Probiotics Show Glimmers of Cognitive Benefit in Dementia Review, but the Evidence Base Is Thin

A systematic review published in the journal Discover Neuroscience has examined whether probiotic dietary interventions can ease the symptoms associated with dementia in adults aged 50 and older, and its verdict is cautiously optimistic but heavily qualified. The review, conducted under PRISMA 2020 guidelines and registered in PROSPERO, identified only six randomized clinical trials, comprising 579 randomized participants in total, that met the authors’ strict eligibility criteria. Within that small pool of evidence, probiotic supplementation consistently yielded cognitive improvements, particularly in individuals who already had deficits such as mild cognitive impairment or Alzheimer’s disease. Yet the same review also exposed striking gaps: no eligible study investigated neuropsychiatric symptoms or functional dependency at all, and emotional outcomes were measured in just a single trial.

The review team, led by Nerea Carrión and Cristian Cabrera with senior investigators Luis Heredia and Margarita Torrente, searched PubMed, Scopus, Web of Science and the Cochrane Library, with preliminary searches beginning in February 2024 and repeated in February 2025 and March 2026. No restrictions on publication dates or languages were applied. Eligible studies had to be randomized trials in participants aged 50 or older, testing probiotics delivered in capsules, sachets or daily dietary intake, with no restrictions on probiotic type. The authors deliberately excluded crossover designs, arguing that because probiotic interventions can induce long-lasting changes in the gut microbiome and the gut-brain axis, standard washout periods of less than a month are often insufficient to restore baseline cognitive and emotional parameters, biasing the results of a subsequent treatment period.

One of the review’s most distinctive methodological choices was its insistence on rigorous statistical analysis. Studies were included only if they reported efficacy data using an intention-to-treat (ITT) approach or a Full Analysis Set (FAS), and crucially, inclusion was based on how the analyses were actually conducted rather than what the authors claimed. If a study stated it used ITT but the full text revealed inappropriate exclusions or reliance on per-protocol methods, it was excluded. This filter proved decisive: during full-text screening, 40 articles were excluded, and 32 of those were rejected solely for not implementing an ITT or FAS analysis. The ITT principle preserves the benefits of randomization by analyzing all participants in the groups to which they were originally assigned, regardless of adherence or protocol deviations, and typically provides more conservative but unbiased efficacy estimates than per-protocol analyses.

The six included trials were conducted between 2016 and 2023 across Asia, in China, Japan, Korea and Iran, with individual sample sizes ranging from 60 to 128 participants and 30 to 65 participants per allocation group. Mean participant ages ranged from approximately 60.9 to 82.0 years. Intervention duration varied enormously, from just 17 days to 24 weeks, and only one study included a two-week pre-intervention baseline period in which participants avoided probiotics and probiotic-containing supplements. The probiotic formulations were strikingly heterogeneous: strains came predominantly from the genera Lacticaseibacillus, including L. paracasei, L. acidophilus, L. casei, L. fermentum and L. plantarum C29, and Bifidobacterium, including B. breve MCC1274, B. longum, B. bifidum and B. breve A1. Every included study used a distinct formulation or strain combination, with estimated daily intakes generally ranging from more than 10 million to more than 10 billion colony-forming units, delivered as beverages, sachets, capsules or probiotic milk.

All six trials evaluated cognitive outcomes, using a wide array of instruments including the Mini-Mental State Examination, the Montreal Cognitive Assessment, the Repeatable Battery for the Assessment of Neuropsychological Status, and the Japanese version of the Alzheimer’s Disease Assessment Scale cognitive subscale, among many others. Three studies targeted older adults with mild cognitive impairment, and all three reported some cognitive benefits. Asaoka and colleagues found no significant between-group differences in overall ADAS-Jcog scores but observed significant improvements in orientation in the probiotic group at week 24, analyzed under a FAS approach with no imputation for missing data. Hwang and colleagues reported greater cognitive improvements in the probiotic group, particularly in attention and prefrontal function, though not in working and verbal memory, using an ITT analysis with a mixed-effects model and maximum likelihood estimation for missing data. Xiao and colleagues observed significant improvement in overall cognitive performance on RBANS total scores, with specific gains in immediate memory, visuospatial and constructional abilities, and delayed memory, but no significant effects in language and attention domains.

The remaining trials extended the picture in different directions. One study of patients with advanced Alzheimer’s disease reported significant cognitive benefits: after a 12-week intervention, MMSE scores increased significantly in the probiotic group while declining in the control group, though missing data were imputed using the last observation carried forward (LOCF) method. A study of individuals with Parkinson’s disease found no significant cognitive improvements, in a trial that suffered a high dropout rate and used ITT analysis with LOCF imputation. Finally, a study of postoperative patients undergoing hip or knee replacement reported no overall group differences, but when participants were stratified post hoc by cognitive decline severity, the incidence of decline in Hopkins Verbal Learning Test-Revised scores and sub-tests was significantly lower in the probiotic group. The review authors flag this kind of retrospective subgrouping as a statistical hazard, since slicing data after seeing non-significant overall results directly increases the risk of a Type I error unless such analyses are pre-specified and adequately controlled.

Emotional outcomes were the province of a single trial. In the Parkinson’s disease study, emotional symptoms were assessed with the Hamilton Depression Scale, the Hamilton Anxiety Scale and the 39-item Parkinson’s Disease Questionnaire, and the authors reported sustained emotional benefits in the probiotic group, including significantly reduced scores on all three measures. Notably, unexpected improvements in anxiety scores were also observed in the placebo group. The review authors describe an intriguing paradox here: no cognitive gains in this population, yet substantial emotional benefits, which cautiously suggests that certain strains, such as Lactobacillus paracasei Shirota, could potentially influence mood-regulating pathways independently of cognitive changes. They are careful to label this interpretation highly speculative and in need of further empirical validation, particularly given the trial’s high dropout rate and the carrying forward of data from 11 placebo participants from week 10 to week 12 due to supply issues.

Beneath the clinical findings, the review mounts a pointed critique of how missing data are handled in this literature. Although all included trials used ITT or FAS principles, many relied on less robust imputation methods such as LOCF, which substitutes missing values with the last available observation. The authors argue this is an unrealistic assumption in progressive conditions like dementia: if a patient in the placebo group drops out, carrying their last observation forward artificially halts their natural cognitive decline, which can falsely minimize the difference between placebo and probiotic groups and underestimate the true treatment effect. Conversely, trials that apply no imputation simply analyze the remaining participants, breaking the initial randomization and frequently inflating the apparent treatment effect, since participants who drop out typically present worse outcomes. Only the Hwang trial employed a robust approach, using maximum likelihood estimation to account for systematic differences between participants with complete and incomplete data. The reviewers also rated studies that failed to adjust for multiple comparisons as high risk under the fifth domain of the Cochrane RoB 2 tool, since conducting many statistical tests simultaneously inflates the probability of a chance significant result.

On safety, the picture is more consistent. The included trials generally reported a favorable profile for probiotic supplementation: several studies reported a complete absence of adverse events or side effects related to administration, the Parkinson’s trial observed no significant differences in adverse events between probiotic and placebo groups, and the Hwang trial strictly monitored safety throughout without highlighting significant untoward medical occurrences. One trial allowed concomitant clinical treatments but did not report specific adverse event outcomes. The review’s broader conclusion is that probiotics could be a promising adjunctive dietary intervention for cognitive decline, presenting a possible window of opportunity during the early and prodromal stages of neurodegeneration, with preliminary patterns hinting at strain-specific benefits across different clinical domains. But the authors are equally clear about what would be needed to turn these hints into clinical guidance: high-quality trials with rigorous missing data handling, head-to-head comparisons of specific bacterial strains, and a broadened focus on neuropsychiatric and functional outcomes, incorporating standardized instruments such as the Neuropsychiatric Inventory and the Alzheimer’s Disease Cooperative Study Activities of Daily Living scale, to fully establish the therapeutic value of probiotics in dementia care.

Subject of Research: A systematic review of randomized trials testing probiotic dietary interventions for dementia-related cognitive and emotional symptoms in adults aged 50 and older.

Article Title: Systematic review of probiotic dietary interventions for symptoms associated with dementia in older adults

Article References: Carrión, N., Cabrera, C., Heredia, L., & Torrente, M. (2026). Systematic review of probiotic dietary interventions for symptoms associated with dementia in older adults. Discover Neuroscience, 21(1), Article 107. https://doi.org/10.1186/s13064-026-00334-w

Image Credits: AI Generated

DOI: 10.1186/s13064-026-00334-w

Keywords: probiotics, dementia, mild cognitive impairment, Alzheimer's disease, gut-brain axis, systematic review, randomized controlled trials, intention-to-treat, missing data, Parkinson's disease, older adults, microbiota