Association Between Intrauterine System Hormone Dosage and Depression Risk
Abstract
Objective:
The authors compared the associated risk of incident depression between first-time users of low-, medium-, and high-dose levonorgestrel-releasing intrauterine systems (LNG-IUSs).
Methods:
This national cohort study was based on Danish register data on first-time users of LNG-IUSs, 15–44 years of age, between 2000 and 2022. Cox regression and a G-formula estimator were used to report 1-year average absolute risks, risk differences, and risk ratios of incident depression, defined as initiation of an antidepressant or receipt of a depression diagnosis, standardized for calendar year, age, education level, parental history of mental disorders, endometriosis, menorrhagia, polycystic ovary syndrome, dysmenorrhea, leiomyoma, and postpartum initiation.
Results:
In total, 149,200 women started using an LNG-IUS, among whom 22,029 started a low-dose one (mean age, 22.9 years [SD=4.5]), 47,712 a medium-dose one (mean age, 25.2 years [SD=6.2]), and 79,459 a high-dose one (mean age, 30.2 years [SD=5.6]). The associated subsequent 1-year adjusted absolute risks of incident depression were 1.21% (95% CI=1.06–1.36), 1.46% (95% CI=1.33–1.59), and 1.84% (95% CI=1.72–1.96), respectively. For the users of high-dose LNG-IUSs, the risk ratios were 1.52 (95% CI=1.30–1.74) and 1.26 (95% CI=1.10–1.41) compared with users of the low- and medium-dose LNG-IUSs, respectively. For users of medium-dose LNG-IUSs, the risk ratio was 1.21 (95% CI=1.03–1.39) compared with users of low-dose LNG-IUSs.
Conclusions:
First-time use of an LNG-IUS was positively associated with incident depression in an LNG-dose-dependent manner across low-, medium-, and high-dose LNG-IUSs. Although the observational design of the study does not permit causal inference, the dose-response relationship contributes to the body of evidence suggesting a relationship between levonorgestrel exposure and risk of depression.
Use of levonorgestrel-releasing intrauterine systems (LNG-IUSs) has been associated with an approximately 40% increased risk of developing depression and initiating antidepressant medication (1–3). An increased risk of depression has also been observed when LNG-IUS users were compared with users of non-hormonal (copper) IUSs (4). However, the literature on mood symptoms related to LNG-IUS use has also shown mixed results (5). Most of the current literature is limited by the observational nature of studies and by vulnerability to confounding, including a healthy user bias (6), which may explain the conflicting results in the literature. Therefore, research is needed that addresses the potential relationship between exogenous hormone exposure from LNG-IUSs and risk of developing depression.
LNG-IUSs are available in over 120 countries worldwide, and they are used by about 14% of women of reproductive age in Denmark, equivalent to one in three hormonal contraceptive users (7, 8). Today, LNG-IUSs are available in three formulations, differentiated by the total LNG content and daily release dose: the low-dose LNG-IUS contains 13.5 mg, with a 14 μg/day initial release dosage and an 8 μg/day average over the first year (9); the medium-dose LNG-IUS contains 19.5 mg, with a 17.5 μg/day initial release dosage and a 12.6 μg/day average over the first year (10); and the high-dose LNG-IUS contains 52 mg, with a 21 μg/day initial release dosage and a 20 μg/day average over the first year (11). All three types are used as birth control, and the high-dose LNG-IUS is also approved for treating heavy menstrual bleeding (menorrhagia) (11). The mechanism of action of LNG-IUS as a contraceptive method is exerted via local effects in the uterus by inducing endometrial thinning and increased cervical mucus, which inhibits sperm survival and prevents fertilization and implementation of the oocyte (12). However, among users of LNG-IUSs, only about 50%–76.5%, 88%, and 97% of high-, medium-, and low-dose users, respectively, have ovulation during the first year after insertion; the rates increase to about 80%–100% the following year, which indicates some systemic effects of LNG-IUS, particularly early after insertion (13–15). Such effects are in line with a higher daily release dose of LNG during the early treatment phase, which declines with duration of use; the daily release doses decline to 6 μg/day, 10 μg/day, and 19 μg/day for low-, medium-, and high-dose LNG-IUSs, respectively, 1 year after insertion (9–11). LNG is able to inhibit ovulation because it is a potent progesterone receptor agonist, by which it can mimic endogenous progesterone effects and provide negative feedback to the hypothalamic-pituitary-gonadal axis (16). However, it also activates other receptors, including androgen, glucocorticoid, and mineralocorticoid receptors, with the potential to affect brain biology (17).
A recent observational study (18) that compared the risk of antidepressant use within 2 years after LNG-IUS use between the medium-dose and high-dose LNG-IUSs found an increased risk among high-dose LNG-IUS users. So far, no study has investigated the existence of a dose-response relationship across all three dosages while considering a close temporal relationship between the start of the hormone exposure and the subsequent risk of developing depression. This approach could further elucidate the relationship between progestogen exposure and subsequent development of mood symptoms. In this study, our aim was to compare the 1-year associated risk of incident depression between first-time users of low-, medium-, and high-dose LNG-IUSs, and we hypothesized that the associated risk would increase in a dose-dependent manner.
Methods
Study Design
This was a population-based cohort study based on Danish registry data. The registers and variables used are listed in Table S1 in the online supplement . Data were provided by the Danish e-Health Authority, hosted by Statistics Denmark, and were linked via the unique personal identification number given to Danish residents at birth or immigration. The study period was from 1995 to 2022, in which the first 5 years were used to distinguish incident from prevalent LNG-IUS and antidepressant users.
Study Population
All women living in Denmark born in 1978 or after (ensuring that there was sufficient prescription history to identify incidence use) who started use of an LNG-IUS for the first time between ages 15 and 44 between January 1, 2000, and December 31, 2022, were eligible for inclusion. Women were excluded if they immigrated or emigrated after turning 15, if they had previously been diagnosed with a mental disorder in a psychiatric inpatient or outpatient clinic (ICD-8 codes 290–315 or ICD-10 codes F00–F99), or if they had previously used a psychotropic drug (Anatomical Therapeutic Chemical [ATC] codes N05* and N06*).
Exposure
The exposure of interest was incident use of an LNG-IUS (ATC code G02BA03), as identified via the Danish Prescription Register (19), with different initial hormone-release dosages: low-dose (14 μg/day), medium-dose (17.5 μg/day), or high-dose (21 μg/day). The date of filling the first prescription was used as the index date.
Outcome
The outcome was incident depression within 12 months after initiating an LNG-IUS. Depression was defined as incident use of antidepressant medication (ATC codes N06A*) or incident diagnosis of depression at an inpatient or outpatient psychiatric department (ICD-10 codes F32–F34, F38, F39).
Covariates
We considered the following potential confounders, defined at the index date: age (in 5-year periods), highest education level (below high school, high school/vocational education, or bachelor’s degree or above), parental history of mental disorders, and postpartum LNG-IUS use, defined as use within 6 months after delivery. In addition, we accounted for potential confounding by indication by including the following medical diagnoses, given that high-dose LNG-IUS may be preferred for women diagnosed with these conditions: menorrhagia (20), leiomyoma (21), polycystic ovary syndrome (22), dysmenorrhea, and endometriosis (23). Finally, we controlled for time trends by including calendar year in 5-year periods.
Statistical Analysis
Women were followed for 12 months from first-time use of an LNG-IUS or until incident depression, emigration, death, or end of study period, whichever came first. A single Cox regression model (given that too few study subjects died during follow-up to allow estimation of the parameters of a cause-specific Cox model relative to death) was used to estimate cause-specific hazard rates between low-, medium-, and high-dose LNG-IUS use, adjusted for the listed covariates. Time in study was used as underlying time scale. Estimated hazard rates were used as inputs in a G-formula estimator to determine the associated average absolute risks, risk differences, and risk ratios after 12 months of use (24).
Several sensitivity analyses were conducted. First, depression diagnosis and antidepressant use were used separately as outcome, and the latter also after excluding prescriptions specifically addressed for other indications unless depression was confirmed by diagnosis. Second, the study period was restricted to 2017–2022, during which all three LNG-IUSs were concurrently prescribed in Denmark. Third, the population was restricted to nulliparous first-time users under age 30, as the frequency of prescribed high-dose LNG-IUSs is particularly high in the population over age 30 compared with low- and medium-dose LNG-IUSs (8). Fourth, potential confounding by indication was further handled by adjusting for menorrhagia as LNG-IUS prescription indication and for number of previous hormonal contraceptive types used (specified in Table S1 in the online supplement ). Missing prescription indications were handled by performing multiple imputations with the MICE package in R, where prescription indication was imputed using logistic regression and education level using polytomous logistic regression based on the listed covariates and outcome with 10 iterations and five imputations (25). Fifth, E-values were estimated as a quantitative summary of the strength of the association of an unmeasured confounder with the exposure and the outcome required to solely explain the observed associations (26). Sixth, an alternative modeling approach for handling confounding was assessed using a propensity score–weighted analysis in which the covariate effects are not assumed to be time constant and the additivity assumption of the covariate effects are made on the logit scale for the probability of treatment, and not on the log hazard scale. This was performed on first-time users during the 2017–2022 period (to ensure overlap of the distribution of calendar periods across all three groups) using multinomial propensity scores based on the listed covariates by use of the mnps function from the twang package in R (27, 28). And seventh, the potential influence of prior mental comorbidity excluded in the main analysis was assessed by restricting the population to women with a prior mental disorder but no potentially ongoing depression, defined as having received a depression diagnosis or filled a prescription for antidepressant medication within 2 years prior to filling the LNG-IUS prescription.
The proportional hazard assumption was assessed for each dependent variable by evaluating the correlation between scaled Schoenfeld residuals against time with the survival package in R (29). In case of a violation of the proportional hazard assumption, a stratified Cox proportional hazard model was used. Outcome estimates were calculated with 95% confidence intervals, and the null hypothesis was rejected if the confidence interval did not overlap with zero for the absolute risk difference and 1.00 for the risk ratio. Analyses were performed using R, version 4.2.2, and the riskRegression package was used to compute average treatment effects (30).
Approval to access national health register data was granted through the Danish Data Protection Agency. According to Danish law, no ethical approval or informed consent is needed for Danish register studies. The study was reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (31).
Results
The study population included 149,200 first-time LNG-IUS users (see Figure S1 in the online supplement ), of whom 22,029 (14.8%) used low-dose, 47,712 (32.0%) medium-dose, and 79,459 (53.3%) high-dose LNG-IUSs (Table 1). First-time users of low-dose LNG-IUSs were younger on average (mean age, 22.9 years [SD=4.5]) than first-time users of medium-dose (mean age, 25.2 years [SD=6.2]) and high-dose LNG-IUSs (mean age, 30.2 years [SD=5.6]). Correspondingly, their education level was lower on average, and they were more often nulliparous (89.7%, 69.8%, and 17.9% for the low-, medium-, and high-dose groups, respectively). Compared with first-time users of low-dose LNG-IUSs, high-dose LNG-IUS users were more often diagnosed with potential medical indications for LNG-IUS use, but this was not the case for medium-dose LNG-IUS, except for the diagnosis of polycystic ovary syndrome. Notably, rates of having menorrhagia as a prescription indication were 3.4%, 6.1%, and 10.9% in the low-, medium-, and high-dose LNG-IUS groups.
| Measure | LNG-IUS Exposure | |||||
|---|---|---|---|---|---|---|
| Low Dose (N=22,029) (14.8%) | Medium Dose (N=47,712) (32.0%) | High Dose (N=79,459) (53.3%) | ||||
| N | % | N | % | N | % | |
| Age group (years) | | | | | | |
| 15–19 | 6,362 | 28.9 | 10,498 | 22.0 | 3,839 | 4.8 |
| 20–24 | 9,707 | 44.1 | 17,175 | 36.0 | 10,799 | 13.6 |
| 25–29 | 4,173 | 18.9 | 9,282 | 19.5 | 21,675 | 27.3 |
| 30–34 | 1,403 | 6.4 | 6,429 | 13.5 | 27,681 | 34.8 |
| 35–39 | 357 | 1.6 | 3,348 | 7.0 | 12,847 | 16.2 |
| 40–44 | 27 | 0.1 | 980 | 2.1 | 2,618 | 3.3 |
| Calendar period | | | | | | |
| 2000–2004 | 0 | 0.0 | 0 | 0.0 | 419 | 0.5 |
| 2005–2009 | 0 | 0.0 | 0 | 0.0 | 4799 | 6.0 |
| 2010–2014 | 2,009 | 9.1 | 0 | 0.0 | 22,858 | 28.8 |
| 2015–2019 | 14,296 | 64.9 | 14,603 | 30.6 | 34,727 | 43.7 |
| 2020–2022 | 5,724 | 26.0 | 33,109 | 69.4 | 16,656 | 21.0 |
| Education levelb | | | | | | |
| Below high school | 6.374 | 28.9 | 10,484 | 22.0 | 9,358 | 11.8 |
| High school/vocational education | 10,274 | 46.6 | 21,506 | 45.1 | 30,783 | 38.7 |
| Bachelor’s degree or above | 5,368 | 24.4 | 15,706 | 32.9 | 39,252 | 49.4 |
| Postpartum initiation | 1,355 | 6.2 | 7,135 | 15.0 | 28,404 | 35.7 |
| Nulliparous | 19,765 | 89.7 | 33,311 | 69.8 | 14,216 | 17.9 |
| Parental history of mental disorders | 3,044 | 13.8 | 6,930 | 14.5 | 10,223 | 12.9 |
| Endometriosis | 68 | 0.3 | 186 | 0.4 | 1,273 | 1.6 |
| Polycystic ovary syndrome | 104 | 0.5 | 372 | 0.8 | 1,268 | 1.6 |
| Dysmenorrhea | 162 | 0.7 | 314 | 0.7 | 747 | 0.9 |
| Leiomyoma | 12 | 0.1 | 54 | 0.1 | 242 | 0.3 |
| Menorrhagia | 327 | 1.5 | 694 | 1.5 | 2,442 | 3.1 |
| Menorrhagia as prescription indicationc | 753 | 3.4 | 2,926 | 6.1 | 8.658 | 10.9 |
| Previous hormonal contraceptive use | | | | | | |
| None | 2,582 | 11.7 | 4,742 | 9.9 | 2,166 | 2.7 |
| One type | 15,257 | 69.3 | 32,871 | 68.9 | 52,671 | 66.3 |
| Two types | 3,778 | 17.2 | 9,232 | 19.3 | 21,718 | 27.3 |
| Three or more types | 412 | 1.9 | 867 | 1.8 | 2,904 | 3.7 |
a
LNG-IUS=levonorgestrel-releasing intrauterine system.
b
Data on education level were missing for 0.1%, 0.0%, 0.1% of low-, medium-, and high-dose LNG-IUS users, respectively.
c
Data on prescription indication were missing for 11.5%, 4.7%, and 12.0% of low-, medium-, and high-dose LNG-IUS users, respectively.
Within 12 months after initiating low-dose LNG-IUS use, 279 women were registered with incident depression, compared with 633 among medium-dose users and 1,346 among high-dose users. This corresponded to adjusted absolute risks of 1.21% (95% CI=1.06–1.36), 1.46% (95% CI=1.33–1.59), and 1.84% (95% CI=1.72–1.96), respectively. The risk ratios of developing depression for high-dose LNG-IUS use were 1.52 (95% CI=1.30–1.74) and 1.26 (95% CI=1.10–1.41) compared with low-dose and medium-dose LNG-IUS use, respectively. The risk ratio for medium-dose LNG-IUS use was 1.21 (95% CI=1.03–1.39) compared with low-dose LNG-IUS use (Table 2, Figure 1A–B). The estimated hazard ratios for the complete Cox proportional hazard model are shown in Figure S2 in the online supplement . Similar trends were found when the outcome was analyzed as incident antidepressant use and depression diagnosis, separately (see Table S2 in the online supplement ). However, given the lower number of depression diagnoses, these risks were estimated with larger uncertainty. When prescriptions for antidepressant use with indications other than depression were excluded, the estimates supported the main finding (see Table S3 in the online supplement ).
| Measure | LNG-IUS Exposure | |||||
|---|---|---|---|---|---|---|
| Low Dose | Medium Dose | High Dose | ||||
| % | 95% CI | % | 95% CI | % | 95% CI | |
| Absolute riskb | 1.19 | 1.04–1.33 | 1.43 | 1.30–1.56 | 1.88 | 1.75–2.00 |
| Absolute riskc | 1.21 | 1.06–1.36 | 1.46 | 1.33–1.59 | 1.84 | 1.72–1.96 |
| Absolute risk differencec | Reference | 0.26 | 0.06–0.45 | 0.63 | 0.43–0.83 | |
| Risk ratioc | Reference | 1.21 | 1.03–1.39 | 1.52 | 1.30–1.74 | |
| Absolute risk differencec | | | Reference | 0.37 | 0.17–0.58 | |
| Risk ratioc | | | Reference | 1.26 | 1.10–1.41 | |
a
LNG-IUS=levonorgestrel-releasing intrauterine system.
b
Standardized over calendar period and age.
c
Standardized over calendar period, age, education level, parental history of mental disorders, endometriosis, polycystic ovary syndrome, dysmenorrhea, leiomyoma, menorrhagia, and postpartum incident LNG-IUS use. Data on education level were missing for 0.1%, 0.0%, 0.1% of low-, medium-, and high-dose LNG-IUS users, respectively. These were imputed to a separate group in the analyses.
FIGURE 1. Depression among first-time users of low-, medium-, and high-dose LNG-IUSsa
aPanel A shows depression-free survival curves for the different LNG-IUS dosages during the year following initiation. Shading represents 95% confidence intervals. Panel B shows the 1-year estimated depression risks for the different LNG-IUS dosages, standardized over calendar period, age, education level, parental history of mental disorders, endometriosis, polycystic ovary syndrome, dysmenorrhea, menorrhagia, leiomyoma, and postpartum incident LNG-IUS use. Error bars represent 95% confidence intervals. LNG-IUS=levonorgestrel-releasing intrauterine system.
*<0.05. ***<0.001.
When the population was restricted to first-time users between 2017 and 2022 (for demographic and clinical characteristics, see Table S4 in the online supplement ), the risk ratios were 1.37 (95% CI=1.10–1.64) and 1.17 (95% CI=1.01–1.34) between high-dose and low-dose and between high-dose and medium-dose LNG-IUS use, respectively (see Table S5 in the online supplement ). When the population was restricted to nulliparous first-time users under age 30, the groups were more comparable in terms of education level and age, with mean ages of 21.7 years [SD=3.3], 21.7 years [SD=3.2], and 22.3 years [SD=3.3] for low-, medium-, and high-dose LNG-IUS users, respectively (see Table S6 in the online supplement ). The results based on this population (see Table S7 in the online supplement ) were also in line with the main analysis, with risk ratios of 1.42 (95% CI=1.16–1.68) and 1.21 (95% CI=0.98–1.44) between high-dose and low-dose and between high-dose and medium-dose LNG-IUS use, respectively. When menorrhagia as prescription indication and previous hormonal contraceptive use were adjusted for, the risk ratios were 1.50 (95% CI=1.28–1.71) and 1.24 (95% CI=1.08–1.40) between high-dose and low-dose and between high-dose and medium-dose LNG-IUS users, respectively (see Table S8 in the online supplement ). The risk ratios of depression due to menorrhagia as diagnosis and prescription indication were 1.52 (95% CI=1.20–1.85) and 1.19 (95% CI=1.02–1.36), respectively. Based on the estimated risk ratios from the main analysis, E-values were estimated at 2.41 (with lower border of 1.92) and 1.71 (with lower border of 1.21) for the association of depression in high- and medium-dose compared with low-dose LNG-IUS users, respectively. The propensity score–weighted analysis, which showed balanced standardized mean differences of covariates (see Table S9 in the online supplement ) and overlapping propensity score distributions (see Figure S3 in the online supplement ), showed risks comparable to those obtained with the G-formula: 1.16%, 1.47%, and 1.62% in absolute risk in low-, medium-, and high-dose LNG-IUS users and hazard ratios of 1.40 (95% CI=1.10–1.77) and 1.27 (95% CI=1.02–1.59) for high- and medium-dose compared with low-dose LNG-IUS users. In the final sensitivity analysis on women with prior mental disorders (N=48,937), the absolute risks were 4.80%, 5.22%, and 5.91% for low-, medium-, and high-dose LNG-IUS users, respectively, with risk ratios of 1.23 (95% CI=1.07–1.40) and 1.13 (95% CI=1.01–1.25) for high-dose compared with low- and medium-dose LNG-IUS users (see Table S10 in the online supplement ).
Discussion
This population-based cohort study demonstrated a positive association between LNG dosage and risk of incident depression across initial LNG release dosages of 14 μg/day, 17.5 μg/day, and 21 μg/day. Notably, the 1-year absolute risks of depression were relatively small, at 1.21%, 1.46%, and 1.84% in first-time users of low-, medium-, and high-dose LNG-IUSs, respectively.
These results are in line with findings from an international multicenter randomized study linking LNG exposure and risk of developing depressive symptoms (32). In that study, a higher frequency of self-reported depressive symptoms in the initial treatment phase and higher rates of discontinuation reported as being due to depression were observed among women allocated to LNG-IUS compared with women allocated to copper IUS. Furthermore, our results are consistent with those from a recent observational study (18) that showed a 13% higher risk of antidepressant use among high-dose compared with medium-dose LNG-IUS users.
A recent study comparing the incidence of starting antidepressant medication associated with first-time use of the three different dosages of LNG-IUS in a population restricted to nulliparous women (33) similarly showed an association with increased 1-year risk among high-dose (34.8 per 1000 person-years) compared with medium-dose (18.6 per 1000 person-years) and low-dose (14.0 per 1000 person-years) LNG-IUS users. However, the study found no statistically significant difference between low- and medium-dose LNG-IUS users, which was explained by a lack of difference in LNG plasma levels among users of these two LNG-IUSs. This explanation, however, is not supported by studies demonstrating that hormone exposure increases with increasing release dosage (15, 34). The present study excluded women who had prior mental disorders or had ever used a psychotropic drug. Prior mental disorders appeared to attenuate the estimated relative risk, despite the presence of absolute differences, but these were not sufficiently large to follow the relative higher absolute risk in this population. This observation has also been reported in other studies (4, 35). It does not necessarily contradict the notion that having a prior mental disorder increases the likelihood of experiencing mood symptoms after initiation of hormonal contraception (36). However, the relatively lower risk associated with hormonal contraceptive use or with the higher LNG-IUD dosage in the high-risk population could be explained by a lower representation of the more hormone-susceptible population when different etiologies are likely to be highly represented. Alternatively, it could also be explained by a potential selection bias if women with prior mental disorders are more likely to prefer lower hormone exposure because of concern about the potential mood side effects.
Intriguingly, our study provides evidence of a dose-dependent association between LNG exposure and risk of subsequent depression across three dosages, which was consistent after considering potential confounders, such as menstrual bleeding indications for high-dose LNG-IUS use. As addressed in a recent meta-analysis (37), the role of the hormonal contraceptive formulation in mood symptoms has been inconclusive; however, the present findings add new evidence on the potential role of the progestogen component of hormonal contraception in the development of mood disorders (38). The knowledge on how progestogens affect the brain and mood and whether the route of administration plays a role is still limited (5). Some evidence points to a potential mechanism through modulating effects on the regulation of stress reactivity via the hypothalamic-pituitary-adrenal axis (39), and preclinical evidence points to a link between LNG-induced alterations in brain γ-aminobutyric acid receptor plasticity and anxiety-like behavior in rats (40).
Even though the 21% and 52% higher relative risks associated with medium- and high-dose compared with low-dose LNG-IUSs amounted to absolute differences of only 0.27% and 0.66%, they highlight the clinical relevance of considering LNG hormone exposure dosage in contraceptive counseling, which may prove relevant together with information on prior reactions to hormonal contraceptives or postpartum depressive episodes (41). These risk differences should be weighed against beneficial medical aspects, including better bleeding control when using high-dose LNG-IUS (20), but also against possible negative outcomes, including an increased risk of ectopic pregnancy with lower LNG dosage (42, 43). It also highlights the importance of educating women to be aware of potential mental health side effects and of clinical evaluation of mood symptoms at follow-up visits after LNG-IUS insertion.
The strengths of the present study include the use of national register data, enabling the study of an unselected nationwide population of first-time users of the different LNG-IUS dosages who had never used any form of psychotropic medication or had any prior mental disorders. Furthermore, it is a strength that the study compares the risk between users of different LNG-IUS dosages and not to a non-user population, which makes it less prone to a surveillance bias (44) or behavioral confounders related the point of starting contraception (45). Another strength is that the study used both prescriptions with information about prescription indication and depression diagnoses to best capture depressive episodes. Last, and importantly, the use of a relatively short follow-up period increases the likelihood that the event of starting an LNG-IUS and the event of depression could be causally linked, and it reduces the risk of healthy survivor bias and the number of women who discontinued within the follow-up period. Discontinuation most often goes undetected in the registers, yet, even if censoring at time of discontinuation had been possible, discontinuation could be due to mood symptoms, and thus it would be a violation of discontinuation being independent of the outcome.
The study also has limitations. First and foremost, an observational study does not allow direct causal inference, although the choice of a close temporal window between incident use of an LNG-IUS and registration of a depressive episode and the demonstration of a dose-response relationship jointly add indirect evidence supporting a causal relationship. Second, prescription patterns of antidepressants and LNG-IUSs changed during the study period, which may introduce bias; however, the sensitivity analysis restricting the study period to 2017–2022 supported the main finding. Third, first-time users of the different LNG-IUSs differed in age and parity, and higher age and recent delivery may introduce confounding by indication, as recent delivery and some chronic diseases would make women choose a progesterone-only contraceptive. However, the sensitivity analysis restricted to nulliparous first-time users under age 30, where the groups were more comparable in age, confirmed the main results. Fourth, potential confounders include undiagnosed medical conditions and unregistered off-label medical indications for the use of the high-dose LNG-IUS. However, women who had an LNG-IUS inserted at a hospital for treatment of a medical condition or who bought LNG-IUSs directly in a specialist clinic were not included in the study. Further, we did account for various diagnosed medical indications and menorrhagia as prescription indication, and as much as 4.5%, 7.3%, and 12.4% were known to have menorrhagia (diagnosed or via prescription indication) among first-time users of low-, medium-, and high-dose LNG-IUSs, respectively. In comparison, the prevalence of menorrhagia is estimated to be 9%–14%, although self-reported estimates reach 27% (46, 47). Nonetheless, this had little impact on the results, which may be related to the fact that high-dose LNG-IUSs are effective in treating these medical disorders, resulting in improvement in quality of life (48, 49). Additionally, based on the E-values, an unmeasured confounder should be associated with the exposure and the outcome with a strength of 2.41 and 1.71 to explain the increased risk associated with high- and medium-dose compared with low-dose LNG-IUSs, which is larger than the estimated risk ratios of menorrhagia as diagnosis and as prescription indication. This indicates that the confounding would need to be strong for the observed association to hold under the null hypothesis (i.e., a false positive result).
As hypothesized, this study provides evidence of a dose-dependent association between intrauterine levonorgestrel exposure and incident depression; that is, high-dose LNG-IUS use was associated with a 26% higher relative risk than medium-dose LNG-IUS use, which was associated with a 21% higher relative risk than low-dose LNG-IUS use. These differences remained after controlling for medical diagnoses and indications for prescribing high-dose LNG-IUSs. The 1-year absolute risks were 1.21%, 1.46%, and 1.84% for low-, medium-, and high-dose LNG-IUSs, emphasizing that increasing dosages were associated with only a small proportion of women developing depression to a degree that resulted in medical attention. These findings should be interpreted in light of the limitations of an observational study design with risk of residual confounding, and the observed risk differences should be weighed against potential benefits as well as other side effects of LNG-IUS use when providing personalized contraceptive counseling.
Supplementary Material
File (appi.ajp.20230909.ds001.pdf)
- View/Download
- 492.40 KB
References
1.
Skovlund CW, Mørch LS, Kessing LV, et al: Association of hormonal contraception with depression. JAMA Psychiatry 2016; 73:1154–1162
2.
Lundin C, Wikman A, Lampa E, et al: There is no association between combined oral hormonal contraceptives and depression: a Swedish register-based cohort study. BJOG 2022; 129:917–925
3.
Stenhammar E, Wikman P, Gemzell Danielsson K, et al: Levonorgestrel intrauterine device and depression: a Swedish register-based cohort study. Int J Psychophysiol 2023; 193:112230
4.
Slattery J, Morales D, Pinheiro L, et al: Cohort study of psychiatric adverse events following exposure to levonorgestrel-containing intrauterine devices in UK general practice. Drug Saf 2018; 41:951–958
5.
Bürger Z, Bucher AM, Comasco E, et al: Association of levonorgestrel intrauterine devices with stress reactivity, mental health, quality of life and sexual functioning: a systematic review. Front Neuroendocrinol 2021; 63:100943
6.
Johansson T, Vinther Larsen S, Bui M, et al: Population-based cohort study of oral contraceptive use and risk of depression. Epidemiol Psychiatr Sci 2023; 32:e39
7.
Buhling KJ, Zite NB, Lotke P, et al: Worldwide use of intrauterine contraception: a review. Contraception 2014; 89:162–173
8.
Kristensen SI, Lidegaard Ø: Hormonal contraceptive use in Denmark 2010–2019. Dan Med J 2021; 68:A08200599
9.
Bayer Group: Jaydess 13.5 mg intrauterine contraceptive device. 2014. https://resources.bayer.com.au/resources/uploads/DataSheet/file10780.pdf
10.
Bayer HealthCare Pharmaceuticals Inc: Kyleena product information. 2016. https://labeling.bayerhealthcare.com/html/products/pi/Kyleena_PI.pdf
11.
Bayer HealthCare Pharmaceuticals Inc: Mirena product information. 2000. https://labeling.bayerhealthcare.com/html/products/pi/Mirena_PI.pdf
12.
Erkkola R, Landgren BM: Role of progestins in contraception. Acta Obstet Gynecol Scand 2005; 84:207–216
13.
Xiao BL, Zhou LY, Zhang XL, et al: Pharmacokinetic and pharmacodynamic studies of levonorgestrel-releasing intrauterine device. Contraception 1990; 41:353–362
14.
Xiao B, Zeng T, Wu S, et al: Effect of levonorgestrel-releasing intrauterine device on hormonal profile and menstrual pattern after long-term use. Contraception 1995; 51:359–365
15.
Apter D, Gemzell-Danielsson K, Hauck B, et al: Pharmacokinetics of two low-dose levonorgestrel-releasing intrauterine systems and effects on ovulation rate and cervical function: pooled analyses of phase II and III studies. Fertil Steril 2014; 101:1656–1662.e1–4
16.
Bashour NM, Wray S: Progesterone directly and rapidly inhibits GnRH neuronal activity via progesterone receptor membrane component 1. Endocrinology 2012; 153:4457–4469
17.
Pletzer B, Winkler-Crepaz K, Maria Hillerer K: Progesterone and contraceptive progestin actions on the brain: a systematic review of animal studies and comparison to human neuroimaging studies. Front Neuroendocrinol 2023; 69:101060
18.
Roland N, Baricault B, Weill A, et al: Association between doses of levonorgestrel intrauterine systems and subsequent use of psychotropic drugs in France. JAMA 2023; 329:257–259
19.
Wallach Kildemoes H, Toft Sørensen H, Hallas J: The Danish national prescription registry. Scand J Public Health 2011; 39:38–41
20.
Goldthwaite LM, Creinin MD: Comparing bleeding patterns for the levonorgestrel 52 mg, 19.5 mg, and 13.5 mg intrauterine systems. Contraception 2019; 100:128–131
21.
Xie ZW, Zhang YN, Wan S, et al: Levonorgestrel-releasing intrauterine device is an efficacious contraceptive for women with leiomyoma. J Int Med Res 2012; 40:1966–1972
22.
Oguz SH, Yildiz BO: An update on contraception in polycystic ovary syndrome. Endocrinol Metab (Seoul) 2021; 36:296–311
23.
Bahamondes L, Petta CA, Fernandes A, et al: Use of the levonorgestrel-releasing intrauterine system in women with endometriosis, chronic pelvic pain and dysmenorrhea. Contraception 2007; 75:S134–S139
24.
Ozenne BMH, Scheike TH, Stærk L, et al: On the estimation of average treatment effects with right-censored time to event outcome and competing risks. Biom J 2020; 62:751–763
25.
van Buuren S, Groothuis-Oudshoorn K: mice: multivariate imputation by chained equations in R. J Stat Softw 2011; 45:1–67
26.
VanderWeele TJ, Ding P: Sensitivity analysis in observational research: introducing the E-value. Ann Intern Med 2017; 167:268–274
27.
Burgette L, Griffin BA, McCaffrey D. Propensity scores for multiple treatments: a tutorial for the mnps function in the twang package. 2023. https://cran.r-project.org/web/packages/twang/vignettes/mnps.pdf
28.
McCaffrey DF, Griffin BA, Almirall D, et al: A tutorial on propensity score estimation for multiple treatments using generalized boosted models. Stat Med 2013; 32:3388–3414
29.
Therneau T: A package for survival analysis in R (R Package Version 3.5-7), 2023. https://CRAN.R-project.org/package=survival
30.
Ozenne B, Sørensen A, Scheike T, et al: riskRegression: predicting the risk of an event using Cox regression models. R J 2017; 9:440–460
31.
von Elm E, Altman DG, Egger M, et al: The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. J Clin Epidemiol 2008; 61:344–349
32.
Andersson K, Odlind V, Rybo G: Levonorgestrel-releasing and copper-releasing (Nova T) IUDs during five years of use: a randomized comparative trial. Contraception 1994; 49:56–72
33.
Skovlund CW, Møller AL, Torp-Pedersen C, et al: Depression risk in users of different doses of levonorgestrel intrauterine systems: a nationwide prospective cohort study. Lancet Reg Health Eur 2024; 38:100813
34.
Hofmann BM, Apter D, Bitzer J, et al: Comparative pharmacokinetic analysis of levonorgestrel-releasing intrauterine systems and levonorgestrel-containing contraceptives with oral or subdermal administration route. Eur J Contracept Reprod Health Care 2020; 25:417–426
35.
Anderl C, de Wit AE, Giltay EJ, et al: Association between adolescent oral contraceptive use and future major depressive disorder: a prospective cohort study. J Child Psychol Psychiatry 2022; 63:333–341
36.
Bengtsdotter H, Lundin C, Gemzell Danielsson K, et al: Ongoing or previous mental disorders predispose to adverse mood reporting during combined oral contraceptive use. Eur J Contracept Reprod Health Care 2018; 23:45–51
37.
Kraft MZ, Rojczyk P, Weiss T, et al: Symptoms of mental disorders and oral contraception use: a systematic review and meta-analysis. Front Neuroendocrinol 2024; 72:101111
38.
Stefaniak M, Dmoch-Gajzlerska E, Jankowska K, et al: Progesterone and its metabolites play a beneficial role in affect regulation in the female brain. Pharmaceuticals (Basel) 2023; 16:520
39.
Aleknaviciute J, Tulen JHM, De Rijke YB, et al: The levonorgestrel-releasing intrauterine device potentiates stress reactivity. Psychoneuroendocrinology 2017; 80:39–45
40.
Porcu P, Mostallino MC, Sogliano C, et al: Long-term administration with levonorgestrel decreases allopregnanolone levels and alters GABA(A) receptor subunit expression and anxiety-like behavior. Pharmacol Biochem Behav 2012; 102:366–372
41.
Larsen SV, Mikkelsen AP, Lidegaard Ø, et al: Depression associated with hormonal contraceptive use as a risk indicator for postpartum depression. JAMA Psychiatry 2023; 80:682–689
42.
Meaidi A, Torp-Pedersen C, Lidegaard Ø, et al: Ectopic pregnancy risk in users of levonorgestrel-releasing intrauterine systems with 52, 19.5, and 13.5 mg of hormone. JAMA 2023; 329:935–937
43.
Elgemark K, Graner S, McTaggart J, et al: The 13.5-mg, 19.5-mg, and 52-mg levonorgestrel-releasing intrauterine systems and risk of ectopic pregnancy. Obstet Gynecol 2022; 140:227–233
44.
Haut ER, Pronovost PJ: Surveillance bias in outcomes reporting. JAMA 2011; 305:2462–2463
45.
Mendle J, Ferrero J, Moore SR, et al: Depression and adolescent sexual activity in romantic and nonromantic relational contexts: a genetically-informative sibling comparison. J Abnorm Psychol 2013; 122:51–63
46.
Fraser IS, Langham S, Uhl-Hochgraeber K: Health-related quality of life and economic burden of abnormal uterine bleeding. Expert Rev Obstet Gynecol 2009; 4:179–189
47.
Fraser IS, Mansour D, Breymann C, et al: Prevalence of heavy menstrual bleeding and experiences of affected women in a European patient survey. Int J Gynaecol Obstet 2015; 128:196–200
48.
Flores RC, Lara EB, Corral LCQ, et al: Quality of life in women with endometriosis pelvic pain treated with the levonorgestrel-releasing intrauterine system. Open J Obstet Gynecol 2015; 5:167–172
49.
Hurskainen R, Teperi J, Rissanen P, et al: Quality of life and cost-effectiveness of levonorgestrel-releasing intrauterine system versus hysterectomy for treatment of menorrhagia: a randomised trial. Lancet 2001; 357:273–277
Information & Authors
Information
Published In
History
Received: 9 November 2023
Revision received: 13 February 2024
Accepted: 20 March 2024
Published online: 10 July 2024
Keywords
Authors
Author Contributions
Drs. Lidegaard and Frokjaer contributed equally as senior authors.
Competing Interests
Dr. Munk-Olsen has served as a speaker for Lundbeck. Dr. Lidegaard has served as a speaker for, and his institution has received research funding from, Exeltis. Dr. Frokjaer has served as a consultant for Sage Therapeutics and as a speaker for Gedeon-Richter, Janssen-Cilag, and Lundbeck Pharma. The other authors report no financial relationships with commercial interests.
Funding Information
Supported by the Independent Research Fund Denmark (grants 0134-00278B and 7025-00111B).
Metrics & Citations
Metrics
Citations
Export Citations
If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download.
For more information or tips please see 'Downloading to a citation manager' in the Help menu.
There are no citations for this item
View Options
View options
PDF/ePub
View PDF/ePubGet Access
Login options
Already a subscriber? Access your subscription through your login credentials or your institution for full access to this article.
Personal login Institutional Login Open Athens loginNot a subscriber?
PsychiatryOnline subscription options offer access to the DSM-5-TR® library, books, journals, CME, and patient resources. This all-in-one virtual library provides psychiatrists and mental health professionals with key resources for diagnosis, treatment, research, and professional development.
Need more help? PsychiatryOnline Customer Service may be reached by emailing [email protected] or by calling 800-368-5777 (in the U.S.) or 703-907-7322 (outside the U.S.).