**7. Limitations/impediments for use of DXA in clinical practice**

Even though there are a variety of important applications of DXA to veterinary research and clinical practice, there are a number of logistical issues that preclude its widespread use. These limitations include the expense to purchase, operate, and maintain DXA equipment, the space required to house a unit, the time to scan a subject, the need to restrain the test subjects during scanning, and the potential for certain confounding variables to influence accurate/precise estimates of body composition as a result of technological limitations of DXA. In this section we discuss these limitations and offer some potential solutions.

#### **7.1 Expense**

106 A Bird's-Eye View of Veterinary Medicine

exception is with turtles. DXA's poor ability to estimate fat content is a result of the relatively

Nagy, 2003 Snake Whole body Euthanasia No (2) 0.6 9.2 1.0 NR

2004b Iguana Femur1 Anesthesia Yes (5) NR NR NR 1.7

2004b Iguana Head Anesthesia Yes (5) NR NR NR 1.3

et al., 2010 Turtle Whole body Anesthesia No (2) 1.05 28.54 1.00 0.97

Rat Whole body Anesthesia Yes (3) 2.88 12.16 6.34 NR

Rat Humerus Excised bone No (6) NR NR 0.90 0.76

Rat Humerus Excised bone Yes (6) NR NR 1.32 0.86

Vole Whole body Euthanized Yes (>5) 1.6 6.8 2.3 3.6

et al.,1999 Pig Whole body Anesthesia (3) 0.72 2.37 1.12 NR

et al., 2004 Bird Whole body Euthanized ? (2-3) 0.47 1.71 NR NR

carcass Euthanized No (3) 0.94 13.51 1.91 NR

spine Anesthesia Yes (3) NR NR NR 7.8

Monkey Whole body Anesthesia Yes (5) 2.3 10.3 1.2 NR

<sup>1994</sup>Dog Whole body Anesthesia No (6) 0.51 1.55 1.40 0.79

Table 1. Literature review of the precision, as determined by mean intraindividual coefficients of variation, of DXA estimates of lean tissue mass, fat mass, bone mineral content, and bone mineral density in various non-human vertebrates. NR = not reported.

restraint ? (2-3) 1.28 4.92 NR NR

Euthanized ? (2-3) 0.16 2.06 NR NR

Subject moved between scans? (#scans)

Spine Anesthesia Yes (5) NR NR NR 1.6

Lean Tissue Mass

Anesthesia Yes (3) 0.86 2.20 1.60 0.84

CV%

Bone Mineral Content

Bone Mineral Density

Fat Mass

high proportion of bone in turtles (Stone et al., 2010, discussed below).

Treatment

Animal Location Subject

Whole body excluding head

Source Study

2004b Iguana Lumbar

Zucker

Lewis

Lewis

al., 1998 Pig Processed

et al., 2006 Rabbit Lumbar

Rhesus

Table adapted from Stone (2009).

et al., 2004 Bird Whole body Physical

Whole body (feathers removed)

Secor &

Zotti et al.,

Zotti et al.,

Zotti et al.,

Stone

Nagy & Clair, 2000 Mouse

Rose et al., 1998

Kastl et al., 2002

Kastl et al., 2002

Stevenson & van Tets, 2008

Elowsson et

Lukaski

Korine

Korine

Korine et al., 2004 Bird

Castaneda

Black et al., 2001

Toll et al.,

In most veterinary practices the purchase of DXA equipment is likely to be cost prohibitive. The cost of a new unit averages \$35,000 USD (Walpert, 2000). Additionally, there are a number of hidden costs such as software upgrades, equipment repair and maintenance, technician training, and remodeling costs associated with installation (e.g. electrical, space, etc.) Also, considering that the majority of diagnosis in a veterinary clinic would utilize traditional X-ray units, it is unlikely that clinics will purchase both traditional X-ray equipment and a dual-energy X-ray absorptiometer. Even though the costs of purchasing a DXA unit might be prohibitive, alternatives might exist. Potential users might be able to contract DXA services from local clinical or academic institutions.

#### **7.2 Size**

Space limitations are potentially major impediments for use of DXA in small animal practice. Since most DXA scanners are designed to be large enough to perform a whole body scan of an adult human, the housing of DXA equipment necessitates a dedicated room, which will likely deter or preclude its use in most small animal veterinary businesses. Some manufacturers, after recognizing this limitation, have designed DXA models that double as an exam table when not in use. Smaller models exist (e.g. PIXImus, GE Medical Systems), but are limited to small rodent-sized species, and not likely useful for most veterinary applications. Despite the space demands that a full-size DXA scanner necessitates, the size offers potential for its use with large-animal practices and has already proven useful for food-industry research. Even though there are benefits of the large scanner size for these applications, there are upper limits in body size that DXA can handle. Although DXA has been used previously with horses and cattle, its uses have been limited to analyses of bone density on excised bones (Secombe et al., 2002; Zotti et al., 2010). Currently, scanners are not large enough to allow for full-body scanning of larger animals, with the exception of carcass analysis.

#### **7.3 Time**

The use of DXA incurs a variable, and in some cases a significant, time component; however, the total time involved might not vary much more than it takes to produce a traditional X-ray radiograph. The total time it takes to scan an individual will vary depending on the type of equipment, its application, and the species under investigation. The time involved in producing a completed DXA scan of a patient involves four separate phases, with the vast majority of time allocated to machine and subject preparation. Phase 1 includes unit powering, calibration, and quality control. Phase 2 includes subject

Use of Dual-Energy X-Ray Absorptiomtetry (DXA) with Non-Human Vertebrates:

collected to confirm the presence of sand in the intestine.

Application, Challenges, and Practical Considerations for Research and Clinical Practice 109

nutrients (Walde et al., 2007). In this study we investigated whether ingestion of intestinal

To determine the effects of intestinal artifacts on DXA estimates of body composition, we performed DXA analysis on 55 captive box turtles, *Terrapene carolina*. This captive population of turtles was housed outdoors in a pen with a sand substrate. Prior to scanning, straight carapace length (SCL) and mass were recorded for each individual. All specimens, who were not fasted prior to analysis, were then cooled for a minimum of eight hours at 4oC to prevent movement artifacts while scanning. Following the same procedure for turtles as Stone et al. (2010), scanning was performed using a Hologic QDR-4500A fan-beam scanner equipped with a small-animal software program on three occasions: 13 May 2004, 15 October 2005, and 10 May 2005. During each scanning day, duplicate scans, without movement between scans, were performed on each individual. If the same individual was scanned on multiple dates, only scans from the first day were used in analyses. Prior to statistical analyses, duplicate DXA estimates of body composition (bone mineral content (BMC), fat mass (FM), lean tissue mass (LTM), and body mass (BM)) were averaged for each individual. All turtles were qualitatively scored based on the amount of sand in the gastrointestinal tract. Turtles were scored as having zero (n=27), moderate (n=20), or heavy levels (n=8) of intestinal artifacts (Figure 7). Following DXA analyses, fecal samples were

Fig. 7. DXA scans of box turtles (*Terrapene carolina*) showing, from left to right, (a.) zero, (b.)

To examine the effects of intestinal artifacts on DXA estimates of bone mineral content and lean tissue mass, we performed analysis of covariance (ANCOVA). Prior to analyses, we log10 transformed bone mineral content and lean tissue mass (dependent variables) to linearize the relationship between covariate (straight carapace length) and the response variable. Prior to analysis, all assumptions of ANCOVA were tested and not violated, including the assumption of parallelism (BMC, F2,48=0.24, p>0.05; LTM, F2,48=0.11, p=0.89). To compare fat mass among turtles with different artifact scores we used ANOVA, because body size did not significantly influence fat mass in this study. Data were left untransformed. For all analyses, when significant differences among intestinal artifact levels were found, Tukey multiple-comparisons were used to determine where differences occurred. Statistical analyses were performed using the general linear model procedure on

moderate, and (c.) heavy levels of intestinal artifacts (yellow circles).

Minitab version 13.2.

artifacts (sand in this case) impacts DXA estimates of body composition in turtles.

preparation, handling, orientation, and restraint. Phase 3 includes subject scanning time. Phase 4 is computational analyses. Subject scanning time can vary depending on a variety of factors including scanner brand, beam type (pencil vs. fan-beam), scanner resolution, and subject size/region of interest. Phase 1 is typically performed once each day that scanning takes place. The total time to power the unit and perform calibrations and quality control will likely vary depending on the unit manufacturer. In our experience this takes approximately 20-30 minutes. Phase 2 is likely to vary drastically with the patient under investigation due to the nuances on each organism's physiology and their body size. For instance, larger organisms pose more of a challenge to manipulate and orient on the scanner bed and may require more advanced planning. Phase 3 will vary primarily with patient size. For small patients such as rodents, where high-resolution small-animal software is necessary, scan time will take up to a few minutes. Scan time will increase with body size of the subject. The completion of DXA computational analyses (Phase 4) exemplifies the power of this tool. DXA nearly instantaneously calculates body composition at the completion of scanning. The relatively rapid scanning time combined with quick analyses makes DXA a potentially powerful tool in a clinical setting.

#### **7.4 Need for chemical/physical restraint**

Despite the relatively rapid scanning time of DXA, the accuracy and precision of DXA estimates of body composition are sensitive to the movements of subjects on the scanner bed (Cawkwell, 1998; Engelke et al., 1995). Due to this caveat, chemical or physical restraint of subjects is typically required with the use of DXA. The choice and efficacy of a particular method of restraint will vary among taxonomic groups. Use of restraint for full - body scanning of mammals and birds will invariably require general anesthesia to prevent movement artifacts. In birds, other methods of restraint have been used. Korine et al. (2004) covered the heads of small birds to immobilize them during scanning, but they found this method decreased precision in estimating lean tissue and fat mass compared to estimates determined after euthanasia (Table 1). In other exotic animals, reptiles in particular, additional methods are available. Cooling of body temperature is an effective method to immobilize subjects during scanning (Stone et al., 2010). This method requires planning on the part of the practitioner because safe cooling of core body temperature can take several hours for large reptiles. Despite its potential as a viable source of restraint, cooling of body temperature in reptiles tended to result in less precise estimates of body composition compared to anesthesia and euthanasia. (Stone et al., 2010)

#### **7.5 Intestinal artifacts: A case study on the ingestion of foreign particles and their associated impacts on DXA estimates of body composition**

DXA estimates of body composition from whole body scans are likely to be influenced by superimposing the contents of the gastrointestinal tract with the composition of the subject's tissues. Even though the effects of intestinal contents are thought to have a negligible impact on DXA estimates of body composition in humans, the consumption of certain items in animals might impact DXA estimates. For instance, the consumption of calcified particles such as sand, stones, or bone might influence estimates of bone mineral content because they will result in similar beam attenuation as bone. In reptiles and birds the consumption of calcified particles is common as these items might aid in digestion or provide particular

preparation, handling, orientation, and restraint. Phase 3 includes subject scanning time. Phase 4 is computational analyses. Subject scanning time can vary depending on a variety of factors including scanner brand, beam type (pencil vs. fan-beam), scanner resolution, and subject size/region of interest. Phase 1 is typically performed once each day that scanning takes place. The total time to power the unit and perform calibrations and quality control will likely vary depending on the unit manufacturer. In our experience this takes approximately 20-30 minutes. Phase 2 is likely to vary drastically with the patient under investigation due to the nuances on each organism's physiology and their body size. For instance, larger organisms pose more of a challenge to manipulate and orient on the scanner bed and may require more advanced planning. Phase 3 will vary primarily with patient size. For small patients such as rodents, where high-resolution small-animal software is necessary, scan time will take up to a few minutes. Scan time will increase with body size of the subject. The completion of DXA computational analyses (Phase 4) exemplifies the power of this tool. DXA nearly instantaneously calculates body composition at the completion of scanning. The relatively rapid scanning time combined with quick analyses makes DXA a

Despite the relatively rapid scanning time of DXA, the accuracy and precision of DXA estimates of body composition are sensitive to the movements of subjects on the scanner bed (Cawkwell, 1998; Engelke et al., 1995). Due to this caveat, chemical or physical restraint of subjects is typically required with the use of DXA. The choice and efficacy of a particular method of restraint will vary among taxonomic groups. Use of restraint for full - body scanning of mammals and birds will invariably require general anesthesia to prevent movement artifacts. In birds, other methods of restraint have been used. Korine et al. (2004) covered the heads of small birds to immobilize them during scanning, but they found this method decreased precision in estimating lean tissue and fat mass compared to estimates determined after euthanasia (Table 1). In other exotic animals, reptiles in particular, additional methods are available. Cooling of body temperature is an effective method to immobilize subjects during scanning (Stone et al., 2010). This method requires planning on the part of the practitioner because safe cooling of core body temperature can take several hours for large reptiles. Despite its potential as a viable source of restraint, cooling of body temperature in reptiles tended to result in less precise estimates of body composition

**7.5 Intestinal artifacts: A case study on the ingestion of foreign particles and their** 

DXA estimates of body composition from whole body scans are likely to be influenced by superimposing the contents of the gastrointestinal tract with the composition of the subject's tissues. Even though the effects of intestinal contents are thought to have a negligible impact on DXA estimates of body composition in humans, the consumption of certain items in animals might impact DXA estimates. For instance, the consumption of calcified particles such as sand, stones, or bone might influence estimates of bone mineral content because they will result in similar beam attenuation as bone. In reptiles and birds the consumption of calcified particles is common as these items might aid in digestion or provide particular

potentially powerful tool in a clinical setting.

**7.4 Need for chemical/physical restraint** 

compared to anesthesia and euthanasia. (Stone et al., 2010)

**associated impacts on DXA estimates of body composition** 

nutrients (Walde et al., 2007). In this study we investigated whether ingestion of intestinal artifacts (sand in this case) impacts DXA estimates of body composition in turtles.

To determine the effects of intestinal artifacts on DXA estimates of body composition, we performed DXA analysis on 55 captive box turtles, *Terrapene carolina*. This captive population of turtles was housed outdoors in a pen with a sand substrate. Prior to scanning, straight carapace length (SCL) and mass were recorded for each individual. All specimens, who were not fasted prior to analysis, were then cooled for a minimum of eight hours at 4oC to prevent movement artifacts while scanning. Following the same procedure for turtles as Stone et al. (2010), scanning was performed using a Hologic QDR-4500A fan-beam scanner equipped with a small-animal software program on three occasions: 13 May 2004, 15 October 2005, and 10 May 2005. During each scanning day, duplicate scans, without movement between scans, were performed on each individual. If the same individual was scanned on multiple dates, only scans from the first day were used in analyses. Prior to statistical analyses, duplicate DXA estimates of body composition (bone mineral content (BMC), fat mass (FM), lean tissue mass (LTM), and body mass (BM)) were averaged for each individual. All turtles were qualitatively scored based on the amount of sand in the gastrointestinal tract. Turtles were scored as having zero (n=27), moderate (n=20), or heavy levels (n=8) of intestinal artifacts (Figure 7). Following DXA analyses, fecal samples were collected to confirm the presence of sand in the intestine.

Fig. 7. DXA scans of box turtles (*Terrapene carolina*) showing, from left to right, (a.) zero, (b.) moderate, and (c.) heavy levels of intestinal artifacts (yellow circles).

To examine the effects of intestinal artifacts on DXA estimates of bone mineral content and lean tissue mass, we performed analysis of covariance (ANCOVA). Prior to analyses, we log10 transformed bone mineral content and lean tissue mass (dependent variables) to linearize the relationship between covariate (straight carapace length) and the response variable. Prior to analysis, all assumptions of ANCOVA were tested and not violated, including the assumption of parallelism (BMC, F2,48=0.24, p>0.05; LTM, F2,48=0.11, p=0.89). To compare fat mass among turtles with different artifact scores we used ANOVA, because body size did not significantly influence fat mass in this study. Data were left untransformed. For all analyses, when significant differences among intestinal artifact levels were found, Tukey multiple-comparisons were used to determine where differences occurred. Statistical analyses were performed using the general linear model procedure on Minitab version 13.2.

Use of Dual-Energy X-Ray Absorptiomtetry (DXA) with Non-Human Vertebrates:

the gastrointestinal tract. Data are presented as mean ± 1SE.

of the gastrointestinal tract prior to DXA scanning.

**7.6 Taxon-specific considerations** 

**8. Conclusions** 

Application, Challenges, and Practical Considerations for Research and Clinical Practice 111

Fig. 10. Mean fat mass estimates for turtles with no sand, moderate sand, and heavy sand in

The results of this study suggest that DXA estimates of bone mineral content are influenced by ingestion of mineralized foreign particles. DXA tended to overestimate bone mineral content in individuals that ingested sand. Therefore, it is important to consider whether individuals have access to foreign particles when utilizing DXA to quantify bone mineral content in exotic animal practice or wildlife research involving birds or reptiles. In cases where ingestion of foreign particles is possible, fasting may be required to ensure evacuation

Technological limitations exist with the use of DXA with certain taxa. As we discussed previously, DXA is unable to directly estimate the proportion of lean, fat, and bone when they occur concurrently at a single pixel on the scan. This is a limitation with the use of two peak energies. To distinguish all three tissue components a third peak energy would be required (Swanpalmer et al., 1998). DXA overcomes this limitation by using neighboring pixels without bone to estimate the proportion of fat and lean tissue and then applies them to pixels containing bone. This limitation presents a serious impediment for the use of this technique with species that contain a high proportion of bone. Stone et al. (2010) found that DXA was unable to accurately or precisely predict fat mass in turtles as a result of the high proportion of pixels containing bone. Despite the limitation, DXA was able to effectively estimate bone and lean tissue mass, and so this methodology could be useful in a clinical setting for predicting and monitoring metabolic bone diseases in turtles (Stone et al, 2010).

Dual-energy X-ray absorptiometry can effectively quantify lean tissue, fat, and bone mineral mass in humans and most animals. As a result of its high degree of precision and accuracy, it has proven useful for research in fracture risk and healing, obesity, metabolism, pathology, and nutrition. Despite the plethora of potential and realized uses for DXA in veterinary research, the use of DXA remains restricted, primarily, to the human health industry or animal research. Currently, a number of limitations exist that prevent the

DXA estimates for bone mineral content were significantly different among turtles with intestinal artifact scores (ANCOVA; F2,50=3.12, p=0.05). Bone mineral content was not different between turtles with zero and moderate levels of intestinal sand (T=1.12, p=0.52), as well as between turtles with moderate and heavy levels (T=1.62, p=0.25). A significant difference was found in mean BMC between turtles with zero and heavy levels of intestinal artifacts (T=2.48, p=0.04; Figure 8).

Fig. 8. Least square means of log-transformed DXA estimates of bone mineral content for turtles with no sand, moderate sand, and heavy sand in the gastrointestinal tract. Significant differences (P<0.05) are indicated by different letters. Data are presented as mean ± 1SE.

DXA estimates for lean tissue mass were not significantly different among turtles with different intestinal artifact scores (F2,50=1.73, p=0.19; Figure 9).

Fig. 9. Least square means of log-transformed DXA estimates of lean tissue mass for turtles with no sand, moderate sand, and heavy sand in the gastrointestinal tract. Data are presented as mean ± 1SE.

The effects of intestinal artifacts on estimates of fat mass followed similar trends to lean tissue mass. We observed a no significant effect of intestinal contents on DXA estimates of fat mass (F = 1.76, P=0.18; Figure 10).

Fig. 10. Mean fat mass estimates for turtles with no sand, moderate sand, and heavy sand in the gastrointestinal tract. Data are presented as mean ± 1SE.

The results of this study suggest that DXA estimates of bone mineral content are influenced by ingestion of mineralized foreign particles. DXA tended to overestimate bone mineral content in individuals that ingested sand. Therefore, it is important to consider whether individuals have access to foreign particles when utilizing DXA to quantify bone mineral content in exotic animal practice or wildlife research involving birds or reptiles. In cases where ingestion of foreign particles is possible, fasting may be required to ensure evacuation of the gastrointestinal tract prior to DXA scanning.

#### **7.6 Taxon-specific considerations**

Technological limitations exist with the use of DXA with certain taxa. As we discussed previously, DXA is unable to directly estimate the proportion of lean, fat, and bone when they occur concurrently at a single pixel on the scan. This is a limitation with the use of two peak energies. To distinguish all three tissue components a third peak energy would be required (Swanpalmer et al., 1998). DXA overcomes this limitation by using neighboring pixels without bone to estimate the proportion of fat and lean tissue and then applies them to pixels containing bone. This limitation presents a serious impediment for the use of this technique with species that contain a high proportion of bone. Stone et al. (2010) found that DXA was unable to accurately or precisely predict fat mass in turtles as a result of the high proportion of pixels containing bone. Despite the limitation, DXA was able to effectively estimate bone and lean tissue mass, and so this methodology could be useful in a clinical setting for predicting and monitoring metabolic bone diseases in turtles (Stone et al, 2010).
