The study was conducted at Tufts Medical Center, an urban tertiary care academic hospital in Boston, Massachusetts. It received ethics approval from the Tufts Medical Center/Tufts University Human Institutional Review Board. A detailed version of the study protocol was reported in 2008 (13).

Patients with knee OA were recruited from the greater Boston area. To ensure adequate enrollment of underrepresented groups, we placed advertisements in local media. We also used the rheumatology clinic patient database at Tufts Medical Center to identify patients with knee OA. For interested respondents, we determined eligibility through a brief, scripted interview that posed questions with predictive values for knee OA that were known from population-based data. Applicants who screened positive on the telephone interview were scheduled for eligibility visits, when written informed consent was obtained.

The eligibility criteria consisted of age ≥55 years, body mass index (BMI) ≤40 kg/m², Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain subscale score (visual analog version) >40 (range 0–500), and fulfillment of the American College of Rheumatology criteria for knee OA (14) with radiographic Kellgren/ Lawrence scale knee OA grade ≥2 (15). We excluded individuals who had prior Tai Chi training or similar types of alternative medicine like Qi Gong or yoga; individuals with serious medical conditions, limiting their ability for full participation as determined by primary care physicians; individuals with intraarticular steroid injections in the previous 3 months, or reconstructive surgery on the affected knee and any intraarticular hyaluronate injections in the previous 6 months; and individuals unable to pass the Mini-Mental examination (score <24) (16).

To avoid bias in favor of the Tai Chi intervention, we informed participants that we were studying the effects of 2 different types of exercise training programs, one of which was combined with nutrition education. Participants were allowed to continue routine medications and maintain their usual visits with their primary care physicians or rheumatologists throughout the study.

Participants were randomly assigned to Tai Chi (n = 20) or to an attention control group (n = 20). Randomization assignments were designated by the statistician (CHS), using computer-generated numbers to randomize permuted blocks of sizes 2 and 4 so that each block was complete. They were provided in sealed, opaque envelopes and opened on the patient's agreement to participate. The block size was randomly assigned to minimize correct prediction of assignments, while preserving approximate balance between the groups. Outcomes that required an analysis of knee strength were based on evaluation of the knee reported as most painful at baseline. If both knees were equally affected (which occurred in 2 participants), one knee was chosen at random as the affected knee.

Subjects participated in 60-minute Tai Chi sessions twice weekly for 12 weeks instructed by a Tai Chi master (Ramel Rones) with more than 20 years of teaching experience. In the first session, we explained Tai Chi theory and procedures and provided the patients with printed teaching materials, including a well-tested, validated Institutional Review Board–approved Tai Chi program that described Tai Chi principles, practicing techniques, and safety precautions for knee OA (13, 17). For the remaining sessions, each subject practiced Tai Chi under the instruction of the Tai Chi master. Every session included 1) 10-minute self-massage and a review of Tai Chi principles, 2) 30 minutes of Tai Chi movement, 3) 10 minutes of breathing technique, and 4) 10 minutes of relaxation. The program consisted of 10 forms from classic Yang style Tai Chi (18), with minor modifications that were suitable for people with knee pain. This involved eliminating stances that require greater than 90° knee flexion and can cause excess knee joint stress (19). We also provided a Tai Chi DVD published by Ramel Rones. Patients were instructed to practice Tai Chi at least 20 minutes per day at home and encouraged to maintain their usual physical activities, but not to participate in additional new strength training or exercise programs other than Tai Chi. After completing the 24 treatment sessions, we instructed the subjects to continue practicing Tai Chi at home following the DVD and handouts until the 48-week followup visit.

The wellness education and stretching program provided an active control for the attention being paid to the Tai Chi group (13, 20). The control group attended two 60-minute class sessions per week for 12 weeks. In the first session, research staff explained the program and procedures. A variety of health professionals provided nutrition and medical information in the following sessions. Every session included 40 minutes of didactic lessons on 1) OA as a disease, 2) diet and nutrition, 3) therapies to treat OA, or 4) physical and mental health education (e.g., recognizing and dealing with stress). The nutrition education was based on Dietary Guidelines for Americans (21) and focused on general knowledge of nutrition, cooking, and shopping, but not on specific nutrients and supplements. The final 20 minutes consisted of stretching exercises involving the upper body, trunk, and lower body, with each stretch being held for 10–15 seconds. Participants were instructed to practice at least 20 minutes of stretching exercises per day at home. They were encouraged to maintain their usual physical activities, but not to participate in additional strength and mind–body exercise programs other than their stretching exercise. The stretching and health information was compiled using materials from our previous program (17) and the Web site of the National Institute of Arthritis and Musculoskeletal and Skin Diseases at the National Institutes of Health (22).

Throughout the 12-week period, we tracked the reasons for missed sessions and the number of missed sessions, and asked the subjects to complete daily logs indicating the amount of time they practiced Tai Chi or stretching exercises.

Our knee OA outcome measurements were drawn from the core set recommended by the Osteoarthritis Research Society International (23), and focused on pain, physical function, and patients' overall assessment of their knee OA severity. The primary outcome measure was the change in the WOMAC pain subscale between baseline and 12 weeks. The WOMAC is a validated, self-administered instrument specifically designed to evaluate knee and hip OA (24, 25). It has 3 subscales that we analyzed separately: pain (score range 0–500), stiffness (score range 0–200), and function (score range 0–1,700), with higher scores indicating more severe disease. Secondary outcomes included weekly WOMAC pain scores during the 0–12-week intervention and assessments of WOMAC function, WOMAC stiffness, and global knee pain status assessed separately by the participant and a study physician (RK) who was blinded to group assignment (visual analog scale score range 0–10, where 0 = no pain) (26). We evaluated physical performance using the timed chair stand (measured in seconds) (27), 6-minute walk test (measured in yards) (28), and standing balance (score range 0–5, where 0 = worst balance) (29). Additional measures included the Center for Epidemiologic Studies Depression Scale (CES-D; score range 0–60, where 0 = no dysphoria) (30), outcome expectations for exercise (score range 1–5, where 1 = no outcome expectations) (31), self-efficacy (score range 1–5, where 1 = no self-efficacy) (32), and the physical component summary and mental component summary of the Short Form 36 to assess quality of life (score range 0–100, where 0 = worst health state) (33). Adherence and occurrence of adverse events were also assessed.

We instructed participants to maintain their regular medications, including nonsteroidal antiinflammatory drugs (NSAIDs) and acetaminophen. We recorded any change of medication use at each assessment. To test durability of response, we repeated outcome measures at the 24- and 48-week followups.

We determined the sample size of 40 patients using the results of an RCT conducted at Tufts University that tested an exercise intervention among older adults with knee OA (19). That study enrolled 46 patients and randomized them to either 4-month home-based progressive strength training or an attention control group. The strength training group experienced a 36% decrease in WOMAC pain (the primary outcome; mean ± SD change of 79 ± 91) compared with an 11% decrease (mean ± SD change of 20 ± 77) in the attention control group. Based on those numbers, a sample size of 20 per group at an alpha level of 0.05 would have 60% power to detect an expected difference of 59 points. Although we recognized that our study was underpowered for a definitive comparison, we were primarily directed toward gathering preliminary data in order to evaluate this research direction.

We analyzed the data on an intent-to-treat basis. We compared between-group changes in outcomes across all times at 0, 12, 24, and 48 weeks with mixed models, using time and group as categorical fixed factors with random intercepts and first-order autocorrelation of the errors. Similar mixed models were used to examine weekly WOMAC pain analyses from baseline to 12 weeks. Also, we evaluated for potential effects of confounding or interaction with treatment by covariates, including age, sex, BMI, disease duration, disease severity (pain, function, and radiograph score), comorbidities, health status, and use of pain medications. A 2-sided P value less than 0.05 was considered to indicate statistical significance. Results are shown as the between-group differences with 95% confidence intervals (95% CIs) of the differences.

Table 1 shows the baseline data of the 40 participants before randomization to the intervention groups. The participants had a mean age of 65 years and were mainly women (75%) and white (70%). On average, participants had 10 years of knee pain and a BMI of 30.0 kg/m². There were 2 patients, 1 in each group, who were enrolled with a self-reported asymptomatic total knee replacement (one knee), but also with a symptomatic nonoperated knee (study knee). Baseline characteristics were reasonably well balanced between the groups, although there was somewhat greater knee OA severity and comorbidity in the control group.

Attendance for the interventions was 85% for Tai Chi and 89% for attention control over 12 weeks. No patient withdrew from the study (40 of 40 participants completed followup at 12, 24, and 48 weeks). We followed our rigorous study protocol to achieve high levels of adherence and attendance (14). We built a highly experienced and dedicated research team, selected participants who were interested and reliable, accommodated patient preference when scheduling evaluation visits, randomized patients after the baseline evaluation to have a large enough pool for replacements, engaged in friendly personal contact with patients, organized interesting classes, conducted a rigorous quality control program, and prepared a detailed description of the training sessions that we pilot tested prior to the initiation of this study. We offered subjects a monetary incentive to maximize participation and rigorously adhered to the Manual of Operations to enhance compliance.

At the 24-week followup, 13 (65%) of 20 participants in the Tai Chi group continued to practice Tai Chi, and 12 (60%) of 20 in the control group continued to practice stretching exercises. At the 48-week followup, these rates were 9 (45%) of 20 for Tai Chi and 8 (40%) of 20 for the control group.

At 12 weeks, the Tai Chi arm exhibited a statistically significant decrease in knee pain, as measured by the WOMAC pain scale, compared with the attention control group. The between-group mean difference was −118.80 (95% CI −183.66, −53.94; P = 0.0005) (Table 2).

Figure 2 shows that the between-group mean difference gradually increased during the 12-week intervention, based on the nearly linear decline in pain with Tai Chi. Table 2 compares changes in all of the secondary outcomes between groups from baseline to 12, 24, and 48 weeks. These changes are illustrated in Figure 3.

One participant in the Tai Chi group reported an increase in knee pain at the 2-week assessment. This was resolved following modification of that participant's Tai Chi technique. One participant in each assignment group reported newly diagnosed cancer (1 breast cancer, 1 colon cancer) during the 12-week intervention period (the patient in the Tai Chi group missed 4 intervention visits but completed her followup evaluations). No other adverse events were reported.

At baseline, 11 (55%) of 20 participants in the Tai Chi group were receiving analgesics and NSAIDs compared with 14 (70%) of 20 in the attention control group. The numbers decreased to 6 (30%) of 20 for the Tai Chi group and 10 (50%) of 20 for the control group at the 12-week assessment (Table 3). We found no statistically significant differences between the groups who were using any of the 2 categories of medications.