Recurrent use of newer oral contraceptives and the risk of venous thromboembolism

S. Suissa1,6, W.O. Spitzer1, B. Rainville2, J. Cusson3, M. Lewis4 and L. Heinemann5

1 Departments of Epidemiology and Biostatistics, and of Medicine, McGill University, Montreal, H3A 1A2, 2 Division of Clinical Epidemiology, Royal Victoria Hospital, Montreal, H3A 1A1, 3 Department of Medicine, Hôtel-Dieu Hospital, University of Montreal, Montreal, Canada, 4 EPES Research Centre, Berlin and 5 ZEG Research Centre, Berlin, Germany


    Abstract
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
The epidemiological studies that assessed the risk of venous thromboembolism (VTE) associated with newer oral contraceptives (OC) did not distinguish between patterns of OC use, namely first-time users, repeaters and switchers. Data from a Transnational case–control study were used to assess the risk of VTE for the latter patterns of use, while accounting for duration of use. Over the period 1993–1996, 551 cases of VTE were identified in Germany and the UK along with 2066 controls. Totals of 128 cases and 650 controls were analysed for repeat use and 135 cases and 622 controls for switching patterns. The adjusted rate ratio of VTE for repeat users of third generation OC was 0.6 (95% CI:0.3–1.2) relative to repeat users of second generation pills, whereas it was 1.3 (95% CI:0.7–2.4) for switchers from second to third generation pills relative to switchers from third to second generation pills. We conclude that second and third generation agents are associated with equivalent risks of VTE when the same agent is used repeatedly after interruption periods or when users are switched between the two generations of pills. These analyses suggest that the higher risk observed for the newer OC in other studies may be the result of inadequate comparisons of pill users with different patterns of pill use.

Key words: deep venous thrombosis/epidemiological study/pulmonary embolism/second and third generation oral contraceptives


    Introduction
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Several recent epidemiological studies assessed the risk of venous thromboembolism (VTE) associated with the use of newer oral contraceptives (OC), that is low-dose oestrogen pills containing the newer progestagens desogestrel and gestodene. Some studies found that these so-called third generation OC had a higher risk of VTE relative to the second generation agent that contains levonorgestrel, with relative risks ranging from 1.5 to 2.2 (Bloemenkamp et al., 1995Go; Jick et al., 1995Go; WHO Collaborative Study, 1995a; Spitzer et al., 1996Go). Other studies, using alternative adjustments for confounders, observed lower and non-significant relative risks ranging from 0.8 to 1.4 (Farmer et al., 1997Go, 1998Go; Lidegaard et al., 1998Go). Although all studies performed careful statistical adjustment for potential confounders, little attention was given to the patterns and duration of OC use. Indeed, the minimal analyses of these profiles of OC use produced inconsistent results, even within the same study (WHO Collaborative Study, 1995a; Poulter et al., 1996Go). Nevertheless, one consistent observation in the studies that presented such data has been that the risk of VTE appears to vary by duration and pattern of OC use.

In the few studies that considered these patterns, however, distinctions between first-time users and recurrent users were either made broadly or not at all in the data analysis. This distinction is important because we postulate that not only can duration of OC use be a principal source of bias in these studies, but these effects can also vary depending on whether the user is a new starter or has already used OC in the past. Indeed, we consider that the observed difference between the second and third generation OC risks of VTE may be largely due to the fact that users of these two OC types vary by duration and pattern of use. It is expected that users of the newer pills in 1995 had a shorter duration of use than users of older pills, a distinction that becomes crucial if the risk varies by duration of OC use. To address this point, we previously used the Transnational Study data (Spitzer et al., 1996Go) to analyse the risk of VTE in first-time ever users of second and third generation pills as a function of the duration of use (Suissa et al., 1997Go). We found similar risk profiles for both types of pills, namely that the risk, relative to non-users, increases sharply to 10 in the first year of continuous use and decreases to around two thereafter for both types of pills. Similar re-analyses were subsequently performed on the WHO data (WHO, 1995a) with somewhat different findings (Farley et al., 1998Go), although several methodological reasons were offered for these differences (Suissa and Spitzer, 1998Go). It was impossible, with the data available in the Transnational study at the time, to estimate these risks in the subjects who were not first-time users. These data are now available and are the object of this article.

In this paper, we use additional data on the history of OC use from the Transnational study (Spitzer et al., 1996Go) to assess the risks of VTE for second and third generation OC in recurrent users who either used these products intermittently or switched between products.


    Materials and methods
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
The data for this study come from the Transnational study (Spitzer et al., 1996Go), which used a case–control design with subjects taken from five European countries (Germany, UK, Austria, France and Switzerland), although only data from the first two countries were used in the principal analyses. Between 1993 and 1995, 471 cases of VTE in women aged 16–44 years were identified in participating hospitals of the two countries, diagnosed clinically and confirmed by imaging procedures. For each case, up to four controls were obtained, for a total of 1772. While these 471 cases and 1772 controls were the object of the initial publication, data collection continued until 1996 and the final study size based on Germany and the UK now included 551 cases and 2066 controls. All subjects were interviewed as of the index date, which was the date of the event for the cases and the corresponding date for the matched controls. They were questioned on their health history and their current use of OC, with the clinical data verified against medical records. Additional data on the chronology of OC use, with confirmation for the brand by inspection of individual patients' pill packets, were collected but not coded or prepared for data analysis until recently. Potential confounding variables used were socio-demography, lifestyle, medical history and family history of disease.

The third generation OC were defined as those combining low-dose oestrogen (<35 µg ethinyl oestradiol) with either of the desogestrel or gestodene progestagens. The second generation agents contained low-dose oestrogen with levonorgestrel or norgestimate; the latter was included in this group to remain consistent with previous studies (WHO Collaborative Study, 1995a, b; Spitzer et al., 1996Go). Current use of OC was defined as use in the 3 month period prior to the index date.

The present analysis involved only second and third generation agents, so that all subjects who used first generation high-dose oestrogen pills were excluded from the analyses. We defined recurrent use of OC as use which was either interrupted for at least one cycle or, if not, which involved agents from the two generations. We distinguished between two types of recurrent OC users, the repeaters and the switchers. The exposed repeaters were current users who had, in the past, interrupted their use of the same generation pill. The exposed switchers were current users who switched OC, from second to third generation agents or vice-versa. The unexposed reference group for both analyses was made up of all former users who interrupted their use for >90 days before the index date, and are thus not current users. This reference group of non-current users included only former users, and not `never' users, so as to make it more comparable to the current user group, which is made up only of former users, by definition. Initially, we distinguished between non-current users on the basis of the type of pill they had used in the past, so as to compare current users to non-current users who had previously used the same pill generation. However, since the results were similar with the two different reference groups, we combined both control groups into a single one, which had the added advantages of increasing the power of all analyses and of eliminating additional sources of bias.

To analyse the duration of current use, it was stratified for current users as <1, 1–2 and >2 years. The cumulative duration of prior OC use, for all prior episodes of OC use excluding the current one, as well as the duration of the last interruption period just prior to the current episode of use, were considered as covariates, as they were likely to differ among the three comparison groups and duration of OC use was known to have an effect on VTE risk. In particular, the duration of the last interruption period may act as a washout period of the previous effect, thereby possibly leading to higher risks with shorter interruption periods. Other covariates used in the analyses as potential confounders were the same factors used in the original study, namely linear age, smoking, alcohol use, study centre and body mass index (Spitzer et al., 1996Go). Rate ratios were estimated from odds ratios by logistic regression. Crude and adjusted odds ratios were obtained for current use of any duration as well as by duration of OC use. The SAS statistical software package was used for all analyses (SAS/STAT Software, 1996Go).


    Results
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Of the 551 cases, we excluded 229 who used first generation pills or whose pill history data were unknown or missing, 66 first-time users and 47 `never' users, leaving 209 cases for analysis. Of the 2066 controls, 786 were excluded who used first or unknown generation pills, 180 first-time users and 271 `never' users, leaving 829 control subjects for analysis.

Table IGo displays the characteristics of the 128 cases and 650 controls that form the study population for the analysis of repeat use, along with those of the 135 cases and 622 controls used for the analysis of switching effect. Some control subjects were used in both groups. For repeat use, the cases had slightly higher body mass index (BMI), were more likely to be smokers but drink less alcohol than the controls. The cases and controls were the same age and mostly came from the UK. The cases had used, previous to their current episode of OC use, OC for shorter periods of time than controls. For the switching analysis, the cases were the same age, were more likely to be smokers but drink less alcohol than the controls. The cases and controls had the same BMI and came mostly from the UK. The cases had been using the current OC for a longer time than controls, after similar interruption periods, than controls.


View this table:
[in this window]
[in a new window]
 
Table I. Potential confounding characteristics of cases and controlsa for repeat use and switching analyses
 
To identify the potential confounding factors, Tables II and IIIGoGo provide the characteristics of the controls by pill exposure status, for repeat use and switching. Control users of third generation pills were much younger and had a lower BMI than second generation current users or non-current users. Only 7% of the controls using third generation pills were from Germany compared to 39% for second generation and 12% for the currently unexposed. Users of second generation pills had longer periods of prior use and of current use than users of third generation pills. Table IIIGo, describing the switcher controls, indicates that switchers to second generation pills were younger and slightly less likely to smoke than third generation current users or non-current users. Switchers to third generation pills had shorter periods of current OC use, prior use and of interruption than users of second generation pills. Thus, age, country, durations of prior OC use and of the last interruption period were considered to be confounders and adjusted for in the estimation of the rate ratios (RR). Duration of current use was controlled for by stratification.


View this table:
[in this window]
[in a new window]
 
Table II. Potential confounding characteristics of repeat use controls by current oral contraceptive use
 

View this table:
[in this window]
[in a new window]
 
Table III. Potential confounding characteristics of switcher controls by current oral contraceptive use
 
Table IVGo provides the RR, estimated from odds ratios, for repeat users of third and second generation pills compared with non-current use along with the 95% confidence intervals (CI). The adjusted RR were 4.7 (95% CI 2.8–7.9) and 2.9 (95% CI 1.4–5.8) for second and third generation pills respectively. The direct comparison of the two pill classes gave an adjusted RR of 0.6 (95% CI:0.3–1.2) for third relative to second generation pills. Table VGo displays the RR for repeat users of third and second generation pills compared with non-current use according to the duration of the current episode of OC use. The RR appeared to be slightly higher in the first 2 years, although this effect is minimal and the CI were wide.


View this table:
[in this window]
[in a new window]
 
Table IV. Crude and adjusted rate ratios (RR) of venous thromboembolism for repeaters by current use of second and third generation oral contraceptives
 

View this table:
[in this window]
[in a new window]
 
Table V. Crude rate ratios (RR) of venous thromboembolism for repeat users by duration of use of second and third generation oral contraceptives (OC)
 
The adjusted RR, given in Table VIGo, after switching from third to second generation pills compared to non-current use was 5.3 (95% CI 2.7–10.5), while after switching from second to third was 7.7 (95% CI 4.7–12.8). The direct comparison of the two pill classes resulted in an adjusted RR of 1.3 (95% CI 0.7–2.4) for switchers from second to third generation pills relative to switchers from third to second. To better understand this risk, we stratified our analyses by the number of prior switches. For the majority of women who only switched once, the RR was 1.1 (95% CI 0.5–2.3), while among the few multiple switchers, the RR was 1.8 (95% CI 0.2–16.8). The RR did not vary according to the duration of the current episode of OC use (data not shown).


View this table:
[in this window]
[in a new window]
 
Table VI. Crude and adjusted rate ratios of venous thromboembolism for switchers by current use of second and third generation oral contraceptives and by number of prior switches
 

    Discussion
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
An important finding of this study is that the risk of VTE is no different for repeat users of second and third generation OC, after adjustment for all confounding factors including the durations of the interruption period and of prior OC use. These subjects never switched products; they either always used a second or a third generation agent, albeit intermittently. Similarly, we found that women who switched from second to third generation agents had a similar risk of VTE as those who switched from third to second.

No other study has ever assessed the risks according to these profiles of OC use. The WHO study (WHO, 1995a) was the only one in which the mix of recurrent users was investigated. It found that the risk of VTE among what were called `former users' of OC was 1.8-fold higher for third than for second generation agents during the first year of use, and it was 2.3-fold higher thereafter (Poulter et al., 1996Go). It is not clear how these former users were defined, although we can assume that they included both repeaters and switchers as well as users of first generation pills, while the reference group was ambiguous. Moreover, the analyses did not adjust for durations of prior OC use and of interruption periods that we found to be relevant confounders. These two factors varied appreciably between the three comparison groups, which is expected because second and third generation pills appeared on the market at different times. And since the risk of VTE varies by duration of OC use, these factors turned out to be relevant confounders. Unlike our findings for first-time users (Suissa et al., 1997Go), however, the risk did not vary by duration of current OC use, and was particularly not higher in the first year of use. In our study, we observed that among repeaters the rate of VTE was lower, although not significantly, for third than for second generation agents during the first year of use as well as subsequently. On the other hand, the switchers from second to third generation agents had a slightly higher, although not significant, rate of VTE than those who switched from third to second, for both the first year of use after the switch as well as subsequently. Comparisons with the WHO study results are not possible because of its unclear definition of `former users' and of the reference group, as well as the lack of adjustment for profile of use.

Our prior findings for first-time users (Suissa et al., 1997Go) show that the risks were identical for first-time users of second and third generation pills after duration of use is accounted for. Along with the present analyses, they suggest that all studies of the risks associated with OC use may have been biased because data analyses have been based simply on current OC use, and have not considered the history of OC use. Indeed, when the history of use is taken into account, both in terms of duration and profile, the risks of the two generations of pills are equivalent. Thus analyses of such epidemiological studies can only be valid if the comparisons are made between subjects with similar durations and profiles of history of OC use. Besides the profile of first-time uninterrupted use, which was considered in a prior paper, repeat but interrupted use of the same product and switching between products must be considered separately. In addition, we also adjusted for durations of prior use and of interruption periods to make these profiles as similar as possible, and excluded subjects who had used first generation agents. This profile of use being an effect-modifier of the risk, it must be stratified for and analyses must be performed within strata. (Guess, 1989Go; Miettinen and Caro, 1989Go). Thus, the overall results of all prior studies may be biased because they were based on uneven mixes of first-time users, repeaters and switchers, of second and third generation OC.

Our study has limitations. First, unlike first-time users, repeaters and switchers are peculiar so that the findings could be affected by several biases, including confounding by indication and selection bias. The repeaters have already been tested with one of these agents and have discontinued their use for a variety of usual reasons. However, the fact that they restarted on the same agent indicates that the current users were able to tolerate well the agent they were on, with no need to change. On the other hand, the reference group of current non-users was made up of subjects who, like the exposed subjects, stopped using OC but, unlike the exposed subjects, did not restart for reasons unknown to us. Indeed, if we had instead used as reference the `never' users, the reference group we had used for first-time users (Suissa et al.,1997), the case to control odds would have been 47:271 instead of 54:443 for former users, which would reduce the RR by 30%. All the same, the RR for repeat use remain high, which does not concur with our findings for first-time users that indicated that the risk essentially disappeared after a few years of uninterrupted use both for second and third generation OC. We surmise that these odds ratios relative to non-users may be biased for the unknown reasons that make repeaters start, interrupt and restart OC use. This could also be explained by selection bias, as subjects who were restarted on OC may have been more likely to be referred for a VTE examination and may have been more likely to have VTE diagnosed. Nevertheless, although comparisons between the exposed and unexposed repeaters may be biased, the direct comparison between repeaters of second and third generation agents are probably valid. For this latter comparison, we found that the risks of VTE associated with second and third generation pills were not statistically different. Second, subjects who switched to another generation may have done so because of a lack of tolerance, and were channelled to a safer agent, or because of marketing effects. Channelling patients to safer agents, leading to confounding by indication, is common in non-experimental risk assessment of drugs (Petri and Urquhart, 1991Go). Although VTE risk factors are not as well known as the arterial ones, a recent survey showed that women at higher risk of VTE were more likely to be switched from second to third generation pills (Heinemann et al., 1996Go). This may explain, at least in part, the RR of 1.3 among switchers, which was in any case not statistically significant. Thirdly, our sample size, although large overall, was not very large for certain analyses, with only 19 third generation repeater cases and 22 second generation switcher cases.

Because comparative non-experimental studies of OC risks are not randomized, their analysis cannot be as simple as that of randomized trials. The intention of randomization must be substituted with careful study design and data analysis that account for the complete history of OC exposure ever since the first time they were used by each study subject. It is because previous studies of OC risk did not aspire to replace randomization by the type of analysis we presented, that the findings are probably inaccurate. By using simple comparative analyses based mostly on current OC use, irrespective of the entire history of OC use, these previous studies did not recognize the complexity of the comparisons performed on groups of subjects from a different mix of history of OC use which has a bearing on the findings. This is particularly important in a field where the results of studies yet to be completed, peer-reviewed, published or discussed are used prematurely for regulatory decisions and by the media, potentially causing hysteria and poor patient management, in the absence of solid scientific evidence and debate (Mills, 1997Go; Benagiano, 1998Go).

In conclusion, we found that second and third generation agents are associated with equivalent risks of VTE among women who had previously used OC, whether they repeated use of the same generation OC agent after an interruption period or switched to another generation of pill.


    Acknowledgments
 
Funded in part by grants from Organon, Schering AG and the Medical Research Council of Canada (MRC). Samy Suissa is the recipient of a Senior Scientist award from the MRC. The McGill Pharmacoepidemiology Research Unit is funded by the Fonds de la recherche en Santé du Québec (FRSQ).


    Notes
 
6 To whom correspondence should be addressed at: Division of Clinical Epidemiology, 687 Pine avenue west, Ross 4.29, Montreal, Québec, Canada, H3A 1A1 Back


    References
 Top
 Abstract
 Introduction
 Materials and methods
 Results
 Discussion
 References
 
Benagiano, G. (1998) Venous thromboembolism and the pill: learning from the past — an endless saga. Hum. Reprod., 13, 1115–1116.[Free Full Text]

Bloemenkamp, K.W.M., Rosendaal, F.R., Hemerhosrt, F.M. et al. (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.[ISI][Medline]

Farley, T.M.M., Meirik, O., Marmot, M.G. et al. (1998) Oral contraceptives and risk of venous thromboembolism: impact of duration of use [Letter]. Contraception, 57, 61–65.[ISI][Medline]

Farmer, R.D.T., Lawrenson, R.A., Thompson, C.R. et al. (1997) Population-based study of risk of venous thromboembolism associated with various oral contraceptives. Lancet, 349, 83–88.[ISI][Medline]

Farmer, R.D.T., Todd, J.C., Lewis, M.A. et al. (1998) The risks of venous thromboembolism disease among German women using oral contraceptives: a database study. Contraception, 57, 67–70.[ISI][Medline]

Guess, H.A. (1989) Behaviour of the exposure odds ratio in a case-control study when the hazard function is not constant over time. J. Clin. Epidemiol., 42, 1179–1184.[ISI][Medline]

Heinemann, L.A.J., Lewis, M.A., Abmann, A. et al. (1996) Could preferential prescribing and referral behaviour of physicians explain the elevated thrombosis risk found to be associated with third-generation oral contraceptives? Pharmacepidemiol. Drug Safety, 5, 285–294

Jick, H., Jick, S.S., Gurewich, V. et al. (1995) Risk of idiopathic cardiovascular death and nonfatal venous thromboembolism in women using oral contraceptives with differing progestagen components [See comments]. Lancet, 346, 1589–1593.[ISI][Medline]

Lidegaard, O., Edstrom, B. and Kreiner, S. (1998) Oral contraceptives and venous thromboembolism: A case control study. Contraception, 57, 291–301.[ISI][Medline]

Miettinen, O.S. and Caro, J.J. (1989) Principles of nonexperimental assessment of excess risk, with special reference to adverse drug reactions. J. Clin. Epidemiol., 42, 325–331.[ISI][Medline]

Mills, A. (1997) Combined oral contraception and the risk of venous thromboembolism. Hum. Reprod., 12, 2595–2598.[Free Full Text]

Petri, H. and Urquhart, J. (1991) Channelling bias in the interpretation of drug effects. Statist. Med., 10, 577–581.[ISI]

Poulter, N.R., Farley, T.M.M., Chang, C.L. et al. (1996) Safety of combined oral contraceptive pills [Letter]. Lancet, 347, 547.[ISI]

Rosenberg, L., Bégaud, B., Bergman, U. et al. (1996) What are the risks of third-generation oral contraceptives? Are third-generation oral contraceptives safe? Hum. Reprod., 11, 687–693.[ISI][Medline]

SAS/STAT Software (1996) Changes and Enhancements. Release 6.11 edn. SAS Institute, Inc., Cary, NC.

Spitzer, W.O., Lewis, M.A., Heinemann, L.A. et al. (1996) Third generation oral contraceptives and risk of venous thromboembolic disorders: an international case-control study. Transnational Research Group on Oral Contraceptives and the Health of Young Women [see comments]. Br. Med. J., 312, 83–88.[Abstract/Free Full Text]

Suissa, S. and Spitzer, W.O. (1998) Methodological limitations in comparing the risks of newer oral contraceptives [Letter]. Contraception, 57, 61–65.[ISI][Medline]

Suissa, S., Blais, L., Spitzer, W.O. et al. (1997) First-time use of newer oral contraceptives and the risk of venous thromboembolism. Contraception, 56, 141–146.[ISI][Medline]

WHO Collaborative Study of Cardiovascular Disease and Steroid Hormone Contraception (1995a) Venous thromboembolic disease and combined oral contraceptives: results of international multicentre case-control study. World Health Organization Collaborative Study of Cardiovascular Disease and Steroid Hormone Contraception [see comments]. Lancet, 346, 1575–1582.[ISI][Medline]

WHO Collaborative Study of Cardiovascular Disease and Steroid Hormone Contraception (1995b) Effect of different progrestagens in low oestrogen oral contraceptives on venous thromboembolic disease. Lancet, 346, 1582–1588.[ISI][Medline]

Submitted on April 26, 1999; accepted on December 13, 1999.