• Incorporate the use of a placebo treatment or trial to overcome the psychological effect of supplementation. It is also interesting, if practical, to add a control (no treatment) trial so that the magnitude of the placebo effect can be determined.
• Where possible, use a repeated-measures or crossover design, in which each participant acts as his own control by undertaking both the treatment and placebo trials. This is a stronger statistical design than an experimental-placebo design (two separate groups of participants who receive either the treatment or the placebo) and requires a smaller sample size.
• Randomly assign participants to treatment and placebo groups, balancing for participant features (e.g., sex, age, fitness, or training characteristics) that could interact with the treatment.
• In a crossover study, provide each of the treatments to participants in a randomized counterbalanced order to remove the effect of time or training on study outcomes. In other words, have equal numbers of participants receive all treatments in the various possible sequences of order. Allow a suitable washout period in a crossover-designed study so that the effects of a treatment are removed before the next trial begins.
• Where possible, use a double-blind allocation of treatments to remove the subjective bias of both researcher and participants. The placebo effect has been most identified in terms of participant expectations. It is not always possible to blind the treatment to participants. When this is the case, use a single-blind presentation in which key researchers who measure performance outcomes are not aware of the treatment received by participants. Blinding of the researchers will help to control the occurrence of the "halo effect," where an observer who believes an effect is likely "marks up" or encourages the performance of participants.
• Choose the sample size after considering the likely range of changes in the measurements of interest. Power analysis of changes in outcome measures will determine the minimum number of participants needed to detect changes that are worthwhile.