Study design and setting
This was a single-blinded randomized controlled trial with two study arms, including a supervised Baduanjin exercise group and a wait-list control group. Participants were recruited from the three affiliated hospitals of Shanghai University of Traditional Chinese Medicine, including Yueyang Hospital of Integrated Traditional Chinese and Western Medicine, Seventh People's Hospital, and Longhua Hospital. The outcome measures were evaluated in a quiet room in the hospital to avoid interference.
Participants were randomly assigned to an intervention group or a control group—in a 1:1 ratio using SPSS 24.0 software. Researchers screened patients who met the inclusion and exclusion criteria and provided relevant information, as well as possible benefits of the exercise and the safety. After obtaining consent, the research team examined the candidates with patients' questionnaire. The inclusion criteria were showed as follows: (1) female patients newly diagnosed with stage I to III breast cancer; (2) scheduled to receive adjuvant chemotherapy; (3) aged 40 to 75 years; (4) using WeChat; (5) sufficient fluency of the Chinese language; (6) willing to participate in the study and be randomly assigned. Exclusion criteria included participants with (1) disease recurrence or metastasis; (2) known conditions and/or diseases that impact cognition (e.g., Alzheimer's disease, dementia, or other psychological diagnoses); (3) disorders that might preclude exercise participation (e.g., fracture, arthritis); (4) current participation in mindfulness-based exercise programs (e.g., TaiChi or Qigong). Due to the nature of exercise studies, only results evaluators were blinded. More details were presented in this previous protocol of study .
Sample size estimation
The sample size was estimated in this study by using perceptive cognitive impairment as the main effect indicator, which refers to the results of previous study . Using the G*Power 3.1 software, 58 cases of the total sample were calculated with a test power of 0.80. Considering the 20% dropout rate, the total sample size for this study was about 70 patients randomized to the intervention group or the control group.
Baduanjin exercise was conducted according to the standards promulgated by the General Administration of Sport in China in 2003. A professionally trained Qigong specialist uniformly trained subject researchers to learn the exercise training. Researchers then conducted on-site instruction to guide theses participants and correct their movements. Participants also were provided with video demonstration to promote daily practice. Before initiating the first cycle of chemotherapy, the patients had the first session in the hospital, then following the video workout at home. Recommended training time was 5 times a week for half an hour each time during the 12-week exercise period. The program began with stretching the joints, inhalation and exhalation for 2 min respectively, and two 12 min Baduanjin sessions, followed by 2 min of muscle relaxation exercises. Each patient's log was recorded during exercise at home, including any obstacles that affected exercise training. When the patient went to the hospital to receive the next cycle of chemotherapy, the feedback of exercise log was collected and checked.
Patients randomized to the control group were given a face-to-face health education. Disease-related questions raised by the patients were directly communicated or answered through WeChat online or oral communication. Patients improved the affected limb through exercises, including abduction, forward flexion, backward extension, internal rotation, and wall climbing according to the guideline and standard for the diagnosis and treatment of breast cancer in China . In addition, the patients were not asked to conduct other aerobic activities. Researchers asked about their conditions twice a week during the 12 weeks.
At baseline, a self-designed demographic data sheet was used to collect the social-demographic data, including age, BMI, marital status, income, education and living condition, as well as relevant disease information from electronic medical system.
The Functional Assessment of Cancer Therapy-Cognitive Function (FACT-Cog) was developed by Wagner et al  to assess subjective cognitive function.The Chinese versions of the FACT-Cog was used; The Multidimensional Fatigue Symptom Inventory-short form (MFSI-SF) was used to assess the fatigue of patients in multiple aspects . Taiwanese scholar then revised the Chinese version of the Multidimensional Fatigue Symptom Inventory-Short Form (MFSI-SF-C) , which was used in this study; The Hospital Anxiety and Depression Scale (HADS) was used for the assessment of patients' anxiety and depression ; The Functional Assessment of Cancer Therapy-Breast (FACT-B), a 44-item self-report instrument was well-known and widely used to measure multidimensional quality of life (QoL) in patients with breast cancer .
Procedure and ethical considerations
The protection of human subject was approved by the Ethics Committee of the Seventh People's Hospital of Shanghai University of Traditional Chinese Medicine (No. 2020-7th-HIRB-011). The purpose and process were explained to potential participants. Written informed consent was obtained from each participant; each participant was assured of confidentiality and the option to decline participation or to withdraw from the trial at any time without penalty. Baseline data collection was conducted by two trained research assistants. After the pre-test data collection, the participants were randomly allocated to receive the Baduanjin exercise programme or the wait-list control group. The post-test data was collected at 4 weeks (T1), 8 weeks (T2), and 12 weeks (T3) during intervention.
Demographics and other characteristics were reported descriptively via the statistics software (SPSS 24.0). The means and SD were calculated for continuous normally distributed variables and medians and ranges for non-normally distributed variables. Categorical variables were presented by absolute numbers and percentages. We performed complete case analyses after describing the pattern of missing data.
According to the intention-to-treat principle, a linear mixed-effect model for repeated-measures analysis was used, in which group, time point, and the interaction between group and time point were set as fixed effects; individual patient was set as the random effects; and age was adjusted as a covariate in the model. Adherence to the Baduanjin exercise was noted. Two-sided significance level for all tests was 0.05. The PROCESS macro  was employed to examine the mediating effects of fatigue and anxiety on the relationship between the exercise and subjective cognition. 95% confidence intervals for direct and indirect effects were generated via bias-corrected bootstrapping with 10,000 resamples. The nonparametric bootstrapping procedures are superior to traditional regression methods for testing the mediating effects, because the former does not make assumptions regarding the shape of the distribution of the variables or the sampling distribution . Confidence intervals without zero showed significant mediation.