Content uploaded by Edzard Ernst
Author content
All content in this area was uploaded by Edzard Ernst on Aug 16, 2014
Content may be subject to copyright.
Cause and effect relationships
..........................................................................................
The ‘‘dirty tricks’’ experience can play
on us
E Ernst
......................................................................................
The therapeutic effect of a medical intervention can be due to the
specific effects of a therapy. In addition, there is a multitude of other
determinants. The totality of their impact can be such that even a
treatment causing no or negative specific effects can be followed by
positive perceived therapeutic response.
‘‘At least treatment x does not
harm my patient.’’ How often
do clinicians think along these
lines? In my field of complementary/
alternative medicine, it is arguably the
most common reason for using this or
that therapy: there is usually little ‘‘hard’’
evidence to suggest harm (by ‘‘harm’’ I
mean a negative effect on the disease, not
a simple adverse effect). So, if treatment x
does not make the condition worse and
the patient is keen to try it, we may well
decide to condone its use. There is
nothing wrong with such a decision—or
is there?
If reliable data are missing, how do we
know treatment x does not worsen the
condition? For one, we have our experi-
ence. Then there is the fact that this
treatment may have been around for
decades or even centuries. And perhaps
a few observational studies are also
available—of course, this type of evidence
is not all that reliable but, in total, it must
amount to something. Perhaps we cannot
conclude that treatment x is effective, but
surely we can be quite certain that it does
not make matters worse? I beg to differ.
Experience, the ‘‘test of time’’, and
observational studies all have one thing
in common: the lack of a control. If we
want to draw conclusions about cause
and effect (and ‘‘this treatment does not
harm my patient’’ is such a conclusion)
we need a positive or negative control.
Case reports and observational data are,
by definition, open to confounding and
bias and therefore unreliable. As a con-
sequence, causal inferences are proble-
matic.
Unfortunately medicine has a long
tradition of disregarding this rather
obvious fact. Whenever in clinical practice
doctors administer a treatment, they are
likely to attribute any ensuing clinical
improvement to the specific effects of
their intervention. In other words, we
regularly make definite conclusions about
cause and effect on less than solid
grounds. So we observe some improve-
ment and we believe it was caused by our
therapy. Or we observe no improvement
but no harm either and conclude ‘‘at least
it does no harm’’.
Let’s try to get some conceptual clarity
about what really is going on in such a
situation. Figure 1 schematically depicts
the case of a patient (or a group of
patients) receiving treatment x. Over
time, the symptoms improve and we
therefore perceive a therapeutic effect.
The assumption therefore is that this
‘‘perceived therapeutic effect’’ is due to
the specific effect of the intervention.
0
5
10
15
20
25
30
35
40
45
50
Symptom
Time
PTE = perceived therapeutic effect
Treatment period
PTE
Figure 1 Schematic analysis of a typical treatment situation.
Figure 2 Schematic differentiation of factors contributing to the perceived therapeutic effect.
PERSONAL VIEW 287
www.postgradmedj.com
In reality, the ‘‘perceived therapeutic
effect’’ can be caused by a multitude of
effects. Figure 2 shows schematically the
range of factors which can be involved. It
is easy to see that, even if the specific
therapeutic effects were negative (ie, the
treatment is harmful), the total perceived
therapeutic effect could still be positive. It
follows that ineffective and even harmful
interventions can be falsely associated
with overall improvement. In other
words, the fact that our patient gets
better or not worse does not mean the
treatment was effective or harmless.
This analysis is, of course, only sche-
matic and therefore has its limitations.
But it outlines how complex cause–effect
relationships can be in clinical medicine. I
believe that conceptual clarity is essential
for recognising what ‘‘dirty tricks’’ experi-
ence can play on us. If nothing else, it
teaches us to be (self) critical and to insist
on reliable evidence—that is, on results
which rigorously control for the multi-
tude of confounding factors and biases.
Postgrad Med J 2007;83:287–288.
doi: 10.1136/pgmj.2007.057521
Correspondence to: Prof Edzard Ernst,
Complementary Medicine, Peninsula Medical
School, Universities of Exeter & Plymouth, 25
Victoria Park Road, Exeter EX2 4NT, UK; Edzard.
Ernst@pms.ac.uk
Competing interests: None
288 PERSONAL VIEW
www.postgradmedj.com
