Relaxation Training for Anxiety: A Ten-Years Systematic Review With Meta-Analysis

Gian Mauro Manzoni; Francesco Pagnini; Gianluca Castelnuovo; Enrico Molinari

Disclosures

BMC Psychiatry 

In This Article

Methods

The overall objective of study selection was to collect published journal articles that examined anxiety level before and after relaxation training for reduction of anxiety both in clinical and non-clinical population.

We searched the following databases: PsycINFO, MEDLINE and the Cochraine Central Register of Controlled Trials. The searches were restricted to the past ten years (1997-2007) and included the following terms: relaxation training, relaxation exercise(s), relaxation therapy, autogenic training, relaxation AND meditation, relaxation. These words were searched as key words, title, abstract, and MeSH subject heading terms. Also, citation maps were examined and the "cited by" search tools was used. These findings were cross referenced with references from reviews. Findings were limited to human adults and English language studies. We didn't consider unpublished works.

Two reviewers (GMM and FP) screened the abstracts of all publications obtained by the search strategy. Studies meeting the following inclusion criteria were selected for the meta-analysis: (a) at least one relaxation training condition (no matter if it was the object of the paper or the treatment of the control group), (b) reporting of interval or ratio data, (c) use of psychometrical questionnaires; (d) anxiety level data presented before and after relaxation training, (e) sufficient reporting of study results (e.g. means and standard deviations) to allow for effect size computation. It is important to note that some studies were both repeated measure designs (before and after relaxation training), as well as comparisons (relaxation training versus control or other conditions). A distinction was not made among studies in which relaxation training alone or in combination was compared to a comparison/control group, studies in which relaxation training was examined without a control element and observational trials. The absence of a control group was not an exclusion criterion, because effect size's calculation can be done also on pre-post modifications. However, since between group and within group analyses are methodologically different, two separated analyses were conducted.

For all papers selected, the full articles were obtained and inspected to assess their relevance, based on the preplanned criteria for inclusion. Data were independently extracted by two reviewers (FP and GM) using a predesigned data collection form: (1) number of subjects; geographic origin of the study; relaxation training type; (4) subjects typology; (5) mean age and women percentage; (6) assessment measures; (7) homework; (8) protocol length; (9) trial context; (10) summary statistics required for computation of effect sizes. Any disagreements were discussed with a third reviewer (GC).

Besides computing the total average effect size relaxation training has on anxiety (separately for controlled and non-controlled studies), also the specific average effect sizes related to the different approaches considered were computed. The relaxation methods included are: autogenic training, Jacobson progressive relaxation, meditation and Benson's technique (considered together, given their similarities), applied relaxation, a combination of two or more methods (i.e. autogenic training in combination with visualization), other techniques.

In controlled studies, comparison condition consists in waitlist, simply laying down on a relaxing chair or on a bed, non-specific relaxing activities (i.e. reading a newspaper).

Specific average effect sizes were calculated also for type of subjects, who have been divided into three large groups, since the sample wasn't broad enough to conduct an higher diagnostic differentiation. The first category is represented by volunteers (i.e. workers) or students (high school or academic). The second one is composed by patients with medical diseases (i.e. irritable bowel syndrome). The last one represents patients with psychological or psychosomatic disorders. Participants have been inserted as psychosomatics patients only if well specified.

A number of moderators were considered: age, gender, context of training (individual or group), duration (expressed in days), use of homework (repetition of relaxation exercises with or without audiotapes), psychometric questionnaire used and studies geographical provenience.

When age was not reported, it was estimated on the base of other data (i.e. the year of school). Ambiguity concerning studies sample sizes (i.e. unspecified attrition) was solved with the conservative approach of using the smallest number for which there was clear documentation.

Anxiety is the dependent variable and only subjective assessments were considered. From the studies selected for the meta-analysis, we decided to extract psychometric data mainly from three questionnaires: the Spielberger's STAI (State-Trait Anxiety Inventory),[35] the HADS (Hospital Anxiety and Depression Scale)[2,36] and the BAI (Beck Anxiety Inventory).[37] Considered studies used also other questionnaires assessing anxiety, but we chose to exclude them from the categorization because of their paucity in our sample of studies.

Effect sizes were calculated for all studies, both within and between group (where possible). This means that some researches contribute to both the between groups and the within group meta-analyses.

Between groups effect sizes for studies with control or comparison group were computed using Cohen's d.[38,39] When the necessary data were available, all effect sizes were calculated directly using the following formula: d = (M1 - M2)/S, where M1 is the mean of the treatment group, M2 the mean of the comparison group and S is the standard deviation for the pooled sample, calculated with the following formula: √ (n1-1) s12+(n2-1)s22)/n1+n2-2, where n1 is the number of subjects in the experimental group, n2 that of the control group and s is the standard deviation of groups.

For within group studies without control group, baseline scores have been used instead of control group in the above formulas.

If these data were not provided, d was estimated using conversion equations for significance tests.[40] These effect sizes may then be interpreted with Cohen's convention[38] of small (0.2), medium (0.5) and large (0.8) effects. The overall mean effect size for all of the studies combined was weighted by the variance of the studies, considering both standard deviations and subjects number.

Prior to combining studies in the meta-analysis, we assessed the homogeneity of the effect sizes.[39] Cochran's Q-statistic[41] was computed by summing the squared deviations of each study's estimate from the overall meta-analytic estimate, weighting each study's contribution in the same manner as in the meta-analysis.[39]

The fail-safe N[42] for ES was also calculated. This is a hypothetical estimator dealing with the problem of an incomplete retrieval of studies. The fail-safe N demonstrates how many file-drawer studies with an assumed ES of zero are necessary to reduce the ES of the meta-analysis to a given level.

Comments

3090D553-9492-4563-8681-AD288FA52ACE
Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.

processing....