Quantitative computed tomographic assessment of the effects of 24 months of teriparatide treatment on 3D femoral neck bone distribution, geometry, and bone strength: Results from the EUROFORS study

Abstract

We studied the changes in bone distribution, geometry, and bone strength based on 3D quantitative computed tomography (QCT) of the femoral neck (FN) in subjects receiving teriparatide (TPTD). Fifty-two postmenopausal women with severe osteoporosis were analyzed. Patients were divided into three subgroups based on their prior treatment with osteoporosis drugs: treatment-naive (Tx-naive; n = 8), pretreated (pre-Tx; n = 12), and pretreated showing an inadequate response to treatment (inad. pre-Tx; n = 32). QCT scans were performed at baseline and after 6, 12, and 24 months of treatment and were analyzed with Mindways QCT-PRO BIT software. Minimum and maximum section modulus, buckling ratio (BR), and cross-sectional area (CSA) were calculated as measurements of bending strength, risk of buckling, and bone apposition, respectively. After 24 months of TPTD treatment, areal and volumetric FN BMD increased significantly by 4.0% and 3.0%, respectively, compared with baseline. Decreases in cortical volumetric BMD occurred in locations not adversely affecting minimum bending strength indicators. Cortical CSA increased by 4.3%, whereas total CSA remained unchanged over the study duration, indicating that endosteal but no periosteal growth was observed. Strength parameters for buckling did not change at 6 and 12 months but improved significantly at 24 months. Measures of bending strength showed a trend toward improvement. Changes tended to be larger in individuals at higher risk of buckling failure. Prior antiresorptive treatment may delay response to TPTD, but based on the small magnitude of the mostly insignificant changes at 6 months, this does not appear to lead to an interim phase of reduced bone strength. In summary, FN QCT provides a tool for detailed longitudinal investigation of bone strength indices in vivo for different loading modes, yields insight into underlying structural changes, and provides relevant mechanostructural information beyond dual-energy X-ray absorptiometry. Continuous TPTD treatment for 24 months improves FN bone strength parameters. © 2010 American Society for Bone and Mineral Research

Introduction

Dual X-ray absorptiometry (DXA)–based analyses of the areal bone mineral density (aBMD) of the proximal femur and lumbar spine are the most widely used methods to evaluate bone mass because of their accessibility, ease of use, and good fracture prediction in untreated patients.1, 2 However, especially for follow-up measurements at the proximal femur, DXA has limitations owing to the complex 3D geometry of this anatomic area, its low spatial resolution, and the 2D imaging approach. At the femoral neck, the reproducibility of defining the volume of interest (VOI) in longitudinal studies is reduced and user-dependent, and the interpretation of standard DXA results is limited to the bone width and aBMD of the structure. Moreover, the projectional aBMD may not appropriately reflect bone strength.3, 4

To overcome these limitations, researchers have tried to derive more complex biomechanical indices such as the one based on hip structure analysis (HSA).5 However, these methods suffer from inherent DXA limitations: DXA-based mechanical indices rely on questionable assumptions (e.g., the cylindrical shape of the femoral neck), and they are as susceptible to partial volume-associated edge-detection errors as aBMD itself. Indeed, despite interesting studies using HSA,3, 4, 6 it still remains unclear if independent and biomechanically relevant information can be derived that improve fracture risk prediction.7, 8

As an alternative to DXA, quantitative computed tomography (QCT) offers complete 3D information, high in-plane resolution, and separate assessment of cortical and trabecular bone of the femur.9, 10 This allows evaluation of cross-sectional mineral distribution along the femoral neck for an in-depth characterization of bone strength for different loading modes. Furthermore, the 3D nature of CT permits high precision in VOI definition, which is of particular importance in longitudinal studies assessing structural effects of bone active drugs at the femoral neck.

We have applied this technique to assess the effects of teriparatide (TPTD), a bone-forming agent. The aim was to investigate TPTD-induced changes on volumetric bone density, bone geometry, and bone strength parameters of the femoral neck in humans. By separation of the effects in the trabecular and cortical bone envelopes and calculation of mechanically relevant properties of bone strength, we aimed to provide a detailed mechanical assessment of TPTD-induced changes in femoral neck bone strength indicators. In addition, we addressed the clinically important question of whether patients may undergo an interim period of reduced bone strength within the early phases of TPTD treatment, temporarily putting them at high risk of fracture until the long-term benefits of the treatment become manifest.

Materials and Methods

Subjects

This analysis included a subgroup of patients from 12 investigative centers in Germany and Spain who participated in the QCT addendum of the EUROFORS Study. Details of this study have been published previously.11 In brief, 868 women 55 years of age or older with a BMD T-score of −2.5 or less at the lumbar spine, total hip, or femoral neck who had at least one documented preexisting fragility fracture in the last 3 years were eligible for enrollment in the study. Women with a known history of metabolic bone disease other than postmenopausal osteoporosis; elevated serum levels of calcium, alkaline phosphatase, or parathyroid hormone; significantly impaired renal or hepatic function; or pretreatment with fluorides in the past 12 months or with any other bone anabolic agent in the past 6 months were excluded. Prior treatment with antiresorptive (AR) drugs had to be discontinued at baseline. A detailed documentation of the prior AR therapy and the patient's clinical response to it was required. Based on these data, patients were grouped into three subsets: (1) treatment-naive patients without any previous antiosteoporosis therapy (Tx-naive), (2) AR pretreated patients (pre-Tx), and (3) AR pretreated patients with documented inadequate clinical outcome (inad. pre-Tx). Women were classified as inadequate AR responders if they met at least one of the following criteria: (1) sustained at least one new fragility fracture despite prior prescription of an AR therapy for at least 12 months prior to fracture, (2) had a lumbar spine, total hip, or femoral neck BMD T-score of −3.0 or less after documented prior AR treatment for at least 24 months, and/or (3) experienced a decrease of greater than 3.5% in BMD at any one of the skeletal sites measured despite documented continuous prescription of an AR agent in the preceding 24 months. The specifications of the DXA assessments have been described in detail previously.11, 12

Of the 78 patients who participated in this addendum and who were assigned to 24 months of continuous TPTD treatment, 52 completed all four scheduled visits with proximal hip measurements and are included in this report.

Institutional review board approval was obtained from each of the clinical study sites, and written informed consents for the general study and the QCT substudy were obtained from each participant.

QCT assessment

Spiral CT scans were acquired from the acetabulum directly above the femoral head down to 1 cm below the lesser trochanter, resulting in 25 to 35 slices, with 3 mm slice thickness, over a range of 8 to 12 cm. All scans were performed at 120 kV and 70 to 200 mAs depending on height and weight of the patient according to the CTXA Hip Module Exposure Tables (Mindways Software, Inc., Austin, TX, USA). All patient QCT data were processed using QCT-pro Software Version 4.1.3 and the QCT-pro Bone Investigational Toolkit Version 2.0 (BIT, Mindways). CT scanner types included Siemens Somatom 16 and Volume Zoom, General Electric Lightspeed 16, and Toshiba Asteion.

The CT values were converted to bone mineral scale using a solid calibration phantom (Mindways) placed underneath the patients during the scan. Quality assurance (QA) scans with a Type 3 Mindways Phantom were performed on each day when study measurements took place at the clinical site in order to adjust for longitudinal changes of the detector. QA measurements were evaluated according to the “QCT-pro QA Guide” from Mindways.

Patient data from all four visits were evaluated in parallel with the CTXA Hip Exam Analysis protocol by Mindways. In the process, the exact 3D rotation of the femur and the threshold setting for defining the bone contours appeared to be the two most critical steps for the achievement of accuracy and reproducibility of the automated procedure. The outer cortical volumetric BMD (vBMD) thresholds had to be adapted individually to each scan. The variable pixel size of the scan (owing to different degrees of rotation adjustment) would have caused a bias because of different partial-volume effects if a constant threshold had been used. We additionally evaluated the entire population a second time with the outer cortical vBMD threshold set at a constant level. This caused more artefacts and required more user interaction.

The femoral neck axis was identified automatically with the Mindways Optimize FN Axis algorithm. A VOI consisting of six reformatted 1 mm slices oriented perpendicular to the neck axis was positioned along the neck axis according to the “eccentricity registration method” (see next section). QCT BIT processing then was performed with a fixed bone threshold for cortical separation set to 300 mg/cc for all patients and visits.

Bone strength was assessed for forces leading to bending or buckling. The corresponding relevant structural measures are section modulus (Z) and buckling ratio (BR). Z is inversely related to maximum stresses exerted by bending loads. QCT as a 3D measurement allows evaluating Z along the strongest (Z_max) and weakest axes (Z_min). When Z_max and Z_min are both considered, they reflect strength in torsion of the structure. Z is defined as Z = CSMI/r, where CSMI (cross sectional moment of inertia) measures the mass distribution relative to the geometric center, and r is the maximal distance from the geometric center to the periosteal surface for Z_max or the corresponding periosteal distance in the respective orthogonal direction for Z_min. BR is a measure of cortical elastic instability as a result of excessive cortical thinning. BR relates the cortical thickness to the width of the femoral neck and is defined as BR = r/ct, where r is the radius and ct is the corresponding cortical thickness. The bone is considered to be vulnerable to the failure mode of buckling when this ratio exceeds 10:1; that is, BR >10.13

VOI placement

Longitudinal evaluation of changes in vBMD and geometry requires careful matching of the VOI evaluated by QCT using image-registration methods. For many applications, the location of the minimal cross-sectional area of the femoral neck provides a good basis for placement of the VOI at a consistent position along the femoral neck. However, geometric properties might change in the course of TPTD treatment, and using them might bias the outcome of the study, specifically if periosteal apposition is to be investigated. Ideally, one would like to use landmarks outside the femoral neck; however, these also may be affected by treatment. Besides, it is questionable whether the variability of the cross-sectional area along the femoral neck axis around the location of minimum cross-sectional area is large enough to represent a sensitive criterion for reproducible placement of the VOI.

In order to select the best methods for VOI placement, we tested the performance of different registration methods in a subgroup of 11 randomly selected patients with all four QCT follow-up measurements. The reproducibility of the cross-sectional images was compared visually in sagittal and cross-sectional projections. Using an automatic algorithm for setting of femoral neck axis implemented in the Mindways software, we tested and compared the following image registration methods: (1) manual distal femoral neck placement of the VOI, flanking the trochanter major (comparable with DXA femoral neck box placement on Hologic devices), (2) automated placement based on minimal volume (comparable with the femoral neck box placement on GE Lunar devices), (3) automated placement based on a midshaft algorithm (femoral neck box placement by length measures of the femoral neck), and (4) automated placement based on eccentricity (“eccentricity registration method”). The latter method is based on variations in eccentricity of the bone cross-sectional area along the neck axis; the cross-sectional structure changes from circular at the head-neck junction to elliptical near the greater trochanter,14 where the ratio of the two principal axes of the ellipse approaches an eccentricity ratio of approximately 2:1 in our observations. These changes of eccentricity account for more than 100% along the femoral neck, whereas changes of cross-sectional area are relatively small, especially in the midshaft region. Crabtree and colleagues15, 16 showed in QCT ex vivo studies that a VOI definition along the femoral neck to the point of eccentricity 1.4:1 most closely matches the smallest volume of the neck, as defined by Kuiper and colleagues.17

Using the eccentricity registration method, a series of 10 reformatted 1 mm slices (i.e., the standard size of the Mindways neck VOI) was positioned perpendicularly to the neck axis. We observed that such a VOI setting is likely to include parts of the trochanteric region. The VOI therefore was reduced from 10 mm to the proximal 6 mm that included six slices. All steps were compared visually across all visits and repeated if positioning did not appear to be accurate.

Statistical analysis

JMP 5.0.1 Software (SAS Institute, Inc., Cary, NC, USA) was used for statistical analyses. Student's t tests and paired t tests were performed for comparative analyses where appropriate. Results for the series of six reformatted slices were averaged. Correlations were based on Pearson's correlation coefficient.

Results

Patients had a mean (SD) age of 67.8 (15) years, weight of 64.0 (24) kg, and height of 158 (15) cm. Thirty-two (61.5%) patients belonged to the inad. pre-Tx group, 12 (23.1%) had been assigned to the pre-Tx group, and 8 (15.4%) to the Tx-naive group.

VOI placement

The comparison of the different methods for VOI positioning showed that the placement based on minimum cross-sectional area did not yield satisfactory results. When tracked along the femoral neck axis, the cross-sectional area varied less than eccentricity, in particular around the location of minimum cross-sectional area. Varying the location of the VOI by about 1 cm around this location resulted in variations in minimum cross-sectional area of about 10% (Fig. 1). This is much less than the variability in eccentricity of about 40%. As a consequence, the criteria of minimal cross-sectional area appear to be too insensitive to provide a robust landmark for a consistent placement (“registration”) of the VOI. An example of the problems encountered is displayed in Fig. 2A: The cross-sectional images vary notably from visit to visit. Based on visual comparison, the automated eccentricity registration method and the distal femoral neck method of locating the femoral neck VOI appeared to produce more consistent results and to be superior to the other two automated registration methods. We decided to use the eccentricity method because it was less dependent on operator interaction. Figure 2B provides examples of consistent cross-sectional images across study visits.

Figure 1. Schematic depiction of the variation in eccentricity (1) of the cross section and cross-sectional area total (cm²) along the femoral neck axis based on patient data. While eccentricity changes rapidly in the center of the femoral neck, cross-sectional area varies less in this region.

Figure 2. Comparison of different methods for consistent placement of the VOI across study visits. Follow-up of 24 months is shown for the same patient and a fixed axis setting. VOI placement was performed based on (A) the method with femoral neck box placement by smallest cross-sectional area and (B) the eccentricity method. The eccentricity method yielded consistent results, whereas the smallest cross-sectional area showed varying shapes across the four visits (round shape close to the femoral head at 24 months and oval shape closer to the trochanter at baseline) indicating inconsistent placement of the VOI.

Response to teriparatide treatment

Baseline characteristics and the detected changes under TPTD treatment are summarized in Table 1. At baseline, the correlation between BR and total vBMD at the femoral neck was higher (r² = 0.83) than the correlations between BR and aBMD (r² = 0.37) and vBMD and Z_min or Z_max (r² range 0.24–0.25).

Table 1. Baseline Data and Percent Changes at 6, 12, and 24 Months After Baseline

	Baseline mean ± SD	6 months	12 months	24 months
aBMD (DXA) = areal bone mineral density; vBMD = volumetric bone mineral density; vBMC = volumetric bone mineral content; CSA = cross-sectional area; Zmin = minimal section modulus; Zmax = maximal section modulus. ° p < .1; * p < .05; ** p < .01; *** p < .001 compared with baseline.
aBMD (DXA) (g/cm^2)	0.56 ± 0.07	−0.9% ns	+2.5%*	+4.0%***
vBMD total (mg/cm^3)	180.5 ± 27.3	−1.4%*	+0.5% ns	+3.0%***
vBMD cortical (mg/cm^3)	553.0 ± 36.0	−2.7%***	−2.6%***	−2.0%***
vBMD trabecular (mg/cm^3)	80.5 ± 12.8	+2.1%*	+2.6%**	+5.2%***
vBMC total (mg/cm)	10378.1 ± 1532.9	−1.9%**	− 1.2%°	+2.2%*
vBMC cortical (mg/cm)	6718.9 ± 1422.8	−3.8%***	−1.7%°	+2.3%°
vBMC trabecular (mg/cm)	3659.2 ± 690.7	+1.9% ns	+0.7% ns	+2.6%**
CSA total (cm^2)	9.6 ± 1.0	−0.5% ns	−1.4%°	−0.6% ns
CSA cortical (cm^2)	2.0 ± 0.38	−1.2% ns	+0.9% ns	+4.3%**
Buckling Ratio (1)	9.52 ± 2.1	+0.7% ns	−1.3% ns	−4.3%***
Zmin (cm^3)	0.47 ± 0.01	±0.0% ns	−0.72% ns	+2.3%°
Zmax (cm^3)	0.50 ± 0.01	−1.6%°	−2.7%*	+1.9%°

Total CSA did not change significantly over the study duration (Fig. 3A), indicating that there was no detectable periosteal apposition. In contrast, the cortical area started to increase after 12 months and showed a significant increase of 4.4% compared with baseline after 24 months of treatment (p < 0.01; see Fig. 3A). Since there was no increase in total area, this finding can be interpreted as the result of apposition at the endosteal surface.

Figure 3. Changes after 6, 12, and 24 months of teriparatide treatment in density, geometry, and strength variables. °p < .1; ^*p < .05; ^**p < .01; ^***p < .001 versus baseline.

QCT revealed different temporal response patterns of cortical and trabecular bone envelopes that cannot be separated by DXA (see Fig. 3B). At 6 months of treatment, cortical vBMD decreased significantly, whereas trabecular vBMD showed a significant increase. This led to a significant decrease in total vBMD of −1.4% (p < .05). In the remaining 18 months, cortical vBMD remained stable, whereas cortical areas increased, which led to a net gain of 6.7% (p < .0001) in cortical bone mineral content (BMC). At 24 months of treatment, the changes in trabecular vBMD (5.2%, p < .0001) and cortical volumetric BMC (2.3%, ns) produced a total vBMD increase of 3.0% (p < .001) compared with baseline. The results of aBMD were similar to the total vBMD findings throughout the treatment period (see Fig. 3B).

The effects on the biomechanical strength parameters are shown in Fig. 3C. The temporary early decrease in cortical and total vBMD had no significant detrimental effects on the strength parameters of the patients analyzed. A transient decrease in Z was observed for the stronger plane (Z_max) but not for Z_min. This trend became significant for Z_max after 12 months of treatment (−2.2%, p < .01), but it was reversed with longer treatment duration. In the second year, buckling strength and bending strength indices increased significantly (BR −4.6%, p < .0001; Z_min +2.7%, p < .05; Z_max +3.9%, p < .01). The improvements over 2 years were significant for the BR (−4.3%, p < .001) and were borderline nonsignificant for Z_min (2.3%, p = .06) and for Z_max (1.9%, p = .08) compared with baseline.

At 6 months, the three predefined subgroups showed some differences in their responses to TPTD treatment (Fig. 4). Patients in the Tx-naive group showed significant increases for both Z_min and Z_max (4.4%, p < .05, for both variables) and a trend toward endosteal expansion of the cortical area. In contrast, patients in the inad. pre-Tx group presented a significant decrease in bending strength indicators in the stronger plane (Z_max) but no significant change in the weaker (Z_min) plane. This analysis revealed significant differences in the Z_min and BR values between the Tx-naive and inad. pre-Tx groups after 6 months of treatment (see Fig. 4).

Figure 4. Changes in density, geometry, and strength variables from baseline to 6-month visit (A) as well as between 6 and 24 months (B) of teriparatide treatment evaluated separately for the three pretreatment groups. Values for CSA cortical and Z_max show significant differences for the pretreatment groups using the Student's t test. °p < .1; ^*p < .05; ^**p < .01; ^***p < .001 versus baseline.

In the period from 6 to 24 months, the patients in the inad. pre-Tx group showed little increase in cortical vBMD (0.73%, ns), but there was a significant endosteal expansion (cortical CSA 6.1%, p < .001). Simultaneously, there were improvements in Z_min (3.3%, p = .07), Z_max (4.9%, p < .01), and BR (5.0%, p < .05). The equivalent results for the smaller pre-Tx group were not significant. However, the endosteal expansion over the 24-month period tended to be lower in both the inad. pre-Tx (cortical CSA 3.6%, ns) and pre-Tx groups (cortical CSA 2.3%, p < .05) compared with the Tx-naive group (9.1%, p < .05). These differences were not large enough to result in significant differences between pretreatment groups at 12 and 24 months of TPTD treatment.

Baseline values of Z and BR were associated with the magnitude of the changes under TPTD treatment; thus the patients with poorest mechanical competence at baseline showed the largest improvement over time. Baseline cortical CSA, Z_min, and Z_max were negatively correlated with their corresponding changes after 24 months of treatment (Fig. 5). Changes of absolute values for BR over 2 years were significantly larger in patients with BR > 10 (−0.8) than in patients with BR < 10 (−0.2, p < .05). The associations of baseline levels and later changes remained significant when adjusting for pretreatment. There was no correlation between baseline vBMD and later improvements in vBMD (r² = .005, ns).

Figure 5. (A) Baseline Z_min values show negative correlation with the outcome of Z_min values after 24 month of TPTD treatment (r = −0.30, p < .05). (B) Baseline Z_min values show negative correlation with the outcome of Z_max values after 24 month of TPTD treatment (r = −0.30, p < .05). (C) The cortical CSA values show negative correlation (r = −0.23, p < .05) between baseline values of cortical CSA and outcome of cortical CSA values after 24 month of TPTD treatment.

Discussion

We present a QCT-based method for investigation of structural, densitometric, and biomechanical changes associated with TPTD treatment at the femoral neck. The 3D nature of QCT allows performing a detailed assessment of changes in area, periosteal and endosteal cortical border, and mechanically relevant measures of bone strength beyond the information obtainable from DXA.6

Figure 6. Differences of density, geometry, and strength variables after 24 months of TPTD treatment for patients with baseline buckling ratio values larger and smaller than 10.13 °p < .1; ^*p < .05; ^**p < .01; ^***p < .001 versus baseline.

During 24 months of TPTD treatment, we observed significant increases in total (3.0%) and trabecular vBMD (5.2%), along with decreased cortical vBMD (−2.0%) at the femoral neck. This resulted in an increase in bending strength parameters that was borderline nonsignificant for Z_min. Over the 2-year period, cortical bone formation took place at the endosteal surface. The resulting increase in cortical area led to a significant increase in the measured buckling strength.

Our results in this group of women with severe osteoporosis, most of them previously treated with AR agents, are similar to previous QCT findings by McClung and colleagues in osteoporosis treatment–naive women.18 These authors found a 4.9% increase in trabecular vBMD as well as a significant 1.2% decrease in cortical vBMD at the femoral neck over an 18-month period of TPTD treatment.18 Our aBMD results, which showed a 4% increase compared with baseline, are also in agreement with earlier studies on TPTD treatment, reporting increases of 2% to 4% after 18 months for both pretreated19 and osteoporosis treatment–naive patients.4

The mechanism and mechanical relevance of the transient decrease in femoral neck BMD reported in DXA-based studies on TPTD treatment for AR-pretreated patients is difficult to assess with DXA. It has been hypothesized that these changes may have been caused by periosteal apposition of newly synthesized, non-fully mineralized bone20 and contrary that this might describe a real BMD loss associated with increased cortical porosity.4 Lindsay and colleagues21 showed in human bone biopsies that daily treatment with human parathyroid hormone(1-34) [hPTH(1-34)] is capable of stimulating new bone formation on the endosteal and periosteal surfaces in humans, providing a mechanism for increased cortical diameter. Similar results have been published in animal biopsy studies.22 Increase in the periosteal circumference as an indication for periosteal apposition has been reported for patients on TPTD treatment at the distal radius.23 In subjects not receiving TPTD, periosteal apposition has been observed at the aging femoral neck for men24–27 and women4, 24–26 and in femoral specimen studies.28, 29 However, a large DXA substudy of the Fracture Prevention Trial using hip structure analysis (HSA) reported that periosteal apposition appeared to be reduced in patients receiving TPTD in comparison with a placebo-treated group.4

In agreement with this study,4 we could not find evidence of periosteal apposition at the femoral neck in our TPTD-treated patients. Based on our findings, it seems unlikely that periosteal apposition can explain the early-phase vBMD loss in the inad. pre-Tx group of TPTD-treated patients. Instead, a real transient bone mineral loss takes place during the first months of treatment. Our QCT data document increases in trabecular density along with decreases in cortical density in the early phase of treatment. This decrease in total BMD is likely related to the increased endocortical porosity that results from the increased bone turnover during the first months of TPTD treatment,18 a well-described phenomenon in primates.30, 31 Our findings support the analysis done concerning the hip and humerus of ovariectomized monkeys showing that TPTD increased the rate of bone turnover and cortical porosity.30, 31 The improvement in total density that is observed in the second year of TPTD treatment is caused by increased trabecular density and, as the data from this and the primate study31 show, by the endosteal expansion of the cortical envelope.

To judge the effect of the observed areal and vBMD changes on the mechanical strength of the femoral neck, we specifically addressed three questions: (1) To what extent does the early bone loss compromise bone strength? (2) For which types of impacting forces does TPTD yield improvements in strength indicators at the end of the 24-month treatment period? and (3) Which patients benefit most/least from TPTD treatment?

Regarding the phase of early bone loss, we observed stable levels of strength indices with regard to both bending and buckling. The decrease in Z_max that accompanies the stable level of Z_min does not affect breaking strength because it only weakens the strongest bending mode of the bone, which is less important than Z_min. This led us to hypothesize that the calcium needed to mineralize bone in the trabecular bone compartment of the femoral neck may be taken in part from those parts of the cortex where this is strongest, that is, the regions affecting Z_max. This hypothesis fits well with the above-mentioned animal studies by Sato and colleagues31 and Burr and colleagues,30 who observed increased porosity under TPTD treatment of ovariectomized monkeys without significant alteration of bone mass or biomechanical strength compared with either ovariectomized or sham controls. Part of the explanation for this finding was that intracortical porosity was spatially localized along the endocortical surface, where the mechanical effect is much smaller than mineral loss along the periosteal region.30, 31 Therefore, the enhanced endocortical porosity was compensated by an increase in cortical area.30, 31 Our results indicate that the changes appear to occur at those cross-sectional sectors of the femoral neck where it is thickest and loss of strength is least critical.

There is still debate about whether the femoral neck fails mostly in bending or also in buckling mode.4, 32 We observed a significant improvement in buckling strength parameters, whereas increase in indices of bending strength (based on Z_min) did not quite reach significance. However, our study may lack the power to detect increases of bending strength indices on the order of around 2%. Age-associated bone adaptation by periosteal apposition leads to better maintenance of bending strength but fails to account for buckling risk.4 Especially the superolateral part of the femoral neck appears to be at risk for fracture in the mode of buckling.32 Reduced periosteal cortical expansion and endosteal growth as induced by TPTD therefore appear to be beneficial for both failure modes.

A limitation of our study is the lack of a control group. However, we were able to stratify our patient cohort according to different criteria, for example, baseline bone strength and prior osteoporosis treatment. We observed that the subjects who had the weakest bones at baseline benefited the most from TPTD treatment. This holds true for strength indicators of bending and buckling but not for plain BMD changes. For the buckling mode, a BR > 10 has been postulated to represent a high risk for failure in buckling mode.13 When we stratified our patients according to this criterion, we observed strong and significant differences in the increases in buckling strength measures during TPTD treatment. Twenty-four percent of patients with an initial BR > 10 improved below this threshold during the course of the study. It is unlikely that this effect is solely due to regression to the mean. When comparing the regression of changes in BR up to months 6, 12, and 24 versus baseline levels (or, similarly, for Z_min), the negative slope increased steadily over the three time points. If the effect were due to regression to the mean, no such increase in slope would be expected. In agreement with the concept of the mechanostat,33, 34 bone tissue needs mechanical loading to respond to treatment. These findings demonstrate that the bone makes very efficient use of the mineral pool available, even under TPTD treatment conditions.

The TPTD treatment response is affected by pretreatment with AR drugs, as shown by Ettinger and colleagues35 and in the EUROFORS Study.12 When investigating differences between the pretreatment groups, we found cortical density to be reduced at 6 months in all pretreatment groups, but only Tx-naive patients already showed trends toward endosteal apposition. As a consequence, a significant increase in strength indicators was observed. Pre-Tx patients, on the other hand, and inad. pre-Tx patients in particular did not show early compensatory apposition at 6 months. Although the measured loss of bending strength was only of 2.2% in the stronger plane (Z_max), this change reached statistical significance. The weaker plane (Z_min) was not significantly affected by the reduction of bending strength parameters even in this subgroup of patients, which is important to note because it challenges the conclusion of compromised bone strength that one would draw from the BMD values alone. In the later course of the study toward the 24-month visit, the pre-Tx and inad. pre-Tx patients showed significant increases in cortical CSA, with improvements in Z_min, Z_max, and BR. These increases tended to be lower than those observed in the treatment-naive group. Therefore, the increase in the strength parameters in this period could be interpreted as a delayed response to treatment following an early phase of inhibition of endosteal expansion in pre-Tx and inad. pre-Tx groups. The increased endosteal density might, for example, represent true endosteal apposition or filling of endosteal remodeling space. However, the analysis of the differential effects related to pretreatment has to be interpreted with caution given the limited size of the Tx-naive and AR-responder groups.

The investigation of the change in geometry, structure, and distribution of vBMD required a refinement of the standard QCT method. This concerns optimized scanning protocols and careful operator training. A slice thickness of 3 mm limits radiation exposure, but compared with 1 to 2 mm slices, larger partial-volume effects can be expected. Thus one could expect to yield even more accurate results on structural changes if smaller slice thickness were selected.

The eccentricity registration method previously used in ex vivo studies16, 32 worked well in this in vivo follow-up setting, better than the standard method of placing the VOI at the location of the narrowest cross section of the femoral neck. This is likely due to the small changes in cross-sectional area along the femoral neck.17 Eccentricity, on the other hand, varies by 100% along the neck axis, and it is both more sensitive and robust. Moreover, the method is largely automated, and for the most part, it reduces the impact of operator errors. This leads to better comparability between patients in comparison with the manual VOI positioning method. There are other steps in the analysis procedure that had to be adapted. Owing to partial-volume effects, we needed to employ lower thresholds for outlining the periosteal border of the cortex. This may have lead to underestimation of cortical expansion under TPTD treatment. However, QCT-pro software corrects for partial-volume effects to some extent, and reanalysis with higher constant thresholds did not lead to relevant differences in the results, which also shows the robustness of the method.

As shown recently by Rivadeneira and colleagues,36 aBMD and DXA HSA-derived BR values are closely correlated and appear to be stronger predictors for hip fracture risk than Z. Our data, however, show rather weak correlation between QCT-derived BR and aBMD (r² = 0.37) but higher correlation of QCT-based BR and total vBMD (r² = 0.83). Both, aBMD and vBMD do not correlate well with Z_min or Z_max (r² range 0.22–0.28). Therefore, it might be adequate to draw conclusions from vBMD on the risk of buckling to some extent, but neither vBMD nor aBMD values provide information on bending strength of the femoral neck cross section. Therefore, with the calculation of Z_min and Z_max values, QCT BIT offers additive information to vBMD and BR for the estimation of fracture risk. Furthermore, it has the potential to provide information on a more localized fracture risk in eight sectors of the femoral neck cross section in vivo32 and to assess the clinically important intertrochanteric region. At this location, CT-derived evaluation of trabecular density values in combination with CSA total values have been shown to correlate very well with fracture loads of the proximal femur in a study on human cadaver bones.37

We conclude that QCT analysis with the eccentricity registration method proved to be a sensitive method for longitudinal assessment of treatment effects in different bone compartments. Changes in femoral neck density, structure, and strength indicators in various failure modes can be evaluated in a semiautomatic fashion. Twenty-four months of TPTD treatment leads to increased trabecular density and cortical thickness that causes increases in bending and buckling strength at the femoral neck. TPTD seems to inhibit or at least reduce age-related periosteal apposition. Therefore, in patients on TPTD treatment, the improvement in bending strength measures does not come at the expense of greater susceptibility to buckling, as would be the case in aging. Temporary reductions in aBMD and vBMD during the first 6 months of TPTD treatment do not impair bone strength in bending and buckling, which can be explained by a redistribution of the mineral between stronger and weaker areas of the femoral neck.

Disclosures

TNN and FM are full-time employees of Eli Lilly & Company. CCG is a consultant for and has received research funding from Eli Lilly & Company. The rest of the authors state that they have no conflicts of interest.

Acknowledgements

The EUROFORS Study was funded by Eli Lilly & Company, Europe. Contributing investigators include: Germany: Bruno Allolio, Eberhard Blind, and Franz Jakob, Klinikum der Universität Würzburg; Klaus Badenhoop, Klinikum der Johann Wolfgang Goethe-Universität Frankfurt; Reimar R. Fritzen, Med. Klinik für Endokrinologie des Universitätsklinikums Düsseldorf; Thorsten Hennings, Frankfurt; Christian Kasperk, Universitätsklinikum Heidelberg; Jörn Kekow, Fachkrankenhaus für Rheumatologie und Orthopädie, Vogelsang-Gommern; Hans-Peter Kruse, Universitäts-Krankenhaus Eppendorf, Hamburg; Heiner Moenig, Universitätsklinikum Schleswig-Holstein, Campus Kiel; Rüdiger Möricke, Magdeburg; Helmut Radspieler, München; Jutta Semler, Immanuel Krankenhaus Berlin-Wannsee; Wolfgang Spieler, Zerbst; Nikolaus Vollmann, München; Andreas Wagenitz, Berlin. Spain: César Díaz-López, Hospital Santa Creu i Sant Pau, Barcelona; Jordi Farrerons, Hospital Santa Creu i Sant Pau, Barcelona; José Andrés Roman Iborra, Hospital Universitario Dr. Pesset, Valencia; Javier del Pino, Hospital Clínico, Salamanca.