databases the rates amongst users of levonorgestrel products were lower than those amongst users of desogestrel and gestodene products. A case fatality rate ...
Human Reproduction Update 1999, Vol. 5, No.6 pp. 688-706
©
European Society of Human Reproduction and Embryology
Oral contraceptives and venous thromboembolic disease. Analyses of the UK General Practice Research Database and the UK MediPIus Database R.D.T.Farmer1, R.A.Lawrenson, J-C.Todd, TJ.Williams and K.MacRae Pharmacoepidemiology Unit, European Institute of Health and Medical Sciences, University of Surrey, Stirling House, Surrey Research Park, Guildford GU2 5RF, Surrey, UK
Key words: combined oral contraceptive formulations/observational- studies/oral contraceptives/progestogens/ venous thromboembolism
TABLE OF CONTENTS Introduction Study populations Results of the new GPRD and MediPIus studies Discussion on the implications of the new GPRD and MediPIus studies References .
'To whom correspondence should be addressed
Introduction 688 692 693 700 705
In 1995 the UK Committee on the Safety of Medicines (CSM) advised doctors and pharmacists that combined oral contraceptives (COC) containing desogestrel or gestodene were associated with a 2-fold increase in risk of venous thromboembolic disease, compared with those containing other progestogens (CSM, 1995).
Downloaded from humupd.oxfordjournals.org by guest on July 17, 2011
The results of three independent studies of venous thromboembolic disease (VTE) and oral contraceptives are reviewed together with two further cohort/case-control studies which we conducted using the MediPIus and General Practice Research Database (GPRD) databases. These latter studies jointly involved 395 cases and uniquely examined the association between VTE and individual combined oral contraceptive (COC) formulations. The two studies yielded very similar results. Crude incidence rates for idiopathic VTE of 4.6 and 3.8 were found per 10 000 exposed woman-years (EWY), in the MediPIus and GPRD studies respectively. Incidence rates increased markedly with age, and in both databases the rates amongst users of levonorgestrel products were lower than those amongst users of desogestrel and gestodene products. A case fatality rate of 3% and a mortality rate of 10 per million EWY were estimated. Odds ratios (OR) were calculated for confounding variables and different COC formulations. Both database studies indicated an excess of current smokers and women with high body mass indices amongst cases. There were significantly more cases with asthma in the GPRD study and cases who had been using their COC for less than a year. No statistically significant differences between COC formulations were found in the analyses where controls were matched to cases by practice and year of birth in both the MediPIus and GPRD studies. In the GPRD study we also ran a study where controls were matched by practice and within 5 year age bands. In this study the OR were consistently higher for the newer or 'third generation' products than when controls were matched by year of birth. However only the acne formulation/OC containing cyproterone acetate and 35 (ig ethinyloestradiol yielded a significant OR of 23. It may be concluded that improvements in prescribing are paramount as the results strongly indicate that overweight women and those who smoke are at a greater risk of VTE. Further study is required to elucidate the possibility that asthma or its treatment may predispose to VTE, alone or in combination with other risk factors. However, neither the MediPIus nor GPRD studies indicate that any one COC formulation poses a greater risk of VTE than another.
Oral contraceptives and VTE
689
Table 1. Results from the World Health Organization (WHO), Transnational, Boston Collaborative Drug Surveillance Programme (BCDSP) and UK MediPlus studies Study
Product
Exposed cases
WHO non-Oxford Centres
Levonorgestrel
97
(Farley era/., 1995)
Desogestrel
WHO Oxford Region Hospital
95% Cl
OR Reference group
7
4.8
0.5, 43.4
Gestodene
16
5.3*
1.8,15.5
Levonorgestrel
40
Reference group
28
2.3*
1.1,4.9
20
2
0.8, 4.7
WHO Oxford Region Community
Levonorgestrel
36
Reference group
Controls (Farley era/., 1995)
Desogestrel
27
1.8
0.7, 4.8
Gestodene
16
0.9
0.3, 2.8
Transnational Germany
2nd generation
68
Reference group
(Spitzerefa/., 1996)
Desogestrel
12
1.5
0.8, 3.1
Gestodene
10
2.6*
1,7.2
Transnational UK
2nd generation
64
Reference group
(Spitzerefa/., 1996)
Desogestrel
53
1.6
1,2.5
Gestodene
45
1.4
0.9, 2.3
BCDSP-GPRD
Levonorgestrel
23
Reference group
(Jickefa/., 1995)
Desogestrel
30
2.2*
1.1,4.4
Gestodene
23
2.1*
1,4.4
Levonorgestrel
24
Reference group
Desogestrel
32
0.9
0.4, 1.8
Gestodene
22
0.8
0.4,1.9
UK MediPlus (1997) (Farmer era/., 1997)
Downloaded from humupd.oxfordjournals.org by guest on July 17, 2011
Desogestrel Gestodene
Controls (Farley et al., 1995)
'Statistically significant odds ratio (OR) (P < 0.05). Cl = confidence interval; GPRD = General Practice Research Database.
Table II. Most frequently prescribed combined oral contraceptives in the General Practice Research Database (GPRD) and MediPlus databases Oestrogen
GPRD
Average
MediPlus
dose per cycle
Womanyears (xiO3)
%
age (years)
Womanyears (xiO3)
%
Average age (years)
630
190
24
27.4
49
23
27.2
630
153
20
25.8
40
18
25.7
630
144
18
25.9
42
19
25.7
Triphasic levonorgestrel
680
103
13
28.0
28
13
28.2
Desogestrel 150 jig
420
38
4.8
28.8
10
4.8
28.8
755
40
5.2
25.8
14
6.7
25.8
Formula
Levonorgestrel 150 ng + EE 30 ng Desogestrel 150 (ig + EE 30 ng Gestodene 75 ng + EE 30 ug
+ EE 20 jig Norgestimate 250 ug + EE 35 ug
EE = ethinyloestradiol.
690
R.D.T.Farmer et al. products had a significantly elevated OR of 2.3 (1.1,4.9) using hospital controls, but not using community controls: 1.8 (0.7, 4.8). When desogestrel and gestodene products were combined there was a significant OR of 2.2 (1.1, 4.2) using hospital controls and a non-significant OR of 1.4 (0.6, 3.1) using the general practice controls. This study did not show consistency between centres, and within centres there were inconsistencies in ORs using different control groups. The Transnational Study was designed to emulate the WHO study but was restricted to European countries (Spitzer et al., 1996). In contrast to the WHO study, it used a combination of hospital and community controls in all centres and was restricted to women aged 16-44 years. The OR for gestodene-based products in the UK centres using 'second generation' COCs as the reference group (essentially products with a low dose oestrogen combined with levonorgestrel) was not significant 1.4 (0.9,2.3). The centres in Germany (the only other country for which the results were reported in detail) produced a significantly elevated OR of 2.6 (1.0, 7.2) for gestodene-based products. Desogestrel-based products yielded a non-significant elevation in the OR of 1.5 (0.8, 3.1) using German data and a marginally significant OR of 1.6 (1.0, 2.5) from UK data. The OR became statistically significant 1.5(1.1, 2.1) when data from the two countries were combined and the desogestrel and gestodene products were aggregated to form a category called 'third generation'. The third study (Jick et al., 1995) was carried out by the Boston Collaborative Drug Surveillance Program (BCDSP). It was based on the UK General Practice Research Database (GPRD). The investigators found an absolute incidence of 1.61 per 10 000 exposed woman-years (EWY), for VTE amongst users of levonorgestrel-based products. The authors stated that this was in line with previous findings and cited their earlier papers (Porter etal., 1982,1985). In the first study the rate was calculated from seven exposed cases in 36 428 EWY; and in the second, from three cases in 37 807 EWY. The estimated rates were 1.9 and 0.8 cases per 10000 EWY respectively. Both estimates are substantially lower than rates derived from other and larger studies. For example, Gerstman reported a rate of 4.2 per 10 000 EWY of exposure to COC containing 25 years. The users of levonorgestrel 250 fig + ethinyloestradiol 30 ug [Eugynon 30 (Schering Health Care Ltd)/Ovran 30 (Wyeth Laboratories)] tended to be older than the users of levonorgestrel 150 ug + ethinloestradiol 30 fig. Desogestrel 150 ug + ethinyloestradiol 20 ug [Mercilon (Organon Laboratories Ltd)] had the highest proportionate use amongst women aged 40-44 years. With the exception of desogestrel 150 ug + ethinyloestradiol 20 ug the age distribution of users is consistent with the hypothesis
695
that new users tend to be prescribed the newer products. It follows that the users of the older products are likely to be 'survivors' of women who started the older OC some years previously. The population utilization rates in the GPRD database, for any COC for each year between 1992 and 1997 indicate the highest rates amongst 20-24 year olds, peaking in 1993 at 345 years of use per 1000 observed woman-years. Women aged 15-19 years showed a decline in use each year of -4% from 1992, with a sharp decline of 12% from 163 to 143 per 1000 observed woman-years between 1995-1996. There was a sharp fall in use amongst the 20-24 year olds (16%) and a less marked decline amongst the 25-29 (12%) and 30-34 year olds (12%) between 1995 and 1996. The fall was sustained in 1997. Women aged >35 years were less affected by the 1995 announcement.
Table IV. Age-specific incidence rates for venous thromboembolism in the General Practice Research Database (GPRD, 1999) and MediPlus (1999) studies GPRD
MediPlus
(years)
Cases
EWYxiO
15-19 20-24 25-29 30-34 35-39 40-44 45-49 Total
24 54 82
8.6 22.0 22.9 15.4 6.9 2.3 0.6 78.7
73 31 22 10 296
4
RatexiO
4
Cases
.
10 24 24
2.8 2.5 3.6 4.7
EWYx10 4
RatexiO 4
2.7 6.1 6.0 4.1
3.7 3.9 4.0 4.6
4.5 9.6 16.7
19 12 6 4
1.9 0.6 0.2
3.8
99
21.6
6.3 10.0 20.0 4.6
EWY = exposed woman-years.
Table V. Product-specific crude incidence rates for venous thromboembolism between 1992 and 1997 in the General Practice Research Database (GPRD, 1999) and MediPlus (1999) studies Product
MediPlus
GPRD Cases
No. years woman years
Crude rate (per 10 000 EWY)
Cases
No. woman years
Crude rate (per 10 000 EWY)
Levonorgestrel 150 ug + EE 30 ug
64
190191
3.4
22
49 484
4.4
Desogestrel 150 ug + EE 30 ug Gestodene 75 ug + EE 30 )ig
65
152 524
4.3
23
39 679
5.8
63
143 581
4.4
21
41 947
5.0
Triphasic levonorgestrel Desogestrel 150 ug + EE 20 ug
26 18
102 787 37 584
2.5
10
4.8
9
27 512 10 426
3.6 8.6
Norgestimate 250 ug + EE 35 ug
15
40 440
3.7
4
13 681
2.9
CPA + EE 35 ug
16
25 709
2.2
4
8090
4.9
EWY = exposed woman-years; EE = ethinyloestradiol; CPA cyproterone acetate.
Downloaded from humupd.oxfordjournals.org by guest on July 17, 2011
Age
696
R.D.T.Farmer et al.
Cohort findings in the GPRD and MediPlus studies A total of 296 cases of idiopathic VTE disease amongst women exposed to a COC were identified in the GPRD. In the MediPlus database, 99 cases were identified. Incidence rates derived from MediPlus tended to.be slightly higher than from the GPRD. (Table IV). The incidence rates rise with age in both databases: steeply after the age of 30 years. Case exposure and the product-specific crude incidence rates are shown in Table V. In both databases the crude rates of VTE amongst users of levonorgestrel products were lower than those amongst users of desogestrel- or gestodene-based products. The crude incidence rate amongst users of the triphasic levonorgestrel compound were lower than those using the main monophasic preparation. There were 9 deaths amongst the cases in the GPRD giving an overall case fatality rate of 3% and a mortality rate of 10 per million EWY.
Nested case-control studies in the GPRD and MediPlus studies The principal characteristics of cases and controls in the study are summarised in Table VI. In both data sets the proportion of women with high BMI and those who had asthma were greater amongst cases than controls, these differences being statistically significant in the. GPRD data. There was a significant excess of smokers amongst the cases compared with the controls in both data sets. In both data sets a higher proportion of cases had no record of a blood pressure measurement in their records than controls — significant in the GPRD analysis. For both data sets there was an excess of cases compared with controls who had been prescribed drugs other than OC and specific asthma treatments, in the 6 months preceding the event date.
GPRD year-of-birth controls
MediPlus year-of-birth controls
Cases(%)
Cases(%)
Controls (%)
Body mass index category (kg/m2) 155(14.1) 35 20(1.8) Unknown 51 (17.9) 160(14.6) Significance X2 = 35.75; df = 5;P35 kg/m2 Smoker Smoking status unknown >3 non-OC/non-asthma scripts Levonorgestrel Desogestrel Gestodene Norgestimate Asthma No record of BP BMI 30-35 kg/m2 BMI >35 kg/m2 Smoker Smoking status unknown >3 non-OC/non-asthma scripts
Pulmonary embolism
704
R.D.T.Farmer et al.
The ORs in the BCDSP study are higher than those found in the study of the GPRD reported here (and higher than the WHO). Although the two studies were conducted on the same database there were differences in the study design that limit their comparability. The main differences are as follows, (i) This study used data accruing between 1992 and mid 1997. The BCDSP used data from 1991 to 1994. (ii) The BCDSP study was restricted to desogestrel, gestodene or levonorgestrel users. The present study included women who were using any COC on the event day, or equivalent in the controls. Table XIII. Differences in the proportionate use of combined oral contraceptives in England and Wales during the periods covered by the World Health Organization (WHO) and Transnational studies WHO 1989-1993
Transnational 1993-1995
23.6 16.6
18.5 12.4
Desogestrel 150 Hg + EE 30 ng
21.8
23.0
Desogestrel 150 fig + EE 20 ng
3.5
6.0
17.5
22.4
Gestodene 75 (ig + EE 30 (ig Norgestimate 250 ng + EE 35 ng Others
1.5
4.4
15.5
13.3
EE = ethinyloestradiol.
(iii)
(iv)
(v)
The BCDSP study excluded fatalities and analysed them in a separate study. The study reported here included fatalities. The BCDSP study was restricted to women aged
