Relationships among Dual-energy X-ray Absorptiometry, Bioelectrical Impedance and Ultrasound Measurements of Body Composition of Swine

In three separate studies (156 pigs total), dual-energy X-ray absorptiometry (DXA), bioelectrical impedance (BIA) and ultrasound were compared as methods for measuring live body composition of pigs at 60 and 100-110 kg BWt. DXA measured total body fat and lean content, BIA measurements of resistance (Rs) and reactance (Xc) were used to calculate total body lean mass and ultrasound measurements of backfat (BF) depth and longissimus muscle area (LMA) were used to calculate total carcass lean mass. Following the 100-110 kg measurements, the pigs were slaughtered and the half-carcass analyzed chemically for fat and water content. At 110 kg both DXA and ultrasound measurements were significantly correlated with the percentages of carcass fat and water, although correlations were higher for DXA. The correlations between DXA and BF measurements were higher at 110 kg than at 60 kg, whereas they were lower for DXA and LMA. For pigs measured at 100 kg there were high correlations between the DXA values and the BIA estimates for both percentage of fat-free lean mass (FFM %) and FFM kg. Furthermore, the correlations between the BIA estimates of FFM and carcass fat and water content were similar to those for DXA and the same carcass values. This study also provided a side-by-side comparison of the BIA and ultrasound lean measurements relative to DXA and carcass composition. The BIA lean measurement correlated more highly with both DXA and carcass composition than did the ultrasound lean measurement.


Introduction
In recent years, a variety of approaches has been used to probe the animal in an attempt to gather information on in vivo body composition.These include the use of ultrasound, x-rays, gamma rays, near-infrared rays, nuclear magnetic resonance, electrical impedance, electromagnetic conductivity, and neutron activation.Techniques that utilize these approaches include dual-energy X-ray absorptiometry (DXA), bioelectrical impedance (BIA) and real-time ultrasound (US).The purpose of the present study is to provide a sideby-side comparison of these three techniques for measuring body composition of swine.
The DXA technique has been used to measure the body composition of live pigs (MITCHELL & SCHOLZ 1997, 2008, MITCHELL et al. 2002, SCHOLZ & FÖRSTER 2006, BEE et al. 2007) and pig carcasses (MITCHELL et al. 1998).The total body DXA scan provides measurements of total body fat, lean, bone mineral and bone mineral density.The DXA measurement of fat content of the live pig is highly correlated to the fat content (r =0.915 for percentage fat and 0.989 for fat weight) of the chemically analyzed carcass MITCHELL et al. 1996).
In limited studies, BIA has been used to predict the fat-free mass of live pigs and carcasses (SWANTEK et al. 1992) and the Boston butt portion of pork carcasses (MARCHELLO & SLANGER 1992).It has also been tested for measuring the fat-free mass of lambs and lamb carcasses (BERG & MARCHELLO 1994, BERG et al. 1996, 1997, HEGARTY 1998, SÜSS et al. 2001), and the composition of steer carcasses (VELAZCO et al. 1999).The BIA procedure consists of placing two sets of electrodes at defined locations on either the live pig or carcass.Once the electrodes are in place, the impedance readings (Rs and Xc) are taken.The study by SWANTEK et al. (1992) reported correlations of −0.56 and −0.63 between Rs and fat-free mass (FFM) and 0.64 and 0.70 between Rs and percentage fat for live and carcass measurements, respectively.The correlations were not as good for Xc (−0.11, −0.08, −0.11, and 0.17, respectively).
Historically, ultrasound has become the most common in vivo technology in swine body composition assessment (KLIESCH et al. 1957, STOUFFER et al. 1961, HORST 1971).Several studies have reported equations for the prediction of percentage lean or lean cuts utilizing ultrasound measurements of live pigs (TERRY et al. 1989, GRESHAM et al. 1992, CISNEROS et al. 1996, MÜLLER & POLTEN 2004).These equations based on transverse or longitudinal scans include anywhere from one to four fat depth readings from a variety of locations, most include either LM depth or area and some include body weight.The reported accuracy of these equations ranges from an R² of 0.36 and residual standard deviation (RSD) of 3.17 to an R² of 0.83 and RSD of 1.67.
The advantages and disadvantages of these and other methods for measuring the body composition of swine have been discussed previously (MITCHELL & SCHOLZ 2004).In studies of human body composition there are numerous reports where techniques have been cross-validated, especially with DXA and BIA, however with swine there has been no prior study that provides a side-by-side comparison of the techniques evaluated here-in.

Material and methods
A total of 156 pigs were used in three separate studies.The pigs were of mixed genetic background and on standard diets.In the first study 99 pigs were measured by DXA and ultrasound at 60 kg; of those, 93 were measured again at 110 kg.Following the measurements at 110 kg, the pigs were euthanized and the half-carcass analyzed chemically for fat and water content.The second study consisted of 33 pigs; each pig was measured by DXA and BIA at 60 kg, 18 of the pigs were fed at maintenance for eight weeks then measured again (a total of 51 measurements at 60 kg) and finally all pigs were measured at 100 kg.After the measurement at 100 kg, the pigs were euthanized and the half-carcass analyzed chemically for fat and water content.In addition, for each of the half-carcasses the area of the longissimus muscle (LM) was determined at the level of the 10th rib and fat thickness (P2BF) over the LM was measured at 65 mm from the midline.The third study consisted of 24 pigs; each pig was measured by DXA, ultrasound, and BIA at 60 kg, 14 of the pigs were fed at maintenance for eight weeks then measured again (a total of 38 measurements at 60 kg) and finally all pigs were measured at 100 kg.After the measurement at 100 kg, the pigs were euthanized and the half-carcass analyzed chemically for fat and water content.

DXA measurements
Each pig was scanned by DXA for body composition analysis as described by Mitchell et al. (1996).The DXA scans were performed using either the Lunar (GE-Lunar, Madison, WI) DPXL (1st study) or the Lunar Prodigy (2nd and 3rd studies) densitometer.Pigs were fasted overnight and then anesthetized (500 mg ketamine, 80 mg tiletamine, 80 mg zolazepam and 333 mg xylazine per 100 kg body weight, i.m.) to prevent movement during the scanning procedure.The DXA scans provided measurements of total body fat, lean, and bone mineral content (BMC).Each pig was placed on the instrument in a prone position and a total body scan was performed.

Ultrasound measurements
Following the DXA scan, while the pig was still anesthetized and lying in the prone position, ultrasound images were obtained using an Aloka Model SSD-500V.The hair was clipped and the transducer, fitted to a stand-off guide, was placed at the level of the 10th rib.The ultrasound image was used to measure thickness of the backfat (BF, cm) layer and area of the longissimus muscle (LMA, cm 2 ).Fat-free lean mass was calculated using the NPPC (1999) formula:

BIA measurements
Following the DXA scan, also while the pig was still anesthetized and lying in the prone position, measurements of resistance (Rs, Ω) and reactance (Xc, Ω) were made using a four-terminal plethysmograph (BIA; Model BIA-101, RJL Systems, Inc.).The four needle electrodes were implanted as described by SWANTEK et al. 1992).Fat-free lean mass (FFM, kg) was calculated using Rs, Xc, live weight (LWt, kg), and body length (L, cm), based on the model reported by SWANTEK et al. (1992):

Chemical analysis
At slaughter, the head and viscera were removed, and the carcass was split at the midline.
The hair and feet remained on the carcass.The right half-carcass was analyzed chemically for lipid (FOLCH et al. 1957) and water (lyophilization) content.

Statistical analysis
Data were analyzed using Statgraphics Plus (Ver.5.1) multi-variable analysis which generated the mean and standard deviation for each variable and the Pearson product moment correlation for each pair of variables.Statistical significance of the correlation coefficients was based on P-values.

Results
The body weight and composition measurements of the pigs in the three studies are shown in tables 1-3.Within weight groups, there was close agreement for the DXA results among the three studies.The DXA measurements for percentage of body fat at 60 kg were 11.8, 12.1, and 11.7 and at 100-110 kg they were 19.0, 18.4, and 17.5 for the 1st, 2nd, and 3rd studies, respectively.Likewise, for pigs in the 100-110 kg group, there was close agreement based on chemical analysis of the half-carcass.The percentages of fat in the half-carcass for the 1st, 2nd, and 3rd studies were 25.9, 24.6, and 25.1, respectively.In addition to the DXA measurements, results of the ultrasound measurements of BF and LMA are shown in Tables 1 and 3 and results of the BIA measurements are shown in Tables 2 and 3. Rs BIA resistance measurement, Xc BIA reactance measurement, BIA-FFM fat free lean mass calculated using the BIA model of Swantek (1992), LMA longissimus muscle area, BF P2 backfat depth, SD standard deviation Correlation coefficients for body composition measurements using DXA, ultrasound and BIA are shown in tables 4-6.In the 1st study, there were significant correlations for all DXA and ultrasound measurements at both 60 kg (Table 4a) and at 110 kg (Table 4b).At 110 kg both DXA and ultrasound measurements were significantly correlated with the percentages of carcass fat and water.At both 60 and 110 the lowest correlations were for LMA.The correlations between DXA and BF measurements were higher at 110 kg than at 60 kg, whereas they were lower for DXA and LMA.In the 2nd study, at 60 kg (Table 5a) DXA measurements (with the exception of DXA lean kg) were significantly correlated with BIA reactance (Xc), but not resistance (Rs).However, at 110 kg (Table 5b) the DXA measurements were significantly correlated with Rs, but not Xc.At both weights the DXA measurements were significantly correlated with the BIA estimate of fat free lean %, although the correlations were higher at 110 kg.At 110 kg both DXA and BIA measurements were significantly correlated with the percentages of carcass fat and water.In the 3rd study, at 60 kg (Table 6a) there were significant correlations among the DXA, ultrasound, and BIA measurements, with exceptions for both Rs and Xc.At 110 kg (Table 6b) there were significant correlations among the DXA, ultrasound, BIA, and carcass measurements, with the exception of both LMA and Xc.

Discussion
The estimates of percentage of lean varied considerably, depending on the method of measurement.Based on averages for the three studies, for pigs at 60 kg the lean content measured by DXA was 85.8 %, by BIA (BIA-FFM) 61.6 % and by ultrasound (US-FFM) 40.65 %.For pigs measured at 100 to 110 kg the lean content by DXA was 79.7 %, by BIA 62.8 %, and by ultrasound 39.4 %.By chemical analysis, the average water content of the carcasses of pigs measured at 100 to 110 kg was 53.4 %.Assuming a protein content of 16 % (MITCHELL et al. 1998), this would translate to a lean content of 69 % for the carcass.
The reason for the discrepancies among these methods is that they each take a different approach for measuring »lean«.DXA, based on a three compartment model, measures total body fat and bone mineral content and the lean component includes »all other«.BIA is based on a two-compartment model consisting of fat mass and fat-free mass (FFM) as described by LUKASKI (1987).The formula (SWANTEK et al. 1992) for calculating FFM of pigs using BIA measurements was derived by subtracting the product of live body mass and the percentage of fat in the cold carcass from the live body mass.This calculation assumed that fat was distributed uniformly throughout the animal and was the same percentage found in non-carcass body parts as in the carcass.The formula for calculating carcass fat free lean (US-FFM) using live ultrasound measurements of BF depth and LM area (NPPC 1999) is based on total lipid-free (chloroform/ methanol extracted, FOLCH et al. 1957) lean that was dissected from the carcass.The DXA measurement of total body lean can be adjusted for bone content rather than BMC (PURSEL et al. 2004) by assuming a 24.14 % ash content for pork bones (FIELD et al. 1974).Furthermore, it has been demonstrated that there is a close relationship between live DXA measured lean and carcass lean (SCHOLZ et al. 2007, SUSTER et al, 2003), indicating that the live DXA lean measurement can be used to predict carcass lean.Also, SWANTEK et al. (1992) reported an equation for their BIA measurements that adjusted for live weight to offset discrepancies between carcass weights and live weights using average head and viscera weights and estimated blood loss.
It should be noted that with both DXA and ultrasound measurements, as expectedwith an increase in fat percentage, the percentage of lean decreased from the 60 kg to the 100-110 kg measurements, whereas with BIA there was an increase in the percentage of lean (fat free mass) in both studies (Tables 2 and 3).A later study by SWANTEK et al. (1999), using the same prediction equation as used here, reported high correlations between actual and BIA predicted FFM, but that BIA underestimated FFM in pigs ranging from 50 to 130 kg.The degree of underestimation was greater with smaller pigs, for example, for barrows at 50 kg BIA-FFM was underestimated by 16 %, but only 3.4 % at 130 kg and the percentage of BIA-FFM increased from 61.8 for 50 kg barrows to 64.7 for 90 kg barrows.This could explain the increase in percentage in BIA-FFM for 100 kg pigs compared to 60 kg pigs that was observed in the present study and suggests the need for separate prediction equations based on body size.
In previous studies DXA measurements of total body fat and lean were found to correlate with the percentages of carcass fat and lean with R 2 values ranging from 0.78 to 0.85 (SCHOLZ et al. 2007).The formula used in this study for calculating fat-free lean based on ultrasound measurements has a reported R 2 value of 0.777 (NPPC 1999).However, the use of ultrasound for assessing live animal composition is subject to a number of measurement errors (HOUGHTON and TURLINGTON 1992).For ultrasound measurements, the correlations between lean percentage and ultrasound fat depth readings range from −0.44 to −0.63 (ISLER & SWIGER 1968, ANDERSON & WAHLSTROM 1969).MERSMANN (1982) reported very low correlations between ultrasound LMA measurements and the percentages of either fat or nitrogen in the carcass.In the studies reported here where DXA was compared with ultrasound measurements (1st and 3rd studies, Tables 4 and 6) the DXA measurements were more highly correlated with BF than with LMA.The correlation between DXA and BF was higher for the 100-110 kg pigs than with the 60 kg pigs, whereas, the correlation between DXA and LMA was higher with the 60 kg pigs.Both DXA fat and BF measurements correlate negatively with lean measurements.The correlation between DXA measured percentage lean and the NPPC calculated percentage US-FFM (using ultrasound BF and LMA measurements) ranged from 0.41 to 0.73.In both studies the DXA measurements of body composition were more highly correlated with carcass fat and water content than were the measurements based on ultrasound readings.Similarly, in the 2nd study (Table 5b) there were higher correlations between DXA measurements and carcass fat and water content compared to the correlations between carcass BF and LMA measurements and carcass fat and water content.Likewise, SUSTER et al. (2003) reported that DXA values were more strongly related with chemically-determined carcass values than were carcass P2 BF measurements.In a study where DXA was used to measure the half-carcass, the DXA measure of fat percentage in the half-carcass correlated with average BF and P2 BF measurements with R 2 values of 0.64 and 0.42, respectively (MITCHELL et al. 1998).
Comparing DXA and BIA (tables 5a and 6a), for pigs measured at 60 kg there did not appear to be a consistent relationship between the DXA values and the BIA measurements of resistance (Rs), reactance (Xc), or the BIA estimate for kg of FFM.However, there were consistently high correlations between the DXA values and the BIA estimates for percentage of FFM.For pigs measured at 100 kg (Tables 5b and 6b) there were positive correlations between the BIA Rs values and both DXA fat (0.69 to 0.86) and carcass fat content (0.61 to 0.84).Conversely there were high negative correlations between BIA Rs values and both DXA lean (−0.72 to −0.86) and carcass water content (−0.58 to −0.83).Similarly, the study by SWANTEK et al. (1992) found moderately high correlations between the BIA Rs values and carcass fat (0.54 to 0.56) and carcass lean (−0.52 to −0.56), but low correlations between the BIA Xc values and both carcass fat (−0.11 to −0.22) and carcass lean (−0.11 to −0.14).For pigs measured at 100 kg there were high correlations between the DXA values and the BIA estimates for both FFM % and FFM kg; and the correlations between the BIA estimates of FFM and carcass fat and water content were similar to those for DXA and the carcass values.The 3rd study provided the only comparisons between BIA and ultrasound measurements (Table 6a and  b).For pigs measured at 60 kg there was a higher correlation between BIA reactance (Xc) and ultrasound measurements compared to BIA resistance (Rs) and ultrasound.Whereas, as observed for BIA (Xc and Rs) and DXA in the 2nd study (Table 5a and b) the opposite was true for pigs measured at 100 kg.At both 60 and 100 kg there were high correlations between the BIA and ultrasound estimates of fat-free lean.This study also provided the only side-by-side comparison of the BIA and ultrasound lean measurements relative to DXA and carcass composition.In both cases, the correlation coefficients were higher for BIA compared to the ultrasound values.
In conclusion, these studies found higher correlations among the measurements with pigs at 100-110 kg compared to pigs at 60 kg.The exceptions to this were the ultrasound LMA measurement and the BIA Xc measurement.For measurements at 100 -110 kg; with ultrasound, BF correlated more highly with other parameters than did LMA; and with BIA, Rs correlated more highly with other parameters than did Xc.Also at 100 -110, DXA (fat and lean) and BIA-FFM correlated more highly with chemical analysis of the half-carcass than did ultrasound (US-FFM).

Table 3
Composition measurements of pigs at 60 and 100 kg BW using DXA, BIA, ultrasound and chemical analysis as well physical measurements of half-carcass of 100 kg pigs (3rd study) Körperzusammensetzung von Schweinen bei 60 und 100 kg Lebendmasse aus DXA, BIA bzw.Ultraschall und chemischer bzw.planimetrischerAnalyse einer Schlachthälfte der