Rapid responses are electronic comments to the editor. They enable our users
to debate issues raised in articles published on bmj.com. A rapid response
is first posted online. If you need the URL (web address) of an individual
response, simply click on the response headline and copy the URL from the
browser window. A proportion of responses will, after editing, be published
online and in the print journal as letters, which are indexed in PubMed.
Rapid responses are not indexed in PubMed and they are not journal articles.
The BMJ reserves the right to remove responses which are being
wilfully misrepresented as published articles or when it is brought to our
attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not
including references and author details. We will no longer post responses
that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
Editor – Farmer et al (1) have cited several “crucial differences”
between a study they published (2) and our recent study on the risk of
venousthromboembolism (VTE) associated with third generation oral
contraceptives
compared with those containing levonorgestrel (3) using the same data
source, the General Practice Research Database (GPRD). We agree that it
is worthwhile to consider how the differences Farmer et al alluded to
might affect the findings of our studies.
Farmer et al (1) first point to the fact that over the years of our using
the GPRD to conduct pharmacoepidemiologic research at the Boston
Collaborative Drug Surveillance Program we have restricted our studies to
a subset of practices that have provided reliably high-quality data.
Farmer et al imply that this may introduce selection bias (at the level of
generalpractice rather than individual subjects). However, this is not a
plausible explanation for the difference between our findings. If there
were truly no
difference in the risk of VTE between the third generation oral
contraceptives and those containing levonorgestrel, our finding of a two-
fold risk in the practices we studied means that we must have excluded
practices that collectively would show an equally strong association
between VTE incidence and third generation oral contraceptives compared
with those
containing levonorgestrel but in the opposite direction from what we
found.
No one has postulated that third generation oral contraceptives might be
associated with a lower risk of VTE compared with those containing
levonorgestrel; moreover, it is difficult to see how we could have
identified practices in which such a supposed protective effect would be
observed before carrying out our study (so that these practices could be
systematically excluded). On the other hand, when the exposure under
study is dichotomous (as in our study) it is relatively easy to obscure a
true
association – that is, to produce bias toward the null – by introducing
misclassification of either the exposure or the outcome, even when such
misclassification is non-differential. To minimize this common potential
bias, we always exclude practices that we have found to provide incomplete
or otherwise inadequate data.
Second, Farmer and his coworkers (1) assert that they reviewed
individually all records of potential cases and suggest that some of the
discrepancy between our reports may be due to our “underidentification” of
cases. We
studied cases that were well documented and idiopathic (not due to an
identifiable proximal cause such as recent surgery or fracture) among
women currently exposed to one of the classes of oral contraceptive under
study.
If there were additional similar cases that we did not study, this would
not bias the effect estimate we reported, although an effect estimate
calculated from a subset of cases would be less precise (that is, it would
have wider
confidence intervals). However, our study (3) showed a statistically
significant association while that of Farmer et al. (2) did not. Hence,
an explanation of the difference between our findings must lie elsewhere.
Third, Farmer's group (1) note that we studied women aged 15 to 39 (3)
while they studied women up to age 49 (2). Again, this would not explain
why we observed an association between a dichotomous exposure variable and
the outcome (VTE) unless one supposes that there exists an equally strong
inverse association (that is, a protective effect for VTE of third
generation oral contraceptives compared with those with levonorgestrel)
among women aged 40-49, who were included in the study of Farmer et al (2)
but not ours (3). In fact, the number of women age 40 to 49 who had an
episode of idiopathic VTE while taking oral contraceptives in the GPRD is
too small for us to be able to estimate reliably the relative risk of
different classes of oral contraceptives among women in this age range.
Finally, Farmer et al (1) seem to acknowledge that their time-series
analysis (2) did not control adequately for confounding due to obesity and
smoking. As pointed out in our discussion (3), more precise control of
confounding is possible in a nested matched case-control study than in a
cohort study, especially when prescription patterns change rapidly over
time (as was the case after the Committee on the Safety of Medicines
issued its warning in 1995 about the higher risk of VTE associated with
third generation oral contraceptives). This is exactly why we conducted
and
presented the results of a matched case-control analysis in addition to
reporting a cohort study.
We leave it to the epidemiologic and medical scientific communities to
assess the relative merits of our case-control study (3) and the time-
series (cohort) study of Farmer et al (2).
James A. Kaye, epidemiologist
Catherine Vasilakis-Scaramozza, epidemiologist
Susan S. Jick, associate professor of epidemiology and biostatistics
Hershel Jick, associate professor of medicine
(1) Farmer R, Williams T, Nightingale A. Pitfalls of
pharmacoepidemiology. BMJ 2000; 321:1352
(2) Farmer RDT, Williams TJ, Simpson EL, Nightingale AL. Effect
of 1995 pill scare on rates of venous thromboembolism among women taking
combined oral contraceptives: analysis of General Practice Research
Database. BMJ 2000;321:477-479
(3) Jick H, Kaye JA, Vasilakis-Scaramozza C, Jick SS. Risk of
venous thromboembolism among users of third generation oral contraceptives
compared with users of oral contraceptives with levonorgestrel before and
after 1995:
cohort and case-control anlaysis. BMJ 2000;321:1190-1195
Competing interests:
No competing interests
01 December 2000
James A Kaye
Catherine Vasilakis-Scaramozza, Susan S Jick, Hershel Jick
Response to letter by Professor Farmer et al
Editor – Farmer et al (1) have cited several “crucial differences”
between a study they published (2) and our recent study on the risk of
venousthromboembolism (VTE) associated with third generation oral
contraceptives
compared with those containing levonorgestrel (3) using the same data
source, the General Practice Research Database (GPRD). We agree that it
is worthwhile to consider how the differences Farmer et al alluded to
might affect the findings of our studies.
Farmer et al (1) first point to the fact that over the years of our using
the GPRD to conduct pharmacoepidemiologic research at the Boston
Collaborative Drug Surveillance Program we have restricted our studies to
a subset of practices that have provided reliably high-quality data.
Farmer et al imply that this may introduce selection bias (at the level of
generalpractice rather than individual subjects). However, this is not a
plausible explanation for the difference between our findings. If there
were truly no
difference in the risk of VTE between the third generation oral
contraceptives and those containing levonorgestrel, our finding of a two-
fold risk in the practices we studied means that we must have excluded
practices that collectively would show an equally strong association
between VTE incidence and third generation oral contraceptives compared
with those
containing levonorgestrel but in the opposite direction from what we
found.
No one has postulated that third generation oral contraceptives might be
associated with a lower risk of VTE compared with those containing
levonorgestrel; moreover, it is difficult to see how we could have
identified practices in which such a supposed protective effect would be
observed before carrying out our study (so that these practices could be
systematically excluded). On the other hand, when the exposure under
study is dichotomous (as in our study) it is relatively easy to obscure a
true
association – that is, to produce bias toward the null – by introducing
misclassification of either the exposure or the outcome, even when such
misclassification is non-differential. To minimize this common potential
bias, we always exclude practices that we have found to provide incomplete
or otherwise inadequate data.
Second, Farmer and his coworkers (1) assert that they reviewed
individually all records of potential cases and suggest that some of the
discrepancy between our reports may be due to our “underidentification” of
cases. We
studied cases that were well documented and idiopathic (not due to an
identifiable proximal cause such as recent surgery or fracture) among
women currently exposed to one of the classes of oral contraceptive under
study.
If there were additional similar cases that we did not study, this would
not bias the effect estimate we reported, although an effect estimate
calculated from a subset of cases would be less precise (that is, it would
have wider
confidence intervals). However, our study (3) showed a statistically
significant association while that of Farmer et al. (2) did not. Hence,
an explanation of the difference between our findings must lie elsewhere.
Third, Farmer's group (1) note that we studied women aged 15 to 39 (3)
while they studied women up to age 49 (2). Again, this would not explain
why we observed an association between a dichotomous exposure variable and
the outcome (VTE) unless one supposes that there exists an equally strong
inverse association (that is, a protective effect for VTE of third
generation oral contraceptives compared with those with levonorgestrel)
among women aged 40-49, who were included in the study of Farmer et al (2)
but not ours (3). In fact, the number of women age 40 to 49 who had an
episode of idiopathic VTE while taking oral contraceptives in the GPRD is
too small for us to be able to estimate reliably the relative risk of
different classes of oral contraceptives among women in this age range.
Finally, Farmer et al (1) seem to acknowledge that their time-series
analysis (2) did not control adequately for confounding due to obesity and
smoking. As pointed out in our discussion (3), more precise control of
confounding is possible in a nested matched case-control study than in a
cohort study, especially when prescription patterns change rapidly over
time (as was the case after the Committee on the Safety of Medicines
issued its warning in 1995 about the higher risk of VTE associated with
third generation oral contraceptives). This is exactly why we conducted
and
presented the results of a matched case-control analysis in addition to
reporting a cohort study.
We leave it to the epidemiologic and medical scientific communities to
assess the relative merits of our case-control study (3) and the time-
series (cohort) study of Farmer et al (2).
James A. Kaye, epidemiologist
Catherine Vasilakis-Scaramozza, epidemiologist
Susan S. Jick, associate professor of epidemiology and biostatistics
Hershel Jick, associate professor of medicine
(1) Farmer R, Williams T, Nightingale A. Pitfalls of
pharmacoepidemiology. BMJ 2000; 321:1352
(2) Farmer RDT, Williams TJ, Simpson EL, Nightingale AL. Effect
of 1995 pill scare on rates of venous thromboembolism among women taking
combined oral contraceptives: analysis of General Practice Research
Database. BMJ 2000;321:477-479
(3) Jick H, Kaye JA, Vasilakis-Scaramozza C, Jick SS. Risk of
venous thromboembolism among users of third generation oral contraceptives
compared with users of oral contraceptives with levonorgestrel before and
after 1995:
cohort and case-control anlaysis. BMJ 2000;321:1190-1195
Competing interests: No competing interests