Question 14 of 20 Intermediate Concept
When is a crossover design appropriate?
Short answer
A crossover design suits a chronic, stable condition where the drug's effect appears fully within a period and wears off completely during washout. Each participant serves as their own control, so fewer people are needed. It fails when effects carry over, the disease changes, or many participants drop out.
Why
In a crossover trial each participant is randomized to a sequence of treatments, for example drug then placebo, or placebo then drug. Because comparisons are within a person, between-person variability drops out and ICH E9 notes the sample size can fall sharply. The main threat is carryover, a lingering effect of the first treatment into the next period.
In the simple two period, two treatment design, carryover cannot be separated statistically from a treatment by period interaction, and the test for it has little power. So it must be prevented by design, not detected afterwards. E9 lists the conditions: a chronic and stable disease, an effect that develops fully within the period, and a washout long enough for complete reversal.
A cure, a disease that progresses, or an outcome like death rules crossover out. Dropouts hurt more because a participant with one period gives little information. The classic good use is bioequivalence in healthy volunteers, where washout is set from the drug's half life.
Symptom relief in stable chronic pain or asthma can also fit.
A strong answer covers
- Each participant as their own control, so fewer participants
- Carryover as the main threat
- Chronic stable disease, effect within period, adequate washout
- Not for curative treatments, progressive disease or terminal outcomes
- Bioequivalence as the standard use
Common mistakes
- Relying on a statistical test to detect carryover in a two period design.
- Using crossover for a disease that improves or worsens over months.
- Ignoring that adverse events in period two may belong to period one treatment.
The interviewer might ask next
How would you choose the washout length for a crossover study?