Effects of study design and allocation on self-reported alcohol consumption: randomized trial

Background: What participants think about the nature of a study might affect their behaviour and bias findings. We tested two hypotheses: (1) participants told they were in an intervention trial would report lower alcohol consumption at follow-up than those told they were in a cohort study; (2) participants told they were in the intervention group in a trial would have lower alcohol consumption at follow-up than those told they were in the control group. Methods: Students from four universities (N = 72,903) were invited to participate in a 'research project on student drinking'. Of 10,415 respondents, 6,788 were moderate to heavy drinkers and were randomized. Group A ('cohort') were informed their drinking would be assessed at baseline and again in one month. Group B ('control') were told the study was an intervention trial and they were in the control group. Group C ('intervention') were told the study was an intervention trial and they were to receive the intervention. All were assessed and directed to read identical online alcohol education material. Whether and how long they accessed the material were recorded. One month later, alcohol intake was reassessed. Results: In relation to hypothesis 1, there were no differences between the groups on the prespecified outcome measures. In relation to hypothesis 2, there were no differences though all point estimates were in the hypothesized direction (that is, 'intervention' < 'control'). The 'cohort' and 'control' groups accessed the material to a similar extent (59% versus 57%) while the 'intervention' group were more likely to access it (78%) and to read it for longer (median 35 s (25th and 75th percentiles: 6, 97) versus medians of 7 s (0, 28) and 8 s (4, 42) for the 'cohort' and 'control' groups, respectively). Conclusions: Although the context given to the research participants significantly influenced access to the online information and reading time, this did not translate into any effect on drinking behaviour, for either hypothesis. This might be because of failure in the experimental paradigm or the possibility of weaker effects using the online approach.