Seventy-three studies were summarised of which 60, including 35 RCTs, were included in the meta-analyses.
Pre-post AT.
The pooled estimates of the main effects in 33 RCTs (117 individual ESs) were ES 0.68 for both physiological and behavioural/psychological outcomes, and ES 0.75 (95% CI: 0.65, 0.85) overall. The fail-safe N to reduce the pooled ES to 0.2 was 91. For unspecific effects the pooled ES was 0.73. Pooling data from 19 non-randomised studies (61 ESs) showed a small ES (0.36 to 0.43) for the main outcomes; a variable ES (0.25 to 0.81) for unspecific effects was shown in 6 studies.
AT versus real controls.
The pooled estimates of the main outcomes in 25 RCTs (88 individual ESs) were ES 0.63 for physiological outcomes and 0.59 for behavioural/psychological outcomes, and ES 0.61 (95% CI: 0.49, 0.74) overall. The fail-safe N to reduce the pooled ES to 0.2 was 49. For unspecific effects the pooled ES was 0.67. Pooling data from 16 non-randomised studies (65 ES) showed a medium ES (0.62 to 0.74) for the main outcomes; a variable ES (0.00 to 1.10) for unspecific effects was shown in 5 studies.
AT versus other psychological treatments.
The pooled estimates of the main outcomes in 18 RCTs (62 individual ESs) indicated small effects in favour of other psychological treatments: ES -0.46 for physiological outcomes, ES -0.25 for behavioural/psychological outcomes, and ES -0.28 (95% CI: -0.40, -0.16) overall. There was wide variation in the pooled ES for unspecific effects (ES: -1.23 to 0.38) based on 7 trials. Pooling data from 9 non-randomised studies (42 ESs) showed no effect of AT on the main outcomes; a small ES (0.34) for unspecific effects was shown in 4 studies.
AT versus medical treatments.
The pooled estimates of the main outcomes in 4 RCTs showed no effects on physiological or behavioural/psychological outcomes. Pooled data from 3 non-randomised studies showed a large effect on the main behavioural/psychological outcome (ES 0.82), but no effect of AT on physiological outcomes.
The pattern of results for main effects was stable at follow-up for all comparisons. Meta-analyses of RCTs in psychosomatic disorders (31 trials) and psychological disorders (4 trials) showed medium effects of AT on the main outcomes compared with real control conditions, and small effects in favour of other treatments. The full report provided a further breakdown of the results by specific disease.