Oral Contraceptives and the Risk of Psychiatric Side Effects: A Review

Abstract

Background

Oral contraceptives (OCs) are an essential medicine used by millions of people every day. Given the widespread usage of these medicines, even a small increase in psychiatric risk could be of clinical significance. Although mood-related side effects are a common reason for OC hesitancy and discontinuation, studies investigating psychiatric responses to OC treatment have had inconsistent results.

Summary

While OCs are beneficial for most users, there is evidence that a subgroup of users is susceptible to mood side effects. Randomized controlled trials have generally failed to find differences in mood symptoms between OC and placebo users, but observational studies comparing OC users to non-users have reported increases in symptoms of depression, anxiety, and eating disorders. Additionally, observational evidence suggests that OC users may be more likely to use prescription psychotropic medications and to attempt or die by suicide. However, responses to OC treatment are highly heterogeneous, and some users report mood improvement. A variety of factors may increase the likelihood of negative psychiatric side effects, including younger age, previous experience of side effects from OCs, and preexisting psychiatric disorders. Progestin-only pills may confer a higher psychiatric risk than combination pills.

Key Messages

Further research investigating factors that contribute to susceptibility to the mood-related side effects of OCs is clearly warranted. Genomic approaches may provide insight as to why some users experience side effects while others do not. Research elucidating who is most at risk and why will be essential to addressing prevalent concerns about the psychiatric risk of OCs.

Introduction

A total of 150 million people worldwide use oral contraceptive (OC) pills on a daily basis [1]. In addition to the personal and social benefits of OCs, providing users with straightforward and reliable options to prevent unintended pregnancies, OCs are also prescribed for several non-contraceptive purposes, including to treat symptoms of endometriosis [2], polycystic ovarian syndrome [3], and premenstrual dysphoric disorder (PMDD) [4]. However, fear of mood-related side effects is a common reason for OC hesitancy, and experience of these side effects contributes to the high rate of discontinuation among new users [5, 6]. This review evaluates the present state of knowledge about some of the potential psychiatric side effects of OC use and their risk factors.

Psychological and behavioral responses to OC treatment are highly heterogeneous. While there is little consistent evidence of mood-related side effects resulting from OC use in the general population, a generally consistent finding across studies is that a subgroup of susceptible individuals reports adverse mood side effects with OC use [7, 8]. Understanding the risk factors which predispose this group to side effects will be key to better clinical predictions and management of adverse outcomes. In particular, we note that few studies have used genetic approaches to investigate this question, and this could be a promising avenue for future research.

In this review, we focus on three groups of psychiatric disorders with elevated incidence in women compared to men, which have typical ages of onset during life stages with changing reproductive hormones [9] (i.e., puberty): depression, anxiety, and eating disorders (EDs) [10, 11]. Depression and anxiety disorders are also known to worsen during hormonal transitions and fluctuate in severity over the course of the menstrual cycle, suggesting that endogenous ovarian hormones may play a role in the development of depression and anxiety symptoms in women who menstruate [9]. We note that OC side effects are not unique to these disorders; rather, OC side effects include a range of psychiatric disorders. In addition, we examine increased rates of psychotropic medications among OC users and increased rates of suicide ideation, attempt, and death.

While many of the studies we include discuss OC use in relation to women, the affected population can be more accurately described as people who menstruate, including most (but not all) cisgender women and girls of reproductive age, as well as many transgender men. Criteria for study selection are outlined in online supplementary Table 1 (for all online suppl. material, see https://doi.org/10.1159/000539515). Inclusion and exclusion criteria for each study have been summarized in online supplementary Table 2, although not every study clarifies these criteria.

Methods

A digital search of the literature was performed on PubMed using the search query (“Contraceptives, Oral/adverse effects” [Mesh]) AND {(“Behavioral Symptoms” [Mesh]) OR “Mental Disorders” (Mesh)}, which returned 570 results. After filtering for articles in English with available full text published in the year 2000 or later, 146 results remained. Additional studies were identified by screening reference lists in each of these studies for relevant articles. Inclusion and exclusion criteria for study selection are outlined in online supplementary Table 1. Although performing a meta-analysis was considered, the clinical and statistical heterogeneity of study measures rendered this inappropriate.

Evidence for Adverse Outcomes

Depression and Anxiety in OC Users

Depression is the most common type of mood disorder, and it is the only psychiatric disorder listed by name in most OC patient leaflets as a possible side effect [5]. While several randomized controlled trials (RCTs) have found no significant differences in levels or types of depressive symptoms between OC and control groups [12, 13], results have at times been conflicting. For example, one placebo-controlled RCT found that combined oral contraceptive (COC) use was associated with a significant improvement during the premenstrual phase in depression (mean difference = −0.33, 95% CI: −0.62 to −0.05, p = 0.049) and a worsening during the intermenstrual phase in anxiety (0.22, 95% CI = 0.07–0.37, p = 0.003), irritability (0.23, 95% CI: 0.07–0.38, p = 0.012), and mood swings (0.15, 95% CI: 0.00–0.31, p = 0.047) [14]. Another placebo-controlled RCT found no change in scores on the Beck Depression Inventory or the Psychological General Well-Being Index (PGWBI) dimensions of depressed mood, anxiety, and general health, but reductions were observed in the dimensions of positive well-being (−3.90, 95% CI: −7.78 to −0.01; p = 0.0492), self-control (−6.63, 95% CI: −11.20 to −2.06; p = 0.0046), and vitality (−6.84, 95% CI: −10.80 to −2.88; p = 0.0008) [15].

It is possible that null results stem from response heterogeneity, rather than a complete lack of response; certain OC users did become depressed, in some cases severely (e.g., Lundin et al. [14]). Since individuals with side effects are usually outnumbered by users who are unaffected and occasionally even those who experience mood improvement, their scores typically fail to shift the mean, leading researchers to conclude that the depressive side effects of OCs are minimal or non-existent [8]. However, for the individuals who are affected by negative side effects, the impact is not insignificant.

Observational investigations into the association between OCs and depression have produced mixed findings. While a Swedish national register-based study reported COC users had a lower relative risk (RR) for first-time depression diagnosis than non-users (RR = 0.86, 95% CI = 0.82 to 0.89) [16], a similar Denmark-based study found that COC users were more at risk (RR = 1.1, 95% CI = 1.08–1.14) [17]. Both studies showed that progestin-only pill (POP) users had a higher depression risk than non-users (RR = 1.11, 95% CI = 1.07 to 1.14; RR = 1.3, 95% CI: 1.27 to 1.40, respectively), and both reported higher estimates in adolescents (aged 15–19) [16, 17]. The discrepant results for COCs are likely due to residual confounding. Several other prospective cohort studies have reported an elevated risk for depression in OC users, especially among adolescents [18–20].

Few epidemiological studies have investigated anxiety as an outcome of OC use, but in one cross-sectional study of sexually active American women aged 20–39, a significant inverse association was reported between OC use and diagnosed or subthreshold panic disorder in the past year as well as a non-significant trend toward reduced odds of diagnostic or subthreshold generalized anxiety disorder [21]. Since the evidence regarding OC use and anxiety is relatively scarce and the existing literature is conflicting, more research into this relationship is clearly warranted.

Several observational studies have provided evidence that OC users may be more likely to redeem prescriptions for psychotropic drugs. OC users in the Danish national registers were at a higher risk for first use of antidepressants than non-users (COC RR = 1.2, 95% CI = 1.22–1.25; POP RR = 1.3, 95% CI = 1.27–1.40), and age-stratified analyses showed that RRs were notably higher for adolescents (aged 15–19) (COC RR = 1.8, 95% CI = 1.75–1.84; POP RR = 2.2, 95% CI = 1.99–2.52) [17]. Another study using the Swedish national registers reported that hormonal contraception (HC) users had an elevated risk of psychotropic drug use (including prescriptions for anxiolytics, hypnotics and sedatives, or antidepressants) (adjusted odds ratio (OR) = 1.34, 95% CI: −1.30–1.37), and this effect was driven by the youngest age group (aged 12–14: adjusted OR = 3.46, 95% CI = 3.04–4.94) [22]. In a separate study of Swedish national registers, progestin-only users had higher odds of antidepressant use than non-users in all age groups studied (aged 16 to 31), while this was only true for the youngest combination HC users (aged 16 to 19) [23]. While it is true that all users of a medication are more likely to be prescribed psychoactive drugs due to presumably increased healthcare access, one retrospective cohort study addressed this confounder by drawing a sample from the US Military Healthcare System and employing two control groups, one with all HC non-users enrolled in the system and one with only enrollees that accessed medical services during the study period [24]. Compared to the enrolled control group, women using HC containing norgestimate, levonorgestrel, etonogestrel, norethindrone, or norelgestromin had a higher risk of depression diagnosis, but compared to the control group that accessed care, risk was elevated only for levonorgestrel or norelgestromin, while risk was reduced for norethindrone. While the persisting results align with previous studies, the attenuation of effects sheds light on the importance of controlling not only for healthcare access, but also for utilization.

Dysregulation of Eating Behaviors and Self-Image in OC Users

To date, there are no population-based studies comparing the risk of any EDs in OC users compared to non-users. In one prospective cohort study, adolescent OC users were more likely to report “eating problems” than non-users, which included eating too much or too little [20].

Even though weight gain is one of the most reported side effects of OC use, decades of research have failed to produce evidence of any association [25]. However, several studies have suggested that users’ perception of their weight may be affected by OC use. According to a US-based cross-sectional study, a person who has ever used OCs has 11% higher odds of perceiving themselves as overweight compared to someone with the same measured BMI who has never used OCs [26]. Additionally, one RCT comparing a COC (20 μg ethinylestradiol and 100 μg levonorgestrel) to placebo found that while participants did not experience any significant weight changes, weight gain was reported as a side effect by participants in both groups, even in some cases where their weight decreased [13].

Suicidal Behaviors in OC Users

Several epidemiological studies have pointed to a significantly higher risk of suicidal behaviors among OC users than non-users. A recent meta-analysis of three British and North American cohorts (n = 184,721) found an elevated risk of suicide death in OC users (overall risk ratio = 1.36, 95% CI = 1.06–1.75) compared to non-users [27]. A recent study of Swedish national register data found that OC users were at an elevated risk for suicidal behavior (attempted suicide or suicide death) after 1 month of use (COC hazard ratio (HR) = 1.73, 95% CI = 1.44–2.08; POP HR = 2.78, 95% CI = 2.14–3.62), but after 1 year of use, only POP users remained at an elevated risk (HR = 1.82, 95% CI = 1.44–2.31) [28]. Likewise, in the Danish national registers, risk for first suicide attempt was elevated in OC users compared to never users (COC RR = 1.91, 95% CI = 1.79 to 2.03; POP RR = 2.29, 95% CI = 1.77 to 2.95) [29]. These findings were further corroborated by a study reporting a small but significant increase in risk for suicidal ideation or attempt in a Korean cohort (OR = 1.13, 95% CI = 1.00–1.19) [30]. One contradictory cohort study of sexually active American women aged 25–34 (n = 6,654) found a lower risk of past-year suicide attempt in HC users than women using low-efficacy methods (periodic abstinence, spermicides, contraceptive film) or no contraception (OR = 0.37, 95% CI = 0.14–0.95) [31]. However, the authors noted many differences between the contraception use groups, including that HC users were younger and tended to have lower depression scores at prior interviews compared to the other groups. Additionally, the authors did not exclude participants with past suicide attempts, as the national register studies did.

Differences in the methods and samples used in these studies render comparison difficult. Participants in the studies meta-analyzed by Pérez-López et al. [27] as well as the study by Keyes et al. [21] were in their late 20s and 30s, while participants in the Danish and Swedish national register studies were in their late teens and early 20s. Data were also drawn from different time periods when different types of OCs were available to the public. The meta-analyzed studies included earlier COC formulations with high doses of ethinylestradiol, while the others included third- and fourth-generation COCs and POPs.

Prediction of Adverse Outcomes

While the prevalence of adverse reactions to OC use remains unclear, several factors have been identified that help predict who will be affected. Most of these factors are inherent to the user, such as age, previous experience with HC, reasons for using OCs, psychiatric history, and genetics. However, the type of pill prescribed may also have an impact, as OCs vary widely in their composition and regimen.

Younger Users Are at Higher Risk for Adverse Outcomes

Observational studies have suggested that adolescents may be more at risk for developing psychiatric side effects with OC use than adults. In the prospective cohort study by de Wit et al. [20], differences in depressive scores between OC users and non-users were only apparent in the 16-year-old age group and were not significant in the cohort as a whole (aged 16 to 25). Another study looked at a smaller portion of the same cohort and found an association between adolescent OC use (16–19 years) and a subsequent episode of MDD in early adulthood (20–25 years), which was primarily driven by young women with no prior history of MDD [19]. In the Swedish national registers, POP use was associated with a decreased depression risk for adults aged 20–25, but an increased risk for adolescents aged 15–19 (RR = 1.13, 95% CI = 1.07–1.19) [16]. COC use was also associated with decreased depression risk for adults (RR = 0.80, 95% CI = 0.78–0.82), but did not significantly impact risk for adolescents (RR = 0.96, 95% CI = 0.93–0.98). In the Danish national registers, risk for first depression diagnosis and first use of antidepressants decreased with increasing age for both COC and POP users, and adolescents aged 15–19 showed notably higher risks than the sample as a whole (aged 15 to 34) [17]. Risk for first suicide attempt or suicide death was also elevated in Danish adolescents using HC (RR = 2.06, 95% CI = 1.92–2.21) as compared to adult users (aged 20–24: RR = 1.61, 95% CI = 1.39–1.85; aged 25–33: RR = 1.64, 95% CI = 1.14–2.36) [29].

Only one placebo-controlled RCT to date has investigated psychiatric side effects of OCs in adolescent users below age 19 [13]. Depressive symptoms at the end of the 3-month study period were not significantly different between the OC and placebo groups (independent sample t test, t = 0.184, p = 0.86) [13]. However, since the study only enrolled adolescents with moderate-to-severe dysmenorrhea, the results cannot be generalized to a broader adolescent population [13]. The increase in depression risk for adolescents observed in cohort studies could be due to the attrition of susceptible women, but without further investigation, the possibility that adolescents are particularly vulnerable to psychiatric effects of OCs cannot be ruled out.

Women with a History of OC-Related Mood Changes Are More at Risk

At least two placebo-controlled RCTs reported that women with a previous experience of OC-related mood symptoms are more likely to report mood symptoms when re-exposed to OCs. In the first trial, women who reported previous adverse mood symptoms while using COCs experienced significantly greater mood worsening in the intermenstrual phase of active treatment compared to women with no such history [14]. Another RCT including 34 women who had all reported a history of mood symptoms during COC use found that those randomized to COC (30 μg ethinylestradiol and 0.15 mg levonorgestrel) had higher scores of depressed mood (T = −2.3, p < 0.05), mood swings (T = −3.4, p < 0.01), and fatigue (T = −3.1, p < 0.01) as assessed by the Cyclicity Diagnoser (CD) scale than those in the placebo group [32].

Preexisting Psychiatric Conditions Affect Outcomes following OC Use

Research has shown that preexisting conditions and the reasons for using OCs may have an impact on psychiatric outcomes. People with conditions that are treated with OCs such as PMDD or endometriosis may experience an improvement in their symptoms and quality of life, improving overall mental health. For example, in a large prospective cohort study of young Australian women (aged 22–30), participants who were prescribed an OC for reasons unrelated to contraception were 1.32 times more likely to report depressive symptoms than those using OCs for contraceptive purposes [33]. In a nested case-control study, early-onset premenstrual mood disturbance and dysmenorrhea were significant predictors of mood improvement with OC use [34].

PMDD, a severe form of premenstrual mood disturbance, is the only mental health condition thought to be alleviated by OC use. One particular COC (24/4 20 μg ethinyl estradiol and 3 mg drospirenone) has been approved by the FDA to treat PMDD. In an RCT, women with PMDD taking this COC experienced significantly greater improvement on all 11 items on the Daily Record of Severity of Problems than placebo users, and COC use was associated with a 49% decrease in premenstrual depression [35].

Other preexisting psychiatric conditions may also impact risk of OC side effects, although in the opposite direction. In one RCT, women with ongoing or previous mood, anxiety, and EDs who received a COC (1.5 mg estradiol and 2.5 mg nomegestrol acetate) had higher total DRSP scores during the intermenstrual phase than corresponding women who received a placebo (p = 0.008), with increases in the specific domains of anxiety (p = 0.049), mood swings (p = 0.005), and irritability (p = 0.002) [36]. Women without mental health problems randomized to COC saw an increase in anxiety scores (p = 0.028), but no change in overall DRSP scores (p = 0.5). A cross-sectional study (N = 976, aged 36–45) found that women with a history of depression before initiating OC treatment were twice as likely to experience premenstrual mood deterioration with OC use [34]. A prospective cohort study in the Swedish national registers found that COC users (aged 15–24) with attention-deficit/hyperactivity disorder had a sixfold higher risk of depression than COC users without attention-deficit/hyperactivity disorder (aHR = 6.10, 95% CI = 5.79 to 6.43) [37]. Studies indicating elevated psychiatric risk in POP users have led some researchers to advise caution for POP use in those with current or past depression, but such suggestions have not been widely clinically implemented and psychiatric disorders are currently not considered contraindications for OC use [5].

Adverse Outcomes Are Less Common with Certain Types of OCs

OCs vary widely by their formulation (i.e., type of progestin), dosage, and regimen (i.e., monophasic or multiphasic). Several epidemiological studies comparing the prevalence of side effects in COC and POP users have reported higher rates of psychiatric symptoms in POP users [17]. Evidence from both RCTs and epidemiological studies has shown that the different generational classes of progestins may have differing effects on mood. RCTs comparing COCs with different classes of progestins have suggested that in terms of mood and affect, third- and fourth-generation COCs may be preferable to second-generation COCs [38, 39]. The newer generational classes of progestins have been suggested to have a lower likelihood of inducing psychiatric side effects [40].

Potential Neurobiological Mechanisms

The mechanisms underlying OC-induced mood changes remain unclear, but biological effects of OC use have been documented that may help explain why certain women experience mood deterioration with OC use while most do not. HC use is known to suppress the natural production of estrogen and progesterone by introducing their synthetic versions to circulation [41]. Additionally, COCs reduce circulating levels of total and free testosterone, and increase sex hormone-binding globulin [42]. While few studies have examined the potential psychiatric effects of these hormonal changes in OC users, fluctuations in these hormones have been linked to affective changes in non-users. It is well established that levels of endogenous estrogen and progesterone fluctuate over the course of the menstrual cycle in naturally cycling women, and for a subgroup of susceptible women, these fluctuations are accompanied by exacerbations in psychiatric symptoms, especially in the premenstrual phase when estrogen and progesterone are low [43]. Further, periods of reproductive transition, such as puberty and menopause, are associated with an increased risk of psychiatric disorders [44, 45]. These findings indicate a role of sex hormones in the pathophysiology of psychiatric disorders.

Numerous neuroimaging studies have reported differences in functional connectivity between the brains of COC users and naturally cycling non-users. In one RCT comparing a COC (21/7, 30 μg ethinyl estradiol and 0.15 mg levonorgestrel) to placebo in women with a self-reported history of OC-induced side effects (n = 34), the COC group had higher scores for depressed mood and mood swings, and these changes were accompanied by altered reactivity in the right amygdala and left insula [32]. In a follow-up study, the COC group displayed increased within-network connectivity in the default mode network over the course of treatment [46]. Increased connectivity in the default mode network has been linked to major depressive disorder [47], and this network was recently reported to be modulated by estrogen and progesterone [48]. The COC group also showed decreased connectivity in the salience network and executive control network [46]. Mood lability was the most reported side effect in the COC group, and connections associated with increased mood lability increased in the COC group, while those associated with decreased mood lability decreased. Furthermore, several of the observed connectivity changes with the highest effect sizes, including that between the dorsal anterior cingulate cortex and the posterior cingulate cortex, were able to predict participants’ treatment groups above chance [46]. These results demonstrate that HC use clearly causes changes in brain connectivity, although more research is needed to clarify the psychiatric impact of such changes.

To date, there have been no genome-wide association studies (GWAS) investigating the genetic contribution to variation in responses to OC use, and this could be a promising avenue for future research. In the only study to investigate heritability for OC-related mood side effects, an analysis of twin pairs implicated genetic factors in OC-related depression [49]. Though studies using candidate gene approaches have been conducted on the subject, this approach is limited in scope because the analysis is restricted to a few genes of interest, and selection is limited by current knowledge. Recent pharmacogenomic studies are increasingly using unbiased approaches like GWAS, although obstacles include the need for large numbers of subjects and the choice of appropriate matched control groups [50]. While obtaining a sample large enough for an OC response GWAS would be challenging, other genetic approaches such as Mendelian randomization could prove fruitful. Mendelian randomization exploits the natural randomization of genetic material that occurs at conception in order to assess the causal effect of an environmental exposure on a clinical outcome, and has been posited as a novel approach for predicting adverse drug reactions [51]. It is possible that there are specific genetic variants that contribute to susceptibility for psychiatric side effects of OCs, perhaps in genes involving drug metabolism, transporters, or hormone receptor density.

Discussion

The evidence refuting a link between OC treatment and psychiatric outcomes relies largely on RCTs. In this case, a major limitation of RCTs is their relatively brief time course. While epidemiological studies can track participant outcomes over many years, RCTs investigating OC side effects generally only observe subjects for at most 4 months. Results from observational studies suggest that this may be too early in the treatment regimen to detect potential psychiatric side effects [17]. Additionally, there may be a more prolonged impact of OC use across the lifespan, as highlighted in the prospective cohort study by Anderl et al. [19] that reported OC use during adolescence is associated with an increased risk of developing MDD in early adulthood.

Most of the evidence in support of an association between OC use and negative psychiatric outcomes relies on observational studies, which cannot establish cause and effect. These studies are limited by the presence of many confounding factors that may not be fully accounted for, making it easy to discover spurious associations. Many of the studies reported that the risk of particular psychiatric side effects was higher in adolescents than adults, but this may be due at least in part to a survivorship bias. These results may be explained by the fact that susceptible adults have already tried and ceased OC use due to side effects. Additionally, the role of healthcare utilization is not addressed in many of the studies. These findings highlight the “healthy user effect,” a source of bias that refers to the tendency of people receiving one preventive therapy (i.e., OCs) to also take other preventive health measures [52].

A major limitation of most of the epidemiological studies is the broad categories used to describe HC type. Since it has been shown that different types and doses of estrogen or progestin may produce differing side effects, future studies should aim to stratify OC types as much as possible. Another limitation in many of these studies is that all types of women are grouped together. In a recent prospective cohort study, the greatest difference in antidepressant use between HC users and non-users was seen in teenagers (aged 12–17) and immigrant women with low income [53]. This study demonstrates that the intersectional context of the individual may be relevant to side effect susceptibility.

Several studies have suggested that experience of OC side effects may be indicative of a broader gonadal hormone sensitivity. In a cross-sectional study of current and former OC users, those who reported a previous experience of mood and/or physical side effects from OC use were at an increased risk of experiencing ED symptoms, particularly body dissatisfaction and drive for thinness [54]. Additionally, a recent prospective cohort study using Danish national registry data found that first-time mothers with prior HC-associated depression were at an elevated risk for postpartum depression compared to those with prior depression unrelated to HC use (adjusted OR = 1.35, 95% CI = 1.17–1.56) [55]. These noteworthy findings support the theory that there is a subgroup of women vulnerable to mood changes during hormonal transitions.

Conclusion

Given the high degree of heterogeneity in psychological responses to OC treatment, further research into the sources of this variation is clearly warranted. Experience of OC side effects may indicate a broader sensitivity to hormonal transitions that could be of clinical significance for other aspects of health and well-being. Future research elucidating who is most at risk of psychiatric OC side effects would advance patient-centered care by allowing providers to make more effective assessments of the benefit/risk balance for individual patients.

Conflict of Interest Statement

We declare no conflicts of interest.

Funding Sources

L.M.H. acknowledges funding from NIMH (R01MH124839, R01MH118278, R01MH125938, RM1MH132648, R01MH136149), NIEHS (R01ES033630), and the Department of Defense (TP220451).

Author Contributions

J.C. and L.M.H. conceptualized and designed the review and edited and refined the manuscript together. J.C. wrote the initial draft and had primary responsibility for literature searches.

Funding Statement

L.M.H. acknowledges funding from NIMH (R01MH124839, R01MH118278, R01MH125938, RM1MH132648, R01MH136149), NIEHS (R01ES033630), and the Department of Defense (TP220451).