Long-Term Effects on Cognitive Function of Postmenopausal Hormone Therapy Prescribed to Women Aged 50 to 55 Years

Importance  Postmenopausal hormone therapy with conjugated equine estrogens (CEEs) may adversely affect older women’s cognitive function. It is not known whether this extends to younger women.

Objective  To test whether prescribing CEE-based hormone therapy to postmenopausal women aged 50 to 55 years has longer-term effects on cognitive function.

Design  Trained, masked staff assessed participants with an annual telephone-administered cognitive battery that included measures of global and domain-specific cognitive functions. Cognitive testing was conducted an average of 7.2 years after the trials ended, when women had a mean age of 67.2 years, and repeated 1 year later. Enrollment occurred from 1996 through 1999.

Setting  Forty academic research centers.

Participants  The study population comprised 1326 postmenopausal women, who had begun treatment in 2 randomized placebo-controlled clinical trials of hormone therapy when aged 50 to 55 years.

Intervention  The clinical trials in which the women had participated had compared 0.625 mg CEE with or without 2.5 mg medroxyprogesterone acetate over a mean of 7.0 years.

Main Outcomes and Measures  The primary outcome was global cognitive function. Secondary outcomes were verbal memory, attention, executive function, verbal fluency, and working memory.

Results  Global cognitive function scores from women who had been assigned to CEE-based therapies were similar to those from women assigned to placebo: mean (95% CI) intervention effect of 0.02 (−0.08 to 0.12) standard deviation units (P = .66). Similarly, no overall differences were found for any individual cognitive domain (all P > .15). Prespecified subgroup analyses found some evidence that CEE-based therapies may have adversely affected verbal fluency among women who had prior hysterectomy or prior use of hormone therapy: mean treatment effects of −0.17 (−0.33 to −0.02) and −0.25 (−0.42 to −0.08), respectively; however, this may be a chance finding.

Conclusions and Relevance  CEE-based therapies produced no overall sustained benefit or risk to cognitive function when administered to postmenopausal women aged 50 to 55 years. We are not able to address whether initiating hormone therapy during menopause and maintaining therapy until any symptoms are passed affects cognitive function, either in the short or longer term.

Trial Registration  clinicaltrials.gov Identifier: NCT01124773

The Women’s Health Initiative Memory Study (WHIMS) demonstrated that postmenopausal hormone therapy with conjugated equine estrogens (CEEs), when prescribed to women 65 years and older, produced deficits in global and domain-specific cognitive functioning.^1-3 On average, these were small; however, deficits persisted for years after cessation of hormone therapy.³ They occurred in conjunction with decreases in brain volume linked to increased incidence of cognitive impairment.^4,5

In contrast, observational and cohort studies and considerable basic science research suggest that there may be a “window of opportunity,” perhaps coincident with the loss of ovarian function during menopause, when hormone therapy may promote or preserve brain health.^6-9 Meta-analyses of clinical trials and systematic literature reviews do not find consistent evidence of benefit^10,11; however, the window-of-opportunity hypothesis remains of great interest and public health importance because hormone therapy continues to be widely prescribed for managing menopausal symptoms.¹²

The Women’s Health Initiative Memory Study of Younger Women (WHIMSY) tested whether prescribing CEE-based hormone therapy to postmenopausal women aged 50 to 55 years has longer-term effects on cognitive function. We present its primary findings.

The Women’s Health Initiative (WHI) included 2 parallel placebo-controlled trials of CEE-based regimens.¹³ Volunteers were postmenopausal and appropriate candidates to receive these medications. Women currently using hormone therapy were eligible after a 3-month washout period. Enrollment occurred from 1996 through 1999 at 40 academic research centers. Active therapies were 0.625-mg/d CEE in women who had undergone hysterectomy and 0.625-g/d CEE combined with 2.5-mg/d medroxyprogesterone acetate (MPA) in women with a uterus and were compared with matching placebos. The trial among women who had not undergone hysterectomy (CEE+MPA) was terminated in July 2002¹⁴; the trial among women who had undergone hysterectomy (CEE-alone) was terminated in February 2004.¹⁵ Administration of study therapies was stopped at these times. Group assignment was revealed, but follow-up continued.

The WHIMSY volunteers had begun screening for WHI enrollment when aged 50 to 54 years (and initiated their assigned WHI treatment when aged 50 to 55 years), were currently followed by the WHI, and had hearing acuity adequate for telephone interviews. All provided written informed consent; protocols were approved by local institutional review boards.

Trained, masked staff collected cognitive data with telephone-administered assessments that have been shown to be valid.¹⁶ The primary outcome was global cognitive function, assessed with the Telephone Interview for Cognitive Status–modified (TICS-m), a 14-question test with scores ranging from 0 to 50.¹⁷ Its selection paralleled use of global cognitive function as the primary cognitive outcome in WHIMS.¹⁸ Secondary outcomes included the following:

• •Immediate and delayed verbal memory, assessed with the 12-point East Boston Memory Test¹⁹;

• •Attention and executive function, assessed with the Oral Trail Making Test, a modification of the Trail Making Test,²⁰ a validated measure of attention (Part A) and executive function (Part B),²¹ scored as time in seconds;

• •Verbal fluency, assessed with the Verbal Fluency–Animals test, scored as the number of unique spontaneously named animals in 1 minute²²; and

• •Working memory, assessed with the Digit Span subtest of the Wechsler Adult Intelligence Scale–Revised, which measures attention and working memory with the longest correct span length recalled for digits forward and backward.²³

The WHI had collected baseline demographic, lifestyle, and clinical data related to the risk of cognitive impairment via self-report and standardized assessments.¹³ Adherence was computed as the mean proportion of assigned study medication use, based on pill counts. Years of on-trial exposure were computed by summing each woman’s adherence rates across years of trial follow-up.

Cognitive measures from 2 annual assessments were analyzed as repeated data to estimate women’s average level of cognitive function. General linear models with covariate adjustment were used to assess mean differences between intervention groups,²⁴ as prespecified in the protocol. Results from generalized estimating equations were similar and are not reported. To facilitate comparisons among tests, measures were normalized by dividing the difference between individual scores and the cohort-wide mean by the scores’ standard deviation (SD) and ordered so that higher values reflected better performance. A composite measure was computed by averaging normalized scores across tests. Prespecified subgroup analyses were performed using tests of interactions. Primary analyses followed intention-to-treat, with women grouped according to treatment assignment.

A total of 1326 women enrolled in WHIMSY (of 1372 potentially eligible women who agreed to be contacted). They had a mean (range) duration of follow-up of 7.0 (3.9-10.1) years during the WHI trials, which ended 7.2 (5.4-10.1) years prior to WHIMSY enrollment (Figure). The mean (range) age of participants was 67.2 (62.9-73.5) years at their first assessment. The second assessment was conducted with the participation of 1168 women (88.1%) with mean age of 68.1 years. Times between assessments for treatment groups were similar (P = .64).

There was reasonable balance in important potential confounders at WHI enrollment between women who had been assigned active vs placebo therapy (all P > .05) (Table 1). Markers of exposure to therapy, based on average pill counts and the sum of pill counts across trial follow-up, were also similar between arms (P > .20).

Table 2 presents mean cognitive function scores averaged over time, with adjustment for age and visit year. For TICS-m, there was essentially no difference in the mean scores between women who had been assigned to active vs placebo therapy (P = .66). This finding was consistent for both CEE+MPA and CEE-alone therapies (P = .23). Similarly, there were no overall treatment differences for any other measure of cognitive function, including the composite score. This held for CEE+MPA and CEE-alone therapies and for all cognitive measures, except verbal fluency. Assignment to CEE-alone therapy was associated with 0.17 SD worse mean scores on verbal fluency with a 95% confidence interval that excluded zero (−0.33 to −0.02); CEE+MPA was associated with 0.07 SD better mean scores on this test; however, its confidence interval included zero (−0.06 to 0.19). Covariate adjustment for the risk factors for cognitive impairment in Table 1 did not materially alter findings (data not shown).

Adherence and overall exposure were weakly correlated with higher executive function scores (partial r = 0.06, P = .003; r = 0.05, P = .02) but had little correlation with scores from any other domains or the composite score. Adherence and overall exposure were not related to the size of the treatment effect for any measure of cognitive function, based on tests of interaction (all P > .30).

The WHIMSY protocol prespecified 3 subgroup analyses to compare treatment effects: women were grouped by assignment to unopposed or opposed CEE therapy (ie, hysterectomy status), self-reported age at last menstrual period, and prior use of hormone therapy. Table 2 describes the subgroup analyses related to the type of CEE regimen; Table 3 summarizes the other 2 analyses. There was little evidence of differential effects for any measure of cognitive function, with 1 exception. For verbal fluency, worse treatment-related performance was seen among women reporting prior hormone therapy use that had ceased before WHI enrollment. Prior hormone therapy use was associated with longer time since last menstrual period (P < .001): compared with nonusers, these times averaged 2.1 years longer for prior users and 0.2 years longer for current users. Because prior use of hormone therapy more often occurred among women who had undergone hysterectomy, we fitted a model that included treatment interactions with both hysterectomy status and prior use of hormone therapy. Both interactions were independently statistically significant: women reporting prior hormone therapy use (interaction P = .01) and those who had undergone hysterectomy (interaction P = .03) appeared to have treatment-attributable deficits in verbal fluency that were not apparent, on average, in other women.

Among women assigned to hormone therapy during the WHI, 28 (4.0%) reported use at some time during posttrial follow-up, compared with 24 (3.8%) of those who had been assigned to placebo (P = .82). Posttrial use of hormone therapy had no associations with any cognitive function measure (all P > .18).

Power projections for WHIMSY were based on the WHIMS Modified Mini–Mental State Exam global cognitive scores.² The recruitment goal of 2240 women was projected to provide 91% power to detect a mean difference of 0.5 units in this test across 2 examinations, which corresponds to 0.10 to 0.15 SDs for these test scores as collected at baseline in WHIMS. Ultimately, WHIMSY fell short of this recruitment goal, enrolling 1326 women. Post hoc power projections based on observed data yielded 80% (90%) power to detect a mean difference of 0.15 (0.18) SDs, which translates to 0.65 (0.75) TICS-m units.

In a large heterogeneous cohort of postmenopausal women aged 50 to 55 years, WHIMSY tested whether random assignment to a mean 7-year prescription of CEE therapies produced long-term cognitive benefits or deficits compared with placebo. For the primary outcome of global cognitive function, and for specific cognitive domains and a composite of individual tests, no evidence for overall benefit or harm was found. There was some evidence that assignment to hormone therapy was associated with relatively poorer performance on verbal fluency among prespecified subgroups of prior hysterectomy or prior use of hormone therapy; however, type I error was not controlled for across the several domain-specific measures. There was also no evidence for differential treatment effects related to on-trial adherence or years of exposure. The original balance between treatment groups afforded by randomization did not appear to be eroded. Although WHIMSY fell short of its recruitment goal, it maintained adequate power to detect the relatively small mean differences targeted during its design.

The results of WHIMS² showed that prescribing 4 to 5 years of CEE-based therapy to women older than 65 years produced a mean (SE) relative decrement of 0.07 (0.03) SDs in global cognitive function, as assessed with the Modified Mini–Mental State Examination. In the 2304 of its participants who enrolled in the Women’s Health Initiative Study of Cognitive Aging (WHISCA), this on-trial relative deficit was maintained during a mean (SD) of 2.4 (1.1) and 4.0 (1.3) years after the termination of the WHI CEE-alone and CEE+MPA trials, respectively, averaging 0.07 (0.03) SDs during posttrial follow-up.³ The endurance of this effect on global cognitive function supports the choice of the TICS-m as the primary outcome measure for WHIMSY.

The WHISCA data revealed modest decrements in other domains that WHIMSY assessed. A test of verbal memory (California Verbal Learning Test²⁵) had a mean (95% CI) decrement of 0.039 (−0.028 to 0.106) on-trial and 0.013 (−0.056 to 0.082) posttrial SDs, neither statistically significant. A test of attention and working memory (Digit Span forward and backward) had mean (95% CI) decrements of 0.064 (−0.009 to 0.146) on-trial and 0.039 (−0.034 to 0.112) posttrial SDs, also not significant. A test of semantic verbal fluency similar to the measure used in WHIMSY had a larger on-trial mean (95% CI) decrement of 0.083 (0.016-0.150) SDs but little posttrial decrement: 0.006 (−0.063 to 0.075) SDs. For each of these domains, posttrial relative decrements were smaller than on-trial decrements, only reaching nominal significance for verbal fluency (P = .009). Because WHIMSY had fewer participants and less follow-up, it cannot rule out deficits (or benefits) as small as in WHISCA.

In general, the absence of differences in cognitive function between women assigned to active vs placebo therapy was consistent between CEE-alone and CEE+MPA regimens and for subgroups based on time since last menstrual period or prior hormone therapy use. The 1 exception was verbal fluency, which seemed to be adversely affected among women assigned to CEE-alone therapy compared with CEE+MPA therapy and among women with prior (but not current) hormone therapy use compared with no prior use. WHISCA also found that posttrial differences between women who had been assigned to active vs placebo therapy in the CEE-alone and CEE+MPA trials were similar for global cognitive function, verbal memory, and attention and working memory. Similar to WHIMSY, it also found marginal differences in the posttrial effects of CEE-alone vs CEE+MPA therapy on verbal fluency: women assigned to CEE-alone therapy had mean (SE) scores 0.092 (0.060) SDs worse than those assigned to placebo, whereas women who had been assigned to CEE+MPA therapy had mean (SE) scores 0.039 (0.044) SDs better than those assigned to placebo (interaction P = .08).³ Although the magnitudes of these possible treatment-related differences in verbal fluency are small, the similarity in the trends across the trials raises the possibility that CEE-alone therapy may be associated with small longer-term adverse effects on verbal fluency. However, this finding could have resulted by chance.

Others have found verbal fluency to be improved,²⁶ unchanged,^27,28 or harmed²⁹ by hormone therapy. Higher levels of endogenous estrogens have been associated with greater declines in verbal fluency in older women.³⁰ Because all women receiving CEE-alone therapy had undergone hysterectomy, which may be a risk factor on its own for cognitive impairment,³¹ it may be that women’s response to hormone therapy depends on whether loss of endogenous estrogens is gradual or precipitous.³²

The WHIMSY trial had sufficient power to rule out mean treatment effects of 0.15 SDs, within its original design specifications, supporting the use of its telephone-based battery. Telephone-based cognitive assessments are becoming more widely used in trials and cohort studies.³³

The larger WHIMS and WHISCA trials, which featured more cognitive assessments over time, detected CEE-related mean decrements of 0.06 to 0.08 SDs in cognitive function.^1-3 Despite these relatively small mean differences, CEE-based therapy among women at least 65 years of age resulted in a 75% increase in the hazard for dementia and significant decrements in brain volumes.^4,34 It is possible that hormone therapy could have had a similar small effect on cognitive function in younger women that may have clinical significance but for which WHIMSY was underpowered to detect. Two findings argue against this. First, both the primary outcome—a measure of global cognitive function—and the composite outcome formed by averaging all test scores had essentially no treatment effects. Second, there was no evidence that differences between intervention groups varied depending on markers of adherence or on-trial exposure. There was, across both arms, a trend for better adherence among women with higher levels of executive function; we interpret this as reflecting an increased ability to adhere to the study protocol rather than a treatment effect.

The WHIMSY trial does not address whether initiating hormone therapy during menopause and maintaining therapy until symptoms pass affects cognitive function, either in the short or longer term. All enrollees received no hormone therapy for at least 3 months prior to randomization; their last menstrual period had occurred a mean of 4 (no prior hysterectomy) to 8 (prior hysterectomy) years before WHI enrollment. As volunteers for a clinical trial and posttrial follow-up, these women may not represent more general populations.³⁵ Participants had been unmasked to their treatment assignment in the WHI, which could have influenced their willingness to participate in WHIMSY and their performance on cognitive tests; however, good balance was maintained between treatment groups for important risk factors for cognitive dysfunction. Pretreatment levels of cognitive function were not assessed; however, the WHIMSY cohorts were well balanced with respect to pretreatment risk factors for cognitive impairment; covariate adjustment for these did not materially affect estimated treatment effects.

Our findings provide reassurance that CEE-based therapies when administered to women earlier in the postmenopausal period do not seem to convey long-term adverse consequences for cognitive function. Although we cannot rule out acute benefits or harm, these do not appear to be present to any degree a mean of 7 years after cessation of therapy. One exception may be for minor longer-term disturbances of verbal fluency for women prescribed CEE alone; however, this may be a chance finding.

Corresponding Author: Mark A. Espeland, PhD, Department of Biostatistical Sciences, Wake Forest School of Medicine, Medical Center Boulevard, Winston-Salem, NC 27157 (mespelan@wakehealth.edu).

Accepted for Publication: April 5, 2013.

Published Online: June 24, 2013. doi:10.1001/jamainternmed.2013.7727.

Author Contributions:Study concept and design: Espeland, Shumaker, Manson, Rapp, Wactawski-Wende, Stefanick.

Acquisition of data: Shumaker, Manson, Robinson, Rapp, Lane, Wactawski-Wende, Stefanick.

Analysis and interpretation of data: Espeland, Shumaker, Leng, Manson, Brown, LeBlanc, Vaughan, Robinson, Rapp, Goveas, Wactawski-Wende, Li, Resnick.

Drafting of the manuscript: Espeland, Shumaker, Leng, Vaughan, Li.

Critical revision of the manuscript for important intellectual content: Espeland, Shumaker, Manson, Brown, LeBlanc, Vaughan, Robinson, Rapp, Goveas, Lane, Wactawski-Wende, Stefanick, Li, Resnick.

Statistical analysis: Espeland, Leng.

Obtained funding: Espeland, Shumaker, Lane, Wactawski-Wende, Stefanick, Resnick.

Administrative, technical, and material support: Shumaker, Manson, Vaughan, Rapp, Lane, Wactawski-Wende, Resnick.

Study supervision: Shumaker, Rapp, Wactawski-Wende.

Conflict of Interest Disclosures: Dr Robinson has received research grants to her institution from Amarin, Amgen, Daiichi-Sankyo, Esperion, Genentech/Hoffman–La Roche, GlaxoSmithKline, Merck, and Zinfandel/Takeda.

Funding/Support: This study was supported by the National Institute on Aging, Contract No. HHSN-271-2011-00004C. WHIMSY is funded by the National Institute on Aging, Contract No. HHSN-271-2011-00004C, and the WHI program is funded by the National Heart, Lung, and Blood Institute, US Department of Health and Human Services.

WHIMSY Study Group Members and Centers: WHIMSY Clinical Coordinating Center: Sally Shumaker, Mark Espeland, Steven Rapp, Claudine Legault, Laura Coker, Maggie Dailey, Michelle Naughton, Beverly Snively, Iris Leng, Kaycee Sink, and Anne L. Vaughan (Wake Forest School of Medicine, Winston-Salem, North Carolina). WHIMSY Clinical Centers: Sylvia Wassertheil-Smoller (Albert Einstein College of Medicine, Bronx, New York), Aleksandar Rajkovic (Baylor College of Medicine, Houston, Texas), JoAnn E. Manson (Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts), Lawrence Phillips (Emory University, Atlanta, Georgia), Andrea LaCroix (Fred Hutchinson Cancer Research Center, Seattle, Washington), Richard Katz (George Washington University Medical Center, Washington, DC), Rowan Chlebowski (Harbor-UCLA Research and Education Institute, Torrance, California), Frances Lynch (Kaiser Permanente Center for Health Research, Portland, Oregon), Bette Caan (Kaiser Permanente Division of Research, Oakland, California), Jane Morley Kotchen (Medical College of Wisconsin, Milwaukee), Barbara Howard (MedSTAR Research Institute, Washington, DC), Charles Eaton (Memorial Hospital of Rhode Island, Pawtucket), Sandra Weintraub (Northwestern University, Chicago, Illinois), Lynda Powell (Cook County Hospital, Rush-Presbyterian St Luke's Medical Center, Chicago, Illinois), Marcia L. Stefanick (Prevention Research Center, Stanford University), Dorothy Lane (State University of New York at Stony Brook, Stony Brook), Rebecca Jackson (Ohio State University, Columbus), Beth Lewis (University of Alabama at Birmingham, Birmingham), Cynthia Thompson (University of Arizona, Tucson/Phoenix, Tucson), Jean Wactawski-Wende (University of Buffalo, Buffalo, New York), John Robbins (University of California at Davis, Sacramento), Allan Hubbell (University of California at Irvine, Orange), Lauren Nathan (University of California at Los Angeles), Margery Gass (University of Cincinnati, Cincinnati, Ohio), Marian Limacher (University of Florida, Gainesville/Jacksonville, Gainesville), Kamal Masaki (University of Hawaii, Honolulu), Jennifer Robinson (University of Iowa, Iowa City/Davenport, Iowa City), Judith Ockene (University of Massachusetts, Worcester), Norman Lasser (University of Medicine and Dentistry of New Jersey, Newark), John Kostis (University of Medicine and Dentistry of New Jersey, New Brunswick), Richard Grimm (University of Minnesota, Minneapolis), Robert Brunner (University of Nevada, Reno), Carol Murphy (University of North Carolina, Chapel Hill), Lewis Kuller (University of Pittsburgh, Pittsburgh, Pennsylvania), Karen Johnson (University of Tennessee at Memphis, Memphis), Donald Royall (University of Texas Health Science Center, San Antonio), Shirley Beresford (University of Washington, Seattle), Gloria Sarto (University of Wisconsin, Madison), Mara Vitolins (Wake Forest University Health Sciences, Winston-Salem, North Carolina), and Michael Simon (Wayne State University School of Medicine/Hutzel Hospital, Detroit, Michigan).

Additional Information: The following is a short list of WHI investigators: Program Office: Jacques Rossouw, Shari Ludlam, Dale Burwen, Joan McGowan, Leslie Ford, and Nancy Geller (National Heart, Lung, and Blood Institute, Bethesda, Maryland). Clinical Coordinating Center: Garnet Anderson, Ross Prentice, Andrea LaCroix, and Charles Kooperberg (Fred Hutchinson Cancer Research Center, Seattle, Washington). Investigators and Academic Centers: JoAnn E. Manson (Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts), Barbara V. Howard (MedStar Health Research Institute/Howard University, Washington, DC), Marcia L. Stefanick (Prevention Research Center, Stanford University), Rebecca Jackson (Ohio State University, Columbus), Cynthia A. Thomson (University of Arizona, Tucson/Phoenix), Jean Wactawski-Wende (University at Buffalo, Buffalo, New York), Marian Limacher (University of Florida, Gainesville/Jacksonville), Robert Wallace (University of Iowa, Iowa City/Davenport), Lewis Kuller (University of Pittsburgh, Pittsburgh, Pennsylvania), and Sally A. Shumaker (Wake Forest University School of Medicine, Winston-Salem, North Carolina). Women’s Health Initiative Memory Study: Sally A. Shumaker (Wake Forest University School of Medicine, Winston-Salem, North Carolina). For a list of all the investigators who have contributed to WHI science, please visit https://cleo.whi.org/researchers/Documents%20%20Write%20a%20Paper/WHI%20Investigator%20Long%20List.pdf.