| |
Intent to Treat
Analyses
As we have previously
discussed, intent to treat primary analyses are a virtual requirement
for publication in peer-reviewed journals, again, because they
are least likely to be biased and are usually conservative in
the magnitude of any treatment effects measured.
There are newer alternatives
to intent to treat analyses (Goetghebeur
and Loeys, 2002), and many investigators now believe it is
acceptable to present analyses that show both 1) intent to treat
and 2) "per the protocol", that is, analyses reflecting
what patients actually received (Peduzzi et al., 2002; Whitney,
C.W. and Dworkin, S.F. Practical implications of noncompliance
in randomized clinical trials for temporomandibular disorders.
Journal of Orofacial Pain, 11:130-138, 1997). In this way, the
readers can decide for themselves the effect of participants not
completing the original treatment that was assigned.
Even so, the preferred
approach is to do everything possible to maximize retention of
all patients who were randomized into receiving the correct interventions,
collect data on them through all the follow-ups, and report intent-to-treat
analyses.
|