Fourteen RCTs (n=3,682 participants) were included in the review. The number of participants ranged from 11 to 877 per condition per comparison. Trial quality was varied with only three reporting independent participant allocation. Intention-to-treat analysis was possible in six trials; completers’ only data was used in the remaining eight trials. Losses to follow-up ranged from 1.1 to 36.7%.
Personalised feedback interventions were associated with a significant reduction in mean alcohol consumption (mean ES (d) 0.22, 95% CI 0.16 to 0.29). The number-needed-to-treat was 8.06 and the area under the curve was 0.562, further confirming the intervention’s effectiveness. There was no evidence of statistical heterogeneity (p=0.69, I2=0).
Sensitivity analyses showed that the overall mean effect sizes were robust and did not change with different trial designs and sample sizes. Similarly, meta-regression analyses showed that effect sizes were robust and not influenced by time (p=0.12) or form of personalised-feedback intervention (p=0.22).
No evidence of publication bias was found.