Twenty-two controlled trials were included. Of these, 12 randomised controlled trials (n=623), 5 crossover studies (n=326) and 2 CCTs (n at least 62) assessed GLA. Four randomised controlled trials (n=328), of which two had a treatment arm with GLA treatment, and 1 CCT (n=48) assessed fish oils.
GLA (19 studies).
There was no significant difference in AD severity between GLA and placebo; the ES (11 trials) was 0.15 (95% CI: -0.02, 0.32).
The results among the 8 studies that reported inadequate data for meta-analysis were mixed: 2 studies found GLA significantly improved some aspect of AD compared with placebo, two found a non-statistically significant improvement with GLA, and two found no statistically significant difference between the treatments. Two studies reported no conclusions.
The meta-regression showed no influence on AD severity of study design (P=0.25), competing interest (P=0.21), age of the patients (P=0.44), duration of supplementation (P=0.33), or source of GLA (P=0.83).
Fish oil (4 studies).
There was no significant difference in AD severity between fish oil and placebo; the ES (3 trials) was -0.01 (95% CI: -0.37, 0.30).
Studies assessing component subscales generally found no statistically significant improvements with treatment compared with placebo.
The funnel plot was mildly suggestive of publication bias (under-representation of high precision studies), but Egger's test showed no evidence for publication bias (P=0.15).
The omission of each study in turn did not influence the results.