to define phenotype so as to maximize relevant information while ensuring as many studies can provide data as possible. In general, there is a trade off between more accurate (but more expensive and perhaps more invasive) measurements on fewer people and less accurate (but cheaper) measurements on more people. For example, although the Fagerstom Test may be a “gold standard” measure of nicotine dependence, currently only a few studies with available genome-wide genotype data have collected data on this test; on the other hand, many studies have collected information about the number of cigarettes smoked per day (a component of the Fagerstrom score). [44] To maximize sample size, investigators may agree to analyze cigarettes per day (which then raises further issues such as what scale to use; whether and how to transform the raw data; how to reconcile continuous with categorical data; etc.). Prospective meta-analyses for height, BMI, and fasting glucose have dealt with the issue of phenotype harmonization in a trait-by-trait basis. [45–48] Other consortia and projects such as the Public Population Project in Genomics (http://www.p3gconsortium.org/) and PhenX (www.phenx.org) aim to facilitate broad collaboration among existing and future genome-wide association studies by making recommendations for standard phenotyping protocols for
