The impact of cholinesterase inhibitors on cognitive trajectories in mild cognitive impairment patients based on amyloid beta status

Abstract

INTRODUCTION

This study examines whether cholinesterase inhibitors (ChEIs) influence the progression to Alzheimer's disease (AD) dementia and cognitive trajectories in amnestic mild cognitive impairment (MCI) patients, considering their amyloid beta (Aβ) status.

METHODS

Kaplan–Meier and time‐varying Cox models evaluated ChEI use and different Aβ status on MCI‐to‐AD progression. Linear mixed‐effects models assessed cognitive trajectories. Locally estimated scatterplot smoothing regression analyzed cognitive changes before and after ChEI initiation.

RESULTS

Among 558 amnestic MCI participants (168 ChEI users), ChEI users exhibited higher risk of progression to AD dementia (hazard ratio = 1.77, 95% confidence interval: 1.15 to 2.73, p = 0.001). Both ChEI use and Aβ burden independently accelerated MCI progression and cognitive decline. Cognitive trajectories demonstrated decline before ChEI initiation and continued to decline after treatment began.

DISCUSSION

The association between ChEI treatment and accelerated progression to AD dementia and cognitive decline, independent of Aβ status, emphasized the need to reconsider optimal timing for ChEI initiation in MCI.

Highlights

• ChEI use in MCI was associated with increased risk of progression to AD dementia.

• ChEI use in MCI was associated with accelerated longitudinal cognitive decline.

• Cognitive decline persisted after ChEI initiation rather than reversing.

• ChEI effects on MCI progression to AD dementia were independent of Aβ status.

1. INTRODUCTION

Alzheimer's disease (AD) is the most common cause of dementia in older adults. ¹ , ² Studies have demonstrated that early cholinergic deficiency, ³ , ⁴ , ⁵ , ⁶ , ⁷ originating from the basal forebrain to the neocortex, significantly contributes to AD‐related cognitive deficits. ⁸ , ⁹ , ¹⁰ This understanding provides the rationale for using cholinesterase inhibitors (ChEIs) in clinical practice. ³ Clinical trials have confirmed that ChEIs are well tolerated and effective in treating mild to moderate AD. ¹¹ , ¹² AD represents a pathophysiological continuum across different stages, with amnestic mild cognitive impairment (MCI) serving as a transitional phase between normal aging and early AD. ¹³ While AD is characterized by reduced cholinergic activity, MCI shows an upregulation of choline acetyltransferase activity, which may contribute to different treatment responses to ChEIs. ¹⁴ Fewer than half of amnestic MCI patients report taking ChEIs, ¹⁵ , ¹⁶ primarily motivated by concerns about progression to dementia, despite the lack of strong evidence supporting their efficacy in this context. ¹⁷ , ¹⁸

Several studies have attempted to assess the effects of ChEI treatment on cognition changes in MCI patients, but the results have been inconsistent. ¹⁹ , ²⁰ , ²¹ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ Previous randomized, placebo‐controlled studies have reported favorable effects, ²⁴ , ²⁵ no effects, ²⁶ , ²⁷ or even disadvantageous effects. ¹⁵ , ²⁸ Currently, there are 47 registered clinical trials, yet some findings do not demonstrate efficacy for ChEIs in MCI. ¹⁶ , ²⁵ , ²⁶ , ²⁹ , ³⁰ A placebo‐controlled trial found that donepezil treatment for the amnestic MCI population resulted in slight improvements in memory performance during the first 12 months of the study; however, these effects were not maintained over the 3‐year follow‐up. ²⁶ In contrast, two observational studies indicated that the use of ChEIs and memantine was associated with a greater decline in cognitive function in amnestic MCI. ¹⁵ , ¹⁶ A possible explanation for this variability is that many of these studies included participants based solely on a clinical diagnosis of amnestic MCI, without confirming underlying AD pathology, resulting in a heterogeneous participant pool and potentially leading to inconsistent findings. Amyloid beta (Aβ) deposition on amnestic MCI progression may differ; studies have shown that Aβ‐positive (Aβ+) subjects with amnestic MCI are significantly more likely to convert to AD than Aβ‐negative (Aβ−) patients. ³¹ Previous studies have also shown that only 40% to 50% of amnestic MCI patients are Aβ−, ³² , ³³ and these patients with different levels of Aβ deposition likely possess distinct pathophysiological characteristics that may influence their response to ChEI treatment. The relationship between cholinergic system and amyloid pathology is bidirectional. Animal studies demonstrate that cholinergic deficits promote Aβ generation and cognitive impairment, while Aβ accumulation impairs cholinergic synaptic function and modulates ChEI efficacy through multiple mechanisms. ³⁴ , ³⁵ , ³⁶ Evidence from clinical and pathological studies further illustrates this complexity. Among the limited studies examining this interaction, a cohort study of amnestic MCI patients found no significant interaction between ChEI use and amyloid positivity on cognitive outcomes. ³² , ³³ , ³⁷ One post mortem study showed that AD patients treated with ChEIs exhibited higher cortical phosphorylated‐tau levels but slightly reduced Aβ accumulation compared to untreated patients. ³⁸ However, existing studies are limited by small sample sizes and relatively short follow‐up periods, potentially hampering their ability to detect meaningful treatment effects. Several key questions remain unanswered, including whether ChEIs affect the progression from amnestic MCI to AD dementia and the rate of cognitive decline. Furthermore, if such effects exist, it remains unclear whether there are significant interaction effects between ChEIs and Aβ burden on cognitive outcomes.

To address these questions, this study evaluated the long‐term effects of ChEI treatment on progression to AD dementia and cognitive trajectories in amnestic MCI patients in a real‐world longitudinal observational cohort. It examined changes in cognitive trajectories following ChEI treatment initiation across different Aβ status. To further explore potential effect modifiers of ChEI treatment, the study investigated interactions between ChEI use and both Aβ burden and apolipoprotein E (APOE) ε4 status, with the aim of informing more personalized treatment approaches for amnestic MCI patients.

2. METHODS

2.1. Study design and inclusion criteria

The participants were recruited from the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort (ADNI 1, 2, 3, 4, GO) (http://www.adni‐info.org), a longitudinal, natural‐history, observational study conducted at more than 60 clinical sites across the United States and Canada. After obtaining Institutional Review Board approval from each participating site and informed consent from all participants, individuals with amnestic MCI were enrolled. ³⁹ Participants were required to be between 55 and 90 years of age, have a study partner capable of providing independent functional evaluation, be fluent in either English or Spanish. The ADNI data used in this study were downloaded on February 8, 2025. In the ADNI database, participants were classified as normal cognition (NC), amnestic MCI, or AD based on standard diagnostic criteria from ADNI. NC participants were required to have Mini‐Mental State Examination (MMSE) scores of 24 to 30, a Clinical Dementia Rating (CDR) score of 0, normal memory function as measured by the Wechsler Memory Scale Logical Memory II (WMS‐II), no memory complaints, and no significant cognitive or functional impairments. Amnestic MCI participants were characterized by MMSE scores of 24 to 30, reported memory complaints, objective memory loss as measured by education‐adjusted WMS‐II scores, a CDR score of 0.5, absence of significant impairment in other cognitive domains, and preserved activities of daily living. AD dementia diagnosis during follow‐up was made according to the National Institute of Neurological and Communicative Disorders and Stroke (NINCDS)‐Alzheimer’s Disease and Related Disorders Association (ADRDA) criteria (NINCDS/ADRDA) criteria for probable AD, ⁴⁰ with participants showing a CDR score of 1 or higher.

A total of 3167 participants with medication records were initially analyzed from the ADNI database. Among these, 1063 were excluded due to insufficient or inconsistent sociodemographic or medical data. Thirty‐three participants were excluded for memantine monotherapy use (over 1 year), and 457 were excluded for either using ChEIs combined with memantine at the baseline or for ChEI use less than 1 year. Among the remaining participants, 993 were excluded due to a baseline diagnosis of NC or AD dementia. There were 63 participants who were not using ChEIs at baseline but initiated treatment during follow‐up with at least 1 year of continuous use and two subsequent visits. These participants were analyzed separately as the ChEI initiation group, allowing us to examine cognitive progression before and after ChEI initiation. A total of 558 participants with amnestic MCI were included in the final analysis (Figure 1). The study analyzed longitudinal data from cognitive assessments and florbetapir positron emission tomography standardized uptake value ratio (AV45‐PET SUVR) imaging, with a maximum follow‐up duration of 120 months. This study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.

FIGURE 1.

Flow chart. The flow diagram shows the selection process starting with 3167 individuals with medication records, through various exclusion criteria, ultimately resulting in a study population of 558 MCI patients (390 non‐users and 168 ChEI users). AD, Alzheimer's disease; ChEI, cholinesterase inhibitor; MCI, mild cognitive impairment; NC, normal cognition.

2.2. Drug exposure

Drug information was obtained from pharmacy data in the ADNI database, including drug name and dates of prescription (http://adni.loni.usc.edu/). ChEI users were defined as participants who had continuously used donepezil, rivastigmine, or galantamine for more than 12 months with at least two follow‐up assessments. ²⁶ , ³⁰ ChEI non‐users were defined as who had never used ChEIs but had completed a minimum of two follow‐up assessments. The final sample comprised 168 ChEI users and 390 non‐users. Based on the median ChEI treatment duration (3.11 years), ChEI users were further stratified into two groups: short‐term users (< 3.11 years) and long‐term users (≥ 3.11 years). This stratification enabled the assessment of duration‐dependent effects of ChEI use on AD dementia progression and cognitive outcomes. Detailed medication information is presented in Table S1 in the supporting information.

RESEARCH‐IN‐CONTEXT

1. Systematic review: A literature review through 2025 in PubMed and Google Scholar revealed inconsistent findings regarding ChEI effects in amnestic MCI patients, with studies reporting varying outcomes. Most studies were limited by short follow‐up periods, small sample sizes, and lack of amyloid status consideration. Consequently, the long‐term impact of ChEI therapy on the risk of AD dementia across different Aβ status in amnestic MCI patients requires further investigation.

2. Interpretation: Using the ADNI database with long‐term follow‐up, we found ChEI therapy in amnestic MCI patients was associated with faster progression to AD dementia and more rapid cognitive decline, independent of amyloid status. This suggests that initiating ChEIs during amnestic MCI may not provide therapeutic benefits.

3. Future directions: Long‐term randomized controlled trials in larger, more diverse populations are needed to evaluate ChEI efficacy in amnestic MCI.

2.3. Cognitive assessment

The CDR‐SB, MMSE,  and Alzheimer's Disease Assessment Scale‐Cognitive Subscale (ADAS cog11 and ADAS cog13) were used to assess the global cognition of participants. Higher scores on the CDR‐SB and ADAS cog11 and cog13, along with lower scores on the MMSE, indicated impaired cognition.

2.4. Method of Aβ PET analysis

The protocols for acquiring and preprocessing AV45 data are available on the ADNI database (http://adni.loni.usc.edu/). Amyloid status was determined using PET imaging with the standardized tracer[18F] florbetapir, smoothed to 6‐mm resolution. PET images were registered to MRI scans (https://www.fil.ion.ucl.ac.uk/spm/), with ROI‐based AV45 standardized uptake value ratios (SUVRs) calculated using the whole cerebellum as a reference. A detailed description of the data processing methods can be found in the PDF “UC Berkeley‐AV45 Analysis Methods” (https://ida.loni.usc.edu/pages/access/studyData.jsp). Participants were determined to be Aβ+ based on pre‐established cutoffs (global florbetapir SUVR > 1.11). ⁴¹ , ⁴²

2.5. Covariates and outcome

Baseline covariates included age, gender, race, ethnicity, education, baseline MMSE, baseline CDR‐SB, and baseline AV45 SUVR. The primary outcome was the risk of progression from MCI to AD dementia, as defined by the NINCDS/ADRDA criteria. ⁴⁰ Secondary outcomes included the likelihood of conversion from MCI to NC and cognitive changes assessed by CDR‐SB, MMSE, and ADAS cog13.

2.6. Statistical analysis

Missing baseline data were handled using multiple imputation by chained equations with predictive mean matching. To minimize potential selection bias and confounding, inverse probability of treatment weighting (IPTW) was applied. Propensity scores were estimated using logistic regression, which included age, gender, education, ethnicity, race, APOE ε4 status, baseline ADAS cog13, CDR‐SB, MMSE, baseline AV45 SUVR, and medical history (hypertension, stroke). Weights were trimmed at the 1st and 99th percentiles and normalized to maintain the original sample size. Balance between groups was assessed using standardized mean differences (SMD). ⁴³ For analyses stratified by amyloid status, subgroup‐specific IPTW calculations were performed for Aβ+ and Aβ− subgroups to ensure optimal covariate balance within each stratum. Baseline characteristics were compared using chi‐squared tests for categorical variables and Mann–Whitney U tests for continuous variables.

The risk of amnestic MCI progression to AD dementia and the likelihood of reversion to NC were estimated using the Kaplan–Meier method. Cox proportional hazards models were used to estimate hazard ratios (HRs) for the effects of ChEIs on both MCI progression to AD dementia and reversion to NC, with adjustment for potential confounders including age, education, gender, baseline MMSE, baseline CDR‐SB, baseline AV45 SUVR, and APOE ε4 status. The models also evaluated interaction terms between ChEI use and AV45 SUVR and APOE ε4 status. When the proportional hazards assumption was violated, as indicated by significant Schoenfeld residual tests, time‐dependent Cox regression models were subsequently implemented.

Longitudinal changes in cognitive measures (CDR‐SB, MMSE, and ADAS cog13) were analyzed using linear mixed‐effects models with random intercepts and slopes. Change scores were calculated as the difference from baseline for each cognitive measure. To assess cognitive trajectories before and after ChEI initiation in the subgroup of 63 participants who started ChEI during follow‐up, locally estimated scatterplot smoothing (LOESS) regression was applied. Individual trajectories and smoothed curves were plotted against time relative to drug initiation, with time 0 representing the start of ChEI treatment. To explore whether the annual changes in AV45 SUVR mediated the association between ChEI treatment and AD dementia conversion, mediation analysis was conducted using the R package mediation. The analysis was performed in three steps: (1) examine effect of ChEI treatment on annual changes in AV45 SUVR; (2) assess effect of annual changes in AV45 SUVR on AD dementia conversion while controlling for ChEI treatment; and (3) calculate direct effect (ADE), indirect effect (ACME), and total effect. The significance of the mediation effect was tested using bootstrapping with 1000 replications. All statistical analyses were performed using R version 4.4.2 statistical software. Statistical significance was set at a two‐sided p value < 0.05.

3. RESULTS

3.1. Baseline characteristics

A total of 558 eligible participants were included in the cohort, with a mean (standard deviation [SD]) age of 72.10 (7.64) years and 238 (42.65%) female participants. Specifically, there were 168 participants in the ChEI user group and 390 in the non‐user group. Table 1 presented the baseline characteristics of this cohort. After IPTW adjustment, all baseline covariates met the threshold for SMD of 0.2 or less, ⁴⁴ indicating that between‐group differences in covariates were effectively eliminated, and any differences in outcomes were attributed to the different therapeutic approaches (Table 1). The ChEI initiation group comprised 63 participants who were not using ChEIs at baseline but began treatment during the follow‐up period. Based on baseline Aβ status, the group was further divided into Aβ‐/ChEI initiation and Aβ+/ChEI initiation subgroups (see Table S2 in the supporting information).

TABLE 1.

Details of covariates before and after IPTW adjustment.

	Before IPTW				After IPTW
Covariate	Non‐users	ChEI users	p	SMD	Non‐users	ChEI users	p	SMD
Age, mean (SD), years	71.70 (7.78)	72.90 (7.26)	0.05	0.16	72.14 (7.67)	72.72 (7.64)	0.42	0.08
Gender, n (%)
Male	214 (54.90)	106 (63.10)	0.09	−0.17	320.25 (56.90)	302.58 (57.38)	0.93	−0.02
Female	176 (45.10)	62 (36.90)	0.09	−0.17	242.61 (43.10)	224.76 (42.62)	0.93	−0.02
Education, mean (SD), years	16.20 (2.69)	15.80 (2.80)	0.06	−0.17	16.13 (2.76)	16.14 (2.65)	0.95	0.003
APOE ɛ4 status, n (%)
non‐E4	246 (63.08)	83 (49.40)	0.004	−0.28	325.27 (57.79)	311.18 (59.01)	0.96	0.006
E4 heterozygous	119 (30.51)	64 (38.10)	0.10	0.16	182.77 (32.47)	155.25 (29.44)	0.62	−0.05
E4 homozygous	25 (6.41)	21 (12.50)	0.03	0.21	54.81 (9.74)	60.91 (11.55)	0.58	0.07
Ethnicity, n (%)
Not Hispanic/Latino	371 (95.10)	164 (97.60)	0.26	0.13	539.93 (95.93)	510.05 (96.72)	0.71	0.04
Others	19 (4.87)	4 (2.38)	0.26	0.13	22.93 (4.07)	17.30 (3.28)	0.71	0.04
Race, n (%)
White	353 (90.50)	162 (96.40)	0.03	0.24	521.05 (92.57)	503.68 (95.51)	0.27	0.12
Asian	7 (1.79)	2 (1.19)	1.00 a	−0.05	8.17 (1.45)	3.90 (0.74)	0.36	−0.06
Black	22 (5.64)	2 (1.19)	0.09 a	−0.25	23.57 (4.19)	10.52 (1.99)	0.26	−0.12
Others	8 (2.05)	2 (1.19)	1.00 a	−0.07	10.07 (1.79)	9.25 (1.75)	1.00	0.0003
Baseline ADAS cog11, mean (SD)	8.17 (3.70)	10.60 (3.98)	<0.001	0.63	8.83 (3.92)	9.17 (3.71)	0.30	0.09
Baseline ADAS cog13, mean (SD)	13.00 (5.69)	17.60 (6.10)	<0.001	0.79	14.30 (6.12)	14.98 (5.68)	0.29	0.12
Baseline CDR‐SB, mean (SD)	1.23 (0.74)	1.77 (0.93)	<0.001	0.64	1.41 (0.86)	1.46 (0.85)	0.51	0.06
Baseline MMSE, mean (SD)	28.20 (1.63)	27.40 (1.77)	<0.001	−0.48	27.99 (1.73)	28.00 (1.75)	0.90	0.01
Baseline AV45 SUVR, mean (SD)	1.15 (0.23)	1.30 (0.28)	<0.001	0.59	1.20 (0.27)	1.22 (0.27)	0.73	0.06
History of hypertension, n (%)	179 (45.90)	92 (54.8)	0.07	0.18	279.08 (49.58)	268.52 (50.92)	0.81	0.02
History of stroke, n (%)	7 (1.79)	2 (1.19)	1.00 a	−0.05	9.65 (1.71)	9.52 (1.80)	0.94	0.01

^a Fisher's exact test was used to assess racial distribution differences due to low frequencies in race and history of stroke. Values are presented as n (%).

Abbreviations: ADAS‐cog11, Alzheimer's Disease Assessment Scale‐Cognitive Subscale (11‐item); ADAS‐cog13, Alzheimer's Disease Assessment Scale‐Cognitive Subscale (13‐item); APOE, apolipoprotein E; AV45 SUVR, florbetapir positron emission tomography standardized uptake value ratio; CDR‐SB, Clinical Dementia Rating Sum of Boxes; ChEI, cholinesterase inhibitor; IPTW, inverse probability of treatment weighting; IPTW, inverse probability of treatment weighting; MMSE, Mini‐Mental State Examination; no., number; SD, standard deviation; SMD, standardized mean difference.

3.2. Effects of ChEIs on progression from MCI to AD dementia and cognitive trajectories

Of the 558 MCI participants, 138 individuals progressed to AD dementia during the follow‐up period. Kaplan–Meier survival analysis before IPTW showed that ChEI users had a significantly higher cumulative probability of progressing to AD dementia compared to non‐users (log‐rank χ² = 127.35, p < 0.001) (Figure 2A). This association remained significant after IPTW adjustment (log‐rank χ² = 14.17, p < 0.001) (Figure 2B). The residual test revealed significant time‐varying effects of APOE ɛ4 status (p = 0.003); therefore, APOE ɛ4 was incorporated as a time‐dependent variable in the Cox proportional hazards model. The IPTW‐adjusted model revealed that ChEI users had a significantly higher risk of progression to AD dementia compared to non‐users (HR = 1.77, 95% CI: 1.15 to 2.73, p = 0.001) (Table 2). Baseline Aβ burden measured by AV45‐PET was significantly associated with increased risk of MCI to AD dementia progression (HR = 12.34, 95% CI: 5.83 to 26.14, p < 0.001). Time‐dependent APOE ε4 showed a trend toward increased risk, though this did not reach statistical significance (HR = 1.27, 95% CI: 0.95 to 1.71, p = 0.11) (Table 2). Regarding conversion to NC, the IPTW‐adjusted Kaplan–Meier analysis showed that ChEI users had a significantly lower cumulative probability of reverting to NC compared to non‐users in the MCI population (log‐rank χ² = 10.45, p < 0.001) (Figure S1 in the supporting information). The IPTW‐adjusted Cox proportional hazards model confirmed this finding, with ChEI users showing a significantly lower likelihood of conversion to NC compared to non‐users in the MCI population (HR = 0.17, 95% CI: 0.06 to 0.52, p = 0.002, Table S3 in the supporting information).

FIGURE 2.

Effects of ChEIs on progression from MCI to AD dementia and cognitive trajectories with IPTW adjustment. Kaplan–Meier survival curves showing progression from MCI to AD dementia in unadjusted analysis (A) and after IPTW adjustment (B), with numbers at risk shown below. Longitudinal trajectories of cognitive measures including CDR‐SB score changes (C), MMSE score changes (D), and ADAS cog13 score changes (E) over time, adjusted using IPTW‐weighted LME models. Blue lines represent non‐users, and orange lines represent ChEI users. Shaded areas indicate 95% confidence intervals. All analyses were adjusted for age, gender, education, and APOE ε4 status using IPTW. The β coefficients and p values indicate the group differences in cognitive decline rates. ADAS cog13, Alzheimer's Disease Assessment Scale‐Cognitive Subscale (13‐item version); APOE ε4, apolipoprotein E ε4; CDR‐SB, Clinical Dementia Rating Sum of Boxes; ChEI, cholinesterase inhibitor; IPTW, inverse probability of treatment weighting; LME, linear mixed effects; MCI, mild cognitive impairment; MMSE, Mini‐Mental State Examination.

TABLE 2.

Time‐varying Cox proportional hazards model for effects of cholinesterase inhibitors on progression from mild cognitive impairment to Alzheimer's disease dementia with IPTW adjustment.

Variable	Hazard ratio (HR)	95% CI (lower)	95% CI (upper)	p‐value
ChEI treatment (Ref = non‐users)	1.77	1.15	2.73	0.001
Age	1.01	0.99	1.04	0.41
Gender (Ref = male)
Female	1.55	0.99	2.42	0.05
Education	1.06	0.99	1.14	0.08
Baseline MMSE	0.86	0.77	0.97	0.01
Baseline CDR‐SB	1.87	1.54	2.27	<0.001
Baseline AV45 SUVR	12.34	5.83	26.14	<0.001
Time‐dependent APOE ε4 carriers (Ref = APOE ε4 non‐carriers)	1.27	0.95	1.71	0.11

Note: The time‐varying Cox model included APOE ε4 as the time‐dependent variable and was adjusted for age, gender, education, baseline MMSE, baseline CDR‐SB, and baseline AV45 PET. Statistical significance was set at p < 0.05.

Abbreviations: AD, Alzheimer's disease; APOE, apolipoprotein E; AV45 SUVR, florbetapir positron emission tomography standardized uptake value ratio; CDR‐SB, Clinical Dementia Rating Sum of Boxes; ChEI, cholinesterase inhibitor; CI, confidence interval; HR, hazard ratio; IPTW, inverse probability of treatment weighting; MCI, mild cognitive impairment; PET, positron emission tomography.

Furthermore, longitudinal analysis using linear mixed‐effects models demonstrated that ChEI users experienced greater cognitive decline over the follow‐up period compared to non‐users. This was evidenced by more pronounced annual changes in CDR‐SB (0.35 points increase per year; 95% CI: 0.31 to 0.39, p < 0.001), MMSE (−0.38 points decrease per year; 95% CI: −0.44 to −0.32, p < 0.001), and ADAS cog13 (0.80 points increase per year; 95% CI: 0.64 to 0.96, p < 0.001) (Figure 2C‐E; Table S4 in supporting information). There was no significant difference between short‐term and long‐term ChEI use on MCI progression (Figure S2 and Table S5 in the supporting information). While both groups showed cognitive decline, long‐term ChEI users demonstrated a trend toward greater cognitive deterioration; it was not statistically significant for CDR‐SB and ADAS cog13, though it was for MMSE (−0.14 points decrease per year; 95% CI: −0.28 to −0.003, p = 0.04) (Figure S2 and Table S6 in the supporting information).

3.3. Cognitive trajectory relative to ChEI initiation in the MCI population

Figure 3 illustrates the longitudinal changes in CDR‐SB, MMSE, and ADAS cog13 scores relative to ChEI initiation (time 0). The smoothed trajectories revealed a gradual cognitive decline in the pretreatment period (−5 to 0 years). Following ChEI initiation, rather than showing stabilization or improvement, cognitive function continued to deteriorate, as evidenced by increasing CDR‐SB and ADAS cog13 scores and decreasing MMSE scores (Figure 3A–C). When stratified by baseline amyloid status, both Aβ− and Aβ+ subgroups showed continued cognitive decline after ChEI initiation. However, the Aβ+/ChEI initiation group demonstrated notably steeper trajectories of cognitive deterioration compared to the Aβ−/ChEI initiation group, particularly in the post‐treatment period. This pattern was consistent across all three cognitive measures (Figure 3D–F).

FIGURE 3.

Cognitive trajectory relative to ChEI initiation in MCI population. LOESS regression was used to examine non‐linear patterns in cognitive changes during pre‐ and post‐treatment periods. (A–C) Overall ChEI initiation group trajectories. (D–F) Trajectories stratified by baseline Aβ status. Changes in CDR‐SB, MMSE, and ADAS cog13 scores are plotted relative to ChEI initiation time (time 0 = treatment start). Individual trajectories (thin lines) and LOESS‐smoothed curves (thick lines) with 95% confidence intervals (shaded areas) are shown. Orange lines: overall ChEI initiation group (A–C) or Aβ+ subgroup (D–F); blue lines: Aβ− subgroup (D–F). Aβ, amyloid beta; ADAS cog13, Alzheimer's Disease Assessment Scale‐Cognitive Subscale (13‐item version); CDR‐SB, Clinical Dementia Rating Sum of Boxes; ChEI, cholinesterase inhibitor; LOESS, locally estimated scatterplot smoothing; MMSE, Mini‐Mental State Examination.

3.4. Effects of ChEI treatment across different Aβ status on progression from MCI to AD dementia and cognitive trajectories

To investigate the effects of ChEI treatment in different amyloid status populations, participants were first stratified by Aβ status. Within each Aβ status group (Aβ− and Aβ+), the IPTW was separately recalculated to compare ChEI users and non‐users (ChEI−/Aβ− vs ChEI+/Aβ− in the amyloid‐negative group; ChEI−/Aβ+ vs ChEI+/Aβ+ in the amyloid‐positive group, see details in Table S7 and Table S8 in supporting information). IPTW‐adjusted Kaplan–Meier analysis revealed that the ChEI+/Aβ+ group had a significantly higher risk of progression to AD dementia compared to the ChEI−/Aβ+ group (log‐rank χ² = 16.03, p < 0.001), while no significant difference was observed between ChEI−/Aβ− and ChEI+/Aβ− groups (log‐rank χ² = 0.33, p = 0.56) (Figure 4A–D). The IPTW‐adjusted Cox proportional hazards model showed that ChEI+/Aβ+ participants had a significantly higher risk of conversion to AD dementia compared to ChEI−/Aβ+ participants (HR = 2.18, 95% CI: 1.31 to 3.63, p = 0.003) (Table S9 in the supporting information). However, no significant difference was observed in progression to AD dementia between ChEI−/Aβ− and ChEI+/Aβ− groups (HR = 1.57, 95% CI: 0.71 to 3.47, p = 0.26) (Table S10 in the supporting information). When examining reversion to NC, both ChEI users and non‐users showed similar patterns of cognitive improvement regardless of their Aβ status (Figure S3 in the supporting information). Longitudinal analysis demonstrated that the ChEI+/Aβ+ group experienced more rapid cognitive decline than the ChEI−/Aβ+ group, as evidenced by greater annual changes in CDR‐SB (0.48 points increase per year; 95% CI: 0.40 to 0.56, p < 0.001), MMSE (−0.61 points decrease per year; 95% CI: −0.71 to −0.51, p < 0.001), and ADAS cog13 (1.01 points increase per year; 95% CI: 0.70 to 1.32, p < 0.001) (Figure 4E–G, Table S11 in the supporting information). A similar pattern of accelerated decline was observed in Aβ− participants, where the ChEI+/Aβ− group showed more rapid cognitive deterioration than the ChEI−/Aβ− group, with significant annual changes in CDR‐SB (0.21 points increase per year; 95% CI: 0.17 to 0.25, p < 0.001), MMSE (−0.15 points decrease per year; 95% CI: −0.21 to −0.09 p < 0.001), and ADAS cog13 (0.37 points increase per year; 95% CI: 0.21 to 0.53, p < 0.001) (Table S12 in the supporting information).

FIGURE 4.

Effects of ChEI treatment across different Aβ status on progression from MCI to AD dementia and cognitive trajectories with IPTW adjustment. Kaplan–Meier survival curves for MCI to AD dementia progression in unadjusted analysis (A and B) and after IPTW adjustment (C and D), stratified by both ChEI use and baseline Aβ status, with numbers at risk shown below. Longitudinal trajectories of cognitive measures including CDR‐SB score changes (E and H), MMSE score changes (F and I), and ADAS cog13 score changes (G and J) over time, adjusted using IPTW‐weighted LME models. Green lines: ChEI−/Aβ−, blue lines: ChEI+/Aβ−, pink lines: ChEI−/Aβ+, orange lines: ChEI+/Aβ+. Shaded areas: 95% confidence intervals. Individual trajectories are shown as thin lines with corresponding group‐specific smoothed curves. All analyses were adjusted for age, gender, education, and APOE ε4 status using IPTW. Aβ, amyloid beta; AD, Alzheimer's disease; APOE ε4, apolipoprotein E ε4; CDR‐SB, Clinical Dementia Rating Sum of Boxes; ChEI, cholinesterase inhibitor; IPTW, inverse probability of treatment weighting; LME, linear mixed effects; MCI, mild cognitive impairment; MMSE, Mini‐Mental State Examination.

To further explore potential effect modifiers of ChEI treatment, interaction analyses were conducted with baseline Aβ burden and APOE ɛ4 status (Table 3). The interaction analysis revealed that both ChEI use (HR = 11.25, 95% CI: 1.71 to 73.96, p = 0.01) and baseline AV45 SUVR (HR = 27.83, 95% CI: 7.26 to 106.62, p < 0.001) were independently associated with increased risk of progression to AD dementia, while their interaction was not significant (HR = 0.27, 95% CI: 0.07 to 1.04, p = 0.06). Additionally, APOE ε4 status showed no significant interaction with ChEI treatment (HR = 0.91, 95% CI: 0.57 to 1.47, p = 0.71). The mediation analysis revealed no significant indirect effect of ChEI treatment on AD dementia conversion through annual changes of AV45 SUVR (ACME = −0.0002, p = 0.99) (Figure S4 in the supporting information). The path from ChEI treatment to annual changes of AV45 SUVR was not significant (β = −0.002, p = 0.64), and there was a significant direct effect of ChEI treatment on AD dementia conversion (ADE = 0.28, p < 0.001), suggesting that the effect was primarily driven by the direct effect rather than the mediation pathway through AV45 SUVR changes.

TABLE 3.

Interaction effects between cholinesterase inhibitor treatment and potential effect modifiers in Cox proportional hazards models with IPTW adjustment.

Variable	Hazard ratio (HR)	95% CI (lower)	95% CI (upper)	p
Model 1
ChEI group	11.25	1.71	73.96	0.01
AV45 SUVR	27.83	7.26	106.62	<0.001
ChEI group×AV45 SUVR	0.27	0.07	1.04	0.06
Model 2
ChEI group	1.91	1.12	3.24	0.02
APOE ɛ4	1.35	0.94	1.93	0.10
ChEI group×APOE ɛ4	0.91	0.57	1.47	0.71

Note: All models were weighted using IPTW. Model 1 examined the interaction between ChEI treatment and baseline AV45, with APOE ε4 status incorporated as a time‐dependent variable and adjusted for age, gender, education, baseline MMSE, and baseline CDR‐SB. Model 2 investigated the interaction between ChEI treatment and time‐dependent APOE ε4 status, adjusted for age, gender, education, baseline MMSE, baseline CDR‐SB, and baseline AV45. HR with 95% CI), and p values are presented for main effects and interaction terms.

Abbreviations: APOE ε4, apolipoprotein E ε4; AV45 SUVR, florbetapir positron emission tomography standardized uptake value ratio; CDR‐SB, Clinical Dementia Rating Sum of Boxes; ChEI, cholinesterase inhibitor; CI, confidence interval; HR, hazard ratio; IPTW, inverse probability of treatment weighting; MMSE, Mini‐Mental State Examination; PET, positron emission tomography.

4. DISCUSSION

This study demonstrated that ChEI use in patients with amnestic MCI was associated with an increased risk of progression to AD dementia and more pronounced cognitive decline, with the initiation of ChEI treatment failing to slow the overall rate of cognitive deterioration. Interaction analyses revealed that ChEI use and Aβ burden were independently associated with increased risk of progression to AD dementia, without a significant interaction between them. The independence of these effects was further supported by mediation analyses showing no significant indirect effect of ChEI treatment through annual changes in AV45 SUVR, suggesting that the impact of ChEIs on cognitive decline operated through mechanisms distinct from amyloid accumulation.

Although fewer than 50% of MCI patients receive ChEI treatment in clinical practice, ¹⁶ , ²⁰ such prescriptions are typically off‐label, emphasizing the importance of understanding their longitudinal effects in real‐world care settings. This study demonstrated that amnestic MCI participants who received ChEIs had an increased risk of progressing to AD dementia and experienced greater cognitive decline in CDR‐SB (0.35 points increase per year) compared to those who did not receive treatment. Furthermore, results suggested that cognitive decline in individuals with MCI began before the initiation of ChEI treatment and continued during the following 5 years (0 to 5 years) rather than halting after treatment began. Previous research using the National Alzheimer's Coordinating Center's Uniform Data Set demonstrated that MCI due to AD experienced more pronounced cognitive decline after ChEI initiation, with CDR‐SB scores showing an acceleration from 0.03 points increase year before initiation to 0.61 points increase year after initiation. ¹⁵ A meta‐analysis of 10 AD clinical trials (n = 2714) demonstrated that ChEI users exhibited a substantial decline of 1.4 points increase per year on the ADAS‐cog compared to non‐users, further supporting these observations. ⁴⁵ Nevertheless, conflicting evidence exists in the literature. For instance, a 48‐week randomized placebo‐controlled trial of donepezil in individuals with MCI demonstrated a small but significant improvement in ADAS‐cog scores. ³⁰ Additionally, while another clinical trial indicated initial benefits of donepezil in reducing progression to AD during the first 12 months in MCI patients, these advantages dissipated by the 3‐year follow‐up, showing no significant differences compared to placebo. ²⁶ In this study, while initial conversion rates were similar between groups, ChEI users demonstrated significantly higher rates of progression from MCI to AD dementia compared to non‐users after 6 years of follow‐up. Compared with short‐term users, long‐term ChEI users demonstrated a trend toward greater cognitive deterioration. These findings suggested that ChEIs did not confer long‐term protective effects against AD progression, and prolonged use may be associated with accelerated cognitive deterioration.

A critical limitation in previous studies has been the reliance on clinical diagnosis of MCI without confirming underlying AD pathology, potentially contributing to heterogeneous findings. The current study addressed this limitation by incorporating amyloid PET imaging, revealing that ChEI users with positive Aβ status experienced the highest risk of progression to AD dementia and more rapid cognitive decline over the 10‐year follow‐up period, with continued deterioration following ChEI initiation. Interaction analyses revealed that ChEI use and baseline amyloid burden were independently associated with increased risk of progression to AD dementia. This independence was further corroborated by mediation analyses showing no significant indirect effect of ChEI treatment through annual changes in AV45 SUVR, suggesting that ChEIs influenced cognitive decline through mechanisms distinct from amyloid accumulation. These results align with previous 2‐year investigations demonstrating that ChEI use was associated with more pronounced cognitive decline regardless of amyloid status. ³⁷ These factors appear to operate through distinct pathological pathways: ChEIs primarily affect the cholinergic system with rapid alterations in synaptic function, while Aβ accumulation impacts multiple neural systems through gradual neurodegenerative changes. ⁴⁶ Consistent with these findings, animal studies demonstrated that ChEI treatment did not slow the progression of amyloid pathology. ⁴⁷ Furthermore, post mortem analyses revealed that ChEI treatment was associated with increased phosphorylated‐tau accumulation while having limited impact on Aβ burden. ³⁸

Several limitations of this study warrant consideration. First, the observational nature of this ADNI cohort study, without randomization or double‐blind medication administration, precluded definitive conclusions about causality, allowing only for the assessment of associations. Second, the ADNI database did not systematically collect mortality data to address death as a potential competing risk for AD progression, and not all participants completed the full 10‐year follow‐up period. Although IPTW was employed to balance between‐group differences and common demographic and clinical confounders were accounted for, potential bias may persist due to unmeasured confounders. Third, the reliance on retrospective medication records limited the assessment of treatment adherence and dosage effects. Moreover, without documentation of clinical decision‐making, it remained unclear whether cognitive decline preceded and prompted ChEI initiation or resulted from the treatment itself, complicating the interpretation of ChEI efficacy. Fourth, from a diversity, equity, and inclusion (DEI) perspective, an important limitation must be acknowledged. Despite careful matching of demographic characteristics between groups at baseline, the study sample was predominantly composed of highly educated, White, non‐Hispanic individuals, constraining the generalizability of findings across diverse populations.

In conclusion, this study demonstrated that ChEI use in amnestic MCI participants was associated with increased risk of progression to AD dementia and accelerated cognitive decline, independent of Aβ status. While cognitive deterioration may have triggered ChEI initiation, the treatment appeared unable to halt or slow the progression of cognitive decline. These findings call for a critical reevaluation of current ChEI prescription practices in amnestic MCI patients. Future research priorities should focus on conducting long‐term randomized controlled trials in larger, more diverse populations to evaluate the efficacy of ChEI treatment in amnestic MCI.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Author disclosures are available in the supporting information.

CONSENT STATEMENT

ADNI was approved by the Institutional Review Boards of all participating institutions. All ADNI participants provided written informed consent according to the Declaration of Helsinki before study enrollment.

Supporting information

Supporting Information

Supporting Information

Liu W, Li Y, Qin W, et al. The impact of cholinesterase inhibitors on cognitive trajectories in mild cognitive impairment patients based on amyloid beta status. Alzheimer's Dement. 2025;21:e70193. 10.1002/alz.70193