Highlights
• Patients with depression show reduced variability in pro-inflammatory immune measures.
• Patients with depression show increases in pro-inflammatory immune markers mean levels, and reductions in anti-inflammatory IL-4.
• Depression appears characterised by a homogenously pro-inflammatory state.
Abstract
Importance: The magnitude and variability of cytokine alterations in depression are not clear.
Objective: To perform an up to date meta-analysis of mean differences of immune markers in depression, and to quantify and test for evidence of heterogeneity in immune markers in depression by conducting a meta-analysis of variability to ascertain whether only a sub-group of patients with depression show evidence of inflammation.
Data Sources: Studies that reported immune marker levels in peripheral blood in patients with depression and matched healthy controls in the MEDLINE database from inception to August 29th 2018 were examined.
Study Selection: Case-control studies that reported immune marker levels in peripheral blood in patients with depression and healthy controls were selected.
Data Extraction and Synthesis: Means and variances (SDs) were extracted for each measure to calculate effect sizes, which were combined using multivariate meta-analysis.
Main Outcomes and Measures: Hedges g was used to quantify mean differences. Relative variability of immune marker measurements in patients compared with control groups as indexed by the coefficient of variation ratio (CVR).
Results: A total of 107 studies that reported measurements from 5,166 patients with depression and 5,083 controls were included in the analyses. Levels of CRP (g=0.71; 95%CI: 0.50-0.92; p<0.0001); IL-3 (g=0.60; 95%CI: 0.31-0.89; p<0.0001); IL-6 (g=0.61; 95%CI: 0.39-0.82; p<0.0001); IL-12 (g=1.18; 95%CI: 0.74-1.62; p<0.0001); IL-18 (g=1.97; 95%CI: 1.00-2.95; p<0.0001); sIL-2R (g=0.71; 95%CI: 0.44-0.98; p<0.0001); and TNFα (g=0.54; 95%CI: 0.32-0.76; p<0.0001) were significantly higher in patients with depression. These findings were robust to a range of potential confounds and moderators. Mean-scaled variability, measured as CVR, was significantly lower in patients with depression for CRP (CVR=0.85; 95%CI: 0.75-0.98; p=0.02); IL-12 (CVR=0.61; 95%CI: 0.46-0.80; p<0.01); and sIL-2R (CVR=0.85; 95%CI: 0.73-0.99; p=0.04), while it was unchanged for IL-3, IL-6, IL-18, and TNF α.
Conclusions and Relevance: Acute depression is confirmed as a pro-inflammatory state. Some of the inflammatory markers elevated in depression, including CRP and IL-12, show reduced variability in patients with depression, therefore supporting greater homogeneity in terms of an inflammatory phenotype in depression. Some inflammatory marker elevations in depression do not appear due to an inflamed sub-group, but rather to a right shift of the immune marker distribution.
Keywords: DepressionInflammationMeta-analysisHeterogeneityCytokineCRP
4. Discussion
Our meta-analysis finds evidence that mean-scaled variability, measured as CVR, is reduced in patients with depression for CRP, IL-12 and sIL-2R, while it is unchanged for IL-3, IL-6, IL-18 and TNF α. In the same sample, we also find that blood levels of CRP, IL-3, IL-6, IL-12, IL-18, sIL-2R and TNF α are significantly elevated in patients with depression with medium-large effect sizes (range 0.54-1.97), and that these findings are robust to a range of potential confounds and moderators. See Table 1 for a summary of our findings.
Our study is, to our knowledge, the first meta-analysis of variability of immune parameters in individuals with depression compared to matched controls. Mean differences in inflammatory markers in depression have been meta-analysed before (Howren et al., 2009, Dowlati et al., 2010, Haapakoski et al., 2015, Goldsmith et al., 2016, Köhler et al., 2017). However, as shown in Supplementary Table 1, this study is by far the largest meta-analysis of immune markers in depression, including a sample 1.48 times larger than the largest previous one. In addition to this, this is one of the first studies to systematically consider the effect on immune markers of excluding patients not experiencing an active depressive episode (previously only considered in a much smaller study by Goldsmith et al), duration of illness (previously only considered descriptively), study quality (previously only considered in a smaller study by Haapakoski et al), and smoking (previously only considered by Kohler et al). Furthermore, our findings of increased mean levels of CRP, IL-6, IL-12 and TNF α in depression replicate previous meta-analytical findings; the same can be said of no changes in levels of TGF β (Table 2). Reductions in IL-4, found in our study with an effect size of -0.73 and resistant to most sensitivity analyses, were not significant in Kohler et al, 2017 nor in (Dowlati et al., 2010), however both these studies were based on considerably smaller samples, which could explain the difference. More controversial is the result for IFNγ, which we find not significantly altered in our main analysis and increased in patients when excluding studies not matched for smoking levels between cases and controls. Given that previous, smaller meta-analyses were also non-concordant with regards to IFNγ (Dowlati et al., 2010, Goldsmith et al., 2016, Köhler et al., 2017), we believe therefore that more research is needed to establish the relationship between IFNγ levels and depression.
4.1. Interpretations and Implications
4.1.1. Meta-analysis of Heterogeneity
In a previous study we have shown that patients with depression show a proportion of high CRP levels at different cut-offs (CRP >1mg/L, >3mg/L and >10mg/L) that is similar to matched controls (Osimo et al., 2019); this supported the hypothesis that the shape of the CRP distribution curve is similar in patients and controls. In this study we find that mean-scaled variability of CRP and of a number of other immune markers is either reduced or unchanged in patients with depression as compared to healthy controls. A reduced variability implies a narrower distribution in patients than in controls, and possibly even a greater homogeneity in the inflammatory phenotype in depression. Therefore, the findings to date, at least for markers that show elevations of the mean and reductions in heterogeneity such as CRP, support a narrower distribution that is shifted to the right in depression. This is important as in the past there have been suggestions that inflammation in depression could be due to a sub-group of “inflamed and depressed” subjects, who might potentially be part of a separate sub-group of the depressed population (Miller and Cole, 2012). Our findings, instead, point in the direction of a continuous distribution of inflammatory markers in the depressed population, which is more homogenous than the healthy population.
The reduction in variability in CRP is worthy of a special mention here, as CRP is the main inflammatory marker routinely measured in clinical practice (Yeh, 2004;109(21_suppl_1):), and it is commonly used to stratify patients based on peripheral inflammatory levels in immunopsychiatric studies. Activation of the inflammatory system is thought to underlie antidepressant resistance (Chamberlain et al., 2018, Benedetti et al., 2002, Lanquillon et al., 2000, Carvalho et al., 2013), highlighting a potential involvement in treatment response (Carvalho et al., 2013, Maes et al., 1997, O’Brien et al., 2007, Yoshimura et al., 2009). Therefore, whether targeting inflammatory cytokines could provide therapeutic benefit for patients with depression is a key question that is being investigated in ongoing trials (e.g. NCT02473289; ISRCTN16942542). Our findings will be relevant for future studies assessing inflammation in depression, especially those recruiting patients based on their baseline inflammatory status.
4.1.2. Meta-analysis of mean differences
We found increases in the average levels of type I and other pro-inflammatory cytokines such as IL-3, IL-6, IL-12, IL-18 and TNF α; we also found reductions in IL-4, one of the main anti-inflammatory and immune-modulatory cytokines; finally, we found mean increases in CRP, which is one of the best characterised inflammatory markers in medical (Danesh et al., 2000, Visser et al., 1999) and psychiatry conditions (Fernandes et al., 2016, von Känel et al., 2007, Fernandes et al., 2016). Taken together, these results confirm that acute depression is associated with a pro-inflammatory state.
CRP is one of the best studied inflammatory markers in the field of medicine. Higher levels of CRP have been consistently found in cross-sectional studies and in population-based longitudinal studies of depression, often preceding the onset of illness (Gimeno et al., 2009, Khandaker et al., 2014, Wium-Andersen et al., 2013, Zalli et al., 2016), suggesting that inflammation could be a cause rather than simply a consequence of the illness; supporting this hypothesis, recently Mendelian randomization analyses of the UK Biobank sample found that IL-6 and CRP are likely to be causally linked with depression (Khandaker et al., 2019). Furthermore, elevated peripheral CRP levels have been found to correlate with its level in the central nervous system, with a strong correlation between plasma and CSF CRP (r = 0.855, p < 0.001) (Felger et al., 2018).
TNF α is one of the major pro-inflammatory cytokines; it is produced by dendritic cells and macrophages and is a major activator of downstream inflammatory cascades with multiple effectors (Abbas et al., 2014). During acute infection dendritic cells and macrophages also produce IL-6 and IL-12; both are type I cytokine family members, secreted in response to an acute inflammatory stimulus (Abbas et al., 2014). IL-12 plays a central role in responses to active infection promoting Th1 responses and, hence, cell-mediated immunity (Stern et al., 1996). TNF α, IL-6 and IL-12 increases in current depressive episodes underline the systemic nature of the inflammatory status, showing some similarity to the immune reaction to an active infection.
For markers found to be overall not different between patients and controls, but with variable results in sensitivity analyses (IL-5, IFNγ and TGF β), our results encourage further research, aiming to disentangle their potential tole in mediating effects of treatment (IL-5), smoking (IFNγ) or BMI differences (TGF β).
Finally, IL-2 and IL-8 were found to be increased in patients in our main analysis, but produced discordant results in sensitivity analyses due to the effect of BMI-matching; future studies should carefully match participants for BMI as this appears to be a particularly relevant factor affecting immune status.
4.2. Strengths and Limitations
The main strength of this work is the use of the largest sample of studies of inflammatory markers in depression to date; the same large sample was used to study heterogeneity and mean differences in patients as compared to controls. Even if we could not make inferences on the shape of the distribution, such as modality, as this would require individual subject data, we were able to obtain the first measure to date of the variability of inflammatory markers in depression.
A further strength of this paper is the employment of a systematic approach to the analysis of potential confounds. Given the large number of studies that focussed on inflammatory markers in depression, we were able to investigate the effect of potential psychiatric (e.g. treatment status, current depressive episode at time of sampling and duration of illness) and lifestyle confounds (e.g. age, BMI and smoking status), as well as statistical and sampling confounds (e.g. data skew and study quality). Sensitivity analyses focussing on studies with strict environmental and physiological matching provided us with greater confidence that depression is associated with the elevation of some immune parameters. Use of a multivariate meta-analytic approach to reduce the influence of multiplicity is a further strength.
Among our limitations, we included cross-sectional studies which used different tools to diagnose depression, even if only studies using ICD or DSM diagnostic criteria were included. Inconsistency between studies was moderate to high. This could reflect methodological factors, e.g. differences in assay sensitivity. However, the random-effects model used is robust to inconsistency, and would not explain our variability findings, because these reflect within-study variation (with methodologic factors common to patient and control groups in any given study). Due to data unavailability, some sensitivity analyses might be subject to type II error, i.e. false negatives; for example, BMI-matched sensitivity analyses often included samples much smaller than that of the main analysis. Furthermore, sensitivity analyses of antidepressant naïve and treatment resistant patients were not possible owing to insufficient studies.
Although all studies included in analyses used well validated quantification techniques, insufficient assay sensitivity may have limited the ability to detect subtle differences in immune parameters between patients and controls, particularly for titres beneath the limit of assay detection. Unfortunately, very few studies (2 out of 106) reported the number of samples below the limit of assay detection, so this factor could not be taken into account. Positive data skew can inflate standard deviation due to outliers within the ‘tail’ of the data (Fayers, 2011). However, we demonstrated no significant difference in the proportion of skewed data sets between patients and controls, suggesting that influence of skew was equal. Thus, excessive skew in healthy controls compared with patients was not likely contributing to results.
We excluded papers that only included patients and controls presenting the same co-morbidity or physiological state in addition to depression (such as studies in autoimmune disorders or pregnancy) to reduce the risk of bias. Most included studies excluded participants with co-morbid medical conditions, and the presence of co-morbidity in participants was assessed as one of the items of our quality assessment of papers. It was not possible to exclude all co-morbidity due to original data quality, but we are confident this issue is not going to significantly affect results as a) we used random effect models to account for additional variation; b) co-morbidity is likely to be equally distributed between cases and controls; and c) our large sample (the largest to date) allows for more individual variation without affecting results.
A very limited number of studies on CRP excluded participants presenting with an acute infection (CRP >10 mg/L); we decided to include these studies because we previously found that the odds ratio of inflammation in patients vs controls is very similar if considering all patients (OR=1.46) or excluding patients and controls with CRP >10 mg/L (OR=1.44) (Osimo et al., 2019), thus suggesting that an equal proportion of patients and controls present with acute inflammation.
4.3. Conclusions and future directions
In this study we found a reduction in mean-scaled variability in CRP, IL-12 and sIL-2R. We found increases in the mean levels of CRP, IL-3, IL-6, IL-12, IL-18, sIL-2R and TNF α in patients with depression. These results survived sensitivity analyses for psychiatric and lifestyle predictors, influence of skew, influence of poor-quality studies and publication bias.
Our results confirm that acute depression is a pro-inflammatory state, and lend support to the hypothesis that inflammatory marker elevations in depression are not due to an inflamed sub-group, but rather to a right shift of the immune marker distribution. However, future research should specifically address the inflammatory sub-group hypothesis of depression, which can only be directly tested in an individual-patient meta-analysis.
Posts
No comments:
Post a Comment