Table 1.

Guidelines for statistical analysis and reporting of microarray studies for epidemiology

DoClearly state the causal contrasts being studied and how exposures were defined.Consider and discuss potential exposure misclassification and its impact on findings.
DoIdentify potential confounders using appropriate causal diagrams and adjust for them using statistical methods.Age may be an important confounder of exposure-gene expression associations.
DoDiscuss how the genomic methods used may have created selection biases.
DoUse class prediction on independent test sets and report sensitivity and specificity of predictors.
Don'tMisuse standard terminology such as “test set” or unsupervised.
DoReport measures of internal reproducibility.Report intraclass correlation coefficients for duplicates.
DoSupplement technical validation (e.g., quantitative PCR) with external validation where possible.Use publicly deposited data to confirm pathways affected by exposure.
DoDeposit data in public databases at the time of review and before publication.
  • NOTE: This table is formatted as a supplement to a table presented by Dupuy and Simons (13).