CPI Data Management & Analysis 2 — Questions and Answers
Question 1: A researcher discovers mid-study that a database field was miscoded for 15% of participants. What is the most appropriate first step?
- Delete all affected records and continue
- Document the error and assess its impact on study outcomes (Correct answer)
- Revert to paper records and abandon the database
- Notify the IRB immediately and halt the study
Correct answer: Document the error and assess its impact on study outcomes
Documenting the error and assessing its impact is the first step before deciding on corrective actions.
Question 2: Which statistical method is most appropriate for analyzing survival time data in a clinical trial?
- Chi-square test
- Kaplan-Meier analysis (Correct answer)
- Pearson correlation
- One-way ANOVA
Correct answer: Kaplan-Meier analysis
Kaplan-Meier analysis is specifically designed to estimate survival functions from time-to-event data.
Question 3: Under FDA 21 CFR Part 11, which of the following is required for electronic records to be considered trustworthy?
- Records must be printed and signed in ink within 24 hours
- System must include audit trails that are computer-generated and tamper-evident (Correct answer)
- Data must be stored on servers located within the United States
- Electronic signatures must include a photograph of the signer
Correct answer: System must include audit trails that are computer-generated and tamper-evident
21 CFR Part 11 requires computer-generated, time-stamped audit trails that cannot be modified without detection.
Question 4: A dataset contains multiple measurements of the same variable collected at different time points. Which analysis approach best handles the within-subject correlation?
- Independent samples t-test
- Mixed-effects models or repeated measures ANOVA (Correct answer)
- Simple linear regression
- Fisher's exact test
Correct answer: Mixed-effects models or repeated measures ANOVA
Mixed-effects models and repeated measures ANOVA account for the correlation structure of longitudinal within-subject data.
Question 5: What does 'data provenance' refer to in research data management?
- The geographic location where data was collected
- The chronological documentation of data origins and transformations (Correct answer)
- The statistical method used to analyze the data
- The funding source that supported data collection
Correct answer: The chronological documentation of data origins and transformations
Data provenance tracks where data came from and every transformation it has undergone, ensuring reproducibility.
Question 6: Which approach to missing data is generally considered superior to complete-case analysis?
- Last observation carried forward (LOCF)
- Multiple imputation (Correct answer)
- Replacing missing values with the mean
- Removing variables with any missing data
Correct answer: Multiple imputation
Multiple imputation preserves statistical power and produces unbiased estimates under missing-at-random assumptions.
Question 7: In a data safety monitoring board (DSMB) review, unblinded interim data is shared. What is the primary data management concern?
- Ensuring data is presented in color-coded charts
- Maintaining confidentiality to prevent unblinding of the ongoing trial (Correct answer)
- Reducing the sample size to save costs
- Converting all continuous variables to categorical
Correct answer: Maintaining confidentiality to prevent unblinding of the ongoing trial
DSMB members receive unblinded data, so strict confidentiality procedures must prevent premature unblinding of investigators.
CPI Data Management & Analysis 2
A researcher discovers mid-study that a database field was miscoded for 15% of participants.
What is the most appropriate first step?