Systematic review and meta-analysis of the association between peripheral inflammatory cytokines and generalised anxiety disorder

Objective Inflammation has been implicated in the aetiology of mental illness. We conducted the first systematic review and meta-analysis of the association between peripheral markers of inflammation and generalised anxiety disorder (GAD). Design Systematic review and meta-analysis of studies measuring peripheral cytokine levels in people with GAD compared with controls. Data sources MEDLINE (1950–), EMBASE (1947–), PsycINFO (1872–) and Web of Science (1945–) databases up until January 2018. Eligibility criteria Primary, quantitative research studies of people with a diagnosis of GAD assessed using a standardised clinical interview that measured peripheral inflammatory markers. Data extraction and synthesis Two independent reviewers extracted data and assessed study quality. Meta-analysis using a random-effects model was conducted for individual cytokines where data from three or more studies were available. Results 14 of 1718 identified studies met the inclusion criteria, comprising 1188 patients with GAD and 10 623 controls. In total 16 cytokines were evaluated. Significantly raised levels of C reactive protein (CRP), interferon-γ and tumour necrosis factor-α were reported in patients with GAD compared with controls in two or more studies. Ten further proinflammatory cytokines were reported to be significantly raised in GAD in at least one study. However, 5 of 14 studies found no difference in the levels of at least one cytokine. Only CRP studies reported sufficient data for meta-analysis. CRP was significantly higher in people with GAD compared with controls, with a small effect size (Cohen’s d=0.38, 0.06–0.69), comparable with that reported in schizophrenia. However, heterogeneity was high (I2=75%), in keeping with meta-analyses of inflammation in other psychiatric conditions and reflecting differences in participant medication use, comorbid depression and cytokine sampling methodology. Conclusion There is preliminary evidence to suggest an inflammatory response in GAD, but it remains unclear whether inflammatory cytokines play a role in the aetiology. GAD remains a poorly studied area of neuroinflammation compared with other mental disorders, and further longitudinal studies are required.


Introduction
There is growing evidence for immune mediated pathogenic mechanisms in several psychiatric disorders with discrete profiles of inflammatory mechanisms 1 . Epidemiological evidence has shown an increased risk of mood disorders and psychosis in people with a history of severe infection or autoimmune conditions 2, 3 . This has been supported by genome wide association studies implicating multiple immune signalling pathways 4 , and altered profiles of pro-inflammatory cytokines and acute phase reactants in schizophrenia 5 , depression 6 , obsessive compulsive disorder (OCD) 7 and bipolar disorder 8 . However, the relationship between inflammation and mental illness remains poorly understood and controversial, with a number of proposed potential neuropathological mechanisms 9,10 , including changes in microglial function 1 , glutamatergic excitotoxicity 11 , synaptic plasticity 12 and reduced hippocampal neurogenesis 13 .
Despite increasing interest in the role of inflammation in mental illness, relatively little research has focused on potential associations with anxiety disorders 14 . These are common, with an estimated lifetime prevalence of 7.3% to 28.8%, are associated with substantial functional impairment and are estimated to cost between 42-47 billion dollars to the U.S economy each year 15,16 . However, only 60% of patients are thought to respond to pharmacological and psychological treatments and understanding of the underlying pathophysiological mechanisms of anxiety disorders remains poor 17 .
Generalised anxiety disorder (GAD) is the most common anxiety disorder, with a degree of associated disability equivalent to that of major depressive disorder (MDD) 18 . Despite psychopharmacological 19 and psychological 20 treatments showing effectiveness in GAD, 42% of people living with GAD experience ongoing symptoms after 12 years and half of remitted patients experience recurrence 18 . GAD is more prevalent in those with inflammatory conditions such as rheumatoid arthritis (RA) 21,22 , with case series studies suggesting symptoms are less common with immune modulating treatment targeting specific inflammatory cytokines 23 . The chronic clinical course and relatively high probability of recurrence in GAD, in addition to preliminary evidence of an inflammatory component in other anxiety disorders 7,24 , suggest that inflammation could be an important neurobiological mechanism in the aetiology of this disorder.
To date, two previous reviews of inflammatory biomarkers in GAD have been conducted. Of these, however, one was a narrative review 25 and the other was restricted to literature published within the last decade 14 , and with a focus on all anxiety disorders. Both reviews reported that there was preliminary evidence for inflammatory changes in GAD. However, only three studies were identified by the systematic review reporting cytokine changes in GAD and no meta-analysis was performed. No study to date has conducted a comprehensive systematic review and meta-analysis of all current literature focusing on GAD or commented on the longitudinal association between inflammation and GAD.
We aimed to systematically review the cross-sectional and longitudinal associations between inflammatory biomarkers and GAD, and perform the first meta-analysis of inflammatory biomarkers in GAD.

Method
We conducted a systematic review of studies that had included people with GAD who had undergone peripheral cytokines measurement and a between group meta-analysis of cytokine F o r p e e r r e v i e w o n l y levels in people with GAD compared to controls. We conducted the study according to Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines 26 .
We searched MEDLINE (1950-), EMBASE (1947-), PsycINFO (1872-) and Web of Science (1945-) databases up until January 2018. Reference lists of eligible studies were then searched for further ones that met eligibility criteria.
Our search terms were: (inflammat* or cytokine or interferon or IFN or interleukin or "translocator protein" or TSPO or "tumour necrosis factor" or "tumor necrosis factor" or TNF or IL-1 or IL-2 or IL-4 or IL-7 or IL-6 or IL-8 or IL-10 or migroglia or t-cell or lymphocyte or "C-reactive protein" or "C reactive protein" or CRP or "acute phase protein" or "fibrinogen") and ("generalised anxiety disorder" or "generalized anxiety disorder" or GAD or worry).
We included primary, quantitative research studies (including unpublished theses and dissertations), written in any language, that included people with a diagnosis of GAD assessed using standardised clinical interview (e.g. Structured Clinical Interview for DSM 27 ) or standardised psychometric instruments. Studies reported cross-sectional or longitudinal data in clinical or community populations. Cross-sectional studies measured inflammatory biomarker concentrations in anxious people versus non-anxious healthy controls, while longitudinal studies measured inflammatory biomarker concentrations at baseline and anxiety scores at follow-up. Inflammatory markers were measured in the unstimulated state (no antigen induced stimulation of cytokine production) and sampled from peripheral blood, CSF, or saliva at any time of day. Exclusion criteria included studies with less than 5 participants, studies in animals and studies where subjects were participants in the treatment arms of clinical trials.

Data extraction and quality assessment
Data were extracted and quality assessed for all studies that met eligibility criteria by two independent raters (HC, EA) with disagreements settled by consensus and discussion. For each cytokine, we extracted the means, variance estimates or 95% confidence intervals (CIs) and sample size for GAD and control groups. We also extracted demographic data (e.g. age, sex) and clinical data (e.g. medication use, co-morbid depression, severity) where available.
Authors were contacted for further information, where necessary.
Risk of bias and study quality were evaluated using the Newcastle-Ottawa Quality Assessment Scale 28 . Other potential confounding factors (including assay type and sensitivity, inflammatory marker analysis and recruitment methods) were also examined to allow more detailed bias and quality analysis of studies.

Strategy for data synthesis
Separate meta-analyses were performed for individual biomarkers in GAD versus controls if sufficient data were available from a minimum of three studies. Due to different measurement methods and anticipated high heterogeneity, we estimated a standardised mean difference (SMD) for each inflammatory marker, and used a random effects model for metaanalysis, conducted using Revman 5. Heterogeneity across studies was quantified with the I 2 statistic, with a value of 25% typically regarded as low, 50% as medium, and 75% as high 29 . If studies were longitudinal or trials of interventions with multiple data collection points, we examined baseline data only to avoid skewed meta-analysis from inclusion of more than one effect size from the same study.
Characteristics of the 14 included studies are shown in Tables 1 and 2. Studies comprised a total of 1188 people with a diagnosis of GAD and 10,623 controls, with a further 116 participants from a study that did not report GAD and control group sizes.
Twelve studies (85.7%) reported the assay method used, all of which were versions of an enzyme-linked immunosorbent assay (ELISA) or enzyme immunoassay (EIA). However, only 7 studies (50%) reported assay sensitivity. All but one study used blood component samples to assess inflammatory marker levels, with the most common sample type being serum (N=6, 42.9%) and plasma (N=4, 28.6%).

Risk of bias and quality in individual studies
All included studies had adequate case definition, with participants meeting diagnostic criteria for GAD according to DSM or ICD, with 12 studies (85.7%) using a structured clinical interview for assessment (see Table 1).
Most (71.4%) studies included people aged 18- known to correlate with inflammation, and only half of these accounted for BMI differences in analysis of group differences 44 (see Table 2).
Use of psychotropic medication and presence of comorbid major depressive disorder are important moderators of inflammation in other psychiatric disorders 24 . Six studies (42.8%) excluded patients who used psychiatric or other immune-modulating medication, though only two studies (4.3%) reported medication use. The majority of studies (64.2%) either excluded patients with co-morbid MDD or adjusted for this in analyses.
Concurrent physical illness is clearly an important determinant of inflammatory cytokine levels and this was accounted for by the majority of included studies by either excluding participants with co-morbidities (5 studies, 35.7%) or adjusting for chronic physical illness in group comparisons (6 studies, 42.8%), though two studies specifically only included participants with co-morbid cardiovascular disease. Use of a pre-determined cut-off value for cytokine levels was employed by three studies (21.4%) to ensure that cases with acute infection were excluded from the sample.
Many inflammatory markers exhibit a diurnal pattern of expression and are affected by consumption of food, thus time of day of sampling and whether the sample was taken in a fasted state are important factors to consider in analysing relative levels of cytokines 45 . However, time of day of sampling was only recorded in a minority of studies (6 studies, 42.8%), and the same number of studies recorded whether fasted samples were taken.
The overall quality of studies included in the review varied significantly, with Newcastle-Ottawa scale scores ranging from 2 to 9 (see Table 3). The area in which most studies were inadequate was in reporting non-response rate and detailing recruitment methods (see Table  3). Lowest quality studies were abstracts or dissertations, and two studies lacked control groups as only GAD patients were sampled (see Table4).

C-reactive protein
CRP is a critical early pro-inflammatory surveillance molecule involved in the activation of the complement system and both innate and adaptive immune systems 46 . We identified nine studies that investigated the association between GAD and CRP, comprising a total of 11,486 participants (see Table 5).
Four studies, involving 578 GAD patients and 4046 controls, provided sufficient information to conduct a meta-analysis of CRP levels in GAD 33,39,41,43 (see Figure 2). This was the only inflammatory marker for which meta-analysis was possible. Meta-analysis showed significantly raised CRP in GAD compared to controls (SMD 0.38, 95% CI 0.06-0.69; Z=2.36, p=0.02) (see Figure 2). However, there was a large and statistically significant degree of heterogeneity between studies (Chi2=12.0; df=3; p=0.007; I 2 = 75%). Given the high heterogeneity and inclusion of less than 10 studies in the meta-analysis we did not have sufficient power to examine publication bias 47 . Five studies 30,33,39,41,43 (n=4669), one of which was conducted in 16 year olds 39 and two in participants with co-morbid heart disease 30 , reported significantly higher CRP levels in participants with a diagnosis of GAD. The largest study 31 (n=5810) examining CRP in GAD examined CRP levels in children from baseline measurement aged 9-16 years to follow up aged 19-21.This was the only study to examine the longitudinal association between GAD and CRP, and found a bivariate association both cross-sectionally and over time between GAD and elevated CRP , however, this was accounted for by potential co-variates including BMI and medication use. The only study 40 to find an inverse correlation between CRP and GAD was conducted in non-smoking women from a longitudinal study in Finland and did not specify the numbers of participants with a diagnosis of GAD or group differences.
No difference was found in a cohort study (n=821) that used a combined inflammatory index consisting of CRP, IL-6 and TNF-α in 93 patients with a diagnosis of GAD and controls with a history of cardiovascular disease (CVD) 35 . Subgroup analysis examining differences in individual inflammatory markers was not reported 35 .
We found two studies 36,43 (n= 196) that examined the association of severity of GAD symptoms with CRP level. One found a significant positive correlation between CRP level and GAD-7 scores 43 , and the other reporting CRP differences in 70 GAD patients with and without a diagnosis of alexithymia found a significant association between higher CRP and suicidal ideation 36 .
Although the meta-analysis and the majority of included studies reported raised CRP in GAD, there was wide variation in reporting and adjustment for important potential moderators, including co-morbid MDD, use of medications, assay used and time of day of blood collection, all of which likely contributed to the high degree of heterogeneity between studies. Of the nine studies to analyse CRP, four (44.4%) did not exclude or adjust for medication use by participants 30,35,39,40 . Co-morbid MDD was not adjusted for in analysis by two studies 31,35 , one of which was included in the meta-analysis 31 . Only three of the nine studies reported time of sample collection 33,36,43 or whether this was in a fasted state 33,35,36 , and though all studies utilised a similar assay method, different assay types were used in every study.
In summary, of the nine studies to have examined differences between GAD and controls, the majority reported raised CRP in GAD and meta-analysis found significantly raised CRP in GAD with a small effect size. However, there was wide variation in study methods including variable adjustment for mediators of inflammation such as co-morbid MDD, medication use and sampling methods. Only one study examined CRP in GAD longitudinally, reporting a bivariate association accounted for by health seeking behaviours 31 .

Interleukins
Seven studies examined the association between interleukins and GAD (see Table 3). IL-6 is a mediator of T-cell and B-cell activation and induces acute phase proteins in hepatocytes,  46 . Pharmacological blockade of IL-6 action is used to treat several auto-immune conditions including RA, and raised IL-6 has been associated with a number of psychiatric conditions including depression, schizophrenia and PTSD 46,48 . We found IL-6 was the most frequently measured interleukin, with five studies (n=2066) examining changes in GAD patients compared to controls [33][34][35]37,43 . The largest study investigated differences between 454 participants with a diagnosis of GAD and 556 controls from The Netherlands Study of Depression and Anxiety (NESDA) cohort 33 . Though analysis was conducted on anxiety disorders as a whole, mean difference in IL-6 in people with GAD compared to controls obtained through direct communication with the author showed significantly higher levels in GAD. However, it is unclear whether these differences remain significant after adjustment for group differences and no associations were found between IL-6 and participants who had all types of anxiety disorder 33 . Two studies 34,43 (n=165), one of which used saliva samples 34 , reported significantly higher IL-6 in medication naïve participants with a diagnosis of GAD compared to age and sex matched healthy controls.
No difference was found in a combined inflammatory index consisting of CRP, IL-6 and TNF-α in a study of 93 patients with GAD and co-morbid ischaemic heart disease 35 . One case-controlled study 43 of 48 Chinese outpatients presenting for the first time with a diagnosis of GAD and 48 age, sex and education matched controls accounted for all results for IL-1α, 5, 8, 12p70. This study found significantly higher levels of IL-1α, -8 and -12p70 in GAD patients, in addition to higher levels of IL-1α and IL-8 with increased severity of GAD (as measured by GAD-7 scale), but did not account for chronic physical co-morbidities during recruitment or in analysis. Both IL-1α and IL-8 have pro-inflammatory functions as chemoattractants for leukocytes and haematopoiesis, and have been targeted for treatments in a number of auto-immune conditions 46 . However, there was no association between GAD and IL-5, which is thought to predominantly mediate myeloid cell activation, and is a target of treatment in asthma 46 . The same study 43 also examined IL-2, which has a major role in T-cell mediated autoimmune and inflammatory conditions 46 . Results showed significantly higher IL-2 in GAD patients, however, this conflicted with results from a smaller study (n=24) that found no significant difference between medication naïve GAD patients and controls, though few details of the characteristics of participants, sampling or analysis were reported in this abstract 32 .
One study that measured IL-1 using sputum analysis found significantly higher levels in GAD compared to controls in 69 participants recruited from the same Chinese hospital 34 . Though IL-1 is pro-inflammatory, there are differences in function dependent on the class of IL-1 protein measured which was not reported in this study 34 .
IL-4 has several pro-inflammatory functions including immunoglobulin-E (Ig-E) class switching, expression of MHC class II and acts as a survival factor for T and B cells 46 . The only study to measure IL-4, found no differences between 54 GAD patients recruited from community mental health teams and primary care after controlling for age, sex, BMI, smoking, alcohol consumption and co-morbid depression 38 . This study 38 also investigated IL-10, which was the only cytokine with an anti-inflammatory function to be measured and is This found significantly lower levels of IL-10 (OR 0.35 p = 0.003) in GAD patients. However, this opposed findings from a smaller study (n=24) which reported significantly higher levels of IL-10 in GAD patients compared to controls, though it was not reported whether this association remained significant after controlling for group differences 32 .
In summary, IL-6 was the most commonly measured interleukin raised in GAD compared to controls in the majority of studies, however, no study examined the longitudinal association with GAD. Other interleukins were examined by relatively few studies that examined small numbers cross-sectionally with mixed findings.

IFN-γ
IFN-γ has anti-viral roles including promoting cytotoxic activity, MHC class I and II upregulation, NK cell activation, and is a treatment target in inflammatory conditions such as Crohn's disease 46 . Three studies investigated IFN-γ levels in GAD (n= 330) 38,40,43 . The largest study (n=118) found higher IFN-γ in GAD patients from the UK that remained significant after adjustment for age, gender, BMI, smoking, alcohol and co-morbid depression, but did not adjust for anxiolytic medication use in analysis 38 . This finding was supported by a study of 96 participants which reported higher IFN-γ levels in GAD and a significant positive correlation between anxiety severity and IFN-γ 43 . Conflicting findings were reported by a Finnish study of 116 participants, which found significantly lower IFN-γ in GAD patients. However, the number of participants with a diagnosis of GAD, differences between groups and adjustment for potential confounders were not reported 40 . In summary, only a few small cross-sectional studies have examined differences in IFN-γ between GAD and control groups, and their findings were mixed.

TNF-α
TNF-α has a wide array of roles in host defence, including initiating a strong acute inflammatory response but limiting duration of inflammatory activation, and is the target of blocking monoclonal antibodies in the treatment of a wide array of autoimmune conditions including Crohn's disease and RA 46 . Six studies (n=2300) investigated TNF-α in GAD, with mixed findings. Three studies (n=303) found TNF-α significantly raised in GAD patients compared to controls 34,38,40 . However, the largest study to measure TNF-α (n=1010) found no difference between participants with GAD and controls, and no correlation between TNF-α and anxiety symptoms 33 . This finding was supported by a study of 93 patients with GAD and co-morbid ischaemic heart disease using a combined inflammatory index of CRP, IL-6 and TNF-α which reported no difference compared to controls 35 . In summary, though the majority of studies to measure differences in TNF-α between GAD and controls reported significantly raised levels, these comprised small cross-sectional studies and the largest study reported no difference.

Other cytokines
One study compared levels of the pro-inflammatory cytokines CCL-5, MCP-1 and SDF-1 in 120 medication naïve physically well patients with a diagnosis of GAD and co-morbid  42 . Significantly higher levels of MCP-1 and SDF-1 were reported in both men and women, and higher CCL-5 in men but not women with a diagnosis of GAD compared to controls 42 .

Discussion
To our knowledge, this is the first systematic review and meta-analysis focusing on inflammatory cytokines in GAD. Using a range of databases we identified 14 studies, comprising 1188 participants with GAD and which measured 16 cytokines. We found significantly raised levels of CRP, IFN-γ and TNF-α in people with GAD compared to controls which were findings replicated in two or more studies. A further 10 proinflammatory cytokines were reported to be significantly raised in GAD in at least one study, however, 5/14 studies found no difference in at least one cytokine.
Despite substantial efforts to acquire data by contacting authors, it was only possible to conduct a meta-analysis of CRP. This identified significantly higher levels in GAD compared to controls with a small effect size (SMD: 0.38), though there was evidence of significant heterogeneity across studies (I 2 = 75%). This effect size in CRP is greater than has been reported in other anxiety disorders (PTSD: SMD = -0.14) 24 or MDD (SMD= 0.14) 6 , and is similar to that reported in schizophrenia (SMD= 0.45) 49 . Though we were only able to metaanalyse CRP, meta-analyses of different cytokines in other anxiety disorders have been conducted with larger effect sizes. A meta-analysis of inflammatory markers in PTSD identified 20 studies which reported increased interleukin 6 (IL-6), interleukin 1β (IL-1β), tumour necrosis factor-α (TNFα), and interferon γ (IFN γ) levels with effect sizes ranging from small (IFN γ: SMD 0.49) to large 1.42 (IL-1β: SMD 1.42) 24 . However, a systematic review and meta-analysis of pro-inflammatory cytokines in OCD identified 12 studies, and concluded that there was a significant reduction in IL-1β with moderate effect size (SMD: -0.60, p<0.001), and only IL-6 levels were significantly increased after subgroup analysis in medication-free adults with OCD 7 . It is unclear whether this profile of inflammatory marker changes would follow a similar pattern in GAD if future studies enabled further metaanalysis.
However, our findings should be interpreted with caution, due to the high heterogeneity among studies, low participant numbers and inconsistent reporting and adjustment for known confounding factors such as BMI, smoking, medication use and co-morbidities. It was not possible to analyse the cause of the degree of heterogeneity due to the paucity of studies.
Other known mediators of inflammation 24 such as physical activity, raised blood pressure and genetics were not accounted for. Furthermore, reporting of GAD severity and duration of symptoms was generally poor, preventing detailed analysis of whether inflammatory markers predicted outcomes and quality of life. We also found limitations in inclusion of specific demographics of participants with GAD. For example, despite GAD in older adults being  50,51,52 , only two studies included participants over the age of 65, both of which only included patients with co-morbid ischaemic heart disease.
We are beginning to understand the interplay between cytokines, the immune system and mental health 1,53 . At a molecular level we are aware that pro-inflammatory cytokines, including IFN, IL-1 and TNF, can reduce the availability of monoamines by inducing expression of pre-synaptic reuptake pumps and inhibiting enzymes involved in monoamine synthesis 54 , linking the monoamine theory of anxiety with inflammatory mechanisms. There is also a growing understanding of the relationship between systemic inflammation and the central nervous system (CNS) 1,55 . Microglial activation has been shown to be mediated by peripheral cytokines, and increased activation has been found in post-mortem studies of patients with MDD and schizophrenia 1 . No study we identified correlated inflammatory marker changes with in vivo microglial activation imaging in GAD and to our knowledge no research on post-mortem microglial changes in GAD has been conducted. Increased neuronal activity has also been shown to induce inflammatory and vascular changes in the brain, suggesting that psychological stress can not only be induced by inflammation but perpetuate chronic low grade inflammation seen in other vascular and neurodegenerative disorders 55 .
Understanding interactions between the CNS and immune system, and identifying biomarkers of GAD offers potential for novel therapeutic approaches. The revolution of development of monoclonal-antibody therapies for inflammatory disorders 56 raises the possibility of repurposing these medications for trials in treatment resistant GAD if specific and consistent profiles of inflammatory biomarkers are identified.
However, it remains unclear as to whether inflammation plays a causal role in GAD 48,54 . For example, although IL-6 is a successful target for treatment in a number of auto-immune conditions and raised IL-6 is implicated in several psychiatric disorders, it also acts to reduce other pro-inflammatory cytokines such as TNF via negative feedback and is induced by physical exercise, hyperthermia, fasting, sleep deprivation and sunlight exposure without activation other pro-inflammatory cytokines 48 . This raises the question as to whether inflammation in GAD is a consequence rather than cause of symptoms. This will only be answered by large prospective longitudinal studies, better characterising the relationship between inflammation and GAD, and our review identified only one published study that examined inflammation in GAD patients longitudinally in a cohort of adolescents.

Conclusion
There is some preliminary evidence to suggest a raised inflammatory response in GAD, although it is unclear whether inflammatory cytokines play a role in aetiology. GAD remains a poorly studied area of psychiatric neuroinflammatory research compared to other mental illnesses such as MDD and schizophrenia. While we are a long way from using inflammatory cytokines as a biomarker or treatment target in GAD, current findings reflect inflammatory changes seen in other mental illnesses and highlight the importance of ongoing investigation of the role inflammation plays in the development and course of GAD. Further,     ↑ IL-1α in GAD patients compared to controls and ↑ with increased severity of GAD.

Interleukin 5 (IL-5)
Tang et al. 2017 48 48 ↔ in GAD patients compared to controls, or association with severity of GAD.

Chemokine C-C motif ligand 5 (CCL-5) / Regulated on activation, normal T cell expressed and secreted (RANTES)
Oglodek ↑ = statistically significant increase in inflammatory marker in people with GAD compared to controls (p< 0.05), ↓ = statistically significant decrease in inflammatory marker in people with GAD compared to controls (p< 0.05), ↔ = no statistically significant difference in inflammatory marker in people with GAD compared to controls (p> 0.05). RCT = randomised controlled trial.

Study selection
17 Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally with a flow diagram.

5
Study characteristics 18 For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and provide the citations.

5
Risk of bias within studies 19 Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12).

5-6
Results of individual studies 20 For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention group (b) effect estimates and confidence intervals, ideally with a forest plot.

DISCUSSION
Summary of evidence 24 Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., healthcare providers, users, and policy makers).

6-11
Limitations 25 Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified research, reporting bias).

6-11
Conclusions 26 Provide a general interpretation of the results in the context of other evidence, and implications for future research.

Abstract
Objective: Inflammation has been implicated in the aetiology of mental illness. We conducted the first systematic review and meta-analysis of the association between peripheral markers of inflammation and generalised anxiety disorder (GAD).
Design: Systematic review and meta-analysis of studies measuring peripheral cytokine levels in people with GAD compared to controls. Eligibility criteria: Primary, quantitative research studies of people with a diagnosis of GAD assessed using a standardised clinical interview that measured peripheral inflammatory markers.
Data extraction and synthesis: Two independent reviewers extracted data and assessed study quality. Meta-analysis using a random-effects model was conducted for individual cytokines where data from three or more studies were available.
Results: 14/1718 identified studies met inclusion criteria, comprising 1188 patients with GAD and 10,623 controls. In total 16 cytokines were evaluated. Significantly raised levels of CRP, IFN-γ and TNF-α were reported in GAD patients compared to controls in two or more studies. Ten further pro-inflammatory cytokines were reported to be significantly raised in GAD in at least one study. However, 5/14 studies found no difference in levels of at least one cytokine. Only CRP studies reported sufficient data for meta-analysis. CRP was significantly higher in people with GAD compared to controls, with a small effect size (Cohen's d = 0.38, 0.06-0.69), comparable to that reported in schizophrenia. However, heterogeneity was high (I 2 = 75%), in keeping with meta-analyses of inflammation in other psychiatric conditions and reflecting differences in participant medication use, co-morbid depression and cytokine sampling methodology.
Conclusion: There is preliminary evidence to suggest an inflammatory response in GAD, but it remains unclear whether inflammatory cytokines play a role in aetiology. GAD remains a poorly studied area of neuroinflammation compared to other mental disorders and further longitudinal studies are required.

Strengths and limitations of this study
 This is the first study to conduct a comprehensive systematic review and metaanalysis of peripheral inflammatory markers in generalised anxiety disorder.  A wide range of databases were searched, and a large number of papers screened for inclusion in the study, 14 of which were subjected to quality assessment and detailed critical appraisal.  It was only possible to conduct a meta-analysis of C-reactive protein, and it was not possible to examine publication bias due to the limited number of studies identified for inclusion in the meta-analysis.  The high levels of heterogeneity across studies mean that findings should be interpreted with caution.

Introduction
There is growing evidence for immune mediated pathogenic mechanisms in several psychiatric disorders with discrete profiles of inflammatory mechanisms 1 . Epidemiological evidence has shown an increased risk of mood disorders and psychosis in people with a history of severe infection or autoimmune conditions 2,3 . This has been supported by genome wide association studies implicating multiple immune signalling pathways 4 , and altered profiles of pro-inflammatory cytokines and acute phase reactants in schizophrenia 5 , depression 6 , obsessive compulsive disorder (OCD) 7 and bipolar disorder 8 . However, the relationship between inflammation and mental illness remains poorly understood and controversial, with a number of proposed potential neuropathological mechanisms 9,10 , including changes in microglial function 1 , glutamatergic excitotoxicity 11 , synaptic plasticity 12 and reduced hippocampal neurogenesis 13 .
Despite increasing interest in the role of inflammation in mental illness, relatively little research has focused on potential associations with anxiety disorders 14 . These are common, with an estimated lifetime prevalence of 7.3% to 28.8%, are associated with substantial functional impairment and are estimated to cost between 42-47 billion dollars to the U.S economy each year 15,16 . However, only 60% of patients are thought to respond to pharmacological and psychological treatments and understanding of the underlying pathophysiological mechanisms of anxiety disorders remains poor 17 .
Generalised anxiety disorder (GAD) is the most common anxiety disorder, with a degree of associated disability equivalent to that of major depressive disorder (MDD) 18 . Despite psychopharmacological 19 and psychological 20 treatments showing effectiveness in GAD, 42% of people living with GAD experience ongoing symptoms after 12 years and half of remitted patients experience recurrence 18 . GAD is more prevalent in those with inflammatory conditions such as rheumatoid arthritis (RA) 21,22 , with case series studies suggesting symptoms are less common with immune modulating treatment targeting specific inflammatory cytokines 23 . The chronic clinical course and relatively high probability of recurrence in GAD, in addition to preliminary evidence of an inflammatory component in other anxiety disorders 7,24 , suggest that inflammation could be an important neurobiological mechanism in the aetiology of this disorder.
To date, two previous reviews of inflammatory biomarkers in GAD have been conducted. Of these, however, one was a narrative review 25 and the other was restricted to literature published within the last decade 14 , and with a focus on all anxiety disorders. Both reviews reported that there was preliminary evidence for inflammatory changes in GAD. However, only three studies were identified by the systematic review reporting cytokine changes in GAD and no meta-analysis was performed. No study to date has conducted a comprehensive systematic review and meta-analysis of all current literature focusing on GAD or commented on the longitudinal association between inflammation and GAD. We aimed to systematically review the cross-sectional and longitudinal associations between inflammatory biomarkers and GAD, and perform the first meta-analysis of inflammatory biomarkers in GAD.

Method
We conducted a systematic review of studies that had included people with GAD who had undergone peripheral cytokines measurement and a between group meta-analysis of cytokine levels in people with GAD compared to controls. We conducted the study according to Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines 26 .
We Our search terms (see appendix 1 for further details) were: (inflammat* or cytokine or interferon or IFN or interleukin or "translocator protein" or TSPO or "tumour necrosis factor" or "tumor necrosis factor" or TNF or IL-1 or IL-2 or IL-4 or IL-7 or IL-6 or IL-8 or IL-10 or migroglia or t-cell or lymphocyte or "C-reactive protein" or "C reactive protein" or CRP or "acute phase protein" or "fibrinogen") and ("generalised anxiety disorder" or "generalized anxiety disorder" or GAD or worry).
We included primary, quantitative research studies (including unpublished theses and dissertations), written in any language, that included people with a diagnosis of GAD assessed using standardised clinical interview (e.g. Structured Clinical Interview for DSM 27 ) or standardised psychometric instruments. Studies reported cross-sectional or longitudinal data in clinical or community populations. Cross-sectional studies measured inflammatory biomarker concentrations in anxious people versus non-anxious healthy controls, while longitudinal studies measured inflammatory biomarker concentrations at baseline and anxiety scores at follow-up. Inflammatory markers were measured in the unstimulated state (no antigen induced stimulation of cytokine production) and sampled from peripheral blood, CSF, or saliva at any time of day. Exclusion criteria included studies with less than 5 participants, studies in animals and studies where subjects were participants in the treatment arms of clinical trials.

Patient and public involvement
There was no patient or public involvement in the study.

Data extraction and quality assessment
Data were extracted and quality assessed for all studies that met eligibility criteria by two independent raters (HC, EA) with disagreements settled by consensus and discussion. For each cytokine, we extracted the means, variance estimates or 95% confidence intervals (CIs) and sample size for GAD and control groups. We also extracted demographic data (e.g. age, Risk of bias and study quality were evaluated using the Newcastle-Ottawa Quality Assessment Scale 28 . Other potential confounding factors (including assay type and sensitivity, inflammatory marker analysis and recruitment methods) were also examined to allow more detailed bias and quality analysis of studies.

Strategy for data synthesis
Separate meta-analyses were performed for individual biomarkers in GAD versus controls if sufficient data were available from a minimum of three studies. Due to different measurement methods and anticipated high heterogeneity, we estimated a standardised mean difference (SMD) for each inflammatory marker, and used a random effects model for metaanalysis, conducted using Revman 5. Heterogeneity across studies was quantified with the I 2 statistic, with a value of 25% typically regarded as low, 50% as medium, and 75% as high 29 . If studies were longitudinal or trials of interventions with multiple data collection points, we examined baseline data only to avoid skewed meta-analysis from inclusion of more than one effect size from the same study.

Systematic review
We identified 1718 papers, excluded 1598 of these by titles and abstracts, and retrieved the remaining 120 papers, of which 14 met eligibility criteria and were included in the final systematic review (see Figure 1). The primary reasons for rejection were that no diagnosis of GAD was recorded or no inflammatory marker was measured.
Characteristics of the 14 included studies are shown in Tables 1 and 2. Studies comprised a total of 1188 people with a diagnosis of GAD and 10,623 controls, with a further 116 participants from a study that did not report GAD and control group sizes.
Twelve studies (85.7%) reported the assay method used, all of which were versions of an enzyme-linked immunosorbent assay (ELISA) or enzyme immunoassay (EIA). However, only 7 studies (50%) reported assay sensitivity. All but one study used blood component samples to assess inflammatory marker levels, with the most common sample type being serum (N=6, 42.9%) and plasma (N=4, 28.6%).

Risk of bias and quality in individual studies
All included studies had adequate case definition, with participants meeting diagnostic criteria for GAD according to DSM or ICD, with 12 studies (85.7%) using a structured clinical interview for assessment (see Table 1).
Most (71.4%) studies included people aged 18-65, but two studies (14.3%) only included participants over the age of 50 and a further two studies (14.3%) used adolescent participant cohorts. The majority (78.6%) of studies accounted for age and sex differences in their analyses. Only 8/14 studies (57.1%) recorded participants' body mass index (BMI) which is known to correlate with inflammation, and only half of these accounted for BMI differences in analysis of group differences 30 (see Table 2).
Use of psychotropic medication and presence of comorbid major depressive disorder are important moderators of inflammation in other psychiatric disorders 24 . Six studies (42.8%) excluded patients who used psychiatric or other immune-modulating medication, though only two studies (4.3%) reported medication use. The majority of studies (64.2%) either excluded patients with co-morbid MDD or adjusted for this in analyses.
Concurrent physical illness is clearly an important determinant of inflammatory cytokine levels and this was accounted for by the majority of included studies by either excluding participants with co-morbidities (5 studies, 35.7%) or adjusting for chronic physical illness in group comparisons (6 studies, 42.8%), though two studies specifically only included participants with co-morbid cardiovascular disease. Use of a pre-determined cut-off value for cytokine levels was employed by three studies (21.4%) to ensure that cases with acute infection were excluded from the sample.
Many inflammatory markers exhibit a diurnal pattern of expression and are affected by consumption of food, thus time of day of sampling and whether the sample was taken in a fasted state are important factors to consider in analysing relative levels of cytokines 31 . However, time of day of sampling was only recorded in a minority of studies (6 studies, 42.8%), and the same number of studies recorded whether fasted samples were taken.
The overall quality of studies included in the review varied significantly, with Newcastle-Ottawa scale scores ranging from 2 to 9 (see Table 3). The area in which most studies were inadequate was in reporting non-response rate and detailing recruitment methods (see Table  3). Lowest quality studies were abstracts or dissertations, and two studies lacked control groups as only GAD patients were sampled (see Table4).

C-reactive protein
CRP is a critical early pro-inflammatory surveillance molecule involved in the activation of the complement system and both innate and adaptive immune systems 32 . We identified nine studies that investigated the association between GAD and CRP, comprising a total of 11,486 participants (see Table 5).
Four studies, involving 578 GAD patients and 4046 controls, provided sufficient information to conduct a meta-analysis of CRP levels in GAD [33][34][35][36] (see Figure 2). This was the only inflammatory marker for which meta-analysis was possible. Meta-analysis showed significantly raised CRP in GAD compared to controls (SMD 0.38, 95% CI 0.06-0.69; Z=2.36, p=0.02). However, there was a large and statistically significant degree of heterogeneity between studies (Chi2=12.0; df=3; p=0.007; I 2 = 75%). Given the high heterogeneity and inclusion of less than 10 studies in the meta-analysis we did not have sufficient power to examine publication bias 37 . Two out of four studies were high quality, scoring 9 on the Newcastle-Ottawa scale, and examined large sample sizes 33,36 (see Table 4). However, in each of the four meta-analysed studies different assay methods were used, and sampling methods varied significantly (see Table 2). The lowest quality study to be included in the meta-analysis did not report mental health co-morbidities and recruited participants from an inpatient setting 34 . There was also a wide range in age of participants included in the four studies, with the largest study examining CRP levels in adolescents, that would likely contribute to high heterogeneity.
Five studies [33][34][35][36]38 (n=4669), one of which was conducted in 16 year olds 33 and two in participants with co-morbid heart disease 38 , reported significantly higher CRP levels in participants with a diagnosis of GAD. The largest study 39 (n=5810) examining CRP in GAD examined CRP levels in children from baseline measurement aged 9-16 years to follow up aged 19-21.This was the only study to examine the longitudinal association between GAD and CRP, and found a bivariate association both cross-sectionally and over time between GAD and elevated CRP , however, this was accounted for by potential co-variates including BMI and medication use. The only study 40 to find an inverse correlation between CRP and GAD was conducted in non-smoking women from a study in Finland and did not specify the numbers of participants with a diagnosis of GAD or group differences.
No difference was found in a cohort study (n=821) that used a combined inflammatory index consisting of CRP, IL-6 and TNF-α in 93 patients with a diagnosis of GAD and controls with a history of cardiovascular disease (CVD) 41 . Subgroup analysis examining differences in individual inflammatory markers was not reported 41 .
We found two studies 35,42 (n= 196) that examined the association of severity of GAD symptoms with CRP level. One found a significant positive correlation between CRP level and GAD-7 scores 35 , and the other reporting CRP differences in 70 GAD patients with and without a diagnosis of alexithymia found a significant association between higher CRP and suicidal ideation 42 .
Although the meta-analysis and the majority of included studies reported raised CRP in GAD, there was wide variation in reporting and adjustment for important potential moderators, including co-morbid MDD, use of medications, assay used and time of day of blood collection, all of which likely contributed to the high degree of heterogeneity between studies. Of the nine studies to analyse CRP, four (44.4%) did not exclude or adjust for medication use by participants 33,38,40,41 . Co-morbid MDD was not adjusted for in analysis by two studies 39,41 , one of which was included in the meta-analysis 39 . Only three of the nine studies reported time of sample collection 35,36,42 or whether this was in a fasted state 36,41,42 ,  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   8 and though all studies utilised a similar assay method, different assay types were used in every study.
In summary, of the nine studies to have examined differences between GAD and controls, the majority reported raised CRP in GAD and meta-analysis found significantly raised CRP in GAD with a small effect size. However, there was wide variation in study methods including variable adjustment for mediators of inflammation such as co-morbid MDD, medication use and sampling methods. Only one study examined CRP in GAD longitudinally, reporting a bivariate association accounted for by health seeking behaviours 39 .

Interleukins
Seven studies examined the association between interleukins and GAD (see Table 3). IL-6 is a mediator of T-cell and B-cell activation and induces acute phase proteins in hepatocytes, among other functions 32 . Pharmacological blockade of IL-6 action is used to treat several auto-immune conditions including RA, and raised IL-6 has been associated with a number of psychiatric conditions including depression, schizophrenia and PTSD 32,43 . We found IL-6 was the most frequently measured interleukin, with five studies (n=2066) examining changes in GAD patients compared to controls 35,36,41,44,45 . The largest study investigated differences between 454 participants with a diagnosis of GAD and 556 controls from The Netherlands Study of Depression and Anxiety (NESDA) cohort 36 . Though analysis was conducted on anxiety disorders as a whole, mean difference in IL-6 in people with GAD compared to controls obtained through direct communication with the author showed significantly higher levels in GAD. However, it is unclear whether these differences remain significant after adjustment for group differences and no associations were found between IL-6 and participants who had all types of anxiety disorder 36 . Two studies 35,45 (n=165), one of which used saliva samples 45 , reported significantly higher IL-6 in medication naïve participants with a diagnosis of GAD compared to age and sex matched healthy controls.
No difference was found in a combined inflammatory index consisting of CRP, IL-6 and TNF-α in a study of 93 patients with GAD and co-morbid ischaemic heart disease 41 . One case-controlled study 35 of 48 Chinese outpatients presenting for the first time with a diagnosis of GAD and 48 age, sex and education matched controls accounted for all results for IL-1α, 5, 8, 12p70. This study found significantly higher levels of IL-1α, -8 and -12p70 in GAD patients, in addition to higher levels of IL-1α and IL-8 with increased severity of GAD (as measured by GAD-7 scale), but did not account for chronic physical co-morbidities during recruitment or in analysis. Both IL-1α and IL-8 have pro-inflammatory functions as chemoattractants for leukocytes and haematopoiesis, and have been targeted for treatments in a number of auto-immune conditions 32 . However, there was no association between GAD and IL-5, which is thought to predominantly mediate myeloid cell activation, and is a target of treatment in asthma 32 . The same study 35 also examined IL-2, which has a major role in T-cell mediated autoimmune and inflammatory conditions 32 . Results showed significantly higher IL-2 in GAD patients, however, this conflicted with results from a smaller study (n=24) that found no significant difference between medication naïve GAD patients and controls, though  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   9 few details of the characteristics of participants, sampling or analysis were reported in this abstract 46 .
One study that measured IL-1 using sputum analysis found significantly higher levels in GAD compared to controls in 69 participants recruited from the same Chinese hospital 45 . Though IL-1 is pro-inflammatory, there are differences in function dependent on the class of IL-1 protein measured which was not reported in this study 45 .
IL-4 has several pro-inflammatory functions including immunoglobulin-E (Ig-E) class switching, expression of MHC class II and acts as a survival factor for T and B cells 32 . The only study to measure IL-4, found no differences between 54 GAD patients recruited from community mental health teams and primary care after controlling for age, sex, BMI, smoking, alcohol consumption and co-morbid depression 47 . This study 47 also investigated IL-10, which was the only cytokine with an anti-inflammatory function to be measured and is involved in immunosuppression of T cell subsets and B cell immunoglobulin production. This found significantly lower levels of IL-10 (OR 0.35 p = 0.003) in GAD patients. However, this opposed findings from a smaller study (n=24) which reported significantly higher levels of IL-10 in GAD patients compared to controls, though it was not reported whether this association remained significant after controlling for group differences 46 .
In summary, IL-6 was the most commonly measured interleukin raised in GAD compared to controls in the majority of studies, however, no study examined the longitudinal association with GAD. Other interleukins were examined by relatively few studies that examined small numbers cross-sectionally with mixed findings.

IFN-γ
IFN-γ has anti-viral roles including promoting cytotoxic activity, MHC class I and II upregulation, NK cell activation, and is a treatment target in inflammatory conditions such as Crohn's disease 32 . Three studies investigated IFN-γ levels in GAD (n= 330) 35,40,47 . The largest study (n=118) found higher IFN-γ in GAD patients from the UK that remained significant after adjustment for age, gender, BMI, smoking, alcohol and co-morbid depression, but did not adjust for anxiolytic medication use in analysis 47 . This finding was supported by a study of 96 participants which reported higher IFN-γ levels in GAD and a significant positive correlation between anxiety severity and IFN-γ 35 . Conflicting findings were reported by a Finnish study of 116 participants, which found significantly lower IFN-γ in GAD patients. However, the number of participants with a diagnosis of GAD, differences between groups and adjustment for potential confounders were not reported 40 . In summary, only a few small cross-sectional studies have examined differences in IFN-γ between GAD and control groups, and their findings were mixed.

TNF-α
TNF-α has a wide array of roles in host defence, including initiating a strong acute inflammatory response but limiting duration of inflammatory activation, and is the target of blocking monoclonal antibodies in the treatment of a wide array of autoimmune conditions  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   10 including Crohn's disease and RA 32 . Six studies (n=2300) investigated TNF-α in GAD, with mixed findings. Three studies (n=303) found TNF-α significantly raised in GAD patients compared to controls 40,45,47 . However, the largest study to measure TNF-α (n=1010) found no difference between participants with GAD and controls, and no correlation between TNF-α and anxiety symptoms 36 . This finding was supported by a study of 93 patients with GAD and co-morbid ischaemic heart disease using a combined inflammatory index of CRP, IL-6 and TNF-α which reported no difference compared to controls 41 . In summary, though the majority of studies to measure differences in TNF-α between GAD and controls reported significantly raised levels, these comprised small cross-sectional studies and the largest study reported no difference.

Other cytokines
One study compared levels of the pro-inflammatory cytokines CCL-5, MCP-1 and SDF-1 in 120 medication naïve physically well patients with a diagnosis of GAD and co-morbid personality disorder to 40 controls 48 . Significantly higher levels of MCP-1 and SDF-1 were reported in both men and women, and higher CCL-5 in men but not women with a diagnosis of GAD compared to controls 48 .

Discussion
To our knowledge, this is the first systematic review and meta-analysis focusing on inflammatory cytokines in GAD. Using a range of databases we identified 14 studies, comprising 1188 participants with GAD and which measured 16 cytokines. We found significantly raised levels of CRP, IFN-γ and TNF-α in people with GAD compared to controls which were findings replicated in two or more studies. A further 10 proinflammatory cytokines were reported to be significantly raised in GAD in at least one study, however, 6/14 studies found no difference in at least one cytokine.
Despite substantial efforts to acquire data by contacting authors, it was only possible to conduct a meta-analysis of CRP. This identified significantly higher levels in GAD compared to controls with a small effect size (SMD: 0.38), though there was evidence of significant heterogeneity across studies (I 2 = 75%). This effect size in CRP is greater than has been reported in other anxiety disorders (PTSD: SMD = -0.14) 24 or MDD (SMD= 0.14) 6 , and is similar to that reported in schizophrenia (SMD= 0.45) 49 . However, the effect size of our meta-analysis was driven by findings in poorer quality studies with small sample sizes. The two higher quality, larger studies reported a smaller effect size and no significant difference between groups respectively. As a result, further high quality studies are required to confirm our findings of raised CRP in GAD.
Though we were only able to meta-analyse CRP, meta-analyses of different cytokines in other anxiety disorders have been conducted with larger effect sizes. A meta-analysis of inflammatory markers in PTSD identified 20 studies which reported increased interleukin 6 (IL-6), interleukin 1β (IL-1β), tumour necrosis factor-α (TNFα), and interferon γ (IFN γ) levels with effect sizes ranging from small (IFN γ: SMD 0. 49 24 . However, a systematic review and meta-analysis of pro-inflammatory cytokines in OCD identified 12 studies, and concluded that there was a significant reduction in IL-1β with moderate effect size (SMD: -0.60, p<0.001), and only IL-6 levels were significantly increased after subgroup analysis in medication-free adults with OCD 7 . It is unclear whether this profile of inflammatory marker changes would follow a similar pattern in GAD if future studies enabled further meta-analysis.
In light of the high heterogeneity among studies, low participant numbers and inconsistent reporting and adjustment for known confounding factors such as BMI, smoking, medication use and co-morbidities our findings should be interpreted with caution. It was not possible to analyse the cause of the degree of heterogeneity due to the paucity of studies. Other known mediators of inflammation 24 such as physical activity, raised blood pressure and genetics were not accounted for. Furthermore, reporting of GAD severity and duration of symptoms was generally poor, preventing detailed analysis of whether inflammatory markers predicted outcomes and quality of life. We also found limitations in inclusion of specific demographics of participants with GAD. For example, despite GAD in older adults being prevalent and often treatment resistant 50,51,52 , only two studies included participants over the age of 65, both of which only included patients with co-morbid ischaemic heart disease.
We are beginning to understand the interplay between cytokines, the immune system and mental health 1,53 . At a molecular level we are aware that pro-inflammatory cytokines, including IFN, IL-1 and TNF, can reduce the availability of monoamines by inducing expression of pre-synaptic reuptake pumps and inhibiting enzymes involved in monoamine synthesis 54 , linking the monoamine theory of anxiety with inflammatory mechanisms. There is also a growing understanding of the relationship between systemic inflammation and the central nervous system (CNS) 1,55 . Microglial activation has been shown to be mediated by peripheral cytokines, and increased activation has been found in post-mortem studies of patients with MDD and schizophrenia 1 . No study we identified correlated inflammatory marker changes with in vivo microglial activation imaging in GAD and to our knowledge no research on post-mortem microglial changes in GAD has been conducted. Increased neuronal activity has also been shown to induce inflammatory and vascular changes in the brain, suggesting that psychological stress can not only be induced by inflammation but perpetuate chronic low grade inflammation seen in other vascular and neurodegenerative disorders 55 .
Understanding interactions between the CNS and immune system, and identifying biomarkers of GAD offers potential for novel therapeutic approaches. The revolution of development of monoclonal-antibody therapies for inflammatory disorders 56 raises the possibility of repurposing these medications for trials in treatment resistant GAD if specific and consistent profiles of inflammatory biomarkers are identified.
However, it remains unclear as to whether inflammation plays a causal role in GAD 43,54 . For example, although IL-6 is a successful target for treatment in a number of auto-immune conditions and raised IL-6 is implicated in several psychiatric disorders, it also acts to reduce other pro-inflammatory cytokines such as TNF via negative feedback and is induced by physical exercise, hyperthermia, fasting, sleep deprivation and sunlight exposure without activation other pro-inflammatory cytokines 43 . This raises the question as to whether inflammation in GAD is a consequence rather than cause of symptoms. This will only be answered by large prospective longitudinal studies, better characterising the relationship between inflammation and GAD. However, remarkably our review identified only one longitudinal study of inflammation in GAD patients that examined a cohort of adolescents until the age of 21 and only investigated CRP.
Recent studies using Mendelian randomisation in depression have suggested that cytokines such as IL-6 are causal risk factors for depression 57 , and trials of immunotherapy in psychosis are already underway 58 . Our study suggests that GAD is an important candidate for future similar future studies exploring causality of inflammation and potentially novel drug trials.

Conclusion
There is some preliminary evidence to suggest a raised inflammatory response in GAD, although it is unclear whether inflammatory cytokines play a role in aetiology. GAD remains a poorly studied area of psychiatric neuroinflammatory research compared to other mental illnesses such as MDD and schizophrenia. While we are a long way from using inflammatory cytokines as a biomarker or treatment target in GAD, current findings reflect inflammatory changes seen in other mental illnesses and highlight the importance of ongoing investigation of the role inflammation plays in the development and course of GAD. Further, methodologically consistent, prospective, longitudinal studies examining the mechanisms and relationship between inflammation and GAD, while accounting for known mediators of cytokine production, are required.
Acknowledgements/ Funding Sources HC and EA are NIHR Academic Clinical Fellows (ACF). This research was supported by the NIHR Biomedical Research Centre at University College London/University College Hospital London.

Declaration of interests
All authors declare no competing interests.

Contributors
RH, RG and HC were involved in initial design of the research. EA and HC performed data extraction of included studies and quality analysis. HC wrote the initial draft of the manuscript and did the statistical analysis with supervision from RG and RH. RH, RG and HC participated in the critical revision of the Article and all authors approved the final Article.

2-3
Objectives 4 Provide an explicit statement of questions being addressed with reference to participants, interventions, comparisons, outcomes, and study design (PICOS).

3-4 METHODS
Protocol and registration 5 Indicate if a review protocol exists, if and where it can be accessed (e.g., Web address), and, if available, provide registration information including registration number.

3-4
Eligibility criteria 6 Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered, language, publication status) used as criteria for eligibility, giving rationale.

4
Information sources 7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and date last searched. 4 Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated.

4
Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable, included in the meta-analysis).

4
Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators.

4
Data items 11 List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made.

Risk of bias in individual studies
12 Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis.

Study selection
17 Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally with a flow diagram.

5
Study characteristics 18 For each study, present characteristics for which data were extracted (e.g., study size, PICOS, follow-up period) and provide the citations.

5
Risk of bias within studies 19 Present data on risk of bias of each study and, if available, any outcome level assessment (see item 12).

5-6
Results of individual studies 20 For all outcomes considered (benefits or harms), present, for each study: (a) simple summary data for each intervention group (b) effect estimates and confidence intervals, ideally with a forest plot.

DISCUSSION
Summary of evidence 24 Summarize the main findings including the strength of evidence for each main outcome; consider their relevance to key groups (e.g., healthcare providers, users, and policy makers).

6-11
Limitations 25 Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified research, reporting bias).

6-11
Conclusions 26 Provide a general interpretation of the results in the context of other evidence, and implications for future research.