Study Quality

Study quality was assessed based on study design, sample size, intervention protocol, and statistical analysis. Studies were classified as low, moderate, or high. These criteria are based on categories set forth by the US Department of Health and Human Services 2002 report. Funding source was not included as a category because most studies did not report a funding source, although the available information about funding was described. Because this review includes both randomized and observational trials, categories were adapted for both ( Table 2 ).

Study Design

Of the 10 studies included in this review, 4 were randomized controlled trials (RCTs) ; 2 compared people with arthritis to healthy controls, matching for age and sex ; 2 were non-RCTs (NRCTs) ; and 2 were cohort studies. Among the 4 RCTs, 3 had a waiting-list control and the other 2 were usual care. The NRCTs assigned participants to control if they were unable to attend the first class session. None of the studies had an active control group. Of the reviewed studies, 6 were reported as journal articles, 1 was a letter to the editor, and 3 were presented as abstracts at annual research meetings. One study was presented as an abstract at an annual meeting, followed by publication in a journal that did not include a process of peer review ; therefore, only the abstract was included in this review.

Sample Size

Sample sizes ranged from 3 to 26 intervention completers, with similar numbers of comparator groups. Only 1 study had as many as 20 persons per group. The necessary sample size to detect differences between groups was not generally described. Generally, a subject to variable item ratio of 10 to 1 is recommended in multivariable regression analysis to avoid type I errors, although this ratio depends on variable distribution.

There were 6 studies reported on attrition, with rates of 0%, 9%, 22%, 36%, and 37% each, with the 3 most rigorous studies having the lowest rates of attrition. The 2 cohort trials and 1 trial with healthy matched controls had the highest rates of attrition. The greatest retention was from the NRCTs and 2 RCTs. Most studies analyzed data for completers only. Only 1 study reported the consideration of attrition in final analysis, excluding 1 dropout before baseline. Remaining studies did not report attrition.

Intervention Protocol

Intervention protocols varied widely. The “dose” of yoga varied substantially between studies and was often inconsistent within studies. For example, the study with the greatest dose included 120 minutes of practice 5 times per week followed by once per week for 3 months with 10 to 30 minutes of daily home practice. In contrast, the lowest dose included 60 minutes once per week for 8 weeks. Yet another study was only 15 days long but included daily practice in a retreat setting. Some studies required daily home practice, some weekly, and some had no element of home practice. Although many protocols were developed and/or taught by licensed or certified yoga professionals (teachers, therapists, or scholars), some did not describe the intervention development or delivery. This is further complicated because requirements and regulation of yoga instruction differ by jurisdiction and culture, and credentials of the yoga professionals are not always standardized. Some studies used a style of yoga with a long history and published texts describing teaching methods and practice, whereas others developed a new protocol for the population under investigation. Some studies failed to describe the protocol in any detail.

Three studies, an RCT for hand OA, a cohort study for knee OA, and an NRCT of young adults with RA, used an Iyengar-based yoga program. This style is known for using props (blocks, straps, bolsters) adjusting to individual anatomy. The program for both the OA studies was developed by one of the authors who is a senior certified yoga instructor, and the RA protocol was devised by an experienced Iyengar yoga (IY) teacher. The hand OA trial included 10 weeks of “stretching and strengthening exercises emphasizing extension and alignment, group discussion, supportive encouragement and general questions and answers.” Poses emphasized respiration and upper body alignment. The protocol is described generally with reference to a previous publication. The knee OA study described a 15-pose series and prop modifications, which could be easily replicated. The RA study by Evans and colleagues listed examples of poses. The IY-based programs were conducted for 6, 8, and 10 weeks, meeting once or twice weekly for 60 to 90 minutes.

An NRCT for RA used a program developed in consultation with rheumatologists and a certified yoga therapist. This program, conducted by Badsha and colleagues, included stretches, strengthening, meditation, and deep breathing of biweekly classes for 6 weeks. A study with healthy matched control by Dash and Telles included poses, breathing practices, meditation, lectures, and joint loosening exercises in a 14-day yoga training camp. The RCT by Haslock and colleagues used gentle tailored poses, breath control, meditation, lectures, and discussions, with the intention to soften emotions. For the first 3 weeks, 120-minute sessions were held 5 days per week, followed by weekly 120-minute sessions for 3 months.

Abstracts from an RCT of RA and OA discussed the use of a gentle yoga program developed by rheumatologists, psychologists, and a registered yoga therapist, incorporating poses, breathing practices, relaxation, meditation, chanting, and supplemental reading. A study with age- and sex-matched controls taught a social cognitive theory–based Kundalini yoga intervention to those diagnosed with arthritis. Kundalini yoga concentrates on the spine, with a focus on raising energy and awareness. The study included poses, breathing techniques, meditation, and relaxation. In 2 studies, the reader is referred elsewhere for description of the practice.

Data Collection and Analysis

Well-validated instruments were administered by blinded assessors in 6 studies. These included anatomic changes, biomarkers, performance outcomes, and clinical assessment. The inclusion of unmasked assessors in one study was found to be its greatest limitation.

All but 2 studies measured baseline variables and outcomes recommended by ACR or OARSI. None used sham yoga to blind participants. Consequently, all self-report data sustain possible expectation bias. However, the chosen self-report instruments are commonly used for persons with arthritis and known for strong psychometric properties.

One study with healthy controls created a new assessment tool to measure intervention efficacy and participant perceptions. The investigators had previously used some of the questions in this population and demonstrated strong validity and reliability. Additions to the tool were checked for face and content validity by 3 academics.

Only 2 trials (an NRCT and a cohort study) reported efforts to ensure that data characteristics supported the methods (such as assuming a normal distribution) and adjusted the statistical plan as necessary. Eight articles and abstracts described hypotheses up front and linked outcomes to those hypotheses. The other 2 listed feasibility as their primary outcome. However, some outcomes were not well explained in the study’s context. For example, a study comparing patients with RA to healthy controls hypothesized that yoga would result in increased strength. However, this study measured pre- and postintervention nonsteroidal antiinflammatory drug (NSAID) dose, without assessing analgesic or other medication use and included no pain measures.

Overall Study Quality

Of 8 possible points, studies ranged from 3 to 6 in overall study quality ( Table 3 ). Future expansion from pilot studies and abstracts may include greater rigor. Although available information is limited, the strongest studies can point toward associations that may be confirmed with additional trials.