1、Original ResearchPhantoms for Cross-Calibration of Dual Energy X-rayAbsorptiometry Measurements in InfantsMouhanad Hammami, MD, Jean-Charles Picaud, MD, PhD, Christoph Fusch, MD, Elaine M. Hockman, PhD,Jacques Rigo, MD, and Winston W.K. Koo, MBBS, FACNDepartments of Pediatrics, Obstetrics and Gyneco
2、logy (M.H., W.W.K.K.), Computing and Information Technology (E.M.H.),Wayne State University, Detroit, Michigan, Neonatal Unit and Human Nutrition Research Center, University of Lyon, Lyon,FRANCE (J.-C.P.), Neonatal Unit, University Childrens Hospital, Greifswald, GERMANY (C.F.), Neonatal Unit, Unive
3、rsity ofLiege, Liege, BELGIUM (J.R.)Key words: phantom, body composition, fat, lean, bone, infantObjective: To test the suitability of phantoms to cross-calibrate body composition measurements in smallsubjects among different dual energy X-ray absorptiometry (DXA) instruments.Methods: A set of four
4、phantoms with total weights 1520g, 3140g, 4650g and 7490g were made with lowcost and easily available materials. Each phantom was made from assembling polyethylene bottles (100 to 1000mL) filled with either pure olive oil or electrolyte solution in different combinations, and borosilicate tubes (3an
5、d 5 mL) and flexible polypropylene tubing filled with calcium carbonate. Triplicate measurements of each ofthe four phantoms were performed with three pencil beam densitometers made by the same manufacturer(Hologic Inc., Waltham, MA): two QDR 2000 (University of Liege, Liege, Belgium, and Wayne Stat
6、eUniversity, Detroit, Michigan) and a QDR1500 (University Childrens Hospital, Greifswald, Germany) usinginfant whole body-scanning mode and analyzed with software V5.73P.Results: DXA measured total weight, or bone, lean and fat masses, from one center were highly predictiveof DXA measurements from t
7、he other centers with an adjusted r2of 0.94 to 1.00, p H11021 0.001. This was the casewhether the measurements from single scan or from average of triplicate scans were used in the analysis.Conclusions: Systematic corrections, in the form of linear transformations, are possible to allow comparisonof
8、 clinical data generated from different centers. Different size phantoms can be made to accommodate thevarying range of weights and body composition of study subjects.INTRODUCTIONDual energy X-ray absorptiometry (DXA) is the recognizedstandard for bone measurements and is increasingly being usedfor
9、the measurement of soft tissue body composition in adults1 and children 2 including infants 3. However, it is wellknown that DXA measurements varied with instruments fromdifferent manufacturers 46 and even within the same man-ufacturer 7 because of differences in scanner design, materialsused for ca
10、libration and analysis algorithms. Thus, in order todetermine the comparability of results generated from differentcenters, it is essential to test all the properties of scannerperformance as a whole with the use of phantoms and/or humansubjects. With increasing availability of DXA technique forstud
11、ies in infants, it is imperative to determine whether differ-ent instruments used to measure body composition in infantscan be cross-calibrated to allow meaningful comparison of datagenerated from different institutions, although no such studyhas been reported. We therefore aim to test the suitabili
12、ty of aset of phantoms to cross-calibrate body composition measure-ments in small subjects among different DXA instruments.METHODSPhantomsPure olive oil (Salov North America Corp, Hackensack,NJ), an electrolyte solution containing a mixture of sodiumAddress reprint requests to: Dr. Winston Koo, Depa
13、rtment of Pediatrics, Hutzel Hospital, 4707 St Antoine Blvd, Detroit, MI 48201. E-mail: wkoowayne.eduJournal of the American College of Nutrition, Vol. 21, No. 4, 328332 (2002)Published by the American College of Nutrition328chloride and potassium mono-basic phosphate (Sigma AldrichInc., St. Louis,
14、MO) and calcium carbonate powder (SigmaAldrich Inc, St. Louis, MO) were used to mimic fat, lean andbone. Polyethylene bottles (Nalge Nunc International, Roches-ter, NY) of different shapes and capacities (100 to 1000 mL)were filled with either pure olive oil or electrolyte solution.Different size bo
15、rosilicate tubes (3 and 5 mL, Becton DickinsonVacutainer Systems, Rutherford, NJ) and flexible polypro-pylene tubing (Nalge Nunc International, Rochester, NY) werefilled with calcium carbonate. Bottles and tubes were tapedtogether in layers to form nine blocks. Each block containeddifferent quantiti
16、es of oil, electrolyte solution and calciumcarbonate. The blocks were assembled contiguously with oneanother in a predetermined fashion to form four phantoms withtotal weights 1520g, 3140g, 4650g and 7490g as determined byan electronic scale (Seca model 727, Toledo Scale Corp.,Toledo, OH). The maxim
17、um dimensions of the four phantomsvaried between 28 to 58 cm in length, 12 to 32 cm in width and11 to 15 cm thick. All blocks were kept in a box at roomtemperature between DXA measurements.DXA ScansThree densitometers from three centers were assessed in thisstudy. All densitometers were from the sam
18、e manufacturer(Hologic Inc., Waltham, MA): two QDR 2000 (one located atUniversity of Liege (UL), Liege, Belgium, and the other lo-cated at Wayne State University (WSU), Detroit, MI, USA) anda QDR1500 located at the Neonatal unit of University Chil-drens Hospital (UCH), Greifswald, Germany. Quality c
19、ontrolscans for each densitometer were performed daily using amanufacturer-supplied anthropomorphic spine phantom. The invitro coefficients of variation (CV) for H110221 year for the deter-mination of bone mineral content, bone area and bone mineraldensity were 0.43%, 0.42% and 0.46%, respectively a
20、t UL,were 0.38%, 0.30% and 0.34% at WSU, and were 0.35%,0.35% and 0.31%, respectively at UCH.All densitometers were operated in the pencil beam mode,the only technique freely available for body composition stud-ies in infants. The four phantoms were scanned in triplicate oneach densitometer using in
21、fant whole body-scanning mode andanalyzed with manufacturer-supplied software V5.73P. Oneinvestigator (J.-C.P.) familiar with the agreed layout of thephantoms was present at each site to insure the correct assem-bly and placement of the phantoms for DXA measurements.Each center used its own operator
22、 for scan acquisition andanalysis. Phantoms were transported personally or shippedbetween centers via commercial courier.Statistical AnalysisDXA measured total weight, lean mass, fat mass, bonemineral content, bone area and bone mineral density were usedin data analysis. Percent of fat was presented
23、 as descriptive dataand not analyzed further since it was calculated from fat massand total weight. Repeated measures analysis of variance wasused to determine the equivalence of the triplicate DXA mea-surements (within subject factor) among the four phantoms(between subject factor) and whether ther
24、e was interactionbetween DXA measurements from different size phantomsfrom different instruments.Regression analyses were performed to determine the abil-ity of DXA measurements from UL and UCH to predict theDXA measurements of the same phantoms at WSU. Univariateanalysis of variance with Helmert co
25、ntrasts was used to analyzecomparability of residuals from each prediction equation basedon UL and UCH data respectively. The same procedures wererepeated to determine the regression equation for prediction ofDXA measurements of the same phantoms at the other centersfrom WSU DXA measurements.The sam
26、e procedures were repeated using the first of thetriplicate measurements to mimic the clinical situation of gen-erating one satisfactory scan per subject. This was done todetermine whether the same relationships exist with the use ofdata from one or three DXA scan. All statistical tests wereperforme
27、d with SPSS 10.0 (SPSS Inc., Chicago, IL) for win-dows at an adopted significance level of 0.05.RESULTSA representative phantom and its corresponding scan areshown in Fig. 1. DXA measurements of the four phantomsfrom the study sites are shown in Table 1. DXA measurementswere highly correlated (adjus
28、ted r2H11005 0.96 to 1.00) with weightof the components and total weight of each phantom. There wasno significant difference among triplicate DXA measurementsof the phantoms. Therefore averages across the three measure-ments were used in further analyses. There was no interactionamong DXA measuremen
29、ts from different size phantoms usingthe three instruments.DXA measurements from UL and UCH were highly pre-dictive (adjusted r2H11005 0.94 to 1.00, p H11021 0.001) of DXAmeasurements of the same phantoms at WSU (Table 2).Fig. 1. A 5 kg phantom assembled from various blocks (left) and theresultant d
30、ual energy X-ray absorptiometry scan (right).Phantoms for Dual Energy X-ray AbsorptiometryJOURNAL OF THE AMERICAN COLLEGE OF NUTRITION 329Helmert contrasts in analysis of variance confirmed that theresiduals from UL and UCH taken together did not differ fromWSU (as zero), nor did UL and UCH residual
31、s differ from eachother.Similar results were found when using the first value insteadof the average of the triplicate DXA measurements for theprediction of DXA measurements among the three institutionswith an adjusted r2H11005 0.94 to 1.00, p H11021 0.001. These findingswere applicable to the determ
32、ination of predicted mean, stan-dard error of estimate and the slope of the prediction equationsregardless of the DXA parameter measured. The differences inthe slope of the predicted equation were H110211% in all caseswhether the first or average of triplicate measurements wereused in the prediction
33、 equations.DISCUSSIONTo advance the understanding of physiologic and pathologiceffects on body composition in infants and young children, it isimperative that a system exists to determine the validity of datagenerated from different centers using different instruments. Itwould be useful to have in v
34、ivo studies to cross-calibratedifferent instruments as has been done with human adults47. However, the use of human infants for this purpose isimpractical under most circumstances, especially since thethree instruments employed for this study are located in widelyseparated geographic regions. Thus,
35、the use of phantoms wouldbe the most practical means to cross-calibrate different instru-ments in the study of body composition in small subjectsparticularly infants. However, phantoms suitable for use inolder children 8 and adults 47 are inappropriate for theassessment of body composition in small
36、subjects because ofmajor differences in the acquisition and analysis of scans.Our goal for this study was to develop a set of phantoms thatcan be made easily and inexpensively and have sufficientflexibility for any investigator to further modify or adjust thecomponents to create different size phant
37、oms with varied bodycomposition. The primary purpose for the use of these phan-toms is to determine the interrelationship of the DXA measure-ments among different densitometers rather than the compari-son of absolute accuracy of the DXA densitometers. This isTable 1. Triplicate Measurements* of Phan
38、toms Using Dual Energy X-ray AbsorptiometryUL UCH WSUMean SD CV% Mean SD CV% Mean SD CV%Phantom 1.5 kgTotal g 1507 1.2 0.08 1499 2.3 0.15 1506 1.6 0.11Lean Mass g 982 6.9 0.70 1034 21.4 2.07 1037 15.1 1.46Fat Mass g 490 6.4 1.30 432 22.3 5.15 435 16.1 3.71BMC g 34.4 0.15 0.42 32.9 1.39 4.23 34.4 1.7
39、9 5.20Bone Area cm2170 2.2 1.32 165 7.0 4.23 163 8.2 5.05BMD g/cm20.202 0.002 0.76 0.199 0.001 0.58 0.211 0.003 1.25Phantom 3 kgTotal g 3156 5.2 0.16 3147 4.9 0.15 3157 3.1 0.10Lean Mass g 1965 25.5 1.30 2005 67.1 3.35 2049 89.4 4.37Fat Mass g 1116 23.9 2.14 1074 70.7 6.58 1037 93.0 8.97BMC g 75.1 1
40、.08 1.44 68.9 1.24 1.80 72.1 3.37 4.67Bone Area cm2285 4.6 1.60 271 2.7 1.00 277 5.9 2.13BMD g/cm20.264 0.004 1.58 0.254 0.004 1.64 0.260 0.009 3.63Phantom 5 kgTotal g 4664 9.1 0.19 4677 3.6 0.08 4670 7.7 0.16Lean Mass g 3178 61.2 1.93 3174 100.5 3.17 3349 71.0 2.12Fat Mass g 1373 68.9 5.02 1395 101
41、.1 7.25 1218 81.4 6.68BMC g 113.2 1.19 1.05 107.4 0.94 0.87 102.9 7.43 7.22Bone Area cm2364 4.4 1.21 341 5.1 1.49 350 3.8 1.09BMD g/cm20.311 0.001 0.32 0.316 0.003 0.80 0.294 0.019 6.59Phantom 7 kgTotal g 7513 9.5 0.13 7521 9.9 0.13 7563 5.5 0.07Lean Mass g 5364 45.7 0.85 5600 99.8 1.78 5695 19.3 0.
42、34Fat Mass g 1991 36.3 1.82 1755 91.2 5.20 1700 19.9 1.17BMC g 182.2 1.54 0.85165.7 1.58 0.96 168.3 0.87 0.51Bone Area cm2631 3.1 0.49 588 9.5 1.61 590 4.6 0.77BMD g/cm20.290 0.002 0.730.282 0.004 1.44 0.285 0.002 0.61Abbreviation: UL H11005 University of Liege, Belgium, UCH H11005 University Childr
43、en Hospital, Germany, and WSU H11005 Wayne State University, USA, BMC H11005 bone mineralcontent, BMD H11005 bone mineral density.* Average of triplicate measurements.Excluded one BMC and one BMD value at H1101150% lower than the value generated from the other two scans.Phantoms for Dual Energy X-ra
44、y Absorptiometry330 VOL. 21, NO. 4consistent with the means to obtain standardized DXA mea-surements of the spine using instruments from different man-ufacturers that are known to provide different values for thesame subject 9. The design of our phantoms also satisfied therecommendations of the Inte
45、rnational DXA StandardizationCommittee that cross-calibration among different instrumentsshould not be based on the use of a single phantom 9.We have independently reported 1012 the validity of thepencil beam DXA technique for the measurement of bodycomposition using instruments from the same manufa
46、cturerbased on animal tissue studies, and it was not our intention toreproduce the anatomically correct or exact duplication of bodycomposition of infants, since there are great differences amonginfants and it would be prohibitively expensive and time con-suming to achieve these goals. In any case,
47、the physical dimen-sions and body composition values of our phantoms can bemodified to accommodate the wide range of weights and bodycomposition in clinical subjects, thus allowing cross compari-son of any clinical studies involving small subjects.In this study, the strongly predictive relationships
48、 of DXAmeasurements among the three instruments would support thatdata generated from different densitometers made by the samemanufacturer employing the same DXA pencil beam techniqueand the same software are comparable. Furthermore, systematiccorrections in the form of linear transformations are po
49、ssible toallow comparison of clinical data generated from differentstudies. It is also possible that our system of phantoms can beused to determine whether these relationships remain true fordata generated from the use of other DXA techniques or the useof instruments from different manufacturers.That the intercepts of the regression equations for the pre-diction of DXA measurements among various centers were notsignificantly different from zero would support the absence ofsystematic difference among the densitometers tested, althoughthe intercept for BMD predi
