Garfinkel 1998.

Methods	Randomised, single‐blind, controlled trial Blinded assessors Ethics approval and informed consent obtained Randomisation occurred at the level of participants, where participants with bilateral CTS received the same intervention for both affected wrists.
Participants	Total N = 51 participants randomised Intervention group N = 26 participants randomised; 22 participants (35 wrists) completed Control group N = 25 participants randomised; 20 participants (32 wrists) completed 13 males; 28 females* Mean age: (SD not reported) Intervention 49 yrs Control 49 yrs Inclusion criteria: 1. Presence of 2 or more of the following: positive Tinel's test; positive Phalen's test; pain in median nerve distribution; sleep disturbance due to hand; numbness/paraesthesias in median nerve distribution 2. Abnormal electrophysiological findings 3. Subject agrees not to change medications, receive other new treatments or change work duties during trial Exclusion criteria: 1. Previous surgery for CTS 2. Rheumatoid arthritis or other recognised inflammatory arthritis 3. CTS related to systemic disease (hypothyroidism) 4. Pregnancy
Interventions	Intervention: Yoga for 1 to 1.5 hours twice weekly for eight weeks Control: Wrist splint to supplement current treatment for eight weeks
Outcomes	Outcomes assessed at the end of eight weeks of treatment: 1. Pain severity using visual analogue scale (0 to 10, with 10 denoting greatest level of pain) 2. Nocturnal wakening using ordinal scale (rated as worsened, same, improved) 3. Phalen's test (rated as worsened, same, improved) 4. Tinel's test (rated as worsened, same, improved) 5. Grip strength in mmHg using sphygmomanometer cuff (mean of 3 trials) 6. Nerve conduction: median motor and sensory distal latencies (in ms) 7. Patterns of paraesthesia and numbness (recorded on hand diagram)**
Notes	*1 missing participant for demographic data **No data reported for this outcome The results of pain severity, grip strength and neurophysiologic parameters were reported as both endpoint and change from baseline values. There were no differences between endpoint and change from baseline values in terms of statistical significance, so we chose to be consistent and only include endpoint values (as we included endpoint values where available for all other studies included in the review). Analysis was undertaken at the wrist‐level for all outcomes, though some participants in each group had bilateral CTS. Bilateral cases had the same intervention applied to each wrist. The trialists did not report how the correlation between both wrists was accounted for in the analysis, and attempts to clarify this information from the trialists were unsuccessful. Therefore, it is not clear whether a unit of analysis error occurred. No attempt was made to adjust outcome data.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "Subjects were randomised into 2 groups by having them select sealed envelopes containing a group assignment". Comment: No information on how the random sequence was generated prior to putting these into envelopes was reported.
Allocation concealment (selection bias)	Unclear risk	Comment: Sealed envelopes were used however  they may not have been distributed according to a randomised sequence and it is unclear whether opaque envelopes were used. It is unclear whether participants or trial personnel could predict assignments.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Comment: Participants were aware of group assignments therefore self‐reported outcomes such as pain, nocturnal wakening, and patterns of paraesthesia and numbness may be biased.
Blinding of outcome assessment (detection bias) All outcomes	High risk	Quote: "The assessments all were conducted by 1 physician who was blinded to the patient's group assignment and the intervention". Comment: Participants were aware of group assignments which may have influenced their performance when outcomes such as grip strength, Phalen's test, and Tinel's test were measured however, nerve conduction studies were less likely to be compromised.
Incomplete outcome data (attrition bias) 3 months or less	High risk	Quote: "9 dropped out or were excluded". Comment: Four participants from the treatment group and five from the control group were not included in the analysis. No reasons were provided to explain these drop‐outs or exclusions
Selective reporting (reporting bias)	High risk	Comment: Patterns of paraesthesia and numbness were recorded on hand diagrams but no results of these measurements were reported. Results were reported for all other measurements.
Other bias	Low risk	Comment: The unit of analysis was at the level of the individual. It is apparent that some participants had bilateral involvement however these cases were equally distributed across the groups [intervention group: N = 22 (35 wrists), control group: N = 20 (32 wrists)].