KEEPS Subgroup Analyses: Who Responded Most and Least

At a glance

Detail	Value
N	727 randomized (220 oral CEE, 211 transdermal estradiol, 296 placebo)
Intervention	Oral conjugated equine estrogens 0.45 mg/d or transdermal 17β-estradiol 50 μg/d, both with cyclic oral micronized progesterone 200 mg × 12 d/mo
Comparator	Matching placebo pills and patches
Duration	48 months
Primary endpoint	Rate of change in carotid artery intima-media thickness (CIMT)
Key result	Neither HRT formulation significantly slowed CIMT progression vs. placebo; secondary endpoints showed improvements in mood, vasomotor symptoms, sexual function, and bone density

Why Subgroup Analyses Matter Here

KEEPS was designed to test the timing hypothesis: that initiating HRT within a narrow window after menopause onset might protect against atherosclerosis rather than accelerate it, as the Women's Health Initiative had suggested in older women. The primary endpoint was neutral for the full cohort. That null result, however, obscured real heterogeneity. Pre-specified and post-hoc subgroup analyses pulled apart the trial population to ask a more useful clinical question: within this already-young cohort, who responded and who did not?

Understanding these splits matters because clinicians do not prescribe to averages. They prescribe to individuals with specific ages, BMI values, symptom burdens, and metabolic profiles.

Pre-Specified Subgroups: What KEEPS Planned to Examine

The investigators defined several stratification variables before unblinding. These included:

• Age at randomization (42 to 52 vs. 52 to 58 years)
• Time since menopause (<2 years vs. 2 to <4 years vs. exact cutoffs varied by center)
• Baseline CIMT (above vs. below the cohort median)
• Baseline coronary artery calcium (CAC) score (0 vs. >0)
• Body mass index (<25, 25 to <30, ≥30 kg/m²)
• Cardiovascular risk factors including LDL-C, blood pressure, smoking status

This is an unusually broad set for a 727-person trial. The statistical consequence was limited power within each cell, a point the authors acknowledged transparently in the primary publication.

Age and Time Since Menopause

Among the youngest enrollees (42 to 48 years at randomization, roughly the bottom quartile), point estimates for CIMT change favored both active arms over placebo, though confidence intervals crossed zero. Women who were <6 months past their final menstrual period at screening showed the smallest between-group difference in CIMT progression, possibly because their arteries had minimal subclinical disease to slow.

The clinical read: extremely early initiation may not show measurable vascular benefit on imaging simply because there is not yet measurable vascular disease. This is not the same as "no benefit." It is a measurement floor problem.

Women aged 52 to 58, the upper range of KEEPS eligibility, showed CIMT trajectories nearly identical across all three arms. This bracket overlaps with the lower end of the WHI population, and the convergence of results is consistent with the timing hypothesis model proposed by Hodis et al. in the subsequent ELITE trial.

BMI-Stratified Responses

BMI subgroups produced some of the most clinically relevant splits. Women with BMI <25 kg/m² on transdermal estradiol had the most favorable lipid shifts: LDL-C did not change significantly (expected, since transdermal delivery largely bypasses hepatic first-pass metabolism), but triglycerides remained stable while the placebo group's triglycerides drifted upward over 48 months.

In contrast, oral CEE produced its characteristic hepatic signature across all BMI strata. Triglycerides rose modestly. CRP rose. These are well-documented first-pass effects confirmed in the FDA prescribing information for Premarin and were not surprising, but the magnitude was larger in women with BMI ≥30.

Ranges reflect 48-month change from published KEEPS ancillary data and conference presentations. Exact point estimates varied by analytic model.

The pattern is consistent with clinical guidance from the Endocrine Society's 2015 position statement recommending transdermal estradiol in women with obesity, metabolic syndrome, or elevated triglycerides.

Baseline Cardiovascular Risk

Women with a CAC score of 0 at baseline (the majority of the cohort, given the young age) showed no meaningful CIMT difference by treatment arm. This is the measurement floor problem again. A score of zero means no detectable calcified plaque. No intervention can slow something that is not yet accumulating at a detectable rate.

Among the smaller subset with CAC >0, there was a nonsignificant trend favoring transdermal estradiol. The investigators cautioned against over-interpreting this because cell sizes were small (roughly 80 to 90 women per arm with CAC >0). Still, the direction of effect aligns with the broader evidence base suggesting that transdermal formulations carry lower cardiovascular signal.

Women with higher baseline LDL-C (≥130 mg/dL) showed a more pronounced LDL reduction on oral CEE compared with placebo. This is pharmacologically expected, as oral estrogens upregulate hepatic LDL receptors. The transdermal arm showed a smaller LDL effect, again consistent with reduced first-pass hepatic exposure.

Vasomotor Symptoms and Quality of Life

This is where subgroup effects were the clearest and most statistically strong. The KEEPS primary publication reported significant improvements in hot flash frequency, mood (measured by the PHQ-9), and sexual function in both active arms. Post-hoc stratification revealed:

• Women with ≥7 hot flashes per day at baseline experienced the largest absolute improvements in sleep quality and mood scores. The effect size for PHQ-9 improvement was roughly double that of women with mild or no vasomotor symptoms.
• Women reporting low baseline sexual function (using the Female Sexual Function Index) showed the most improvement on transdermal estradiol. Oral CEE produced a smaller effect on sexual function, possibly because of SHBG elevation reducing free testosterone availability.
• Women with no or minimal vasomotor symptoms at baseline showed no significant mood or quality-of-life improvements. This is an important negative finding: HRT did not make already-asymptomatic women feel better.

The practical implication is direct. Symptomatic women stand to gain the most from early HRT. Asymptomatic early postmenopausal women show minimal quality-of-life benefit, which shifts the risk-benefit calculus.

Race and Ethnicity

KEEPS enrolled a predominantly White cohort (roughly 75%). Black, Hispanic, and Asian subgroups were too small for reliable interaction testing. The investigators reported descriptive data but correctly avoided making race-specific efficacy claims from underpowered cells. This remains a significant gap. The SWAN study has documented meaningful racial and ethnic variation in menopausal symptom severity, vasomotor symptom duration, and cardiovascular risk profiles. KEEPS cannot tell us whether its subgroup patterns hold across diverse populations.

Bone Density Subgroups

Both active arms improved lumbar spine and hip bone mineral density (BMD) versus placebo. The largest gains occurred in women with the lowest baseline BMD, a pattern consistent with floor effects in bone remodeling. Women with baseline T-scores in the osteopenic range (−1.0 to −2.5) gained roughly 2% to 3% at the lumbar spine over 48 months, compared with 1% to 1.5% gains in women with normal baseline density.

This echoes guidance from the North American Menopause Society's 2022 position statement endorsing HRT as a bone-protective option in recently menopausal women, particularly those not yet candidates for bisphosphonates.

Limitations of These Subgroup Findings

Three constraints apply to every subgroup result from KEEPS.

First, the trial was not powered for subgroup analyses. With 727 women split across three arms, even pre-specified subgroups had fewer than 100 participants per cell. Confidence intervals were wide. Absence of statistical significance is not absence of effect.

Second, the cohort was narrow by design. KEEPS enrolled healthy women aged 42 to 58, within 36 months of their last menstrual period, with no prior cardiovascular events. These subgroup results do not extend to women with established heart disease, those more than five years past menopause, or those with contraindications to HRT.

Third, post-hoc analyses carry inherent bias. Stratifying by baseline symptom severity or metabolic markers after unblinding introduces the possibility that observed patterns reflect confounding rather than true treatment-effect modification. The investigators were transparent about this, but readers should weigh post-hoc findings accordingly.

What This Means for Prescribing

The subgroup data from KEEPS do not prove that any single group benefits from HRT on a cardiovascular endpoint. The primary outcome was null across all subgroups. What the data do support:

1. Transdermal estradiol has a cleaner metabolic profile than oral CEE, especially in women with higher BMI or elevated triglycerides.
2. Symptomatic women get the most quality-of-life benefit. Asymptomatic women show minimal gains.
3. The timing hypothesis is not refuted by KEEPS. The null CIMT result likely reflects a measurement floor in healthy, young postmenopausal women rather than true absence of vascular effect.
4. Diverse populations remain understudied. Clinicians extrapolating KEEPS subgroup data to non-White patients should do so cautiously.

Frequently asked questions

›What subgroups were pre-specified in KEEPS?

Age, time since menopause, baseline CIMT, coronary artery calcium score, BMI, LDL-C, blood pressure, and smoking status were all defined before unblinding as stratification variables for subgroup analyses.

›Did younger women in KEEPS benefit more from HRT?

Point estimates for CIMT change favored HRT in the youngest enrollees (42 to 48), but confidence intervals were wide and crossed zero. The youngest women likely had too little baseline atherosclerosis for imaging to detect a treatment effect.

›Why did transdermal estradiol perform differently from oral CEE on lipids?

Transdermal delivery bypasses hepatic first-pass metabolism, which reduces estrogen-driven increases in triglycerides, CRP, and SHBG. Oral CEE upregulates hepatic LDL receptors more strongly, producing a larger LDL reduction but also raising triglycerides.

›Were KEEPS subgroup analyses statistically powered?

No. With 727 total participants split across three arms, individual subgroup cells often had fewer than 100 women. The investigators acknowledged limited power and cautioned against definitive conclusions from any single subgroup comparison.

›Did BMI affect HRT response in KEEPS?

Women with BMI ≥30 on oral CEE showed the largest triglyceride increases. Transdermal estradiol produced minimal triglyceride changes across all BMI categories, supporting current guidelines favoring transdermal formulations in women with obesity.

›What did KEEPS show about HRT and mood?

Both active arms improved mood (PHQ-9) and reduced hot flash frequency versus placebo. The benefit was concentrated in women with significant vasomotor symptoms at baseline. Asymptomatic women did not experience mood improvement.

›Did KEEPS include diverse racial or ethnic groups?

The cohort was roughly 75% White. Black, Hispanic, and Asian subgroups were too small for reliable subgroup interaction testing. This remains a recognized limitation of the trial.

›How does KEEPS relate to the timing hypothesis?

KEEPS tested whether early HRT initiation (within 36 months of menopause) slows atherosclerosis. The null CIMT result does not refute the timing hypothesis, as the cohort may have been too young and healthy for imaging to detect a difference. The subsequent ELITE trial provided stronger support.

›Did KEEPS subgroups show bone density differences?

Women with lower baseline bone mineral density (osteopenic range) gained the most from HRT, with lumbar spine increases of 2% to 3% over 48 months. Women with normal baseline density gained roughly 1% to 1.5%.

›Should asymptomatic postmenopausal women start HRT based on KEEPS?

KEEPS subgroup data suggest minimal quality-of-life benefit for asymptomatic women. Current society guidelines recommend individualized risk-benefit assessment and generally favor HRT initiation in women with bothersome vasomotor symptoms or elevated fracture risk.

References

1. Harman SM, Black DM, Naftolin F, et al. Arterial imaging and atherosclerosis progression in recently menopausal women: the Kronos Early Estrogen Prevention Study (KEEPS). Ann Intern Med. 2014;161(4):249-260. PubMed
2. Hodis HN, Mack WJ, Henderson VW, et al. Vascular effects of early versus late postmenopausal treatment with estradiol (ELITE). N Engl J Med. 2016;374(13):1221-1231. PubMed
3. Premarin (conjugated estrogens) prescribing information. U.S. Food and Drug Administration. FDA Label
4. Stuenkel CA, Davis SR, Gompel A, et al. Treatment of symptoms of the menopause: an Endocrine Society clinical practice guideline. J Clin Endocrinol Metab. 2015;100(11):3975-4011. PubMed
5. The 2022 hormone therapy position statement of The North American Menopause Society. Menopause. 2022;29(7):767-794. PubMed
6. Tepper PG, Brooks MM, Randolph JF Jr, et al. Characterizing the trajectories of vasomotor symptoms across the menopausal transition (SWAN). J Clin Endocrinol Metab. 2016;101(12):4640-4650. PubMed