Rheumatoid Arthritis and Cognitive Impairment or Dementia: An Updated Review of Epidemiological Data

Abstract

Rheumatoid arthritis (RA) is hypothesized to be associated with cognitive impairment and dementia, including Alzheimer’s disease, through shared biological processes related to inflammation. It is important to elucidate this potential relationship as both conditions confer increased morbidity and even mortality among older adults. This narrative review provides a survey of recent epidemiologic studies, examining the association between rheumatoid arthritis and either dementia or cognitive impairment. Sixteen studies were included after searching in PubMed and EMBASE. All were published between 2012 and 2022 and were characterized as epidemiologic studies (either cohort, cross-sectional, or case-control). Studies varied in location, design, measures of exposure and outcome, and covariates considered. Of the 16 studies included, only five found statistically significant positive associations between RA and dementia or cognitive impairment. One study found an inverse relationship, while five studies found no associations at all. The remaining five studies found variable statistically significant associations between demographic or RA disease characteristics and cognitive measures. Given these mixed findings, further studies at both the mechanistic and population level are needed to clarify the possible shared biological underpinnings of these two conditions.

Keywords

Aging Alzheimer’s disease cognition dementia rheumatoid arthritis

INTRODUCTION

In considering prevalent conditions affecting older individuals, there have been many efforts to elucidate the potential association between autoimmune pathology and dementia or cognitive impairment— including Alzheimer’s disease (AD). AD is the most common dementia, with an estimated 6.5 million Americans age 65 and older affected in 2022 [1]. It is pathophysiologically characterized by the extracellular accumulation of amyloid-β plaques, and intracellular accumulation of tau neurofibrillary tangles, leading to progressive neuronal damage over time with corresponding loss of function [1].

Some hypotheses suggest that systemic inflammatory pathology may be related to neuroinflammation through shared underlying biological processes. One such autoimmune entity associated with high levels of inflammation is rheumatoid arthritis (RA). RA is the most common form of inflammatory arthritis, characterized by autoimmune-mediated joint destruction and synovitis, resulting in both articular and extra-articular manifestations [2].

Mechanistically, both RA and AD are characterized by inflammation [3]. Both conditions are thought to involve hyperactivity of certain inflammatory cytokines, including IL-6, IL-1, and TNFα [3]. RA is also known to have significant effects on the vasculature, which may in turn contribute to vascular dementia through alterations in cerebral blood flow and increased risk for cerebral infarcts [4].

From an epidemiological perspective, case-control, cohort, and cross-sectional studies have been conducted to assess whether RA predisposes individuals to develop cognitive impairment or dementia. A previous systematic review and meta-analysis by Ungprasert et al. in 2016 revealed a higher pooled risk of dementia among patients with RA, with a relative risk of 1.61 and 95% confidence interval of 1.10-2.37 [5]. However, this review included only five studies in its meta-analysis. Another systematic review and meta-analysis with Mendelian randomization by Policicchio et al. in 2017 explored studies on RA in relation to AD specifically, with contradicting results. Meta-analysis revealed lower risk of AD among those with RA (odds ratio 0.60, 95% confidence interval 0.46-0.77) [6]. It is possible that other factors, such as RA treatment, may affect this relationship and account for the mixed associations in prior studies.

Existing reviews focus more on mechanistic links between RA and AD or cognition, and tend to include older studies from the 1990 s and early 2000 s. In this narrative review of more current literature, we provide an updated summary of epidemiological data related to the association between RA and both dementia (including AD) as well as general cognitive impairment among various populations.

METHODS

Both PubMed and EMBASE were utilized to search for epidemiological studies (either case-control, cross-sectional, or cohort) within the last 10 years (2012-2022) using the following search strategy: “Rheumatoid arthritis AND (dementia OR Alzheimer’s disease).” The timeframe of the last 10 years was chosen in order to provide a more updated review of the subject; as mentioned previously, prior reviews included older studies, or only few studies within the past decade [5, 6], The above search yielded 445 results in PubMed, and 1,042 results in EMBASE. Articles were included if they met the criteria noted above based on title and abstract, and if they included odds ratios (OR), hazard ratios (HR), relative risk (RR), or other statistical estimates related to the specific exposure of RA in relation to dementia or other cognitive measure. Studies that explored other autoimmune conditions in conjunction with RA were also included (unless they did not report a separate, RA-only estimate). With regards to the outcome, studies that examined all dementias (including AD) as well as other general measures of cognitive impairment were included. Studies that specifically examined the role of treatment were not a focus of our primary review, but several are described in the Discussion with a more detailed summary represented in Supplementary Table 1.

RESULTS

See Tables 1 and 2 for details on individual studies examined. Table 1 contains studies with an outcome measure of dementia; Table 2 contains studies with a more general outcome measure of cognitive performance or cognitive impairment.

Table 1

Main summary of studies considered with dementia as primary outcome

Authors	Year	Country	Study Population	Study Design	Measure of RA	Measure of dementia or cognitive impairment	Covariates or other features examined	Main association for RA and dementia/CI
Kronzer et al. [7]	2021	USA	897 participants with incident RA between 1980 and 2009, 885 controls; matched on age, sex, and calendar year, via Rochester Epidemiology Project (REP). 67% seropositive (positive RF or anti-CCP) among RA cases. Among RA cases (among all decades), 36% with MTX use in first year, 45% with hydroxychloroquine use, 15% with other csDMARD use, 4% with biologic use.	Retrospective cohort	RA diagnosis by ACR criteria (1987), between 1980 and 2009	2 codes for AD or ADRD noted at least 30 days apart. Follow-up time: median of 15 years for RA participants and 16 years for non-RA participants.	Examined demographic features, as well as smoking status, treatment history, comorbidities, biomarkers (ESR, anti-CCP, RF) Cox proportional hazard models for incident ADRD by decade, adjusting for age and sex.	Overall adjusted HR of incident ADRD among those with RA: 1.37 (1.04-1.81) Increased risk for ADRD among those with RA in 1980 s and 1990 s, but not 2000 s.
Kodishala et al. [13]	2022	USA	886 patients with RA diagnosed between 1980-2014, via Rochester Epidemiology Project (REP). 62% RF or anti-CCP positive. 47.5% with MTX use, 40.1% with hydroxychloroquine use, 15.8% with other DMARD use, 6.3% with biologic use.	Retrospective cohort	Diagnosis of RA by ACR (1987) criteria between 1980 and 2014	2 relevant ICD 9 or 10 codes for dementia, at least 30 days apart. Follow-up time: median of 8.5 years.	RA disease characteristics examined: erosions, swelling, surgeries, extraarticular features, seropositivity (RF or anti-CCP). Also examined treatment history, CVD risk factors/events, depression/anxiety. 2 Cox proportional hazard models for each risk factor: Model 1 adjusted for age, sex, and calendar year of RA incidence; Model 2 adjusted additionally for smoking, obesity, diabetes, hypertension, hyperlipidemia, CVD.	Adjusted HRs for incident dementia related to RA-specific characteristics: rheumatoid nodules at baseline, 2.02 (1.17-3.50), large joint swelling at baseline, 2.26 (1.35-3.79) or ever, 2.41 (1.22-4.73). Other significant positively associated risk factors: age, baseline hypertension, CVD conditions/risk factors (overall, ischemic stroke, and heart failure), anxiety and depression.
Lin et al.^* [8]	2018	Taiwan	19,556 patients in RA subgroup, from Taiwan National Health Insurance (NHI) database between 2001-2012. Comparison cohort from general population via random frequency-matching by age, sex, and index year of rheumatic disease diagnosis.	Retrospective cohort	RA diagnosis by 1987 American Rheumatism Association criteria and ICD-9 code	Dementia diagnosis by a neurologist during two outpatient visits or one hospital admission. Dementia types included AD, arteriosclerotic, and unspecified dementia (with corresponding ICD-9 codes). Follow-up time: median of 5.97 years for autoimmune rheumatic disease cohort; 6.37 years for non-rheumatic disease cohort.	Cox proportional hazards model with adjustment for age, sex, and comorbidities. Separate stratified analysis with age groups < 65 and > 65.	Adjusted HR for dementia incidence in RA-subgroup compared to comparison group: 1.14 (1.06-1.32). Non-significant association in < 65 age group; HR 1.15 (1.06-1.26) in > 65 age group.
Wotton et al.^* [9]	2017	UK	National hospital and mortality administrative data between 1999-2012; 316,043 patients in RA cohort (with corresponding control cohort)	Retrospective cohort	First recorded day case or inpatient admission for RA coded in any diagnostic field on discharge summary (ICD-10 codes)	Subsequent hospital care for or death from dementia based on ICD-9 or ICD-10 codes on discharge summary; further subdivision of AD versus vascular dementia. Follow-up time: median of 5.3 years.	Stratified and standardized by age (5-year groups), sex, calendar year of first recorded admission, region of residence, and quintile of Index of Deprivation score	Stratum-standardized adjusted rate ratio for vascular dementia post-RA admission: 1.16 (1.12-1.20) Rate ratio for AD post-RA: 0.89 (0.86-0.93)
Kao et al. [12]	2016	Taiwan	2,271 patients with AD who had received prescriptions for acetylcholinesterase inhibitors, 6,813 controls without AD. Cases and controls matched on age group, sex, and year of index date (date of first receiving treatment)	Case-control nested in a cohort	RA diagnosis by ICD-9 codes (two or more RA diagnoses, including by rheumatologist), and prescription of at least one DMARD	AD diagnosed between 2001-2013 (ICD-9 codes), with prescriptions for acetylcholinesterase inhibitors	Stratification by age group, sex, year of index date. Adjustment for geographic location, urbanization level, comorbidities.	Adjusted OR for prior RA among AD cases compared to controls: 0.73 (0.55-0.98).
Min et al. [33]	2020	Korea	Study 1: participants with dementia (11,443) matched to controls (45,772), analyzed history of RA. Study 2: participants with RA (4,420) matched to controls (17,680), analyzed occurrence of dementia. Cases and controls matched on age, sex, income, region of residence, medical history. Used Korean National Health Insurance Service sample cohort. 99.1% of RA patients prescribed DMARDs.	Nested case-control (Study 1) and retrospective follow-up study (Study 2)	RA diagnosis by ICD-10 codes and medication history (prescription for biologics or DMARD)	Neurodegenerative dementia diagnosed by ICD-10 codes (AD or dementia in AD) and treatment 2 or more times	Logistic regression (Study 1) or Cox proportional hazards (Study 2) with stratification by age, sex, income, region of residence, hypertension, diabetes and dyslipidemia. Adjustment for history of ischemic heart disease, cerebral stroke, depression, Sjogren’s syndrome. Subgroup analyses based on age and sex.	Study 1 adjusted OR for RA in dementia group versus controls: 0.96 (0.78-1.16). Study 2 adjusted HR for dementia in RA group versus controls: 0.91 (0.76-1.10).
Park et al.^* [34]	2021	Korea	6028 patients with dementia, 30,140 controls (matched by age and sex), from Korean Health Insurance Service claims database	Case-control nested in a cohort	RA diagnosis by ICD code (at least 6 months prior to dementia diagnosis) and seropositivity (RF or anti-CCP)	Dementia defined as acetylcholinesterase inhibitor prescription, along with symptoms meeting AD ICD-10 code or vascular dementia code	Multivariate logistic regression with adjustment for income, residence, comorbidities, age, and sex. Subgroup analyses with age groups, dementia subtype, and dementia severity.	Adjusted OR for dementia in RA-subgroup: 0.96 (0.83-1.11)
Chen et al.* [10]	2018	Taiwan	10,180 patients with dementia, 61,080 controls (matched on age, gender, index date of first dementia diagnosis); from Taiwan National Health Insurance (NHI) Research Database (3516 patients with RA)	Case-control	RA diagnosis by ICD-9-CM code, and at least one outpatient visit for RA prior to index date of dementia.	Dementia diagnosed with ICD-9 code between 2000-2010, with 2 or more outpatient visits or one or more hospital admission.	Multivariable conditional logistic regression adjusting for age, sex, and comorbidities.	Adjusted OR for association between rheumatic disease and dementia in RA subgroup 1.11 (1.10-1.22) Non-significant when stratified by age groups (40-64 versus 65 or older) or gender.
Vassilaki et al. [18]	2022	USA	Mayo Clinic Study of Aging: participants with RA (104) were matched to controls (312) on age, sex, education, and baseline cognitive diagnosis. RA cases with available MRI also matched with non-RA cases with available MRI. 70% RF-positive or anti-CCP-positive among RA cases. 55% with MTX use, 42% with hydroxychloroquine use, 23% with biologic DMARD use, 24% with other csDMARD use.	Longitudinal cohort study	RA diagnosis based on ACR criteria (1987 and/or 2010), or physician diagnosis between 2005-2020	Nine neuropsychological tests to assess 4 domains: memory, language, attention/recall, visuospatial. Consensus determination of cognitively unimpaired (CU), MCI, or dementia based on study coordinator, physician, and neuropsychologist. Follow-up time: mean of 3.9 years for RA patients and 4.3 years for non-RA participants.	Examined neuroimaging studies (PET, MRI), RA biomarkers (ESR, RF, anti-CCP), RA treatment history. Cox proportional hazards adjusted for age, sex, education, baseline cognitive status for association of RA/activity markers with progression to dementia. Linear mixed-effects model (adjusted for age, sex, education, baseline cognitive diagnosis and cognitive test-naïve status) to examine RA and change in global cognition among CU participants.	Adjusted HR for incident dementia in RA patients versus non-RA (in those who were cognitively unimpaired/MCI at study baseline): 0.96 (0.5-1.7). Adjusted HR for incident dementia in RA patients versus non-RA (in those who were cognitively unimpaired at study baseline): 1.0 (0.4- 2.3).
Booth et al.^* [35]	2021	USA	2842 adults over 50 with Medicare Health and Retirement Study data, from 2006-2014 (73 subjects in RA subgroup)	Longitudinal cohort	RA diagnosis from Medicare claims using ICD9-CM codes (two codes at least 1 day apart), from 2006-2009. “RA” included all RA-related diagnoses (including Felty’s, inflammatory polyarthropathy, etc.)	Incident AD diagnosis from Chronic Conditions Warehouse (CCW) identifiers— single ICD9-CM diagnosis within a 3-year observation window. Average follow-up time: 4.93 years.	No adjusted Cox model for RA-specific analyses due to small subgroup sizes.	Unadjusted HR for AD in RA-subgroup: 1.45 (0.77-2.74)

^aStudies noted with “*” included multiple autoimmune/rheumatic conditions as exposures, but reported RA-specific estimates and were therefore included. ^bRA, rheumatoid arthritis; AD, Alzheimer’s disease; ADRD, Alzheimer’s disease related dementias; ESR, erythrocyte sedimentation rate; CCP, cyclic citrullinated peptide; RF, rheumatoid factor; HR, hazard ratio; ACR, American College of Rheumatology; ICD, International Classification of Diseases; CVD, cardiovascular disease; DMARD, disease-modifying anti-rheumatic drug; OR, odds ratio; MTX, methotrexate; MRI, magnetic resonance imaging; MCI, mild cognitive impairment; PET, positron emission tomography.

Table 2

Main summary of studies considered with general cognition/cognitive impairment as primary outcome

Authors	Year	Country	Study Population	Study Design	Measure of RA	Measure of dementia or cognitive impairment	Covariates or other features examined	Main association for RA and dementia/CI
Gwinnutt et al. [14]	2021	UK	38 people with RA (from the Norfolk Arthritis Register), 28 age-matched healthy controls (no cognitive impairment)	Cross-sectional	RA diagnosis based on 2010 ACR/EULAR criteria	Addenbrooke’s Cognitive Examination (ACE) III, Institute of Cognitive Neurology Frontal Screening (IFS), mini-Social Cognition and Emotion Assessment (SEA), Rey Complex Figure Test (RCFT)	Age, sex, BMI, smoking status, education, anxiety, and depression	Adjusted standard mean difference between RA and controls, ACE-III score: -0.9 (-1.2, -0.6) IFS score: -1.1 (-1.5, -0.7) Mini-SEA, facial recognition: -0.7 (-1.2, -0.2) No statistically significant difference in RCFT.
Lee et al. [16]	2018	Korea	70 RA patients, 40 healthy controls. Among RA patients: 27% treated with MTX, 27% with sulfasalazine, 39% with hydroxychloroquine, 20% with leflunomide, 20% with TNF inhibitors, 7% with tocilizumab, and 1% with tofacitinib.	Cross-sectional	RA diagnosis based on ACR criteria	CERAD-K neuropsychological battery total scores (consists of 9 other tests encompassing various cognitive domains)	Examined carotid plaques and IMT; DAS28-ESR; biomarkers (RF, anti-CCP, hsCRP)	Multivariable regression analysis: negative correlation between age, DAS28 scores and CERAD-K total scores among those with RA. Overall RA patients showed more cognitive impairment compared to controls on various components of CERAD-K (mean scores for verbal fluency: 12.97 in those with RA versus 15.48 in controls; for Boston Naming Test: 11.50 versus 12.30; for total CERAD-K score: 70.93 versus 75.73; all p-values<0.05). No association between cognitive function and carotid atherosclerotic changes.
Olah et al. [15]	2019	Hungary	99 female patients (60 with steroid, MTX, or biologic-managed RA, 39 healthy controls). 70% RF-positive; 65% ACPA-positive among all RA patients.	Cross-sectional	Diagnosed RA patients with regular follow-up and at least 6 months of therapy prior to study	MoCA, TMT, VST, WAIS, BVRT	Examined RA disease activity (biomarkers, disease duration), education, smoking, alcohol consumption, as well as general health. Also investigated depression/anxiety, cerebrovascular pathology.	Multiple linear regression analysis of RA patients: age, school years, MRI lesions, and use of biologics versus MTX were independently associated with various cognitive tests. Overall, RA patients with cognitive impairment on multiple tests (mean scores for MoCA: 23.3 in those with RA versus 25.6 in controls; for WAIS: 25.4 versus 28.5; for BVRT: 6.58 versus 7.67; all p-values<0.05).
Vitturi et al. [17]	2019	Brazil	210 patients with RA (followed from 2016-2018 at a rheumatology clinic), 70 controls. 76% RF-positive among RA patients. Among RA patients, 28% treated with biologics, 73% with MTX, 7% with sulfasalazine, 44% with leflunomide, 16% with chloroquine.	Cross-sectional/ case-control	RA diagnosis based on ACR criteria	MMSE, MoCA	Examined neuropsychiatric impairment (HADS), baseline functional limitation (HAQ), education, comorbidities, smoking, alcohol, RA disease activity (DAS28 ESR, CRP, medications), RF.	Multivariate analysis, MMSE: biological therapy associated with decreased odds of impairment (OR = 0.5, 95% confidence interval 0.3-0.9); MoCA: oral glucocorticoids and positive RF with higher odds of impairment (ORs/95% confidence interval 7 (1.1-43.6), 13.8 (1.5-126.8) respectively). Logistic regression: older patients with first symptoms at older age more likely to have cognitive impairment on MoCA.
Booth et al. [36]	2021	USA	4,462 participants in the 2016 wave of US Health and Retirement Study with linked Medicare claims	Cross-sectional	RA diagnosis based on ICD-9CM or ICD-10 codes (minimum two Medicare billing diagnoses)	Langa-Weir classification or proxy questionnaire with classifications of normal, cognitive impairment without dementia, and dementia	Adjustment for age, gender, education, and race	Adjusted OR of cognitive impairment among those with RA versus controls: 1.08 (0.74-1.59)
Wallin et al. [11]	2012	Finland	1,449 participants in the Cardiovascular risk factors, Aging and Incidence of Dementia (CAIDE) study, evaluated between 1972-1987. Analyses performed on 1,368 participants with no missing data. 13% of participants with NSAID or steroid use.	Prospective cohort	Baseline survey inquiring about RA or other joint disorder diagnosis within the prior year	Three-stage cognitive assessment at re-examination in 1998: including screening phase (MMSE), further medical/ neuropsychological/ neuroimaging evaluation, and final re-analysis and diagnosis by a review board. MCI identified with modified Mayo Clinic AD Research Center criteria, dementia diagnosed with DSM-IV criteria, AD diagnosed with US NINCDS-ADRDA criteria. Mean follow-up time: 20.9 years.	Ordinal logistic regression; Model 1 adjusted for age, gender and follow-up time, Model 2 additionally for education and at least one APOEɛ4 allele, Model 3 additionally for BMI, smoking, and the use of anti-inflammatory medications, Model 4 additionally for later-life diabetes.	Adjusted OR (Model 4) of worse cognitive status later in life among those reporting midlife RA: 2.77 (1.26-6.10). Similar results when considering AD specifically instead of dementia; adjusted OR (Model 4): 2.49 (1.09-5.67)

^aStudies noted with “*” included multiple autoimmune/rheumatic conditions as exposures, but reported RA-specific estimates and were therefore included. ^bRA, rheumatoid arthritis; CI, cognitive impairment; ESR, erythrocyte sedimentation rate; CCP, cyclic citrullinated peptide; RF, rheumatoid factor; ACR, American College of Rheumatology; ICD, International Classification of Diseases; EULAR, European Alliance of Associations for Rheumatology; BMI, body mass index; OR, odds ratio; MTX, methotrexate; TNF, tumor necrosis factor; ACPA, anti-citrullinated protein antibodies; MoCA, Montreal cognitive assessment; TMT, Trail making test; VST, Victoria Stroop test; WAIS, Wechsler adult intelligence scale; BVRT, Benton visual retention test; MRI, magnetic resonance imaging; CERAD, Consortium to establish a registry for Alzheimer’s disease; IMT, intima media thickness; DAS28, Disease activity score; MMSE, Mini-Mental Status Examination; HADS, Hospital anxiety and depression scale; HAQ, Health assessment questionnaire; CRP, C-reactive protein; MCI, mild cognitive impairment; AD, Alzheimer’s disease; DSM, Diagnostic and statistical manual of mental disorders; NINCDS-ADRDA, National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association; NSAID, non-steroidal anti-inflammatory drug.

Studies were epidemiological in nature and published between 2012-2022; 16 studies in total were selected. Notably, studies differed in their design, definitions for both exposure (RA diagnosis) and outcome (dementia diagnosis or measure of cognitive impairment), and covariates. With respect to study design, of the studies included, four were determined to be “retrospective cohort studies,” one was a “prospective cohort,” three were “case-control,” four were “cross sectional,” one was “cross-sectional/case-control,” two were “longitudinal” cohort studies, and one included both a “case-control” and “retrospective” study component. With respect to the exposure measure (RA), five studies utilized the American College of Rheumatology (ACR) criteria, two utilized ACR criteria in addition to another measure (ex. physician diagnosis or International Classification of Diseases, ICD, code), three utilized ICD-9 or ICD-10 codes exclusively, four utilized ICD codes in addition to another measure (ex. medication history, seropositivity, or outpatient visit), one used a survey and self-report of RA diagnosis, and one did not specify the RA diagnostic measure. With respect to the outcome measure, nine studies identified cases of dementia using ICD codes and a combination of either hospital admissions/outpatient visits or treatment history. Of these, four studies specifically included cases of AD or Alzheimer’s disease and related dementias (ADRD), whereas the others included other types of dementia (including vascular). Six studies utilized cognitive assessments as their primary outcome measure. One study included both cognitive assessments and formal diagnostic criteria to characterize degrees of cognitive impairment.

Of the 16 studies included, only five found statistically significant positive associations between RA and the outcome measure of dementia. For example, Kronzer et al [7] found an adjusted HR (95% confidence interval) of ADRD incidence to be 1.37 (1.04-1.81) among those with RA compared to those who did not have RA, while Lin et al. [8] found an adjusted HR (95% CI) of dementia incidence to be 1.14 (1.06-1.32) associated with RA. Wotton et al. found an increased adjusted rate ratio of 1.16 (1.12-1.20) for vascular dementia post-RA admission, but interestingly found an inverse association for AD (adjusted rate ratio of 0.89, 95% CI 0.86-0.93) [9]. Chen et al. found a statistically significant positive association between RA and dementia, with an adjusted OR of 1.11 (1.10-1.22) [10]. Finally, Wallin et al. [11] found increased odds of worse cognitive outcomes later in life among those with reported RA at midlife (adjusted OR 2.77 (1.26-6.10)). This association remained when the outcome was specific to AD; adjusted OR 2.49 (1.09-5.67).

Of the studies reviewed, only one study conducted by Kao et al. found an inverse relationship between dementia and RA, with an OR (95% CI) of 0.73 (0.55-0.98) for prior RA among those with dementia [12].

Kodishala et al. examined various RA disease characteristics in relation to dementia, with positive associations found between the presence of rheumatoid nodules or large-joint swelling and the outcome [13].

Other studies focused more on a more general outcome measure of cognitive function or cognitive impairment. When exploring mean differences on various cognitive tests, including the Addenbrooke Cognitive Examination, the Institute for Cognitive Neurology Frontal Screening, and the mini-Social Cognition and Emotion Assessment, Gwinnutt et al. found statistically significant decreases in scores among individuals with RA compared to controls [14]. Olah, Lee, and Vitturi et al. conducted similar studies examining various demographic factors, RA disease-related characteristics, and treatment histories in relation to cognitive tests, all with significant findings [15 –17].

The remaining studies (five) found no apparent association between RA or RA-related factors and dementia or cognitive outcomes.

To briefly address the role of RA treatment on dementia or cognition, seven epidemiological studies exploring this association were summarized and included in the Discussion. Of note, the selection of these studies was largely separate from that of the main studies reviewed. These chosen studies do not represent a comprehensive summary of this particular topic and did not adhere to a particular search strategy. Their specific results are described in Supplementary Table 1.

DISCUSSION

Broad differences in studies examined

As noted previously, studies included in the review differed in several ways. One major difference was study design— while all were epidemiological, some were characterized as cohort studies (either retrospective or prospective), while others were structured as case-control or cross-sectional studies. Cohort studies may afford stronger evidence than cross-sectional study designs, since the exposures are typically measured over an extended period of time prior to disease occurrences. Prospective cohort studies in particular, if well-conducted, can provide findings with high level of validity due to the temporal sequence to observe effects after exposure. Among cohort studies included in this review, follow-up periods were variably documented. Often times, the exposure (RA) was simply noted as existing prior to documented index date of the outcome (AD, dementia). However, some studies, including Kodishala et al., Vassilaki et al., Wallin et al., and Kronzer et al., more explicitly noted follow-up periods (in years) among RA and non-RA individuals [7 , 18]. Follow-up time is an important parameter, as cognitive impairment has a long preclinical stage (up to decades) and insidious disease onset prior to overt clinical symptoms [19]. In addition, greater RA symptom duration and/or management with treatment may affect the magnitude of its possible association with dementia and cognition [3].

Studies also differed in their measurement of the exposure (RA) and outcome (dementia, including AD, versus cognition or cognitive impairment in general). Considering the exposure measure, RA, most studies used the American College of Rheumatology (ACR) criteria, which includes symptoms such as morning stiffness, arthritis of three or more joint areas, arthritis of hand joints, symmetric arthritis, rheumatoid nodules, serum rheumatoid factor (RF), and radiographic changes— with satisfaction of 4/7 criteria warranting a diagnosis of RA [20]. One study (Gwinnutt et al.) used the updated 2010 ACR/European League Against Rheumatism (EULAR) criteria, which includes broad categories of joint involvement, serology, acute phase reactants, and duration of symptoms [20]. Several studies utilized the ACR criteria in addition to ICD codes, while others utilized ICD codes exclusively. Still other studies additionally included outpatient visits, seropositivity, or prescribed medications as components of determining RA status. Although most studies were overall fairly similar in their method for defining subjects with RA, in cohort studies, it is possible that differences in treatment history or other disease characteristics (such as comorbid conditions) may be related to both dementia and treatment, leading to potential selection bias or detection bias away from the null. In case-control studies, if both cases of dementia and controls were ascertained from hospital records, there is a risk of Berkson’s bias that could be away or towards the null. These potential biases may also contribute to some of the differential associations between RA and dementia as evidenced between the studies [21].

Considering the outcome measure further, some studies measured dementia, whereas others utilized cognition/cognitive impairment as evidenced by various tests. One consideration in defining this outcome is the fact that changes in cognition exist across a wide spectrum, influenced by practical factors in studies such as follow-up time. This is because dementia, as mentioned, can be an insidious process, with pathological changes often occurring several years before evidence of overt clinical disease [19]. One can argue that it is therefore important to consider even prodromal changes during the follow-up period, including mild cognitive impairment and even subjective memory complaints, as possible outcome measures that may be affected more gradually or subtly by systemic autoimmunity. Future studies may therefore consider preclinical changes in cognition, with greater attention to follow-up time between RA diagnosis and cognitive changes. Furthermore, some studies utilized other variables in addition to ICD codes or physician diagnosis to qualify the dementia outcome measure, including treatment history, hospital admissions, or outpatient visits. Such factors may introduce an additional component of dementia severity that may contribute to the variation in findings and associations. Including other objective measures, such as biomarkers and imaging findings, may also help to better quantify the outcome measure [19, 22]. One study, the Mayo Clinic Study of Aging, included imaging modalities, with Pittsburgh compound-B positron emission tomography and magnetic resonance imaging to examine global amyloid-β and cerebrovascular pathology, respectively [18].

Finally, studies varied in the relative incorporation of covariates in regression models or other statistical analyses. Most studies adjusted for demographic characteristics such as age, sex, and comorbidities (often including vascular pathology, as well as anxiety/depression); several also adjusted for education level, and socioeconomic factors including area of residence. The differences in covariates included may therefore also explain some of the variance in findings between the studies, and the differential susceptibility of some studies more than others to residual confounding. It is also important to note that some studies had the added strength of matching cases and controls on various characteristics— allowing for better mitigation of confounding from the start.

THE ROLE OF RA TREATMENT

While RA treatment was not a primary focus of the studies summarized in this review, given the recognition of inflammation as a key component of both RA and dementia, there have been several theories about the role of anti-inflammatory agents (including non-steroidal anti-inflammatory drugs or NSAIDs, disease-modifying antirheumatic drugs or DMARDs, biologics, and steroids) in mediating this primary relationship. Certain existing studies have presented mixed findings (Supplementary Table 1). For example, one case-control study examining the risk of AD among RA patients prescribed various modalities of RA treatment found a lower risk of AD among RA patients with exposure to anti-tumor necrosis factor (anti-TNF) agents as a class (adjusted OR 0.45, 95% CI 0.23-0.90), with subgroup analyses specifically finding a decreased AD risk among those using etanercept. This association remained even with adjustment for comorbidities that my confer increased risk for vasculopathy, including coronary artery disease, diabetes, and peripheral vascular disease [23]. These results were supported by another recent retrospective case-control study, with decreased risk of AD among those prescribed etanercept (adjusted OR 0.34, 95% confidence interval 0.25-0.47), adalimumab (0.28, 0.19-0.39), and infliximab (0.52, 0.39-0.69). Protective effects were also noted among those treated with methotrexate (0.64, 0.61-0.68) [24].

Yet another retrospective cohort study utilizing two separate insurance claims databases and directly comparing TNFα inhibitors to methotrexate among RA patients found no apparent difference between the two groups in terms of risk for AD or dementia [25]. With regards to methotrexate specifically, a UK-based cohort study exploring various conventional DMARDs (cDMARDs) found that cDMARD use was associated with a general reduction in dementia risk; furthermore, this effect was most pronounced among methotrexate users (overall HR with cDMARD use: 0.52, 95% CI 0.42-0.85) [26]. DMARD users were propensity score-matched to non-users on several confounders, including patient characteristics, comorbidities, and medications [26]. Yet another study by Chou et al. (2017) utilizing the Taiwan National Health Insurance (NHI) data showed an increase in dementia risk (specifically, vascular dementia) among RA patients who used conventional synthetic DMARDs (csDMARDs) (OR 1.63, 95% CI 1.33-2.00); such an association was not found among those who used biologic DMARDs (bDMARDs) or NSAIDs [27]. Interestingly, this study specifically found an increased risk of dementia among those who used methotrexate, in contrast to the aforementioned UK study.

NSAIDs have specifically been an area of curiosity, given their role in managing systemic inflammation in RA and reducing symptom burden. For example, a Taiwanese retrospective cohort study examining the risk of dementia in patients with RA found no statistically significant difference between RA and non-RA patients in terms of dementia incidence, but did find an apparent decreased risk of dementia among RA patients using NSAIDs for > 2191 days compared to a reference group of < 730 days (adjusted HR 0.56, 95% CI 0.45-0.68, with adjustment for age, sex, and various comorbidities) [28]. In contrast to NSAIDs, steroids (another mainstay of RA management) have been shown to have primarily negative effects on cognition, with a recent meta-analysis showing modest deleterious effects on executive function, recent memory, and very long-term memory, specifically [29].

In an effort to combat many of the biases to which observational pharmacoepidemiology studies are prone— including immortal time bias, reverse causation, and confounding by indication— the recent Drug Repurposing for Effective Alzheimer Medicines (DREAM) initiative has specifically sought to consider these limitations when investigating the role of RA treatment on dementia. As part of this initiative, a US cohort study by Desai et al. (2022) examined Medicare fee-for-service data on patients with RA, comparing three DMARDs to an active comparator of abatacept (a T-cell activation inhibitor) with respect to risk of ADRD. The three DMARDs chosen included a Janus-kinase (JAK) inhibitor (tofacitinib), an IL-6 inhibitor (tocilizumab), and TNF inhibitors. Four separate analytic schemes were chosen to separately mitigate the effects of the aforementioned biases, with no reduction in the risk of ADRD for patients prescribed any of the three DMARD classes compared to abatacept across all analyses [30].

Finally, another important consideration when evaluating the effect of RA treatment on dementia or cognitive outcomes is blood-brain barrier permeability; for example, TNF inhibitors are too large to cross the blood-brain barrier, while methotrexate has little penetrance [23, 26]. It is possible that peripheral actions and inflammatory regulation may be the primary method of action for these agents, though further mechanistic studies are certainly warranted to fully examine this hypothesis.

As mentioned previously, the supplemental studies (see Supplementary Table 1) specifically examining RA treatment included in this review were not selected through a standardized search process, and therefore do not represent a comprehensive survey of the literature on this particular subject.

CONSIDERING COMORBIDITIES

Apart from shared inflammatory mechanisms, conditions comorbid with RA may also impact cognition and dementia risk, several of which were included as covariates or other variables of interest in the summarized studies in this review. For example, it is known that RA is associated with increased cardiovascular (CV) risk and mortality. In turn, CV pathology and risk factors themselves (including stroke, coronary artery disease, hypertension, and diabetes) increase risk for dementia, both AD and vascular [31]. RA patients with CVD or CVD risk factors may be more susceptible to dementia. A recent study by Sattui et al. (2021) examined this association using claims-based data comparing RA patients without cardiovascular disease (CVD) or CV risk factors to those with them, exploring the incidence of all-cause dementia [31]. CVD was defined as myocardial infarction, heart failure, stroke, or CVD procedures such as coronary artery bypass graft, percutaneous coronary intervention, and carotid endarterectomy. CV risk factors included diabetes, hyperlipidemia, hypertension, and obesity at baseline. The study found an increased risk of dementia among RA patients with CV risk factors (but without CVD) compared to the reference group of those without either CVD or CV risk factors (adjusted HR 1.03, 95% CI 1.00-1.11). Similarly, there was an increased risk of dementia among RA patients with CVD (adjusted HR 1.18, 95% CI 1.04-1.33) compared to the reference group of those without either CVD or CVD risk factors— in models adjusted for demographics, comorbidities, medications, and ambulatory clinical encounters. Notably, this association was only observed among the 65-74 age group, and was not statistically significant among the 75-84 or 85 + age groups [31].

POSSIBLE BIASES AND ADDITIONAL CONSIDERATIONS

Given the differences in the studies, it is important to consider various biases and additional confounders that may have affected results and interpretation. One consideration is the possible impact of detection bias. This relates to the differential measure of the outcome (cognitive impairment or dementia) between the various studies. It also relates closely to the exposure, as there is a possibility that closer follow-up for RA management or treatment may confer a greater possibility of also detecting incident dementia or changes in cognition, for example [32].

Another consideration is the issue of confounding in relation to treatment for RA. It is possible that those receiving treatment may have a more severe form of disease that could artificially exaggerate the nature/directionality of the association observed between the exposure and outcome. Additionally, treatment may affect the etiology of observed associations— for example, patients requiring treatment for more severe RA may also be predisposed to greater vascular insults (in addition to inflammatory-mediated pathology) leading to impaired cognition. This may complicate the hypothesized pathophysiologic relationship between the exposure and outcome, even if it does not affect the underlying nature of the association. Conversely, treatment may also improve control of the disease and actually mitigate some effects on cognition that may have otherwise been observed.

CONCLUSION

Overall, the studies included in this review reveal a mixed picture of the association between RA and dementia or general cognitive impairment. While many studies did observe statistically significant positive associations between the variables of interest, others found inverse or no associations. Still others found apparent differences based on stratification by different parameters, including age and gender, or when considering specific disease-related or demographic risk factors. Some of this heterogeneity may be explained by intrinsic differences in study design, covariates examined, and definitions of both the exposure and outcome. Future studies should seek to expand on existing work through incorporation of more standardized objective measures of RA and dementia, with further accounting for measures of disease severity and treatment. The use of biomarkers and imaging may also provide an area of continued growth in this work. Additionally, the breadth of our outcome measure is limited by our search strategy focusing on dementia and AD, and we might not have gathered representative studies of other cognitive measures. Finally, while our narrative review provides a comprehensive qualitative summary of this important subject, more rigorous study designs, including meta-analyses, can shed more light on quantitative results.

As both RA and dementia are critical issues affecting older individuals— with a great degree of associated morbidity and even mortality— continued efforts to elucidate the possible relationship between these conditions will yield an enormous benefit for a large proportion of our aging population. [1, 2]

Footnotes

ACKNOWLEDGMENTS

We would like to acknowledge Caitlin Plovnick, an Assistant Curator for the Medical Library at the NYU Grossman School of Medicine, for her guidance in developing an appropriate search strategy for this narrative review.

FUNDING

The authors have no funding to report.

CONFLICT OF INTEREST

The authors have no conflicts of interest to report.

DATA AVAILABILITY

Data sharing is not applicable to this article as no datasets were generated or analyzed during this study.

The supplementary material is available in the electronic version of this article: .

References

(2022) 2022 Alzheimer’s disease facts and figures. Alzheimers Dement 18, 700–789.

Lee

, Weinblatt

(2001) Rheumatoid arthritis. Lancet 358, 903–911.

Trzeciak

, Herbet

, Dudka

(2021) Common factors of Alzheimer’s disease and rheumatoid arthritis-pathomechanism and treatment. Molecules 26, 6038.

Sangha

, Thakur

, Akhtar

, Ramani

, Gyamfi

(2020) The link between rheumatoid arthritis and dementia: A review. Cureus 12, e7855.

Ungprasert

, Wijarnpreecha

, Thongprayoon

(2016) Rheumatoid arthritis and the risk of dementia: A systematic review and meta-analysis. Neurol India 64, 56–61.

Policicchio

, Ahmad

, Powell

, Proitsi

(2017) Rheumatoid arthritis and risk for Alzheimer’s disease: A systematic review and meta-analysis and a Mendelian Randomization study. Sci Rep 7, 12861.

Kronzer

, Crowson

, Davis

, Vassilaki

, Mielke

, Myasoedova

(2021) Trends in incidence of dementia among patients with rheumatoid arthritis: A population-based cohort study. Semin Arthritis Rheum 51, 853–857.

Lin

, Chen

, Sheu

, Chen

, Chang

(2018) Autoimmune rheumatic diseases increase dementia risk in middle-aged patients: A nationwide cohort study. PLoS One 13, e0186475.

Wotton

, Goldacre

(2017) Associations between specific autoimmune diseases and subsequent dementia: Retrospective record-linkage cohort study, UK. J Epidemiol Community Health 71, 576–583.

10.

Chen

, Chen

, Fan

, Lin

, Wang

, Lin

, Chiou

, Wei

(2018) Rheumatic diseases are associated with a higher risk of dementia: A nation-wide, population-based, case-control study. Int J Rheum Dis 21, 373–380.

11.

Wallin

, Solomon

, Kareholt

, Tuomilehto

, Soininen

, Kivipelto

(2012) Midlife rheumatoid arthritis increases the risk of cognitive impairment two decades later: A population-based study. J Alzheimers Dis 31, 669–676.

12.

Kao

, Kang

, Lin

, Huang

, Lee

, Chung

(2016) Rheumatoid arthritis was negatively associated with Alzheimer’s disease: A population-based case-control study. PLoS One 11, e0168106.

13.

Kodishala

, Hulshizer

, Kronzer

, Davis

, Ramanan

, Vassilaki

, Mielke

, Crowson

, Myasoedova

(2023) Risk factors for dementia in patients with incident rheumatoid arthritis: A population-based cohort study. J Rheumatol 50, 48–55.

14.

Gwinnutt

, Toyoda

, Jeffs

, Flanagan

, Chipping

, Dainty

, Mioshi

, Hornberger

, MacGregor

(2021) Reduced cognitive ability in people with rheumatoid arthritis compared with age-matched healthy controls. Rheumatol Adv Pract 5, rkab044.

15.

Olah

, Kardos

, Andrejkovics

, Szarka

, Hodosi

, Domjan

, Sepsi

, Sas

, Kostyal

, Fazekas

, Florian

, Lukacs

, Miksi

, Barath

, Kerekes

, Pentek

, Valikovics

, Tamasi

, Bereczki

, Szekanecz

(2020) Assessment of cognitive function in female rheumatoid arthritis patients: Associations with cerebrovascular pathology, depression and anxiety. Rheumatol Int 40, 529–540.

16.

Lee

, Kim

, Lee

(2018) Cognitive function of patients with rheumatoid arthritis is associated with disease activity but not carotid atherosclerotic changes. Clin Exp Rheumatol 36, 856–861.

17.

Vitturi

, Nascimento

BAC

, Alves

, de Campos

FSC

, Torigoe

(2019) Cognitive impairment in patients with rheumatoid arthritis. J Clin Neurosci 69, 81–87.

18.

Vassilaki

, Crowson

, Davis

, Duong

, Jones

, Nguyen

, Mielke

, Vemuri

, Myasoedova

(2022) Rheumatoid arthritis, cognitive impairment, and neuroimaging biomarkers: Results from the Mayo Clinic Study of Aging. J Alzheimers Dis 89, 943–954.

19.

Porsteinsson

, Isaacson

, Knox

, Sabbagh

, Rubino

(2021) Diagnosis of early Alzheimer’s disease: Clinical practice in 2021. J Prev Alzheimers Dis 3, 371–386.

20.

American College of Rheumatology (2023) Criteria. https://www.rheumatology.org/Practice-Quality/Clinical-Support/Criteria, Accessed in January, 2023.

21.

Westreich

(2012) Berkson’s bias, selection bias, and missing data. Epidemiology 23, 159–164.

22.

O’Brien

, Thomas

(2015) Vascular dementia. Lancet 386, 1698–1706.

23.

Chou

, Kane

, Ghimire

, Gautam

, Gui

(2016) Treatment for rheumatoid arthritis and risk of Alzheimer’s disease: A nested case-control analysis. CNS Drugs 30, 1111–1120.

24.

Zhou

, Xu

, Kaelber

, Gurney

(2020) Tumor necrosis factor (TNF) blocking agents are associated with lower risk for Alzheimer’s disease in patients with rheumatoid arthritis and psoriasis. PLoS One 15, e0229819.

25.

Kern

, Lovestone

, Cepeda

(2021) Treatment with TNF-alpha inhibitors versus methotrexate and the association with dementia and Alzheimer’s disease. Alzheimers Dement 7, e12163.

26.

Judge

, Garriga

, Arden

, Lovestone

, Prieto-Alhambra

, Cooper

, Edwards

(2017) Protective effect of antirheumatic drugs on dementia in rheumatoid arthritis patients. Alzheimers Dement 3, 612–621.

27.

Chou

, Wang

, Lin

, Chung

(2017) DMARD use is associated with a higher risk of dementia in patients with rheumatoid arthritis: A propensity score-matched case-control study. Toxicol Appl Pharmacol 334, 217–222.

28.

Chang

, Hsu

, Lin

, Hsu

, Lee

, Chong

, Liu

, Lin

, Kao

(2016) Prolong exposure of NSAID in patients with RA will decrease the risk of dementia: A nationwide population-based cohort study. Medicine 95, e3056.

29.

Prado

, Crowe

(2019) Corticosteroids and cognition: A meta-analysis. Neuropsychol Rev 29, 288–312.

30.

Desai

, Varma

, Gerhard

, Segal

, Mahesri

, Chin

, Horton

, Kim

, Schneeweiss

, Thambisetty

(2022) Comparative risk of Alzheimer disease and related dementia among Medicare beneficiaries with rheumatoid arthritis treated with targeted disease-modifying antirheumatic agents. JAMA Netw Open 5, e226567.

31.

Sattui

, Rajan

, Lieber

, Lui

, Sterling

, Curtis

, Mandl

, Navarro-Millan

(2021) Association of cardiovascular disease and traditional cardiovascular risk factors with the incidence of dementia among patients with rheumatoid arthritis. Semin Arthritis Rheum 51, 292–298.

32.

Catalogue of bias collaboration, Aronson

, Bankhead

, Mahtani

, Nunan

(2018) Confounding by indication. https://catalogofbias.org/biases/confounding-by-indication, Accessed January, 2023.

33.

Min

, Bang

, Kim

, Oh

, Choi

(2020) Rheumatoid arthritis and neurodegenerative dementia: A nested case-control study and a follow-up study using a national sample cohort. Clin Rheumatol 39, 159–166.

34.

Park

, Yim

, Ochirpurev

, Eom

, Choi

, Ju

, Kim

(2021) Association between dementia and systemic rheumatic disease: A nationwide population-based study. PLoS One 16, e0248395.

35.

Booth

, Kobayashi

, Janevic

, Clauw

, Piette

(2021) No increased risk of Alzheimer’s disease among people with immune-mediated inflammatory diseases: Findings from a longitudinal cohort study of U.S. older adults. BMC Rheumatol 5, 48.

36.

Booth

, Janevic

, Kobayashi

, Clauw

, Piette

(2021) No association between rheumatoid arthritis and cognitive impairment in a cross-sectional national sample of older U.S. adults. BMC Rheumatol 5, 24.