Original research articleCase-control studies on venous thromboembolism: bias due to design? A methodological study on venous thromboembolism and steroid hormone use☆
Introduction
The recent “pill crisis” in Europe has centered on the controversial finding that newer combined oral contraceptives (OCs) that contain the progestins desogestrel and gestodene (called third-generation OCs) appear to be associated with a higher risk of venous thromboembolism (VTE) than older products containing primarily levonorgestrel (known as second-generation OCs). The ensuing debate between those who think the risk estimates reflect the association between VTE and OCs correctly and those who believe them to be due to biases that produce overestimates of risk is still unresolved and was the subject of various extensive reviews and consensus papers [1], [2], [3].
The issues concerning the methods used to measure the risk of VTE in users of OCs raised in this context include many technical issues of concern for epidemiologists, but also some clinical features that deserve wider appreciation and that will be addressed in this article. One such clinical problem relates to the emergence of diagnostic suspicion and the decision to refer patients to further diagnostics based on their risk profile. We have shown, for example, that both diagnostic suspicion and referral decisions were influenced by whether the patient with a suspect diagnosis uses an OC or not [4]. Two earlier articles, however, found no evidence for this of detection bias [5], [6].
Epidemiologic studies in general require rather strict criteria for endpoints. This is appropriate when such endpoints can be clearly defined. In the case of the detection and treatment of VTE, however, there are many decisional elements in medical practice, which may or may not lead patients with nearly identical features to hospital. Thus, of the large number of potential events, only a small proportion might satisfy the diagnostic criteria for inclusion into the previous case-control studies. These criteria are that the patient has been admitted to a hospital in the first place and has then been subjected to the diagnostic tests set forth for inclusion in the study. Furthermore, to achieve an unconfounded estimate of the association between an event (i.e., VTE) and an exposure (i.e., OC use), epidemiologic studies wish to exclude all risks that might compete with the risk factor under investigation. This leads to the exclusion of women who might have had surgery, trauma, or prolonged bed rest at some specified time before the event, and it also excludes individuals who have previously had VTE. For study purposes, therefore, an artificial category of “idiopathic” VTE is created. These are patients for whom “no other (apparent) cause” but the exposure under study may be held responsible for the event. These study criteria cannot, of course, exclude patients with unknown risk factors, such as one of the several genetic factors identified in the last decade.
In this article we examine the difficulties inherent in investigating the association of VTE and OCs by means of (hospital-based) case-control studies, explore the influence of the use of various population groups on the estimates of risk, and further show the effect of exclusion criteria corresponding to those that define “idiopathic” VTE. Thus, we address important methodological issues in the design and how design decisions can affect the result. We conclude with a discussion of the general appropriateness of such a diagnostic entity and of how much the notion of “idiopathic” VTE really contributes to our understanding of VTE and its diagnostic pathways.
Section snippets
Methods
All women aged 15 to 49 years with signs or symptoms consistent with VTE were accrued in 21 physician practices (general practitioners, internists, surgeons, gynecologists, or orthopedists), diagnostic centers and hospitals in four of the Federal States of Germany (Bavaria, Thuringia, Saxonia-Anhalt, and Berlin) between January 1994 and July 1999. These women were followed-up until the diagnosis of VTE was either confirmed or dismissed.
Diagnostic categories for VTE were centrally assigned on
Results
Of 1068 women who entered the initial study, 385 patients from hospital and 221 patients from general practice were diagnosed as VTE and were used as population case group (all 606) and hospital cases (385). A total of 362 hospital controls were collected at the same time from the same hospitals as those of the hospital cases, and 2580 population controls were drawn from the same areas of residence as the population cases. The characteristics of the case and control groups are shown in Table 1.
Discussion
The debate on biases in case-control studies on VTE risk and OC use prompted us to conduct a methods-oriented case-control study in Germany that was capable of exploring the impact of various suspected biases [7], [8], [9]. The basic premise of this project was to begin data collection at the time VTE was first suspected, so that all information potentially related to this debate might be evaluated. It was the intent to trace the population of VTE cases from first suspicion to their final
Conclusions
The risk of VTE associated with OC use measured with hospital-based case-control studies is likely to be systematically overestimated because of methodological reasons. This bias is also found to work differentially in favor of OCs of the second generation. Moreover, the concept of so-called “idiopathic” VTE, which leads to exclusion of a high proportion of VTE cases, needs further discussion and probably a revision.
Current case-control studies of VTE disregard important information in favor of
References (13)
The epidemiology of oral contraceptive usea critical review of the studies on oral contraceptives and the health of young women
Am J Obstet Gynecol
(1998)- et al.
Transnational Research Group on Oral Contraception and the health of young women. The increased risk of venous thromboembolism and the use of third-generation progestagens—role of bias in observational research
Contraception
(1996) Cardiovascular diseases and steroid hormone use
WHO Technical Report Series 877
(1998)Reproductive choices in 2000the relative safety of current oral contraceptives
Hum Reprod Update
(1999)- et al.
Venous thromboembolism and oral contraceptive usea methodological study of diagnostic suspicion and referral bias
Eur J Contracept Reprod Hlth Care
(2000) - et al.
Risk of venous thrombosis with use of current low-dose oral contraceptives is not explained by diagnostic suspicion and referral bias
Arch Intern Med
(1999)
Cited by (32)
Combined hormonal contraception and the risk of venous thromboembolism: a guideline
2017, Fertility and SterilityPrevention and treatment of hormone-associated venous thromboembolism: A patient management approach
2010, Hematology/Oncology Clinics of North AmericaUse of oral contraceptives containing gestodene and risk of venous thromboembolism: outlook 10 years after the third-generation "pill scare"
2010, ContraceptionCitation Excerpt :A subanalysis comparing the use of gestodene-containing OCs and second-generation OCs containing LNG was also performed. The design of the study, which was conducted in Austria in 2005 and 2006, was consistent with earlier VTE case-control studies described elsewhere [8,11,12]. The study protocol was approved by independent ethics committees or institutional review boards and was conducted in accordance with the ethical principles of the Declaration of Helsinki.
Gender Issues in Venous Thromboembolism
2010, Principles of Gender-Specific MedicineThe safety of a drospirenone-containing oral contraceptive: final results from the European Active Surveillance study on Oral Contraceptives based on 142,475 women-years of observation
2007, ContraceptionCitation Excerpt :Whether the VTE risk associated with gestodene and desogestrel is real or the result of bias and confounding has been the subject of heated scientific debate. This debate, which has yet to be conclusively resolved [5,6], is due to the fact that robust safety data that could either substantiate or refute a higher risk of VTE were not available at the time of the “pill scare.” In order to avoid a comparable situation for the market introduction of an OC with a novel progestin [30 μg EE and 3 mg drospirenone (DRSP)] [7–11], the manufacturer of this OC commissioned the Center for Epidemiology and Health Research in Berlin to perform a large, prospective, controlled cohort study as part of a phase IV commitment.
- ☆
This study was supported over several years by Schering AG Berlin by an unconditional grant.