Letters to the editor
To the Editor:
In the July issue (Contraception 2003;68:11–17), you
published a meta-analysis by Dr. Khader et al. on the risk of
myocardial infarction associated with the use of oral con-
traception [1]. The article has a nearly one-page long de-
scription of the methods, i.e., how articles were identified by
several approaches, how two reviewers independently re-
viewed studies and how the data from 19 studies were
extracted by no less than three investigators using a standard
protocol.
Among these 19 studies I find, listed with their study
characteristics in Table 1, one study by “Frits et al.” and one
study by “Tanis et al.” In the table, these two studies have
different reference numbers and, indeed, in the reference
list, we find one by “Tanis et al.,” and one by “Frits R.”
Interestingly, the first one has only two authors, the second
one being “Rosendaal FR.” Both references cite the same
article title and the same journal and source (N Engl J Med
2001;345:1887–1793) [2].
Nevertheless, in the table the article by Frits and the
article by Tanis appear to have different characteristics: one
included women aged 18 – 49, and the other women aged
24 – 49. Under the column that gives details about adjust-
ment, it says Frits et al. adjusted, “but factors are unknown,”
while Tanis et al. adjusted for smoking, diabetes, hypercho-
lesterolemia, hypertension, obesity, family history of CVD,
education and alcohol intake. The studies apparently also
had different locations, since one (Tanis) was conducted in
the Netherlands, whereas Frits et al. performed his study in
“Nationwide.” Closer inspection of the article yields more
inaccuracies, e.g., a US article is cited to include data on
prothrombin 20210A mutation, while it only relied on in-
terviews, and the information is not consistent over the
tables.
Readers of meta-analyses rely on the integrity and accu-
racy of those who performed and reported the study. In
theory, it is easier to spot unreliable analyses in meta-
analyses than in original research, since one only needs a
library to repeat the exercise. This article shows that it may
not be so easy after all, since the independent reviewers and
abstractors did not notice that they included the same article
twice and abstracted it incorrectly, nor did the editors and
reviewers of Contraception. It is needless to say that this
particular article is completely invalidated by the inclusion
of nonexistent data, but there are also wider lessons to be
learned; first, that a meta-analysis or systematic review
should be viewed as seriously as an original study, e.g.,
when performed by junior researchers they should be ade-
quately supervised and second, that even when a label of Ia
Level of Evidence is attached to a study, readers should not
believe all they read.
Frits R. Rosendaal
Clinical Epidemiology, C9-P
Leiden University Medical Center
P.O. Box 9600
2300 RC Leiden
The Netherlands
E-mail address: F.R.Rosendaal@lumc.nl
doi:10.1016/j.contraception.2003.12.018
References
[1] Khader YS, Rice J, John L, Abueita O. Oral contraceptives use and the
risk of myocardial infarction: a meta-analysis. Contraception 2003;68:
11–7.
[2] Tanis BC, van den Bosch MA, Kemmeren JM, Manger Cats V,
Helmerhorst FM, Algra A, van der Graaf Y, Rosendaal FR. Oral
contraceptives and the risk of myocardial infarction. N Engl J Med
2001;345:1787–93.
Response to Letter to the Editor
Dear Editor:
We agree with Frits R. Rosendaal that we described
the study by Tanis et al. [1] twice using different author
names and somewhat different characteristics and refer-
ences. The correct reference is reference [21], in our
study that refers to the study by Tanis et al. (the correct
age group is 18 – 49). However, this study was used once
in our data analyses and not repeated at all. Also, it is
clear from graphs (Forest plots) that nonexistent data
were not included in our data analysis. It is obvious in our
study that the incorrect reference (i.e., reference [19])
was neither shown in the main analysis nor in the sub-
group analysis.
The mistakes that happened in: (a) enumerating the study
Contraception 70 (2004) 85– 86
0010-7824/04/$ – see front matter © 2004 Elsevier Inc. All rights reserved.