A systematic review and meta-analysis of the impact of transcranial direct current stimulation on cognitive function in older adults with cognitive impairments: the influence of dosage parameters

Cognitive decline among older adults is a growing public health concern, linked to age-related brain changes such as synaptic loss and neuronal dysfunction, leading to memory and learning deficits, reduced attention, and slower processing speed. Common etiologies include Alzheimer’s disease (AD), vascular and Lewy body dementias, mild cognitive impairment (MCI), and frontotemporal degeneration. Pharmacological treatments (e.g., AChEIs, memantine, levetiracetam) may attenuate decline but can have adverse effects and increase polypharmacy risks. Nonpharmacological strategies, particularly non-invasive brain stimulation like transcranial direct current stimulation (tDCS), have gained attention for cognitive support with a favorable safety profile (mild transient side effects). tDCS delivers low-intensity direct current (0.5–2 mA) via scalp electrodes to modulate cortical excitability and synaptic plasticity. However, tDCS effects depend on parameters (intensity, duration, electrode montage/size, sessions) and may be used alone or adjunctive to training. Prior reviews suggest potential benefits but are limited by few studies, insufficient long-term data, and incomplete parameter evaluation. This study systematically reviews and meta-analyzes RCTs to determine tDCS effects on cognition in older adults with cognitive impairments and to identify optimal dosage/parameter configurations for clinical application.

Previous systematic reviews and meta-analyses reported that tDCS can improve global cognition in MCI and mild-to-moderate AD, and possibly slow cognitive decline, especially when combined with aerobic exercise. However, these reviews noted limitations: small numbers of studies, inadequate assessment of long-term effects, limited exploration of tDCS parameters (e.g., intensity, duration, montage, session frequency), and insufficient subgroup analyses for combinations with cognitive training. Recent reviews emphasized the need to investigate dosage and montage variables and to expand sample sizes for better generalizability. Addressing these gaps, the present review updates evidence through November 2023, examines parameter influences, and compares tDCS alone versus tDCS combined with training.

Protocol and reporting followed Cochrane MECIR and PRISMA 2020, prospectively registered on PROSPERO (CRD42023418267; 09/03/2023). Literature search: PubMed (2003–2023), Scopus (1994–2023), EMBASE (2017–2023), Cochrane Library (1995–2023), and EBSCO (1954–2023), for English-language publications up to November 7, 2023, using population, impairment, intervention, and outcome terms with MeSH and Boolean operators; reference lists of relevant articles were screened. Inclusion criteria (PICOS): - Population: older adults (≥60 years or mean age ≥65) with cognitive impairments (AD, MCI, dementia), diagnosed per DSM-IV/V, NINCDS-ADRDA, Petersen criteria, or cognitive assessments (MoCA, CDR). - Intervention: active tDCS alone or combined with additional interventions; unilateral (anodal/cathodal), bilateral/dual/bihemispheric configurations acceptable. - Comparator: sham tDCS (alone or combined with additional interventions). - Outcomes: at least one objective cognitive measure. - Study design: randomized controlled trials (parallel or crossover), English language. Exclusions: non-RCTs (e.g., case studies, protocols, single-group pre-post), non-English, abstracts/posters without full text, clinical guidelines, and studies with other neurological/psychiatric conditions (e.g., Parkinson’s, stroke, MS, depression). Screening and data extraction: Two independent reviewers screened titles/abstracts in Covidence, then full texts; discrepancies resolved by a third author. Data extracted: study characteristics (authors, year, geography), sample size and participant demographics (age, sex, diagnosis, education, disease duration), intervention parameters (program, montage, electrode size, intensity, duration, sessions, target area, follow-up), outcome measures, and effect estimates (group sizes, means/SDs or CIs). Authors were contacted for missing data where needed. Quality assessment: RoB 2 tool used by two independent reviewers to rate bias domains; overall judgments as low risk, some concerns, or high risk. Statistical analysis: RevMan 5.4 used for meta-analyses. Continuous outcomes pooled with mean differences or standardized mean differences (SMD) depending on measures. Fixed-effects model used when heterogeneity was low (p>0.05 and I²<50%); random-effects when heterogeneity was high (p<0.05 or I²≥50%). Heterogeneity explored via subgroup analyses (diagnosis, outcome instrument, tDCS configuration: combination vs alone; current density ≤/>0.06 mA/cm²; session duration 20 vs >20 min; sessions ≤/>15; total charge density </>0.50 mAh/cm²; target region temporal vs left DLPFC; montage extracephalic vs intracephalic). Sensitivity analyses assessed robustness by removing studies at high risk (e.g., selective reporting). Publication bias assessed via funnel plots, Egger’s test, and trim-and-fill.

- Study pool: 19 RCTs (17 parallel, 2 crossover), 945 participants (mean age 71.7±5.9 years), diagnoses: AD (8 studies), MCI (6), vascular dementia (1), dementia (1), executive dysfunction (1), mixed AD/MCI/AD-VD (2). - Global cognition (immediate post-intervention): overall significant improvement (SMD=0.45; 95% CI 0.10–0.80; p=0.01; I²=80%). Instrument subgroups: MMSE significant (SMD=0.51; 95% CI 0.15–0.86; p=0.005; I²=24%); ADAS-Cog non-significant (SMD=0.70; 95% CI −0.31–1.71; p=0.17; I²=90%); MoCA non-significant (SMD=0.28; 95% CI −0.28–0.84; p=0.32; I²=83%). - Global cognition (follow-up 1–8 weeks): overall significant (SMD=0.91; 95% CI 0.35–1.38; p=0.001; I²=90%). Instrument subgroups: MMSE significant (SMD=2.29; 95% CI 1.04–3.55; p=0.0003; I²=94%); ADAS-Cog non-significant (SMD=−0.06; 95% CI −0.30–0.17; p=0.60; I²=0%); MoCA non-significant (SMD=0.04; 95% CI −0.29–0.49; p=0.61; I²=0%). - Diagnosis subgroup (MMSE-based): significant in AD (SMD=0.91; 95% CI 0.13–1.70; p=0.02; I²=87%); non-significant in MCI (SMD=−0.07; 95% CI −0.34–0.20; p=0.62; I²=23%) and dementia (SMD=0.26; 95% CI −0.23–0.75; p=0.30; I²=0%). - Memory function: immediate effects non-significant overall (SMD=0.01; 95% CI −0.17–0.20; p=0.89; I²=30%); verbal memory non-significant (SMD=0.14; 95% CI −0.10–0.38; p=0.25; I²=0%); working memory non-significant (SMD=−0.17; 95% CI −0.46–0.12; p=0.24; I²=50%). Follow-up working memory non-significant (SMD=−0.01; 95% CI −0.25–0.22; p=0.92; I²=0%). - Executive function: immediate effects non-significant (Stroop color SMD=0.44; 95% CI −1.71–2.58; p=0.69; I²=0%; Stroop word SMD=0.03; 95% CI −5.44–5.50; p=0.99; I²=69%; Stroop color-word SMD=3.07; 95% CI −0.35–6.49; p=0.08; I²=0%). - Parameter influences on MMSE (immediate): tDCS alone significant (SMD=0.39; 95% CI 0.11–0.67; p=0.006; I²=7%); tDCS + training not significant (SMD=0.36; 95% CI −0.25–0.97; p=0.25; I²=88%). Current density ≤0.06 mA/cm² significant (SMD=0.25; 95% CI 0.02–0.49; p=0.04; I²=13%); >0.06 mA/cm² not significant. Session duration >20 min significant (SMD=0.89; 95% CI 0.18–1.60; p=0.01; I²=80%); 20 min not significant. Sessions ≤15 significant (SMD=0.28; 95% CI 0.07–0.50; p=0.009; I²=5%); >15 not significant. Total charge density <0.50 mAh/cm² significant (SMD=0.28; 95% CI 0.07–0.50; p=0.009; I²=5%); >0.50 mAh/cm² not significant. Target region: temporal areas significant (SMD=0.33; 95% CI 0.06–0.61; p=0.02; I²=69%); left DLPFC not significant (SMD=0.24; 95% CI −0.04–0.51; p=0.09; I²=0%). Montage: extracephalic vs intracephalic both non-significant. - Safety: Minor transient adverse effects reported (tingling, scalp burning/redness, headache, sleepiness), with good tolerability across protocols (up to 2 mA, 30–40 min, ≤36 sessions). - Publication bias: Egger’s test indicated asymmetry (p=0.026); trim-and-fill imputed 5 studies, yielding adjusted effect size 0.703 (95% CI 0.306–1.100). - Sensitivity: Removing studies with high risk of selective reporting (n=4) did not alter pooled conclusions substantially.

The review demonstrates that tDCS can enhance global cognition in older adults with cognitive impairments, especially in AD, with effects most consistently captured by MMSE both immediately and at short-term follow-up. The lack of significant changes in ADAS-Cog and MoCA suggests instrument-specific sensitivity differences, with MMSE potentially more responsive to broad cognitive shifts induced by tDCS. Absence of benefits in memory and executive functions indicates that general cognitive improvements may not translate to domain-specific gains under the tested protocols, or that tasks used were insufficiently sensitive or underpowered. Parameter subgroup findings highlight actionable guidance: tDCS alone, lower current density (≤0.06 mA/cm²), longer single-session duration (>20 min), limited total sessions (≤15), lower total charge density (<0.50 mAh/cm²), and temporal lobe targeting were associated with better MMSE outcomes; montage location of the reference electrode did not yield clear differences. The differential efficacy by diagnosis (benefit in AD, not MCI/dementia) may reflect greater room for improvement in more severe impairment or disease-specific neurophysiology, and uneven sample sizes. Despite heterogeneity, sensitivity analyses and random-effects modeling support robustness. Potential publication bias suggests observed effects may be conservative. Clinically, tDCS appears safe and tolerable. Future work should refine protocols, align cognitive measures with targeted domains, and expand long-term follow-up.

tDCS can significantly improve global cognitive performance in older adults with cognitive impairments, with strongest evidence in AD and when assessed by MMSE. Optimal conditions identified include current density ≤0.06 mA/cm², session duration >20 minutes, ≤15 sessions and total charge density <0.50 mAh/cm², and targeting temporal areas; tDCS alone outperformed combinations with training in this synthesis. No significant benefits were found for memory or executive function. Further high-quality RCTs should standardize tDCS parameters, evaluate domain-specific outcomes with sensitive instruments, include diverse severities and etiologies, and assess durable long-term effects.

- Mixed severities/types of cognitive impairment were synthesized without stratification by severity or MCI subtype due to insufficient studies. - English-only inclusion introduces language bias. - Limited sample sizes for some diagnoses (especially dementia) and outcomes reduce power and generalizability. - High heterogeneity across studies in design, populations, and tDCS protocols; although addressed with random-effects and subgroup analyses, precision is limited. - Limited number of studies with long-term follow-up for global cognition and even fewer for memory/executive outcomes. - Evidence of publication bias (Egger’s test p=0.026). - Some studies exhibited high risk of bias (e.g., selective reporting), though sensitivity analyses suggested robustness.