Article Text

Download PDFPDF

Research paper
Peripheral inflammatory markers in Alzheimer’s disease: a systematic review and meta-analysis of 175 studies
  1. Ka Sing P Lai1,2,
  2. Celina S Liu1,3,
  3. Allison Rau1,3,
  4. Krista L Lanctôt1,3,4,
  5. Cristiano A Köhler5,
  6. Maureen Pakosh6,
  7. André F Carvalho5,
  8. Nathan Herrmann1,3,4
  1. 1 Neuropsychopharmacology Research Group, Sunnybrook Health Sciences Centre, Toronto, Ontario, Canada
  2. 2 Faculty of Medicine, University of Toronto, Toronto, Ontario, Canada
  3. 3 Department of Pharmacology and Toxicology, University of Toronto, Toronto, Ontario, Canada
  4. 4 Departments of Psychiatry, University of Toronto, Toronto, Ontario, Canada
  5. 5 Translational Psychiatry Research Group and Department of Clinical Medicine, Faculty of Medicine, Federal University of Ceará, Fortaleza, Ceará, Brazil
  6. 6 Toronto Rehabilitation Institute, University Health Network, Toronto, Ontario, Canada
  1. Correspondence to Dr Nathan Herrmann, Neuropsychopharmacology Research Group, Sunnybrook Health Sciences Center, 2075 Bayview Avenue, Suite FG08, Toronto, Ontario M4N 3M5, Canada; nathan.herrmann{at}


Objectives Increasing evidence suggests that inflammation is involved in Alzheimer’s disease (AD) pathology. This study quantitatively summarised the data on peripheral inflammatory markers in patients with AD compared with healthy controls (HC).

Methods Original reports containing measurements of peripheral inflammatory markers in AD patients and HC were included for meta-analysis. Standardised mean differences were calculated using a random effects model. Meta-regression and exploration of heterogeneity was performed using publication year, age, gender, Mini-Mental State Examination (MMSE) scores, plasma versus serum measurements and immunoassay type.

Results A total of 175 studies were combined to review 51 analytes in 13 344 AD and 12 912 HC patients. Elevated peripheral interleukin (IL)-1β, IL-2, IL-6, IL-18, interferon-γ, homocysteine, high-sensitivity C reactive protein, C-X-C motif chemokine-10, epidermal growth factor, vascular cell adhesion molecule-1, tumour necrosis factor (TNF)-α converting enzyme, soluble TNF receptors 1 and 2, α1-antichymotrypsin and decreased IL-1 receptor antagonist and leptin were found in patients with AD compared with HC. IL-6 levels were inversely correlated with mean MMSE scores.

Conclusions These findings suggest that AD is accompanied by a peripheral inflammatory response and that IL-6 may be a useful biological marker to correlate with the severity of cognitive impairment. Further studies are needed to determine the clinical utility of these markers.

  • cytokines
  • chemokines
  • adipokines
  • inflammation
  • alzheimer’s disease

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Alzheimer’s disease (AD) is the most common form of dementia accounting for up to 70% of all cases, either alone or in combination with cerebrovascular disease.1 AD currently affects approximately one in nine people aged 65 years and older worldwide and is expected to triple by 2050.2 AD is one of the most disabling degenerative dementias leading to poor quality of life and eventually death.3

Aberrant amyloid beta (Aβ) accumulation in plaques and the hyperphosphorylation of tau protein forming neurofibrillary tangles are hallmark features of the AD neurodegenerative cascade.4 However, interventions targeting these specific elements of the cascade have found limited success in modulating the clinical trajectory,5 indicating the need to uncover additional pathological mechanisms that may be involved in the development and progression of AD. Therefore, detection of additional biomarkers of AD may allow for the development of novel therapeutics for the prevention and treatment of AD.

Increasing evidence suggests that inflammation is involved in AD pathology and plays a crucial role in the early stages of disease when intervention may be most beneficial.6 Immunohistochemical evidence has demonstrated that brains in the early stages of AD are associated with amyloid plaques that are colocalised with complement factors, acute-phase proteins and proinflammatory cytokines.7 For example, interleukin (IL)-6 and monocyte chemoattractant protein (MCP)-1 are inflammatory markers that have been associated with neuronal injury through increased Aβ deposition,8 demyelination, white matter hyperintensities and neurodegeneration.9 10 They have also been shown to increase blood–brain barrier permeability to leucocytes and introduce neurotoxic cytokines and chemokines into the brain to alter neuronal structure.11–13 In preclinical studies, administration of antitumour necrosis factor alpha (TNF-α) monoclonal antibody has been shown to reduce amyloid plaques and tau hyperphosphorylation.14 Clinical studies investigating the effects with the administration of interferon (IFN)-1β1α antagonists in patients with AD have shown significant functional improvements, suggesting that reducing proinflammatory cytokine concentrations may have beneficial effects on AD symptomology.15

In recent years, studies have begun to investigate chemokines and adipokines as inflammatory mediators in the central nervous system. Evidence suggests that chemokines may play a role in neuron–microglia interactions underlying AD pathogenesis.16 Peripheral leptin and adiponectin have been shown to exert neuroprotective properties through regulation of brain metabolism and inflammation.17 18 Leptin receptors have also been found to be located on the hippocampus, suggesting that metabolic dysfunction may impair memory and learning processes.19 20 Higher leptin levels have been shown in animal models to reduce beta-secretase activity in neuronal cells and modify Aβ levels in brain.20

Peripheral cytokine measurements are less invasive and more readily accessible than other techniques for AD diagnosis, which includes cerebrospinal fluid (CSF) Aβ and tau levels, as well as structural and functional MRI and amyloid imaging.21 Through regulated transport processes, peripheral inflammatory markers are able to cross the blood–brain barrier22 and have neuromodulatory effects. Therefore, peripheral cytokine measurements may also be useful biomarkers in detecting AD progression.23 24 A meta-analysis assessing the association between peripheral IL-6 and C reactive protein (CRP) levels in healthy adults found that higher inflammatory marker concentrations conferred increased risk of developing AD.1 A previous meta-analysis by our group also found elevated peripheral concentrations of IL-6, TNF-α, IL-1β, transforming growth factor-β (TGF-β), IL-12 and IL-18 in patients with AD compared with healthy controls (HCs)25; however, that meta-analysis did not include reports of chemokines, adipokines or cellular adhesion molecules.

This meta-analysis adds to the current literature by reviewing recent studies investigating the association between cytokines and AD, with a focus on exploring emerging peripheral inflammatory markers. Therefore, this meta-analysis aims to provide an updated and more comprehensive analysis of cytokines (including peripheral chemokines, colony-stimulating factors and adipokines), acute phase reactant proteins, cellular adhesion molecules and fibrinogens.


Data sources

Methodology recommended by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines was followed for this review.26 Articles published before September 2016 were searched using MEDLINE, PsycINFO, Embase, Cochrane Database of Systematic Reviews, Cochrane Central Register of Controlled Trials and CINAHL databases for original reports containing measurements of inflammatory factors in AD patients and HCs in peripheral blood. A sample search strategy (for Embase) is detailed in online supplementary table 1. Three independent reviewers were involved for assessment and data extraction of each retrieved reference.

Supplementary file 1

Study selection

Inclusion criteria consisted of: (1) original peer-reviewed clinical studies reporting levels of inflammatory markers in serum or plasma, (2) clinical diagnosis of AD (based on criteria from the National Institute of Neurological and Communicative Diseases and Stroke/Alzheimer’s Disease and Related Disorders Association or the Diagnostic and Statistical Manual of Mental Disorders) and (3) inclusion of a medically healthy and cognitively intact control group. No language restrictions were applied. Studies measuring cytokine concentrations from peripheral blood cells following stimulation were excluded as these assays can introduce variability secondary to differential gene expression of cells.27 28 Preclinical studies and studies measuring inflammatory marker concentrations in postmortem samples of patients with AD were also excluded. Mild cognitive impairment (MCI) was also excluded as it was recently reviewed in a meta-analysis.29 Final decisions regarding inclusion were reached by consensus between reviewers.

Data extraction

The mean (±SD) peripheral inflammatory marker concentrations for AD and control groups were extracted for each study. In addition, participant characteristics (mean age, gender proportion, years of education, body mass index, cognitive test scores, apolipoprotein E genotype) and study characteristics (inclusion criteria, diagnosis method, cytokine assay methodology) for each study were extracted. Corresponding authors of publications were contacted for missing data.

Methodological quality assessment

Study quality and risk of bias was assessed using items from the Newcastle Ottawa Scale30 and the Cochrane Collaboration’s risk of bias assessment tool.31 Results were compared between raters, and final decisions regarding inclusion were reached by consensus.

Statistical analyses

Standardised mean differences (SMDs) and 95% CIs were calculated for each outcome using a random effects model.32 The random effects model was selected as variability in absolute biomarker concentrations was anticipated between assays from different laboratories.33 Random effects models are also preferable when significant heterogeneity is expected as it assumes genuine diversity across studies and incorporates a between-study variance within the calculations. An effect size (ES) of 0.2 was regarded as small, ES of 0.5 was moderate and an ES ≥0.8 was high.

Heterogeneity across studies was evaluated using the Cochran Q test, a weighted sum of the squares of the deviations in individual study ES estimates from the summary ES estimate. A p value of <0.05 was indicative of significant heterogeneity. The degree of heterogeneity was estimated with the I2 statistic to determine the impact of heterogeneity34: I2 >50% suggested large heterogeneity, and I2 >75% suggested very large heterogeneity.

To explore heterogeneity, inverse-variance weighted-meta regression analyses were used to regress the SMD against year of publication, mean age, gender proportion, and mean Mini-Mental State Examination (MMSE) scores35 if at least nine studies measured a particular inflammatory factor. Subgroup analyses were performed to determine whether type of blood sample (ie, plasma, serum or whole blood) or type of assay contributed to heterogeneity if data from at least three datasets were available.

Evidence of publication bias was assessed qualitatively using funnel plots and quantitatively with the Egger’s test.36 All statistical analyses were conducted using Stata software (V.12).


Literature search findings

Search criteria returned 5687 unique records of inflammatory marker studies in living subjects with AD (online supplementary figure 1). A total of 329 studies measuring peripheral blood inflammatory marker concentrations were reviewed in full text, and a total of 175 studies were included in this meta-analysis. Characteristics of included studies are detailed in online supplementary table 2 and assessment of study quality in online supplementary table 3.

A total of 71 analytes were retrieved from literature search and were considered for inclusion. Seventeen analytes had fewer than three datasets available and therefore were excluded from meta-analysis. In addition, copper and ceruloplasmin were excluded as they may be involved in other pathological processes, and brain-derived neurotrophic factor was excluded as it is beyond the scope of this meta-analysis. Therefore, a total of 51 analytes were studied combining results for 13 344 AD and 12 912 HC patients (online supplementary table 2).

Differences in peripheral marker concentrations

Peripheral marker concentrations were found to be significantly different between AD and HC in 17 analytes (table 1). Compared with HC, patients with AD had significantly higher peripheral blood cytokine concentrations for IL-1β, IL-2, IL-6, IL-18, α−1 antichymotrypsin, C-X-C (chemokine ligand)-10 (CXCL-10), epidermal growth factor (EGF), homocysteine, high sensitivity CRP (hsCRP), IFN-γ, soluble TNF-receptors 1 and 2, TNF-α converting enzyme (TACE) and vascular cell adhesion molecule-1 (VCAM-1) (table 1, figure 1 and figure 2). IL-1 receptor antagonist, leptin and transferrin were found to be higher in the HC group compared with AD (table 1). SMDs were not significant for other analytes studied (ie, IL-1α, IL-3, IL-4, IL-8, IL-10, IL-11, IL-12, α1-antitrypsin, adiponectin, ANG-2I, CC ligand (CCL)-2, CCL-3, CCL-5, CRP, E-selectin, fibrinogen, granulocyte-colony stimulating factor, intracellular adhesion molecule-1, insulin-like growth factor-1, insulin, macrocyte colony stimulating factor, MCP-3, macrophage inflammatory protein (MIP)-1δ, platelet-derived growth factor-BB, stem cell factor (SCF), stromal cell-derived factor (SCF)-1α, TNF-α, TGF-β, vascular endothelial growth factor, total cholesterol, total triglyceride, high-density lipoproteins, low-density lipoprotein and ceruloplasmin) (online supplementary table 4). Additional forest plots are included as supplementary figures.

Figure 1

Forest plot displaying peripheral blood interleukin (IL)-1β and IL-6 concentrations in AD and control subjects. Shown are the standardised mean differences and 95% CIs. Positive values denote higher in Alzheimer’s dementia patients, while negative values denote higher in healthy control subjects. SMD, standardised mean difference.

Figure 2

Forest plot displaying peripheral blood tumour necrosis factor (TNF)-α, soluble TNF receptor-1, soluble TNF-receptor-2 and TNF-α converting enzyme concentrations in AD and control subjects. Shown are the standardised mean differences and 95% CIs. Positive values denote higher in Alzheimer’s dementia patients, while negative values denote higher in healthy control subjects. SMD, standardised mean difference.

Table 1

Meta-analyses of studies measuring peripheral markers in individuals with AD versus HC

Investigation of heterogeneity and publication bias

Significant heterogeneity was found in most comparisons (table 1). Meta-regression revealed a significant association between MMSE scores and IL-6 concentrations (p=0.009, I 2=88.3%, R 2=29.66%) (figure 3). After using meta-regression to control for variations in MMSE scores across studies, the difference in IL-6 concentrations between AD and HC groups remained significant (p=0.004). Meta-regression analyses demonstrated no significant effect of MMSE scores on heterogeneity in the remaining analytes. Furthermore, there was no significant effect of publication year, mean age or gender on heterogeneity in the remaining analytes. Subgroup analyses evaluating the differences in bioassay techniques or plasma versus serum measurements did not significantly reduce heterogeneity in any of the analytes. Publication bias was not detected by funnel plots, Egger’s or trim and fill tests.

Figure 3

Meta-regression demonstrating inverse correlation between mean Mini-Mental State Examination (MMSE) scores and standardised mean differences (SMD) between Alzheimer’s dementia and healthy control subjects. Larger circles indicate a larger sample size included in that study.


Inflammatory marker concentrations

Brain inflammation is one of the hallmarks of AD and originates in the central nervous system, where several inflammatory products, including cytokines, are formed and quickly removed into the blood stream. The passage of cytokines through the blood–brain barrier allows for the ability to peripherally measure them.37

Numerous proinflammatory markers were found to be elevated in patients with AD compared with HC indicating chronic inflammation in AD. IL-1β, IL-2, IL-18, homocysteine, hsCRP and IFN-γ, all pro-inflammatory molecules, were found to be significantly elevated in patients with AD. IL-1β has been suggested to be secreted from microglia following Aβ deposition in the brain leading to chronic neuroinflammation, neuronal dysfunction and ultimately accelerating neurodegenerative processes in AD.38 39 IL-2 and IL-18 are also important proinflammatory molecules in the recruitment of leukocytes, lymphocytes and macrophages to stimulate an inflammatory response. In addition, α1-antichymotrypsin enzyme levels were elevated in AD patients compared to HC, and transferrin was found to be decreased in AD patients compared to HC. Both α1-antichymotrypsin and transferrin are proteins well established as markers of an elevated inflammatory state. Homocysteine and hsCRP have also been suggested to be involved in the pathogenesis of AD through pathways of inflammation, cerebral microangiopathy and endothelial dysfunction.40 41

IL-6 has been extensively investigated in both preclinical and clinical studies as a proinflammatory cytokine that can accelerate ongoing neurodegenerative processes in AD.38 IL-6 can be produced by lymphoid cells and macrophages in the periphery,42 43 which in turn communicates with the brain through complex transport mechanisms and cellular activation.44 45 As a result, there is local activation of the immune system in the cortex, and the microglia and astrocytes release IL-6 in the brain under Aβ deposition.46 These cells are also known to release cytokines in response to Aβ deposition. This study found that levels of IL-6 were elevated in the AD population compared with HC and inversely correlated with mean MMSE scores. This finding is substantiated by neuroimaging correlations between IL-6 levels and ventricular volumes in patients with AD,47 suggesting that IL-6 plays a critical role in the inflammatory cascade and may clinically correlate with the severity of AD. Some studies have suggested that AD is primarily an immunologically driven process as IL-6 is associated with altered amyloid precursor protein metabolism.48–50 Since serum IL-6 levels have been found to correlate with matched CSF samples in patients with AD,51 peripheral IL-6 levels may lend utility as a surrogate marker for AD severity. Although advanced age has been found to be associated with higher IL-6 levels,52 this meta-analysis did not find age to be a significant contributor to heterogeneity suggesting that age alone does not explain for the observed elevation in blood IL-6.

TNF-α is a molecule that is elevated in many inflammatory disease states although it was found to be similar between AD and HC groups. However, as inflammation is a dynamic process and TNF-α is a marker that has been found to be elevated in many disease states, soluble TNF receptor (sTNF-R) levels may be a more reliable indicator of chronic TNF-α system activity as the receptor half-lives are longer than TNF-α itself.53 54 This study found that sTNF-R1 and sTNF-R2 were significantly elevated in the AD group compared with HC, indicating that these markers provided good discrimination between AD and HC patients.55 Evidence has also suggested that elevated levels of sTNF receptors are associated with higher incidence of conversion from MCI to AD.54 56 sTNF receptors stimulate Aβ production through activation of transcription factors such as nuclear factor-κβ and promotes cellular apoptosis.54 As TACE produces soluble TNF-receptors (sTNF) in the blood, elevated TACE levels in AD compared with HC were expected.

Previous reports have demonstrated that IL-12 and TGF-β are elevated in peripheral blood, although this study showed no differences between patients with AD versus HCs. In a previous meta-analysis by our group, levels of IL-12 and TGF-β in peripheral blood of patients with AD were found to be elevated compared with HCs.25 This may be secondary to methodological differences in inclusion criteria, and/or a result of new studies that have been published since the previous review.

Chemokine concentrations

The results of this study demonstrated increased peripheral concentrations of CXCL-10, consistent with the finding that IFN-γ was also elevated as IFN-γ stimulates CXCL-10 release. In AD, the binding of CXCL-10 to C-X-C chemokine receptor-3, a chemokine receptor involved in T-cell priming and the maintenance of natural killer cells in the body, has been hypothesised to induce the extracellular signal-regulated kinase pathway resulting in neuronal dysfunction and apoptosis.57–59 It has also been suggested that elevations in CXCL-10 may be a result of Aβ deposition60 leading to astrocyte aggregation and migration, ultimately further propagating the inflammatory process. Although previous studies have reported elevated concentrations of CCL-261 in the CSF of patients with AD, the results of this meta-analysis did not find a statistically significant difference in peripheral CCL-2 concentrations between AD and HCs. This suggests that peripheral CCL-2 may not be a strong marker of AD pathology. Expression of CCL-3 and CCL-5 in AD has also been shown to be elevated in previous studies,62 but the results of this study did not demonstrate a statistically significant difference between AD patients and HCs. This may be attributed to the small number of studies comparing peripheral concentrations of CCL-3 and CCL-5 in these populations.

Adipokines and cellular growth factors

Leptin, an adipokine, was significantly lower in patients with AD compared with HCs. Synthesised by adipocytes, leptin has been shown to have neuroprotective functions and plays a role in the hippocampus and long-term potentiation.63 64 Therefore, decreased concentrations in AD may result in a loss of neuroprotective effects and an increase in neurodegeneration and cognitive deficits witnessed in patients with AD. Previous studies have demonstrated that peripheral adipokine concentrations reflect cerebral spinal fluid levels in both AD and control groups.65 It is possible that the differences in patient body fat composition may explain the heterogeneity and variability of leptin synthesis in this population. These findings may also explain and provide a biological mechanism linking exercise and cognitive protection from the development of AD. Adiponectin is an adipokine that confers neuroprotective properties, although peripheral blood levels were not found to be different between AD patients and HC subjects.

This study found that patients with AD had increased levels of plasma EGF compared with HC subjects. EGF is a mitogenic polypeptide that stimulates cellular growth, differentiation and survival through its binding to the EGF receptor, resulting in multiple downstream effects.66–68 Few studies have explored the significance of EGF and receptor overexpression in AD pathology.69 A study evaluating the levels of EGF receptors has found increased expression in astrocytes, particularly near damaged regions in patients with AD,70 with subsequent upregulation of Aβ precursor protein.71 Therefore, circulating levels of EGF in combination with receptor expression and activity may be involved in the pathology of AD.

VCAM-1 is a cell adhesion molecule and part of the immunoglobulin superfamily that is expressed on endothelial cells and is involved in the regulation of microvasculature permeability.72 Cytokines such as TNF-α, IL-1β and IFN-γ are released and induces VCAM-1 expression.73 74 VCAM-1 is then involved in leucocyte migration and downstream signalling cascades.74 Given the significant inflammatory component in the pathogenesis of AD, the increased inflammatory markers (eg, IL-1β and IFN-γ) may be responsible for the increase in VCAM-1 in AD.73 75

Limitations and future directions

Although meta-regression and subgroup analyses were performed, significant heterogeneity remained for most analytes suggesting that other sources of heterogeneity could not be assessed systematically among the included studies. For example, studies did not consistently report disease severity. In clinical studies, individuals with severe AD have been shown to have greater levels of TNF-α compared with those with mild-to-moderate AD.76–78 Certain medical conditions such as inflammatory bowel disease, rheumatoid disease or cancer are also known to be involved with elevated inflammatory marker concentrations and patients with rheumatoid arthritis can be at an elevated risk of AD.79 It is also possible that additional factors such as acute infections may have elevated inflammatory markers (eg, CRP and hsCRP) and were not reported in these studies. Concomitant medications including non-steroidal anti-inflammatory drugs and corticosteroids are also involved in the modulation of neuroinflammation80 81 and were not consistently reported. Therefore, a more thorough report of disease severity, comorbidities and concomitant medications is needed to determine the effect of these variables on heterogeneity. Furthermore, numerous studies have found associations between CSF neuroinflammatory markers (eg, TNF-α, IL-1β and IL-6) with amyloid-β protein and neurofibrillary tangles (NFTs) in patients with AD.82 83 Future studies would be needed to evaluate how these peripheral markers would correlate with CSF inflammatory markers such as amyloid-tau, and if these markers can identify pathologically confirmed cases of AD would be of great importance. The present meta-analysis was also limited by the small number of studies included in the analyte comparisons (eg, CXCL-10, TACE, IL-3, IL-12, ANG-2I, CCL-3, MIP-1δ and SDF-1α). In addition, as the studies were cross-sectional, this only allowed for correlations to be evaluated and may not represent a continuous elevation in the disease state. Further studies would be needed to elucidate the mechanisms underlying the immunological processes of AD more clearly.

The clinical significance of peripheral inflammatory markers remains unclear as elevated peripheral inflammatory marker concentrations may not fully represent inflammatory activity within the central nervous system. However, given the ability for central nervous system molecules to disrupt the blood–brain barrier and enter the periphery, the inflammatory analytes investigated in this meta-analysis may provide utility as biomarkers to identify AD pathology. Future studies should consider using a metabonomics approach84 to identify a ‘fingerprint’ of AD to allow for the development of targeted therapeutics.


This meta-analysis found evidence to suggest elevated peripheral levels of IL-1β, IL-2, IL-6, IL-18, sTNF-R1, sTNF-R2, homocysteine, hsCRP, IFN-γ, CXCL-10, EGF, VCAM-1, α1-antichymotrypsin and transferrin and decreased levels of IL-1Ra transferrin and leptin in patients with AD compared with HCs, emphasising the role of peripheral inflammation in AD pathology. However, it should be noted that there was significant heterogeneity in most comparisons. It remains to be investigated whether inflammatory markers are associated with particular clinical characteristics or disease stages that may contribute to heterogeneity within AD cohorts. Further studies are needed to appropriately determine the clinical utility of these inflammatory markers in the diagnosis, prognostication, treatment and management of AD.


We would like to acknowledge the following authors for contributing their data for this meta-analysis: Professor Chafia Touil-Boukoffa; Dr Jianping Jia; Drs DoHoon Kim and Kyung Chan Choi; Drs Kang Soo Lee and Chang Hyung Hong; Drs Rufina Leung and Simon Lovestone; Dr Christian Humpel and Josef Marksteiner; Dr L. Malaguarnera; Professor Aynur Özge; Dr Myron F. Weiner; Dr Michael McGrath; Drs Daniel A. Llano and Viswanath Devanarayan; Dr Vijendra Singh; Dr Masaaki Waragai.


View Abstract


  • Contributors KSPL, CSL and AR were involved in the planning, data extraction, data analysis, and writing of the manuscript. MP was involved in data acquisition for the manuscript. CAK and AFC were involved in editing the manuscript. KLL and NH were involved in planning, editing and overseeing completeness of the manuscript.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.