Crossover Versus
Parallel Design
While a crossover design
is intuitively appealing and has been successfully used in a variety
of fields such as pain research as explained in Chapter
1, a parallel design should be chosen for antiemetic studies.
Three "desirable"
features of the crossover design can be problematic for studies
of nausea and vomiting.
- Crossover allows
for patient preference.
- Crossover avoids
inter patient variability.
- Crossover typically
requires a smaller sample size for a given power and
significance level.
Patient preference
is challenged by the fact that a sizable proportion of cancer
patients will not change an antiemetic that is being helpful to
them and will drop off study rather than switch arms at the crossover.
This means that the second wave of assessment may be inappropriately
biased by first wave failures. And it is much more likely that
success in wave one is followed by success in wave two, than that
failure in wave one is followed by success in wave two. The assumption
of independent events required for the design is not met with
chemotherapy treatments as it may be for chronic pain or other
symptoms like hot flashes from menopause or chemotherapy.
Furthermore, since
the second course of treatment may be different from the first
course; the within-patient variability may be no less than
the between patient variability.
And, finally, since
in a crossover antiemetic trial it is advisable to test for period
and carryover effects, the planned sample size may be greater
than with a parallel study.
Additional challenges
to study integrity in a crossover design include:
- The high probability
(due to many factors) of a substantial patient loss between
subsequent chemotherapy cycles.
- The inability to
evaluate the antiemetic efficacy over multiple cycles with a
crossover design.
- The ethical and
clinical concerns about changing a patient's treatment that
has been successful in the first period.
|