Twenty-five studies were included in the analyses (total number of participants unknown): 16 studies had a comparison group; the other nine studies assessed a single group before and after treatment only. Only four studies were deemed to be of adequate methodological rigour. Mean attrition rate over the study period was 14% (range 0% to 45.7%).
CBT was significantly effective at reducing overall gambling behaviour at three months (ES -0.72, 95% CI -0.96 to -0.49, p<0.0001; 20 studies), six months (ES -0.58, 95% CI -0.84 to -0.32, p<0.0001; 13 studies), 12 months (ES -0.40, 95% CI -0.7 to -0.08, p<0.02; six studies) and 24 months or beyond (ES -0.81, 95% CI 1.16 to 0.47, p<0.0001; three studies).
CBT was significantly more effective than control groups for reducing gambling behaviours when compared to no control group (ES -0.57, 95% CI -0.88 to -0.26, p<0.0001; six studies), waiting list (ES -1.06, 95% CI -1.47 to -0.65, p<0.0001; five studies) and either another treatment as control or group delivery of treatment as control (ES -0.41, 95% CI -0.69 to -0.13, p<0.01; 10 studies).
Subgroup analyses indicated statistically significant effect sizes for most outcome measures; exceptions were duration of gambling bout and SOGS (South Oaks Gambling Screen) and the gambling type of scratch/lottery cards (results were reported in the review). There were significant results for comparison of one version of CBT with another (ES -0.27, 95% CI -0.45 to -0.09; seven studies).
Meta-regression analyses reported that poorer quality studies and those with a greater percentage of men in the sample had a greater effect size. Sensitivity analyses did not significantly alter the results at six months (analyses were not possible at 12 months or 24 months and beyond). There was no evidence of publication bias.
Other results were reported in the review.