Analysis of low-dose estrogen on fracture healing measured by pQCT in postmenopausal women

The skeletal disease osteoporosis affects primarily postmenopausal women. It is characterized by increased fracture risk due to defective remodeling of bone tissue and a pathological bone microarchitecture in response to the decreased levels of estrogen. Osteoporosis is clinically silent up until a rst fracture occurs. Therefore, the diagnosis of the disease is usually late. Recent preclinical studies with mice, rats and rabbits suggested that estrogen treatment enhances the healing effect after bone fractures. Here we asked whether transdermally applied estrogen, which restores the systemic premenopausal level of estrogen, has a benecial effect on fracture healing of postmenopausal women as observed in preclinical models. Methods


Abstract Background
The skeletal disease osteoporosis affects primarily postmenopausal women. It is characterized by increased fracture risk due to defective remodeling of bone tissue and a pathological bone microarchitecture in response to the decreased levels of estrogen. Osteoporosis is clinically silent up until a rst fracture occurs. Therefore, the diagnosis of the disease is usually late. Recent preclinical studies with mice, rats and rabbits suggested that estrogen treatment enhances the healing effect after bone fractures. Here we asked whether transdermally applied estrogen, which restores the systemic premenopausal level of estrogen, has a bene cial effect on fracture healing of postmenopausal women as observed in preclinical models.

Methods
To investigate whether estrogen has a bene cial effect on bone fracture healing of postmenopausal patients, we performed a prospective double-blinded randomized study with 76 patients suffering from distal radius fractures. A total of 31 patients (71.13 years ± 11.99) were treated with estrogen and 45 patients (75.62 years ± 10.47) served as untreated controls. Calculated bone density as well as cortical bone density were determined by peripheral quantitative computed tomography (pQCT) prior to and six weeks after the surgery.

Results
We found that unlike with preclinical models, bone fracture healing of human patients was not improved in response to estrogen treatment. Furthermore, we observed no dependence between age-dependent bone tissue loss and constant callus formation in the patients.

Conclusions
Transdermally applied estrogen to postmenopausal women, which results in estrogen levels similar to the systemic level of premenopausal women, has no signi cant bene cial effect on bone fracture healing.

Background
The distal radius fracture is one of the ten most expensive medical incidents worldwide and the most common osteoporotic fracture in women over 50 years of age as announced by the World Health Organization [1,2,3]. The most important risk factor for postmenopausal osteoporosis appears to be ovarian hormone de ciency [4]. Premenopausal women experience bone loss after ovariectomy [5] suggesting that reduced ovarian hormone levels participate in the cause of the disease. In fact, morphology ical and metabolic studies have shown that one of the ovarian hormones, estrogen, has both anabolic and anticatabolic effects on the osteoblastic and osteoclastic processes of bone remodeling [6,7]. In preclinical model organisms, such as mice, rats and rabbits, estrogen was shown to markedly affect all phases of bone fracture healing, in particular during the later stages of the healing process [4,8,9]. In mice for example, estrogen de ciency causes an impaired fracture repair with decelerated remodeling of the periosteal callus and formation of a porous neocortex [4]. Conversely, estrogen treatment of mice with bone fractures led to enhanced healing, i.e. the chondrocyte areas were larger, callus mineralization was increased, and the neocortex was thicker when compared to untreated control animals [4,8,9,11,12,13,14]. Furthermore, estrogen treatment also led to bene cial effects on the biomechanical properties of the bones [4]. These observations are in line with the observation that osteoporotic patients sustain a prolonged and impaired healing process after bone fracture when compared to non-osteoporotic patients [10].
Encouraged by these preclinical results and the knowledge of reduced sensitivity of estrogen in old age, we decided to critically evaluate the in uence of estrogen on bone fracture healing in a prospective randomized double-blinded and placebo-controlled trial in postmenopausal woman who suffered from a distal radius fracture. To avoid possible side effects of hormone replacement therapy (HRT) treatment, we chose a low-dose regimen and examined the effect on bone healing as measured by peripheral quantitative computed tomography (pQCT). We have chosen pQCT for this analysis because it allows the differentiation between cortical and trabecular bone. This distinction was necessary because the trabecular bone, which covers a larger area, reacts faster to changes in bone metabolism than the cortical bone. Also, with pQCT bone loss or the success of a therapy can be demonstrated at an earlier stage and with greater signi cance than with methods that cannot distinguish between cortex and cancellous bone [15]. We therefore applied the pQCT technology to score the effect of the post-surgical treatment with transdermally applied estrogen.

Patient collective for study
Inclusion criteria were de ned as female postmenopausal patients at least 50 years of age who sustained a single distal radius fracture (AO-type A3 or C1-3; see Fig. 1a) requiring extrafocal k-wiring and mounting of an external xateur. Exclusion criteria were malignoma or medical treatment with bisphosphonates or other drugs affecting bone turnover.
The participants were assigned to placebo or verum group following the randomization sequence created by the study biostatistician. All study personnel remained blinded to the subject treatment assignment. Randomized subjects were assigned to receive either estrogen or a corresponding placebo plaster.
Estrogen was applied via a transdermal plaster containing 4 mg estradiol (Estramon® Uno 50, Hexal AG, 83607 Holzkirchen, Germany). Four plasters were handed over to the patient. Each plaster was applied for one week (this corresponds to a daily dose of 0,571 mg) resulting in a total treatment time of four weeks.

Page 4/22
The proper application was demonstrated to the patient when the rst plaster was set. The patients of the control group received the corresponding placebo plaster which could not be distinguished from the estradiol plasters.
Before surgical treatment of the fracture, and four weeks later, blood samples were collected and estrogen levels were measured. After taking the rst blood sample, the patients both groups received daily 1,000 mg calcium and 880 IU vitamin D (25-[OH]-vitamin D) over a period of six weeks to avoid any in uence of a lack of calcium or vitamin D. The calcium, parathyroid hormone (PTH), vitamin D and alkaline phosphatase serum levels were analyzed before surgical treatment and six weeks postoperatively. The calcium, parathyroid hormone (PTH) and vitamin D values in the serum of patients of both groups have already been published in a study by Kolb et al. (2013) [16].
The present study was approved by the ethics committee of the University Medical Center Hamburg-Eppendorf (approval number: AZ 2252).

Surgical intervention
Within an interval of 48 h after trauma, patients were surgically treated by k-wiring (two 1.6 mm titanium k-wires) and mounting of an external xateur (HoffmannTM Fixateur, Stryker Howmedica, Mahwah, NJ, USA). According to our standard clinical care, X-ray controls in anterior-posterior and lateral projections were acquired prior surgery, postoperatively and after two and six weeks (see Fig. 1b and c). The X-ray controls only served to verify the correct implant position and not for tracking any kind of callus index.
The external xateur and the k-wires were removed six weeks after the surgery.
pQCT Measurements pQCT allows the differentiation between cortical and trabecular bone. Because of the larger area, the trabecular bone reacts faster to changes in bone metabolism than the cortical bone. Bone loss or the success of a therapy can therefore be demonstrated at an earlier stage and with greater signi cance with pQCT than with methods that cannot distinguish between cortex and cancellous bone [15]. We applied therefore the pQCT technology to score the success of the post-surgical treatment with transdermally applied estrogen.
pQCT scans, (XCT-2000; Stratec Medizintechnik, Pforzheim, Germany) of the fractured and contralateral non-fractured distal radius were performed within de ned intervals: postoperatively, after two and four weeks as well as after removal of the external xateur and k-wires after six weeks. The device was calibrated using a standard phantom and a cone phantom provided by the manufacturer. The settings selected for the pQCT analysis were adopted according to the manufacturer's recommendations as previously described [16,17,18,19,20].
The forearm length was de ned as the distance between the ulna styloid process and olecranon. Measurements were taken while forearm was supinated and elbow exed at 90 degrees. A 2-mm-thick single tomographic slice with pixel size 0.59 × 0.59 mm was taken directly in the fracture plane of the distal radius using the standard mode (see Fig. 1d). Image processing and calculation were performed using the manufacturers software package CXCT550 (version 5.50D). The bone mineral density (BMD) was de ned as mean density of the total cross section.
The k-wires which were used to position and x the bone fracture after surgery interfered with the pQCT measurements. Control measurements at the − 4% position from the fractured site could also not been taken into account, since the fractures of the patients were at variable positions of the radius, and thus the − 4% sites were at different positions dependent on the fractured site. Thus, the k-wire xatives precluded a reliable quantitative analysis during the initial and early bone healing process. pQCT measurements at the fractured site and at the − 4% control position were therefore rst examined six weeks post-surgery and after the k-wires were removed and analyzed together with the corresponding Xray images.

Determination of Calculated Bone Density (CALCBD)
Based on the protocol of the manufacturer, the calculated bone density (CALCBD), which determines the total and cancellous bone density was acquired in three steps: 1. A region of interest (ROI) was de ned, which was the region tight to the outer contour of the radius and was individually determined by us on the device. 2. Within this ROI, based on a soft tissue threshold (280 mg/cm³), each pixel outside the bone window was de ned and removed. This left the bone part of the ROI from which the total area and the total density (CALCBD) was derived. 3. The total area was concentrically divided in the outer 55%, which was de ned as the cortical and subcortical area, and the inner 45%, de ned as trabecular area.
The total CALCBD levels were determined at different times and at different locations of the arm. The measurements on the healthy side were performed directly postoperatively (zero weeks) and after six weeks at -4% of the total length of the radius. The measurements on the fractured side were carried out postoperatively and at the time points four and six weeks at two different heights -namely standardized at -4% of the total length of the radius above the fracture gap and directly on the fracture height (fracture gap), which was determined on the basis of the X-ray images.

Determination of Cortical Bone Density (CORTBD)
The cortical bone density (CORTBD) is a measure of cortical bone structure in the diaphysis. For its determination, all pixels within the ROI with a lower density than the threshold (here 710 mg/cm³) were removed. From the remaining pixels, the cortical area and cortical mineral density were computed.

Examination of the serum levels of diagnostic molecules
Bone-speci c alkaline phosphatase activity (U/l) in the blood samples of patients was measured immediately after the surgery and six weeks later (see Fig. 2a), at the Central Laboratory of the University Medical Center Hamburg-Eppendorf as described before [21]. Calcium (mmol/l), parathyroid hormone (PTH) (ng/ml) and vitamin D (µg/l) serum levels were analyzed before surgical treatment and six weeks postoperatively [21].

Statistics
Statistical analysis was performed using a student's t-test for age comparison and for the data comparing estrogen-treated and placebo patients. Dependent on the outcome variable, either linear regression or mixed-effects models were used to analyze the effect of the different independent variables.
The signi cance level was set to alpha = 5% for all statistical tests. All analyses were performed with the statistics software R (version 3.4.0, www.r-project.org) using the R-package lme4 for the mixed-effects models.

Results
A total of 76 patients with complete clinical datasets were included in this double-blinded randomized study. Of those, 31 patients (mean age: 71.13 years ± 11.99 years) were treated with estrogen and 45 patients (mean age: 75.62 years ± 10.47 years) were left in the placebo group. The age difference between patients of the two groups was not statistically signi cant (p = 0.1006; see Table 1). The dominant arm was fractured in 29.03% of the patients of the estrogen treated group, and in 53.33% of the patients of the placebo group (see Table 1). To determine the effect of transdermal estrogen treatment, we rst examined the estrogen levels in blood samples of patients immediately after the surgery and four weeks post-surgery, respectively. Figure 2b shows box plots of the estrogen levels in the blood of estrogen-treated and placebo-treated patients immediately after surgery and four weeks later. The estrogen values of all patients prior to the estrogen application or the placebo-treatment were in the range of ≤ 30 ng/l, indicating that all individuals were in post menopause [21]. Four weeks after the surgery, the estrogen concentration was signi cantly increased only in the estrogen-treated patients (16.7 ng/l ± 4.9 ng/l vs. 34.5 ng/l ± 5.1 ng/l; p *** = 0.0003; Fig. 2b), while no signi cant increase was observed in the patients of the placebo group (10.0 ng/l ± 4.8 ng/l vs. 10.2 ng/l ± 4.3 ng/l; p = 0.8071; Fig. 2b). The results indicate that the four weeks of transdermal estrogen-treatment increased the estrogen levels signi cantly (increase between 12.0 and 25.0 ng/l) in 18 patients (58.06%) and only slightly in the others. In contrast, the estrogen values of the untreated control patients remained low, i.e. <30 ng/l. In the estrogen-treated group, 28 out of 31 patients had estrogen blood levels in a range commonly observed in premenopausal women. Two patients had signi cantly higher estrogen levels than normal and in one case, the levels were lower because the patch had been fallen off during the process of the study (see Fig. 2b).
Even though transdermal estrogen treatment increased the systemic estrogen levels of the patients, we did not observe an effect of this treatment on bone fracture healing in postmenopausal women. We compared the healing process by measuring the total calculated bone density (CALCBD) and cortical bone density (CORTBD) by peripheral quantitative computed tomography (pQCT) at three different sites of the patients. These sites were the fractured site, a position which is 4% of the bone length of the radius above the fracture gap (-4% level) and the corresponding position of the non-fractured healthy arm. Measurements were taken from estrogen treated and placebo treated control patients..Measurements at the − 4% position of the fractured arm were only taken into account six weeks after the surgery and after both k-wires and xative, since they interfere with measurements (see Methods).
Calculated bone density and cortical bone density in estrogen versus placebo treated patients The k-wires used to x the bone fracture after surgery interfered with the pQCT measurements and thus, they precluded a reliable quantitative analysis during the initial and early bone healing process. For the fractured side, pQCT measurements were therefore only performed six weeks post-surgery when both the external xateur and the k-wires were removed. At that time point, we measured the total calculated bone density (CALCBD) of the estrogen and placebo treated patients. Figure 3a1 shows total CALCBD content of the healthy arm at different time points of the estrogen-treated patients and the patients of the placebo group (Fig. 3a1) Figure 3a2 shows the CALCBD content at the fractured site of the patients six weeks after the surgery (see also  Figure 3b1 shows the CORTBD density also from the healthy arm at different time points, postoperatively and after 6 weeks of surgery at the − 4% position.     Figure 3c2 shows the total CALCBD area from the fractured side after six weeks of surgery at the fracture gap (see also 103.06 mg/cm³) of the estrogen-treated patients. The corresponding p = 0.8036 value indicates that there is no statistically signi cant difference with respect to bone fracture healing between the two patient groups. The data obtained from fractured and non-fractured arms of estrogen treated and placebo treated patients suggest nevertheless a tendency towards better healing in response to the transdermally applied estrogen.
We next asked whether there is an age-dependent effect of estrogen on bone fracture healing because of the reduced sensitivity to estrogen with age [22]. Figure 4 summarizes the data from CALCBD (Fig. 4a1 and 4b1) and the CORTBD analyses ( Fig. 4a2 and 4b2) of the healthy (Fig. 4a1 and a2) and fractured ( Fig. 4b1 and b2) arm of postmenopausal women of different ages. Figure 4a1 shows the total CALCBD levels of the healthy arm of individual patients of different ages. Mixed effects model calculations indicated that callus formation tends to decline in both estrogen treated and untreated postmenopausal patients during ageing. Although the scatter range was large, this effect was statistically signi cant (see Fig. 4a1; p ** = 0.0014; placebo: r = -0.49 and R 2 = 0.24; E2: r = -0.28 and R 2 = 0.08) but independent of the estrogen treatment ( Fig. 4a2; placebo: r = -0.29 and R 2 = 0.09; E2: r = -0.15; R 2 = 0.02). Figure 4b1 shows the total CALCBD level at the fracture gap of individual patients of different ages, indicating that callus formation at the fractured site follows a declining trend in the placebo group, whereas a trend in the estrogen treated patients towards increased callus formation was again notable (placebo: r = -0.03 and R 2 = 0.00; E2: r = 0.10; R 2 = 0.01). Furthermore, CORTBD density measurements at the fracture gap showed a tendency to decrease with the age of the placebo group patients, whereas a stable and consistent status prevails in the estrogen treated patients (Fig. 4b2; placebo: r = -0.03 and R 2 = 0.00; E2: r = 0.16; R 2 = 0.03).
Alkaline phosphatase serum levels are not affected by estrogen treatment of the patients Bone-speci c alkaline phosphatase (ALP) activity is a diagnostic maker for neotissue mineralization during bone growth and fracture healing [23]. Thus, in addition to the morphological studies on bone healing, we also measured ALP activity in the blood serum of the patients. These measurements were performed on patients with and without estrogen treatment both immediately after surgery and six weeks later. In agreement with morphological studies reported above, there was no signi cant effect of the estrogen treatment. Untreated patients of the placebo group had an average ALP value of 51.37 U/l (standard deviation: 53.49 U/l) and 49.53 U/l (standard deviation: 52.47 U/l) after surgery and six weeks later, respectively. The corresponding ALP values of the estrogen treated patients were 28.15 U/l (standard deviation: 33.88 U/l) and 30.12 U/l (standard deviation: 32.19 U/l). The similar ALP values suggest that neotissue mineralization during bone fracture healing is not affected in response to the transdermally applied estrogen. Taken together, both the morphological measurements and the enzymatic ALP activity showed no signi cant effect of estrogen treatment on bone fracture healing.

Discussion
Several lines of preclinical evidence suggest that estrogen plays an essential physiological role during bone development and the bone remodeling process [4]. The administration of estrogen in a sustained way was shown to be able to reverse the decline in bone strength and to re-establish bone quality in rats [8,24,25]. Furthermore, long-term ovariectomized rats showed impaired fracture healing [26]. Moreover, histomorphological analyses demonstrated a delay in fracture callus healing with poor development of mature bone in these animals [27]. These ndings suggested a possible role of the hormone in bone fracture healing, as the remodeling of the fracture callus is essential during this process [4,28]. It has also been demonstrated that cells involved in fracture healing such as osteoblasts, osteocytes and osteoclasts express functional estrogen receptors, ER-alpha and ER-beta, and that ER alpha supports callus formation and thus fracture healing [29]. These receptors are also expressed in bone marrow stromal cells, the precursors of osteoblasts. In addition, Boden et al. (1989) [30] reported that rat estrogen receptor genes are expressed in fracture callus, implying that estrogen could play a role in the normal fracture healing process. Finally, it was shown in a mouse model that estrogen plays a crucial role in fracture healing [4], and that estrogen may enhance the fracture healing process in animal models such as rabbits [9]. Collectively, these ndings suggested that estrogen treatment may bene t postmenopausal patients with osteoporotic fractures [31].
However, there are reports on severe side effects of estrogen treatment of human. High doses of estrogen can cause incident coronary heart disease, stroke, venous thromboembolism, breast cancer, colorectal cancer, endometrial cancer or hip fractures [32,33,34]. It has recently been shown that estrogen-alone hormone therapy does not increase the risk of breast cancer in postmenopausal women as outlined in an updated report of the Women's Health Initiative (WHI) estrogen-alone trial [35]. Doses of 0.625 mg/day of conjugated equine estrogen (CEE) were applied to 10,739 post hysterectomy women showing that they caused other adverse reactions [36]. This dose was about twice as high as applied in the animal model organisms where bene cial effects on bone fracture healing were reported, but as observed in humans, strong side effects were also observed [4].
Studies analyzing the dose responses of estrogen with respect to bone healing, its dose-dependent side effects and clinical studies with lower doses of estrogen on post hysterectomy women have not yet been reported. To avoid any potential estrogen-dependent complications [32,33,34] in our study, we used a su ciently low estrogen dose to reach the endogenous systemic level of estrogen equivalent to the level in premenopausal women. This was achieved by applying a weekly transdermal dose of 4 mg estradiol over a period of four weeks. This treatment increased the endogenous estrogen level in the blood of the patients from about 16 ng/ml to a blood concentration of ≤ 30 ng/ml estrogen, corresponding to the endogenous estrogen level of women during the rst half of the menopause cycle [37].
Our ndings showing that there is no statistically signi cant bene cial effect of estrogen treatment on bone fracture healing are in apparent contrast to observations with preclinical models. They show that estrogen administration, although at much higher doses however, strengthens indeed the properties of bones of ovariectomized animals after femoral fracture [4,8,9,11,12,13,14,38]. However, there are no studies yet which determine the minimal dose of estrogen that causes a bene cial effect on bone healing in the preclinical model organisms without producing the known side effects of estrogen. It could well be that a high estrogen dose as applied to the preclinical models would also have bene cial effects on human bone fractures, but causes otherwise adverse side reactions. It is noteworthy that mechanostimulation by low-magnitude high frequency vibration was shown to provoke anabolic effects on the intact skeleton of both mice and human. This effect is estrogen-dependent and shown to be mediated primarily via the estrogen receptor alpha in mice, and vibration-induced effects on fracture healing in combination with estrogen treatment have also been shown to improve the osteopenic bone structure and increased the bone stiffness in preclinical studies [14,39,40,41]. Thus, it is possible that such a combination of treatments could improve the healing of bone fractures rather than estrogen treatment alone.
We observed that callus formation after bone fracture was age-dependent. The CALCBD levels decreased with the age of the patients, i.e. the on average 60-year-old patients had a CALCBD level of 250 mg/cm 3 , whereas the CALCBD level of 90-year-old patients was about 20% lower. This nding suggests a tendency for the reduction of callus formation with increasing age.

Conclusion
Our study shows that increasing the levels of estrogen in postmenopausal women to premenopausal levels have no signi cant effect on their bone fracture healing. Possible reasons for this observation could be insu cient acuity of the chosen measurement protocol, and/or an insu cient dose of estrogen applied. Alternatively or in addition, there could be an age-dependent reduction of estrogen receptors in the hormone-receiving bone cells, representing a limiting factor for bone regeneration in postmenopausal women. One additional reason could be that estrogen was transdermally administered and that a more local administration may support a better bone fracture healing.
Overall, of course, this study has limitations. According to animal data that showed a signi cant effect of estrogen for fracture healing, our goal was to nd a method to study this effect in humans. This turned out to be di cult and one way to examine the fracture healing in detail was to use pQCT in radial fractures. Even after a very extensive clinical study with 76 patients, which meant a lot of effort, we could not prove any clear effect of estrogens on fracture healing. However, we believe that from this data the message can be drawn that estrogens do not have the maximum effect on fracture healing and that we could already have recognized a possible effect on fracture healing with this method. The present study was approved by the ethics committee of the University Medical Center Hamburg-Eppendorf (approval number: AZ 2252). Written informed consent will be obtained from all participants.

Consent for publication
Not applicable.
Availability of data and material "The datasets supporting the conclusions of this article are included within the article (and its additional les."

Competing interests
The authors declare that they have no nancial and non-nancial competing interests.   Measurements at -4% of the total length of the radius on the healthy side. Measurements were taken after zero and six weeks. b2: Shows the measurements at six weeks after removal of the external xateur and the K-wires on the fractured side (n (E2) = 31; n (placebo) = 45). The isolated points represent three outliers. c1 and c2: CALCBD area levels at different time points and at different heights of the radius. c1: Measurements at -4% of the total length of the radius on the healthy side. Measurements were taken after zero and six weeks. The isolated points represent three outliers. c2: Shows the measurements at six weeks after removal of the external xateur and the K-wires on the fractured side (n (E2) = 31; n (placebo) = 45). The isolated point represent an outlier.

Figure 4
Shows scatterplots representing measurements of the CALCBD content ( g. 4a1 and 4b1) as well as the cortical bone density (CORTBD) ( g. 4a2 and 4b2) of the healthy arm ( g. 4a1 and a2) compared to the fractured arm at the height of the fracture gap of the radius (see g. 4b1 and b2); the scatterplots represent the measurements with (right side) and without treatment (left side) of estrogen. a1: shows total CALCBD levels of the healthy side of individual patients of different ages with bone fractures on the opposite side. a2: shows the CORTBD density measurements during bone healing. b1: shows total CALCBD levels of the fracture gap of individual patients of different ages. b2: shows CORTBD density in the fracture gap. The gray shading shows the 95% con dence interval. n (E2) = 31; n (placebo) = 45. R2 represents the coe cient of determination; r represents the correlation coe cient.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. CONSORTchecklist.docx Linkfortrialregistration.docx