with insulin-glucose disregulation, increased

risk for depression, and cardiovascular complications.

Its premorbid presence would have complicated our

analyses and subsequent inferences.

METHODS

Participants

Caregiver couples were recruited from 1) mailings to physicians,

2) the University of Washington Alzheimer’s disease (AD) registry,

3) the AD Association, and 4) printed and electronic media. Criteria

for care recipient inclusion were 1) living with a spouse who is the

primary caregiver, 2) �60 years old, and 3) diagnosis of possible or

probable primary degenerative dementia based on the Diagnostic

and Statistical Manual of Mental Disorders (DSM) III (66), the Na-

tional Institute of Neurological and Communicative Disorders and

Stroke and the Alzheimer’s Disease and Related Disorders Associa-

tions (NINCDS-ADRDA) (67), and exclusion criteria, which are pro-

vided in much greater detail elsewhere (68). Noncaregivers were

recruited from senior centers, retirement organizations, and the me-

dia. They and their spouses (noncare recipients) had to be �60 years

old and functioning independently (as did the caregivers). No ex-

clusion criteria for major illnesses were used a priori for caregivers

and noncaregivers. This provided a more representative sample of

older adults.

Because Alzheimer’s disease is a progressive degenerative disor-

der, we used the course of this disease as a chronic stressor for

spouse caregivers. We assessed their temporal psychophysiological

reactions to such stress relative to noncaregivers at two points in

time (time 1 was study entry, and time 2 was 15–18 months later).

One year after time 2 (or 27–30 months after study entry), we also

obtained these subjects’ medical records. Finally, at times 1 and 2 we

examined cognitive and behavioral measures in care recipients and

noncare recipients.

At time 1, the samples consisted of 90 spouse caregivers and their

spouses (care recipients with AD) and 88 noncaregiver spouses (and

their spouses). From time 1 to time 2, we had attrition of 13 care-

givers because of the following reasons: 3 “care recipients” did not

decline and were thought not to have AD, 5 caregivers died (3 from

heart attacks and 2 from strokes), 1 got divorced, 1 noncaregiver

moved, and 3 refused to continue. Five noncaregivers were lost to

follow-up because of the following reasons: one noncaregiver died of

a stroke, two spouses of noncaregivers died (one from cancer and

one from CHD), and two noncaregivers moved. Five caregivers and

three noncaregivers with verified diabetes (see medical records be-

low) were excluded in the current analyses. Finally, although an

attempt was made to group-match the caregivers and noncaregivers

on age and gender, the percentage of women using HRT in each

group was left to chance. Complete data were available on 72 care-

givers (24 men, 28 women not using HRT, and 20 women using

HRT) and 80 noncaregivers (23 men, 36 women not using HRT, and

21 women using HRT).

Measures Used to Assess Constructs

Measures of chronic stress. Caregiver status (caregiver vs. noncar-

egiver) was assessed according to the above inclusion and exclusion

criteria. Care recipient mental status was measured by the Mini-Mental



Status Exam (69). It assesses orientation, memory, etc. Care recipient

functioning was measured by the Record of Independent Living (70). It

assesses competence in maintenance (eg, feeding and washing) and

higher functioning (eg, reading and recreation) (Figure 1, top).

Measures of personal resources. The Revised Ways of Coping

Checklist (71) assessed problem-focused coping (eg, came up with a

between MVs.

1

To address whether LVs were separate constructs,

we examined the r values of the residuals. Interpretation of the LVs

was based on 1) the r and � values between LVs, 2) the amounts of

variability in the endogenous LVs explained by their predictors, 3)

the magnitude of the r

Record of Independent Living at time 2. Tables 1 and

2 contain health-related, psychosocial, and demo-

graphic data for the MVs used to assess the LVs in

Figure 1, bottom, and for other variables of interest.

MVs are compared primarily for caregivers and non-

caregivers (or for their spouses) stratified by gender

and HRT (men, women not using HRT, and women

using HRT). At each time, male caregivers were more

obese and reported more depressed mood, more bur-

den, fewer uplifts, fewer social supports, less prob-

lem-focused coping, and less education than did

male noncaregivers. Caregiver women not using

HRT reported less exercise than their noncaregiver

counterparts as well as more depressed mood, sleep

problems, and burden at both times. Caregiver

women using HRT reported more depressed mood,

sleep problems, and burden and fewer uplifts than

their noncaregiver counterparts at both times. Male

caregivers reported less burden at time 1 and fewer

sleep problems at times 1 and 2 than did each group

of female caregivers. Finally, cognitive and func-

tional impairments were greater for AD victims than

for spouses of noncaregivers (Table 2).

Relationships of Caregiving With CHD

Point prevalence. Dates of diagnoses were used to

estimate CHD at or before study entry and in the 27- to

TABLE 2. Psychosocial and Demographic Variables Used to Measure Latent Variables at Times 1 and 2, Stratified byCaregivers/Noncaregivers, Gender, and HRTa

Variables Time

Men Women Not Using HRT Women Using HRT

Caregivers(N � 24)

Noncaregivers(N � 23)





Mini-Mental Status T1 19.0 (5.0)c 28.8 (1.5)** 20.9 (5.4) 28.8 (1.3)** 18.4 (5.8) 28.9 (1.2)**T2 13.8 (6.7) 29.0 (1.2)** 14.1 (7.5) 29.0 (1.2)** 10.4 (7.6) 28.8 (1.3)**

Maintenance functioningb T1 1.0 (0.8) 0.0 (0.0)** 1.2 (1.0) 0.0 (0.0)** 1.7 (1.2) 0.0 (0.0)**T2 1.9 (1.5) 0.0 (0.0)** 2.1 (1.4) 0.0 (0.1)** 2.8 (1.6) 0.0 (0.0)**

Higher functioningb T1 1.9 (1.4) 0.0 (0.1)** 2.0 (1.4) 0.1 (0.1)** 3.0 (1.1) 0.0 (0.0)**T2 2.9 (1.4) 0.1 (0.1)** 3.1 (1.1) 0.0 (0.1)** 3.6 (1.4) 0.0 (0.1)**

Depression T1 6.8 (4.1) 3.6 (2.3)** 8.3 (4.2) 5.7 (4.1)* 8.7 (3.4) 3.4 (2.5)**T2 6.2 (2.9) 4.5 (2.6)* 7.5 (4.1) 4.7 (2.9)** 7.9 (3.2) 4.7 (3.4)**

Sleep problems T1 7.2 (4.1) 6.2 (4.0) 11.0 (4.7) 7.6 (3.2)** 10.9 (5.2) 7.5 (4.1)*T2 7.6 (3.6) 6.5 (4.5) 10.7 (4.1) 7.5 (4.1)** 11.5 (6.1) 7.9 (3.5)*

Burden T1 32.3 (6.6) 26.3 (1.8)** 39.3 (10.3) 27.1 (2.7)** 40.2 (10.0) 26.0 (1.2)**T2 34.3 (8.8) 25.7 (0.96)** 38.0 (7.1) 27.1 (2.5)** 40.2 (10.7) 25.9 (1.5)**

Uplifts T1 61.4 (27.9) 70.8 (26.2) 58.1 (23.9) 61.7 (19.3) 62.4 (18.5) 81.0 (22.8)**T2 53.2 (23.1) 68.8 (23.2)* 58.1 (19.0) 57.8 (19.0) 64.4 (24.5) 78.9 (20.1)*

Trait anger T1 28.0 (6.5) 27.6 (5.6) 28.3 (7.0) 27.3 (6.9) 27.2 (5.4) 26.9 (6.3)T2 27.3 (7.3) 26.8 (5.2) 28.6 (6.7) 26.0 (6.8) 25.4 (5.0) 25.7 (6.2)

Anger-out T1 12.3 (4.4) 12.3 (2.6) 13.2 (4.8) 11.9 (3.3) 11.9 (2.0) 11.9 (3.1)T2 12.0 (2.4) 11.5 (2.3) 12.1 (3.5) 11.3 (2.6) 11.2 (1.7) 12.0 (2.9)

Anger control T1 26.4 (3.5) 26.4 (4.0) 24.8 (6.2) 26.5 (4.1) 24.0 (4.6) 27.5 (4.2)*T2 25.9 (4.8) 27.5 (4.0) 25.3 (4.9) 26.7 (3.9) 26.1 (4.3) 26.6 (4.3)

ISEL Tangible Support T1 31.8 (4.9) 32.0 (4.8) 28.2 (6.2) 30.2 (5.3) 29.4 (4.3) 35.5 (4.2)**T2 30.6 (3.8) 32.6 (3.9) 28.4 (7.0) 31.8 (4.7)* 29.4 (7.0) 35.2 (4.1)**

ISEL Self-Esteem T1 29.4 (3.9) 28.8 (3.6) 26.2 (4.7) 27.6 (5.1) 28.1 (5.2) 29.9 (3.6)T2 27.9 (3.4) 30.1 (2.9)* 27.4 (6.1) 28.3 (3.5) 27.4 (3.9) 31.5 (4.2)**

ISEL Belonging T1 30.3 (4.3) 30.2 (5.8) 27.9 (5.5) 29.5 (4.0) 29.6 (4.4) 33.0 (3.3)**T2 28.6 (4.3) 30.7 (2.7)* 28.8 (6.2) 30.1 (4.5) 30.7 (4.9) 33.3 (3.7)

ISEL Appraisal T1 27.8 (5.5) 29.1 (5.7) 26.3 (6.0) 27.1 (4.6) 24.9 (8.4) 30.4 (5.4)*T2 28.1 (6.6) 30.1 (4.5) 25.2 (7.3) 26.8 (4.7) 27.2 (6.7) 30.3 (4.7)

Social support satisfaction T1 5.6 (0.7) 5.9 (0.2)* 5.4 (1.2) 5.6 (0.7) 5.4 (1.0) 5.7 (0.6)T2 5.4 (0.9) 6.0 (0.0)** 5.3 (1.1) 5.5 (0.8) 5.5 (1.2) 5.6 (0.6)

Problem-focused coping T1 14.4 (4.4) 18.1 (4.9)* 14.7 (3.5) 13.1 (5.4) 16.8 (3.8) 16.0 (5.6)T2 15.4 (4.1) 16.1 (5.9) 15.4 (4.1) 13.8 (4.7) 17.0 (5.0) 16.8 (7.4)

Education (yr) T1/T2 14.0 (2.3) 17.4 (3.7)** 13.5 (2.1) 13.8 (2.4) 14.5 (2.2) 15.0 (2.8)Income (median) T1/T2 $35,000 $45,000 $25,000 $35,000 $30,000 $35,000

aAll comparisons are between caregivers and noncaregivers (or their spouses) within each stratum.

bValues are for spouses of caregivers or noncaregivers.

cMean (SD).

* p � .05; ** p � .01.

P. P. VITALIANO et al.


30-month interval thereafter. At study entry (time 1), no

differences existed in the point prevalence of CHD for

caregivers and noncaregivers (Figure 2); however, 27 to

30 months later, the point prevalence of CHD in caregiver

men (54%; 13/24) was higher than in noncaregiver men

(26%; 6/23) (�2

(1) � 3.85; p � .05) (Figure 2, top).

Incidence. In the 27- to 30-month interval after

block (in which the LV is viewed as an effect rather than

a cause of the MVs) to estimate the poor health habits LV.

Here, LVPLS estimates the LV as a linear combination of

the MVs to maximize the relationship with other LVs

(85).

4

Despite this procedure, poor health habits was still

quite variable. Based on the regression weights, poor

health habits were defined by diet in men at time 1.

However, over time, exercise became increasingly im-

18-path coefficients and 18-R

2

values from chronic

stress, vulnerability, and resources to distress. In 14 of

18 cases (in bold) these predictors explained signifi-

cant total variance (R

2

) in distress

6

; namely 46.1% to

70.0%. In all strata, positive relationships between

chronic stress and distress were very strong (R

2 �30%–45%, Table 3). Predictors were also related—

vulnerability and social resources were related (nega-

tively) in women, and negative relationships between

social resources and distress were larger in women not

using HRT (R

2 � 20.4% and 28.4%) than in women

using HRT (R

2 � 13.4% and 15.0%) and two to five

times larger than in men. Also, the positive relation-

ship of vulnerability with distress was significant in

men at time 1, significant in women not using HRT at

both times, and never significant in women using

HRT.

Is variability in poor heath habits influenced bychronic stress and distress? Depending on the stratum

and time of study, six of the 12 paths from chronic

stress and distress to health habits were significant

(total R

2 � 5.8% to 32.8%) (Table 3).

Is variability in the MS explained by distress andpoor health habits? Depending on the stratum and

time of study, six of the 12 paths from distress and

poor health habits to the MS were significant (total R

2

� 6.8% to 23.0%). Figure 3 contains the final models

for the strata. Emboldened paths were significant at

times 1 and 2, dotted paths were only significant at

time 1, dashed paths were only significant at time 2,

and deleted paths were nonsignificant at both times.

The pathway from distress to the MS was significant at

both times in men, significant at time 2 in women not

using HRT, and never significant in women using

HRT. The relationship of poor health habits with the

MS, although insignificant in men, is significant in

women not using HRT at time 1 and significant in

women using HRT at both times (explaining 15.2%

and 20.7% of the variance, respectively).

Is CHD prevalence explained by elevated MS val-ues? The pathway from the MS to CHD is significant in

men (R

2 � 14.8%) and women not using HRT at time

2 (R

2 � 7.3%) and never significant in women using

HRT.

Do intermediate pathways mediate relationships ofchronic stress with CHD? In men, the relationship of

chronic stress with CHD at time 2 (Figure 2, top) may

have been mediated because it became insignificant in

the presence of significant pathways (Figure 3, top)

from chronic stress to distress, distress to the MS, and

the MS to CHD.

Do the relationships determined in the previous fiveresearch questions become stronger over time? In

women not using HRT, several pathways are signifi-

cant at time 2 that were not significant at time 1. These

include chronic stress to health habits, distress to the

MS, and the MS to CHD. In men, pathways from social

resources to distress and from the MS to CHD were

significant at time 2 but not at time 1.

Are the above relationships modified by gender andHRT use? Extensive modification occurred for poor

health habits with the MS. In men, this relationship

was driven primarily by associations of sedentary be-

haviors with all variables in the MS. In women not

using HRT, this relationship was driven primarily by

associations of sedentary behaviors with insulin, glu-

cose, and obesity. However, in women using HRT, this

relationship was driven primarily by the association of

fat intake with glucose.

Evaluation of Overall Cross-Sectional Models

To evaluate models, six criteria are recommended

(83, 85): 1) LVs should have more than three MVs—in

our models, all LVs had three to five MVs except CHD;

2) measurement loadings should be �0.55—here, 98 of

120 reflective paths met this criterion; 3) the total R

2

for endogenous LVs (ie, resources, poor health habits,

distress, MS, and CHD) should be �0.10—here, 21 of

30 LVs had total R

2 � 0.10 (this is calculated by adding

over the R

2

values for the paths leading to each LV);

and 4) each path should explain �1.5% of the variance

in a predicted LV—here, 52 of 66 paths met this cri-

terion. Criteria 5 and 6 involve comparing the ob-

served model to the null (no paths among LVs) and

saturated (paths between all exogenous and endoge-

nous LVs) models. An observed model should be par-

simonious and closer to the saturated than the null

model. Criterion 5 uses the average R

2

values for en-

dogenous LVs. Criterion 6 uses the root mean square

Cov(E, U). The RMS of the covariance between the MV

residuals compares the relative goodness of fit of the

observed to the saturated and null models.

Table 4 shows the average R

2

values and RMS

Cov(E, U) for all strata as well as comparisons of null

and saturated models. At time 1, the models for each

stratum were similar in overall predictiveness (average

R

2 � 19% to 22%). However, by time 2 the R

2

for men

increased from 19.4% to 32.2%, and the model for

men explained 9% more variance than did the time 2

model for women not using HRT. The average R

2

val-

ues and the RMSs of all models were close to the

saturated models, except for the time 1 RMS in women

using HRT.

Prospective PLS Results: Predicting CHD in Men

In men, we were able to do prospective analyses

with time 1 LVs “predicting” new cases of CHD after



time 1. Figure 4 illustrates the models for men free of

CHD at time 1 (N � 39) for whom 11 new cases had

been diagnosed in the following 27 to 30 months.

Although vulnerability and resources were also used

in these analyses, for parsimony they are not de-

picted in the Figures. In Figure 4, top, significant

paths (R

2 � 9.9%) occurred from chronic stress to

distress to the MS (all at time 1) and then to new

CHD cases in the following 27 to 30 months. Al-

though a path occurred from distress to poor health

habits, the latter was not related to the MS at time 1.

In Figure 4, bottom, chronic stress, distress, and

poor health habits at time 1 and the MS at time 2

were used to explain and predict CHD 27 to 30

months after time 1. Here, a path occurred from

chronic stress to distress to health habits at study

entry, then to the MS assessed 15 to 18 months later,

and from the MS to CHD. Although all psychosocial

measures were assessed before CHD, three of the 11

new cases of CHD in men occurred before the MS at

Fig. 3. Path model that results from each stratum. Top, men; middle, women not using HRT; bottom, women using HRT. Bold line �significant path at both time 1 and time 2; dotted line � significant path at time 1; dashed line � significant path at time 2; no line

� no significant path at either times.



time 2. Hence, part of this analysis is not prospec-

tive, but as noted below, it is unlikely that the MS

followed the development of CHD.

DISCUSSION

Three groups of older adults were used to test a

theoretical stress model and to examine cross-sec-

tional and prospective interrelationships of chronic

stress, psychobehavioral constructs, the MS, and CHD.

Although some paths were consistent across men,

suggesting that women not using HRT had pathways

intermediate between men and women using HRT.

In contrast to men and women not using HRT, at

both times, women using HRT showed a complete

absence of pathways from distress to the MS and from

the MS to CHD (Figure 3, bottom). However, because

this study lacked random assignment to HRT, it is

unclear whether weak psychophysiological pathways

habits to the MS was large and significant at both times

(Table 3).

It is very important to note that the prospective

analyses of CHD incidence in men yielded results that

were quite different from the cross-sectional analyses.

These results suggest different interpretations than

those afforded by cross-sectional work. Cross-section-

ally, distress directly influenced variability in the MS,

Advances, Conclusions, and Implications

Despite these limitations, we believe this analysis

has advantages. The theoretical stress model chain of

pathways from psychosocial constructs to the MS and

on to CHD has rarely been examined in one sample. It

allowed us to assess the relative importance of vari-

ables relating chronic stress to psychophysiological

variables in older adults. In contrast to checklists of

life events, we studied a prototypic chronic stressor

(caregiving) that reflects real life experiences and has

important relevance to society. Moreover, caregivers

and noncaregivers were nondiabetic, thus yielding re-

sults unconfounded by this disease. By examining

men, women not using HRT, and women using HRT,

we were able to compare the potential importance of

gender and HRT in psychophysiological processes.

Men (N � 47) had the strongest relationship between

distress and the MS, and women using HRT (N � 41)

had the weakest relationship, with women not using

HRT (N � 64) between these extremes. By measuring

anger, hostility, and distress, we were able to show

that anger and hostility contributed to distress, but that

distress had a more direct relationship with metabo-

lism. This suggests that the interface of such effects

may be important to cardiovascular and metabolic dis-

regulation. In men, the pathway from distress to the

MS was one of the largest pathways from caregiving to

CHD. Hence, although caregiving may activate dis-

tress, the distress reaction may precipitate metabolic

reactions. This was seen when we reanalyzed the

model without chronic stress. The use of two time

points showed that relationships grew stronger over

time and that some variables may have lagged effects.

In pathways connecting chronic stress, distress, MS,

and CHD, 8 of 9 associations across the three strata

showed increases from time 1 to time 2, suggesting that

reactions to chronic stress may accumulate in older

adults. The cross-sectional and prospective results for

men suggest that distress may have immediate associ-

ations with both the MS and poor health habits, but

health habits may not have an immediate association

with the MS. Conversely, prior distress does not di-

rectly predict future MS, but it does predict metabolic

changes through alterations in health habits measured

15 to 18 months before the MS. The longitudinal de-

sign also allowed us to observe that the model pre-

dicted CHD in the 27- to 30-month interval after time 1

in men who had no record of CHD at time 1. Therefore,

it was less likely that relationships of distress and the

MS were due exclusively to preexisting CHD

problems.

In the absence of protective behaviors (eg, good diet

and exercise), even healthy older adults experience

changes that increase their vulnerability to CHD (3,

100), namely, greater insulin resistance from sedentary

behavior and greater adipose tissue (47, 101). Aging,

poor health habits, and chronic stress may jointly ex-

acerbate pathophysiology and lead to even greater

health risks, particularly if chronic stress and CHD

have been present for many years. Caregiving is a

situation of high demand, low control, and psycholog-

ical challenges. Such situations may trigger CHD

events and/or result in CHD progression (11). In this

study, by time 2, spouse caregivers had already pro-

vided full-time care for an average of 53 months, and

by the time medical records were obtained, they had

been caregiving for an average of 5.4 years (53 � 12 �65 months). In some of these caregivers, a physiologi-

cal load threshold may have been reached because in

only 27 to 30 months, the point prevalence of CHD

increased by 19% in caregivers and 8% in noncaregiv-

ers (33% in male caregivers and 13% in male noncar-

egivers). This is provocative because caregivers may be

unable to provide home care if they become ill, and in

response to this, society will incur tremendous costs.

In 1996, approximately 15 billion dollars was spent on

AD patients in nursing homes (102). Moreover, in

1996, for every extra month that persons with AD were

cared for in the community, $1.35 billion in institu-

tional costs of care nationwide were saved (102). Thus,

identifying individuals who are most vulnerable to the

ill effects of caregiving may be a first step to targeting

interventions with the greatest benefits, both from a

humanitarian and fiscal perspective. We hope that the

current work will begin to accomplish this goal.

Drs. Vitaliano, Scanlan, Savage, and Zhang weresupported by National Institute of Mental HealthGrant RO1 MH57663, National Institute on AgingGrant RO1 AG10760, and National Institutes ofHealth, Clinical Research Center Grant M01-RR00037;Dr. Siegler was supported by National Heart, Lung,and Blood Institute Grant RO1 HL55356, National In-stitute on Aging Grants 1RO1 AG12458 and RO1AG19605, and National Cancer Institute GrantPO1CA72099. We thank Drs. Karen Moe and NancyWoods for comments on the manuscript and RoslynSiegel for clerical support.

REFERENCES

1. Greenwood DC, Muir KR, Packham CJ, Madeley RJ. Coronary

heart disease: a review of the role of psychosocial stress and

social support. J Public Health Med 1996;18:221–31.

2. Pickering TG, Devereax RB, James GD, Gerin W, Landsbergis P,

Schnall PL, Schwartz JE. Environmental influences on blood

pressure and the role of job strain. J Hypertens 1996;14:179–85.

3. Schneider EL, Rowe JW, editors. Handbook of the biology of

aging. 3rd ed. San Diego (CA): Academic Press; 1990.



83. Lohmoeller JB. Latent variable path modeling with partial least

squares. New York: Springer-Verlag; 1989.

84. Chin WW, Newsted PR. Structural equation modeling analysis

with small samples using partial least squares. In: Hoyle RH,

editor. Statistical strategies for small sample research. Thou-

sand Oaks, (CA): Sage; 1999. p. 307–41.

85. Fornell C, Bookstein F. A comparative analysis of two struc-

tural equation models: LISREL and PLS applied to market data.

In: Fornell C, editor. A second general of multivariate analysis.

Vol. 1. New York: Praeger; 1982. p. 289–323.

86. Bansal A, Monnier J, Hobfoll SE, Stone B. Comparing men’s

and women’s loss of perceived social and work resources fol-

lowing psychological distress. J Soc Pers Relationships 2000;

17:265–81.

87. Riegel B, Gocka I. Gender differences in adjustment to acute

myocardial infarction. Heart Lung 1995;24:457–66.

88. Lovallo WR. Stress and health. Thousand Oaks (CA): Sage;

1997.

89. den-Tonkelaar I, Seidell JC, vanNoord PA, Baanders-van-

Halewijn EA, Ouwehand IJ. Fat distribution in relation to age,

degree of obesity, smoking habits, parity and estrogen use: a

cross-sectional study in 11,825 Dutch women participating in

the DOM-project. Int J Obesity 1990;14:753–61.

90. Menozzi R, Cagnacci A, Zanni AL, Bondi M, Volpe A, Del Rio

G. Sympathoadrenal response of postmenopausal women prior

and during prolonged administration of estradiol. Maturitas

2000;34:275–81.

91. Matthews KA, Wing RR, Kuller LH, Meilahn EN, Owens JF.

Menopause as a turning point in midlife. In: Manuck SB, Jen-

nings R, Rabin BS, Baum A, editors. Behavior, health and

aging. Mahwah (NJ): Lawrence Erlbaum Associates; 2000. p.

43–57.

92. Sung BH, Ching M, Izzo JL, Dandona P, Wilson MF. Estrogen

improves abnormal norepinephrine-induced vasoconstriction

in postmenopausal women. J Hypertens 1999;17:523–8.

93. Mitrakou A, Ryan C, Veneman T, Mokan M, Jenssen T, Kiss I,

Durrant J, Cryer P, Gerich J. Hierarchy of glycemic thresholds

for counterregulatory hormone secretion, symptoms, and cere-

bral dysfunction. Am J Physiol 1991;260:67–74.

94. Bjorntorp P. Metabolic implications of body fat distribution.

Diabetes Care 1991;14:1132–43.

95. Brindley DN, McCann BS, Niaura R, Stoney CM, Suarez EC.

Stress and lipoprotein metabolism: modulators and mecha-

nisms. Metabolism 1993;42:3–15.

96. Kawakami N, Araki S, Hayashi T, Masumoto T. Relationships

between perceived job-stress and glycosylated hemoglobin in

white-collar workers. Ind Health 1989;27:149–54.

97. Zamboni M, Armellini F, Turcato E, DePergola G, Todesco T,

Bissoli L, Bergamo-Andreis IA, Bosello O. Relationships be-

tween visceral fat, steroid hormones and insulin sensitivity in

premenopausal obese women. J Intern Med 1994;236:521–7.

98. Hinrichsen GA, Ramirez M. Black and white dementia

caregivers: a comparison of their adaptation, adjustment, and

service utilization. Gerontologist 1992;32:375–81.

99. Svec F, Nastasi K, Hilton C, Bao W, Srinivason SR, Berenson

GS. Black-white contrasts in insulin levels during pubertal

development: the Bogalusa Heart Study. Diabetes 1992;41:

313–7.

100. Bennett PH. Diabetes in the elderly: diagnosis and epidemiol-

ogy. Geriatrics 1984;39:37.

101. Goldfine ID. The insulin receptor: molecular biology and trans-

membrane signaling. Endocr Rev 1987;8:235.

102. Leon J, Cheng CK, Neumann PJ. Alzheimer’s disease care: costs

and potential savings. Health Aff 1998;17:206–16.



vitaliano et al 2002

Documents