DXA reference values of the humanoid sheep model in preclinical studies

Animal husbandry and feeding

The animals were kept under the supervision of experienced animal keepers and veterinarians, who guaranteed species-appropriate husbandry. The sheep were kept in their primary flock and allowed to free range on the premises of the Central Research Facility of Goethe University Frankfurt am Main, Germany. Water and food were regularly available. Since DXA measurement can only be performed on the resting patient/animal, a general anesthesia was required for the animals. Before the measurements, two animals were placed in a common stall box. In the last twelve hours before anesthesia, they received only water and no food. The premedication for the succeeding DXA procedure was also given in the boxes. Before each bone scan the weight of the animals was determined. The sheep average weight was 65.5 (41–111) kg.

Experimental design

The current study investigated control groups (Sham animals) from two multicenter research projects (SFB TRR79) that aimed to study osteoporosis in sheep model. Therefore, the first group was named osteoporosis (T1) and the second osteoporosis (T2). However, no animal had any treatment that would induce osteoporotic bone status or other metabolic discrepancy in sheep bone. The third group utilized the data from a different project that aimed to investigate a novel interference screw for ACL reconstruction. Therefore, bilateral DXA scans were obtained before the first operation, and a unilateral scan was taken after 46 weeks. The animals from mentioned study are referred as (HA/Biocomposite) group. The DXA scans were performed on healthy Merino sheep, which can be considered representative of their species. No specific diet or surgery was conducted. In each study group scans were performed at the beginning of the study (0 months = 0M). Additionally, in group 1 (T1) after three and eight months (3M; 8M) and in group 3 (HA) after ten months (10M). In group 2 (T2) no additional scans were performed according to the study design (Table 2).

In total, 116 scans were taken on 74 Merino sheep using the device Lunar Prodigy (GE Healthcare, Little Chalfont, United Kingdom). The scans were analyzed by enCORE Software (v13.40, GE Healthcare). Each day the device was calibrated using the enCORE Software and the manufacturer provided phantom bodies, which were used for calibration.

The animals were sedated by intramuscular application of 10 mg/kg Ketaminhydrochlorid (Ketavet^® 10 mg/ml, Bela- Pharm GmbH und Co.KG, Vechta), 0,01 ml/kg Xylazin (Rompun^® 2%, Bayer AG, Leverkusen), 0,3 mg/kg Midazolam (Midazolam Rotexmedica 5 mg/ml, Rotexmedica GmbH, Trittau) and 0,01 mg/kg Atropin (Atropinsulfat B. Braun 0,5 mg/ml, B. Braun Melsungen AG, Melsungen). Afterwards, the sedation was kept upright by intravenous application of Propofol (Propofol 2% (20 mg/ml), Fresenius Kabi Deutschland GmbH, Bad Homburg) using a perfusor (50 ml/h). The sedated animals were placed on the DXA-device according to the manufacturer‘s guidelines.

DXA scans focused on BMD, BMC as well as absolute fat content (fat (g)) and relative fat content (fat (%)) in 4 clinically relevant regions of interest (ROIs), namely whole body (ROI 1), upper lumbar spine (ROI 7), left and right femoral head (ROI 8 and 9).

Independent reference group

For each merino land sheep, a T-score was calculated based on the BMD and the PBM values (Formula 1). Using Sheep from control groups from different studies enables researchers to obtain a baseline as a reliable alternative to the use of a new control group in future experiments. Such procedure is implemented in the clinical practice and if followed will reduce the number of animals required in new bone experiments. This study utilized scan measurements from animals between 48 and 72 months old to determine the baseline5 values.

Statistical analysis

The statistical evaluation of the bone scans was performed descriptively and quantitatively by using the software SPSS (V. 24.0, IBM, CA, ). The measured values of the parameters of BMD, BMC, age, weight, fat (%) and fat (g) were evaluated. The analysis focused on the correlation between age and BMD of the animals, PBM was calculated out of the measured BMDs of all scans. Data were examined for normal distribution and was found non-parametric. Therefore, Mann–Whitney U test was used to examine the significance considering a cutoff of p ≤ 0.05. Bar graphs were plotted as Mean ± Standard Error of Mean (Mean ± SEM) (Fig. 2).

Further analysis were carried out using the statistical software R (The Comprehensive R Archive Network (https://cran.r-project.org/)). The measured values were examined in relation to the various ROIs and study groups. Using the ‘ggplot2’ package built in R, scatter plots were created (Yavorska & Burgess, 2017). Furthermore, the corrplot package in R (https://cran.r-project.org/web/packages/corrplot/index.html) was used to display the correlation matrix in the form of a heat map (Fig. 3). Using the core Package in R (Cores of Recurrent Events) (Yavorska & Burgess, 2017) (Version 3.4.0, R Core Team, 2018) to assess regression line, quadratic regression, as well as non-parametric regression and the correlation coefficient according to Spearman’s rank correlation coefficient (Figs. 4A and 4B).

Descriptive statistics

BMD represents bone mineralization and is the most important value of osteodensitometry. In 682 BMD measurements, the average BMD was 1.2 (0.79–2.09) g/cm² (Table 3).

BMC as an area-independent value is another important parameter in osteodensitometry. For 682 measurements, the mean BMC value was 341.81 (4.1–2973.2) g.

In addition, the relative and absolute fat content within the respective ROI was determined from the DXA scans. Fat is found both in the soft tissue surrounding the bone and in the bone marrow. The values were used to check whether a high fat content has an influence on the measured bone density and is associated with the development of osteoporosis. The mean value of the relative fat content was 6.94 (4–29) % based on 673 measurements (Table 3). For the absolute fat percentage, based on 672 measurements, the mean value was 1613.35 (2–31978) g.

BMD measurements in different ROIs

The BMD values were considered separately for each ROI.

In ROI 1 (whole body), the BMD values averaged 1.15 (0.98–1.38) g/cm² (minimum-maximum) (Table 4). The values showed a normal distribution (skewness: 0.27; kurtosis: −0.35).

In ROI 7 (upper lumbar spine), the mean value was 1.2 g/cm². BMD values ranged between 0.92 g/cm² (minimum) and 1.68 g/cm² (maximum). The measured values were normally distributed (skewness: 0.37; kurtosis: −0.49).

For ROI 8 (left femoral head) the mean value was 1.26 (0.87–2.0) g/cm². The values were normally distributed (skewness: 0.55; kurtosis: 0.19) (Table 4).

Evaluation of BMD in ROI 9 (right femoral head) yielded values between 0.79 g/cm² (minimum) and 2.09 g/cm² (maximum). The mean value was 1.21 g/cm². The values were not normally distributed (skewness: 0.95; kurtosis: 1.62).

Differences between the experimental groups

The BMD values of the measured ROIs (1,7,8,9) were collected and compared separately for each group. In group 3 animals, HA and biocomposite were analyzed separately. Bonferroni correction was applied as a post-hoc procedure (Fig. 2), with which the BMD values of the experimental groups were checked for significant differences (p-value ≤ 0.05).

In ROI 1 (whole body) (Fig. 2A) BMD values differed significantly between the groups T1 and T2 {p < 0.001} and T1 and biocomposite {p = 0.016}. Also, between group T2 and biocomposite {p < 0.001}, T2 and HA {p < 0.001}.

In ROI 7 (upper lumbar spine) (Fig. 2B) the analysis showed significant differences between groups T1 and T2 {p < 0.001}, T2 and biocomposite {p < 0.001} and T2 and HA {p < 0.001}.

In ROI 8 (left femoral head) (Fig. 2C) the BMD values showed only significant differences between the values of groups T1 and biocomposite {p < 0.001}, and groups biocomposite and HA {p = 0.001}.

Statistical analysis of ROI 9 (right femoral head) (Fig. 2D) revealed significant differences between the results of groups T2 and HA {p = 0.016} and biocomposite and HA {p = 0.031}.

Correlations between BMD and other bone density parameters

Bone density cannot be considered independently of the other measured parameters. Spearman rank correlation coefficient was determined to assess the relationship between BMD and the parameters weight, age, BMC and fat. This was initially recorded independently of the animal groups for the region’s whole body, upper lumbar spine, and left and right femoral head (Table 5 and Heatmap: Fig. 3). The significance level was p ≤ 0.05. It was then also collected separately for each group.

In ROI 1: whole body, (Table 6; Fig. 2) significant correlation between BMD and body weight was observed (p ≤ 0.001). The correlation between BMD and BMC was significant and also for BMD with the absolute and relative fat content (p ≤ 0.001; p ≤ 0.001). However, the correlation of BMC and BMD with age were not significant. In the upper lumbar spine area, significant correlations were found between BMD and all other parameters surveyed (Table 6).

The evaluation of the ROI of the left femoral head showed a significant correlation between BMD and BMC values (Table 6). The correlations between BMD and the other measured values were not significant.

Also, in the ROI of the right femoral head, a significant correlation was seen between BMD and BMC (Table 6).

To calculate the T-score, a specific reference group was formed (see Table 5). For each of the four relevant ROIs, 48 measurements of animals aged between 48 and 72 months could be evaluated.

Regression models for the age-dependent course of bone density parameters

Using regression models, the age-dependent course of BMD and BMC was examined. For BMD, the linear regression model showed a significant steady increase in ROI 7, upper lumbar spine p = 0.04. For BMC, significant increases were found in the linear regression models of regions 1 (whole body), 7 (upper lumbar spine) and 8 (left femoral head). In all regions, there was a wide range of values (Table 7).

In the quadratic regression model, no significance was found in the determined regression curves for either BMD or BMC. Here, too, there was a widespread value. For BMD, the values ranged between (R² = 0.021) in ROI 1 and ROI 9 and (R² = 0.0696) in ROI 7. The BMC yielded coefficients of determination between (R² = 0.035) in ROI 9 and (R² = 0.14) in ROI 7.

Exemplary consideration of the course of BMD and BMC in ROI 7, upper lumbar spine

Due to higher incidence of osteoporotic fracture in lumbar vertebrae, the course of BMD and BMC in ROI 7: upper lumbar spine was seen, as an example (Fig. 4).

In the age-dependent view of BMD (Fig. 4A), a significant correlation was found in the Pearson method. Higher BMD values are more frequently observed from the age of about 50 months, while the dispersion of values above the linear regression line increases up to the age of 90 months. The maximum value is observed between 60 and 80 months of age. From an age of about 90 months, a wider dispersion of BMD values below the linear regression line is observed, although the average BMD value continues to rise.

The age-dependent pattern of the BMC also shows a significant correlation of the slope of the linear regression line in the Pearson method (Fig. 4B). From the 40th month onwards, the dispersion of values above the linear regression line increases. The maximum value is reached at 90 months. After that, a wider dispersion below the linear regression line is observed, while the average BMC value continues to rise, similar to the BMD.

The distribution of the quadratic regression curves of BMD and BMC was not significant.

Control groups in current large animal studies

Currently, studies using large animal models often do not use independent reference values. The implementation of the ARRIVE guidelines has contributed to a standardized documentation and publication of results from animal models (Kilkenny et al., 2010). Initial values must be redefined and/or an untreated control group must be used at the beginning of each study (Zarrinkalam et al., 2009). By applying the data from this study, the number of sheep used in animal models can be reduced by up to 50% (reduction). Nevertheless, the goal remains a complete replacement of animal experiments. In addition, the establishment of reference values can significantly improve the meaningfulness and transferability of results through a more targeted selection of test animals.

In the present study, significant differences in the BMD of the individual groups in each of the four ROIs were observed (Fig. 2). We conclude that individual control or sham groups per study are not sufficient. The large animal model of sheep does not represent a uniform genetic population, therefore the variation in values of the animals is greater than in rodents. To improve the comparability between the results of different bone density studies, it is necessary to establish an independent reference group.

Factors influencing BMD

In earlier studies a connection between BMD, osteoporosis and increased fat content in the bone could be observed (Heiss et al., 2017; Rosen & Bouxsein, 2006; Sekiya et al., 2004). Age also influences the bone density of the animals (Reinwald & Burr, 2011).

In DXA scans, therefore, in addition to BMD and BMC, the relative and absolute fat content in the ROI investigated was also measured (Miller et al., 2018).

As with BMD, significant differences were found between the four ROI compared to the other parameters. The large interindividual differences between the animals could not provide comparable results, so this should be regarded as further evidence of the need to form a reference group.

The influence of age on bone density parameters

The physiological course of bone metabolism in merino land sheep is still largely unexplored. In the literature, the age of three to four years is given for the sheep to reach skeletal maturity. The BMD and BMC values measured in this study were age correlated to determine the time of PBM and its further course. Our data show the first peak around the age of four years. Unlike humans, bone density does not remain on a plateau for several years after this first peak but increases steadily with age in sheep. Around the seventh year of life a further peak can be seen, followed by a plateau phase.

Influence on the design of future studies

Osteoporosis is usually a disease of old age (EPOS et al., 2002), which occurs decades after patients have reached skeletal maturity. Therefore, a suitable animal model should also have already reached skeletal maturity in order to allow a better transferability of the results to humans. The data of this study on the age-dependent course of the disease suggest that merino land sheep reach a plateau phase of bone density around the age of seven. In order to avoid unnecessary experiments on unsuitable animals, animals aged seven years or older should therefore be used in future studies on osteoporosis in merino land sheep. In addition to the four ROIs investigated here, the publication of the raw data from this study will allow reference values to be established for each bone region of the animals. For example, sheep are often used in neurosurgical preclinical research as a model of the human cervical spine (Long et al., 2018). Independent reference values of bone density in the cervical spine region could be obtained quickly and easily using our raw data.

For this study, female bone-healthy merino land sheep were examined without exception. The data are also made publicly available as open-source documents at http://DXAdb.glycosciences.de/. A database with reference values for male animals and/or animals with osteoporosis would simplify the planning of future studies, reduce possible sources of error and thus lead to more effective use of the animal models.

In terms of refinement, clinical studies can be planned more effectively, using animals who represent the bone characteristics of the mainly affected age group in humans.

Future studies can benefit from the results by choosing animals in specific age groups according to their study aims. Furthermore, the number of animals in studies can be reduced by sparing reference groups in osteoporotic studies according to 3R (Hooijmans, Leenaars & Ritskes-Hoitinga, 2010; Russell & Burch, 1959).