Data Management and Statistical Considerations
Data management and statistics ensure that what is collected can be analyzed credibly. Candidates must understand the data lifecycle - CRF design, database build, data entry, edit checks, query resolution, medical coding (e.g., MedDRA, WHO Drug), database lock, and data transfer. Know the statistical analysis plan (SAP), analysis populations (intention-to-treat, per-protocol, safety), randomization and blinding, missing data handling, and interim analyses.
Core concepts include data quality dimensions (accuracy, completeness, consistency, timeliness), audit trails and ALCOA+ principles, CDISC standards (SDTM for tabulation, ADaM for analysis), and the relationship between the protocol, the SAP, and the database. Understand that the ITT principle preserves randomization and is generally the primary analysis for superiority, while per-protocol is supportive.
Traps: confusing SDTM (submission tabulation) with ADaM (analysis-ready datasets); assuming database lock can occur with open queries; and misapplying analysis populations - for example, analyzing safety in the ITT population when the safety population is defined by treatment received. Another frequent trap is a question about missing data where the answer is "prespecify the handling in the SAP," not "impute after seeing results."
Rule of thumb: any analysis decision made after unblinding must be justified as prespecified or clearly labeled exploratory - the SAP is the contract.
Worked example: A subject randomized to Drug A received Drug B. For ITT, the subject is analyzed as randomized; for the safety population, as treated. Both analyses are correct for their respective populations.