Palgrave Handbook of Econometrics: Applied Econometrics

138 Metastatistics for the Non-Bayesian Regression Runner

medication overuse was acauseof migraine pain “became well-established” in the
early 1980s (Capobiancoet al.,2001). It was first officially defined by the Inter-
national Headache Society (1988) – the international association of neurologists
with a specialty in headache – as “drug induced headache” in the International
Classification of Headache Disorders (ICHD-1) (Obermann and Katsarava, 2007).
The view that medication usecausedhead pain was developed during a period
of time when it was widely held that “migraine was a disorder of neurotic women”
(Silberstein, 2004).^45

3.6.2.2 The evidence

While a complete review of the evidence is not possible, let me take one represen-
tative example: Mathewet al.(1990).^46 Figure 3.5 is a modified version of a table
from Mathewet al.The title of the table is also from the original article. Patients
were assigned^47 to different treatment groups and their progress was observed.^48
As Mathewet al.report, the data in the figure were based only on those patients
who remained in the study – 90% in the group which continued to receive medica-
tion but only 50% in the group which had the medication withdrawn. With slight
variations, Mathewet al.’s conclusions have become standard.^49 As far as I have

90

Percent^50

remaining in

study

Percentage of Improvement in the headache indices. Note the

58% improvement in group Ib by mere discontinuation of

symptomatic medications

Group Ib symptomatic

medication discontinued

21

90

58

0 20 40 60 80 100

Percentage

improvement:

initial versus

end of three

months

Group Ia symptomatic

medication continued

Figure 3.5 The evidence in a nutshell