Computerized and manual searches of the literature were used to locate published reports that met the following criteria: (1) randomized controlled trial design, (2) blood pressure as a primary outcome, and (3) studies that compared the effects of the TM program with those of other behavioral control interventions. Two trials met the search criteria and provided data for the present pooled analysis.6–8
The methods of the trial by Alexander et al8
have been described in detail. In brief, the 77 white participants (18% men and 82% women; mean age 81 ± 9.3 years) were residents of homes for the elderly in the Boston area. Baseline systolic blood pressure averaged 138 ± 15.9 mm Hg. Outcome measurements in the original study were cognitive functioning, mental health, and systolic blood pressure after 3 months and mortality rate after 3 years. Subjects were randomized to 4 treatment groups: a TM group (n = 21), a mindfulness training group (n = 23), a mental relaxation group (n = 22), and a usual-care control group (n = 11). In this study, the 3 active interventions (TM, mindfulness training, and mental relaxation) were matched to each other for teaching format, instructional time, home practice requirements (20 minutes 2 times a day), and expectancy of beneficial health outcomes. None of the interventions required other changes in lifestyle or personal beliefs.
The distinctive features of the 3 active treatments were as follows, The TM program was described as the principal mind-body technique of Maharishi Consciousness-Based Health Care, a comprehensive traditional system of natural medicine derived from the ancient Vedic tradition.11,12
The TM program was reported to include a simple, yet precise technique whereby the ordinary thinking process becomes quiescent and a unique psychophysiologic state of “restful alertness” appears to be gained.13
Details of the instructional protocol have been published.14
Mindfulness training, developed by Alexander et al,8
involved learning a guided attention technique to engage participants in novel and creative mental activities.8
The mental relaxation technique was designed to simulate generic features of the TM program, hypothesized by some researchers to be responsible for its relaxation effects.15
The mental relaxation technique required subjects to repeat silently to themselves familiar and brief mental stimuli (e.g., a phrase or verse) during the relaxation period. Home practice compliance rates after 3 months for each of the 3 active treatment groups were 80% for TM subjects who practiced regularly compared with 48% for the mindfulness training group and 47% for the mental relaxation group. Systolic blood pressure was significantly lower in the TM group than in the control group at 3 months.
Methods of the second trial have been published in detail.6,7
Briefly, this was a randomized controlled trial of community-dwelling older African-Americans (43% men and 57% women; mean age 67 ± 7.6 years) who had stage I and II hypertension (stages I and II; baseline blood pressure averaged 147/92 ± 13.2/8.7 mm Hg) and lived in Oakland, California. The 125 subjects were randomized to a TM group (n = 40), a progressive muscle relaxation group (PMR; n = 42), or a health education control group (n = 43). The 2 active interventions (TM and PMR) were matched for expectancy of benefits, teaching format, instructional time, and home practice (20 minutes 2 times a day). The TM program was taught according to the same standardized procedures as in the Alexander et al8
trial. The PMR technique followed previously standardized and validated procedures of Bernstein and Borkovic,16
based on a classic muscle relaxation program by Jacobson.17
This technique involved systematically relaxing the major muscle groups of the body with the goal of physical and mental relaxation. The health education control group received lifestyle modification information consistent with national guidelines.18
The education control intervention was designed to reflect usual-care community practice. Original outcomes in this study included systolic and diastolic blood pressures and other cardiovascular risk factors for cardiovascular disease (diet, exercise, weight, and substance use). The duration of the intervention was 3 months. Home practice regularity rates for TM and PMR after 3 months were 97% and 81%, respectively. There were no significant changes in diet, exercise, or blood pressure medications. Systolic and diastolic blood pressures were significantly lower in the TM group than in the PMR and education control groups.6,7
Determination of mortality:
Data from 202 subjects from the 2 clinical trials were pooled.19
Vital status was obtained from the National Death Index (National Center for Health Statistics, Research Triangle Park, North Carolina), which provided an estimated accuracy of 93% to 98% in identifying vital status of patients.20
The database of subjects’ records was searched against the National Death Index database through December 31, 1998, the most recent year available as of autumn 2000. There were 101 mortalities. Causes of deaths were obtained by using the National Death Index Plus service, which provided coded causes of death according to the International Classification of Diseases
, Ninth Revision. Cause-specific mortality was based on the underlying cause of death. Codes 401 to 438 were defined as cardiovascular diseases, and codes 140 to 239.9 were defined as cancer.
Survival analyses were performed on the pooled individual patient data from the 2 trials. This approach used meta-analysis of individual patient data.19
Survival time was measured starting from the date of randomization in the original studies. The primary outcome was all-cause mortality after completion of maximum follow-up. Secondary outcomes were mortality due to cardiovascular disease and cancer. Maximum follow-up periods were 18.8 years in trial 1 and 9.0 years in trial 2. In addition, secondary analyses were conducted of survival during 3 years (for comparison with the short-term published findings from trial 1) and during 10 years.
Statistical comparisons were made across the following treatment categories: the TM program (used in the 2 trials), “other active behavioral treatments” (mindfulness training and mental relaxation in trial 1 and PMR in trial 2), and the “usual-care” conditions (usual care in trial 1 and education control in trial 2). The usual-care groups constituted minimally controlled conditions for comparison with the 2 active treatment categories. Comparisons of the TM program with the other active behavioral therapies allowed control of nonspecific therapeutic effects, such as motivational factors and generalized effects of relaxation therapies (e.g., effects of sitting with eyes closed and novel deployment of attention). This was feasible because all active therapies were taught and practiced in structurally similar formats.
Survival analyses were performed with the SAS statistical package (SAS Institute, Cary, North Carolina). Data were checked for validity of the proportional hazards assumption.21
This was done using a time-dependent covariate, Z × ln(t), where Z is the indicator variable for TM, and t is the time since randomization, in Cox’s regression analysis of survival times. The time-dependent covariate was highly statistically significant in this analysis (p <0.0001), indicating that there was a significant violation of the proportional hazards assumption. Therefore, survival times of treatment groups were statistically compared using the generalized Wilcoxon’s test that has been recommended for analysis of nonproportional hazards.21
Stratification on study location was used to adjust for demographic differences between studies. The effect of treatment on survival was quantified as relative risk that was estimated from Cox’s regression analysis of survival times. Two-sided p values ≤0.05 were considered statistically significant. Survival analyses were conducted with “intention-to-treat” and “on-trial” principles. Subjects who were randomly allocated and completed post-testing at 3 months were included in the on-trial analyses. Missing data were handled in the intention-to-treat analysis by the method of multiple imputation.22