Hormonal contraception increases the risk of psychotropic drug use in adolescent girls but not in adults: A pharmacoepidemiological study on 800 000 Swedish women.

In clinical practice and among the general public, psychological side effects are a well-known factor for dissatisfaction with hormonal contraception (HC) [ 1 ]. Discontinuation rates of HC are high, especially among adolescents, and mood complaints are one of the most frequently stated reasons for discontinuation [ 2 , 3 ]. Depression and anxiety disorders are twice as prevalent in women as in men; a difference that is yet to be fully understood [ 4 – 6 ] and rates of psychiatric diagnosis and visits to mental health professionals are increasing in young Swedish women [ 7 ]. The gender disparity in psychological health does not exist in children but starts at the onset of puberty, rendering female sex hormones one plausible explanation [ 8 ]. Several mechanisms have been suggested to explain the mood-altering effect of these hormones. Estrogen is a regulator of serotonin levels, a known factor in depression [ 9 , 10 ]. Progesterone most likely influences mood through the action of its neuroactive metabolites, to which not all women respond in the same way despite similar levels [ 11 ]. The scientific literature on psychological side effects of HC is relatively scant and inconclusive. Few randomized controlled trials have been performed, but most existing studies show a protective association between the most prevalent HC, combined oral contraception (COC), and anxiety, mood disturbances and depressive symptoms [ 12 , 13 ]. The field is, however, disputed, and recent studies have also found adverse mental health effects [ 14 , 15 ]. A new Danish register-based cohort study found an association between HC and subsequent use of antidepressants and a hospital diagnosis of depression, which was higher in adolescents [ 16 ]. This could be due to a healthy survivor effect , or more correctly, a selective discontinuation bias , where only women content with their treatment continue, creating a false protective association in older women. Against the above background we aimed to investigate a possible adverse effect of HC on psychological health, using a nationwide cohort of 815 662 women during 2010–2011. Our main hypothesis was that this association would be stronger in adolescents since, being new users of HC, this population should be less affected by a selective discontinuation bias .

The study population consisted of 815 662 women aged 12–30 years (mean age 20.44 years; standard deviation (SD) 5.3) on 31 December 2010, residing in Sweden since at least four years and having no previous psychiatric morbidity. Overall, 49.5% (n = 411 559) of the women were users of HC. Psychotropic drugs were dispensed to 3.1% (n = 24 973) of the women during the follow-up. Among HC users the incidence of psychotropic drug use was 3.7%, while this figure was 2.5% for non-users. For detailed information on prevalence and types of HC included, see S1 Table . The age-stratified baseline characteristics of the population are presented in Table 1 . The mean age for non-users of HC was 19.3 years (SD 5.8), and users were somewhat older, with a mean age of 21.6 years (SD 4.4). We did not find any major differences between the groups at baseline. Health care utilization was higher among HC users (21.7% vs. 17.9% had been hospitalized), and so was the frequency of menstrual disorders (4.1% vs. 2.8%). Use of hormonal contraception, socioeconomic characteristics, contact with health care, and previous diagnoses at baseline (2010–2011) by age groups and use of hormonal contraceptives in our cohort of 815 662 Swedish women. Values are percentages, unless otherwise indicated. Patterns of HC use varied with age. In the youngest age stratum, 7.77% were HC users, while as many as 48.8% of the girls aged 15–17 years and 68.2% of the women aged 18–20 years used HC during 2010–2011. In users of HC the absolute risk of using psychotropics was around 4% for all age groups ( Fig 1 ). In non-users of HC, however, this risk was very low in young adolescents and increased with age, to reach the level of HC users at the age of 21 or older. Percentage (i.e. absolute risk) of first-time use of psychotropic drugs during a one-year follow-up from baseline (2010–2011) by age, in users (gray line with circles) and non-users (black line with squares), of hormonal contraceptives in 815 662 Swedish women. Overall, compared to non-users, users of HCs had an adjusted OR of a first-time use of psychotropic drugs of 1.34 (95% CI: 1.30–1.37). Adjustments were made for age, family income, educational level, health care utilisation, and having a diagnosis of thromboembolism, epilepsy or migraine, or menstrual disturbances. However, the age-stratified analysis ( Fig 2 ) revealed a strong association in adolescent girls, which decreased with age to disappear after adolescence. We observed this pattern for all groups of HC. The highest ORs were found in the youngest age group of 12- to 14-year-olds, with an adjusted OR of 3.46 (95% CI: 3.04 to 3.94). Age-stratified odds ratios with 95% confidence intervals (black lines) for the association between use of hormonal contraceptives and use of psychotropic drugs within a one-year follow-up after baseline (2010–2011) in 815 662 Swedish women. Crude (black circles) and adjusted (gray squares) values. Adjustments were made for age, family income, highest educational level in family, previous hospitalizations, outpatient hospital visits, and having a diagnosis of thromboembolism, epilepsy or migraine, or menstrual disturbances including endometriosis. In the analysis differentiating between oral and non-oral forms of combined and progesterone-only methods, we observed a pattern where use of HC in young adolescents displayed strong associations with psychotropic drug use ( Table 2 ). The strongest association was found among 12–14 year olds using a non-oral progesterone-only method such as a skin patch or intravaginal ring, with an of OR 4.47 (95% CI: 2.08–8.78). The non-oral methods generally had higher ORs than the oral methods, as did the progesterone-only methods compared to combined methods regardless of administrative route. For example, users of the vaginal ring/patch in the age group 15–17 years had an adjusted OR of 2.27 (95% CI: 1.85–2.79) for a first use of psychotropic drugs compared to non-users, while COC users had an OR of 1.52 (95% CI: 1.41–1.64) in the same age group. Adjusted age-stratified odds ratios (OR) with 95% confidence intervals (CI) for the association between oral and non-oral hormonal contraception (HC) methods and use of psychotropic drugs within one-year follow-up after baseline (2010–2011) in 815 662 Swedish women. Non-users used as reference within each age group. The analyses distinguish between combined and progesterone-only methods within the oral and non-oral forms of contraceptives. Adjustments were made for age, family income, highest educational level in family, previous hospitalizations, outpatient hospital visits, and having a diagnosis of thromboembolism, epilepsy or migraine, or menstrual disturbances including endometriosis. HC; Hormonal contraceptive, COC; combined oral contraceptives, POP; progesterone-only pills, Patch; skin patch (Evra), Ring; intravaginal ring (NuvaRing), IUD; Intrauterine device, Inj.; injection (Depo-Provera), Impl.; implant (Implanon, Jadelle)* a Non-oral combined methods. b Non-oral progesterone-only methods The analysis of separate HC types showed distinct differences between adolescents (12–19 years) and adult women (20–30 years) ( S2 Table ). While the majority of HC substances where positively associated with a first-time use of psychotropic drugs in adolescents, we observed few conclusive associations in adult users. Non-oral HC forms did however display stronger associations in both adults and adolescent users. Use of the depot medroxyprogesterone acetate (DMPA) injection (ATC G03AC06), a high dose progesterone-only method, had the strongest association with subsequent use of psychotropic drugs in adult women, with an adjusted OR of 1.56 (95% CI: 1.34–1.82). For adolescents, the levonorgestrel containing IUD had the strongest association with use of psychotropic drugs, with an adjusted OR of 2.90 (95% CI: 2.22–3.79), followed by DMPA (OR: 2.37; 95% CI: 1.46–3.84). We also performed a sensitivity analysis excluding women diagnosed with a number of illnesses (i.e., thromboembolism, endocrine disorders, migraine or epilepsy, disorders of the female genital tract, and malignant diseases) that could influence both use of HC and of psychotropic drugs, and the results were very similar ( S3 Table ). Finally, calculations of the AUC showed that the ability to categorize users of psychotropic drugs based on information on HC use was very low. In women over 18 years, the AUCs ranged between 0.54 and 0.60 for unadjusted and adjusted values, respectively ( S2 Table ). For adolescents the AUC was somewhat higher with values between 0.62 and 0.68 in the youngest age stratum. We also ran a model with all the explanatory variables except HC, and then added use of HC in a consecutive model, to be able to assess how much knowledge of HC use added to the AUC ( S4 Table ). This analysis showed that HC added nothing to the AUC in women over 18 years of age. In adolescent girls, knowledge of HC added between 0.02 and 0.04 units to the AUC of the model containing only individual predictors.

We find an association between hormonal contraception and subsequent use of psychotropic drugs for women of reproductive age. The association is large for young adolescents, and insignificant in adult women, pinpointing the need to bear population heterogeneity in mind and identifying adolescent girls using hormonal contraception as a vulnerable population. Our also study adds to a growing body of evidence that hormonal contraception could adversely affect psychological health in certain girls and women, and warrants further investigation of the influence of different hormonal contraception on psychological health, particularly in young women.

We designed a nationwide database by record linkage of several Swedish registers, using the official unique personal identification number assigned to every individual residing in Sweden [ 17 ]. The database was constructed by Statistics Sweden and the Swedish National Board of Health and Welfare, and it was delivered to us after replacing the personal identification number by an arbitrary one to ensure anonymity of subjects. Based on the Total Population Register, we defined an initial cohort containing all 1 094 069 women aged 12–30 years residing in Sweden on 31 December 2010 [ 18 ]. We obtained information on medication use from the Swedish Prescribed Drug Register (SPDR), which contains individual level data on all dispensed drug prescriptions at Swedish pharmacies (excluding dispensation at hospitals and nursing homes) for the whole population in Sweden since 2006 [ 19 ]. We assigned a baseline date to every woman, defined by the first dispensed prescription of an HC drug (Anatomical Therapeutic Chemical (ATC) codes G02BA, G02BB, G03AA, G03AB, G03AC) between 1 January 2010 and 31 December 2011. The baseline date was set to 31 December 2010 if the woman did not have any HC prescription during these two years. Each woman, user and non-user, thereby got an individual baseline date based on her first prescription fill or the assigned date of 31 December 2010. From the individual baseline date, a one-year follow-up and a previous period of four years were recorded, meaning all women were followed for exactly one year. With this definition, we can furthermore be sure that the non-users of HC were true non-users during four years before, and one year after their baseline date of 31 December 2010. After calculating the individual baseline date, we excluded 99 052 women who were not residing in Sweden for four years prior to, and one year after their own baseline date, since information on diagnosis and medication were lacking in those cases. Women with pre-existing psychiatric disorders were excluded by using information on diagnoses from hospital discharges and outpatient visits recorded at the Swedish Patient Register. Through the International Classification of Diseases, 10 th version (ICD-10), women with a psychiatric diagnosis (ICD-10 codes F00–F99) in the previous four years were excluded. Since diagnosis from primary care visits are not available in Swedish registers, we also used information from the SPDR, which includes prescriptions from primary care as well as specialised care to identify women with less severe mental health issues. Women having a prescription fill of a psychotropic drugs (ATC N05 or N06) in the previous four years were therefore also excluded (n = 156 302). The period of four years was determined because of restrictions in the available information on medications. Finally, we excluded 23 053 women who had a diagnosis of child delivery (ICD-10 O80-84) at a hospital during follow-up. The final cohort consisted of 815 662 women. We defined non-users of HC as those who did not fill a prescription for HC between 1 January 2010 and 31 December 2011, and took this category as reference in all comparisons with users of HC . Users of HC were identified according to the first HC prescription fill during baseline. Since HC can only be acquired via a prescription from a physician or midwife in Sweden all users were included. The exception is emergency contraceptives that are mainly dispensed over the counter and are therefore not registered in the SPDR. A small number, 288 prescription fills of emergency contraceptives, were recorded in the SPDR, but were excluded from the analysis. HC prescriptions in Sweden are filled every three months or annually. The outcome of our study was use of psychotropic medication defined as filling at least one prescription with anxiolytics, hypnotics and sedatives or anti-depressants (ATC N05B, N05C or N06A) during the one-year follow-up, from the first dispensation of HC or 31 December 2010, if the individual did not use HC at baseline. Use of psychotropic medication can be considered as a proxy for impaired psychological health [ 20 ]. We obtained individual and family level data on socioeconomic factors as of 31 December 2010 from Statistics Sweden’s Longitudinal integration database for health insurance and labour market studies and operationalized SEP by means of education and income [ 21 ]. Education was categorized according to the number of years into low (less than 14 years) and high (14 years or more) level of educational achievement, with a category for missing values. We used the highest educational level achieved by any member of the woman’s current household, since many individuals in our cohort were children. Individualized disposable family income was calculated by dividing the total disposable income of the family by the number of family members, taking into account the different consumption weights of adults and children determined by Statistics Sweden. Thereafter, we created three categories (i.e., low, medium, and high) of income using the total Swedish population aged 18–80 years. We considered the high income category as the reference in the comparisons. We categorized age into the following groups; 12–14, 15–17, 18–20, 21–25 and 26–30 years, as well as into adolescents (12–19 years) and adult women (20–30 years) and used these age categorizations to perform different stratified analyses. We created dichotomous variables to adjust for the presence or absence of any previous diagnosis related to thromboembolism (ICD-10 I26, I80 and I82), epilepsy or migraine (ICD-10 G40-47) and menstrual disturbances including abnormal bleeding, premenstrual disorders and endometriosis (ICD-10 N80 and N91-94) in the previous four years. Information on diagnosis were not available from primary health care clinics. Contraindication of certain HCs for women with these diagnoses, and presence of menstrual disturbances, could affect both HC prescription and mood and therefore confound the results. Thereafter, we adopted a more exhaustive approach and adjusted for the presence or absence of any hospitalization or contact with an outpatient hospital clinic in the past four years. In the case that a potential association between HC use and use of psychotropic medication were mediated by an increased contact with health care services, our results would underestimate this association. Since the use of psychotropic drugs rises with age, we also adjusted for age as a continuous variable within each age stratum. Since the follow-up was short and complete, we applied logistic regression to estimate odds ratios (OR) and 95% confidence interval (CI) of psychotropic drug use in relation to the different HC categorizations, with non-users of HC as the reference group. We performed simple and multiple age-stratified logistic regression models. In the multiple regression models, 2545 cases were dropped due to missing information on family education level. We obtained the predicted probabilities from the models to calculate the area under the operating receiver operator characteristics curve (AUC) [ 22 ]. The AUC measures the ability of the model to correctly classify those using or not using psychotropic medication, based on knowledge of HC use. The AUC assumes a value between 0.5 and 1, where 1 represents perfect discrimination and 0.5 corresponds with a predictive value equal to that obtained by flipping an unbiased coin. We performed all analyses using SPSS 22.0 (Statistical Package for Social Sciences, SPSS Inc., Chicago, IL, USA) software. The database was approved by the Regional Ethical Review Board in Lund, Sweden, the Data Safety Board at Statistics Sweden and the National Board of Health and Welfare (Dnr: 2014/856, 2015/341).

Prevalence of different hormonal contraceptive methods in 2010–2011 by contraceptive type in our cohort of 815 662 Swedish women aged 12–39. All combined HC contain estrogen in addition to progesterone, except for G03AA14 (Zoely), which contains estradiol. (DOCX) Click here for additional data file. Odds ratios (OR) with 95% confidence intervals (CI) and area under the curve (AUC), stratified on adolescents and adults, for the association between use of different hormonal contraceptives and a first time use of psychotropic drugs within a one-year follow-up after baseline (2010–2011) in 815 662 Swedish women. (DOCX) Click here for additional data file. Sensitivity analysis excluding women with certain diagnosis . Odds ratios (OR) with 95% confidence intervals (CI) for the use of hormonal contraceptives (HC) (ATC G02BA, G02BB, G03AA, G03AB, G03AC) and use of psychotropic drugs (ATC N05B, N05C, N06A) within a one-year follow-up after baseline (2010–2011) in 815 662 Swedish women. A sensitivity analysis excluding women with different illnesses (delivery: ICD-10 O80-84, pregnancy and delivery complications: ICD-10 O00-O99, menstrual disturbances; ICD-10: N80 and/or N91- 94, endocrine disorders, epilepsy and migraine, malignant disorders, thrombosis or disorders of the female genital tract: E00E3, G40G43, C00C97, I80, I82, I26 and/or N70N98). Both crude and adjusted values shown. Adjustments were made for age, family income, educational level and health care utilization. (XLSX) Click here for additional data file. Change in area under the receiver operating curve (AUC) with addition of hormonal contraceptive use to a model containing individual predictors . Age stratified change in the area under the curve for use of psychotropic drugs when adding information on use of hormonal contraceptives (ATC G02BA, G02BB, G03AA, G03AB, G03AC) to a model already including age, family income, highest educational level in family, previous hospitalizations, outpatient hospital visits and having a diagnosis of thromboembolism, epilepsy or migraine, or menstrual disturbances including endometriosis in the 815 662 Swedish women. (DOCX) Click here for additional data file.