Components evaluation of a web-based personalized normative feedback intervention for alcohol use among college students: a pragmatic randomized controlled trial with a dismantling design.

Aims: To evaluate the effects of the two main components of a personalized normative feedback (PNF) [normative feedback only (NFO); and consequences feedback only (CFO)] compared with the full intervention (PNF) in reducing alcohol use and consequences.

Design: Three-arm pragmatic randomized controlled trial with dismantling design and 1-, 3- and 6-month follow-ups.

Setting: Web-based among Brazilian college students.

Participants: College students (aged 18-30 years) who reported alcohol use in the last 3 months (n = 5476).

Interventions: (1) Full PNF (a) drinking profile; (b) normative comparisons; (c) practical costs; (d) alcohol consequences; (e) strategies to decrease risks; (2) NFO components (a), (b) and (e); or (3) CFO components (c), (d) and (e).

Measurements: The primary outcome was change in Alcohol Use Disorders Identification Test (AUDIT) score; secondary outcomes were the number of alcohol consequences, drinking frequency and typical/maximum number of drinks. We used mixed models with multiple imputation and a pattern-mixture model to account for attrition. Subgroup analyses considered participant motivation to know more about their drinking (less motivated versus motivated).

Findings: Dismantled components reduced rather than increased AUDIT score compared to full PNF, with significant effects for NFO at 1 month [b = -0.23, 95% confidence interval (CI) = -0.46; -0.002] and for CFO at 3 months (b = -0.33, 95% CI = -0.62; -0.03). Compared with PNF, NFO reduced the number of alcohol consequences at 1 month (b = -0.16, 95% CI = -0.25; -0.06) and drinking frequency at 3 months (b = -0.42, 95% CI = -0.79; -0.05), but increased the number of typical drinks at 6 months (b = 0.38, 95% CI = 0.04; 0.72). CFO reduced drinking frequency at 3 months (b = -0.37, 95% CI = -0.73; -0.01). Attrition models confirmed all results, except for the NFO effect on typical drinks and drinking frequency. Subgroup analyses indicated superiority of dismantled components among the students less motivated in knowing more about their drinking.

Conclusions: There was no evidence that either the normative or the consequences components of a web-based personalized normative feedback intervention to reduce alcohol use and its consequences contributed to intervention effects. There was some evidence of adverse effects of personalized normative feedback, and these results were driven by 20% of participants who were less motivated in knowing more about their drinking.