Health Psychology

2010, Vol. 29, No. 4, 446 453

2010 American Psychological Association

0278-6133/10/$12.00 DOI: 10.1037/a0020061

Associations Between Adult Attachment Ratings and Health Conditions:

Evidence From the National Comorbidity Survey Replication
Lachlan A. McWilliams and S. Jeffrey Bailey
Acadia University
Objective: Attachment insecurity has been hypothesized to be a risk factor for the development of disease
and chronic illness. This study was the first to investigate associations between adult attachment ratings
and a wide range of health conditions. Design: Cross-sectional data from the National Comorbidity
Survey Replication (N 5645) were used. Measures: Participants completed Hazan and Shavers (1987)
measure of adult attachment and provided reports regarding 15 health conditions. Results: Logistic
regression analyses that adjusted for demographic variables indicated that avoidant attachment ratings
were positively associated with conditions defined primarily by pain (e.g., frequent or severe headaches).
Anxious attachment ratings were positively associated with a wider range of health conditions, including
several involving the cardiovascular system (i.e., stroke, heart attack, high blood pressure). Secure
attachment ratings were unrelated to the health conditions. Additional analyses investigated whether the
attachment ratings accounted for unique variance in the health conditions beyond that accounted for by
lifetime histories of depressive, anxiety, and alcohol- or substance-related disorders. In these analyses,
anxious attachment ratings continued to have significant positive associations with chronic pain, stroke,
heart attack, high blood pressure, and ulcers. Conclusion: The findings were generally supportive of the
theory that insecure attachment is a risk factor for the development of disease and chronic illness,
particularly conditions involving the cardiovascular system. Further research regarding the role of
attachment in the development of specific health conditions is warranted.
Keywords: attachment theory, cardiovascular disease, pain, ulcer, psychopathology

attachment, which involve ratings of particular attachment styles

(Hazan & Shaver, 1987), or scales assessing attachment dimensions (Bartholomew & Horowitz, 1991) thought to underlie attachment styles.
Attachment theory has increasingly been applied to understanding disease and chronic illness. Much of this work has been based
on a pathoplasty model (Clark, Watson, & Mineka, 1994) that
suggests that a variable (viz., insecure attachment) influences the
expression or course of a disorder. For example, using a mail
survey of patients with diabetes, Ciechanowski et al. (2004) found
evidence that self-reports of insecure attachment were associated
with poorer diabetes self-management (e.g., lower adherence to
recommendations related to diet, exercise, foot care, oral hypoglycemic medications, and smoking) and negative outcomes (e.g.,
elevated glycosylated hemoglobin levels). This research did not
address the question of causality but rather suggested that attachment influences coping responses to diabetes and possibly the
disorders severity. Similarly, in the context of chronic pain, ratings of insecure adult attachment have been found to be positively
associated with disability levels (McWilliams, Cox, & Enns, 2000)
and with depressive symptoms (Ciechanowski, Sullivan, Jensen,
Romano, & Summers, 2003; Meredith, Strong, & Feeney, 2007).
Predisposition models posit that vulnerability factors play a
causal role in the development of a disorder (Clark et al., 1994).
Related to this model, Maunder and Hunter (2001) delineated three
mechanisms that could lead those with insecure attachment to have
elevated rates of disease. First, those with insecure attachment
have an increased susceptibility to stress, such as the tendency to
perceive more stress and have more extreme physiological re-

Attachment theory (Bowlby, 1969) has been a major influence

in social and developmental research over the past few decades
(Roisman et al., 2007). In short, it posits an evolutionarily based
system designed to ensure infants maintain proximity to caregivers
during times of threat. On the basis of interactions with primary
caregivers, children are thought to develop enduring cognitive
schemas that continue into adulthood and guide behavior and
expectations in other relationships. Responsive caregiving tends to
result in the development of comfort with interpersonal closeness
and a willingness to depend on others, which is referred to as
secure attachment. Insecure attachment develops in response to
inconsistent or unresponsive caregiving and can involve a tendency to be anxious regarding possible rejection, discomfort with
close relationships, or both. Two approaches to adult attachment
research have evolved (Bartholomew & Shaver, 1998). The developmental approach has used the Adult Attachment Interview to
infer states of mind regarding childhood experiences with caregivers. The second approach was developed within social and personality research and uses self-report measures of attachmentrelated thoughts and feelings in adult relationships (e.g., degree of
anxiety and avoidance). Health-related research regarding attachment has relied almost exclusively on self-report measures of adult

Lachlan A. McWilliams and S. Jeffrey Bailey, Department of Psychology, Acadia University.

Correspondence concerning this article should be addressed to Lachlan
A. McWilliams, Department of Psychology, Acadia University, Wolfville,
Nova Scotia, Canada B4P 2R6. E-mail: Lachlan.Mcwilliams@acadiau.ca
446

ATTACHMENT AND HEALTH

sponses to stress. Second, those with insecure attachment have a

greater tendency to use external methods of regulating affect, such
as substance use and food consumption, which could lead to health
problems. Finally, those with insecure attachment have less effective help-seeking behavior, such as the underuse of social support
and difficulties using medical assistance effectively. This model
has growing support. For example, self-reports of attachment
avoidance have been found to be associated with altered autonomic functioning (Maunder, Lancee, Nolan, Hunter, & Tannenbaum, 2006). However, it is important to note that these mechanisms could play a role in both the development of a disorder and
adjustment to that disorder. For example, an unwillingness to
follow medical recommendations regarding diet and exercise
would be expected to influence the development of conditions,
such as heart disease and diabetes, and adjustment to these conditions.
Although a growing body of research clearly supports the possibility that insecure attachment could be a risk factor for a variety
of health conditions, a paucity of research has investigated the
hypothesis that those with insecure attachment actually experience
more disease or illness. One small study (Tacon, 2003) did find
that women with a history of breast cancer (N 52) had higher
scores on a measure of attachment avoidance relative to a control
sample of women with no history of breast cancer (N 52). Aside
from this study, no other published research exists regarding
associations between adult attachment and health conditions.
In this study, we used data from the National Comorbidity
Survey Replication (NCSR; Kessler & Merikangas, 2004) to
investigate associations between attachment and 15 health conditions. The NCSR was a large investigation of the prevalence and
correlates of psychiatric disorders in the general U.S. civilian
population. Most of the NCSR interview involved the administration of a comprehensive diagnostic interview based on criteria
of the fourth edition of the Diagnostic and Statistical Manual of
Mental Disorders (DSMIV; American Psychiatric Association,
1994). However, a large subsample used in the present study (N
5,692) was also administered a series of questions regarding lifetime histories of health problems and completed ratings of attachment. The attachment measure was based on Hazan and Shavers
(1987) original self-report measure of adult attachment that includes single-item ratings of secure, avoidant, and anxious attachment. The NCSR does have some methodological limitations,
such as a cross-sectional design and the use of brief measures of
attachment and health conditions. Nonetheless, it provides a rare
opportunity to investigate attachment and a wide range of health
conditions in a community sample representative of the U.S. adult
population.
Research with nonclinical samples has suggested that both attachment anxiety and avoidance may influence physiological responses to stress that could determine later health outcomes. For
example, Gallo and Matthews (2006) found that in conjunction
with particular social interactions, both forms of insecurity were
associated with ambulatory blood pressure readings obtained from
adolescents. During blood pressure monitoring periods involving
recent or current interactions with friends, attachment anxiety was
associated with augmented diastolic and systolic blood pressure.
During periods involving social conflict, avoidance was associated
with augmented diastolic blood pressure. However, some studies
have found only one type of insecure attachment to be related to

447

health-relevant variables. For example, lower natural killer cell

cytotoxicity was found to be associated with avoidance but not
with anxiety (Picardi et al., 2007). In light of these diverse findings, it was not possible to develop hypotheses regarding associations between specific attachment ratings and specific health
conditions, so we simply hypothesized that attachment insecurity
would be positively associated with the health conditions. Numerous studies have demonstrated that psychopathology is positively
associated with both ratings of insecure attachment (e.g., Mickelson, Kessler, & Shaver, 1997) and a wide range of health conditions (Scott et al., 2007). Thus, it is possible that associations
between attachment ratings and health conditions may simply be
the result of the variance these variables share with psychopathology. A secondary goal of this study was to determine whether
ratings of adult attachment account for unique variance in health
conditions beyond that contributed by several forms of psychopathology.

Method
Participants and Procedures
We used Part 2 of the NCSR public-use data set for this study.
The NCSR is a probability sample of the U.S. noninstitutionalized civilian population, age 18 or older, and reported a response
rate of 70.9%. Administration of the interview consisted of two
parts. Part 1 consisted primarily of a diagnostic assessment of
psychiatric disorders and was administered to 9,282 participants.
Part 2 included questions regarding potential risk factors and was
administered to 5,692 respondents. The second group of respondents included all of the Part 1 respondents who reported a lifetime
disorder along with a probability subsample of the others. A
weighting procedure was used to adjust for differential probabilities of selection and nonresponse and to adjust the sample to reflect
the U.S. populations demographics (sex, race, marital status,
education, living arrangements, region, urbanicity). All analyses
conducted in this study are based on Part 2 data and used these
weights. Ethical approval for the primary data collection of the NCS
was provided by the University of Michigan. Further details of the
NCSR methodology are available in Kessler, Berglund, et al. (2004).
The interview materials described in the next section are available at
http://www.hcp.med.harvard.edu/ncs/replication.php.

Measures
Demographic variables. Participants provided information
regarding their demographic characteristics. This information was
used to calculated odds ratios that adjusted for gender, marital
status (e.g., never married, married, or divorced, separated, or
widowed), race (e.g., Caucasian, Hispanic, Black, or other), age
(e.g., 18 29, 30 44, 4559, or 60 and above), and education level
(0 11 years, 12 years, 1315 years, 16-plus years).
Attachment style. Adult attachment style was measured by
presenting Hazan and Shavers (1987) attachment style measure in an
interview format. It included brief statements describing adult versions of the three attachment styles (secure, avoidant, and anxious)
originally identified in children (Ainsworth, Blehar, Waters, & Wall,
1978). Secure attachment was assessed with this statement: I find
it relatively easy to get close to other people. I am comfortable

MCWILLIAMS AND BAILEY

448

depending on others and having them depend on me. I dont worry

about being abandoned or about someone getting too close to me.
Avoidant attachment was assessed with the statement I am somewhat uncomfortable being close to others; I find it difficult to trust
them completely and difficult to depend on them. I am nervous
when anyone gets too close to me. Anxious attachment was
assessed with the statement I find that others are reluctant to get
as close as I would like. I often worry that people who I care about
do not love me or wont want to stay with me. I want to merge
completely with another person, and this desire sometimes scares
people away. Similar to Shaver and Brennans (1992) modification of this measure, each respondent provided a self-rating on
each of these statements using a 4-point scale ranging from 1 (not
at all like me) to 4 (a lot like me).
Chronic health conditions. Lifetime experiences with
chronic conditions were measured using a yesno format. Seven
questions assessed the participants in terms of lifetime histories of
arthritis, chronic back or neck problems, frequent or severe headaches, other forms of chronic pain, seasonal allergies, stroke, and
heart attack. The remaining eight questions asked whether participants had ever been told by their doctor that they had the health
problem (i.e., Did a doctor or other health professional ever tell
you that you had any of the following illnesses: heart disease?).
These questions assessed the participants in terms of lifetime
histories of heart disease, high blood pressure, asthma, chronic
lung disease, diabetes or high blood sugar, ulcers, epilepsy or
seizures, and cancer.
Psychiatric disorders. The World Health Organization Composite International Diagnostic Interview (CIDI; Robins et al.,
1988) was used to assess a wide range of DSMIV (American
Psychiatric Association, 1994) disorders. The CIDI is a fully
structured lay-administered diagnostic interview. In this study, we
used composite variables representing lifetime histories of depressive disorders (major depressive disorder and dysthymia), anxiety

disorders (generalized anxiety disorder, panic disorder, agoraphobia, social phobia, simple phobia, and posttraumatic stress disorder), and alcohol- and substance-related disorders (alcohol abuse,
alcohol dependence, substance abuse, and substance dependence).
A clinical reappraisal study supportive of the CIDIs validity
(Kessler, Abelson et al., 2004) found good concordance between
CIDI diagnoses and the research nonpatient version of the Structured Clinical Interview for DSMIV (First, Spitzer, Williams, &
Gibbon, 1995).

Results
Because of the complex sample design and weighting, all analyses were calculated using STATA (StataCorp, College Station,
TX), which used the Taylor series linearization method and the
stratification and weighting information available in the NCSR
data to factor design effects into its variance estimates. Given the
large number of analyses, we used a critical p value of .01
(two-tailed) to reduce the chance of Type I error.
A small number of participants (n 47) did not complete each
attachment rating and were excluded from all analyses, resulting in
a final sample of 5,645. Secure attachment was rated most strongly
by the participants (M 2.90, 95% CI [2.86 2.93]), followed by
avoidant attachment (M 1.77, 95% CI [1.74 1.81]) and anxious
attachment (M 1.31, 95% CI [1.28 1.33]). Secure ratings were
negatively associated with the avoidant (r .29, p .001) and
anxious (r .07, p .001) ratings. The two insecure ratings
were positively associated with each other (r .34, p .001). The
number of participants reporting each of the 15 health conditions
are also reported in Table 1. The most common conditions were
seasonal allergies (37.6%), chronic back or neck problems
(29.3%), arthritis (27.%), and high blood pressure (24.0%).
We used logistic regression analyses to determine whether the
attachment ratings were associated with each health condition. The

Table 1
Bivariate and Multivariate Associations Between Attachment Ratings and Health Conditions (N 5645)
Secure
a

Health condition

Arthritis (1,531)
Back or neck problem
(1,651)
Headaches (1,275)
Other chronic pain (546)
Allergies (2,127)
Stroke (150)
Heart attack (207)
Heart disease (281)
High blood pressure (1,349)
Asthma (653)
Lung disease (124)
High blood sugar (406)
Ulcer (526)
Seizures (99)
Cancer (371)

OR

Avoidant
Adjusted OR

0.94 [0.871.03]

0.89 [0.830.96]
0.88 [0.810.95]
0.97 [0.881.07]
1.02 [0.961.08]
0.86 [0.721.03]
1.03 [0.831.28]
1.00 [0.831.19]
1.00 [0.921.08]
1.03 [0.941.13]
0.92 [0.741.14]
0.89 [0.771.02]
0.90 [0.811.01]
0.89 [0.741.07]
1.22 [1.111.33]

0.94 [0.851.03]
0.91 [0.840.98]
0.91 [0.831.00]
1.02 [0.911.14]
1.00 [0.931.08]
0.86 [0.711.05]
1.06 [0.861.31]
0.97 [0.801.19]
1.00 [0.911.10]
1.10 [1.001.20]
0.97 [0.811.16]
0.89 [0.781.02]
0.91 [0.801.02]
0.98 [0.781.22]
1.13 [1.011.27]

OR

Anxious

Adjusted OR

1.10 [1.001.21]

1.19 [1.101.28]
1.23 [1.161.31]
1.26 [1.141.38]
1.07 [0.991.16]
1.05 [0.861.29]
1.20 [1.001.43]
1.12 [0.971.28]
1.06 [0.981.15]
1.12 [1.011.25]
1.17 [0.971.40]
1.06 [0.911.24]
1.17 [1.081.27]
1.21 [0.971.52]
0.96 [0.821.11]

OR

1.15 [1.041.28]

1.16 [1.081.25]
1.14 [1.041.24]
1.19 [1.071.32]
1.10 [1.001.21]
0.93 [0.741.17]
1.12 [0.921.36]
1.17 [0.991.38]
1.06 [0.961.18]
1.09 [0.981.22]
1.06 [0.841.32]
1.03 [0.891.20]
1.06 [0.951.17]
1.10 [0.871.38]
1.08 [0.911.27]

Adjusted OR

1.06 [0.961.18]

1.13 [1.031.24]
1.32 [1.191.47]
1.39 [1.231.56]
0.94 [0.851.04]
1.41 [1.111.80]
1.43 [1.101.87]
1.03 [0.881.20]
1.19 [1.071.32]
1.20 [1.031.41]
1.39 [1.081.79]
1.14 [0.951.38]
1.36 [1.231.50]
1.27 [1.021.59]
0.93 [0.781.10]

1.07 [0.931.23]
1.08 [0.981.19]
1.27 [1.101.47]
1.31 [1.141.51]
0.96 [0.851.08]
1.55 [1.172.05]
1.47 [1.131.91]
1.05 [0.851.32]
1.29 [1.131.46]
1.13 [0.951.33]
1.32 [0.941.86]
1.17 [0.941.47]
1.35 [1.171.54]
1.11 [0.881.39]
1.06 [0.861.31]

Note. Odds ratios (OR) were adjusted for gender, marital status, education level, race, age, and the other attachment style ratings; 95% confidence intervals
are included in brackets.
a
Numbers in parentheses indicate the number of participants reporting each health condition.

p .01. p .001.

ATTACHMENT AND HEALTH

first set of analyses investigated bivariate associations between

each attachment rating and each health condition. The odds ratios
from these analyses are presented in Table 1. Eight of the health
conditions were associated with at least one attachment variable.
Ratings of secure attachment were positively associated with cancer and negatively associated with chronic back or neck problems
and headaches. Ratings of avoidant attachment were positively
associated with chronic back and neck problems, frequent or
severe headaches, other forms of chronic pain, and ulcers. Ratings
of anxious attachment were positively associated with chronic
back or neck problems, frequent or severe headaches, other forms
of chronic pain, stroke, heart attack, high blood pressure, and
ulcers.
The second set of analyses adjusted for (i.e., included as independent variables) demographic variables including gender, marital status, race, age, and education level. The demographic variables selected were similar to those used in numerous previous
studies using the NCSR data set. The second set of analyses also
simultaneously entered the three attachment ratings into the logistic regression analyses. This approach was designed to investigate
whether the attachment ratings could account for unique variance
in the health conditions after accounting for (a) demographic
variables that could be confounded with either attachment ratings
or the health conditions and (b) responses to the other attachment
ratings. The odds ratios from these analyses are also presented in
Table 1. In these multivariate analyses, eight of the health conditions were associated with at least one attachment variable. Ratings
of secure attachment were not significantly associated with any of
the health conditions. Ratings of avoidant attachment were positively associated with arthritis, back and neck problems, headaches, and other chronic pain. Ratings of anxious attachment were
positively associated with headaches, other chronic pain, stroke,
heart attack, high blood pressure, and ulcer.
In several cases, the findings of the multivariate analyses were
inconsistent with those of the bivariate analyses. In most cases,
associations that were initially significant were no longer significant after adjusting for the demographic variables (i.e., the association between anxious attachment ratings and back or neck pain).
The association between avoidant attachment ratings and arthritis
was the only one in which a statistically significant finding was
obtained after adjusting for the demographic variables. Detailed
examination of the findings from this multivariate logistic regression analysis indicated that women were more likely to report
arthritis (OR 1.41, 95% CI [1.16 1.73], p .001) and that age,
marital status, and education level were also related to arthritis. In
terms of age, the youngest group was treated as the reference
group. Relative to the 18- to 29-year-olds, those in the 30 44 age
group (OR 2.36, 95% CI [1.70 3.30], p .001), the 4559
age group (OR 8.72, 95% CI [5.60 13.56], p .001), and the
60 and older age group (OR 16.08, 95% CI [10.29 25.12], p
.001) were all found to have a significantly increased adjusted odds
of having arthritis. For marital status, the married group was
treated as the reference category and those who were divorced,
separated, or widowed were more likely to report arthritis (OR
1.38, 95% CI [1.111.73], p .005). For education, those with 12
years of education were the reference group. Fewer than 12 years
of education was positively associated with arthritis (OR 1.50,
95% CI [1.19 1.91], p .001) and 16 or more years of education
was negatively associated with arthritis (OR .58, 95% CI

449

[0.47 0.70], p .001). As noted earlier and reported in Table 1,

avoidant attachment ratings were also associated with arthritis.
This pattern of findings indicates that an association between
ratings of avoidant attachment and arthritis exists but that it is only
observable after accounting for several demographic variables also
associated with arthritis.
Also noteworthy were that the bivariate findings regarding
cancer were in the direction opposite to expectations. In the case of
the insecure ratings, these findings were not statistically significant. However, secure attachment did have a positive and statistically significant association with cancer. After adjusting for the
demographic variables, this positive association was only marginally significant (OR 1.13, 95% CI [1.011.27], p .034).
Because this finding was so counterintuitive, we explored it in
more detail. In the multivariate analyses, we found age, race, and
marital status to be associated with cancer. Relative to the 18 29
age group, being in the 30 44 age group had a marginally significant association with cancer (OR 2.37, 95% CI [1.19 4.71],
p .016). The 4559 age group (OR 4.99, 95% CI [2.44
10.23], p .001) and the 60 and older group (OR 15.88, 95%
CI [7.3734.20], p .001) both had a positive association with
cancer. In terms of race, the Caucasian group was treated as the
reference category and Black race was found to be negatively
associated with cancer (OR 0.38, 95% CI [0.21 0.71], p .01).
As well, relative to the married group, never being married had a
marginal negative association with cancer (OR 0.44, 95% CI
[0.23 0.84], p .014). It is possible that the initial significant
positive association between secure attachment ratings and cancer
was the result of a confound involving one or more of these
demographic variables. Consistent with this possibility, secure
attachment ratings had a small but significant positive association
with age (r .06, p .01) and a negative association with Black
race (B .35, p .001) and never-married marital status (B
.17, p .01).
The second aim of the study was to determine whether the
attachment ratings could account for unique variance in the health
conditions beyond that accounted for by psychopathology. As a
preliminary step, we investigated associations between each attachment rating and three forms of psychopathology. These findings are presented in Table 2. Secure attachment ratings were
negatively associated with lifetime history of depressive disorders,
anxiety disorders, and alcohol- and substance-related disorders. As
expected, the two insecure attachment ratings had positive associations with each of these forms of psychopathology.

Table 2
Associations (Odds Ratios and 95% Confidence Intervals)
Between Attachment Ratings and Three Types of Lifetime
Psychiatric Disorders
Type of psychiatric disorder
Attachment
ratings

Depressive

Anxiety

Alcohol and
substance

Secure
Avoidant
Anxious

0.86 [0.800.93]
1.48 [1.381.58]
1.49 [1.331.66]

0.80 [0.740.86]
1.68 [1.561.81]
1.67 [1.491.88]

0.82 [0.740.91]
1.51 [1.411.63]
1.54 [1.411.68]

Note.

All odds ratios were significant at p .001.

MCWILLIAMS AND BAILEY

450

Given that the attachment ratings shared variance with each

form of psychopathology, it was appropriate to determine whether
the attachment ratings could account for unique variance in the
health conditions. We used additional multivariate analyses to
investigate this possibility. These analyses adjusted for the same
demographic variables as the earlier multivariate analyses and
included three variables representing lifetime histories of depressive disorders, anxiety disorders, and alcohol- and substancerelated disorders. The dependent variables were the conditions
positively associated with the attachment ratings in the earlier
multivariate analyses. The first set of these analyses included the
pain-related conditions and are reported in Table 3. The depressive
and anxiety disorder variables were positively associated with each
pain condition. The alcohol- and substance-related disorders variable was only associated with chronic back or neck problems. In
the earlier analyses, avoidant attachment ratings, anxious attachment ratings, or both were positively associated with the pain
conditions, but in these analyses the only significant finding regarding attachment was the positive association between anxious
attachment and other forms of chronic pain.
The second set of analyses adjusting for lifetime psychiatric
disorders included three health conditions involving the cardiovascular system as the dependent variables. These findings are reported in Table 4. Lifetime history of an anxiety disorder had a
significant positive association with high blood pressure, but the
psychiatric disorder variables were not otherwise significantly
associated with the cardiovascular conditions. In the earlier multivariate analyses, ratings of anxious attachment had significant
positive associations with stroke, heart attack, and high blood
pressure, and these association remained significant in these analyses.
The final analysis included ulcers as the dependent variable.
Depressive disorders (OR 1.36, 95% CI [1.131.63], p .01),
anxiety disorders (OR 1.55, 95% CI [1.212.00], p .01), and
alcohol- or substance-related disorders (OR 1.50, 95% CI [1.13
1.99], p .01) were all positively associated with ulcers. Anxious
attachment ratings were positively associated with ulcers in the
earlier analyses and continued to be associated with ulcers in the
final analysis (OR 1.27, 95% CI [1.121.44], p .001). Also
consistent with the earlier analyses, secure (OR .92, 95% CI
[.811.04], ns) and avoidant (OR .98, 95% CI [.88 1.10], ns)
attachment ratings were not significantly associated with ulcers.

Discussion
This study is the first to investigate associations between adult
attachment ratings and a wide range of health conditions. Consistent with the studys general hypothesis, both the bivariate and the
initial multivariate logistic regression analyses indicated that insecure attachment was positively associated with approximately half
of the health conditions investigated.
We made no specific hypotheses regarding associations between
the attachment ratings and particular health conditions. Relative to
avoidant attachment, anxious attachment ratings were more
strongly associated with poor health because they were associated
with more health conditions, and these associations were generally
larger than those involving avoidant attachment ratings. Beyond
this, two other trends emerged. First, when considering the initial
multivariate analyses, avoidant attachment ratings were significantly associated with only those conditions that primarily involve
symptoms of pain (i.e., arthritis, back or neck problems, frequent
or severe headaches, and other forms of chronic pain). In contrast,
anxious attachment ratings were associated with a wider range of
conditions. Second, anxious attachment ratings were associated
with several conditions involving the cardiovascular system, including stroke, heart attack, and high blood pressure.
Previous research has already found insecure attachment ratings
to be positively associated with physical symptoms (Feeney,
2000). The findings of this study go beyond these earlier findings
by considering specific health conditions such as stroke, heart
attack, high blood pressure, and seizures. However, it is important
to note that some of the health conditions (e.g., back or neck
problems) in this study overlap with those included in the physical
symptom ratings used in earlier studies. As well, in some cases the
distinction between a condition and symptoms is not clear. For
example, frequent or severe headaches could have been endorsed by individuals with specific conditions, such as migraine,
and by those who subjectively rated their headaches as severe or
frequent.
A second aim of this study was to determine whether ratings of
adult attachment account for unique variance in health conditions
beyond that contributed by three forms of psychopathology. In the
initial multivariate analyses regarding the four pain-related conditions, six significant positive associations involved insecure ratings
of attachment. In the analyses that adjusted for lifetime histories of
psychiatric disorders, only the association between anxious attach-

Table 3
Multivariate Odds Ratios (and 95% Confidence Intervals) From Logistic Regression Models Predicting Pain Conditions
Pain conditions
a

Select independent variables

Arthritis

Back and neck problem

Headaches

Other chronic pain

Depressive disorders
Anxiety disorders
Alcohol- or substance-related disorders
Secure attachment
Avoidant attachment
Anxious attachment

1.28 [1.071.53]
1.55 [1.271.89]
1.18 [0.931.51]
0.95 [0.861.05]
1.09 [0.991.21]
1.02 [0.891.16]

1.63 [1.431.84]
1.61 [1.411.83]
1.50 [1.261.79]
0.92 [0.851.00]
1.07 [0.991.16]
1.01 [0.921.10]

1.98 [1.712.28]
1.81 [1.492.19]
1.15 [0.951.40]
0.94 [0.851.03]
1.04 [0.941.14]
1.17 [1.031.34]

1.88 [1.492.38]
1.63 [1.312.02]
1.37 [1.011.86]
1.04 [0.931.17]
1.09 [0.971.21]
1.22 [1.061.40]

All analyses included independent variables to adjust for gender, marital status, education level, race, and age.
p .01. p .001.

ATTACHMENT AND HEALTH

451

Table 4
Multivariate Odds Ratios (and 95% Confidence Intervals) From Logistic Regression Models
Predicting Cardiovascular Conditions
Cardiovascular conditions
Select independent variablesa

Stroke

Heart attack

High blood pressure

Depressive disorders
Anxiety disorders
Alcohol- or substance-related disorders
Secure attachment
Avoidant attachment
Anxious attachment

1.34 [0.971.87]
1.45 [0.992.11]
1.07 [0.601.90]
0.86 [0.711.05]
0.88 [0.701.11]
1.49 [1.111.99]

1.00 [0.671.48]
1.20 [0.821.77]
1.55 [1.022.35]
1.07 [0.861.32]
1.08 [0.881.33]
1.45 [1.111.91]

1.16 [0.911.48]
1.44 [1.231.69]
0.98 [0.861.13]
1.01 [0.921.11]
1.03 [0.921.14]
1.24 [1.091.41]

All analyses included independent variables to adjust for gender, marital status, education level, race, and age.
p .01. p .001.

ment and other forms of chronic pain remained statistically significant. These findings raise the possibility that psychopathology,
particularly depressive and anxiety disorders, may mediate the
relationship between insecure attachment and pain conditions.
Given evidence that in many cases psychopathology develops
subsequent to the onset of chronic pain (Fishbain, Cutler, Rosomoff, & Rosomoff, 1997) and the cross-sectional nature of the
data, we did not test mediational models regarding attachment,
psychopathology, and the pain conditions.
Neuroticism is a personality construct that refers to the tendency
to experience negative affect. It is positively associated with exaggerated reporting of physical symptoms (Johnson, 2003), perceptions of poor health (Goodwin & Engstrom, 2002), and numerous health conditions included in the original NCS (Goodwin,
Cox, & Clara, 2006). Ratings of avoidant and anxious attachment
are also moderately associated with neuroticism (Mickelson et al.,
1997; Shaver & Brennan, 1992). Given this overlap, it is possible
that the associations between insecure attachment ratings and some
of the health conditions could have been inflated because of a bias
toward reporting or identifying symptoms. This issue is particularly relevant to the conditions defined by symptoms of pain.
Unfortunately, the NCSR did not include a measure of neuroticism, so we could not directly examine this possibility. However,
neuroticism overlaps conceptually and empirically with the psychiatric disorders included in this study (see Enns & Cox, 1997),
so adjusting for psychiatric disorders in this study could arguably
be considered similar to adjusting for neuroticism. Thus, the initial
positive associations between the insecure attachment ratings and
the pain-related conditions may have been the result of their shared
variance with neuroticism. Consistent with this possibility, all but
one of the significant associations between insecure attachment
ratings and the pain-related conditions were no longer significant
after adjusting for the psychopathology variables.
The findings regarding the cardiovascular-related conditions
were much different than those obtained regarding the pain-related
conditions. First, the psychopathology variables were, with one
exception, unrelated to the cardiovascular conditions. Second, the
associations between anxious attachment ratings and each of these
conditions remained significant after adjusting for psychopathology. Thus, anxious attachment had robust associations with these
conditions that cannot be accounted for by shared variance with
psychopathology. As noted earlier, anxious attachment is associ-

ated with elevated blood pressure during social interactions (Gallo

& Matthews, 2006). It is possible that our findings regarding the
cardiovascular-related conditions reflect the long-term health consequences of this relationship.
After ulcers were found to involve infectious disease, there was
a rapid decline in research regarding the psychological aspects of
this condition (Levenstein, 2000). However, several recent studies
with large community samples (e.g., Goodwin, Keyes, Stein, &
Talley, 2009; Goodwin & Stein, 2002) have found psychiatric
disorders to be positively associated with ulcers. The present
findings are consistent with this research and suggest that anxious
attachment may also be an additional risk factor for ulcers. Levenstein (2000) suggested several psychophysical processes that
could be involved in the development of ulcers, such as the effects
of hypothalamicpituitaryadrenal axis activation on healing, that
bear similarity to the heightened physiological response to the
stress mechanism that Maunder and Hunter (2001) hypothesized as
a link between insecure attachment and poor health.
The studys cross-sectional design is one of its main limitations
because it precluded the directions of the associations from being
investigated. Although insecure attachment is theorized to lead to
the onset of various health conditions, it is certainly possible that
the experience of poor health could lead to relationship distress
and heightened relationship insecurity. The cross-sectional design
also has implications for interpreting several of the unexpected
findings. Although not a focus of the study, the psychopathology
variables were largely unrelated to the cardiovascular-related conditions. Previous longitudinal research has found positive associations between psychopathology and cardiovascular events including stroke (Jonas & Mussolino, 2000; Larson, Owens, Ford, &
Eaton, 2001) and heart attacks (Rutledge et al., 2009). These
studies captured incident cases including those resulting in death.
In contrast, the cases captured in this study included only the
survivors of such events who were able to participate in a community survey. Thus, a selection bias related to survival of these
conditions may be responsible for the unexpected null findings
regarding psychopathology and the cardiovascular-related conditions. It is also possible that the associations between the attachment ratings and the health conditions with high rates of mortality
could have been attenuated as a result of such a selection bias.
However, prospective studies would be required to clarify the

MCWILLIAMS AND BAILEY

452

relationships among attachment, incident health conditions, survival, and adjustment to such health conditions.
One finding was opposite to what was hypothesized. In the
bivariate analysis, secure attachment was positively associated
with cancer. This positive finding became marginally significant
( p .034) in the multivariate analyses that adjusted numerous
demographic variables. Given that this positive association was
only marginally significant, not overinterpreting it is important.
Nonetheless, this positive association may reflect a form of posttraumatic growth among those who have experienced cancer (see
Costanzo, Ryff, & Singer, 2009). Alternatively, attachment may be
unrelated to the development of cancer, but a positive association
between secure attachment ratings and cancer could have emerged
because those with secure attachment may be more likely to
survive cancer and be captured in a cross-sectional survey of the
general population. This explanation has strong potential because
psychological distress is associated with insecure attachment
(Mickelson et al., 1997) and appears to be a predictor of cancer
prognosis (i.e., cancer mortality among those with a current cancer
diagnosis or a history of cancer) rather than of incident cases of
cancer (see Brown, Levy, Rosberger, & Edgar, 2003; Hamer,
Chida, & Molloy, 2009).
We should note two other limitations. Self-reports of physician
diagnoses are appropriate for conditions that can only be identified
with diagnostic procedures, but they also pose methodological
problems because there is evidence that diagnostic testing is influenced by patients personality. For example, associations between neuroticism and self-reported high blood pressure appear to
be the result of a confounding between neuroticism and exposure
to medical attention that can increase the detection and awareness
of this condition (Irvine, Garner, Olmstead, & Logan, 1989). Thus,
it is possible that a similar process is responsible for the present
finding of an association between anxious attachment and high
blood pressure. A final limitation of the study is that attachment
was assessed with ratings of three attachment styles. Theory (Bartholomew & Horowitz, 1991) and research (Fraley & Waller,
1998) have indicated that attachment is most appropriately conceptualized as involving two dimensions (anxiety or model of self
and avoidance or model of others). The use of continuous measures of these two dimensions would have been ideal, but such
measures were not available in the NCSR. Despite the limitations
noted here, the findings provide important preliminary support for
Maunder and Hunters (2001) model and indicate that further
research examining the role of insecure attachment in the development of health conditions is warranted. The findings of earlier
research on attachment and physiological responses to stress
(Gallo & Matthews, 2006) and the present findings suggest that
research regarding the role of anxious attachment in the development of cardiovascular-related conditions holds particular promise.

References
Ainsworth, M. D. S., Blehar, M. C., Waters, E., & Wall, S. (1978). Patterns
of attachment: A psychological study of the strange situation. Hillsdale,
NJ: Erlbaum.
American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington, DC: Author.
Bartholomew, K., & Horowitz, L. M. (1991). Attachment styles among
young adults: A test of a four-category mode. Journal of Personality and
Social Psychology, 61, 226 244.

Bartholomew, K., & Shaver, P. R. (1998). Methods of assessing adult

attachment: Do they converge? In J. A. Simpson & W. S. Rholes (Eds.),
Attachment theory and close relationships (pp. 25 45). New York, NY:
Guilford Press.
Bowlby, J. (1969). Attachment and loss. London, England: Penguin Books.
Brown, K. W., Levy, A. R., Rosberger, Z., & Edgar, L. (2003). Psychological distress and cancer survival: A follow-up 10 years after diagnosis. Psychosomatic Medicine, 65, 636 643.
Ciechanowski, P., Russo, J., Katon, W., Von Korff, M., Ludman, E., Lin,
E., . . . Bush, T., (2004). Influence of patient attachment style on selfcare and outcomes in diabetes. Psychosomatic Medicine, 66, 720 728.
Ciechanowski, P., Sullivan, M., Jensen, M., Romano, J., & Summers, H.
(2003). The relationship of attachment style to depression, catastrophizing, and health care utilization in patients with chronic pain. Pain, 104,
627 637.
Clark, L. A., Watson, D., & Mineka, S. (1994). Temperament, personality,
and the mood and anxiety disorders. Journal of Abnormal Psychology,
103, 103116.
Costanzo, E. S., Ryff, C. D., & Singer, B. H. (2009). Psychosocial adjustment among cancer survivors: Findings from a national survey of health
and well-being. Health Psychology, 28, 147156.
Enns, M. W., & Cox, B. J. (1997). Personality dimensions and depression:
Review and commentary. Canadian Journal of Psychiatry, 42, 274 284.
Feeney, J. A. (2000). Implications of attachment style for patterns of health
and illness. Child: Care, Health and Development, 26, 277288.
First, M. B., Spitzer, R. L., Williams, J. B. W., & Gibbon, M. (1995).
Structured clinical interview for DSMIV (SCID). Washington, DC:
American Psychiatric Press.
Fishbain, D., Cutler, R., Rosomoff, H., & Rosomoff, R. (1997). Chronic
pain associated depression: Antecedent or consequence of chronic pain?
A review. Clinical Journal of Pain, 13, 116 137.
Fraley, R. C., & Waller, N. G. (1998). Adult attachment patterns: A test of
the typological model. In J. A. Simpson & W. S. Rholes (Eds.), Attachment theory and close relationships (pp. 25 45). New York, NY:
Guilford Press.
Gallo, L. C., & Matthews, K. A. (2006). Adolescents attachment orientation influences ambulatory blood pressure responses to everyday social
interactions. Psychosomatic Medicine, 68, 253161.
Goodwin, R. D., Cox, B. J., & Clara, I. (2006). Neuroticism and physical
disorders among adults in the community: Results from the National
Comorbidity Survey. Journal of Behavioral Medicine, 29, 229 238.
Goodwin, R., & Engstrom, G. (2002). Personality and the perception of
health in the general population. Psychological Medicine, 32, 325332.
Goodwin, R. D., Keyes, K. M., Stein, M. B., & Talley, N. J. (2009). Peptic
ulcer and mental disorders among adults in the community: The role of
nicotine and alcohol use disorders. Psychosomatic Medicine, 71, 463
468.
Goodwin, R. D., & Stein, M. B. (2002). Generalized anxiety disorder and
peptic ulcers among adults in the United States. Psychosomatic Medicine, 64, 862 866.
Hamer, M., Chida, Y., & Molloy, G. J. (2009). Psychological distress and
cancer mortality. Journal of Psychosomatic Research, 66, 255258.
Hazan, C., & Shaver, P. (1987). Romantic love conceptualized as an
attachment process. Journal of Personality and Social Psychology, 52,
511524.
Irvine, M. J., Garner, D. M., Olmstead, M. P., & Logan, A. G. (1989).
Personality differences between hypertensive and normotensive individuals: Influence of knowledge of hypertension status. Psychosomatic
Medicine, 51, 537549.
Johnson, M. (2003). The vulnerability status of neuroticism: Overreporting or genuine complaints? Personality and Individual Differences, 35, 877 887.
Jonas, B. S., & Mussolino, M. E. (2000). Symptoms of depression as a

ATTACHMENT AND HEALTH

prospective risk factor for stroke. Psychosomatic Medicine, 62, 463
471.
Kessler, R. C., Abelson, J., Demler, O., Escobar, J. I., Gibbon, M., Guyer,
M. E., . . . Zheng, H. (2004). Clinical calibration of DSMIV diagnoses
in the World Mental Health (WMH) version of the World Health
Organization (WHO) Composite International Diagnostic Interview
(WMH-CIDI). International Journal of Methods in Psychiatry Research, 13, 122139.
Kessler, R. C., Berglund, P., Chiu, W. T., Demler, O., Heeringa, S., Hiripi,
E., . . . Zheng, H. (2004). The US National Comorbidity Survey replication (NCS-R): Design and field procedures. International Journal of
Methods in Psychiatric Research, 13, 69 92.
Kessler, R. C., & Merikangas, K. R. (2004). The National Comorbidity
Survey replication (NCS-R): Background and aims. International Journal of Methods in Psychiatric Research, 13, 60 68.
Larson, S. L., Owens, P. L., Ford, D., & Eaton, W. (2001). Depressive
disorder, dysthymia, and risk of stroke: Thirteen-year follow-up from the
Baltimore epidemiological catchment area study. Stroke, 32, 1979 1983.
Levenstein, S. (2000). The very model of a modern etiology: A biopsychosocial view of peptic ulcer. Psychosomatic Medicine, 62, 176 185.
Maunder, R. G., & Hunter, J. J. (2001). Attachment and psychosomatic
medicine: Developmental contributions to stress and disease. Psychosomatic Medicine, 63, 556 567.
Maunder, R. G., Lancee, W. J., Nolan, R. P., Hunter, J. J., & Tannenbaum,
D. W. (2006). The relationship of attachment insecurity to subjective
stress and autonomic function during standardized acute stress in healthy
adults. Journal of Psychosomatic Research, 60, 283290.
McWilliams, L. A., Cox, B. J., & Enns, M. W. (2000). Impact of adult
attachment styles on pain and disability associated with arthritis in a
nationally representative sample. Clinical Journal of Pain, 16, 360 364.
Meredith, P. J., Strong, J., & Feeney, J. A. (2007). Adult attachment

453

variables predict depression before and after treatment for chronic pain.
European Journal of Pain, 11, 164 170.
Mickelson, K. D., Kessler, R. C., & Shaver, P. R. (1997). Adult attachment
in a nationally representative sample. Journal of Personality and Social
Psychology, 73, 10921106.
Picardi, A., Battisti, F., Tarsitani, L., Baldassari, M., Copertaro, A., Mocchegiani, E., & Boindi, M. (2007). Attachment security and immunity in
healthy women. Psychosomatic Medicine, 69, 40 46.
Robins, L. N., Wing, J., Wittchen, H. U., Helzer, J. E., Babor, T. F., Burke,
J., . . . Towle, L. H. (1988). The Composite International Diagnostic
Interview: An epidemiologic instrument suitable for use in conjunction
with different diagnostic systems and in different cultures. Archives of
General Psychiatry, 45, 1069 1077.
Roisman, G. I., Holland, A., Fortuna, K., Fraley, C., Clausell, E., & Clarke,
A. (2007). The Adult Attachment Interview and self-reports of attachment style: An empirical rapprochement. Journal of Personality and
Social Psychology, 92, 678 697.
Rutledge, T., Linke, S. E., Krantz, D. S., Johnson, D., Bittner, V., Eastwood, J.-A., . . . Merz, C. N. (2009). Comorbid depression and anxiety
symptoms as predictors of cardiovascular events: Results from the
NHLBI-sponsored Womens Ischema Syndrome Evaluation (WISE)
study. Psychosomatic Medicine, 71, 958 964.
Scott, K. M., Bruffaerts, R., Tsang, A., Ormel, J., Alonso, J., Angermeyer,
M. C., . . . Von Korff, M. (2007). Depression-anxiety relationships with
chronic physical conditions: Results from the World Mental Health
surveys. Journal of Affective Disorders, 103, 113120.
Shaver, P. R., & Brennan, K. A. (1992). Attachment styles and the Big
Five personality traits: Their connections with each other and with
romantic relationships outcomes. Personality and Social Psychology
Bulletin, 18, 536 545.
Tacon, A. M. (2003). Attachment experiences in women with breast
cancer. Family Community Health, 26, 147156.