Skip to main content
NCBI home page
Alzheimer's & Dementia : Translational Research & Clinical Interventions logoLink to Alzheimer's & Dementia : Translational Research & Clinical Interventions
. 2025 Dec 29;12(1):e70193. doi: 10.1002/trc2.70193

Critical evaluation of real‐world evidence of repurposable medicines in the Alzheimer's disease drug development pipeline using a target trial emulation

Reina Tonegawa‐Kuji 1,2, Ehud Karavani 3, Michael Danziger 3, Pengyue Zhang 4, Yuan Hou 1,2, Yadi Zhou 1,2, Marina Bykova 1, Andrew A Pieper 5,6,7,8,9,10, Michal Rosen‐Zvi 3,11, Jeffrey Cummings 12, Feixiong Cheng 1,2,13,✉
PMCID: PMC12746425  PMID: 41473419

Abstract

INTRODUCTION

Repurposing Food and Drug Administration (FDA)‐approved drugs could accelerate treatment development for Alzheimer's disease (AD).

METHODS

Using the MarketScan claims database (2011 to 2020), we applied a trial emulation approach in two base cohorts: (1) individuals with mild cognitive impairment (MCI cohort) and (2) individuals aged ≥70 years (over‐70 cohort). We evaluated drugs represented in clinical trials for AD, comparing them with same‐class or active comparators. Covariate‐adjusted hazard ratios (HRs) were estimated using inverse‐probability‐weighted Cox models.

RESULTS

A total of 6 out of 38 (16%) drugs in the MCI cohort and 10 out of 53 (19%) drugs in the over‐70 cohort were associated with a lower AD incidence versus same‐class comparators. Active comparator analyses indicated that bupropion (vs escitalopram; HR 0.57, 95% confidence interval [CI] [0.49, 0.66]), trazodone (vs sertraline; HR 0.82, 95% CI [0.74, 0.91]), venlafaxine (vs escitalopram; 0.72, 95% CI [0.62, 0.84]), and zolpidem (vs lorazepam; HR 0.69, 95% CI [0.56, 0.85]) were associated with a lower AD incidence in the MCI cohort; these four plus liraglutide were associated with a lower incidence of AD dementia in the over‐70 cohort (vs metformin; HR 0.74, 95% CI [0.59, 0.93]).

DISCUSSION

This is the first comprehensive set of trial emulations for FDA‐approved drugs that are represented in AD trials. Findings may inform future trial designs.

Highlights

  • Repurposing FDA‐approved drugs originally developed for other diseases could accelerate treatment development for AD.

  • We identified repurposable drugs that are in current or complete clinical trials of AD and emulated trials for these agents using a large‐scale insurance claims‐based database.

  • Among 54 drugs evaluated, 6/38 (16%) drugs in the MCI cohort and 10/53 (19%) in the over‐70 cohort were associated with reduced AD incidence versus same‐class comparators. Active comparator analyses indicated that bupropion, trazodone, venlafaxine, and zolpidem were associated with reduced AD incidence in the MCI cohort; these four plus liraglutide were associated with a lower incidence of AD dementia in the over‐70 cohort.

  • A minority of repurposed table drugs that are in current or completed clinical trials for AD and meet criteria for inclusion in this study showed no effect in our trial emulation studies.

  • This is the first comprehensive set of trial emulations for FDA‐approved drugs that are represented in AD trials. Building on our findings, integrating real‐world evidence can inform future trials and accelerate drug development for AD.

Keywords: Alzheimer's disease, drug repurposing, liraglutide, real‐world evidence, target trial emulation

1. INTRODUCTION

Alzheimer's disease (AD) is a progressive neurodegenerative disorder, affecting ∼6.5 million individuals in the United States. 1 Mild cognitive impairment (MCI), a common precursor to AD, affects approximately 22% of individuals over the age of 65, with nearly half progressing to AD. 2 , 3

Numerous clinical trials have been conducted to develop therapies for AD, leading to recent approvals of anti‐amyloid antibodies and brexpiprazole for agitation associated with AD. 4 , 5 , 6 , 7 , 8 , 9 Repurposing Food and Drug Administration (FDA)‐approved drugs originally developed for other indications may accelerate treatment development. There are 46 repurposed agents in the 2025 AD drug development pipeline, comprising 33% of drugs in trials. 4 Despite the fact that many trials have tested repurposable agents, few have meaningfully modified cognitive function, and none has been approved aside from brexpiprazole for agitation.

Evidence informing clinical trials for drug repurposing typically derives from observational or experimental studies but is scattered across sources, limiting robust inference. To address this gap, we leveraged a large‐scale claims database to emulate clinical trials, aiming to systematically assess the potential of FDA‐approved drugs being evaluated in AD clinical trials for their associations with AD dementia incidence and disease progression. Similar high‐throughput target trial emulations for AD drug repurposing have been reported. 10 This approach also informs pragmatic randomized controlled trial (RCT) redesigns and sample size estimations. 11 Using real‐world evidence, we aimed to prioritize candidates and advance therapeutic development.

2. METHODS

2.1. Drugs of interest and comparators

We systematically searched ClinicalTrials.gov 12 for active and completed clinical trials for AD drugs (detailed in Text S1). RxNorm Concept Unique Identifier (RxCUI) was used to extract cohorts. 13

Comparators were drugs in the same Anatomical Therapeutic Chemical (ATC) second‐level class, which categorizes drugs by therapeutic use (e.g., A10 for diabetes). For drugs assigned to multiple ATC‐2 classes, comparisons were conducted separately for each class. Drugs associated with a significantly lower incidence of AD dementia (defined as an upper 95% confidence interval [CI] < 0 for differences, <1 for risk/hazard ratios [HRs], and q value < 0.1) underwent additional active comparator analyses mirroring the design commonly used in RCTs using commonly prescribed drugs for the same indication but with different biological mechanisms.

2.2. Study design, data sources, and cohorts

Using a target trial emulation framework, we conducted a population‐based comparative effectiveness research using the MarketScan Research Databases of health insurance claims (Figure 1A, Table S1), 14 which includes records on >171 million individuals across the United States enrolled in Medicare, Medicaid, or commercial insurance from 2011 to 2020.

FIGURE 1.

FIGURE 1

Open in a new tab

Overview of study design. (A) Data source, cohorts, and comparators. Using the MarketScan data, we created two cohorts: (1) MCI cohort, which included individuals diagnosed with MCI regardless of age; and (2) over‐70 cohort, which included individuals aged 70 years or older. From each cohort, individuals who had taken drugs tested in past AD‐related trials or corresponding controls were included. (B) Study timeline. The index date is defined as the first day the drug was prescribed. The baseline period is the 1‐year period preceding the index date, during which baseline characteristics were collected. Following the index date, a follow‐up period of up to a maximum of 3 years was used to track whether a diagnosis of AD was made. (C) The left panel illustrates the hypothesized relationship between drug use (treatment), the suppression of AD onset (outcome), and covariates. The arrows indicate the hypothesized causal pathways. The right panel shows a conceptual Love plot illustrating covariate balance before and after weighting; it was considered well balanced when the weighted absolute mean standardized difference was <0.2 across all covariates. *Events refer to cognitive function‐related events. (D) An example of the inverse‐probability‐weighted cumulative curves. When the weighted cumulative probability for the drug group and control group was d% and c%, respectively, the cumulative difference was calculated as (d−c)%, and the ratio was calculated as d/c. A weighted Cox proportional hazards model was also used to estimate the hazard ratio. AD, Alzheimer's disease; MCI, mild cognitive impairment.

We constructed two base cohorts: the mild cognitive impairment (MCI) cohort, to evaluate associations with AD progression, and the over‐70 cohort, to evaluate associations with AD dementia incidence. For the MCI cohort, we included individuals with diagnoses of MCI regardless of their age and excluded individuals with less than 1 year of coverage and those without prescription coverage (Figures 1A and S1). For the over‐70 cohort, insurance holders were required to be aged >70 years at any point during the observation period, regardless of their comorbidities (Figure S2). Individuals with less than 4 years of coverage and without prescription coverage or prior AD diagnosis were excluded.

Finally, from each cohort, individuals prescribed a drug of interest or a comparator drug were included in each simulated trial. International Classification of Diseases (ICD) codes used for cohort definitions are listed in Table S2.

2.3. Index date, follow‐up, and outcomes

The index date (time zero) was defined for the treatment group as the first prescription of the study drug after eligibility criteria were met (Figure 1B). For the control group, the index date was based on the first prescription of a randomly selected drug from the same ATC second‐level group or an active comparator drug.

RESEARCH IN CONTEXT

  1. Systematic review: The authors reviewed active and completed clinical trials for AD using ClinicalTrials.gov and traditional literature sources. Repurposing Food and Drug Administration (FDA)‐approved drugs originally developed for other diseases could accelerate treatment development for AD. Despite numerous trials, few agents have shown meaningful effects on cognitive decline, and none have been approved. Furthermore, evidence informing trial designs is often fragmented and scattered.

  2. Interpretation: Bupropion, trazodone, venlafaxine, and zolpidem were associated with a lower rate of AD progression, while these four drugs in addition to liraglutide were associated with reduced AD incidence. A minority of repurposable drugs tested in clinical trials for AD showed no effect in our trial emulations.

  3. Future directions: Trial emulation may help identify repurposed drugs more likely to succeed and guide more practical clinical trial designs. Building on our findings, integrating real‐world evidence can inform future trials and accelerate drug development for AD.

Participants were followed from their index date until their first diagnosis of AD dementia, end of healthcare enrollment, or 3 years, whichever occurred first. We selected a 3‐year follow‐up period, longer than the 18 months commonly used in RCTs evaluating cognitive function, 6 , 7 to keep the analysis feasible and informative.

The primary endpoint was first AD dementia diagnosis.

2.4. Adjustment set

Our primary estimand was the intention‐to‐treat effect. We adjusted for >600 potential confounders recorded in the year before the index date (Figure 1C). These included age at index date, biological sex; 18 Charlson Comorbidity Indices 15 ; >500 Clinical Classification Software Revised diagnosis code indications 16 ; prescriptions across 86 ATC second‐level drugs; and 18 cognitive‐related event indicators (codes in Table S2).

2.5. Statistical analyses

Categorical variables are presented as counts and percentages, and continuous variables as means with standard deviations (SDs). For each comparison, the treatment assignment indicator was regressed against the baseline covariates using logistic regression twice. First, to satisfy the covariate overlap assumption required for valid comparisons, 17 observations with non‐overlapping propensity scores were excluded using the dynamic optimal threshold. 18 A second logistic regression model was then fit on the remaining sample to derive inverse probability (IP) weights to adjust for confounding in treatment assignment. Covariate balance after weighting was evaluated using standardized mean difference (SMD) and the difference from the IP‐weighted area under the receiver operating characteristic curve (AUROC). Adequate balance required all weighted |SMD| < 0.20 and the difference from IP‐weighted AUROC within 0.01 of 0.50 (detailed in Text S2). 19 , 20 , 21 Only well‐balanced analyses were considered.

Using the IP weights, the time to first diagnosis of AD dementia was compared between the treatment and control groups in two ways. First, IP‐weighted cumulative incidence curves were estimated using a weighted Kaplan–Meier estimator (Figure 1D), 22 and the difference and ratio in cumulative incidence between groups at 3 years were calculated. Second, to estimate covariate‐adjusted HRs, we fit an IP‐weighted Cox regression on a univariable treatment indicator. The methodology used to obtain confidence intervals is detailed in Text S2. Multiplicity was controlled using the Benjamini–Hochberg method 23 for ATC level‐2 comparisons, with q values < 0.1 considered statistically significant.

2.6. Sample size calculation

To demonstrate one practical application of our findings, we calculated the sample size required for a clinical trial based on the observed difference in event rates from this study. 24 Methods are detailed in Text S3.

3. RESULTS

3.1. Drug identification and filtering

A total of 109 FDA‐approved drugs tested in prior AD trials were initially identified. After applying filters based on data availability and covariate balance, we conducted 49 studies in 38 drugs in the MCI cohort and 73 studies for 53 drugs in the over‐70 cohort (Figure 2 and Tables S3–S6, filtration process detailed in Text S4). The SMDs and differences from the IP‐weighted AUROC from 0.5 for each cohort are provided in Table S3 (MCI cohort) and Table S5 (over‐70 cohort). The drugs investigated in this study spanned 25 pharmacological/therapeutic subgroups (ATC second level) in the MCI cohort and 36 second‐level ATC classifications in the over‐70 cohort.

FIGURE 2.

FIGURE 2

Open in a new tab

Drug attrition flow chart. From ClinicalTrials.gov, we identified 109 drugs previously tested in AD trials. After excluding agents without codes and over‐the‐counter products not captured in claims, 85 drugs remained. Applying balance criteria further reduced the set to 49 comparisons for 38 drugs in the MCI cohort and 72 for 53 drugs in the over‐70 cohort. AD, Alzheimer's disease; MCI, mild cognitive impairment

3.2. Target trial emulation

When compared with other drugs within the same second‐level ATC category, 8 of 38 (21%) drugs in the MCI cohort were associated with lower incidence of AD dementia, and 16 of 53 (30%) drugs in the over‐70 cohort were associated with a lower rate of AD progression, based on at least one estimate–risk difference, risk ratio, or HR (Figures 3A,B, and S3; Tables S7 and S8). Among these, six drugs (prazosin, zolpidem, bupropion, mirtazapine, trazodone, and venlafaxine) in the MCI cohort were consistently associated with lower incidence of AD dementia across all three models: risk difference, risk ratio, and IP‐weighted HR (Figure 3C,D). In the over‐70 cohort, 10 drugs (liraglutide, amlodipine, valacyclovir, zolpidem, bupropion, caffeine, escitalopram, mirtazapine, trazodone, and venlafaxine) were consistently associated with lower rate of AD progression across all three models. Across both cohorts, 37 of 54 (69%) drugs showed no statistically significant association (q < 0.1) with incidence or progression of AD dementia in any of the three statistical models compared to other drugs in the same ATC second level. Even using a nominal significance threshold (p < 0.05), eight drugs (cromoglicic acid, itraconazole, cilostazol, guanfacine, insulin glulisine, leuprolide, rotigotine, and salsalate) showed no evidence of association in any model.

FIGURE 3.

FIGURE 3

Open in a new tab

Results comparing patients who took a target drug versus those who took other drugs in the same ATC subcategory. (A and B) The figure displays forest plots for studies showing at least one significant result across three statistical models in the (A) MCI cohort and (B) over‐70 cohort. Forest plots include cumulative difference and hazard ratios in a Cox proportional hazards model. Results with a 95% confidence interval upper limit <0 for differences, <1 for hazard ratios, and q value <0.1 are highlighted in red. An asterisk (*) on drug names indicates significant results across all three models. (C and D) Inverse‐probability‐weighted cumulative incidence curves of AD diagnosis in the MCI cohort. Shading around the cumulative curves represents the 95% confidence intervals. Number at risk in each group is shown at the bottom of the curves. (C) Bupropion versus ATC‐N06. (D) Venlafaxine versus ATC‐N06. AD, Alzheimer's disease; ATC, anatomical therapeutic chemical classification; CI, confidence intervals; HR, hazard ratio; MCI, mild cognitive impairment.

Drugs that showed consistent associations across all three estimates were further analyzed against active comparators, except for prazosin in the MCI cohort, for which no well‐balanced active comparator was available. Experiment and cohort‐specific baseline characteristics for active comparator designs are in Tables S9–S12; forest plots for the MCI and the over‐70 cohorts are shown in Figures 4 and 5, respectively.

FIGURE 4.

FIGURE 4

Open in a new tab

Forest plots comparing patients who took a target drug versus active comparators in MCI cohort. The figure displays results for studies showing at least one significant result across three statistical models in the MCI cohort. Results with a 95% confidence interval upper limit <0 for differences or <1 for HRs are highlighted in red. An asterisk (*) on the drug names indicates significant results across all three statistical models. HR, hazard ratio; MCI, mild cognitive impairment.

FIGURE 5.

FIGURE 5

Open in a new tab

Forest plots comparing patients who took a target drug versus active comparators in the over‐70 cohort. The figure displays results for studies showing at least one significant result across three statistical models in the over‐70 cohort. Results with a 95% confidence interval upper limit <0 for differences or <1 for HRs are highlighted in red. An asterisk (*) on the drug names indicates significant results across all three statistical models. HR, hazard ratio; MCI, mild cognitive impairment.

In the MCI cohort, bupropion was consistently associated with lower risk of progression to AD compared to selective serotonin reuptake inhibitors (SSRIs), including escitalopram (difference: −0.06 [−0.08, −0.05], p = 0.002; HR: 0.57 [0.49, 0.66], p < 0.001), fluoxetine (difference: −0.03 [−0.05, −0.01], p = 0.002; HR: 0.75 [0.61, 0.91], p = 0.004), and sertraline (difference: −0.06 [−0.08, −0.04], p = 0.002; HR: 0.62 [0.53, 0.71], p < 0.001) (Figures 4, 6, S4, and S5; Table S13). While the observed difference was smaller than that seen with SSRIs, the findings continue to indicate a potential beneficial association of bupropion compared to trazodone and venlafaxine (vs trazodone: difference −0.02 [−0.04, −0.004], p = 0.002; HR 0.74 [0.63, 0.87], p < 0.001; vs venlafaxine: difference −0.02 [−0.04, 0.00], p = 0.01; HR: 0.80 [0.67, 0.97], p = 0.02). Trazodone and venlafaxine were consistently associated with lower risk of progression to AD dementia compared to escitalopram or sertraline but not when compared to fluoxetine (Figures 4, 6B,C, S4, S6, and S7).

FIGURE 6.

FIGURE 6

Open in a new tab

Cumulative incidence curves and mechanism of action of atypical antidepressants. (A–D) Inverse‐probability‐weighted cumulative incidence curves of AD diagnosis in the MCI cohort. Shading around the cumulative curves represents the 95% confidence intervals. Number at risk in each group is shown at the bottom of the curves. (A) Bupropion versus escitalopram in MCI cohort. (B) Trazodone versus sertraline in MCI cohort. (C) Venlafaxine versus escitalopram in MCI cohort. (D) Liraglutide versus metformin in over‐70 cohort. (E) MoA of atypical antidepressants. The left panel illustrates the MoA of each drug at the synapse. The table on the right specifies which MoAs are associated with each drug. AD, Alzheimer's disease; MCI, mild cognitive impairment; MoA, mechanism of action.

In the over‐70 cohort, bupropion and trazodone were associated with lower incidence of AD dementia compared to escitalopram or sertraline (Figures 5 and S8; Table S14). Venlafaxine was also associated with a lower risk compared to sertraline. Although trazodone showed a slightly more favorable association than venlafaxine in two of the three models (difference: −0.01 [−0.02, 0.000], p = 0.03; HR: 0.97 [0.90, 1.04], p = 0.39), no significant risk differences were observed between bupropion and either trazodone or venlafaxine (Figures S9–S11).

In the MCI cohort, zolpidem use was associated with lower risk of progression of AD compared to lorazepam (difference: −0.04 [−0.06, −0.01], p = 0.002; HR: 0.69 [0.56, 0.85], p < 0.001) (Figure S12).

In the over‐70 cohort, zolpidem use was associated with lower incidence of AD dementia compared to lorazepam (difference: −0.03 [−0.04, −0.03], p = 0.002; HR: 0.65 [0.60, 0.70], p < 0.001) and ramelteon (difference: −0.04 [−0.07, −0.01], p = 0.002; HR: 0.57 [0.43, 0.76], p < 0.001) (Figure S13). However, this association was not observed when compared to suvorexant.

In the over‐70 cohort, liraglutide use was associated with lower incidence of AD dementia compared with metformin (difference: −0.01 [−0.02, −0.003], p = 0.002; HR: 0.75 [0.59, 0.93], p = 0.010) (Figure 6D). The results consistently suggested a beneficial association of liraglutide with reduced incidence of AD, even when compared with conventional antidiabetic agents such as glipizide and sitagliptin (Figure S14).

3.3. Sample size calculation

As an example, we estimated the sample size that a RCT would require to detect differences similar to the associations observed in this study comparing bupropion and escitalopram in the MCI cohort (detailed in Text S5). Based on the calculation, the required sample size was approximately 385 participants per group.

4. DISCUSSION

A key strength of our study is the use of advanced statistical methods designed to reduce confounding in an observational setting, including IP weighting adjusting for >600 potential confounders. Prespecified diagnostics (trimming for common support, weighted |SMD|< 0.2, and IP‐weighted AUROC near 0.50) confirmed acceptable balance and weight stability despite the high‐dimensional adjustment. We emulated 49 trials in a cohort of individuals diagnosed with MCI and 72 trials in a general insured population aged 70 years or older, covering 54 FDA‐approved drugs previously tested for AD dementia across 37 ATC second‐level categories. Bupropion was associated with a lower rate of both AD progression and incidence, followed by trazodone and venlafaxine. Liraglutide was associated with a reduced AD dementia incidence.

To our knowledge, no prior study systematically conducted trial emulations for all FDA‐approved drugs previously included in trials for AD. By revisiting these repurposable agents within a rigorous observational framework, our findings provide real‐world evidence that not only supports or complements the rationale for conducting such trials but also informs the redesign of future clinical trials, including setting realistic target differences and sample size estimations. These findings should, however, be interpreted as hypothesis‐generating and do not replace the need for fully powered and well‐designed RCTs.

A minority of repurposable drugs that are in current or completed clinical trials for AD and meet criteria for inclusion in this study showed no effect in our trial emulation studies. Most of these drugs have failed to demonstrate a drug‐placebo difference in clinical trials. Trial emulation may represent one opportunity to determine which repurposed drugs are more likely to succeed in clinical trials.

Many AD trials target neuropsychiatric symptoms rather than cognitive change. However, even when targeting such symptoms, the use of agents might still affect cognitive decline. Therefore, assessing the potential impact of these agents on cognitive function through observational studies is a valuable approach.

4.1. Atypical antidepressant

Prior clinical trial results for AD‐associated drugs identified in this study are summarized in Table S15. Bupropion, trazodone, and venlafaxine, generally classified as atypical antidepressants due to mechanisms of action beyond serotonin reuptake inhibition, showed favorable associations in this study (Figure 6E). Bupropion was associated with a lower rate of AD progression compared to three commonly prescribed SSRIs – escitalopram, sertraline, and fluoxetine – with trazodone and venlafaxine also showing benefits against some SSRIs. In a head‐to‐head comparison, bupropion appeared slightly more effective, while trazodone and venlafaxine appeared to have comparable effects. As new onset depression in elderly patients can precede memory impairment in AD, these insights might help guide selection of antidepressant treatment in elderly patients. 25

In individuals aged 70 and older, all three were also associated with a lower incidence of AD dementia compared to some SSRIs. Although trazodone appeared slightly more beneficial than venlafaxine in two of the three models, no significant differences were observed between bupropion and the other two drugs. Taken together, these findings suggest that the three may have comparable associations with AD incidence in older adults.

Bupropion, a norepinephrine‐dopamine reuptake inhibitor and a nicotinic acetylcholine receptor antagonist, is a component of AXS‐05, a combination drug with dextromethorphan, which has demonstrated preliminary efficacy in treating agitation in patients with AD. 26 , 27 , 28 However, bupropion has not shown clear benefits for cognitive impairment in prior studies. 26 , 27 , 28 , 29 Several trials have not demonstrated significant benefits of these agents on cognitive function, possibly due to inadequate sample sizes not specifically calculated to detect cognitive outcomes. Based on the difference observed in the trial comparing bupropion and escitalopram in the MCI cohort, a conventional randomized trial with approximately 770 participants (385 per arm) would be required to detect a statistically significant association. 24 Notably, this comparison was conducted against escitalopram, a proper placebo‐controlled trial would also require consideration of the expected differences between the test agent and placebo. Nevertheless, given that previous trials evaluating bupropion or related compounds enrolled fewer participants and had shorter follow‐up periods, it is likely that these studies were underpowered to detect effects on cognitive function. 26 , 27 , 28 , 29

Trazodone acts by inhibiting serotonin reuptake, blocking 5‐HT2A/2C receptors, and antagonizing histamine and alpha‐1‐adrenergic receptors, contributing to its sedative effects. 30 Two small trials investigated its use in AD dementia. In the first, trazodone showed no significant benefit over placebo or behavior management therapy (BMT) for agitation, and post hoc analysis suggested greater cognitive decline with trazodone. 31 This unexpected finding reflects the absence of diagnostic biomarkers and different rates of biological AD dementia in the different arms. In the second trial, trazodone increased sleep duration compared to placebo but demonstrated no significant effect on cognitive function. 32

Venlafaxine, a serotonin‐norepinephrine reuptake inhibitor, has shown limited evidence for cognitive effects in the literature, likely due to its infrequent use as a first‐line antidepressant and the lack of large‐scale trials. Nonetheless, our observational findings suggest it may merit further investigation regarding its potential associations with AD progression or dementia incidence.

4.2. Selective serotonin reuptake inhibitors

In older patients with major depressive disorder, SSRIs are the most commonly recommended pharmacological treatments. 33 In this study, three commonly used SSRIs – escitalopram, fluoxetine, and sertraline – were compared with atypical antidepressants as active comparators. Overall, the findings suggest that fluoxetine may show more favorable associations with lower incidence or slower progression of AD compared with escitalopram or sertraline. Given its relatively activating profile – attributed to increased dopamine and norepinephrine in the prefrontal cortex via 5‐HT2C receptor antagonism and low affinity for histamine H1, muscarinic, and alpha1‐adrenergic receptors – fluoxetine may be a reasonable choice for older adults with depression who require SSRI treatment.

4.3. GLP‐1 agonists

Glucagon‐like peptide 1 (GLP‐1) analogs are being investigated as disease‐targeted therapies with the potential to slow the progression of cognitive decline. 34 , 35 , 36 , 37

In our study, liraglutide, approved for treatment of type 2 diabetes mellitus and obesity, was associated with lower incidence of AD dementia when compared to conventional antidiabetic drugs such as metformin. The risk difference between liraglutide and empagliflozin was close to zero, raising the possibility that their associations with AD dementia incidence may be comparable, although further studies are needed to confirm this. In the MCI cohort, liraglutide was associated with a reduced risk of AD dementia progression compared with other anti‐diabetic drugs in the Cox proportional hazards model, but the association did not remain significant after false discovery rate correction. Liraglutide can cross the blood–brain barrier 38 and demonstrates favorable effects in mouse models of AD 39 , 40 or in humans. 41 , 42 , 43

In the present study, semaglutide, another GLP‐1 analog, did not show a significant association with reduced AD incidence in the over‐70 cohort compared with other anti‐diabetic drugs and could not be evaluated in the MCI cohort due to insufficient covariate balance. The discrepancy with a prior analysis in a larger electronic health record (EHR) network that reported an association among patients with type 2 diabetes 44 may reflect a smaller available sample as well as differences in population and comparator definitions. Two phase III clinical trials are ongoing, 45 and re‐evaluation with more recent data is warranted.

4.4. Non‐benzodiazepine sleep aids

In this study, zolpidem was consistently associated with lower incidence and progression of AD compared to lorazepam. Comparisons with more modern hypnotics such as ramelteon and suvorexant could not be conducted in the MCI cohort due to insufficient covariate balance but were feasible in the over‐70 cohort. In that cohort, zolpidem was also associated with lower AD incidence than ramelteon, but not suvorexant. Therapies such as zolpidem and suvorexant may improve sleep, and it is increasingly recognized that sleep accelerates the removal of amyloid beta protein and tau protein from the brain. 46 While pharmacologic treatments for insomnia are generally given reduced recommendations in clinical guidelines due to concerns about dependence and cognitive side effects, 47 , 48 , 49 , 50 our findings suggest possible differences in associations with cognitive outcomes across hypnotics, and further investigation is warranted.

4.5. Limitations

First, although we performed robust adjustments, there might have been residual biases. Second, since the database captures prescribed drugs, results of the drugs that are available over the counter are not reflected in the data. Third, we were unable to estimate the effects of taking the drug versus not taking the drug, nor evaluate the severity of neuropsychiatric symptoms due to database constraints. The methodology employed here cannot distinguish between slowing of disease progression and delay of onset of symptoms. It is possible that the findings related to antidepressants represent a symptomatic effect and delay of onset of symptoms rather than an effect on the underlying biology of the disease. The diagnosis of AD dementia was used as the outcome, preventing assessment of clinical progression after AD diagnosis. Therefore, the findings of this study should be interpreted as reflecting the potential associations of the drugs during the relatively early stages of the AD time course. The diagnosis of AD in this study was based on clinical assessment and was not confirmed by amyloid‐related biomarkers. Furthermore, our analyses targeted an intention‐to‐treat framework defining exposure by the first prescription and not modeling post‐baseline adherence or duration. This estimand may differ from a per‐protocol effect, but a unified adherence rule was not feasible across heterogeneous drug regimens; the intention‐to‐treat approach enabled a consistent, scalable analysis across all comparisons. Moreover, some drugs were not evaluable across both cohorts or specific comparator groups due to inadequate covariate balance. In addition, we did not directly compare candidate drugs with FDA‐approved AD agents, as these are preferentially used in more symptomatic patients and detailed cognitive measures are unavailable in claims data, making such comparisons difficult to interpret. Finally, while we maximized internal validation within the database using both same‐class and active comparator designs and multiple effect measures, replication in independent EHR database will be an important future direction.

5. CONCLUSION

We systematically emulated trials for 54 drugs previously evaluated in AD‐related trials using a real‐world large‐scale database. Bupropion, trazodone, venlafaxine, and zolpidem were consistently associated with favorable outcomes in both AD progression and incidence, while liraglutide showed a potential association with lower incidence. These findings demonstrate the utility of large‐scale claims data and trial emulation methods in identifying repurposing opportunities for AD.

CONFLICT OF INTEREST STATEMENT

Dr. Cummings has provided consultation to Acadia, Acumen, ALZpath, Annovis, Aprinoia, Artery, Biogen, Biohaven, BioXcel, Bristol‐Myers Squib, Eisai, Fosun, GAP Foundation, Green Valley, Janssen, Karuna, Kinoxis, Lighthouse, Lilly, Lundbeck, LSP/eqt, Merck, MoCA Cognition, New Amsterdam, Novo Nordisk, Optoceutics, Otsuka, Oxford Brain Diagnostics, Praxis, Prothena, ReMYND, Roche, Scottish Brain Sciences, Signant Health, Simcere, Sinaptica, TrueBinding, and Vaxxinity pharmaceutical, assessment, and investment companies. Dr. Karavani (E.K.), Dr. Danziger (M.D.), and Dr. Rosen‐Zvi (M.R.Z.) are employees of IBM Research. The MarketScan data were supplied by Merative L.P. as part of one or more Merative MarketScan Research Databases. Any analysis, interpretation, or conclusion based on these data is solely that of the authors and not Merative L.P. and its subsidiaries. The other authors have declared no competing interests. Author disclosures are available in the Supporting Information.

CONSENT STATEMENT

Because the data (MarketScan) are fully de‐identified and compliant with the Health Insurance Portability and Accountability Act, informed consent was not required.

Supporting information

Supporting Information

TRC2-12-e70193-s003.pdf (2.9MB, pdf)

Supporting Information

TRC2-12-e70193-s002.xlsx (2.5MB, xlsx)

Supporting Information

TRC2-12-e70193-s004.pdf (57KB, pdf)

Supporting Information

TRC2-12-e70193-s005.pdf (1.3MB, pdf)

Supporting Information

TRC2-12-e70193-s001.docx (29.4KB, docx)

ACKNOWLEDGMENTS

This work was primarily supported by the National Institute on Aging (NIA) under Award Numbers U01AG073323, R01AG066707, R01AG084250, R01AG076448, R01AG082118, R01AG092462, R01AG092591, RF1AG082211, and R33AG083003; the National Institute of Neurological Disorders and Stroke (NINDS) under Award Number RF1NS133812; the Alzheimer's Association award ALZDISCOVERY‐1051936; the Alzheimer's Disease Drug Discovery Foundation (ADDF); and Dr. Keyhan and Dr. Jafar Mobasseri Endowed Chair for Innovative Research  to F.C. This work was supported in part by the Cleveland Alzheimer's Disease Research Center (National Institutes of Health [NIH]/NIA: P30AG072959) to F.C., A.A.P., and J.C. A.A.P. was supported by the Valour Foundation, by Department of Veterans Affairs Merit Award I01BX005976, and as the Rebecca E. Barchas, MD, Professor in Translational Psychiatry of Case Western Reserve University and the Morley‐Mather Chair in Neuropsychiatry of University Hospitals of Cleveland Medical Center, as well as NIH/NIA 1R01AG071512, NIH/NIA RO1AG066707, NIH/NIA 1U01AG073323, the Louis Stokes VA Medical Center, and an anonymous donor. This work was supported in part by Keep Memory Alive (KMA), National Institute of General Medical Sciences (NIGMS) grant P20GM109025, NINDS grant U01NS093334, NIA grants R01AG053798 and R35AG071476, and the Alzheimer's Disease Drug Discovery Foundation (ADDF) to J.C.

Tonegawa‐Kuji R, Karavani E, Danziger M, et al. Critical evaluation of real‐world evidence of repurposable medicines in the Alzheimer's disease drug development pipeline using a target trial emulation. Alzheimer's Dement. 2026;12:e70193. 10.1002/trc2.70193

DATA AVAILABILITY STATEMENT

The data in this study from MarketScan Research Databases contain private health information, for which access to the study team was provided by Merative L.P., and therefore public redistribution of the data is prevented.

REFERENCES

  • 1. 2023 Alzheimer's disease facts and figures. Alzheimers Dement. 2023;19(4):1598‐1695. [DOI] [PubMed] [Google Scholar]
  • 2. Jansen WJ, Ossenkoppele R, Knol DL, et al. Prevalence of cerebral amyloid pathology in persons without dementia: a meta‐analysis. JAMA. 2015;313(19):1924‐1938. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3. Manly JJ, Jones RN, Langa KM, et al. Estimating the prevalence of dementia and mild cognitive impairment in the US: The 2016 Health and Retirement Study Harmonized Cognitive Assessment Protocol Project. JAMA Neurol. 2022;79(12):1242‐1249. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4. Cummings J, Zhou Y, Lee G, Zhong K, Fonseca J, Cheng F. Alzheimer's disease drug development pipeline: 2025. Alzheimers Dement (NY). 2025;11(2):e70098. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5. Budd Haeberlein S, Aisen PS, Barkhof F, et al. Two randomized phase 3 studies of aducanumab in early Alzheimer's disease. J Prev Alzheimers Dis. 2022;9(2):197‐210. [DOI] [PubMed] [Google Scholar]
  • 6. Sims JR, Zimmer JA, Evans CD, et al. Donanemab in early symptomatic Alzheimer disease: The TRAILBLAZER‐ALZ 2 randomized clinical trial. JAMA. 2023;330(6):512‐527. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7. van Dyck CH, Swanson CJ, Aisen P, et al. Lecanemab in early Alzheimer's disease. N Engl J Med. 2023;388(1):9‐21. [DOI] [PubMed] [Google Scholar]
  • 8. Lee D, Slomkowski M, Hefting N, et al. Brexpiprazole for the treatment of agitation in Alzheimer dementia: a randomized clinical trial. JAMA Neurol. 2023;80(12):1307‐1316. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9. Cummings JL. Maximizing the benefit and managing the risk of anti‐amyloid monoclonal antibody therapy for Alzheimer's disease: strategies and research directions. Neurotherapeutics. 2025;22:e00570. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10. Zang C, Zhang H, Xu J, et al. High‐throughput target trial emulation for Alzheimer's disease drug repurposing with real‐world data. Nat Commun. 2023; 14(1):8180. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11. Cook JA, Julious SA, Sones W, et al. DELTA(2) guidance on choosing the target difference and undertaking and reporting the sample size calculation for a randomised controlled trial. BMJ. 2018;363:k3750. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12. ClinicalTrials.gov: National Library of Medicine, Available from: https://clinicaltrials.gov/
  • 13. RxNav: National Library of Medicine, Available from: https://mor.nlm.nih.gov/RxNav/ [DOI] [PubMed]
  • 14. Hernán MA, Wang W, Leaf DE. Target trial emulation: a framework for causal inference from observational data. JAMA. 2022;328(24):2446‐2447. [DOI] [PubMed] [Google Scholar]
  • 15. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373‐83. [DOI] [PubMed] [Google Scholar]
  • 16. Clinical classifications software refined (CCSR): Agency for Healthcare Research and Quality, Available from: https://hcup‐us.ahrq.gov/toolssoftware/ccsr/ccs_refined.jsp
  • 17. Westreich D, Cole SR. Invited commentary: positivity in practice. Am J Epidemiol. 2010;171(6):674‐677. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18. Crump RK, Hotz VJ, Imbens GW, Mitnik OA. Dealing with limited overlap in estimation of average treatment effects. Biometrika. 2009;96(1):187‐199. [Google Scholar]
  • 19. Harder VS, Stuart EA, Anthony JC. Propensity score techniques and the assessment of measured covariate balance to test causal associations in psychological research. Psychol Methods. 2010;15(3):234‐249. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20. Normand ST, Landrum MB, Guadagnoli E, et al. Validating recommendations for coronary angiography following acute myocardial infarction in the elderly: a matched analysis using propensity scores. J Clin Epidemiol. 2001;54(4):387‐398. [DOI] [PubMed] [Google Scholar]
  • 21. Yishai Shimoni EK, Ravid S, Bak P, et al. An evaluation toolkit to guide model selection and cohort definition in causal inference 2019. Available from: https://arxiv.org/abs/1906.00442
  • 22. Cole SR, Hernán MA. Adjusted survival curves with inverse probability weights. Comput Methods Programs Biomed. 2004;75(1):45‐49. [DOI] [PubMed] [Google Scholar]
  • 23. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Series B Methodol. 2018;57(1):289‐300. [Google Scholar]
  • 24. Kieser M. Methods and Applications of Sample Size Calculation and Recalculation in Clinical Trials. Springer Cham; 2020. [Google Scholar]
  • 25. Lyketsos CG, Lopez O, Jones B, Fitzpatrick AL, Breitner J, DeKosky S. Prevalence of neuropsychiatric symptoms in dementia and mild cognitive impairment: results from the cardiovascular health study. JAMA. 2002;288(12):1475‐1483. [DOI] [PubMed] [Google Scholar]
  • 26. Axsome Therapeutics Announces AXS‐05 Achieves Primary Endpoint in the ADVANCE‐1 Pivotal Phase 2/3 Trial in Alzheimer's Disease Agitation 2020. Accessed September 24th, 2025. Available from: https://axsometherapeuticsinc.gcs‐web.com/news‐releases/news‐release‐details/axsome‐therapeutics‐announces‐axs‐05‐achieves‐primary‐endpoint‐2
  • 27. Axsome Therapeutics Announces AXS‐05 Achieves Primary Endpoint in the ACCORD Phase 3 Trial in Alzheimer's Disease Agitation 2022. Accessed September 24th, 2025. Available from: https://axsometherapeuticsinc.gcs‐web.com/news‐releases/news‐release‐details/axsome‐therapeutics‐announces‐axs‐05‐achieves‐primary‐endpoint‐3
  • 28. Axsome Therapeutics Announces Successful Completion and Results of Phase 3 Clinical Program of AXS‐05 in Alzheimer's Disease Agitation. Accessed September 24th, 2025. Available from: https://www.biospace.com/press‐releases/axsome‐therapeutics‐announces‐successful‐completion‐and‐results‐of‐phase‐3‐clinical‐program‐of‐axs‐05‐in‐alzheimers‐disease‐agitation
  • 29. Maier F, Spottke A, Bach JP, et al. Bupropion for the treatment of apathy in Alzheimer disease: a randomized clinical trial. JAMA Netw Open. 2020;3(5):e206027. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30. Stahl SM. Mechanism of action of trazodone: a multifunctional drug. CNS Spectr. 2009;14(10):536‐546. [DOI] [PubMed] [Google Scholar]
  • 31. Teri L, Logsdon RG, Peskind E, et al. Treatment of agitation in AD: a randomized, placebo‐controlled clinical trial. Neurology. 2000;55(9):1271‐1278. [DOI] [PubMed] [Google Scholar]
  • 32. Camargos EF, Louzada LL, Quintas JL, Naves JO, Louzada FM, Nóbrega OT. Trazodone improves sleep parameters in Alzheimer disease patients: a randomized, double‐blind, and placebo‐controlled study. Am J Geriatr Psychiatry. 2014;22(12):1565‐1574. [DOI] [PubMed] [Google Scholar]
  • 33. McQuaid JRLE, Barber JP, Breland‐Noble AM, et al. APA clinical practice guideline for the treatment of depression across three age cohorts: American psychological association; 2019. Accessed September 25th, 2025. Available from: https://www.apa.org/depression‐guideline
  • 34. Nørgaard CH, Friedrich S, Hansen CT, et al. Treatment with glucagon‐like peptide‐1 receptor agonists and incidence of dementia: data from pooled double‐blind randomized controlled trials and nationwide disease and prescription registers. Alzheimers Dement (NY). 2022;8(1):e12268. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35. Akimoto H, Negishi A, Oshima S, et al. Antidiabetic drugs for the risk of Alzheimer disease in patients with Type 2 DM using FAERS. Am J Alzheimers Dis Other Demen. 2020;35:1533317519899546. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36. Wium‐Andersen IK, Osler M, Jørgensen MB, Rungby J, Wium‐Andersen MK. Antidiabetic medication and risk of dementia in patients with type 2 diabetes: a nested case‐control study. Eur J Endocrinol. 2019;181(5):499‐507. [DOI] [PubMed] [Google Scholar]
  • 37. Seminer A, Mulihano A, O'Brien C, et al. Cardioprotective glucose‐lowering agents and dementia risk: a systematic review and meta‐analysis. JAMA Neurol. 2025;82(5):450‐460. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38. Hunter K, Hölscher C. Drugs developed to treat diabetes, liraglutide and lixisenatide, cross the blood brain barrier and enhance neurogenesis. BMC Neurosci. 2012;13:33. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39. Vargas‐Soria M, Carranza‐Naval MJ, Del Marco A, Garcia‐Alloza M. Role of liraglutide in Alzheimer's disease pathology. Alzheimers Res Ther. 2021;13(1):112. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40. Salles GN, Calió ML, Hölscher C, Pacheco‐Soares C, Porcionatto M, Lobo AO. Neuroprotective and restorative properties of the GLP‐1/GIP dual agonist DA‐JC1 compared with a GLP‐1 single agonist in Alzheimer's disease. Neuropharmacology. 2020;162:107813. [DOI] [PubMed] [Google Scholar]
  • 41. Gejl M, Gjedde A, Egefjord L, et al. In Alzheimer's disease, 6‐month treatment with GLP‐1 analog prevents decline of brain glucose metabolism: randomized, placebo‐controlled, double‐blind clinical trial. Front Aging Neurosci. 2016;8:108. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42. Edison P, Femminella GD, Ritchie CW, et al. Evaluation of liraglutide in the treatment of Alzheimer's disease. Alzheimers Dement. 2021;17(S9):e057848. [Google Scholar]
  • 43. GLP‐1 drug liraglutide may protect against dementia. News provided by Alzheimer's Association. 2024. Accessed September 25th, 2025. Available from: https://prnmedia.prnewswire.com/news‐releases/glp‐1‐drug‐liraglutide‐may‐protect‐against‐dementia‐302209255.html
  • 44. Wang W, Wang Q, Qi X, et al. Associations of semaglutide with first‐time diagnosis of Alzheimer's disease in patients with type 2 diabetes: target trial emulation using nationwide real‐world data in the US. Alzheimers Dement. 2024; 20: 8661‐8672. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45. Cummings JL, Atri A, Feldman HH, et al. Evoke and evoke+: design of two large‐scale, double‐blind, placebo‐controlled, phase 3 studies evaluating efficacy, safety, and tolerability of semaglutide in early‐stage symptomatic Alzheimer's disease. Alzheimers Res Ther. 2025;17(1):14. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46. Wang C, Holtzman DM. Bidirectional relationship between sleep and Alzheimer's disease: role of amyloid, tau, and other factors. Neuropsychopharmacology. 2020;45(1):104‐120. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47. Edinger JD, Arnedt JT, Bertisch SM, et al. Behavioral and psychological treatments for chronic insomnia disorder in adults: an American Academy of Sleep Medicine Clinical Practice Guideline. J Clin Sleep Med. 2021;17(2):255‐262. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48. Sateia MJ, Buysse DJ, Krystal AD, Neubauer DN, Heald JL. Clinical practice guideline for the pharmacologic treatment of chronic insomnia in adults: an American Academy of Sleep Medicine Clinical Practice Guideline. J Clin Sleep Med. 2017;13(2):307‐349. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49. Glass J, Lanctôt KL, Herrmann N, Sproule BA, Busto UE. Sedative hypnotics in older people with insomnia: meta‐analysis of risks and benefits. BMJ. 2005;331(7526):1169. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50. Hauglund NL, Andersen M, Tokarska K, et al. Norepinephrine‐mediated slow vasomotion drives glymphatic clearance during sleep. Cell. 2025;188(3):606‐622. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supporting Information

TRC2-12-e70193-s003.pdf (2.9MB, pdf)

Supporting Information

TRC2-12-e70193-s002.xlsx (2.5MB, xlsx)

Supporting Information

TRC2-12-e70193-s004.pdf (57KB, pdf)

Supporting Information

TRC2-12-e70193-s005.pdf (1.3MB, pdf)

Supporting Information

TRC2-12-e70193-s001.docx (29.4KB, docx)

Data Availability Statement

The data in this study from MarketScan Research Databases contain private health information, for which access to the study team was provided by Merative L.P., and therefore public redistribution of the data is prevented.

Articles from Alzheimer's & Dementia : Translational Research & Clinical Interventions are provided here courtesy of Wiley

RESOURCES