John DiNardo 139been able to determine, the literature has not concerned itself with the problem of
non-random attrition. One might expect that a patient who fails to improve after
stopping the offending medication might be more likely to drop out than one who
has improved (for some reason possibly unrelated to treatment). One possible rea-
son for improvement might be mere “regression to the mean.” For evidence that
strongly suggests this is a problem see Whitney and Von Korff (1992).
3.6.2.3 First criticism
In one of the first (and extremely rare^50 ) criticisms of research in this area, Fisher
(1988) observed:
As I understand it, analgesics beget more headache by making more brain sero-
tonin available, which paradoxically increases the pain. My question is whether
it holds only for headache or for other pains as well? In our arthritic clinic aspirin
was used in doses of 8 to 14 tablets a day for about 15 years. I asked physicians
who attended that clinic in those years whether they had ever noticed [the devel-
opment of headache pain]...in any of the patients and they never had. Also 3 to
5 million people in the United States are taking aspirin daily to prevent arterial
thrombosis. Should we expect headache...under these circumstances?^51As Fisher understood, the answer to his questions was “no” and advocated a ran-
domized trial on the effect of withdrawal from headache medications where the
control group would be subject to sham double-blind withdrawal. One can think
of this as a proposal for a “severe test.”
The response in the literature was to maintain that the theory was, in the main,
correct and to merely amend the theory to accommodate the troubling fact high-
lighted by Fisher.^52 A typical amendment stated that “‘analgesic abuse headache’
may be restricted to those patients who are already headache sufferers [and
that] individuals with...migraine, are predisposed to developing chronic daily
headache in association with regular use of analgesic” (Bahraet al.,2003).
3.6.3 Redefining MOH to avoid a severe test
The process of defining MOH provides a clear example of researchers avoiding
a severe test. A useful place to start is theinitialInternational Classification of
Headache Disorders – 2 (ICHD-2).^53 The initial ICHD-2 definition of “medication
overuse headache” is displayed in Table 3.3 (Headache Classification Commit-
tee of the International Headache Society, 2006).^54
A key aspect of the definition is criterion C: the patient’sdecisionto continue
using analgesics at more than the approved rate.^55 This was immediately recog-
nized to be a problem: existing standards of treatment for other forms of migraine,
such as “menstrual migraine,”requiredthe use of analgesics at a rate which could
then (inappropriately) be described as MOH.^56
Another important aspect of the definition of MOH that was the subject of great
dispute was item D – the requirement that, after removing the patient from the