Kedokteran

Scandinavian Journal of Public Health, 2009; 37: 801819
REVIEW ARTICLE
Do exercise and fitness protect against stress-induced health

complaints? A review of the literature
HSE
MARKUS GERBER & UWE PU
Institute of Exercise and Health Sciences, University of Basel, Basel, Switzerland
Abstract
Background: Understanding how exercise influences health is important in designing public health interventions. At present,
evidence suggests that there is a positive relationship between exercise and health. However, whether this relationship is
partly due to the stress-moderating impact of exercise has been less frequently investigated although more and more people
are taxed by stressful life circumstances. Methods: A comprehensive review of studies testing the potential of exercise as a
stress-buffer was conducted (including literature from 1982 to 2008). The findings are based on a narrative review method.
Specific criteria were taken into account to evaluate causality of the evidence. Results: About half of the studies reported at
least partly supportive results in the sense that people with high exercise levels exhibit less health problems when they
encounter stress. The causality analyses show that stress-moderation effects were consistently found in different samples and
with different methodological approaches. Although more support results from cross-sectional studies, exercise-based stressbuffer effects were also found in prospective, longitudinal and quasi-experimental investigations. Conclusions: This review
underscores the relevance of exercise as a public health resource. Recommendations are provided for future
research. More prospective and experimental studies are needed to provide insight into how much exercise is
necessary to trigger stress-buffer effects. Furthermore, more information is warranted to conclude which sort of
exercise has the strongest impact on the stress-illness-relationship.
Key Words: Buffer, exercise, fitness, health, physical activity, review, stress
Introduction
In industrialized countries, many people feel stressed
[1]. National health surveys show that the levels of
perceived stress have increased considerably over the
past decades [2,3]. Even children and adolescents
report high amounts of psychological stress [4].
According to the World Health Organization, stress
is among the leading causes for the global burden of
disease, entailing heavy costs both for national health
systems and the private economy [5,6].
Despite considerable heterogeneity in defining and
measuring stress, research leaves no doubt that stress
plays an important role in the aetiology of both
somatic and psychological diseases [4,711]. Human
stress regulation systems have evolved over thousands of years to deal primarily with short-term,
mostly somatic stressors. Hence, the physiological
stress reactions of the human organism to address

psychosocial and chronic stressors prevailing in
todays society often seem inappropriate [12].
Moreover, stress is linked with unhealthy behaviours
such as snacking, smoking, drinking and substance
use [1315]. As a result, stress prevention represents
a major public health concern in many modern
societies [16]. At the same time, considerable efforts
are undertaken to develop effective strategies to
reduce the adverse effects of stress [6,17].
Buffering stress: Cultivating health resources
Public health concepts emphasize the importance of
personal and social resources to buffer stress. Thus,
stress alone is not sufficient to explain interindividual
differences
in
peoples
health.
Consequently, stress research has shifted from the
Correspondence: Markus Gerber, Institute of Exercise and Health Sciences, University of Basel, Birsstrasse 320, CH-4052 Basel, Switzerland.
Tel: 0041 61 377 87 83. Fax: 0041 61 377 87 89. E-mail: markus.gerber@unibas.ch
(Accepted 14 September 2009)
2009 the Nordic Societies of Public Health
DOI: 10.1177/1403494809350522
Downloaded from sjp.sagepub.com at UNIVERSITY OF ALBERTA LIBRARY on April 25, 2015
802
M. Gerber & U. Puhse
analysis of main-effects to the investigation of interaction-effect models [18]. In this context, a multitude of cognitive, psychosocial and behavioural
variables were identified that are able to protect
against the debilitating effects of chronic stress
(i.e. hardiness, sense of coherence, mastery, optimism, self-esteem, social support) [19,20].
contributes to lowered stress perceptions, for

instance because of reduced feelings of loneliness
(i.e. broader social networks) or because of a reduced
susceptibility when confronted with external
demands (higher stress tolerance). Results obtained
from empirical studies show that exercise is generally
associated with lower stress rates. Aldana and
colleagues [26] concluded that after adjustment for
age and gender, highly active individuals were more
than half as likely to report high perceived stress
compared with those who had low exercise levels.
Similar findings were reported in other crosssectional studies [2733], although significant relationships were not supported in all of them [3436].
Few studies have used longitudinal designs.
However, Schnohr and colleagues [37] found evidence for a causal link between exercise and stress, in
the sense that increased exercise was followed by
decreased levels of stress. Alternatively, some
researchers explored whether high stress leads to
lowered exercise. For instance, Steptoe and colleagues [38] illustrated that college students became
increasingly sedentary when an academic exam
approached. Oaten and Cheng [39] reported comparable findings, while Stetson and colleagues [40]
indicated that women significantly reduced their
fitness training when they had increased levels of
stress. In sum, these studies indicate that exercise and
stress influence each other in reciprocal ways.
Beside preventive effects, exercise may strengthen
other personal (i.e. self-esteem) and social resources
(i.e. social support), which in their turn have the
potential to influence the stresshealth relationship
(path 3). As an example, adopting an active lifestyle
or experiencing motor competence may lead to
feelings of mastery and, thus, confidence in overcoming everyday problems and barriers [41,42].
Furthermore, exercise may protect against stressinduced losses of personal and social resources (path 4).
Physical activity, exercise and fitness as stress-buffers

In the past, epidemiological studies have shown that
daily physical activity, regular exercise and a sufficient level of physical fitness (in the remainder of the
article, exercise is used to refer to all of these
constructs [21]) benefit health and protect against a
variety of physical and psychological conditions [22].
Some researchers assume that the potential role of
exercise as a health resource stems at least partly from
the fact that exercise buffers against the deleterious
effects of stress on health [23]. In 1982, Kobasa
and colleagues [24] first explored whether exercise
moderates the stressillness relationship. In their
pioneer work, they found convincing evidence that
those male managers who engaged in exercise on a
regular basis reported less symptoms when exposed
to high stress.
Theoretically, exercise may influence the transactional stress process by means of short- and longterm effects as well as moderating and mediating
processes [6,20,25]. Figure 1 illustrates the most
important pathways of how exercise may affect the
interplay between stress, resources and health.
The direct relationship between exercise and
health is well established for variables such as coronary heart disease, diabetes mellitus, bone mineral
density, colon cancer, life-satisfaction or depression
(path 1) [22]. These effects, however, occur independent of the amount of stress a person perceives.
A preventive effect (path 2) exists whenever exercise
Personal & social

resources
(4) Protection of
resources
(3) Cultivation of
resources
(1) Direct effect
Exercise
(2) Preventive
effect
Stress-event
(5) Stress-buffering
effect
Figure 1. Potential influence of exercise on the interplay between stress, resources and health [69].
Health
Exercise, fitness and stress-induced health complaints: a review

However, empirical evidence for the latter mechanism is sparse [43].
The protective impact that exercise may offer in
times of heightened stress is referred to as the stressbuffering effect (path 5). Several mechanisms may
explain how exercise alleviates the pathological
effects of stress. First, it can be assumed that exercise
results in reduced arousal (i.e. mood enhancement
due to cognitive distraction or biochemical changes)
or more positive health behaviours during periods of
high stress (i.e. decreased smoking, healthier eating
habits). Second, exercise may bring about higher
levels of fitness and as an indirect consequence a
more efficient physiological stress regulation (i.e.
reduced secretion of hormones, lowered blood
pressure) or enhanced recovery processes.
Several studies have empirically tested whether
exercise moderates the stressillness relationship.
A systematic review, however, is still missing. The
primary purpose of this paper is to provide a comprehensive and critical overview of the body of
literature. A narrative review was conducted by
referring to seven of the nine criteria outlined by
Hill [44] that should be considered in order to infer
causal interpretation of empirical evidence.
Review method
Databases searched for this review include Psychinfo/
Psyndex, Eric, Pubmed, Medline, Web of Science,
SportDiscus and Spolit. The last search was conducted in May 2008. Keywords used were physical
activity, fitness, exercise, stress, stress reactivity and
health. Moreover, citation lists from papers, book
chapters and books were cross-referenced. Not taken
into account were unpublished doctoral dissertations
[4548] and studies focusing on clinical populations.
In addition, articles selected for inclusion in this
review met the following criteria: (1) They were
written in English, German or French; (2) They
incorporated (a) a measure of naturally occurring
stress as an independent variable (no laboratory
stressors), (b) physical activity, exercise, sport participation or fitness as a moderating variable and (c)
one or several health outcome(s) as a dependent
variable; (3) They provided sufficient statistical
information to conclude whether the two- or threeway interaction effects were significant or not. A
distinction was made between studies that supported, partly supported or did not support the
notion of exercise as a stress-buffer. Commonly,
multiple indicators were used to assess stress, exercise and, particularly, health. Accordingly, many
studies provided more than a single interaction
term. Studies in which more than two thirds of all
803
interaction effects were significant were classified as

fully supportive. If more than one third of the
analyses supported a moderator effect, studies were
considered partly supportive. If less than one third of
the statistics were significant, the studies were classified as non-supportive. Despite the fact that this
categorization is based on the 5% level of significance
as a cut-off criterion, this review does not purely rely
on a vote-counting procedure. Vote-counting may be
a problematic approach that may entail misleading
conclusions (e.g. no validity concerns, reliance on
sample size). Accordingly, we opted against such an
approach and, instead, decided to refer to Hills [44]
criteria to determine more solidly to what degree the
influence of exercise on the stresshealth relationship
can be regarded as causal.
Results
Studies included in this review
Thirty-one journal articles, book sections and books
dealing with the role of exercise as a stress-buffer
were found that met all the inclusion criteria
described above. Some articles were based on the
same cohort, which reduced the number of independent datasets to 27. Detailed information about the
samples, sampling methods, research designs, control of confounding variables, measurements and
data analysis techniques and the results is provided in
Tables I to IV (separately for adolescents and adults
and studies with cross-sectional and prospective,
longitudinal or experimental design).
In sum, 12 studies fully supported the validity of an
exercise-based stress-buffer hypothesis [24,4959].
In addition, four studies showed partial support
[3,6062]. In contrast, 15 studies did not confirm
the stress-buffering hypothesis [25,6375], although
some significant interactions were found in these
studies, too [25,57,64,73,76]. This global analysis
points to a great variability regarding the potential of
exercise to buffer stress. Certainly, given the diversity
of sample characteristics, study designs and measurements, this variability is not unexpected. However,
the inconsistent findings raise the question whether
we should consider the glass as half-full or halfempty, that is to say, whether this pattern of results
(about half of the studies providing support) is to be
interpreted as good evidence in support of, or as a
basis for discounting the original hypothesis. This
question, however, must be answered with neither
nor because the presence (and consistency: see
below) of supportive findings is a necessary, but still
insufficient, condition for causality. Thus, further
criteria need to be taken into account.
Subjects, n, age, sample

characteristics
Male managers (n 137,

M 52.1 years, predominantly protestant, white and
married)
University students (n 373,

163m, 210f, M 21.7
years)

37m, 73f)
Male law enforcement officers

(n 412, M 35.9 years,
predominantly white)
US National Health Interview

Survey respondents
(n 31,868, 419 years)

61m, 74f, M 22.0 years,
predominantly white)
Authors, year,
location
Kobasa et al.
1982, USA
Roth et al. 1989,

USA
Brown 1991,
USA
Young 1994,
USA
Zuzanek et al.
1998, USA
Carmack et al.
1999, USA
Cross-sectional (nonrandom sample, no

gender, age or ethnicity
differences in dependent
variables, controlled for
life events)
Cross-sectional (multistage
probability sampling,
controlled for age and
education)
Cross-sectional and prospective (prospective

data only for one outcome, missing data analysis, controlled for
psychological distress)
Cross-sectional (nonrandom sample, return
rate 59%)
Cross-sectional (random
sample, return rate
67.2%, controlled for
age, marital status, job
level and years of
service)
Cross-sectional (controlled
for gender)
Study design (timeframe,

sampling, return rate,
confounders)
Table I. Studies with adult samples and cross-sectional design.
Regression analyses
1. WSIe
2a. Profile of Mood States
2b. Health complaints
3a. Maximal fitness test
3b. Exercise (multiple items)
Regression analysis
1. LESc
2. SIRS
3a. Perceived fitness (12 items)
3b. Exercise (FDIT, 15 items)
4. Hardiness
Regression analyses
1. LESc
2a. SIRSb
2b. Health centre visits (prosp.)
3a. Exercise (TD, 15 items)
3b. Sub-maximal fitness test
Partial correlation analyses
1a. Job stress
1b. PSSd
2. Coronary heart risk factors
3. Maximal fitness test
Regression analyses
1. General stress (1 item)
2. Health perception (1 item)
3a. Exercise (general, 1 item)
3b. Exercise (TF, 22 items)
ANOVA (median split)

1. SRRSa
2. SIRSb
3. Exercise (TDI, 4 items)
4. Hardiness
Data analysis, measures

1. Stress, 2. Health,
3. Exercise, 4. Moderators
Stress-buffer: Partial support

Comment: Significant regression
weights: from 0.18 to
0.21. Interaction explained
between 3% and 4% of additional
variance.
Stress-buffer: No support
Stress " <4 Health #

Fitness <//4 Health
Fitness <//4 Stress
Exercise <//4 Health
Exercise <//4 Stress

Fitness <//4 Health

Exercise " <4 Health "
Exercise (3a) " <4 Stress #
Exercise (3b) " <4 Stress "
(continued )

Comment: Buffer effect for males
only. Significant regression
weights: from 0.06 to 0.08.
Interactions explained 0.1% of
additional variance.
Comment: Buffer effect for 3b only.
Significant regression weights:
from 0.34 to 0.46.
Interaction explained 4.3% of
additional variance.
Stress-buffer: No support (neither

for exercise nor fitness)

Fitness " <4 Health "
Fitness " <4 Stress #
Stress <//4 Health

Fitness " <4 Health "
Fitness <?4 Stress
Stress-buffer: Full support

Comment: No information about the
strength of association. Three-way
interaction not examined.
Two- and three-way interactions,

comments

Bivariate relationships
804
Employees in health promotion

programme (n 192, 65m,
127f, M 35.3 years, all
hierarchical levels, predominantly white)
Three cohorts of male managers and MBA students

(n1 280, n2 192,
n3 162, M1 34.6,
M2 33.6, M3 38.7 years)
Varsity athletes and non-athletes (n 270, 136m, 134f,

1724 years)
Employees of a university
medical centre (n 1,720,
540m, 880f, different job
levels)
Community health survey

respondents (n 1,261,
52% females, 18 years,
representative for Upstate
New York, predominantly
white)

125m, 145f, M 20.9
years)
Manning &
Fusilier,
1999, USA
Siu et al. 2000,

Hong Kong
Skirka 2000,
USA
Kaluza et al.
2001, 2002,
Germany
Ensel & Lin

2004, USA
Lochbaum et al.
2004, USA
Regression analyses
1. LEIf
2a. Health costs
2b. Healthcare use
3. Exercise (FDT)
4a. Hardiness
4b. Social support
Regression analyses
1. OSI-2g
2a. Job satisfaction
2b. Mental well-being
2c. Physical well-being
3. Exercise (F, 1 item)
Correlation analyses
1. DHSh
2. Profile of Mood States
3. Athletes vs. non-athletes
4a. Hardiness
4b. Sense of coherence
ANCOVA and Chi2-test
1. Occupational stress
2a. Health complaints
2b. Psychological distress
2c. General well-being
2d. Backache
Regression analyses
1. Social life events
2a. Depression
2b. Health complaints
3. Exercise
4a. Social support
4b. Self esteem
Regression analyses
1. PSSd
2. Health complaints
3. Strenuous exercise (F, 1 item)
4. Personality traits

rate 60%, controlled
for healthcare use,
gender, age, race, education and income)
Cross-sectional (nonrandom sample, controlled for gender)
sample, controlled for
gender, marital status,
age and income)

rate 33%, controlled
for sex, age, education,
BMI and smoking
habits)
Cross-sectional (nonrandom sample, nonathletes: not member of

a varsity team and 2
exercise episodes/wk)
Cross-sectional (nonrandom sample, controlled for age, gender,

marital status, job level
and tenure)


sampling, return rate,
confounders)
Comment: Three-way interactions
not examined.
Comment: Buffer effects found in one
cohort for 2c. Regarding 2a, exercise was a positive buffer in one
cohort, but had detrimental effects
in another.
not examined.

Comment: Stress-buffer effects found
in all variables except for 2c.

not examined. Significant regression weights: from 0.10 to
0.18.

Comment: Significant regression
weight: 0.31. Three-way
interaction not examined.

Exercise <?4 Health
Exercise <?4 Stress

Exercise " <4 Stress #


Exercise <?4 Stress

Exercise " <4 Health #
Exercise <?4 Stress

comments

a
SRRS Social Readjustment Rating Scale. bSIRS Seriousness of Illness Rating Scale. cLife Experiences Survey. dPSS Perceived Stress Scale. eWSI Weekly Stress Inventory. fLEI Life
Events Inventory. gOSI-2 Occupational Stress Indicator-Version 2. hDHS Daily Hassles Scale. D Duration, F Frequency, I Intensity, T Type.

characteristics
Authors, year,
location
Table I. Continued.

805
East German migrants

before and after the fall
of the Wall (a,b: n 330,
186m, 144f, Media
n 30 years; c: n 235,
126m, 109f, M 31.0
years; d: n 110, males
only)
Longitudinal study of aging

respondents (n 5,014,
70 years, representative sample of non-institutionalized population)
Fuchs & Leppin

1992(a); Fuchs &
Hahn 1992(b);
Fuchs & Appel
1999(c); Fuchs
et al. 1999(d);
Germany
Unger et al. 1997,

USA
Longitudinal (6 years, exercise and

social interactions only measured at T1, controlled for age,
gender, education, income,
minority status and chronic
diseases)
Longitudinal (a,b: baseline, 8

months later, non-random
sample, dropout rate 61.8%,
dropouts more often males,
younger and with lower education, controlled for initial
mental health, migration, age,
gender; c,d: T3 after 20
months, controlled for age and
gender)
Longitudinal (entry, two and four

years later, non-random
sample, controlled for participation status, return rate
63.7%, dropout rate 34.5%,
controlled for initial symptoms)
Prospective (one-year interval, illness measured at t1 and t2, all
the other variables at t1 only,
dropout rate 17.6%)
Study design (timeframe, sampling, return rate, confounders)
Individual growth curve models

1. Widowhood
2. Physical functioning
3. Exercise (general, 5 items)
4. Social interactions
Regression analyses
1. SRSSa
2. SIRSb
3. Exercise (TDI, 4 items)
4a. Hardiness
4b. Social support at work
Regression analyses (a) and
ANCOVAs (b,c,d)
1. Unemployment
2a. Mental health (a,b)
2b. Health complaints (b,d)
2c. Depression (c)
2d. Self-efficacy (c,d)
3. Exercise (DF, 1 item)
4a. Social support (a)
4b. Anxiety (d)
Regression analyses
1. SRRSa
2. Somatic complaints
3. Exercise (?)


No information about strength of
association.

(only for 2c and 2d)
indicate that among stressed males,
exercise has a positive impact when
social support is lacking (same
finding for 4b); among females,
exercise compensates for lacking
support only when unstressed (a).
No information about strength of
association.
Comment: Although two-way interactions predominantly insignificant, exercise buffered stress among
less healthy individuals.

Exercise <?4 Stress



Comment: Regression weights:
0.06. Interaction explained
between 0.3% and 1.9% of additional variance.

comments

Exercise " <4 Health #
SRRS Social Readjustment Rating Scale. bSIRS Seriousness of Illness Rating Scale. D Duration, F Frequency, I Intensity, T Type.
Highly stressed male business executives (n 85,

scoring above median on
stressful life events)
Kobasa et al. 1985,

USA
Male employees (n 278,

M 44.0 years, middle
and top management,
risk factor assessment as
incentive)

characteristics
Howard et al. 1984,

Canada
Authors, year,
location
Table II. Studies with adult samples and prospective, longitudinal or experimental design.
806
Secondary school students

(n 3,426, 1,769m,
1,657f, M 15.6 years)
Adolescents attending
annual laboratory visits
(n 303, 50% boys, 12
24 years, 53% white,
47% black, family
hypertension
background)
(n 407, 213m, 194f,
M 14.0 years)
Gogoll 2004,
Germany
Yin et al. 2005,

USA
Cross-sectional (non-random
sample, controlled for sex,
age and nationality)
Cross-sectional (controlled for

age, race, gender family
socioeconomic status and
parental smoking habits)
Cross-sectional (systematic
cluster sampling, return rate
76%, controlled for gender
and perceived family
wealth)
sample, representative for
two German states, controlled for place of residence, sex, grade and
education)
Cross-sectional (non-random
sample, return rate 89%,
comparison of students with
the highest (n 55) and
lowest stress scores (n 55)

sampling, return
rate, confounders)
Regression analyses
1. Multiple objective and subjective stress indicators
2. Multiple health outcomes
3. Multiple exercise indicators
(related to youth sports)
4a. Self-esteem
4b. Family support
Regression analyses
1a. ARCSd
1b. Community stresse
2a. BMI
2b. Sum of skin folds
2c. Waist circumference
3. Exercise (FI, 1 item)
ANCOVAs
1. School-based stress
4. Self-esteem
ANCOVA
1. School-based stress
ANOVA
1. SRRSa
2a. SIRSb
2b. Depressed affect
3. Exercise (FDI, 4 items)

Comment: Although few significant
two-way interactions were found,
the majority of the analyses did not
support a stress-buffer effect.
Similarly, the three-way-interactions generally remained
insignificant.
Comment: Stress-buffer effect more
consistent for 1a than 1b.
from 0.46 to 0.49.
Interaction explained between 2%
and 3% of additional variance.
Comment: No significant three-wayinteraction found between stress x
exercise x self-esteem.

Comment: Given the large
number of variables, the
associations indicate general trends

Exercise <?4 Stress


Comment: Interaction explained
1.3% of additional variance.

Comment: No information about
strength of association.

comments

Exercise <?4 Stress

Exercise <//4 Stress (2a)
Exercise " <4 Health (2b)"
SRRS Social Readjustment Rating Scale. bSIRS Seriousness of Illness Rating Scale. cHealth Behavior of School-Aged Children. dARCS Adolescent Resources Challenge Scale.
Monthly rent or mortgage in the community. D Duration, F Frequency, I Intensity, T Type.
Gerber & Puhse,

2008,
Switzerland
Haugland et al.
2003, Norway
Female secondary school

students (n 220,
Median 14 years, private school, predominantly white, upper
middle-class)
in HBSCc study
(n 1,577, 50.3% boys,
15 years old)

characteristics
Brown & Lawton

1986, USA
Authors, year,
location
Table III. Studies with adolescent samples and cross-sectional design.

807

27m, 28f, M 18.9
years, from an original
sample of 1,051 individuals, all reporting high
rates of stress and no
regular exercise/
relaxation)
Female secondary students

(n 212, M 13.8
years, 711 grade, private school, predominantly white, high
socioeconomic status)

(study 1: n 147, 70m,
77f, M 14.3 years;
study 2: n 60, 31m,
29f, M 16.516.8
years, diverse socioeconomic status)
Secondary and vocational

school students
(n 367, 60% boys,
M 16.5 years)
Brown & Siegel

1988, USA
Norris et al.
1992, UK
Rothlisberger
et al. 1997,
Switzerland
University students
(n 112, 40m, 72f,
M 18.9 years)

characteristics
Roth & Holmes

1985b, USA
Roth & Holmes

1985a, USA
Authors, year,
location
Longitudinal (five measurements

over two years, each variable
assessed at least two times,
classes randomly selected,
n 264 with complete data over
five measurements, controlled
for baseline scores)
Cross-sectional (study 1) and

quasi-experimental (study 2)
(controlled for pre-intervention
scores, high intensity training
(30min/2d/wk) results in
increased fitness at T2)
Prospective (two measurements

with 9 months interval, T1:
n 364, no differences due to
dropout status, controlled for
baseline age and illness)
Quasi-experimental (comparison
of two types of interventions
(30min/3d/wk) and one control
group, baseline, after 5 and 11
weeks and 2-months follow-up,
no baseline differences in stress,
controlled for baseline scores)
Prospective (T1: Stress during the

previous year, T2: Health
record during the course of the
investigation, T3: health problems during the past 2
months)
ANCOVAs
1a. LEQc
1b. PSSd
2a. SIRSb
2b. Mental health
3a. Exercise (FDI, 4 items)
3b. High intensity, n 22
3c. Moderate intensity, n 16
3d. Flexibility training, n 19
3e. Control group, n 16
ANCOVAs
1a. Daily hassles
1b. Life events
1c. Combination 1a/1b
2a. Life satisfaction
2b. Health satisfaction
3. Exercise (FDIT, 13 items,
vigorous)
4. Emotional support
Regression analyses
1. LESa
2a. Health record form
2b. Depression
2c. Anxiety
2d. Psychological functioning
3. Sub-maximal fitness test
ANCOVAs
1. LESa
2a. Health record form
2b. Health problems
2c. Depression
2d. Anxiety
2e. Health complaints
3a. Exercise: n 18
3b. Relaxation: n 19
3c. No treatment: n 18
Regression analyses
1. LESa
2. SIRSb
3. Exercise (DT, 14 items)

Table IV. Studies with adolescent samples and prospective, longitudinal or experimental design.
Stress (1a) <//4 Health (2a/b)

Stress (1b)" <4 Health (2a) #
Stress (1b) <//4 Health (2b)
(continued )
Comment: Three-way interactions yielded non-significant
results.

Comment: Buffer effect if crosssectional and prospective data
analysis (interaction stronger
for prospective analysis).
from 0.32 to 0.35.
Study 2:
Comment: Stress-health relationship weakened from T1 to T2
in high intensity training group
(but not in the other groups).

Study 1:
Exercise " <4 Health (2a)"
Exercise <//4 Health (2b)
Exercise <//4 Stress (1a)
Exercise " <4 Stress (1b) #
Comment: Students reported
recovery from stress across all
treatment conditions.

Comment: Fitness buffers against
stress only when severity of
total health problems was
examined. No information
about strength of association.
Two- and three-way interactions, comments
Stress <?4 Health

Exercise <?4 Stress

Fitness " <4 Health (2a)"
Fitness <//4 Health (2b/c)
Exercise <?4 Stress
808
LES Life Experiences Survey. bSIRS Seriousness of Illnesses Rating Scale. cLEQ Life Events Questionnaire. dPSS Perceived Stress Scale. eSEM Structural Equation Modelling.
COR-E-Y Conservation of Resources Evaluation for Youth. D Duration, F Frequency, I Intensity, T Type.
a
Comment: No stress-buffer effect
with cross-sectional and longitudinal analyses. Three-way
interactions yielded non-significant results.
Cross-sectional and longitudinal
(the latter with n 281 students, T2 after 1-year, dropouts
analysis on main study variables, controlled for gender,
school type, age, status and
place of residence)
Secondary and vocational
school students
(n 1,183, 537m, 646f,
M 17.16 years)
Gerber 2008,
Switzerland
SEMe and ANCOVAs

1. COR-E-Yf
2. Latent health construct
3a. Exercise (FDIT, 12 items,
moderate versus vigorous)
3b. Physical ability self-concept
4a. Academic self-concept
4b. Self-efficacy
4c. Optimism
4d. Social support

characteristics
Authors, year,
location
Table IV. Continued.

Two- and three-way interactions, comments
809
Criterion 1: Consistency of evidence

Hill [44] argues that an association between constructs can be viewed as more evidence-based if it has
been observed repeatedly over time by different
persons, in different places, in different circumstances and times, and by using different methodological approaches. Accordingly, consistency refers
to external validity issues, that is, to the degree of
generalizability of stress-buffer effects across individuals and methodologies. Thus, in this section, we
address the question whether evidence in support of
exercise-based stress-buffer effects was found (a) in
different countries, (b) in different age groups, (c) in
females versus males, (d) in normally stressed people
versus individuals at risk for heightened stress, (e) in
groups with minority versus majority ethnicity, (f) in
different moments in time, (g) using different measures of exercise, (h) applying different measures of
naturally occurring stress, (i) including populations
with small versus large sample sizes, and (j) using
different methods of data analysis.
Location: Exercise-based stress-buffer effects have
been investigated in the USA, Canada, Germany,
Norway, the UK, Switzerland and Hong Kong.
About 60% of all North American publications at
least partly supported the stress-buffer hypothesis. In
European research, evidence was found as well, but
to a more limited extent (in about 40%). In other
parts of the world, few attempts have been made to
tackle this issue.
Participants age: Most research has been conducted with adults. Approximately one third of the
studies were done with adolescent samples, which is
an indication for experts beliefs that most (psycho)
pathological diseases have their onset early in life and
that prevention should already start in childhood and
adolescence [20]. If studies are grouped by subjects
age, about half of the studies highlight moderator
effects among adults and adolescents. Studies with
children (below 12 years) do not exist. Similarly,
investigations with elderly people are scarce.
However, Unger et al. [71] showed that among
elderly people exercise can lessen the negative
impact of stress, especially among the less healthy
ones. Studies comparing different age groups are
missing, most likely because stressors and cognitive
capacities to make accurate stress appraisals vary
considerably with age [4].
Gender differences: More than half of the studies did
not consider gender differences despite the fact that
most researchers used mixed samples. Two studies
incorporated purely female samples (adolescent girls)
providing support for a stress-buffer effect [56,57].
In six studies, the data analysis was restricted to male
810
subjects, resulting in inconsistent results [24,53,54,

64,66,69]. When gender-unspecific analyses were
performed, four out of ten studies provided at least
partly supportive results. This ratio was similar when
gender-sensitive analyses were conducted.
Stress levels: With regard to the subjects stress
levels, most research has focused on normally
stressed individuals. Only two studies have included
people with explicitly high stress levels. While Kobasa
et al. [54] assessed highly stressed male business
executives, Roth and Holmes [74] targeted college
students. Further, Yin et al. [59] focused on
adolescents with a family history of coronary disease.
Despite the limited number of studies with at-risk
populations, it appears that exercise may be particularly helpful if people are in difficult life circumstances. While two studies found significant
stress-buffer effects [54,59], no moderation occurred
in the third study (perhaps because all individuals
recovered well from baseline stress).
Ethnicity: Race/ethnicity has received little attention. In many studies, no information was provided
about the participants ethnic background. If more
detailed information was available, most studies
comprised predominantly Caucasian or majority
group samples. Only two studies were based on
samples with a high ratio of ethic minority individuals
[59,67]. These studies, however, provided inconsistent results. The fact that exercise protects minority
youth from getting overweight [59] is, however,
an important discovery since race/ethnicity is generally considered as a risk factor in the aetiology of
obesity [77].
Date and source of publication: If the number of
yearly published articles is considered as an indicator
for the popularity of the topic, no significant increase
is apparent since the initial work of Kobasa and
colleagues [24]. Evidence was found in the 1980s,
the 1990s and since the turn of the millennium.
About two thirds of all articles were published in non
exercise-specific journals, which underlines the social
relevance of the topic.
Exercise measurement: Regarding the use of exercise
versus fitness as moderator variables, exercise clearly
prevailed, while studies examining objective physical
fitness [61,62,64,73] or perceived physical fitness
[78] were less common. With exercise, 14 studies
used simple measures (consisting of one item), while
nine administered activity indices, which mostly
assessed frequency, but in some cases also duration,
type and intensity. In seven studies, exercise was
measured with a range of items including more
general questions about the subjects exercise levels
(i.e. comparison with peers). In sum, it can be
concluded that the various methods of exercise
assessment produced similar results. However, the

high portion of studies using simple and nonvalidated measures underlines the need to attach
more importance to accuracy issues regarding data
assessment. Furthermore, this review shows less
empirical evidence when physical fitness was used
as a moderator of the stressillness relationship.
Despite that, some studies produced conflicting findings when fitness and exercise were directly compared. In one study [62], exercise (but not fitness)
resulted in significant interaction effects, whereas in
two studies, the opposite was the case [61,62].
Stress measurement: Fourteen studies used life event
inventories or general stress scales to gather information about participants stress levels. Among the
most popular instruments were the Social
Readjustment Rating Scale [79], the Life
Experiences Survey [80], and the Perceived Stress
Scale [81]. Only three studies administered a daily
hassle scale, while eight studies used single stressors
such as school stress [58,82], occupational stress
[49,50,66], unemployment [55,6870] or widowhood [71]. Moreover, one study used resource loss to
operationalize stress [25] and three studies simultaneously tested daily hassles and critical life events.
As far as minor stressors were concerned, only one
study supported the stress-moderation hypothesis
[65], while the other two failed to establish significant
interactions for both types of stress [67,75]. To
summarize, this review shows that stress-buffer
effects were most frequently found when life events
were used to measure stress (in about 60%).
Compared to that, the ratio found in studies with
single stressors and daily hassles was lower (about
one third of studies).
Sample size: Sample size affects most tests of
significance in the sense that using large samples
increases the likelihood of finding significant relationships. Thus, a standardized regression weight (e.g.
b 0.15) is more likely to be significant in a large
(e.g. n 1,000) than a small sample (e.g. n 100).
Accordingly, one might expect that the use of small
sample sizes results in an underestimation of (interaction) effects. This assumption was partly confirmed as large-scale studies (with more than 500
participants) more often produced supportive
results. A prime example for a large-scale study is
the one conducted by Zuzanek and colleagues [3]
including more than 30,000 individuals. Although
significant interaction effects between stress and
exercise (p < 0.05) were found, the magnitude of
this interaction was rather marginal (b 0.06; 0.1%
explained variance). Evidence for exercise-based
stress-buffer effects, however, was also found in
studies with small sample sizes (<200 participants).

Method of data analysis: As far as data analysis was
concerned, analyses of (co)variance (ANCOVA) and
hierarchical regression analyses were the most popular methods. Structural equation modelling with
multiple group comparisons is a newer method and
has, therefore, been less frequently applied. In contrast to AN(C)OVAs and hierarchical regression
analyses, studies using SEM (thus, controlling for
measurement error) have, not yet been able to
demonstrate significant interactions between stress
and exercise on health [25].
To summarize the findings with regard to the first
quality criterion (consistency), it might be inferred
that evidence was found in North American and
European samples, adolescents and adults, men and
women, normally and highly stressed individuals
and ethnic minority and majority samples.
Moreover, evidence seems to be independent of
date of publication, the assessment of stress, sample
size and the method of data analysis. This pattern of
results indicates that stress-buffer effects are relatively generalizable across different groups of people
and methodological approaches. On the other hand,
it is difficult to explain why only half of the studies
have supported the existence of exercise-based stressbuffer effects. However, with one exception [66],
consistency was also found as far as the direction of
the association was concerned. According to our
review, exercise can be viewed as a protective resource that allows more constructive coping with stress.
This is worth mentioning since investing energy in
exercise in times of heightened stress might be
considered as a waste of resources. Similarly, exercise
does not result in additional stress during stressful life
circumstances.
Criterion 2: Strength of association
According to Hill [44], the strength of association is
an important indicator of causality because it shifts
attention from interpreting levels of significance
towards pondering meaning and relevance of empirical data. To provide an indication of the strength of
exercise as a stress buffer, the amount of explained
variance by the interaction term beyond that of the
main effects was examined. In regression analyses,
this information is provided by the R2-increment,
while the partial eta-squared coefficient (g2) is used
as an indicator in AN(C)OVA. For AN(C)OVA,
Cohen [83] generally recommended considering
effect sizes between 0.01 g2 < 0.06 as small, between 0.06 g2 < 0.14 as moderate, and g2 0.14 as
large. For multiple regression analyses, the recommended benchmark is 0.02 R2 < 0.15 (small),
0.15 R2 < 0.35 (moderate) and R2 0.35 (large).
811
Against general recommendations [84], many studies

have provided insufficient information to gauge the
magnitude of the effects [24,51,52,5457,73]. To
illustrate, some authors who have applied hierarchical
regression analyses displayed standardized regression
weights, but did not report R2-increments
[51,52,57]. In other studies, the stress x exercise
interaction was entered together with the main effects
into the regression model, which, again, precludes
from determining the exact increment of explained
variance attributable to the stress moderation [53].
Those studies that provided sufficient information,
found relatively small effects with levels of additionally explained variance ranging between 0.1% [3] and
4.3% [62]. Most effects sizes varied between 1% and
3% of explained variance [3,53,59,61,85]. However,
as outlined by McClelland and Judd [86] and Jaccard
et al. [87], the detection of interaction effects is
difficult, especially in field studies, which seem to
have less than 20% of the efficiency of optimal
experimental tests. Important reasons for this difficulty are (i) problems of multicollinearity (which can
be resolved by centering variables), (ii) measurement
error (e.g. unreliable measures, non-optimal distribution of variables, treatment of ordinal data as
interval data), (iii) non-linear forms of regression
(e.g. curvilinear interactions), and (iv) adjustment for
covariates (diminishing the statistical power due to
reduced residual variance).
In summary, it can be said that many researchers
have not paid sufficient attention to the second
quality criterion (strength of association), with an
overemphasis put on levels of significance as the
exclusive reference point [88]. Apart from that, the
effect of exercise was relatively weak. However, as
Hill [44] states, we must not be too ready to dismiss
a cause-and effect hypothesis merely on the grounds
that the observed association appears to be slight
(p. 8). As shown earlier, this applies particularly for
interaction models [86,87].
Criterion 3: Specificity of effects
Following Hill [44], specificity of effects refers to the
issue whether a specific disease (e.g. lung cancer) is
limited to specific conditions thought to be responsible for this disease (e.g. cigarette smoking).
Hill claims that such specificity strengthens the
plausibility of cause-effect relationships. However,
for several reasons, specificity cannot be claimed
regarding the association between exercise, stress and
health. First, stressors mostly lead to unspecific
adaptation processes regulated by the sympathetic
nervous system and hypothalamo-pituitary adrenal
axis [89,90], despite a certain degree of specificity
812
existing for some events [91]. On the other hand,

stress appraisals are complex and depend both on
personal, social and environmental factors [92,93].
Additionally, research has shown that exercise protects against a myriad different diseases [22,94]. As
far as stress-buffer research is concerned, this section
explores whether stress-buffer effects were found for
different health outcomes. Furthermore, the question is addressed whether stress-buffer effects depend
on the interplay with other salutogenetic resources.
In sum, the present review shows that exercise buffers
both against mental and physical health problems.
For instance, evidence was found when health was
operationalized by means of the Seriousness of Illness
Rating Scale [95], psychosomatic complaints, obesity
and backache. With respect to mental health, exercise
buffered primarily against depression and anxiety. In
summary, evidence is slightly stronger for physical
health compared to psychological outcomes.
Moreover, there is only limited evidence that exercise-based stress-buffer effects depend on the interplay with other resources. Although several studies
included additional health resources such as hardiness [24,54,65,67,78], self-esteem [25,51,76,82],
academic self-concept [25], coping ability [50],
sense of coherence [67] and social support
[25,49,51,54,65,70,71,75,76], these resources were
mostly considered as independent factors. Thus,
three-way interactions were rarely analyzed.
Unexpectedly, Fuchs and Hahn found that exercise
benefits only individuals with low stress and high
anxiety score [68]. Regarding social support, similar
findings emerged among women. In turn, the original
hypothesis that exercise compensates for a lack of
social support in times of high stress was corroborated among male subjects [70].
To summarize, both mental and physical health
appear less affected in stressful life circumstances
among exercisers. Evidence is somewhat stronger for
physical outcomes, which is an important finding
since some authors attribute the relationship between
stress and (mental) health partly to methodological
artefacts (e.g. shared method variance) [96,97].
Criterion 4: Temporality
The criterion of temporality (or temporal sequencing) is closely connected with issues about the
internal validity of the study design. Following Hill
[44], temporal sequence is particularly important
with diseases of slow development because it
addresses the question of which is the cart and
which the horse (p. 297). As is generally accepted
[98], temporal sequence can only be tested by means
of prospective or (quasi-)experimental studies,
preferably using long timeframes. Regarding studies

included in this review, 21 provided exclusively crosssectional analyses, whereas 12 used prospective and
longitudinal designs (we distinguish between prospective (only health outcome repeatedly measured)
and longitudinal studies (all study variables repeatedly measured)). In contrast, the number of quasiexperimental studies is very limited [60,74].
Prospective and longitudinal studies included timeframes between eight months [68,70] and six years
[71]. Most prior research, however, encompassed
periods between eight months and one year.
Generally, exercise-based stress-buffer effects are
best confirmed in cross-sectional studies (evidence
found in almost 70% of the publications). However,
prospective and longitudinal research supports the
moderation hypothesis as well, with larger timeframes producing more supportive findings. Among
adults, prospective evidence is provided by Howard
et al. [53] who followed up male employees over a
period of four years. In this study, exercise buffered
against stress even when baseline symptoms were
controlled. Another study showed that the initial
exercise levels interacted with baseline stress in
explaining health after 12 months among highly
stressed males [54]. Among adolescents, research
showed similar effects within a female sample [57].
In sum, prior research has found significant stressbuffer effects in both cross-sectional, prospective and
longitudinal studies. However, supporting evidence
is stronger in studies with cross-sectional design.
Given that stress-related diseases have no clearly
defined onset and often develop slowly over time, the
limited support found in prospective studies is not a
surprise since timeframes were relatively short
(mostly less than one year). Perhaps using longer
follow-up periods would have resulted in stronger
effects. Finally, it is important to note that apart from
two exceptions [25,71], prospective and longitudinal
studies have generally operated with small samples.
Criterion 5: Biological gradient
What Hill [44] called a biological gradient is
currently known as a dose-response relationship. In
this review, two aspects are of particular interest, that
is, whether (i) a linear increase in exercise results in a
linearly increasing stress-buffer effect and (ii)
whether vigorous intensity exercise has a stronger
influence than moderate intensity exercise. Basically,
AN(C)OVAs are well suited to address the first issue
because they juxtapose groups with different exercise
levels. However, many researchers only compared
two groups [24,56,75], which makes it impossible to
examine a gradient. Despite that, Kaluza and

colleagues [49,50] found a rather linear increase of
exercise in buffering stress among participants who
felt moderately stressed. Among adolescents,
Haugland et al. [85] showed that stress-buffer effects
do not linearly increase with augmenting exercise
involvement. Rather, it is the group with the lowest
exercise level that is least able to cope with stress. As
the second issue is concerned (moderate versus
vigorous intensity exercise), very little is known so
far. For instance, Norris and co-workers [60] showed
that among highly stressed youth exercise enhances
health only when participants engaged in high intensity exercise training (compared to moderate intensity exercise and flexibility training).
In sum, a conclusion about the biological gradient
seems premature, as this issue has been largely
ignored in previous research. While one study
showed that physical inactivity precipitates stressrelated health problems [85], another study found
that exercise training produced positive findings
only when relatively high intensity levels were
applied [60].
Criterion 6: Plausibility
The concept of plausibility refers to the question
whether or not biological, psychological and behavioural mechanisms can convincingly explain
exercise-based stress-buffer effects. To provide evidence for the biological plausibility, exercise physiologists have generally looked at the phenomenon
using a laboratory paradigm. In sum, this line of
research has shown that exercise can indeed be
understood as a physical stressor that disturbs baseline homeostasis and elicits typical stress reactions
(e.g. secretion of norepinephrine, ACTH, cortisol)
[99]. Whether exercise leads to a cross-stressor effect,
that is, whether or not exercise related adaptation
processes transfer to other, non-exercise stressors
(particularly psychosocial and cognitive stressors)
[23,100], is less clear. Research exploring whether
chronic exercise training or heightened fitness levels
contribute to a suppression of individuals stress
reactions to, or boost recovery from, experimentally
induced stress, provided only limited support [23].
Although it was suggested in an early review that
exercisers have a significantly lower reactivity [101],
more recent reviews (using more specific inclusion
criteria) came to less supportive conclusions. In a
meta-analytic study, Jackson and Dishman [102]
found that trained individuals show slightly increased
stress reactions. In turn, they seemed to recover faster
from laboratory stressors. The latter finding, however, was not supported by another meta-analysis.
Regarding research examining the effects of acute
813
bouts of exercise on experimentally induced stressor

tasks (following immediately after), a majority of
previous studies showed that individuals react less
after exercise sessions [103,104]. Stronger effects
were generally observed among at-risk populations
(e.g. with high blood pressure), for high intensity
exercise (460% VO2max) and longer exercise bouts
(420 minutes). In sum, biological mechanisms only
partly supported exercise-based stress-buffer effects.
However, it must be kept in mind that laboratory
stressors are not equivalent with real life stress. First,
laboratory stressors are often short-term and, therefore, less powerful. Second, subjective appraisals are
ignored. Third, laboratory tasks are prone to experimenter effects. And lastly, experimental stressors
only allow a limited range of coping strategies [105].
In turn, the psychological foundation of a stressbuffer effect is relatively well documented. Many
studies found that exercise enhances mood and that
the effects are particularly strong among individuals
who initially felt low or irritated [106], which
reinforces the idea that exercise allows stress-sufferers to take time-out from everyday problems. In
contrast, it is still not clear whether exercisers and
non-exercisers differ in terms of preferred coping
styles [107]. Furthermore, epidemiological studies
show that exercise is positively related with quality of
sleep [108]. However, the specific impact of exercise
on sleep in stressful life circumstances is still
unexplored. Finally, research concerned with
health-related behaviours (nutrition, alcohol, smoking) has provided inconsistent evidence [109]. And
again, the impact of exercise on these behaviours
needs to be explored under high stress.
Taken together, exercise physiology has provided
only limited support for the biological plausibility of
stress-buffer effects. However, laboratory and naturally occurring stress are not synonymous and are
difficult to compare. Moreover, new methodologies
will be available to assess stress reactions.
Accordingly, we warn against drawing hasty conclusions. In contrast, exercise psychologists provided
compelling evidence that exercise buffers stress
because it enhances mood. Additionally, our recommendation for future research are to combine mediator and moderator models to find out (i) whether
stress affects health because it contributes to poor
sleep or unhealthy behaviours (mediator model) and
(ii) whether the interplay between these variables
depends on participants exercise levels (moderator
model). Finally, it seems useful to look at the
association between exercise, stress and health at a
micro-level, for example, by collecting data with
exercise-, sleep- and stress-logs [110] or by using the
experience sampling approach [111].
814
Criterion 7: Experimental evidence

As Biddle and Mutrie [94] have emphasized, the best
evidence for a causal link comes under Hills criterion
of experimental evidence. According to Hill [44], it is
possible to observe (experimentally induced or naturally occurring) circumstances that lead to changed
patterns between two variables. With humans,
manipulating stress (in terms of creating stressful
circumstances) is difficult and except for naturally
occurring stressors (e.g. exams, relocation, redundancy) unethical. In contrast, experimentally
manipulating exercise levels is feasible. However,
studies with (quasi-)experimental design are rare
exceptions. Thus, only two studies inform about
the effects of particular exercise training interventions. Roth and Holmes [74] compared 11 weeks
exercise or relaxation training with control conditions
focusing on highly stressed university students.
However, no stress-buffer effects occurred. Norris
et al. [60] compared moderate and high intensity
fitness training, flexibility training to no treatment
and found that only students in the highly intensive
training programme felt less affected by stress after
completion of the programme.
To summarize, the limited number of studies using
a quasi-experimental design does not allow a final
conclusion despite promising results in one study.
This lack of research clearly indicates the necessity
for more experimental research in general. In particular, the following aspects should be taken into
account: (i) Whenever possible, participants should
be randomly assigned to different conditions (as random sampling seems to facilitate the detection of
moderation effects [86]), (ii) sample sizes should be
sufficiently large (which was not the case in the two
studies described above), (iii) stress and health
should be assessed at every measurement (so that
reciprocal influences can be tested empirically), and
(iv) baseline-exercise and stress levels should be
taken into account (since the highest effects can be
expected among completely inactive or highly
stressed participants).
Discussion
This review supports the view that stress is negatively
associated with good health. Only one study did not
show a significant relationship between stress and
health [64]. In four studies, only some (but not all)
health indicators were negatively associated with
stress [57,62,73,75]. In all other investigations,
individuals with high stress levels reported significantly more health troubles.
Support for the validity of an exercise-based stressbuffer hypothesis was found in 16 studies. In
contrast, 15 investigations did not report significant
moderation effects. Given the status quo between
supportive and non-supportive studies, drawing a
conclusion for or against the validity of the stressbuffer hypothesis seems difficult unless other criteria
are taken into account. Accordingly, Hills [44]
quality criteria were used to obtain a more solid
and accurate interpretation of the evidence. The first
criterion was consistency, which was strongly supported. Stress-buffer effects were found among different groups of people, at different places and
moments in time and with a wide spectrum of
methodologies. From the inclusion of additional
moderator variables can be inferred that significant
relationships occurred slightly more often when
studies were conducted in North America.
Furthermore, physical fitness had a more limited
potential to balance stress-induced health complaints
compared with exercise. In turn, there is preliminary
evidence that exercise facilitates coping among at-risk
populations. The second criterion (strength of association) provided less support for a causal influence
as the magnitude of stress-buffer effects found in
most studies was relatively small. However, it must be
remembered that interaction effects are difficult to
identify and that, therefore, also small effects can be
of practical significance [86,87]. The third criterion
was specificity. However, specificity is not a strong
criterion since (i) stress has been defined as a general
adaptation syndrome and (ii) exercise influences a
myriad mental and physical diseases [22,94].
Regardless of that, moderator effects were most
often significant when health was defined as a
physical outcome. Moreover, little evidence was
found that moderator effects depend on the interplay
with other resources. As the fourth criterion (temporal sequence) was concerned, this review shows
that a bare majority of significant interactions
resulted from cross-sectional studies. In these studies, significant interactions might mean (a) that
exercisers are less affected by stress or (b) that people
who feel unaffected by stressful life circumstances are
more likely to maintain high exercise levels. However,
prospective studies also produced significant results,
especially when longer timeframes were used. The
fifth criterion (dose-response relationship) was rarely
addressed in past research. Studies provided mixed
messages about whether moderate or vigorous exercise leads to more beneficial results. To address the
sixth criterion (plausibility), we referred to studies
using a laboratory stress paradigm to find out
whether stress-buffer effects may be attributable to
biological mechanisms. However, exercise physiology

did not provide strong support for the biological
plausibility. Psychological aspects seem to be of
greater importance. However, more research is
warranted, particularly concerning the interplay
between stress, sleep, health behaviours and health.
Finally, experimental evidence (the seventh criterion)
is very limited since only two studies have used
a quasi-experimental approach.
In sum, the following conclusions seem legitimate:
First, an exercise-based stress-buffer effect was
consistently found in previous studies. Second,
some prospective evidence was found, also to a very
limited extent. Given that the direction of the interaction effects was generally consistent, showing that
exercise alleviates stress, we feel confident to advertise exercise as a stress-management strategy.
Although exercise may not always help, high exercise
levels in periods of increased stress do not generate
additional stress. Third, it needs to be remembered
that exercise has direct health effects, although this
review exclusively focused on the stress-buffering
impact. Fourth, little knowledge exists about how
exercise should be designed in order to promote
coping. Accordingly, efforts should be made to find
out how much exercise (e.g. dose, intensity) is
necessary to trigger stress-buffer effects. Moreover,
it should be clarified what exercise environments
(e.g. type, setting) may be helpful for different groups
of people (e.g. adolescents, adults, people at-risk) to
alleviate stress.
Strengths and limitations of this review

The findings of this review are based on a comprehensive literature review in the most widely used
databases. Thirty-one published articles met the
inclusion criteria. Seven criteria were used to interpret the causality of the relationships and several
moderating factors (e.g. age, gender, study design)
were taken into account. The limited number of
studies, however, did not allow higher-order interactions (i.e. gender x age) to be controlled. Since this is
a narrative review, no global effect size was obtained,
nor did the procedure permit to pool out each studys
sample size or to provide information about possible
file-drawer effects.
Recommendations for future research

This review demonstrates that the role of exercise in
preventing psychopathology and chronic diseases is
not limited to a direct relationship. Instead, the
findings show that exercise may contribute to peoples well-being also by protecting against the
815
negative health outcomes of chronic stressors.

At present, however, several issues are still unresolved. The following research recommendation can
be made: First, several studies have looked at the
main effects between exercise and stress [2633].
Indeed, whether exercisers experience less stress is
interesting. However, from a public health perspective, interaction effects between exercise and stress
are far more relevant. Beside that, studies should use
at least longitudinal research designs to determine
whether stress influences exercise or vice versa. If
only cross-sectional data are gathered, data analyses
should provide insight into possible dose-reponse
relationships. Distinguishing more clearly between
daily physical activity, exercise and sport participation is also to be recommended. Second, several
studies have simultaneously assessed stress, exercise,
other resources and health. Most of the analyses are,
however, restricted to isolated moderator effects.
Whether the interplay between exercise and other
resources entails adding or compensating effects
deserves more attention. Third, to explore the
causal links between stress, exercise and health,
there is great need for experimental studies.
Randomized controlled trials could provide insight
into which types and forms of physical activities
display particularly high buffering effects. Moreover,
the setting and circumstances under which exercise
takes place should be considered more closely.
Additionally, more knowledge is necessary about
how exercise can buffer stress together with other
leisure behaviours. Fourth, future studies should
focus more on at-risk populations. Such groups
include people who are exposed to high levels of
physical and psychological stress (i.e. police officers,
managers, teachers, bus drivers) or who are already
suffering from stress-related impairments (i.e. burnout patients, people in cardiac rehabilitation). Fifth,
there might be cross-cultural differences regarding
the impact of exercise on the stressillness relationship. To find out whether real intercultural differences exist, studies seem warranted, in which
samples from different countries are assessed with
the same instruments. Similarly, more emphasis
should be put on gender and age sensitive analyses.
Sixth, future research should be extended to other
health parameters. For instance, it would be interesting to know how exercise affects the link between
stress and risk behaviour, aggression, delinquency,
substance abuse and social integration among youth.
Similarly, more information is needed on how exercise affects the relationship between stress and diabetes type II and cancer in adults. Finally, diligence
seems necessary as far as the valid assessment of
stress and exercise are concerned. Particularly, future
816
studies should attempt to assess exercise with more

objective data assessment techniques (i.e. pedometers) as the effects of measurement errors are
multiplied if the product between two variables is
formed.
Acknowledgments
We would like to thank Dr. Dean Barker (Basel,
Switzerland) for proofreading the manuscript.
References
[1] Ramaciotti D, Perriard J. Les couts du stress en Suisse
[Stress-related costs in Switzerland]. Bern: Staatssekretariat
fur Wirtschaft (Seco); 2001.
[2] Cohen S, Williamson GM. Perceived stress in a probability
sample of the United States. In: Spacapan S, Oskamp S,
editors. The social psychology of health. Newbury Park:
Sage; 1988. pp. 3167.
[3] Zuzanek J, Robinson JP, Iwasaki Y. The relationships
between stress, health, and physically active leisure as a
function of life-cycle. Leis Sci 1998;20(4):25375.
[4] Grant KE, Compas BE, Thurm AE, McMahon SD,
Gipson PY. Stressors and child and adolescent psychopathology: measurement issues and prospective effects. J Clin Child
Adolesc Psychol 2004;33(2):41225.
[5] Tangri RP. What stress costs. Halifax: Chrysalis Performance
Strategies Inc.; 2003.
[6] Leka S, Griffiths A, Cox T. Work organisation and
stress. Systematic problem approaches for employers,
managers and trade union representatives (Protecting
Workers Health Series No. 3). Geneva: Word Health
Organization; 2003.
[7] Franco GP, de Barros AL, Nogueira-Martins LA, Michel JL.
Stress influence on genesis, onset and maintenance of
cardiovascular diseases: literature review. J Adv Nurs
2003;43(6):54854.
[8] Nielsen NR, Gronbaek M. Stress and breast cancer: a
systematic update on the current knowledge. Nat Clin
Pract Oncol 2006;3(11):61220.
[9] Siervo M, Wells JC, Cizza G. The contribution of psychosocial stress to the obesity epidemic: an evolutionary
approach. Horm Metab Res 2009;41(4):26170.
[10] Shiloah E, Rapoport MJ. Psychological stress and new onset
diabetes. Pediatr Endocrinol Rev 2006;3(3):2725.
[11] Stetz MC, Thomas ML, Russo MB, Stetz TA,
Wildzunas RM, McDonald JJ, et al. Stress, mental health,
and cognition: a brief review of relationships and countermeasures. Aviat Space Environ Med 2007;78(5 Suppl)
S25260.
[12] Carlson NR. Physiologische Psychologie [Physiological
psychology]. Munchen: Pearson Studium; 2004.
[13] Park CL, Armeli S, Tennen H. The daily stress and coping
process and alcohol use among college students. J Stud
Alcohol 2004;65(1):12635.
[14] Nishitani N, Sakakibara H. Relationship of obesity to job
stress and eating behavior in male Japanese workers. Int J
Obes 2006;30(3):52833.
[15] Byrne DG, Mazanov J. Adolescent stress and future smoking

behaviour. A prospective investigation. J Psychosom Res
2003;54:31321.
[16] Antonovsky A. Health, stress, and coping. San Francisco:
Jossey-Bass; 1979.
[17] Frydenberg E, Brandon CM. The best of coping.
Melbourne: Camberwell; 2002.
[18] Schwarzer R. Stress, Angst und Handlungsregulation
[Stress, anxiety, and behavioural regulation]. Stuttgart:
Kohlhammer; 2000.
[19] Schwarzer R. Manage stress at work through preventive and
proactive coping. In: Locke EA, editor. The Blackwell
handbook of principles of organizational behavior. Oxford;
2000. pp. 34255.
[20] Grant KE, Compas BE, Thurm AE, McMahon SD,
Gipson PY, Campbell AJ, et al. Stressors and child and
adolescent psychopathology: evidence of moderating and
mediating effects. Clin Psychol J 2006;26(3):25783.
[21] Caspersen CJ, Powell KE, Christenson GM. Physical
activity, exercise, and physical fitness: definitions and
distinctions for health-related research. Public Health
Reports 1985;100:12631.
[22] Sallis JF, Owen N. Physical activity and behavioral medicine.
Thousand Oaks: Sage; 1999.
[23] Sothmann MS. The cross-stressor adaptation hypothesis and
exercise training. In: Acevedo EO, Ekkekakis P, editors.
Psychobiology of physical activity. Champaign: Human
Kinetics; 2006. pp. 14960.
[24] Kobasa SC, Maddi SR, Puccetti MC. Personality and
exercise as buffers in the stress-illness-relationship. J Behav
Med 1982;5:391404.
[25] Gerber M. Sport, Stress und Gesundheit bei Jugendlichen
[Exercise, stress and health among adolescents]. Schorndorf:
Hofmann; 2008.
[26] Aldana SG, Sutton LD, Jacobson BH. Relationship between
leisure time physical activity and perceived stress. Percept
Mot Skills 1996;82(1):31521.
[27] Melamed S, Kushnir T, Strauss E, Vigiser D.
Negative association between reported life events and cardiovascular disease risk factors in employed men: the
CORDIS Study. Cardiovascular Occupational Risk Factors
Determination in Israel. J Psychosom Res 1997;
43(3):24758.
[28] Kornitzer M, Kittel F. How does stress exert its effects
smoking, diet and obesity, physical activity? Postgrad Med J
1986;62:6956.
[29] Wemme KM, Rosvall M. Work related and non-work related
stress in relation to low leisure time physical activity in a
Swedish population. J Epidemiol Community Health
2005;59(5):3779.
[30] Kouvonen A, Kivimaki M, Elovainio M, Virtanen M,
Linna A, Vahtera J. Job strain and leisure-time physical
activity in female and male public sector employees. Prev
Med 2005;41(2):5329.
[31] Bennett GG, Wolin KY, Avrunin JS, Stoddard AM,
Sorensen G, Barbeau E, et al. Does race/ethnicity moderate
the association between job strain and leisure time physical
activity? Ann Behav Med 2006;32(1):607.
[32] Sundblad GB. Perceived health in Swedish school students.
A longitudinal prevalence study. Stockholm: Karolinska
Institutet; 2006.
[33] Moksnes UK, Espnes GA. The connection between stress
and physical activity in an adolescent population in Norway.
In: Theodorakis Y, Goudas M, Papaioannou A, editors.
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
[49]
[50]
Sport and exercise psychology: bridges between disciplines

and cultures. Halkidiki: University of Thessaly; 2007.
Plante TG, LeCaptain SE, McLain HC. Perceived fitness
predicts daily coping better than physical activity. J Appl
Biobehav Res 2000;2000(5):1.
Bond DS, Lyle RM, Tappe MK, Seehafer RS, DZurilla TJ.
Moderate aerobic exercise, Tai Chi, and social problemsolving ability in relation to psychological stress. Int J Stress
Manag 2002;9(4):32943.
Cramer SR, Nieman DC, Lee JW. The effects of moderate
exercise on psychological well-being and mood state in
women. J Psychosom Res 1991;35:43749.
Schnohr P, Kristensen TS, Prescott E, Scharling H. Stress and
life dissatisfaction are inversely associated with jogging and other
types of physical activity in leisure time The Copenhagen City
Heart Study. Scand J Med Sci Sports 2005;15:10712.
Steptoe A, Wardle J, Fuller R, Holte A, Justo J,
Sanderman R, et al. Leisure-time physical exercise: prevalence, attitudinal correlates, and behavioral correlates
among young Europeans from 21 countries. Prev Med
1997;26:84554.
Oaten M, Cheng K. Academic examination stress impairs
self-control. J Soc Clin Psychol 2005;24(2):25479.
Stetson BA, Rahn JA, Dubbert PM, Wilner BI,
Mercury MG. Prospective evaluation of the effects of stress
on exercise adherence in community-residing women.
Health Psychol 1997;16(6):51520.
Fox KR. Self-esteem, self-perceptions and exercise. Int J
Sport Psychol 2000;31:22840.
Ekeland E, Heian F, Hagen KB. Can exercise improve selfesteem in children and young people? A systematic review of
randomised controlled trials. Br J Sports Med
2005;39:7928.
Stoll O. Wirkt sportliche Aktivitat ressourcenprotektiv?
[Does exercise protect against the loss of resources?].
Lengerich: Pabst; 2001.
Hill AB. The environment and disease: association or
causation? Proc R Soc Med 1965;58:295300.
Dineen AC. The relationship of physical activity to
self-concept and perceived stress in adolescent females.
Dissertation Abstracts International: Section B 1999;
59(8-B):4507.
Ellefsen KA. Physical activity as a moderator of life stress in
elementary school children. Dissertation Abstracts International: Section A 1998;59(2-A):417.
Hughes DC. Physical activity and stress in Hispanic breast
cancer survivors. Unpublished doctoral dissertation,
University of Houston; 2004.
Scarinci IC. Leisure-time physical activity as a moderator
of the association between stress and depression among lowincome primary care female population. Dissertation
Abstracts International: Section B 1998;59(6-B):3074.
Kaluza G, Hanke C, Keller S, Basler H-D. Salutogene
Faktoren bei chronischen Ruckenschmerzen: Moderieren
soziale Unterstutzung, Arbeitszufriedenheit und sportliche
Aktivitat den Zusammenhang zwischen Arbeitsbelastung
und Ruckenschmerzaktivitat? [Salutogenetic factors and
backache: do social support, job-satisfaction and exercise
buffer the relationship between work-stress and backache?].
Zeitschrift fur Klinische Psychologie und Psychotherapie
2002;31(3):15968.
Kaluza G, Keller S, Basler H-D. Beanspruchungsregulation
durch Sport? [Stress-regulation through exercise?].
Zeitschrift fur Gesundheitspsychologie 2001;9(1):2631.
817
[51] Ensel WM, Lin N. Physical fitness and the stress process.
J Community Psychol 2004;32(1):81101.
[52] Lochbaum MR, Lutz RS, Sells S, Ready A, Carson T.
Perceived stress and health complaints: an examination of
the moderating roles of personality and physical activity.
Percept Mot Skills 2004;99(3):90912.
[53] Howard JH, Cunningham DA, Rechnitzer PA. Physical
activity as a moderator of life events and somatic complaints:
s longitudinal study. Can J Appl Sport Sci 1984;
9(4):194200.
[54] Kobasa SC, Maddi SR, Puccetti MC, Zola MA.
Effectiveness of hardiness, exercise and social support as
resources against illness. J Psychosom Res 1985;29:52533.
[55] Fuchs R, Appel E. Belastungsregulation durch Sport.
In: Schwarzer R, Jerusalem M, editors. Gesellschaftlicher
Umbruch als kritisches Lebensereignis. Psychosoziale
bersiedlern und Ostdeutschen.
Krisenbewaltigung von U
[Social change as a critical life event coping of East
German migrants and people living in East Germany].
Weinheim: Juventa; 1994. pp. 22740.
[56] Brown JD, Lawton M. Stress and well-being in adolescents:
the moderating role of physical exercise. J Human Stress
1986;12:12531.
[57] Brown JD, Siegel JM. Exercise as a buffer of life stress: a
prospective study of adolescent health. Health Psychol
1988;7:34155.
[58] Haugland S, Wold B. Subjective health complaints in
adolescence - reliability and validity of survey methods.
J Adolescence 2001;24:61124.
[59] Yin Z, Davis CL, Moore JB, Treiber FA. Physical activity
buffers the effects of chronic stress on adiposity in youth.
Ann Behav Med 2005;29(1):2936.
[60] Norris R, Carroll D, Cochrane R. The effects of physical
activity and exercise training on psychological stress and
well-being in an adolescent population. J Psychosom Res
1992;36(1):5565.
[61] Brown JD. Staying fit and staying well: physical fitness as a
moderator of life stress. J Pers Soc Psychol 1991;60:55561.
[62] Carmack CL, Boudreaux E, Amaral-Melendez M,
Brantley PJ, De Moor C. Aerobic fitness and leisure physical
activity as moderators of the stress-illness relation. Ann Beh
Med 1999;21(3):2517.
[63] Roth DL. Acute emotional and psychophysical effects of
aerobic exercise. Psychophysiol 1989;26:593602.
[64] Young DR. Can cardiorespiratory fitness moderate the
negative effects of stress on coronary artery disease risk
factors. J Psychosom Res 1994;38(5):4519.
[65] Manning MR, Fusilier MR. The relationship between stress
and health care use: an investigation of the buffering roles of
personality, social support and exercise. J Psychosom Res
1999;47(3):15973.
[66] Siu O, Cooper CL, Leung T. Three-wave trend study of
managerial stress in Hong Kong: the role of type A behavior
and exercise. Int J Stress Manag 2000;7(2):1537.
[67] Skirka N. The relationship of hardiness, sense of coherence,
sport participation, and gender to perceived stress and
psychological symptoms among college students. J Sports
Med Phys Fitness 2000;40(1):6370.
[68] Fuchs R, Hahn A. Physical exercise and anxiety as moderators of the stress-illness relationship. Anxiety, Stress, and
Coping 1992;5:13949.
[69] Fuchs R, Hahn A, Schwarzer R. Effekte sportlicher Aktivitat
auf Selbstwirksamkeits-Erwartung und Gesundheit in
einer stressreichen Lebenssituation [Effects of exercise on
818
[70]
[71]
[72]
[73]
[74]
[75]
[76]
[77]
[78]
[79]
[80]
[81]
[82]
[83]
[84]
[85]

self-efficacy and health in a stressful life situation].
Sportwissenschaft 1994;24(1):6781.
Fuchs R, Leppin A. Sportliche Aktivitat, sozialer Ruckhalt
und Lebensstress als Determinanten der psychischen
Gesundheit [Exercise, social support and life stress as
determinants of mental helath]. Sportpsychologie
1992;6:1319.
Unger JB, Anderson Johnson C, Marks G. Functional
decline in the elderly: evidence for direct and stress-buffering
protective effects of social interactions and physical activity.
Ann Beh Med 1997;19(2):15260.
Gerber M, Puhse U. Dont crack under pressure!
The influence of leisure time physical activity and selfesteem on the stress-illness-relationship among adolescents.
J Psychosom Res 2008;65(4):3639.
Roth DL, Holmes DS. Influence of physical fitness in
determining the impact of stressful life events on physical
and psychological health. Psychosom Med 1985a;
47(2):16473.
Roth DL, Holmes DS. Influence of aerobic exercise training
and relaxation training on physical and psychologic health
following stressful life events. Psychosom Med
1985b;49:35565.
Rothlisberger C, Calmonte R, Seiler R. Sport, Stress und
emotionaler Ruckhalt als Determinanten von Gesundheitsund Lebenszufriedenheit bei Adoleszenten. Eine zweijahrige
Longitudinalstudie [Exercise, stress and emotional support
as determinants of health and life satisfaction among
adolescents. A two-year longitudinal study]. Psychologie
und Sport 1997;4(3):92101.
Gogoll A. Belasteter Geist Gefahrdeter Korper. Sport,
Stress und Gesundheit im Kindes- und Jugendalter
[Threatened mind endangered body. Exercise, stress and
health during childhood and adolescence]. Schorndorf:
Hofmann; 2004.
Singh GK, Kogan MD, Yu SM. Disparities in obesity and
overweight prevalence among US immigrant children and
adolescents by generational status. J Comm Health
2009;34(4):27181.
Roth DL, Wiebe DJ, Fillingim RB, Shay KA. Life events,
fitness, hardiness, and health: a simultaneous analysis of
proposed stress-resistance effects. J Pers Soc Psychol
1989;57:13642.
Holmes TH, Rahe RH. The Social Readjustment Scale.
J Psychosom Res 1967;11:21318.
Sarason IG, Johnson JH, Siegel JM. Assessing the impact of
life changes: development of the Life Experiences Survey.
J Consult Clin Psychol 1978;46(5):93246.
Cohen S, Kamarck T, Mermelstein R. A global measure of
perceived stress. J Health Soc Behav 1983;24:38596.
Gerber M, Puhse U. Dont crack under pressure
The influence of leisure time physical activity and selfesteem on the stress-illness-relationship among adolescents.
J Psychosom Res 2008;65(4):3639.
Cohen J. Statistical power analysis for the behavioral
sciences. Mahwah: Erlbaum; 1988.
Trusty J, Thompson B, Petrocelli JV. Practical guide
for reporting effect size in quantitative research in the
Journal of Counseling & Development. J Couns Dev
2004;82:10710.
Haugland S, Wold B, Torsheim T. Relieving the pressure?
The role of physical activity in the relationship between
school-related stress and adolescent health complaints. Res
Q Exerc Sport 2003;74(2):12735.
[86] McClelland GH, Judd CM. Statistical difficulties of

detecting interactions and moderator effects. Psychol Bull
1993;114(2):37690.
[87] Jaccard J, Wan CK, Turrisi R. The detection and interpretation of interaction effects between continuous variables in
multiple regression. Multivariate Behav Res 1990;
25(4):46778.
[88] Kellow TJ. Beyond statistical significant tests: the importance of using other estimates of treatment effects to
interpret evaluation results. Am J Eval 1998;19(1):12334.
[89] Selye H. The general adaptation syndrome and the diseases
of adaptation. J Endocrinol 1946;6:117230.
[90] Dishman RK, Jackson EM. Exercise, fitness, and stress. Int
J Sport Psychol 2000;31:175203.
[91] Mason JW. A reevaluation of the concept of nonspecificity in stress theory. J Psychiatr Res 1971;
80:32333.
[92] Lazarus RS, Folkman S. Stress, appraisal, and coping.
New York: Springer; 1984.
[93] Hobfoll SE. Stress, culture, and community. The psychology and philosophy of stress. New York: Plenum Press;
1998.
[94] Biddle SJ, Mutrie N. Psychology of physical activity:
Determinants, well-being and interventions. London:
Routledge; 2006.
[95] Wyler AR, Masuda M, Holmes TH. Seriousness of Illness
Rating Scale. J Psychosom Res 1968;11:36375.
[96] Maddi SR, Bartone PT, Puccetti MC. Stressful
events are indeed a factor in physical illness: reply to
Schroeder and Costa (1984). J Pers Soc Psychol
1987;52(4):83343.
[97] Dohrenwend BS, Dohrenwend BP, Dodson M, Shrout PE.
Symptoms, hassles, social support, and life events: problem
of
confounded
measures.
J
Abnorm
Psychol
1984;93:22230.
[98] Kumar R. Research methodology. A step-by-step guide for
beginners. London: Sage; 2005.
[99] Meeusen R. Physical activity and neurotransmitter release.
In: Acevedo EO, Ekkekakis P, editors. Psychobiology of
physical activity. Champaign: Human Kinetics; 2006.
pp. 12944.
[100] Kvetnansky R. Recent progress in catecholamines under
stress. In: Usdin ER, Kvetnansky R, Kopin IJ, editors.
Catecholamines and stress: recent advances. New York:
Elsevier; 1980. pp. 17.
[101] Crews DJ, Landers DM. A meta-analytic review of aerobic
fitness and reactivity to psychosocial stressors. Med Sci
Sports Exerc 1987;19:11430.
[102] Jackson EM, Dishman RK. Cardiorespiratory fitness and
laboratory stress: a meta-regression analysis. Psychophysiol
2006;43:5772.
[103] Taylor AC. Physical activity, anxiety, and stress.
In: Biddle SJ, Fox KR, Boutcher SH, editors. Physical
activity and psychological well-being. London: Routledge;
2000. pp. 1045.
[104] Boutcher SH, Hamer M. Psychobiological reactivity, physical activity, and cardiovascular health. In: Acevedo EO,
Ekkekakis P, editors. Psychobiology of physical activity.
Champaign: Human Kinetics; 2006. pp. 16176.
[105] Laux L, Vossel G. Theoretical and methodological
issues in achievement-related stress and anxiety research.
In: Krohne HW, Laux L, editors. Achievement,
stress, and anxiety. Washington: Hemisphere; 1982.
pp. 318.

[106] Ekkekakis P, Acevedo EO. Affective responses to acute
exercise: toward a psychobiological dose-response model.
In: Acevedo EO, Ekkekakis P, editors. Psychobiology of
physical activity. Champaign: Human Kinetics; 2006.
pp. 91110.
[107] Antonini PR, Seiler R, Mengisen W. Relationship of
coping style with type of sport. Percept Mot Skills
2004;98:47986.
[108] Youngstedt DF, Freelove-Charton JD. Exercise and sleep.
In: Faulkner GEJ, Taylor AH, editors. Exercise,
health and mental health. London: Routledge; 2005.
pp. 15989.
819
[109] Seelig H. Korperliche Aktivitat im Zusammenspiel mit

anderen Gesundheitsverhaltensweisen [The interplay
between exercise and other health-related behaviours].
In: Fuchs R, Gohner W, Seelig H, editors. Aufbau eines
korperlich-aktiven Lebensstils: Theorie, Empirie und
Praxis. Gottingen: Hogrefe; 2007. pp. 11737.
[110] Steptoe A, Kimbell J, Basford P. Exercise and the experience and appraisal of daily stressors. A naturalistic study.
J Behav Med 1998;21:36374.
[111] Csikszentmihalyi M, Larson R. Validity and reliability of
the experience sampling method. J Nerv Ment Dis 1987;
175(9):52636.

Kedokteran

Încărcat de

Informații document

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

Kedokteran

Încărcat de

Drepturi de autor:

Formate disponibile

Scandinavian Journal of Public Health, 2009; 37: 801819

Do exercise and fitness protect against stress-induced health

stress reactions of the human organism to address

Downloaded from sjp.sagepub.com at UNIVERSITY OF ALBERTA LIBRARY on April 25, 2015

M. Gerber & U. Puhse

contributes to lowered stress perceptions, for

Physical activity, exercise and fitness as stress-buffers

Personal & social

Downloaded from sjp.sagepub.com at UNIVERSITY OF ALBERTA LIBRARY on April 25, 2015

Exercise, fitness and stress-induced health complaints: a review

interaction effects were significant were classified as

Downloaded from sjp.sagepub.com at UNIVERSITY OF ALBERTA LIBRARY on April 25, 2015

Subjects, n, age, sample

Male managers (n 137,

University students (n 373,

University students (n 110,

Male law enforcement officers

US National Health Interview

University students (n 135,

Roth et al. 1989,

Downloaded from sjp.sagepub.com at UNIVERSITY OF ALBERTA LIBRARY on April 25, 2015

Cross-sectional (nonrandom sample, no

Cross-sectional and prospective (prospective

Study design (timeframe,

Table I. Studies with adult samples and cross-sectional design.

ANOVA (median split)

Data analysis, measures

Stress-buffer: Partial support

Stress " <4 Health #

Stress " <4 Health #

Stress " <4 Health #

Stress-buffer: Partial support

Stress-buffer: No support (neither

Stress " <4 Health #

Stress <//4 Health

Stress-buffer: Full support

Two- and three-way interactions,

Stress " <4 Health #

Downloaded from sjp.sagepub.com at UNIVERSITY OF ALBERTA LIBRARY on April 25, 2015

Employees in health promotion

Three cohorts of male managers and MBA students

Varsity athletes and non-athletes (n 270, 136m, 134f,

Community health survey

University students (n 270,

Siu et al. 2000,

Ensel & Lin

Cross-sectional (nonrandom sample, return

Cross-sectional (nonrandom sample, controlled for gender)

Cross-sectional (nonrandom sample, return

Cross-sectional (nonrandom sample, nonathletes: not member of

Cross-sectional (nonrandom sample, controlled for age, gender,

Data analysis, measures

Study design (timeframe,

Stress-buffer: Full support

Stress-buffer: Full support

Stress-buffer: Full support

Stress " <4 Health #

Stress " <4 Health #

Stress " <4 Health #

Stress " <4 Health #

Stress " <4 Health #

Two- and three-way interactions,

Stress " <4 Health #