Randomised controlled trial of homoeopathy versus placebo in perennial allergic rhinitis with overview of four trial seriesCommentary: Larger trials are needed

If double blind trials really show homeopathic doses to
be effective, physicists and chemists should fall over
themselves trying to find out how a solution so dilute as
to contain not a single molecule of the alleged agent
can still be active. It would certainly be worth a Nobel
Prize, for it would have to involve a fundamentally new
principle of Physics.

It seems, on the face of it, more probable that an
honest mistake has been made in the experimental
design. I have one immediate suggestion. If I am right,
the way to control for it would be to perform the
successive dilution procedure separately for every
patient. It is not clear whether Taylor et al’s French
suppliers did this. I suspect not, since the process
would have been tedious and time-consuming, and
well worth mentioning alongside all the other
exhaustive precautions so laudably taken.

The 24 experimental patients received a homeopathic
dose of their specific allergens, successively diluted 30
times by a factor of 99:1 each time. This is enough to
achieve a dilution of 10 to the minus 60 -- the chance
of any one patient encountering as much as a single
molecule of active ingredient is therefore vanishingly
small. The 27 control patients received a placebo
made by passing the same diluent (30% alcohol)
through the same procedure of 30 stages of
successive ‘dilution’.

The intention was that the only difference between
experimental and control doses should be the
presence or absence of a single drop of allergen
solution in the first of the 30 successive dilution stages.
But consider the solution finally administered to the 27
placebo patients. On the assumption that it was a bulk
preparation administered to all of them, it would have
been dominated, 99:1, by the last tranche of diluent
used. One would normally assume that this was pure
diluent, just 30% alcohol. Where homeopathy is
concerned, however, things are by no means so clear.
There must be trace impurities in the tank of diluent,
and these are unlikely to be completely homogenously
distributed. Trivial as this heterogeneity might seem to
an ordinary chemist, its effects will be massive when
compared with homeopathic dilutions of the allergens.

The upshot is that different ‘brews’ of placebo are likely
to be subtly different (and not so subtly, by homeopathic
standards) from each other. If all patients were given
their own individual brew of successively diluted
solvent, the only consistent difference between
experimental patients and controls would indeed be the
single drop of allergen which initiated the process for
experimentals. In fact, however, if all control patients
were given a dose taken from the same ‘brew’, the data
obtained from different patients is not independent. All
would have received the same ‘batch’ of whatever trace
impurities were heterogeneously distributed in the last
tranche of diluent. This vitiates the
independence-of-data criterion for statistical
comparison.

My criticism is quantitatively mitigated by the fact that not
all experimental patients received the same
homeopathic dose: each was given an individually
tailored allergen. To that extent, separate dilution
series must have been run. The authors do not say
how many separate allergens were used. But the mere
fact that (as I presume) only one batch of placebo was
prepared, is enough to invalidate the conclusion.
Admittedly, normal chemists are quite happy to
assume that successive samples from the same bottle
are the same substance. But we are talking
homeopathy here, and successive samples from the
same bottle are likely to differ from each other by much
more than a homeopathic dose differs from a control.

Richard Dawkins