Effect of Adjunctive Estradiol on Schizophrenia Among Women of Childbearing Age: A Randomized Clinical Trial

Key Points

Question  What is the effect of transdermal estradiol added to antipsychotic treatment in women of childbearing age with schizophrenia?

Findings  In this 8-week, double-blind, randomized, placebo-controlled parallel-group study of 200 women with schizophrenia randomized to a 200-μg estradiol patch or placebo added to antipsychotics, participants receiving transdermal estradiol patches had statistically significant improvements in the primary outcome measure, Positive and Negative Syndrome Scale positive subscale, as well as Positive and Negative Syndrome Scale negative subscale, general psychopathology subscale, and Positive and Negative Syndrome Scale total scores.

Meaning  Estradiol might be an effective treatment for schizophrenia; these results should be viewed in the context of the differences in the natural course of schizophrenia between females and males.

Importance  Several lines of evidence suggest that estradiol influences the course of schizophrenia, and a previous randomized controlled trial demonstrated that transdermal estradiol improved symptoms in female patients of childbearing age. However, many initial positive findings in schizophrenia research are not later replicated.

Objective  To independently replicate the results of the effect of estradiol on schizophrenia in women of childbearing age.

Design, Setting, and Participants  An 8-week randomized, placebo-controlled trial performed in the Republic of Moldova between December 4, 2015, and July 29, 2016, among 200 premenopausal women aged 19 to 46 years with schizophrenia or schizoaffective disorder as defined by the DSM-5.

Intervention  Patients were randomized to receive a 200-μg estradiol patch or placebo patch changed twice a week added to their antipsychotic treatment.

Main Outcomes and Measures  The primary outcome was the positive subscale of the Positive and Negative Syndrome Scale (PANSS; lower scores indicated fewer symptoms and higher scores indicated more symptoms), analyzed with mixed models for repeated measures on an intention-to-treat basis.

Results  A total of 100 women (median age, 38 years; interquartile range, 34-42 years) were randomized to receive an estradiol patch and 100 women (median age, 38 years; interquartile range, 31-41 years) were randomized to receive a placebo patch; the median age at baseline for the entire group of 200 women was 38.0 years (range, 19.5-46.0 years). At baseline, the mean positive PANSS score was 19.6 for both groups combined; at week 8, the mean positive PANSS score was 14.4 in the placebo group and 13.4 in the estradiol group. Compared with placebo, participants receiving add-on estradiol patches had statistically significant improvements in the primary outcome measure, PANSS positive subscale points (–0.94; 95% CI, –1.64 to –0.24; P = .008; effect size = 0.38). Post hoc heterogeneity analyses found that this effect occurred almost entirely in 100 participants older than 38.0 years (46 in placebo group vs 54 in estradiol group; difference, –1.98 points on the PANSS positive subscale; 95% CI, –2.94 to –1.02; P < .001). Younger participants did not benefit from estradiol (difference, 0.08 points on the PANSS positive subscale; 95% CI, –0.91 to 1.07; P = .87). Breast tenderness was more common in the estradiol group (n = 15) than in the placebo group (n = 1) as was weight gain (14 in estradiol group vs 1 in placebo group).

Conclusions and Relevance  The results independently replicate the finding that transdermal estradiol is an effective add-on treatment for women of childbearing age with schizophrenia and extend it, finding improvements in negative symptoms and finding that the effect could be specific to those older than 38 years. The results should be viewed in the context of the differences in the natural course of schizophrenia between females and males.

Trial Registration  ClinicalTrials.gov identifier: NCT03848234

Although antipsychotic drugs are of value in reducing the symptoms of schizophrenia,^1,2 mortality, hospitalizations, and suicide³ as well as treating agitation, hallucinations, and delusions, they do not help negative symptoms or cognitive impairments⁴ and cause significant adverse effects.⁴ New directions in treatment of schizophrenia are urgently needed, but as in other fields of medicine,⁵ many promising new directions for treatment have been reported, but not later replicated.⁶ Well-replicated sex differences in the age of onset and course of schizophrenia indicate that estradiol might influence the course of schizophrenia.⁷ Several randomized clinical trials have been conducted on the administration of oral estradiol to female patients with schizophrenia; most, but not all, trials reported a beneficial effect.⁸ Three studies on estradiol patches in women with schizophrenia have been published. An initial 4-week pilot study in 36 women showed a decrease in positive symptoms using 100-μg estradiol patches⁹; this study was followed by a larger (N = 102) 4-week study showing improvements in positive symptoms and general psychopathology with the use of 100-μg estradiol patches in women with schizophrenia.¹⁰ A larger 8-week study of 183 women of childbearing age administered 100-μg and 200-μg estradiol patches vs placebo and showed significant improvements in positive symptoms (effect size, 0.44), general psychopathology subscale, and total Positive and Negative Syndrome Scale (PANSS) scores, with the higher dose showing greater effects.¹¹ All studies used add-on designs in which estradiol or placebo was added to an antipsychotic drug; estradiol doses above 200 μg have not been tested.

This study aimed to independently replicate the results of the study by Kulkarni et al¹¹ using the most efficacious dose of estradiol (200 μg) in a different country by different investigators with nonprofit funding.

The trial protocol can be found in Supplement 1. The study was originally planned to be performed in Romania and the Republic of Moldova, but approval by the Romanian regulatory authorities was substantially delayed; hence, the study was performed only in the Republic of Moldova (eAppendix 1 in Supplement 2). Two hundred women, either inpatients (≥3 days after admission) or outpatients, were recruited from the Psychiatric Hospital of Chisinau in the Republic of Moldova between December 4, 2015, and July 29, 2016. Inclusion and exclusion criteria were identical to those in the study by Kulkarni et al¹¹: participants had schizophrenia or schizoaffective disorder as defined by the DSM-5, were aged 18 to 45 years, were not postmenopausal, were receiving an antipsychotic medication, and had a total PANSS score of 60 or above (lower scores indicated fewer symptoms and higher scores indicated more symptoms). The sample size calculation was based on the results of Kulkarni et al.¹¹ The study was approved by the Ministry of Health of the Republic of Moldova—National Ethics Committee for clinical research of drugs and new methods of treatment. All individuals provided written informed consent before participating in the study.

Quiz Ref IDThis trial was a randomized, double-blind, placebo-controlled parallel-group study. Study medication (200-μg estradiol patches and identical placebo patches) were purchased, packaged, and centrally randomized (1:1) by Sharp (http://www.sharpservices.com/our-facilities/sharp-clinical-services-wales/). Patches were changed twice weekly for 8 weeks. The randomization list was computer generated and provided by Sharp, and centrally and sequentially numbered containers were provided. All patients, investigators, and raters were blinded. Both study groups continued to receive antipsychotics and other medications they were taking when entering the study.

Based on the study by Kulkarni et al,¹¹ the a priori defined primary outcome measure was the PANSS positive subscale score.¹² Secondary outcome measures included total PANSS and subscale scores, the Clinical Global Impression Scale (CGI)–Severity score, the Brief Assessment of Cognition in Schizophrenia (BACS),¹³ the Montgomery Depression Scale (MADRS), the Simpson-Angus Scale,¹⁴ and the Udvalg for Kliniske Undersogelser Side Effect Rating scale.¹⁵ Clinical outcomes were assessed at baseline and weeks 1, 2, 4, and 8 (end of study).

Quiz Ref IDData analyses were conducted using Stata, version 15 (StataCorp Inc). The analysis code is available upon request. Intention-to-treat analysis of PANSS subscale scores, CGI score, MADRS score, and BACS score used mixed models for repeated measures¹⁶ with visit, treatment, and group-by-visit interaction as fixed effects; an unstructured covariance matrix; and specified restricted maximum likelihood (no separate imputation for missing values was made). As a sensitivity analysis for mixed models for repeated measures modeling assumptions, analysis of covariance with last observation carried forward and observed case (completers) cohorts were performed. A post hoc analysis for heterogeneity using a median split of baseline age, PANSS scores, MADRS scores, antipsychotic medications, and plasma levels of hormones was performed by adding an interaction term in a mixed models for repeated measures regression model and then using a likelihood ratio test.

In addition, the role of plasma estradiol was assessed in relation to age and PANSS score at the end of study. Group differences in adverse events were assessed. All P values were from 2-sided tests, and results were deemed statistically significant at P < .05. No adjustment for multiple testing was made.

A meta-analysis was performed combining the results of the current study together with all previous studies that tested estradiol patches in women with schizophrenia. The protocol is available at https://www.crd.york.ac.uk/prospero/display_record.asp?ID=CRD42017073342.

A mixed (random) model grand mean was computed for estradiol patch vs placebo patch using Comprehensive Meta-Analysis, version 3.0 software (Biostat). Risk-of-bias assessments, investigations for publication bias, and assessments of heterogeneity were also completed.

One hundred participants were randomized to receive an estradiol patch and 100 to receive a placebo patch; all randomized participants were included in the intention-to-treat analysis (mixed models for repeated measures). A total of 12 participants (6.0%; 7 placebo and 5 estradiol) did not complete the study (Figure 1). The median age at baseline was 38.0 years (range, 19.5-46.0 years), and the mean PANSS total score at baseline was 82.3 in the placebo group and 83.4 in the estradiol group (mean baseline PANSS total score for both groups combined, 82.8) (Table 1). Baseline demographic and clinical characteristics were well balanced between the groups (eTables 1-3 in Supplement 2).

Quiz Ref IDParticipants in both groups showed considerable symptomatic improvement. At baseline, the mean PANSS positive subscale score was 19.6 for both groups combined and at week 8, the mean PANSS positive subscale score was 14.4 in the placebo group and 13.4 in the estradiol group. In mixed-model analyses, the estradiol group showed statistically significant improvement in the primary outcome measure, the difference in PANSS positive subscale scores (–0.94; 95% CI, –1.64 to –0.24; P = .008; effect size, 0.38), and in the secondary outcome measures of PANSS total score (difference, –4.10; 95% CI, –6.73 to –1.47; P = .002; effect size, 0.45) as well as the PANSS subscales for negative symptoms and general psychopathology compared with the placebo group (Table 2 and Figure 2). There were also significantly greater improvements with the estradiol patch relative to placebo in the CGI-Severity score and MADRS score. These findings were maintained in the sensitivity analyses using analysis of covariance (Table 2). There were no statistically significant differences in the BACS composite z score or in the Simpson-Angus Scale.

In post hoc analyses, we tested whether the improvement seen in PANSS positive subscale scores in the treatment vs placebo group at week 8 was moderated by a median split of baseline PANSS score, MADRS score, antipsychotic treatment, or age (eAppendix 2, eFigure 1, and eAppendix 3 in Supplement 2). Tests of interaction for week 8 PANSS positive subscale scores revealed evidence of an interaction with baseline age but not with baseline total PANSS score, MADRS scores, or type or dosage of antipsychotic medications.

Quiz Ref IDThe beneficial effect of estradiol vs placebo was evident only in participants older than 38.0 years (n = 100; 46 placebo and 54 estradiol), on PANSS positive subscale scores (effect size, 0.79; difference, –1.98 PANSS points; 95% CI, –2.94 to –1.02; P < .001), total PANSS scores (effect size, 0.83; difference, –7.94 PANSS points; 95% CI, –11.74 to –4.14; P < .001), and PANSS negative subscale, CGI, and MADRS scores (Table 2 and Figure 3; eFigure 2 in Supplement 2). Younger participants did not benefit from estradiol (effect size, 0.05; difference, 0.08 points on the PANSS positive subscale; 95% CI, –0.91 to 1.07; P = .87). We also examined the interaction of continuous age and treatment group and found that a quadratic age interaction best fit the data, with women in the higher age range of the sample (between 35 and 45 years) having better PANSS scores in the estradiol group and poorer PANSS scores if they were in the placebo group (eFigure 3 in Supplement 2). In addition, in the placebo group, older women showed a smaller improvement compared with younger women.

Plasma estradiol levels increased significantly in the estradiol group compared with the placebo group (eTable 4 in Supplement 2). No significant differences were found in week 8 levels of luteinizing hormone, follicle-stimulating hormone, progesterone, or prolactin between the estradiol group and the placebo group (eFigure 5 and eAppendix 5 in Supplement 2).

More women in the estradiol group reported breast discomfort (15 vs 1 in placebo group) and weight gain (14 vs 1 in placebo group) (eTable 5 and eAppendix 4 in Supplement 2). No other adverse event differed significantly between the estradiol and placebo groups.

Four studies that used estradiol patches in schizophrenia were included in the meta-analysis: 3 by Kulkarni et al^9-11 and the current study. The meta-analysis showed an effect size of –0.318 (95% CI, –0.504 to –0.132; P = .001) on the PANSS total scale, favoring estradiol vs placebo (eAppendix 6 and eFigure 7 in Supplement 2). Results were similar across subscales. In the risk-of-bias analysis, it was found that the 2002 study by Kulkarni et al⁹ added a 100-μg group post hoc; those patients were excluded from the quantitative analyses. In the 2008 study by Kulkarni et al,¹⁰ 3 times as many patients in the placebo group had incomplete outcome data, which gave a high risk of attrition bias. Complete results, including further exploration of PANSS subscales, heterogeneity, PRISMA diagram, risk of bias, publication bias, and subgroup analysis, are presented in eAppendices 6 and 7, eTables 7 to 11, eFigures 6 to 14 in Supplement 2.

The current study found significant beneficial effects of transdermal estradiol on total PANSS score, PANSS positive subscale score, PANSS negative subscale score, general psychopathology subscale, and depressive symptoms in women with severe schizophrenia (mean baseline PANSS total score, 82.8) receiving antipsychotics. The estradiol patch effect size of 0.45 for PANSS total score and 0.38 for PANSS positive subscale score is relatively large for augmentation studies in schizophrenia¹⁷ and might be considered somewhat significant from a clinical perspective. The effect of estradiol patches was driven by consistent effects on a large number of symptoms rather than large effects on a few symptoms.

Our study replicates and expands on the results of the 3 previous randomized clinical trials administering estradiol patches to women with schizophrenia, which reported improvements in positive symptoms. The largest and most recent of these studies also reported a numerical improvement in negative symptoms, more pronounced with the 200-μg dose, which did not reach statistical significance.¹¹ Because estradiol significantly improves positive and affective symptoms and had a smaller effect on extrapyramidal symptoms, the changes in negative symptoms might reflect any or all of these, and no inference should be made about primary negative symptoms. Baseline dose or type of antipsychotic did not have a differential effect on response to estradiol. All patients entering the study were already taking antipsychotics, so their baseline PANSS scores already reflected residual symptoms. During the study they continued to receive antipsychotics, many participants were hospitalized, and all received more intense care as part of the study. This treatment, in addition to the placebo effect, might have contributed to the improvement. The addition of estradiol improved symptoms further.

In heterogeneity analyses using a median split of possible moderator variables, we found a strikingly better response in women older than 38.0 years, with an effect size of 0.79 on the primary outcome measure of positive PANSS subscale score, in contrast to estradiol having no significant effect in younger women (≤38.0 years of age), with an effect size of 0.05. We also examined the interaction of continuous age and treatment group and found that a quadratic age interaction best fit the data, with women in the higher age range of the sample (between 35 and 45 years) having better PANSS scores in the estradiol group and poorer PANSS scores if they were in the placebo group (eFigure 3 in Supplement 2), consistent with the finding using the median split cutoff for age. This finding suggests that exogenous estrogen might become effective only when endogenous estrogen starts to diminish or when estrogen receptors become less sensitive in the menopausal transition, consistent with the finding using the median split cutoff for age as an approximation. In addition, in the placebo group, older women showed a smaller improvement compared with younger women. This outcome might be due to either older women having a smaller placebo response compared with younger women or to older women responding less to baseline antipsychotics.

We had not anticipated this age outcome; however, it does fit with the epidemiologic data finding that the onset of schizophrenia is 2 to 4 years later in women than in men and that women have a lower incidence of schizophrenia until ages 35 to 44 years, after which women start to have an increased incidence of schizophrenia.^7,18,19 Moreover, women are more likely to have their first episode of schizophrenia during an estradiol trough in the menstrual cycle.²⁰

We do not have specific data explaining the effect of age on the efficacy of estradiol because plasma estradiol levels measured at the beginning and end of the study did not differ between the older and younger participants. Bergemann et al²¹ assessed symptoms of schizophrenia in correlation with endogenous estrogen. Our results regarding lack of a connection between plasma estradiol levels and improvement in symptoms cannot address the effect of endogenous estrogen because our study administered exogenous estrogen. The same is true for a study reporting a positive correlation between endogenous estradiol levels and cognitive performance, which also included postmenopausal women.²² Other factors related to age and the approach of menopause might contribute to this finding.

The lack of improvement in cognition is not definitive or generalizable; this study was not designed to test improvements in cognition, which was a secondary outcome measure. In addition, many of the participants were uneducated and not used to taking tests.

As expected, patients receiving estradiol reported more weight gain and breast discomfort.¹ It is unlikely that these adverse events caused functional unblinding of the study, as the beneficial effects of estradiol remained after controlling for this possibility. No other adverse event differed significantly between the estradiol and placebo groups. The participants were examined before and after the study by a gynecologist, including gynecological ultrasonography, with no cases of cancer reported. Our study cannot address risk beyond 2 months.

Quiz Ref IDOur study tested the effects of estradiol only on premenopausal women receiving antipsychotics and is not applicable to potential effects of estradiol on older women, men,²³ or as monotherapy. As data were not collected after the estradiol patches were stopped, we do not know how long, if at all, the effects of estradiol remain after cessation of treatment. Performing the study in a single center might introduce site-specific biases, unknown to us, that influence the results. The doses used were relatively low; hence, these results should be replicated in women receiving higher doses of antipsychotics with plasma levels monitored. Our inclusion criteria did not differentiate between patients with acute, chronic, and/or treatment-resistant schizophrenia. Future studies might contemplate addressing this differentiation.

Blood levels of estradiol were measured only by a single sample at the beginning and the end of the study, and women were not queried about their last menstrual period (eFigure 4 in Supplement 2). For these reasons, the lack of association between plasma estradiol levels and efficacy should not be regarded as definitive. A more precise interpretation of the results could be achieved by taking weekly hormone measurements and inquiring about the last menstrual period, allowing calculation of the area under the estradiol curve as moderated by the menstrual cycle during the treatment period, and assessment of smoking status, as cigarette smoking inhibits estradiol synthesis in the brain and lowers circulating estradiol levels.^24,25

The findings of this study are encouraging because they are is a replication of previous seminal studies by Kulkarni et al^9-11 showing significant effects of estradiol patches on women with schizophrenia receiving antipsychotics. Nonreplication of research findings is a common phenomenon in medicine,⁵ causing increasing concern in scientific circles.^26,27 This study, performed by independent investigators in a different country with nonprofit funding, shows strong support for the efficacy of estradiol patches in woman with schizophrenia, which is supported by the results of the meta-analysis. The Stanley Medical Research Institute has funded many independent replication studies of initial studies with positive results on other drugs, most of which failed to replicate previous findings, except for this study.

First, the finding of differential effects by age should be replicated. As the drug-placebo difference in this study increased over time, other studies might administer estradiol patches for a longer period and study dose-response relationships, the addition of progestin, and the effect of estradiol patches as monotherapy on negative symptoms, in postmenopausal women, for other indications, and in men (if the issue of feminization can be addressed adequately).

Future studies must also deal with the safety of administration of estradiol to premenopausal women. As estradiol increases the risk of thromboembolic events, which are rare but serious complications,² the potential use of estradiol should be approached cautiously in women with preexisting cardiovascular risks. The possibility of an increase in the risk of breast cancer and related mortality due to exogenous estrogen has been the subject of extensive studies in the context of oral contraceptive use,^3,4 which reported no overall association between oral contraceptive use and risk of breast cancer or a small increase in current users, no increase in breast cancer mortality, and decreased mortality from other forms of cancer^28,29; however, as oral contraceptives contain progestins, generalization is limited to estrogen without progestins. The levels of estradiol in the blood of the participants in this study receiving estradiol were increased, and increased levels of endogenous estradiol are associated with a slightly increased risk of breast cancer.³⁰ Premenopausal breast cancers are relatively rare but likely to express estrogen receptors,³⁰ suggesting that estradiol administration should be avoided in women with a past or present diagnosis of breast cancer. Future research might also consider the role of estrogen receptors, as suggested by our preliminary genetic analysis in eAppendix 8, eTable 12, and eFigure 15 in Supplement 2.

The current study found significant beneficial effects of adjunctive transdermal estradiol on a range of symptoms in women with severe schizophrenia, particularly in those older than 38 years. The accompanying meta-analysis of 4 randomized clinical trials suggests that an estradiol patch added to antipsychotics is effective at improving schizophrenia symptoms in women. Although promising, the results of the meta-analysis are not robust enough to support clinical application at this point. US Food and Drug Administration–level studies should be performed and submitted for approval with schizophrenia as an indication.

Accepted for Publication: May 14, 2019.

Corresponding Author: Mark Weiser, MD, Stanley Medical Research Institute, 10605 Concord St, Ste 206, Kensington, MD 20895 (weiserm@stanleyresearch.org).

Published Online: July 31, 2019. doi:10.1001/jamapsychiatry.2019.1842

Author Contributions: Dr Weiser had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Weiser, SanGiovanni, Davidson, Burshtein, Gonen, Slobozean Pavalache, Ryan, Davis.

Acquisition, analysis, or interpretation of data: Weiser, Levi, Zamora, Biegon, SanGiovanni, Radu, Slobozean Pavalache, Nastas, Hemi, Ryan, Davis.

Drafting of the manuscript: Weiser, Levi, Zamora, Biegon, Davidson, Burshtein, Gonen, Slobozean Pavalache, Hemi, Davis.

Critical revision of the manuscript for important intellectual content: Weiser, Zamora, Biegon, SanGiovanni, Davidson, Radu, Nastas, Ryan, Davis.

Statistical analysis: Levi, Zamora, Biegon, SanGiovanni, Ryan, Davis.

Obtained funding: Weiser.

Administrative, technical, or material support: Weiser, SanGiovanni, Davidson, Burshtein, Gonen, Radu, Hemi, Davis.

Supervision: Weiser, Davidson, Gonen, Radu, Nastas, Davis.

Conflict of Interest Disclosures: Dr Davidson reported that, at the time of the trial, he owned stocks in the site management organization that supported the site conducting the trial. No other disclosures were reported.

Funding/Support: This study was funded by the Stanley Medical Research Institute (http://www.stanleyresearch.org/), a nonprofit organization based in Maryland supporting research on the causes of and treatments for schizophrenia and bipolar disorder.

Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Sharing Statement: See Supplement 3.

Additional Contributions: Mirit Hanochi, MSc, Institute of Endocrinology, Sheba Medical Center, assisted in measuring the levels of hormones in the sera. Mark Horowitz, MS, National Institute on Aging, National Institutes of Health, assisted with statistical programming. They were not compensated for their contributions.