Use of contraceptives and risk of inflammatory bowel disease: a nested case–control study

Summary Background How contraceptive formulation, dose, duration of therapy and mode of delivery affects the risk of inflammatory bowel disease (IBD) is poorly described. Aim To examine associations between types of hormonal contraception and development of IBD. Methods This was a nested case-control study using IQVIA Medical Research Data. Women aged 15-49 years with a new diagnosis of IBD were matched with up to six controls by age, practice and year. Odds ratios (OR) and 95% confidence intervals (95% CI) for incident IBD and use of contraception were calculated. Results 4932 incident cases of IBD were matched to 29 340 controls. Use of combined oral contraceptive pills (COCPs) was associated with the development of Crohn's disease and ulcerative colitis (OR 1.60 [1.41-1.82] and 1.30 [1.15-1.45], respectively). Each additional month of COCP exposure per year of follow-up increased risk of Crohn's disease by 6.4% (5.1%-7.7%) and ulcerative colitis by 3.3% (2.1%-4.4%). Progestogen-only pills had no effect on Crohn's disease risk (OR 1.09 [0.84-1.40]) but there was a modest association with ulcerative colitis (OR 1.35 [1.12-1.64]). Parenteral contraception was not associated with the development of Crohn's disease or ulcerative colitis (OR 1.15 [0.99-1.47] and 1.17 [0.98-1.39], respectively). Conclusions We observed an increase in the risk of IBD with increasing duration of exposure to COCPs. Progestogen-only pills were not associated with Crohn's disease but there was a modest association with ulcerative colitis. There was no association between parenteral progestogen-only contraception and IBD. These findings are broadly consistent with a hypothesis that the oestrogen component of contraception may drive IBD pathogenesis.


| INTRODUC TI ON
Changes in the epidemiology of inflammatory bowel disease (IBD) across geographical location and time suggest that environmental risk factors play a major role in disease development. 1 In the UK, approximately 26% of women of reproductive age use hormonal contraception 2 and combined oral contraceptive pills (COCPs) which work by releasing an oestrogen and a progestogen are the most popular method. An increased risk of development of IBD in association with oral contraceptive pill exposure has been shown in numerous studies. [3][4][5] However, the precise biological mechanism remains unknown. A number of proposed theories exist, largely relating to the effect of exogenous oestrogen on immunomodulation, intestinal wall function, gut microbiome and hypercoagulability.
Oestrogen has been linked to inhibition of TH1 mediated cytokines and stimulation of TH2 mediated cytokines. 6 Additionally, oestrogen has been implicated in the pathogenesis and disease progression in a number of TH2 mediated inflammatory conditions. 7,8 This would support a relationship between exogenous oestrogens and development of UC but not CD.
It is established that exogenous oestrogen affects oral and vaginal microbiota. 9,10 However, more recent research has implicated the oestrogen-gut microbiome axis as playing a crucial role in the pathogenesis of several oestrogen-mediated diseases. 11 If a complex relationship exists between oestrogen levels and the gut microbiome then one could hypothesise that changing circulating levels of oestrogen may, in turn, disrupt gut flora and precipitate gastrointestinal disease.
Oestrogen has been shown to modulate intestinal wall barrier function 12,13 and individuals who have an episode of bacterial gastroenteritis have been shown to be fourfold more likely to develop IBD in the following year. 14 Therefore, if the barrier function of the intestinal wall is compromised by exogenous oestrogen then this may potentially increase the risk of triggering IBD in a genetically susceptible individual. Additionally, some enteric infections can be sexually acquired and one could hypothesise that women taking contraception may be at greater risk of exposure. 15 Some have theorised that IBD development may be related to micro-ischaemia within the vasculature of the gut 16 and it is established that COCPs are associated with thromboembolic disease. 17 How hormone formulation, dose and duration of contraceptive pill exposure relate to IBD risk is poorly characterised. Additionally, there is a paucity of literature on how progestogen-only and parenteral preparations of contraception affect IBD risk.
We hypothesise that oestrogen-containing contraceptives are associated with an increased risk of IBD and progestogen-only methods are not. We aimed to examine the association between various types of contraception and development of IBD. We were particularly interested in the impact of hormone formulation, dose and duration of therapy on subsequent IBD.

| Study population
A cohort of women aged 15-49 years who were registered with study practices contributing to IMRD for the period 1 January 2000-31 December 2018 was identified. Women were required to be registered with the practice for at least 9 months prior to cohort entry to avoid misclassifying prevalent IBD as incident disease. 27 GP practices were required to meet acceptable standards of electronic data quality prior to cohort entry. 25,26 Women were censored from the cohort at the first recording of a condition which would usually preclude future contraceptive use (bilateral salpingo-oophorectomy, hysterectomy, sterilisation) or the first prescription of hormone replacement therapy (supplementary code lists).
Within the cohort, we designed two nested case-control studies, one for CD and one for UC. Cases were those diagnosed with incident CD or UC during study follow-up. Case definition was taken from our previously published incidence study of IBD; In order to qualify as a case, an individual had to have either (a) two codes for IBD at different time points, (b) one code for IBD plus one prescription for a drug commonly used to treat IBD 23 (supplementary code lists). Eligible cases were required to have at least one year of prescribing history prior to the date of diagnosis. One year was selected because prescriptions for contraceptive pills are typically not longer than one year in length.
Each case was matched with up to six controls by year of birth and GP practice using incidence density sampling. Each control was allocated an index date which was the date of diagnosis for their matched case. Each control was required to have the same (or greater) prescribing history prior to the index date as their matched case. Any additional prescribing history that a control may have had did not contribute towards the analysis (ie all controls contributed the same amount of prescribing history as their matched cases over the same calendar period). The lookback period was defined as the period between the start of the prescribing history and the IBD diagnosis date (or matched index date for controls).

| Exposures
Exposure to contraceptives was based on the total lookback period. COCPs were subdivided by pill generation. Pill generation is the standard four-level classification system used for COCPs as they were rolled out chronologically, first-generation pills being the oldest and fourth generation the newest. Most pills contain ethinylestradiol and the difference between the generations of pill is the type of progestogen that is included. First-generation pills were not included as they had all been discontinued by the early 1990s.
Co-cyprindiol, a pill containing ethinylestradiol and cyproterone acetate which is used as a treatment for acne and as a contraceptive was also included.
For the primary analysis, women were categorised as either noncontraceptive users (no prescribed contraceptive use during the lookback period), second-generation COCP users, newer generation COCP users (including third-generation and fourth-generation COCPs in addition to co-cyprindiol), POP users, long-acting reversible contraception users (these are parenteral progestogen-only methods including intrauterine systems, contraceptive implants and contraceptive intramuscular injections) or mixed contraceptive users (any combination of contraceptives during the lookback period) (supplementary code lists).
Specifically for contraceptive pills, women were classed as current users if their most recent prescription would finish ≤28 days before (or after) the index date. Twenty-eight days was selected because contraceptive pills come in boxes which last 28 days.
For those pills containing mestranol, we treated 50 µg mestranol as bioequivalent to 35 µg ethinylestradiol. 28 For those pills-containing estradiol, we treated 200 µg estradiol as bioequivalent to 1 µg ethinylestradiol. 29,30 "Average months of contraceptive pill exposure per year of follow-up" was calculated and treated as both a continuous variable and separately as categorical variable in quantiles of three months per year to check for evidence of non-linearity with the development of IBD. This was done separately for COCPs and POPs. We also calculated "Average daily dose of oral oestrogen over follow-up" and similarly analysed as both a continuous variable and a categorical variable in quantiles of 5 µg ethinylestradiol per day (or equivalent).

| Covariates and confounding factors
History of endometriosis, acne and polycystic ovarian syndrome were included as covariates because they are all commonly treated with COCPs and are also potentially linked to the development of IBD (supplementary code lists); increased risk of IBD has been shown in women with endometriosis in a nationwide Danish cohort study, 31 severe acne can be a feature of IBD 32,33 and polycystic ovarian syndrome has been shown to be associated with reduced biodiversity in the gut microbiome. 34 We adjusted for smoking status treating smoking as a categorical variable with the levels "never smoker," "ex-smoker" and "current smoker" (supplementary code lists). Smoking was included as it is an established risk factor for CD and may decrease the risk of developing UC. 35 Additionally, smoking is a relative contraindication to the prescription of COCPs. 36,37 We adjusted for body mass index (BMI) as a categorical variable using the levels "underweight" (BMI <18), "normal weight" (BMI 18-25), "overweight" (BMI 25-30) and "obese" (over 30) for the primary analysis and as a continuous variable in a sensitivity analysis. BMI was included as CD often presents with weight loss and BMI is an important factor to consider when choosing appropriate contraception. 37 Social deprivation as measured by Townsend score 38 was included as we found there to be an association between Townsend score and risk of UC in a previous study. 23 Additionally, contraceptive uptake is lower in more deprived socio-economic groups. 39 Evidence of pregnancy during follow-up was included as a yes/no binary variable; pregnancy would usually preclude the use of contraception and women may be less likely to conceive if they are unwell and developing a chronic inflammatory illness (supplementary code lists).
Data on BMI and smoking were captured using the earliest value recorded during the lookback period. If data were missing during this period then the latest value recorded prior to the start of the lookback period was substituted.

| Statistical analysis
Crude incidence estimates per 100 000 person-years at risk were calculated for the source cohort. Ninety-five per cent confidence intervals (95% CI) were then calculated assuming a Poisson distribution.
Conditional logistic regression was used to analyse the nestedcase control studies and obtain odds ratios (OR) for each exposure with 95% CI. The Wald test was used to test for the significance of exposures and categorical variables in the regression model and to test for multiplicative interactions. We were particularly interested in an interaction between contraceptive pill exposure and smoking as it was reported in a large cohort study that the increased risk of UC with contraceptive pills was exclusive to smokers. 40

| Patient and public involvement
We involved representatives from the University College Hospitals NHS Foundation Trust IBD patient panel in refining the research question and designing the study protocol.

| Crohn's disease
Use of COCPs was associated with an increased risk of CD  Figure 1, Table S1).
The risk of CD went up with increasing duration of exposure to COCPs (Figure 2). When treating "average months of COCP exposure per year" as a continuous linear variable, each additional month per year of COCP exposure, increased risk of CD by 6.4% (95% CI 5.1-7.7) compared to non-users. When treating average daily dose of oral oestrogen over follow-up as a continuous linear variable, CD risk increased by 3.1% (95% CI 2.5-3.7) per µg/day of ethinylestradiol (or equivalent) compared to non-users. Longer durations of exposure to POPs had no effect on CD risk (OR 0.99 [95% CI 0.97-1.02]).
We found no evidence of an interaction between smoking and contraceptive pill exposure on risk of CD (Tables S2-S4). We found no evidence of temporal changes in the relationship between OCP exposure and CD (Table S5).

| Ulcerative colitis
We found use of all types of contraceptive pills to be associated an increase in risk of UC; OR 1.27 (95% CI 1.12-1.44) for second-  (Table S2).
However, we found that the development of UC was slightly more associated with non-smokers taking COCPs than smokers taking COCPs (P = 0.03) (Table S3). We found no evidence of temporal changes in the relationship between OCP exposure and UC (Table S5).

| Sensitivity analysis
When treating BMI as a continuous variable and excluding those with missing BMI, results were similar to the primary analysis across all methods of contraception for both CD and UC. However, confidence intervals were wider and crossed the null value for CD and newer generation COCPs (Table S6).

| D ISCUSS I ON
This is the first study to describe IBD diagnosis in relation to a range of different contraceptives including progestogen-only methods.
We observed an increase in the risk of CD with increasing durations of exposure to COCPs but not to POPs. We observed a more modest increase in the risk of UC with exposure to COCPs and POPs. There  although IMRD includes detailed prescribing data, we were unable to capture information on patient adherence. It has been reported that up to 52% of women miss their contraceptive pill once or more per month with 14% missing twice or more per month. 42 These factors could potentially result in a shift in the ORs towards unity and an underestimate in the effect of contraceptives on IBD risk. Secondly, BMI and smoking data was unavailable for a slightly larger proportion of controls than cases (Table 1). Thirdly, although our sample size was large, we lacked statistical power to analyse newer classes of COCPs separately; third generation, fourth generation and co-cyprindiol were grouped together. Fourthly, we were not able to confirm our cases with radiological, endoscopic or histological findings. Therefore, it is possible that a small number were misclassified. Finally, although a validation paper has shown that median time between IBD diagnosis and the electronic recording in the primary care records is only eight days, 22 there are bound to be delays in IBD diagnosis for a number of other reasons such as misdiagnosis or extended wait times for colonoscopy. This could introduce bias if F I G U R E 1 Adjusted odds ratios for Crohn's disease and ulcerative colitis exposed to contraceptives compared with non-use. Odds ratios with 95% confidence intervals are adjusted for Townsend score, body mass index, smoking status and history of polycystic ovarian syndrome, endometriosis, acne and pregnancy. *Abbreviation: COCP, combined oral contraceptive pill 0.5 1 1. 5 2 Odds raƟos for exposure to contracepƟves compared to non-use we have included contraceptive exposure after a woman has already developed IBD; being diagnosed with a chronic illness may influence contraceptive uptake.
In keeping with published literature, we found an association between contraceptive pill use and risk of IBD. 3  We observed a more than doubling in risk of CD in those taking COCPs continuously throughout follow-up. Contrary to a large US cohort study, we found that the development of UC was slightly more associated with non-smokers taking COCPs. 40 However, we did not observe this effect for POPs or contraceptive pills overall (Tables S3-S5). As the effect was small, this may represent a chance finding.
No previous studies have looked at IBD risk specifically in relation to progestogen-only contraceptive methods and our finding that increased CD risk was isolated to oestrogen-containing contraception is novel. Of note, a study exploring associations between contraceptive pills and disease outcomes in CD found that there was an increased risk of surgery in those taking COCPs but not progestogen-only methods. 44  shed some light on the potential biological mechanisms involved in the pathogenesis of these two diseases, highlighting the importance of future studies focusing on specific exogenous sex hormones.

ACK N OWLED G EM ENTS
This study was funded by The Charles Wolfson Charitable Trust

Guarantor of the article: Thomas Joshua Pasvol
Author contributions: The lead author (Thomas Joshua Pasvol) confirms the independence of researchers from funders and that all authors, external and internal, had full access to all of the data (including statistical reports and tables) in the study and can take responsibility for the integrity of the data and the accuracy of the data analysis. The lead author affirms that this manuscript is an honest, accurate and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as originally planned have been explained.
TJP, SB, AWS, GR and LH designed the study. TJP extracted the data.
TJP and LH performed the statistical analysis. TJP, SB, AWS, GR and LH interpreted the results. TJP wrote the manuscript. SB, AWS, GR and LH revised the manuscript for important intellectual content.
TJP confirms that all authors approved the final version of the article, including the authorship list.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data may be obtained from a third party and are not publicly avail-