Impact of Different Diagnostic Measures on Drug Class Association with Dementia Progression Risk: A Longitudinal Prospective Cohort Study

Abstract

Background:

The Clinical Dementia Rating Scale Sum of Boxes (CDRSOB) score is known to be highly indicative of cognitive-functional status and is regularly employed for clinical and research purposes.

Objective:

Our aim is to determine whether CDRSOB is consistent with clinical diagnosis in evaluating drug class associations with risk of progression to mild cognitive impairment (MCI) and dementia.

Methods:

We employed weighted Cox regression analysis on longitudinal NACC data, to identify drug classes associated with disease progression risk, using clinical diagnosis and CDRSOB as the outcome.

Results:

Aspirin (antiplatelet/NSAID), angiotensin II inhibitors (antihypertensive), and Parkinson’s disease medications were significantly associated with reduced risk of progression to MCI/dementia and Alzheimer’s disease medications were associated with increased MCI-to-Dementia progression risk with both clinical diagnosis and CDRSOB as the outcome. However, certain drug classes/subcategories, like anxiolytics, antiadrenergics, calcium (Ca²⁺) channel blockers, and diuretics (antihypertensives) were associated with reduced risk of disease progression, and SSRIs (antidepressant) were associated with increased progression risk only with CDRSOB. Additionally, metformin (antidiabetic medication) was associated with reduced MCI-to-Dementia progression risk only with clinical diagnosis as the outcome.

Conclusions:

Although the magnitude and direction of the effect were primarily similar for both diagnostic outcomes, we demonstrate that choice of diagnostic measure can influence the significance of risk/protection attributed to drug classes and consequently the conclusion of findings. A consensus must be reached within the research community with respect to the most accurate diagnostic outcome to identify risk and improve reproducibility.

Keywords

Alzheimer’s disease anticoagulants antidepressants antihypertensives CDRSOB dementia diagnosis metformin mild cognitive impairment risk factors

INTRODUCTION

Dementia is a complex disease with several subtypes and etiologies, and few effective therapeutics [1]. Therefore, extensive research has been carried out analyzing factors that may influence disease incidence and progression [2, 3]. Medications, in particular, may affect cognition in older adults [4, 5] due to increased drug sensitivity associated with age-related factors such as impaired liver metabolism and decreased renal function [4]. Moreover, factors like potentially inappropriate prescribing, drug interactions, and polypharmacy, which are common in dementia patients, complicate the assessment of specific drug classes on cognition [6].

Several pharmacoepidemiology studies have analyzed the relationship between medications and dementia risk with conflicting findings. Researchers have acknowledged that methodological differences substantially contribute to the variation in risk attributed to different medications. Study design, inclusion criteria, data preparation, and especially the diagnostic criteria used can influence outcomes [7 –9]. One study examining antidiabetic medications found protective effects of metformin on dementia risk [10], whereas another reported increased risk of cognitive impairment associated with metformin use [11]. For antihypertensives, one systematic review reported a significant association between reduced dementia risk and use of diuretics and angiotensin-converting enzyme (ACE) inhibitors [12], whereas a study by Rouch et al. [13] showed that Ca²⁺ channel blockers and renin–angiotensin system blockers were associated with prevention of dementia. Similarly, discrepant findings have been reported across studies regarding the association between anxiolytics, such as benzodiazepines, and cognitive decline [14, 15]. This is also the case for other drug classes including antidepressants, antipsychotics, hypolipidemic drugs, and non-steroidal anti-inflammatory drugs (NSAIDs) [16 –20]. We note significant variability in the diagnostic criteria used across studies, which undoubtedly impacts upon risk attribution leading to conflicting findings.

Clinical assessment of dementia involves detailed examination of medical history, cognitive tests, laboratory assays, psychiatric evaluation, and brain imaging to identify dementia subtype. Studies have shown that subtle differences in classification can arise from variations in diagnostic guidelines provided by the National Institute of Neurological and Communicative Disorders and Stroke-Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA), Diagnostic and Statistical Manual Of Mental Disorders (DSM) criteria, including DSM–III–R, DSM–IV, DSM-V, International Classification of Diseases (ICD) criteria, including ICD-9, ICD-10, ICD-11, and Cambridge Mental Disorders of the Elderly Examination (CAMDEX) criteria, and can influence clinical diagnosis leading to variation in dementia prevalence [7 , 22]. As a result of this variation, researchers often employ cognitive scores for analysis, especially in longitudinal studies, as they allow for identification of small changes in cognition over time. However, in clinical settings cognitive tests are mainly used to aid the diagnostic decision-making process.

The Clinical Dementia Rating (CDR^® Dementia Staging Instrument) scale, a cognitive assessment tool, is regularly used in clinical and research settings to gauge dementia severity. It provides a global score, and a more detailed CDRSOB score obtained through patient and informant interview based on the following cognitive and functional domains: memory, orientation, judgement & problem solving, community affairs, home & hobbies, and personal care [23]. Studies have reported moderate [24, 25], to good [22 , 27] correlation of the CDR scale with DSM and McKhann [28] diagnostic criteria, with one study reporting efficiency of CDRSOB in distinguishing MCI from dementia, for patients with CDR global score 0.5 [29]. However, it is unclear whether such cognitive-functional assessments are consistent with clinical diagnosis in terms of evaluating the benefits or risks of medications.

In this study, we investigated how the associations between drug exposure and risk of progressing to MCI or all-cause dementia vary between clinical diagnosis and CDRSOB scores. We analyzed several drug classes available in the National Alzheimer’s Coordinating Center (NACC) dataset, that are commonly prescribed to dementia patients for treatment of various comorbidities associated with the disorder [30]. These include antihypertensives, lipid lowering medication, NSAIDs, anticoagulants, antidepressants, anxiolytics, antidiabetics, Alzheimer’s disease (AD) medication, Parkinson’s disease (PD) medication, and subcategories of drugs within each class, to identify drugs significantly associated with progression to MCI and dementia.

METHODS

Data source

Archival data from the NACC, consisting of over 500 variables on genetic, lifestyle, and clinical features for 48,605 individuals was used in this study. Details about the NACC, recruitment of participants, and assessment process has been previously described [31, 32].

The NACC Uniform Data Set (UDS), comprising of data collected from September 2005 until June 2023 was used in our analysis. The NACC is approved by the University of Washington Institutional Review Board. Written, informed consent from all participants and co-participants included in the UDS was obtained by the Alzheimer’s disease research centers (ADRCs) where they completed their study visits. For the purpose of publication, patient consent was not required. The following drug classes were analyzed: antihypertensives, lipid lowering medication, anxiolytics/sedatives/hypnotics, antidepressants, NSAIDs, diabetes medication, anticoagulants/antiplatelets, AD medication, and PD medication.

The diagnostic category of participants was determined based on both clinical diagnosis and CDRSOB scores. In NACC-UDS Version 1 and 2, the process of clinical diagnosis for all-cause dementia relied on the diagnostic protocol of the ADRC, with centers generally using DSM-IV [33] or NINCDS-ADRDA guidelines. In NACC-UDS Version 3, the criteria for all-cause dementia were modified from McKhann et al. (2011) [34]. Diagnoses of MCI were established using the modified Petersen criteria [35]. Individuals included in the analyses were aged≥40 years.

Data preparation

A multi-time-point data preparation approach was employed to conduct progression analysis for reducing bias associated with variables measured at a single time point [36]. Clinical diagnosis and CDRSOB scores were identified across all visits for each participant to determine changes in cognitive status over time. Subsequently, the following five groups were identified; participants that were healthy across all visits (Remained Healthy), participants that were MCI across all visits (Remained MCI), those who progressed from Healthy-to-MCI, from Healthy-to-Dementia, and from MCI-to-Dementia. Individuals with single visits or a fluctuating diagnosis were excluded and only those with a clear progression trend were analyzed. CDRSOB scores were categorized as healthy: 0, MCI: 0.5–4.0 and dementia: 4.5–18 [29].

Since covariates like diabetes and traumatic brain injury can occur after baseline, observations across all visits before the individual progressed to MCI/dementia were examined to determine their presence/absence. In the case of education and number of years smoked, the maximum value across all visits (for stable groups) or before the individual progressed to MCI/dementia was used for analysis. Body mass index (BMI) was categorized as underweight-1 (<18.5 kg/m2), normal-2 (18.5–24.99 kg/m2), or overweight-3 (>24.99 kg/m2). Subsequently, transitions in BMI over time were determined by calculating the average of BMI categories (underweight, normal, or overweight) across all visits (for stable groups) or before the individual progressed to MCI/dementia, and qualitatively comparing this to the baseline category to determine increase/decrease/stable progression [36]. Figure 1 depicts the sample selection process for the different progression groups analyzed. Details about the multi-time-point analysis approach have been published previously [36].

Fig. 1

Data preparation process for progression groups within the NACC dataset. CDRSOB, Clinical Dementia Rating Scale Sum of Boxes; MCI, mild cognitive impairment; NACC, National Alzheimer’s Coordinating Center.

The major drug classes analyzed in this study, and subcategories of antihypertensives, are available in the UDS dataset as variables. Next, we used the researcher’s data dictionary (RDD) [37] to identify the common medication names stored in the UDS for the frequently prescribed medication subcategories within each drug class. Subsequently, we searched for specific drug terms (Supplementary Table 1) across all visits (for stable groups) or before the individual progressed to MCI/dementia, to determine prescription data for each individual. To ascertain drug exposure over time, those who reported taking a specific medication during at least two separate visits were categorized as medication users. All participants included in the analyses were using the respective medications for at least six months (self-reported). The medications were then categorized based on their mechanism of action. The analysis excluded medications that were administered via ophthalmic or topical routes.

Statistical analysis

D’Agostino-Pearson test was employed to assess normality of data for demographic analysis. In the case of continuous variables, significance of differences was evaluated using the Welch two sample t-test for normally distributed data and the Mann-Whitney U test for non-normally distributed data. For categorical variables, Pearson’s chi-square test with Yate’s continuity correction was applied to compare differences.

To explore the relationship between different drug classes and time to MCI/dementia incidence, survival analysis was employed in RStudio using the ‘coxphw’ package [38]. This package implements a weighted estimation in Cox regression as proposed by Schemper et al. [39], to provide estimates of average hazard ratios while accounting for non-proportionality. The Cox regression analysis accounts for bias associated with loss to follow-up by treating non-converters as censored observations. Additionally, logistic regression was employed to calculate balancing weights using the inverse probability weighting (IPW) method for drug classes to account for treatment imbalances in the data [40]. The analyses were adjusted for important confounders and risk factors of dementia including age, sex, race, years of education and smoking, alcoholism, hypertension, diabetes, cardiovascular disorders, depression, BMI, traumatic brain injury, and hearing impairment [2].

Further, to determine whether significant associations occur due to the medication or the underlying condition, respective comorbidities were also included in the model while analyzing subcategories of drug classes, for, e.g., PD in analysis of PD medications. Drug classes/subcategories with low sample sizes were not analyzed in the study. Adjustment for multiple hypothesis testing was achieved by applying the False Discovery Rate (FDR) using the Benjamini-Yekutieli correction method [41] to all covariates in each Cox regression model. FDR adjusted p-values (FDR p)<0.01 were considered statistically significant.

Statistical analyses were performed in RStudio (2022.07.2 + 576) on a Windows machine with eight memory cores. Codes are available on GitHub (https://github.com/DamanKaurT/NACC-medication-analysis-Oct23).

RESULTS

General characteristics

The average follow-up time for the groups analyzed using clinical diagnosis (and CDRSOB) was 5.5 (5.4) years for stable healthy, 4.9 (4.8) years for Healthy-to-MCI, 2.9 (3.1) years for stable MCI, 5.8 (5.9) years for Healthy-to-Dementia, and 2.5 (2.6) years for MCI-to-Dementia progression group.

Analysis revealed that individuals who progressed from Healthy-to-MCI and from Healthy-to-Dementia were less likely to be married (p < 0.0001; Table 1), a higher proportion of them suffered from cardiovascular disorders (p < 0.0001), they were relatively less educated (p < 0.0001), and they had a higher average of total years smoked compared to stable healthy individuals (p < 0.01; Table 1). Stable healthy individuals were less likely to suffer from hypertension (p < 0.01) and depression (p < 0.0001) compared to those who progressed to dementia (Table 1). Those who progressed from Healthy-to-Dementia were more likely to experience a decrease in BMI over time compared to stable healthy individuals (p < 0.0001; Supplementary Table 2).

Table 1

Demographic characteristics of study participants

	Stable healthy versus healthy-to-MCI	Stable healthy versus healthy-to-dementia	Stable MCI versus MCI-to-dementia
Clinical diagnosis (n)	9706 vs. 941	9706 vs. 781	2358 vs. 1318
CDRSOB (n)	8407 vs. 1461	8407 vs. 575	4778 vs. 2392
Age at baseline, mean (SD)
Clinical diagnosis	69.5 (9.4) vs. 75.3 (8.2)^****	69.5 (9.4) vs. 77.3 (8.6)^****	72.7 (8.6) vs. 74.6 (8.2)^****
CDRSOB	69.2 (9.3) vs. 73.9 (8.8)^****	69.2 (9.3) vs. 76.8 (9.0) ^****	72.3 (9.1) vs. 73.0 (8.9)^***
Years of education, mean (SD)
Clinical diagnosis	16.0 (2.8) vs. 15.5 (2.9)^****	16.0 (2.8) vs. 15.5 (2.9)^****	15.4 (3.2) vs. 15.7 (3.1)^*
CDRSOB	16.1 (2.7) vs. 15.9 (2.9)^*	16.1 (2.7) vs. 15.6 (2.8)^****	15.3 (3.3) vs. 15.6 (3.1)^***
Female (%)
Clinical diagnosis	66.9 vs. 61.8^**	66.9 vs. 65.4	49.4 vs. 44.9^*
CDRSOB	67.4 vs. 62.4^***	67.4 vs. 67.5	49.5 vs. 44.9^***
Married (%)
Clinical diagnosis	60.3 vs. 47.4^****	60.3 vs. 57.1^****	64.2 vs. 73.6^****
CDRSOB	61.0 vs. 55.6^****	61.0 vs. 57.0^****	66.2 vs. 76.9^****
Diabetes (%)
Clinical diagnosis	15.2 vs. 15.8	15.2 vs. 16.1	19.0 vs. 14.1^***
CDRSOB	14.7 vs. 14.1	14.7 vs. 16.2	19.1 vs. 11.5^****
Hypertension (%)
Clinical diagnosis	58.4 vs. 65.4^****	58.4 vs. 68.5^****	64.5 vs. 55.5^****
CDRSOB	58.0 vs. 60.5	58.0 vs. 64.2^**	63.1 vs. 53.6^****
Cardiovascular diseases (%)
Clinical diagnosis	37.0 vs. 47.1^****	37.0 vs. 51.9^****	44.7 vs. 42.6
CDRSOB	35.1 vs.45.0^****	35.1 vs. 47.8^****	46.5 vs. 40.1^****
Depression (%)
Clinical diagnosis	31.4 versus 33.5	31.4 vs.42.8^****	42.7 vs. 44.5
CDRSOB	29.4 versus 34.2^***	29.4 vs. 45.6^****	47.3 vs. 48.4
Number of years smoked, mean (SD)
Clinical diagnosis	8.6 (13.7) vs. 9.7 (14.6)^**	8.6 (13.7) vs. 10.5 (15.8)^***	10.1 (15.3) vs. 9.5 (14.5)
CDRSOB	8.5 (13.7) vs. 9.9 (14.7)^***	8.5 (13.7) vs. 10.2 (15.0)^**	10.1 (15.2) vs. 9.2 (14.1)

CDRSOB, Clinical dementia rating scale sum of boxes; MCI, mild cognitive impairment; SD, standard deviation. (^*p < 0.05, ^**p < 0.01, ^***p < 0.001, ^****p < 0.0001).

Compared to the stable MCI group, those who progressed from MCI-to-Dementia were more likely to be married (p < 0.0001), a higher proportion of them were males (p < 0.05), a lower proportion had a diagnosis of diabetes (p < 0.001) and hypertension (p < 0.0001; Table 1), and a larger percentage of them had a first degree relative with cognitive impairment (p < 0.001; Supplementary Table 2). Detailed results can be found in Supplementary Table 2.

Medication analysis

Using Cox regression with inverse probability of treatment weighting, we identified drug classes associated with risk of progression from Healthy-to-MCI, Healthy-to-Dementia, and MCI-to-Dementia. The hazard ratios were calculated based on drug exposure and time to MCI or dementia incidence/last recorded observation for non-converters using both CDRSOB and clinical diagnosis as the diagnostic outcome. Complete regression results can be found in Supplementary Table 3.

Drug classes associated with risk of progression from healthy to MCI

Comparing stable healthy individuals to those who progressed from Healthy-to-MCI, anticoagulants/antiplatelets in general were significantly associated with decreased progression risk with CDRSOB as the outcome (hazard ratio (HR): 0.75, 95% confidence interval (CI) 0.64–0.89, FDR p < 0.01; Table 2). With clinical diagnosis, similar effect sizes were observed, however with a lower significance level (HR: 0.75, 95% CI 0.60–0.93, FDR p < 0.05). While analyzing subclasses of antiplatelets, aspirin, which is also classified as an NSAID, was associated with decreased risk of progression with both clinical diagnosis and CDRSOB as the outcome (FDR p < 0.01; Table 2). While analyzing NSAIDs, significantly reduced progression risk was observed with both clinical diagnosis and CDRSOB as the outcome (FDR p < 0.01). Additionally, propionic acid derivatives (NSAID) and diuretics (antihypertensive) were associated with reduced progression risk with lower significance (FDR p < 0.05). Metformin, an antidiabetic drug, was associated with increased Healthy-to-MCI progression risk with lower statistical significance (FDR p < 0.05).

Table 2

IPW HR with 95% CIs for the effect of drug classes on progression from cognitively healthy to MCI

Stable healthy versus healthy-to-MCI
Drug class	Clinical diagnosis	FDR	CDRSOB	FDR
	HR (95% CIs)	p	HR (95% CIs)	p
Anxiolytics	1.32 (0.99–1.76)		1.15 (0.91–1.46)
Anticoagulants/Antiplatelets	0.75 (0.60–0.93)	^*	0.75 (0.64–0.89)	^**
Aspirin	0.67 (0.54–0.83)	^**	0.76 (0.65–0.89)	^**
NSAIDs	0.69 (0.55–0.86)	^**	0.74 (0.63–0.87)	^***
Propionic acid derivatives	0.86 (0.67–1.12)		0.74 (0.60–0.91)	^*
Antidepressants	1.04 (0.83–1.30)		1.01 (0.82–1.23)
SSRIs	1.05 (0.76–1.47)		1.06 (0.79–1.44)
Antihypertensives	1.15 (0.85–1.55)		0.78 (0.59–1.04)
Antiadrenergics	0.82 (0.54–1.23)		0.71 (0.51–0.98)
ACE inhibitors	0.90 (0.69–1.18)		0.84 (0.64–1.12)
Angiotensin II inhibitors	1.01 (0.64–1.59)		0.95 (0.60–1.49)
Beta blockers	0.81 (0.65–1.01)		1.00 (0.81–1.23)
Ca²⁺ channel blockers	0.73 (0.57–0.94)		0.97 (0.71–1.33)
Diuretics	0.78 (0.61–0.99)		0.77 (0.63–0.93)	^*
Antihypertensive combination therapy	0.72 (0.48–1.09)		0.82 (0.56–1.19)
Antidiabetic drugs	2.08 (0.74–5.81)		2.52 (1.31–4.83)	^*
Metformin	1.99 (0.94–4.23)		2.20 (1.31–3.72)	^*
Lipid lowering drugs	0.92 (0.66–1.27)		1.03 (0.81–1.30)
Statins	0.98 (0.71–1.35)		1.07 (0.84–1.35)

Ca²⁺, calcium; CDRSOB, Clinical dementia rating scale sum of boxes; CIs, confidence intervals; FDR, false discovery rate; HR, hazard ratio; MCI, mild cognitive impairment; NSAIDs, non-steroidal anti-inflammatory drugs; SSRIs, selective serotonin reuptake inhibitors. (^*p < 0.05, ^**p < 0.01, ^***p < 0.001, ^****p < 0.0001). Drug classes associated with risk of progression from healthy to dementia

Anxiolytics, lipid lowering medications, antidepressants, and other subclasses of antihypertensives (antiadrenergics, ACE inhibitors, angiotensin II inhibitors, beta blockers, Ca²⁺ channel blockers, combination therapy) were not significantly associated with risk of progression from healthy to MCI.

Next, we focused on analysis of participants that progressed from healthy to dementia. In the case of subcategories of antidepressants, selective serotonin reuptake inhibitors (SSRIs) were associated with increased risk of progression from Healthy-to-Dementia with only CDRSOB as the diagnostic outcome (HR: 1.95, 95% CI 1.36–2.79, FDR p < 0.01; Table 3). In the case of antihypertensives, angiotensin II inhibitors were associated with reduced risk of progression to dementia with both clinical diagnosis and CDRSOB scores (FDR p < 0.0001; Table 3). Antihypertensive combination therapy was also associated with a reduced progression risk with CDRSOB as the outcome, albeit with lower significance (FDR p < 0.05). Anxiolytics, anticoagulants/antiplatelets, NSAIDs, antidiabetic drugs, lipid lowering drugs, and other subcategories of antihypertensives (antiadrenergics, ACE inhibitors, beta blockers, Ca²⁺ channel blockers, diuretics) were not associated with risk of progression from Healthy-to-Dementia (Table 3).

Table 3

IPW HR with 95% CIs for the effect of drug classes on progression from cognitively healthy to dementia

Stable healthy versus healthy-to-dementia
Drug class	Clinical diagnosis	FDR	CDRSOB	FDR
	HR (95% CIs)	p	HR (95% CIs)	p
Anxiolytics	1.07 (0.77–1.48)		1.30 (0.88–1.92)
Anticoagulants/Antiplatelets	0.92 (0.72–1.17)		0.97 (0.73–1.29)
Aspirin	0.83 (0.65–1.05)		0.84 (0.65–1.10)
NSAIDs	0.85 (0.64–1.13)		0.77 (0.58–1.03)
Propionic acid derivatives	0.89 (0.67–1.17)		1.01 (0.75–1.36)
Antidepressants	1.09 (0.87–1.37)		1.35 (1.05–1.74)
SSRIs	1.29 (0.90–1.84)		1.95 (1.36–2.79)	^**
Antihypertensives	0.93 (0.66–1.31)		0.88 (0.64–1.20)
Antiadrenergics	0.79 (0.54–1.15)		0.73 (0.47–1.14)
ACE inhibitors	0.87 (0.66–1.16)		0.90 (0.64–1.27)
Angiotensin II inhibitors	0.50 (0.40–0.62)	^****	0.46 (0.35–0.59)	^****
Beta blockers	0.91 (0.65–1.26)		0.81 (0.47–1.40)
Ca²⁺ channel blockers	0.96 (0.72–1.28)		1.08 (0.77–1.52)
Diuretics	0.83 (0.64–1.08)		0.78 (0.61–1.01)
Combination therapy	1.00 (0.66–1.52)		0.61 (0.43–0.86)	^*
Antidiabetic drugs	0.76 (0.52–1.10)		0.93 (0.49–1.76)
Metformin	0.91 (0.68–1.22)		1.33 (0.93–1.92)
Lipid lowering drugs	1.25 (0.92–1.71)		0.98 (0.67–1.45)
Statins	1.12 (0.79–1.58)		1.02 (0.67–1.54)

Ca²⁺, calcium; CDRSOB, Clinical dementia rating scale sum of boxes; CIs, confidence intervals; FDR, false discovery rate; HR, hazard ratio; NSAIDs, non-steroidal anti-inflammatory drugs; SSRIs, selective serotonin reuptake inhibitors. (^*p < 0.05, ^**p < 0.01, ^***p < 0.001, ^****p < 0.0001).

Drug classes associated with risk of progression from MCI to dementia.

With respect to analysis of MCI-to-Dementia progression group, anxiolytics were significantly associated with reduced progression risk only with CDRSOB as the outcome (HR: 0.64, 95% CI 0.50–0.83, FDR p < 0.01). However, similar effect sizes were observed using clinical diagnosis (HR: 0.69, 95% CI 0.51–0.95; Table 4). In accordance with analysis comparing stable healthy individuals versus those who progressed from Healthy-to-MCI (Table 2), use of anticoagulants/antiplatelets, including aspirin specifically, was associated with reduced MCI-to-Dementia progression risk with both clinical diagnosis and CDRSOB as the outcome (FDR p < 0.0001; Table 4).

Table 4

IPW HR with 95% CIs for the effect of drug classes on progression from MCI to dementia

Stable MCI versus mci-to-dementia
Drug class	Clinical diagnosis	FDR	CDRSOB	FDR
	HR (95% CIs)	p	HR (95% CIs)	p
Anxiolytics	0.69 (0.51–0.95)		0.64 (0.50–0.83)	^**
Anticoagulants/Antiplatelets	0.71 (0.61–0.82)	^****	0.75 (0.67–0.84)	^****
Aspirin	0.66 (0.58–0.75)	^****	0.75 (0.68–0.84)	^****
NSAIDs	0.57 (0.50–0.65)	^****	0.66 (0.59–0.73)	^****
Propionic acid derivatives	0.70 (0.56–0.88)	^*	0.63 (0.53–0.74)	^****
Antidepressants	0.83 (0.70–0.98)		0.84 (0.74–0.96)	^*
SSRIs	0.84 (0.66–1.07)		0.95 (0.82–1.10)
Antihypertensives	0.84 (0.64–1.09)		0.77 (0.64–0.91)	^*
Antiadrenergics	0.74 (0.48–1.16)		0.56 (0.40–0.79)	^**
ACE inhibitors	0.67 (0.49–0.92)		0.92 (0.74–1.16)
Angiotensin II inhibitors	0.64 (0.48–0.87)	^*	0.59 (0.41–0.84)	^*
Beta blockers	0.86 (0.71–1.04)		0.84 (0.73–0.97)	^*
Ca²⁺ channel blockers	0.71 (0.52–0.98)		0.54 (0.43–0.69)	^****
Diuretics	0.69 (0.52–0.93)		0.69 (0.57–0.84)	^**
Combination therapy	1.34 (0.89–2.00)		1.20 (0.67–2.13)
Antidiabetic drugs	0.62 (0.38–1.01)		0.76 (0.37–1.58)
Metformin	0.47 (0.29–0.74)	^**	0.63 (0.28–1.39)
Lipid lowering drugs	0.67 (0.53–0.83)	^**	0.76 (0.63–0.91)
Statins	0.72 (0.59–0.89)	^**	0.78 (0.65–0.93)	^*
Alzheimer’s disease medication	1.62 (1.42–1.85)	^****	2.27 (2.04–2.54)	^****
Donepezil	1.50 (1.31–1.72)	^****	1.84 (1.67–2.04)	^****
Memantine	1.39 (1.15–1.68)	^**	1.74 (1.56–1.94)	^****
Parkinson’s disease medication	0.49 (0.31–0.78)	^**	0.52 (0.39–0.70)	^***
Levodopa	0.99 (0.59–1.67)		0.67 (0.41–1.09)

Overall, NSAIDs were associated with a reduced MCI-to-Dementia progression risk with both clinical diagnosis and CDRSOB as the outcome (FDR p < 0.0001). Propionic acid derivatives, a class of NSAIDs that include several non-prescription medications like ibuprofen and naproxen, were associated with reduced progression risk with CDRSOB (FDR p < 0.0001) and clinical diagnosis, with a lower level of significance. (FDR p < 0.05; Table 4).

In the case of antihypertensives, antiadrenergics (FDR p < 0.01), Ca²⁺ channel blockers (FDR p < 0.0001), and diuretics (FDR p < 0.01) were associated with reduced progression risk with only CDRSOB as the diagnostic outcome (Table 4). Antidiabetic and lipid lowering drugs, metformin (FDR p < 0.01) and statins (FDR p < 0.01), were associated with reduced progression risk only with clinical diagnosis as the outcome. In the case of statins, similar effect sizes were observed with CDRSOB scores as the outcome, albeit with lower significance (FDR p < 0.05; Table 4). Corresponding to results from Healthy-to-Dementia progression analysis (Table 3), angiotensin II inhibitors were associated with reduced MCI-to-Dementia progression risk, with both clinical diagnosis and CDRSOB as the outcome, but with lower significance (FDR p < 0.05). Additionally, with CDRSOB as the diagnostic outcome, antidepressants and beta blockers were associated with reduced progression risk with lower significance (FDR p < 0.05; Table 4).

Antihypertensive combination therapy and ACE inhibitors were not significantly associated with risk of MCI-to-Dementia progression.

Commonly prescribed AD medications—donepezil, a cholinesterase inhibitor (ChEI) and memantine, an N-methyl-D-aspartate (NMDA) receptor antagonist—were also analyzed while comparing stable MCI and MCI-to-Dementia groups. The results were consistent for both diagnostic outcomes with significantly increased risk associated with progression from MCI-to-Dementia (Table 4). Finally, analysis of PD medications revealed significantly reduced risk associated with MCI-to-Dementia progression with both clinical diagnosis (FDR p < 0.01) and CDRSOB score (FDR p < 0.001) as the outcome. However, while analyzing subcategories of PD medications, Levodopa (dopamine precursor) was not significantly associated with MCI-to-Dementia progression risk.

DISCUSSION

In this study, we demonstrate that the choice of diagnostic measures for MCI and dementia, i.e., clinical diagnosis and CDRSOB scores, can impact the significance of the observed risks or protective effects associated with drug classes in relation to disease progression. Among the drug classes analyzed, aspirin (antiplatelet/NSAID) was significantly associated with reduced Healthy-to-MCI and MCI-to-Dementia progression risk with both diagnostic outcomes. Angiotensin II inhibitors were linked to a reduced Healthy-to-Dementia progression risk with both diagnostic outcomes. The analysis of MCI-to-Dementia progression groups indicated an increased risk associated with AD medications and a reduced risk of progression associated with PD medications, considering both CDRSOB and clinical diagnosis as outcomes. However, SSRIs showed an increased risk of progression from Healthy-to-Dementia only with CDRSOB scores. Anxiolytics in general, antiadrenergics, Ca2 + channel blockers, and diuretics were associated with reduced MCI-to-Dementia progression risk only with CDRSOB scores. Additionally, metformin and lipid lowering drugs were significantly associated with reduced MCI-to-Dementia progression risk using clinical diagnosis as the diagnostic outcome. Overall, the magnitude and direction of effect for different drug classes was similar with both diagnostic variables, however, in several cases, the significance of association was considerably influenced by the choice of diagnostic criteria used.

In the NACC dataset, consensus-based clinical diagnosis was made using the criteria proposed by McKhann et al. [28, 34]. A previous study, analyzing the NACC dataset, reported diagnostic accuracy with 0.71 sensitivity and 0.81 specificity when comparing CDRSOB scores to dementia diagnosis [29]. We report that despite good correlation between the two diagnostic measures [26], with hazard ratios consistently in the same direction and similar effect sizes, the statistical significance of results was inconsistent between the two measures.

The CDRSOB scores offer a wider score range allowing better demarcation of subtle changes between different stages of disease progression. In the NACC cohort, a lower number of individuals are classified into Healthy-to-MCI group by clinical diagnosis (n = 941), compared to CDRSOB score (n = 1,461; Fig. 1). CDRSOB is therefore associated with more ‘diagnoses’ of MCI. In terms of a clinical diagnosis, there are no specific tests that confirm MCI; judgement is based on clinical evaluation and the exclusion of other causative factors such as hypothyroidism or vitamin B12 deficiency. These factors not only influence the prevalence of MCI and dementia within a cohort but may also introduce variation in risk analysis.

Anticoagulants/antiplatelets are usually prescribed for treatment and prevention of blood clots, strokes, and related cardiovascular diseases. These were significantly associated with reduced Healthy-to-MCI and MCI-to-Dementia risk with both diagnostic outcomes. Although several studies have found reduced risk of cognitive impairment associated with oral anticoagulants [42, 43], randomized controlled trials of antiplatelet therapy in general have not shown significant protective effects against cognitive impairment and dementia [44]. In particular, aspirin was associated with reduced progression risk to MCI and dementia in our analysis. In addition to its antiplatelet properties, aspirin is also classified as a salicylate (NSAID), which has been associated with protective potential against AD and vascular dementia [45]. In relation to NSAIDs in general and propionic acid derivatives, our study yielded similar results, where the drug classes were associated with reduced Healthy-to-MCI and MCI-to-Dementia risk. In the literature, conflicting outcomes have been reported in observational and experimental studies regarding the association between NSAIDs and dementia risk [46 –48]. However, it is suggested that NSAID use may modulate inflammatory pathways, which could be beneficial given that inflammation plays a role in AD development [49].

Analysis of PD medications in our study revealed a significantly reduced MCI-to-Dementia progression risk associated with this drug class (FDR p < 0.01; Table 4). Previous studies have shown that treatment of PD can have beneficial effect on cognition and activities of daily living [50, 51]. Additionally, PD-related MCI is clinically distinct and has a longer conversion period to dementia compared to AD-related MCI [52]. Anxiolytics in general were significantly associated with a reduced MCI-to-Dementia progression risk with CDRSOB (HR: 0.64, 95% CI 0.50–0.83, FDR p < 0.01). A similar but insignificant effect size was observed with clinical diagnosis as the outcome (HR: 0.69, 95% CI 0.51–0.95; Table 4). However, the limited sample sizes within subcategories of anxiolytics prevented further analysis.

In our study, angiotensin II inhibitors were associated with a reduced Healthy-to-Dementia risk with both diagnostic outcomes. Antiadrenergic agents, Ca²⁺ channel blockers, and diuretics were significantly associated with reduced MCI-to-Dementia risk with CDRSOB scores as the outcome (Table 4). Several studies have reported an increased risk of cognitive impairment associated with hypertension, and a subsequent reduced risk with antihypertensive treatment [12, 53]. However, there is ambiguity regarding the duration, specific drug subclass, and optimal dosage to prevent cognitive impairment [54]. A previous study analyzing the NACC dataset reported lower cognitive scores associated with lower systolic BP [55]. Hypertension is known to be associated with impaired cognition. The difference in significance observed between CDRSOB and clinical diagnosis for respective drug classes may suggest improved cognition associated with drug treatment; however, these medications may not necessarily reduce the risk of dementia. Other subclasses of antihypertensives, namely ACE inhibitors, beta blockers, and antihypertensive combination therapy were not found to be significantly associated with disease progression in our study.

The analysis of antidepressants revealed a significantly increased risk of Healthy-to-Dementia progression associated with SSRIs when CDRSOB scores were used as the diagnostic measure (Table 3). This finding is consistent with previous meta-analysis studies of observational data that reported an increased risk of dementia associated with SSRI treatment [56, 57].

We found that metformin, used for the treatment of type 2 diabetes, was associated with a reduced MCI-to-Dementia risk with only clinical diagnosis as the outcome (FDR p < 0.01). Despite the association of diabetes with lower cognitive function, several studies have reported a reduced dementia risk in individuals taking oral hypoglycemic medication, particularly metformin [58, 59].

Lipid lowering drugs, in particular statins, were associated with a reduced MCI-to-Dementia progression risk with clinical diagnosis as the outcome (FDR p < 0.01). These results are supported by systematic reviews and meta-analysis suggesting a protective role of statins against dementia [60]. Similar effect sizes for statins were also observed with CDRSOB as the outcome in our study, albeit with lower significance.

While focusing on specific AD drugs for MCI-to-Dementia progression, donepezil and memantine were significantly associated with an increased risk of disease progression with both clinical diagnosis and CDRSOB as the outcome. Previous studies have shown faster cognitive decline with ChEI therapy in individuals with MCI and early AD [61]. These associations could be due to confounding by indication, as patients with cognitive impairment are more likely to be prescribed these drugs, especially those with amnestic MCI, considered to be at higher risk of conversion to dementia. This also reflects prescribing patterns across clinics, where physicians are known to prescribe ChEIs and memantine during early stages of impairment, contrary to US FDA guidelines [61]. Surprisingly, it was observed in the dataset that some cognitively healthy individuals (n = 91) were taking AD medications. Further investigation revealed that these included participants with higher CDRSOB scores (>0) who had not been clinically diagnosed with dementia.

This study undertook a comprehensive evaluation of various drug classes using population-based data from NACC to investigate the relationship between medication exposure and cognitive impairment/dementia. This is an important research topic because medications as risk factors for dementia are generally poorly understood. The data preparation methodology adopted in this study addresses bias associated with using a single time point measurement of risk factors. Therefore, changes in comorbidity status and trends in factors such as BMI and smoking over time were analyzed, providing a more accurate representation of patient status. Additionally, the statistical methods used, namely Cox regression analysis with IPW, address bias associated with unbalanced drug exposure and loss to follow-up.

There were certain limitations in this study since exposure to the drug classes was analyzed irrespective of duration, dosage, and potential drug-drug interactions, as such data was unavailable. Therefore, our findings could be vulnerable to unmeasured confounders or confounding by indication, with AD drugs, for example. Prescribing patterns can also affect regression results. For instance, if certain drug classes are more or less likely to be prescribed, this can affect the significance of association with disease progression. Alternatively, the observed associations might also be influenced by low adherence to medications, which is common among dementia patients. If individuals stopped taking medications prior to the timeline covered by the NACC data, that could also affect the results. Data on adherence to therapy in follow-up studies is essential.

High levels of comorbidity and polypharmacy, in aged individuals, is a reality in clinical practice that adds complexity to studies analyzing progression to dementia. To address these issues, there is a need for more detailed longitudinal studies involving larger cohorts, with data on dosage and duration of exposure to each drug class. Such studies will allow analysis of the temporal relationship between disease development and prescription of drugs and will permit use of comprehensive drug interaction databases like DDInter [62]. Undoubtedly, further studies are required to identify the most relevant diagnostic measure and to inform prescribing patterns aimed at reducing modifiable risks of dementia. Additionally, participants with a fluctuating diagnosis were excluded from analysis to ensure a homogeneous study population and to maintain the consistency and reliability of data. Future studies exploring factors associated with alternating diagnoses are important and may be compared to findings from this study.

In summary, our study demonstrates that the significance of the association between different medications and dementia progression can vary depending on the diagnostic measure employed. CDRSOB scores, although comparatively less comprehensive, are part of the clinical diagnosis process, leading to a high correlation between CDRSOB and clinical diagnosis. Due to the unavailability of clinical diagnosis in some datasets, or a preference for more quantitative analysis to increase precision and reduce subjectivity, researchers often opt for cognitive scores. Researchers must therefore be aware of the limitations of choosing a particular diagnostic measure and exercise caution in interpreting the results.

AUTHOR CONTRIBUTIONS

Daman Kaur (Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Resources; Software; Validation; Writing – original draft); Magda Bucholc (Conceptualization; Data curation; Methodology; Project administration; Software; Supervision; Writing – review & editing); David P. Finn (Investigation; Project administration; Supervision; Writing – review & editing); Stephen Todd (Investigation; Methodology; Supervision; Writing – review & editing); KongFatt Wong-Lin (Conceptualization; Investigation; Methodology; Project administration; Resources; Supervision; Writing – review & editing).

Paula L. McClean (Conceptualization; Investigation; Methodology; Project administration; Resources; Supervision; Writing – review & editing).

Footnotes

ACKNOWLEDGMENTS

This work was supported by the European Union’s INTERREG VA Programme, managed by the Special EU Programmes Body (SEUPB (Centre for Personalised Medicine, IVA 5036)), with additional support by the Northern Ireland Functional Brain Mapping Project Facility (1303/101154803), funded by invest Northern Ireland and the University of Ulster (K.W.-L.), Alzheimer’s Research UK (ARUK) NI Pump Priming (M.B.,S.T.,K.W.-L.,P.L.M.), Ulster University Research Challenge Fund (M.B.,S.T.,K.W.-L.,M.B.), and the Dr George Moore Endowment for Data Science at Ulster University (M.B.). The views and opinions expressed in this paper do not necessarily reflect those of the European Commission or the SEUPB.

The NACC database is funded by NIA/NIH Grant U01 AG016976. NACC data are contributed by the NIA-funded ADCs: P30 AG019610 (PI Eric Reiman, MD), P30 AG013846 (PI Neil Kowall, MD), P50 AG008702 (PI Scott Small, MD), P50 AG025688 (PI Allan Levey, MD, PhD), P50 AG047266 (PI Todd Golde, MD, PhD), P30 AG010133 (PI Andrew Saykin, PsyD), P50 AG005146 (PI Marilyn Albert, PhD), P50 AG005134 (PI Bradley Hyman, MD, PhD), P50 AG016574 (PI Ronald Petersen, MD, PhD), P50 AG005138 (PI Mary Sano, PhD), P30 AG008051 (PI Thomas Wisniewski, MD), P30 AG013854 (PI M. Marsel Mesulam, MD), P30 AG008017 (PI Jeffrey Kaye, MD), P30 AG010161 (PI David Bennett, MD), P50 AG047366 (PI Victor Henderson, MD, MS), P30 AG010129 (PI Charles DeCarli, MD), P50 AG016573 (PI Frank LaFerla, PhD), P50 AG005131 (PI James Brewer, MD, PhD), P50 AG023501 (PI Bruce Miller, MD), P30 AG035982 (PI Russell Swerdlow, MD), P30 AG028383 (PI Linda Van Eldik, PhD), P30AG053760 (PI Henry Paulson, MD, PhD), P30 AG010124 (PI John Trojanowski, MD, PhD), P50 AG005133 (PI Oscar Lopez, MD), P50 AG005142 (PI Helena Chui, MD), P30 AG012300 (PI Roger Rosenberg, MD), P30 AG049638 (PI Suzanne Craft, PhD), P50 AG005136 (PI Thomas Grabowski, MD), P50 AG033514 (PI Sanjay Asthana, MD, FRCP), P50 AG005681 (PI John Morris, MD), P50 AG047270 (PI Stephen Strittmatter, MD, PhD).

FUNDING

This work was supported by the European Union’s INTERREG VA Programme, managed by the Special EU Programmes Body (SEUPB (Centre for Personalised Medicine, IVA 5036)), with additional support by the Northern Ireland Functional Brain Mapping Project Facility (1303/101154803), funded by invest Northern Ireland and the University of Ulster (K.W.-L.), Alzheimer’s Research UK (ARUK) NI Pump Priming (M.B.,S.T.,K.W.-L.,P.L.M.), Ulster University Research Challenge Fund (M.B.,S.T.,K.W.-L.,M.B.), and the Dr George Moore Endowment for Data Science at Ulster University (M.B.). The funders had no role in study design, data collection, and interpretation, or the decision to submit the work for publication.

CONFLICT OF INTEREST

The authors have no conflict of interest to report.

DATA AVAILABILITY

The processed data can be found in results section in the manuscript and Supplementary Tables. R codes for data preparation and analysis are available on GitHub (https://github.com/DamanKaurT/NACC-medication-analysis-Oct23). NACC asks investigators to not share the data with individuals who are not collaborators on the project for which the data was requested. This is partially because they have distinctions between commercial and non-commercial recipients in place due to the option for NACC participants to elect to decline sharing of their data with commercial entities. Additionally, NACC has a data use agreement in place to help prevent misuse of the data. Lastly, this is helpful in their tracking of proposals, publications and data requests. NACC data is available through request to any interested researcher regardless of commerciality. However, commercial requests will only be able to receive a subset of the data due to the consent restrictions mentioned above. At the time of the request submission, investigators will be asked to provide details of the proposal and will need to submit a data use agreement. Requests can be submitted on their website ().

Previously Published Datasets: National Alzheimer’s Coordinating Center: Besser et al. 2018, Morris et al. 2006, since 2005, , The NIA Alzheimer’s Disease Research Centers Program.

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

References

Arvanitakis

, Shah

, Bennett

(2019) Diagnosis and management of dementia: Review. JAMA 322, 1589–1599.

Livingston

, Huntley

, Sommerlad

, Ames

, Ballard

, Banerjee

, Brayne

, Burns

, Cohen-Mansfield

, Cooper

, Costafreda

, Dias

, Fox

, Gitlin

, Howard

, Kales

, Kivimäki

, Larson

, Ogunniyi

, Orgeta

, Ritchie

, Rockwood

, Sampson

, Samus

, Schneider

, Selbæk

, Teri

, Mukadam

(2020) Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. Lancet 396, 413–446.

Peters

, Booth

, Rockwood

, Peters

, D’Este

, Anstey

(2019) Combining modifiable risk factors and risk of dementia: A systematic review and meta-analysis. Open 9, e022846.

Barton

, Sklenicka

, Sayegh

, Yaffe

(2008) Contraindicated medication use among patients in a memory disorders clinic. Am J Geriatr Pharmacother 6, 147–152.

Nikaido

, Ellinwood

Jr , Heatherly

, Gupta

(1990) Age-related increase in CNS sensitivity to benzodiazepines as assessed by task difficulty. Psychopharmacology (Berl) 100, 90–97.

Delgado

, Bowman

, Clare

(2020) Potentially inappropriate prescribing in dementia: A state-of-the-art review since. BMJ Open 10, e029172.

Erkinjuntti

, Ostbye

, Steenhuis

, Hachinski

(1997) The effect of different diagnostic criteria on the prevalence of dementia. N Engl J Med 337, 1667–1674.

Jamsen

, Ilomáki

, Hilmer

, Jokanovic

, Tan

ECK

, Bell

(2016) A systematic review of the statistical methods in prospective cohort studies investigating the effect of medications on cognition in older people. Res Social Adm Pharm 12, 20–28.

Wancata

, Börjesson-Hanson

, Ostling

, Sjögren

, Skoog

(2007) Diagnostic criteria influence dementia prevalence. Am J Geriatr Psychiatry 15, 1034–1045.

10.

Chin-Hsiao

(2019) Metformin and the risk of dementia in type 2 diabetes patients. Aging Dis 10, 37–48.

11.

Moore

, Mander

, Ames

, Kotowicz

, Carne

, Brodaty

, Woodward

, Boundy

, Ellis

, Bush

, Faux

, Martins

, Szoeke

, Rowe

, Watters

, AIBL Investigators (2013) Increased risk of cognitive impairment in patients with diabetes is associated with metformin. Diabetes Care 36, 2981–2987.

12.

Shah

, Qureshi

, Johnson

, Parikh

, Schulz

, Kunik

(2009) Does use of antihypertensive drugs affect the incidence or progression of dementia? A systematic review. Am J Geriatr Pharmacother 7, 250–261.

13.

Rouch

, Cestac

, Hanon

, Cool

, Helmer

, Bouhanick

, Chamontin

, Dartigues

, Vellas

, Andrieu

(2015) Antihypertensive drugs, prevention of cognitive decline and dementia: A systematic review of observational studies, randomized controlled trials and meta-analyses, with discussion of potential mechanisms. CNS Drugs 29, 113–130.

14.

Salzman

(2020) Do benzodiazepines cause Alzheimer’s disease? Am J Psychiatry 177, 476–478.

15.

Verdoux

, Lagnaoui

, Begaud

(2005) Is benzodiazepine use a risk factor for cognitive decline and dementia? A literature review of epidemiological studies. Psychol Med 35, 307–315.

16.

Biringer

, Rongve

, Lund

(2009) A review of modern antidepressants effects on neurocognitive function. Curr Psychiatry Rev 5, 164–174.

17.

Hill

, Bishop

, Palumbo

, Sweeney

(2010) Effect of second-generation antipsychotics on cognition: Current issues and future challenges. Expert Rev Neurother 10, 43–57.

18.

Imbimbo

, Solfrizzi

, Panza

(2010) Are NSAIDs useful to treat Alzheimer’s disease or mild cognitive impairment? Front Aging Neurosci 2, 19.

19.

Schultz

, Patten

, Berlau

(2018) The role of statins in both cognitive impairment and protection against dementia: A tale of two mechanisms. Transl Neurodegener 7, 5.

20.

Zhang

, Wang

, Zhang

(2018) NSAID exposure and risk of Alzheimer’s disease: An updated meta-analysis from cohort studies. Front Aging Neurosci 10, 83.

21.

Berman

, Bursztajn

(1999) Clinical criteria for three types of dementia had low sensitivity and high specificity. Evid Based Ment Health 2, 91.

22.

Chaves

, Camozzato

, Godinho

, Kochhann

, Schuh

, de Almeida

, Kaye

(2007) Validity of the clinical dementia rating scale for the detection and staging of dementia in Brazilian patients. Alzheimer Dis Assoc Disord 21, 210–217.

23.

Morris

(1993) The clinical dementia rating (CDR): Current version and scoring rules. Neurology 43, 2412–2414.

24.

Forsell

, Fratiglioni

, Grut

, Viitanen

, Winblad

(1992) Clinical staging of dementia in a population survey: Comparison of DSM-III-R and the Washington University clinical dementia rating scale. Acta Psychiatr Scand 86, 49–54.

25.

Juva

, Sulkava

, Erkinjuntti

, Ylikoski

, Valvanne

, Tilvis

(1994) Staging the severity of dementia: Comparison of clinical (CDR, DSM-III-R), functional (ADL, IADL) and cognitive (MMSE) scales. Acta Neurol Scand 90, 293–298.

26.

Ding

, Bucholc

, Wang

, Glass

, Wang

, Clarke

, Bjourson

, Dowey

LRC

, O’Kane

, Prasad

, Maguire

, Wong-Lin

(2018) A hybrid computational approach for efficient Alzheimer’s disease classification based on heterogeneous data. Sci Rep 8, 9774.

27.

Lima

APV

, Castilhos

, Chaves

MLF

(2017) The use of the clinical dementia rating scale sum of boxes scores in detecting and staging cognitive impairment/dementia in Brazilian patients with low educational attainment. Alzheimer Dis Assoc Disord 31, 322–327.

28.

McKhann

, Drachman

, Folstein

, Katzman

, Price

, Stadlan

(1984) Clinical diagnosis of Alzheimer’s disease: Report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer’s disease. Neurology 34, 939–944.

29.

O’Bryant

, Lacritz

, Hall

, Waring

, Chan

, Khodr

, Massman

, Hobson

, Cullum

(2010) Validation of the new interpretive guidelines for the clinical dementia rating scale sum of boxes score in the national Alzheimer’s coordinating center database. Arch Neurol 67, 746–749.

30.

Vargese

, Halonen

, Raitanen

, Forma

, Jylhä

, Aaltonen

(2021) Comorbidities in dementia during the last years of life: A register study of patterns and time differences in Finland. Aging Clin Exp Res 33, 3285–3292.

31.

Besser

, Kukull

, Knopman

, Chui

, Galasko

, Weintraub

, Jicha

, Carlsson

, Burns

, Quinn

, Sweet

, Rascovsky

, Teylan

, Beekly

, Thomas

, Bollenbeck

, Monsell

, Mock

, Zhou

, Thomas

, Robichaud

, Dean

, Hubbard

, Jacka

, Schwabe-Fry

, Wu

, Phelps

, Morris

, Neuropsychology Work Group, Directors, Clinical Core leaders of the National Institute on Aging-funded US Alzheimer’s Disease Centers (2018) Version 3 of the national Alzheimer’s coordinating center’s uniform data set. Alzheimer Dis Assoc Disord 32, 351–358.

32.

Morris

, Weintraub

, Chui

, Cummings

, Decarli

, Ferris

, Foster

, Galasko

, Graff-Radford

, Peskind

, Beekly

, Ramos

, Kukull

(2006) The uniform data set (UDS): Clinical and cognitive variables and descriptive data from Alzheimer disease centers. Alzheimer Dis Assoc Disord 20, 210–216.

33.

Bell

(1994) DSM-IV: Diagnostic and statistical manual of mental disorders. JAMA 272, 828–829.

34.

McKhann

, Knopman

, Chertkow

, Hyman

, Jack

Jr , Kawas

, Klunk

, Koroshetz

, Manly

, Mayeux

, Mohs

, Morris

, Rossor

, Scheltens

, Carrillo

, Thies

, Weintraub

, Phelps

(2011) The diagnosis of dementia due to Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7, 263–269.

35.

Petersen

, Morris

(2005) Mild cognitive impairment as a clinical entity and treatment target. Arch Neurol 62, 1160–1167.

36.

Kaur

, Bucholc

, Finn

, Todd

, Wong-Lin

, McClean

(2020) Multi-time-point data preparation robustly reveals MCI and dementia risk factors. Alzheimers Dement (Amst) 12, e12116.

37.

National Alzheimer’s Coordinating Center (2015) Researchers Data Dictionary Version 3.0. https://files.alz.washington.edu/documentation/uds3-rdd.pdf, Last updated March, 2015, Accessed on April 20, 2023.

38.

Dunkler

, Ploner

, Schemper

, Heinze

(2018) Weighted cox regression using the R package coxphw. J Stat Softw 84, 1–26.

39.

Schemper

, Wakounig

, Heinze

(2009) The estimation of average hazard ratios by weighted Cox regression. Stat Med 28, 2473–2489.

40.

Chesnaye

, Stel

, Tripepi

, Dekker

, Fu

, Zoccali

, Jager

(2021) An introduction to inverse probability of treatment weighting in observational research. Clin Kidney J 15, 14–20.

41.

Benjamini

, Yekutieli

(2001) The control of the false discovery rate in multiple testing under dependency. Ann Stat 29, 1165–1188.

42.

Cheng

, Liu

, Li

(2018) Relationship of anticoagulant therapy with cognitive impairment among patients with atrial fibrillation: A meta-analysis and systematic review. J Cardiovasc Pharmacol 71, 380–387.

43.

Zeng

, Jiang

, Su

, Tan

, Wu

(2019) Anticoagulation in atrial fibrillation and cognitive decline: A systematic review and meta-analysis. Medicine (Baltimore) 98, e14499.

44.

Kitt

, Murphy

, Clarke

, Reddin

, Ferguson

, Bosch

, Whiteley

, Canavan

, Judge

, O’Donnell

(2023) Antiplatelet therapy and incident cognitive impairment or dementia-a systematic review and meta-analysis of randomised clinical trials. Age Ageing 52, afad197.

45.

Nguyen

TNM

, Chen

, Trares

, Stocker

, Holleczek

, Beyreuther

, Brenner

, Schöttker

(2022) Long-term low-dose acetylsalicylic use shows protective potential for the development of both vascular dementia and Alzheimer’s disease in patients with coronary heart disease but not in other individuals from the general population: Results from two large cohort studies. Alzheimers Res Ther 14, 75.

46.

Ali

, Ghouri

, Ans

, Akbar

, Toheed

(2019) Recommendations for anti-inflammatory treatments in Alzheimer’s disease: A comprehensive review of the literature. Cureus 11, e4620.

47.

Breitner

, Haneuse

, Walker

, Dublin

, Crane

, Gray

, Larson

(2009) Risk of dementia and AD with prior exposure to NSAIDs in an elderly community-based cohort. Neurology 72, 1899–1905.

48.

Côté

, Carmichael

, Verreault

, Lindsay

, Lefebvre

, Laurin

(2012) Nonsteroidal anti-inflammatory drug use and the risk of cognitive impairment and Alzheimer’s disease. Alzheimers Dement 8, 219–226.

49.

Newcombe

, Camats-Perna

, Silva

, Valmas

, Huat

, Medeiros

(2018) Inflammation: The link between comorbidities, genetics, and Alzheimer’s disease. J Neuroinflammation 15, 276.

50.

Ikeda

, Kataoka

, Ueno

(2017) Can levodopa prevent cognitive decline in patients with Parkinson’s disease? J Neurodegener Dis 6, 9–14.

51.

Koch

, Motta

, Bonnì

, Pellicciari

, Picazio

, Casula

, Maiella

, Di Lorenzo

, Ponzo

, Ferrari

, Scaricamazza

, Caltagirone

, Martorana

(2020) Effect of rotigotine vs placebo on cognitive functions among patients with mild to moderate Alzheimer disease: A randomized clinical trial. JAMA Netw Open 3, e2010372.

52.

Besser

, Litvan

, Monsell

, Mock

, Weintraub

, Zhou

, Kukull

(2016) Mild cognitive impairment in Parkinson’s disease versus Alzheimer’s disease. Parkinsonism Relat Disord 27, 54–60.

53.

Ding

, Davis-Plourde

, Sedaghat

, Tully

, Wang

, Phillips

, Pase

, Himali

, Gwen Windham

, Griswold

, Gottesman

, Mosley

, White

, Guðnason

, Debette

, Beiser

, Seshadri

, Ikram

, Meirelles

, Tzourio

, Launer

(2020) Antihypertensive medications and risk for incident dementia and Alzheimer’s disease: A meta-analysis of individual participant data from prospective cohort studies. Lancet Neurol 19, 61–70.

54.

, Tan

, Shen

, Xu

, Hou

, Dong

, Tan

, Yu

(2020) Blood pressure and risks of cognitive impairment and dementia: A systematic review and meta-analysis of 209 prospective studies. Hypertension 76, 217–225.

55.

Fiford

, Nicholas

, Biessels

, Lane

, Cardoso

, Barnes

(2020) High blood pressure predicts hippocampal atrophy rate in cognitively impaired elders. Alzheimers Dement (Amst) 2, e12035.

56.

Wang

, Tai

, Poly

, Islam

, Yang

, Wu

, Li Y

(2018) Increased risk of dementia in patients with antidepressants: A meta-analysis of observational studies. Behav Neurol 2018, 5315098.

57.

Wang

, Chen

, Chang

, Zhang

, Shao

, Guo

, Lo-Ciganic

(2023) Association between first-line antidepressant use and risk of dementia in older adults: A retrospective cohort study. Res Sq, doi:https://doi.org/10.21203/rs.3.rs-3266805/v1 [Preprint]. Posted September 13, 2023.

58.

Hsu

, Wahlqvist

, Lee

, Tsai

(2011) Incidence of dementia is increased in type 2 diabetes and reduced by the use of sulfonylureas and metformin. J Alzheimers Dis 24, 485–493.

59.

Campbell

, Stephenson

, de Courten

, Chapman

, Bellman

, Aromataris

(2018) Metformin use associated with reduced risk of dementia in patients with diabetes: A systematic review and meta-analysis. J Alzheimers Dis 65, 1225–1236.

60.

Olmastroni

, Molari

, De Beni

, Colpani

, Galimberti

, Gazzotti

, Zambon

, Catapano

, Casula

(2022) Statin use and risk of dementia or Alzheimer’s disease: A systematic review and meta-analysis of observational studies. Eur J Prev Cardiol 29, 804–814.

61.

Schneider

, Insel

, Weiner

, Alzheimer’s Disease Neuroimaging Initiative (2011) Treatment with cholinesterase inhibitors and memantine of patients in the Alzheimer’s disease neuroimaging initiative. Arch Neurol 68, 58–66.

62.

Xiong

, Yang

, Yi

, Wang

, Zhu

, Wu

, Lu

, Chen

, Liu

, Hou

, Cao

(2022) DDInter: An online drug-drug interaction database towards improving clinical decision-making and patient safety. Nucleic Acids Res 50, D1200–D1207.