Table 1. Phenotype harmonization and analysis plan check list.

Items may be addressed concurrently and not necessarily in this order.

Establish a group and group leader

Identify a leader interested and willing to drive the process forward Include representatives from all studies and groups contributing data and any other groups (such as a coordinating center) contributing to the work

Identify common phenotypes of interest (including covariates)

Review available data—what was collected and what can be shared Review data collection forms and questionnaires Create a spreadsheet of phenotypes showing which studies collect which variables Create a web-based or other survey, if appropriate

Determine feasibility of cross-study analyses

Identify participating studies Are the subjects representative of the cohort Identify approximate numbers of subjects to be included in cross-study analyses Is the combined total number of subjects sufficient to detect a significant association Review consent status of each contributing study Do planned analyses fall within the scope of the data use statements of the consent forms used by each study Review each study's inclusion and exclusion criteria Are there studies for which the inclusion/exclusion criteria make their inclusion in the cross-study analysis not useful Review data definitions from each contributing study What exactly was asked (or measured)—compare the wording of the actual questions Are response options comparable Compare data distributions If data distributions are not similar, i.e. if there is little overlap, the data may not be comparable Confirm that the data are not being used for similar analyses by another collaboration or consortium

Prepare common definitions

Prepare phenotype definition Define the outcome of interest and values Define the outcome type—whether discrete, ordinal, continuous

Code variables

Create algorithms for converting each study's raw data to conform to the agreed upon definitions

Draft an analysis plan

Describe the phenotype, covariates, and inclusion/exclusion criteria List participating studies and anticipated number of subjects contributing to the analyses Describe the planned subgroup analyses Describe the analysis approach How will outliers be handled Will outliers be Windsorized (i.e. equated to the next highest/lowest value) Will extreme values be truncated (e.g. to 4 standard deviations of sample mean) How will missing values be handled How will covariates be adjusted for How will data from longitudinal studies (studies with data from multiple time points) be combined with data from cross-sectional studies Will a date range be chosen for which data values to include Determine which statistical model fits best; often several models are needed Identify type of analysis, i.e. meta-analyses of summary statistics or analyses of pooled individual data Describe the statistical support required and who will provide it Describe the imputation plan Who will be responsible for imputation, i.e. will this be done by individual studies or will it be done centrally Identify any candidate genes that can be examined Identify key personnel and their roles Who at each site and/or the CC will be doing analyses Who will be the primary author Will any assistance be required from the CC What will be the role of the CC Develop a timeline—genotype and phenotype data for different studies may not be ready at the same time When can analyses start and what studies will be included How will studies whose data will be ready later be incorporated into the analysis Relationship to work being done in other consortia/collaborations How do these analyses fit in with what other consortia are doing Can collaborations be established with another consortium to combine data for a primary analysis Can the other consortium's data serve as a replication study What specific associations and G*E analyses can be performed that have not been already examined