The aim of this study was to examine the extent of recanting (inconsistencies in reporting of lifetime alcohol use) and its impact on the assessment of primary outcomes within a large-scale alcohol prevention trial. One hundred and five post-primary schools in were randomised to receive either the intervention or education as normal. Participants (N = 12,738) were secondary school students in year 8/S1 (mean age 12.5) at baseline. Self-report questionnaires were administered at baseline (T0) and at T1 (+ 12 months post-baseline), T2 (+ 24 months) and T3 (+ 33 months). The primary outcomes were (i) heavy episodic drinking (consumption of ≥ 6 units in a single episode in the previous 30 days for males and ≥ 4.5 units for females) assessed at T3 and (ii) the number of alcohol-related harms experienced in the last 6 months assessed at T3. Recanting was defined as a negative report of lifetime alcohol consumption that contradicted a prior positive report. Between T1 and T3, 9.9% of students recanted earlier alcohol consumption. Recanting ranged from 4.5 to 5.3% across individual data sweeps. While recanting was significantly associated (negatively) with both primary outcomes, the difference in the rate of recanting across trial arms was small, and adjusting for recanting within the primary outcome models did not impact on the primary outcome effects. Males were observed to recant at a greater rate than females, with a borderline small-sized effect (V = .09). While differential rates of recanting have the potential to undermine the analysis of prevention trial outcomes, recanting is easy to identify and control for within trial primary outcome analyses. Adjusting for recanting should be considered as an additional sensitivity test within prevention trials.Trial Registration: ISRCTN47028486 ( http://www.isrctn.com/ISRCTN47028486 ). The date of trial registration was 23/09/2011, and school recruitment began 01/11/2011.
Keywords: Alcohol; Measurement error; RCT, prevention trial; Recanting; Self-report.