Article https://doi.org/10.

1038/s41467-023-38602-6

Association of gout with brain reserve and

vulnerability to neurodegenerative disease

Received: 12 December 2022 Anya Topiwala 1 , Kulveer Mankia2, Steven Bell 3, Alastair Webb4,
Klaus P. Ebmeier 5, Isobel Howard 1, Chaoyue Wang 6,7,
Accepted: 9 May 2023
Fidel Alfaro-Almagro6, Karla Miller 6, Stephen Burgess 8,9, Stephen Smith6 &
Thomas E. Nichols 1,6

Check for updates

Studies of neurodegenerative disease risk in gout are contradictory. Rela-
tionships with neuroimaging markers of brain structure, which may offer
1234567890():,;
1234567890():,;

insights, are uncertain. Here we investigated associations between gout, brain

structure, and neurodegenerative disease incidence. Gout patients had smaller
global and regional brain volumes and markers of higher brain iron, using both
observational and genetic approaches. Participants with gout also had higher
incidence of all-cause dementia, Parkinson’s disease, and probable essential
tremor. Risks were strongly time dependent, whereby associations with inci-
dent dementia were highest in the first 3 years after gout diagnosis. These
findings suggest gout is causally related to several measures of brain structure.
Lower brain reserve amongst gout patients may explain their higher vulner-
ability to multiple neurodegenerative diseases. Motor and cognitive impair-
ments may affect gout patients, particularly in early years after diagnosis.

Gout is the most common inflammatory arthritis affecting ~1–4% of the Clarifying the impact on the brain is vital given that hyperuricaemia is a
population1. The clinical syndrome is characterised by acute flares of treatable target.
joint pain and swelling, resulting from deposition of monosodium Insights into relationships between gout and neurodegenerative
urate crystals in joints and peri-articular tissues, initiating an acute disease could result from examining links with brain structure, as yet
inflammatory cascade. In contrast to multiple other organ systems2, unexplored. MRI markers provide quantitative, sensitive intermediate
classically the brain is not thought to be affected. However, emerging endophenotypes for neuropsychiatric disease10. A few studies have
studies have cited contradictory associations between hyperuricaemia investigated associations between serum urate and a handful of bio-
and neurodegenerative disease. Observational studies have reported a markers for stroke and dementia. Most have found no association11,12.
lower risk of dementia, particularly Alzheimer’s disease, in However, to date, no studies have examined gout. Urate analyses have
hyperuricaemia3–5. Antioxidant properties of uric acid have been pro- been small (n < 2500), not accounted for many potential confounding
posed as a potential mechanism for this neuroprotection6. Mendelian variables, and explored only a few aspects of brain structure, while
randomisation studies, which can offer insights into causal relation- using solely observational approaches.
ships, have offered conflicting results in Alzheimer’s disease7,8. We performed the first investigation of neuroimaging markers in
Hyperuricaemia and gout have also been linked to higher stroke risk9. patients with gout. Observational and Mendelian randomisation (MR)

1
Nuffield Department of Population Health, Big Data Institute, University of Oxford, Oxford, UK. 2Leeds Institute of Rheumatic and Musculoskeletal Medicine,
University of Leeds and NIHR Leeds Biomedical Research Centre, Leeds Teaching Hospitals NHS Trust, Chapel Allerton Hospital, Leeds, UK. 3Department of
Clinical Neurosciences, School of Clinical Medicine, University of Cambridge, Cambridge, UK. 4Wolfson Centre for Prevention of Stroke and Dementia,
Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, UK. 5Department of Psychiatry, University of Oxford, Warneford Hospital,
Oxford, UK. 6Wellcome Centre for Integrative Neuroimaging (WIN FMRIB), Nuffield Department of Clinical Neurosciences, University of Oxford, Oxford, UK.
7
SJTU-Ruijin-UIH Institute for Medical Imaging Technology, Shanghai, China. 8MRC Biostatistics Unit, School of Clinical Medicine, University of Cambridge,
Cambridge, UK. 9Department of Public Health and Primary Care, School of Clinical Medicine, University of Cambridge, Cambridge, UK.
e-mail: anya.topiwala@bdi.ox.ac.uk

Nature Communications | (2023)14:2844 1

Article https://doi.org/10.1038/s41467-023-38602-6

approaches were combined for stronger causal inference. Further- (beta = −5.63E-04, T = −6.23, P = 4.95E-10) and midbrain (beta = −4.00
more, we explored relationships between gout and relevant neuro- E-04, T = −5.15, P = 2.67E-07) in gout and high urate (Fig. 2). Cerebellar
degenerative diseases. The purpose of this study was to assess whether differences were focused on the posterior lobe, particularly inferiorly.
associations between gout and brain structure would provide insights Subcortical differences centred on the nucleus accumbens, putamen
into relationships with neurodegenerative disease risk. and caudate. Significant differences were also observed in white mat-
ter tract microstructure in the fornix, the major output of the hippo-
Results campus. These included lower fractional anisotropy (beta = −2.81E-4,
Participant characteristics T = −2.83, P = 4.73E-03) and higher mean diffusivity. Notably, there
In total, 11,735 participants (1165 within the imaging subset) had a were no significant associations with hippocampal volume or WMH
diagnosis of gout. Medical professionals made the majority of diag- volumes. Gout and higher urate significantly associated with markers
noses (31.1% primary care, 18.5% hospital admission). A minority (15.2%) of higher iron deposition (lower T2* and higher magnetic suscept-
were solely self-reported. Overall, 30.8% of gout patients reported ibility) of several basal ganglia structures, including bilateral putamen
current use of urate-lowering therapy (ULT) at assessment. Gout suf- (beta=9.50E-04, T = 9.25, P = 2.38E-20) and caudate (7.62E-4, T = 7.12,
ferers were older and comprised a greater proportion of males. Rela- P = 1.13E-12). Whilst associations with urate were generally robust after
tionships between baseline urate and demographic variables differed additional adjustment for urate-lowering therapy and consequences of
by sex (Supplementary Data 1). In males, urate was positively corre- hyperuricaemia, many with gout were not. Further exploration of these
lated with alcohol intake and lower socioeconomic status. This was not models clarified that blood pressure was responsible for a loss of
the case in females. Mean baseline serum urate amongst asymptomatic association with cerebellar and striatal volumes, whereas markers of
individuals was 305.0 ± 76.7 μmol/L. During follow-up 3126 partici- renal function (creatinine and cystatin C) reduced associations with
pants developed dementia and 16,422 individuals died. Deaths basal ganglia QSM susceptibility measures. Higher urate (but not gout)
amongst gout patients were more than double those of controls was positively associated with cuneal and frontal gyrus volumes in
(11% vs. 5%). model 1, but not in model 2 or in stratified analyses amongst solely the
highest earners (Supplementary Data 2).
Observational analyses Gout associated with a higher incidence of dementia (average
Gout and higher urate were associated with multiple measures of brain over study HR = 1.60 [1.38–1.85]). The risk was time-varying, highest in
structure (Fig. 1). There were no marked sex differences in associa- the first 3 years after gout diagnosis (HR = 7.40 [4.95–11.07]) and then
tions. Urate inversely associated with global brain volume (beta = decreasing (Fig. 3). We explored possible reasons for this temporal
−1.66E-4, T = 3.60, P = 2.30E-4), and also grey and (separately) white gradient. Accounting for the competing risk lowered hazard ratios but
matter volumes, with corresponding higher cerebrospinal fluid did not alter the pattern of associations (first 3 years HR = 3.53
volumes (beta=4.93E-4, T = 4.94, P = 7.84E-4). To contextualise the [2.31–5.38]) (Supplementary Data 3). Excluding controls with asymp-
effect size, we related the differences to the cross-sectional effects of tomatic hyperuricaemia (average HR = 1.59 [1.42–1.78]) or additionally
age (discounting non-linear effects of age). The effect of gout on the controlling for possible consequences of hyperuricaemia (average
global grey matter was equivalent to a 2-year greater age, assuming all HR = 1.54 [1.27–1.87]) made little impact. Receiving a gout diagnosis
other potentially confounding factors were held constant. later in the study (independently from age) was associated with a
There were highly significant differences in regional grey slightly higher dementia incidence (average HR = 1.03 [1.01–1.05]). ULT
matter volumes, particularly of mid- and hindbrain structures, at baseline assessment was not linked to dementia incidence (HR −0.25
such as cerebellum (beta = −9.91E-04, T = −9.26, P = 2.25E-20), pons [0.59–1.04]).

Associations between brain structure image-derived phenotypes and serum urate in asymptomatic participants

20

15
-log10(p value)

10

5
Bonferroni

FDR

Brain structure measures

Global volumes Cortical volumes Tissue intensity Iron markers (T2* and susceptibility)

Subcortical volumes Cortical areas Grey-white matter contrast Diffusion MRI measures

Regional volumes Cortical thicknesses White matter hyperintensities

Fig. 1 | Associations between serum urate and brain structure measures in ratio, alcohol intake, smoking status, diuretic use. Multiple comparison correction:
asymptomatic participants (n = 33,367). Multiple linear regression models blue line represents False Discovery Rate (FDR) two-sided P value threshold
adjusted for: all image-related confounds, age, age2, sex, educational qualifications, (5%) = 9.58 × 10−3, red line represents Bonferroni threshold on 2138
Townsend Deprivation Index, household income, historical job code, waist-hip- tests = 2.34 × 10−5). Source data are provided as a Source Data file.

Nature Communications | (2023)14:2844 2

Article https://doi.org/10.1038/s41467-023-38602-6

0
5% FDR-thresholded T statistic
P A R L R L R L

-5.9
Lorem ipsum
MNI coordinates: 7.2,14,-8.2 MNI coordinates: -9.9x105, -60.3, -47.1

Fig. 2 | Differences in regional grey matter volume between participants with threshold on uncorrected P values). Models adjusted for: age, age2, alcohol units
gout (n = 1165) and controls (n = 32,202), as analysed by voxel-based morpho- weekly, imaging site, smoking status, waist-hip-ratio, total household income. FDR
metry. Blue regions represent areas where participants with gout had significantly false discovery rate, L left, R right, A anterior, P posterior, MNI Montreal Neuroi-
less grey matter. T statistics are thresholded at a 5% false discovery rate (0.0013 maging Institute. Source data are provided as a Source Data file.

Time after exposure

0-3 years
3-6 years
>6 years

Hazard ratio

Fig. 3 | Hazard ratios of incident dementia for gout (N = 303,149, 3126 dementia measurement, generated from Cox proportional hazards models, adjusted for: age,
cases) and serum urate in asymptomatic participants (N = 247,328, 2454 age2, sex, Townsend Deprivation Index, educational qualifications, household
dementia cases). Estimates (points) and their 95% confidence intervals (bars) income, historical job code, smoking, alcohol intake, waist-hip-ratio, diuretic use.
represent hazard ratios over different time periods after gout diagnosis or urate Source data are provided as a Source Data file.

Risks were higher for vascular dementia (average HR = 2.41 and incident Parkinson’s disease (HR = 0.89 [0.84–0.95], P for non-
[1.93–3.02]) compared to all-cause dementia, but not for Alzheimer’s linearity=0.3), but not with probable essential tremor (HR = 0.95
disease (average HR = 1.62 [1.30–2.02]). Again, there was a strong time [0.89–1.02]).
dependence to the risks, particularly for vascular dementia. Amongst
asymptomatic individuals, in the linear model, there was an inverse Genetic analyses
association between urate and dementia incidence (HR = 0.85, 95% CI: Four of the urate-associated SNPs and one of the gout-associated SNPs
0.80–0.89), with no time dependence. Using restricted cubic splines a were unavailable in the outcome datasets. Proxy SNPs were found for
non-linear relationship between urate and dementia incidence was three urate SNPs (Supplementary Data 4). Genetic associations mostly
observed (P for non-linearity <0.0001). Both low and high urate levels mirrored observational ones. Both genetically predicted gout and
associated with a higher dementia incidence (Fig. 2). serum urate significantly associated with regional grey matter
Given the associations with cerebellum and striatum measures, we volumes, including cerebellar (gout IVW beta = −0.05 [−0.08 to −0.03],
examined relevant phenotypes post hoc. Gout was associated with a P = 1.50E-04; urate IVW beta = −0.05 [−0.09 to −0.01], P = 7.71E-03),
higher incidence of both Parkinson’s disease (HR = 1.43 [1.15–1.79]) and midbrain (gout IVW beta = −0.07 [−0.12 to −0.03], P = 1.55E-03; urate
probable essential tremor (HR = 6.75 [5.59–8.00]). Amongst asympto- IVW beta = −0.06 [−0.10 to −0.02], P = 7.31E-03), pons (gout IVW
matic individuals, there was a linear inverse relationship between urate beta = −0.11 [−0.16 to −0.06], P = 3.01E-05; urate IVW beta = −0.07

Nature Communications | (2023)14:2844 3

Article https://doi.org/10.1038/s41467-023-38602-6

Outcome Estimate
Grey matter volumes

Global

Cerebellar

Pons

Brainstem
Exposure
Midbrain
Gout
Striatum
Serum urate
Susceptibility

Putamen, left

Putamen, right

Caudate, left

Caudate, right

-0.2 -0.1 -00 0.1 0.2

Beta

Fig. 4 | Mendelian randomisation estimates for the association of genetically asymptomatic hyperuricaemia. Magnetic susceptibility measures are derived from
predicted gout and serum urate and brain imaging phenotypes. Genetically quantitative susceptibility maps, with higher values indicating higher iron. IDP
predicted gout was instrumented using 12 SNPs (one-sample MR). Genetically image-derived phenotype, SNPs single-nucleotide polymorphism, LCI lower con-
predicted serum urate was instrumented using 182 SNPs (two-sample MR). Beta fidence interval, UCI upper confidence interval. Source data are provided as a
estimates (points) and their 95% confidence intervals (bars) represent inverse- Source Data file.
variance weighted estimates for a 1 mg dl−1 increase in serum urate, or gout versus

[−0.11 to −0.02], P = 2.10E-03) and brainstem (gout IVW beta = −0.11 reported higher white matter atrophy with hyperuricaemia14, and
[−0.16 to −0.06], P = 3.78E-05; urate IVW beta = −0.07 [−0.10 to −0.02], another a null association with hippocampal volume11. We replicated
P = 2.48E-03). There were also significant associations with higher these findings, but did not duplicate reported positive relationships
magnetic susceptibility in the putamen and caudate, markers of higher between urate and WMH15. Methodological differences may be
iron (Fig. 4). However, whilst genetically predicted gout was sig- responsible. In UKB, WMH volumes were calculated automatically
nificantly associated with global grey matter volume (IVW beta = −0.05 whereas previous studies used visual ratings. We found associations
[−0.08 to −0.01], P = 7.75E-03), urate did not IVW beta = −0.05 [−0.29 with global grey matter and cerebellar (motor and non-motor
to −0.20], P = 0.71]. All associations survived FDR correction for mul- regions16), brainstem and striatal brain volumes. Gout patients had a
tiple testing. Pons, brainstem and the magnetic susceptibility asso- higher incidence of probable essential tremor. The neurobiological
ciations additionally survived Bonferroni correction. Robust MR correlates of essential tremor are unclear, although links have been
methods including the weighted median and mode gave broadly made with cerebellar atrophy17,18. As this likely comprises a highly
consistent estimates for gout (Supplementary Data 5). However, for heterogeneous category we are cautious in our interpretations of
urate only IVW estimates were significant. Whilst there was little evi- these results. Associations with markers of higher iron in basal ganglia
dence of horizontal pleiotropy on the MR Egger intercept test, there with hyperuricaemia mirror findings in aging, Alzheimer’s and Par-
was significant heterogeneity between IVW estimates, particularly kinson’s diseases19,20, and recently alcohol consumption21. Our MR
for urate. findings provide support for causal relationships underlying the
associations. Taken together, these brain findings lend further support
Discussion to recent claims of a causal impact of hyperuricaemia on
In this UK prospective cohort study, participants with a history of gout Alzheimer’s7,22.
had smaller global and regional brain volumes and higher brain iron. The mechanism(s) underlying how gout affects brain volume is
MR analyses suggested gout was causally related to brain structure. unclear. Hyperuricaemia has been linked to arterial stiffness23, and
Gout was associated with a higher incidence of several neurodegen- associated with brain microvascular damage24, which may improve
erative diseases, particularly in the first three years after diagnosis. with allopurinol treatment25. Alternatively, toxic metabolic pathways
Lower brain reserve in gout patients may explain their vulnerability to may be responsible26,27. Interestingly, we found that controlling for
dementia. blood pressure reduced associations between gout and brain volumes,
To our knowledge, there are no previous neuroimaging studies of suggesting this as a potential mediating mechanism. Whilst urate
gout. Investigations examining urate are small and, with two excep- associated with measures of white matter microstructure, no relation
tions, examined solely markers of cerebrovascular disease13. One study was observed with WMHs. One explanation is that due to selection

Nature Communications | (2023)14:2844 4

Article https://doi.org/10.1038/s41467-023-38602-6

biases, the healthier UKB imaging subset may not have yet manifested Alzheimer’s disease7,22. Differing choice of SNPs as instrumental vari-
WMHs. Of note, the strongest associations of hyperuricaemia were ables and outcome GWAS between studies may explain the contra-
with posterior brain regions, but the calculation of WMH volume only dictory results. Drawing together the recent MR studies and the
includes the anterior circulation. It is unclear whether, at least in consistent brain associations we observed in urate and gout, we
healthy individuals, urate can cross the blood brain barrier28. High hypothesise collider bias may be causing a spurious protective asso-
levels of iron in basal ganglia could result from inflammatory processes ciation for urate40. Both higher urate (in affluent individuals), and
in gout29,30 or poorer urinary iron excretion as suggested by our absence of dementia increase likelihood of UKB recruitment.
observed reduction in association after controlling for renal function. Taken together, these results lend support not only for a causal
Conversely, higher ferritin (a blood measure of iron) secondary to diet, role of gout in Alzheimer’s disease, but more widely in neurodegen-
could lead to higher urate levels29. erative disease. Furthermore, the brain associations offer a potential
In this sample gout associated with higher incidence of dementia, mechanism through which this risk is conferred. This is an area ripe for
a finding at odds with previous claims of protective effects in obser- further exploration in the future with mediation MR, when large-scale
vational studies13,31. A recent small meta-analysis found no evidence for GWAS of dementia subtypes become available. These findings are
an association with gout overall, but cited a possible protective effect relevant for gout patients, clinicians and researchers. There are also
on Alzheimer’s disease based on two studies5. To our knowledge, there wider implications for our understanding of mechanistic pathways in
have been just two MR studies of gout and dementia, both reporting neurodegenerative disease that could lead to novel intervention tar-
no significant associations with Alzheimer’s disease22,32. However, low gets. Prophylaxis for gout flares may be justified on the basis of brain
statistical power, driven by weak instrument bias, limits their inter- health in addition to joint and cardiovascular health, although inter-
pretation. Causal relationships with other dementia subtypes, such as ventional trials will be needed to test this. Patients with gout should be
vascular dementia, has been limited by the lack of available large-scale monitored for cognitive and motor symptoms of neurodegenerative
GWAS. Whilst we found gout was associated with smaller global brain disease given their increased risk, especially in the early period after
volumes, associations were not observed with classical imaging mar- diagnosis.
kers of Alzheimer’s disease33 or vascular dementia34 such as hippo-
campal volume35 or white matter hyperintensities. Instead as gout Strengths and limitations
played a causative role in multiple neurodegenerative pathologies, we The large size and statistical power enabled triangulation of observa-
propose a brain vulnerability model (Fig. 5). A similar phenomenon is tional and MR approaches. Consistent directions of association
well recognised with delirium36. Global brain volumes are commonly between different methods strengthen causal inference41. A compre-
conceived as a proxy for brain reserve, structural characteristics of the hensive range of brain metrics were examined, whilst adjusting for
brain that allow some people to better cope with brain pathology multiple confounds and multiple testing. Other key strengths include
before cognitive changes emerge37. The strong temporal gradient in the prospective cohort, examination of relevant clinical phenotypes in
risk of neurodegenerative disease after gout is interesting. We propose the same sample and time-to-event analyses accounting for
three hypotheses. First, death is a competing risk. Second, there may competing risks.
be a selective detection bias. Receiving a gout diagnosis may result in This study has several limitations. First, UKB is a self-selective
more frequent medical review initially, when cognitive problems could cohort42. Second, the measurement of serum urate was cross-sectional
be noticed. Third, the inflammatory response in a gout flare may and may not reflect chronic exposure43. Third, MRI data was at a single
induce a stepwise global decline that may precipitate acute cognitive time point so we cannot infer changes in brain structure. Fourth, case
decline. identification may be subject to ascertainment bias, likely skewed to
The relationship between baseline urate and dementia incidence more severe cases, which could bias associations towards the null.
was non-linear. As others have previously raised, associations with a Fifth, there may be diagnostic confounding between Parkinson’s dis-
single urate measurement, which may not reflect long-term ease and essential tremor. Sixth, gout covaries with obesity and alco-
exposure38, should be interpreted cautiously. Reduced risks of hol. However, we controlled for these factors, performed MR analyses,
dementia with hyperuricaemia have been previously reported in and we have previously reported brain structure associations with
observational studies4, including in an earlier analysis of UKB3. Whilst alcohol in a markedly different spatial distribution in the same
earlier Mendelian randomisation studies using genetically proxied sample44. Seventh, MR relies on a set of unverifiable assumptions,
urate, reflective of lifelong exposure, were not significant8,39, two stu- which we tried to assess where possible. Whilst we found no evidence
dies published this year found support for a harmful causal effect on of horizontal pleiotropy (untestable Inside assumption), there was

neurodegenerative
pathology
Reduced brain reserve

gout smaller brain Resilience factors

volume

clinical disease

accelerates
decelerates

Fig. 5 | Hypothesised model for the impact of gout on neurodegenerative disease. Gout is causally related to smaller brain volume, a proxy for cognitive reserve.
Reduced cognitive reserve indicates a reduced ability for individuals to tolerate neurodegenerative pathology before symptoms ensue.

Nature Communications | (2023)14:2844 5

Article https://doi.org/10.1038/s41467-023-38602-6

heterogeneity of MR estimates. This may be due to failure of the Neurodegenerative disease. Cases were algorithmically defined from
assumptions of homogeneity of the genetic association with the risk UKB’s baseline assessment data collection (including self-report),
factor or linearity of the effect of the risk factor on the outcome, or linked data from Hospital Episode Statistics, primary care and death
plausibly, differing biological causal mechanisms between variants. A records (https://biobank.ndph.ox.ac.uk/showcase/refer.cgi?id=460).
minority of genetic variants used to instrument urate may have suf- Incident cases (diagnosed any time after baseline) were included.
fered from weak instrument bias. However, post hoc choice of Prevalent cases (diagnosis before baseline) and those who developed
instruments based on measured F-statistics can exacerbate bias45. neurodegenerative disease prior to gout were excluded, to lessen
Eighth, the strongest genetic associations were between gout and reverse causation. The primary outcome was all-cause dementia. In
brain markers rather than urate. Sample overlap between gout and planned subgroup analyses, Alzheimer’s and vascular dementia were
brain imaging GWAS may bias genetic associations towards observa- examined.
tional associations. The interpretation of MR estimates with binary
exposures needs care. The underlying latent continuous trait (urate), Covariates
although alone is insufficient to cause gout, may influence the brain Selected a priori confound covariates were identified on the basis of
independently of gout and therefore bias MR estimates [exclusion literature supporting their impact on both gout and MRI markers: all
restriction assumption]46. As such, IV estimates from the gout MR image-related confounds, age, age2, sex (self-reported/NHS records),
analyses should be interpreted as an estimate for genetic compliers educational qualifications, Townsend Deprivation Index, household
(likely to be a small subgroup of the population). Finally, this was not a income, historical job code, waist-hip-ratio, alcohol, smoking, and
bespoke study with optimal measures for testing cerebellar function, diuretic use (Model 1). Additional adjustments for potential con-
for example, measured gait speed or eye movements. sequences of hyperuricaemia (potentially on the causal pathway)
In this population-based sample, participants with gout had included: blood pressure, cholesterol, cystatin C, creatinine, diabetes
smaller global and regional grey matter volumes. Lower brain reserve mellitus, urate-lowering therapy (ULT), and chronic kidney disease
may explain the vulnerability of gout patients to neurodegenerative (Model 2). Supplementary Methods describes measurement details.
disease. Clinicians should be vigilant for motor and cognitive problems
in gout patients. Genetic variants
Mendelian randomisation was performed for two related exposures:
Methods (1) serum urate (two-sample), and (2) gout (one-sample). Genetic
Study population instruments for urate (n = 183) were identified based on the sentinel
Participants were from UK Biobank (UKB) study47, which recruited variants at genome-wide significant loci (P < 5 × 10−8) reported in the
volunteers aged 40–69 years in 2006–2010. UK Biobank has approval largest publicly available genome-wide association study53 (Supple-
from the North West Multi-centre Research Ethics Committee (MREC) mentary Data 4). Genetic instruments for gout (n = 13) were identified
as a Research Tissue Bank (RTB) approval. All participants gave from the largest GWAS of gout versus asymptomatic hyperuricaemia
informed consent. A subset underwent imaging which included brain (aiming to distinguish high urate from inflammatory components of
MRI (~50,000 scanned to date). Participants underwent imaging at gout)54. Clumping was performed in the original GWAS, using a 1-Mb
three centres (Newcastle, Stockport or Reading) with identical Sie- locus. Genetic associations with brain phenotypes were obtained from
mens Skyra 3T scanners (software VD13) using a standard 32-channel the largest GWAS of brain imaging phenotypes55. Proxy SNPs were
head coil. Details of image pre-processing and quality control pipelines selected (LD R2 > 0.9) where available.
are described elsewhere48. Participants with complete data were
included (Supplementary Fig. S1). Statistical analysis
All analyses were performed in R (version 4.1.2) unless otherwise
Exposure measurements stated.
Gout diagnoses were algorithmically defined from UKB’s baseline
assessment data collection, linked data from Hospital Episode Statis- Observational analyses. Multivariable linear regression models
tics, primary care and death records (https://biobank.ndph.ox.ac.uk/ assessed the relationship between gout/urate and each of 2138 IDPs as
showcase/refer.cgi?id=460). Individuals who developed gout after outcomes. Brain measures and urate were quantile normalised
dementia were excluded to minimise reverse causation. Serum urate resulting in Gaussian distributions with mean zero and standard
(in μmol/L) was measured using a Beckman Coulter AU5800 at study deviation one. To account for multiple testing, Bonferroni and (sepa-
baseline in all participants. rately) 5% false discovery rate (FDR) multiple comparison corrections
were applied (2138 tests). Effect sizes were compared with those of age
Outcome measurements (Supplementary Methods). For VBM, the Big Linear Model toolbox
Brain markers. We used 2138 summary image-derived phenotypes (Supplementary Methods) was used to perform mass univariate
(IDPs) representing distinct measures of brain structure from the fol- ordinary least squares regression (parametric inference) voxelwise.
lowing modalities: T1-weighted and T2-weighted-FLAIR structural Cox proportional hazards models were used to estimate associa-
imaging, susceptibility-weighted MRI, diffusion MRI. These are outputs tions between gout/urate and neurodegenerative disease incidence.
of a dedicated processing pipeline and are available from UKB49. IDPs The length of follow-up was calculated as the interval between the
included: whole brain and cerebrospinal fluid (CSF) volumes, cortical origin and either date of event (first disease code) or censoring (date of
volumes, surface area, thickness, T2-FLAIR white matter hyper- death or last date of data collection—January 2, 2022). Origin was the
intensities (WMH) volumes50, brain iron deposition metrics (T2* and baseline date, except for participants who developed gout after
magnetic susceptibility51), white matter microstructural measures baseline, where the gout diagnosis date was used. The impact of time
(Supplementary Methods). to gout diagnosis on neurodegenerative disease incidence was
The precise spatial distribution of associations between gout and explored by including the interval between study baseline and gout
grey matter volume (T1-weighted images) across the brain was inves- diagnosis date as a fixed covariate. Relationships between the urate
tigated using FSL-VBM52 (Supplementary Methods), a method to level and dementia were assessed using linear (fixed effects) and non-
compare grey matter volume in each 3D volume element (voxel) of the linear regression models. For the latter urate was categorised into
structural image, after adjusting between individuals for estimated quintiles and restricted cubic splines (5 knots) applied56. Non-linearity
total intracranial volume. was formally tested (H0: β2 = β3 = …=βk − 1 = 0) with an F test.

Nature Communications | (2023)14:2844 6

Article https://doi.org/10.1038/s41467-023-38602-6

Assumptions were checked as described in Supplementary Methods. 3. Cao, Z. et al. Associations of BMI and serum urate with developing
Associations between gout and neurodegenerative disease incidence dementia: a prospective cohort study. J. Clin. Endocrinol. Metab.
accounting for the competing risk of death were assessed using the 105, e4688–e4698 (2020).
subdistribution method57. In sensitivity analyses, participants with 4. Scheepers, L. E. et al. Urate and risk of Alzheimer’s disease and
asymptomatic hyperuricaemia (defined by serum urate levels vascular dementia: a population-based study. Alzheimer’s Dement.
>357 μmol/L in females and >428 μmol/L in males58) were excluded 15, 754–763 (2019).
from the control group in gout analyses. 5. Pan, S.-Y., Cheng, R.-J., Xia, Z.-J., Zhang, Q.-P. & Liu, Y. Risk of
dementia in gout and hyperuricaemia: a meta-analysis of cohort
Genetic analyses. We used MR to investigate whether causal rela- studies. BMJ Open 11, e041680 (2021).
tionships could underpin the observational associations we found with 6. Ames, B. N., Cathcart, R., Schwiers, E. & Hochstein, P. Uric acid
brain structure (Supplementary Methods for further details, including provides an antioxidant defense in humans against oxidant-and
power calculation). One- (gout) and two-sample (urate) linear MR radical-caused aging and cancer: a hypothesis. Proc. Natl Acad. Sci.
analyses using summary statistics from European participants were USA 78, 6858–6862 (1981).
conducted using R packages MendelianRandomization (version 0.5.1) 7. Şanlı, B. A. et al. Unbiased metabolome screen links serum urate to
and TwoSampleMR (version 0.5.6). Variant harmonisation ensured that risk of Alzheimer’s disease. Neurobiol. Aging 120, 167–176 (2022).
association between SNPs and exposures, and between SNPs and 8. Efstathiadou, A., Gill, D., McGrane, F., Quinn, T. & Dawson, J.
outcomes, reflected the same allele. Several robust MR methods were Genetically determined uric acid and the risk of cardiovascular and
used to evaluate the consistency of the causal inference. To adjust for neurovascular diseases: a Mendelian randomization study of out-
multiple testing, Bonferroni and false discovery rate (FDR, 5%) cor- comes investigated in randomized trials. J. Am. Heart Assoc. 8,
rected p values were calculated across 20 tests. e012738 (2019).
9. Seminog, O. O. & Goldacre, M. J. Gout as a risk factor for myocardial
Post hoc analyses. To investigate whether positive associations infarction and stroke in England: evidence from record linkage
between urate and a few IDPs could be the result of residual con- studies. Rheumatology 52, 2251–2259 (2013).
founding, models were stratified by income. After finding associations 10. Braskie, M. N., Ringman, J. M. & Thompson, P. M. Neuroimaging
between gout and cerebellum, midbrain and striatum, additional measures as endophenotypes in Alzheimer’s disease. Int. J. Alzhei-
relevant clinical outcomes were also examined post hoc: incidence of mer’s Dis. 2011, 490140 (2011).
Parkinson’s disease and incidence of other extrapyramidal movement 11. Latourte, A. et al. Uric acid and incident dementia over 12 years of
disorders (ICD10 G25). The latter is likely dominated by essential tre- follow-up: a population-based cohort study. Ann. Rheum. Dis. 77,
mor given its prevalence but also includes drug-induced tremor, 328–335 (2018).
myoclonus, and akathisia. We refer to this diagnosis henceforth as a 12. Kim, J. W. et al. Serum uric acid, Alzheimer-related brain changes,
probable essential tremor, although we recognise the heterogeneous and cognitive impairment. Front. Aging Neurosci. 12, 160 (2020).
nature of this entity. 13. Latourte, A., Dumurgier, J., Paquet, C. & Richette, P. Hyperuricemia,
gout, and the brain—an update. Curr. Rheumatol. Rep. 23,
Reporting summary 1–10 (2021).
Further information on research design is available in the Nature 14. Verhaaren, B. F. et al. The relation of uric acid to brain atrophy and
Portfolio Reporting Summary linked to this article. cognition: the Rotterdam Scan Study. Neuroepidemiology 41,
29–34 (2013).
Data availability 15. Tang, X., Song, Z.-H., Cardoso, M. A., Zhou, J.-B. & Simó, R. The
Full pseudonymized participant data cannot be openly shared under relationship between uric acid and brain health from observational
the material transfer agreement with UK Biobank and ethics approval. studies. Metabolic Brain Dis. 37, 1989–2003 (2022).
Other researchers can apply for UK Biobank data to answer specific 16. Guell, X., Gabrieli, J. D. & Schmahmann, J. D. Triple representation
research questions. Further information about applying for data of language, working memory, social and emotion processing in the
access can be obtained from the UK Biobank website (https://www. cerebellum: convergent evidence from task and seed-based rest-
ukbiobank.ac.uk) or by emailing UK Biobank (ukbiobank@ukbio- ing-state fMRI analyses in a single large cohort. Neuroimage 172,
bank.ac.uk). Genetic summary statistics for serum urate, gout and 437–449 (2018).
brain imaging measures are freely available. Source data are provided 17. Bagepally, B. S. et al. Decrease in cerebral and cerebellar gray
with this paper. matter in essential tremor: A voxel‐based morphometric analysis
under 3T MRI. J. Neuroimaging 22, 275–278 (2012).
Code availability 18. Pietracupa, S., Bologna, M., Tommasin, S., Berardelli, A. & Pantano,
No custom computer code or algorithm was used to generate the P. The contribution of neuroimaging to the understanding of
results reported in this paper R version 4.1.2 was used for most ana- essential tremor pathophysiology: a systematic review. Cerebellum
lyses (standard packages—survival v3.5-5, cmprsk v2.2-11, Mende- 21, 1–23 (2021).
lianRandomization v0.7.0, TwoSampleMR v0.4.26, forester 0.5.0, 19. Ayton, S. et al. Brain iron is associated with accelerated cognitive
ggforestplot available here: https://github.com/NightingaleHealth/ decline in people with Alzheimer pathology. Mol. Psychiatry 25,
ggforestplot). For VBM analyses, BLM was used (details in Supple- 2932–2941 (2020).
mentary Methods). Code for the VBM analyses performed is available 20. Thomas, G. E. C. et al. Brain iron deposition is linked with cognitive
at https://github.com/TomMaullin/BLMM. severity in Parkinson’s disease. J. Neurol. Neurosurg. Psychiatry 91,
418–425 (2020).
21. Topiwala, A. et al. Associations between moderate alcohol con-
References sumption, brain iron, and cognition in UK Biobank participants:
1. Kuo, C.-F., Grainge, M. J., Zhang, W. & Doherty, M. Global epide- observational and Mendelian randomization analyses. PLoS Med.
miology of gout: prevalence, incidence and risk factors. Nat. Rev. 19, e1004039 (2022).
Rheumatol. 11, 649–662 (2015). 22. Ou, Y.-N. et al. The association of serum uric acid level, gout, and
2. Borghi, C. et al. Hyperuricaemia and gout in cardiovascular, meta- Alzheimer’s disease: a bidirectional Mendelian randomization
bolic and kidney disease. Eur. J. Intern. Med. 80, 1–11 (2020). study. J. Alzheimer’s Dis. 89, 1–11 (2022).

Nature Communications | (2023)14:2844 7

Article https://doi.org/10.1038/s41467-023-38602-6

23. Albu, A., Para, I. & Porojan, M. Uric acid and arterial stiffness. Ther. 46. Burgess, S. & Labrecque, J. A. Mendelian randomization with a
Clin. risk Manag. 16, 39 (2020). binary exposure variable: interpretation and presentation of causal
24. Mitchell, G. F. et al. Arterial stiffness, pressure and flow pulsatility estimates. Eur. J. Epidemiol. 33, 947–952 (2018).
and brain structure and function: the Age, gene/environment 47. Allen, N. E., Sudlow, C., Peakman, T., Collins, R. & biobank, UK. UK
susceptibility–Reykjavik study. Brain 134, 3398–3407 (2011). biobank data: come and get it. Sci. Transl. Med. 6, 224ed224 (2014).
25. Ng, K. P. et al. Allopurinol is an independent determinant of 48. Miller, K. L. et al. Multimodal population brain imaging in the UK
improved arterial stiffness in chronic kidney disease: a cross- Biobank prospective epidemiological study. Nat. Neurosci. 19,
sectional study. PLoS ONE 9, e91961 (2014). 1523–1536 (2016).
26. Pierce, D. R., Williams, D. K. & Light, K. E. Purkinje cell vulnerability to 49. Alfaro-Almagro, F. et al. Image processing and Quality Control for
developmental ethanol exposure in the rat cerebellum. Alcohol the first 10,000 brain imaging datasets from UK Biobank. Neuro-
Clin. Exp. Res. 23, 1650–1659 (1999). image 166, 400–424 (2018).
27. Manto, M. Toxic agents causing cerebellar ataxias. Handb. Clin. 50. Griffanti, L. et al. BIANCA (Brain Intensity AbNormality Classification
Neurol. 103, 201–213 (2012). Algorithm): a new tool for automated segmentation of white matter
28. McFarland, N. R., Burdett, T., Desjardins, C. A., Frosch, M. P. & hyperintensities. Neuroimage 141, 191–205 (2016).
Schwarzschild, M. A. Postmortem brain levels of urate and pre- 51. Wang, C. et al. Phenotypic and genetic associations of quantitative
cursors in Parkinson’s disease and related disorders. Neurodegener. magnetic susceptibility in UK Biobank brain imaging. Nat. Neurosci.
Dis. 12, 189–198 (2013). 25, 818–831 (2021).
29. Fatima, T. et al. The relationship between ferritin and urate levels 52. Douaud, G. et al. Anatomically related grey and white matter
and risk of gout. Arthritis Res. Ther. 20, 1–9 (2018). abnormalities in adolescent-onset schizophrenia. Brain 130,
30. Yuan, S. & Larsson, S. Causal associations of iron status with gout 2375–2386 (2007).
and rheumatoid arthritis, but not with inflammatory bowel disease. 53. Tin, A. et al. Target genes, variants, tissues and transcriptional
Clin. Nutr. 39, 3119–3124 (2020). pathways influencing human serum urate levels. Nat. Genet. 51,
31. Min, K. H., Kang, S. O., Oh, S. J., Han, J. M. & Lee, K. E. Association 1459–1474 (2019).
between gout and dementia in the elderly: a nationwide population- 54. Sandoval-Plata, G., Morgan, K. & Abhishek, A. Variants in urate
based cohort study. Am. J. Geriatr. Psychiatry 29, 1177–1185 (2021). transporters, ADH1B, GCKR and MEPE genes associate with transi-
32. Lee, Y. H. Gout and the risk of Alzheimer′ s disease: a Mendelian tion from asymptomatic hyperuricaemia to gout: results of the first
randomization study. Int. J. Rheum. Dis. 22, 1046–1051 (2019). gout versus asymptomatic hyperuricaemia GWAS in Caucasians
33. McKhann, G. M. et al. The diagnosis of dementia due to Alzheimer’s using data from the UK Biobank. Ann. Rheum. Dis. 80,
disease: recommendations from the National Institute on Aging- 1220–1226 (2021).
Alzheimer’s Association workgroups on diagnostic guidelines for 55. Smith, S. M. et al. An expanded set of genome-wide association
Alzheimer’s disease. Alzheimer’s Dement. 7, 263–269 (2011). studies of brain imaging phenotypes in UK Biobank. Nat. Neurosci.
34. Wardlaw, J. M., Valdés Hernández, M. C. & Muñoz‐Maniega, S. What 24, 737–745 (2021).
are white matter hyperintensities made of? Relevance to vascular 56. Herndon, J. E. & Harrell, F. E. Jr The restricted cubic spline as
cognitive impairment. J. Am. Heart Assoc. 4, e001140 (2015). baseline hazard in the proportional hazards model with step func-
35. Duguid, J., De La Paz, R. & DeGroot, J. Magnetic resonance imaging tion time‐dependent covariables. Stat. Med. 14, 2119–2129 (1995).
of the midbrain in Parkinson’s disease. Ann. Neurol. 20, 57. Fine, J. P. & Gray, R. J. A proportional hazards model for the sub-
744–747 (1986). distribution of a competing risk. J. Am. Stat. Assoc. 94,
36. Fong, T. G., Davis, D., Growdon, M. E., Albuquerque, A. & Inouye, S. 496–509 (1999).
K. The interface between delirium and dementia in elderly adults. 58. Li, Q. et al. Diagnosis and treatment for hyperuricemia and gout: a
Lancet Neurol. 14, 823–832 (2015). systematic review of clinical practice guidelines and consensus
37. Stern, Y. et al. Whitepaper: defining and investigating cognitive statements. BMJ Open 9, e026677 (2019).
reserve, brain reserve, and brain maintenance. Alzheimer’s Dement.
16, 1305–1311 (2020). Acknowledgements
38. Merriman, T. R. & Joosten, L. A. CHIP and gout: trained immunity? We wish to thank Maria Christodoulou and the University of Oxford
Blood J. Am. Soc. Hematol. 140, 1054–1056 (2022). Statistics Consulting service for providing advice on the survival model
39. Yuan, H. & Yang, W. Genetically determined serum uric acid and analyses. We also thank the Wellcome Trust—216462/Z/19/Z (A.T.),
Alzheimer’s disease risk. J. Alzheimer’s Dis. 65, 1259–1265 (2018). 225790/Z/22/Z (S.Bu.), 206589/Z/17/Z (A.W.), 215573/Z/19/Z (S.S.),
40. Munafò, M. R., Tilling, K., Taylor, A. E., Evans, D. M. & Davey Smith, G. 202788/Z/16/Z (K.Mi.), 100309/Z/12/Z (T.N.); the UK MRC—G1001354
Collider scope: when selection bias can substantially influence (K.P.E.), G1001354 (K.P.E.); the European Commission—Horizon 2020
observed associations. Int. J. Epidemiol. 47, 226–235 (2018). 732592 (K.P.E.); the British Heart Foundation—RG/16/4/32218 (S.Be.); the
41. Munafò, M. R. & Davey Smith, G. Robust research needs many lines NIH—R01EB026859 (T.N.); EPSRC—EP/S02428X/1 (I.H.); and the NIHR—
of evidence. Nature 553, 399–401 (2018). BRC-1215-20014 (S.Be.); NIHR203312 (S.Bu.); BRC-1215-20014 (T.N.). The
42. Fry, A. et al. Comparison of sociodemographic and health-related funders had no role in study design, data collection and analysis, deci-
characteristics of UK Biobank participants with those of the general sion to publish or preparation of the manuscript.
population. Am. J. Epidemiol. 186, 1026–1034 (2017).
43. Shaffer, A., Rahn, E., Saag, K., Mudano, A. & Gaffo, A. Variation in Author contributions
serum urate levels in the absence of gout and urate lowering A.T. conceived of the study and carried out imaging and MR analyses.
therapy. BMC Rheumatol. 5, 1–7 (2021). C.W., F.A.A. and K.Mi. created relevant imaging data and confounds. T.N.
44. Topiwala, A., Ebmeier, K. P., Maullin-Sapey, T. & Nichols, T. E. and S.S. co-supervised the imaging analyses. S.Bu. and S.Be. co-
Alcohol consumption and MRI markers of brain structure and supervised the MR analyses. K.Ma., A.W., K.P.E., I.H., S.S. and T.N. con-
function: cohort study of 25,378 UK Biobank participants. Neuro- tributed to the interpretation of results. A.T. wrote the paper. All authors
Image Clin. 35, 103066 (2022). revised the paper.
45. Burgess, S. & Thompson, S. G., collaboration, C.C.g. Avoiding bias
from weak instruments in Mendelian randomization studies. Int. J. Competing interests
Epidemiol. 40, 755–764 (2011). The authors declare no competing interests.

Nature Communications | (2023)14:2844 8

Article https://doi.org/10.1038/s41467-023-38602-6

Additional information Open Access This article is licensed under a Creative Commons
Supplementary information The online version contains Attribution 4.0 International License, which permits use, sharing,
supplementary material available at adaptation, distribution and reproduction in any medium or format, as
https://doi.org/10.1038/s41467-023-38602-6. long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons license, and indicate if
Correspondence and requests for materials should be addressed to changes were made. The images or other third party material in this
Anya Topiwala. article are included in the article’s Creative Commons license, unless
indicated otherwise in a credit line to the material. If material is not
Peer review information Nature Communications thanks Chang-Fu Kuo, included in the article’s Creative Commons license and your intended
Johnathan Cooper-Knock and the other, anonymous, reviewer(s) for use is not permitted by statutory regulation or exceeds the permitted
their contribution to the peer review of this work. A peer review file is use, you will need to obtain permission directly from the copyright
available. holder. To view a copy of this license, visit http://creativecommons.org/
licenses/by/4.0/.
Reprints and permissions information is available at
http://www.nature.com/reprints © The Author(s) 2023

Publisher’s note Springer Nature remains neutral with regard to

jurisdictional claims in published maps and institutional affiliations.

Nature Communications | (2023)14:2844 9