Effects of Estrogen Replacement on Bone Geometry and Microarchitecture in Adolescent and Young Adult Oligoamenorrheic Athletes: A Randomized Trial

Abstract

Oligoamenorrheic athletes (OAs) have lower bone mineral density (BMD) and greater impairment of bone microarchitecture, and therefore higher fracture rates compared to eumenorrheic athletes. Although improvements in areal BMD (aBMD; measured by dual-energy X-ray absorptiometry) in OAs have been demonstrated with transdermal estrogen treatment, effects of such treatment on bone microarchitecture are unknown. Here we explore effects of transdermal versus oral estrogen versus no estrogen on bone microarchitecture in OA. Seventy-five OAs (ages 14 to 25 years) were randomized to (i) a 100-μg 17β-estradiol transdermal patch (PATCH) administered continuously, with 200 mg cyclic oral micronized progesterone given for 12 days of each month; (ii) a combined 30 μg ethinyl estradiol and 0.15 mg desogestrel pill (PILL); or (iii) no estrogen/progesterone (NONE) and were followed for 12 months. Calcium (≥1200 mg) and vitamin D (800 IU) supplements were provided to all. Bone microarchitecture was assessed using high-resolution peripheral quantitative CT at the distal tibia and radius at baseline and 1 year. At baseline, randomization groups did not differ by age, body mass index, percent body fat, duration of amenorrhea, vitamin D levels, BMD, or bone microarchitecture measurements. After 1 year of treatment, at the distal tibia there were significantly greater increases in total and trabecular volumetric BMD (vBMD), cortical area and thickness, and trabecular number in the PATCH versus PILL groups. Trabecular area decreased significantly in the PATCH group versus the PILL and NONE groups. Less robust differences between groups were seen at the distal radius, where percent change in cortical area and thickness was significantly greater in the PATCH versus PILL and NONE groups, and changes in cortical vBMD were significantly greater in the PATCH versus PILL groups. In conclusion, in young OAs, bone structural parameters show greater improvement after 1 year of treatment with transdermal 17β-estradiol versus ethinyl estradiol–containing pills, particularly at the tibia.

Introduction

The prevalence of amenorrhea in female athletes ranges from 3.4% to 66%, varying by exercise type, intensity and duration, and nutritional status,^(1–3) compared with only 3% to 4% in the general population.⁽⁴⁾ Athletes who participate in endurance activities in which leanness is believed to confer a performance advantage are especially at risk for developing Female Athlete Triad (Triad), described as decreased energy availability (EA), menstrual dysfunction, and low bone mineral density (BMD).^(5,6) Low EA occurs when caloric intake does not keep pace with caloric expenditure, and may be inadvertent or purposeful. Such low EA, even while allowing for a normal body mass index (BMI), has negative effects on menstrual function and bone.^(7–9) In addition to the detrimental effects of hypogonadism to bone,^(10,11) low EA has negative effects on other hormones important for bone metabolism, such as insulin-like growth factor 1 (IGF-1) and leptin,^(12,13) even when weight is in the normal range.

Our studies and those of others have shown that the beneficial effects of weight-bearing exercise on bone during adolescence/young-adulthood are lost in oligoamenorrheic athletes (OAs), who have lower areal BMD (aBMD) and greater impairment of bone geometry and microarchitecture compared with eumenorrheic athletes,^(11,14,15) increasing their risk for fractures.⁽¹⁶⁾ Up to 24% of adolescent athletes experience oligoamenorrhea,⁽¹⁷⁾ with 78% of high school female athletes reporting at least one component of the Triad.⁽¹⁸⁾ Because adolescence is a critical time for bone acquisition, with 90% of peak bone mass typically achieved by 18 years,⁽¹⁹⁾ preventing and treating the Triad is critical for preventing future fractures.

Although combined oral contraceptive pills (COCs) are frequently prescribed to treat oligoamenorrhea and/or low BMD in adolescents and young women, there are little data to support this common practice.⁽²⁰⁾ A systematic review of 75 studies of varying sample sizes and sample characteristics showed variably an increase, but more commonly a decrease or no change, in BMD with COCs.⁽²¹⁾ Ethinyl estradiol is the typical form of estrogen in COCs and is a nonphysiologic form of estrogen. Additionally, oral estrogen has hepatic first-pass effects, known to downregulate IGF-1, a hormone anabolic to bone.^(22,23) Thus, the type of estrogen and its oral delivery may explain its lack of efficacy in improving low BMD associated with low EA in individuals with oligoamenorrhea.⁽²⁴⁾ We have shown that 17β-estradiol (17β-E2) (the physiologic form of estrogen), administered via the transdermal route (avoiding hepatic effects), increases aBMD in low-weight girls with anorexia nervosa and normal-weight OA.^(24,25) However, effects on volumetric BMD (vBMD) and bone geometry and microarchitecture have not been reported.

Our objective was to determine the impact of a transdermal and physiologic form of estrogen replacement compared to a commonly used COC or no estrogen on changes in vBMD, bone geometry, and microarchitecture at the distal tibia and radius over 1 year in young normal-weight OAs engaged in weight-bearing activity of the lower extremities. We hypothesized that these bone parameters would improve over 12 months of transdermal 17β-E2 administration compared to COCs or no estrogen treatment.

Subjects and Methods

Subject selection

The study was performed at the Massachusetts General Hospital Clinical Research Center. Participants were recruited through medical clinics, advertisements in local newspapers, and fliers at colleges and sporting events in the community. We screened 140 girls and young women 14 to 25 years for participation in our parent study, randomizing eligible OAs to transdermal estrogen with cyclic progesterone, COC, or no treatment for 12 months, and assessed aBMD as the primary study outcome as reported.⁽²⁴⁾ A total of 121 OAs met inclusion criteria and participated in the main study,⁽²⁴⁾ of whom 75 had high-resolution peripheral quantitative computed tomography (HRpQCT) measurements performed at baseline and 12 months, and are reported here for the first time. Participants were required to have a bone age of ≥14 years, because statural growth is mostly complete at this time, and estrogen administration in full replacement doses is unlikely to accelerate epiphyseal fusion and impact adult height.⁽²⁶⁾ Subjects were also required to have a BMI that was >10th percentile for age, >85% of the median BMI (mBMI) for age, or >17.5 kg/m² to avoid enrolling young women with active anorexia nervosa (per Diagnostic and Statistical Manual of Mental Disorders, 4th Edition [DSM-IV]⁽²⁷⁾ criteria, because the 5th edition of the DSM [DSM-5]⁽²⁸⁾ criteria were not published at the time of enrollment). We used these parameters as a “low weight threshold.”

Oligoamenorrhea was defined as the absence of menses for ≥3 months within a period of oligomenorrhea (cycle length >6 weeks) for ≥6 months, or absence of menarche at ≥15 years, as ascertained by self-report. All subjects were screened by a study psychologist/psychiatrist, and those meeting criteria for current active anorexia nervosa were excluded. Of note, all participants were above the “low weight threshold.”

Participants were all endurance athletes involved in ≥4 hours of aerobic weight-bearing exercise of the legs (eg, basketball, soccer, dance) and/or ≥20 miles of running weekly for ≥6 preceding months. Gymnasts, rowers, cyclists, and swimmers were excluded to minimize weight-bearing variability. Exclusion criteria also included conditions other than functional hypothalamic amenorrhea and endurance training that may cause oligoamenorrhea (eg, pregnancy, hyperprolactinemia, polycystic ovarian syndrome, premature ovarian failure, and thyroid dysfunction), use of medications affecting bone (eg, estrogen, progesterone, anabolic steroids, glucocorticoids, bisphosphonates, and teriparatide) in the 3 months preceding study enrollment, and other medical conditions that may affect bone. We did not exclude potential participants if they were taking calcium or vitamin D supplements. For subjects randomized to estrogen and progesterone treatment, exclusion criteria included complex migraines, a first-degree relative with estrogen-dependent cancer (eg, breast cancer), and a family or personal history of conditions that may increase the risk of thromboembolism.

The study was approved by the Institutional Review Board of Partners Healthcare. Informed consent was obtained from participants ≥18 years old and parents of participants <18 years old. Informed assent was obtained from participants <18 years old.

Experimental protocol

Screening visit

At the screening visit, individuals underwent a complete history and physical examination. Height was measured on a wall-mounted stadiometer to the nearest millimeter (average of three measurements), and weight was measured on an electronic scale to the nearest 0.1 kg. BMI was calculated as weight (in kilograms) divided by height (in meters) squared. Height and BMI Z-scores were obtained using Centers for Disease Control and Prevention (CDC) tables.⁽²⁹⁾ A detailed exercise history was obtained to confirm that participants met endurance criteria. Average hours per week of different types of weight-bearing activity over the preceding 6 months were calculated. A blood draw was performed to rule out other causes of oligoamenorrhea. A urine pregnancy test was performed. Subjects younger than 21 years old had a wrist and hand radiograph to determine bone age, assessed by both a radiologist and clinical endocrinologist to confirm that the bone age was ≥14 years.

Baseline visit

The baseline visit, which occurred within 8 weeks of the screening visit, included an interval history and physical examination with anthropometric measurements; fasting levels of 25(OH) vitamin D, estradiol, sex hormone binding globulin (SHBG), and IGF-1; HRpQCT scans at the distal tibia and radius to assess vBMD, bone geometry, and bone microarchitecture; and assessment of resting energy expenditure (REE) using indirect calorimetry (VMAX Encore 29 metabolic cart; Viasys Healthcare, Carefusion, San Diego, CA, USA).⁽³⁰⁾ Data for dual-energy X-ray absorptiometry (DXA) measures of BMD, lean mass, and fat mass and biochemical parameters have been reported in detail.^(24,31)

OAs were randomized to physiological estrogen replacement via the transdermal 17-β E2 patch (PATCH), estrogen replacement via a COC (PILL), or to the no estrogen (NONE) arm by our institution’s research pharmacy based on a predetermined computer-generated randomization sequence. The PATCH group received the 100-μg transdermal 17-β E2 patch applied twice weekly continuously and cyclic oral micronized progesterone (200 mg) for 12 days of each month. Subjects were counseled that PATCH is not an effective form or dose of contraception and were reminded to use other nonhormonal forms of contraception as needed. The PILL group received a COC containing 30 μg ethinyl estradiol with 0.15 mg desogestrel. Doses of transdermal or oral estrogen were based on recommended replacement doses in this age group for hypogonadal individuals.⁽³²⁾ The NONE group received no estrogen or progesterone. All subjects were given calcium and vitamin D supplementation and instructed to take at least 1200 mg elemental calcium in divided doses and 800 IU vitamin D daily. This study was not blinded given concerns that girls and women randomized to PATCH or NONE groups may assume they were on an oral contraceptive pill, and may not use appropriate contraception when necessary.

Follow-up visits

Follow-up visits included an interval history and physical examination, pregnancy test, and fasting blood draw every 3 months for the 12-month duration of the study to review adverse events and rule out pregnancy. Menstruating subjects in the NONE group were scheduled within 10 days of their menses to ensure blood specimens were drawn within the follicular phase. At each visit, participants were asked about any changes in their exercise habits. At these visits, we assessed compliance with study medications with verbal questionnaires and by collecting calendars provided to subjects for recording missed medication doses. We also collected used and unused patches and pills. Estradiol levels were assessed to confirm compliance in the transdermal PATCH group and SHBG levels were assessed as an objective measure of compliance in the PILL group, because levels of SHBG increase with oral ethinyl estradiol and desogestrel administration.^(33,34) Repeat HRpQCT scanning was performed at 12 months.

HRpQCT

vBMD, bone geometry, and microarchitecture were measured at the distal tibia and radius using HRpQCT (XtremeCT; Scanco Medical AG, Brüttisellen, Switzerland), as described.⁽¹¹⁾ Scans were acquired at the nondominant leg and wrist, unless there was a prior fracture in that limb, in which case the contralateral limb was used. Manufacturer’s protocols were followed, including acquisition of a 2D scout view followed by CT slices obtained 22.5 and 9.5 mm from a reference line manually placed at the endplate of the tibia and radius, respectively.⁽³⁵⁾ We chose fixed sites because most of their linear growth was complete in our participants. Scans used 60 kVp effective energy and 100 ms integration time, obtaining 110 CT slices (9.02 mm) with an isotropic voxel size of 82 μm³. At our institution, root mean squares of coefficients of variation (RMS-CV) are 0.2% to 0.8% for density parameters (eg, total, trabecular, and cortical density), 3% to 7% for trabecular microarchitecture, and 3% to 8% for cortical microstructure. 2D image registration based on the total cross-sectional area was performed to align the baseline and follow-up scans, and the degree of overlap was 78% or higher for all subjects.

Outcomes at both the tibia and radius included total vBMD, cortical vBMD, cortical perimeter, cortical area, cortical thickness, trabecular vBMD, trabecular area, trabecular number, trabecular separation, and trabecular thickness. Effective total radiation dose from HRpQCT at the tibia and radius at baseline and 12 months was 20 μSv per participant. One participant each in the PILL and NONE groups did not have data for the tibia.

Biochemical analysis

25(OH) vitamin D was measured using an immunochemiluminometric assay (LabCorp Esoteric Testing, Burlington, NC, USA; sensitivity 4.0 ng/mL; intraassay CV 4.8% to 7.7%). An electrochemiluminescence immunoassay was used to measure 17β E2 (LabCorp Esoteric Testing; sensitivity 25.0 pg/mL; intraassay CV 1.2% to 6.7%), and SHBG (LabCorp Esoteric Testing; 2.00 nmol/L; intraassay CV 1.1% to 1.7%), IGF-1 was measured using mass spectrometry (Quest Diagnostics, Nichols Institute, San Juan Capistrano, CA, USA; sensitivity 15.6 ng/mL; intraassay CV 3.5% to 15%); age-adjusted IGF-1 Z-scores were provided by Quest Diagnostics.

Statistical methods

We used SAS and JMP Statistical version 13.0 (SAS Institute, Inc., Cary, NC, USA) for all analyses and report data as means ± SE. For baseline characteristics, our data were mostly normally distributed (except for estradiol levels) and did not require transformations. For comparison of baseline characteristics (except estradiol) among the three randomization groups, and changes in biochemical variables over time, we performed an overall ANOVA, followed by a Tukey-Kramer analysis to assess between-group differences while controlling for multiple comparisons. For comparison of estradiol levels, which were not normally distributed, the Kruskal-Wallis test was used. A two-tailed p < 0.05 was considered significant. For the randomized controlled trial, we compared changes in HRpQCT measures among PATCH, PILL, and NONE groups over 12 months by performing multivariable regression analysis followed by the Dunnett’s test to assess differences between PATCH versus PILL and PATCH versus NONE, while controlling for age, height, race, and ethnicity (consistent with the planned approach at study outset). Race and ethnicity were based on self-report. These data are reported as least square means ± SE. Further, given differences among groups for menarcheal age and certain HRpQCT endpoints at baseline, we performed additional analyses also controlling for menarcheal age and for baseline measures. Our power calculation was based on preliminary data indicating that the estimated SD for percent change in aBMD over a year is 2.1% in those receiving transdermal estrogen versus no therapy. Specifically, using Dunnett’s test, with a total of at least 72 OA (24 each in the two treatment arms and in the no-treatment arm), the study would have 80% power (using a two-sided alpha level of 0.05) to detect a difference of at least 1.7% in mean percent change in key HRpQCT parameters, such as vBMD, in groups receiving transdermal versus oral estrogen or no estrogen (assuming that the oral estrogen arm would perform no better than no therapy). To determine associations between changes in hormone levels over the study duration in relation to changes in HRpQCT parameters over the same duration, Pearson correlations were performed.

Results

Baseline characteristics

Baseline characteristics of participants randomized to PATCH, PILL, or NONE groups are shown in Table 1. The subset of participants who completed HRpQCT assessment at baseline and 12 months and are reported here did not differ from the larger group in the published parent study⁽²⁴⁾ for clinical characteristics, and only relevant data are presented in this work (Fig. 1). Randomization groups did not differ for age, BMI, percent body fat, months since last menses, biochemical parameters, or resting energy expenditure. Type of menstrual dysfunction (amenorrhea, oligomenorrhea, irregularity) did not differ among groups (Table 1). Baseline aBMD measures have been reported,⁽²⁴⁾ and did not differ among groups in this subset (data not shown). HRpQCT measurements did not differ among groups at baseline, except for slightly lower cortical vBMD and thickness at the distal radius in the PATCH versus PILL groups (p = 0.034 and p = 0.033, respectively). Due to previous fractures, 12 contralateral radii were scanned (PATCH: 6, PILL: 4, NONE: 2) and 11 contralateral tibias (PATCH: 3, PILL: 2, NONE: 6); these were not different across groups (data not shown). Changes in height, weight, BMI, fat mass, and lean mass did not differ among PATCH, PILL, and NONE groups.

Table 1.

Clinical Characteristics and HRpQCT Measures of Oligoamenorrheic Athletes Randomized to PATCH, PILL, or NONE (Baseline Data for Completers)

Clinical characteristics	PATCH (n = 24)	PILL (n = 24)	None (n = 27)	ANOVA p
Age (years), mean ± SE	19.8 ± 0.6	20.1 ± 0.6	19.2 ± 0.5	0.510
Height (cm), mean ± SE	165.8 ± 1.5	164.6 ± 1.2	163.2 ± 0.9	0.302
Height Z-score, mean ± SE	0.57 ± 0.21	0.40 ± 0.21	0.39 ± 0.22	0.810
Weight (kg), mean ± SE	55.06 ± 1.72	55.37 ± 1.31	55.50 ± 1.12	0.974
BMI (kg/m2), mean ± SE	19.94 ± 0.41	20.42 ± 0.41	20.85 ± 0.42	0.299
BMI Z-score, mean ± SE	−0.43 ± 0.17	−0.22 ± 0.16	0.05 ± 0.16	0.109
Fat mass (kg), mean ± SE	13.66 ± 0.84	14.47 ± 0.81	14.34 ± 0.58	0.716
Lean mass (kg), mean ± SE	41.36 ± 0.11	39.80 ± 0.82	40.15 ± 0.78	0.453
Percent body fat, mean ± SE	23.56 ± 0.90	25.43 ± 1.06	25.26 ± 0.72	0.274
Duration since last menses (months), mean ± SE	11.1 ± 2.6	13.7 ± 4.1	7.6 ± 2.8	0.401
Race (white/Asian/More than one race) (%)	83.3/12.5/4.2	95.8/0.0/4.2	92.6/0.0/7.4	0.159
Ethnicity (Non-Hispanic/Hispanic) (%)	100.0/0.0	91.7/8.3	100.0/0.0	0.113
Age of menarche (years), mean ± SE	14. ± 0.4*	12.7 ± 0.5	13.7 ± 0.3	0.024*
Percent with primary amenorrhea/secondary amenorrhea/irregular menses (%)	4.2/62.5/33.3	8.3/41.7/50.0	3.7/40.7/55.6	0.466
Type of predominant sports activity (running/team sport/other) at baseline, n	16/3/5	14/5/5	18/7/2	0.512
Hours/week of weight bearing exercise at baseline, mean ± SE	9.4 ± 1.0	11.8 ± 1.3	9.8 ± 1.0	0.293
Hours/week of weight bearing exercise at 12 months, mean ± SE	8.1 ± 0.8	8.2 ± 1.5	9.0 ± 0.8	0.761
25 (OH) vitamin D (ng/mL), mean ± SE	36.0 ± 2.4	39.3 ± 2.7	37.0 ± 1.9	0.613
Estradiol (pg/mL), median (IQR)	21 (11.1–32.8)	26.6 (13.0–68.3)	37.5 (23–56)	0.061
SHBG (nmol/L), mean ± SE	64.2 ± 6.5	59.7 ± 4.7	56.9 ± 4.5	0.613
IGF-1 (ng/mL), mean ± SE	233.0 ± 17.0	226.6 ± 17.9	254.5 ± 15.6	0.458
IGF-1 Z-score, mean ± SE	−0.64 ± 0.23	q0.51 ± 0.15	−0.40 ± 0.12	0.600
REE (kcal), mean ± SE	1177.1 ± 33.6	1193.6 ± 26.3	1202.3 ± 29.3	0.833
HRpQCT parameters (distal tibia), mean ± SE
vBMD
Total vBMD (mg HA/cm3)	314.1 ± 10.0	332.3 ± 10.2	336.3 ± 9.6	0.245
Cortical vBMD (mg HA/cm3)	856.1 ± 7.6	876.7 ± 7.7	868.0 ± 7.3	0.170
Trabecular vBMD (mgHA/cm3)	195.8 ± 6.4	204.8 ± 6.5	201.6 ± 6.1	0.608
Bone geometry
Cortical area (mm2)	117.9 ± 4.7	120.4 ± 4.8	126.1 ± 4.5	0.435
Cortical thickness (mm)	1.17 ± 0.05	1.22 ± 0.05	1.29 ± 0.05	0.210
Trabecular area (mm2)	562.9 ± 19.6	529.1 ± 20.0	508.0 ± 18.8	0.134
Bone microarchitecture
Trabecular number (1/mm)	1.86 ± 0.06	1.95 ± 0.04	1.89 ± 0.05	0.458
Trabecular separation (mm)	0.46 ± 0.01	0.43 ± 0.01	0.45 ± 0.01	0.278
Trabecular thickness (mm)	0.089 ± 0.003	0.088 ± 0.003	0.089 ± 0.002	0.940
HRpQCT parameters (distal radius), mean ± SE
vBMD
Total vBMD (mg HA/cm3)	273.4 ± 11.9	310.6 ± 11.9	304.2 ± 11.2	0.067
Cortical vBMD (mg HA/cm3)	778.2 ± 14.9	831.0 ± 14.9	820.2 ± 14.0	0.034*
Trabecular vBMD (mg HA/cm3)	159.3 ± 6.5	168.9 ± 6.5	162.3 ± 6.1	0.569
Bone geometry
Cortical area (mm2)	40.6 ± 2.7	48.6 ± 2.7	48.1 ± 2.6	0.712
Cortical thickness (mm)	0.60 ± 0.04	0.74 ± 0.04	0.73 ± 0.04	0.033*
Trabecular area (mm2)	224.9 ± 8.7	204.5 ± 8.7	199.4 ± 8.2	0.089
Bone microarchitecture
Trabecular number (1/mm)	1.95 ± 0.05	2.01 ± 0.05	1.91 ± 0.05	0.350
Trabecular separation (mm)	0.45 ± 0.01	0.43 ± 0.01	0.46 ± 0.01	0.269
Trabecular thickness (mm)	0.070 ± 0.002	0.070 ± 0.002	0.070 ± 0.002	0.716
12-Month change in body composition measures, mean ± SE
Change in height (cm)	0.37 ± 0.30	0.10 ± 0.14	0.30 ± 0.11	0.613
Change in weight (kg)	1.94 ± 0.86	0.86 ± 0.68	2.05 ± 0.53	0.427
Change in BMI (kg/m2)	0.62 ± 0.27	0.32 ± 0.25	0.70 ± 0.19	0.504
Change in fat mass (kg)	0.45 ± 0.59	1.29 ± 0.47	1.33 ± 0.40	0.396
Change in lean mass (kg)	0.86 ± 0.33	0.01 ± 0.34	0.80 ± 0.33	0.151

BMI = body mass index; HA = hydroxyapatite; SHBG = sex hormone binding globulin; IGF-1 = insulin-like growth factor-1.

^*p < 0.05 for PATCH versus PILL.

Fig. 1.

Flowchart of oligoamenorrheic athletes recruited for this study and attrition according to randomization group.

Changes in aBMD over 12 months

Change and percent change in lumbar spine, femoral neck, and total hip BMD in this subset reflected reported changes at these sites in the parent study⁽²⁴⁾ and are not reported again.

Changes in HR-pQCT parameters over 12 months

Weight-bearing distal tibia

Table 2 shows changes at 12 months in HRpQCT results at the distal tibia after controlling for age, height, race, and ethnicity. The PATCH versus PILL group demonstrated greater percent increases in total vBMD (2.4% ± 1.0% versus 0.5% ± 1.0%, p = 0.010), trabecular vBMD (1.6% ± 1.1% versus 0.1% ± 1.1%, p = 0.069 [p = 0.037 for absolute change]), cortical area (3.1% ± 1.8% versus −0.3% ± 1.8%, p = 0.007), cortical thickness (0.03% ± 0.02% versus −0.00% ± 0.02%, p = 0.020), and trabecular number (5.1% ± 4.3% versus −0.7% 4.3%, p = 0.076 [p = 0.043 for absolute change]) (Fig. 2A,B). Further, PATCH had greater percent reductions than PILL in trabecular area (−0.5% ± 0.3% versus −0.10% ± 0.3%, p = 0.040) and separation (−5.4% ± 4.3% versus 1.2% ± 4.3%, p = 0.040). PATCH versus NONE had greater percent increase in cortical area (3.1% ± 1.8% versus 0.5% ± 1.9%, p = 0.044) and decrease in trabecular area (−0.5% ± 0.3% versus −0.2% ± 0.3%, p = 0.076 [p = 0.036 for absolute change]) (Fig. 2B).

Table 2.

Twelve-Month Changes in Tibia HRpQCT Bone Microarchitecture in Participants Randomized to the PATCH, PILL, or NONE (Controlled for Age, Height, Race, and Ethnicity)

Distal tibia	LSC	PATCH (n = 24)	PILL (n = 24)	NONE (n = 27)	p (whole model)	PATCH versus PILL	PATCH versus NONE
Absolute changes in vBMD
Change in total vBMD (mg HA/cm3)	1.8	7.01 ± 3.06	1.17 ± 3.05	3.71 ± 3.33	0.018	0.009	0.160
Change in cortical vBMD (mg HA/cm3)	9.3	8.30 ± 5.06	3.00 ± 5.05	3.95 ± 5.52	0.238	0.194	0.295
Change in trabecular vBMD (mg HA/cm3)	2.2	3.27 ± 2.17	−0.11 ± 2.17	2.01 ± 2.37	0.061	0.037	0.554
Absolute changes in bone geometry
Change in cortical thickness (mm)	0.01	0.03 ± 0.02	−0.00 ± 0.02	0.01 ± 0.02	0.037	0.023	0.123
Change in cortical area (mm2)	1.0	2.93 ± 1.73	−0.46 ± 1.73	0.72 ± 1.89	0.013	0.007	0.084
Change in trabecular area (mm2)	1.1	−2.74 ± 1.50	−0.17 ± 1.50	−0.47 ± 1.64	0.021	0.021	0.036
Absolute changes in bone microarchitecture
Change in trabecular number (1/mm)	0.43	0.08 ± 0.08	−0.03 ± 0.08	0.06 ± 0.08	0.049	0.043	0.876
Change in trabecular separation (mm)	0.13	−0.027 ± 0.02	0.000 ± 0.02	−0.021 ± 0.02	0.112	0.085	0.822
Change in trabecular thickness (mm)	0.02	−0.004 ± 0.004	0.000 ± 0.004	−0.003 ± 0.004	0.172	0.132	0.858

Values are least square means ± SE. Bold values are significant. Comparisons (PATCH versus PILL) and (PATCH versus NONE) performed using Dunnett’s Values are least square means ± SE. Bold values are signi adjustment for multiple comparisons.

HA = hydroxyapatite; LSC = least significant change.

Fig. 2.

Percent change in compartmental volumetric BMD and bone geometry parameters at the distal tibia (A and C) and distal radius (B and D) in oligoamenorrheic athletes randomized to the PATCH, PILL, or no-estrogen (NONE).

After also controlling for baseline measures (in addition to age, height, race, and ethnicity), differences between PATCH and PILL groups persisted for changes in total vBMD (p = 0.02), cortical area (p = 0.01), trabecular area (p = 0.02), and cortical thickness (p = 0.04), but differences were attenuated to a trend level for trabecular vBMD and trabecular number (p < 0.10). Differences between PATCH and NONE groups persisted for change in trabecular area (p = 0.03).

Similarly, when we added menarcheal age to the multivariable regression model, differences between PATCH and PILL groups for total vBMD (p = 0.04), cortical and trabecular area (p = 0.02 and p = 0.04), and cortical thickness (p = 0.048) persisted, as did the differences for percent change in total vBMD, cortical area, and thickness. However, differences between groups for trabecular vBMD and number were lost after controlling for menarcheal age.

Non–weight-bearing radius

Table 3 illustrates the changes over 12 months in HRpQCT parameters at the distal radius. After controlling for age, height, race, and ethnicity, percent changes in cortical vBMD (4.6% ± 1.5% versus 2.4% ± 1.5%, p = 0.047), area (22.2% ± 12.4% versus 2.4% ± 12.4%, p = 0.030), and thickness (20.9% ± 9.4% versus 5.6% ± 9.4%, p = 0.025) were greater in the PATCH versus PILL groups (Fig. 2). Percent changes in cortical area (22.2% ± 12.4% versus 3.3% ± 13.5%, p = 0.032) and thickness (20.9% ± 9.4% versus 5.8% ±10.2%, p = 0.022) were greater in the PATCH versus NONE groups (Fig. 2C,D). Baseline cortical vBMD and thickness at the radius differed among groups (Table 1). Controlling for baseline, cortical vBMD attenuated differences among groups for changes in cortical vBMD over time. However, increases in cortical thickness trended higher in the PATCH versus PILL group, even after controlling for baseline cortical thickness (p = 0.067). These differences were lost after controlling for menarcheal age.

Table 3.

Twelve-Month Changes in Radius HRpQCT Bone Microarchitecture in Participants Randomized to the PATCH, PILL, or NONE (Controlled for Age, Height, Race and Ethnicity)

Distal radius	LSC	PATCH (n = 24)	PILL (n = 23)	NONE (n = 26)	p (whole model)	PATCH versus PILL	PATCH versus NONE
Absolute changes in vBMD
Change in total vBMD (mg HA/cm3)	13.0	13.03 ± 4.92	6.72 ± 4.92	9.43 ± 5.35	0.145	0.090	0.393
Change in cortical vBMD (mg HA/cm3)	27.0	31.48 ± 9.50	17.79 ± 9.49	20.29 ± 10.33	0.066	0.052	0.112
Change in trabecular vBMD (mg HA/cm3)	6.0	1.78 ± 2.34	−0.71 ± 2.34	−0.20 ± 2.54	0.227	0.177	0.299
Absolute changes in bone geometry
Change in cortical thickness (mm)	0.06	0.083 ± 0.03	0.026 ± 0.03	0.038 ± 0.04	0.023	0.017	0.058
Change in cortical area (mm2)	3.7	6.07 ± 4.04	0.45 ± 4.04	1.35 ± 4.39	0.076	0.062	0.115
Change in trabecular area (mm2)	2.5	88.0 ± 37.7	37.6 ± 37.7	50.0 ± 41.1	0.106	0.076	0.190
Absolute changes in bone microarchitecture
Change in trabecular number (1/mm)	0.31	−0.028 ± 0.08	−0.061 ± 0.08	−0.049 ± 0.09	0.831	0.767	0.888
Change in trabecular separation (mm)	0.08	0.007 ± 0.02	0.011 ± 0.02	0.005 ± 0.02	0.869	0.904	0.989
Change in trabecular thickness (mm)	0.01	0.0015 ± 0.003	0.0016 ± 0.003	0.0013 ± 0.003	0.984	0.997	0.991

Data are reported as least square means ± SE. Bold values are significant. Comparisons (PATCH versus PILL) and (PATCH versus NONE) performed using Data are reported as least square means ± SE. Bold values are signi Dunnett’s adjustment for multiple comparisons.

HA = hydroxyapatite; LSC = least significant change.

Changes in biochemical parameters over 12 months

Changes in vitamin D, estradiol, and SHBG levels and IGF-1 Z-scores have been reported for the full cohort,^(24,31) and are reported here for the subset. Changes in 25(OH) vitamin D levels over 12 months in this subset did not differ among groups (p = 0.55). The PATCH group had greater increases in estradiol compared to the PILL and NONE groups (mean [IQR], 42.0 [7.0, 57.7) versus −17.5 [−81.8, 0.0] pg/mL and −5.0 [−14.0, 52.7] pg/mL, p < 0.05 for both), lesser increases in SHBG (1.0 ± 3.8 versus 155.6 ± 22.1 nmol/L, p < 0.0001), and lesser reductions in IGF-1 Z-scores (−0.02 ± 0.17 versus −0.61 ± 0.22, p = 0.050) compared with the PILL group. The PILL group had greater increases in SHBG than both other groups, similar to the parent study (p < 0.0001).⁽²⁴⁾ Together this suggests more bioavailable estradiol in the PATCH group compared to both PILL and NONE groups during the study, and overall, good compliance with study medications in the PATCH and PILL groups.

At 12 months, at the distal tibia, changes in estradiol levels were associated with changes in total vBMD (r = 0.35, p = 0.005) (Fig. 3A), trabecular vBMD (r = 0.28, p = 0.027) (Fig. 3B), cortical thickness (r = 0.41, p = 0.0009) (Fig. 3C), cortical area (r = 0.43, p = 0.0005) (Fig. 3D), trabecular area (r = −0.43, p = 0.0004) (Fig. 3E), trabecular number (r = 0.32, p = 0.009), and trabecular separation (r = −0.28, p = 0.024) (Fig. 3F).

Fig. 3.

Associations of changes in estradiol with changes in volumetric BMD (A and B), bone geometry (C, D, and E), and microarchitectural parameters (F) at the distal tibia.

Adverse events

Subjects were asked about adverse events throughout the study, during phone calls and at follow-up visits. There were no significant differences among the groups for adverse events, including no differences in headache, breast tenderness, bloating, skin rash, or fracture (Table 4).⁽²⁴⁾

Table 4.

Reported Adverse Events Among Oligoamenorrheic Athletes Randomized to the Estrogen PATCH, or PILL, or NONE (for Those Who Had an HRpQCT at Baseline)

Adverse events	PATCH (n = 24) (%)	PILL (n = 24) (%)	NONE (n = 27) (%)	Fisher’s exact test p
Likely/probably related to study medications or procedures
Headaches	2.4	2.8	0	0.628
Dizziness	2.4	0	0	0.426
Nausea	0	5.6	0	0.113
Fatigue	0	2.8	0	0.340
Breast tenderness	2.4	2.8	0	0.628
Fluid retention	0	2.8	0	0.340
Bloating	2.4	0	0	0.426
Early menses	0	2.8	0	0.340
Yeast infection	0	2.8	0	0.340
Worsening depression	2.4	0	0	0.426
Mood swings	0	2.8	0	0.340
Generalized discomfort	2.4	0	0	0.426
Vasovagal response	0	5.6	0	0.113
Syncopal episodes	0	2.8	0	0.340
Irritation from the patch	7.1	0	0	0.074
Skin rash	2.4	0	0	0.426
Possibly related to study medications or procedures
Decreased libido	0	2.8	0	0.340
Diarrhea	0	0	2.9	0.325
Gastritis	2.4	0	0	0.426
Unrelated to study medications or procedures
Mononucleosis	0	2.8	0	0.340
Fracture	2.4	0	0	0.426
Iliofemoral sprain	2.4	0	0	0.426
Labrum tear	2.4	0	0	0.426
Sinus surgery	0	0	2.9	0.325
Upper respiratory symptoms	0	0	2.9	0.325
Skin tags	0	2.8	0	0.340
Electrolyte changes	0	2.8	0	0.340

Discussion

This is the first study to demonstrate the efficacy of transdermal 17β-E2 replacement in improving vBMD, bone geometry, and bone microarchitecture in OA, a population at high risk for stress fractures from overuse and hormonal suppression, and long-term consequences of suboptimal peak bone mass acquisition.^(6,16) In fact, it is the first study to investigate the effects of estrogen administration in any form on HRpQCT measurements in a cohort of premenopausal women. Our previously published work demonstrated that OAs receiving transdermal estradiol with cyclic oral progesterone (PATCH) had an increase in lumbar spine, femoral neck, and total hip aBMD at 6 and 12 months (as measured by DXA) compared to those receiving ethinyl estradiol and a progestogen via a standard COC (PILL).⁽²⁴⁾ The PATCH group also had greater increases in lumbar spine and femoral neck BMD than the no-estrogen group at 12 months.⁽²⁴⁾ Our current study builds on that work to show superior effects of transdermal 17β-E2 replacement compared to COC on peripheral bone density and microarchitecture, particularly at the weight-bearing distal tibia.

Our study shows that at the weight-bearing tibia, the PATCH compared to the PILL group had greater increases in total and trabecular vBMD, cortical area and thickness, and trabecular number. The PATCH versus PILL group also had a greater percent decrease in trabecular separation, consistent with an increase in trabecular number. The PATCH group did have an increase in cortical area, but a decrease in trabecular area compared to the other two groups.

The decrease in trabecular area was expected given that estradiol reduces endosteal bone resorption.⁽³⁶⁾ Animal and human studies support that cortical and trabecular bone are regulated differently by various cell types (eg, osteoblast progenitors and osteoclasts), resulting in contrasting responses to hormonal and mechanical signaling.⁽³⁷⁾ Of note, we expected trabecular area to decrease more in the PILL group as well. The lack of this effect in the PILL group needs to be studied further, but may reflect significant increases in SHBG in the PILL group that were not seen in the other two groups, resulting in a reduction in the amount of bioavailable estrogen.⁽³⁸⁾ In general, effects of estrogen appear to be attenuated in adolescents/young adults compared to older women based on prior studies.^(39,40) This may explain why we did not see more robust effects on trabecular vBMD and other trabecular components. We also found reductions in preadipocyte factor-1 and brain-derived neuro-rophic factor within the PATCH group that were not observed in the PILL group.⁽³¹⁾ It is unclear if these changes have differential effects on cortical versus trabecular bone.

We previously found that in contrast to the non–weight-bearing radius, the weight-bearing tibia was slightly more protected from bone deficits in amenorrheic exercisers.^(11,14) Of the improvements in tibial HRpQCT parameters observed in our current study, only total vBMD, trabecular number, and trabecular spacing differed cross-sectionally in amenorrheic versus eumenorrheic athletes and non-athletes.⁽¹¹⁾ Improvements in multiple tibial parameters with physiologic estrogen replacement has implications for reduced fracture risk.^(14,16,41)

At the non–weight-bearing radius, the PATCH group had greater gains over 12 months in cortical area and thickness compared to the other two groups. The more pronounced effects of PATCH at the tibia compared to the radius likely reflect the interaction between mechanical loading and estrogen on the skeleton, with estrogen having permissive effects on the skeletal impact of mechanical loading.⁽³⁶⁾ In a previously reported cross-sectional study, radial cortical area and thickness were noted to be lower in amenorrheic athletes than eumenorrheic athletes and non-athletes.⁽¹¹⁾ Our present study suggests that physiologic estrogen and progesterone replacement may improve such differences. These results differ somewhat from those reported by Farr and colleagues⁽⁴²⁾ in a different population: recently postmenopausal women. Subjects received placebo, 0.45 mg/d equine estrogens orally, or 50 μg/d 17β-E2 transdermally. Both estrogen groups received 200 mg micronized progesterone for 12 days of each month. After 4 years, cortical vBMD decreased and porosity increased in women receiving placebo. Estradiol given orally or transdermally prevented these changes. However, there were no differences in changes in radial cortical area or thickness across groups.⁽⁴²⁾ This is the only other study to date that has examined effects of hormonal treatment on HRpQCT measured parameters in an estrogen-deficient population, with obvious differences in study design (including type and dose of estrogen), study duration, and study population. Beyond this, we are only aware of 12 studies that have examined effects of any treatment on HRpQCT-measured bone outcomes in women.⁽⁴³⁾

As we have previously reported for the effects of PATCH versus PILL or NONE on DXA measures, the fact that the PATCH is more beneficial for HRpQCT parameters can likely be explained by the fact that (i) transdermal 17β-E2 does not experience first-pass metabolism at the liver and thus does not suppress IGF-1 secretion, a bone-trophic hormone downregulated by COCs⁽⁴⁴⁾ (as also noted in the current sample); (ii) higher SHBG levels in the PILL group^(34,45) may further reduce bioavailable estradiol, essential for bone accrual and maintenance⁽³⁶⁾; and (iii) the formulation of COC used for this study (30 μg ethinyl estradiol, 0.15 mg desogestrel) resulted in a decrease in serum estradiol versus the PATCH group, which had an increase in estradiol compared to baseline. The last likely reflects the estradiol assay, which measures17β-E2, and not ethinyl estradiol, the estrogen formulation in the COC; however, 17β-E2 may be more effective in optimizing bone outcomes. A study of premenopausal women who received a COC containing 17β-E2 and nomegestrol acetate versus a COC containing ethinyl estradiol/levonorgestrel attempted to address this. After 2 years, there were no differences between groups for bone outcomes.⁽⁴⁶⁾ In our study, we found significant associations of changes in serum estradiol levels over time with changes in HRpQCT measures of vBMD, bone geometry, and structure.

Further studies are necessary to determine whether lower doses of ethinyl estradiol may cause lesser increases in SHBG, and therefore cause different impacts on bone. Cibula and colleagues⁽⁴⁷⁾ reported lower BMD with the 15-μg vs. the 30-μg ethinyl estradiol COC formulation in adolescent girls. Scholes and colleagues⁽⁴⁸⁾ found a trend for lower BMD with longer duration of use and lower estradiol dose formulations in young women. However, others have not found BMD effects of different COC estradiol dosages.^(49,50) More work is needed comparing COCs with various estradiol formulations and doses, but similar progesterone/progestogens, to elucidate the impact of estradiol formulation and dose on bone.

Of note, the PATCH group received 17β-E2 continuously, while the COC group received pills in the standard recommended fashion with 21 days of active and 7 days of placebo pills. Overall greater estrogen exposure in the PATCH versus PILL groups may have impacted outcomes more favorably. However, worse outcomes for changes in cortical area at the tibia and radius in PILL versus PATCH and NONE groups suggest possibly detrimental effects of COCs on bone microarchitecture, which may have been worse with continuous use of active pills.

Although it is commonly discussed that estrogen acts on bone by inhibiting bone resorption, animal cell culture studies have demonstrated that 17β-E2 may enhance bone formation at physiologic doses, and potentially bone resorption at higher doses.^(51,52) However, thus far, ethinyl estradiol has been shown to inhibit osteoclastic bone resorption in some, but not all animal tissue and human studies, and has not demonstrated an effect on bone formation.^(53–55) In two human studies, ethinyl estradiol (via COC and via transdermal patch) attenuated bone accrual in adolescents/young women.^(56,57) In our study, as we previously reported, over the 12 months, N-terminal propeptide of type 1 procollagen, a marker of bone formation, decreased the most in the PILL group, but again, this was associated with a decrease in IGF-1 levels.⁽³¹⁾ Thus, not only route, but type of estrogen replacement may be important for metabolic bone effects.

Studies suggest a possible role of progesterone/progestins on bone metabolism. In a meta-analysis, Seifert-Klauss and Prior⁽⁵⁸⁾ indicated that regular ovulation (associated with increases in progesterone) likely contributes to peak BMD. Additionally, oral cyclic medroxyprogesterone acetate (MPA) increases BMD in healthy, normal-weight, physically active premenopausal women with oligoamenorrhea or subclinical ovulatory disturbances.^(58,59) However, in a small, 1-year study of oligoamenorrheic teenagers who were underweight and/or had anorexia nervosa, oral cyclic MPA did not prevent bone loss.^(58,60) Further, depot MPA leads to bone loss in adolescent and adult premenopausal women.^(61–67) No studies to date have examined effects of oral progesterone alone on BMD in young women. Thus, it is unclear whether 12 days of micronized progesterone each month in the PATCH group may have impacted bone differently versus 21 days of desogestrel every month in the COC group. It is also important to note that we used two different types of progesterone; micronized progesterone (PATCH) commonly prescribed for hormonal replacement therapy and desogestrel (PILL) a type of progesterone found in some COC pills. The PATCH micronized progesterone dose (200 mg) was much higher than the PILL desogestrel dose (0.15 mg), but the active form of desogestrel has a higher affinity for the progesterone receptor than micronized progesterone.^(68,69) It is unclear if type and potency of progesterone contributed to our findings.

Although it is important for patients with Triad to focus on increasing EA to improve hormonal function (which will result in improving bone health), many patients are resistant to such changes in lifestyle. Triad treatment typically involves a team of specialists (eg, physician, dietitian, psychologist) to aid in such behavioral changes.⁽⁶⁾ A “normal weight” athlete may find it just as disturbing to gain weight as a traditionally lower-weight patient with anorexia nervosa due to body image issues, and societal and sport pressures.^(70,71) To decrease bone loss and perhaps enhance bone accrual, physiologic transdermal 17β-E2 with cyclic progesterone may be an appropriate adjunct treatment for these young women. This is particularly important during the adolescent years of peak bone accrual.⁽⁷²⁾

There are limitations to our study. Trends in differences in various HRpQCT parameters were noted, as expected, but often did not reach a significance of p < 0.05. More participants in each treatment arm may have improved the significance of some findings as may a longer duration of study. Additionally, although BMD improvements with PATCH have been reported,⁽²⁴⁾ the relatively new usage of HRpQCT in premenopausal women makes interpreting the bone microarchitecture changes in our current study slightly more difficult. However, based on existing literature, we suspect that it is an improvement to see increases in cortical area and thickness and decreases in trabecular area.^(16,73) Our data for strength estimates are pending micro-finite element analysis using registered images. In older adults, Samelson and colleagues⁽⁷³⁾ demonstrated correlations between HRpQCT measurements and fracture risk, independent of DXA-measured BMD. Our group found correlations with history of stress fractures and BMD and bone microarchitecture in OAs.⁽¹⁶⁾ Further prospective studies are needed to determine if the changes in bone microarchitecture from PATCH correlate with decreased future fracture risk. Additionally, we utilized one dose of transdermal 17β-E2 and one dose and type of COC. Estradiol levels measured during treatment in the PATCH group were in the range that one aims for with hormonal replacement therapy,⁽⁷⁴⁾ but other doses may need to be considered. The ethinyl estradiol dose in the PILL was a standard dose that is reasonably representative of COC use in the general population,⁽⁷⁵⁾ but again, different doses and formulations may yield slightly different results. Further studies are necessary to compare other transdermal 17β-E2 and COC formulations and determine the role of progesterone (if any) on bone endpoints. Another consideration is that although we required study participants to be off COCs or other hormonal contraception for at least 3 months before study entry, we do not have the history of lifelong exposure to these substances in our participants. However, given the randomized nature of treatment allocation, we do not expect this to differ among groups. Finally, we do not have data for fracture prevalence, and larger studies are necessary to address this limitation.

Strengths of this work include the following: (i) participants were all endurance, weight-bearing OAs, who met specific activity type and volume criteria (to limit lifestyle confounding effects); (ii) we used a state-of-the art modality (HRpQCT) to assess small differences in bone to better understand treatment effects; and (iii) we monitored compliance and effects with questionnaires and laboratory testing.

In conclusion, this is the first study to examine effects of transdermal physiologic estrogen replacement with cyclic progesterone on HRpQCT measures in OA. We compared this treatment, not only to no treatment, but also to a typically prescribed COC. We found that transdermal 17β-E2 was superior to a typical COC in improving many bone outcomes and may be superior in some aspects compared to no therapy as well. These results, along with the improvements in aBMD⁽²⁴⁾ and bone turnover markers⁽³¹⁾ we have reported in PATCH versus PILL and PATCH versus NONE, suggest an overall superiority of transdermal 17β-E2 as an adjunct treatment to improve skeletal health in female athletes with oligoamenorrhea.