Statistical design considerations applicable to clinical trials of iodine supplementation in pregnant women who may be mildly iodine deficient.

No large, randomized, placebo-controlled trial of iodine supplementation in pregnant women in a region of mild or moderate iodine deficiency has been completed in which a primary outcome measure was an assessment of the neurobehavioral development of the offspring at age ≥2 y. In this article, I discuss considerations for the design of such a trial in a region of mild iodine deficiency, with a focus on statistical methods and approaches. Exposure and design issues include the ethics of using a placebo, the potential for overexposure to iodine, and the possibility of community randomization. The main scientific goal of the trial is important in determining the follow-up period. If the goal is to determine whether iodine supplementation during pregnancy improves neurobehavioral development in the offspring, then follow-up should continue until a reasonably reliable assessment can be conducted, which might be at age ≥2 y. Once the timing of assessment is decided, the impact of potential loss to follow-up should be considered so that appropriate statistical methods can be incorporated into the design. The minimum sample size can be calculated by using a sample size formula that incorporates noncompliance and assumes that a certain proportion of study participants do not have any outcome observed. To have sufficient power to detect a reasonably modest difference in neurobehavioral development scores using an assessment tool with an SD of 15, a large number of participants (>500/group) is required. The minimum adequate number of participants may be even larger (>1300/group) depending on the magnitude of the difference in outcome between the supplementation and placebo groups, the estimated proportion of the iodine-supplementation group that fails to take the supplement, and the estimated proportion of pregnancies that do not produce outcome measurements.