Human Reproduction Vol.18, No.3 pp. 515±521, 2003

DOI: 10.1093/humrep/deg090

Differences in the use of combined oral contraceptives amongst women with and without acne H.E.Seaman1, C.S.de Vries and R.D.T.Farmer Pharmacoepidemiology Unit, Postgraduate Medical School (University of Surrey), Guildford, Surrey, UK 1

To whom correspondence should be addressed at: Pharmacoepidemiology Unit, Postgraduate Medical School (University of Surrey), Stirling House, Stirling Road, Surrey Research Park, Guildford, Surrey GU2 7DJ, UK. E-mail: [email protected]

BACKGROUND: Cyproterone acetate combined with ethinyl estradiol (CPA/EE) provides a treatment option for women with acne, hirsutism or polycystic ovary syndrome (PCOS). CPA/EE may be prescribed as an oral contraceptive (OC), but is not licensed as such in the UK. The use of CPA/EE steadily increased after its introduction to the UK market in 1987, but there was a marked increase in its share of the OC market after 1995. METHODS: Using the General Practice Research Database, utilization patterns of CPA/EE and conventional oral contraceptives were compared in women aged 15±39 years, with or without acne or PCOS. RESULTS: Between 1994 and 1998, CPA/EE accounted for an increasing proportion of all OC use. The proportion of CPA/EE prescribed to women with acne declined between 1994 and 1998, whereas that prescribed to women with PCOS remained constant. The age-speci®c use of CPA/EE by women with acne or PCOS almost doubled. After 1995, there was a marked increase in the use of products containing levonorgestrel by women with acne or PCOS. CONCLUSIONS: A large proportion of CPA/EE is prescribed to women with acne and/or PCOS, although this proportion decreased between 1992 and 1998. This has important implications in CPA/EE risk assessment studies. Key words: cyproterone acetate/ethinyl estradiol/oral contraception/utilization

Introduction Acne is generally associated with the onset of puberty and increased sebum production from the hair follicles on the face, chest and back. Propionibacterium acnes may proliferate within blocked pilosebaceous units and cause comedones and in¯amed pustules. The severity of acne varies, as does the tolerance to the condition demonstrated by affected individuals. Many women rely on over-the-counter preparations and/or cosmetics, and not all women with acne will consult a general practitioner (GP). The pattern of consultation with a GP is therefore not a reliable measure of acne severity. Conventional medical treatments for acne act by reducing sebum production, unblocking pores, delivering antibiotics or suppressing the effects of androgens. Some combined oral contraceptives (COCs) may be helpful in alleviating acne. Whilst none of the progestogens in conventional COCs is antiandrogenic, desogestrel has low androgenic activity, as does gestodene (Speroff, 1993). Consequently, in women with acne, COC formulations containing desogestrel or gestodene are likely to be prescribed in preference to COCs containing the more androgenic progestogen, levonorgestrel. In the UK, the anti-androgen cyproterone acetate (CPA 2 mg) combined with ethinyl estradiol (EE 35 mg) is licensed for the treatment of women with acne that is refractory to prolonged antibiotic therapy and women with moderately ã European Society of Human Reproduction and Embryology

severe hirsutism; it also provides contraception. Acne and hirsutism are commonly associated with polycystic ovary syndrome (PCOS) (Rittmaster, 1997; Guzick, 1998). Other common features of PCOS include obesity, secondary infertility, dyslipidaemia and insulin resistance, independent of obesity (Taylor, 1998; Solomon, 1999; Atiomo et al., 2000; Kelly et al., 2000). Some women with acne will exhibit other PCOS co-morbidities and thus present an intrinsically adverse cardiovascular risk pro®le. In October 1995, the UK Committee on Safety of Medicines informed doctors and pharmacists that the risk of venous thromboembolism (VTE) amongst women using COCs containing desogestrel or gestodene was twice that amongst women using COCs containing other progestogens (Committee on Safety of Medicines, 1995). The use of COCs containing desogestrel and gestodene consequentially declined (Figure 1). Figure 2 illustrates a coincident increase in the COC market share occupied by CPA/EE (as a proportion of all COC cycle packs sold) from 1995. Trend lines for the periods 1987 to 1995 and 1995 to mid-year 2000 indicate a marked increase in CPA/ EE market share after 1995. In the present report, data from the General Practice Research Database (GPRD) are used to compare patterns of the use of conventional COCs and CPA/EE in young women by the presence or absence of acne and/or PCOS. In addition, the 515

H.E.Seaman, C.S.de Vries and R.D.T.Farmer

Figure 1. Trends in oral contraceptive market share in the UK by progestogen. (Source: company sales data, Schering AG, personal communication.)

Figure 2. Trends in UK CPA/EE sales. (Source: company sales data, Schering AG, personal communication.)

characteristics of CPA/EE users and COC prescribing practice in women with or without acne or PCOS during the years 1992 to 1998 are described. Materials and methods The GPRD contains the anonymized patient records for about 4% of the UK population at any one time. Data have been collected from 1987, and since the beginning of 1992 most contributing practices have provided good quality data. The validity and utility of this database for studies of drug safety have been published elsewhere (Van Staa and Abenhaim, 1994; Hollowell, 1997; Walley and Mantgani, 1997; Lawrenson et al., 2000). The number of practices contributing data to the GPRD has changed in recent years: over the past 5 years about 300 of the initial 750 practices have ceased to contribute, either because of concerns about the quality of their data or because they have changed to another computer system. This report is restricted to practices that have contributed data deemed to be of research standard by the owners of the database (the UK Medicines Control Agency). In addition to demographic information, the GPRD contains three types of coded data. Medical diagnoses are coded to an OXMIS (Oxford Medical Indexing System) or Read code. Prescribed drugs are recorded with reference to Prescription Pricing Authority (PPA) drug codes, dosage and quantity. Prescribing information is around 95% complete (Walley and Mantgani, 1997). Other information related to smoking habits, alcohol consumption, height and bodyweight are recorded with some incompleteness that varies by practice.

516

The study population comprised women aged between 15 and 39 years who were registered with a general practitioner who provided data of research quality to the GPRD at any time during the study period (January 1, 1992 to December 31, 1998). Patients could join the study cohort (by registering with one of the contributing practices) or leave (by changing their doctor, as a result of death or by the doctor ceasing to contribute to the GPRD) at any time during the study period. For each woman, her age (calculated as observation year minus year of birth) and the number of days on which was she was registered with the practice during each calendar year were calculated. The agespeci®c number of woman-years of observation [observed women years (OWY)] was aggregated in 5-year age bands (by summing the number of days on which each woman was registered during each calendar year of observation within each year of age and dividing by 365). COC and CPA/EE prescriptions for all women with at least 6 months' prior data were identi®ed and used to map drug use to each day in the study period (Farmer et al., 1999). Where there were prescriptions for the same product and the periods covered by the two prescriptions overlapped, the duration of overlap was added to the logical end date of the last overlapping script. Overlapping prescriptions were de®ned as two prescriptions where the start date of the second was within 14 days of the logical end-date of the ®rst (represented by the start date plus duration of exposure). The 14-day period was used because any new prescription could ordinarily be started any time between 1 and 28 days after the day of issue. Nonoverlapping scripts with a gap between them of more than 14 days were represented in the ®le as discontinuous periods of exposure, whether or not there was a change in product. In cases where there were overlapping prescriptions for different products, the exposure to the ®rst product was truncated to the day of the switch and a new period of exposure to the switched product assumed from the subsequent day. Thus, all logically unused packs of oral contraceptives were discounted in the calculation of exposed women years (EWY). Women with a diagnosis of acne, polycystic ovaries or PCOS were identi®ed using appropriate OXMIS and Read codes. Because of its chronic nature, women were deemed to have acne in the year of an acne diagnosis, in the subsequent year, and in the 2 years prior to diagnosis. For women with more than one diagnosis of acne the acne period extended from 2 years prior to the ®rst diagnosis until 1 year after the last. All women with a diagnosis of polycystic ovaries or PCOS at any time were included in the PCOS cohort. Also included were those women who at any time had three or more markers for PCOS (acne, hirsutism/alopecia, anovulation/infertility, amenorrhoea/ oligomenorrhoea, obesity or endocrinological abnormalities). A cohort of CPA/EE users was described by age and by presence of acne and/or PCOS as well as the proportionate use of CPA/EE amongst all COC use. Amongst all women with or without acne or PCOS, age-speci®c use of the different COCs was calculated.

Results Table I shows the patterns of CPA/EE use for the years 1992 to 1998. In all age groups the use of CPA/EE increased throughout the period covered, with the highest overall use among women aged 20±24 years. CPA/EE use as a percentage of all COC preparations increased in all age groups between 1992 and 1998; for all ages combined, CPA/EE use almost tripled from 2.4 to 7.0% during the 7-year period. At the same time, the proportion of all CPA/EE that was prescribed to women with acne was nearly halved. The proportion of CPA/

2.5 4.4 7.1 8.5 10.3 5.8 3.0 4.7 7.8 8.4 9.8 6.1 3.6 4.9 7.4 8.7 7.4 6.1

PCOS

Ages

1992

1994

1992

1994

1996

1998

15±19 20±24 25±29 30±34 35±39 All ages

133.4 71.8 56.8 47.9 35.9 66.2

135.1 70.1 52.3 44.6 34.9 63.7

7.6 10.1 10.5 7.7 3.6 8.1

5.9 10.2 12.3 10.1 5.4 9.1

4.3 10.0 12.9 12.5 7.9 9.9

2.7 8.6 13.2 13.8 9.9 10.1

9.5 8.1 6.3 4.8 3.7 7.0

69.6 56.5 51.9 51.2 44.6 56.0

66.2 48.2 42.7 40.1 38.7 47.8

62.6 41.7 37.9 35.0 37.0 43.0

51.2 30.8 26.9 24.9 22.3 32.3

3.3 5.0 5.2 5.5 5.8 4.9

aPrevalence calculated as: acne/PCOS observed women years/1000 observed women years.

CPA = cyproterone acetate; EE = ethinylestradiol; OWY = observed women years.

7.0 5.9 4.5 3.5 2.9 5.1 4.6 3.6 3.0 2.4 2.3 3.3 3.1 2.5 2.2 1.9 1.6 2.4 1.5 2.2 1.6 0.9 0.4 1.3 1.1 1.7 1.2 0.7 0.3 1.0 0.6 0.8 0.7 0.4 0.1 0.5 15±19 20±24 25±29 30±34 35±39 All ages

0.8 1.2 0.9 0.5 0.3 0.7

1996 1994 1992 1996 1994 1992 1992 Age (years)

1994

1996

1998

1992

1994

1996

1998

Percentage of CPA/EE prescribed to women with acne CPA/EE as % of all COC use Age-speci®c use of CPA/EE (years of use/100 OWY)

Table I. Patterns of CPA/EE use, 1992±1998 (Source: GPRD)

Table II. Estimated prevalencea of acne and polycystic ovary syndrome (PCOS), 1992±1998 (Source: GPRD) Acne

1998

Percentage of CPA/EE prescribed to women with PCOS

1998

Use of combined oral contraceptives in acne

EE prescribed to women with PCOS increased with age and remained relatively constant between 1992 and 1998, except in older women for whom there was a 45% increase in the proportion of CPA/EE prescribed to women with PCOS. Table II shows the estimated prevalence of acne and PCOS by year. The study period was truncated at the end of 1998, so women with a ®rst diagnosis of acne after that time would not have contributed to the acne cohort. Only women with acne diagnoses after 1991 and before the end of 1997 might have contributed the full complement of acne time to the numerator according to our de®nition of the `acne period', although that period would have been truncated at 1998 for any woman with a further diagnosis of acne in 1998. An additional complication arises from the decline in the number of practices contributing data to the GPRD during the study period and the consequent right-truncation of data for almost half of the study population. Consequently, in Table II acne prevalence data are presented for the years 1992 and 1994 only. Prevalence estimates for PCOS would be unaffected by these right-censoring issues as the numerator was derived from the total time registered with the contributing practice. Thus, PCOS prevalence data for 1992 to 1998 are presented in Table II. The prevalence of acne decreased with increasing age, whilst the prevalence of PCOS was highest among women aged between 25 and 34 years; increased case ascertainment would result when women with menstrual or fertility problems present to the general practitioner during childbearing years. The overall prevalence of acne and PCOS appeared to have remained fairly stable during the study period although the prevalence of PCOS declined in younger women (aged 15±24 years) and increased in older women (aged 25±39 years). A total of 32 889 women who had been prescribed CPA/EE between the ages of 15 and 39 years was identi®ed. The medical record at the time of ®rst prescription implied oral contraception as the indication in almost 44% of women; in about 35% the implied indication at the time of ®rst prescription for CPA/EE was acne or disorders suggestive of PCOS. However, 66% of CPA/EE users had a diagnosis of acne on or at some time before the date of their ®rst prescription for CPA/EE. Some 23% of women prescribed CPA/EE had no recorded indication for CPA/EE (an acne 517

H.E.Seaman, C.S.de Vries and R.D.T.Farmer Table III. Use of different formulations of COC and CPA/EE amongst women with and without acne by progestogen, 1992±1998 (Source: GPRD) Age (years)

Progestogen

Age-speci®c use of COCs (years of use per 100 women-years) Women with acne 1992

1994

Women without acne 1996

1998

1992

1994

1996

1998

15±19

Levonorgestrel Desogestrel Gestodene Cyproterone acetate Othersa

3.2 4.9 3.8 3.0 1.3

2.0 4.7 4.1 4.0 1.6

7.2 0.5 0.3 5.4 3.3

7.2 0.6 0.2 8.9 3.2

4.6 6.8 5.6 0.2 1.5

2.7 6.7 6.2 0.3 1.8

9.6 0.9 0.8 0.5 3.9

10.2 0.5 0.3 0.8 3.9

20±24

Levonorgestrel Desogestrel Gestodene Cyproterone acetate Othersa

11.0 9.0 7.0 6.6 3.6

6.5 9.6 8.4 8.1 3.7

14.0 1.8 1.5 10.9 6.4

11.3 1.4 1.3 15.2 5.8

12.6 10.0 7.2 0.4 3.8

7.9 11.0 9.4 0.7 3.5

15.6 2.6 2.2 1.0 6.5

14.7 1.9 1.7 1.6 6.9

25±29

Levonorgestrel Desogestrel Gestodene Cyproterone acetate Othersa

11.8 7.1 4.6 5.9 3.4

8.8 8.3 5.9 7.2 3.8

13.6 2.0 1.5 9.2 6.0

11.3 1.9 1.3 13.1 5.1

13.5 6.6 4.8 0.3 3.7

10.4 7.9 6.4 0.5 3.8

15.3 2.2 1.7 0.8 5.8

13.7 2.1 1.7 1.2 5.9

30±34

Levonorgestrel Desogestrel Gestodene Cyproterone acetate Othersa

7.9 4.5 3.2 3.9 2.0

6.7 5.5 4.2 4.5 2.7

9.6 1.2 0.8 5.9 4.0

8.5 1.0 0.5 8.5 3.5

9.5 4.2 3.1 0.2 2.2

8.4 5.2 4.2 0.3 2.5

11.8 1.4 1.0 0.5 4.0

11.3 1.3 1.0 0.7 4.2

35±39

Levonorgestrel Desogestrel Gestodene Cyproterone acetate Othersa

3.5 2.3 1.7 1.8 0.8

3.5 3.0 2.3 2.8 1.2

5.7 0.8 0.4 3.6 1.9

5.7 1.0 0.4 4.3 2.3

4.2 2.2 1.4 0.1 1.0

4.2 3.0 2.0 0.2 1.2

6.4 1.0 0.6 0.2 2.2

6.7 0.9 0.6 0.3 2.3

All ages

Levonorgestrel Desogestrel Gestodene Cyproterone acetate Othersa

7.5 6.0 4.4 4.4 2.3

5.2 6.4 5.3 5.5 2.6

10.1 1.2 0.9 7.2 4.4

8.8 1.1 0.7 10.3 4.0

9.4 6.0 4.4 0.2 2.6

7.1 6.8 5.6 0.4 2.7

11.9 1.7 1.3 0.6 4.5

11.3 1.4 1.1 0.9 4.6

aOther

progestogens include ethynodiol, lynoestrenol, mestranol, norethisterone, norgestimate and norgestrel.

diagnosis, prescription for an acne treatment, PCOS-like condition) before or on the date of their ®rst prescription for CPA/EE, and 21% had no recorded indication for CPA/EE treatment anywhere in the records. These data are limited by the data censoring imposed by registration periods on the GPRD, and may also be limited by the quality of recording. To explore the possibility that these observations might be affected by the data censoring issue, women prescribed CPA/ EE for whom there was at least 2 years' data prior to the ®rst prescription for CPA/EE and at least 1 year's data after the ®rst prescription for CPA/EE were identi®ed (n = 11 994). Of these women, 10 190 (85%) had a diagnosis of acne, PCOS or hirsutism at some time. Altogether, 21% of women with a diagnosis of acne had been prescribed CPA/EE at some time (19 782/94 717). Of those women, 90% had a diagnosis of acne on or before the date of the ®rst CPA/EE prescription, and the remainder had been prescribed an acne-speci®c treatment prior to the ®rst CPA/EE prescription. In total, 9629 women had PCOS, and 2082 of these (22%) had been prescribed CPA/EE at some time between 1992 and 1998. Table III shows the changes in the patterns of use of the major COC formulations and CPA/EE amongst women with 518

and without acne. In 1994, the use of products containing levonorgestrel declined, particularly in younger women. Amongst women with and without acne there were similar patterns of use of products containing desogestrel and gestodene in all age groups. The use of CPA/EE in women with acne reached levels comparable with desogestrel and gestodene, although in women without acne the use of CPA/EE was much less. By 1996 there had been a sharp fall in the use of both gestodene- and desogestrel-based COCs by women both with and without acne and of all ages. In women with acne there was a rise in the use of levonorgestrel and CPA/EE, but the proportionate increase in the use of levonorgestrel was greater than that for CPA/EE. For example, there was a 3.6-fold increase in levonorgestrel use among 15- to 19-year-olds with acne between 1994 and 1998 compared with a 2.2-fold increase in the use of CPA/EE during the same period (P < 0.001). The proportionate increase in the use of levonorgestrel-based products was less amongst the women aged >20 years with a history of acne, but in each age group the proportionate increase in levonorgestrel use was greater than that for CPA/EE (P < 0.01). Between 1996 and 1998, the pattern changed: the use of levonorgestrel products altered little amongst women with acne, whereas the use of CPA/EE increased. In women

Use of combined oral contraceptives in acne Table IV. Use of different formulations of COC amongst women with and without PCOS by progestogen, 1992±1998 (Source: GPRD) Age (years)

Progestogen

Age-speci®c use of COCs (years of use per 100 women-years) Women with PCOS

Women without PCOS

1992

1994

1996

1998

1992

1994

1996

1998

15±19

Levonorgestrel Desogestrel Gestodene Cyproterone acetate Others

3.9 6.4 5.5 2.5 1.3

1.8 7.8 6.3 4.9 1.8

8.4 0.5 0.3 7.9 4.8

5.7 0.2 1.0 14.4 3.9

4.5 6.6 5.3 0.6 1.5

2.6 6.4 5.9 0.8 1.8

9.3 0.9 0.7 1.1 3.8

10.0 0.5 0.3 1.5 3.9

20±24

Levonorgestrel Desogestrel Gestodene Cyproterone acetate Others

7.1 8.0 5.3 4.1 2.3

3.4 8.2 6.3 5.7 2.2

8.7 1.3 0.6 7.8 5.1

8.3 0.8 0.7 11.3 3.9

12.5 10.0 7.2 0.8 3.8

7.9 11.0 9.4 1.1 3.5

15.6 2.6 2.1 1.6 6.5

14.6 1.9 1.7 2.1 6.9

25±29

Levonorgestrel Desogestrel Gestodene Cyproterone acetate Others

6.0 5.6 4.5 3.2 1.9

4.3 6.3 4.6 5.3 1.7

8.3 1.2 0.9 7.0 3.4

6.5 0.9 0.6 8.4 4.5

13.4 6.7 4.8 0.6 3.7

10.4 7.9 6.4 0.8 3.8

15.3 2.2 1.7 1.1 5.9

13.7 2.2 1.7 1.5 5.9

30±34

Levonorgestrel Desogestrel Gestodene Cyproterone acetate Others

3.2 3.4 2.0 2.6 0.8

2.2 4.2 3.1 4.3 1.4

6.0 0.7 0.6 4.4 2.4

5.8 0.8 0.7 5.4 2.1

9.5 4.2 3.1 0.3 2.2

8.4 5.3 4.2 0.5 2.5

11.8 1.4 1.0 0.6 4.1

11.3 1.3 0.9 0.8 4.2

35±39

Levonorgestrel Desogestrel Gestodene Cyproterone acetate Others

1.2 1.2 1.2 2.3 0.5

1.3 2.7 0.9 3.4 0.5

2.2 0.8 0.2 3.8 0.8

2.5 0.7 0.3 4.2 1.0

4.2 2.2 1.4 0.1 1.0

4.2 3.0 2.1 0.2 1.3

6.4 0.9 0.6 0.3 2.2

6.8 0.9 0.6 0.4 2.4

All ages

Levonorgestrel Desogestrel Gestodene Cyproterone acetate Others

5.0 5.6 4.1 3.2 1.6

3.0 5.9 4.3 4.9 1.6

6.8 1.0 0.6 6.0 3.1

5.7 0.8 0.6 7.3 2.9

9.3 6.0 4.4 0.5 2.6

7.0 6.8 5.6 0.7 2.7

11.9 1.6 1.3 0.9 4.5

11.3 1.4 1.0 1.2 4.6

EE = ethinylestradiol.

without acne, similar increases in the use of products containing levonorgestrel by all age-groups were sustained through to 1998. The use of CPA/EE by women without acne increased only fractionally. Table IV shows the changes in the patterns of use of the major COC formulations and CPA/EE in women with and without PCOS. The use of CPA/EE and of products containing levonorgestrel increased over time in women with and without PCOS, whilst the use of products containing desogestrel or gestodene decreased considerably. There was greater use of CPA/EE by women with PCOS than by those without PCOS. Overall, the use of COCs (including CPA/EE) by women without acne fell following the `pill scare' (Figure 3), but among women with acne it remained largely unchanged (Figure 4). A similar pattern of all COC use was observed in women with and without PCOS (data not shown). Sensitivity analysis The de®nition of an `acne period' as 2 years prior to ®rst acne diagnosis until 1 year after last diagnosis arose through a considered, though arguably arbitrary, process. To assess the impact of an alternative de®nition the data were re-analysed based on an `acne period' commencing 1 year prior to the ®rst

Figure 3. Use of all combined oral contraceptives (including CPA/ EE) by women without acne. (Source: GPRD.)

acne diagnosis and ending 2 years after the last acne diagnosis. Acne prevalence was comparatively lower in the younger women and higher in the older women. A decline in the percentage of CPA/EE prescribed to women with acne similar to that based on the original `acne period' was observed. Agespeci®c rates of use among women without acne were 519

H.E.Seaman, C.S.de Vries and R.D.T.Farmer

Figure 4. Use of all combined oral contraceptives (including CPA/ EE) by women with acne. (Source: GPRD.)

unchanged for all products. In 1992, in younger women (aged 15±24 years) with acne, the age-speci®c use of most products was decreased slightly by the change in acne period de®nition, although the use of CPA/EE increased slightly. In 1994 and 1996, a slight increase in the age-speci®c use of all products by the youngest women was apparent with the change in de®nition, but in women aged 20±24 years the age-speci®c utilization rates decreased a little or remained constant for all products except CPA/EE. By 1998, the use of all products was increased slightly or unaffected by the changed de®nition, although the use of CPA/EE declined. The patterns of change in CPA/EE use according to the de®nition of the `acne period' must be due in part to the prescription of CPA/EE to treat diagnosed acne. Discussion CPA/EE is not licensed for use as an oral contraceptive in the UK, although the present analysis shows that for some women prescribed CPA/EE, no legitimate indication for CPA/EE treatment was recorded. CPA/EE accounts for a relatively small market share of oral contraceptives. Since the introduction of CPA/EE in the UK in 1987, sales have steadily increased, and after the ``pill scare'' in 1995 the market share of COCs occupied by CPA/EE more than doubled (from 4% in 1995 to nearly 9% by June 2000). Using data extracted from the GPRD, an increase in use of CPA/EE has been described both in terms of use as a proportion of all COCs and in agespeci®c use between 1992 and 1998. Several studies have found an apparent increased risk of VTE associated with exposure to CPA/EE (Farley et al., 1995; Farmer et al., 1999; Lidegaard et al., 2002; Parkin et al., 2000; Vasilakis-Scaramozza and Jick, 2001), although efforts to adjust those risk estimates for confounding have been mostly inadequate. Only in the last-mentioned study (VasilakisScaramozza and Jick, 2001) did the authors adjust for the presence of acne, hirsutism and PCOS. Because the investigators considered all CPA/EE users and a sample of levonorgestrel users, however, the population will have been a mixture of CPA/EE users more likely to have acne, hirsutism and PCOS than women prescribed COCs containing the more androgenic progestogen levonorgestrel. It is dif®cult to tell whether the adjusted analysis in that study would have adequately 520

controlled for all potential confounding by acne, hirsutism and PCOS. The present observations on changes in the pattern of CPA/ EE sales and the issue of VTE risk raise important questions. First, is CPA/EE associated with an increased risk of VTE? If this is the case, then increased use of CPA/EE as a result of the `pill scare' would have been counterproductive. Second, is CPA/EE selectively prescribed to women with a higher intrinsic risk of VTE, associated with underlying co-morbidity and if so, will the results of the study by Vasilakis-Scaramozza and Jick have been adequately adjusted? The present analysis of secular trends in CPA/EE prescribing indicates that the proportion of CPA/EE prescribed to women with acne declined between 1992 and 1998. At the same time, the age-speci®c use of CPA/EE by the youngest women with a diagnosis of acne almost tripled. The proportion of CPA/EE prescribed to women with PCOS has remained fairly constant (except in older women), although the age-speci®c use of CPA/EE has increased considerably in younger women with PCOS. Of particular interest are the secular trends in the use of COCs containing different progestogens in women with acne or PCOS. In younger women with acne in 1994, the agespeci®c use of CPA/EE and COCs containing desogestrel and gestodene was greater than the use of COCs containing levonorgestrel. In 1996, however, subsequent to the `pill scare', the use of products containing levonorgestrel amongst those women with acne increased to levels comparable with CPA/ EE. A similar trend was observed for women with PCOS. The observation that in later years more CPA/EE was being prescribed to women without acne could be because the recording of acne was declining or it could also be an effect of the `pill scare'Ðwomen with troublesome skin that had not been recorded as acne who, prior to October 1995 had been using the less androgenic COC formulations containing gestodene and desogestrel, switched to CPA/EE as a consequence of the `pill scare'. It appears, therefore, that as a result of the `pill scare' women may have stopped using COCs containing desogestrel and gestodene in favour of COCs containing the more androgenic progestogen, levonorgestrel or CPA/EE. CPA/EE use has been associated with an increased risk of VTE compared with levonorgestrel (Vasilakis-Scaramozza and Jick, 2000), but it remains unclear whether this is a drug effect or because the women prescribed CPA/EE in this study have an inherently adverse VTE risk pro®le. The results of the present study clearly demonstrate that a large proportion of CPA/EE is prescribed to women with acne and/or PCOS, although this proportion appears to decrease with calendar time. This has implications for the evaluation of VTE risk associated with CPA/EE use: analyses should be strati®ed by use before versus after the pill scare, and the results should be adjusted for the presence of acne and/or PCOS. Investigators should be aware of the potential for residual confounding given that acne is probably under-recorded. In conclusion, the morbidity of CPA/EE users needs to be more fully explored in order to understand the underlying risk pro®le of women prescribed CPA/EE and anticipate any future risk assessment of CPA/EE.

Use of combined oral contraceptives in acne

References Atiomo, W.U., Fox, R., Condon, J.E., Shaw, S., Friend, J., Prentice, A.G. and Wilkin, T.J. (2000) Raised plasminogen activator inhibitor-1 (PAI-1) is not an independent risk factor in the polycystic ovary syndrome (PCOS). Clin. Endocrinol., 52, 487±492. Committee on Safety of Medicines (1995) Combined Oral Contraceptives and Thromboembolism. Committee on Safety of Medicines, London. Farley, T.M.M, Meirik, O., Chang, C.L., Marmot, M.G. and Poulter, N.R. (1995) World Health Organization Collaborative Study of Cardiovascular Disease and Steroid Hormone Contraception. Effect of different progestagens in low oestrogen oral contraceptives on venous thromboembolic disease. Lancet, 346, 1582±1588. Farmer, R.D.T., Lawrenson, R.A., Todd, J.-C., Williams, T.J. and MacRae, K.D. (1999) Oral contraceptives and venous thromboembolic disease. Analysis of the UK General Practice Research Database and the UK MediPlus Database. Hum. Reprod. Update, 5, 688±706. Guzick, D. (1998) Polycystic ovary syndrome: symptomatology, pathophysiology, and epidemiology. Am. J. Obstet. Gynecol., 179, S89±S93. Hollowell, J. (1997) The General Practice Research Database: quality of morbidity data. Popul. Trends, 87, 36±40. Kelly, C.J., Connell, J.M., Cameron, I.T., Gould, G.W. and Lyall, H. (2000) The long term health consequences of polycystic ovary syndrome. Br. J. Obstet. Gynaecol., 107, 1327±1338. Lawrenson, R., Todd, J.C., Leydon, G.M., Williams, T.J. and Farmer, R.D.T. (2000) Validation of the diagnosis of venous thromboembolism in general practice database studies. Br. J. Clin. Pharmacol., 49, 591±596.

Lidegaard, O., EdstroÈm, B. and Kreiner, S. (2002) Oral contraceptives and venous thromboembolism: a ®ve-year national case-control study. Contraception, 65, 187±196. Parkin, L., Skegg, D.C.G., Wilson, M., Herbison, G.P. and Paul, C. (2000) Oral contraceptives and fatal pulmonary embolism. Lancet, 355, 2133±2134. Rittmaster, R.S. (1997) Hirsutism. Lancet, 349, 191±195. Solomon, C.G. (1999) The epidemiology of polycystic ovary syndrome. Prevalence and associated disease risks. J. Endocrinol. Metab. Clin. North Am., 28, 247±263. Speroff, L. (1993) Evaluation of a new generation of oral contraceptives. Obstet. Gynecol., 81, 1034±1047. Taylor, A.E. (1998) Understanding the underlying metabolic abnormalities of polycystic ovary syndrome and their implications. Am. J. Obstet. Gynecol., 179, S94±S100. Van Staa, T.P. and Abenhaim, L. (1994) The quality of information on a UK database of primary care records: a study of hospitalizations due to hypoglycaemia and other conditions. Pharmacoepidemiol. Drug Safety, 3, 15±21. Vasilakis-Scaramozza, C. and Jick, H. (2001) Risk of venous thromboembolism with cyproterone or levonorgestrel contraceptives. Lancet, 358, 1427±1429. Walley, T. and Mantgani, A. (1997) The UK General Practice Research Database. Lancet, 350, 1097±1099. Submitted on May 10, 2002; resubmitted on September 3, 2002; accepted on November 7, 2002

521