The RESIST Study: Examining Cognitive Change in Rheumatoid Arthritis Patients with Mild Cognitive Impairment Being Treated with a TNF-Inhibitor Compared to a Conventional Synthetic Disease-Modifying Anti-Rheumatic Drug

Abstract

Background:

Evidence suggests that TNF inhibitors (TNFi) used to treat rheumatoid arthritis (RA) may protect against Alzheimer’s disease progression by reducing inflammation.

Objective:

To investigate whether RA patients with mild cognitive impairment (MCI) being treated with a TNFi show slower cognitive decline than those being treated with a conventional synthetic disease-modifying anti-rheumatic drug (csDMARD).

Methods:

251 participants with RA and MCI taking either a csDMARD (N = 157) or a TNFi (N = 94) completed cognitive assessments at baseline and 6-month intervals for 18 months. It was hypothesized that those taking TNFis would show less decline on the primary outcome of Free and Cued Selective Reminding Test with Immediate Recall (FCSRT-IR) and the secondary outcome of Montreal Cognitive Assessment (MoCA).

Results:

No significant changes in FCSRT-IR scores were observed in either treatment group. There was no significant difference in FCSRT-IR between treatment groups at 18 months after adjusting for baseline (mean difference = 0.5, 95% CI = –1.3, 2.3). There was also no difference in MoCA score (mean difference = 0.4, 95% CI = –0.4, 1.3).

Conclusions:

There was no cognitive decline in participants with MCI being treated with TNFis and csDMARDs, raising the possibility both classes of drug may be protective. Future studies should consider whether controlling inflammatory diseases using any approach is more important than a specific therapeutic intervention.

Keywords

Alzheimer’s disease inflammation mild cognitive impairment rheumatoid arthritis tumor necrosis factor-alpha

INTRODUCTION

There is growing research interest into the role of inflammation in the progression of Alzheimer’s disease (AD). Studies have shown that microglia in the central nervous system interact with soluble amyloid-β (Aβ) oligomers and Aβ fibrils via cell-surface receptors, resulting in an inflammatory response mediated in part by the production of proinflammatory cytokines, including tumor necrosis factor (TNF)-α [1, 2]. This response is thought to be beneficial in the early stages of AD, aiding the clearance of Aβ. However, sustained activation of microglia over time, persistent exposure to proinflammatory cytokines and microglial process retraction can cause functional and structural changes that result in neuronal dysfunction, followed by degeneration [3, 4].

There is evidence that peripheral blood concentrations of TNF-α are significantly higher in AD patients compared to healthy controls [5]. Furthermore, Holmes et al. [6] demonstrated an increased rate of cognitive decline in AD patients with high systemic levels of TNF-α. These findings have led to the hypothesis that reduction in circulating levels of TNF-α could lead to reduced cognitive decline in AD patients. To investigate this, a six-month pilot study administered a weekly dose of the TNF inhibitor (TNFi) etanercept to 15 AD patients [7]. Results showed a significant improvement in cognitive function among the patients, although there was no control group for comparison. A small randomized, placebo-controlled trial designed to assess tolerability that allocated 41 AD patients to receive either 50 mg of etanercept or a placebo over 24 weeks found trends that favored etanercept but no significant between-group differences in cognitive decline [8].

Rheumatoid arthritis (RA) is a chronic, systemic inflammatory condition primarily affecting the joint synovium. Evidence suggests that dementia risk is greater among RA patients [9]. It has been suggested that RA patients may represent a useful model population to study the association between inflammation and dementia, given that they exhibit higher levels of systemic inflammation and thus processes driven by inflammation (such as cognitive decline) may be accelerated in these individuals [10]. Furthermore, existing immunosuppressant medications used to treat RA may offer a therapeutic approach to AD by reducing systemic inflammation.

The typical first line of treatment for RA are conventional synthetic disease-modifying anti-rheumatic drugs (csDMARDs) such as methotrexate. Methotrexate treatment has previously been associated with both lower [11] and higher risk [12] of dementia in RA patients. If patients do not respond to csDMARDs or are unable to tolerate them, they may be prescribed a biological drug, such as a TNFi [13, 14]. Several large-scale cross-sectional studies have reported a lower risk of developing AD among individuals with RA on TNFi treatment [15 –18] (example OR 0.45 [15]). Another study of 210 RA patients without diagnosed dementia found that those being treated with a TNFi had a 50% reduced risk of cognitive impairment [19]. This finding indicates that TNFi may be protective before the onset of AD, for example among individuals with mild cognitive impairment (MCI). Patients with MCI demonstrate memory impairment but do not have clinical symptoms of dementia [20]. The risk of developing AD is greater among people diagnosed with MCI, with evidence suggesting this risk is even greater among individuals carrying the APOE ɛ4 allele [21, 22].

Previous research therefore suggests that dementia risk may be lower among RA patients taking a TNFi. However, there is a lack of longitudinal studies that provide evidence for a protective effect against cognitive decline over time. A twelve-week pilot study of 15 RA patients without diagnosed cognitive impairment found that TNFi treatment was associated with improvement in cognitive function [23], but further evidence with larger samples and longer follow-up times is necessary to draw conclusions. The rheumatoid arthritis medication and memory study (RESIST) examined whether treatment with a TNFi may be associated with reduced cognitive decline in RA patients with MCI. It was hypothesized that RA patients treated with a TNFi would show less decline over an 18-month period compared to those taking a csDMARD.

METHODS

Design and participants

RESIST was a longitudinal, observational study examining cognitive change in older adults with both RA and MCI. Data were collected during four visits to participants’ homes over an 18-month period. Study eligibility criteria were: a) age 55 and over; b) fluent in English; c) able to give informed consent to participate; d) willing and able to participate for the duration of the study; e) diagnosis of RA according to EULAR/ACR criteria [24]; f) being treated with either a csDMARD (leflunomide, methotrexate, azathioprine, cyclosporin, cyclophosphamide, sulfasalazine or hydroxychloroquine sulfate) or a TNFi (etanercept, infliximab, golimumab, adalimumab or certolizumab; participants may have also been taking a csDMARD in conjunction); g) diagnosis of MCI according to Petersen criteria [20]. Exclusion criteria were: a) significant visual/hearing impairment that would interfere with ability to complete cognitive tests; b) participation in another research study involving administration of any investigational drug; c) any previous or current medical condition that may impact on cognitive performance (according to Principal Investigator’s judgement); d) taking a non-TNFi biologic immunosuppressant (tocilizumab, abatacept, rituximab, baricitinib, sarilumab, belimumab, secukinumab, ustekinumab, ixekizumab or tofacitinib); e) taking a cholinesterase inhibitor (donepezil, rivastigmine or galantamine) or memantine.

Participants were recruited from specialist rheumatology clinics across three UK healthcare trust areas. Three clinics were located within the Belfast Health and Social Care Trust (Belfast City Hospital, the Royal Group of Hospitals and Musgrave Park Hospital), one within the Northern Health and Social Care Trust (Antrim Area Hospital), and one within the University Hospital Southampton NHS Foundation Trust (University Hospital Southampton). Specialist nurses or rheumatology consultants from these clinics identified potentially eligible participants during regularly scheduled outpatient check-up appointments and referred them to members of the study team for screening, either immediately following these appointments or at a later date (usually within two weeks).

The screening process involved a short interview. After obtaining written informed consent, participants provided demographic information along with a list of medication they were currently taking. Participants then completed the Montreal Cognitive Assessment (MoCA) [25]. Individuals who met the eligibility criteria outlined above were formally recruited into the study and booked in for the first home visit. Recruitment and screening were carried out between April 2018 and March 2020; the final follow-up visit was conducted in October 2021.

Power analysis indicated that a sample size of 324 participants (108 TNFi and 216 csDMARD) was necessary for over 80% power at an alpha level of.05 (based on the assumption of a decline of 3.2 points [SD = 4.7] on the primary outcome measure in the untreated group [i.e., participants not on a TNFi], derived from [26], and a 50% reduced decline of 1.6 points [SD = 4.7] in the treated group [i.e., participants on a TNFi]). Assuming a 10% rate of drop-out in both groups, a target sample size of 360 was set (120 TNFi, 240 csDMARD).

This study was conducted in accordance with ethical standards as laid down in the Declaration of Helsinki. This study received approval from the West Midlands –Black County Research Ethics Committee (REC reference: 17/WM/0161). Research governance permission was also granted by local research and development departments in both Northern Ireland (BHSCT, NHSCT) and Southampton (UHS). All participants provided written informed consent to take part.

Materials and measures

Medication

Participants were asked to list all medications that they were currently taking at study screening; this information was used to categorize them into either the csDMARD group or the TNFi group. The TNFi group included all participants currently taking a TNF inhibitor, either solely or in conjunction with a csDMARD. Corticosteroid (prednisolone) use was also recorded.

Outcome measures

The primary outcome of the study was performance on the Free and Cued Selective Reminding Test with Immediate Recall (FCSRT-IR) [27]. The FCSRT is designed to measure memory under conditions that ensure appropriate attentional and semantic processing of stimuli, allowing for the measurement of any specific memory impairment absent of other impairments in attention or semantic ability. Participants were presented with a card containing four words. During the study phase, participants were asked to search for each word (e.g., ‘grapes’) using a unique category cue (e.g., ‘fruit’). Once all four items were identified, the card was removed and immediate cued recall was tested. Participants were asked to identify each item using the category cue (e.g., ‘tell me the name of the fruit’). The study phase was repeated for any items not retrieved during immediate recall, until all four items were retrieved. This procedure was then repeated with three other cards, until a total of 16 items had been identified and retrieved during immediate recall. Following 20 s of interference (counting backwards or forwards depending on the participant’s ability), there were three recall trials, consisting of free recall followed by cued recall for any items not retrieved during free recall. Selective reminding was conducted for any items not retrieved during cued recall. There were 20 seconds of interference between each trial. Two scores were recorded, both out of 48: free recall (the total number of items retrieved over the three free recall trials) and total recall (total number of items retrieved over the three free recall trials + the total number of items retrieved over the three cued recall trials). Different word lists were used for each visit to reduce practice effects.

The secondary outcome of the study was MoCA score. The MoCA is a ten-minute, cognitive screening tool commonly used to detect MCI. The test assesses cognitive function across several domains, including memory, visuospatial ability, executive function, attention, concentration, working memory, language, and orientation. The maximum possible score is 30. To reduce any possible practice effects, version 7.1 was used for screening, and version 7.3 was used for main study visits. Scores were adjusted for education level, with one point added for participants with twelve or fewer years of education.

Covariates

Covariates included a range of potential confounders that may account for differences in cognitive outcomes among RA patients. Participants’ level of education was measured as a binary variable (≤/>12 years). RA disease duration was measured as number of years since diagnosis at visit 1 (the year of diagnosis was reported by participants and confirmed in clinical records, if available).

The DAS-28 ESR [28] was used to assess RA disease activity. The DAS-28 ESR provides a composite score based on three measures: a count of swollen and/or tender joints (out of 28), erythrocyte sedimentation rate and self-rating of general wellbeing on a visual analogue scale (VAS; i.e., participants marked a line labelled ‘Best Imaginable Health State’ and ‘Worst Imaginable Health State’ at either extreme, scored 0–100). DAS-28 scores can range from 0 to 9.4. High RA disease activity has previously been associated with cognitive impairment [29].

The Geriatric Depression Scale-Short Form (GDS) [30] was used to measure depressive symptoms. The questionnaire includes 15 yes/no items. The maximum score is 15, with higher scores representing a greater number of depressive symptoms. Higher GDS score has previously been linked to greater cognitive decline [31].

The short version of the Health Assessment Questionnaire (HAQ-DI) [32] was used to measure participants’ overall functional status. The HAQ-DI includes 20 items covering a range of functional activities across eight categories: dressing/grooming, rising, eating, walking, hygiene, reach, grip and usual activities. Response options for each item are 0 = ‘without any difficulty’, 1 = ‘with some difficulty’, 2 = ‘with much difficulty’, and 3 = ‘unable to do’. The total score was calculated as the average of the highest score from within each category, ranging from 0 to 3 (scores are adjusted according to the use of aids/devices/physical assistance); higher HAQ scores have been linked to poorer cognitive function in RA patients [19]. The short version of the Health Assessment Questionnaire also included a visual analogue scale measurement of pain. Participants indicated their current overall level of pain by marking a 10-cm line labelled ‘no pain at all’ and ‘my pain is as bad as it could possibly be’ at either extreme. Scores can range from 0 to 100.

Blood collection

Blood samples were obtained by venepuncture of a peripheral vein. At baseline blood samples were taken for DNA analysis of APOE ɛ4 genotype. Serum and plasma samples were initially stored at –80°C onsite at either the Centre for Public Health, Queen’s University Belfast or the Memory Assessment and Research Centre Southampton. APOE genotyping was performed at the QUB Core Genomics Facility using Taqman Real Time PCR Assays (Thermo Fisher Scientific).

Procedure

Study assessments were conducted in participants’ homes by trained members of the study team (either a specialist dementia research nurse, a PhD student, or a post-doctoral researcher). Four home visits took place at six-month intervals. At the first visit, participants first provided written informed consent. The original screening procedure was then repeated (i.e., participants indicated any changes to their medication and completed the MoCA); at this point any participants who no longer met the eligibility criteria were excluded from the study. Those who were still eligible then provided a detailed medical history, and completed the FCSRT-IR, DAS-28, GDS and HAQ-DI. A blood sample was also collected. This procedure was repeated at each follow-up visit (with the exception that participants were no longer excluded based on their MoCA score at visit 2, 3, or 4).

COVID-19

The COVID-19 pandemic had a substantial impact on the study. Study recruitment and fieldwork was halted when England and Northern Ireland entered a period of lockdown in March 2020, and did not resume until September 2020. At this point the opportunity to complete scheduled follow-ups had been missed for many participants. Once fieldwork had resumed, many participants chose not to continue with the study. Although specific reasons were not recorded, some expressed that they were not comfortable with a researcher visiting them at their home. It is important to note that given the sample in this study were aged 55 and over and immunosuppressed they were likely to be at greater risk of severe disease. In order to mitigate risk of COVID-19 infection, once data collection resumed all visits were conducted with full personal protective equipment (face mask, apron, and gloves). There was an additional pause in data collection during a period of increased restrictions between January and May 2021. A decision was made to focus on completing fourth visits. As such, numbers of participants who completed visit 4 were slightly higher than those who completed visits 2 and 3.

The pandemic also had an impact on the duration of time between study visits. Descriptive statistics for the duration of each of these intervals is presented in Supplementary Table 1. As per study protocol follow-up visits were scheduled to take place at 6-month intervals (+/–2 weeks). Due to lockdown-related delays, the average interval between visit 1 and visit 2 was approximately 6 and a half months (198.6 days, SD = 43.8). The average interval between visit 1 and visit 3 was 13 and a half months (408.3 days, SD = 77.4). The average interval between visit 1 and visit 4 was approximately 23 months (701.9 days, SD = 170.7).

Statistical analysis

All analyses were conducted in R (version 4.2.2) [33]. Due to larger rates of attrition at visit 2 and visit 3, analyses focused on visit 4 as the primary study end-point. Scores at visit 1 were treated as the baseline. To examine any possible systematic patterns of study withdrawal, Welch’s t-tests were conducted to compare baseline characteristics of those who did and did not complete the primary study end-point visit on all study outcomes and covariates (chi-square tests were used for categorical variables). Comparisons of baseline characteristics between the TNFi group and the csDMARD group were also conducted.

Paired t-tests were used to test for significant within-group changes in cognitive outcomes and measures of disease activity over time for both the csDMARD group and the TNFi group.

To examine group differences in cognitive change over time, analysis of covariance (ANCOVA) was conducted to calculate the difference in mean FCSRT-IR scores between the TNFi and csDMARD groups at study end-point while controlling for baseline scores [34]. Adjusted models repeated the above analysis controlling for additional covariates (age, gender, education, RA disease duration, baseline VAS pain, baseline VAS wellbeing, baseline DAS-28 score, baseline GDS score and baseline HAQ-DI score).

To examine the possible moderating effect of APOE genotype, models were re-run including the between-participants factor of APOE ɛ4 genotype (ɛ4 positive or ɛ4 negative) and the interaction between medication group and ɛ4 status. ANCOVA models comparing medication groups were then run separately for each APOE genotype subgroup.

Analyses were also repeated with MoCA as the secondary outcome variable. All analyses were conducted on available complete cases. Complete case analysis can be biased when data are missing not at random (MNAR; i.e., missingness in a given variable is related to the unobserved values in that variable). Given the possibility that missing values on the cognitive outcome were more likely among participants who experienced cognitive decline, a sensitivity analysis was conducted on the primary outcome of FCSRT-IR free recall score to examine whether a decline among those with missing data affected results. This involved conducting δ-based multiple imputation, in which missing values are imputed before an offset value (δ) is added to the imputed values; the offset value represents a specified difference in mean outcome among missing cases [35]. In this case, δ was set to be –4 (i.e., the mean free recall score among missing cases was set to be four points lower), which would represent a reasonable decline.

RESULTS

Study attrition

The study recruitment process is summarized in Fig. 1. Seven hundred and twenty-eight individuals were screened for eligibility. Of these, 262 were either ineligible or chose not to participate in the study. The remaining 466 individuals were recruited onto the study. One hundred and fifty-two participants withdrew prior to the first study visit. A further 63 participants were excluded from the study at the first visit as they no longer met the study eligibility criteria (either due to changes in medication or scoring outside the accepted cut-offs on the MoCA during the visit). Therefore, the final eligible sample who completed the first visit of the study comprised 251 participants (mean age = 69.1 years).

Fig. 1

Study participation flowchart. Exclusions due to medication included participants who were no longer taking either a csDMARD or a TNFi, and participants who were taking a medication listed in the exclusion criteria.

Of the 251 participants who completed the first visit, 143 (57.0%) completed visit 2, 112 (44.6%) completed visit 3 and 149 (59.4%) completed visit 4. Follow-up completion rates were largely comparable in the csDMARD group (visit 2 : 60.5%; visit 3 : 43.3%; visit 4 : 56.7%) and the TNFi group (visit 2 : 51.1%; visit 3 : 46.8%; visit 4 : 63.8%). Comparisons of baseline characteristics between visit 4 completers and non-completers are presented in Supplementary Material (Supplementary Table 2). At baseline, study completers scored significantly higher than non-completers in FCSRT-IR free recall (mean difference = 2.5, 95% CI = 0.9, 4.2, p = 0.003), and on the MoCA (mean difference = 0.5, 95% CI = 0.0, 1.0, p = 0.039). All other comparisons were non-significant.

3.2 Baseline characteristics

A summary of baseline characteristics for the complete sample and each individual treatment group are presented in Table 1. Participants in the TNFi group had significantly longer RA disease duration and scored significantly higher on FCSRT-IR free recall and the HAQ-DI than those in the csDMARD group. There were no between-group differences in gender, education, smoking history, APOE status, self-rated pain, wellbeing, disease activity, depressive symptoms, FCSRT-IR total recall score or MoCA score at baseline.

Table 1

Baseline characteristics for complete analytic sample

Variable	Complete Sample		csDMARD		TNFi		p ^b
	N	Mean (SD)/% ^a	N	Mean (SD)/% ^a	N	Mean (SD)/% ^a
Age	144	69.1 (7.9)	86	69.4 (8.9)	58	68.6 (6.3)	0.527
Gender							0.170
Male	44	30.6	30	34.9	14	24.1
Female	100	69.4	56	65.1	44	75.9
Education							0.591
≤12 y	87	60.4	53	61.6	34	58.6
>12 y	55	38.2	31	36.0	24	41.4
Missing	2	1.4	2	2.3	–	–
Smoking							0.347
Smoker	18	12.5	12	14.0	6	10.3
Ex-smoker	52	36.1	27	31.4	25	43.1
Non-smoker	74	51.4	47	54.7	27	46.6
Taking methotrexate							0.026
Yes	101	70.1	66	76.7	35	60.3
No	42	29.2	19	22.1	23	39.7
Missing	1	0.7	1	1.2	–	–
Taking sulfasalazine							0.091
Yes	21	14.6	16	18.6	5	8.6
No	122	84.7	69	80.2	53	91.4
Missing	1	0.7	1	1.2	–	–
Taking hydroxychloroquine sulfate							0.002
Yes	21	14.6	19	22.1	2	3.4
No	122	84.7	66	76.7	56	96.6
Missing	1	0.7	1	1.2	–	–
Taking other csDMARD^c							0.223
Yes	6	4.2	5	5.8	1	1.7
No	137	95.1	80	93.0	57	98.3
Missing	1	0.7	1	1.2	–	–
Taking prednisolone							0.116
Yes	11	7.6	9	10.5	2	3.4
No	132	91.7	76	88.4	56	96.6
Missing	1	0.7	1	1.2	–	–
APOE							0.945
ɛ4+	28	19.4	17	19.8	11	19.0
ɛ4–	83	57.6	51	59.3	32	55.2
Missing	33	22.9	18	20.9	15	25.9
Disease duration (y)	128	15.4 (15.2)	75	11.2 (14.4)	53	21.3 (14.6)	<0.001
VAS Wellbeing	144	36.7 (24.9)	86	35.5 (22.5)	58	38.5 (28.1)	0.500
VAS Pain	143	39.1 (27.6)	86	38.1 (28.4)	57	40.5 (26.6)	0.607
DAS-28	136	3.3 (1.4)	82	3.3 (1.5)	54	3.4 (1.4)	0.600
GDS	142	3.4 (3.6)	85	3.7 (3.8)	57	3.1 (3.2)	0.278
HAQ-DI	144	1.4 (0.9)	86	1.2 (0.9)	58	1.6 (0.7)	0.003
FCSRT-IR Free Recall	144	26.6 (5.4)	86	25.5 (5.4)	58	28.2 (5.0)	0.002
FCSRT-IR Total Recall	144	47.1 (1.8)	86	47.0 (2.0)	58	47.4 (1.5)	0.183
MoCA	144	24.8 (1.8)	86	24.6 (1.8)	58	25.0 (1.8)	0.147

Analytic sample comprises N = 144 participants with complete data on primary outcome (FCSRT-IR) at baseline and primary study end-point (visit 4). VAS, visual analogue scale; DAS-28, disease activity scale 28; FCSRT-IR, Free and Cued Selective Recall Test –Immediate Recall; GDS, Geriatric Depression Scale; HAQ-DI, Health Assessment Questionnaire (short version); MoCA, Montreal Cognitive Assessment. ^aContinuous variables are summarized as mean (SD) and categorical variables are summarized as frequency (%). ^bContinuous variables were compared using Welch’s t-test and categorical variables were compared using chi-square tests. ^cLeflunomide, azathioprine, or cyclophosphamide.

In the complete sample, 70.1% of participants were taking methotrexate at study screening. This proportion was significantly greater in the csDMARD group (76.7%) compared to the TNFi group (60.3%). Eleven participants (7.6%) were taking a corticosteroid (prednisolone); this proportion did not differ between groups.

Cognitive change over time

Table 2 presents mean scores on all outcome measures in each medication group at baseline and at the primary study end-point. Within the csDMARD group, paired-sample t-tests revealed no evidence of a significant change between baseline and end-point in FCSRT-IR free recall (25.5 versus 26.2, mean difference = 0.7, 95% CI = –0.4, 1.8, p = 0.219), FCSRT-IR total recall (47.0 versus 47.3, mean difference = 0.4, 95% CI = –0.2, 0.9, p = 0.172), or MoCA score (24.5 versus 24.9, mean difference = 0.5, 95% CI = –0.2, 1.1, p = 0.141). Within the TNFi group, there was also no evidence of a significant change in FCSRT-IR free recall (28.2 versus 29.0, mean difference = 0.7, 95% CI = –0.7, 2.2, p = 0.307), FCSRT-IR total recall (47.4 versus 47.1, mean difference = –0.3, 95% CI = –0.9, 0.3, p = 0.275), or MoCA score (25.0 versus 25.6, mean difference = 0.6, 95% CI = –0.1, 1.3, p = 0.073). Mean scores on measures of disease activity at baseline and end-point are presented in Supplementary Table 3; there was no evidence of significant changes in DAS-28 score, or self-rated pain or wellbeing.

Table 2

Mean scores and adjusted between-group differences on all outcome variables at baseline and primary study end-point (18 months)

Outcome	csDMARD				TNFi				Mean difference^a (95% CI)	p	Adjusted mean difference (95% CI)^b	p
	N	Baseline mean (SD)	N	End-point Mean (SD)	N	Baseline mean (SD)	N	End-point Mean (SD)
FCSRT-IR Free Recall	86	25.5 (5.4)	86	26.2 (6.5)	58	28.2 (5.0)	58	29.0 (7.3)	0.5 (–1.3, 2.3)	0.578	1.3 (–1.0, 3.6)^c	0.253
FCSRT-IR Total Recall	86	47.0 (2.0)	86	47.3 (1.5)	58	47.4 (1.5)	58	47.1 (3.3)	–0.5 (–1.2, 0.2)	0.182	–0.3 (–1.2, 0.6)^c	0.516
MoCA	88	24.5 (1.9)	88	24.9 (2.8)	60	25.0 (1.8)	60	25.6 (2.5)	0.4 (–0.4, 1.3)	0.308	0.8 (–0.3, 1.8)^d	0.151

Analyses conducted using complete cases. FCSRT-IR, Free and Cued Selective Reminding Test with Immediate Recall; MoCA, Montreal Cognitive Assessment. ^aMean difference between csDMARD and TNFi groups at primary end-point, adjusted for baseline score, from ANCOVA. ^bMean difference between csDMARD and TNFi groups at primary end-point, adjusted for baseline score, age, gender, education, disease duration, baseline VAS Pain, baseline VAS Wellbeing, baseline DAS-28, baseline GDS and baseline HAQ-DI, from ANCOVA. ^cComplete cases due to missing covariate data: N csDMARD = 71, N TNFi = 47. ^dComplete cases due to missing covariate data: N csDMARD = 73, N TNFi = 49.

Group differences

As shown in Table 2, when adjusting for baseline scores, there were no significant differences between those on csDMARDs and those on TNFis in FCSRT-IR free recall score or total recall score at the primary study end-point. This remained the case when models were adjusted for age, gender, education, years diagnosed, VAS Pain, VAS Wellbeing, DAS-28 score, GDS score, and HAQ-DI score. Results were the same for the secondary outcome of MoCA score, with no significant differences observed in the baseline adjusted or fully adjusted models. Repeating analyses for the primary outcome of FCSRT-IR free recall score using δ-based MI (δ= –4) did not appreciably alter the results (adjusted mean difference = 0.1, 95% CI = –2.4, 2.6).

Additional models were created including the interaction between APOE ɛ4 status and medication type. p values for the interaction terms in each ANCOVA model are presented in Table 3. For both FCSRT-IR free recall score and MoCA score at study end-point, there was no significant ɛ4-medication interaction when adjusting for baseline score only or when additionally adjusting for all other covariates. For FCSRT-IR total recall score at study end-point, there was a significant ɛ4-medication interaction when adjusting for baseline score only, but this interaction became non-significant when adjusting for all other covariates. ANCOVA models comparing those on csDMARDs and those on TNFi in each APOE ɛ4 subgroup (ɛ4 positive and ɛ4 negative) are presented in Table 3. For all outcomes, there was no evidence of a significant difference between medication groups in either APOE ɛ4 subgroup, in both baseline-adjusted and fully adjusted models.

Table 3

Mean scores and adjusted between-group differences on all outcome variables at baseline and primary study end-point for participants with and without the APOE ɛ4 allele

Outcome	csDMARD				TNFi				ɛ4×med interaction p^a	Mean difference^b (95% CI)	p	Adjusted ɛ4×med interaction p^c	Adjusted mean difference^d (95% CI)	p
	N	Baseline mean (SD)	N	End-point Mean (SD)	N	Baseline mean (SD)	N	End-point Mean (SD)
FCSRT-IR Free Recall									0.651			0.767
ɛ4–	51	25.1 (5.8)	51	26.3 (7.2)	32	29.0 (4.5)	32	29.0 (6.4)	–0.3 (–2.9, 2.3)	0.813	1.8 (–1.5, 5.1)^e	0.277
ɛ4+	17	25.1 (4.9)	17	25.2 (5.6)	11	25.5 (5.7)	11	26.1 (10.7)	0.4 (–4.0, 4.7)	0.867	2.2 (–4.4, 8.9)^f	0.469
FCSRT-IR Total Recall									0.039			0.129
ɛ4–	51	46.9 (2.2)	51	47.2 (1.8)	32	47.5 (1.1)	32	47.4 (1.6)	0.2 (–0.6, 1.0)	0.642	0.4 (–0.7, 1.5)^e	0.463
ɛ4+	17	46.9 (1.6)	17	47.5 (1.5)	11	46.2 (2.7)	11	44.7 (6.9)	–1.7 (–4.3, 0.9)	0.184	–1.0 (–5.6, 3.6)^f	0.642
MoCA									0.554			0.380
ɛ4–	52	24.4 (1.9)	52	24.9 (2.5)	33	25.3 (1.6)	33	25.3 (2.6)	–0.1 (–1.2, 1.1)	0.918	0.1 (–1.5, 1.6)^g	0.915
ɛ4+	18	24.8 (1.8)	18	24.9 (3.5)	11	24.8 (2.0)	11	25.5 (2.7)	0.6 (–1.8, 3.0)	0.624	–0.4 (–2.5, 1.8)^h	0.706

Analyses conducted using complete cases. FCSRT-IR, Free and Cued Selective Reminding Test with Immediate Recall. MoCA, Montreal Cognitive Assessment. ^ap value for interaction between APOE ɛ4 status (positive vs negative) and medication group, in model adjusted for baseline score only ^bMean difference between csDMARD and TNFi groups at primary end-point, adjusted for baseline score, from ANCOVA ^cp value for interaction between APOE ɛ4 status (positive vs negative) and medication group, in model adjusted for baseline score, age, gender, education, disease duration, baseline VAS Pain, baseline VAS Wellbeing, baseline DAS-28, baseline GDS and baseline HAQ-DI ^dMean difference between csDMARD and TNFi groups at primary end-point, adjusted for baseline score, age, gender, education, disease duration, baseline VAS Pain, baseline VAS Wellbeing, baseline DAS-28, baseline GDS and baseline HAQ-DI, from ANCOVA ^eComplete cases due to missing covariate data: N csDMARD = 47, N TNFi = 26. ^fComplete cases due to missing covariate data: N csDMARD = 12, N TNFi = 9. ^gComplete cases due to missing covariate data: N csDMARD = 48, N TNFi = 27. ^hComplete cases due to missing covariate data: N csDMARD = 13, N TNFi = 9.

DISCUSSION

The RESIST study compared cognitive change between RA patients with MCI who were being treated with a TNFi and those being treated with a csDMARD. It was hypothesized that participants taking a TNFi would show less cognitive decline than those taking a csDMARD. Results did not support this hypothesis. No significant declines were observed on any cognitive outcomes over the duration of the study. Additionally, there were no significant differences between the treatment groups at the end of the study when adjusting for baseline score and other covariates. In other words, participants being treated with a TNFi exhibited no difference in cognitive change over time compared to those being treated with a csDMARD.

Previous studies examining longitudinal cognitive change in RA patients being treated with a TNFi have reported mixed results. A pilot study reported an improvement in global cognitive function among RA patients (N = 15) initiating adalimumab [23]. However, this study did not include a control group for comparison, so the observed improvement could not be definitively linked to TNFi treatment. Another study reported no evidence of global cognitive decline over an average of 3.9 years in a sample of cognitively unimpaired RA patients taking biological drugs (including TNFi) [36]. The present study used a larger sample and included a comparison group who were being treated with csDMARDs. The lack of any observed cognitive change in either group, and, importantly, the absence of any between-group differences in cognitive change, suggests that, compared to csDMARDs alone, TNFi treatment offers no additional protective effect on cognition in RA patients. Similarly, a large scale cohort study of RA patients comparing TNFi treatment to treatment with a T-cell activation inhibitor (abatacept) reported no difference in risk of dementia [37].

Previous studies that have examined whether TNFi treatment slows cognitive decline in AD patients without RA did not suggest any specific benefits. A pilot study reported a significant improvement in cognitive function after six months taking etanercept, but this study employed a small sample (N = 15) and lacked a control group [7]. The only randomized controlled trial of etanercept in AD patients found trends that favored etanercept compared to a placebo group, but no significant between-group differences in cognitive decline [8]. While the present results build on this finding by suggesting that TNFi treatment has no effect on cognitive change in patients with MCI, it is important to emphasize that, unlike previous studies, participants in this study had a diagnosis of RA. Furthermore, those in the TNFi group were being compared to patients on a csDMARD rather than a placebo group.

Given that both MCI and RA are linked to greater risk of cognitive decline and AD [9, 21], it might be expected that the recruited sample would show some degree of cognitive decline over the duration of the study (roughly eighteen months). Indeed, power calculations for this study were based on an observed decline of 3.2 points on the FCSRT-IR derived from a previous study of participants with MCI over the same duration [26]. However, for both study groups (csDMARD and TNFi), performance on cognitive assessments did not show any significant change. One possible reason for the lack of differences observed between groups is that the majority of both medication groups were taking methotrexate (76.7% in the csDMARD group and 60.3% in the TNFi group). Two cross-sectional studies of large health record databases have reported significantly lower risk of AD [17] and dementia (AD, vascular and mixed cause) [11] in RA patients being treated with methotrexate. Newby et al. examined dementia risk in a large multinational sample of RA patients aged 50 and over; they reported a significantly lower risk of dementia among patients taking methotrexate [38]. This risk was lowest among those on methotrexate therapy for longer than 4 years. Conversely, Chou et al. reported a higher risk of dementia in participants taking methotrexate, regardless of the exposure duration [12]. Overall, while results are somewhat mixed, the weight of evidence currently suggests that methotrexate may reduce the risk of dementia. It is possible that methotrexate may also protect against cognitive decline prior to the onset of dementia; the high rates of methotrexate use in the current sample may therefore partly explain the general lack of decline observed.

A recent retrospective cohort study compared dementia risk in RA patients taking TNFi to those taking methotrexate [39]. Both groups had no prior exposure to either drug. No difference in dementia risk was found, suggesting that TNFi treatment may offer no additional protection compared to methotrexate. Only around 40% of those taking TNFi in the present study were not also currently taking methotrexate. In addition, data regarding historical medication use was not collected, and it is possible that those not currently taking methotrexate may have taken it in the past. Therefore, our analysis was not able to examine cognitive change among those who initiated TNFi treatment without previous methotrexate exposure. Future studies should consider the possible confounding effect of methotrexate use when evaluating the possible benefits of TNFi drugs.

The present findings raise the possibility that controlling inflammation through a range of possible treatments may have cognitive benefits for RA patients. Adopting a treat-to-target approach has previously been recommended in the treatment of RA [40]. This approach involves regularly measuring disease activity and monitoring and adjusting drug therapy to achieve a desired outcome (ideally disease remission). Importantly the approach does not focus on recommending one specific type of therapy, but rather using any necessary therapy to achieve the therapeutic target. The absence of any difference in cognitive change between the two therapies observed in this study suggest that controlling inflammation by any means may ameliorate cognitive decline. Future studies could test this theory by examining cognition in RA patients being treated using therapeutic targets.

Analyses also examined whether the differences between treatment groups may have varied according to APOE ɛ4 status. Previous evidence suggests that the association between chronic systemic inflammation and AD risk is significantly greater for e4 allele carriers [41], indicating that treatment targeting inflammation in these individuals may be particularly beneficial. This study found no interaction between treatment group and ɛ4 status, suggesting that there was no difference in cognitive change among TNFi users with or without the ɛ4 allele. However, the number of ɛ4 carriers was relatively small (indeed, only 11 participants in the TNFi group possessed the ɛ4 allele); future studies with larger samples are encouraged.

A major strength of the present study was its longitudinal design, which allowed us to evaluate the potential benefits of TNFi treatment over time, unlike previous studies which were predominantly cross-sectional. However, longitudinal studies are subject to attrition bias, something that was exacerbated in the present study by the impact of the COVID-19 pandemic. Power analysis was calculated assuming a 10% dropout rate. In practice, dropout rates were higher and there was some evidence that those who did not complete visit 4 had lower scores on the cognitive assessments at baseline (specifically FCSRT-IR free recall and MoCA). In other words, those with initially lower levels of cognitive function were less likely to complete the study. It is thus important to note that results may not be fully generalizable beyond higher functioning individuals. Furthermore, there may be other causes of missing data that were not measured, leading to biased model estimates. Future studies should endeavor to collect data on a wide range of potential causes of missing data in order to make a missing at random assumption more plausible.

Another possible explanation for the lack of observed decline is a general practice effect for all participants. Assessments were repeated four times over the duration of the study (five times, including screening, for the MoCA), so participants were likely to be more familiar and comfortable with the procedures by the end of the study. To account for this, different versions of the MoCA were used for screening and home visits, and different versions of the FCSRT-IR were used at each administration (as in previous longitudinal studies, e.g., [26]). Nevertheless, familiarity with the study procedure and interviewers may have made participants more relaxed during later assessments, which may have led to improvements in performance.

Analyses were adjusted for a number of covariates that could have influenced cognitive performance, including age, pain level, mood and functional status. However other possible confounders were not measured, such as medication adherence and length of time on medication. In practice this meant that length of time on treatment likely varied among participants within a given treatment group. It is likely that some participants had only recently commenced TNFi treatment, and thus potential effects may not have been evident. Additionally, it is likely that those participants on TNFi treatment initiated this treatment due to more severe prior disease activity. There is evidence that higher levels of disease activity may be linked to cognitive impairment [29]; a such, including participants who had just commenced treatment may have introduced some uncontrolled bias into the analyses. Collecting data on these unmeasured variables would allow future observational studies to control any influence they may have on cognitive change.

Additional research will be necessary to fully evaluate the long-term effects of TNFi treatment on cognitive function in RA patients. The 18-month duration of the present study did not allow for the tracking of long-term changes typically observed in people with AD. Longitudinal studies conducted over a greater length of time may be necessary to track meaningful differences in cognitive decline. Additionally, the inclusion of a control group of individuals without RA would allow a direct comparison of observed change in RA patients to that of a non-RA population over the same duration. Such studies could also consider progression from MCI to AD as a measurable outcome, and examine whether participants on a TNFi are at greater risk of developing AD. Larger sample sizes would also allow for the separation of participants taking a TNFi in combination with a csDMARD versus those taking just a TNFi.

In summary, no evidence was found to suggest that TNFi treatment was associated with reduced cognitive decline in RA patients with MCI when compared to csDMARD treatment alone. The lack of significant decline observed in both groups suggests both drugs may be protective. Given the urgent need to develop effective treatments for AD, further investigation of the beneficial effects of anti-inflammatory drugs is necessary.

AUTHOR CONTRIBUTIONS

Calum Marr (Data curation; Formal analysis; Writing –original draft; Writing –review & editing); Bethany McDowell (Conceptualization; Data curation; Investigation; Methodology; Project administration; Writing –review & editing); Clive Holmes (Conceptualization; Funding acquisition; Investigation; Methodology; Resources; Supervision; Writing –review & editing); Christopher J. Edwards (Conceptualization; Funding acquisition; Investigation; Methodology; Resources; Supervision; Writing –review & editing); Christopher Cardwell (Conceptualization; Data curation; Formal analysis; Funding acquisition; Investigation; Methodology; Supervision; Writing –review & editing); Michelle McHenry (Conceptualization; Investigation; Resources; Writing –review & editing); Gary Meenagh (Conceptualization; Investigation; Resources; Writing –review & editing); Jessica L. Teeling (Conceptualization; Investigation; Resources; Writing –review & editing); Bernadette McGuinness (Conceptualization; Funding acquisition; Investigation; Methodology; Project administration; Resources; Supervision; Writing –review & editing).

Footnotes

ACKNOWLEDGMENTS

We thank all participants for taking part in this study, as well as all doctors, nurses, and psychologists within BHSCT, NHSCT, UHS, Southern Health Foundation NHS Trust, and Northern Ireland Clinical Research Network (NICRN) who contributed to recruitment. We thank Dr Lena Azbel-Jackson, Patricia Quinn, Michael McAlinden, Sylvia Clyde, and Craig Woods for their contributions to data collection and collation. We also thank the research network volunteers assigned to the RESIST study: Gordon Kennedy, Bill Megraw, and Kieran Hanna, for their contributions to the development and design of the study.

This work was submitted as part of Bethany McDowell’s PhD thesis ‘TNF inhibitors for the prevention of Alzheimer’s disease’. .

FUNDING

This study was supported by funding from the Alzheimer’s Society (TNF Inhibitors for the prevention of Dementia: 318 [AS-PG-16-023]) and Health and Social Care Research and Development Northern Ireland. Research network volunteers were involved in the design of the study proposal. The funders were not involved in the collection, analysis or interpretation of data, the writing of this manuscript or in the decision to submit this article for publication. Queen’s University Belfast agreed to act as sponsor for this study.

CONFLICT OF INTEREST

CE reports grants for research support from Abbvie, Biogen, and Samsung for work that was not directly related to this research but was in the field of biological therapies. He was also paid consulting fees from Abbvie, GSK, and Gilead for work that was not directly related to this research. CE discloses receiving payments from Abbvie, Pfizer, Biogen, Fresenius, Galapagos, Lilly, and Janssen for lectures/presentations in the last 3 years. MMcH reports receiving funding to attend conferences from a number of companies who produce biologics over the last 3 years and discloses that registration fees were covered by UCB. BMcG discloses receiving honorarium from Nutricia, Biogen, Roche, and Eisai for consultancy work. All other authors have no conflicts of interest to disclose.

DATA AVAILABILITY

The data supporting the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

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

References

Stewart

, Stuart

, Wilkinson

, van Gils

, Deng

, Halle

, Rayner

, Boyer

, Zhong

, Frazier

, Lacy-Hulbert

, Khoury

, Golenbock

, Moore

(2010) CD36 ligands promote sterile inflammation through assembly of a Toll-like receptor 4 and 6 heterodimer. Nat Immunol 11, 155–161.

El Khoury

, Moore

, Means

, Leung

, Terada

, Toft

, Freeman

, Luster

(2003) CD36 mediates the innate host response to β-amyloid. J Exp Med 197, 1657–1666.

Kinney

, Bemiller

, Murtishaw

, Leisgang

, Salazar

, Lamb

(2018) Inflammation as a central mechanism in Alzheimer’s disease. Alzheimers Dement (N Y) 4, 575–590.

Heneka

, Carson

, Khoury

, Landreth

, Brosseron

, Feinstein

, Jacobs

, Wyss-Coray

, Vitorica

, Ransohoff

, Herrup

, Frautschy

, Finsen

, Brown

, Verkhratsky

, Yamanaka

, Koistinaho

, Latz

, Halle

, Petzold

, Town

, Morgan

, Shinohara

, Perry

, Holmes

, Bazan

, Brooks

, Hunot

, Joseph

, Deigendesch

, Garaschuk

, Boddeke

, Dinarello

, Breitner

, Cole

, Golenbock

, Kummer

(2015) Neuroinflammation in Alzheimer’s disease. Lancet Neurol 14, 388–405.

Swardfager

, Lanctôt

, Rothenburg

, Wong

, Cappell

, Herrmann

(2010) A meta-analysis of cytokines in Alzheimer’s disease. Biol Psychiatry 68, 930–941.

Holmes

, Cunningham

, Zotova

, Woolford

, Dean

, Kerr

, Culliford

, Perry

(2009) Systemic inflammation and disease progression in Alzheimer disease. Neurology 73, 768.

Tobinick

, Gross

, Weinberger

, Cohen

(2006) TNF-alpha modulation for treatment of Alzheimer’s disease: A 6-month pilot study. MedGenMed 8, 25.

Butchart

, Brook

, Hopkins

, Teeling

, Püntener

, Culliford

, Sharples

, Sharif

, McFarlane

, Raybould

, Thomas

, Passmore

, Perry

, Holmes

(2015) Etanercept in Alzheimer disease: A randomized, placebo-controlled, double-blind, phase 2 trial. Neurology 84, 2161

Ungprasert

, Wijarnpreecha

, Thongprayoon

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

10.

Mason

, Holmes

, Edwards

(2018) Inflammation and dementia: Using rheumatoid arthritis as a model to develop treatments? Autoimmun Rev 17, 919–925.

11.

Judge

, Garriga

, Arden

, Lovestone

, Prieto-Alhambra

, Cooper

, Edwards

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

12.

Chou

M-H

, Wang

J-Y

, Lin

C-L

, Chung

W-S

(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.

13.

Singh

, Furst

, Bharat

, Curtis

, Kavanaugh

, Kremer

, Moreland

, O’Dell

, Winthrop

, Beukelman

, Bridges

Jr , Chatham

, Paulus

, Suarez-almazor

, Bombardier

, Dougados

, Khanna

, King

, Leong

, Matteson

, Schousboe

, Moynihan

, Kolba

, Jain

, Volkmann

, Agrawal

, Bae

, Mudano

, Patkar

, Saag

(2012) 2012 Update of the 2008 American College of Rheumatology recommendations for the use of disease-modifying antirheumatic drugs and biologic agents in the treatment of rheumatoid arthritis. Arthritis Care Res (Hoboken) 64, 625–639.

14.

Yee

, Filer

, Pace

, Douglas

, Situnayake

, Rowe

, on behalf of the West Midlands Rheumatology Services and Training Committee (2003) The prevalence of patients with rheumatoid arthritis in the West Midlands fulfilling the BSR criteria for anti-tumour necrosis factor therapy: An out-patient study. Rheumatology (Oxford) 42, 856–859.

15.

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.

16.

McGuinness

, Holmes

, Mirakhur

, Kearsley-Fleet

, Vieira

, Watson

, Hyrich

(2018) The influence of TNF inhibitors on dementia incidence in patients with rheumatoid arthritis; an analysis from the BSRBR-RA. Int J Geriatr Psychiatry 33, 556–558.

17.

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.

18.

Zheng

, Fillmore

, Ramos-Cejudo

, Brophy

, Osorio

, Gurney

, Qiu

, Au

, Perry

, Dubreuil

, Chen

, Qi

, Davis

, Do

, Xu

(2022) Potential long-term effect of tumor necrosis factor inhibitors on dementia risk: A propensity score matched retrospective cohort study in US veterans. Alzheimers Dement 18, 1248–1259.

19.

Vitturi

, Nascimento

BAC

, Alves

, de Campos

FSC

, Torigoe

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

20.

Petersen

, Morris

(2005) Mild cognitive impairment as a clinical entity and treatment target. Arch Neurol 62, 1160–1163.

21.

Mitchell

, Shiri-Feshki

(2009) Rate of progression of mild cognitive impairment to dementia –meta-analysis of 41 robust inception cohort studies. Acta Psychiatr Scand 119, 252–265.

22.

J-Q

, Tan

, Wang

H-F

, Tan

M-S

, Tan

, Xu

, Zhao

Q-F

, Wang

, Jiang

, Yu

J-T

(2016) Risk factors for predicting progression from mild cognitive impairment to Alzheimer’s disease: A systematic review and meta-analysis of cohort studies. J Neurol Neurosurg Psychiatry 87, 476.

23.

Raftery

, He

, Pearce

, Birchall

, Newton

, Blamire

, Isaacs

(2012) Disease activity and cognition in rheumatoid arthritis: An open label pilot study. Arthritis Res Ther 14, R263.

24.

Aletaha

, Neogi

, Silman

, Funovits

, Felson

, Bingham Iii

, Birnbaum

, Burmester

, Bykerk

, Cohen

, Combe

, Costenbader

, Dougados

, Emery

, Ferraccioli

, Hazes

JMW

, Hobbs

, Huizinga

TWJ

, Kavanaugh

, Kay

, Kvien

, Laing

, Mease

, Ménard

, Moreland

, Naden

, Pincus

, Smolen

, Stanislawska-Biernat

, Symmons

, Tak

, Upchurch

, Vencovský

, Wolfe

, Hawker

(2010) 2010 Rheumatoid arthritis classification criteria: An American College of Rheumatology/European League Against Rheumatism collaborative initiative. Arthritis Rheum 62, 2569–2581.

25.

Nasreddine

, Phillips

, Bédirian

, Charbonneau

, Whitehead

, Collin

, Cummings

, Chertkow

(2005) The Montreal Cognitive Assessment, MoCA: A brief screening tool for mild cognitive impairment. J Am Geriatr Soc 53, 695–699.

26.

Sussams

, Schlotz

, Clough

, Amin

, Simpson

, Abbott

, Beardmore

, Sharples

, Raybould

, Brookes

, Morgan

, Culliford

, Holmes

(2020) Psychological stress, cognitive decline and the development of dementia in amnestic mild cognitive impairment. Sci Rep 10, 3618.

27.

Grober

, Buschke

(1987) Genuine memory deficits in dementia. Dev Neuropsychol 3, 13–36.

28.

Prevoo

MLL

, Van’T Hof

, Kuper

, Van Leeuwen

, Van De Putte

LBA

, Van Riel

PLCM

(1995) Modified disease activity scores that include twenty-eight-joint counts development and validation in a prospective longitudinal study of patients with rheumatoid arthritis. Arthritis Rheum 38, 44–48.

29.

McDowell

, Marr

, Holmes

, Edwards

, Cardwell

, McHenry

, Meenagh

, McGuinness

(2022) Prevalence of cognitive impairment in patients with rheumatoid arthritis: A cross sectional study. BMC Psychiatry 22, 777.

30.

Yesavage

, Brink

, Rose

, Lum

, Huang

, Adey

, Leirer

(1982) Development and validation of a geriatric depression screening scale: A preliminary report. J Psychiatr Res 17, 37–49.

31.

Yaffe

, Blackwell

, Gore

, Sands

, Reus

, Browner

(1999) Depressive symptoms and cognitive decline in nondemented elderly women: A prospective study. Arch Gen Psychiatry 56, 425–430.

32.

Bruce

, Fries

(2005) The Health Assessment Questionnaire (HAQ). Clin Exp Rheumatol 23, S14–S18.

33.

R Core Team (2022) R: A language and environment for statistical computing, R Foundation for Statistical Computing, Vienna, Austria.

34.

Vickers

, Altman

(2001) Analysing controlled trials with baseline and follow up measurements. BMJ 323, 1123.

35.

Cro

, Morris

, Kenward

, Carpenter

(2020) Sensitivity analysis for clinical trials with missing continuous outcome data using controlled multiple imputation: A practical guide. Stat Med 39, 2815–2842.

36.

Vassilaki

, Crowson

, Davis Iii

, 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.

37.

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.

38.

Newby

, Prieto-Alhambra

, Duarte-Salles

, Ansell

, Pedersen

, van der Lei

, Mosseveld

, Rijnbeek

, James

, Alexander

, Egger

, Podhorna

, Stewart

, Perera

, Avillach

, Grosdidier

, Lovestone

, Nevado-Holgado

(2020) Methotrexate and relative risk of dementia amongst patients with rheumatoid arthritis: A multi-national multi-database case-control study. Alzheimers Res Ther 12, 38.

39.

Kern

, Lovestone

, Cepeda

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

40.

Smolen

, Aletaha

, Bijlsma

JWJ

, Breedveld

, Boumpas

, Burmester

, Combe

, Cutolo

, de Wit

, Dougados

, Emery

, Gibofsky

, Gomez-Reino

, Haraoui

, Kalden

, Keystone

, Kvien

, McInnes

, Martin-Mola

, Montecucco

, Schoels

, van der Heijde

(2010) Treating rheumatoid arthritis to target: Recommendations of an international task force. Ann Rheum Dis 69, 631.

41.

Tao

, Ang

TFA

, DeCarli

, Auerbach

, Devine

, Stein

, Zhang

, Massaro

, Au

, Qiu

(2018) Association of chronic low-grade inflammation with risk of Alzheimer disease in ApoE4 carriers. JAMA Netw Open 1, e183597–e183597.