Randomized controlled trial of yoga and exercise in multiple sclerosis

Study design.

This was a 6-month parallel-group, randomized controlled trial performed in adults with MS that had the approval of the OHSU Institutional Review Board. All subjects provided written informed consent. After completion of the baseline evaluations, subjects were randomized to one of the three experimental groups lasting 6 months: yoga class, exercise class, or wait-list control group. To ensure acceptance of the protocol, wait-list subjects were told they could enroll in either a yoga or exercise class after the 6-month period at no cost.

Subjects were randomly assigned to treatment groups in this study using a modified minimization scheme in such a way as to maintain balance across multiple stratification variables with relatively small numbers of subjects. Cohorts were recruited and then treatment group assignments were made for the entire cohort at one time. A cohort is a group of subjects enrolled within a 1-month period that allowed for active intervention subjects to begin their exercise and yoga classes at the same time. The stratification variables were age group (less than 50 years old and 50 years or older), sex, and baseline Expanded Disability Status Scale (EDSS) (less than or equal to 2.5 and greater than 2.5). Treatment assignment was made by the project statistician who was otherwise uninvolved with the assessments. Initially, all the subjects in each of the cohorts were randomly ordered ensuring this assignment scheme is, in fact, random. Then, the minimization approach of Taves13 was used as a starting point. This process minimizes the absolute differences between the groups for each possible assignment of the next patient to treatment groups. Where these absolute differences are equal in two or three groups, the subject was assigned randomly (with probability of being assigned to one group equal across groups). The minimization approach was modified to account for randomization within cohorts prior to treatment assignment.

Subjects.

Subjects were recruited through the local newspaper, the OHSU newsletter Web site, the newsletter of the local MS Society, and through the OHSU MS Center. Recruitment began January 1999 and the last cohort of subjects had outcome assessments in June 2002. A neurologist reviewed medical records for diagnostic criteria for MS.14 In order to create a consistent exercise and yoga intervention, we enrolled only subjects with an EDSS of 6.0 or less, i.e., able to walk 100 meters with at most unilateral support.15

Prospective participants were screened for other major medical problems with medical history, physical examination, and EKG to ensure the safety of the intervention and to exclude subjects with an underlying medical illness that may impair cognition. We excluded subjects with: insulin-dependent diabetes; uncontrolled hypertension; liver or kidney failure; symptomatic lung disease; alcoholism/drug abuse; symptoms or signs of congestive heart failure, ischemic heart disease, or symptomatic valvular disease; or corrected visual acuity worse than 20/50 binocularly. Color vision was intact to color dot perception on the Stroop with 100% accuracy. We excluded subjects if they had performed yoga or tai-chi in the last 6 months or were regularly performing aerobic exercise more than 30 minutes per day. Subjects spoke English as their primary language.

Initially, as for many studies with cognitive outcome measures, we planned to not include subjects taking any medications known to affect CNS function or subjects with significant psychiatric diseases including major affective disorder. However, these exclusions were immediately eliminated since we were unable to recruit subjects with these exclusions. This recruitment problem was not dissimilar to the mentioned difficulties of a recent multicenter trial of donepezil in MS.16 We simply encouraged subjects to minimize changes in CNS-active medications (e.g., modafinil and antidepressants) during the course of the study. For CNS-active medications taken on an as needed basis, subjects were asked to not take the medications within 24 hours of the assessments. The original targeted enrollment was 150 subjects with a projected power of 0.8 for the Stroop test. This was based on an estimated moderate effect size and a reasonably low dropout rate. However, it was decided to stop the enrollment at the halfway point, prior to any analysis, because these drug and psychiatric issues would add significant variance to the primary outcome measures even with, at best incomplete, statistical correction. To maintain adequate power for the primary outcome measures once these exclusions were removed would have required randomizing significantly more subjects than feasible with the budget.

Interventions.

Yoga classes were 90 minutes in duration once per week. The yoga class was set up following discussions among certified Iyengar yoga teachers and a neurologist. The details of the design of the yoga class have been previously reported.17 Briefly, the modifications to a usual Iyengar yoga class had to take into account fatigue as well as spasticity and cerebellar dysfunction. Essentially all poses were supported, either with a chair or having the subject on the floor or against the wall. Within that framework, 19 poses were instructed, although not all each week. The sequence of poses minimized exertion in getting up or down. Each pose was held for approximately 10 to 30 seconds with rest periods between poses lasting 30 seconds to 1 minute. Participants were encouraged to honor individual limits and hold the pose for less time if necessary. All poses were adapted to suit individual needs and modifications of some of the poses were taught for periods of lowered ability, e.g., during an exacerbation. There was an emphasis on breathing for concentration and relaxation during the session. Each class ended with a 10-minute deep relaxation with the subject lying supine. Progressive relaxation, visualization, and meditation techniques were introduced during this time. Daily home practice was strongly encouraged. Subjects were given a booklet demonstrating the specific poses practiced to assist in their home practice.

The aerobic exercise intervention arm was directed by a physical therapist with extensive experience with the MS population. The intervention was analogous to the yoga intervention with one class per week along with home exercise. The aerobic exercise consisted of bicycling on recumbent or dual-action stationary bicycles. The weekly exercise class began and ended with about 5 minutes of stretching of cycling muscles. Participants were instructed to stretch to a gentle pull but not to the point of pain and hold it for 15 to 30 seconds while breathing. Subjects monitored cycling intensity using the modified Borg Rate of Perceived Exertion scale.18 Subjects were instructed to exercise at the 2 to 3 or very light to moderate intensity on the scale, i.e., they were able to converse during the sessions. There was no monitoring of heart rate. Although cycling was the usual mode of exercise, periodically participants were given the option of exercising on a Swiss ball. Occasional variety was provided by batting a balloon among participants while cycling and adding some arm, trunk, and balance work. Subjects continued bicycling until they were ready to stop because of fatigue, onset of other MS symptoms, or they reached their personal goal (e.g., 1 hour for several subjects). Subjects were given an exercise bicycle for home use if they did not already have one. Subjects were encouraged to exercise regularly at home (on bicycles and any other modes of exercise of choice) in addition to the weekly class session.

Compliance with the interventions was assessed by study participants daily filling out 2-week log sheets that recorded whether they exercised or practiced yoga and for how long. Class attendance was also recorded.

Assessments.

After screening medical history, physical examination, and routine EKG, baseline assessments of outcome measures were performed. Baseline assessments were performed before subjects were randomized and occurred 1 to 30 days before the classes started. The practical limitations on the rate of testing participants necessitated three separate cohorts of subjects. Each cohort contained 23 subjects, each of whom was allocated to one of the three intervention arms. Outcome assessments were done at baseline and 6 months. There was also a 3-month visit although not all outcome measures were obtained at this mid-study visit and these data were not included in the analysis. On the baseline visit, demographic data were recorded and the oral reading on the Wide Range Achievement Test (WRAT) 3rd edition19 was administered to assess equality of educational achievement in the three intervention groups.

It was important to plan carefully to maintain blinding of the assessors generating the outcome measures, since the subjects were non-blinded. Only a single liaison person who was responsible for direct phone calls to subjects was unblinded, and this person did not participate in one on one testing after randomization, i.e., after the baseline assessment. The 3- and 6-month assessments were planned carefully to ensure continued blinding. Some of the outcome measures were done independently of any assessor (self-rating forms). For the in-person evaluations, the liaison person scheduling the appointment instructed the subjects to not tell the assessor what intervention group they were in. A reminder call the day prior to the assessment was made and subjects reminded to not speak about their intervention group. Even with these precautions there were rare instances of unblinding. However, the assessments were objective and many were computer based and scored. The blinded research staff maintained equipoise about potential results of the study. The data analysis was blinded to intervention group.

No assessments were performed within 50 days of an exacerbation, defined as new, recurrent, or worsening neurologic symptoms present for more than 24 hours, documented by neurologic examination and not associated with a febrile illness. Subjects with MS with a recent exacerbation were retested as soon as they were more than 50 days post exacerbation onset. During this time, they continued the intervention to which they were initially randomized to whatever degree they could.

The baseline and outcome sets of cognitive assessments were performed at the same time of day for each subject, at their preferred time either in the morning or afternoon. The subjects were instructed to refrain from alcohol consumption for 24 hours prior to the testing. They were allowed to ingest their usual dose of morning caffeine.

Cognitive measures.

The cognitive assessments focused on aspects of attention (focusing attention, shifting attention, dividing attention, and sustaining attention) that may be impaired in subjects with MS and were also thought to be most likely to be improved with the intervention. The Stroop Color and Word Test20 color-word interference was used as a measure of ability to focus attention. The covert orienting of spatial attention task compares reaction time (RT) when targets are validly cued, neutrally cued, invalidly cued, or not cued.21–23⇓⇓ Median RTs were calculated for the four cue conditions and the change in RT to validly cued circles from the beginning to last quartile was used as a measure of mental fatigue or vigilance. The attentional shifting task utilized was adapted from that used in the Cambridge Neuropsychological Test Automated Battery and is related to the Wisconsin Card Sorting Test. It allows attentional shifting to be broken down into three types: intradimensional, reversal, and extradimensional.24 The outcome measure was the percentage of error trials and numbers of shifts correctly performed prior to completion. A modified Useful Field of View task was chosen as a divided attention test since it has been used in MS and has ecologic validity in relationship to driving ability25,26⇓ and has been previously shown to be changed in people with MS.27 Our modification determined a precise temporal threshold. Simple visual RT was measured at the beginning and end of the test session. The difference in median RT between the end and beginning of the session was used as a measure of vigilance or mental fatigue. The Paced Auditory Serial Addition Test (PASAT) was also administered.28,29⇓ In case there were effects of the intervention on alertness and attention, we performed other cognitive tasks to determine the specificity of the effect: the Wechsler Memory Scales III Logical Memory30 (delayed memory adjusted for immediate recall) and the Wechsler Adult Intelligence Scale III Similarities.31

Alertness, mood, fatigue, and quality of life.

Alertness was measured with two subjective scales, the Stanford Sleepiness Scale (SSS)32 and the Profile of Mood States (POMS)33 subscales, and an objective measure based on EEG frequency analysis as we have previously done.23 The only EEG frequency analysis measures used for this analysis were posterior median power frequency and relative alpha activity recorded from the eyes closed rest and eyes closed attentive state. We have previously shown these alertness measures to be sensitive to drug effects.23,34⇓ The SSS was administered at the beginning and end of the cognitive testing sessions. Subjects took the POMS once at the end of the session. We used the POMS mood subscales as well since it has been reported to show improvement with an exercise intervention in MS study5 and with a Hatha yoga intervention in young adults.8

Mood, including measures of fatigue and vigor, was assessed using the POMS.33 Fatigue was also assessed using the Multidimensional Fatigue Inventory (MFI).35 Depression was assessed by the POMS and the CESD-10.36 Stress was assessed using the State Trait Anxiety Inventory (STAI).37 Health-related quality of life was assessed by the Short Form (SF)-36.38 The MFI, POMS, STAI, and SF-36 were filled out by the subjects at home and reviewed by the research assistant at the time of the cognitive testing to help minimize the duration of the assessment session.

Physical measures.

As part of the standard MSFC, 25-foot timed walk and the 9-Hole Peg Test were performed. Subjects performed a measure of forward bend flexibility: the chair sit and reach.39 Subjects were asked to stand as long as they could on one leg with and without their eyes open. Data from subjects requiring an aid to stand were eliminated from this particular outcome analysis.

Data analysis.

The analysis used all randomized subjects who completed the 6-month study and no attempt was made to impute missing variables.

Comparisons of the baseline factors age, MSFC, and EDSS among the three intervention groups were done using an analysis of variance. The outcome data were analyzed using an analysis of covariance (ANCOVA) approach with baseline value as the covariate, indicator variables for each of the two active groups (i.e., yoga and exercise), and the interactions of the indicators with baseline. In addition, three baseline factors (age, EDSS, and sex) were evaluated as potential confounding variables. The numeric values of age and of EDSS were included in these models rather than grouping the numeric values into categories.

The following approach was used to determine the “best” ANCOVA model for each response. Backward variable elimination determined which among baseline, age, EDSS, and an indicator for age were significant predictors. Any predictor significant at 0.10 was included in the next stage of model fitting. Second, indicator variables for the two active groups and the interactions of these indicators with baseline were added to the best model above. Partial F-tests were used to test whether the two interaction terms were simultaneously equal to zero. If the interactions were not significant, partial F-tests were also used to simultaneously test whether the two group indicators were simultaneously equal to zero. If either hypothesis was rejected, backward elimination was used to eliminate any individual terms that were not significant (using a significance level of 0.05). Residuals from the best model were assessed to determine if normality was violated (in particular if there was substantial skewness). If so, one or more transformations (the natural logarithm, square root, or the rank transformation, in order) were evaluated by following this same approach.

The primary outcome measures were assessment of alertness based on EEG median power frequency and color-word interference on the Stroop Color and Word Test.20 No Bonferroni adjustments were made for multiple outcome measures. The secondary measures were the rest of the cognitive assessments, self-rated scales (MFI, POMS, CESD-10, SF-36), and the physical measures.