27-hydroxycholesterol, an endogenous SERM, and risk of fracture in postmenopausal women: A nested case-cohort study in the Women’s Health Initiative

Abstract

27-hydroxycholesterol (27HC) is a purported, novel endogenous SERM. In animal models, 27HC has an anti-estrogen effect in bone, and 17β-estradiol mitigates this effect. 27HC in relation to fracture risk has not been investigated in humans. Depending on the level of bioavailable 17β-estradiol (bioE₂), 27HC may increase fracture risk in postmenopausal women and modify the fracture risk reduction from menopausal hormone therapy (MHT). To test these a priori hypotheses, we conducted a nested case-cohort study of 868 postmenopausal women within the Women’s Health Initiative Hormone Therapy trials (WHI-HT). The WHI-HT tested conjugated equine estrogens versus placebo and separately conjugated equine estrogens plus progestin versus placebo. Fracture cases were 442 women who had an adjudicated incident hip or clinical vertebral fracture during the WHI-HT follow-up. The sub-cohort included 430 women randomly selected at WHI-HT baseline, 4 of whom had a subsequent fracture. Of 868 women, 266 cases and 219 non-cases were assigned to the placebo arms. Cox models estimated hazard ratios for incident fracture in relation to pre-randomization circulating levels of 27HC and 27HC/bioE₂ molar ratio. Models adjusted for age, race/ethnicity, total cholesterol, bioE₂, sex hormone-binding globulin, 25-hydroxyvitamin D, diabetes, osteoporosis, prior MHT use, BMI, falls history and prior fracture. In women assigned to placebo arms, those in the middle and the highest tertiles of 27HC/bioE₂ had an up to 1.9-fold (95% confidence intervals: 1.25–2.99) greater risk of fracture than women in the lowest tertile. In women assigned to MHT arms, fracture risk increased with continuous 27HC/bioE₂ levels but not with categorical levels. 27HC levels alone were not associated with fracture risk. 27HC and 27HC/bioE₂ did not modify the fracture risk reduction from MHT. In postmenopausal women, circulating levels of 27HC relative to bioE₂ may identify those at increased risk of fracture.

INTRODUCTION

Approximately one in two non-Hispanic white women ages 50 or older experience a fragility fracture during her lifetime.⁽¹⁾ The link between bone loss and low estrogen, the most established risk factor for fracture in postmenopausal women, was reported as early as 1941.⁽²⁾ Low circulating endogenous 17β-estradiol (E₂), including free E₂ and bioavailable E₂ (bioE₂), are associated with increased risk of hip fracture in postmenopausal women.^(3–6) Short-acting exogenous estrogens, e.g. conjugated equine estrogens (CEE), prevent bone mineral loss at the hip and spine.^{(7, 8)} In the Women’s Health Initiative hormone therapy (WHI-HT) trials, CEE reduced the risk of fragility fractures in postmenopausal women, the majority of whom had no diagnosis of low bone mineral density (BMD) or osteoporosis.^(9–11) Furthermore, CEE provides comparable protection against hip fracture for postmenopausal women, regardless of circulating levels of total E₂ or bioE₂.⁽¹²⁾

Selective estrogen receptor modulators (SERMs) are pharmacological agents that act at the estrogen receptor (ER) to modulate estrogen’s effects systemically but differentially as a relative agonist or antagonist, depending on tissue site, including bone.⁽¹³⁾ 27-hydroxycholesterol (27HC) is a purported novel endogenous SERM identified in mice.^(14–19) In animal and in vitro studies, 27HC functions as a negative ER regulator that reduces the favorable effects of estrogen in bone;^{(15, 16)} an ER antagonist in vascular endothelium and smooth muscle cells;⁽¹⁷⁾ a partial agonist in ER-positive breast cancer in mice;^{(18, 19)} and recently an ER agonist in prostate cancer cell.⁽²⁰⁾ 27HC is the most abundant circulating oxysterol in humans and a primary cholesterol metabolite produced by the action of the enzyme CYP27A1.^{(14, 21)} Circulating levels of 27HC have shown to be regulated by vitamin D supplements in women with breast cancer.⁽²²⁾

To our knowledge, the effect of 27HC on the risk of fragility fracture in humans has not been reported. Based on findings in animal studies,⁽¹⁵⁾ we postulated that an effect of 27HC on bone depends on the amount of estrogen present in women, since 27HC is likely to compete with E₂ for ER-binding and may thereby reduce the protective effect of estrogen and increase fracture risk. In this study, we hypothesized that 1) postmenopausal women who have a high circulating level of 27HC relative to bioE₂ (27HC/bioE₂ molar ratio) have an increased risk of incident fracture than those who do not, and 2) the circulating levels of 27HC and 27HC/bioE₂ molar ratio modify the effect of menopausal hormone therapy (MHT) on fracture risk in postmenopausal women. To test these hypotheses, we measured baseline (pre-randomization) plasma 27HC levels in a prospective case-cohort study of 868 postmenopausal women nested within the WHI-HT trials.

MATERIALS AND METHODS

Women’s Health Initiative Hormone Therapy (WHI-HT) Trials

The detailed information of study design, survey methodology, laboratory techniques, and major findings for two WHI-HT trials have been published.^{(9, 10, 23)} The WHI-HT trials consisted of 27,347 postmenopausal women aged 50 years or older (mean 63 years). Women with hysterectomy were randomly assigned to receive 0.625 mg CEE or placebo in the “E-alone trial”,⁽⁹⁾ and women with intact uteri were randomly assigned to receive 0.625 mg CEE plus 2.5 mg medroxyprogesterone acetate or placebo in the “E+P trial”.⁽¹⁰⁾ Hip and clinical vertebral fractures were secondary outcomes of WHI-HT trials, adjudicated by centrally trained local physician adjudicators who reviewed radiology reports.⁽²⁴⁾ Hip fractures underwent a second central adjudication. The agreement between two adjudications was 94%.^{(25, 26)} Effects of the two MHT interventions compared with placebo groups on the risk of such fractures were evaluated during a mean follow-up of 7.2 years (E-alone trial) and 5.6 years (E+P trial).^(9–11)

Prospective Case-Cohort Study Nested within WHI-HT Trials

We conducted a prospective case-cohort study of 868 postmenopausal nested within the WHI-HT trials; 485 women were randomized to the placebo arms and 383 women to the MHT arms. We randomly selected 430 women who had had sufficient volumes of pre-randomization plasma sample as the random sub-cohort, irrespective of whether they subsequently had a fracture (Figure 1). The fracture cases included 442 women who had had an incident hip fracture (all 242 such women) or an incident clinical vertebral fracture (200 randomly selected from 231 such women) during the WHI-HT trials. Among these 442 women with incident fracture, 4 women were also sampled into the random sub-cohort. Therefore, 426 of 430 women in the sub-cohort who did not experience an incident hip or clinical vertebral fracture were analyzed as “non-cases”. Among 485 women randomized to the placebo arms, 266 were incident fracture cases and 219 were non-cases. We censored women at the date of adjudicated incident fracture or the last follow-up date of the WHI-HT trials. The median follow-up for the current study was 6.7 years. The Research Compliance Office at Stanford University waived the IRB review because the data and specimens were collected for purposes other than the current research and researchers cannot identify study subjects.

Figure 1.

Flow chart of the case-cohort study nested within the two WHI-HT trials.

Baseline Participant Characteristics

Trained interviewers of the WHI-HT trials conducted baseline questionnaires, which gathered information of age, self-reported race/ethnicity, reproductive and medical history, lifestyle habits, and fracture-related factors (such as ever broke a bone and number of falls in past 12 months). Interviewers also examined use of bone-altering drugs (bisphosphonate, calcitonin, estrogen, glucocorticoids, and supplements of calcium and vitamin D) and lipid-lowering drugs (statins, fibric acid derivatives, and niacin) in pill bottles.⁽²⁷⁾ Certified clinic staff conducted anthropometric measures with participants lightly dressed without shoes. Body mass index (BMI) was calculated by weight (kg) divided by the square of height (m²).

Measurement of Baseline Plasma 27HC

Blood samples were processed and stored at −80°C.⁽²⁷⁾ Banked EDTA plasma samples (250 μL) were shipped to UT Southwestern Department of Molecular Genetics (Dallas, TX). 27HC concentrations were measured using high-performance liquid chromatography-mass spectrometry (HPLC-MS).^{(21, 28)} A modified Bligh-Dyer extraction was performed to isolate lipids in samples mixed with hydrolysis solution (0.5M KOH in MeOH), methanolic KOH, and deuterated standards (Avanti Polar Lipids; Alabaster, AL).⁽²⁸⁾ Aminopropyl solid-phase extraction (Biotage; Charlotte, NC) was used to purify lipids. 27HC was quantitatively analyzed using a tertiary Shimadzu LC-20XR HPLC system (Shimadzu Scientific Instruments; Columbia, MD), coupled to a Sciex API-5000 triple quadrupole MS equipped with Turbo V ESI source (Framingham, MA). 27HC was resolved with a binary solvent gradient, using a Kinetex C₁₈ HPLC column (150×2.1 mm, 1.7 μm particle size; Phenomenex; Torrance, CA). Mobile phase A was 70% acetonitrile, mobile phase B was 1:1 (v/v) acetonitrile:isopropanol, and mobile phase C was dichloromethane. Mobile phase A and B contained 5mM ammonium acetate. The correlation of 27HC levels among 44 blind duplicates was 0.85 and the average coefficient of variation was 9.5% (max. 36.6%).

Measurement of Baseline Plasma 25-hydroxyvitamin D (25OHD)

Pre-randomization levels of 25OHD were the sum of 25-hydroxyvitamin D₂ and 25-hydroxyvitamin D₃ levels, determined by HPLC-MS following the same process of 27HC measurement at the UT Southwestern Department of Molecular Genetics (Dallas, TX).^{(21, 28)} Among 44 blind duplicates, the correlation of 25-hydroxyvitamin D₂ levels was 0.95 and the correlation of 25-hydroxyvitamin D₃ levels was 0.98. The average coefficient of variation for 25-hydroxyvitamin D₂ and D₃ levels was 14.8% (max. 26.9%) and 4.3% (max. 16.6%), respectively.

Pre-existing Baseline Blood Measurements

Pre-randomization total E₂ and sex hormone-binding globulin (SHBG) were measured previously in serum at the Reproductive Endocrine Research Laboratory, University of Southern California (Los Angeles, CA).^{(4, 12)} Total E₂ concentrations were quantified using indirect radioimmunoassay with an initial column chromatographic extraction step, with a lower detection limit of 3 pg/mL (11 pmol/L), intra-assay CV of 7.9% at 34 pg/mL (124 pmol/L), and inter-assay CV of 8.0% at 16 pg/mL (58.7 pmol/L) and 12.0% at 27 pg/mL (99.1 pmol/L).⁽¹²⁾ SHBG was measured by solid-phase two-site chemiluminescent immunoassay (Immulite Analyzer; Diagnostic Products; Los Angeles, CA) with a lower detection limit of 0.02 μg/mL (0.2 nmol/L), intra-assay CVs of 4.1% to 7.7%, and inter-assay CVs of 5.8% to 13%.^{(4, 12)} Free E₂ (non-albumin-, non-SHBG-bound E₂) and bioE₂ (non-SHBG-bound E₂) were calculated based on the mass action equations,^(29–31) using concentrations of measured total E₂ and SHBG with an assumed normal albumin level. Fasting lipids were measured previously in WHI-HT.⁽²⁷⁾ LDL-C levels were calculated using the Friedewald formula in specimens having triglycerides <400 mg/dL.⁽³²⁾

Statistical Analysis

All descriptive analyses were performed separately for women in the placebo and the MHT arms. The χ² test (categorical variables) and Kruskal-Wallis test (continuous variables) were conducted to compare baseline characteristics between incident fracture cases and non-cases. We calculated Pearson correlation coefficients (r) in non-cases to examine correlations between levels of 27HC, 27HC/bioE₂, total E₂, free E₂, bioE₂, SHBG, 25OHD, total cholesterol, triglycerides, LDL-C HDL-C, and BMI. We presented results for bioE₂ since this is the bioactive form of E₂ and was highly correlated with total E₂ and free E₂.

We performed Cox regression modeling modified for the nested case-cohort study using “unweighted” approach for pseudo-likelihood estimation.^{(33, 34)} We constructed separate models to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for incident fracture in relation to pre-randomization plasma levels of 27HC and 27HC/bioE₂ molar ratio. 27HC and 27HC/bioE₂ were each included as a continuous variable or as a categorical variable using the tertiles defined in non-cases. Cox models adjusted for age, race/ethnicity (non-Hispanic white, black, other), total cholesterol level, BMI, diabetes, bioE₂ level, SHBG level, 25OHD level, ever use MHT, osteoporosis, history of any fracture, and falls in past 12 months. These covariates were considered to be potential confounders (BMI, diabetes, and SHBG) or fracture risk factors (ever use MHT, osteoporosis, history of any fracture, and falls history). Models included bioE₂ to evaluate the 27HC/bioE₂ molar ratio in relation to risk of incident fracture, while holding constant circulating bioE₂ levels. Models also adjusted for 25OHD level because use of vitamin D supplements may decrease circulating 27HC levels and reduce fracture risk in postmenopausal women of the WHI-HT trials.^{(22, 35)} To maintain parsimony, the model excluded physical activity, smoking, and alcohol use since the estimates of HRs and CIs did not materially change when the model included these covariates. Baseline fracture risks differed between women of two WHI-HT trials but did not violate proportionality within each trial (Figure S1); therefore, we stratified Cox models on WHI-HT trials (E-alone or E+P trial). In sensitivity analyses, we performed analyses that included the overall study population.

To investigate whether circulating levels of 27HC and 27HC/bioE₂ modify the MHT’s effect on fracture risk, we constructed separate Cox models, which included the trial arm assignment (MHT or placebo) as a covariate and a product-term of MHT×27HC or MHT×27HC/bioE₂. All statistical analyses were performed in SAS v9.4 (Cary, North Carolina).

RESULTS

The average (± standard deviation, SD) follow-up time was 6.0 (±2.8) years for 868 women of the current study population, 5.7 (±2.9) years for 485 women in the placebo arms and 6.3 (±2.8) years for 383 women in the MHT arms. Regardless of trial arms, women with incident fracture were older, were more likely to be non-Hispanic white, and had a lower BMI compared with non-cases (all p-value <0.001, Table 1). In women randomized to placebo arms, history of osteoporosis and history of any fracture were more prevalent in incident fracture cases than non-cases (p<0.001). Less than 10% of our study population regularly used lipid-lowering medications or bone-altering drugs.

Table 1.

Baseline characteristics of postmenopausal women without incident fracture (non-cases) and those with incident fracture (cases) in the placebo (n=485) and the menopausal hormone therapy (MHT) intervention (n=383) arms of the WHI-HT trials

Baseline (pre- randomization) Characteristics in the WHI-HT trials	Placebo Arms				MHT Arms
	Overall (n= 485)	Non-cases (n=219)a	Incident fracture cases (n=266)	P- valueb	Overall (n= 383)	Non-cases (n=207)a	Incident fracture cases (n=176)	P- valueb
	n (%) or median (interquartile range)
Age (years) c	69 (64, 73)	67 (62, 71)	70 (65, 74)	<.001	69 (62, 73)	66 (60, 71)	71 (65, 75)	<.001
50–64	132 (27)	76 (35)	56 (21)	<.001	131 (34)	90 (43)	41 (23)	<.001
65–69	130 (27)	66 (30)	64 (24)		72 (19)	45 (22)	27 (15)
70 and older	223 (46)	77 (35)	146 (55)		180 (47)	72 (35)	108 (61)
Race/ethnicity				<.001				<.001
Non-Hispanic white	416 (86)	164 (75)	252 (95)		321 (84)	160 (77)	161 (91)
Black	40 (8)	34 (15)	6 (2)		31 (8)	25 (12)	6 (3)
Others	29 (6)	21 (10)	8 (3)		31 (8)	22 (11)	9 (5)
BMI (kg/m2)	27.2 (23.7, 31.2)	28.6 (24.6, 32.0)	26.2 (22.7, 30.0)	<.001	27.8 (24.5, 31.0)	28.3 (24.7, 31.6)	26.9 (24.3, 30.5)	0.037
MET-hr/wk	5.0 (1.2, 14.2)	5.6 (1.5, 16.0)	5.0 (0.7, 13.5)	0.394	5.0 (0.7, 13.0)	5.0 (0.7, 12.7)	5.0 (1.0, 14.2)	0.697
Ever use MHTd	242 (49)	104 (47)	137 (51)	0.407	172 (45)	100 (49)	72 (42)	0.186
Diabetes	44 (9)	24 (11)	20 (8)	0.190	29 (8)	17 (8)	12 (7)	0.606
CVD	79 (16)	36 (16)	43 (16)	0.867	66 (17)	29 (15)	37 (24)	0.053
Osteoporosis	56 (12)	14 (6)	42 (16)	<.001	46 (12)	20 (10)	26 (15)	0.138
Any fracture	214 (44)	70 (32)	144 (54)	<.001	165 (43)	70 (37)	95 (60)	<.001
Self-report falls in last 12 months	162 (33)	62 (28)	100 (37)	0.150	118 (30)	52 (27)	66 (40)	0.028
Lipid-lowering druge	44 (9)	20 (9)	24 (9)	0.967	33 (9)	19 (9)	14 (8)	0.670
Statin	40 (8)	18 (8)	22 (8)	0.984	27 (7)	15 (7)	12 (7)	0.870
Bone-altering drugf	22 (5)	5 (2)	17 (6)	0.025	17 (4)	10 (5)	7 (4)	0.685
Current tobacco smoking	42 (9)	12 (5)	30 (11)	0.065	40 (10)	18 (9)	22 (13)	0.495
Alcohol use				0.905				0.876
Non-drinker	53 (11)	25 (11)	28 (11)		40 (10)	22 (11)	18 (10)
Past drinker	115 (24)	54 (25)	61 (23)		97 (25)	51 (25)	46 (26)
<1 drink/week	150 (30)	68 (31)	82 (30)		123 (32)	70 (34)	53 (30)
1 or more drinks/week	163 (33)	70 (32)	93 (34)		119 (31)	62 (30)	57 (32)
Lipid Levels (mg/dL)
Total cholesterol	225 (202, 254)	231 (206, 255)	221 (200, 252)	0.138	228 (208, 256)	229 (210, 258)	225 (203, 256)	0.189
LDL-C	147 (123, 172)	148 (127, 179)	143 (121, 167)	0.096	148 (127, 171)	148 (127, 171)	148 (127, 171)	0.496
HDL-C	50 (42, 62)	49 (43, 59)	53 (42, 64)	0.043	51 (43, 62)	52 (42, 62)	51 (43, 61)	0.809
Triglyceride	116 (85, 168)	117 (85, 167)	116 (84, 169)	0.734	119 (88, 178)	117 (90, 181)	120 (86, 174)	0.673

WHI-HT = Women’s Health Initiative Hormone Therapy; MET = metabolic equivalent.

^aThe non-cases excluded the 4 fracture cases who were in the random sub-cohort.

^bP-value for comparing fracture cases with non-cases within the placebo and the interventions arms.

^cP-value <0.01 for the distribution of categorical age comparing 485 women in the placebo arms with 383 women in the intervention arms; P-values were >0.1 for other variables.

^dEver use MHT, excluded birth control pill use before age 50.

^eLipid-lowering drugs included statins, fibric acid derivatives, and niacin.

^fBone-altering drugs included bisphosphonate, calcitonin, calcium supplement, estrogen, glucocorticoids, and vitamin D supplement.

Circulating median levels of 27HC/bioE₂ were higher for fracture cases than non-cases in both women in the placebo arms and the MHT arms (p<0.05; Table 2). Approximately 75% of the study population had a total E₂ below 15 pg/mL. Fracture cases had a lower median total E₂, free E₂ and bioE₂, and a higher median SHBG level when compared with non-cases (all p-value <0.05). In non-cases, 27HC levels positively correlated with levels of total cholesterol, LDL-C, triglycerides, and 27HC/bioE₂ (r ranging from 0.30 to 0.53). 27HC/bioE₂ levels positively correlated with total cholesterol, HDL-C, and SHBG (r: 0.20 to 0.32) and negatively with BMI, total E₂, free E₂, and bioE₂ (r: −0.24 to −0.40, all p-values <0.001). BioE₂ levels correlated strongly with levels of total E₂ and free E₂ (both r > 0.98) and less so with BMI (r=0.31). 25OHD levels did not differ between fracture cases and non-cases (p-values >0.1), and 25OHD levels correlated moderately with 27HC/bioE₂ ratio among women in the placebo arms (r =0.15, p <0.001). We did not observe significant correlations of 25OHD to levels of 27HC or bioE₂.

Table 2.

Circulating levels of 27HC, 27HC/bioE₂ molar ratio, and estradiol by incident fracture status in the placebo and the menopausal hormone therapy (MHT) intervention arms of the WHI-HT trials

Biomarkers, median (interquartile range)	Placebo arms (n= 485)		P- valuea	MHT arms (n= 383)		P- valuea	P- valueb
	Non-cases (n= 219)	Fracture cases (n=266)		Non-cases (n=207)	Fracture cases (n=178)
27HC (ng/mL)c	149 (123, 176)	151 (124, 177)	0.977	156 (129, 186)	150 (122, 180)	0.057	0.064
27HC/bioE2 (nmol/L:pmol/L)d	13.8 (8.9, 20.3)	17.8 (11.9, 28.4)	<.001	14.2 (9.5, 22.2)	17.3 (11.4, 25.6)	0.021	0.216
Total E2 (pg/mL)e	11.0 (8.2, 15.9)	9.3 (6.3, 13.0)	<.001	11.1 (8.5, 15.4)	9.4 (7.0, 13.5)	0.003	0.767
Free E2 (pg/mL)e	0.29 (0.20, 0.44)	0.22 (0.13, 0.32)	<.001	0.28 (0.21, 0.40)	0.23 (0.17, 0.32)	<.001	0.563
BioE2 (pg/mL)e	7.3 (5.1, 11.2)	5.6 (3.4, 8.3)	<.001	7.1 (5.3, 10.1)	5.8 (4.1, 8.0)	<.001	0.576
SHBG (nmol/L)f	39.2 (29.1, 57.7)	50.5 (35.3, 72.2)	<.001	42.2 (29.9, 58.4)	49.7 (36.0, 62.0)	0.002	0.611
25OHD (ng/mL)	24.2 (15.3, 30.0)	24.9 (19.1, 30.8)	0.138	23.4 (16.1, 29.3)	23.4 (17.8, 28.7)	0.899	0.627

27HC = 27-hydroxycholesterol; BioE₂ = bioavailable E₂; E₂ = 17β-estradiol; SHBG = sex hormone-binding globulin; 25OHD = 25-hydroxyvitamin D.

^aP-value for Kruskall-Wallis test comparing fracture cases with non-cases in the placebo and in the MHT arms.

^bP-value for Kruskall-Wallis test comparing 219 non-cases in the placebo arms with 207 non-cases in the MHT arms.

^cTo convert 27HC in ng/mL to nmol/L, multiply by 2.48.

^d27HC/bioE₂ molar ratio represents the moles of 27HC relative to bioE₂ (or the number of 27HC molecules relative to bioE₂ molecules) in a given volume of blood.

^eTo convert pg/mL to pmol/L, multiply by 3.67. Free E₂ and bioE₂ were estimated from the mass action equations,^(28–30) using measured concentrations of total E₂ and SHBG with assumed constant for albumin.

^fTo convert nmol/L to μg/mL, divide by 8.896.

In women assigned to the placebo arms, the risk of incident fracture increased with pre-randomization 27HC/bioE₂ level (Table 3). Women who had 27HC/bioE₂ in the middle (10–16 nmol/L:pmol/L) and the highest tertiles (>16 nmol/L:pmol/L) had an up to 1.9-fold (95 CI%: 1.25 to 2.99) greater risk of fracture, compared with women in the lowest tertile (<10 nmol/L:pmol/L). In women assigned to the MHT arms, pre-randomization 27HC/bioE₂ in tertiles were not associated with incident fracture risk. Analyzing the overall study population, irrespective of the WHI-HT trials’ assignment, women in the highest tertile of 27HC/bioE₂ had a 96% greater risk of incident fracture than those in the lowest tertile (adjusted HR: 1.96, 95% CI: 1.39–2.78). Circulating levels of 27HC alone did not show associations with fracture risk in women assigned to either the placebo arms or the MHT arms.

Table 3.

Hazard ratio estimates for incident fracture associated with circulating levels of 27HC and 27HC/bioE₂ molar ratio in the placebo and the menopausal hormone therapy (MHT) intervention arms of the WHI-HT trials

	Placebo arms (n=485)	MHT arms (n= 383)
	Adjusted Hazard Ratio (95% CI)a
27HC (ng/mL)
Per 1-SD (44 ng/mL)	1.07 (0.91–1.28)	0.99 (0.78–1.25)
P for interactionb	0.983	0.847
Tertile categoriesc
<132	1 (reference)	1 (reference)
132–164	1.04 (0.74–1.46)	0.64 (0.29–1.39)
>164	0.94 (0.63–1.40)	0.63 (0.26–1.50)
P for interactionc	0.248	0.377
27HC/bioE2 (nmol/L:pmol/L)
Per 1-IQR (12 nmol/L:pmol/L)	1.09 (0.97–1.24)	1.62 (1.16–2.25)
P for interactionb	0.020	0.270
Tertile categoriesc
<10	1 (reference)	1 (reference)
10 – 16	1.93 (1.25–2.99)	1.14 (0.40–3.26)
>16	1.61 (1.01–2.56)	2.93 (0.92–9.33)
P for interaction†	<.001	0.425

27HC = 27-hydroxycholesterol; BioE₂ = bioavailable E₂; CI = confidence interval; IQR = interquartile range

^aModel stratified on WHI-HT trials (E-alone or E+P trial) and adjusted for age, race/ethnicity, total cholesterol, bioE₂, SHBG, BMI, diabetes, osteoporosis, 25OHD, prior MHT use, any fracture history, and history of falls in past 12 months.

^bP for interaction refers to 27HC×WHI-HT trials (E+P or E-alone trial) or 27HC/bioE₂×WHI-HT trials (E+P or E-alone trial) in models that did not stratify on WHI-HT trials.

^cTertile category for 27HC or 27HC/bioE₂ was derived from the women who did not have an incident fracture (non-cases).

In our study population, MHT reduced fracture risk in postmenopausal women by 36% (adjusted HR: 0.64, 95% CI: 0.45–0.92, MHT arms versus placebo arms) with adjustment for age, race/ethnicity, history of any fracture, history of falls in the last 12 months, bioE₂ level and 25OHD level. Circulating levels of 27HC and 27HC/bioE₂ did not modify such an association (p-values for MHT×27HC and MHT×27HC/bioE₂ were >0.20, Figure 2).

Figure 2.

Hazard ratios of incident fracture comparing women assigned to the MHT arms with the placebo arms for a given level of 27HC (panel A) and 27HC/bioE₂ molar ratio (panel B) in product-term models that included MHT×27HC (panel A) and MHT×27HC/bioE₂ (panel B). 27HC levels and 27HC/bioE₂ molar ratio did not modify MHT’s effect on incident fracture risk (p-value = 0.298 for MHT×27HC and 0.883 for MHT×27HC/bioE₂). Hazard Ratios were adjusted for age, race/ethnicity, history of any fracture, total cholesterol, 12-months history of falls, and bioE₂ level. MHT = menopausal hormone therapy. Dotted lines in black = 95% Confidence Intervals. Dotted lines in blue = Null HR.

DISCUSSION

The current prospective study, the first in humans to our knowledge, showed that postmenopausal women with higher levels of 27HC, relative to estradiol, had an increased risk of incident fracture. During a mean follow-up of 6.0 years, postmenopausal women in the placebo arms who had 27HC/bioE₂ molar ratio in the middle or the highest tertile had an adjusted 61% to 93% greater risk of incident fracture than those in the lowest tertile. Circulating levels of 27HC and 27HC/bioE₂ molar ratio did not appear to alter the fracture risk reduction from MHT. Depending on bioE₂ level, 27HC may be detrimental to bone in postmenopausal women, resulting in an increased fracture risk.

In mice and in vitro studies, 27HC competes with E₂ for binding to the ER and negatively regulates bone, as well as ER-positive breast tumors and the cardiovascular system.^(15–19) In wild-type mice, 27HC reduced BMD at the spine and femur, modified trabecular architecture, decreased trabecular bone volume, reduced osteocalcin level, and increased expression of OPG and RANKL.^{(15, 16)} Ovariectomized Cyp7b1^−/− mice, which had higher amounts of 27HC than wild-type, had a significantly lower cortical BMD than ovariectomized wild-type mice, while supplementation with E₂ reversed this effect on bone.⁽¹⁵⁾ 27HC appears to bind preferentially to ERβ over ERα, and 27HC may not able to fully compete with E₂ binding, suggesting the two may bind at different sites.⁽³⁶⁾ The current study results do not appear to contradict these observations at the receptor level and in the animal models. However, this study cannot comment on how 27HC may bind to ERβ in an environment of different circulating bioE₂ levels.

The current study is the first to report the relationship of circulating 27HC levels to incident fracture risk, accounting for bioE₂ level. A handful of studies have focused on cardiovascular disease and breast cancer in humans.^{(18, 22, 37)} In the WHI-HT trials, circulating levels of 27HC were not associated with incident coronary heart disease (CHD) and did not modify the effect of MHT on CHD risk among 350 CHD cases and 813 matched controls.⁽³⁷⁾ However, 27HC levels were measured by a method different from that in the current study, and E₂ levels were not available. Using the same measurement method as the current study, serum 27HC levels in 3230 participants (56% female) from the Dallas Heart Study were positively correlated with fasting levels of total cholesterol, triglycerides, and LDL-C (r: 0.30 to 0.50), and negatively with BMI (r: −0.10),⁽²¹⁾ which were similar to findings in the current study. In women with ER-positive breast cancer who received high dose vitamin D supplements, change in circulating 27HC levels has an inverse relationship with change in 25OHD levels.⁽²²⁾ We did not observe correlations between levels of 27HC and 25OHD in postmenopausal women.

The current study had limitations. Firstly, unlike hip fractures, vertebral fractures are often asymptomatic, and our sub-cohort and hip fracture cases may have unknowingly included women with such vertebral fractures. However, we would expect that this potential misclassification, if present, would be non-differential between exposure groups and bias risk estimates conservatively toward the null. Secondly, BMD was unavailable in this study. Thirdly, we cannot rule out residual confounding although the multivariable models adjusted for many potential confounders and putative fracture risk factors. Fourthly, the CEE contains mostly estrone and we did not measure estrone levels. We cannot know how the interaction of estrone, E₂ and 27HC may influence the risk of fracture. Lastly, the current study had limited power to detect associations within each of the trial arms.

This study has several strengths. All incident fractures were adjudicated, and the participant population was characterized in detail. Endogenous circulating levels of 27HC and E₂ were obtained in blood drawn prior to WHI-HT randomization, ensuring that measured levels of 27HC and E₂ in the current study were not affected by the MHT intervention. The relatively large sample size gave the study 80% power to detect a minimum HR between 1.20 and 1.25. We were able to adjust for multiple fracture risk factors and confounding factors, including 25OHD levels. Levels of 27HC and 25OHD were quantitatively measured by mass spectrometry.

CONCLUSION

Circulating 27HC in relation to the risk of incident fracture depends on the relative amount of circulating endogenous E₂ in postmenopausal women. Specifically, the risk of incident fracture increased with circulating levels of 27HC to bioE₂ (molar ratio) among postmenopausal women. On the other hand, 27HC levels and 27HC/bioE₂ levels do not modify the fracture risk reduction from MHT. If findings are confirmed, circulating levels of 27HC relative to bioE₂ may identify a subgroup of postmenopausal women who are at particularly high risk of fracture. This study supports the possibility that interventions that reduce the production of 27HC in cholesterol metabolism may benefit bone health in postmenopausal women.⁽¹⁶⁾ The effects of 27HC on bone and on other estrogen-influenced tissue sites warrant further study in the context of circulating bioE₂ level, to weigh the potential risks and benefits of this newly purported endogenous SERM.