Journal of Psychosomatic Research 98 (2017) 27–33

Contents lists available at ScienceDirect

Journal of Psychosomatic Research

journal homepage: www.elsevier.com/locate/jpsychores

Effects and mediators of psychosocial work characteristics on somatic MARK

symptoms six years later: Prospective findings from the Mannheim
Industrial Cohort Studies (MICS)
Raphael M. Herra,b,⁎,1, Jian Lib,1, Adrian Loerbroksb, Peter Angererb, Johannes Siegristc,
Joachim E. Fischera
a
Mannheim Institute of Public Health, Social and Preventive Medicine, Medical Faculty Mannheim, Heidelberg University, Germany
b
Institute of Occupational, Social and Environmental Medicine, Centre for Health and Society, Faculty of Medicine, Heinrich-Heine-University of Düsseldorf, Germany
c
Senior Professorships on Work Stress Research, Life Science Center, Faculty of Medicine, Heinrich-Heine-University of Düsseldorf, Germany

A R T I C L E I N F O A B S T R A C T

Keywords: Objective: Ample evidence documented the adverse health effects of work stressors, and recent research has
Effort-reward imbalance increasingly focused on somatic symptoms which are very common and costly. Prospective evidence is however
Mental health sparse and yielded mixed findings. Furthermore, there is reason to assume that depression and anxiety might
Cohort study mediate the effects of adverse psychosocial work conditions on somatic symptoms. This study aimed to
Somatic symptoms
investigate longitudinal effects of work stressors on somatic symptoms and the potential mediation by anxiety
Psychosocial work characteristics
and/or depression.
Methods: Six year follow-up data from 352 individuals – free of potentially stress-related chronic disease – were
utilized. Somatic symptoms were assessed by 19 items of an established list of complaints at baseline and follow-
up. The effort-reward-imbalance (ERI) model measured adverse psychosocial work conditions and over-
commitment (OC). Linear regressions adjusted for socio-demographics, social status, lifestyle, and baseline
symptoms estimated the effects of the ERI ratio, effort, reward, OC, and the ERI ratio × OC interaction on
somatic symptoms six years later. Furthermore, single and multiple mediation by anxiety and/or depression was
investigated.
Results: There was a strong longitudinal effect of the ERI ratio, as well as of its subcomponents, and OC on
somatic symptoms (all Bs ≥ | 0.49|; p-values ≤0.004). Moreover, the ERI ratio × OC interaction was significant
(p-value = 0.047). Multiple mediation analyses revealed especially anxiety to mediate the effect of work
stressors on somatic symptoms (Sobel test = 0.007).
Conclusion: Adverse psychosocial work conditions seem to longitudinally affect somatic symptoms, potentially
moderated by OC, and mediated by anxiety.

1. Introduction Disorders (4th Edition) and the International Classification of Diseases

The effects of work stressors have been extensively studied and
include health-related consequences such as cardiovascular disease [1],
depression [2], and musculoskeletal complaints [3]. More recently,
research has begun to focus on the association of work stressors with
somatic symptoms and other health complaints, such as pain [4,5],
burnout [6], and sleep disturbances [7,8]. Single or multiple somatic
symptoms may progress to or signal pre-stages of somatoform disorders
(SFD), as represented in the Diagnostic and Statistical Manual of Mental
(10th Edition) [9]. Somatic symptoms “are characterized by patterns of
persistent bodily complaints for which adequate examination does not
reveal sufficiently explanatory structural or other specified pathology”
([10]; page 946). Usually these symptoms are medically unexplained
and cannot be attributed to clinical diseases with clear diagnostic
criteria. Somatic symptoms are very frequent in primary care settings,
incur high societal cost, elicit extensive diagnostic examinations and
treatment [11], contribute to sickness absence due to reduced work
ability [12], and result in substantial economic loss [13].

⁎
Corresponding to: R. Herr, Mannheim Institute of Public Health, Social and Preventive Medicine, Medical Faculty Mannheim, Heidelberg University, Ludolf-Krehl-Strasse 7-11, 68167
Mannheim, Germany.
E-mail addresses: Raphael.Herr@medma.uni-heidelberg.de (R.M. Herr), Jian.Li@uni-duesseldorf.de (J. Li), Adrian.Loerbroks@uni-duesseldorf.de (A. Loerbroks),
Peter.Angerer@uni-duesseldorf.de (P. Angerer), Johannes.Siegrist@med.uni-duesseldorf.de (J. Siegrist), Joachim.Fischer@medma.uni-heidelberg.de (J.E. Fischer).
1
Equal contribution.

http://dx.doi.org/10.1016/j.jpsychores.2017.05.003
Received 3 January 2017; Received in revised form 29 March 2017; Accepted 3 May 2017
0022-3999/ © 2017 Elsevier Inc. All rights reserved.
R.M. Herr et al. Journal of Psychosomatic Research 98 (2017) 27–33

The causes of somatic symptoms are not fully understood [10]. It

has been suggested that psychological stress generally contributes to
psychophysiological changes that alter perception, that promotes a
biased appraisal of symptoms [9]. Among the potential stressors,
adverse psychosocial working conditions, i.e., work stressors, have
received special attention [14,15].
Ample evidence from cross-sectional studies has documented that
adverse psychosocial work characteristics are associated with somatic
complaints. This evidence covers various types of adverse psychosocial
work conditions, including those conceptualized by the job-demand
control (JDC) (support) model [16–23], and the effort reward imbal-
ance (ERI) model [21,24–28]. Studies were conducted in various
occupational subpopulations – ranging from Asian garment workers
[28,29] to managers in Germany [22].
This cross-sectional evidence is supplemented by four prospective
studies. Leitner and Resch [30] evaluated individual stressors quantita-
tively in terms of extra work (minutes/week) and found them to be
correlated with somatic complaints eight years later. Applying a
standardized questionnaire, Parkes and colleagues found somatic
symptoms to be predicted after two months by a three way de-
mand × skill discretion × support interaction at baseline [31]. Godin
and colleagues [17] reported onset of job stressors, and one-year
cumulative job stressors measured by the effort-reward model to be
associated with somatization after one year. Furthermore, Pereira and
Elfering [32] found social stressors at work (i.e., social climate and
conflict with co-workers) to predict somatic complaints six weeks later.
One of the internationally established psychosocial work stress
model is the effort-reward imbalance (ERI) model. It is both theoreti-
cally sound and empirically corroborated and validated [33,34]. The
ERI model conceptualizes work situations to be stressful in which
efforts are insufficiently reciprocated. Given its focus on the work
contract it can be applied to all types of employment contracts and
occupational groups, while other theoretical models of work stress are
assumed to be particularly useful for distinct occupational groups (e.g.,
the job-demand-control model for blue-collar workers, or organiza-
tional injustice for white-collar occupations) [5]. It is based on three
main assumptions [35,36]: first, high efforts (i.e., demands and
obligations) in combination with low rewards (i.e., money, esteem,
and job security/career opportunities) increase the risk of poor health
(extrinsic ERI hypothesis). Second, high levels of over-commitment
(OC) independently increase the health risk (intrinsic over-commitment
hypothesis), and, third, the combination of high OC and ERI carries the
highest risk of adverse health (interaction hypothesis). The ERI model Fig. 1. Sampling procedure.
repeatedly confirmed its validity by predicting poor self-reported or
medically recorded health in a variety of studies (for details see reviews
between psychosocial work characteristics and somatic symptoms.
[35–37]).
Third, we will provide preliminary evidence of the mediating role of
Mental health seems to be of particular importance for the relation-
mental health in the association of adverse psychosocial work char-
ship of work stressors with somatic symptoms. Psychosocial work
acteristics and somatic symptoms.
characteristics are predictive of anxiety and depression [2,38], which
in turn are risk factors for somatic symptoms [39–41]. Thus, anxiety
and depression might mediate the effect of work stressors on somatic 2. Methods
symptoms.
Our study will make novel contributions to the current literature by 2.1. Participants
addressing three specific aims. First, we aim to provide the first long-
term longitudinal data (i.e., covering six years) examining an estab- This study used longitudinal data from the Mannheim Industrial
lished work stressor model – i.e., the ERI model, its sub-components Cohort Studies (MICS). Specifically, we linked combined assessments in
(i.e., effort and reward), and OC – as potential predictors of somatic 2000 and 2002 (baseline; T1) to subsequent assessment in 2007
symptoms. In addition, the moderating effect of OC on the ER ratio (follow-up; T2), as we have done previously [42]. Briefly, in 2000, a
upon somatic symptoms will be investigated. Among the four long- stratified random sample of 647 individuals was drawn from the total
itudinal studies, the measures of work stressors were not theory-based workforce at one production site of an airplane manufacturing company
in two studies [30,32]. The other two studies [17,31], however, in Germany (n = 1760; Fig. 1). In total, 537 (83%) of those individuals
examined the short-term longitudinal effects on somatic symptoms participated. At the same production site, an additional representative
only (no more than one year). Thus, our study with a standard measure sample of 1117 employees was invited in 2002 and 816 (73%) persons
of work stressors, aims at examining the long-term effects on somatic participated. Data from both surveys were combined into a single
symptoms (i.e., six years). Second, to improve conclusions about sample (2000/02), comprising 1353 individuals. Of these, 676 persons
causality a cross-lagged model examines bidirectional relationships (50%) could be followed up until 2007 (mean follow-up = 6.3 years).

28
R.M. Herr et al.
Four hundred eight persons (69%) had no missing values on relevant
variables. The present analyses comprised 352 persons, who were free
of major chronic disease at baseline (i.e., myocardial infarction or other
heart problems, asthma or chronic lung disease, chronic pains).
Comparing follow-up participants and non-participants with regard to
all baseline characteristics, we found participants to be more frequently
male, younger, and to smoke. Regarding ERI, its components, and
depression and anxiety there were no differences. The study sample
reported less somatic symptoms at baseline, compared to the drop-outs,
which could be explained by the exclusion criteria. When we compared
participants and drop-outs without chronic disease at baseline, no
significant differences regarding somatic symptoms were found. Parti-
cipants provided written informed consent, and information was
provided according to the Declaration of Helsinki. The local institu-
tional review board approved the study (2007- 009E-MA).
2.2. Procedures
At baseline (2000/02) participants received a standardized oral
briefing and completed the questionnaire in groups of 10 to 25 persons
in a quiet room away from the workplace. No incentives were offered
but time to participate in the study was recorded as working time. At
follow-up (2007) participants completed the questionnaire at home and
returned it the next day.
2.3. Measures
2.3.1. Adverse psychosocial work conditions
Psychosocial work characteristics were assessed by the validated
ERI questionnaire, measuring effort with six items, reward with 11
items (consisting of four items esteem, two items job security, and four
items job promotion), and OC with six items [34]. Responses to the
items of ‘effort’ and ‘reward’ were scored on a 5-point scale where a
value of 1 indicates no respective stressful experience, and a value of 5
indicates a highly stressful experience. An example of effort is 'I am
often pressured to work overtime', and of reward is 'Considering all my
efforts and achievements, my salary/income is adequate'. The ‘over-
commitment’ items are scored on a 4-point scale (1 = full disagree-
ment, 4 = full agreement with the respective statement). An example of
over-commitment is: 'Work rarely lets me go, it is still on my mind when
I go to bed'. Cronbach's alphas were 0.70, 0.85, and 0.82 for effort,
reward and over-commitment at baseline, and 0.75, 0.85, and 0.81 at
follow-up, respectively. Sum scores were computed with higher scores
reflecting higher effort, reward and OC with potential scoring ranges of
6–30, 11–55 and 6–24, respectively. Furthermore, a weighted ratio
between the effort and reward was calculated (ER ratio) [34].
2.3.2. Somatic symptoms
Somatic symptoms were measured by items from an established
questionnaire (B-L Beschwerdeliste [43]). This instrument uses 24 items
to assess the extent to which respondents suffer from specific com-
plaints. The checklist of complaints covers different domains, these are,
cardiovascular symptoms (two items); pain at the head and neck (two
items); gastrointestinal complaints (three items); sensory issues in terms
of sensations of heat (two items) and cold (one item); motor function
complaints (two items); musculoskeletal pain (two items); sleep
problems (two items); emotional exhaustion (two items); respiratory
symptoms (one item); psychological complaints (three items); and two
specific items measuring weight loss and “tiredness in the legs”,
respectively [44]. In 2000/02 (T1) all 24 items were included and
rated by each respondent. In 2007 (T2) the items on sleep (two items)
and psychological complaints (three items) were not included however,
as we decided to use more extended instruments to asses those types of
complaints in greater detail. Therefore, this analysis was restricted to
19 items measured in both waves. Response categories ranged from
0 = not at all to 2 = often. Items were categorized as 0 = symptom
29
Journal of Psychosomatic Research 98 (2017) 27–33
does not apply and 1 = symptom applies. Cronbach's alpha were 0.87
for baseline, and 0.80 for follow up. Sum scores calculated a somatic
symptoms score ranging from 0 to 19.
2.3.3. Mental health
Mental health was assessed using the 14 items of the Hospital
Anxiety and Depression Scale (HADS) [45] at baseline, measuring
anxiety (Cronbach's alpha = 0.78) and depression (Cronbach's al-
pha = 0.76) with seven items each on a 4 point scale (scored as 0 to
3). An example item for depression is “I feel cheerful” (reverse), and for
anxiety is “I get a sort of frightened feeling as if something awful is
about to happen”. A sum score was computed for both scales, ranging
potentially from 0 to 21, with higher values indicative more severe
symptoms of anxiety and depression.
2.4. Statistical analysis
Linear regression models estimated the effect of the ER ratio, effort,
reward, and OC on somatic complaints six years later. In addition,
adding a multiplicative interaction term to the model, which also
involved both main effects, assessed if OC moderated the effect of the
ER ratio on somatic symptoms. ER ratio was logarithmically trans-
formed to approach normal distribution and all scales were standar-
dized (z-transformation). Unstandardized (B) and standardized (Beta)
regression coefficients, and proportion of explained variance (R2) were
reported. To avoid floor or ceiling effects, all analyses were adjusted for
baseline values of the somatic symptoms score. Three regression models
with gradual adjustment were calculated. Adjustment was a-priori
defined based on established associations of covariates with adverse
working conditions or somatic symptoms (e.g., [46,47]). The first
model controlled for age and sex. The second model additionally
adjusted for socioeconomic status (i.e., professional education, job
position, and income), and the third model also adjusted for lifestyle
factors (i.e., smoking status, alcohol consumption, physical exercise).
Structural equation modelling to test cross-lagged relationships
between psychosocial work characteristics and somatic symptoms was
applied using IBM SPSS Amos Version 23. Estimates based on the
maximum likelihood method. A reciprocal model taking forward and
reverse relationship into account was fitted. This model was adjusted
for age, sex, and marital status (Model 1 in linear regressions).
Measurement errors were allowed to correlate to improve model fit.
Model fit was assessed by χ2 test and χ2/DF, the comparative fit index
(CFI), and the root mean square error of approximation (RMSEA).
Potential mediation by depression and anxiety was examined by the
PROCESS macro (release 2.16.1) for SPSS by Hayes [48,49]. Direct and
indirect effects were estimated applying bias-corrected bootstrap 95%
confidence intervals, using 5000 bootstrap samples. In a first step,
potential mediation of the effect of work stressors (baseline) on somatic
symptoms (follow-up) was estimated separately for anxiety and depres-
sion (baseline). In a second step, both mediating variables were
assessed simultaneously (multiple mediation). Models were fully ad-
justed (Model 3) and variables were z-transformed. In a second set of
analyses the index of moderated mediation tested potential moderation
of the mediation by over-commitment [50]. These analyses were
performed using IBM SPSS Statistics Version 22.
3. Results
The baseline characteristics of the study population can be found in
Table 1. The majority was male (88%), with a mean age of 37 years. On
average 7.80 (SD = 4.86) somatic symptoms were reported at baseline,
and 6.55 (SD = 3.71) at follow up.
ER ratio, effort, reward, and OC were significantly associated with
somatic symptoms score at baseline and follow-up (r ≥ |0.221 |; p-
values < 0.001; Table 2). The baseline somatic symptoms score was
highly correlated with the follow-up score (r = 0.609; p-value <
R.M. Herr et al. Journal of Psychosomatic Research 98 (2017) 27–33

Table 1 p = 0.109, χ2/DF = 1.45, CFI = 0.989, RMSEA = 0.036). In line with
Sample characteristics at baseline. proceeding linear regression analyses, adverse psychosocial work
characteristics at T1 affected somatic symptoms at T2 (β = 0.180;
% or mean n or SD
p = 0.001; Fig. 2). Somatic symptoms at T1 were not related to
Gender: male (% n) 88.4% 311 psychosocial work characteristics at T2 however (β = −0.057;
Age (years, mean, SD) 36.89 9.99 p = 0.373).
Marital status (%, n)
In single mediation analyses, anxiety and depression mediated the
Single 32.4% 114
Married 61.9% 218 effect of the ER ratio on somatic symptoms, indicated by a significant
Divorced 5.1% 18 indirect effect (bias-corrected bootstrap 95% confidence intervals:
No answer 0.6% 2 anxiety = 0.0367–0.1249; depression = 0.0206–0.0919, respectively;
Professional education (%, n) Table 4). Multiple mediation revealed especially anxiety to mediate the
Apprentice 36.1% 127
effect of ER-ratio on somatic symptoms (bias-corrected bootstrap 95%
Technical college 16.2% 57
Academic degree 8.5% 30 confidence intervals: anxiety = 0.0208–0.1173 vs. depres-
No answer 39.2% 138 sion = −0.0109–0.0603).
Job position (%, n) Neither the single nor the multiple index of moderated mediation
Area manager 4.0% 14
was significant (95% bias-corrected bootstrap confidence intervals
Process owner 14.2% 50
Skilled worker 71.0% 250 included zero) [50]. The mediated association of the ER ratio with
Worker 4.0% 14 somatic symptoms by mental health seems not to be moderated by over-
Apprentice 5.4% 19 commitment.
No answer 1.4% 5
Monthly net income (%, n)
< 2000 € 22.7% 80 4. Discussion
2000 € to < 3000 € 20.5% 72
3000 € and more 13.6% 48
No answer 43.2% 152 This study aimed to explore the association between work stressors,
Smoking status: yes (%, n) 31.3% 110 – as assessed by the effort-reward-imbalance model, – and somatic
Alcohol consumption (gr/day, mean, SD) 14.02 15.85 complaints six years later and to test the potential mediating role of
Physical exercise (h/week, %, n)
anxiety and depression. There was a robust positive association of the
> 2 h/w 21.6% 76
1–2 h/w 13.9% 49 ER ratio, but also of its components effort, reward, and OC, with
< 1 h/w 14.2% 50 somatic complaints. Furthermore, we observed an interaction effect of
No physical exercise 10.8% 38 the ER ratio and OC with regard to the prediction of somatic symptoms.
No answer 39.5% 139 In addition, cross-lagged model supported the causality of the effect.
HADS-anxiety (mean, SD) 5.86 3.35
HADS-depression (mean, SD) 4.14 2.93
Finally, the association of ER ratio with somatic symptoms appeared to
Somatic symptoms T1 (mean, SD) 7.80 4.86 be mediated by mental health, in particular by anxiety. The moderated
Somatic symptoms T2 (mean, SD) 6.55 3.71 mediation analysis provided no evidence for moderation by over-
Effort reward ratio (mean, SD) 0.65 0.30 commitment, however.
Effort (mean, SD) 14.84 4.12
Multiple pathways may underlie the observed associations. The
Reward (mean, SD) 44.85 7.84
Over-commitment (mean, SD) 13.70 4.00 proposed mechanisms cover various domains, which are neither
exhaustive nor mutually exclusive and include cognitive mediators
SD = standard deviation. (e.g., increased attention to symptoms and dysfunctional attribution
styles), behavioral mediators (e.g., physical inactivity in order to avoid
0.001). experience of cardiovascular responses and pain), and physiological
In fully adjusted regression analyses, the ER ratio at baseline was arousal (i.e., neuro-endocrine responses under control of the hypotha-
significantly related to somatic symptoms six years later (B = 0.662; p- lamic-pituitary-adrenal (HPA) axis and the (para)sympathetic nervous
value < 0.001; Table 3). Effort showed the expected positive effect system) [9,51,52]. Those pathways are not mutually exclusive, but may
(B = 0.498; p-value = 0.002), while the effect of reward was negative in fact interact to translate adverse psychosocial work characteristics
(B = − 0.493; p-value = 0.004). In addition, levels of OC had a strong into somatic symptoms (e.g., catastrophic cognitive attributions that
effect on the somatic symptoms score (B = 0.694; p-value < 0.001) increase physiological arousal, thereby inducing a vicious circle).
and the interaction with ER ratio was statistically significant (p- The three main assumptions of the ERI model were confirmed in this
value = 0.047). Additional adjustment for lifestyle at follow-up did study. The ER ratio showed a marked effect on somatic symptoms
not change the pattern of associations (data not shown). (extrinsic ERI hypothesis), as did OC (intrinsic over-commitment
The cross-lagged model between psychosocial work characteristics hypothesis). It is worth highlighting that a significant interaction term
and somatic symptoms revealed a good fit to the data (χ2 = 23.19, with OC revealed a particularly adverse effect of the ER ratio on

Table 2
Pearson's correlations between study variables.

ERI Effort Reward OC Somatic symptoms T1 Somatic symptoms T2 Anxiety

Effort 0.801
Reward − 0.777 −0.378
OC 0.457 0.552 − 0.262
Somatic symptoms T1 0.275 0.290 − 0.221 0.267
Somatic symptoms T2 0.287 0.278 − 0.246 0.307 0.609
Anxiety 0.430 0.404 − 0.339 0.516 0.445 0.465
Depression 0.405 0.359 − 0.371 0.454 0.355 0.365 0.654

All correlations were significant (p < 0.001).

ERI = effort-reward imbalance.
OC = over-commitment.

30
R.M. Herr et al. Journal of Psychosomatic Research 98 (2017) 27–33

Table 3
Effects of effort-reward imbalance on somatic symptoms score six years later.

Model 1 Model 2 Model 3

B Beta P R2 B Beta P R2 B Beta P R2

ER ratio 0.633 0.171 < 0.001 0.404 0.657 0.177 < 0.001 0.414 0.662 0.179 < 0.001 0.415
Effort 0.463 0.128 0.004 0.393 0.489 0.135 0.003 0.403 0.498 0.138 0.002 0.404
Reward − 0.488 − 0.127 0.003 0.394 −0.489 −0.127 0.004 0.402 − 0.493 − 0.128 0.004 0.402
Over-commitment (OC) 0.625 0.173 < 0.001 0.414 0.671 0.185 < 0.001 0.426 0.694 0.192 < 0.001 0.427
ER ratio × OC − 1.419 − 0.383 0.090 0.425 −1.631 −0.441 0.053 0.437 − 1.692 − 0.457 0.047 0.439

ER ratio = ratio of effort-reward.

Model 1: adjustment for age, sex, marital status, and baseline somatic symptoms.
Model 2: Model 1 + additional adjustment for socioeconomic status (education, occupation, income).
Model 3: Model 2 + additional adjustment for lifestyle (smoking, alcohol consumption and physical exercise).

as quality of life and turnover behaviors). Among these studies, only

three reports [56–58] lent support to the interaction hypothesis.
Clearly, more research is needed to further test the interaction
hypothesis.
We also explored the potential mediating role of depression and
anxiety and thereby provided, to our knowledge, the first longitudinal
evidence on this specific research questions. Multiple mediation
analyses revealed anxiety to mediate the effect of ERI on somatic
symptoms, indicating adverse work stressors might result in higher
levels of anxiety, which in turn triggers higher levels of reported
somatic symptoms. It is of interest that we did not observe significant
mediating effect of depression in multiple mediation analysis. Here, the
high correlation between anxiety and depression (r = 0.65; Table 2)
deserves attention. It has been pointed out that the indirect effects
Fig. 2. Simplified illustration of reciprocal structural model between psychosocial work would be attenuated in multiple statistical modelling [59]. Thus, it
characteristics and somatic symptoms. Standardized regression coefficients (β); might be assumed that the potential mediating role of depression might
ERI = Effort reward imbalance; ***p ≤ 0.001; Adjusted for age, sex and marital status be masked by simultaneous consideration of anxiety.
(Model 1). Some methodological limitations and strengths of our study should
be taken into account when interpreting our findings. With regard to
Table 4 our measurement of work stressors the risk of information bias is
Single and multiple mediation of anxiety and depression of the effect of work stressors on
somatic symptoms.
limited as the ERI model was assessed by a standardized, psychome-
trically validated questionnaire [37]. Our assessment of somatic
Effect 95% C.I. p-Value symptoms was based on an abbreviated version of a German-language
instrument [43]. While the original questionnaire is well-established we
Lower Upper
cannot rule out that our decision to slightly reduce the number of items
Total effect 0.1418 0.056 0.2277 0.0013 has affected its validity to some extent. A further limitation concerns
Mediation by anxiety the sample attrition. Regrettably, only 50% of the baseline respondents
Direct effect 0.0668 −0.0236 0.1572 0.1471 participated at follow-up assessment, however, participants did not
Indirect effecta 0.0750 0.0367 0.1249 0.0002b
differ regarding ERI, its components, and depression and anxiety, as
Mediation by depression well as somatic symptoms (see drop-out analyses in methods section).
Direct effect 0.0922 0.0028 0.1817 0.0433 The long-term effects of six years, an extensive time period never
Indirect effecta 0.0496 0.0206 0.0919 0.0033b
explored before in this context, represent an advantage of this study.
Multiple mediation While it can be interpreted as the result of a chronically stressful
Direct effect 0.0601 − 0.031 0.1512 0.1953
experience we cannot rule out that unobserved factors occurring
Indirect effecta between the two measurement waves exerted some impact. As a
Total 0.0817 0.0421 0.1395
limitation at the methodological level, it should be mentioned that a
Anxiety 0.0613 0.0208 0.1173 0.0070b
Depression 0.0204 −0.0109 0.0603 0.2613b two- wave design does not allow an adequate test of mediation Finally,
our study population consisted mainly of men and was recruited from
5000 bootstrap samples. Adjusted for age, sex, marital status, baseline somatic symptoms, one single industry branch. Consequently, we were unable to conduct
socioeconomic status, and lifestyle (Model 3). gender-specific analyses, and these shortcomings restrict the general-
a
95% C.I. = bias-corrected bootstrap confidence interval.
b
ization of our findings to women and other occupational settings.
p-Value = normal-theory tests for indirect effect (Sobel test).
These limitations are balanced by several strengths. First, we were
able to adjust for a range of important confounders. Nevertheless,
somatic symptom reporting. This latter finding is of interest in view of
unmeasured or residual confounding (i.e., due to imperfect measure-
the fact that the interaction hypothesis was less often investigated in
ment of confounders) cannot be ruled out. The prospective design is a
earlier research on the ERI model. A recent systematic review [36]
further strength of our study as it introduces a temporal and potential
reveals that, in the past two decades, only 14 out of 27 studies reported
cause-effect sequence into our analyses. This constitutes a sine qua non
significant evidence for the interaction hypothesis. When focusing on
to establish causality of associations [60]. In addition, as demonstrated
the longitudinal evidence, six studies [53–58] were identified covering
by structural equation modelling, this study points to the direction of
various health-related outcomes (ranging from coronary health disease,
the effect under study: adverse psychosocial work conditions appear to
blood pressure, to mental disorders such as depression, burnout, as well

31
R.M. Herr et al.
affect somatic symptoms, but not vice versa. With respect to practical
implications our findings provide a solid entry point for stress manage-
ment interventions as they are based on a theoretical model that has
already demonstrated its effectiveness in reducing work-related stres-
sors [61,62].
In conclusion, this study provides evidence of direct and indirect
effects of an adverse psychosocial work environment on somatic
symptoms. Indirect effects are due to reduced mental health, in
particular anxiety. Moreover, a personal pattern of coping with work
demands, over-commitment, acts as a moderator of the association for
stressful work with somatic symptoms. The findings lend support to
theory-based intervention measures at the organizational as well as
individual level.
Competing interests
Prof. Fischer has received royalties for lectures regarding occupa-
tional health from various companies and public agents. Until 2012
Prof. Fischer was CEO and major shareholder of Health Vision GmbH,
who organized the data collection. The other authors have no compet-
ing interests to report.
Funding
This research did not receive any specific grant from funding
agencies in the public, commercial, or not-for-profit sectors.
References
[1] P.L. Schnall, M. Dobson, P. Landsbergis, Globalization, work, and cardiovascular
disease, Int. J. Health Serv. 46 (2016) 656–692.
[2] T. Theorell, A. Hammarstrom, G. Aronsson, L. Traskman Bendz, T. Grape,
C. Hogstedt, et al., A systematic review including meta-analysis of work environ-
ment and depressive symptoms, BMC Public Health 15 (2015) 738.
[3] S. Kraatz, J. Lang, T. Kraus, E. Munster, E. Ochsmann, The incremental effect of
psychosocial workplace factors on the development of neck and shoulder disorders:
a systematic review of longitudinal studies, Int. Arch. Occup. Environ. Health 86
(2013) 375–395.
[4] P. Koch, A. Schablon, U. Latza, A. Nienhaus, Musculoskeletal pain and effort-reward
imbalance—a systematic review, BMC Public Health 14 (2014) 37.
[5] R.M. Herr, J.A. Bosch, A. Loerbroks, A.E. van Vianen, M.N. Jarczok, J.E. Fischer,
et al., Three job stress models and their relationship with musculoskeletal pain in
blue- and white-collar workers, J. Psychosom. Res. 79 (2015) 340–347.
[6] A. Seidler, M. Thinschmidt, S. Deckert, F. Then, J. Hegewald, K. Nieuwenhuijsen,
et al., The role of psychosocial working conditions on burnout and its core
component emotional exhaustion - a systematic review, J. Occup. Med. Toxicol.
(London, England) 9 (2014) 10.
[7] S.J. Linton, G. Kecklund, K.A. Franklin, L.C. Leissner, B. Sivertsen, E. Lindberg,
et al., The effect of the work environment on future sleep disturbances: a systematic
review, Sleep Med. Rev. 23 (2015) 10–19.
[8] R. Rugulies, M. Norborg, T.S. Sorensen, L.E. Knudsen, H. Burr, Effort-reward
imbalance at work and risk of sleep disturbances. Cross-sectional and prospective
results from the Danish Work Environment Cohort Study, J. Psychosom. Res. 66
(2009) 75–83.
[9] M. Witthoft, W. Hiller, Psychological approaches to origins and treatments of
somatoform disorders, Annu. Rev. Clin. Psychol. 6 (2010) 257–283.
[10] P. Henningsen, S. Zipfel, W. Herzog, Management of functional somatic syndromes,
Lancet (London, England) 369 (2007) 946–955.
[11] A.J. Barsky, E.J. Orav, D.W. Bates, Somatization increases medical utilization and
costs independent of psychiatric and medical comorbidity, Arch. Gen. Psychiatry 62
(2005) 903–910.
[12] A. Aamland, K. Malterud, E.L. Werner, Phenomena associated with sick leave
among primary care patients with Medically Unexplained Physical Symptoms: a
systematic review, Scand. J. Prim. Health Care 30 (2012) 147–155.
[13] A. Konnopka, C. Kaufmann, H.H. Konig, D. Heider, B. Wild, J. Szecsenyi, et al.,
Association of costs with somatic symptom severity in patients with medically
unexplained symptoms, J. Psychosom. Res. 75 (2013) 370–375.
[14] M. Nakao, Work-related stress and psychosomatic medicine, BioPsychoSocial Med.
4 (2010) 4.
[15] M.K. Shoss, B.L. Shoss, Check-up time: a closer look at physical symptoms in
occupational health research, Stress. Health 28 (2012) 193–201.
[16] H.R. Eriksen, H. Ursin, Subjective health complaints: Is coping more important than
control? Work Stress. 13 (1999) 238–252.
[17] I. Godin, F. Kittel, Y. Coppieters, J. Siegrist, A prospective study of cumulative job
stress in relation to mental health, BMC Public Health 5 (2005) 67.
[18] M. Nishikitani, M. Nakao, K. Karita, K. Nomura, E. Yano, Influence of overtime
work, sleep duration, and perceived job characteristics on the physical and mental
32
Journal of Psychosomatic Research 98 (2017) 27–33
status of software engineers, Ind. Health 43 (2005) 623–629.
[19] H.R. Eriksen, C. Ihlebaek, J.P. Jansen, A. Burdorf, The relations between psycho-
social factors at work and health status among workers in home care organizations,
Int. J. Behav. Med. 13 (2006) 183–192.
[20] K. Nomura, M. Nakao, M. Sato, H. Ishikawa, E. Yano, The association of the
reporting of somatic symptoms with job stress and active coping among Japanese
white-collar workers, J. Occup. Health 49 (2007) 370–375.
[21] S. Yu, G. Gu, W. Zhou, S. Wang, Psychosocial work environment and well-being: a
cross-sectional study at a thermal power plant in China, J. Occup. Health 50 (2008)
155–162.
[22] M.C. Gadinger, J.E. Fischer, S. Schneider, D.D. Terris, K. Kruckeberg, S. Yamamoto,
et al., Gender moderates the health-effects of job strain in managers, Int. Arch.
Occup. Environ. Health 83 (2010) 531–541.
[23] N. Fischer, C. Degen, J. Li, A. Loerbroks, A. Muller, P. Angerer, Associations of
psychosocial working conditions and working time characteristics with somatic
complaints in German resident physicians, Int. Arch. Occup. Environ. Health 89
(2016) 583–592.
[24] N. van Vegchel, J. de Jonge, T. Meijer, J.P. Hamers, Different effort constructs and
effort-reward imbalance: effects on employee well-being in ancillary health care
workers, J. Adv. Nurs. 34 (2001) 128–136.
[25] L. Xu, J. Siegrist, W. Cao, L. Li, B. Tomlinson, J. Chan, Measuring job stress and
family stress in Chinese working women: a validation study focusing on blood
pressure and psychosomatic symptoms, Women Health 39 (2004) 31–46.
[26] I. Godin, F. Kittel, Differential economic stability and psychosocial stress at work:
associations with psychosomatic complaints and absenteeism, Soc. Sci. Med. (1982)
58 (2004) 1543–1553.
[27] S. Weyers, R. Peter, H. Boggild, H.J. Jeppesen, J. Siegrist, Psychosocial work stress
is associated with poor self-rated health in Danish nurses: a test of the effort-reward
imbalance model, Scand. J. Caring Sci. 20 (2006) 26–34.
[28] A. Buapetch, S. Lagampan, J. Faucett, S. Kalampakorn, The Thai version of Effort-
Reward Imbalance Questionnaire (Thai ERIQ): a study of psychometric properties
in garment workers, J. Occup. Health 50 (2008) 480–491.
[29] M. Steinisch, R. Yusuf, J. Li, O. Rahman, H.M. Ashraf, C. Strumpell, et al., Work
stress: its components and its association with self-reported health outcomes in a
garment factory in Bangladesh-findings from a cross-sectional study, Health Place
24 (2013) 123–130.
[30] K. Leitner, M.G. Resch, Do the effects of job stressors on health persist over time? A
longitudinal study with observational stressor measures, J. Occup. Health Psychol.
10 (2005) 18–30.
[31] K.R. Parkes, C.A. Mendham, C. von Rabenau, Social support and the demand–dis-
cretion model of job stress: tests of additive and interactive effects in two samples,
J. Vocat. Behav. 44 (1994) 91–113.
[32] D. Pereira, A. Elfering, Social stressors at work, sleep quality and psychosomatic
health complaints—a longitudinal ambulatory field study, Stress. Health 30 (2014)
43–52.
[33] J. Siegrist, Adverse health effects of high-effort/low-reward conditions, J. Occup.
Health Psychol. 1 (1996) 27–41.
[34] J. Siegrist, D. Starke, T. Chandola, I. Godin, M. Marmot, I. Niedhammer, et al., The
measurement of effort-reward imbalance at work: European comparisons, Soc. Sci.
Med. (1982) 58 (2004) 1483–1499.
[35] D. Montano, J. Li, J. Siegrist, The measurement of Effort-Reward Imbalance (ERI) at
work, in: J. Siegrist, M. Wahrendorf (Eds.), Work Stress and Health in a Globalized
Economy - The Model of Effort-Reward Imbalance, Springer, 2016, pp. 21–42.
[36] J. Siegrist, J. Li, Associations of extrinsic and intrinsic components of work stress
with health: a systematic review of evidence on the effort-reward imbalance model,
Int. J. Environ. Res. Public Health 13 (2016) 432.
[37] J. Siegrist, M. Wahrendorf, Work Stress and Health in a Globalized Economy: The
Model of Effort-Reward Imbalance: Springer, (2016).
[38] L.B. Finne, J.O. Christensen, S. Knardahl, Psychological and social work factors as
predictors of mental distress: a prospective study, PLoS One 9 (2014) e102514.
[39] T.T. Haug, A. Mykletun, A.A. Dahl, The association between anxiety, depression,
and somatic symptoms in a large population: the HUNT-II study, Psychosom. Med.
66 (2004) 845–851.
[40] I. Hassan, R. Ali, The association between somatic symptoms, anxiety disorders and
substance use. A literature review, Psychiatry Q. 82 (2011) 315–328.
[41] S. Kohlmann, B. Gierk, A. Hilbert, E. Brähler, B. Löwe, The overlap of somatic,
anxious and depressive syndromes: a population-based analysis, J. Psychosom. Res.
90 (2016) 51–56.
[42] P.M. Mommersteeg, R. Herr, J. Bosch, J.E. Fischer, A. Loerbroks, Type D personality
and metabolic syndrome in a 7-year prospective occupational cohort, J. Psychosom.
Res. 71 (2011) 357–363.
[43] D. Zerssen, The List of Complaints (BL) [Die Beschwerdeliste (BL)], Weinheim,
ManuaL Beltz, 1975.
[44] U. Baumann, R.-D. Stieglitz, The comparison of four lists of complaints [Ein
Vergleich von vier Beschwerdenlisten], Arch. Psychiatr. Nervenkr. 229 (1980)
145–163.
[45] A.S. Zigmond, R.P. Snaith, The hospital anxiety and depression scale, Acta
Psychiatr. Scand. 67 (1983) 361–370.
[46] J. Siegrist, Effort-reward imbalance at work and cardiovascular diseases, Int. J.
Occup. Med. Environ. Health 23 (2010) 279–285.
[47] K.H. Ladwig, B. Marten-Mittag, B. Formanek, G. Dammann, Gender differences of
symptom reporting and medical health care utilization in the German population,
Eur. J. Epidemiol. 16 (2000) 511–518.
[48] A.F. Hayes, Introduction to Mediation, Moderation, and Conditional Process
Analysis: A Regression-based Approach, Guilford Press, New York, 2013.
[49] A.F. Hayes, Beyond Baron and Kenny: statistical mediation analysis in the new
R.M. Herr et al.
millennium, Commun. Monogr. 76 (2009) 408–420.
[50] A.F. Hayes, An index and test of linear moderated mediation, Multivar. Behav. Res.
50 (2015) 1–22.
[51] K. Kozlowska, Stress, distress, and bodytalk: co-constructing formulations with
patients who present with somatic symptoms, Harv. Rev. Psychiatry 21 (2013)
314–333.
[52] V. Deary, T. Chalder, M. Sharpe, The cognitive behavioural model of medically
unexplained symptoms: a theoretical and empirical review, Clin. Psychol. Rev. 27
(2007) 781–797.
[53] M. Gilbert-Ouimet, C. Brisson, M. Vezina, A. Milot, C. Blanchette, Repeated
exposure to effort-reward imbalance, increased blood pressure, and hypertension
incidence among white-collar workers: effort-reward imbalance and blood pressure,
J. Psychosom. Res. 72 (2012) 26–32.
[54] H. Derycke, P. Vlerick, N. Burnay, C. Decleire, W. D'Hoore, H.M. Hasselhorn, et al.,
Impact of the effort–reward imbalance model on intent to leave among Belgian
health care workers: a prospective study, J. Occup. Organ. Psychol. 83 (2010)
879–893.
[55] C. Aboa-Eboule, C. Brisson, E. Maunsell, R. Bourbonnais, M. Vezina, A. Milot, et al.,
Effort-reward imbalance at work and recurrent coronary heart disease events: a 4-
year prospective study of post-myocardial infarction patients, Psychosom. Med. 73
(2011) 436–447.
[56] M. Wahrendorf, G. Sembajwe, M. Zins, L. Berkman, M. Goldberg, J. Siegrist, Long-
33
Journal of Psychosomatic Research 98 (2017) 27–33
term effects of psychosocial work stress in midlife on health functioning after labor
market exit—results from the GAZEL study, J. Gerontol. Ser. B Psychol. Sci. Soc. Sci.
67 (2012) 471–480.
[57] B. Buddeberg-Fischer, R. Klaghofer, M. Stamm, J. Siegrist, C. Buddeberg, Work
stress and reduced health in young physicians: prospective evidence from Swiss
residents, Int. Arch. Occup. Environ. Health 82 (2008) 31–38.
[58] T. Feldt, M. Huhtala, U. Kinnunen, K. Hyvönen, A. Mäkikangas, S. Sonnentag, Long-
term patterns of effort-reward imbalance and over-commitment: Investigating
occupational well-being and recovery experiences as outcomes, Work Stress. 27
(2013) 64–87.
[59] K. Preacher, A. Hayes, Asymptotic and resampling strategies for assessing and
comparing indirect effects in multiple mediator models, Behav. Res. Methods 40
(2008) 879–891.
[60] K.J. Rothman, S. Greenland, T.L. Lash, Modern Epidemiology: Lippincott
Williams & Wilkins, (2008).
[61] H. Limm, H. Gundel, M. Heinmuller, B. Marten-Mittag, U.M. Nater, J. Siegrist, et al.,
Stress management interventions in the workplace improve stress reactivity: a
randomised controlled trial, Occup. Environ. Med. 68 (2011) 126–133.
[62] C. Brisson, M. Gilbert-Ouimet, C. Duchaine, X. Trudel, M. Vézina, Workplace
interventions aiming to improve psychosocial work factors and related health, in:
J. Siegrist, M. Wahrendorf (Eds.), Work Stress and Health in a Globalized Economy -
The Model of Effort-reward Imbalance, Springer, 2016, pp. 333–363.