This pilot study used a RCT design with two arms: an immediate treatment group who received a 12-week Hatha yoga program and a wait-list control group. The allocation ratio was 50/50. Outcomes were assessed at multiple time points: baseline prior to initiation of the yoga intervention program, 12 weeks upon the completion of the intervention program, and 6 months post intervention. Because participants in the wait-list control group received the same intervention program after the treatment group completed their program, the wait-list control group had two baseline measurements: at the beginning of study (first) and at 12 weeks prior to their intervention program (second). A 24-week data was collected from the wait-list group upon the completion of the intervention program. There was a total of four data collection points for the wait-list group. The research protocol was approved by the University of Minnesota Institutional Review Board.
Participants and randomization
The CONSORT flow diagram (Fig. 1) illustrates the recruitment and retention process for this study. The sample size was determined by the feasibility of having no more than 10 participants in class with one yoga teacher and a research assistant (RA). Participants were recruited from clinics via flyers, through local and national PD networks such as PD support groups and PD community events, and through the study website or were referred by an investigator from his neurological practice. Inclusion criteria were as follows: individuals diagnosed with mild to moderate idiopathic PD (Hoehn and Yahr stages I–III) , age 45–75 years, on stable dopaminergic therapy for 4 weeks prior to enrollment if taking medication, and able to ambulate 6 m with/without assistive device. Individuals were excluded if they had atypical parkinsonism or other significant brain conditions such as a stroke, had any medical condition that prohibited safe exercise as assessed by the Exercise Assessment and Screening for You Questionnaire , had significant cognitive impairment as indicated by scoring less than 26 in the Montreal Cognitive Assessment (MoCA) , had a decline in immune function such as pneumonia or systemic infection, had spinal fusion or other orthopedic surgery in the past 6 months, had a significant psychiatric disease, needed greater than minimal assistance for gait and transfers, were already practicing yoga regularly, or were unable to commit to attend scheduled yoga sessions. The number of potential participants assessed for eligibility was documented to give an indication of the appeal of the yoga program. Informed written consent was obtained just prior to the initial assessment.
Each participant was assigned a study identification (ID) number from 1 to 20 based on the order of enrolment. A computer-generated random assignment list prepared by a statistician was used to randomize participants. Group allocations were contained in a set of sealed envelopes, each bearing on the outside only the study ID number. The envelopes were distributed by a RA at the end of baseline data collection. Participants assigned to the treatment group participated immediately in a 12-week Hatha yoga group intervention program. Participants in the wait-list group served as control during the first 12 weeks and received the same intervention afterward. The RA who collected the data, the scientists who performed the laboratory analyses, and the statistician who performed the data analysis were blinded to group assignment.
Interventions and setting
The yoga for PD intervention program was initially designed by the lead yoga instructor based on a focused literature review on relevant yoga programs. The draft program was reviewed by an expert panel composed of six yoga experts who specialized in teaching individuals with musculoskeletal and neurological disorders. The expert panel met for a 2-h meeting to discuss the feasibility and intended effects of the program. The final program was approved by all the experts and implemented in the study . The yoga intervention sessions were held twice weekly for 60 min each session at a local yoga studio which was conveniently located at street level with ample parking space. For safety reasons, home practice was not prescribed in this pilot project.
The primary outcome of the study was the alterations in oxidative stress measures in resting blood samples from baseline to 12 weeks. To assess the effects of yoga on oxidative stress, we measured intracellular total glutathione (GSH) levels and glutathione redox status, which is the ratio of reduced to oxidized glutathione (GSH:GSSG) as previously described  in red blood cells (RBCs) using the Thermo Scientific TSQ Quantum MAX, a triple quadrupole mass spectrometer (MS-MS) with electrospray mode (ESI). The liquid chromatography system consisted of Dionex Ultimate 3000 and a Zorbax Eclipse XDB C18 (3.0 × 150 mm × 3.0 μm) column. In addition, we measured malondialdehyde (MDA), a lipid peroxidation marker, using TBARS (TCA method) and protein carbonylation levels in plasma using commercial kits (Cayman Chemical, Ann Arbor, MI). Activity of intracellular antioxidant enzymes, superoxide dismutase (SOD), catalase, and glutathione peroxidase (GPx) was analyzed in RBCs using commercial colorimetric assay kits (Cayman Chemical, Ann Arbor, MI) as per manufacturer’s instructions.
Motor function was examined using the motor portion of the Unified Parkinson’s Disease Rating Scale (UPDRS), i.e., mUPDRS , which has high internal consistency and construct validity . The mUPDRS includes a total of 14 items that provide 27 scores: speech, facial expression, body bradykinesia, posture, gait, and tremors. The scoring range for each item is from 0 (normal) to 4 (severe). Physical activity was measured by the 31-item Longitudinal Aging Study Amsterdam Physical Activity Questionnaire (LAPAQ) [26, 27]. The questionnaire is highly correlated with the 7-day diary (r = 0.68, p < .001) and moderately with the pedometer (r = 0.56, p < .001). The repeatability of the LAPAQ was reasonably good (weighted kappa, 0.65–0.75) in older adults .
Non-motor symptoms including cognitive function, mood, sleep quality, and quality of life were measured using standardized survey instruments. The MoCA  was used to assess participants’ cognitive function including domains of attention and concentration, executive function, conceptual thinking, calculations, visuospatial, memory, language, and orientation. This scale can detect mild cognitive impairment with 90–96% range sensitivity and specificity of 87% with 95% confidence interval .
The Beck Depression Inventory (BDI), a 21-item questionnaire, was used to measure participants’ characteristic attitudes and symptoms of depression . A meta-analysis of the BDI’s internal consistency estimates yielded a mean coefficient alpha of 0.86 for psychiatric patients and 0.81 for non-psychiatric subjects .
Sleep quality was assessed by the Parkinson’s Disease Sleep Scale (PDSS) . This 15-item questionnaire includes items that measure the overall quality of a night’s sleep, sleep onset and maintenance insomnia, nocturnal restlessness, nocturnal psychosis, nocturia, nocturnal motor symptoms, sleep refreshment, and daytime dozing. The scale demonstrated a high intraclass correlation coefficient (ICC) and a good discriminatory power between PD and healthy controls .
The 33-item Parkinson’s Disease Quality of Life Questionnaire (PDQUALIF) was used to assess quality of life in seven domains: social and role function, self-image/sexuality, sleep, outlook, physical function, independence, and urinary function, plus one item of global health-related quality of life . Cronbach’s α of the scale was reported to be 0.89 and the ICC was 0.88 .
Feasibility was measured by the number of eligible subjects, number and type of yoga-related adverse events, retention rate, and reasons for not participating and for withdrawing from the study. Acceptability was evaluated at 12 weeks using an investigator-developed questionnaire. Participants were asked to rate their satisfaction through perceived enjoyment of class, the ease of class, and intention to continue use of the program using a 4-point Likert scale with “1” representing “not at all” and “4” representing “definitely.” Questions on whether participants were satisfied with the frequency and duration of the intervention program were also included. Program adherence was determined by class attendance during the intervention period and whether participants continued to practice yoga at 6-month post-intervention.
The follow-up survey was developed to examine the frequency and duration of yoga practice, factors that influence yoga adherence, beneficial yoga poses, and self-report PD symptoms 6 months after the intervention program. Participants received an email invitation that included a link to complete the follow-up survey via research electronic data capture (REDCap), a secure web application for building and managing online surveys and database .
Demographic information (e.g., age, race/ethnic background, education level, annual household income, marital status, and living arrangement), weight, height (for body mass index calculation), and comorbidities were collected from all participants.
Descriptive statistics were presented as mean (SD) for continuous variables or count (percent) for categorical variables. All the clinical assessments were summarized by time points (at baseline, 12 weeks, and 6 months post-intervention). Data from the first baseline from the wait-list control group were used for between-group comparisons (baseline to 12 week). Data from the second baseline from the wait-list control were used for within-group analyses (12 to 24 weeks).
Data were assessed for whether they conformed reasonably to the normal distribution and to identify outliers. For unadjusted analysis of continuous variables, two-sample t tests were used for between-group comparisons and paired t tests for within-group comparisons; for categorical variables, Fisher’s exact test was used. To compare treatment effect between groups at 12 weeks, linear regression was applied adjusting for the corresponding baseline value and l-dopa; to assess the treatment effect within wait-list control group, a linear mixed model was applied adjusting for baseline and l-dopa value. Change scores (95% confidence interval for mean differences) were calculated for all the efficacy outcomes. All analyses used SAS (V9.4; SAS Institute, Cary, NC). No multiple comparison adjustment is done because this is a pilot study, and no definitive findings are claimed.