Rheumatoid arthritis (RA) is a destructive autoinflammatory arthritis characterized by the overproduction of certain proinflammatory cytokines. The disease is destructive in nature, leading to pain, stiffness, bone resorption, and systemic inflammation. Treatment with monoclonal antibodies or antibody derivatives (biologicals) that block specific cytokines has been very successful. Many of these biologicals prevent bone destruction and have a substantial impact on disease activity [
Patients with rheumatoid arthritis have a higher prevalence of depression and anxiety than the general population [
In order to investigate the role of cytokines in depression associated with rheumatoid arthritis, we studied the relationship between depression, clinical markers, and serum cytokine levels in rheumatoid arthritis patients, patients being treated for primary depression, healthy controls, and lupus patients. We further hypothesized that cytokine inhibition via biological agents would have an impact on depressive symptoms. To test this hypothesis, a subcohort of rheumatoid arthritis patients being treated with biological therapeutics to inhibit specific cytokines was examined, to determine the effect of cytokine inhibition on depression.
A cohort of 209 individuals was recruited for this study. This group included 82 RA patients, 22 healthy control subjects, 32 subjects with primary depression, and 73 subjects with systemic lupus erythematosus as a control group for autoimmune arthritis (Table
Sociodemographics of cohort.
All subjects | RA subjects | Healthy subjects | Depressed subjects | SLE subjects | ||||
---|---|---|---|---|---|---|---|---|
( |
( |
( |
( |
( |
||||
Gender, number (%) | ||||||||
Female | 192 (92) | 69 (84) | 21 (95) | 0.291 | 29 (91) | 0.550 | 73 (100) | <0.001 |
Age, mean ± SD | 49.27 ± 11.99 | 54.99 ± 10.08 | 43.14 ± 11.54 | < 0.001 | 49.25 ± 14.03 | 0.040 | 44.70 ± 10.38 | <0.001 |
Education (years), mean ± SD | 8.43 ± 4.45 | 6.85 ± 3.99 | 13.55 ± 3.10 | < 0.001 | 8.91 ± 4.31 | 0.017 | 8.88 ± 4.14 | 0.002 |
Education level, number (%) | < 0.001 | 0.200 | <0.001 | |||||
Primary (or less) | 97 (46) | 53 (65) | 1 (5) | 15 (47) | 28 (38) | |||
Middle school | 34 (16) | 14 (17) | 2 (9) | 7 (22) | 11 (15) | |||
High school | 44 (21) | 7 (9) | 8 (36) | 5 (16) | 24 (33) | |||
College | 32 (15) | 7 (9) | 10 (45) | 5 (16) | 10 (14) | |||
Graduate school | 2 (1) | 1 (1) | 1 (5) | 0 (0) | 0 (0) | |||
Marriage status, number (%) | 0.024 | 0.428 | 0.026 | |||||
Unmarried | 33 (16) | 9 (11) | 8 (36) | 6 (19) | 10 (14) | |||
Married | 135 (65) | 50 (61) | 9 (41) | 22 (69) | 54 (74) | |||
Divorced | 28 (13) | 14 (17) | 4 (18) | 2 (6) | 8 (11) | |||
Widowed | 10 (5) | 8 (10) | 0 (0) | 2 (6) | 0 (0) | |||
Cohabiting | 3 (1) | 1 (1) | 1 (5) | 0 (0) | 1 (1) | |||
Employment status, number (%) | 0.002 | 0.137 | 0.452 | |||||
Employed | 88 (42) | 28 (34) | 16 (73) | 16 (50) | 28 (38) |
The study was submitted and approved by the Ethical Committee of the São João Hospital IRB (EPE) in accordance with the Declaration of Helsinki. The nature and the purpose of the study were explained to all participants who signed the informed consent form before they entered the study.
Sociodemographic characteristics were included in the questionnaire. Education level was given as years of school. Age and marital status were listed. Socioeconomic class evaluation was included using employment status and years of education as proxies. Psychological evaluations were performed using several standardized instruments.
The short form of the FSS was used for this study. This form allows the participant to report their perceived level of fatigue [
The Cronbach
The Hospital Anxiety and Depression Scale (HADS) measures anxiety and depression, especially in people with physical illness [
The PSQI is a measurement of self-assessed sleep quality. High PSQI scores show low sleep quality, and low PSQI scores mean high sleep quality. Reliability of the PSQI is high, with a Cronbach alpha of 0.83. Seven components of sleep are evaluated by this instrument. These include sleep latency, sleep disturbances, sleep duration, sleep quality, sleep efficiency, use of sleep medications, and daytime dysfunction. The scores for each of these components are combined for a global score that can range from 0 to 21 [
The RAS [
The Portuguese version of the DAS 28 [
Pain was self-reported to the telephone interviewer using a numerical scale from 1 to 10.
ELISA kits were purchased from eBiosciences, and serum cytokine levels were measured according to the manufacturer’s directions. Peripheral blood samples were always collected in the morning (9–11 a.m.) into plastic tubes containing K3EDTA as an anticoagulant for plasma determinations or a gel separator for serum assays (BD Vacutainer, Franklin Lakes, NJ). Blood was centrifuged at 2000
Significant differences in the demographical, clinical, and psychological variables between the SLE subjects, healthy controls, RA subjects, and depressed control subjects were determined using the independent
Univariate analysis was accomplished by using generalized linear or logistic regression, using the Poisson function when appropriate, due to the Poisson distribution observed in the data. HADS depression scores were utilized as the dependent variable and with the individual variable suspected of showing a correlation to depression as the independent variable. Statistical analysis was performed using the statistical software R and SPSS (IBM). The multivariate analysis of the RA cohort was accomplished using logistic regression utilizing the Poisson function. HADS depression was used as the dependent variable. Only independent variables that were not co-related and had demonstrated a significant association in univariate analysis were used in the multivariate models to generate the best fit model. The best fit model was chosen based on the model demonstrating the lowest AICc value with the highest pseudo
To compare the rheumatoid arthritis patients that began biological therapy during the second month of the study, the appropriate paired
A post hoc power analysis was performed to test for the likelihood of type 2 errors, in which true correlations are missed due to inadequate sample size. The advantage of a post hoc power analysis is that it can consider the actual variation seen in the sample, as opposed to a projected variance, and thus provide a realistic analysis (Supplementary Table
Four different groups were recruited for this study: rheumatoid arthritis patients, lupus patients, primary depression patients, and healthy controls. The rheumatoid arthritis patients were significantly older than the other groups. They also had fewer years of education than the other groups. The RA patients were more likely to be married than the healthy controls, but less likely than the lupus patients. They were also less likely to be employed than the healthy controls, but not than the primary depression or lupus patients (Table
Depressive symptoms correlated with medication in rheumatoid arthritis.
Medication | Pseudo |
AICc | Coefficient | Odds ratio (95% CI) | |
---|---|---|---|---|---|
Biological medications | 0.154 | 572.98 | |||
None | 0.059 | 1.06 (0.83–1.37) | 0.643 | ||
Abatacept | −0.143 | 0.87 (0.45–1.52) | 0.642 | ||
Adalimumab | −0.143 | 0.87 (0.45–1.52) | 0.642 | ||
Etanercept | 0.011 | 1.01 (0.72–1.41) | 0.948 | ||
Golimumab | 0.011 | 1.01 (0.43–2.03) | 0.978 | ||
Infliximab | 0.113 | 1.12 (0.73–1.66) | 0.588 | ||
Rituximab | 0.145 | 1.16 (0.85–1.57) | 0.361 | ||
Tocilizumab | −0.351 | 0.70 (0.52–0.95) | 0.023 | ||
Classic medications | 0.829 | 478.81 | |||
Leflunomide | −0.284 | 0.75 (0.56–1.01) | 0.061 | ||
Methotrexate | −0.207 | 0.81 (0.62–1.06) | 0.125 | ||
Leflunomide and Methotrexate | 0.301 | 1.35 (0.82–2.12) | 0.209 | ||
Sulfasalazine | −0.546 | 0.58 (0.37–0.87) | 0.012 | ||
Hydroxychloroquine + Leflunomide + Methotrexate | −0.441 | 0.64 (0.31–1.17) | 0.184 | ||
Hydroxychloroquine + Methotrexate | 0.090 | 1.09 (0.73–1.60) | 0.650 | ||
Cyclosporine | 0.029 | 1.03 (0.46–1.98) | 0.937 | ||
Hydroxychloroquine + Leflunomide | −0.258 | 0.77 (0.30–1.61) | 0.539 | ||
Sulfasalazine + Methotrexate | 0.147 | 1.16 (0.70–1.86) | 0.565 | ||
Hydroxychloroquine | 0.147 | 1.16 (0.68–1.86) | 0.565 | ||
Sulfasalazine + Hydroxychloroquine | 0.147 | 1.16 (0.68–1.86) | 0.565 | ||
Sulfasalazine + Leflunomide | −1.357 | 0.26 (0.04–0.81) | 0.057 |
Correlations between depressive symptoms and serum cytokine levels.
Pseudo |
AICc | Coefficient | Odds ratio (95% CI) | |||
---|---|---|---|---|---|---|
Rheumatoid arthritis | ( | |||||
IL-6 (pg/mL), mean ± SD | 44.04 ± 104.39 | 1.000 | 73.80 | 0.002 | 1.00 (1.00-1.00) | 0.132 |
IL-10 (pg/mL), mean ± SD | 4.12 ± 2.97 | 1.000 | 68.99 | 0.121 | 1.13 (1.03–1.22) | 0.005 |
TNF-alpha (pg/mL), mean ± SD | 109.92 ± 151.85 | 1.000 | 75.19 | 0.001 | 1.00 (1.00-1.00) | 0.430 |
Lupus | ( | |||||
IL-6 (pg/mL), mean ± SD | 2.81 ± 2.18 | 1.000 | 83.69 | 0.055 | 1.06 (0.96–1.15) | 0.229 |
IL-10 (pg/mL), mean ± SD | 6.76 ± 4.06 | 1.000 | 76.47 | −0.088 | 0.92 (0.86–0.97) | 0.006 |
TNF-alpha (pg/mL), mean ± SD | 60.66 ± 102.38 | 1.000 | 83.56 | −0.001 | 1.00 (1.00-1.00) | 0.243 |
Depressive disorder | ( | |||||
IL-6 (pg/mL), mean ± SD | 1.30 ± 2.23 | 0.970 | 143.99 | 0.070 | 1.07 (1.01–1.13) | 0.011 |
IL-10 (pg/mL), mean ± SD | 9.18 ± 15.66 | 0.962 | 146.13 | 0.008 | 1.01 (1.00-1.02) | 0.059 |
TNF-alpha (pg/mL), mean ± SD | 38.43 ± 67.75 | 0.962 | 146.01 | 0.002 | 1.00 (1.00-1.00) | 0.048 |
Healthy controls | ( | |||||
IL-6 (pg/mL), mean ± SD | 1.11 ± 1.34 | 0.653 | 102.02 | 0.014 | 1.01 (0.84–1.18) | 0.869 |
IL-10 (pg/mL), mean ± SD | 3.07 ± 3.78 | 0.712 | 98.70 | 0.047 | 1.05 (1.00-1.09) | 0.047 |
TNF-alpha (pg/mL), mean ± SD | 33.77 ± 53.08 | 0.706 | 99.05 | −0.005 | 1.00 (0.99–1.00) | 0.109 |
Correlations between depressive symptoms and clinical assessments in RA subjects.
( |
Pseudo |
AICc | Coefficient | Odds ratio (95% CI) | ||
---|---|---|---|---|---|---|
Disease duration (years), mean ± SD | 17.56 ± 11.48 | 0.252 | 461.64 | −0.040 | 0.96 (0.88–1.05) | 0.401 |
BMI, mean ± SD | 26.46 ± 5.49 | 0.144 | 557.56 | 0.028 | 1.03 (1.01–1.04) | < 0.001 |
Pain score, median | 6 | 0.411 | 527.31 | 0.114 | 1.12 (1.08–1.16) | < 0.001 |
Smoking, number (%) | 0 (0) | < 0.001 | 570.11 | −0.029 | 0.97 (0.77–1.22) | 0.806 |
Drinking, number (%) | 13 (16) | 0.014 | 569.01 | −0.129 | 0.88 (0.69–1.11) | 0.290 |
Physical activity, number (%) | 13 (16) | 0.256 | 546.22 | −0.668 | 0.51 (0.38–0.68) | < 0.001 |
Biologic medication, number (%) | 42 (51) | 0.027 | 570.09 | −0.082 | 0.92 (0.73–1.18) | 0.503 |
Classic medication, number (%) | 55 (67) | 0.030 | 569.89 | −0.062 | 0.94 (0.75–1.19) | 0.601 |
Disease activity (patient), mean ± SD | 54.04 ± 30.16 | 0.880 | 426.21 | 0.002 | 1.01 (1.01–1.02) | < 0.001 |
Disease activity (doctor), mean ± SD | 35.44 ± 27.03 | 0.890 | 419.81 | 0.013 | 1.01 (1.01–1.02) | < 0.001 |
Sedementation velocity, mean ± SD | 26.59 ± 22.27 | 0.834 | 448.13 | 0.010 | 1.01 (1.01–1.01) | < 0.001 |
C-reactive protein, mean ± SD | 11.11 ± 17.43 | 0.778 | 468.09 | 0.006 | 1.01 (1.00-1.01) | 0.007 |
DAS28, mean ± SD | 5.04 ± 5.65 | 0.830 | 449.81 | 0.038 | 1.04 (1.02–1.05) | < 0.001 |
Brain leukocytes, mean ± SD | 8.09 ± 3.40 | 0.707 | 484.34 | 0.010 | 1.01 (0.98–1.04) | 0.425 |
Brain lymphocytes, mean ± SD | 27.10 ± 9.84 | 0.713 | 482.76 | −0.007 | 0.99 (0.98–1.00) | 0.139 |
Anti-dsDNA, mean ± SD | 10.66 ± 5.30 | 0.999 | 248.28 | 0.017 | 1.02 (1.00-1.04) | 0.069 |
Correlations between depressive symptoms and SLE clinical assessments.
SLE subjects | ||||||
---|---|---|---|---|---|---|
( |
Pseudo |
AICc | Coefficient | Odds ratio (95% CI) | ||
Disease duration (years), mean ± SD | 17.81 ± 8.85 | 0.021 | 467.95 | 0.006 | 1.01 (1.00-1.02) | 0.211 |
BMI, mean ± SD | 24.51 ± 5.60 | 0.103 | 461.59 | 0.021 | 1.02 (1.01–1.03) | 0.003 |
Smoking, number (%) | ||||||
Yes | 13 (18) | 0.032 | 467.14 | −0.186 | 0.83 (0.65–1.05) | 0.133 |
Drinking, number (%) | ||||||
Yes | 7 (10) | 0.006 | 469.03 | 0.100 | 1.11 (0.82–1.45) | 0.489 |
Physical activity, number (%) | ||||||
Yes | 24 (33) | 0.054 | 465.43 | −0.196 | 0.82 (0.68–0.99) | 0.047 |
Sedimentation velocity, mean ± SD | 23.48 ± 20.96 | 0.355 | 440.12 | 0.003 | 1.00 (1.00-1.01) | 0.220 |
C-reactive protein, mean ± SD | 0.44 ± 0.57 | 0.191 | 454.67 | 0.120 | 1.13 (0.97–1.29) | 0.094 |
Leukocytes, mean ± SD | 6.73 ± 2.13 | 0.160 | 457.26 | −0.003 | 1.00 (0.96–1.04) | 0.884 |
Brain leukocytes, mean ± SD | 6.22 ± 2.20 | 1.000 | 91.44 | 0.121 | 1.13 (1.04–1.23) | 0.005 |
Lymphocytes, mean ± SD | 28.89 ± 8.09 | 0.160 | 457.27 | 0.001 | 1.00 (0.99–1.01) | 0.929 |
Brain lymphocytes, mean ± SD | 29.05 ± 6.77 | 1.000 | 96.64 | −0.022 | 0.98 (0.95–1.01) | 0.117 |
Anti-dsDNA, mean ± SD | 98.79 ± 111.01 | 0.829 | 370.64 | < 0.001 | 1.00 (1.00-1.00) | 0.990 |
Brain anti-dsDNA, mean ± SD | 635.54 ± 1504.14 | 1.000 | 96.28 | < 0.001 | 1.00 (1.00-1.00) | 0.081 |
Correlations between depression, pain, and psychosocial values.
Pseudo |
AICc | Coefficient | Odds ratio (95% CI) | |||
---|---|---|---|---|---|---|
Rheumatoid arthritis | ( | |||||
Pain score, median | 6 | 0.411 | 527.31 | 0.114 | 1.12 (1.08–1.16) | <0.001 |
PSQI, mean ± SD | 8.09 ± 4.29 | 0.442 | 523.54 | 0.065 | 1.07 (1.05–1.09) | <0.001 |
Relationship Assessment Scale, mean ± SD | 28.74 ± 9.54 | 0.989 | 328.73 | −0.041 | 0.96 (0.95–0.97) | <0.001 |
Fatigue Severity Scale, mean ± SD | 4.19 ± 1.57 | 0.627 | 491.36 | 0.261 | 1.30 (1.22–1.38) | <0.001 |
Lupus (SLE) | ( | |||||
Pain score, median | 5 | 0.328 | 440.46 | 0.078 | 1.08 (1.05–1.11) | <0.001 |
PSQI, mean ± SD | 8.99 ± 2.97 | 0.081 | 463.29 | 0.037 | 1.04 (1.01–1.07) | 0.012 |
Relationship Assessment Scale, mean ± SD | 25.45 ± 4.19 | 0.818 | 370.67 | 0.014 | 1.01 (0.99–1.04) | 0.252 |
Fatigue Severity Scale, mean ± SD | 4.26 ± 1.38 | 0.551 | 410.99 | 0.268 | 1.31 (1.22–1.41) | <0.001 |
Depressive disorder | ( | |||||
Pain score, median | 5.5 | 0.600 | 185.10 | 0.132 | 1.14 (1.09–1.20) | <0.001 |
PSQI, mean ± SD | 8.48 ± 2.61 | 0.782 | 173.46 | 0.067 | 1.07 (1.02–1.12) | 0.008 |
Relationship Assessment Scale, mean ± SD | 22.52 ± 7.21 | 0.983 | 129.17 | 0.012 | 1.01 (0.99–1.03) | 0.250 |
Fatigue Severity Scale, mean ± SD | 3.83 ± 1.49 | 0.876 | 158.17 | 0.216 | 1.24 (1.13–1.36) | <0.001 |
Healthy controls | ( | |||||
Pain score, median | 0 | 0.331 | 112.07 | −17.810 | 0 (0–0) | 0.993 |
PSQI, mean ± SD | 5.18 ± 2.36 | 0.136 | 117.70 | 0.079 | 1.08 (0.99–1.18) | 0.072 |
Relationship Assessment Scale, mean ± SD | 28.06 ± 4.46 | 0.917 | 81.43 | −0.066 | 0.94 (0.89–0.98) | 0.009 |
Fatigue Severity Scale, mean ± SD | 2.78 ± 1.05 | 0.311 | 112.71 | 0.291 | 1.39 (1.10–1.65) | 0.005 |
Multivariate Model for RA subject cohort.
Variables | Coefficient | Odds ratio (95% CI) | |
---|---|---|---|
Sex | 0.164 | 1.18 (0.81–1.75) | 0.401 |
Age (years) | −0.001 | 1.00 (0.98–1.02) | 0.970 |
Education (years) | −0.071 | 0.93 (0.82–1.05) | 0.273 |
Marital status | −0.016 | 0.98 (0.82–1.16) | 0.854 |
Physical activity | −0.808 | 0.45 (0.25–0.74) | 0.003 |
Relationship Assessment Scale | −0.041 | 0.96 (0.94–0.98) | < 0.001 |
Fatigue Severity Scale | 0.206 | 1.23 (1.10–1.37) | < 0.001 |
Model fit summary | Pseudo |
||
0.999 | 247.280 |
Depressive symptoms were assessed for all participants using the HADS instrument. Therapeutics, clinical indicators, and psychological factors were all examined for correlation with symptoms of depression. Multiple correlations were found, encompassing all three areas. The use of the four different participant groups (rheumatoid arthritis, lupus (SLE), depressive disorder, and healthy controls) allowed the evaluation of the relative importance of each factor in rheumatoid arthritis. For example, if a particular condition correlated with depression in RA but not SLE, that would indicate that the condition was specific for RA and not a general feature of autoimmune arthritis.
In RA patients, the use of Tocilizumab, an IL-6 inhibitor, was associated with decreased depressive symptoms (
Fourteen RA patients were not initially on a biological therapeutic but were being treated with one at the time of the one-month follow-up. Of these patients, there was no significant change in depressive symptoms, although counter to what might be expected, they reported significantly higher fatigue (
Levels of IL-6, IL-10, and TNF-alpha were measured using ELISA for each participant. IL-10 concentration correlated with depressive symptoms in patients with rheumatoid arthritis (
IL-6 and TNF-alpha correlations were also identified, although only in the primary depression patients. In these patients, IL-6 and TNF-alpha were both positively correlated with depressive symptoms (
Disease activity was strongly correlated with depression in the RA patients in this study. Both patient and physician assessments of overall disease activity strongly correlated with depressive symptoms (
Certain physical factors also correlated with depression. In both the RA patients and the lupus patients, body mass index correlated strongly with depressive symptoms (
Self-reported pain scores correlated with depression in patients with RA (
Multivariate analysis was performed to identify which of the variables was the most predictive for depressive symptoms and to ensure that differing sociodemographic factors between groups were not responsible for the outcomes of the study. This analysis showed that none of the sex, age, marital status, or years of education significantly predicted depressive symptoms. The best correlations with depressive symptoms were found for low physical activity (
It is evident from these results that depression in rheumatoid arthritis is strongly linked to disease activity, as nearly every measure of systemic inflammation and disease activity was correlated with depressive symptoms. This is not the case in systemic lupus, where disease activity often bears little relation to depressive symptoms. Of the clinical indicators, only C-reactive protein levels were associated with active depression in lupus. This difference suggests that the root cause of depression is different in rheumatoid arthritis and lupus.
The results of the cytokine analysis support this finding. IL-10 is associated with depression in both lupus and rheumatoid arthritis, but whereas higher IL-10 levels correlate with depression in RA, lower levels correlate with depression in lupus. Since IL-10 is normally thought of as an anti-inflammatory cytokine, the association between IL-10 levels and depressive symptoms in RA is unexpected. It may be that increased IL-10 is serving as a proxy for increased inflammation. In this case, increased disease activity produces an increase in IL-10 as the normal immune regulatory systems attempt to damp down the inflammation. IL-10 seems to act differently in lupus and rheumatoid arthritis, being disease suppressive in RA and disease promoting in lupus [
IL-6 is also likely to be involved in depression. Treatment with the IL-6 inhibitor Tocilizumab was correlated with decreased depressive symptoms in the RA patients. Furthermore, IL-6 concentration correlated with active depression in the primary depression patients. This supports other studies that have suggested a role for IL-6 in depression [
The relatively small number of patients in the control groups and the demographic differences between groups are the principal limitation of the present study. Although the longitudinal design of the study could be a strength, the short time between evaluations represents another limitation.
The use of biological therapeutics is a powerful tool not only for treatment but also for aid in the understanding of the mechanisms and pathogenesis of rheumatic diseases. The ability of these medications to inhibit cytokines or pathogenic cells provides a mechanism to examine the function of these factors in a very specific way, as was done with IL-6 in this work. Experiments are currently underway to examine the roles of cytokines in RA in an extended longitudinal analysis by examining patients on biologicals. These techniques are capable of being applied to any of the many diseases that are treated with biological therapeutics.
All data used in the generation of this article are available upon request.
The authors declare that they have no conflicts of interest.
This study was supported by Portugal Fulbright Commission (Dr. Brian D. Poole).
The post hoc power analysis of this project is included as Supplementary Table 1. This shows the likelihood of a type 2 statistical error for each comparison, based on the actual variance observed in our study.
The complete database that the study is based on is also included as Supplementary Table 2. This includes the complete data set used for each analysis.