CCDM Statistical Analysis 5 — Questions and Answers
Question 1: Which sensitivity analysis approach tests whether study conclusions hold when missing data are assumed to be Missing Not at Random (MNAR)?
- Complete case analysis
- Pattern mixture models or tipping point analysis (Correct answer)
- Multiple imputation under MAR
- Last observation carried forward
Correct answer: Pattern mixture models or tipping point analysis
Pattern mixture models and tipping point analyses explore how conclusions change under MNAR assumptions to assess robustness of primary results.
Question 2: What is the purpose of a Q-Q (quantile-quantile) plot in clinical data analysis?
- To compare cumulative survival curves between two groups
- To assess whether a dataset follows a specified theoretical distribution such as normal (Correct answer)
- To visualize dose-response relationships
- To detect outliers using interquartile range cutoffs
Correct answer: To assess whether a dataset follows a specified theoretical distribution such as normal
A Q-Q plot compares the quantiles of observed data against those of a theoretical distribution; points falling on a straight line indicate a good fit.
Question 3: In a two-period, two-sequence crossover trial, which statistical effect must be tested before interpreting treatment differences?
- Site effect
- Carryover (sequence-by-period interaction) effect (Correct answer)
- Investigator bias effect
- Stratification effect
Correct answer: Carryover (sequence-by-period interaction) effect
A significant carryover effect in crossover trials means Period 1 treatment influences Period 2 outcomes, invalidating the within-subject comparison.
Question 4: The intraclass correlation coefficient (ICC) is most commonly used in clinical trials to quantify:
- The strength of association between a biomarker and clinical outcome
- The degree of agreement between raters or reliability of repeated measurements (Correct answer)
- The variance explained by the treatment effect
- The correlation between primary and secondary endpoints
Correct answer: The degree of agreement between raters or reliability of repeated measurements
ICC measures the proportion of total variance attributable to between-subject variability, reflecting rater reliability or measurement consistency.
Question 5: When a clinical trial reports a statistically significant result but a very small effect size, the most appropriate interpretation is:
- The result is both statistically and clinically meaningful
- Statistical significance does not guarantee clinical relevance; the effect size must be evaluated (Correct answer)
- The study was underpowered and results should be ignored
- The p-value threshold should be lowered to confirm the finding
Correct answer: Statistical significance does not guarantee clinical relevance; the effect size must be evaluated
Large sample sizes can detect trivially small differences as statistically significant; effect size measures (e.g., Cohen's d, odds ratio) contextualize clinical importance.
Question 6: Which regression diagnostic is used to identify observations that have an undue influence on the estimated regression coefficients?
- Variance inflation factor (VIF)
- Cook's distance (Correct answer)
- Durbin-Watson statistic
- Shapiro-Wilk test
Correct answer: Cook's distance
Cook's distance measures the influence of each observation on all fitted values; high values indicate potentially influential or leverage points.
Question 7: In adaptive trial designs, what is an 'interim look' primarily used for?
- Unblinding all investigators to review individual patient data
- Assessing pre-specified criteria to modify the trial (e.g., sample size, dose selection) while preserving Type I error control (Correct answer)
- Replacing the primary endpoint based on emerging evidence
- Allowing sponsors to terminate the trial for commercial reasons
Correct answer: Assessing pre-specified criteria to modify the trial (e.g., sample size, dose selection) while preserving Type I error control
Planned interim looks in adaptive designs allow pre-specified modifications based on accumulating data while controlling the overall error rate through group sequential methods.
CCDM Statistical Analysis 5
Which sensitivity analysis approach tests whether study conclusions hold when missing data are assumed to be Missing Not at Random (MNAR)?