This meta-analysis was conducted and reported in accordance with PRISMA principles [18]. Up to 19th December 2022, Jinling Y. and Yang L. made literature search from WBM, CNKI, PubMed, Web of Science (WOS), Scopus, Cochrane Library, and EBSCOhost databases by using the Boolean search technique, phrases and fields specific to the database (Appendix Table 3). For instance, the PubMed search string was: [title/abstract] [(“Parkinson disease”[MeSH])] AND (“Tai Ji”[MeSH]) AND (“cognition”[MeSH] OR “cognitive behavioral”) AND (“randomized controlled trial”[MeSH]).

The inclusion and exclusion criteria for the designed study were as follows: (i) Participants: people with PD and no other primary disease; (ii) Intervention and comparison: studies comparing TJ interventions with active and inactive controls were included; studies where TJ was specified as control condition or TJ combined with other practices or exercise forms were excluded; (iii) Outcomes: any data assessing the cognitive function; (iv) Study type: studies with randomized controlled trials (RCTs) were included.

Titles and abstracts were used to filter distinct items (after removing duplicates) retrieved from electronic bibliographic database searches for inclusion. The searched literature was vetted based on inclusion and exclusion criteria. JY and YL organized and executed the research project, designed and executed statistical analysis. WL, conception and organization of research project. YF, conception of research project. Additional papers were regarded as complementary but used in the data collection process. Text reporting the most comprehensive data was chosen as the main text.

The searches used the article quality assessment criteria manual recommended by Cochrane Manual 5.1.0 to assess the quality of included literature [19] and evaluated each sub-item as “low risk”, “unclear” and “high risk”.

The two studies extracted the following information: first author, publication year, sample size, age, disease severity, intervention method, intervention dose, and outcome. The methods used to test various cognitive functions in humans were divided into executive function, global cognition, and cognitive motor [20].

Forest maps were plotted for included publications using Review Manager 5.3 (RevMan 5.3, The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark), and the calculated effect sizes were analyzed through standardized mean differences (SMD). The fixed-effects and random-effects models’ 95% confidence intervals (95% CI) were estimated. I${}^2$ was the main indicator of heterogeneity; If I${}^2$ was 50%, the heterogeneity could be ignored, and a fixed-effect model was applied; if I${}^2$ was between 50% and 75%, heterogeneity was moderate; and if I${}^2$ was $>$75%, the random effect model was applied, and heterogeneity causes were discussed and analyzed. SMD $\ge$0.2 was the modest, $\ge$0.5 was medium, and $\ge$0.8 was large when calculating the effect size. A significant impact was considered with p 0.05.

The databases yielded a total of 43 articles, where 16 were excluded before screening. Thirteen articles were disregarded after reading the titles and abstracts. The remaining 14 were required to be read in their entirety. One article was not retrieved. Seven articles were excluded because of too low quality (n = 2), TJ intervention present in both experimental and control groups (n = 1), outcome indexes did not mention cognitive component (n = 2), and the outcome data unavailable (n = 2). The remaining 6 articles were quantitatively synthesized and analyzed (Fig. 1).

Fig. 1.

Flow diagram summarizing the literature search strategy.

The characteristics of included studies were summarized in Table 1, where 6 RCTs [21, 22, 23, 24, 25, 26] were included in this study. 116 participants (experimental group) completed TJ training, and 135 underwent other training forms, usual healthcare, or no intervention as a control group. The age ranged from 62 to 70 years. The studied number of subjects varied from 21 [23] to 60 [24]. Most studies had a higher ratio of men to women [21, 23, 24, 25] except one study with a higher ratio of women [22] and one study did not mention this ratio [26]. In most studies, the participants had disease duration of ~8 years [21, 22, 23], few ~4 years [24, 25], and the shortest was ~3 years [26]. Studies presented participants’ baseline PD ratings in the same way (Hoehn-Yahr) between 1 and 3. Intervention doses ranged 2–4 times per week, with each lasting 30 to 60 min and the common duration being 60 min [21, 23, 24, 26]. Total intervention time varied from 8 [22] to 24 weeks [26]. In outcomes, most studies used scale reflecting cognition (Parkinson’s Disease Questionnaire 39 (PDQ-39) [21, 23], Korean Version of Mini Mental Status Examination, K-MMSE [22], Montreal Cognitive Assessment (MoCA) [24, 25]), while few employed executive function measures [23, 26].

Five studies (83.3%) reported the random sequence generation methods. Four (66.7%) mentioned the allocation concealment techniques. Three (50%) managed to blind the subjects and staff. They thus had low bias risk because of assessors’ blinding. Regarding bias reporting, every study reported the anticipated findings. The majority of studies lacked a description of additional bias risks. Three studies (50%) were of excellent quality providing an overview of 6 RCTs’ quality ratings. A quality assessment summary of 6 RCTs was presented in Figs. 2,3.

Fig. 2.

The bias risk for included studies.

Fig. 3.

Bias risk summary: authors’ judgments about each bias risk item for included studies.

Disease duration: In the disease duration subgroup analysis (Table 2). TJ had the greatest impact on improving cognitive function in PD patients having the disease of 4 years [SMD = 2.38, 95% CI (1.32, 3.45), p 0.0001]. There was no significant difference in PD patients with a disease of $>$7 years [SMD = 0.25, 95% CI (–0.08, 0.59), p = 0.14].

Intervention dose: The subgroup analysis results for exercise frequency of intervention doses depicted that two TJ interventions per week were the most conducive [SMD = 2.81, 95% CI (1.96, 3.66), p 0.00001]. Three TJ interventions per week decreased [SMD = 0.49, 95% CI (0.52, 1.89), p = 0.0006], while four did not significantly improve the cognitive function in PD patients (p = 0.66). The effect of intervention lasting longer than 12 weeks was the largest [SMD = 2.21, 95% CI (1.45, 2.97), p 0.00001], while TJ interventions of more than 16 weeks did not significantly improve the cognitive function in PD patients (p = 0.26).

A meta-analysis of 6 RCTs was performed with 251 participants. The findings revealed that TJ led to cognition improvements (overall cognitive function, p 0.05). TJ training also improved the global cognitive function (p 0.05) and executive function (p = 0.09) compared with control. However there was no significant improvement of cognitive motor. In subgroup analysis, TJ training improved cognitive function in PD patients with a disease duration of 4 years, while there was slight improvement with a disease duration $>$7 years which was not statistically significant. For TJ training intervention, doses of two times per week with 45–60 min sessions over $\le$12 weeks improved the cognitive function of PD, and other training dose modes had obvious but small effects.

TJ training, compared to control, enhanced cognitive function of PD patients as per this meta-analysis. This was consistent with the findings of Horstink M et al. [27] that TJ training enhanced cognitive function in both normal cognitive function and cognitive impairment. TJ delayed cognitive decline in moderate PD patients. Previous studies revealed that physical activity interventions improved cognitive function in adults regardless of their cognitive status, and TJ significantly improved cognitive function [28, 29, 30, 31]. This study summarized TJ effectiveness in PD patients of various degrees, providing a theoretical foundation for precise exercise prescriptions.

The vital part of a person’s cognition is executive function. These results initially suggested that TJ might not be the best treatment for enhancing the cognitive executive function of PD patients (p 0.1). The brain frontal lobe was recognized for controlling tasks like memory renewal, attention switching, and multi-task coordination, but the executive function was higher cognitive activity connected to it [32]. The prefrontal, parietal, and temporal lobes of the brain could modify their form and function [33]. TJ training as moderate-intensity aerobic exercise thus improved people’s executive function. Considering that one study was included [23], there was a trend of significant differences in outcomes, although not significant. The hypothesis that TJ training improved executive function in PD patients would be clearly demonstrated if more RCTs were analyzed.

Dual tasks were used for the outcomes regarding TJ effect on cognitive motor [26]. A dual-task design was employed to compare the young and older adults executing specific cognitive tasks while standing [34], walking [35], or in response to postural perturbation [36, 37]. It assessed postural control with concurrent cognitive processing. According to the Vergara-Diaz G study, the dual-task gait stride-time variability was sensitive and logical outcome for evaluating the combined cognitive motor was affected by TJ in PD [26]. However, because of the small sample size and high RCT bias in the TJ intervention group of the Vergara-Diaz G study, there was insufficient evidence in our study to support that TJ exercise significantly improved cognitive motor in PD (p = 0.19).

Parkinson’s Disease Questionnaire 39 (PDQ-39), the Trail Making Test A and B, and Montreal Cognitive Assessment (MoCA) were the often utilized tools in evaluating the global cognitive function. Mini-Mental State Examination (MMSE) was employed for cognitive screening. MMSE was a comprehensive instrument for detecting cognitive deficits [38]. MMSE exhibited ceiling effects on mild cognitive impairment and floor effects on severe cognitive impairment [39]. MMSE was less sensitive to mild cognitive impairment in PD [40]. K-MMSE performed like MMSE. MoCA was created as a screening tool for mild cognitive impairments (MIC) and was an often used instrument for these investigations [41]. MoCA assessed the executive and visuospatial abilities, memory, language, and attention. The scale’s psychometric qualities were sufficient for a quick evaluation of global cognition in PD [42]. Trail Making Test (TMT) was another often used instrument. TMT could be affected by cognitive changes in PD [43]. The 39 items on self-administered PDQ-39 were divided into eight categories, one of which was the cognitive domain [44].

In the present study, it was found that signal intensity of global cognition in PD was improved (p 0.05) by TJ intervention. However, a thorough analysis of these findings was required. There was no agreement on ideal intervention frequency and length because of the heterogeneity of cognitive tests employed in evaluating the global cognitive performance in included trials. This did not change the conclusion that TJ exercises considerably enhanced PD global cognitive function.

TJ intervention modalities performed more than 3 times a week for 45 to 60 min over at least 12 weeks are preferred in PD patients of short disease duration.

Despite our clarifications of TJ application in PD patients, some issues need further explorations in future experimental protocols. Firstly, which cognitive domains of PD are better enhanced by TJ cannot be accurately judged because of the limitations of current RCTs. Second, there is a need for accurate exercise prescriptions for PD patients of different disease courses. Third, the number of high qualities RCTs is insufficient, and subjective assessment processes may have confounded the results.