CCDM Statistical Analysis 4 — Questions and Answers
Question 1: A non-inferiority trial is designed to show that a new treatment is:
- Significantly better than placebo
- Not worse than an active comparator by more than a pre-specified margin (Correct answer)
- Equivalent to the active comparator within a symmetric margin
- Superior to the active comparator on a secondary endpoint
Correct answer: Not worse than an active comparator by more than a pre-specified margin
Non-inferiority trials aim to demonstrate that the new treatment is not worse than the comparator by more than a clinically acceptable margin (delta).
Question 2: When analyzing ordered categorical data (e.g., none/mild/moderate/severe), which test is most appropriate?
- Pearson chi-square test
- Wilcoxon rank-sum test (Mann-Whitney U) (Correct answer)
- Student's t-test
- McNemar's test
Correct answer: Wilcoxon rank-sum test (Mann-Whitney U)
The Wilcoxon rank-sum test is suitable for ordinal data as it uses ranks rather than assuming a normal distribution of continuous values.
Question 3: In a randomized controlled trial, what is the primary role of a Data Safety Monitoring Board (DSMB) regarding interim analyses?
- To unblind randomization codes for all investigators
- To review accumulating data and recommend stopping or continuing the trial (Correct answer)
- To approve all statistical analysis plan amendments
- To adjudicate all primary endpoints
Correct answer: To review accumulating data and recommend stopping or continuing the trial
The DSMB independently reviews interim safety and efficacy data and may recommend early stopping for benefit, harm, or futility.
Question 4: Which approach controls the overall Type I error rate when testing a primary and key secondary endpoint in a hierarchical manner?
- Simultaneous testing of all endpoints at alpha=0.05
- Fixed-sequence testing, proceeding to the next endpoint only if the prior is significant (Correct answer)
- Applying the Bonferroni correction to all endpoints equally
- Using a Bayesian framework for all endpoints
Correct answer: Fixed-sequence testing, proceeding to the next endpoint only if the prior is significant
Fixed-sequence (hierarchical) testing preserves the family-wise error rate by requiring each prior test to be significant before proceeding.
Question 5: In logistic regression for a binary clinical endpoint, the exponentiated regression coefficient (e^β) represents:
- The absolute risk difference between groups
- The odds ratio for the outcome per unit increase in the predictor (Correct answer)
- The relative risk reduction attributable to treatment
- The probability of the event occurring in the treated group
Correct answer: The odds ratio for the outcome per unit increase in the predictor
Exponentiating the logistic regression coefficient yields the odds ratio, which quantifies the association between the predictor and the binary outcome.
Question 6: What is the key assumption of a mixed-effects model for repeated measures (MMRM) regarding missing data?
- Data are Missing Completely at Random (MCAR)
- Data are Missing at Random (MAR) — missingness depends only on observed data (Correct answer)
- All missing values must be imputed before analysis
- Missing data are assumed to follow a worst-case scenario
Correct answer: Data are Missing at Random (MAR) — missingness depends only on observed data
MMRM assumes data are MAR, meaning that the probability of missingness may depend on previously observed data but not on unobserved outcomes.
Question 7: In a Phase III clinical trial, what does 'blinding' primarily protect against?
- Protocol deviations due to patient non-compliance
- Bias in outcome assessment and patient behavior due to knowledge of treatment assignment (Correct answer)
- Imbalance in baseline characteristics between groups
- Errors in database entry and data cleaning
Correct answer: Bias in outcome assessment and patient behavior due to knowledge of treatment assignment
Blinding prevents knowledge of treatment assignment from influencing outcome assessment, patient behavior, or clinical decision-making.
CCDM Statistical Analysis 4
A non-inferiority trial is designed to show that a new treatment is: