Thirteen studies (n=397) were included in this review: 11 used random assignment (n=343) and two (n=54) used matching to allocate the groups; nine studies addressed a specific phobia.
VRET versus control: random-effects analysis produced a a mean overall effect size of Hedge's g = 1.08 ( 95% CI: 0.80, 1.35) and of Cohen's d = 1.11 (95% CI: 0.82 - 1.39), indicating a large effect size in favour of VRET compared with control conditions.
When analysed by outcome categories, all comparisons indicated either a medium or large effect size for VRET over control conditions: general subjective distress Hedge's g =0.5 (95% CI: 0.006, 0.95) based on four studies; cognitive measures Hedge's g = 1.30 (95% CI: 0.70, 1.91) based on five studies; behavioural measures Hedge's g = 1.27 (95% CI: 0.66, 1.88) based on two studies; and psychophysiological measures Hedge's g = 0.68 (95% CI: 0.03, 1.34) based on two studies.
VRET versus in vivo exposure: VRET was more effective than in vivo exposure according to random-effects measures of both Hedge's g 0.34 (95% CI: 0.05, 0.63) and Cohen's d 0.35 (95% CI: 0.02, 0.65), indicating there was a small effect in favour of VRET.
Meta-regression found: a non-significant trend towards a dose-response relationship between number of sessions and effect sizes (p=0.06); no significant relationship between sample size and effect size (p=0.10); and no significant relationship between publication year and effect size (p=0.70).
Calculation of the fail-safe N indicated that it would require more than 231 additional studies with an effect size of 0 to reduce any of the primary findings to non-significant levels.