Timing and Dose of Acupuncture Linked to Cognitive Recovery After Stroke, Retrospective Study Finds

Among 226 patients recovering from ischemic stroke who received acupuncture as part of a standardized inpatient rehabilitation program in China, roughly two-thirds showed a meaningful gain on a widely used cognitive screening test, and a new retrospective analysis has identified which patients were most likely to be in that group. The study, published in Frontiers in Neurology, found that younger age, more years of education, milder neurological deficits at admission, absence of diabetes, earlier initiation of acupuncture, higher weekly session frequency, and a longer total treatment course were each associated with a greater likelihood of improving by at least two points on the Montreal Cognitive Assessment (MoCA). The research team, based in the Encephalopathy Department of Shaanxi Provincial Hospital of Traditional Chinese Medicine in Xi'an, reviewed electronic medical records of ischemic stroke patients hospitalized between January 2023 and December 2025.

After applying the criteria, 226 patients formed the analytical cohort. The investigators split the cohort into an improvement group of 148 patients (65.49 percent) whose MoCA score rose by two or more points between admission and discharge, and a non-improvement group of 78 patients (34.51 percent) whose gain fell short of that threshold. The two-point cut-off was drawn from published estimates of the minimal clinically important difference for the MoCA, but the authors stress that it served as an operational stratification variable for this retrospective analysis rather than a validated definition of cognitive recovery.

Measurement error, practice effects from repeated testing, and regression to the mean can all influence such change scores. At baseline, the two groups were closely matched: mean MoCA scores were 19.03 in the improvement group and 19.24 in the non-improvement group, a difference that was not statistically significant. By the end of treatment, however, the trajectories had diverged.

The improvement group's mean MoCA rose to 21.44, while the non-improvement group edged up to 19.85, a difference that was not statistically significant within that group. The between-group difference in post-treatment MoCA was 1.59 points (95 percent confidence interval 0.82 to 2.36). A similar pattern emerged on the Mini-Mental State Examination, where the post-treatment between-group difference reached 2.44 points (95 percent CI 1.69 to 3.19).

Secondary measures told a consistent story. The improvement group showed significant gains on the Loewenstein Occupational Therapy Cognitive Assessment, the Modified Barthel Index of daily living activities, and significant reductions on the Hamilton depression and anxiety rating scales and the Pittsburgh Sleep Quality Index. Notably, even patients in the non-improvement group, who failed to reach the cognitive threshold, still recorded significant improvements in mood, sleep, and daily functioning.

The authors interpret this as evidence that different recovery domains follow distinct trajectories, and that patients can derive meaningful functional benefits from comprehensive rehabilitation even when measurable cognitive gains are absent. To understand what separated the two groups, the team first ran univariate logistic regressions, then built a multivariable model that included clinically relevant variables together with two temporal covariates—time from stroke onset to baseline assessment and the interval between baseline and final assessment—forced into the model to adjust for potential confounding. Because the outcome event was defined as non-improvement, odds ratios above one indicate higher odds of falling short of the improvement threshold, while values below one indicate lower odds.

The adjusted results pointed to four characteristics associated with a lower probability of improvement: each additional year of age (odds ratio 1.208), each one-point increase in the admission National Institutes of Health Stroke Scale score (OR 1.643), each additional day of delay before acupuncture began (OR 1.162), and a diabetes diagnosis (OR 4.737). Conversely, each additional year of education (OR 0.740), each additional weekly acupuncture session (OR 0.271), and each additional week of total treatment course (OR 0.248) were associated with a higher probability of meeting the improvement threshold. Baseline MoCA score, the time from stroke onset to baseline assessment, and the assessment interval were not significantly associated with the outcome after adjustment.

No serious acupuncture-related adverse events, such as pneumothorax, significant hemorrhage, or infection, were recorded in the medical records, though minor, self-limiting events like transient bruising and mild needling pain appeared informally in nursing notes. The authors are careful to flag a critical caveat concerning the treatment-intensity findings. Unlike age, education, or admission stroke severity, treatment frequency and total course were not fixed at baseline; they unfolded over the hospitalization.

Patients who were clinically stable, adherent, or showing early signs of improvement may simply have been more likely to receive more frequent or longer courses, while those with complications or deteriorating status may have had abbreviated regimens. This confounding by indication and possible reverse causation means the strong odds ratios for frequency and duration should be read as exploratory and hypothesis-generating, not as evidence that more acupuncture causes better cognition. Other limitations compound that caution.

The study was single-center and retrospective, with no non-acupuncture or sham-acupuncture control group, so it cannot establish any acupuncture-specific effect; every patient also received pharmacological management, cognitive training, occupational therapy, exercise therapy, and physical therapy, and the records did not disaggregate the cognitive-training component that is most relevant to MoCA outcomes. Twelve patients who interrupted treatment were excluded, potentially introducing attrition bias, and pre-stroke cognitive function was not prospectively measured. With only 78 non-improvement events, the model's events-per-variable ratio was about 7.8 to 1, slightly below the conventional 10 to 1 threshold, which constrained how many covariates could be included.

No internal validation or discrimination metrics were computed, and the model was built for explanatory rather than predictive purposes. Within those constraints, the biological reasoning the authors offer is grounded in established stroke literature. Aging is associated with reduced neuroplasticity, diminished cerebrovascular reserve, and lower expression of brain-derived neurotrophic factor, all of which could narrow the window for recovery.

Education is thought to build cognitive reserve—denser, more redundant neural networks that buffer against ischemic damage—while also correlating with better health literacy and adherence. Higher admission NIHSS scores reflect larger infarcts or damage to cognition-critical regions such as the thalamus and frontotemporal cortex. Diabetes, which showed the strongest association in the model, may impair recovery through advanced glycation end products, oxidative stress, endothelial dysfunction, microvascular injury, and disrupted insulin signaling in the hippocampus.

The timing finding aligns with the view that the acute and early subacute phases after stroke are periods of heightened neuroplasticity, when neurogenesis, synaptic remodeling, and network reorganization are most active. Earlier intervention, on this reasoning, may harness endogenous repair mechanisms more effectively. The dose-related findings echo a meta-analysis of electroacupuncture cited by the authors, in which subgroup heterogeneity in MoCA outcomes was high among four-week trials but nearly absent among eight-week trials, suggesting treatment duration as a source of variability in the literature.

For now, the authors conclude, these results should not guide clinical practice or justify specific acupuncture protocols. Their value lies in sharpening the questions for the next generation of studies: prospective, multicenter randomized controlled trials with fixed-dose acupuncture arms, component-level documentation of concurrent rehabilitation, and follow-up extending beyond discharge. Whether the frequency and duration effects plateau at some point, and whether older patients over 75 would respond differently than the relatively young cohort studied here—mean ages were 58 in the improvement group and 63 in the non-improvement group—remain open questions that only controlled experiments can answer.

Subject of Research: Factors associated with cognitive improvement in post-ischemic stroke patients receiving acupuncture within multimodal inpatient rehabilitation

Article Title: Determinants of MoCA improvement in post-ischemic stroke patients receiving acupuncture: a retrospective study

Article References: Long, H., & Wu, P. (2026). Determinants of MoCA improvement in post-ischemic stroke patients receiving acupuncture: a retrospective study. Frontiers in Neurology, 17, Article 1892576. https://doi.org/10.3389/fneur.2026.1892576

Image Credits: AI Generated

DOI: 10.3389/fneur.2026.1892576

Keywords: acupuncture, post-stroke cognitive impairment, ischemic stroke, MoCA, retrospective cohort study, neurorehabilitation, diabetes, cognitive reserve, NIHSS, treatment frequency, stroke recovery, logistic regression