Introduction

Dementia is a major cause of morbidity and mortality.1 Neurodegenerative dementias, including Alzheimer’s disease (AD), dementia with Lewy Bodies (DLB), Parkinson’s disease dementia (PDD) and frontotemporal dementia, are all associated with abnormal protein folding, protein aggregation and neuronal loss. Their etiology is complex and poorly understood, but increasing evidence suggests that inflammation may represent a common pathway contributing to pathological protein aggregation and neuronal loss across different neurodegenerative dementias.2,3

AD is defined by the presence of amyloid plaques and neurofibrillary tangles and is the most common cause of dementia.4 It is a clinical syndrome of progressive cognitive decline,5 usually beginning with a prodromal Mild Cognitive Impairment (MCI) stage.6 Neuroinflammation is well recognized in AD. In response to amyloid-β plaques, microglia secrete inflammatory cytokines7,8 and upregulate interferon response genes.9 Imaging microglia in vivo with positron emission tomography (PET) of the translocator protein (TSPO) receptor has identified increased ligand binding, interpreted as microglial activation, in AD.10 TSPO binding correlates with amyloid and tau pathology, as measured by 18-[F]-flumetamol and 18-[F]-AV1451 PET imaging.11 TSPO binding has also been associated with tau spreading12 and subsequent cognitive decline.13

DLB, the second most common cause of dementia,14 is characterized by visual hallucinations, parkinsonism, rapid eye movement (REM) sleep behavior disorder and cognitive fluctuations.15 α-synuclein, a hallmark of DLB pathology,16 induces an immune response in vitro and in animal models.17 There is evidence for cortical recruitment of T-lymphocytes,18 higher levels of peripheral cytokines including interleukin (IL)-6 and tumor necrosis factor (TNF)-α,19 and PET imaging of microglial activation in early DLB.20

FTD presents with behavioral-executive and language (primary progressive aphasia, including semantic variant and non-fluent) variants. Peripheral and central inflammation are well recognized across the FTD spectrum, with alterations in serum cytokines, CSF cytokines, and T-cell phenotypes in patients compared to controls.21

Neuropsychiatric symptoms (NPS) are prevalent in dementia and associated with carer stress and poor outcomes. Common NPS include depression, apathy, anxiety, delusions, hallucinations, and agitation among others.22 They often precede dementia diagnosis as they can occur in the prodromal stages of AD23 and DLB.24–26 NPS are associated with patient outcomes. This varies by neurodegenerative disease and symptom domain. For example, in longitudinal studies of Alzheimer’s dementia psychosis, agitation, and depression were associated with cognitive progression and increased dependence.27 Psychosis, affective symptoms, and agitation/aggression were associated with earlier death.28 A recent meta-analysis in PD has found that psychosis and depression are associated with cognition and disease progression.29 In DLB, higher scores on a composite NPS scale were associated with caregiver burden and depression.4 In AD, total neuropsychiatric symptoms are more closely associated with caregiver distress than cognition,30 with studies also reporting a bidirectional relationship between NPS and distress.31 Apathy in gene carriers at the presymptomatic stage of FTD predicts cognitive decline.32 Across dementia subtypes, a factor including a range of BPSD was associated with hospital admission.33 NPS are predictive of nursing home placement in AD,34 specifically with depression, delusions, and agitation increasing the risk of institutionalization across cognitive disorders.35 In PD, hallucinations are a predictor of future nursing home placement.36

NPS have been associated with specific imaging,37 neurotransmitter,38–40 and pathological changes,41,42 but there is a need for a greater understanding of the mechanisms of these symptoms to develop effective therapies,43 especially in dementias apart from AD where the evidence base in more limited.44,45 Current treatments primarily target neurotransmitters and are of limited effectiveness46 or associated with significant risks.47,48

NPS are common in inflammatory diseases.49 There is evidence that anti-inflammatory or immunomodulatory treatments may play a role in relieving depression comorbid autoimmune conditions.50 There is increasing evidence that inflammation plays a causal role in the development51 and symptoms52 of psychiatric disorders, and trials of anti-inflammatory of immune modulatory treatments are ongoing in depression.53 Several mechanistic pathways have been proposed for the relationship between inflammation and NPS. In depression, chronic stress has been linked to the release of pro-inflammatory cytokines, which has been linked to increased glutamate in magnetic resonance imaging (MRS) studies. This leads to decreased activity in corticostriatal reward networks and symptoms such as anhedonia.54 Cytokines also regulate indoleamine 2.3-dioxygenase (IDO), which metabolizes tryptophan to kynurenine, leading to quinolinic acid production, an N-methyl-D-aspartate (NMDA) receptor agonist, with this pathway linked to depression in animal models and clinical studies.55 In NMDA receptor (NMDAR)-antibody encephalitis, antibody mediated internalisation of the NMDAR leads to reduced signals in inhibitory neurons, which has been proposed as a mechanism for the psychotic symptoms.56

There are several potential mechanisms linking Alzheimer’s disease and the immune response. Firstly, as a core part of the disease pathology – microglial genes represent a large proportion of the risk for AD,57 specific microglial response to amyloid-beta and tau pathology associated with cytokine release, and microglial activation in vivo as measured by PET imaging of the translocator protein (TSPO) linked to amyloid-β58 and tau.59 This central inflammation could lead to peripheral inflammation via the release of cytokines that cross the blood–brain barrier, however there is a paucity of studies with paired measures of central and peripheral inflammation.58,60

Secondly, there is evidence that chronic diseases, such as diabetes61 and cardiovascular disease,62 are associated with risk of AD, or more rapid progression. One proposed mechanism is that the chronic low-grade inflammation associated with chronic disease or acute inflammatory events such as sepsis influence central immune responses and increase the risk of neurodegeneration.63

Mid and late life depression are risk factors for dementia.64 The mechanisms of this association are not fully understood. It could represent prodromal symptoms of neurodegenerative disease, or alternatively the chronic inflammation associated with depression could increase dementia risk. In support of this, individuals with MCI and a history of depression treated with serotonin selective reuptake inhibitors (SSRIs) showed reduced rate of progression to AD,65 and meta-analysis has shown that treatment with SSRIs is associated with a reduction in proinflammatory cytokines.66

Whether the inflammatory response is generated as part of the underlying neurodegenerative process, or secondary to other chronic disease or associated with depression, if this response is causally associated with the development of NPS, this could represent a potential therapeutic target for symptomatic relief.

Therefore, a greater understanding of the biological mechanisms underpinning NPS and the associations with inflammation is needed to develop more effective therapies. Identifying immune biomarkers associated with NPS could guide treatment trials, including repurposing of existing immunomodulatory medications and allowing stratification of clinical trials, to identify which patients are most likely to benefit from these drugs.67 Previous reviews have specifically looked at cytokines with behavioral and psychological symptoms in dementia (BPSD) in dementia,68 apathy in AD,69 and depression in Parkinson’s disease.70 We undertook a systematic review exploring peripheral and central measures of inflammation across different neurodegenerative disorders.

As there is evidence that common pathways such as protein aggregation and neuronal death are associated with common immune pathways, such as toll-like receptor signalling3 and microglial activation2 across neurodegenerative diseases. There is also evidence for distinct immune responses that depend on specific protein pathologies – such as microglial states depending on amyloid-β or tau pathology,71 and the association of CD4 cells with α-synuclein.72 For this reason, we were broad in our inclusion of neurodegenerative dementias, including AD, LBD, FTD, PSP, and their prodromes but excluding non-neurodegenerative causes, which lack these shared features.73

In order to capture the full extent of the literature, we were broad in our definition of an inflammatory marker: we aimed to identify studies that measured a biomarker74 that was related to an immune or inflammatory process. This will include cytokines, chemokines, immune cell subsets, and PET imaging of microglia. As there is evidence for both peripheral and central inflammation in dementia,63 our intention is to include markers measured in both compartments. We have also been broad in our definition of NPS as symptoms associated with changes in perception, thought, mood and/or behaviour,75 including psychosis, agitation, apathy, depression, and sleep disturbance.76

Methods

We searched MEDLINE, Embase, and PsycINFO databases through OVID on 25/07/23 with the terms:

((microglia) OR (TSPO) OR (interleukin) OR (“translocator protein”) OR (“tumour necrosis factor”) OR (TNF) OR (“c-reactive protein”) OR (CRP) OR (cytokine)) AND ((Dementia) OR (“Alzheimer’s disease”) OR (Lewy) OR (“Parkinson’s disease”) OR (“mild cognitive impairment”) OR (MCI) OR (“progressive supranuclear palsy”)) AND ((depression) OR (psychosis) OR (apathy) OR (anxiety) OR (hallucination) OR (delusions) OR (neuropsychiatric) OR (agitation))

We included no publication date restrictions but limited the search to articles in the English language. Therefore, the timeframe of the search was as follows: 1966–2023 for MEDLINE, 1947–2023 for EMBASE and 1967–2023 for PsychINFO. We included additional articles identified from the citations or those known to the authors. Our inclusion criteria were as follows: Firstly, studies of a neurodegenerative dementia (including but not limited to Alzheimer’s disease, Lewy body dementias, frontotemporal dementia or progressive supranuclear palsy) or their prodromal stages (such as mild cognitive impairment due to Alzheimer’s disease or mild cognitive impairment due to Lewy Bodies) or Parkinson’s disease. Secondly, an outcome measure of peripheral or central inflammation (such as inflammatory markers in plasma, serum, or cerebrospinal fluid (CSF), immune cell subsets, or PET imaging of immune markers). Thirdly, an outcome measure of neuropsychiatric symptoms, and finally a reported test for association or relationship between the inflammatory markers and neuropsychiatric symptoms.

Our exclusion criteria were case reports (or small cohorts with less than five participants), fundamental science research, animal studies, or non-neurodegenerative dementia. Our primary outcome was the association between an inflammatory or immune marker and neuropsychiatric symptoms in neurodegenerative dementia groups. Duplicates were removed in Endnote, and the search results were then imported into Rayyan. Articles were then screened by title and abstract, then for potentially suitable papers full text of articles was obtained. These were further screened against the inclusion and exclusion criteria. Reasons for exclusion were recorded and reported in Figure 1. From included articles we extracted from the full text into Excel the study design, setting, diagnosis (and diagnostic criteria), number of participants, demographics, inflammatory marker measured, assay source and details, psychiatric symptom measurement and by which scale, and the overall findings of the study. We did not exclude studies based on formal quality criteria (beyond the lower limit of number of participants). Our aim was to be inclusive, and extract data that allowed the quality of studies to be compared qualitatively, including the diagnostic criteria used (case definition), the source of recruitment, inclusion of controls, the specific assays used for biomarker measurement and scales used to measure NPS.

Figure 1 Flow diagram describing the systematic review process.

Notes: **Publication type included abstracts or conference proceedings with insufficient details, book chapters, editorials/letters without new data, or protocol papers. Study design describes reviews with no original data, case reports, animal studies, or fundamental science papers. Study population describes studies with no neurodegenerative disease group. Study outcome describes studies with no psychiatric outcome, no inflammatory measure, or no testing for association between psychiatric and inflammatory measures. This flow diagram is based on PRISMA reporting guidelines.162

Results

Our search identified 3208 articles after duplicates were removed. A total of 3074 articles were excluded by title and abstract, with 135 articles sought for full-text retrieval. These were supplemented with 5 articles known to the authors or identified from the citations. Ninety-nine articles were included in the final review (Figure 1). Many excluded studies recorded an inflammatory measure and psychiatric outcome but did not report testing for an association.

Figure 2 Graphical summary of number of studies by diagnostic group.

Notes: Bar chart showing the number of studies by diagnosis. Different bars represent diagnostic groups and are identified with the color key on the right. The height of the bar represents the number of studies identified by the systematic review studying each diagnostic group.

Firstly, we will describe four key themes that emerged across studies, regardless of diagnosis, inflammatory marker, or neuropsychiatric symptom. These were the number and selection of participants, the study design, the choice of outcome measure, and the quality of the assay.

Secondly, we will report the results from each neurodegenerative diagnosis in order from most to least studied, broken down firstly by neuropsychiatric outcome, and within each outcome, assay type.

Participants

Most articles included participants with either AD (including mild cognitive impairment, n = 58) or Parkinson’s disease (n = 35). Very few articles included participants with DLB (n = 3), FTD (n = 2) or Huntington’s disease (n = 2). One study78 did not specify dementia subtype, and one study included a mixed group.79 Comparator groups included cognitively unimpaired controls, major depressive disorder (MDD), delirium, subjective cognitive impairment (SCI), vascular dementia (VD), or other neurodegenerative diseases. Whilst studies frequently included unspecified MCI or MCI due to AD (MCI-AD), there was only one study that included MCI-LB (alongside MCI-AD).80 A wide range of diagnostic criteria were used. Ten studies did not describe standardized diagnostic criteria. A graphical summary of the distribution of studies by diagnostic group is given in Figure 2. The overall study sizes ranged from 10 participants to 3073, with several large multicenter research cohorts with paired blood and CSF.81–83 However, many studies were small including less than 20 participants per group.84,85 Full descriptions of the diagnostic criteria provided in each study are summarized in Supplementary Table 1.

Study Design

Most studies were cross-sectional (n = 72), with many fewer (n = 13) including longitudinal follow-up. These studies primarily focused on baseline biomarkers on disease progression rather than measuring biomarkers at multiple timepoints.86,87 Studies were focused on outpatients, with five studies including participants as inpatients (including nursing homes).

Outcome

The most common symptom domains studied were depression and anxiety (PD, n = 19, AD/MCI-AD, n = 26, DLB/MCI-LB, n = 1). Less frequently studied were psychosis, apathy, agitation, and sleep disturbance. Many studies used a composite measure of NPS (AD/MCI-AD, n = 20, DLB/MCI-LB n = 3) with fewer studies separating the symptom domains.88 A graphical summary of the number of studies by neuropsychiatric symptom outcome is given in Figure 3. Controlling for disease severity was variable. Those with NPS often had lower cognitive scores,89,90 or studies identified groups with increased inflammation, more psychiatric symptoms, and more cognitive or motor symptoms.91,92

Figure 3 Graphical summary showing the proportions of studies by neuropsychiatric symptom outcome measure.

Notes: Tree plot showing the proportions of studies exploring each neuropsychiatric symptom. The size of the area within the grid represents the number of studies, and each colored area represents a different neuropsychiatric symptom, as described by the key on the right. Composite symptoms represent those using an outcome with multiple symptom domains summed together, like the total NPI score.

Assays

Most studies used blood biomarkers of inflammation (n = 58), with fewer having paired CSF (n = 9), or CSF alone (n = 8). Studies on blood were primarily performed on serum and plasma, with some using peripheral blood mononuclear cells (PBMCs). Less commonly used were genetic, epigenetic or RNA sequencing. There were only two imaging studies, both in AD.93,94 Both used PET imaging of the TSPO receptor (one with 11[C]-DPA-713 and one with 11-[C]-PBR28). A summary of number of studies by assay source and sample size is given in Figure 4. Some studies provided detailed descriptions of the assay used, and the parameters around sample collection,95 whilst others provided minimal details.96 Individual studies mostly measured single or small numbers of select inflammatory markers using separate ELISAs, with multiplex methods becoming more common over time (Figure 5).97 Major immune biomarkers like C-reactive protein (CRP), IL1-β, IL-6 and TNF-α were frequently assayed. Across studies there was a spread of biomarkers assessed, with a range of in-house and commercial assays, and limited consistency between studies. These include, but are not limited to, interleukins (such as IL-2, IL-4, IL-6, IL-8, IL-10, and IL-12), interferons (such as IFN-γ), immune cell subsets as measured by flow cytometry (including CD3, CD4, CD8 subsets, and dendritic cell subsets) either stimulated or unstimulated, and mRNA expression from whole blood. There were several studies that looked at immune gene single nucleotide polymorphisms (SNPs) in association with psychiatric symptoms98–107 in small cohorts. Full descriptions of the assays are provided in Supplementary Table 1.

Figure 4 Bar chart of number of participants in each study.

Notes: Bar chart showing the number of participants in each study, colored by the assay source. The x-axis displays the number of studies, and each bar on the y-axis represents the number of participants in each study. The assay source is shown by the color key on the right.

Figure 5 Bar chart showing the change in assay type for studies measuring fluid inflammatory markers.

Notes: Each bar represents an epoch of years, and the y-axis shows the number of studies in each year measuring fluid inflammatory markers, split into those using single ELISAs (purple) and those using multiplex assays (yellow).

Alzheimer’s Disease and MCI-AD

Studies in AD predominantly examined the relationship between inflammation and depression (n = 26), followed by those using composite scores (n = 20). Few studies looked at other individual symptom domains (agitation, apathy, psychosis, and sleep disturbance). All studies looked at sporadic AD or did not specify whether genetic cases were included. Full descriptions of the studies are included in Supplementary Table 2.

AD-Depression Neuropathology

Two studies used brain tissue. The first108 focused on neutrophil gelatinase-associated lipocalin (NGAL), an innate immune protein associated with Alzheimer’s pathology and depression. They identified increased NGAL levels in the hippocampus of AD patients with depression. This relationship varied by region, with areas of the cingulate having higher NGAL levels in those without depression. This study combined neuropathology with serum and CSF. There were no differences in serum levels of NGAL, but a significant decrease in CSF NGAL. A second study109 used a range of microglial, cell adhesion, and cytokine markers and identified IL-4 as significantly increased in AD-depression compared to AD cases without depression.

Paired CSF and Blood

A large study81 examined a panel of immune biomarkers in CSF, plasma and serum focusing on the link between inflammation and the metabolism of tryptophan via the kynurenine (Kyn) pathway, which has been linked to depression.110 This study identified a higher Kyn/5-HT ratio associated with peripheral immune markers, and this ratio was associated with increased depression and anxiety scores. This study identified a possible shared mechanism between primary depression and depression in AD.

CSF

Total depression scores were not associated with CSF IL-6 or brain-derived neurotrophic factor (BDNF), but there were associations with individual symptom domains (psychomotor slowing and nighttime disturbance).111 Although a single exploratory study, it is of interest as inflammation-driven reductions in BDNF have been implicated in depression112 and psychomotor slowing.113 One study found a micro-RNA (miR-451a) that regulates toll-like receptor (TLR) 4 was lower in AD with depression and correlated with depressive symptom scores.114

Serum and Plasma

There are three studies that measured IL-6 in AD with depression and found an association.89,115,116 This is not without caveats; AD with depression groups often had lower cognitive scores,89,116 or findings had wide confidence intervals and limited details of assays.115 One large study in AD did not find any associations with IL-6 and depression scores,117 and another in MCI118 does not does not report similar findings; however, this was in outpatients with diabetes with MCI diagnosed based on cognitive scores.119 One study measured IL-6 with a composite outcome measure including depression and is detailed in the section below.86 Details of studies measuring either IL-6 or TNF-α in association with depression or anxiety in AD are given in Table 1.

Table 1 Studies Measuring Peripheral IL-6 or TNF-α in Association with Depression or Anxiety in Alzheimer’s Disease. Number of Asterisks Denote Significance Level (* P<0.05, ** P<0.01, *** P<0.001)

Two studies from a large multi-centre research cohort (the Texas Alzheimer’s Research and Care Consortium) have studied the impact of inflammatory biomarkers on depressive symptoms. CRP was a significant predictor of the apathy domain of depression scores, and in combination with MIF and ICAM accounted for 26% of the variance in apathy scores in one study.121 In another study from this cohort, this effect appeared to be driven by female AD cases.122 Another study123 found no association between CRP, IL-1β, or transforming growth factor (TGF)-β and total geriatric depression scale (GDS) scores, despite significant group differences of IL-1β and TGF-β compared to controls, and correlations with cognitive scores.

Two studies89,116 also found higher levels of TNF-α in AD cases with depression, with the second finding a correlation between TNF-α and depressive symptoms, and a further study finding TNF-α contributed to an apathy in men.117 One study did not find an association,120 but this was a small study (n = 20) and did not include details of disease stage or cognitive scores. One study measured TNF-α with a composite outcome measure including depression and is detailed in the section below86 (Table 1 and Figure 6).

Figure 6 Graphical summary of studies testing the association of TNF-α or IL-6 in association with depression or anxiety in AD.

Notes: Studies measuring either TNF-α or IL-6 in association with depression or anxiety in AD. Size of the point and the number represent the study size, and color whether the study demonstrated broadly positive (blue) or negative (red) results. Studies in the figure are Alvarez et al 1996,120 Banerjee et al, 2017,89 Gorska-Ciebiada et al, 2015,118 Hall et al, 2012, Holmes et al, 2011,86 Johnson et al, 2012115 and Khemka et al, 2014116 (see “AD-Depression” section in the main text and Supplementary Table 2 for full details of individual studies).

Studies found higher levels of vascular endothelial growth factor (VEGF)124,125 in AD cases with depression compared to AD cases without depression, or controls, including two large studies81,95 as part of a multiplex panel, and significantly lower levels of the associated protein endocan.126

One study127 found associations between IL-7 and depressive symptoms, but these correlations were inverted depending on the assay source (positive correlation in serum and negative correlation in plasma), and differential effects in males compared to females. This highlights the need to consider assay source. A study that measured messenger ribonucleic acid (mRNA) by reverse transcription polymerase chain reaction (RT-PCR) in PBMCs128 found that whilst a number of cytokines had increased transcripts (interferon (IFN)-γ, TNF-α, IL-1R2) were increased in late life depression, those with AD and depression had downregulation of IL1-β. Measuring triggering receptor expressed on myeloid cells (TREM)1 mRNA in blood identified a significant negative correlation with depressive scores.129 TREM1 is related to TREM2, identified as a key inflammatory risk gene in AD.

AD- Composite Imaging

Two studies93,94 looked at the association between neuropsychiatric inventory (NPI) scores and regional bindings of PET ligands to the TSPO receptor. The first93 found agitation associated with increased TSPO binding in temporal regions, not explained by cognition or AD pathology (although could potentially be explained by differences in TSPO binding between controls and those with MCI/AD). This study was from a large well-characterized cohort with multimodal PET imaging. Interestingly, the second94 also identified agitation associated with TSPO binding, again excluding confounders.

Paired CSF and Blood

A longitudinal study measured CSF and serum inflammatory markers in cognitively unimpaired older adults, those with MCI, and those with AD, in association with NPS measured by the NPI-Questionnaire (NPI-Q).92 This study identified markers in both the CSF (IL-8, IL-15, soluble intracellular adhesion molecule (sICAM)-1 positively, and inducible protein (IP)10 and CRP negatively) and serum (macrophage inflammatory protein (MIP)-1α, IL-6, Eotaxin-3 positively and CRP negatively) associated with the presence of NPS, suggesting different pathways may be involved peripherally and centrally. These relationships varied when controlling for cognitive impairment and AD biomarker profiles, with CRP levels appearing the most robust. sICAM-1 was associated with longitudinal decline, highlighting how central nervous system (CNS) inflammation contributes to both NPS and cognitive decline.

CSF

There was an inverse relationship between IL-10 and NPS in one study.78 This study specifically targeted recruitment towards those with NPS, capturing a greater range of severity. Studies also identified inflammatory markers associated with cognition (IL-1β, IL-6, and IL-12p70)130 or disease state (gasdermin, a pyroptosis marker)131 but not NPS.

Serum and Plasma

A number of small studies found no relationship between single peripheral cytokines measured by ELISA (TNF-α or IL-1β) and NPS.132,133 Others using multiplex assays134 found no significant relationships; however, one study135 found that relationships depending on APOE status, with lower IL-15 being a significant predictor of NPS in APOE4 carriers, and lower IL-18 and TNF-α were significant predictors in non carriers, although this study did not control for disease stage or severity. A large86 (n = 275) longitudinal study with repeat biomarker measurement found that lower TNF-α predicted lower NPI scores over six months, and higher TNF-α was associated with increased agitation, depression and anxiety, with raised IL-6 associated with hallucinations and apathy. There were no associations between CRP and NPI scores.

Homocysteine, involved in a number of biological processes correlated with inflammatory markers, was associated with increased risk of NPS in a single study.136

PBMCs

Two studies used stimulated PBMCs84,137 which overcomes some challenges of measuring cytokines in serum,138 and found that IL-6 expression was associated with total NPI scores, and psychosis subscale scores. An unstimulated PBMC study found an association between NPI depression subscale and myeloid dendritic cells.139

AD-Agitation

Four studies looked specifically at agitation in AD disease, and each of these studies included an inpatient setting. One study included CSF,140 finding a relationship between the TSPO ligand diazepam-binding inhibitor (DBI) and agitation. DBI was correlated between CSF and serum, and this study is particularly notable as agitation and irritability were both the primary NPS associated with PET imaging of the TSPO receptor. Other studies using serum or plasma identified Il1-β, IL-6, natural killer cell activity (NKCA), and TNF-α associated with agitation.141–143

AD-Apathy

One of the largest studies identified82 (n = 1319) measuring a panel of inflammatory markers identified IL-6 and IL-10 as significant predictors of apathy scores, whilst two small studies identified associations with soluble TNF receptors,85 ICAM-1, and IL-1.144

AD-Psychosis

One study identified reduced IL1β from stimulated PBMCs as associated with psychosis in AD,145 and another found significantly higher IL-8146 and decreases in other cytokines in AD patients without psychosis, which may relate to antipsychotic treatment. A large study that included AD among dementia subtypes found no relationship between CRP and psychosis.79 Although this study grouped together a wide range of dementia diagnosis, it was one of the few performed on an inpatient ward.

AD-Sleep

Two studies by the same author96,147 looked at the relationship between inflammation and sleep in AD using actigraphy, and found no associations in MCI, and IL-6 associated with increased total sleep time and reduced sleep latency in dementia. There were limited details described in these studies, but they are unique in using a biosensor to measure outcomes.

AD-Summary

In summary, there are some potential associations between peripheral and central markers of inflammation and NPS in AD, but primarily without replication or with divergent results between studies. There is interest in how these relationships may be affected by the assay source127 or mediated by APOE status.135 Studies included potential mechanisms for associations, such as tryptophan metabolism81 and reductions in BDNF.111 The most studied combination is peripheral IL-6 and depression, with replication across three studies (although with a high degree of uncertainty). A large study found an association with IL-6 and apathy.82 The other largest studies identified positive associations with NPS and markers including homocysteine,136 TNF-a,86 and CRP,148 or combinations of markers in multiplex panels.81,95 Whilst there are only two PET imaging studies, both found associations between microglial activation and agitation93,94 suggesting there may be both central and peripheral immune associations with NPS.

Parkinson’s Disease

Thirty-five studies explored the relationship between inflammation and NPS in PD. Similarly to AD, the majority of studies looked at the relationship with depression and anxiety. A few studies explored hallucinations. The majority were in idiopathic PD, however two studies included participants with the LRRK2 mutation.91,149 The most common diagnostic criteria used was the UK Brain Bank criteria. Full descriptions of the studies are included in Supplementary Table 3.

PD-Depression CSF

There were eight studies testing the relationship between inflammatory markers in the CSF and depression in PD.150–157 A variety of markers were examined; however, the most common were IL-6, TNF-α, and IL-1β. The only markers with replicated findings were CRP and IL-10, showing positive associations in two studies.150,153,154,156

Serum and Plasma

Eight studies examined serum inflammatory markers and symptoms of depression or anxiety87,91,149,157–160 with one161 looking at depression as part of a composite measure. Four studies examined inflammatory markers in plasma.77,162,163 The most consistent positive associations were for IL-6 and TNF-α, although results were generally mixed. Details of studies measuring either IL-6 or TNF-α in association with depression or anxiety in PD are given in Table 2 and displayed in Figure 7.

Table 2 Studies Measuring Peripheral IL-6 or TNF-α in Association with Depression or Anxiety in Parkinson’s Disease. Number of Asterisks Denote Significance Level (* P<0.05, ** P<0.01, *** P<0.001)

Figure 7 Graphical summary of studies testing the association of TNF-α or IL-6 in association with depression or anxiety in PD.

Notes: Studies measuring either TNF-α or IL-6 in association with depression or anxiety in PD. Size of the point and number represent the study size, and color whether the study demonstrated broadly positive (blue) or negative (red) results. Studies in the figure are Fu et al, 2023,158 Karpenko et al, 2018,153 Kim et al, 2022,166 Lindqvist et al, 2012,161 Menza et al, 2010,163 Rastegar et al, 2019,149 Selikhova et al, 2002,162 Veselý et al, 2018,87 Wan et al, 2023,164 and Wang et al, 2016165 (see main text and Supplementary Table 3 for full details of individual studies).

The most studied individual biomarker was IL-6, with nine studies examining the association with peripheral IL-6 and depression or anxiety in PD. Four studies found an association with depression or anxiety. Two studies158,162 found a positive correlation with IL-6 and depression scores, and two longitudinal cohorts87,149 found that IL-6 positively correlated with an increase in depression scores over 2 years, suggesting that increased inflammation predicts subsequent NPS. Other studies165 found no association, or did not directly test the association directly but via miRNA intermediaries.164 The largest studies found positive associations,149,158 except for one which was stratified based on a range of variables beyond NPS.91 The discrepancies between the studies may also be explained by the use of plasma or serum, assay choice, depression scale, and inclusion of covariates. In studies that reported cognitive scores, IL-6 was also associated with worse cognition and more rapid disease progression.158,162

Three studies149,161,163 found an association between depression and peripheral TNF-α, with one showing that along with 13 other cytokines, TNF-α correlated with increases in GDS scores over two years. Four studies found no association between depression and TNF-α. Here there were similar sized studies, with differing assay methods and depression scales finding divergent results,149,153 and differences between studies including both idiopathic PD and LRRK2 cases.91,149

A large study91 stratified participants into three groups based on clinical variables and found that IL-8 (involved in neutrophil recruitment) was higher in a “malignant” subtype of PD patients with the LRRK2 mutation. This subgroup had higher levels of inflammatory markers, including monocyte chemoattractant protein (MCP)-1, and more severe symptoms, including higher levels of depression. One other study found no association with MCP-1149 and another found an association with MCP-1 but in the CSF, but do not report serum results,161 with a separate study finding an association with plasma MCP-1 and depressive symptoms.167

Two studies examined IL-1091,149 finding positive correlations with baseline and longitudinal GDS scores.

Three studies measuring CRP in serum87,160,161 found no association with depression, representing the larger cohorts, whilst two studies77,165 found higher rates of depressive symptoms were associated with elevated CRP. sIL-2r was found to be associated with depression in two studies.161,165 Other studies measuring single inflammatory markers (VIP, S100β, miR-218-5p) found potential associations that would need replication.

PBMCs

Three papers examined PBMCs in relationship to depression in PD.168–170 These studies used different panels looking at different immune cell populations and activation markers. Two168,170 found an association between natural killer (NK) cell populations (one an association between %CD56+ve cells, and one higher levels of p11, a protein involved in endo/exocytosis and linked to mood disorders). The third found an association with depression and myeloid dendritic cells, an association found in AD by the same group.169

Genetics

The largest study identified by our search83 incorporated genetics with disease phenotype in three independent longitudinal cohorts. This identified a cluster where anxiety and depression were associated with a severe non-motor phenotype, rapid disease progression, and AD pathology. This group was associated with increased risk genes expressed in microglia, implicating an inflammatory driver of a severe, neuropsychiatric disease phenotype.

An epigenetics study171 found 35 CpG regions enriched in immune pathways (IFN- γ signaling and T-cell receptor signaling) associated with depression in PD, and the PD with depression group had higher neutrophils, lower CD4 cells, and a higher neutrophil-to-lymphocyte ratio.

PD-Psychosis

Three studies looked at the relationship between inflammatory markers and psychosis in PD. One study172 found increased CRP in PD participants with either illusions or hallucinations, but these participants were also older and had lower cognitive scores, one found no associations with a range of cytokines,173 and one found an association with an IL-6 polymorphism and visual hallucinations. This study looked specifically at the development of visual hallucinations during dopaminergic drug treatment, and measuring a single genetic polymorphism in a small sample size risks a false positive result.

PD-Composite

A longitudinal study using the non-motor symptom scale (NMSS) found no correlations between individual cytokines and symptoms scores either at baseline or follow-up, but using principal components analysis identified a latent variable, representing IL-6 and IL-2, that was associated with worsening NMSS scores, mood, and apathy symptoms over time.166 One study of CSF155 found that lower TNF-α was associated with lower neuropsychiatric symptom scores. Another174 found no association between CRP and non-motor symptom scores.

PD- Fatigue and Sleep

Two studies looked at fatigue in PD, finding associations with sIL-2R161 and IL-6.90 Studies in sleep157 found lower CSF TNF-α and higher prostaglandin E2 (PGE2) in those with sleep disturbance, and no association between CRP174 and sleep quality.

PD - Summary

In summary, most studies measured the association between inflammatory markers and depression in PD. There was a replicated finding of peripheral IL-6 and depression, including increased depression scores at follow-up. IL-6 was also associated with composite NPS and fatigue. A large genetic study identified a cluster that linked inflammation to an overall more severe and rapidly progressive phenotype associated with NPS,83 as did a clinical cohort including LRRK2 participants.91 This suggests peripheral inflammation may be associated with NPS in PD, or may represent a more aggressive phenotype. Further studies are needed to draw conclusions about the impact of inflammation on other neuropsychiatric symptoms, like hallucinations and sleep disturbance, in PD.

Dementia with Lewy Bodies

Two studies specifically focused on DLB, with one study including those with MCI-LB. One used flow cytometry paired with a serum cytokine panel found no relationship between either immune cell subsets or cytokines with depression and NPI scores.175 Another study looked specifically at the relationship of domains of the NPI and cytokines (TNF-α and IL-6) in DLB.88 They identified increased TNF-α (but not IL-6) was associated with overall NPI score. Hallucinations, depression, sleep disturbance and eating disturbance domains were associated with higher TNF-α, although some of these associations could be due to age as a confounder. One study included DLB in a mixed dementia group that, as discussed, found no relationship between CRP and psychosis.79 In the only study of MCI-LB, there were no associations between a multiplex panel of cytokines and NPS.80 Full descriptions of the studies are included in the Supplementary Table 4.

Frontotemporal Dementia

One study looked specifically at the role of inflammation in apathy, a key neuropsychiatric symptom in FTD176 and found an association with B-Cell activating factor (BAFF) with apathy in behavioral variant FTD (bvFTD).

Huntington’s Disease

Two longitudinal studies177,178 investigated the role of peripheral inflammation in apathy, irritability, and depressive symptoms in HD using the problem behavior assessment scale. One found no associations with a cytokine panel, and the other found an association with CRP and apathy, but queried whether this was due to the initiation of antipsychotic medication rather than a disease process.

Full descriptions of the studies including frontotemporal dementia and Huntington’s disease are included in the Supplementary Table 5.

Conclusions and Future Directions

This systematic review identified a number of studies reporting associations between inflammatory biomarkers and NPS in neurodegenerative dementia. However, the results are inconsistent with limited replication. Disease severity is a common confounder, making it difficult to identify whether there is a direct role for inflammation in NPS, or inflammation and NPS are epiphenomena of a more severe disease phenotype. Firstly, we will briefly discuss the heterogeneity, study design and methodological challenges that may contribute to the inconsistent findings and propose areas for further study. Secondly, despite these challenges, there are some tentative mechanisms that may overlap between psychiatric disorders and dementia we shall discuss, and finally we will suggest some recommendations for future research.

Heterogeneity

Our inclusion criteria were broad in order to capture the range of inflammatory markers measured both peripherally and centrally across neurodegenerative diseases, but despite this there was still considerable heterogeneity in study participants, design, outcome measures, and assays. These differences may contribute to the discrepancies in results. For example, a range of different diagnostic criteria were used for AD, with some using standardized criteria including biomarkers,93 and others a single score on a cognitive test.118 Some studies recruit a specific stage of disease, such as mild-moderate,132 although the majority of studies do not specify this. One potential source of heterogeneity is the use of clinical criteria in the selection of participants for cohorts. There is an increasing shift in neurodegenerative diseases to move to biological definitions rather than clinical criteria, where diagnosis depends on the detection of disease-specific biomarkers.179,180 Although diagnostic criteria are validated against neuropathological diagnosis181 and some of the studies included biomarkers of amyloid-beta and tau in AD93 or specific genetic diagnosis of PD,91 a major potential source of bias is the reliance of clinical criteria, unsupported by biomarkers, as case definitions. This may lead to participants with a variety of underlying neuropathology under the same category and fail to capture cases with mixed pathology. As biomarkers for major neurodegenerative diseases become more widely accessible,182,183 future studies could incorporate cohorts with biomarker evidence of pathology, or examine the relationships between pathology, immune response and NPS independently of clinical diagnosis. Measures of outcomes also vary, for instance, depression can be measured with a clinical diagnosis,89 a self-reported measure such as the GDS81 or as a composite in an informant-related scale such as the NPI.135 Finally, for the same inflammatory marker, there are multiple assays used for quantification. TNF-α is measured using ELISAs from different manufacturers,132,147 as part of multiplex panels,175 with a range of preanalytical variables inconsistently reported.

Study Participants

Small studies, without defined diagnostic criteria, can lead to inconsistent results. Larger studies,81,86,92 including participants from multiple sites, and replicating in independent cohorts