By Jeffrey T. Jensen, MD, MPH, Editor

Leon Speroff Professor and Vice Chair for Research, Department of Obstetrics and Gynecology, Oregon Health and Science University, Portland

Dr. Jensen reports he is a consultant for Teva Pharmaceuticals, Microchips, and Evo.

SYNOPSIS: A population-based study suggests that hormonal contraception increases the risk of treatment for depression, but bias provides a more likely explanation for the association.

SOURCE: Skovlund CW, Morch LS, Kessing LV, Lidegaard O. Association of hormonal contraception with depression. JAMA Psychiatry 2016 Sept. 28. doi: 10.1001/jamapsychiatry.2016.2387. [Epub ahead of print].

Little objective information is available to assess the relationship between mood disorders and hormonal contraception. In this manuscript, the group headed by Øjvind Lidegaard used the Danish Sex Hormone Register Study, a nationwide cohort study that includes all women living in Denmark, to evaluate the question. The database links health records and diagnoses to prescriptions. To explore the association of hormonal contraception to depression, the authors used the database to identify incident diagnoses of depression in women aged 15 to 34 years from Jan. 1, 2000, to Dec. 31, 2013. Women with a prior history of depression or a medical contraindication to hormonal contraception were excluded, as were those with a history of cancer or infertility. The authors defined hormonal contraception use as current or recent (within six months) use to categorize women who may have discontinued hormonal contraceptive due to depression as users. The reference group of nonusers included never users and former users (discontinued more than six months). The primary outcome was first diagnosis of depression or prescription of an antidepressant medication in users vs. non-users of hormonal contraception.

After exclusions, the study population included 1,061,997 women with 6,832,938 person-years of observation. Of these, 55.5% were current or recent users of hormonal contraception. A total of 133,178 first prescriptions for antidepressants and 23,077 first diagnoses of depression were identified in the cohort. An increased risk of first use of antidepressants was seen in users of all methods of hormonal contraception: combined oral contraceptives (relative risk [RR], 1.23; 95% confidence interval [CI], 1.22-1.25); progestin-only pills (RR, 1.3; 95% CI, 1.27-1.40); transdermal patch (RR, 2.0; 95% CI, 1.76-2.18); vaginal ring (RR, 1.6; 95% CI, 1.55-1.69); etonogestrel implant (RR, 2.1; 95% CI, 2.01-2.24); levonorgestrel intrauterine system (RR, 1.4; 95% CI, 1.31-1.42); and medroxyprogesterone acetate depot (RR, 2.7; 95% CI, 2.45-2.87). They found similar associations with a first diagnosis of depression.

Of concern, the authors also noted a higher risk of first use of an antidepressant associated with hormonal contraception among adolescents (15-19 years) compared to older women. They also found that the overall risk peaked after six months of use (RR, 1.4; 95% CI, 1.34-1.46) but then declined. Long-term use of hormonal contraception did not increase risk of first use of an antidepressant; the authors attributed this to attrition of susceptibles.

To further explore these relationships, the authors performed sensitivity analyses comparing the first use of an antidepressant in the same women before and one year after starting use of hormonal contraceptives. This strengthened the elevated overall risk (RR, 1.6; 95% CI, 1.58-1.69) with the highest risk in younger women (RR, 1.8; 95% CI, 1.72-1.88).

The authors concluded that depression is a potential adverse effect of hormonal contraceptive use.

COMMENTARY

As you ponder these results, consider two interesting demographic observations. First, maternal mortality is on the rise in the United States. The estimated maternal mortality rate for 48 states and the District of Columbia increased from 2000 to 2014.1 In the other direction, and for the first time in my life, we see a sustained decrease in both teen pregnancy and unintended pregnancy that began in 2008.2 The widespread use of hormonal contraception, and in particular highly effective long-acting reversible contraception methods, explains the success in pregnancy reduction, with health disparities a likely contributor to mortality. Highly effective reversible contraception levels the playing field for women and allows them to achieve their full potential. The most effective methods are hormonal.

Since hormonal methods are used widely, any health benefits or risks associated with their use deserve serious attention. Unfortunately, the media disproportionately report studies suggesting potential health risks and remain silent on benefits.3 Thus, you may have seen news stories or fielded patient calls related to this new manuscript by the Lidegaard group. Before accepting the conclusion of a link between hormonal contraception and depression, consider the following.

The Danish National Database is the gift that keeps on providing concerning information about the safety of hormonal contraception. This comprehensive population-based database provides substantial strengths for epidemiologic study, in particular the linkage of prescription data to clinical outcomes. The large number of individuals in the database yields precise estimates of risk with tight confidence intervals irresistible to journals. But, as I have pointed out in prior reviews, bias must be considered as an explanation of any epidemiologic finding. For example, the numerous publications from Lidegaard using this database that suggest an increased risk of venous thrombosis associated with drospirenone and desogestrel-containing combined pills have not been confirmed by true prospective studies.4,5 Major limitations of the Danish Database include the absence of baseline information concerning important confounders, and the inability to control for prescription bias and the healthy user effect.

The current study on depression puts a new twist on this. The overall effect appears sound and free of prescription bias, as the elevated risk occurred with any use of hormonal contraception. Lidegaard boldly asserted that the sensitivity analysis that compared women to themselves before and after a prescription for hormonal contraception “controlled for all potential confounders, which did not change during the observation period and eliminated healthy-user bias.” However, women who accept the use of hormonal contraception likely differ from nonusers. Most importantly, those who use a prescription medication like hormonal contraception may be more likely to use another prescription medicine, like an antidepressant, or to seek medical care for depression. Furthermore, these attitudes may change over time. Even a small tendency toward this could bias the results. Do not be misled by the tight confidence intervals and statistically significant effect for risk estimated in this range of “weak association.”

What is a weak association? A recent publication by Grimes provides an excellent review for those interested in evaluating database studies.6 The most important consideration is strength of association. For cohort studies like the Danish Database, the lower limit of discrimination for a causal effect is a RR of 2.0. Bias provides the most likely explanation for RRs below 2. Only a properly conducted randomized, controlled trial free of bias can adequately assess smaller effects. Few of the associations in this paper cross the threshold so by definition are weak and suspicious. Absence of a clear dose response also is puzzling (e.g., lowest risk for highest dose levonorgestrel combined oral contraceptives [1.3] with higher risks for the levonorgestrel progestin-only pill [1.7] and levonorgestrel intrauterine system [1.4]).

But what about the transdermal patch (RR 2.1), the implant (2.1), and depo medroxyprogesterone acetate (DMPA, 2.7) with risk estimates that cross this threshold? I suspect misclassification bias, as the study design does not account for prior use. Thus, we cannot determine if a diagnosis of depression or prescription correlated with the current method, the prior method, or unrelated factors. Consider that Denmark is ground zero for the oral contraception thrombosis controversy. Danish providers and many patients are strongly influenced by the first-line recommendation promoted heavily by Lidegaard to use only levonorgestrel pills for initial starts, switching only for intolerance. Pills also represent the most common method prescribed to teens. It would not be surprising if users experiencing mood disturbances might switch to a new method (particularly an estrogen-free method) to control symptoms prior to receiving a diagnosis or treatment for depression. Also, the limited use of DMPA, implants, and the patch suggests that users of these methods may differ in many unmeasured characteristics that could influence risk of depression.

To be fair, Lidegaard performed a sensitivity analysis to address the issue of unmeasured confounding. This included only those women who started hormonal contraception during the study period, comparing the incidence of depression (or first prescription for antidepressants) within one year after initiation of hormonal contraception with the incidence rate during the time prior to hormonal contraceptive use. Overall, this strengthened the relationship observed in the cohort. However, the authors’ claim that this “controlled for all potential confounders” and “eliminated healthy-user bias” deserves careful consideration. Although this approach does control for stable physical attributes, it does not take into account psychological and social factors that change over time and influence risk of depression. A fatal flaw of the analysis is the failure to account for the correlation between first use of hormonal contraception and initiation of sexual activity, a potentially emotionally vulnerable time.7 That the sensitivity analysis supports the main analysis simply demonstrates the problem with precision and accuracy. You can hit the same spot on the target every time with great precision, but always miss the bull’s-eye.

Depression represents an important health concern. Many women report mood changes while on hormonal contraception, and all of us can think of patients who improved after discontinuation. Most of us also have prescribed hormonal contraception to women with premenstrual dysphoric disorder with improvement, a position supported by a recent Cochrane review.8 More research obviously is needed. Unfortunately, this study may discourage vulnerable women from using a hormonal contraceptive. I worry about this, as research supports a link between unintended pregnancy and depression.9 As clinicians, we need to provide an open discussion about the risks and benefits of hormonal methods. No one method will meet the needs of all women. In my experience, women experiencing a mood disturbance while using hormonal contraception will discontinue with or without medical advice. A frank discussion of mood disorders in a welcoming clinic environment will help ensure that the patient receives appropriate guidance in selecting an appropriate new method of birth control.

REFERENCES

  1. MacDorman MF, Declercq E, Cabral H, Morton C. Recent increases in the U.S. maternal mortality rate: Disentangling trends from measurement issues. Obstet Gynecol 2016;128:447-455.
  2. Finer LB, Zolna MR. Declines in unintended pregnancy in the United States, 2008-2011. N Engl J Med 2016;374:843-852.
  3. Jones KP. Oral contraception: Current use and attitudes. Contraception 1999;59:17S-20S.
  4. Lidegaard O, Nielsen LH, Skovlund CW, Lokkegaard E. Venous thrombosis in users of non-oral hormonal contraception: Follow-up study, Denmark 2001-10. BMJ 2012;344:e2990.
  5. Dinger J, Bardenheuer K, Heinemann K. Cardiovascular and general safety of a 24-day regimen of drospirenone-containing combined oral contraceptives: Final results from the International Active Surveillance Study of Women Taking Oral Contraceptives. Contraception 2014;89:253-263.
  6. Grimes DA. Epidemiologic research with administrative databases: Red herrings, false alarms and pseudo-epidemics. Hum Reprod 2015;30:1749-1752.
  7. Lara LA, Abdo CH. Age at time of initial sexual intercourse and health of adolescent girls. J Pediatr Adolesc Gynecol 2016;29:417-423.
  8. Halbreich U, Freeman EW, Rapkin AJ, et al. Continuous oral levonorgestrel/ethinyl estradiol for treating premenstrual dysphoric disorder. Contraception 2012;85:19-27.
  9. McCrory C, McNally S. The effect of pregnancy intention on maternal prenatal behaviours and parent and child health: Results of an irish cohort study. Paediatr Perinat Epidemiol 2013;27:208-215.