The effect of physical activity interventions on cognition function in patients with diabetes: A systematic review and meta‐analysis

Abstract Background In recent years, studies have revealed that cognition may be impaired by glucose metabolism disorder. Meanwhile, physical activity has been demonstrated to maintain blood glucose. This meta‐analysis was conducted to assess the effect of physical activity on cognition in patients with diabetes and provide evidence for the treatment of cognition impairment among them. Methods We searched studies published in five databases from 1 January 1984 to 29 August 2020. A random‐effect or fixed‐effect meta‐analysis was used to estimate the pooled effect of physical activity on the change of cognition throughout intervention duration and post‐intervention cognition scores by standardized mean difference (SMD) and its 95% confidence interval (CI). We used funnel plots to evaluate the publication bias, I 2 statistic to evaluate the heterogeneity and did subgroup analysis stratified by sample size and follow‐up time. Results Five eligible studies involving 2581 patients with diabetes were included. The pooled effect of physical activity on cognition improvement in patients with diabetes was significant (SMD = 0.98, 95% CI: 0.34–1.62), while the effect on post‐intervention cognition scores was not significant (SMD = 0.35, 95% CI: −0.04–0.73). In the subgroup analysis, the pooled effect was significantly higher in studies of follow‐up time less than 1 year (SMD = 2.14, 95% CI: 1.63–2.64), while observing no significant effect in studies of follow‐up time over 1 year (SMD = 0.10, 95% CI: −0.11–0.32). Conclusions Physical activity is beneficial to improving cognition in patients with diabetes. However, the long‐term effect needs to be explored in future studies.


| INTRODUCTION
Diabetes mellitus is a common chronic disease that features raised level of blood glucose due to the deficiency in insulin secretion or the inability for receptor to respond to insulin signal, which results in multiple complications including cardiovascular diseases, nerve damage, kidney damage, eye disease and cognitive dysfunction. 1,2 With the change in people's lifestyle, the global prevalence of diabetes increases year by year. Currently, there is an estimation of 463 million adults aged 20-79 years old who live with diabetes, which constitutes 9.3% of the world's population. 2 It is expected to reach 578 million cases (10.2%) in 2030 and 700 million (10.9%) in 2045, bringing heavy burden to families and the society. 2 Currently, various epidemiological studies and scientific research have discovered the correlation between diabetes and cognitive dysfunction, demonstrating that patients with diabetes have lower cognitive function (including attention, memory and information processing speed) than healthy subjects while having a higher risk of dementia. It indicates that diabetes serves as a risk factor of cognitive impairment, which damages patients' memory and language function, declines their quality of life and leads to severe behaviour disorder. [3][4][5][6][7][8][9][10][11][12][13][14] Previous studies have revealed the positive influence of physical activity on both diabetes and cognitive impairment, showing that physical activity intervention not only contributes to hypoglycaemic control, [15][16][17] but also reduces the risk of dementia and improves cognition function. [18][19][20][21] Furthermore, among studies which focused on the effect of physical activity intervention in patients with diabetes, some of them reported that physical activity had benefits on improving cognitive function in patients with diabetes, 22-25 while others did not find the impact of physical activity. 26 The effect of physical activity intervention in patients with diabetes remains controversial.
Several meta-analysis studies 27,28 had been conducted to examine the effect of physical activity on the cognition of patients with dementia or the risk of cognitive impairment among patients with diabetes. However, no meta-analysis has been made to synthesize the effect of physical activity on cognition in patients with diabetes. Patients with diabetes and complicated with cognitive dysfunction bring a much heavier burden for the care from family members, cause significant economic impact in the society, and increase the difficulty of treatment and compliance. Hence, we performed a systematic review and meta-analysis of randomized control trials and cohort studies to systematically investigate the effect of physical activity on cognition function in individuals with diabetes.

| Data sources and search strategy
We searched for eligible studies published from 1 January 1984 to 29 August 2020, from five databases including PubMed, Embase, Web of Science, Medline and Cochrane Library using the following search term with no limitation of language: ('physical activity', 'exercise', 'sports', 'walk', 'activity', 'danc(e/ing)', 'train', 'yoga', 'Tai Chi', or 'strength') AND ('trial' or 'RCT' or 'cohort') AND ('diabetes' or 'glucose') AND ('cognition', 'cognitive', or 'MMSE'). Records were managed by EndNote X 8.0 software to exclude duplicates. This study was strictly performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) and the PRISMA checklist was also provided in Appendix 1.

| Quality assessment
The quality of included studies was assessed by the criteria developed by Hoy and his colleagues. 29 We assigned the 10 items with a total score of 10, with a score of 1 representing 'yes' while 0 represents 'no'. According to the criteria, we assessed that the studies included had a moderate (6-8 scores) risk of bias. Two investigators (WRT and YWX) assessed the quality of studies independently, with discrepancies solved by a third investigator (DM).

| Data extraction
Two researchers (WRT and YWX) scanned independently titles and abstracts of studies according to the inclusion and exclusion criteria to identify eligible studies, and full text was read if necessary. In the five studies selected, the following data were extracted independently by two investigators (WRT and YWX): (1) basic information including the first author and publication year of each study; (2) characteristics including sample size, mean age, sex ratio, cognition measurement method (scale), type of intervention, frequency, duration and therapy for control group; (3) primary outcomes including the scores and mean difference between control and intervention group's cognition, the mean difference of the change in cognition throughout intervention duration and the corresponding standard deviation.

| Data synthesis and statistical analysis
We used a meta-analysis to summarize data from RCT or cohort research and pooled the study-specific estimates using a randomeffects or fixed-effects model to obtain an overall summary estimate of the effect of physical activity across studies. The primary outcome in this study was the change of cognition throughout intervention duration. The secondary outcome was post-intervention cognitive scores. The intervention effect was measured by the SMD of the change of cognition throughout intervention duration or post-intervention cognition scores between intervention and control groups. The results of the included studies were performed with fixed-effect models or random-effect models in cases of significant heterogeneity between estimates. I 2 statistics was used to assess the magnitude of heterogeneity, with 25%, 50% and 75% representing low, moderate and high degrees of heterogeneity, respectively. 30 The chosen proper effect model was based on the analysis results: the fixed-effects model was used if I 2 ≤ 50% and the random-effects model was used if I 2 > 50%.
If substantial heterogeneity was detected, we did subgroup analysis when possible to investigate the possible sources of heterogeneity using the following grouping variables: sample sizes and follow-up time. Subgroup comparisons used the Q test. We considered a subgroup difference p-value less than 0.05 to be indicative of significant difference between subgroups. Sensitivity analysis was performed by a deleted study with the lowest quality score and by using a different model (fixed-effect or randomeffect model). The effect of physical activity was quantified using the SMD values and the corresponding 95% CIs, and a value of p < 0.05 was deemed significant. We used forest plots to describe the pooled effect of physical activity on related outcomes, and used funnel plots and Egger' publication bias test to assess publication bias. We analysed data using Stata version 16.0.

| Study selection and study characteristics
The systematic literature search identified 4341 articles. After screening the titles and abstracts of all references, 120 reviews, -3 of 9 381 conference paper and 1919 irrelevant studies were excluded, 12 potentially eligible articles were read full text, as 7 articles that failed to provide sufficient data or meet inclusion criteria were excluded. In total, five studies were therefore included (see Figure 1).

| Effects of physical activity interventions on cognitive function in patients with diabetes
The pooled effect of physical activity on the change of cognition throughout intervention duration and post-intervention cognition scores are shown in Table 2. The pooled analysis showed a significant  After excluding the drop-outs.

| Sensitivity analysis and publication bias
In the sensitivity analysis, the pooled results of meta-analysis using random-effect models were consistent with the principal findings of bias on the effect of physical activity interventions on the postintervention cognitive scores (t = 2.57, p = 0.124, Figure 3B).

| Subgroup analysis
In the subgroup analysis, the pooled effect of physical activity on improvement of cognition function was significantly higher in studies of follow-up time less than 1 year (SMD = 2.14, 95% CI: 1.63-2.64, Zhang et al. 28 made a meta-analysis of 17 studies and showed that the risk of Alzheimer's disease (AD) is higher among people with diabetes than in the general population, they suggested that the necessary treatment measures should be taken in order to decrease the risk of AD. This was similar with the results from our study that physical activity positively improved cognition of patients with diabetes. Our findings indicated that physical activity could be a measure to prevent AD and reduce the risk of dementia in individuals with diabetes.
Diabetes is an important public health problem. Uncontrolled diabetes could lead to complications in many organs including brain.
Previous studies reported that the presence of comorbidities in severe acute respiratory syndrome patients increased the risk of death by nearly twofold. 31 Balance and working memory functions were simultaneously impaired in patients with type 2 diabetes. 32 In contrast, physical activity was associated with less microvascular disease in the brain and in other vascular beds 33 and less brain atrophy. 34 Previous studies reported that physical activity could improve cognition in patients with dementia and diabetes. It is noteworthy that individuals with diabetes, who experienced greater cognitive difficulties, were less likely to remain adherent to exercise or diet. 35 So it was recommended for clinical that physical activity intervention should be undertaken as early as possible.
Among the included studies, physical activity intervention

| CONCLUSION
This meta-analysis shows that physical activity interventions including aerobic exercises are beneficial to improving cognitive function in patients with diabetes. Our findings could provide evidence for physicians regarding the efficacy of physical activity in patients with diabetes to improve cognitive function and prevent dementia. However, the long-term effect of physical activity intervention needs to be explored in future studies.

F I G U R E 4
Forest plots for subgroup analysis of the effect of physical activity on the change of cognition function in patients with diabetes WANG ET AL.