|
|
Limitations
of Crossover Studies
Despite
their potential to provide greater statistical power, crossover
designs may be inefficient in certain situations:
- Persistence
of effect of first treatment. If
treatment-induced changes
in the major outcome are not soon reversed when treatment
is withdrawn, crossover designs are inappropriate.
- Rapidly
changing underlying disease. Changes
in the underlying disease over time may introduce great
variability into patient responses, thereby undermining
the major potential advantage of the crossover design.
This necessitates that the total duration of the crossover
study be short enough to ensure that such within-patient
variation will be less than the variation that already
exists between the patients enrolled.
- High
dropout rates. Because
of the added length of crossover studies, changes in the
underlying disease, as well as logistical factors and
voluntary withdrawals, usually cause a higher dropout
rate than in parallel group studies. Although the greater
power of the crossover approach may compensate for a higher
dropout rate, reviewers may doubt the general applicability
of the results of a study completed by a minority of the
patients entered.
Experience with one or two
crossover studies in the population of interest will help
the investigator predict whether a crossover design will
improve efficiency and can suggest its optimal length.
Question
7.1
Would the use of a crossover
design be likely to offer advantages over a parallel group
design in comparing the following treatments to placebo?
 | An antidepressant
drug for the treatment of major depression? |
 
 | A quickly-eliminated
oral drug that blocks pain transmission in patients
with diabetic neuropathy. |
 | A two-week
treatment involving slow titration of a potential morphine
potentiator in patients with advanced cancer. |
 | Postoperative
pain. |
|