Deep Brain Stimulation for Alzheimer's Disease: A Systematic Review and Meta-Analysis

Alzheimer’s disease (AD) accounts for approximately 60–80% of dementia cases worldwide and is characterized by progressive decline in episodic memory and cognition, with later language, visuospatial, and behavioral disturbances. Neurodegeneration and atrophy in medial temporal lobe structures (hippocampus, parahippocampal cortices, amygdala) and in the cholinergic nucleus basalis of Meynert (NBM), as well as degradation of the fornix, are associated with cognitive impairment and conversion from mild cognitive impairment to AD. Given the concept of AD as a disorder of neural circuits affecting cortical–subcortical networks subserving memory and cognition, neuromodulation with deep brain stimulation (DBS) has been explored as a potential therapy. Prior studies have targeted the fornix, NBM, and ventral capsule/ventral striatum (VC/VS). This systematic review and meta-analysis aimed to synthesize existing clinical evidence on the effects of DBS on cognitive function in AD.

Clinical studies of fornix-DBS have yielded mixed results. Laxton et al. (2010) reported potential slowing of cognitive decline over 6–12 months in six patients with mild AD, with better response in milder disease; Smith et al. (2012) observed increased cerebral glucose metabolism after 1 year correlated with better global outcomes, though ADAS-cog scores rose over time. Sankar et al. (2015), analyzing the same cohort, found increased ADAS-cog and decreased MMSE over 1 year, indicating deterioration. The ADvance trial (Leoutsakos et al., 2018) showed no overall clinical benefit at 2 years, with possible benefit in participants over 65 years. Mao et al. (2018) reported partial early improvements on some cognitive tasks in five patients with severe AD. NBM-DBS results are also inconsistent. Kuhn et al. (2015) found an average 3-point worsening on ADAS-cog at 1 year (considered below clinically significant deterioration), with near-stable MMSE compared to typical pharmacotherapy trajectories. Baldermann et al. (2018) reported 1-year stabilization of MMSE and non-significant improvement on ADAS-mem, but worsening on ADAS-cog, associated with a specific cortical thickness pattern. VC/VS-DBS has been studied in a small nonrandomized cohort (Scharre et al., 2018), with three patients showing slower decline on CDR-SB compared with controls over 18–27 months and acceptable tolerability. Rationales for these targets include: fornix as a major outflow tract within the Papez circuit and a carrier of septal cholinergic fibers to the hippocampus, with integrity predicting memory decline; NBM as a key cholinergic hub with tau pathology and atrophy in AD; VC/VS within frontal-limbic circuits related to motivation, initiative, and executive function, domains impacted in AD. Mechanistic hypotheses for DBS effects include modulation of cholinergic transmission, resetting hippocampal theta rhythms, induction of neurotrophic factors (e.g., NGF), regulation of default mode and Papez circuits, reductions in amyloid and tau pathology via glial modulation and autophagy-lysosomal pathways, decreased neuroinflammation, and effects on GABA/glutamate balance, synaptic plasticity, neurogenesis, and trophic factors.

This systematic review and meta-analysis followed PRISMA 2015 guidelines. Databases searched (from inception to November 28, 2021) included PubMed/Medline, Scopus, Embase, Cochrane Library and Center for Reviews & Dissemination, and Web of Science, using terms related to “Alzheimer’s disease/AD” and “deep brain stimulation/DBS” with the strategy: [TITLE-ABS-KEY (Alzheimer’s AND disease) OR TITLE-ABS-KEY (AD) AND TITLE-ABS-KEY (deep AND brain AND stimulation) OR TITLE-ABS-KEY (DBS)]. References from included studies and relevant reviews were screened. Inclusion criteria: human clinical trials/reports of DBS in AD; sample size ≥2; original, peer-reviewed English articles. Case reports were excluded. Two independent reviewers performed screening and data extraction (study design, sample size, demographics, DBS target, cognitive tests, outcomes, adverse events); discrepancies were resolved by discussion or a third reviewer. The primary outcome was cognitive function measured by MMSE and/or ADAS (including subtests). For studies reporting multiple outcomes within a domain, combined effects were calculated using averaging of effect sizes and appropriate variance estimation. Risk of bias was assessed with the Cochrane Collaboration tool (selection, performance, detection, attrition, reporting biases) by two independent reviewers. Statistical analysis used Comprehensive Meta-Analysis software (v2) and SPSS 26. Fixed-effects models were applied due to low heterogeneity (I2 < 25%). Pre–post comparisons used standardized mean differences (SMD). Heterogeneity was quantified with I2 (25%, 50%, 75% as low, moderate, high). Publication bias was evaluated via funnel plots, trim-and-fill, Begg and Mazumdar rank correlation, and Egger’s regression, acknowledging potential distortions when plotting SMD versus SE. Significance threshold p < 0.05.

- Study selection: Of 524 records, after deduplication and screening, five citations (six comparisons) were included, totaling 66 participants. Some reports were excluded due to overlapping cohorts. - Overall cognitive effect of DBS: Pooled analysis found no significant effect on cognitive function in AD (SMD = 0.116; 95% CI −0.236 to 0.469; p = 0.518). Heterogeneity was low (Q=6.23; T2=0.053; df=5; I2=19.76%; p=0.284). - Fornix-DBS subgroup (three comparisons): No significant cognitive effect (SMD = 0.145; 95% CI −0.246 to 0.537; p = 0.467). Heterogeneity not significant (Q=0.685; T2=0.000; df=3; I2=0.0%; p=0.877). - NBM-DBS (one comparison): Non-significant effect (SMD = −0.431; 95% CI −1.318 to 0.465; p = 0.341). - VC/VS-DBS: Only one small nonrandomized study (n=3) reported slower decline on CDR-SB versus controls over 18–27 months with good tolerability. - Adverse events: Reported neurological AEs included altered mental status (n=2), seizures/possible seizures (n=2), agitation (n=1), mild IPG-site pain, incision-site headache, transient visual neglect post-surgery, and depression (frequency not always specified). Non-neurological/stimulation-related autonomic and cardiovascular AEs (warmth, sweating, flushing, increased heart rate and blood pressure) occurred at higher voltages (>7 V) and were mitigated by programming at ~50% of side-effect thresholds. Device-related issues included malfunctioning plug-in connectors requiring surgical revision; inner restlessness at >5 V in one patient. - Publication bias: Funnel plot reasonably symmetric; Begg and Mazumdar p=0.452; Egger’s p=0.407; trim-and-fill found no missing studies. - Study quality: Many studies were non-randomized and unblinded with inadequate allocation concealment, indicating risks of selection, performance, and detection bias.

The meta-analysis did not detect a significant overall benefit of DBS on cognitive outcomes in AD, addressing the primary question by suggesting that, based on current small and heterogeneous studies, DBS has not yet demonstrated efficacy for global cognitive improvement. The lack of effect likely reflects heterogeneity in targets (fornix, NBM, VC/VS), stimulation parameters (frequency, voltage, pulse width), disease stage at implantation, limited sample sizes, and study designs (many pre–post, unblinded). Some individual or subgroup signals—such as potential benefits in older patients (>65 years) in the ADvance trial or slowed CDR-SB decline with VC/VS-DBS—indicate areas worth further exploration. Mechanistic rationales (network modulation, cholinergic enhancement, neurotrophic effects, and potential amyloid/tau modulation) remain compelling, but require confirmation in rigorous trials. Future research should standardize stimulation parameters, carefully select disease stage, and adopt randomized, double-blind, sham-controlled designs to clarify efficacy and identify responders.

DBS for AD is a promising but as yet unproven therapeutic approach. This meta-analysis highlights inconsistent findings, small sample sizes, and methodological limitations across studies, with no significant pooled cognitive benefit observed. Future investigations should employ larger, high-quality randomized controlled trials with standardized targets and stimulation parameters and consider patient selection (e.g., disease stage, age) to determine whether DBS can meaningfully modify cognitive decline in AD.

- Small number of included studies and overall participants. - Predominance of pre–post, non-randomized, and unblinded designs, increasing risk of bias and potential overestimation of effects. - Heterogeneity in stimulation parameters (frequency, pulse width, voltage), targets (fornix, NBM, VC/VS), laterality, start time, and duration across studies. - Potential publication bias cannot be fully excluded despite negative tests; SMD-based funnel plots may be misleading. - Overlapping cohorts in some publications limited independent comparisons.