|
Cross-over
trials are trials in which patients are allocated to sequences
of treatment with the purpose of studying differences between
individual treatments (Senn, 1993).
N-of-1 studies are special cases of cross-over trials in which
the same patient is repeatedly randomised to receive either the
experimental treatment or its control (Senn,
1993). The distinction between these types of trials has more
to do with presumed purpose than with statistics, traditionally
cross-over trials have been seen as efficient alternatives to
parallel group trials for the purpose of investigating typical
effects of treatments, whereas n-of-1 trials have been developed
by medical researchers with the express purpose of extending the
methodology of the clinical trial to the treatment of the individual
patient. These distinctions, however, are less important than
has commonly been supposed. We shall look at cross-over trials
initially and then consider what further matters are raised by
n-of-1 trials.
This chapter
concentrates on practical matters to do with planning, running
and analyzing cross-over trials. Technical advice on analysis
can be found in Cross-over Trials in Clinical Research.
Further discussion will be found in Statistical Issues in Drug
Development and the Wiley Encyclopedia of Statistical Science
(Senn, 1999).
Cross-over
Trials: Examples and Uses
We start
by considering a number of illustrative examples.
Example
1: A cross-over trial was run to compare the effect of tramadol
to placebo in painful polyneuropathy (Sindrup
et al., 1999). Forty-five patients were randomized to one
of two sequences, tramadol followed by placebo or placebo followed
by tramadol. Each treatment was delivered for four weeks. Using
10-point numeric scales, patients rated pain, paresthesia and
touch-evoked pain.
| Figure
1.1: AB/BA Randomization Model
|
|---|
|
|
This is an example
of the most common sort of cross-over, often loosely referred
to as a two period design, but more accurately described by referring
explicitly to the sequences as an 'AB/BA design'. In this case
we have A as placebo and B as tramadol.
|