Association between burnout and circulating levels of pro- and
anti-inflammatory cytokines in schoolteachers
Roland von Känela, Silja Bellingrathb, Brigitte M. Kudielkab,⁎
aDivision of Psychosomatic Medicine, Inselspital, Bern University Hospital and University of Bern, Switzerland
bDepartment of Theoretical and Clinical Psychobiology, Graduate School of Psychobiology, University of Trier, Germany
Received 21 September 2007; received in revised form 5 February 2008; accepted 5 February 2008
Objective: The burnout syndrome has been associated with an
increased risk of cardiovascular disease. The physiological
mechanisms potentially involved in this link are underexplored.
Knowing that a chronic low-grade systemic inflammatory state
contributes to atherosclerosis, we investigated circulating cytokine
levels in relation to burnout symptoms. Methods: We studied 167
schoolteachers (median, 48 years; range, 23–63 years; 67%
women) who completed the Maslach Burnout Inventory with its
three subscales emotional exhaustion (EE), lack of accomplishment
(LA), and depersonalization (DP). Levels of the proinflammatory
cytokine tumor necrosis factor (TNF)-α and of the anti-inflamma-
tory cytokines interleukin (IL)-4 and IL-10 were determined in
fasting morning plasma samples. The TNF-α/IL-4 ratio and the
TNF-α/IL-10 ratio were computed as two indices of increased
inflammatory activity. Analyses were adjusted for demographic
factors, medication, lifestyle factors (including sleep quality),
metabolic factors, and symptoms of depression and anxiety.
Results: Higher levels of total burnout symptoms aggregating the
EE, LA, and DP subscales independently predicted higher TNF-α
levels (ΔR2=.024, P=.046), lower IL-4 levels (ΔR2=.021, P=.061),
and a higher TNF-α/IL-4 ratio (ΔR2=.040, P=.008). Higher levels
of LA predicted decreased IL-4 levels (ΔR2=.041, P=.008) and a
higher TNF-α/IL-4 ratio (ΔR2=.041, P=.007). The categorical
dimensions of the various burnout scales (e.g., burnout yes vs. no)
showed no independent relationship with any cytokine measure.
Conclusion: Burnout was associated with increased systemic
inflammation along a continuum of symptom severity rather than
categorically. Given that low-grade systemic inflammation pro-
motes atherosclerosis, our findings may provide one explanation
for the increased cardiovascular risk previously observed in burned-
© 2008 Elsevier Inc. All rights reserved.
Keywords: Burnout; Cardiovascular disease; Cytokines; Inflammation; Psychological stress
Burnout in the workplace as a consequence of prolonged
occupational stress has recently been associated with an
increased risk of cardiovascular disease [1–3]. The physio-
logical underpinnings linking burnout with atherosclerosis
are not fully understood . Low-grade systemic inflamma-
tion as evidenced by an increase in circulating levels of
proinflammatory cytokines and acute phase reactants on the
one hand and by a decrease in levels of anti-inflammatory
cytokines on the other crucially contributes to atherosclerosis
initiation and progression [4–6]. Such enhanced inflamma-
tion activity might contribute to the increased cardiovascular
risk in burned-out individuals . Plasma levels of the
proinflammatory cytokine tumor necrosis factor (TNF)-α
were higher in apparently healthy female employees with
high burnout than in women with low burnout, whereas
levels of C-reactive protein (CRP) were similar between
groups . Another study found a positive association
between burnout and CRP levels in female employees but
not in their male counterparts . Emotional exhaustion is a
core component of burnout . Feelings of vital exhaustion
Journal of Psychosomatic Research 65 (2008) 51–59
⁎Corresponding author: Department of Theoretical and Clinical
Psychobiology, Graduate School of Psychobiology, University of Trier,
Johanniterufer 15, D-54290 Trier, Germany. Tel.: +49 651 201 2981;
fax: +49 651 201 3690.
E-mail address: firstname.lastname@example.org (B.M. Kudielka).
0022-3999/08/$ – see front matter © 2008 Elsevier Inc. All rights reserved.
positively correlated with circulating levels of interleukin
(IL)-6 in subjects with coronary artery disease  and in
community-dwelling subjects . Elevated plasma levels of
TNF-α, CRP, and IL-6 have been associated with an
increased risk of future myocardial infarction independent
of other risk factors [12–14].
IL-4 and IL-10 are commonly viewed as anti-inflamma-
tory cytokines  since they may, for instance, reduce
production of proinflammatory cytokines, including TNF-α
and IL-6, by human monocytes stimulated with lipopoly-
saccharide . Decreased circulating IL-10 levels were
shown to predict increased risk of myocardial infarction .
Decreased IL-4 might accelerate atherosclerosis by virtue of
reducing matrix metalloproteinase-1 production by smooth
muscle cells, thereby enhancing remodeling of the arterial
We aimed to further explore the relationship between
burnout and cytokine measures. We determined levels of
proinflammatory TNF-α and anti-inflammatory IL-4 and
IL-10 in the circulation of teachers who are a group at
considerably high risk of developing burnout . We
focused our analyses on TNF-α because, during inflamma-
tion, this cytokine is secreted before IL-1 and IL-6;
therefore, TNF-α might be comparably more sensitive to
subtle changes in the inflammation cascade under psycho-
logical stress [20,21]. Although simplified, cytokines have
been functionally divided into those produced by T helper 1
(Th1) cells (e.g., TNF-α) and those produced by T helper 2
(Th2) cells (e.g., IL-4 and IL-10). Whereas Th1 cytokines
activate the cellular immune response resulting in monocyte
and macrophage activation, Th2 cytokines activate the
humoral immune response and antibody production .
Particularly in patients with acute coronary syndromes and
stable coronary artery disease, the Th1-to-Th2 cytokine
ratio was shifted toward Th1 dominance . The TNF-α/
IL-10 balance may play an important role in various
diseases . Elevated plasma TNF-α/IL-4 ratio was
associated with greater proinflammatory activity of circulat-
ing monocytes during acute myocardial infarction  and
implied in the pathophysiology of bipolar disorder .
Given the potentially important role of a shift in the balance
between pro- and anti-inflammatory cytokines toward the
former in different somatic (including cardiovascular) and
psychiatric diseases, we also computed the ratio between
TNF-α and IL-4 and between TNF-α and IL-10. Both ratios
are indices of proinflammatory activity taking into account
the balance between pro- and anti-inflammatory cytokine
activity . We hypothesized that more severe burnout
would be associated with higher TNF-α levels on the one
hand and lower IL-4 and IL-10 levels on the other. We
further hypothesized a direct association between burnout
severity and the TNF-α/IL-4 and TNF-α/IL-10 ratios.
Inflammation is affected by many factors. Therefore, we
controlled in our analysis for demographics, medication,
lifestyle factors, metabolic factors, and symptoms of
depression and anxiety, postulating that the relationship
between burnout and cytokine measures would be inde-
pendent of these covariates.
To measure burnout, we applied the widely used Maslach
Burnout Inventory (MBI), which defines burnout across the
three dimensions emotional exhaustion (EE), lack of
accomplishment (LA), and depersonalization (DP) .
There is an unresolved debate about the pros and cons of
conceptualizing burnout as a unidimensional versus a three-
dimensional construct and, moreover, whether burnout
should be investigated continuously or categorically (i.e.,
assuming a threshold effect) in terms of its impact on
outcomes [9,29–33]. Also, some experts strongly advocate
focusing burnout research on the dimension of exhaustion
, whereas others argue for a concept of burnout that must
equally consider its dimensions LA and DP . We did not
favor any particular conceptualization of burnout given that
the research on biological risk markers for cardiovascular
disease related to burnout is still a nascent field . An even-
handed approach to the different measurements of burnout
offers the advantage of not prematurely neglecting possible
relationships between some burnout concepts and inflamma-
tion. A broad analysis can also contribute to the generation of
hypotheses for future research on cardiovascular risk
markers in burnout. Accordingly, we performed analyses
considering burnout as a unidimensional (i.e., a syndrome
that aggregates the three symptom clusters EE, LA, and DP)
as well as a multidimensional concept (i.e., separate analyses
for each of the underlying dimensions EE, LA, and DP) .
We not only further tested for a continuous association
between burnout (i.e., the degree of burnout severity) and
cytokines  but also considered burnout as a dichotomous
construct (i.e., teachers with burnout vs. those without
burnout), applying a categorization algorithm to the MBI
scales as previously proposed in a nonclinical Dutch nor-
mative sample .
Study participants and recruitment
We approached teachers of all major school types in the
region of Trier (Germany) and Luxembourg by means of
personal visits in local schools and by newspaper announce-
ments. Demographic data (age, gender) and current health
status (including use of beta-blocker, thyroid hormones, and
statins) were assessed during a telephone screening.
Volunteers with a psychiatric disorder, those medicated
with corticosteroids or psychotropic drugs, those with
diabetes, or pregnant women were not eligible. Eligible
teachers received questionnaires via postal mail for the
psychometric assessment as detailed below.
After an overnight fast, they were invited to a morning lab
visit. This included the assessment of lifestyle factors
(smoking status, alcohol consumption), metabolic factors
[resting blood pressure, heart rate, body mass index (BMI)],
52R. von Känel et al. / Journal of Psychosomatic Research 65 (2008) 51–59
and a venous blood draw for the measurement of cytokines,
lipids, and glucose levels. The mean arterial pressure was
computed by the formula (2/3 diastolic blood pressure plus
1/3 systolic BP) and used in statistical analysis.
The ethic committees of the State Medical Association of
Rheinland Pfalz and the University of Trier approved the
study protocol. All participants provided written informed
consent and were paid 50 euros as an incentive after
completion of the study.
The severity of burnout symptoms was assessed with the
widely used 22-item MBI , applying its validated
German version . The scoring of the individual items
for the three dimensions EE (nine items), LA (eight items),
and DP (five items) was anchored on a Likert scale ranging
from 0 (never) to 6 (every day). Accordingly, the maximum
range is 0 to 54 points for the MBI EE subscale, 0 to 48
points for the MBI LA subscale, 0 to 30 points for the MBI
DP subscale, and 0 to 132 points for the MBI total symptom
scale (i.e., summing up individual scores of the three
subscales) . For the categorical definition of burnout,
teachers were regarded to score high (as opposed to low) on
the respective scales if EE≥20 points, LA≥28 points, and
DP≥8 points . A diagnosis of burnout (yes/no) was
assigned if they fulfilled the previously described “exhaus-
tion+1” criterion (i.e., scoring high on EE and at least on
either LA or DP) .
Depression and anxiety
Symptom levels of depression and anxiety were assessed
by the German version  of the Hospital Anxiety and
Depression Scale (HADS)  consisting of seven items for
each dimension. Answers are coded on a 4-point Likert scale
(0=not at all, 3=mostly), giving rise to a total depression and
anxiety score ranging between 0 and 21 points.
Subjective sleep quality was assessed by the Jenkins
Sleep Questionnaire (JSQ) . The JSQ has been designed
to track common sleep problems in clinical populations and
has been successfully applied in a German-speaking
population . The JSQ comprises four items asking into
(a) difficulty in initiating sleep, (b) awakening during the
night, (c) awakening during sleep with difficulty maintaining
sleep, and (d) awakening exhausted in the morning despite
having slept as usual. All items are rated on a Likert scale (0–
5) yielding a maximum score of 20 points. Higher scores
indicate more sleep problems and poorer sleep quality.
Biological data were determined by a commercial labo-
ratory (Synlab, Trier, Germany; blood lipids and glucose), by
the Biochemical Laboratory of the Department of Theoretical
and Clinical Psychobiology at the University of Trier,
Germany (TNF-α), and by the Thrombosis Research Labora-
tory at the University Hospital of Bern, Switzerland (IL-4 and
IL-10). Fasting venous blood was collected into citrate tubes
for cytokine measurements and into serum tubes for blood
lipids and glucose (Sarstedt, Nümbrecht, Germany). Citrate
tubes were instantaneously stored on ice; all samples were
immediately centrifuged at 4°C for 15 min at 2000×g in an
adjacent room and pipetted into aliquots. Within 60 min,
right away or frozen at −80°C until further analysis (IL-4
Plasma levels of TNF-α (sensitivity, 0.5 pg/ml), IL-4
(sensitivity, 0.20 pg/ml), and IL-10 (sensitivity, 0.20 pg/ml)
were determined by duplicates of ultrasensitive enzyme-
linked immunosorbent assays obtained from R&D Systems
Europe, Ltd., Abingdon, UK (kit for TNF-α) and Biosource
International, Camarillo, CA, USA (kits for IL-4 and IL-10)
as per manufacturers' instructions. Approximately 85% of
plasma IL-4 levels were below the sensitivity level of the
assay (b0.20 pg/ml) and were accurately interpolated from
the standard curve. Inter- and intra-assay coefficients of
variation for all measures were b10%.
Data were analyzed using SPSS 13.0 for Windows (SPSS
Inc., Chicago, IL). The significance level was set at P≤.05
(two tailed). Normality of the data distribution was verified
normally distributed and required normalization, which was
achieved by log transformation (sleep quality, total choles-
terol/HDL cholesterol ratio, triglycerides, glucose), square
root transformation (depressive symptoms), and Blom
normal score transformation (age, BMI, all MBI scales, all
cytokine measures) . Data are given as medians with
interquartile range of raw values or as percentages of
proportions. Figures show transformed values to illustrate
the distribution of normalized data.
Student's t test and Pearson's chi-square test (Fisher's
Exact Test where appropriate) were used to test for group
differences of variables. Pearson correlation analysis was
applied to estimate the bivariate correlation coefficient
between two variables. We employed hierarchical linear
regression analysis, using forced entry, to identify a
continuous association between burnout scales and cytokine
measures independent of five blocks of a priori determined
control variables (cf. legend to Table 2). In order to prevent
model overfitting, given our sample size of 167 subjects, we
limited the number of control variables to 16, corresponding
to a minimum of 10 observations per predictor variable; this
technique generally allows for good estimates in multiple
linear regression models . Because previous literature
suggests that burnout-related inflammation might differ
between men and women , we additionally probed
53R. von Känel et al. / Journal of Psychosomatic Research 65 (2008) 51–59
whether gender and burnout measures would significantly
interact in predicting cytokine levels; however, they did not.
Therefore, we report on regression models considering the
of covariance (ANCOVA) was then computed to test for
differences in cytokine measures between categorized
burnout measures (yes/no) independent of covariates. Effect
sizes are expressed as Cohen's d for two independent groups,
R2in regression analysis, and partial eta-squared (ηp
termed small, medium, and large, respectively .
We did not adjust significance level for multiple tests
because of three reasons. Firstly, the function of pro- and
anti-inflammatory cytokines in relation to stress  and the
psychological meaning of the different dimensions of the
MBI  are quite distinct from each other. Secondly, we
preestablished a specific hypothesis of a positive relationship
between burnout severity and inflammation activity .
Thirdly, investigations of an atherosclerotic risk in burnout
are only emerging . If these premises are met, some
authors do not recommend adjustment of P values for
multiple comparisons since this might deem a truly
important association insignificant, thereby unnecessarily
misleading future research in a developing field .
the burned-out and non-burned-out teachers. Twenty-one
percent of teachers met the criteria for a categorical diagnosis
of burnout. Burnout was more frequent among men than
women. Sleep quality was poorer and the level of depressive
andanxietysymptomswas higherinthe teacherswithburnout
relative to their non-burned-out counterparts. As expected,
teachers with burnout scored higher on all burnout scales
than teachers without burnout.
Dimensional association between burnout symptoms
Total burnout symptoms
symptom level and IL-4 (r=−.17, P=.032; Fig. 1A) and
the TNF-α/IL-4 ratio (r=.17, P=.026; Fig. 1B).
Table 2 shows the hierarchical linear regression models
for the relationship between the level of total burnout
symptoms and the five cytokine measures with subsequent
adjustment for five blocks of control variables. In the last
model, total burnout symptoms were directly associated with
TNF-α levels and the TNF-α/IL-4 ratio, explaining 2% and
4% of the respective variance. Also, in the last model, there
was an inverse association between total burnout symptoms
and IL-4 levels showing borderline significance. IL-10 levels
and the TNF-α/IL-10 ratio were not significantly predicted
by the level of total burnout symptoms in any model. Of
note, the independent associations between total burnout
symptoms and TNF-α levels and IL-4 levels reached
significance and borderline significance, respectively, only
with adjustment for symptoms of depression and anxiety in
the last step. In contrast, the independent relationship
between total burnout symptoms and the TNF-α/IL-4 ratio
was significant in virtually all steps, suggesting little
confounding by the control variables.
Characteristics of the 167 teachers studied as per burnout category
Variable Burnout (n=35) No burnout (n=132)P value
Thyroid hormone (%)
Statin drug (%)
Current smoker (%)
Alcohol consumption (days/week)
Sleep quality (score)
Mean arterial blood pressure (mmHg)
Heart rate (beats/min)
Total cholesterol/HDL cholesterol ratio
Serum triglycerides (mg/dl)
Fasting glucose (mg/dl)
HADS depression (score)
HADS anxiety (score)
MBI EE (score)
MBI LA (score)
MBI DP (score)
MBI total symptoms (score)
P values were determined by Student's t test or chi-square test. Values are given as percentages or median with interquartile range (untransformed raw scores).
HDL, high-density lipoprotein.
54R. von Känel et al. / Journal of Psychosomatic Research 65 (2008) 51–59
The full models for TNF-α, IL-4, IL-10, the TNF-α/IL-4
ratio, and the TNF-α/IL-10 ratio explained between 4% and
18% of the total variance; however, only the model for the
TNF-α/IL-4 ratio reached significance. Of all the control
variables, older age independently predicted a higher TNF-α/
IL-4 ratio (P=.013).
Fig. 1. The scatterplots with fit line show the significant associations between total burnout symptoms and IL-4 levels (A) and the TNF-α/IL-4 ratio (B) as well as
between LA and IL-4 levels (C) and the TNF-α/IL-4 ratio (D). All values are given after transformation to normal scores using the Blom formula.
Multivariate relationships between total burnout symptoms and cytokine measures (N=167)
IL-4IL-10 TNF-α/IL-4 ratio
R2of the full model
The columns show the unique variance of cytokine measures explained by the MBI total symptom score (ΔR2) with standardized β coefficients and P values
after subsequent adjustment for age and gender in Block 1 (demographics); beta-blockers, thyroid hormones, and statins in Block 2 (medication); smoking
status, alcohol consumption, and sleep quality in Block 3 (lifestyle factors); mean arterial blood pressure, heart rate, BMI, total cholesterol/high-density
lipoprotein cholesterol ratio, serum triglyceride levels, and fasting glucose levels in Block 4 (metabolic factors); and symptoms of depression and anxiety in
Block 5. Values are given as transformed scores.
.221 (F=1.27).142 (F=1.40).996 (F=0.32).021 (F=1.91) .838 (F=0.66)
55R. von Känel et al. / Journal of Psychosomatic Research 65 (2008) 51–59
There were no significant bivariate correlations between
levels of EE and any cytokine measure. Also, none of the
cytokine measures were significantly predicted by EE in the
last models controlling for all 16 covariates.
Lack of accomplishment
There were bivariate correlations between LA and IL-4
levels (r=−.21, P=.006; Fig. 1C) and the TNF-α/IL-4 ratio
(r=.17, P=.027; Fig. 1D). Taking into account all 16
covariates, higher level of LA was maintained as an
independent predictor of decreased IL-4 levels (ΔR2=.41,
β=−.24, P=.008) and of an elevated TNF-α/IL-4 ratio
(ΔR2=.41, β=.24, P=.007). In these models, a higher
depressive symptom level predicted higher IL-4 levels
(P=.041) and older age predicted a higher TNF-α/IL-4
There were no bivariate correlations between levels of DP
and any cytokine measure. Also, DP did not significantly
predict any cytokine measure in the last model accounting
for all five blocks of covariates.
Categorical association between burnout scales
Table 3 shows the crude differences in cytokine
measures between the different categories of burnout
dimensions. There was a trend for a significantly higher
TNF-α/IL-4 ratio in burned-out teachers relative to their
non-burned-out counterparts (P=.086; Cohen's d=.35). In
addition, IL-4 levels were lower in teachers with LA versus
those without LA (P=.026; Cohen's d=.42). When control-
ling for all covariates, teachers with DP showed a trend
toward statistical significance for higher TNF-α levels
compared to teachers without DP (P=.093; ηp
were no significant differences in any cytokine measure
between the other categories of burnout measures control-
ling for all covariates.
A growing body of research suggests that burnout
emerging from long-standing stress at work is associated
with an increased cardiovascular risk [1–3], whereby
enhanced inflammation might be one psychobiological
pathway linking burnout with atherosclerosis [3,7,8]. Our
study provides further support for this notion. Defining
burnout as a unidimensional construct, we found that more
severe burnout symptomatology was positively associated
with enhanced inflammatory activity in the circulation of
schoolteachers even when controlling for covariates. More
precisely, we found a continuous and independent relation-
ship between total burnout symptoms and elevated levels of
Bivariate relationships between categorical dimensions of burnout symptoms and cytokine measures (N=167)
Burnout (EE+1 criterion)
Values are given as median with interquartile range (untransformed raw scores).
⁎Pb.10 (for comparisons between groups).
⁎⁎Pb.05 (for comparisons between groups).
56 R. von Känel et al. / Journal of Psychosomatic Research 65 (2008) 51–59
TNF-α on the one hand and decreased levels of IL-4 on the
other. Moreover, the TNF-α/IL-4 ratio was also directly
associated with total burnout symptoms. This suggests that
burnout was associated not only with elevated proinflam-
matory activity but also with decreased anti-inflammatory
activity. Accordingly, the variance explained in the TNF-α/
IL-4 ratio — a measure of the balance between pro- and anti-
inflammatory forces — by the total burnout symptom score
was comparably greater than the respective variances
explained in TNF-α and IL-4 levels alone.
In contrast, we did not find an association between
burnout and plasma levels of the anti-inflammatory cytokine
IL-10 and the TNF-α/IL-10 ratio. Somewhat counterintui-
tive, cultured monocytes from individuals with burnout
previously released higher amounts of IL-10 when stimu-
lated with lipopolysaccharide in vitro than did monocytes
from individuals without burnout; in addition, stimulated
TNF-α release was not different between groups .
However, stimulated release of cytokines from cells reflects
dynamics that are likely different from steady-state cytokine
activity in the circulation.
In a post hoc analysis, we found that the continuous
relationship between the severity of burnout and cytokine
measures was vastly confined to the dimension of LA but
less so to the dimensions of EE and DP. Elevated levels of EE
is usually regarded the principal component of the burnout
construct, although some authors feel that restricting burnout
to feelings of exhaustion alone may simply yield burnout as
an alternative label for fatigue with little specific benefit for
research in the work stress area . In light of the
unequivocal importance the exhaustion dimension has
attracted in conceptualizing burnout , the lack of an
association between EE and cytokine measures in our study
seems puzzling and, to some extent, may weaken our
observation of an association between burnout and circulat-
ing levels of cytokines. However, previous studies showing
an association between burnout and TNF-α levels  and
CRP levels  did not specifically investigate whether some
of the different burnout dimensions were particularly related
to inflammation activity. Studies investigating feelings of
exhaustion alone found a positive association between the
construct of vital exhaustion and circulating levels of IL-6
[10,11]. However, we did not measure IL-6, which is
different in its immune function from TNF-α and IL-4
[46,47]. As a whole, we feel that it is premature to favor one
burnout dimension over any other in terms of a potentially
higher explanatory value in predicting inflammation activity.
Based on the current evidence, a parsimonious interpretation
is that unidimensional constructs of burnout relate to
inflammation at least as strong as individual dimensions of
We found that measuring burnout along a continuum of
severity yielded greater evidence for enhanced inflammation
in burned-out individuals than did the analyses in which
burnout was conceptualized as a categorical dimension.
Indeed, the categorical dimensions of the total burnout
symptom scale (i.e., burnout yes vs. no) and of the three
burnout subscales did not significantly differentiate between
cytokine measures when controlling for covariates. Being
aware that the proportion of teachers meeting a categorical
dimension of burnout was comparably small, thereby
limiting statistical power, we suggest, based on our data,
that future research on perturbed physiology in burnout
might particularly profit from modeling burnout as a
continuous measure. This reasoning is supported by our
previous study on traumatized individuals in whom we
found that posttraumatic stress along a continuum of severity
showed more robust associations with TNF-α and IL-4 levels
than did a categorical diagnosis of posttraumatic stress
It is a strength of our study that we were able to control for
several covariates of cytokine function and burnout. For
instance, poor sleep and depression are highly prevalent in
burned-out individuals [48,49] and are both associated with
inflammation [50,51]. In light of studies suggesting that the
vagus nerve modulates the production of proinflammatory
cytokines, including TNF-α , we also controlled for heart
rate, which is a proxy measure of parasympathetic outflow to
the heart . It seemed particularly important to disentangle
the contribution of burnout from that of depression as well as
of anxiety to cytokine measures. Burnout emerged as a
unique predictor of TNF-α and IL-4 levels only after
adjusting for depression and anxiety in the last step of
Our study also has its limitations. A large proportion of
subjects had low IL-4 levels such that we had to interpolate
these from standard curves. However, this source of a
potential unreliability in our IL-4 measures may only
underscore the assumption of a true association between
burnout and IL-4 levels and the TNF-α/IL-4 ratio. Due to
funding restrictions, we were unable to determine levels of
other proinflammatory cytokines, particularly IL-6, which
have been directly related to exhaustion [10,11] and chronic
psychosocial stress [54,55]. The study was cross-sectional;
therefore, our cytokine measures do not mirror the dynamics
of the cytokine network, where, for example, anti-inflam-
matory cytokines increase to counteract an increase in
proinflammatory cytokines. Also, any inference to a causal
link between burnout and enhanced inflammation or vice
versa (i.e., “sickness behavior”) cannot be made. For
statistical analyses, we had to transform several variables
to obtain a normal distribution that precludes the interpreta-
tion of results in reference to the original measures used.
Although dichotomizing of continuous measures should be
avoided because of a loss of power and potential inflation of
false-positive results , it was inevitable to dichotomize
burnout scores in order to investigate burnout as a categorical
construct. Moreover, we did not perform arbitrary categor-
ization of burnout dimensions but applied previously
suggested cutoff points developed in the Netherlands from
a Dutch normative sample . Although cross-national use
of cutoffs to classify individuals by burnout scores has its
57 R. von Känel et al. / Journal of Psychosomatic Research 65 (2008) 51–59
caveats [34,57], German, Luxembourgian, and Dutch
samples might be quite similar in their characteristics.
In summary, our data suggest in broad terms that burnout
is associated with an imbalance between pro- and anti-
inflammatory forces by increasing proinflammatory and
decreasing anti-inflammatory activity of the innate immune
system. We found that burnout was associated with enhanced
inflammation activity in the circulation of burned-out
schoolteachers. This association was particularly seen
when burnout was conceptualized as a unidimensional
construct and along a continuum of severity. Our study
corroborates previous findings of a chronic low-grade
inflammatory state in burnout. The study also provides
further evidence for the assumption that enhanced inflam-
mation is one explanation for the increased atherosclerotic
risk previously observed in burned-out individuals.
The authors express their gratitude to André Haeberli and
Monika Stutz for laboratory assistance. B.M.K. is supported
by Emmy Noether Research Grant KU 1401/4-1 and KU
1401/4-2 of the German Research Foundation. B.M.K. and
S.B. are members of the International Research Training
Group funded by the DFG (GRH 1389/1).
 Honkonen T, Ahola K, Pertovaara M, Isometsa E, Kalimo R, Nykyri E,
Aromaa A, Lonnqvist J. The association between burnout and physical
illness in the general population — results from the Finnish Health
2000 Study. J Psychosom Res 2006;61:59–66.
 Toppinen-Tanner S, Ojajarvi A, Vaananen A, Kalimo R, Jappinen P.
Burnout as a predictor of medically certified sick-leave absences and
their diagnosed causes. Behav Med 2005;31:18–27.
 Melamed S, Shirom A, Toker S, Berliner S, Shapira I. Burnout and risk
of cardiovascular disease: evidence, possible causal paths, and
promising research directions. Psychol Bull 2006;132:327–53.
 Willerson JT, Ridker PM. Inflammation as a cardiovascular risk factor.
 Tziakas DN, Chalikias GK, Kaski JC, Kekes A, Hatzinikolaou EI,
Stakos DA, Tentes IK, Kortsaris AX, Hatseras DI. Inflammatory and
anti-inflammatory variable clusters and risk prediction in acute
coronary syndrome patients: a factor analysis approach. Atherosclero-
 Kilic T, Ural D, Ural E, YumukZ, Agacdiken A, Sahin T, KahramanG,
Kozdag G, Vural A, Komsuoglu B. Relation between proinflammatory
to anti-inflammatory cytokine ratios and long-term prognosis in
patients with non-ST elevation acute coronary syndrome. Heart 2006;
 Grossi G, Perski A, Evengard B, Blomkvist V, Orth-Gomer K.
Physiological correlates of burnout among women. J Psychosom Res
 Toker S, Shirom A, Shapira I, Berliner S, Melamed S. The association
between burnout, depression, anxiety, and inflammation biomarkers:
C-reactive protein and fibrinogen in men and women. J Occup Health
 Cox T, Tisserand M, Taris T. The conceptualization and measurement
of burnout: questions and directions. Work Stress 2005;19:187–91.
 Janszky I, Lekander M, Blom M, Georgiades A, Ahnve S. Self-rated
health and vital exhaustion, but not depression, is related to
inflammation in women with coronary heart disease. Brain Behav
 Sjögren E, Leanderson P, Kristenson M, Ernerudh J. Interleukin-6
levels in relation to psychosocial factors: studies on serum, saliva, and
in vitro production by blood mononuclear cells. Brain Behav Immun
 Ridker PM, Rifai N, Pfeffer M, Sacks F, Lepage S, Braunwald E.
Elevation of tumor necrosis factor-alpha and increased risk of recurrent
coronary events after myocardial infarction. Circulation 2000;101:
 Sakkinen P, Abbott RD, Curb JD, Rodriguez BL, Yano K, Tracy RP.
C-reactive protein and myocardial infarction. J Clin Epidemiol 2002;
 Ridker PM, Rifai N, Stampfer MJ, Hennekens CH. Plasma concentra-
tion of interleukin-6 and the risk of future myocardial infarction among
apparently healthy men. Circulation 2000;101:1767–72.
 Elenkov IJ, Chrousos GP. Stress hormones, proinflammatory and anti-
inflammatory cytokines, and autoimmunity. Ann N YAcad Sci 2002;
 De Waal Malefyt R, Figdor CG, Huijbens R, Mohan-Peterson S,
Bennett B, Culpepper J, Dang W, Zurawski G, de Vries JE. Effects of
IL-13 on phenotype, cytokine production, and cytotoxic function of
human monocytes. Comparison with IL-4 and modulation by IFN-
gamma or IL-10. J Immunol 1993;151:6370–81.
 Seljeflot I, Hurlen M, Solheim S, Arnesen H. Serum levels of
interleukin-10 are inversely related to future events in patients with
acute myocardial infarction. J Thromb Haemost 2004;2:350–2.
 Sasaguri T, Arima N, Tanimoto A, Shimajiri S, Hamada T, Sasaguri Y.
A role for interleukin 4 in production of matrix metalloproteinase 1 by
human aortic smooth muscle cells. Atherosclerosis 1998;138:247–53.
 Unterbrink T, Hack A, Pfeifer R, Buhl-Griesshaber V, Müller U,
Wesche H, Frommhold M, Scheuch K, Seibt R, Wirsching M, Bauer J.
Burnout and effort–reward-imbalance in a sample of 949 German
teachers. Int Arch Occup Environ Health 2007;80:433–41.
 Papanicolaou DA, Wilder RL, Manolagas SC, Chrousos GP. The
pathophysiologic roles of interleukin-6 in human disease. Ann Intern
 von Känel R, Hepp U, Kraemer B, Traber R, Keel M, Mica L,
Schnyder U. Evidence for low-grade systemic proinflammatory
activity in patients with posttraumatic stress disorder. J Psychiatr Res
 Paul WE, Sedar RA. Lymphocyte responses and cytokines. Cell 1994;
Nakken B, Szegedi G, Soltesz P. TH1/TH2 imbalance, measured by
circulating and intracytoplasmic inflammatory cytokines—immunolo-
gical alterations in acute coronary syndrome and stable coronary artery
disease. Scand J Immunol 2006;64:336–44.
 Elenkov IJ, Chrousos GP, Wilder RL. Neuroendocrine regulation of
IL-12 and TNF alpha/IL-10 balance. Clinical implications. Ann N Y
Acad Sci 2000;917:94–105.
 Yao L, Huang K, Huang D, Wang J, Guo H, Liao Y. Acute myocardial
infarction induced increases in plasma tumor necrosis factor-alpha and
interleukin-10 are associated with the activation of poly(ADP-ribose)
 Kim YK, Jung HG, Myint AM, Kim H, Park SH. Imbalance between
pro-inflammatory and anti-inflammatory cytokines in bipolar disorder.
J Affect Disord 2007;104:91–5.
 Gomez D, Correa PA, Gomez LM, Cadena J, Molina JF, Anaya JM.
Th1/Th2 cytokines in patients with systemic lupus erythematosus: is
tumor necrosis factor alpha protective? Semin Arthritis Rheum 2004;
 Maslach C, Jackson S. Maslach Burnout Inventory Manual. 2nd ed.
Palo Alto, CA: Consulting Psychologists Press, 1986.
 Schaufeli WB, Taris TW. The conceptualization and measurement of
burnout: common ground and worlds apart. Work Stress 2005;19:
58R. von Känel et al. / Journal of Psychosomatic Research 65 (2008) 51–59
 Shirom A. Reflections on the study of burnout. Work Stress 2005;19: Download full-text
 Kristensen TS, Borritz M, Villadsen E, Christensen KB. The
Copenhagen Burnout Inventory: a new tool for the assessment of
burnout. Work Stress 2005;19:192–207.
 Brenninkmeijer V, VanYperen N. How to conduct research on burnout:
advantagesanddisadvantagesof a unidimensional approach in burnout
research. Occup Environ Med 2003;60:i16–i20.
 Maslach C, Schaufeli WB, Leiter MP. Job burnout. Annu Rev Psychol
 Schaufeli WB, van Dierendonck D. A cautionary note about the cross-
national and clinical validity of the cut-off points for the Maslach
Burnout Inventory. Psychol Rep 1995;76:1083–90.
 Schwarzer R, Jerusalem M. Skalen zur Erfassung von Lehrer- und
Schülermerkmalen Dokumentation der psychometrischen Verfahren
im Rahmen der Wissenschaftlichen Begleitung des Modellversuchs
selbstwirksame Schulen. Berlin, online web version, 2001.
 Buunk BP, Ybema JF, Gibbons FX, Ipenburg M. The affective
consequences of social comparison as related to professional burnout
and social comparison orientation. Eur J Soc Psychol 2001;31:337–51.
 Herrmann C, Buss U, Snaith RP. Hospital Anxiety and Depression
Scale — Deutsche Version. Ein Fragebogen zur Erfassung von Angst
und Depressivität in der somatischen Medizin. Bern: Huber, 1995.
 Zigmond AS, Snaith RP. The Hospital Anxiety and Depression Scale.
Acta Psychiatr Scand 1983;67:361–70.
 Jenkins CD, Stanton BA, Niemcryk SJ, Rose RM. A scale for the
estimation of sleep problems in clinical research. J Clin Epidemiol
 Kudielka BM, von Känel R, Gander ML, Fischer JE. Effort–reward
imbalance, overcommitment and sleep in a working population. Work
 Blom G. Statistical estimates and transformed beta-variables. New York:
 Babyak MA. What you see may not be what you get: a brief,
nontechnical introduction to overfitting in regression-type models.
Psychosom Med 2004;66:411–21.
 Cohen J. Statistical power analysis for the behavioral sciences. 2nd ed.
Hillsdale, NJ: Lawrence Erlbaum Associates, 1988.
 Perneger TV. What's wrong with Bonferroni adjustments? BMJ 1998;
 Mommersteeg PM, Heijnen CJ, Kavelaars A, van Doornen LJ.
 Chrousos GP. The hypothalamic–pituitary–adrenal axis and immune-
mediated inflammation. N Engl J Med 1995;352:1351–62.
 Nelms K, Keegan AD, Zamorano J, Ryan JJ, Paul WE. The IL-4
receptor: signalling mechanisms and biologic functions. Annu Rev
 Sonnenschein M, Sorbi MJ, van Doornen LJ, Schaufeli WB, Maas CJ.
Evidence that impaired sleep recovery may complicate burnout
improvement independently of depressive mood. J Psychosom Res
 Ahola K, Hakanen J. Job strain, burnout, and depressive symptoms: a
prospective study among dentists. J Affect Disord 2007;104:103–10.
 Mills PJ, von Känel R, Norman D, Natarajan L, Ziegler MG, Dimsdale
JE. Inflammation and sleep in healthy individuals. Sleep 2007;30:
 Pasic J, Levy WC, Sullivan MD. Cytokines in depression and heart
failure. Psychosom Med 2003;65:181–93.
 Tracey KJ. Physiology and immunology of the cholinergic antiin-
flammatory pathway. J Clin Invest 2007;117:289–96.
 Thayer JF, Lane RD. The role of vagal function in the risk for
cardiovascular disease and mortality. Biol Psychol 2007;74:224–42.
 Lutgendorf SK, Garand L, Buckwalter KC, Reimer TT, Hong SY,
Lubaroff DM. Life stress, mood disturbance, and elevated interleukin-
6 in healthy older women. J Gerontol A Biol Sci Med Sci 1999;54:
 Ranjit N, Diez-Roux AV, Shea S, Cushman M, Seeman T, Jackson SA,
Ni H. Psychosocial factors and inflammation in the multi-ethnic study
of atherosclerosis. Arch Intern Med 2007;167:174–81.
 Royston P, Altman DG, Sauerbrei W. Dichotomizing continuous
predictors in multiple regression: a bad idea. Stat Med 2006;25:
 Van Horn J, Schaufeli WB, Greenglass ER, Burke RJ. Burnout among
teachers: a Canadian–Dutch comparison. Psychol Rep 1997;81:
59 R. von Känel et al. / Journal of Psychosomatic Research 65 (2008) 51–59