Introduction

In clinical practice, bone strength is usually measured using dual-energy X-ray absorptiometry (DXA) and expressed as bone mineral density (BMD). However, in earlier reports, it is stated that next to BMD, bone geometry is associated with bone strength as well.^{(1, 2)} Structural bone geometry can be described in terms of periosteal and endocortical width and is, among other factors, determined by sex steroids.⁽³⁾ The main regulators of periosteal and endocortical appositions are thought to be testosterone and estrogen, respectively.^(4-6) Another important mediator in the development of sex-specific bone geometry is the timing of puberty. Therefore, as most bone mass accrual occurs during this period of development, sex-specific changes in bone geometry occur mainly during puberty.⁽⁷⁾

Recently, unique population groups have emerged to study the sex-specific role of sex steroids in bone geometry during puberty. These include transgender adolescents treated with gonadotropin-releasing hormone agonist (GnRHa) to suppress their endogenous puberty, and subsequently, gender-affirming hormones (GAH) to induce puberty toward the experienced gender.

Transgender people experience an incongruence between one's gender assigned at birth and one's experienced gender, which can lead to gender dysphoria (GD). This may have already emerged during early childhood. Therefore, development of secondary sex characteristics during puberty can be a traumatizing event for some transgender children. To avoid this, suppressive treatments for pubertal development were introduced in the Netherlands approximately 20 years ago.^{(8, 9)} The treatment exists as a periodically administered GnRHa.⁽¹⁰⁾ GnRHa overstimulation initiates downregulation of the pituitary GnRH receptor. As a result, production of gonadotropins, the luteinizing hormone and the follicle-stimulating hormone, is reduced. In turn, the body enters a hypogonadotropic hypogonadal state, which ultimately suppresses pubertal development. The current guideline of the Endocrine Society on care for GD people recommends GnRHa initiation at Tanner stage 2.⁽¹¹⁾ Subsequently, after careful evaluation and deliberation, puberty of the experienced gender may be induced with GAH. Trans women, who have a male sex assigned at birth but female gender identity, are treated with oral or transdermal estradiol. Trans men, who have a female sex assigned at birth but male gender identity, are treated with transdermal or intramuscular testosterone. We acknowledge the fact that the described treatment protocol is not globally endorsed. In some countries, GAH is not allowed before the age of 18 years.

Bone geometry has been shown to remain unchanged during the GAH treatment in adult transgender people.⁽¹²⁾ This lack of change may be because adult participants had already achieved final geometrical proportions. However, this assumption has not yet been studied in adolescent transgender populations. Seizing the exceptional opportunity to investigate bone geometry in transgender youth yields valuable knowledge on bone strength development in transgender adolescent health care. When translated to general physiology, it also provides more insight into the effects of sex steroids in bone development during the different phases of puberty. We hypothesized that bone geometry of transgender people who received puberty blockers and subsequent GAH treatment during childhood would show more resemblance to that of the experienced gender than to that of the gender assigned at birth. For trans women receiving estrogen, this implies a smaller endocortical diameter than that of cis men (men whose gender identity matches the sex assigned at birth). For trans men receiving testosterone, this implies a larger periosteal diameter than that of cis women (women whose gender identity matches the sex assigned at birth). This led to the research question at hand: Does bone geometry of transgender people who received GnRHa and GAH treatment during adolescence show more resemblance to that of the experienced gender than to that of the gender assigned at birth after ≥2 years of GAH? Because sex-specific change in bone geometry occurs as a gradual process during puberty, the additional question was if possible change depends on pubertal stage during which GnRHa treatment was initiated.

Materials and Methods

Bone geometry

Conventional DXA measures BMD as two-dimensional average bone density within a particular region but does not consider the exact placement of mineral content. Hence, BMD does not provide information about the potential endurance for mechanical loading or structural bone geometry. In 1989, Beck and colleagues introduced hip structure analysis (HSA) on DXA-derived images to assess biomechanical and geometrical parameters of the proximal femur.⁽¹⁵⁾ The main outcomes of our study were two geometrical parameters: subperiosteal width and endocortical diameter. These were obtained by running an HSA option APEX software version 4.0 on previously produced DXA images of the non-dominant proximal femur. This software calculates HSA parameters through an automated process. DXA images were generated using Hologic Discovery A (Hologic Inc., Bedford, MA, USA). The regions of interest were the narrow neck, the intertrochanteric region, and the femoral shaft (Fig. 1). {FIG1}

At each site, the subperiosteal width and endocortical diameter were measured. The subperiosteal width marks the distance between opposing cortices of the outer bone diameter and was measured in centimeters. The endocortical diameter is an assessment of the distance between the opposing inner edges of the cortex and was also measured in centimeters. The coefficient of variation (CV) was calculated based on repeated phantom measurements. For the subperiosteal width, the CV was 1.1%. For endocortical diameter, the CV was 1.4%. Only participants with at least one scan after February 2011 could be included because the HSA software was not available at our institution before that date.

At the time of DXA, ethnicity and body mass index (BMI) were determined. The DXA software has four possibilities for ethnicity input: white, black, Hispanic, and Asian. BMI was calculated with directly measured height and weight.

Changes in subperiosteal width and endocortical diameter due to natural occurring growth had to be considered. Therefore, sex-specific HSA parameters at different ages in the general population during puberty were obtained from previously published literature.⁽¹⁶⁾ Only reference data measured at the narrow neck were available.

Results

Participant characteristics

In total, 322 participants were included, of whom 106 were trans women and 216 were trans men. A flowchart of the inclusion process is shown in Fig. 2. Furthermore, Tables 1 and 2 show characteristics of trans women and trans men, respectively, subdivided per pubertal stage at the start of the GnRHa treatment. We provided the data on age at the start of the GnRHa treatment, the duration of mono GnRHa treatment, the age at the start of GAH, the age upon DXA after ≥2 years of GAH, and the concentrations of estradiol and testosterone at all three time points. As per treatment protocol, participants became medically eligible for gonadectomy between the periods of DXA performed at the start of GAH and DXA performed after ≥2 years of GAH. Overall, 115 participants underwent gonadectomy before performing DXA at ≥2 years of GAH. Additionally, baseline characteristics concerning ethnicity and BMI were given. There were no significant differences in change in BMI over time between the different puberty groups for either trans women or trans men.

HSA parameters of narrow neck

Figures 2 and 3 show the changes in subperiosteal width and endocortical diameter, respectively, of both trans women and trans men. These results were derived from the linear mixed-model analyses. The total numbers of observations in trans women used in the mixed-model analyses in the early, mid, and late puberty groups were 56, 43, and 51, respectively. In trans men, the total numbers of observations in the early, mid, and late puberty groups were 11, 36, and 248, respectively. In addition, literature-extracted reference values of the cis population are presented in the figures. In Table 3, absolute values for subperiosteal width and endocortical diameter at the start of the GnRHa treatment, changes at the start of GAH, and development after ≥2 years of GAH are summarized. All data shown are based on narrow neck HSA measurements.

Table 3. Change in Centimeters in Subperiosteal Width and Endocortical Diameter in Trans Women and Trans Men, Derived From the Linear Mixed-Model Analyses

	Start of GnRHa	∆ between the start of GnRHa and the start of GAH	∆ between the start of GnRHa and after ≥2 years of GAH	∆ between the start of GAH and after ≥2 years of GAH
Trans women
Early puberty
Subperiosteal width	3.00 (2.81; 3.19)	0.38 (0.16; 0.60)	0.44 (0.23; 0.65)	0.06 (−0.15; 0.27)
Endocortical diameter	2.68 (2.49; 2.87)	0.39 (0.16; 0.61)	0.38 (0.17; 0.60)	−0.00 (−0.21; 0.21)
Mid puberty
Subperiosteal width	3.06 (2.89; 3.23)	0.33 (0.15; 0.50)	0.57 (0.39; 0.75)	0.25 (0.11; 0.38)
Endocortical diameter	2.73 (2.56; 2.89)	0.34 (0.17; 0.51)	0.55 (0.37; 0.72)	0.21 (0.08; 0.34)
Late puberty
Subperiosteal width	3.36 (3.25; 3.47)	0.06 (−0.08; 0.20)	0.27 (0.16; 0.39)	0.21 (0.09; 0.34)
Endocortical diameter	2.98 (2.86; 3.09)	0.08 (−0.06; 0.22)	0.27 (0.15; 0.40)	0.19 (0.06; 0.33)
Trans men
Early puberty
Subperiosteal width	2.71 (2.35; 3.07)	0.63 (0.58; 0.68)	0.79 (0.72; 0.85)	0.15 (0.12; 0.19)
Endocortical diameter	2.39 (2.02; 2.76)	0.62 (0.57; 0.67)	0.73 (0.67; 0.79)	0.11 (0.08; 0.14)
Mid puberty
Subperiosteal width	2.99 (2.81; 3.17)	0.10 (−0.09; 0.29)	0.31 (0.11; 0.50)	0.21 (0.03; 0.38)
Endocortical diameter	2.66 (2.46; 2.85)	0.09 (−0; 11; 0.30)	0.27 (0.06; 0.48)	0.18 (−0.01; 0.36)
Late puberty
Subperiosteal width	3.18 (3.10; 3.27)	0.07 (−0.03; 0.18)	0.15 (0.04; 0.26)	0.07 (−0.04; 0.18)
Endocortical diameter	2.80 (2.71; 2.89)	0.10 (−0.01; 0.21)	0.17 (0.05; 0.28)	0.07 (−0.04; 0.17)

• GnRHa = gonadotropin-releasing hormone analogue; GAH = gender-affirming hormone treatment.
• Data are shown as mean (95% confidence interval).

Marked differences among changes in subperiosteal width (Fig. 3) and endocortical diameter (Fig. 4) during GAH were noted that depended on pubertal stage. It was only in early pubertal trans women and trans men that the curve shifted toward the reference curve of the experienced gender. Participants starting the GnRHa treatment in mid and late puberty remained within the curve for their gender assigned at birth.

The results for the intertrochanteric region and femoral shaft are shown in Supplemental Tables S1 and S2. Reference values of the general population for these regions of interest are not available in the current literature.

Discussion

This study demonstrated that participants starting the GnRHa treatment in early puberty resemble the change in subperiosteal width and endocortical diameter of the experienced gender during GAH. Previous studies on bone health of transgender youth who received puberty blockers have predominantly focused on bone density. In 2015, Klink and colleagues investigated BMD development in 34 transgender adolescents undergoing GnRHa and GAH treatments.⁽¹⁴⁾ They found that the absolute bone mass of both trans women and trans men treated with GnRHa decreased. Although absolute bone mass increased after commencing GAH, the areal BMD Z-score was still below the pretreatment level in both groups after long-term follow-up. In 2017, Vlot and colleagues reported the relation between bone turnover markers and bone mass during GnRHa and GAH treatments.⁽²⁰⁾ During treatment with GnRHa, both bone turnover and bone mineral apparent density (BMAD) Z-scores decreased. After 24 months, bone turnover further decreased, except for the subgroup of trans men with a bone age of ≥14 years at the start of treatment. In contrast, an increase in BMAD Z-scores for all participants was found during the GAH treatment. Consistent with the 2015 study, Z-scores did not reach pretreatment levels. Although knowledge on BMD development of transgender youth is growing steadily, bone geometry in this specific population was still an unexplored area of research.

Participants in our current study, starting in mid or late puberty, acquired bone geometry more closely resembling the reference curve of the gender assigned at birth. This illustrates that the main effect of testosterone and estrogen on periosteal and endocortical bone growth occurs during early puberty. A possible explanation may the phenomenon of “programming.”

The idea of programming entails that exposure to stimuli during a critical window of development in early life can have major consequences throughout one's life span. Previous studies by Barker and colleagues have laid the foundation of this theory.^(21-23) In these studies, it was shown that poor nutrition in prenatal and early life, resulting in low birth weight and small size during infancy, was associated with a higher risk of cardiovascular disease during adulthood. This theory is now known as the “developmental origins of adult health and disease” (DOHaD).

A 1959 concept similar to DOHaD, the organizational-activational hypothesis, is also well established in the behavioral sciences. This hypothesis proposed steroid hormones organize brain structure during a sensitive period in early development and may activate behavior during puberty and adulthood.⁽²⁴⁾ The primary focus of both theories was on the effects of stimuli during fetal and perinatal growth. More recently, it has been suggested that puberty might also be a possible window of opportunity for programming or organization.^(25-28) This latter statement was supported by our results. Additionally, this line of thought was consistent with previous research on bone geometry in adult transgender people.^{(12, 29, 30)} Similar to the mid and late puberty groups in our study, various geometrical parameters of adult transgender participants in these studies did not shift toward the experienced gender. This suggests a closure of developmental window in which modulation of sex-specific bone development was still possible. Alternatively, since all participants in these studies had already attained peak bone mass, bone structure remodeling might no longer be influenced by sex steroids. However, explaining the striking dependence on the pubertal phase of sex-specific bone geometry development was beyond the scope of our study and requires further research.

The role of testosterone and estrogen on bone geometry has been established. However, the specific effects have not fully been elucidated. In the second half of the 20th century Garn explored sexual dimorphism in bone geometry.⁽⁷⁾ Based on two-dimensional radiographic images of the metacarpal bones, they postulated that periosteal apposition occurred during adolescent growth and was greater in cis boys than in cis girls. In contrast, cis girls experienced greater endocortical apposition during puberty than cis boys, which was possibly driven by estrogen.

More recent work by Gabel and colleagues confirmed the effect of testosterone on periosteal growth but showed a different result for the postulated estrogen-driven endocortical apposition.⁽³¹⁾ This research was based on peripheral quantitative computed tomography and showed that endocortical resorption in cis girls was less than that in cis boys. The authors suggest that this discrepancy might be due to different anatomical regions of interest. Bone metabolism can vary by anatomical region owing to different forces experienced at that specific site, whereas Garn's findings were based on non-weight-bearing metacarpal bones, the latter study assessing the tibia. This emphasizes that the impact of mechanical forces should be considered, in our study more so because it was shown that GAH affected muscle size and body weight in transgender adolescents.⁽³²⁾ Shifts in muscle strength could contribute to the changes in bone geometry in accordance to Frost's mechanostat theory, which suggested testosterone enhances muscle mass and thereby increases mechanical loading, stimulating periosteal growth.⁽³³⁾ Therefore, the sex-specific changes in bone geometry were most likely not solely a direct result of sex steroids but rather a combination of mechano- and hormonal regulation. Unfortunately, for instance, data on grip strength at the time of DXA were unavailable to test this hypothesis.

Our study has some limitations. First, a standardized data set on HSA parameters that could serve as reference data was not available. Given this absence, the literature for the most suitable reference data was researched. We found two studies reporting on HSA parameters. One study described HSA parameters of the femoral neck in a Chinese cohort of participants aged 15 to 91 years.⁽³⁴⁾ Possible differences in lifestyle and genetic predisposition and the lack of appropriate age intervals made this study unsuitable for reference. Furthermore, another study based on a Swedish cohort in which lifestyle and genetic predisposition were expected to be more compatible with our study population was researched.⁽¹⁶⁾ Indeed, all participants in this cohort were white, which was very similar to our study population. Owing to lack of data, we were unable to compare other factors that may impact bone geometry, such as nutritional status, physical activity, and calcium or vitamin D intake. However, it has been shown that physical activity is similar among adult transgender people and cisgender controls before starting GAH.^{(30, 35)} Previous studies on vitamin D levels in adult transgender people demonstrated that 25-OH vitamin D levels in trans women are significantly lower than in cis men and that adult trans men had 25-OH vitamin D levels similar to those of cis women.^{(30, 35)}

The DXA device used by Alwis and colleagues was manufactured by a company different from that used in our study (Lunar versus Hologic).⁽¹⁶⁾ The Lunar DPX-L utilizes a pencil beam, whereas the Hologic Discovery A used in our study operates a fan beam. Thus, it was not possible to compare absolute values directly. In the current literature, we did not find studies comparing HSA parameters between these manufacturers. Nevertheless, trends in change, visualized by the slope of the curves, could still be compared. We think that this is very valuable knowledge because the aim of this study was to investigate change in bone geometry over time and not to directly compare absolute numbers. However, results should be interpreted with caution. Furthermore, the Swedish data set provided parameters divided per age bands and not the pubertal stage. Because Tanner stage and its corresponding number of people were listed per age band, we were able to compare both study populations. We found that for the vast majority, these Tanner stages, reported per age bands, corresponded with those in our study groups at start of GnRHa, allowing for age-related comparisons of bone geometry changes.

Second, there are technical limitations inherent to measuring bone geometry with DXA. The HSA parameters are three-dimensional measurements of bone reconstructed from two-dimensional images. Inconsistencies in placement of the hip can result in measurement errors. However, all scans were performed by trained technicians according to the manufacturer's manual. Hence, we expected potential errors to be reduced to a minimum. Furthermore, when compared with quantitative computed tomography (QCT), the gold standard of three-dimensional measuring, numerous studies have found a high degree of correlation with DXA-derived HSA.^{(36, 37)} In these studies, HSA measurements derived from the same model of Hologic device as used in our study are compared with 3D structural measurements with QCT. Both studies can support the validity of HSA.

Third, not all included participants had DXA scans available at all three predetermined occasions. However, by analyzing the data with a mixed model, the missing data were randomly imputed.⁽¹⁹⁾ Therefore, the effect of missing data was minimalized while maintaining the largest study sample.

Lastly, because of the small sample size of early pubertal trans men, it is hard to draw firm conclusions regarding this specific group. The discrepancy between early pubertal trans women and trans men is explained by the fact that trans men, whose gender assigned at birth is female, enter puberty at an earlier age than trans women. By the time the diagnostic process is finished, many trans men already completed the early phase of puberty. Unfortunately, we were unable to increase the number of participants in this particular group. However, we have no indication to assume that the included trans men are an inadequate representation of this group.

As mentioned, the course of BMD and bone turnover markers during the GnRHa treatment and GAH was previously investigated.^{(14, 20)} BMD, however, is just one aspect of bone strength. Bone geometry and material properties of the bone are important as well. Through this study, we therefore add another dimension to the body of knowledge on sex-specific bone development during puberty. We showed that only when the GnRHa treatment was started in early puberty, and GAH subsequently, did changes in the direction of bone geometrical development occur. This could hold important clinical implications in terms of fracture risks. Duan and colleagues previously showed there is a sex-specific difference in hip fracture risk in old age as a result of sexual-dimorphic bone geometry.⁽²⁾ This is attributable to the fact that, compared with cis women, in young cis men the cortex is placed further from the neutral axis of the bone, providing greater bending strength. However, there are no studies on long-term fracture risks in this particular trans population yet. We could speculate that if transgender women, treated with puberty suppression in early puberty, obtain bone geometry similar to that of the cis women, fracture risk could also equal that of the cis women, but further research should address this question. Another potential opportunity for research is exploring the translation of the observed change in HSA to a change in bone turnover markers. This may be done by comparing markers for bone formation and resorption in a similar population. Additionally, our results could raise questions for future research on the programming effects of sex steroids on bone geometry development during puberty in the general population.

In conclusion, development of hip bone geometry in transgender adolescents resembled that of the experienced gender if the GnRHa treatment was initiated during early puberty and was followed by a start of GAH. Only participants starting during early puberty showed more resemblance to the reference curves of their experienced gender. Participants starting GnRHa and GAH treatments during mid or late puberty continued within the curve of their gender assigned at birth.

Disclosures

All authors state that they have no conflicts of interest.