Persistence with antidementia therapy in Germany: A retrospective cohort study of 567,815 patients

Abstract

Background

Despite the availability of these therapies, maintaining long-term adherence remains a significant challenge.

Objective

This retrospective cohort study aimed to investigate 12-month and 5-year persistence with antidementia drug therapy in Germany and to examine the association between demographic and clinical variables and the risk of therapy discontinuation.

Methods

Patients aged 60 years or older from the IQVIA Longitudinal Prescription Database who received an initial prescription for antidementia therapy between 2016 and 2023 (index date) were included. Time to discontinuation was estimated using the Kaplan–Meier method, and a multivariable Cox proportional hazards model was used to assess associations between predefined variables and the risk of discontinuation.

Results

The study included 567,815 patients (mean age: 80.2 years, 59.1% female). Five years after the index date, 19.8% of dementia patients were still receiving therapy, with a 12-month persistence rate of 53.1%. Cox regression models conducted for the total population revealed that younger age (<70 versus ≥90 years; HR: 1.21; 95% CI: 1.19–1.23; 71–80 years versus ≥90 years; HR: 1.13; 95% CI: 1.11–1.14) was significantly associated with an increased risk of therapy discontinuation. Initiating therapy with memantine was associated with a slightly lower risk of discontinuation compared to donepezil (HR: 0.87; 95% CI: 0.86–0.87).

Conclusion

In this study, half of the patients discontinued antidementia therapy within one year and 80% within five years. Younger age was linked to a higher risk of therapy discontinuation, while memantine therapy was associated with improved persistence, potentially reflecting better adherence among patients with more advanced dementia.

Introduction

Dementia primarily affects older adults and is one of the most prevalent neurological conditions associated with aging. As life expectancy increases, the rising prevalence of dementia in aging populations poses a significant challenge to healthcare systems worldwide. In 2019, an estimated 57.4 million people were living with dementia globally, with projections suggesting this number will rise to 153 million by 2050. ¹ In Germany, approximately 1.8 million individuals were living with dementia in 2021. ²

Pharmacological treatments, including cholinesterase inhibitors (ChEIs) and memantine, are commonly initiated to alleviate symptoms and slow disease progression. These therapies have been available for more than two decades. ³ In Germany, three types of ChEIs are available for the treatment of mild to moderate Alzheimer's disease (AD): donepezil, rivastigmine, and galantamine. Memantine, an anti-glutamatergic agent, is typically prescribed for moderate to severe AD. Anti-dementia medications can be prescribed by general practitioners, neurologists, and psychiatrists. ⁴ While evidence supports the effectiveness of these treatments and professional guidelines recommend their use,^5,6 they must be taken consistently over time to be effective. For instance, Oh et al. demonstrated that a 20% increase in medication possession rate during the first year following a dementia diagnosis was associated with reduced healthcare costs due to fewer hospitalizations and emergency room visits. ⁷ Moreover, early treatment with anti-dementia drugs has been shown to delay institutionalization and reduce mortality.^8,9

Despite the availability of these therapies, maintaining long-term adherence remains a significant challenge. Sistanizad et al. conducted a systematic review and meta-analysis of 68 studies involving 684,493 participants, reporting a mean 12-month persistence rate of 49% (95% CI: 42%–56%). The authors noted that medication adherence in older adults is influenced by a range of factors, including side effect profiles, disease severity, cognitive capacity, social support, prescribing practices, and the complexity of navigating the healthcare system. ¹⁰

Bohlken et al. (2015) reported a 12-month persistence rate of 53% in a study including 12,281 patients treated in private practices. ⁴ However, studies examining long-term persistence using longitudinal prescription data over a period of at least five years are scarce. Furthermore, the influence of variables such as age, sex, treatment type, and physician specialty on therapy persistence remains insufficiently explored. To address this gap, the present study utilizes the IQVIA Longitudinal Prescription Database (LRx), which includes a large and representative sample of patients covered by Germany's statutory health insurance, to assess real-world patterns of antidementia drug use over an extended time frame.

Methods

Database

This retrospective cohort study was based on the IQVIA Longitudinal Prescription Database (LRx), which includes approximately 80% of prescriptions reimbursed by statutory health insurance funds in Germany. ¹¹ The database provides patient-level data, including age and sex, while ensuring full anonymization of all personal information in compliance with data privacy regulations. Each prescription entry includes detailed product information (e.g., brand, active substance, package size, and dosage form) and the date of dispensing. However, the database does not contain clinical diagnoses or laboratory results. Consequently, information on established dementia risk factors such as APOE genotype, education, vascular comorbidities (e.g., hypertension, diabetes, dyslipidemia), smoking status, physical activity, and cognitive scores was not available. The LRx has been widely used in previous research on therapy persistence.^12,13

Study population and outcomes

The study included patients aged 60 years and older who received an initial prescription for anti-dementia drugs (donepezil, rivastigmine, galantamine, or memantine). New users were defined as patients initiating anti-dementia therapy between January 1, 2016, and December 31, 2023, with no dispensing of any anti-dementia drug from January 1, 2015, until the index date (baseline washout period). To ensure inclusion of incident users and minimise prevalent user bias, we required a washout period without anti-dementia drug prescriptions prior to the index date. The LRx database contained records from January 2015 onwards, and the inclusion period for new users was January 2016 to December 2023. Therefore, all patients initiating therapy in 2016 had at least 12 months without any anti-dementia prescriptions before the index date, and patients initiating therapy in subsequent years had proportionally longer washout periods (e.g., ≥ 5 years for those starting in 2020) (Figure 1).

Figure 1.

Study design.

The primary outcomes were the therapy persistence rates over a 12-month and a five-year period following the index date. Each patient was followed for up to 60 months, and therapy discontinuation was defined as a gap of at least 90 days without treatment. If patients switched from one anti-dementia drug to another within 90 days after stopping the previous therapy, they were still considered persistent.

Prescription duration was estimated based on package size, the number of packages dispensed, and the defined daily dose. A sensitivity analysis was conducted using a longer gap threshold of 180 days. Importantly, to be classified as a discontinuation, patients had to remain registered in the database after the therapy-free interval (either 90 or 180 days), thereby ruling out death as a reason for discontinuation.

Statistical analyses

Persistence rates were evaluated using Kaplan–Meier curves, which were generated separately by age group, initial drug (donepezil, rivastigmine, galantamine, memantine), and the specialty of the initiating physician (general practitioners or neuropsychiatrists). A multivariable Cox proportional hazards regression model was applied to estimate the association between variables including age, sex, initial drug, physician specialty, and psychiatric co-therapies (antidepressants, antipsychotics, and benzodiazepines), other co-therapies (antihyperglycemic, antihypertensive, and lipid lowering drugs) and therapy discontinuation. Covariates were assessed based on prescription records in the 12 months preceding the index date.

The regression analysis was also stratified by age group (<70, 70–79, 80–89, ≥ 90 years). Finally, a sensitivity analysis was conducted including only individuals with at least five years of available follow-up time to reflect the possible impact of the death on the persistence.

The proportional hazards assumption for the Cox models was evaluated using log-minus-log survival plots for categorical variables and by testing interactions between covariates and the logarithm of time (Grambsch–Therneau method). For age, violations of the PH assumption were addressed by conducting separate models for each age group. For other covariates, the formal tests indicated statistically significant deviations, likely due to the large sample size; however, visual inspection of the plots suggested that these deviations were minor.

To assess the potential impact of unmeasured confounding, we calculated E-values for selected hazard ratios. The E-value represents the minimum strength of association that an unmeasured confounder would need to have with both the exposure and the outcome, above and beyond the measured covariates, to fully explain the observed association.

Given the large sample size, p-values <0.001 were considered statistically significant, and hazard ratios (HRs) > 1.10 or <0.90 were considered clinically relevant. All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA).

Results

Baseline characteristics of study patients

A total of 567,815 patients (mean age: 80.2 years; 59.1% female) were included in the analysis. Table 1 presents the baseline characteristics of the study population. Neurologists initiated therapy more frequently (65.1%) than general practitioners (34.9%). Donepezil was the most commonly prescribed therapy (38.4%), followed by memantine (31.8%) and rivastigmine (23.0%), while galantamine was the least frequently prescribed agent (6.8%). Antidepressants were prescribed to 29.3% of patients, and antipsychotics to 25.6%, whereas benzodiazepines were rarely prescribed (5.7%).

Table 1.

Basic characteristics of the study patients.

Variable	Number and proportion among study patients a
N	567,815
Age (Mean, SD)	80.8 (6.7)
Age <70 years	37,965 (6.2)
Age 70–79 years	200,575 (32.5)
Age 80–89 years	329,154 (53.3)
Age ≥90 years	49,653 (8.0)
Sex
Female	335,298 (59.1)
Male	232,517 (40.9)
Physician who initiated therapy
Neurologist	401,699 (65.1)
General practitioner	215,648 (34.9)
Initial drug
Galantamine	38,478 (6.8)
Donepezil	218,207 (38.4)
Rivastigmine	130,785 (23.0)
Memantine	180,345 (31.8)
Co-therapy
Antidepressants	166,116 (29.3)
Antipsychotics	145,223 (25.6)
Benzodiazepines	32,268 (5.7)
Antihyperglycemic drugs	111,977 (19.7)
Antihypertensive drug	378,813 (66.7)
Lipid lowering drugs	216,559 (38.1)

^a Proportions of patients given in %, unless otherwise indicated. SD: standard deviation.

Persistence rates

Five years after the index date, 19.8% of patients were still receiving therapy when therapy discontinuation was defined as a gap of at least 90 days. When using a 180-day threshold for discontinuation, 33.9% remained on therapy at the five-year mark (Figure 2). The 12-month persistence rate was 53.1% based on the 90-day definition and 63.9% based on the 180-day definition.

Figure 2.

Kaplan–Meier curves for persistence in patients treated with anti-dementia drugs using 90- and 180 days therapy gap.

Persistence rates increased with age—from 48.7% in patients aged <70 to 57.3% in those aged ≥90 after 12 months, and from 16.6% to 22.5%, respectively, after 60 months (Figure 2). The highest persistence rates were observed in patients who started therapy with memantine (57.6% at 12 months and 32.0% at 60 months). In contrast, the lowest 12-month persistence rate was observed with rivastigmine (50.2%), and the lowest 60-month rate with donepezil (17.8%) (Figure 3). No substantial differences in persistence rates were observed between male and female patients or between those treated by neuropsychiatrists versus general practitioners.

Figure 3.

Kaplan–Meier curves for persistence in patients treated with anti-dementia drugs by age group, sex, initial drug, and initiating physician specialty.

Factors associated with anti-dementia drug discontinuation

Cox regression analyses for the total study population showed that younger age was significantly associated with a higher risk of therapy discontinuation (<70 versus ≥90 years: HR 1.21; 95% CI: 1.19–1.23; 71–80 versus ≥90 years: HR 1.13; 95% CI: 1.11–1.14). Initiating therapy with memantine was associated with a slightly reduced risk of discontinuation compared to donepezil (HR: 0.87; 95% CI: 0.86–0.87). Other variables showed either no associations or only weak associations (HRs between 0.90 and 1.10) (Table 2).

Table 2.

Association between predefined variables and risk of therapy discontinuation (multivariable Cox regression models).

Variable	Total population	Age <70 years	Age 70–79 years	Age 80–89 years	Age ≥90 years
Age <70 years	1.21 (1.19–1.23) *
Age 71–80 years	1.13 (1.11–1.14) *
Age 81–90 years	1.09 (1.07–1.10) *
Age ≥90 years	Reference
Sex: Female	1.04 (1.03–1.04) *	1.05 (1.03–1.08) *	1.05 (1.04–1.06) *	1.03 (1.02–1.04) *	1.02 (0.99–1.05) *
Sex: Male	Reference	Reference	Reference	Reference	Reference
Therapy initiator: Neurologist	1.00 (1.00–1.01)	0.99 (0.96–1.01)	1.00 (0.99–1.01)	1.02 (1.01–1.03) *	0.92 (0.90–0.95) *
Therapy initiator: GP	Reference	Reference	Reference	Reference	Reference
Initial drug: Galantamine	0.95 (0.94–0.97) *	0.97 (0.92–1.02)	0.95 (0.93–0.98) *	0.95 (0.94–0.97) *	0.92 (0.87–0.97) *
Initial drug: Donepezil	Reference	Reference	Reference	Reference	Reference
Initial drug: Rivastigmine	1.05 (1.04–1.06) *	0.99 (0.96–1.02)	1.00 (1.00–1.03)	1.07 (1.06–1.08) *	1.21 (1.17–1.26) *
Initial drug: Memantine	0.87 (0.86–0.87) *	0.95 (0.92–0.98) *	0.88 (0.87–0.89) *	0.85 (0.84–0.86) *	0.87 (0.84–0.90) *
Co-therapy: Antidepressants	1.02 (1.01–1.03) *	1.03 (1.01–1.06)	1.04 (1.03–1.05) *	1.01 (1.00–1.02)	0.99 (0.96–1.02)
Co-therapy: Antipsychotics	0.90 (0.89–0.91) *	0.92 (0.90–0.95) *	0.93 (0.92–0.94) *	0.89 (0.88–0.90) *	0.85 (0.83–0.88) *
Co-therapy: Benzodiazepines	1.10 (1.08–1.11) *	1.12 (1.07–1.18) *	1.10 (1.08–1.13) *	1.09 (1.07–1.11) *	1.06 (1.01–1.12)
Co-therapy: Antihypertensive drugs	0.93 (0.93–0.94) *	0.97 (0.95–1.00)	0.94 (0.93–0.95) *	0.92 (0.91–0.93) *	0.93 (0.90–0.96) *
Co-therapy: Antihyperglycemic drugs	0.98 (0.97–0.99) *	1.07 (1.04–1.10) *	0.98 (0.97–1.00)	0.97 (0.96–0.98) *	0.95 (0.92–0.99)
Co-therapy: Lipid-lowering drugs	0.95 (0.95–0.96) *	0.95 (0.92–0.98) *	0.94 (0.93–0.95) *	0.96 (0.95–0.97) *	0.99 (0.96–1.02)

^*p < 0.001.

In age-stratified analyses, memantine was associated with a lower risk of discontinuation across all age groups except those under 70. Conversely, rivastigmine was associated with an increased risk of discontinuation in patients aged ≥90 years (HR: 1.21; 95% CI: 1.17–1.26). Among elderly patients (aged 80–89 and ≥90), co-therapy with antipsychotics was linked to a lower discontinuation risk (HR: 0.89 and HR: 0.85, respectively). Co-therapy with benzodiazepines was associated with a higher discontinuation risk, particularly in patients aged <70 (HR: 1.10), 70–79 years (HR: 1.12), and 80–89 years (HR: 1.10).

Results of sensitivity analysis were similar. Here, younger age was also significantly associated with a higher risk of therapy discontinuation (<70 versus ≥90 years: HR 1.26; 95% CI: 1.12–1.21; 71–80 versus ≥90 years: HR 1.07; 95% CI: 1.03–1.11), however the association became slightly weaker.

The E-value for the HR of 1.22 for patients aged <70 years versus ≥90 years was 1.57, and for the HR of 0.88 for memantine versus donepezil was 1.43. These values indicate that an unmeasured confounder would need to have at least moderate associations with both the exposure and treatment discontinuation to fully account for the observed effects.

Discussion

This large-scale study of more than half a million patients revealed that approximately 50% discontinued anti-dementia drug therapy within one year of initiation, and 80% discontinued within five years. Cox regression models indicated that younger age (<70 and 70–79 versus ≥90 years) was associated with an increased risk of therapy discontinuation, while initial treatment with memantine and co-therapy with antipsychotics in elderly patients were linked to a decreased risk of discontinuation.

The observed 12-month persistence rate (49.9%) is consistent with previous findings, such as the 52.6% reported in a methodologically similar study involving 66,000 Polish patients treated between 2016 and 2018, ¹⁴ and the 51% persistence rate observed in a Canadian population-based cohort of over 24,000 new users of ChEIs between 2009 and 2013. ¹⁵ It is slightly lower than the 60% found in a study of 12,910 German outpatients with dementia treated between 2003 and 2013. ⁴ However, no previous research has evaluated anti-dementia drug persistence over a five-year period, making direct comparisons challenging.

High discontinuation rates in anti-dementia drug therapy may be attributed to a combination of clinical, patient-related, and systemic factors. First, the symptomatic benefits of drugs such as donepezil, rivastigmine, galantamine, and memantine are often modest and may not be immediately apparent, leading patients or physicians to discontinue treatment. Adverse effects, particularly gastrointestinal symptoms, may also contribute to early discontinuation. ¹⁶ Additionally, medication adherence in dementia largely depends on caregivers. Lack of support, caregiver burden, or insufficient understanding of the treatment's importance can result in poor persistence. ¹⁷

On the other hand, some factors can be reasonably excluded. Anti-dementia drugs typically do not involve complex dosing regimens, as they are most often taken once daily, and rarely twice. ¹⁸ Mortality was not considered a reason for discontinuation in this study due to methodological censoring. As the disease progresses to more severe stages, the perceived value of continued pharmacotherapy may decline.^19-21 This argument does not match with our data since older individuals exhibited higher levels of persistence. Cost is also unlikely to be a limiting factor, since anti-dementia medications are covered by statutory health insurance in Germany. ²² Although dementia impairs memory and cognition, which might be assumed to affect adherence, previous studies suggest otherwise. Dementia was associated with better persistence in antihypertensive therapy ²³ and had no significant impact on persistence with oral antidiabetic medications. ²⁴ These findings may be explained by the close involvement of family members or professional caregivers, who often manage medication intake and ensure adherence. ²⁵

In the current study, multivariable regression analysis revealed no strong or moderate associations with therapy discontinuation, except for younger age. Age < 70 years was associated with a 22% higher risk of discontinuation. ²⁶

The higher discontinuation rates observed in younger age groups compared to the ≥90-year reference group may reflect differences in living arrangements, disease stage, and caregiver involvement. Older individuals are more likely to reside in nursing homes or have dedicated caregivers who can ensure regular medication intake, whereas younger patients may live more independently or rely on less structured support. ²⁶ Moreover, younger patients may present with atypical or rapidly progressive dementia subtypes, in which perceived treatment benefits are limited, prompting earlier discontinuation. Physician prescribing behaviors and treatment goals may also differ across age groups, with a greater focus on symptomatic management in older patients and more selective continuation in younger individuals.

Furthermore, initial therapy with memantine, as compared to rivastigmine, was significantly associated with a lower risk of discontinuation. This may be due to the side effect profiles of ChEIs^16,27 or the fact that memantine is typically prescribed for more advanced dementia, where caregiver support is more intensive. Co-therapy with antipsychotics was also associated with lower discontinuation rates among patients aged 80 and above. A recent study using German claims data found that antipsychotic use was 3.4 times more likely in patients with advanced dementia, ²⁸ reinforcing the idea that more severe disease prompts stronger caregiver involvement and better adherence.

Clinical implications

From a public health perspective, anti-dementia pharmacotherapy remains an essential component of comprehensive dementia care. Strategies to reduce discontinuation should focus on greater integration of caregivers into treatment planning and the use of digital tools, such as medication reminder apps, to support adherence.^29-31 It is also important to provide realistic expectations, educate patients and caregivers about potential side effects, tailor drug selection to individual needs, and monitor adherence regularly, switching medications if necessary.

Strengths and limitations

This study benefits from several strengths, including the use of real-world data from a large, representative patient population and a long follow-up period. However, limitations must also be acknowledged, particularly those related to the nature of the database used. ¹¹ First, the data are derived from pharmacy records and lack diagnostic or clinical information, such as dementia severity or subtype. Second, our analyses could not account for several well-established clinical and lifestyle risk factors for Alzheimer's disease and related dementias, including APOE genotype, education level, and physical activity, as these data are not captured in the LRx database. Although we adjusted for age, sex, physician specialty, initial drug choice, and major co-therapies, residual confounding by unmeasured factors is possible.

Third, although residual confounding from unmeasured factors cannot be ruled out, our E-value analysis suggests that an unmeasured confounder would require a risk ratio of approximately 1.4–1.6 with both the exposure and outcome to explain away the observed associations for age and therapy type. This provides some reassurance regarding the robustness of our findings, although the influence of unmeasured confounding remains possible.

Fourth, the LRx database includes only prescriptions reimbursed by statutory health insurance, which excludes approximately 10% of the German population who are privately insured. Fifth, the reasons for discontinuation (e.g., side effects, institutionalization, or perceived lack of benefit) are not recorded and can only be inferred. Sixth, the database reflects medication dispensing, not actual consumption, though prescription records are widely accepted as a valid proxy for adherence and persistence. Finally, this study did not include a negative control exposure or outcome, which can be a valuable tool in observational pharmacoepidemiology for detecting residual confounding, measurement error, or bias. The LRx database contains only anonymized prescription dispensing records and does not capture non-prescription clinical events, making it challenging to identify a suitable negative control outcome. Similarly, defining a negative control exposure that is truly unrelated to anti-dementia therapy persistence would require additional prescription or clinical variables not available in this dataset. Future studies linking prescription data to richer clinical records may incorporate negative control analyses to further assess the robustness of associations.

Conclusions

The findings of this study highlight low persistence rates for anti-dementia drug therapy in Germany, particularly among younger patients. These results suggest there is significant room for improvement in long-term treatment adherence, especially when considering international data that point to more favorable outcomes with consistent pharmacological management of dementia.