Transdermal estradiol for postpartum depression: results from a pilot randomized, double-blind, placebo-controlled study.

The study took place between June 2004 and August 2012. Inclusion criteria were as follows: 1. Age 20–45 years; 2. Delivery of healthy baby within 6 months by uncomplicated vaginal delivery or cesarean section; 3. Onset of depressive symptoms within 3 months of delivery; 4. Current episode meeting criteria for minor or major depression as determined by the Schedule for Affective Disorders and Schizophrenia-Lifetime (SADS-L) ( Endicott and Spitzer 1978 ) and the Structured Clinical Interview for DSM-IV (SCID) ( American Psychiatric Association 2000 ); and 5. Meeting minimum symptom severity criteria: scores ≥ 10 on either the Beck Depression Inventory (BDI) ( Beck et al. 1996 ) or the Center for Epidemiological Studies-Depression (CES-D) ( Radloff 1977 ) scales on ≥ 2 clinic visits during the screening period, as well as a score ≥ 10 on the Hamilton Depression Rating Scale (HAM-D) ( Hamilton 1960 ). These severity criteria were chosen to confirm a stable symptomatic baseline prior to enrollment, consistent with entry criteria of other studies on reproductive mood disorders in our institution. Multiple scales were used to ensure that women met severity criteria based on a clinician-rated measure (HAM-D) and at least one self-report measure (BDI or CES-D). Low severity thresholds corresponding to mild depression were chosen to include lower acuity PPD cases likely to present and be managed in the outpatient setting. Our symptom onset criteria of within 3 months postpartum are broader than the DSM-V criterion of 4 weeks, consistent with clinical experience and studies suggesting that a more inclusive definition of PPD may be warranted ( Forty et al. 2006 ; Deligiannidis et al. 2013 ; Schiller et al. 2015 ). Exclusion criteria were as follows: severe depression (psychotic symptoms, suicidal ideation, or requiring immediate treatment), current antidepressant use, history of psychiatric illness within 2 years, history of mania or psychosis, and medical illness or contraindications to estradiol (e.g., history of thromboembolic disease, recurrent migraines, breast or uterine cancer, endometriosis). Women were recruited via physician referral or local advertisements. All women were assessed during a screening phase of at least 2 weeks. All women provided written informed consent. Women were reimbursed for travel and subsistence consistent with NIH guidelines, but were not be paid for their participation in this treatment trial. The study was terminated prior to completion due to closure of our institution’s pharmacy and consequent unavailability of study medications. This is a double-blind, placebo-controlled trial to study the feasibility and efficacy of TE monotherapy for outpatient PPD management. After a 2-week screening period, women were enrolled by a study coordinator and randomized to 6 weeks of 17β-estradiol skin patch (100 mcg/day, Alora ® , Allergan, Ireland, before September 2008; Vivelle-Dot ® , Novartis, Switzerland after September 2008) or placebo patch. Prior to enrollment, each woman was informed of the possibility of menstrual bleeding during this study. TE and matched placebo patch formulations were switched in September 2008 due to changes in manufacturer supply, but dose and administration schedule were unchanged. Women were instructed to replace patches twice weekly on both formulations. Patch compliance was assessed by clinician follow-up and measurement of plasma estradiol levels. An independent pharmacist supervised the generation of randomization sequences from a computer program, and randomization schedule was stratified by breastfeeding status in blocks of 4. Participants, principal investigator, care providers, and rating clinician were blinded to the treatment group assignment. During the 6-week treatment phase, both groups completed weekly symptom ratings. At the end of the treatment period, women in the TE group who did not resume menses received an additional week of estradiol with oral medroxyprogesterone tablets (5 mg/day, Provera ® , Pfizer, New York City) to induce a withdrawal bleed for endometrial protection, while the placebo group received an additional week of placebo patches and tablets instead of TE and medroxyprogesterone. Following study completion, women responding to estradiol who requested continued treatment or whose clinical status indicated treatment were offered the option to continue TE for a maximum of 8 weeks before referral to a community physician. Symptomatic women were offered standard therapy with an SSRI and referred as clinically indicated to a community provider. Women completed weekly symptom ratings, first at week 0 (before treatment) and then every week during the 6-week treatment phase, on three scales: BDI ( Beck et al. 1996 ), HAM-D ( Hamilton 1960 ), and Edinburgh Postnatal Depression Scale (EPDS) ( Cox et al. 1987 ). The 21-item BDI is scored on a 0–63 scale, the 17-item HAM-D on 0–54, and the 10-item EPDS on 0–30. The BDI and EPDS are based on self-report, while the HAM-D is clinician-rated and administered by a trained clinician. The rating clinician was not involved in medical follow-up of participants to preserve the clinician-blind. Medical follow-up of participants, including evaluation of patch compliance, patch-related concerns, and menstrual symptoms, was performed by a separate clinician. All scales have been validated for PPD in multiple studies ( Appleby et al. 1997 ; Beck and Gable 2001 ; Wisner et al. 2006 ; Yonkers et al. 2008 ; Tandon et al. 2012 ). Primary outcomes were treatment response (> 50% decrease from baseline BDI) and remission (BDI < 10) at 6 weeks. BDI was chosen for its extensive validation and sensitivity to clinical change ( Beck et al. 1996 ; Richter et al. 1998 ), its feasible implementation as a self-report measure in the outpatient setting ( Zimmerman et al. 2006 ), and to allow results to be comparable with current and prior studies at our institution in other reproductive mood disorders utilizing BDI to monitor symptoms in response to estradiol exposure and/or withdrawal. Response and remission criteria were adapted from prior studies defining optimal cutoffs ( Beck et al. 1988 ; Riedel et al. 2010 ; Reeves et al. 2012 ). In secondary analysis, ratings on all scales (BDI, HAM-D, EPDS) were modeled and analyzed for between-group differences at weeks 3 and 6. These time points were chosen based on a preceding trial showing a response after 1–2 months ( Gregoire et al. 1996 ), and a small open-label trial and case series suggesting a response after 1–3 weeks ( Ahokas et al. 1998 , 2001 ). Weekly plasma estradiol was measured to monitor the effect of TE on plasma concentrations. Weekly plasma progesterone levels were measured to monitor ovulation status; progesterone ≥ 2 ng/ml at any point was interpreted as resumption of ovulatory function ( Israel et al. 1972 ). Estradiol and progesterone plasma levels were measured using electrochemiluminescence immunoassay on the Roche Cobas e601 analyzer (Roche Diagnostics, Switzerland), a validated assay method for both estradiol ( Zhang et al. 2019 ) and progesterone ( Patton et al. 2014 ). Estimating an effect size of 0.8, based on prior literature on estrogen’s effects on depressed mood ( Zweifel and O’Brien 1997 ), 15–20 patients were required in each treatment cell to achieve 80% power at a significance level of 0.05 to detect a significant effect of treatment. Anticipating an attrition rate of 10–15%, we aimed to enroll 22 women per treatment group. An intention-to-treat analysis was performed using R version 3.4.1 ( R Core Team 2017 ). Data from all women for whom post-randomization follow-up data existed were included ( Fig. 1 ). In primary outcome analysis, response and remission rates were compared with Fisher’s exact test. In secondary analysis, BDI, HAM-D, and EPDS scores were used to fit a linear mixed-effects (LME) model with treatment group, time, interaction of group and time, ovulation status, and baseline score as fixed effects, and subjects as a random effect using the “lme4” R package ( Bates et al. 2015 ). Because ovulation status correlated closely with breastfeeding as expected in lactational amenorrhea ( Supplemental Table 1 ), breastfeeding status was not included. An LME model for estradiol levels was fitted similarly, with an additional fixed effect for the interaction of treatment with ovulatory status based on exploratory plotting that suggested an effect of this interaction ( Supplemental Fig. 1 ). LME models were analyzed using the least-square means (“lsmeans”) R package ( Lenth 2016 ), and multiple comparisons adjustment in week-by-week analyses was performed using the Bonferroni method. Significance is reported at a level of 0.05.

Fourteen women were recruited before discontinued access to study medications resulted in study termination. One woman from the placebo group dropped out before receiving treatment. One woman from the TE group dropped out before her first follow-up. The remaining 12 women were included in the analysis ( Fig. 1 ). Five women (42%, 3 TE, 2 placebo) attended all 6 post-randomization visits. Ten women (83%, 5 TE, 5 placebo) attended at least 4 visits. There were no significant baseline differences between treatment groups in demographics (age, ethnicity, marital status), body mass index (BMI), obstetrical history, psychiatric history, smoking status, breastfeeding status, estradiol levels, or BDI, HAM-D, or EPDS scores ( Table 1 ). There was also no significant difference in the proportion of women reporting menstrual bleeding during the study period: 3 women in TE group and 1 woman in the placebo group. The TE group had more women with a prior PPD history (2 vs. 0), more women with a prior history of major depression (3 vs. 2), and fewer breastfeeding women (2 vs. 4). Of the two women in the TE group with a prior PPD history, one woman also had a history of major depression. Four women in the TE group had resumed ovulatory function based on progesterone measurements, compared to 3 women in the placebo group ( Supplemental Fig. 1 ). Three women from the placebo group dropped out after weeks 3, 4, and 5, respectively, due to severe depression requiring community provider referral and conventional antidepressant therapy. There were no drop-outs in the TE group beyond the first follow-up visit. No participants reported any issues with patch adhesion. Response and remission rates at week 6 are reported in Table 2 . The TE group had non-significantly higher response and remission rates ( p = 0.24). Interpretation is limited by 3 placebo group drop-outs, as discussed above; for these women, BDI scores from their last visit were carried forward. In LME modeling, there were significant main effects of time on BDI ( p = 0.001), HAM-D ( p = 0.006), and EPDS scores ( p = 0.008). The main effect of treatment group was not significant for BDI ( p = 0.112), HAM-D ( p = 0.082), and EPDS ( p = 0.328) scores. The interaction of group and time were also not significant for any scale ( Table 3 ). Modeled between-group differences at weeks 3 and 6 showed that women receiving TE compared to those receiving placebo were 9.2 points lower on the BDI scale at 3 weeks (95%CI − 19.5 to + 1.0, p = 0.074), and 10.5 points lower at 6 weeks (95%CI − 21.0–0.0, p = 0.049). HAM-D scores were 8.5 points lower at 3 weeks (95%CI − 16.5 to − 0.4, p = 0.040), and 9.6 points lower at 6 weeks (95%CI − 18.0 to − 1.2, p = 0.026). These differences were no longer significant following multiple comparisons correction. The model showed no significant difference in EPDS scores at any week ( Table 4 and Fig. 2 ). Mean baseline plasma estradiol levels for TE and placebo groups were 64.8 ± 30.1 pg/ml and 83.6 ± 55.8 pg/ml, respectively. During the treatment period, mean estradiol levels were 99.6 ± 51.7 pg/ml for the TE group and 98.0 ± 84.4 pg/ml for the placebo group. Mean plasma progesterone levels for TE and placebo groups were 2.6 ± 4.1 ng/ml and 4.9 ± 7.4 ng/ml, respectively, at baseline, and 1.4 ± 2.9 ng/ml and 2.7 ± 5.0 ng/ml during treatment. No comparisons between baseline and treatment period levels or between TE and placebo groups showed a significant difference in hormone levels. In LME modeling of estradiol levels, there was no significant main effect of TE treatment and no significant interactions between treatment group and time or ovulatory status ( Supplemental Table 2 ). No significant differences were found in plasma estradiol at any time, though exploratory analysis suggested a possible interaction of treatment and ovulatory status ( Fig. 3 , Supplemental Fig. 2 ). Modeled levels of plasma estradiol are reported in Supplemental Table 3 . One woman randomized to TE reported a local skin rash that resolved following discontinuation. No serious adverse events were reported.

In this pilot study, TE (100 mcg/day) may be a feasible option for outpatient PPD management, with preliminary evidence (albeit not significant compared with placebo) suggesting efficacy. While the primary outcome of this study was not met, analysis of secondary outcomes of our limited sample suggests that TE monotherapy may significantly reduce mood symptoms in PPD compared to placebo as early as 3 weeks, and may not directly depend on increased plasma estradiol. More and larger studies are needed to further investigate these claims. If effective, TE would broaden the range of treatment options available to women with PPD.

In the last two decades, postpartum hormone fluctuations have been implicated in PPD, with estradiol emerging as one promising treatment option ( Gregoire et al. 1996 ; Dennis et al. 2008 ; Moses-Kolko et al. 2009 ). In this small pilot study, TE was not associated with a significant difference in the primary outcome, treatment response, and symptom remission according to BDI scores. Acknowledging important limitations due to under-enrollment and participant drop-out, this study provides preliminary evidence (based on secondary outcomes) to support the feasibility and potential efficacy of TE for outpatient treatment of PPD. Albeit in a small sample, we observed no drop-outs beyond the first follow-up or any serious adverse events in the TE group. Furthermore, on secondary analysis, we observed significantly reduced mood symptoms as early as 3 weeks compared to placebo, though these differences were no longer significant following multiple comparisons correction. These potentially significant effects were seen on BDI and HAM-D scores, but not EPDS scores, perhaps reflecting different sensitivities of these scales in the setting of an underpowered study ( Harris et al. 1989 ; Lee et al. 2001 ; Beck and Gable 2001 ; Putnam et al. 2017 ). There are three preceding trials of estradiol for PPD. Gregoire et al. randomized women to TE (200 mcg/day, n = 34) or placebo ( n = 27), assessed monthly EPDS ratings, and found significantly reduced scores after 1 month, with an overall effect of − 4.4 points (1996). Comparably, our results show (albeit insignificantly) an effect between − 1.3 and − 7.7 EPDS points. We also find that a lower dose (100 mcg/day) was sufficient to provide benefit as early as 3 weeks—a dosage and response time consistent with that of TE for perimenopausal depression, another disorder associated with estrogen withdrawal ( Schmidt et al. 2000 ; Soares et al. 2001 ). An open-label study by Ahokas et al. treated 23 women with PPD and estrogen deficiency (estradiol < 200 pmol/l) with sublingual 17β-estradiol and found a significant improvement in Montgomery-Asberg Depression Rating Scale (MADRS) scores after 1 week (2001). While these women were recruited based on clinically suspected estrogen deficiency, estrogen deficiency is neither essential nor common in PPD despite relative ovarian quiescence in the weeks or even months following delivery ( Wisner and Stowe 1997 ; Schiller et al. 2015 ). Furthermore, while both trials included women taking antidepressants (approximately half of Gregoire et al. ‘s participants), our study provides evidence for TE as a monotherapy. Recently, a pilot study by Wisner et al. compared TE (50–200 mcg/day), sertraline (50–200 mg/day), and placebo, with symptom monitoring using the Structured Interview Guide for the Hamilton Depression Rating Scale-Atypical Depression Symptoms (SIGH-ADS) for 8 weeks (2015). The trial was terminated after batch analysis found no between-group differences in serum estradiol. Analysis of SIGH-ADS scores for 85 women (TE n = 26, sertraline n = 30, placebo n = 29) found no benefit of either TE or sertraline over placebo, a finding potentially related to a high placebo response: 58.6%, compared to an expected 35–40% ( Furukawa et al. 2016 ). While our study suggests a therapeutic effect similar to that of Gregoire et al., additional studies are needed to build consensus. The magnitude of the differences in the observed responses to TE in this study were comparable to those reported in the brexanolone trials ( Meltzer-Brody et al. 2018 ), a promising new agent for PPD (with the obvious caveat that the latter trials were conducted in much larger samples within a multi-site RCT, and, therefore, the therapeutic potential of brexanolone and TE cannot be further compared). In two phase 3 trials, brexanolone infusion (60–90 mcg/kg/h) was associated with a 2–6 HAM-D point reduction compared to placebo ( Meltzer-Brody et al. 2018 ), comparable to the 3–10 HAM-D point reduction seen in this study. Similarly, in analysis of secondary outcomes in our data, TE was associated with a 8.5 point decrease in HAM-D scores at 3 weeks, and a 9.6 point decrease at 6 weeks ( p = 0.040 and p = 0.026, respectively), though these results do not survive correction for multiple comparisons. While independent studies are needed, brexanolone is distinguished from TE for its stronger evidence and rapid action, but also its higher cost, requirement for intravenous administration (though oral analogs are under study). Compared to SSRIs, TE may be more acceptable to some women due to stigma and patient preference, and offers a more reliable route of administration when psychological and socioeconomic barriers may pose a challenge to timely, daily use of oral medications. If confirmed efficacious, TE could represent an accessible, cost-effective alternative to SSRIs and brexanolone in women without contraindications, and for whom acute inpatient treatment is not indicated. Interestingly, no effect of TE was seen on plasma estradiol levels. We propose three mechanisms that may explain this: First, endogenous estradiol production may outweigh the effect of TE. Recruited up to 6 months postpartum, 7 women (58%) had resumed menstrual cycling. Assuming an expected increase of 100 μg/ml from studies of TE (100 mcg/day) in postmenopausal women ( Hossain et al. 2003 ), this additional load may not be detected in a small sample when menstrual estradiol levels fluctuate 15–350 pg/ml. Exploratory modeling suggested that TE may measurably increase estradiol in anovulatory women, though our sample was too small to conclude this definitively. Second, exogenous estradiol may not be invariably reflected in peripheral measurements. Estradiol rapidly diffuses across cell membranes, forming intracellular receptor-ligand complexes. Long-term intranuclear retention of estradiol has been shown to be necessary for stimulating growth in uterine tissues ( Clark and Peck 1976 ). Thus, although intracellular sequestration is a key mechanism by which estradiol exerts its biological effects, it may also prevent its reliable detection in peripheral assays ( Clark et al. 1977 ; Bergman et al. 1987 ). Failure to deliver exogenous estradiol, as in cases of patch non-adhesion or poor transdermal absorption (due to adiposity, for example), may also be a factor ( Wisner et al. 2015 ). Third, the exogenous addition of estradiol may be attenuated by compensatory mechanisms. Cytochrome p450 and drug-elimination enzymes, upregulated by preceding pregnancy or exogenous estradiol, increase estradiol clearance ( Hebert et al. 2008 ; Chen et al. 2009 ; Choi et al. 2013 ). TE may also suppress ovarian estradiol synthesis ( Crosignani et al. 1996 ; D’Arpe et al. 2016 ). To reconcile the finding that TE may improve mood symptoms independent of plasma estradiol levels, we consider the hypothesis that TE may exert its therapeutic effect not by increasing estradiol levels but by stabilizing estradiol variability. Delivery of estradiol by TE is constant and pharmacokinetically distinct from the pulsatile pattern of ovarian estradiol secretion ( Rossmanith et al. 1990 ). Exploratory analysis of estradiol levels in this small study supports the estradiol-stabilizing effects of TE, in both ovulating and non-ovulating women. Examining estradiol levels across the study period, ovulating women in the placebo group appeared to have overall greater estradiol variability compared to anovulatory women in the placebo group, and to both ovulatory and anovulatory women in the TE group. This mechanism of action would be consistent with some pathophysiological models of PPD and other reproductive mood disorders, which suggest that mood symptoms are due to an abnormal neural response to fluctuations in reproductive hormones, rather than objectively different levels of circulating hormones ( Rubinow and Schmidt 2006 ; Schiller et al. 2015 , 2016 ). For example, mood symptoms are correlated with estradiol variability in perimenopause ( Gordon et al. 2016 ), a reproductive stage associated with hypervariable ovarian activity and transient hypo- and hyper-estrogenic states ( O’Connor et al. 2009 ; Tepper et al. 2012 ). Similarly, the postpartum period is likely also associated with ovarian irregularity, though this has not been specifically studied. Our observations of serum estradiol levels are consistent with those of Wisner et al., who reported no difference between women receiving TE and placebo (2015). Discerning whether this observation is due to shared technical challenges in drug delivery or a reflection of the complex pharmacobiology of estradiol requires further study. This is the second randomized, placebo-controlled trial to study TE as a monotherapy for PPD and the first to suggest efficacy in this setting, observing a therapeutic benefit with a lower dose (100 mcg/day) and shorter course (3–6 weeks) than previously reported. As estrogens may interfere with lactation and carry the small risk of predisposing thromboembolic disease and estrogen-responsive cancers, minimizing the dose and duration of treatment is warranted. The dose studied in this trial (100 mcg/day) is understood to be safe for short-term therapy, with negligible breastmilk transfer or interference with infant nursing or growth ( Perheentupa et al. 2004 ; Pinheiro et al. 2015 ). Our study was underpowered due to its small sample size, as pharmacy closure resulted in premature study termination and failure to reach our target recruitment of 22 women per group. The TE group had non-significantly higher response and remission rates. Using response and remission rates observed in our study, at least 12 women per group would be needed to detect a significant difference with 80% power. Similarly, EPDS scores were non-significantly decreased. Assuming an effect size of TE on week 6 EPDS scores of 0.75, 28 women per group would be needed ( Hedeker et al. 1999 ). Furthermore, differences in BDI and HAM-D scores at weeks 3 and 6 were not significant following multiple comparisons adjustment. Therapeutic effects may also be obscured by a significant effect of placebo, which may be bolstered by the use of multiple symptom rating scales and frequent, weekly clinician contacts during the follow-up period ( Rutherford and Roose 2013 ; Papakostas et al. 2015 ). Nonetheless, we were unable to demonstrate a significant therapeutic benefit of TE compared with placebo in PPD. Though LME modeling is a favored technique for handling missing values ( Baraldi and Enders 2010 ), missing data—especially for later time points in the placebo group—may skew results when significant effects of time and placebo exist, as often seen in mental health studies ( Leber 2000 ). Finally, more women with a prior PPD history were in the TE group ( n = 2) compared to placebo ( n = 0). Outcomes in primary vs. recurrent PPD have not been studied, but in experimental settings, women with prior PPD respond briskly to hormonal manipulation ( Bloch et al. 2000 ). Prior PPD history may be a proxy for a hormone-responsive phenotype. Of the 2 women with prior PPD receiving TE, one achieved remission while the other failed to respond. Also, while participants were blinded to their treatment group, they could not be blinded to their own pattern of menstrual bleeding. Differences in menstrual bleeding may have impacted the integrity of our double-blinding, though menstrual bleeding was present in both TE and placebo arms at non-significantly different rates (three in TE group vs. one in placebo group). Subgroup analysis based on PPD history, ovulation status, or menstrual status was not meaningful at our sample size. More studies are needed. Considering the logistical and ethical challenges of recruiting postpartum women, future studies should invest in efforts to ensure robust recruitment and follow-up. Larger studies would allow for more sophisticated moderation and mediation analysis of other variables of interest (e.g., blood estradiol levels, psychosocial variables, prior medical and psychiatric history).

Postpartum depression (PPD) affects 8–15% of deliveries ( Cooper et al. 1988 ; Cox et al. 1993 ; Gaynes et al. 2005 ), causing significant maternal morbidity and is thus a major source of disability for new mothers and their families ( O’Hara and McCabe 2013 ). While mild cases are frequently responsive to psychosocial interventions, particularly cognitive behavioral therapy (CBT) and interpersonal therapy (IPT) ( Stamou et al. 2018 ; Johansen et al. 2019 ), pharmacologic treatments are considered in severe or refractory PPD. Studied agents include antidepressants such as tricyclics (TCAs) and selective serotonin receptor inhibitors (SSRIs), with SSRIs as preferred first-line agents for their greater safety and non-inferiority to TCAs ( Wisner et al. 2006 ; Lanza di Scalea et al. 2009 ) and demonstrated efficacy in randomized controlled trials ( Appleby et al. 1997 ; Misri et al. 2004 ; Wisner et al. 2006 ; Yonkers et al. 2008 ; Hantsoo et al. 2014 ). SSRIs are poorly tolerated by some patients and not recommended in cases of bipolar disorder, which may also present similarly to PPD ( Sharma et al. 2008 ; Kim et al. 2014 ). The controversial association between fetal exposure to serotonergic medications and adverse neurodevelopmental outcomes may raise concern for similar risks to the nursing neonates ( Sujan et al. 2017 ; Millard et al. 2017 ). While multiple studies support the safety of SSRIs in breastfeeding, individualized consideration of this theoretical risk is still recommended ( Wisner et al. 1996 ; Berle et al. 2004 ; Weisskopf et al. 2015 ). Finally, stigma and patient preference may limit the acceptability of SSRIs in some women. At least in a subset of PPD cases, postpartum withdrawal of estradiol and progesterone may trigger symptoms in susceptible women ( Bloch et al. 2000 ). Psychiatric vulnerability to hormone fluctuations is hypothesized to be the result of several biological and psychosocial factors associated with pregnancy and postpartum period. While estradiol and progesterone directly affect the brain by modulating the amygdala and prefrontal-limbic connectivity and altering the activity of critical molecules such as brain-derived neurotrophic factor (BDNF) and cAMP response element-binding protein (CREB), they also indirectly affect mood through their effects on physiologic systems that are themselves subject to dynamic changes in the puerperium: thyroid function, lactogenic (oxytocinergic) signaling, hypothalamic-pituitary-adrenal (HPA) activity, and inflammation ( Schiller et al. 2015 ). These biological stressors may also interact with known psychosocial risk factors for PPD (e.g., young age, traumatic life events, poor socioeconomic support, marital dissatisfaction, neonatal morbidity) to produce psychiatric decompensation ( Yim et al. 2015 ). For reasons not entirely understood, women at risk for PPD appear to be sensitive to normal changes in estradiol and progesterone levels that would be tolerated in unaffected women ( Schiller et al. 2015 ). Bloch et al. demonstrated that mood symptoms in women with a history of PPD could be reproduced by clinical hormone manipulations mimicking postpartum hormone withdrawal, an effect not seen in women without such history (2000). Several studies have since investigated hormone-based PPD treatments. Brexanolone, a synthetic version of the progesterone metabolite allopregnanolone, was recently FDA-approved for acute management of PPD ( Kanes et al. 2017 ; Meltzer-Brody et al. 2018 ), though its use may be limited by its high cost and requirement for intravenous administration, and may not be appropriate for patients not requiring hospitalization. Oral formulations of other allopregnanolone analogs are currently under study ( Frieder et al. 2019 ). Estradiol, an inexpensive agent with multiple routes of administration, has also been studied. Of two trials demonstrating efficacy for PPD, one was a randomized trial of adjunctive transdermal estradiol (TE) ( Gregoire et al. 1996 ); the other was an open-label study of sublingual estradiol in women later shown to be estrogen-deficient ( Ahokas et al. 2001 ). Another pilot study of TE monotherapy reporting negative results was underpowered due to early study termination ( Wisner et al. 2015 ). As multiple systematic reviews conclude, estradiol is a promising treatment option in need of greater evidence ( Dennis et al. 2008 ; Moses-Kolko et al. 2009 ).