Professional Documents
Culture Documents
Research paper
A R T I C LE I N FO A B S T R A C T
Keywords: Background: The aim was to quantify and to compare the associations between longitudinal changes in pain and
Pain depression in different chronic pain conditions.
Depression Methods: Data were retrieved from 6 observational cohort studies. From baseline to the 6-month follow-up, the
Longitudinal change score changes on the Short Form (36) Health Survey (SF-36) bodily pain (pain) and the SF-36 mental health
Association
(depression) scales (0=worst, 100=best) were quantified, using partial correlations obtained by multiple re-
Correlation
gression. Adjustment was performed by age, living alone/with partner, education level, number of comorbid-
Causality
ities, baseline pain and baseline depression.
Results: Stronger associations were found between changes in levels of pain and depression for neck pain after
whiplash (n = 103, mean baseline pain=21.4, mean baseline depression=52.5, adjusted correlation
r = 0.515), knee osteoarthritis (n = 177, 25.4, 64.2, r = 0.502), low back pain (n = 134, 19.0, 49.4, r = 0.495),
and fibromyalgia (n = 125, 16.8, 43.2, r = 0.467) than for lower limb lipedema (n = 68, 40.2, 62.6, r = 0.452)
and shoulder arthroplasty (n = 153, 35.0, 76.4, r = 0.292). Those correlations were somewhat correlated to
baseline pain (rank r=–0.429) and baseline depression (rank r=–0.314).
Limitations: The construct of the full range of depressive symptoms is not explicitly covered by the SF-36.
Conclusions: Moderate associations between changes in pain and depression levels were demonstrated across 5
of 6 different chronic pain conditions. The worse the pain and depression scores at baseline, the stronger those
associations tended to be. Both findings indicate a certain dose-response relationship – an important char-
acteristic of causal interference. Relieving pain by treatment may lead to the relief of depression and vice versa.
⁎
Corresponding author at: RehaClinic, Quellenstrasse 34, 5330 Bad Zurzach, Switzerland,
E-mail addresses: fangst@vtxmail.ch, f.angst@rehaclinic.ch (F. Angst).
https://doi.org/10.1016/j.jad.2020.05.044
Received 21 November 2019; Received in revised form 21 April 2020; Accepted 10 May 2020
Available online 19 May 2020
0165-0327/ © 2020 Elsevier B.V. All rights reserved.
F. Angst, et al. Journal of Affective Disorders 273 (2020) 508–516
Table 1
Cohort studies, baseline disease-relevant data.
Region Shoulder Lower limb Soft tissue Low back Knee Neck
Diagnosis Osteoarthritis Lipedema Fibromyalgia Chronic pain Osteoarthritis After whiplash
Intervention Arthroplasty Comprehensive Inpatient pain Inpatient pain Comprehensive Inpatient pain
rehabilitation program program rehabilitation program
Patients (n) 153 68 125 134 177 103
Age (years): m (s) 67.3 (10.4) 48.2 (13.5) 45.6 (9.7) 48.3 (12.2) 65.6 (10.3) 38.0 (11.9)
Female (%) 64.5 100.0 89.6 82.1 79.1 79.6
Living alone (%) 26.1 17.6 17.6 23.9 37.9 28.2
Education (%)
Compulsory schooling (8–9 20.3 14.7 32.0 22.4 36.2 2.9
y)
Vocational training 48.6 52.9 48.0 60.4 47.5 19.4
Upper secondary/university 31.2 32.4 20.0 17.2 16.4 77.7
Comorbidities (%)
0 25.4 10.3 9.6 11.9 0.6 16.5
1 23.2 8.8 28.0 23.9 6.8 34.0
2 28.3 19.1 33.6 32.8 18.6 26.2
3 7.2 29.4 16.8 20.7 28.8 14.6
≥4 15.9 32.4 12.0 10.7 45.2 4.9
of adult depression following chronic pain in childhood by an OR=1.38 predictor of subsequent depression and vice versa in all cohorts and in
(Noel et al., 2016). A 10-year survey (n = 5001) found that pain and all analysis models.
depression/anxiety in adolescence were both associated with adverse While many population surveys have documented the association
childhood experiences, such as verbal abuse, sexual abuse, parental loss between baseline pain and future depression and vice versa, very few
or parental psychopathology (Sachs-Ericsson et al., 2017). Anxiety and/ have examined the associations between changes in the severity of the
or depression were found to both moderate and mediate the relation- two outcomes (Chang et al., 2015; Chou, 2007; Erdal et al., 2017;
ship between pain and such adverse childhood experiences. In this Kroenke et al., 2011; Patel et al., 2016). In addition, many studies ex-
context, mediation was defined as the observation that anxiety and amined single pain conditions but did not compare the association of
depression were amplified by adverse childhood experiences – in- pain and depression across different chronic pain syndromes using the
dependently of pain – and increased the risk of developing pain in same measures. After an extensive review of the literature on the topic,
adolescence. Coping was inversely correlated to the number of adverse it became clear that the burden of pain and depression is more salient in
experiences in childhood, leading to the conclusion that developed certain conditions, for example in fibromyalgia than in osteoarthritis
coping skills might reduce the extent to which depression/anxiety in- (Salaffi et al., 2019).
duce pain. Our study aimed 1) to quantify the associations of longitudinal
Most studies have used thresholds to diagnose depression and/or changes in pain and depression with and without adjustment for po-
pain and to determine the frequencies of cases/non-cases and ORs; very tential disease-relevant confounders, 2) to compare the results across 6
few have used fine-graded scales that measure the levels of severity of different chronic pain conditions. As the main outcome our focus was
pain and depressivity and allow quantification of their interrelationship the correlation of “pain change” with “depression change”. As a sec-
by (partial) correlations (Angst et al., 2008c). The well-known US ondary outcome we examined the correlations to baseline pain and
Women's Health Initiative (n = 144,956) found that, compared to no/ depression. It was hypothesized that the strength of the pain-depression
mild pain, 15.8% of the study participants reported moderate to ex- relationship depends on the level of baseline pain or baseline depres-
treme pain, which was associated with the incidence of depressive sion, i.e. that higher levels of baseline symptoms may increase the
symptoms by a relative risk (RR) of 1.82 in non-veterans and 1.77 in strength of the association.
veterans after 18 years’ follow-up (Patel et al., 2016). In persons with
cognitive impairment (n = 931) over an observation period of 4–6
2. Methods
months, a 1-point increase in pain was associated with a 0.48 (score
points) increase in depression (Erdal et al., 2017). It was concluded that
2.1. Data sources
improving the treatment of pain may lower the levels of depression in
the clinical course.
Data were retrieved from the 6 treated cohort studies characterized
In line with this hypothesis, pain relief in young adolescents was
in Table 1. “Treated” means that the data were collected through the
found to be positively associated with the relief of depressive symptoms
observational measurement of individuals consecutively admitted to
and vice versa (n = 95, mean age 15.2 years, 1-year observation time)
specialized interventions who received standard clinical treatment
(Lewandowski Holley et al., 2013). Unfortunately, that study used only
(independently of the assessment and the aim of the study), in contrast
regression coefficients, not correlations, to quantify the associations. In
to artificially designed intervention studies. Detailed information re-
lumbar spinal stenosis (n = 96, 2-year observation time), pain relief
garding the selection of the patients, refusal rates and possible con-
was significantly associated with the relief of depression when arbitrary
founding is to be found in the original reports on the individual studies
relief thresholds on the visual analogue scale (VAS) for pain and the
(Angst et al., 2013, 2010, 2008a, 2006; in review).
Beck Depression Scale were chosen to determine ORs (Sinikallio et al.,
Cohort 1 consisted of shoulder osteoarthritis or rheumatoid arthritis
2010).
patients (n = 153) who received PROMOS shoulder endoprosthesis at
Findings consistent with the above are available for specific mus-
the Wilhelm Schulthess Clinic for orthopedic surgery, Zurich,
culoskeletal conditions, as outlined in the discussion. One of the most
Switzerland (Angst et al., 2008a). The other 5 studies (cohorts 2 to 6)
valid studies in this context examined a depressed (n = 250) and a
were conducted at the Rehabilitation clinic ("RehaClinic"), Bad Zur-
nondepressed (n = 250) cohort in primary care with up to 3 follow-ups
zach, Switzerland. These were: (2) patients with lipedema of the lower
(Kroenke et al., 2011). Change in pain was consistently a strong
extremity (n = 68) who underwent inpatient rehabilitation or
509
F. Angst, et al. Journal of Affective Disorders 273 (2020) 508–516
consultation for outpatient management, including complex physical Various studies have demonstrated that low scores on the SF-36 mental
decongestive therapy (Angst et al., review); patients (3) with fi- health dimension represent high levels of nervousness and depression
bromyalgia (n = 125) and (4) low back pain (n = 134) who took part (Busija et al., 2011). The SF-36 mental health scale has shown high
at the “Zurzach Interdisciplinary Pain (Schmerz) Program (ZISP)”, an responsiveness, i.e. longitudinal content and construct validity in
inpatient, multidisciplinary rehabilitation program (Angst et al., 2006); chronic pain (Angst et al., 2008b). It was statistically more responsive
(5) patients with neck pain after whiplash injury (n = 103) who took than the Hospital Anxiety and Depression Scale (HADS) depression
part at the “Zurzach Interdisciplinary Cervical Spine (Halswirbelsäule) scale and of comparable responsiveness to the MPI negative mood scale.
Concept (ZIHKo)”, a multidisciplinary inpatient rehabilitation program However, the 5 mental health items mentioned do not cover im-
(Angst et al., 2010), and (6) patients who underwent inpatient re- portant symptoms defining depression, namely diminished interest and
habilitation of knee osteoarthritis (n = 177) in the “Zurzach Osteoar- pleasure, agitation or retardation, fatigue or loss of energy, feelings of
thritis Study” (Angst et al., 2013). Except for the shoulder and the worthlessness or guilt, lack of self-esteem, feelings of guilt, diminished
whiplash cohorts (1 and 5), the relevant data sets were updated before ability to think or concentrate, thoughts of death or suicidal ideas as
the present analysis in order to include additional entrants and specific well as changes in sleep and appetite. Moreover, the SF-36 assesses only
follow-ups of the respective programs; those updates supplement the self-rated symptoms.
originally published data (Angst et al., 2013, 2006, 2020). Nevertheless, the validity of the SF-36 mental health scale for the
In all 6 cohorts, baseline scores were collected on the day of entry measurement of depression has been proven to be high by various
into the clinic and/or (outpatients) at the first visit before start of evaluations (Busija et al., 2011). For example, it correlated cross-sec-
treatment. For all cohorts the follow-up assessment was carried out tionally by maximum r = 0.57 to the Symptom Checklist 90-Revised
exactly 6 months later. (SCL-90R) affective & behavioural scale and by maximum r = 0.77 to
Beck Depression Inventory II (BDI-II) scale in individuals with trau-
2.2. Measures matic brain injury (n = 271) (Findler et al., 2001). In rheumatoid ar-
thritis patients (n = 223), the cross-sectional correlations were
Sociodemographic and disease-relevant data were recorded using a r = 0.80 between the SF-36 bodily pain and the Visual Analogue Scale
standardized questionnaire that has proven its worth in several earlier (VAS) of pain, and r = 0.80 between the SF-36 mental health and the
studies (Angst et al., 2013, 2010, 2008a, 2006, 2001, 2020). All re- HADS (Ruta et al., 1998).
quisite medical records were obtained to confirm diagnoses and to When the SF-36 mental health scale was first constructed, special
evaluate inclusion and exclusion criteria and the number of comorbid focus was laid on the validity criterion (Ware et al., 2004, 2000): Cri-
conditions. terion validation was performed using 7 different predefined dichot-
The validated German versions of the Short Form 36 Health Survey omous criteria (present/absent) and 1 (continuous) measure
(SF-36) were used to quantify pain and depression (Bullinger and (Ware et al., 2004, 2000). The criteria were: 1) dissatisfaction with life:
Kirchberger, 1998; Ware et al., 2004, 2000). The SF-36 is the most dissatisfied or very dissatisfied on the Dupuy's quality of life measure,
widely used questionnaire for the self-assessment of generic health and 2) depressive symptom scores on the Centre of Epidemiologic Studies
quality of life and facilitates good comparability among various health Depression scale (CES-D) beyond the cut off for clinical depression but
conditions (Busija et al., 2011). not satisfying DSM-III criteria for major depression, 3) a diagnosis of
The SF-36 bodily pain scale consists of 2 items: 1) How much bodily depression fulfilling the DSM-III criteria for major depression, dys-
pain have you had during the past 4 weeks? (none, very mild, mild, thymia, or both after CES-D score beyond the cut off, 4) suicide ideation
moderate, severe, very severe:6 levels – SF-36 version 1 and 2 iden- defined by the corresponding item of the Mental Health Inventory
tical), 2) during the past 4 weeks, how much did pain interfere with (MHI), 5) outpatient visit(s) for mental health treatment in the past 6
your normal work (outside home and home work) (not at all, a little bit, months, 6) visit(s) at formally trained mental health specialist's office,
moderately, quite a bit, extremely: 5 levels – SF-36 version 1 and 2 7) inpatient stay(s) in a hospital for mental health treatment in the past
identical). The internal constancy of the SF-36 bodily pain scale was 12 months. The average of the percentages of those (out of a total
Cronbach's Alpha (CA)=0.88 in the normative German general popu- n = 2988) who fulfilled criteria 1) to 7) were related to the item re-
lation study (n = 2914) (Bullinger and Kirchberger, 1998). sponse categories of the SF-36 mental health items: “all the time” (score
The content and construct validity of the SF-36 bodily pain scale is 0=most depression) included 58.6% positive cases, “most of the time”
clear and has been proven by numerous studies that have applied it to (score 20) 48.1%, “good bit of the time” (score 40) 36.8%, “some of the
measure outcome (Busija et al., 2011; Ware et al., 2004, 2000). Of time” (score 60) 24.5%, “a little” (score 80) 13.7%, and “none of the
particular interest is the longitudinal validity, which has been tested in time” (score 100=no depression) 5.8%. This results in a Pearson cor-
our fibromyalgia and chronic low back pain setting (n = 273) relation of r = 0.998 (severe to severe). The 8th criterion, the mental
(Angst et al., 2008b). The SF-36 bodily pain scale was slightly less re- health scale average score on the 32-item long-form of the MHI, cor-
sponsive than the Multidimensional Pain Inventory (MPI) pain severity related with the SF-36 mental health score by r = 0.96.
scale and slightly more responsive than the pain Numeric Rating Scale
(NRS); however neither difference was statistically significant. 2.3. Analysis
The SF-36 mental health consists of 5 items as follows: How much
during the past month 1) have you been a very nervous person, 2) have Both the SF-36 bodily pain and the SF-36 mental health were scaled
you felt so down in the dumps that nothing could cheer you up, 3) have from 0=maximum pain/depression to 100=no pain/depression ac-
you felt calm and peaceful, 4) have you felt downhearted and blue, 5) cording to the original scaling rules (Bullinger and Kirchberger, 1998;
have you been a happy person? (Possible responses to each item are: Ware et al., 2004, 2000). While the SF-36 bodily pain scale is scored by
all/most/a good bit/some/a little/none of the time, i.e. 6 levels in SF-36 the mean of the two items, which are complexly recoded depending on
version 1; and in SF-36 version 2: all/most/some/a little/none of the various “if” statements, the SF-36 mental health scale is the un-
time i.e. 5 levels (Bullinger and Kirchberger, 1998; Ware et al., 2000). weighted linear combination of the 5 items, 2 of which have to be in-
The internal constancy of the SF-36 mental health scale was CA=0.80 versely recoded. As measure of reliability / internal consistency, the
(Bullinger and Kirchberger, 1998). Cronbach's alphas were determined between the baseline and the
The validity of the SF-36 mental health scale for the measurement of follow-up scores for each scale and cohort.
self-rated depressivity (the number and level of depressive symptoms) Since for both scales, the baseline, the follow-up scores and the
or depression is less obvious and is presented below. To simplify no- score changes (baseline to 6-month follow-up) showed approximate
menclature, “depressivity” has been replaced by “depression” hereafter. normal Gauss distribution, parametric Pearson correlations were
510
F. Angst, et al. Journal of Affective Disorders 273 (2020) 508–516
computed. The score changes on SF-36 bodily pain and SF-36 mental correlations between pain and depression changes. An analysis of var-
health dimensions were determined by bivariate, non-adjusted corre- iance across the 6 cohorts resulted in p<0.001 for both change on pain
lation. Multiple linear regression was used to determine the adjusted and on depression.
correlations (Backhaus et al., 2016; Hedderich and Sachs, 2016;
Ho, 2013): Change in SF-36 bodily pain (dependent variable) was 3.3. Correlations between pain and depression
modeled by sex, age, living with a partner, education level, number of
comorbidities, baseline SF-36 bodily pain, baseline SF-36 mental The main outcome correlations, those between changes in pain and
health, score change on the SF-36 mental health scale (independent depression, were higher than all the others, including the baseline
variables). The partial correlation of the score change on the SF-36 scores (Table 2). For interest, we have also listed the bivariate, (un-
mental health scale represented the adjusted correlation between pain adjusted) correlation coefficients for all analyses. The adjusted (partial)
and depression. The overall explained variances (%) were depicted to correlations were 0.515 for neck pain, 0.502 for knee osteoarthritis,
quantify the fit of the regression models. Finally, in order to test the 0.495 for low back pain, 0.467 for fibromyalgia, 0.452 for lipedema,
initial hypothesis, the levels of the adjusted correlations were related to and 0.292 for shoulder osteoarthritis. All correlations between changes
the levels of baseline pain and depression by graphic interpretation and in pain and depression were statistically significantly different from
non-parametric rank-correlation analysis. zero (two-sided test, respectively, the 95% confidence intervals ex-
A minimum sample size of n = 50 has been estimated as necessary cluded zero). The rank correlation coefficient between the adjusted
to quantify valid correlation estimates of the basic population with both change correlations and baseline pain was –0.429 (Fig. 1). There was a
normally distributed 95% confidence intervals (Hedderich and similar but weaker relationship between the adjusted change correla-
Sachs, 2016). To illustrate the precision of the parameter estimation, tions and baseline depression: rank correlation –0.314 (Fig. 2). The
the 95% confidence intervals are in the range of ≤0.23 (lower limit) models, which included the confounding co-factors age, sex, living si-
and ≤0.17 (upper limit) for n = 50 to 100; for n ≥ 100 the corre- tuation, education level, the number of comorbidities, and the baseline
sponding levels are ≤0.16 and ≤0.13. scores of pain and depression (see Table 1), explained moderate pro-
portions of variance, ranging from 29.7% for shoulder osteoarthritis to
3. Results 46.7% for knee osteoarthritis (Table 2).
The 3 secondary correlation analyses showed consistently lower
3.1. Baseline socio-demographic and disease-relevant data levels than those of the primary analysis (Table 2). The adjusted cor-
relation between pain change and baseline depression was low for knee
The 6 cohort studies are characterized in Table 1. They are ordered osteoarthritis (adjusted correlation 0.143) and statistically not sig-
from left to right according to the adjusted association between pain nificant but higher for the other 5 cohorts (0.261 to 0.368) and statis-
change and depression change (Table 2). Both SF-36 scores of all 6 tically significant. The corresponding levels of adjusted correlations
cohorts were approximately normally distributed, as described in detail between depression change and baseline pain were very low to low,
in the original reports on the individual studies. The shoulder ar- ranging from 0.083 (shoulder osteoarthritis) to 0.271 (lower limb li-
throplasty patients (n = 153) were relatively elderly, had the highest pedema); 4 out of 6 were statistically significant. Baseline pain corre-
proportion of males, were well-educated, and had few comorbid con- lated to baseline depression by adjusted levels from 0.198 (low back
ditions. The cohort with lower limb lipedema (n = 68), consisted solely pain) to 0.396 (lipedema), all of which are statistically significant.
of women; many were well-educated and, although still in middle age, The corresponding detailed parameters of the bivariate and multiple
comorbid conditions were very frequent. The fibromyalgia cohort (adjusted) regression models can be found in Table 3.
(n = 125) was predominantly female, again middle-aged and with a
lower level of educational attainment. Similar characteristics were 4. Discussion
found in the low back pain sample (n = 134). The knee osteoarthritis
group (n = 177) was elderly and characterized by multiple co- This is the first study that has undertaken to correlate changes in
morbidities. The members of the cohort with neck pain after whiplash pain with changes in depression using fine-graded scales and to com-
injury (n = 103) were under forty and very well-educated. pare the for confounding adjusted results across different chronic pain
conditions. Five of the 6 cohorts analyzed revealed relatively high
3.2. Pain and depression in the clinical course correlations between the changes in pain and depression levels (base-
line to 6-month follow-up). These were: neck pain after whiplash (ad-
As shown in Table 2, the burden of pain at baseline (SF-36 bodily justed, partial correlation r = 0.515), knee osteoarthritis (r = 0.502),
pain) was relatively heavy in fibromyalgia (mean score 16.8, where low back pain (r = 0.495), fibromyalgia (r = 0.467), and lower limb
0=maximum pain), low back pain (19.0), knee osteoarthritis (25.4), lipedema (r = 0.452). These are chronic pain conditions, where pain is
and neck pain (21.4) but markedly lower in shoulder osteoarthritis still present after conservative therapy. In contrast, much of the pain
(35.0) and lower limb lipedema (40.2). Baseline levels of depression experienced in shoulder osteoarthritis was to a large extent/ reduced by
were relatively high in fibromyalgia (mean score 43.2, where arthroplasty, where the corresponding association was weaker
0=maximum depression), low back pain (49.4), and neck pain (52.5) (r = 0.292).
but lower in the 3 other cohorts (62.2 to 76.4). An analysis of variance However, the association between pain change and depression
across the 6 cohorts resulted in p<0.001 for both baseline pain and change was somewhat stronger when pain and depression scores were
baseline depression. higher at baseline, which partly confirmed our initial hypothesis (rank
At the 6-month follow-up, the greatest improvements in pain were correlation to baseline pain r=–0.429, to baseline depression
observed after shoulder arthroplasty (mean change +28.7 score points) r=–0.314). The outlier was knee osteoarthritis with relatively low le-
but were smaller in the other 5 cohorts with conservative therapy vels of depression. Even if those two correlation levels are not high,
(+4.8 to +9.0). On the depression scale, 4 cohorts showed small mean they indicate a possible relationship between the pain-depression as-
changes ranging from +4.1 to +7.1 but in two cohorts (shoulder os- sociation and the baseline syndrome levels, which can be only partly be
teoarthritis and knee osteoarthritis) mean changes were close to zero. captured by the SF-36 scales that were implemented.
However, the standard deviations of the score changes were high, both A correlation of r = 0.515, as in neck pain after whiplash, means
in pain (17.7 to 26.2) and depression (14.5 to 18.5), reflecting wide that r2=26.5% of the variance of pain change is explained by the
variations in improvement or deterioration between baseline and the variance of depression change and vice versa, signifying a moderate
follow-up. This is an important characteristic in determining the level of association (Backhaus et al., 2016). Conversely, 46.9%
511
F. Angst, et al. Journal of Affective Disorders 273 (2020) 508–516
Table 2
Outcome and correlation data.
Region Shoulder Lower limb Soft tissue Low back Knee Neck
Diagnosis Osteoarthritis Lipedema Fibromyalgia Chronic pain Osteoarthritis After whiplash
Legend: n=number of cases, m=arithmetic mean, s=standard deviation, SF-36: Short Form 36, Δ=Change baseline to 6 month follow-up, B=baseline, the construct
of “Pain” was measured by the SF-36 Bodily pain scale, the construct of “Depression” (in fact self-rated depressivity, see in methods) was measured by the SF-36
Mental health scale, p of Δ (t-test)=two-sided type I error of the one-sample t-test of the score change against zero, (%)=explained variance in percent of the
corresponding multiple model (4 models per condition), CA=Cronbach's alpha between the baseline and the follow-up score of the scale, Correlation with (95%
confidence interval): bold are the partial (adjusted) correlation coefficients, exclusion of zero in the 95% confidence interval means statistical significance, i.e., the
correlation coefficient is significantly different from zero with a two-sided type I error of p = 0.050.
Fig. 1. Dependency of the adjusted correlations (pain change with depression change) from baseline pain (n = 6 cohorts).
512
F. Angst, et al. Journal of Affective Disorders 273 (2020) 508–516
Fig. 2. Dependency of the adjusted correlations (pain change with depression change) from baseline depression (n = 6 cohorts).
“explained variance” of the model in knee osteoarthritis means that the between changes in pain and subsequent levels of depression and vice
total model consisting of depression change and the confounders cor- versa. Those t levels ranged from 2.90 to 6.46 and were comparable to
related by r=√0.469=0.685 with pain change. These findings indicate ours (3.54 to 7.52, pain change to depression change, adjusted models).
a first dose-response relationship, which is independent of potential The bivariate correlations between pain/depression change and de-
confounders: Greater changes on pain were associated with greater pression/pain levels at the follow-up ranged between 0.26 and 0.37 in
changes on depression and vice-versa. The confounders were sex, age, the report of Kroenke et al. and were, thus, also comparable to our
living with a partner, education level, number of comorbidities, base- results. Moreover, this valuable study provides a summary of 8 com-
line SF-36 bodily pain, baseline SF-36 mental health. The second dose- parable studies.
response relationship consists in the stronger correlations between pain Again on the basis of assessment by regression coefficients only,
change and depression change where baseline pain and baseline de- remission of depression and anxiety (DSM-IV diagnoses) was statisti-
pression were worse. A dose-response relationship is one of the clearest cally significantly associated with a decrease in the severity of pain and
characteristics of the causality of an association (Rothman and the number of locations in n = 2093 subjects over 4 years follow-up
Greenland, 2005). All the other associations, including baseline pain or (Gerrits et al., 2015). A systematic review of 13 studies of lumbar spinal
baseline depression, were weaker but still present and at levels con- stenosis concluded that preoperative depression was likely to be a
sistent with the literature. prognostic factor for postoperative symptom severity, including pain, in
Our findings of positive associations between changes in levels of lumbar spinal stenosis. There were wide variations, however, in the
pain and depression are supported by the literature. In the general levels of the associations (McKillop et al., 2014), a finding in partial
population of Taiwan (3–11 years of follow-up) the adjusted relative accord with our results.
risk (RR) of developing depression was RR=6.28 for persons suffering In contrast, another systematic review of 10 studies found only
from fibromyalgia (n = 25,969) and, vice versa, RR=7.46 for people moderate evidence that post-injury depression or anxiety were risk
with depression (n = 17,142) of developing fibromyalgia (Chang et al., factors for persistent pain after orthopedic trauma (Clay et al., 2012).
2015). Depression was found to be a partial mediator of the relationship Furthermore, since the association between pain and depression dis-
between pain intensity and physical functioning in fibromyalgia appeared after adjusting for co-twin case-control (n = 1269 twins, 2–4
(n = 216, 36 months of follow-up), indicating that treating depression years observation time), the authors concluded that the relationship
may relieve pain (Steiner et al., 2017). between depression and chronic low back pain is not causal
In early rheumatoid arthritis (n = 520, 2-year observation time), (Fernandez et al., 2017). However, depression in that study was mea-
baseline pain (VAS 0–10) was significantly associated with worsening sured by one single dichotomous item of the Euro Quality Of Life-5
on the SF-36 mental health dimension (scaled 0–100) by 0.08 score Dimensions (EQ-5D), to which limited validity has been attributed
points per 1 point VAS pain, whereas no significant association was (Gignac et al., 2011).
found with any of the scores of the Disease Activity Score (DAS) and the An additional link in establishing causality in the association be-
Health Assessment Questionnaire (HAQ) (Euesden et al., 2017). In a tween pain and depression would be evidence of etiopathophysiological
study of psoriatic arthritis (n = 394, 7.5 years observation time), a pathways (Rothman and Greenland, 2005). Neuroanatomical or neu-
condition which tends to be diagnosed later in the course than early rochemical models describe biochemical pathways. Anatomical and
rheumatoid arthritis, pain relief measured by the HAQ was associated functional areas, neurotransmitters and receptors (especially those for
with improvement in mental health measured by the SF-36 mental Serotonine and Glutamate) that process the perceptions and symptoms
component summary (MCS); however, the study showed only regres- of pain and depression are thought to be shared (Kroenke et al., 2011;
sion coefficients and no partial correlations (Husted et al., 2012). Meerwijk et al., 2013; Sheng et al., 2017; Zis et al., 2017). Functional
The same was true for chronic hip, knee and back pain (n = 500, 1 MRI, for example, has shown that similar brain regions are active for
year observation time) in primary care (Kroenke et al., 2011). In this pain after injuries and for mood management (Adler-Neal et al., 2019;
study, the t (distribution)-values of all types of analysis were very high Meerwijk et al., 2013). Considering the typical direction of causality,
(leading to p<0.001 in most cases) indicating strong relationships chronic pain is thought to lead to depression rather than vice versa
513
F. Angst, et al. Journal of Affective Disorders 273 (2020) 508–516
Legend: n=number of cases, β=standardized regression coefficient of the independent variable, t=Student's t-distribution value of the t-test if β is different from zero, p=two-sided type I error of t (Student's t-test).
(Sheng et al., 2017).
<0.001
<0.001
0.012
0.656
0.638
0.012
0.081
0.092
This hypothesis is supported by the causal characteristics of the
p
relationship of physical and psychological distress, which are provoked
by pain (IsHak et al., 2018). Psychotherapeutic models can also explain
−0.45
the vicious cycle of continuous interaction of cognitions, emotions,
5.29
5.74
2.58
0.47
2.57
1.77
2.22
After whiplash
−0.044
depression and pain remain unclear and need further elucidation
0.466
0.527
0.242
0.047
0.227
0.173
0.235
Neck
103
<0.001
<0.001
0.003
0.062
0.047
0.003 scales for pain and depression, identical observation periods, and ad-
p
−2.00
7.96
7.52
1.88
3.01
5.98
4.11
showed large variance in the baseline and change levels on both out-
Osteoarthritis
−0.149
0.516
0.477
0.126
0.220
0.387
0.284
partial correlations, which are free from confounding by the tested co-
Knee
177
variates. For example, baseline pain and baseline depression, which are
β
<0.001
0.094
0.058
0.233
0.026
0.85
4.42
1.92
1.20
2.26
all 6 cohorts had the same levels of baseline pain and baseline de-
Low back
0.074
0.330
0.152
0.104
0.201
<0.001
<0.001
<0.001
0.865
0.004
that the construct of the full range of depressive symptoms is not ex-
−0.17
plicitly covered by the SF-36. However, the various criteria and other
4.81
5.67
0.98
3.96
2.94
3.66
4.17
mental health scale. This may be due to the high co-prevalence of the
Soft tissue
−0.015
0.088
0.330
0.247
0.265
0.383
125
β
5. Conclusions
<0.001
0.003
0.446
0.024
0.136
0.033
0.023
0.001
−1.51
2.31
2.18
2.32
3.34
−0.094
−0.183
0.289
0.274
0.275
0.385
pain may lead to the relief of depression and vice versa. Our findings
should be further substantiated by specific intervention studies.
<0.001
<0.001
0.007
0.001
0.135
0.427
0.337
0.003
p
1.50
3.93
0.96
3.03
3.67
−0.067
0.224
0.297
0.122
0.334
0.089
0.241
0.292
153
Δ Pain – Δ Depression
Δ Pain – B Depression
Δ Depression – B Pain
B Pain – B Depression
bivariate
bivariate
bivariate
bivariate
adjusted
adjusted
adjusted
adjusted
Region
Table 3
All data and material are freely available. Please contact the cor-
responding author for data requests.
514
F. Angst, et al. Journal of Affective Disorders 273 (2020) 508–516
515
F. Angst, et al. Journal of Affective Disorders 273 (2020) 508–516
veteran postmenopausal women. Gerontologist 56 (Suppl 1), S91–101. https://doi. Sheng, J., Liu, S., Wang, Y., Cui, R., Zhang, X., 2017. The link between depression and
org/10.1093/geront/gnv670. chronic pain: neural mechanisms in the brain. Neural Plast. 2017, 9724371. https://
Rothman, K.J., Greenland, S., 2005. Causation and causal inference in epidemiology. Am. doi.org/10.1155/2017/9724371.
J. Public Health 95 (Suppl 1), S144–S150. https://doi.org/10.2105/AJPH.2004. Sinikallio, S., Lehto, S.M., Aalto, T., Airaksinen, O., Kröger, H., Viinamäki, H., 2010.
059204. Depressive symptoms during rehabilitation period predict poor outcome of lumbar
Ruta, D.A., Hurst, N.P., Kind, P., Hunter, M., Stubbings, A., 1998. Measuring health status spinal stenosis surgery: a two-year perspective. BMC Musculoskelet. Disord. 11, 152.
in British patients with rheumatoid arthritis: reliability, validity and responsiveness https://doi.org/10.1186/1471-2474-11-152.
of the short form 36-item health survey (SF-36). Br. J. Rheumatol. 37, 425–436. Stein, K., Pearson, R.M., Stein, A., Fazel, M., 2017. The predictive value of childhood
https://doi.org/10.1093/rheumatology/37.4.425. recurrent abdominal pain for adult emotional disorders, and the influence of negative
Sachs-Ericsson, N.J., Sheffler, J.L., Stanley, I.H., Piazza, J.R., Preacher, K.J., 2017. When cognitive style. Findings from a cohort study. PLoS ONE 12, e0185643. https://doi.
emotional pain becomes physical: adverse childhood experiences, pain, and the role org/10.1371/journal.pone.0185643.
of mood and anxiety disorders. J. Clin. Psychol. 73, 1403–1428. https://doi.org/10. Steiner, J.L., Bigatti, S.M., Slaven, J.E., Ang, D.C., 2017. The complex relationship be-
1002/jclp.22444. tween pain intensity and physical functioning in fibromyalgia: the mediating role of
Salaffi, F., Di Carlo, M., Carotti, M., Farah, S., Ciapetti, A., Gutierrez, M., 2019. The depression. J. Appl. Biobehav. Res. 22. https://doi.org/10.1111/jabr.12079.
impact of different rheumatic diseases on health-related quality of life: a comparison Ware, J.E., Kosinski, M., Dewey, J.E., 2000. In: Lincoln, R.I. (Ed.), 2nd ed. Quality Metric
with a selected sample of healthy individuals using SF-36 questionnaire, EQ-5D and Incorporated.
SF-6D utility values. Acta Biomed. 89, 541–557. https://doi.org/10.23750/abm. Ware, J.E., Snow, K.K., Kosinski, M., Gandek, B., New England Medical Center Hospital,
v89i4.7298. Health Institute, 2004. SF-36 Health survey: Manual and Interpretation Guide, 3rd
Sanders, J.B., Comijs, H.C., Bremmer, M.A., Deeg, D.J.H., Beekman, A.T.F., 2015. A 13- ed. Health Institute, New England Medical Center, Boston.
year prospective cohort study on the effects of aging and frailty on the depression- Zis, P., Daskalaki, A., Bountouni, I., Sykioti, P., Varrassi, G., Paladini, A., 2017.
pain relationship in older adults. Int. J. Geriatr. Psychiatry 30, 751–757. https://doi. Depression and chronic pain in the elderly: links and management challenges. Clin
org/10.1002/gps.4224. Interv Aging 12, 709–720. https://doi.org/10.2147/CIA.S113576.
516