We searched PubMed and Cochrane libraries for all articles in any language from database inception to January 13, 2016, using search terms developed with a reference librarian (Appendix A). Reference sections of identified articles were reviewed to help identify additional studies.

We included all published clinical and PK studies in any language. Abstracts, conference presentations, dissertations, and other non-published results were excluded for the formal review. Articles were included if they studied women of reproductive age using any method of hormonal contraception [COCs, transdermal patches, or vaginal rings; progestin-only pills (POPs), injectables, implants; emergency contraceptive pills, or levonorgestrel (LNG) intrauterine devices (IUDs)] in combination with any included psychotropic medication. We used the term oral contraceptive (OC) if a study did not specify whether women were using COCs or POPs or both. We used the term hormonal contraceptive (HC) if a study did not specify route or type of HC administration. Psychotropic medications of interest, identified by consultation with women’s mental health experts and review of the American Psychiatric Association’s treatment guidelines, were selective serotonin reuptake inhibitors (SSRIs), serotonin-norepinephrine reuptake inhibitors (SNRIs), tricyclic antidepressants (TCAs), oral benzodiazepines, bupropion, mirtazapine, trazadone, buspirone, hydroxyzine, monoamine oxidase inhibitors (MAOIs), and atypical antipsychotics [3,4]. Clinical outcomes of interest included psychotropic drug effectiveness (e.g. treatment response using depression/anxiety scale scores), contraceptive effectiveness (unintended pregnancy rates, as well as proxy measures including breakthrough bleeding and measures of ovulation) and adverse health effects related to the psychotropic medication or contraceptive method. Studies with PK outcomes of either the psychotropic drug or contraceptive steroid hormone were included.

One author (E.B.B.) performed the database search and screened all titles and abstracts. Two authors (E.B.B. and K.C.) reviewed the full text of all possible articles to determine which articles met inclusion criteria.

One common method for evaluating possible clinical significance of statistically significant differences in PK parameters is to calculate geometric mean ratios for PK parameters [e.g. geometric mean ratio for area under the curve_{drug A} (AUC) = AUC_{drug A} in users of drug B/AUC_{drug A} in non-users of drug B ×100], construct 90% confidence intervals (CIs) around that ratio and set a pre-defined range (typically a 90% CI of 80–125%) that would suggest a lack of interaction [13]. Studies were not required to perform these calculations for inclusion in this review; however, if performed these calculations were used to interpret clinical significance.

Study design, sample size, validity, and generalizability were used to assess study quality. All clinical studies were assigned quality using the three-level United States Preventative Services Task Force grading scale (good, fair, poor) [15]. As no standard guidelines exist to assess quality of PK studies, we designed a 7-item quality rating system and applied it to the identified PK articles (Appendix B). The quality of each study was assigned independently by two authors. Any differences were resolved through discussion with a third author.

The data were assessed and summarized using standard abstraction forms. We constructed separate evidence tables for each class of psychotropic drug. Meta-analysis could not be conducted due to heterogeneity of study design and outcomes, as well as limited data in most classes of drugs.

In a single PK study of women using bupropion and HCs, 12 healthy women (age 20–25) were administered a single dose of bupropion (150 mg) in phase 1 [26]. In phase 2, the women were pretreated with a COC (30 mcg EE and 150 mcg desogestrel) for 10 days and on day 10 administered a single dose of bupropion (150 mg). The AUC_bupropion for the COC phase was 19% less than with bupropion alone (0.72 vs. 0.89 mcg/mL·per hour; pb.001). For the active metabolite, the AUC_{hydroxybupropion} for the COC phase was 31% less than with bupropion alone (11 vs. 16 mcg/mL·per hour; pb.001), and the C_max was also significantly less with COCs compared to bupropion alone (20% decrease; p=.009). The ratio between the AUC of the active metabolite to bupropion was not significantly different between phases [26].

In a double-blind placebo controlled cross-over trial, 19 healthy women were administered a COC (30 mcg EE and 150 mcg LNG) for 21 days plus a placebo or ziprasidone (40 mg/day in divided doses) on days 8–15. [27] No difference in the AUC or C_max for either steroid hormone was seen in the placebo compared with ziprasidone arms. This study also collected limited information on treatment-related adverse events. There were no serious adverse events and most mild–moderate adverse events were typical of those seen on ziprasidone therapy alone; however, the study did not include women using ziprasidone alone [27].

Data from a routine therapeutic drug monitoring center in Norway were examined to investigate potential interactions between HCs and olanzapine [28]. Dose-adjusted olanzapine concentrations (and metabolites) were compared for reproductive aged women on estrogen-containing contraceptives (n=10), progestin-only contraceptives (n=10) or non-contraceptive users (n=129). A 33% decrease in the dose adjusted concentration of the metabolite (N-desmethyl olanzapine) was seen for estrogen containing contraceptive users compared with non-contraceptive users (1.3 vs 1.95; p=.03). No other significant differences were seen in the dose-adjusted olanzapine or metabolite concentrations among groups [28].

In a double-blind placebo controlled cross-over trial, 17 healthy women (age 18–40) were administered a COC (35 mcg EE and 180–250 mcg norgestimate) daily for 28 days plus a placebo or lurasidone (40 mg) on days 12–21 [29]. The AUC and C_max values of both steroid hormones on day 21 were similar between placebo and lurasidone treatment arms. In addition, the 90% CIs for the geometric mean ratios did not exceed the no-effect boundary. One subject experienced dysmenorrhea while taking COC plus lurasidone but no serious treatment-related adverse events were reported [29].

In a non-randomized, non-blinded trial, women (age 19–37) using COCs (50 mcg of EE or less, n=17) and controls not taking COCs (n=14), matched to COC users by smoking status, were administered a single dose of oxazepam (30 mg) [30]. No significant differences were observed in volume of distribution (Vd), elimination half life, total clearance or free fraction of oxazepam in plasma between users and non-users of COCs. AUC, t_max and C_max were not reported.

A smaller non-randomized, non-blinded trial examined PK parameters for a single dose of oxazepam (45 mg) in healthy women (age 21–33) taking a 50 mcg EE COC for at least 6 months (n=5) and controls not using COCs (n=6) [32]. No changes to oxazepam Vd was observed, though clearance increased by 157% (pb.01), and a nonsignificant decrease in elimination half life was reported. AUC, t_max and C_max were not reported.

Two non-randomized, non-blinded parallel studies examined four benzodiazepines administered with COCs [31,35]. Women taking COCs with 35 mcg of EE or less (n=19) and matched controls not taking COCs (n=21) received single doses of either (study 1) temazepam (30 mg) and triazolam (0.5 mg) or (study 2) alprazolam (1 mg) and lorazepam (2 mg) in two sequential COC cycles. In a parallel analysis, they observed no significant difference in any PK parameter between COC and non-COC users during administration of triazolam. However, COC users taking alprazolam demonstrated a higher AUC and lower elimination rate constant than non-users, with no changes to other parameters. COC users taking lorazepam had higher elimination rate constant, while COC users taking temazepam demonstrated a higher elimination rate constant and lower AUC than non-COC users taking either drug. Other parameters were unchanged. For clinical outcomes, COC users experienced greater impairment of psychomotor performance than non-users after taking alprazolam, lorazepam, and triazolam [31]. This increase in psychomotor impairment did not correlate with PK changes. COC users did not differ from non-users in measures of sedation and memory for any of the four benzodiazepines, though the study was underpowered for these outcomes.

One non-randomized, non-blinded study examined PK parameters after a single alprazolam (1 mg) dose in COC users (less than 50 mcg EE, n=16) and non-users (n=23) [33]. AUC, half life, Vd and clearance did not differ between groups; C_max and t_max were not reported.

One observational study reported the incidence of breakthrough bleeding in 72 women taking COCs (EE ≤50 mcg) in combination with an oral sedative (diazepam 5 mg [n=15], chlordiazepoxide 5 mg [n=19], nitrazepam 5–10 mg [n=21], or meprobamate 200 mg [n=17]) [34]. No woman experienced breakthrough bleeding prior to initiation of the sedative, whereas 36.1% of women reported breakthrough bleeding after initiating a sedative, with most breakthrough bleeding occurring in users of chlordiazepoxide and meprobamate (Table 4). Only one of 15 diazepam users and two of 21 nitrazepam users reported breakthrough bleeding. This study did not control for dose or duration of use of the sedatives and no statistics were performed. No pregnancies were observed but the duration of follow up was not reported [34].

Limited data from five studies of co-administration of SSRIs and HCs overall suggest low concern for clinically significant interactions. One study found no difference in pregnancy rates in OC users taking and not taking fluoxetine [16]. Another found no difference in depression scales for HC versus non-HC users treated with citalopram [17]. However, these studies are of fair quality due to several limitations, such as not specifying how unintended pregnancy was measured, lack of information on response and follow-up rates, being unable to correlate time of unplanned pregnancy with OC use/compliance and not controlling for potential confounders [16,17]. Exposure to HCs (or OCs) was inadequately defined and/or measured and a wide range of types and doses of HCs were likely used.

Two PK studies of SSRIs and HCs were of poor quality due to the uncertain clinical relevance of the PK parameters measured, small sample size for parallel designs and a wide range of doses of both drugs [18–20]. One good quality PK study with the SSRI-like antidepressant vortioxetine was the only study to examine the PK parameters of the contraceptive hormones with use of an SSRI, and it reassuringly showed no clinically significant interactions [21].

Limited data from five studies of co-administration of TCAs and HCs overall suggest limited concern for clinically significant interactions. Two poor quality clinical studies of TCAs with HCs showed no increase in adverse events from the TCA. However, these studies were limited by the wide type and variety of OCs used, small sample sizes (n=15–20 per group) and high drop-out rates (N20%) (Table 2) [22,36]. The PK evidence for TCAs with OCs is of fair-poor quality but demonstrates no significant PK interaction with clomipramine, and a possible increase in concentrations of amitriptyline and imipramine among OC users compared with non-OC users [37,38]. However, the two PK studies that demonstrated increased concentrations of the TCA with amitriptyline and imipramine were of poor quality and limited by very small sample sizes (5 or 6 OC users per study), inclusion of a wide range of OC doses and formulations, failure to report AUC and C_max, and failure to state exposure assessment (Table 2). In addition, one study included only women with bulimia, which may have resulted in highly variable intake and absorption of medications [37]. Thus, although these studies raise concern that co-administration of OCs with certain TCAs might lead to increased exposure to that TCA, and thus theoretically have an increased potential for TCA-related adverse events, the data are fair to poor quality and cannot be clinically applied with full confidence. No studies were identified, either PK or clinical, that evaluated whether TCAs induce or inhibit HCs or affect contraceptive effectiveness.

One good-quality PK study demonstrated a modest decrease in the exposure to bupropion in COC users compared with non-users. [26] Strengths of the study included using a crossover design, measurement of appropriate PK parameters, and administration of the same COC to each subject. Given the small reduction in exposure to bupropion, theoretical concern for decreased clinical efficacy of bupropion among COC users exists. However, the clinical effect of a reduction of this magnitude is uncertain.

Although no clinical studies were identified, PK data from three fair to good quality studies examining the co-administration of atypical antipsychotics and HCs suggest low concern for significant interactions. Two good quality studies examined steroid hormone PK parameters, and neither demonstrated any significant changes in the AUC or C_max for the estrogen or progestin component of COCs [27,29]. Both used a crossover design, a single COC type/dose and measured clinically relevant PK parameters. One fair quality study that examined the PK parameters of olanzapine demonstrated no significant difference in the exposure of olanzapine between progestin-based contraceptive users, estrogen containing contraceptive users or non-contraceptive users [28]. It did demonstrate a decreased concentration of the metabolite N-desmethyl olanzapine in users of estrogen-containing contraceptives compared with the other two groups; however, the clinical relevance of this finding is unknown [28]. Though limited to PK data, the lack of interaction between HCs and atypical antipsychotics is more reassuring than data for other classes of psychotropic drugs due to the higher quality of studies.

Data from four fair quality PK studies showed minimal interaction of COCs on oral benzodiazepines and COCs. Two of these studies are limited by use of PK parameters of uncertain clinical significance [30,32]. Those which did address AUC and C_max of the benzodiazepine did not demonstrate any consistent direction of change for this class of drugs when co-administered with COCs [33,35]. While some benzodiazepines are metabolized by the cytochrome P450 system (alprazolam), others such as lorazepam and oxazepam are not, which could explain the heterogeneity in these results. Given the inconsistent and small magnitude of changes in PK, as well as one study which showed no difference in sedation or memory during co-administration of four benzodiazepines with COCs [31], these fair quality studies suggest minimal concern for a clinically significant interaction between COCs and oral benzodiazepines.

As the available published evidence examining drug interactions between HCs and psychotropic drugs used to treat anxiety and depression is limited, it is useful to consider theoretical concerns for possible interactions, i.e. the potential for a psychotropic drug to inhibit or induce the metabolic pathways of HCs and the potential for HCs to inhibit or induce the metabolic pathways of psychotropic drugs. We considered minor inducers or inhibitors as unlikely to cause clinically significant interaction, thus, we looked for at least moderate inhibitors or inducers of relevant CYP P450 enzymes. While several of the psychotropic drugs included in this review are thought to have minor inhibitory effects on the important CYP P450 enzymes in the HC pathway (CYP 3 A4 and CYP 2C9), none of the psychotropic drugs are moderate or strong inhibitors those enzymes [13]. Thus, limited theoretical concern exists for any of the psychotropic drugs to significantly inhibit HCs leading to increased concentrations of steroid hormones and posing contraception-related safety concerns.

Psychotropic drugs may also induce HCs. However, as little is known about the potential for psychotropic agents to induce CYP 450 enzymes, the theoretical concern for psychotropic drugs causing a decrease in steroid hormones concentrations, thus potentially decreasing contraceptive effectiveness, is unknown. This review identified four studies (one clinical and three PK) that addressed this concern and none demonstrated a significant effect of the psychotropic agent on HC concentrations [16,21,27,29]. Thus, limited clinical and PK data do not suggest a concern for decreased contraceptive effectiveness when HCs are co-administered with the psychotropic drugs in this review.

Next, we consider possible effects of HCs on psychotropic drugs. As COCs are considered as a moderate inhibitor of CYP 1 A2, when co-administered with psychotropic drugs metabolized by CYP 1 A2 (duloxetine, clomipramine, imipramine, amitriptyline, olanzapine, clozapine, ziprasidone and mirtazapine), COCs (and perhaps other HCs) may result in increased exposure to the psychotropic agents and potentially pose safety concerns [13]. Fair-to-poor quality studies in this review suggest that clomipramine [22–25] and olanzapine [28] PK are not significantly affected by HCs, but poor quality studies suggested that imipramine and amitriptyline concentrations may be increased by OCs [37,38]. No published articles were identified to address this concern for duloxetine, mirtazapine or clozapine. However, we identified four case reports not meeting criteria for this review that associated COC use with increased systemic exposure of clozapine leading to clinically significant adverse events such as pericarditis and severe fatigue, weakness and dizziness [39–42]. The finding that some TCAs may have increased concentrations when co-administered with HCs is of concern as many TCAs have narrower therapeutic windows than other psychotropic agents. TCAs are less often used to treat depression and anxiety disorders given the more favorable safety profile of newer agents such as SSRIs; however, they are used to treat chronic pain disorders and chronic migraines, which commonly affect women of reproductive age. In summary, although theoretical concern exists for HCs to moderately inhibit the metabolism of certain psychotropic drugs (metabolized by CYP 1 A2), the scant data which explore this concern are reassuring except in the case of amitriptyline, imipramine and clozapine, an atypical anti-psychotic. Further research examining the safety of these drugs with HC is needed.

There is no known theoretical concern for HCs to induce CYP enzymes, thus HCs are not likely to decrease concentrations of psychotropic drugs and lead to treatment failures. In the four clinical studies in this review that examined this question, no significant differences in psychotropic drug efficacy were found [16,17,22,36]. The only PK study to find a significant decrease in psychotropic drug concentrations examined bupropion with COCs [26]. The proposed mechanism of action for this finding is thought to be from EE potentially inhibiting the enzyme CYP 2B6, which is responsible for hydroxylation of bupropion into its active metabolite. Thus, inhibition of CYP 2B6 decreases conversion of bupropion into the active metabolite, resulting in increased concentrations of bupropion (inactive form) but decreased concentrations of the active drug. No clinical data are available to further investigate this interaction.

One strength of this review was our inclusion of all study designs on a wide range of psychotropic agents. However, this review is limited by the scarcity of published evidence, mostly of fair to poor quality, thus limiting definitive conclusions. Additionally, there is virtually no published information on the use of progestin-only methods or non-oral contraceptives (including long-acting reversible methods) with psychotropic drugs. One study attempted to differentiate between combined hormonal contraceptives and progestin-only contraceptives [28] but all other studies examined OCs only, often not specifying combined or progestin-only oral contraceptives. Due to potential differences in drug metabolism and drug interactions with different types of progestins and routes of administration, the findings of the studies included in this review cannot be assumed to apply to progestin-only or non-oral HC formulations. Likewise, many common drugs (e.g. sertraline, mirtazapine) and even entire classes of psychotropic agents (e.g. SNRIs and MAOIs) did not have any published data examining drug interactions with HCs. Given the high frequency of use for both HCs and psychotropic drugs among reproductive age women in the US, this review highlights a great need for further research in this area [43,44]. Well designed, good quality PK and clinical studies of commonly used psychotropic drugs with oral and non-oral HC could add substantially to the field.

The limited evidence on drug interactions between psychotropic drugs used to treat anxiety and depression and HCs suggests low concern for clinically significant interactions. However, theoretical concern, supported by limited PK studies, indicates that concomitant use of COCs and certain TCAs could have the potential to increase exposure to the TCA, potentially posing safety concerns for drugs with narrow therapeutic windows. The metabolism of HCs and psychotropic drugs is complex and often uncertain. No data exists on drug interactions for non-oral formulations of HC or for several classes of psychotropic drugs. Given the public health importance of providing guidance on the safety of contraceptive method use to prevent unintended pregnancies among women with depression and anxiety disorders, future clinical and pharmacokinetic studies are needed to investigate the safety and effectiveness of contraceptive use among women taking psychotropic drugs.