Contemporary hormonal contraception and cervical cancer in women of reproductive age

To determine cervical cancer risk associated with contemporary hormonal contraceptives, we conducted a cohort study of women aged 15 to 49 living in Denmark from 1995 to 2014, using routinely collected information about redeemed prescriptions, incident cancer and potential confounders. Poisson regression calculated adjusted cervical cancer risks among different contraceptive user groups by duration of use, time since last use, hormonal content and cancer histology. During >20 million person‐years, 3643 incident cervical cancers occurred. Ever users of any hormonal contraceptives compared to never users had a relative risk (RR) of 1.19 (95% confidence interval [CI] 1.10‐1.29). Increased risks were seen in current or recent users of any hormonal: RR 1.30 (95% CI 1.20‐1.42) and combined: RR 1.40 (95% CI 1.28‐1.53), but not progestin‐only contraception: RR 0.91 (95% CI 0.78‐1.07). Current or recent users of any hormonal contraception had an increased risk of both adenocarcinoma (RR 1.29, 95% CI 1.05‐1.60) and squamous cancer (RR 1.31, 95% CI 1.19‐1.44). The risk pattern among any hormonal and combined contraceptive users generally increased with longer duration of use and declined after stopping, possibly taking longer to disappear among prolonged users. Combined products containing different progestins had similar risks. Approximately one extra cervical cancer occurred for every 14 700 women using combined contraceptives for 1 year. Most women in our study were not vaccinated against human papillomavirus (HPV) infections. Our findings reinforce the urgent need for global interventions such as systematic screening, treatment of cervical intraepithelial neoplasia and HPV vaccination programmes to prevent cervical cancer, especially among users of combined contraceptives.


| MATERIALS AND METHODS
The previously described Danish Sex Hormone Register Study 12,13 includes all women aged 15

| Statistical analysis
Analyses were conducted using SAS version 9.3 (SAS Institute, Inc, Cary, NC). During the study, women were categorised according to their use of hormonal contraception as current or recent (within 1 year of stopping); former (more than 1 year since stopping) or ever (any hormonal contraceptive use during the study period) users. Never users had no redeemed prescriptions for hormonal contraceptives recorded at study entry or during the study period. If a woman was a never user It is possible that some women begin using, or restart, hormonal contraception because they experience symptoms such as heavy bleeding, which are subsequently attributed to cervical cancer. It is also possible that some women have a cervical smear around the time of beginning or restarting hormonal contraception, providing an opportunity for cervical cancer to be detected. In either situation, a short-term increase in events could be observed among current users of hormonal contraceptives, due to these factors rather than any biological effects of the contraceptives. To investigate whether such protopathic bias might have occurred, we undertook sensitivity analyses in which periods of observation were ignored for 1 year before the date of cervical cancer diagnosis. This resulted in the exclusion of 243 women with less than 1 year of observation before cancer diagnosis.
We did not adjust for multiple comparisons. For the full cohort, we calculated age-standardised absolute risks (incidence exposed − incidence unexposed ) and the number needed to harm (1/incidence exposed − incidence unexposed ).

| RESULTS
A total of 2339 incident cervical cancers occurred during  (Table 1S). The relatively popular levonorgestrel-releasing intrauterine system (LNG-IUS) and desogestrel-containing progestin-only pills tended to be used by parous rather than nulliparous women, unlike other progestin-only products. Age-specific incidence of cervical cancer increased until age 40, whereupon it fell ( Table 1). The age-adjusted incidence of cervical cancer in never users was 14.9 per 100 000 person-years and in ever users of any hormonal contraceptives 17.8 per 100 000 person-years ( In analyses of both any hormonal and combined contraceptives, there was a trend of increasing cervical cancer risk with duration of current use. This relationship was not seen among users of progestinonly products. When the entire dataset was examined, there was no increased risk of cervical cancer among women who were more than 1 year since last current use (Table 2). However, when the data were T A B L E 1 Age-specific incidence per 100 000 of cervical cancer during the period 1995 to 2015 stratified by duration of use and time since last use, there was evidence that the risk among women with prolonged use may take longer to disappear (up to 10 years) than the risk among women with shorter-term use ( Table 3).
The overall risk estimates in the subset of women followed until their first switch in hormonal contraceptive were of similar magnitude to those seen in the full cohort (Table 4). We had insufficient data to calculate risk estimates for some of the products used including vaginal rings and contraceptive patches. Overall, there was little evidence of major differences in risk between combined products containing different progestins. Analyses where 30 to 35 μg ethinylestradiol plus levonorgestrel products formed the referent group (Table 2S), and when product-specific estimates were calculated among the full cohort (Table 3S), also found few differences between products. In both analyses, current or recent users of the LNG-IUS had a reduced cervical cancer risk when compared to current or recent users of 30 to 35 μg ethinylestradiol plus levonorgestrel products.
Larger, but very imprecise, point estimates were observed in the exploratory analysis among women with full contraceptive history (- Table 4S). Former users of any hormonal contraception had an increased risk of cervical cancer in this subset analysis: RR 4.35 (95% CI 1.57-12.00). More than three-quarters of former users had stopped within the previous 5 years (data not shown). In this subset, current or recent use of progestin-only products was associated with increased cervical cancer risk: RR 3.38 (95% CI 1.13-10.10). There was little evidence of differences in the risk estimates of combined products containing different progestins when compared against 30 to 35 μg ethinylestradiol plus levonorgestrel products (Table 4S). More never users in the full contraceptive history subset had received an HPV vaccination than ever users (24.0% vs 9.7% periods of observation, respectively) ( Table 5S). The risk estimates in Table 4S changed very little after also adjusting for HPV vaccination (data not shown).
Approximately three-quarters of the cervical cancers were squamous (Table 5). Current or recent users of any hormonal contraception had an increased risk of both adenocarcinoma and squamous tumour types.
Sensitivity analysis which excluded periods of observation 1 year before diagnosis in the full cohort found that short duration (<1 year) current use of combined or progestin-only hormonal contraceptives was no longer positively associated with cervical cancer (Table 6S).
A similar sensitivity analysis of women followed up until their first switch in hormonal contraception also found short-term current use of any hormonal and combined contraceptives did not increase the risk of cervical cancer (Table 7S).
Age-adjusted absolute risks were calculated for the main patterns of contraceptive use in the entire cohort (

| DISCUSSION
In this cohort study of mostly women unvaccinated for HPV, cur-  The results from both of these studies 8 29 we await the opportunity to determine whether HPV vaccination within our cohort reduces cervical cancer incidence. Such an effect would increase the number needed to harm from combined oral contraceptive use.
In our cohort of women mostly unvaccinated for HPV, we estimate that one extra case of cervical cancer occurred in our cohort for every 14 706 women who used contemporary combined contraception for 1 year. The absolute risk will be higher in countries where cervical cancer is more common than in Denmark; countries where comprehensive screening and HPV vaccination programmes are often absent and where perhaps two-thirds of all hormonal contraceptive users live. Women should be informed of the association between contemporary combined contraception and cervical cancer, an effect which is enhanced by prolonged use but which disappears some years after stopping. Such information should be balanced against high levels of protection against pregnancy and its associated mortality and morbidity, and other important non-contraceptive benefits including large sustained protection against ovarian 30 and endometrial 31 cancer.
Our results indicate that currently available combined contraceptives continue to be positively associated with the risk of cervical cancer, at least among women not vaccinated against HPV. Women wishing to use this method of contraception need to be informed of this risk and encouraged to participate in a cervical screening programme, if available. They should also be alert to any symptoms indicative of cervical cancer, and report these promptly to their health care provider. Our findings also reinforce the urgent public health need for global interventions to prevent cervical cancer.

CONFLICT OF INTEREST
The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT
The study data are held with the secure data depository at Statistics Denmark, permissions to access can only be granted by Statistics Denmark (https://www.dst.dk/en/TilSalg/Forskningsservice). Further details that support the findings of this study are available from the corresponding author upon request.

ETHICS STATEMENT
Ethical approval is not required for register-based studies in Denmark.
Approval was obtained from the Health Data Board and Danish Data Protection Agency. The data were analysed (by SF) and held within the secure data repository at Statistics Denmark, with personal identification numbers encrypted.