The Journal of Clinical Endocrinology & Metabolism Vol. 84, No. 6 1825-1828
Copyright © 1999 by The Endocrine Society
Gestodene, Desogestrel, and Venous Thromboembolism: A Little Risk after a Long Look
Alexander M. Walker
Department of Epidemiology
Harvard School of Public Health
Boston, Massachusetts 02115
 |
Introduction
|
|---|
FOURresearch reports appearing near the end of 1995 held that oral
contraceptives (OCs) containing gestodene and desogestrel double the
risk of venous thromboembolism (VTE) (1, 2, 3, 4). The standard-of-comparison
group comprised low-estrogen OCs with levonorgestrel. Subsequent
studies have been more mixed, but have not changed the overall picture
(5, 6). I have previously reviewed and summarized this research (7), so
in this comment, I will go over the major criticisms that have been
leveled against the original findings. My conclusion is that the
dissenting arguments are coherent, but that they fail on quantitative
grounds. That is, they do not actually account for the published
results. It follows that we should accept that low-estrogen OCs
containing desogestrel and gestodene carry a modestly elevated risk of
VTE in comparison to other low-estrogen OCs in widespread use.
|
Counter arguments
|
|---|
There are three principal arguments against a straightforward
reading of the major studies. 1) Prescription bias: doctors
preferentially give third generation oral contraceptives (TGOCs, oral
contraceptives containing less than 50 mcg of ethinyl estradiol
together with either gestodene or desogestrel) to women at high risk
for VTE. 2) Diagnostic bias: TGOC use increases the
likelihood of diagnosis of an existing VTE. 3) Changing
risk: OCs that have been on the market for a long time tend to be
associated with lower risk than more recently introduced products
because risk drops with prolonged use. I will try to describe each of
these ideas more fully below, and I will address their adequacy as
explanations for a doubling of risk.
|
Prescription bias
|
|---|
Hypothesis.The so-called third generation products are the
latest step in a decades-long advance aimed at reducing OC-induced
thrombogenesis (8, 9). Physicians might therefore be especially
inclined to use TGOCs for women at high risk of VTE. TGOC users
collectively would therefore be at higher risk of VTE, and an
unsuspecting observer might mistakenly attribute the higher risk to the
TGOCs themselves, rather than to the reasons for which the TGOCs were
prescribed.
Comment.Pharmacoepidemiologists call this phenomenon
"confounding by indication" (10). Confounding means that an
estimate of effect is contaminated by differences in the baseline risks
of compared groups. Confounding "by indication" arises when doctors
prescribe differently to high- and low-risk women.
To account for a doubling of risk confounding requires substantial
associations between the confounding factor, the choice of OC, and the
risk of VTE. If all the TGOC users and none of the other OC users were
high-risk women, the high-risk characteristic would have to entail a
doubling of VTE occurrence to account for an overall doubling of risk.
Unfortunately for the hypothesis, this kind of perfect segregation is
unlikely, and incomplete allocation of the treatments to high- and
low-risk women raises enormously the risk differential that needs to be
invoked. If the high-risk factor were present in 20% of the TGOC users
and 5% of the other OC users, the required risk increase for women
bearing the factor jumps to
11-fold.b
What characteristics could play this role? Other than advanced age, the
only clinically apparent feature to elevate the risk of new VTE by an
order of magnitude is a personal history of VTE. However, every one of
the major studies screened out women who had previously experienced a
VTE, and each one undertook statistical control for the effects of age,
so these cannot have created the needed bias. Inherited common
coagulopathies such as factor V Leiden mutation could be candidates, if
doctors were aware of them (11). But such information is generally
unavailable for asymptomatic women and therefore cannot affect
prescribing practice.
TGOCs were not by and large being given to high-risk women. OCs
containing desogestrel and gestodene were most commonly given to young,
short-term, and first-time users (12). TGOCs were also being used in
the oldest contracepting women, but this was numerically a trivial
phenomenon. The great majority of women simply continued to use their
established and apparently satisfactory contra-ceptive.
Studies of doctors’ prescribing attitudes in the wake of the first
public controversy cast an illuminating sidelight on the confounding
argument. Risk-sensitive, selective prescribing did not approach the
powerful associations required to double the apparent risk. Moreover,
physician-recalled behavior was dramatically at variance with actual
practice. German physicians correctly identified a number of settings
in which the risk of VTE might be elevated and reported that they would
have prescribed TGOCs to such women (13). A review of their clinical
records showed that there had been only a modest tendency toward the
asserted practice. English doctors replied to a similar survey the same
way that the Germans had, but their prescription choices ran modestly
to the opposite direction of what they had claimed (14). French,
Swedish, and Dutch attitudes were consonant with those reported from
Germany and Britain, but practice has not been examined (15, 16).
Danish prescribing practice, examined quantitatively, was found to
account potentially for a 15% bias in the TGOC-VTE association
(17).
The doctor attitude surveys should put us strongly on guard. Clinical
intuition may not hold the key to understanding the epidemiologic
studies. Doctors reported and may have believed in retrospect that
bright lines demarcated their prescription choices, but it was not
true. As a result, practitioners’ subjective impressions of an obvious
bias may be thoroughly off the mark.
|
Diagnostic bias
|
|---|
Hypothesis.Worried that a TGOC could cause VTE, doctors might
be especially quick to evaluate ambiguous symptoms in TGOC users. True
cases of VTE with equivocal presentation will thus be counted in the
TGOC group, but not in comparison groups, introducing a false
differential.
Comment.This hypothesis of diagnostic bias is an implausible
explanation for TGOC-specific effects, as TGOCs were believed to be
less, not more, risky than their predecessors. Physicians surveyed did
not claim to pursue TGOC users with selective diagnostic vigor (13, 14). Moreover, the bias would inflate the OC-VTE association in the
research studies only for categories classified as "possible" or
"mild." None of the studies reported stronger associations for
dubious cases than for definite ones. A recent study sought to
eliminate possible effects of diagnostic bias by matching on referring
physicians’ tentative diagnoses (18). The adjusted odd ratios for
TGOCs were 1.5 times those for other low-dose OCs, which is little
different from the findings of studies that did not match on referral
diagnoses. The hypothesis of diagnostic fails because not one of its
empirical implications has been borne out.
|
High early risks stigmatize new products
|
|---|
Hypothesis.Risk for VTE is relatively high for new OC users
and tends to decline with time. At any one time, newer OC products will
represent a larger share of the market in women who have recently begun
oral contraception. Long-term, hypothetically low-risk users will be
continuing on products introduced earlier. A comparison that does not
separate out short- and long-term use will give the impression that
newly marketed agents are riskier. Researchers have invoked a concept
drawn from infectious disease epidemiology, the depletion of a
susceptible pool, to explain the phenomenon of declining risk.
Comment.Several observations initially supported this idea.
When one of the research groups confined attention to women 25–44 yr
of age, they found that there was a regular ordering of
product-specific risks from the most recently introduced down to the
earliest (19). Analysis of first-time users (about 25% of the
population) in the same population showed that risk declined
dramatically from a first-year high, after the same time course in
users of TGOCs and other OCs (20).
Efforts to duplicate the analytic results in other studies have instead
confirmed the relative elevation of risk in TGOC users, even in the
first year of use (21). The observed ordering of risks by date of
introduction did not hold for the full population (age 15–44) in which
it was originally noted (22, 23). Previous studies have noted lower,
not higher, risks with newer OCs, suggesting that there is no general
phenomenon of risks that order themselves in line with dates of product
introduction (24). Declining risk functions therefore explain only a
small part of the available data. Even where they were observed, in a
small proportion of subjects in one study, they do not account for the
findings in the rest of the study.
Depletion of susceptibles is a deceptive analogy that often hides more
than it reveals. In order for their numbers to be depleted, a large
proportion of susceptible women must be struck down. Could the
occurrence of a few cases of VTE per 10,000 women be enough to deplete
a susceptible pool? Known markers of VTE susceptibility are far too
common in European populations to be depleted by a tiny number of VTE
episodes.
|
Conclusions
|
|---|
Reviews and reanalyses outnumber original reports of VTE in
relation to contemporary OC use. At least two editorialists have
proclaimed the end of the TGOC controversy, one finding in favor of a
risk differential, the other finding the opposite (25, 26). None of the
arguments that I have touched on here has the muscle to explain the
consistent and repeated observations from large and well-designed
studies. Yet the debate festers, straining collegial ties, even
erupting into litigation. Why? I think that the problem arises because
the medical profession—practitioners, researchers, regulators,
pundits—lacks a calculus for balancing small risks. Our intuition,
shaped by training and possibly hard-wired by evolution, seems to
require a big-risk model to stir any behavior at all. We either ignore
or over-react.
Two instances of over-reaction, in Britain and Germany, raised the
stakes of the TGOC-VTE debate. When Britain’s Committee on the Safety
of Medicines (CSM) got wind of results from three studies, it requested
copies, including unfinished analyses and preliminary manuscripts. One
of the investigators tells me that he supplied his findings with
hesitation and pleaded to see the CSM’s interpretation before they
went public. Like everyone else, the next he saw was the tabloid
account of a CSM advisory to British doctors, published a day before
the doctors had seen the letter and weeks before the Lancet
could get out the first peer-reviewed reports. Sixty-eight percent of
women who were using TGOCs at the time of the CSM letter switched to
another product, and 18% discontinued contraception altogether (27).
There have been widely varying estimates of the resulting increase in
pregnancies and pregnancy terminations, but the number is almost surely
greater than zero (27, 28). German authorities simply removed TGOCs
from the marketplace, affecting tens of millions of dollars of sales in
Germany and, by a ripple effect, probably hundreds of millions in sales
around the world. Women harmed, pharmaceutical giants gored: both the
social motive to unseat the epidemiologic findings and the material
support to do the heavy investigative research were in place from the
first moment of the controversy. I think that the efforts have been
misguided. Discrediting the research may indeed serve a short-term goal
of setting straight the errors of zealous regulators and predatory
journalists, but it does not further drug safety or women’s health in
the long run.
Observational studies are prone to all sorts of error and bias.
Inescapably, no true experimental studies will ever be done to address
the range of safety concerns that women will properly express. The
consequence is that we will always confront worrisome studies that fail
to meet the highest standard of science. We ignore them at great peril,
but neither can we embrace them.
I believe that the only alternative is to examine epidemiologic studies
with an open mind and quantitative spirit. Qualitative criticism is a
potent rhetorical device, but its persuasiveness outruns its scientific
value. The question for professionals is, "Does the bias account for
the results?" Unless a critic can demonstrate an affirmative answer
to that singular question, the result should stand. The result,
however, is a quantitative one and calls for a measured response.
|
Footnotes
|
|---|
Financial disclosure: supported by the Harvard
Program in Pharmacoepidemiology, which has received gifts, grants, or
contracts from the following pharmaceutical companies or their
associated foundations: Astra*, Berlex*, Boehringer-Ingelheim*, Eli
Lilly, Glaxo-Wellcome*, Roche*, Merck*, Novartis*,
Pasteur-Mérieux-Connaught*, Pfizer, Wyeth-Ayerst*. Dr. Walker has
served as a paid consultant in drug safety issues to the companies
marked with a "*" and to Janssen, Parke-Davis, Rhône-Poulenc
Rorer, and Searle. This paper is an extension of work performed at the
request of the World Health Organization, and the author has received
no outside support for its preparation.
The relative risk (RR) of 11 is the solution to the
equation (0.8 + 0.2 RR) ÷ (0.95 + 0.05 RR) = 2, where the
0.8 and the 0.2 correspond to the postulated fractions of women without
the high-risk factor (0.8) and with the high-risk factor (0.2) in users
of TGOCs, and the 0.95 and 0.05 correspond to the same fractions in
users of other OCs.
|
References
|
|---|
-
World Health Organization Collaborative Study of
Cardiovascular Disease and Steroid Hormone Contraception. 1995 Effect of different progestagens in low oestrogen oral contraceptives
on venous thromboembolic disease. Lancet. 346:1582–1588.[CrossRef][Medline]
-
Jick H, Jick SS, Gurewich V, Myers MW, Vasilakis
C. 1995 Risk of idiopathic cardiovascular death and nonfatal
venous thromboembolism in women using oral contraceptives with
differing progestagen components. Lancet 346:1589–1593.
-
Bloemenkamp KWM, Rosendaal FR, Helmerhorst FM,
Böller HR, Vandenbroucke JP. 1995 Enhancement by factor V
Leiden mutation of risk of deep-vein thrombosis associated with oral
contraceptives containing a third-generation progestagen. Lancet. 346:1593–1596.[CrossRef][Medline]
-
Spitzer WO, Lewis MA, Heinemann LAJ, Thorogood M,
MacRae KD. 1996 Third generation oral contraceptives and risk of
venous thromboembolic disorders: an international case-control study. BMJ. 312:83–88.[Abstract/Free Full Text]
-
Farmer RDT, Lawrenson RA, Kennedy JG, Hambleton
IR. 1997 Population-based study of risk of venous thromboembolism
associated with various oral contraceptives. Lancet. 349:83–88.[CrossRef][Medline]
-
Lidegaard Ø. 1998 Oral contraceptives and venous
thromboembolism. A case-control study. Contraception. 57:291–301.[CrossRef][Medline]
-
Walker AM. 1998 Newer oral contraceptives and the
risk of venous thromboembolism. Contraception. 57:169–181.[CrossRef][Medline]
-
Böttinger LE, Boman G, Eklund G, Westerholm
B. 1980 Oral contraceptives and thromboembolic disease: effects of
lowering oestrogen content. Lancet. i:1097–1101.
-
Meade TW, Greenberg G, Thompson SG. 1980 Progestagens and cardiovascular reactions associated with oral
contraceptives and a comparison of the safety of 50- and 30-
g
preparations. BMJ. 2:1157–1161.
-
Walker AM. 1996 Confounding by indication. Epidemiology. 7:335–336.[Medline]
-
Koster T, Rosendaal FR, Ronde H de, Briöt E,
Vandenbroucke JP, Bertina RM. 1993 Venous thrombosis due to poor
anticoagulant response to activated protein C: Leiden Thrombophilia
Study. Lancet. 342:1503–1506.[CrossRef][Medline]
-
Farmer RDT, Lawrenson R. 1996 Utilization patterns
of oral contraceptives in UK general practice. Contraception. 53:211–215.[CrossRef][Medline]
-
Heinemann LAJ, Lewis MA, Assmann A, Gravens L,
Guggenmoos-Holzmann I. 1996 Could preferential prescribing and
referral behavior of physicians explain the elevated thrombosis risk
found to be associated with oral contraceptives? Pharmacoepidemiol Drug
Safety. 5:285–294.
-
Dunn N, White I, Freemantle S, Mann R. 1998 The
role of prescribing and referral bias in studies of the association
between third generation oral contraceptives and increased risk of
thromboembolism. Pharmacoepidemiol Drug Safety. 7:3–14.
-
Jamin C, de Mouzon J. 1996 Selective prescribing of
third generation oral contraceptives. Contraception. 54:55–56.[CrossRef][Medline]
-
van Lunsen HW. 1996 Recent oral contraceptive use
patterns in four European countries: evidence for selective prescribing
of oral contraceptives containing third-generation progestogens. Eur J
Contracept Reprod Health Care. i:39–45.
-
Lidegaard Ø. 1997 The influence of thrombotic risk
factors when oral contraceptives are prescribed. Acta Obstet Gynecol
Scand. 76:252–260.[Medline]
-
Bloemenkamp KWM, Rosendaal FR, Büller HR,
Helmerhorst FM, Colly LP, Vandenbroucke JP. 1999 Risk of venous
thrombosis with use of current low-dose oral contraceptives is not
explained by diagnostic suspicion and referral bias. Arch Intern Med. 159:65–70.[Abstract/Free Full Text]
-
Lewis MA, Heinemann LAJ, MacRae KD, Bruppacher R,
Spitzer WO. 1996 The increased risk of venous thromboembolism and
the use of third generation progestagens: Role of bias in observational
research. Contraception. 54:5–13.[CrossRef][Medline]
-
Suissa S, Blais L, Spitzer WO, Cusson J, Lewis M,
Heinemann L. 1997 First-time use of newer oral contraceptives and
the risk of venous thromboembolism. Contraception. 56:141–146.[CrossRef][Medline]
-
Farley TMM, Meirik O, Marmot MG, Chang CL, Poulter
NR. 1998 Oral contraceptives and risk of venous thromboembolism:
impact of duration of use. Letter. Contraception. 57:61–65.[CrossRef][Medline]
-
Weiss NS. 1997 Bias in studies of venous
thromboembolism in relation to the use of new formulations of oral
contraceptives. Contraception. 55:189–190. (Letter).[CrossRef][Medline]
-
Vandenbroucke JP, Bloemenkamp KWM, Helmerhorst FM,
Rosendaal FR. 1997 Risk of oral contraceptives and recency of
market introduction. Contraception. 55:191–192. (Letter).[CrossRef][Medline]
-
Gerstman BB, Piper JM, Tomita DK, Ferguson WJ, Stadel
BV, Lundin FE. 1991 Oral contraceptive estrogen dose and the risk
of deep venous thromboembolic disease. Am J Epidemiol. 133:32–37.[Abstract/Free Full Text]
-
Vandenbroucke JP, Rosendaal FR. 1997 End of the
line for "third-generation-pill" controversy? (Editorial). Lancet. 349:1113–1114.[CrossRef][Medline]
-
Jamin C. 1998 Contraception estroprogestative et
pathologies cardio-vasculaires: la fin d’une polémique sur la
troisième génération. Contracept Fertil Sex. 26:331–337.[Medline]
-
Jick SS, Vasilakis C, Jick H. 1998 Pregnancies and
terminations after 1995 warning about third generation oral
contraceptives. Lancet. 351:1404–1405.[Medline]
-
Wood R, Botting B, Dunnel K. 1997 Trends in
conceptions before and after the 1995 pill scare. Population Trends
(Office for National Statistics, London). 89:5–12.
Top
Introduction
Counter arguments
