Effects of exercise interventions in Alzheimer's disease: A meta‐analysis

Abstract Objective The aim of this study was to investigate the clinical efficacy of exercise intervention in the treatment of patients with Alzheimer's disease (AD) by meta‐analysis. Methods From January 2000 to January 2022, PubMed, Web of Science, Embase, CNKI, and WanFang databases were searched for all studies on the clinical efficacy of exercise intervention in the treatment of AD patients. Stata 17.0 statistical software was used for meta‐analysis. Results Specifically, data of 983 patients were subjected to meta‐analysis, including 463 patients in the control group (conventional drug therapy) and 520 patients in the treatment group (physical exercise on the basis of conventional therapy). The results of meta‐analysis showed that Mini‐Mental State Examination (MMSE) score and Activities of Daily Living Scale (ADL) score in the treatment group were significantly higher than those in the control group. Further subgroup analysis of exercise intervention >16 weeks found that MMSE and ADL scores in the treatment group were significantly higher than those in the control group. Subgroup analysis of exercise intervention ≤16 weeks demonstrated that MMSE and ADL in the treatment group were higher than those in the control group. In addition, the treatment group had a significant lower Neuropsychiatric Inventory (NPI) score compared with the control group (SMD = –0.76, 95% CI (–1.37, –0.16), p = .013); subgroup analysis showed that the NPI score in the treatment group were lower than that in the control group when exercise intervention was >16 weeks [SMD = –1.01, 95% CI (–1.99, –0.04), p = .042] and ≤16 weeks [SMD = 0.43, 95% CI (–0.82, –0.03), p = .034]. Conclusion Exercise intervention can improve the neuropsychiatric symptoms, activities of daily living and cognitive function of AD patients, but the improvement is not significant in case of exercise intervention ≤16 weeks.


Inclusion criteria
(1) Study subjects: The diagnosis was in accordance with the diagnostic criteria of AD introduced in Clinical Diagnosis and Efficacy Criteria (Chinese Edition) (Wang, 2010). informed about the study and signed the informed consent.

Exclusion criteria
(1) Subjects with severe critical limb ischemia, severe orthopedic diseases, severe spinal cord injury, severe muscle injury, malignant tumors, lifestyle-limiting claudication or mental disorders caused by congenital genetic diseases.
(3) Articles from which data required for this study could not be obtained were excluded.

Literature screening and data extraction
The relevant articles obtained from the search were imported into NoteExpress5.4 software, and the repeated ones were automatically eliminated by computer, and then manually eliminated by the researchers. Two researchers independently screened the literature and determined the final included literature. Then they extracted the required data by reading the titles and full text in strict accordance with the inclusion and exclusion criteria. In case of disagreement between the two researchers, a third evaluator was consulted to reach a consensus. Relevant data mainly included literature title, author, study subjects, intervention measures, and relevant scales.

Statistical analysis
Stata16.0 statistical software was used for meta-analysis. Standardized mean difference (SMD) and 95% confidence interval (CI) were used to express continuous variables. Heterogeneity was tested using the χ 2 test and I 2 statistics. If I 2 was <50% and p > .05, there was no statistical heterogeneity among the studies, and a fixed-effects model was used to combine effect sizes; otherwise, a random effect model was adopted. Sensitivity analysis was performed for possible causes of heterogeneity, and funnel plot was used for publication bias analysis.
Differences were considered statistically significant when p < .05.

Literature search results
The initial search yielded 523 articles, and 205 duplicates were excluded. Then 158 articles were excluded by reading titles and abstracts, 148 were considered ineligible after reviewing the full text F I G U R E 1 Flow chart of literature selection.

Meta-analysis results of MMSE score
A total of seven (Burgener et al., 2008;Hoffmann et al., 2016;Holthoff et al., 2015;Huang et al., 2019;van Santen et al., 2020;Vreugdenhil et al., 2012;Yang et al., 2022) studies compared MMSE items between the two groups in this study. Significant heterogeneity was found among studies (I 2 = 78.5%, p < .001), and the random-effects model was used to combine the effect sizes. The meta-analysis results showed that the MMSE score in the treatment group was higher than that in the control group, and the difference was statistically significant [SMD = 0.45, 95% CI (0.07, 0.83), p = .019; Figure 2a].
According to the experimental duration, the included studies were divided into two subgroups of ≤16 weeks and >16 weeks for further analysis. There was significant heterogeneity among studies when the exercise intervention time was >16 weeks (I 2 = 82.2%, p < .001), and the MMSE score in the treatment group was markedly higher than that in the control group [SMD = 0.55, 95% CI (0.02, 1.07), p = .041; Sensitivity analysis was used to test the stability of the study results.
The analysis results showed that the newly pooled effect size, which TA B L E 1 The basic characteristics of included literature  was obtained after removing each study one by one, was within the 95% CI of the original result, indicating that the results of this study were robust and credible (Figure 2b).

2022) studies compared ADL items between the two groups in this
study. The random-effects model was adopted to combine the effect sizes because of significant heterogeneity among studies (I 2 = 92.7%, p < .001). The meta-analysis results showed that the ADL score in the treatment group was significantly higher than that in the control group [SMD = 0.90, 95% CI (0.20, 1.60), p = .01; Figure 3a].
Further, the six studies were divided into two subgroups according to different experimental duration of each study, ≤16 weeks and >16 weeks. The results showed that there was significant heterogeneity among studies when the exercise intervention time was >16 weeks (I 2 = 96.0%, p < .001), and the ADL score in the treat-ment group was higher than that in the control group [SMD = 2.33, 95% CI (0.72, 3.94), p = .005; Figure 3a]. By contrast, when the exercise intervention time was ≤16 weeks, no significant heterogeneity was identified among studies (I 2 = 37.7%, p = .201), and the treatment group had higher ADL score, but the difference between the two groups was not statistically significant [SMD = 0.16, 95% CI (−0.16, 0.468), p = .326; Figure 3a].
Sensitivity analysis was carried out to test the stability of the study results. After one-by-one removal of the six studies, the newly pooled effect size was still within the 95% CI of the original result, indicating that the results of this study were robust and credible ( Figure 3b). Begg's test was further used to detect publication bias, and no publication bias was detected (p = .260; Figure 3c).

Meta-analysis results of NPI score
A total of seven (Fleiner et al., 2017;Hoffmann et al., 2016;Holthoff et al., 2015;Huang et al., 2019;Rolland et al., 2007;Stella et al., 2011;Yang et al., 2022) studies compared NPI items between the two groups In consideration of the different experimental duration of each study, ≤16 weeks and >16 weeks, the seven studies were split into two groups for further analysis. There was no significant heterogeneity among the studies when the exercise intervention time was >16 weeks (I 2 = 94.1%, p < .001), and the NPI score in the treatment group was significantly lower than that in the control group [SMD = −1.01, 95% CI (−1.99, −0.04), p = .041; Figure 4a]. When the exercise intervention time was ≤16 weeks, there was significant heterogeneity among the studies (I 2 = 51.6%, p = .151), and although the NPI score in the treatment group was lower than that in the control group, the difference was not statistically significant [SMD = −0.43, 95% CI (−0.82, −0.03), p = .034; Figure 4a].
In order to determine the stability of the study results, sensitivity analysis was carried out by one-by-one removal of studies, and found that the newly pooled effect size was still within the 95% CI of the original result, indicating that the results of this study were robust and credible ( Figure 4b).

DISCUSSION
AD is a neurological disease commonly occurring in older adults.
Reported by Alzheimer's Disease International, one more AD patient is added every 3 s, and AD treatment costs will increase to $818 billion by 2030 (Prince et al., 2015). At present, the pathogenesis of AD is not clear, and the popular hypotheses include the amyloid beta cascade hypothesis, tau hypothesis and neurovascular hypothesis (Bakota & Brandt, 2016). The amyloid beta (Aβ) cascade hypothesis suggests that an imbalance between the production and elimination of Aβ-protein is the initial event leading to neuronal degeneration and dementia (Chong et al., 2018). The tau hypothesis is that hyperphosphorylated tau affects the stability of microtubule-associated proteins, leading to neurofbrillary tangles and dysfunction of neurons and synapses (Chong et al., 2018 (Laurin et al., 2001). It has also been suggested that exercise prevents Aβ accumulation (Carro et al., 2006) and promotes brain-derived neurotrophic factor expression, thereby inhibiting cognitive decline (Erickson et al., 2011). However, there is no definitive evidence to prove whether exercise intervention is effective in treating AD patients, and thus we analyzed existing studies.
We screened the relevant literature published at home and abroad in strict accordance with the inclusion and exclusion criteria, and the relevant statistics were combined for meta-analysis. The results showed that the MMSE, ADL, and NPI scores in the treatment group were significantly better than those in the control group. Using the duration of the study as a criterion, we divided the included studies into two groups for subgroup analysis: exercise intervention time ≤16 weeks and exercise intervention time >16 weeks. The subgroup analysis showed that the treatment group was superior to the control group, but there was no statistically significant difference in MMSE score and ADL score between the two groups in the subgroup of