A total of 39 articles appear to have been included (995 subjects). It was nor clear if all these studies were included since some may have been subsequently omitted using Tukey's procedure.
Publication bias and the fail safe N value indicated lack of serious publication bias.
Inter validity rater agreement ranged from 70% to 85%. Inter coder agreement on data extraction was about 98%.
Treatment effect measured by HDRS within morning light group: Strong light was significantly more effective than medium light (P < 0.05) and medium light was significantly more effective than dim light (P < 0.05) in treating SAD. Effect sizes were as follows: dim light 1.13 (95% CI: 0.82, 1.45); medium light 1.74 (95% CI: 1.56, 1.93); and strong light 2.94 (95% CI 2.30, 3.58).
Treatment effect measured by AS within morning light group: no statistically significant difference was present between light intensity groups. Effect sizes were as follows: dim light 1.32 (95% CI 1.03, 1.60); medium light 1.41 (95% CI: 1.15, 1.68); and strong light 1.29 (95% CI: 0.94, 1.63). Exclusion of a dawn study showed no statistically significant difference between the three mean effect sizes.
Treatment effect measured by HDRS within morning-evening light group: medium light produced a significantly larger effect size than dim light. Effect sizes were as follows: dim light 0.60 (95% CI 0.22, 0.97); medium light 2.09 (95% CI: 1.81, 2.38).