Only Half of Dementia Patients Stay on Their Medications After One Year

Dementia affects approximately 55 million people worldwide, with costs projected to reach $2 trillion annually by 2030. The mainstay pharmacological treatments — cholinesterase inhibitors (ChEIs: donepezil, rivastigmine, galantamine) and the NMDA antagonist memantine — are prescribed to slow cognitive decline and preserve daily functioning, though they are not disease-modifying. Despite their clinical importance, real-world persistence with these medications has been poorly characterized at scale. This meta-analysis, registered with PROSPERO (CRD42022361744), represents the most comprehensive quantitative synthesis to date, covering nearly three decades of observational data.

Researchers searched Medline, Embase, PsycINFO, and CINAHL from January 1995 to February 2024, ultimately including 68 observational studies involving 684,493 patients aged 50 and older. All studies reported real-world persistence — defined as whether a patient was still taking their medication after a predefined period — using pharmacy claims, administrative healthcare databases, or medical records. Persistence was treated as a dichotomous outcome. The primary endpoint was 12-month persistence rate, analyzed using random-effects Mantel–Haenszel models to account for anticipated heterogeneity across studies.

The pooled 12-month persistence rate was 49% (95% CI: 42%–56%), meaning roughly half of all dementia patients had discontinued their anti-dementia medication within one year of initiation. Substantial heterogeneity was observed (I² was high across subgroups), driven significantly by how persistence was defined. Studies that did not require a permissible refill gap — typically those using medical records rather than pharmacy claims — reported markedly higher persistence at 67% (95% CI: 38%–90%), compared to studies using stricter gap-based definitions. This methodological variable was the only independent predictor of between-study heterogeneity in multivariate meta-regression (β = 0.36, 95% CI: 0.18–0.54).

Subgroup analyses revealed meaningful variation by drug class, geography, and publication era. Memantine showed higher pooled persistence at 61% (95% CI: 38%–82%) compared to ChEIs collectively. European studies reported 57% persistence (95% CI: 43%–71%), higher than other regions. Studies published between 2011 and 2015 showed the highest persistence at 54% (95% CI: 41%–68%). Sensitivity analyses using the leave-one-out method confirmed the robustness of the overall estimate. Funnel plot asymmetry and Egger's/Begg's tests were used to assess publication bias, though the authors note this remains a potential concern given the observational nature of included studies.

The clinical implications are substantial. Low persistence means many patients are not receiving the sustained pharmacological benefit these drugs are designed to provide — including slowed cognitive decline, preserved daily functioning, and reduced caregiver burden. The authors emphasize that 'critical junctures' — early in therapy, after hospitalization, or following adverse events — represent key windows for clinical intervention. Targeted adherence support programs, caregiver education, and simplified dosing regimens could meaningfully improve persistence rates. The authors also call for a standardized framework for measuring persistence in future research, as the current variability in permissible gap definitions makes cross-study comparisons unreliable and may be masking the true scale of the problem.