Guide-linked quiz · Medium Level
Synthetic Control Arms in Clinical Trials Quiz
Tests external controls versus randomization, prespecification, comparability, confounding, and limits of AI matching in trials.
Question 1 of 8
Which description matches an external control arm?
Preview every question8 questions
Answers and explanations appear as you take the quiz.
- Which description matches an external control arm?
- A second randomized group in the same trial
- A model-generated outcome with no patient data
- An automatic replacement for eligibility criteria
- A comparator built from people outside a concurrently randomized control group
- Why prespecify the external control before outcomes are known?
- To make eligibility criteria unnecessary
- To reduce the risk of selecting data that favor a desired result
- To guarantee a positive trial
- To avoid defining endpoints
- What should an external dataset match as closely as possible?
- Only the trial’s software vendor
- A matching number of spreadsheet columns
- Eligibility, treatment context, outcome definitions, and follow-up
- Only the same country code
- What can AI matching not fix?
- Unmeasured confounding or variables absent from the data
- Differences in observed measured covariates
- A documented data source
- A prespecified endpoint
- Which condition supports using an external control?
- A high similarity score alone
- A completed single-arm trial with visible results
- A larger synthetic cohort regardless of quality
- A justified disease context and suitable source data for the inference
- What do propensity methods balance?
- Measured factors, subject to overlap and data quality
- All possible causes of treatment outcome
- Every unrecorded variable
- The results after the trial is complete
- Why maintain an audit trail?
- To hide missing data
- To avoid prespecifying methods
- To document source selection, exclusions, transformations, and analyses
- To change the comparator after results are known
- Does a synthetic control automatically equal randomized evidence?
- No external comparison can ever be useful
- No; evidence depends on comparability, bias, and the specific inference
- Yes, if the model is complex
- Yes, if the cohorts have equal size
Keep testing yourself
More quizzes picked for your level.