Impact of dose and compliance of antidementia medications on long-term outcomes in Alzheimer’s disease: a nationwide real-world study

Abstract

Background

Antidementia medications are widely prescribed for Alzheimer’s disease (AD), but their long-term real-world effectiveness remains uncertain. This study investigated whether long-term outcomes differ according to medication dosage and compliance using nationwide data.

Methods

Data from the Korean National Health Insurance Service (NHIS) covering 47 million individuals were analyzed. Prescription data for acetylcholinesterase inhibitors and memantine were analyzed for dosage and compliance. Among 1,704,547 dementia cases (2010–2016), 466,773 patients with clinically diagnosed AD were included. Medication dosage and compliance during the first three years after diagnosis were categorized to define optimal versus suboptimal treatment. Clinical outcomes included progression to moderate to severe dementia, institutionalization, and mortality. Multivariable logistic regression identified factors associated with outcomes.

Results

Patients who maintained optimal dosage and compliance during the first three years after diagnosis showed a lower rate of progression to moderate to severe dementia than those receiving suboptimal treatments consistently across all classification criteria. Regression analyses revealed that optimal compliance and dosage were strongly associated with reduced progression (OR 0.807 and 0.704, respectively; p < 0.0001) and early mortality within five years. In contrast, mortality and institutionalization rates were not significantly different between groups except that mortality within five years.

Conclusions

Both medication dosage and persistence were independently associated with better long-term outcomes in AD. Maintaining optimal treatment during the early disease period may delay disease progression and improve survival within five years. This nationwide real-world study provides robust evidence supporting the importance of sustained, adequate antidementia therapy in clinical practice.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13195-025-01942-0.

Introduction

Antidementia medications, acetylcholinesterase inhibitors (ChEIs) and N-methyl-D aspartate receptor antagonist (memantine), are widely prescribed for patients with Alzheimer’s disease (AD) to improve cognitive and functional outcomes [1]. Short-term clinical trials and open label extension studies have demonstrated modest cognitive and functional benefits, as well as behavioral improvements of antidementia medications [2–8] in AD dementia. However, their efficacy in delaying long-term clinical progression has not been clearly established, as most studies were of relatively short duration below 6 months. Several studies have examined long-term clinical outcomes associated with antidementia medication use, but their results have been inconsistent. A 3-year placebo-controlled trial reported that donepezil treatment did not delay institutionalization or reduce disability, despite improvements in cognition and function, suggesting limited cost-effectiveness [9]. In contrast, several observational studies have reported that antidementia medication use delay nursing home admission [10, 11] and prolong the time to functional endpoint and death [12]. Nevertheless, no studies have comprehensively clarified long-term clinical outcomes in relation to dosage and persistency of antidementia medications in real-world clinical settings. Because dose and persistency of antidementia medications vary considerably in routine practice [13], examining their association with long-term outcomes using large-scale, real-world data could provide valuable insights for optimizing treatment strategies.

The Korean National Health Information Database (NHID) includes nationwide claims data and health examination results supervised by the Ministry of Health and Welfare. Since the Korean National Health Insurance Service (KNHIS) operates as a single insurer that controls the national medical system, the NHID constitutes a comprehensive big data source containing information on insurance coverage, medical check-ups, past medical history and comorbidities, detailed questionnaires on lifestyle factors, medication details, admission history to long-term care hospital, hospitalization history, disease registration, and mortality. We hypothesized that patients with AD who consistently take antidementia medications with standard dosage would show better long-term clinical outcomes compared with those receiving suboptimal treatments. Using NHID data, we aimed to test this hypothesis.

In this study, we aimed to investigate whether long-term clinical outcomes including incidence of clinical progression to moderate to severe dementia, institutionalization, and mortality, differ according to drug compliance and dosage of antidementia medications in AD. We also assessed relevant factors associated with these outcomes.

Methods

Data source

We extracted data from the National Health Insurance Service (NHIS) database, which contains healthcare information for nearly 47 million individuals, focusing on patients aged 60 and above who were diagnosed with dementia between 2010 and 2016. The medication utility information was analyzed with respect to antidementia medication dose and compliance across 12 years.

Data acquisition

This study utilized data from the NHID, spanning from 2009 to 2022. The primary cohort included patients aged 60 years and older who were first diagnosed with dementia between 2010 and 2016. Long-term clinical outcomes (within 5 years of dementia diagnosis, after 5 years of dementia diagnosis, all periods) were assessed based on data between 2010 and 2022. Data collection was performed in November 2024 after obtaining authorization from KNHIS. Dementia due to AD was defined using the International Classification of Diseases (ICD) codes (10th revision, ICD-10): (1) F00 (Dementia in Alzheimer’s disease) and its subtypes including F00.0 (Dementia in Alzheimer’s disease with early onset), F00.1 (Dementia in Alzheimer’s disease with late onset), F00.2 (Dementia in Alzheimer’s disease, atypical or mixed typ10th revision, ICD-10), and F00.3 (unspecified dementia/presenile/senile dementia). (2) G30 (Alzheimer’s disease) and its subtypes including G30.0 (Alzheimer’s disease with early onset), G30.1 (Alzheimer’s disease with late onset), and G30.9 (Alzheimer’s disease, unspecified).

Participants

Individuals who were clinically diagnosed with AD dementia between 2010 and 2016 were eligible for the study. Inclusion criteria were as follows: (1) age 60 years old or older at the time of diagnosis, (2) diagnosed with dementia due to Alzheimer’s disease based on the predefined ICD codes, (3) patients who took antidementia medication (donepezil/galantamine/rivastigmine/memantine) with at least 70% drug compliance during the first year from dementia diagnosis. Drug compliance (%) during the first year was assessed based on the number of days of antidementia medication prescribed over a year (prescription days/365 days). Exclusion criteria were as follows: (1) rapid progression to moderate to severe dementia (defined as Mini-Mental State Examination (MMSE) [14] ≤ 18 and clinical dementia rating (CDR) [15] ≥ 2 under national certification system) or institutionalization within one year after being diagnosed with dementia, (2) brain tumor within 1 year (before and after) of dementia diagnosis, (3) patients with human immunodeficiency virus (HIV) based on the ICD codes, (4) no medication prescription during the first three years after dementia diagnosis, 4) missing baseline or follow-up data. This study was approved by the institutional review board (#UC23ZASI0016).

Outcome measures

Baseline age, sex, comorbidities (existence of diabetes mellitus, hypertension, hyperlipidemia, heart failure, or atrial fibrillation), socioeconomic factors (income, drinking, smoking, and regular exercise by self-reported questionnaires), and physical values (including body mass index, blood pressure, hemoglobin values, fasting glucose, total cholesterol, triglyceride, HDL-cholesterol, LDL-cholesterol, creatinine, waist circumference, emotional stress) were collected from national health checkup data within 1 year of dementia diagnosis.

The study examined three primary outcomes: (1) progression to moderate to severe dementia (using predefined ICD codes) until 2022, (2) Mortality (death events were identified using data provided by the Korean National Statistical Office, which included the date and cause of death). (3) Institutionalization. Progression to moderate to severe dementia was defined using ICD codes G30.1 (Alzheimer’s disease with late onset) and F00.1 (Dementia in Alzheimer’s disease with late onset), the special classification code V810 under the Korean national insurance system, within five years and beyond five years post-diagnosis. The V810 code is a welfare system that reduces medical expenses when a patient progresses to moderate to severe dementia (MMSE score ≤ 18 and CDR score ≥ 2) under the Korean national insurance system, hence all patients who progress to moderate to severe dementia register the special classification code. The physician must submit the results and date of a detailed neuropsychological tests battery along with a brain imaging test performed to the national insurance system to determine eligibility for V810. Institutionalization was identified as admission to nursing or psychiatric hospitals for long-term care without a predefined length of stay based on recorded institutionalization data from KNHIS database. Dose of antidementia medication, component of the antidementia medication, and medication possession ratio (MPR) during the first 3 years from the dementia diagnosis were collected to categorize groups.

We compared incidence of moderate to severe dementia within the first 5 years from the dementia diagnosis, incidence of moderate to severe dementia after 5 years, mortality within the first 5 years, mortality after 5 years, and institutionalization within the first 5 years, institutionalization after 5 years between the two groups.

Group classification

To compare long-term clinical outcomes according to dosage and compliance during the first three years after dementia diagnosis, patients were categorized into two groups based on the following 3 categories: (1) MPR (prescription days/365 days, %): patients with MPR < 70% and those with MPR ≥ 70%, (2) Antidementia medication dosage: patients receiving standard-dose treatment versus those receiving low-dose treatment (≤ 50% of the standard dose) with an MPR ≥ 70%, (3) Optimal treatments (based on both MPR and dosage): patients with optimal dosing (MPR ≥ 70% and standard dosage) versus those with suboptimal treatment (MPR < 70% or low dosage).

Regarding category 1), optimal drug compliance (MPR ≥ 70%) indicated antidementia medication (ChEIs or memantine) at the same dose for at least 70% of the days during the first three years following dementia diagnosis. For category 2), optimal dose intensity was defined as receiving a standard daily dose (≥ 10 mg donepezil, ≥ 12 mg rivastigmine, ≥ 16 mg galantamine, or 20 mg memantine) with persistence ≥ 70% during the same period. For category 3), optimal treatments indicated maintaining both an MPR ≥ 70% and a standard (or higher) dosage for at least three years after diagnosis.

Statistical analysis

All statistical analyses were performed using SAS Enterprise Guide (version 8.3; SAS Institute, Cary, NC, USA). Baseline characteristics and comorbidities were reported as mean ± standard deviation for continuous variables and as proportions for categorical variables. For group comparisons, chi-square tests were used for categorical variables and Student’s t-tests for continuous variables. Because large sample sizes may yield statistically significant p-values even for negligible differences, standardized mean differences (SMDs) were additionally calculated to assess the magnitude of group differences. An SMD < 0.1 was considered to indicate negligible imbalance, whereas an SMD ≥ 0.2 was considered meaningful. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated to evaluate the associations between baseline variables and clinical outcomes (progression to moderate-to-severe dementia and mortality) using multiple logistic regression models. The models were adjusted for age, sex, and income level. Additionally, multivariable analyses with backward elimination were conducted to identify the most relevant factors associated with clinical outcomes.

Results

Study population

From the initial cohort of 1,704,547 patients diagnosed with dementia between 2010 and 2016, several exclusion criteria were applied: Patients who died or progressed to severe dementia within one year after the dementia diagnosis (n = 427,534), brain tumors within one year prior to the index date (n = 3,306), patients diagnosed with HIV during the entire period (n = 2,127), patients who did not take medication within three years after diagnosis (n = 804,807) were excluded as they were likely to have either incomplete diagnostic workups or inaccurate AD diagnosis. After applying these exclusion criteria, 466,773 patients were included in the final analysis (Fig. 1).

Fig. 1.

Study flowchart

During the study period of 12 years, 140,136 (30.02%) patients progressed to moderate to severe dementia state, 295,156 (63.23%) patients died, and 269,004 (57.63%) patients were admitted to nursing hospitals.

Clinical outcomes according to drug compliance

Of 466,773 patients, 130,435 who received standard doses during the first three years after diagnosis were included for compliance analysis. Baseline characteristics were comparable between compliance groups (Table 1).

Table 1.

Clinical outcomes according to drug compliance during the first 3 years

Variables	Low compliance (MPR <70%)		Optimal compliance (MPR ≥70%)		P value	SMD
	n=64,332		n=66,103
	N, Mean	%, SD	N, Mean	%, SD
Age (year)	78.77	7.25	77.48	7.01	<0.0001	0.1809
Women	42142	65.5%	44821	67.8%	<0.0001	0.0492
Diabetes	48273	75.0%	50012	75.7%	0.0093	−0.0152
Hypertension	29867	46.4%	30415	46.0%	0.133	0.008
Hyperlipidemia	35618	55.4%	37360	56.5%	<0.0001	−0.0221
Heart failure	10141	15.8%	8905	13.5%	<0.0001	0.0606
Atrial fibrillation	4711	7.3%	4265	6.5%	<0.0001	0.0313
Drinking	2275	3.5%	2389	3.6%	0.0033	−0.0055
Regular exercise	9252	14.4%	9889	15.0%	<0.0001	−0.0164
Smoking	2209	3.4%	2234	3.4%	<0.0001	0.0000
BMI (body mass index)	23.35	3.33	23.59	3.26	<0.0001	−0.0728
Hemoglobin	13.08	1.63	13.15	1.54	<0.0001	−0.0442
Fasting glucose	110.69	38.04	109.2	34.36	<0.0001	0.0411
Total cholesterol	193.45	43.75	195.34	41.37	<0.0001	−0.0444
Waist circumference	82.64	9.04	82.83	10.69	0.0361	−0.0192
Triglyceride	139.94	88.48	141.18	88.32	0.1356	−0.0140
HDL-cholesterol	53.2	20.58	53.31	33.92	0.6705	−0.0039
LDL-cholesterol	113.94	43.59	115.74	46.12	<0.0001	−0.0401
Creatinine	0.99	0.69	0.98	0.72	0.0393	0.0142
Dose of antidementia medication (mg/day)
Rivastigmine(Mean, %)	14.7	26.38	13.71	19.67	0.1085	0.043
Donepezil(Mean, %)	10.09	4.13	9.86	3.47	<0.0001	0.0603
Galantamine(Mean, %)	15.72	4.64	15.89	5.06	0.2744	−0.035
Memantine(Mean, %)	18.93	57.53	18.81	59.81	0.898	0.002
Clinical outcomes
Moderate to severe  Dementia (progression)	23192	36.1%	9410	14.2%	<0.0001	0.5038
Within 5yrs	17202	26.7%	7529	11.4%	<0.0001	0.3916
After 5 yrs	5990	9.3%	1881	2.8%	<0.0001	0.2715
Mortality	46650	72.5%	44252	66.9%	<0.0001	0.1212
Within 5yrs	32589	50.7%	20557	31.1%	<0.0001	0.3981
After 5 yrs	14061	21.9%	23695	35.8%	<0.0001	−0.3085
Institutionalization	41967	65.2%	43685	66.1%	<0.0001	−0.0179
Within 5 yrs	32790	51.0%	28163	42.6%	<0.0001	0.1677
After 5 yrs	9177	14.3%	15522	23.5%	<0.0001	−0.2352

Abbreviations: MPR Medication possession ratio, SMD Standardized mean differences

Patients with optimal compliance (MPR ≥ 70%) had markedly lower rates of progression to moderate-to-severe dementia and mortality than those with poor compliance. Within five years, progression occurred in 11.4% of the optimal compliance group versus 26.7% in the low compliance group, and mortality in 31.1% versus 50.7%. Institutionalization rates were slightly lower in the optimal group. Overall, optimal compliance was associated with slower disease progression and lower early mortality.

Clinical outcomes according to dose intensity

Among 100,874 patients with MPR ≥ 70%, 66,103 received optimal doses (daily dose ≥ standard dose) and 34,771 received low doses (daily dose ≤ 50% of the standard dose). Baseline features were similar between the two groups (Table 2).

Table 2.

Clinical outcomes according to dose intensity during the first 3 years

Variables	Low dose		Standard dose		P value	SMD
	n = 34,771		n = 66,103
	N, Mean	%, SD	N, Mean	%, SD
Age (year)	77.38	7.18	77.48	7.01	< 0.0001	−0.0141
Women	25,094	72.2%	44,821	67.8%	< 0.0001	−0.0946
Diabetes	27,230	78.3%	50,012	75.7%	0.858	0.0627
Hypertension	16,219	46.6%	30,415	46.0%	0.367	0.0127
Hyperlipidemia	20,128	57.9%	37,360	56.5%	< 0.0001	0.0277
Heart failure	5139	14.8%	8905	13.5%	< 0.0001	0.0378
Atrial fibrillation	2129	6.1%	4265	6.5%	< 0.0001	−0.0135
Drinking	1097	3.2%	2389	3.6%	< 0.0001	−0.0251
Regular exercise	5564	16.0%	9889	15.0%	0.0004	0.0289
Smoking	1027	3.0%	2234	3.4%	< 0.0001	−0.0241
BMI (body mass index)	23.86	3.26	23.59	3.26	0.1803	0.0828
Hemoglobin	13.06	1.5	13.15	1.54	< 0.0001	−0.0590
Fasting glucose	108.21	34.19	109.2	34.36	0.0115	−0.0289
Total cholesterol	195.11	44.94	195.34	41.37	0.239	−0.0054
Waist circumference	83.09	8.7	82.83	10.69	0.1417	0.0259
Triglyceride	143.22	91.9	141.18	88.32	0.0761	0.0228
HDL-cholesterol	52.78	19.49	53.31	33.92	0.0417	−0.0178
LDL-cholesterol	116	50.2	115.74	46.12	0.7883	0.0055
Creatinine	0.97	0.73	0.98	0.72	0.1964	−0.0138
Dose of antidementia medication (mg/day)
Rivastigmine(Mean, %)	9.72	14.12	13.71	19.67	< 0.0001	−0.2223
Donepezil(Mean, %)	5.66	3.97	9.86	3.47	< 0.0001	−1.1507
Galantamine(Mean, %)	11.18	4.97	15.89	5.06	< 0.0001	−0.9365
Memantine(Mean, %)	10	0	18.81	59.81	< 0.0001	−0.1820
Clinical outcomes
Moderate to severe Dementia (progression)	11,890	34.2%	9410	14.2%	< 0.0001	0.4891
Within 5yrs	8009	23.0%	7529	11.4%	< 0.0001	0.3226
After 5 yrs	3881	11.2%	1881	2.8%	< 0.0001	0.3583
Mortality	20,473	58.9%	44,252	66.9%	< 0.0001	−0.1682
Within 5yrs	8774	25.2%	20,557	31.1%	< 0.0001	−0.1291
After 5 yrs	11,699	33.6%	23,695	35.8%	< 0.0001	−0.0461
Institutionalization	20,460	58.8%	43,685	66.1%	< 0.0001	−0.1505
Within 5 yrs	12,691	36.5%	28,163	42.6%	< 0.0001	−0.1244
After 5 yrs	7769	22.3%	15,522	23.5%	< 0.0001	−0.027

Abbreviations: SMD Standardized mean differences

Clinical progression occurred less frequently in the optimal-dose group (11.4% vs. 23.0% within five years), whereas mortality and institutionalization did not differ although mortality and institutionalizations within five years were numerically more in the optimal dose group (Table 2). Among patients with good compliance, dose intensity primarily influenced cognitive progression but not survival or institutionalization.

Clinical outcomes according to optimal treatments

When optimal treatment (optimal dose + optimal compliance, n = 66,103) was compared with suboptimal treatment (low dose or low compliance, n = 400,670), baseline characteristics were similar (Table 3).

Table 3.

Clinical outcomes according to optimal drug treatments during the first 3 years

Variables	Sub-optimal treatment		Optimal treatment		P value	SMD
	n = 400,670		n = 66,103
	N, Mean	%, SD	N, Mean	%, SD
Age (year)	77.91	7.29	77.48	7.01	< 0.0001	0.0593
Women	273,136	68.2%	44,821	67.8%	< 0.0001	−0.0078
Diabetes	298,260	74.4%	50,012	75.7%	0.858	−0.0280
Hypertension	183,581	45.8%	30,415	46.0%	0.367	−0.0039
Hyperlipidemia	222,595	55.6%	37,360	56.5%	< 0.0001	−0.0194
Heart failure	61,244	15.3%	8905	13.5%	< 0.0001	0.0508
Atrial fibrillation	24,879	6.2%	4265	6.5%	< 0.0001	−0.0100
Drinking	14,737	3.7%	2389	3.6%	< 0.0001	0.0034
Regular exercise	64,861	16.2%	9889	15.0%	0.0004	0.0335
Smoking	13,914	3.5%	2234	3.4%	< 0.0001	0.0051
BMI (body mass index)	23.58	3.3	23.59	3.26	0.0892	−0.0030
Hemoglobin	130.62	16.57	130.96	16.72	< 0.0001	−0.0581
Fasting glucose	13.06	1.55	13.15	1.54	< 0.0001	−0.0164
Total cholesterol	108.63	34.73	109.2	34.36	< 0.0001	−0.0096
Waist circumference	194.92	44.17	195.34	41.37	0.123	−0.0039
Triglyceride	82.79	10.31	82.83	10.69	0.1399	0.0074
HDL-cholesterol	141.84	89.61	141.18	88.32	0.0599	−0.0015
LDL-cholesterol	53.27	26.11	53.31	33.92	0.1616	−0.0102
Creatinine	115.26	46.99	115.74	46.12	0.0094	−0.0109
Dose of antidementia medication (mg/day)
Rivastigmine (Mean, %)	12.03	40.34	13.71	19.67	< 0.0001	−0.0441
Donepezil (Mean, %)	7.27	4.59	9.86	3.47	< 0.0001	−0.5822
Galantamine (Mean, %)	11.38	5.01	15.89	5.06	< 0.0001	−0.8989
Memantine (Mean, %)	24.95	131.57	18.81	59.81	< 0.0001	0.0495
Clinical outcomes
Moderate to severe Dementia (progression)	130,726	32.6%	9410	14.2%	< 0.0001	0.4745
Within 5yrs	96,806	24.2%	7529	11.4%	< 0.0001	0.3644
After 5 yrs	33,920	8.5%	1881	2.8%	< 0.0001	0.2636
Mortality	250,904	62.6%	44,252	66.9%	< 0.0001	−0.089
Within 5yrs	147,548	36.8%	20,557	31.1%	< 0.0001	0.1322
After 5 yrs	103,353	25.8%	23,695	35.8%	< 0.0001	−0.2281
Institutionalization	225,319	56.2%	43,685	66.1%	< 0.0001	−0.2123
Within 5 yrs	159,597	39.8%	28,163	42.6%	< 0.0001	−0.0583
After 5 yrs	65,722	16.4%	15,522	23.5%	< 0.0001	−0.2283

Abbreviations: SMD Standardized mean differences

Progression was substantially lower in the optimal treatment group (11.4% vs. 24.2% within five years). In contrast, the mortality and institutionalization showed mixed patterns. Early mortality was also lower, although mortality after five years and institutionalization were higher in the optimal group. Thus, optimal early treatment was associated with reduced early clinical progression and mortality, with divergent long-term mortality and institutionalization patterns.

Factors associated with rapid progression and mortality

In multivariable models, older age, female sex, and hyperlipidemia increased the risk of rapid progression after adjustment for age, sex, and income level (Supplementary Table 1, Fig. 2). No drinking, no smoking, diabetes, heart failure, atrial fibrillation, higher body mass index (BMI), greater waist circumference, higher hemoglobin, and optimal medication persistency and dosage were protective. Drug persistency and optimal dose showed strong associations with reduced progression (OR 0.807 and 0.704, respectively; p < 0.0001). In subgroup analyses, both optimal dose ChEI monotherapy and optimal dose memantine monotherapy were independently associated with a reduced risk of rapid disease progression (ChEI monotherapy: OR, 0.793; p < 0.0001; memantine monotherapy: OR, 0.721; p < 0.0001).Mortality risk was elevated with age, diabetes, hypertension, heart failure, atrial fibrillation, smoking, alcohol consumption, elevated fasting glucose, and high creatinine adjusted for age, sex, and income level (Supplementary Table 1, Fig. 3). In contrast, protective factors included hyperlipidemia, higher BMI, regular exercise, higher hemoglobin, greater waist circumference, and optimal drug compliance and dosage. Particularly, heart failure and atrial fibrillation were associated with a markedly higher mortality risk, whereas no smoking, optimal compliance (OR 0.854), and optimal dosage (OR 0.664; p < 0.0001) were associated with a notably lower risk (Fig. 3). In subgroup analyses, both optimal dose ChEI monotherapy and optimal dose memantine monotherapy were associated with lower mortality (ChEI monotherapy: OR, 0.813; p < 0.0001; memantine monotherapy: OR, 0.784; p < 0.0001).

Fig. 2.

Factors related with rapid progression within 5 years

Fig. 3.

Factors related with mortality

Discussion

In the study, we evaluated long-term clinical outcomes of patients clinically diagnosed with AD to identify whether these outcomes differ according to drug compliance and dosage of antidementia medications. Clinical outcomes included progression to moderate to severe dementia, institutionalization to nursing hospitals, and mortality based on nationwide ‘real world’ data from the national claims database.

Our study yielded several major findings. First, when AD patients were classified into two groups according to drug compliance, dosage, or both, those who maintained optimal treatments during the first three years after diagnosis showed a lower rate of progression to moderate to severe dementia than those who received suboptimal treatments. This trend was consistent across all classification criteria. Particularly, rapid clinical progression within five years were markedly reduced in the optimal treatment groups, regardless of whether the classification criteria were based on compliance or dosage. Consistently, regression analyses identifying factors associated with clinical outcomes revealed that optimal dosage and good compliance were strongly associated with lower risks of disease progression and mortality within five years. These findings suggest that maintaining optimal dosage and treatment persistence may have a protective effect against rapid clinical progression. Our results align with previous studies reporting that persistent use of antidementia medications has positive effects on cognitive, functional, and global outcomes [16, 17], with cognitive benefits sustained over time. Although antidementia medications are known to provide modest cognitive and functional benefits without reducing underlying AD related pathologies, the mechanisms by which they may delay clinical progression remain uncertain. Sustained cognitive and functional benefits might contribute indirectly by enabling patients to maintain social engagement, physical activity, and overall brain function over a longer period.

Second, mortality rates did not generally differ according to drug compliance or dosage intensity. This finding is consistent with a previous study that reported no significant differences in institutionalization or disability progression between patients receiving donepezil and those on placebo over three years [9]. Only mortality within five years occurred less in the optimal drug compliance group than in the low compliance group. This reduction in early mortality may partly reflect the cognitive and functional benefits of antidementia medications. Similarly, prior cohort studies have shown that donepezil use was associated with a lower risk of mortality, possibly due to vagotonic and anti-inflammatory effects that reduce cardiovascular mortality [18, 19]. These authors also suggested that maintaining independence and cognitive function could extend life expectancy by facilitating continued participation in psychosocial and physical activities [18]. Another possible explanation is that patients with good compliance may undergo more regular hospital visits and health check-ups, enabling earlier detection and management of fatal conditions, whereas those with better overall health are also more likely to maintain treatment adherence. In contrast, institutionalization rates and later institutionalization/mortality after five years were higher in the optimal treatment group. These later outcomes and institutionalization may not be directly related to medication compliance or dosage but could instead reflect the accumulation of comorbidities or worsening physical health over time.

Regarding the relationship between drug use and long-term outcomes, a previous observational study found that ChEIs use was associated with delayed progression to functional decline and death, and memantine use was associated with longer survival [12]. These findings suggest potential long-term benefits of ChEIs and memantine on patient outcomes. Our results are consistent with those prior reports but extend them by demonstrating that optimal treatment defined by both persistence and dosage was strongly associated with improved long-term outcomes.

Third, clinical progression rates were lowest among patients who maintained both optimal compliance and dosage during the first three years after diagnosis compared with those in the low-dose or low-compliance groups. When comparing the relative importance of compliance and dosage, good compliance (even at lower doses) appeared to yield better clinical outcomes than optimal dosage without regular medication use, particularly regarding rapid progression, early mortality, or early institutionalization within five years. Although previous studies have shown that higher ChEI doses and longer treatment duration were associated with longer survival in AD [20, 21], our findings suggest that consistent medication adherence and regular clinical follow-up may be more beneficial for long-term outcomes than dosage optimization alone. This is further supported by our finding that patients with optimal dosage but poor compliance exhibited the highest mortality within five years.

Our study has several limitations. First, although the diagnosis was Alzheimer’s disease, it was based on insurance claims data without neuroimaging (e.g., amyloid PET or MRI) or fluid biomarker (CSF or blood) information, making it difficult to exclude mixed or non-AD pathologies. Although biomarker data were not available, dementia diagnosis and staging in South Korea follow a national certification system requiring objective cognitive testing (both MMSE and CDR) for medication reimbursement and V810 registration for moderate to severe dementia. This standardized verification process supports the clinical accuracy of disease identification and progression in our cohort. In a similar context, diagnostic approaches may differ across primary clinics and university-affiliated hospitals, and the NHIS database does not capture their relative proportions. To improve diagnostic accuracy, we excluded patients without any antidementia prescriptions in the first three years and those with rapid progression or conditions unlikely to represent AD (early nursing hospital admission, early progression to moderate to severe dementia, HIV, or brain tumors). Second, although baseline comorbidity and laboratory data were available, information on acute medical events such as stroke, major surgery, or hospitalization—which could accelerate clinical progression—was not included and should be considered when interpreting the results. Third, as with all observational cohort studies, causal inferences cannot be established. The NHID primarily records insurance claims and reimbursements; thus, data on non-reimbursed medications or procedures were unavailable, and the severity of comorbidities was not recorded. In addition, not all eligible individuals undergo regular health examinations, introducing potential selection bias. Fourth, data on psychiatric comorbidities and antipsychotic/antidepressant medication use were not available, although these factors may influence medication compliance and dosage. In this study, we focused on elucidating whether drug compliance and dosage are associated with long-term outcomes. Finally, medication persistency was estimated from prescription refill data, which does not guarantee actual medication adherence.

Notwithstanding the several limitations, our study has a few notable strengths. Most prior studies compared outcomes between treated and untreated patients, whereas our analysis examined whether long-term outcomes differ according to treatment persistency and dosage among patients who were all receiving antidementia medications. In Korea, most patients diagnosed with AD receive treatment under the National Health Insurance Service, making untreated cases rare. Therefore, comparing patients with optimal versus suboptimal treatment provides a more accurate reflection of real-world practice. Furthermore, the large sample size, long follow-up period, and use of a nationwide database enhance the generalizability of our findings and provide real world evidence on the long-term effects of optimal antidementia medication use.

In conclusion, patients who maintained optimal antidementia medication treatment-both in dosage and compliance-during the first three years after diagnosis demonstrated better long-term clinical outcomes, particularly a lower incidence of rapid progression to moderate to severe dementia within five years.

Authors’ contributions

TW Kim, YJ Hong: drafting/revision of the manuscript; study concept and design; analysis and interpretation of data. M Kim: major role in the acquisition of data; drafting of the manuscript for analysis and interpretation of data. Y Bae: drafting of the manuscript for analysis and interpretation of data. SB Lee: drafting of the manuscript for content. SH Kim: drafting of the manuscript for content. MA Lee: drafting of the manuscript for content. E Ko: drafting of the manuscript for content. JW Park: drafting of the manuscript for content. DW Yang: drafting of the manuscript for study concept.

Funding

This study was supported by the Beum Saeng Kim academic research fund, Department of Neurology, The Catholic University of Korea, the Hanmi pharm. Co., and The Uijeongbu St. Mary’s Hospital Clinical Research Laboratory Foundation made in the program of 2022.

Data availability

The datasets used during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Institutional Review Board (UC23ZASI0016). The study was conducted in accordance with the Declaration of Helsinki and principles of Good Clinical Practice.

Competing interests

The authors declare no competing interests.

Consent for publication

Not applicable.