Agreement between dual x-ray absorptiometers using pencil beam and fan beam : indicators of bone health and whole-body plus appendicular tissue composition in adult athletes

1. Department of Medical Imaging and Radiation Therapy, School of Health and Technology, Polytechnic Institute of Coimbra, Coimbra, Portugal; 2. CIDAF UID/DTP/04213/2016, Faculty of Sport Sciences and Physical Education, University of Coimbra, Coimbra, Portugal; 3. Portuguese Foundation for Science and Technology SFRH/BD/101083/2014, Lisbon, Portugal; 4. Portuguese Foundation for Science and Technology SFRH/BD/121441/2016, Lisbon, Portugal; 5. LACAPS, Federal University of Alagoas (Ufal), Arapiraca, Brasil; 6. SHERU Sport, Health & Exercise Research Unit – Polytechnic Institute of Castelo Branco, Portugal; 7. Faculty of Physical Education and Sport, Lusófona University of Humanities and Technologies, Lisbon, Portugal; 8. Portuguese Foundation for Science and Technology SFRH/BPD/100470/2014, Lisbon, Portugal; 9. Instituto Politécnico de Coimbra, Coimbra, Portugal; 10. Faculty of Medicine, University of Coimbra, Coimbra, Portugal; 11. Metabolism, Nutrition and Exercise Laboratory, Physical Education and Sport Center, Londrina State University.


INTRODUCTION
Assessment of body composition corresponds to a branch of human biology 1, 2 and traditionally, body composition models comprise two compartments, 3 dividing body mass into fat and fat free mass.Technological advancements have allowed the appearance of nondestructive methods of evaluating body composition, being, in general, considered in vivo methodologies.Dual energy X-ray absorptiometry (DXA) assessment is based on a model of body composition evaluation, allowing quantification of fat tissue, lean soft tissue and bone mineral content (BMC), both for the whole body and for standardized segments (head; trunk, subdivided into ribs, pelvis and spine; upper limbs and lower limbs), and it is also possible to obtain the aforementioned components in regions of interest, termed ROI.From the clinical point of view, there is considerable interest in assessment of the proximal femoral area, often the target of prosthesis, by surgery, and additionally the lumbar spine (especially L1-L4), a segment with high informative value regarding the general state of the skeletal structure.
DXA technology has gained popularity over other technologies for its speed, low cost and reduced radiation exposure. 4, 5Previously, the available technology would use iodine-125 as a source of radiation, the methodology being single-photon absorptiometry (SPA), aimed to estimate the mineral content of long bones distal portion, such as radius and ulna, being described as valid. 6The SPA technology was used to obtain one of the most popular equations to determine the percentage of fat mass from the triceps and subscapular fat folds. 7DXA technology measures the attenuation of X-rays emitted at frequencies of two different energies, with iodine-125 being replaced by gadolinium-153, which emits radiation at 44 and 100 KeV.There are currently several DXA devices and manufacturers Hologic and Lunar are the most mentioned ones. 8-10DXA has been used mainly in the determination of BMC and, consequently, bone mineral density (BMD), by combining BMC and bone area.In fact, this is a criticism of the method, since the BMD should consider BMC by bone volume and not so much by area, assuming that there are inter-individual variations of those observed in relation to thickness measurement.Each of the manufacturers has developed models for reduction of radiological exposure, scanning time and precision gains [either by improving detec-tors resolution or by the technology of X-ray emission tubes, highlighting procedures "pencil beam" (PB) and "fan beam" (FB)].DXA, like the generality of the technologies, has assumptions, such as the heating of tissues during the procedures of application of the method, and values of water and carbon dioxide are lost, making it necessary to transform the BMC quantified to obtain the bone mineral itself. 5Additionally, it is assumed that fat and lean mass have attenuation constants and the exposure of said tissues to low and high voltages allows determining the proportions per unit area, assuming that the thickness in the antero-posterior plane does not affect the estimates.
Most of the studies focus on the comparisons between equipment from different manufacturers (Hologic vs. Lunar) or between equipment of the same manufacturer adopting different technologies for X-ray emission, i.e., PB as opposed to FB 11 or even among methodologies: DXA vs. computed tomography scan 12-14 or by the methodology of corporal potassium measurement. 15As a general rule, studies point to an error associated with DXA technology ranging from 1% to 3%. 16, 17The interest of Sports Sciences for body composition is due, in large part, to the search for indices of metabolic efficiency.Recently, studies have emerged to express the oxygen uptakes per unit of fat-free body mass 18 , since the fat corresponds to a nonrelevant metabolic tissue, or even considering oxygen consumption in the running pattern expressed per liter of lower limb volumetry (or even mL.L -1 .kg - ).The expression per appendicular mass of the lower limb is especially relevant when the movement pattern does not oblige the athlete under evaluation to support the whole body mass and this happens in the cycle ergometer to obtain maximum and average mechanical power 19 or in the dynamometer isokinetic technique to evaluate the moments of maximum strength (peak torques) of the muscles responsible for knee extension and flexion actions. 20However, the search for informative and noninvasive protocols on limbs volume and body mass has justified a line of study [21][22][23] devoted to more parsimonious methodologies in terms of costs and morosity (anthropometry) based on geometric models. 24The interest in total body composition and of particular segments has been especially emphasized for the lower limb or only for the thigh.Despite the popularity of the DXA technology, studies about com-peting equipment to estimate the determination accuracy of fat tissue, lean soft tissue and BMC are not abundant for the whole body and particular segments.][23] The present study aims to examine agreement among indicators resulting from the application of competing DXA from manufacturer Lunar, namely taking into account Lunar DPX-MD+ (PB technology) and Lunar iDXA (FB technology), having been carried out with healthy young adults from multiple sports.

Procedures and sample
The present study is cross-sectional in nature and adopts procedures recommended for human research 25 , having been approved by a Research Ethics Committee (REC) (CE/FCDEF-UC/00102014).The research design comprises several models of body composition assessment and, in the case of DXA, two pieces of equipment from manufacturer Lunar (DPX-MD+ and iDXA).All measurements were carried out on the same day and by qualified technicians in duly certified laboratory units.Sample subjects were informed about the study objectives and the nature of the procedures, having signed an Informed Consent Form (ICF).The final sample corresponds to 32 moderately active adult males (27.6 ± 10.1 years) practicing various sport modalities.

Anthropometry
Height and sitting height were measured to the nearest 0.1 cm (stadiometer Harpenden, model 98,603, Holtain Ltd., Crosswell, GB, and Harpenden Sitting Height Table ).Lower limbs length was calculated by the difference from previous measurements.Body mass was obtained by a SECA (model 770, Hanover, MD, USA) scale with a reduction of 0.1 kg.All measurements were performed by the same evaluator.

Plethysmography of displaced air
Information on body volume was obtained through displaced air plethysmography (BODPOD composition system, model BODPOD 2006, Life Measurement, Inc, Concord, CA, USA).Participants were evaluated twice consecutively following the manufacturer's instructions.Body density was calculated by dividing body mass (kg) by body volume (L).

Dual energy x-ray absorptiometry (DXA)
The two DXA instruments used were: Lunar DPX-MD+ (Software: enCORE version 4,00,145, GE Lunar Corporation, 726 Heartland Trail, Madison, WI 53717-1915 USA) and Lunar iDXA (enCORE version 13,60,033, GE Medical Systems Lunar3030, Ohmeda Drive, Madison, WI 53718 USA).For each of the equipments, a full body scanner was carried out.Additionally, area near the femur [information on the BMD of the femur neck, Ward triangle, trochanter and body of the femur (or shaft)] was carried out for each equipment.Subsequently, in the processing phase, an ROI (right thigh) was defined, as described in previous studies. 22,23 or the whole body and each and every segment, the information includes BMC and bone area to subsequently determine the BMD and the component of the fat tissue and lean soft tissue.Calibration was performed on the same day, before the first one, using the model ("phantom") and the procedures recommended by the manufacturer.
For the Lunar DPX-MD+ equipment, technical documentation specifies that it is a device with X-ray emission by PB technology, having a potential of 76 kV, with an accuracy of < 1% for BMD (whole body) and also 1% for body composition without specifying whether it applies to both tissues (fat tissue and lean soft tissue).In turn, one of the innovations announced in the iDXA technical leaflet concerns the FB issue.Although iDXA has an FB narrow-angle device and proceeds to multiple passages, the FB technology is understood as having higher error, which is due to interindividual variation of body size of those evaluated, especially in the sagittal measurements.Reduction of the FB angle and overlapping of scans, successively obtained, supposedly mitigates the error associated with FB equipment with wider angle (wide-angle FB).The X-ray in the Lunar iDXA equipment presents potential of 100 kV (slightly higher than Lunar DPX-MD+) and, beyond the FB narrow-angle technology , iDXA is different from previous equipment by having a high resolution detector (CZT-HD).

Statistical analysis
Analyses comprised descriptive statistics (range, mean, standard error of the mean, confidence intervals of the mean and standard deviation) for the entire sample (n = 32), as well as the verification of normality assumptions.Subsequently, intra-individual differences (time 1 and time 2) have been determined, parallel to the calculation of the technical error of measurement (TEM). 26Then, and based on TEM, the coefficient of variation was determined (CV %: expressed as a percentage of the combined mean of repeated measurements).Also based on associative statistics, it was possible to determine the intraclass correlation coefficient (ICC) and its 95% confidence interval.The differences of the means of the repeated measurements were evaluated based on the effect size (Cohen's d), which have been qualitatively interpreted as follows: 27 < 0.2 (trivial); 0.2-0.6 (small); 0.6-1.2(moderate); 1.2-2.0(large); 2.0-4.0 (very large); > 4.0 (extremely large).Statistical procedures were carried out using resources from software IBM SPSS v. 23 for Mac OS software (SPSS Inc., IBM Company, NY, USA).

RESULTS
Table 1 summarizes the characteristics of the sample.BMC, bone area used in determination of BMD, whole body lean soft tissue and whole body Comparison of results by two competing pieces of equipment (Table 3) suggests a substantial intraindividual difference for whole body BMC (mean of intraindividual difference = 110 g, magnitude of the wide effect: d = 1,312) and also for the bone area used in calculating BMD (mean intraindividual difference = 65 cm 2 , effect: d = 1.761).However, for the BMD intraindividual differences were negligible (-0,001 g.cm-2 ) and the magnitude effect was trivial.In addition, magnitude of the intraindividual difference was large for the fat tissue, with the average being 11.87 kg for equipment Lunar DPX-MD+ and 13.56 kg for equipment Lunar iDXA, corresponding to a mean of intraindividual differences of 1.70 kg (d = 1.612, magnitude of the wide effect).Differences for the lean tissue were trivial, that is, 0.04 kg.
BMD values for the proximal femoral area showed that for the femoral neck.The variation associated with the equipment corresponded to a trivial intraindividual difference of 0.001 (d = 0.031) and the magnitude of the differences among intraindividual means for the Ward triangle and trochanter was respectively, d = 0.512 (small magnitude) and d = 0.495 (small magnitude).For the shaft of the femur, the magnitude of the differences was moderate (d=0.656).For lean soft tissue indicators, differences were observed for all segments and ROI: d = 1.687 (large differences) for the trunk; d = 1.237 (large effect size) for upper limbs; d = 1.402 (also large effect) for the lower limbs.For the ROI, the lean soft tissue showed a vary large variation between equipments (d = 4.014).

Equipment
With regard to the ICCs, for all measures above, ICC > 0.900 was obtained.The CV % fluctuated between 2.3% and 2.7% for the measures used in cal-culation of BMD.For tissue, CV % is only 1.5% for the lean soft tissue component and 11.2% for the fat tissue component.For the variables of the proximal femoral area, TEM was always less than 5% of the combined mean [i.e., CV % equal to 2.4% for the femoral neck, 2.5% for the Ward triangle, 2.0% for the trochanter and 4.5% for the shaft], the data quality being corroborated by ICC coefficients always higher than 0.950.For lean soft tissue, CV % = 2.1

LEGENDS: bone mineral content (BMC), bine mineral density (BMD), LST (lean soft tissue)
and ICC = 0.972 for the trunk, CV % = 4.2 and ICC = 0.967 for the upper limbs; CV % = 3.2 and ICC = 0.966 for the lower limbs were observed.For the ROI, the lean soft tissue showed a higher variation between equipment (CV % = 8.43, ICC = 0.931).

DISCUSSION
In the present study, agreement among indicators resulting from the application of competing equipments used in DXA, one being a PB technology (Lunar DPX-MD+) and another, FB (Lunar iDXA) was examined.Regarding BMC, the bone area for determining BMD, BMD, fat tissue and lean soft tissue, healthy adults and sportsmen of various sports were assessed.Negligible differences were found for BMD, despite a trend for Lunar DPX-MD+ to produce higher values for BMC and also for the bone area.In a study of women between the ages of 21 and 80, 11 there was a trend for the FB mode to underestimate (by comparison to the PB mode) the bone area used to calculate BMD.This study, previously mentioned, was carried out with the equipment Hologic QDR-2000, that has the possibility of adopting the two modes mentioned above (FB and PB). 5 However, other studies 28, 29 shoed that among FB technology equipment, the subject's thickness constitutes a source of discrepancy.The Lunar manufacturer's first equipment had a beam angle of about 30 degrees, considering wide-angle FB, 30 having been replaced by narrow-angle FB equipment (in the Lunar Prodigy equipment the angle is 4.5 degrees), and considered several overlapping scans, which takes place in the Lunar iDXA (equally a narrow-angle FB, with the added advantage of being equipped with a CZT-HD high resolution detector).In the present study, a high ICC was always obtained between the aforementioned narrow-angle FB (iDXA) equipment and the PB (Lunar DPX-MD+) equipment.
For measurements of the proximal femoral area, which are widely used in clinical settings, BMD presented differences between equipment which fluctuated between trivial and moderate, although CV % and ICC confirmed a certain idea of data quality, especially for the femur neck, trochanter and Ward triangle, revealing shaft as a more problematic parameter.The literature confirms this trend for variation in the discrepancy between FB and PB modes, namely in a study of 63 women 11 which made it possible to conclude that there was an overestimation of +1.5% by FB in the lumbar spine, in parallel with an underestimation of -0.7% in the femoral neck and -1.8% in the trochanter.In the latter region, a correlation of +0.36 was found among the residuals of the two modes (FB/PB) and the body mass of the women For fat tissue, it is possible that variations in energy and data processing associated with each of the equipments contribute to a substantive intraindividual difference and caution is recommended for the acceptance of fat mass data, expressed in kg, from the DXA technology.In the present study, this variable showed the highest value for CV % (11.2%) and the magnitudes of intraindividual differences were large.Regarding lean soft tissue, the data between DXA-MD+ and iDXA appeared to be in agreement, except for ROI, where it is considered an additional source of error, that is, the error introduced by the observer, also involving data processing and not just acquisition.Use of DXA for determination of trunk fat, combined with measurements and thickness of subcutaneous fat folds, was tested as a protocol to obtain an intra-abdominal fat quantification using computed tomography 14 , having explained 91% of the interindividual variance, although with a CV % of 14.8%.
Appendicular composition has also been of interest in several studies using DXA technology.For example, a study of 41 male rugby players (16.3 to 20.7 years old) calibrated the geometric models by anthropometry to determine the lower limb volumetry, based on data obtained by DXA (Hologic, Explorer W, Waltham, Massachusetts, USA, software QDR version 12.4) to obtain data on fat mass and fat-free mass of the lower limb. 21Thus, it was possible to determine correlation coefficients of 0.81 and 0.90 between the anthropometric method and the DXA reference.More recently, the same researchers 22 has calibrated the geometric models based on two conical structures (only for the thigh) in 168 school-age children using the DXA equipment (Hologic Explor-er W, Waltham, Massachusetts, USA, software QDR version 12.4).In this last study, the appendicular thigh volume corresponded to an ROI defined between the transverse planes that pass between two anthropometric references: ischium and suprapatellar.Finally, another study 23 has been carried out with 42 adolescent volleyball players (14.0-17.9years) aiming at anthropometric calibration of thigh volumes obtained by anthropometry and DXA (Lunar DPX NT/Pro/MD+/Duo/Bravo).However, intra-observer reliability for the same equipment has not been determined, particularly with regard to ROI, which requires more expertise from the observer.

CONCLUSIONS
In general, the various parameters revealed good reproducibility and allowed to confirm a certain idea of the quality of the indicators resulting from the application of competing DXA equipment (Figure 1).Negligible differences were found for BMD, despite a trend for equipment Lunar DPX-MD+ to produce higher numbers for BMC and also for the area.It is recommended, however, that measurements of whole body fat tissue and in the case of lean soft tissue in the thigh, when obtained by DXA, be not taken as a criterion, but rather as a reference.Such an understanding has implications for the interpretation of intraindividual discrepancies that would comprise measurement error in each of the competing variables and not only in the predictive variable.

TABLE 1
DESCRIPTIVE STATISTICS FOR THE TOTAL SAMPLE AND TEST FOR NORMALITY ON ILLUSTRATIVE VARI-ABLES (N=32) tissue, are presented in Table 2. Also the outputs related to BMD in the proximal femoral area are presented in Table 2 [femur neck, triangle of ward, trochanter and shaft].Table 2 includes data from Lunar DPX-MD+ and Lunar iDXA.Violation of assumptions of normal distribution was uniquely noted for fat tissue with regard to the whole body.