Pediatr Blood Cancer

QCT Versus DXA in 320 Survivors of Childhood Cancer: Association of BMD With Fracture History Sue C. Kaste, DO,1,3,4* Xin Tong, MPH,2 Jennifer M. Hendrick,1 Evguenia J. Karimova, MD,1 Deo Kumar Srivastava, PhD,2 Frances A. Tylavsky, PhD,5 Terry L. Snider, RT (R)(CT),1 and Laura D. Carbone, Purpose. To assess agreement on diagnosis of diminished bone mineral density (BMD) and correlation between BMD values obtained by dual-energy X-ray absorptiometry (DXA) and quantitative computed tomography (QCT) in childhood cancer survivors. Patients and Methods. We retrospectively reviewed lumbar spine QCT and DXA studies for BMD in patients who underwent both imaging studies within a 24-hr period. We determined correlation between BMD values and agreement on diagnosis of diminished BMD obtained by both modalities. Diminished BMD was defined as two or more SDs below mean for age- and gender-matched reference values. We evaluated the relationship of BMD values determined by each modality to self-reported fracture history in the 160 (50%) patients with available reports. Results. Of 320 patients, 56% (178) were male; 87% (277) were white. Median age was 16.4 (range,

Key words:

6

5.1–36.0) years. Median BMD Z-score was
1.43 (range,
5.96 to 3.20) by QCT and
1.30 (range,
5.50 to 2.80) by DXA. Correlation between QCT- and DXA-determined BMD values was significant but low, and agreement on diminished BMD was fair (k ¼ 0.32). There was no association between BMD measured by either QCT or DXA and self-reported traumatic fracture history. Male gender was associated with doubling the traumatic fracture risk (P ¼ 0.0499). Conclusions. Quantitative computed tomography and DXA may give discrepant results when used to assess bone health in childhood cancer survivors, especially in those of non-white race. This inconsistency in indicators of BMD deficiency may complicate clinical decision-making. Consecutive use of a single modality is recommended to provide reliable longitudinal information. Pediatr Blood Cancer ß 2006 Wiley-Liss, Inc.

bone density; childhood cancer; DXA; quantitative computed tomography (QCT)

INTRODUCTION The skeletal health of the rapidly growing population of survivors of childhood cancer may be compromised by the effects of disease and therapy [1–3]. Maximal peak bone mass, an important determinant of osteoporosis and fracture associated with aging, may not be achieved. In normal individuals, bone mass rises rapidly during puberty [4] and usually reaches its peak at the end of sexual development. Therefore, children who undergo cancer treatment during the normal period of maximal bone-mass accrual are at risk for future deficits in bone mineral density (BMD) and its sequelae. Consequently, an accurate and consistent assessment of bone mass in this population is important. The leading methods of assessing BMD are quantitative computed tomography (QCT) and dual X-ray absorptiometry (DXA), although DXA is more widely used [3,4]. Both methods are used clinically sometimes in the same patient, making it difficult to interpret longitudinal changes in BMD and assess treatment efficacy. We tested the relationship between BMD values obtained using both DXA and QCT in the same young patients. Both methods provide rapid assessment; each has strengths and weaknesses. For the reason that DXA BMD assessment depends partially on bone size, its use in growing children is especially challenging [3–6]. With both methods, clinical treatment decisions in pediatrics typically are based on standard deviation or Z-scores, which, despite their limitations, allow comparison of individual BMD values to ageand sex-matched normative values. DXA also defines bone mineral status by comparing to race-matched references. ß 2006 Wiley-Liss, Inc. DOI 10.1002/pbc.20854

MD

Bone mineral density values determined by QCT in children are volumetric and independent of bone size [3,4]. QCT also can differentiate trabecular from cortical bone, providing additional information regarding bone health. Affordable QCT software can be installed on most CT scanners, but the initial cost of CT scanners and support staff limits its availability. Dual X-ray absorptiometry is available worldwide and can assess the BMD of the whole body or of individual areas. This two-dimensional method produces values of ‘‘areal’’ BMD; changes in the third dimension are not accounted for, which leads to underestimation of BMD of smaller sized bones by DXA. This disadvantage is an important — ————— 1

Department of Radiological Sciences, St. Jude Children’s Research Hospital, Memphis, Tennessee; 2Department of Biostatistics, St. Jude Children’s Research Hospital, Memphis, Tennessee; 3Department of Hematology-Oncology, St. Jude Children’s Research Hospital, Memphis, Tennessee; 4Department of Radiology, University of Tennessee School of Medicine, Memphis, Tennessee; 5Department of Preventive Medicine, University of Tennessee School of Medicine, Memphis, Tennessee; 6Department of Medicine, University of Tennessee School of Medicine, Memphis, Tennessee Work was performed at St. Jude Children’s Research Hospital. Grant sponsor: National Institutes of Health; Grant numbers: P30 CA21765, P01 CA-20180; Grant sponsor: State of Tennessee; Grant sponsor: American Lebanese Syrian Associated Charities (ALSAC). *Correspondence to: Sue C. Kaste, Department of Radiological Sciences, Division of Diagnostic Imaging, St. Jude Children’s Research Hospital, 332 N. Lauderdale, Memphis, TN 38105-2794. E-mail: [email protected] Received 4 August 2005; Accepted 6 March 2006

2

Kaste et al.

consideration in growing children [3]. When only two dimensions are measured on DXA, the volumetric BMD or ‘‘apparent BMD’’ (BMAD) is calculated using the formula: BMAD ¼ bone mineral content (BMC)/area1.5 [7]. Other formulas for reconstructing DXA-derived volumetric BMAD, which is less dependent on bone size than areal BMD, have been suggested [8,9]. Other disadvantages of DXA include the lack of differentiation between cortical and trabecular bone, attenuating effect of overlying soft tissues and fat on BMD measurement, and sensitivity to artifacts and scoliosis [3,10–12]. Cross-calibration of DXA units is needed to accurately compare values obtained on different machines [13]. PATIENTS AND METHODS We performed an electronic record search of the Diagnostic Imaging database for subjects aged at least 5 years who were evaluated between December 2000 and August 2003 and identified 320 pediatric cancer patients who, for a variety of clinical and research indications, were evaluated for BMD by QCT and DXA within the same 24-hr period. For patients with multiple examinations, the most recent were used in this analysis. This retrospective study was performed with approval by the Institutional Review Board (IRB), waiver of consent, and in compliance with Health Insurance Portability and Accountability Act (HIPAA) regulations. A small subset of the cohort (n ¼ 66) is part of an ongoing IRB-approved prospective clinical trial. Information regarding patients’ primary diagnoses, demographic characteristics, and body-mass indices was extracted from the institutional electronic medical record that is updated with each patient visit. QCT Quantitative computed tomography of the lumbar spine was performed with a Siemens Somatom-Plus spiral CT scanner (Siemens, Iselin, NY) and Mindways QCT Calibration Phantoms and software (Mindways Software, Inc., South San Francisco, CA). BMD was determined by obtaining direct axial images of the centers of the first and second lumbar vertebrae (L1 and L2, respectively) as localized from a sagittal scout image, as previously described [14–16]. BMD (mg/cm3) was recorded for the individual vertebral bodies, and the mean value was calculated. Normative values in the manufacturer’s reference database were used to calculate the Z-score, which was generated by the QCT software program. DXA A Hologic 4500 QDR-A fan beam system (Hologic, Inc., Bedford, MA) was used to measure BMD of the lumbar spine in anterior projections of L1–L4 and lateral projections of L2–L4. All subjects were analyzed with the QDR software Pediatr Blood Cancer DOI 10.1002/pbc

for Windows (version 12.1). Lumbar spine BMAD was calculated using the following formula: BMC  (area of lumbar vertebral bodies in cm2)1.5. Although previously reported studies of BMAD applied this equation to L1–L4 [7–9], we limited our assessment to L1 and L2 to allow direct comparison of DXA-determined BMAD to the volumetric determination of the same vertebral bodies by QCT. Normative values in the manufacturer’s reference database were used to calculate the Z-score, generated by the DXA software program. Z-Scores Z-scores were calculated by the manufacturer’s software for each subject’s BMD value (using the average values from L2–L4 for DXA and of L1–L2 for QCT) as the number of standard deviations from the mean age- and gender-matched reference values provided by the manufacturers of the instruments. To obtain a Z-score, DXA software also matches BMD value with a race-specific database. Diminished BMD was defined as a Z-score at least two standard deviations below the reference mean values. Fracture History Patients or their parents or guardians completed a questionnaire about the patient’s fracture history before the DXA examination. Information included whether the fracture was related to trauma (yes or no) and details about fracture healing (normal healing, prolonged healing, or unknown). Because regular use of these questionnaires was not yet implemented in the beginning of the study period, this information was available for only half of the study cohort. In these patients, we tested the relationship of BMD measurements yielded by both modalities to fracture history. Statistical Methods Correlation between BMD measurements by QCT and DXA was assessed with Pearson correlation coefficients. Agreement between the two methods with respect to Z-score categories of BMD was assessed with kappa statistics. Multiple logistic regression was used to test for associations between patient demographics and identification of diminished BMD by QCT or DXA (Z-score 
2 vs. others). Multiple logistic regression also was used to test for associations between positive traumatic fracture history (yes vs. no) and various other patient characteristics. Multiple linear regression model was used to test for associations between patient demographics and absolute measurements of BMD and Z-score by QCT and DXA. The criterion for significance for all analyses was a P-value significant at the level of a ¼ 0.05. All statistical analyses were done using SAS version 9.1 (SAS Institute, Cary, NC).

QCT- and DXA-BMD in Childhood Cancer Survivors

RESULTS Of the 320 patients, 178 (56%) were male; 277 (87%) were white, 33 (10%) black, and 10 (3%) Hispanic. The median age at time of examination was 16.40 years (range, 5.05–35.98 years) (Table I). Approximately 90% of patients had been treated for leukemia/lymphoma or brain tumor. Correlation Between BMD Measurements by QCT and DXA

3

(k ¼ 0.32) [17]. Agreement was higher for white patients (k ¼ 0.38) but there was no agreement for non-white patients (k ¼ 0.12, 95% CI:
0.1019 to 0.3388, P ¼ 0.29) (Table II). There was no difference between genders in the strength of agreement between DXA and QCT in diagnosis of diminished BMD (k ¼ 0.29 for female and k ¼ 0.35 for male) (Table II). Relationship Between Absolute BMD Values and Patient Characteristics

Significant linear relationship was observed between average BMD of the L1–L2 as measured by DXA and QCT (Pearson correlation coefficient ¼ 0.5215, P < 0.0001). Correlation between DXA-derived BMAD and QCT BMD was somewhat higher (Pearson correlation coefficient ¼ 0.6042, P < 0.0001). Significant linear relationship between DXA and QCT Z-scores was observed (Pearson correlation coefficient ¼ 0.6367, P < 0.0001) (Fig. 1). Agreement Between DXA and QCT Z-Score Categories The agreement of QCT and DXA with the diagnosis of diminished BMD (Z-score 
2) was fair in the whole cohort

Using the multiple linear regression model, we found that measurements of areal BMD by DXA were significantly associated with age and race. Increasing age and nonwhite race were associated with significantly higher BMD in L1 (P < 0.0001 and P ¼ 0.0221, respectively) and L2 (anterior-posterior projection) (P < 0.0001 and P ¼ 0.0370, respectively). There was no association between patient age or gender and QCT-determined BMD of L1, L2, or average BMD of L1 and L2. Similar to DXA, non-white patients had significantly higher QCT BMD values than white patients in L1 and L2: (P < 0.0001). Tables III and IV show the descriptive statistics by groups.

TABLE I. Patient Characteristics (n ¼ 320) Characteristic Age at examination (years) Median Range Race White Black Hispanic QCT Z-score 
2 >
2 QCT Z-score Median Range DXA Z-score 
2 >
2 Sex Female Male Diagnosis Brain tumor Leukemia/lymphoma Solid tumor Fracture (N ¼ 160) Traumatic fracture Atraumatic fracture Unknown fracture No fracture DXA Z-score Median Range

N (%) 16.43 5.05–35.98 277 (86.6) 33 (10.3) 10 (3.1) 96 (30.0) 224 (70.0)
1.43
5.96 to 3.20 89 (27.8) 231 (72.2) 142 (44.4) 178 (55.6) 142 (44.4) 146 (45.6) 32 (10.0) 42 (26.3) 3 (1.9) 12 (7.5) 103 (64.4)
1.30
5.50 to 2.80

DXA, dual X-ray absorptiometry; QCT, quantitative computed tomography. Pediatr Blood Cancer DOI 10.1002/pbc

Relationship Between Demographic Characteristics and QCT- and DXA-Derived Z-Scores Using multiple logistic regression analysis, we found no relationship between age or gender and diminished BMD (Zscore 
2 vs. others) as measured by QCT (P ¼ 0.59 and P ¼ 0.79, respectively) or DXA (P ¼ 0.65 and P ¼ 0.44, respectively). For race, however, results were opposite for DXA and QCT. By QCT, white patients were 2.97 (95% confidence interval [CI], 1.21–7.31, P ¼ 0.018) times as likely as non-white patients to have diminished BMD. By DXA, non-white patients, most of whom were black, were 2.60 (95% CI, 1.34–5.03, P ¼ 0.0046) times as likely as white patients to have diminished BMD. We used the multiple linear regression model to test the association of Z-scores with age at time of examination, gender, and race without reference to the threshold used to define diminished BMD. The QCT-determined Z-scores of white patients were significantly (P ¼ 0.0001) lower than those of non-whites, concordant with results of multiple linear regression of lower BMD among white patients (above) and usage by the QCT software of a non-race specific database. By contrast, the DXA-determined Z-scores were not associated with race (P ¼ 0.40), most likely due to the patients DXA BMD values being compared with the manufacturer’s race-specific reference database. Relationship Between BMD and Fracture History Self-reported fracture history was available for 50% (160 of 320) of the patient cohort (Table V). Fifty-seven patients

4

Fig. 1.

Kaste et al.

Correlation between QCT- and DXA Z-scores in 320 childhood cancer patients (Pearson correlation coefficient ¼ 0.6367, P < 0.0001).

([36%] out of 160 for whom this data was available) reported having had a fracture. Only 3 patients (5%) reported atraumatic fractures; 42 (74%) patients reported fractures associated with trauma, and the mechanism of fracture was unknown for 12 (21%) patients. Only two patients (4%) reported fractures associated with abnormal healing. Multiple logistic regression analysis was conducted to examine the association between the following factors (age, gender, race, DXA Z-score, QCT Z-score and DXA and QCT Z-score interaction) and the traumatic fracture history. Fifteen patients with atraumatic and fractures of nonspecified mechanism were excluded from the analysis. Traumatic fracture history differed between genders. Male patients’ frequency of traumatic fracture was twice that of

females (OR, 2.22; 95% CI, 1.001–4.902; P ¼ 0.0499). No association was found between positive history of traumatic fracture and diminished bone density (Z 
2) determined by QCTor DXA as well as other determinants of bone density or demographic characteristics of patients beyond male gender. We assessed the prevalence of fracture (including atraumatic and fractures of non-specified mechanism) in patients who were classified by either QCT or DXA, but not by both, as having diminished BMD (Z-score 
2). Among the 29 patients classified as having diminished BMD by QCT but not by DXA, 10 (34%) had fractured. Among the 23 patients classified as having diminished BMD by DXA but not by QCT, 10 (43%) reported a positive history of fracture.

TABLE II. Agreement Between DXA and QCT on the Diagnosis of Diminished BMD (Z-score 
2 vs. >
2) DXA Z-scores

Group

QCT Z-scores


2

>
2


2 >
2 
2 >
2 
2 >
2 
2 >
2 
2 >
2

48 41 19 17 29 24 44 25 4 16

48 183 23 83 25 100 46 162 2 21

All patients Female Male White Non-white

Kappa statistics Kappa (95% confidence limit)

P-value (Testing Kappa ¼ 0)

0.3237 (0.2114, 0.4360)

<0.0001

0.2946 (0.1235, 0.4656)

0.0004

0.3453 (0.1968, 0.4938)

<0.0001

0.3781 (0.2609, 0.4953)

<0.0001

0.1185 (
0.1019, 0.3388)

0.2860

DXA, dual X-ray absorptiometry; QCT, quantitative computed tomography; BMD, bone mineral density. Pediatr Blood Cancer DOI 10.1002/pbc

QCT- and DXA-BMD in Childhood Cancer Survivors

5

TABLE III. Patient Characteristics by Sex (n ¼ 320)

Age at examination Male Female QCT Z-score Male Female DXA Z-score Male Female Anterior L1BMD (g/cm2) Male Female Anterior L2BMD (g/cm2) Male Female Anterior average of L1–L2 (g/cm2) Male Female Lateral L2BMD (g/cm2) Male Female DXA L1–L2 BMAD (g/cm2) Male Female QCT BMDL1 (mg/cc) Male Female QCT BMDL2 (mg/cc) Male Female QCT average of L1–L2 (mg/cc) Male Female

N

Mean

SD

Min

Median

Max

178 142

16.48 16.74

5.02 5.40

5.23 5.05

16.25 16.81

35.98 34.86

178 142


1.36
1.35

1.37 1.43


5.16
5.96


1.51
1.40

3.20 1.87

178 142


1.37
1.10

1.25 1.35


4.40
5.50


1.40
1.05

2.80 2.70

178 142

0.77 0.74

0.19 0.19

0.35 0.25

0.78 0.75

1.43 1.28

178 142

0.82 0.82

0.19 0.20

0.40 0.36

0.82 0.84

1.54 1.37

178 142

0.79 0.78

0.19 0.19

0.38 0.31

0.80 0.80

1.48 1.32

171 130

0.74 0.75

0.14 0.19

0.41 0.33

0.73 0.75

1.18 1.61

178 142

0.23 0.24

0.04 0.05

0.14 0.11

0.23 0.24

0.40 0.35

177 142

153.87 151.31

37.51 36.57

54.47 28.48

153.93 150.60

286.35 238.65

178 142

145.17 145.04

37.25 37.26

47.32 25.39

146.04 146.68

272.35 234.22

177 142

149.51 148.18

37.19 36.73

51.92 26.94

150.01 148.38

279.35 235.33

DXA, dual X-ray absorptiometry; QCT, quantitative computed tomography; BMD, bone mineral density; BMAD, bone mineral apparent density.

We found no statistically significant difference between the two groups in the prevalence of fracture (P ¼ 0.57). DISCUSSION Since peak bone density is not attained until late adolescence or early adulthood, Z-scores, rather than Tscores, are used to diagnose abnormal bone mass in children [3,18]. There was previously no consensus on the Z-score threshold that best reflects low BMD in pediatric patients [19,20]. Recently, the International Society for Clinical Densitometry (ISCD) suggested that the term ‘‘low bone density for chronological age’’ be applied to children whose Z-scores are less than
2. However, clinicians are frequently unsure how to compare the Z-scores determined by QCTwith those determined by DXA and may interpret them similarly. In our study, which used the ISCD definition of low bone mass for chronological age, the overall agreement between Z-scores determined by QCT and DXA was poor, especially in non-white patients. One explanation is the use of reference Pediatr Blood Cancer DOI 10.1002/pbc

data for white race for all patients studied by QCT (Mindways Software) and the use of race-specific reference data for DXA studies (Hologic, Corp., Waltham, MA). Because peak BMD and BMD throughout life appear to be higher in blacks than in whites [21,22], the Z-scores of nonwhite patients are likely to be higher on QCT than on DXA, parallel to what was observed in the current study. Clinicians making treatment decisions have to remember this distinction between the two methods and bear in mind that Z-scores of black children might be overestimated by QCT, which are derived using BMD normative data for whites. This practice may have lasting implications for the skeletal health of nonwhite children, as recent studies suggest that blacks suffer more morbidity and mortality than whites after experiencing fracture [23,24]. The extent of disagreement between BMD and Z-scores determined by these two methods can complicate clinical interpretation. We found only moderate correlation between QCT and DXA measurements of the average BMD of the L1–L2 as measured by DXA and QCT (Pearson correlation

6

Kaste et al. TABLE IV. Patient Characteristics by Race (n ¼ 320)

Age at examination Caucasian Others QCT Z-score Caucasian Others DXA Z-score Caucasian Others Anterior L1BMD (g/cm2) Caucasian Others Anterior L2BMD (g/cm2) Caucasian Others Anterior average of L1–L2 (g/cm2) Caucasian Others Lateral L2BMD (g/cm2) Caucasian Others DXA L1–L2 BMAD (g/cm2) Caucasian Others QCT BMDL1 (mg/cc) Caucasian Others QCT BMDL2 (mg/cc) Caucasian Others QCT average of L1–L2 (mg/cc) Caucasian Others

N

Mean

Std

Min

Median

Max

277 43

16.62 16.44

5.30 4.42

5.05 6.44

16.44 16.38

35.98 24.72

277 43


1.47
0.61

1.31 1.66


5.96
5.16


1.52
0.82

1.87 3.20

277 43


1.22
1.42

1.28 1.45


5.50
4.40


1.30
1.80

2.70 2.80

277 43

0.75 0.80

0.19 0.18

0.25 0.44

0.75 0.82

1.28 1.43

277 43

0.82 0.86

0.20 0.19

0.36 0.51

0.83 0.88

1.37 1.54

277 43

0.78 0.83

0.19 0.18

0.31 0.47

0.79 0.84

1.32 1.48

263 38

0.74 0.78

0.16 0.20

0.33 0.39

0.73 0.74

1.21 1.61

277 43

0.23 0.25

0.04 0.04

0.11 0.17

0.23 0.25

0.35 0.40

276 43

149.30 174.76

34.89 43.06

28.48 54.47

149.06 168.76

238.72 286.35

277 43

141.73 166.96

35.31 41.83

25.39 49.36

141.67 161.17

230.13 272.35

276 43

145.50 170.86

34.89 42.28

26.94 51.92

145.93 164.15

231.74 279.35

DXA, dual X-ray absorptiometry; QCT, quantitative computed tomography; BMD, bone mineral density; BMAD, bone mineral apparent density.

coefficient ¼ 0.52). Correlation between the mean QCTderived BMD of L1-L2 and the mean DXA-derived BMAD of L1–L2 was somewhat higher (Pearson correlation coefficient ¼ 0.60). Although we limited BMAD examination to the L1–L2 vertebral bodies to allow direct comparison with QCT values, our findings corroborate the reports of other investigators that BMAD more accurately reflects volumetric BMD than does DXA-determined BMD [16] and therefore should agree more closely with QCT-derived BMD measurement. It should be noted, that standard DXA software calculates Z-scores based on areal BMD measurements, and not on BMAD. The dependence of DXA-derived areal BMD on bone size is a well-known limitation of the method that might become especially important in survivors of childhood cancer, who are at risk for abnormal linear growth [3]. Although areal BMD has a well-documented association with the risk of fracture in postmenopausal women [25], such a relationship has not been definitively established in healthy children [10,26,27]. Most [10,26] but not all [27] studies suggest that BMD as measured by DXA is Pediatr Blood Cancer DOI 10.1002/pbc

lower in children who have low-energy fractures than in others. In contrast, DXA measurements (including absolute BMD, Z-score, and calculated BMAD) are suggested to be a weak predictor of fracture in children with acute lymphoblastic leukemia (ALL) [28–30]. Our study found no evidence that QCT and DXA used in combination provide better association with traumatic fracture than either modality alone. However, the prevalence of fracture in our study group was low, and we did not assess longitudinal change in BMD by either QCT or DXA. Prospective studies are needed to study fractures. In one longitudinal study, reduction of BMC during the first 6 months of chemotherapy in patients with ALL had a 64% positive predictive value for fracture, whereas increased BMC had a negative predictive value of 82% for subsequent fracture [31]. Although our study did not directly address the prediction of fracture, it is possible that studies using DXA, QCT, or both will be able to predict fracture through longitudinal studies. This question will be addressed by ongoing studies at our institution.

QCT- and DXA-BMD in Childhood Cancer Survivors

7

TABLE V. Characteristics of the 160 Patients for Whom Fracture History Was Available

Age at examination Mean SE Median Min Max N QCT Z-score Mean SE Median Min Max N DXA anterior lumbar spine Z-score Mean SE Median Min Max Sex Female Male Race Black Hispanic White Diagnosis Brain tumor Leukemia/ lymphoma Solid tumor

Patients with traumatic fractures

Patients with fractures unrelated to trauma or of unspecified mechanism

Patients who reported no fracture

All patients with fracture information

17.39 0.75 17.89 8.69 35.26 42

17.62 1.34 16.70 9.70 29.27 15

15.83 0.48 15.85 5.46 28.91 103

16.40 0.39 16.66 5.46 35.26 160


1.75 0.17
1.59
5.16
0.04 42


0.89 0.33
1.13
3.35 1.39 15


1.34 0.15
1.47
4.86 3.20 103


1.40 0.11
1.43
5.16 3.20 160


1.48 0.17
1.50
4.40 1.50


1.47 0.26
1.20
4.10
0.20


1.27 0.14
1.50
4.30 2.80


1.34 0.10
1.50
4.40 2.80

N N

13 29

6 9

51 52

70 90

N N N

4 1 37

3 0 12

10 5 88

17 6 137

N N

20 18

5 8

55 40

80 66

N

4

2

8

14

Our study had several limitations. We retrospectively studied a group of children treated for childhood cancer. These children underwent BMD assessment for a variety of clinical and research indications. Therefore, the findings of this study may not be applicable to the general population, although our results provide information possibly useful in assessing skeletal health in a comparable group of patients at high risk of BMD deficiency. Our calculation of Z-scores involved comparison with norms for chronological age and were not adjusted for bone age, sexual maturation, or body size or habitus. Fracture history was based on patient self-report, and low-impact versus high-impact trauma and skeletal site of fracture were not ascertained. Therefore, the proportion of low-impact fractures that may have been related to diminished BMD could not be assessed, although the frequency of fractures in our study group only slightly exceeded fracture prevalence in the general population in whom approximately one-third of boys and girls sustain at least one fracture before 17 years of age [32]. Despite these Pediatr Blood Cancer DOI 10.1002/pbc

limitations, we believe that our study directly comparing quantitative lumbar spine BMD measurements by QCTand DXA in a large group of children exposes a very important problem of pediatric BMD assessment in survivors of childhood cancer. Our findings suggest that clinicians should not assume that Z-scores obtained by QCT and DXA are equivalent in pediatric populations. It is important that clinicians consider the complexity of BMD assessment, particularly in growing children, in order to use the information for the maximal benefit of individual patients. Consecutive use of a single modality can provide reliable longitudinal information for any single patient and avoid the complex interpretations that ensue from changing evaluation methods. The establishment of standard terminology and the cross-referencing of normative data between the two methods of BMD measurement would be helpful, as would a large prospective study designed to assess BMD in the growing population of survivors of childhood cancer.

8

Kaste et al.

ACKNOWLEDGMENT The authors thank Sharon Naron and Sandra Gaither for assistance in manuscript editing and preparation. This study was supported in part by grants P30 CA-21765 and P01 CA20180 from the National Institutes of Health, a Center of Excellence grant from the State of Tennessee, and the American Lebanese Syrian Associated Charities (ALSAC). REFERENCES 1. Arikoski P, Komulainen J, Riikonen P, et al. Reduced bone density at completion of chemotherapy for a malignancy. Arch Dis Child 1999;80:143–148. 2. Kaste SC, Chesney RW, Hudson MM, et al. Bone mineral status during and after therapy of childhood cancer: An increasing population with multiple risk factors for impaired bone health. J Bone Miner Res 1999;14:2010–2014. 3. Leonard MB. Assessment of bone health in children and adolescents with cancer: Promises and pitfalls of current techniques. Med Pediatr Oncol 2003;41:198–207. 4. Gilsanz V. Bone density in children: A review of the available techniques and indications. Eur J Radiol 1998;26:177–182. 5. Bachrach LK. Bare-bones fact—children are not small adults. N Engl J Med 2004;351:924–926. 6. van Rijn RR, van der Sluis IM, Link TM, et al. Bone densitometry in children: A critical appraisal. Eur Radiol 2003;13:700–710. 7. Carter DR, Bouxsein ML, Marcus R. New approaches for interpreting projected bone densitometry data. J Bone Miner Res 1992;7:137–145. 8. Nevill AM, Holder RL, Maffulli N, et al. Adjusting bone mass for differences in projected bone area and other confounding variables: An allometric perspective. J Bone Miner Res 2002;17: 703–708. 9. Kroger H, Kotaniemi A, Kroger L, et al. Development of bone mass and bone density of the spine and femoral neck—a prospective study of 65 children and adolescents. Bone Miner 1993;23:171– 182. 10. Goulding A, Cannan R, Williams SM, et al. Bone mineral density in girls with forearm fractures. J Bone Miner Res 1998;13:143–148. 11. Lequin MH, van der Sluis M, van Rijn RR, et al. Bone mineral assessment with tibial ultrasonometry and dual-energy x-ray absorptiometry in long-term survivors of acute lymphoblastic leukemia in childhood. J Clin Densitom 2002;5:167–173. 12. Leonard MB. Dual energy x-ray absorptiometry: Shortcomings in the assessment of bone health in children. Calcif Tissue Int 2002;70:355–383 (abstract). 13. Lentz H, Samuelson G, Bratteby LE, et al. Differences in whole body measurements by DXA scanning using two Lunar DPX-L machines. Int J Obes Relat Metab Disord 1999;23:764–770. 14. Cann CE. Quantitative CT for determination of bone mineral density: A review. Radiology 1988;1166:509–522.

Pediatr Blood Cancer DOI 10.1002/pbc

15. Cann CE, Genant HK. Precise measurement of vertebral mineral content using computed tomography. J Comput Assist Tomogr 1980;4:493–500. 16. Cann CE, Genant HK, Kolb FO, et al. Quantitative computed tomography for prediction of vertebral fracture risk. Bone 1985; 6:1–7. 17. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:159–174. 18. Harke HT: Pediatric bone densitometry: Technical issues. Semin Musculoskelet Radiol 1999;3:371–378. 19. Fielding KT, Nix DA, Bachrach LK: Comparison of calcaneus ultrasound and dual X-ray absorptiometry in children at risk of osteopenia. J Clin Densitom 2003;6:7–15. 20. Ellis KJ, Shypailo RJ, Hardin DS, et al. Z score prediction model for assessment of bone mineral content in pediatric diseases. J Bone Miner Res 2001;16:1658–1664. 21. Gilsanz V, Roe TF, Mora S, et al. Changes in vertebral bone density in black girls and white girls during childhood and puberty. N Eng J Med 1991;325:1597–1600. 22. Gilsanz V, Gibbens DT, Roe TF, et al. Vertebral bone density in children: Effect of puberty. Radiology 1988;166:847–850. 23. Furstenberg AL, Mezey MD. Differences in outcome between black and white elderly hip fracture patients. J Chronic Dis 1987; 40:931. 24. Kotzan JA, Martin BC, Reeves JH, et al. The impact of race and fractures on mortality in a postmenopausal Medicaid population. Clin Ther 1992;21:1988–2000. 25. Melton LJ III, Atkinson EJ, O’Fallon WM, et al. Long-term fracture prediction by bone mineral assessed at different skeletal sites. J Bone Miner Res 1993;8:1227–1233. 26. Landin LA. Fracture patterns in children. Acta Orthop Scand Suppl 1983;202:1–109. 27. Ma DQ, Jones G. Clinical risk factors but not bone density are associated with prevalent fractures in prepubertal children. J Paediatr Child Health 2002;38:497–500. 28. Van Der Sluis IM, Van Den Heuvel-Eibrink MM, Hahlen K, et al. Altered bone mineral density and body composition, and increased fracture risk in childhood acute lymphoblastic leukemia. J Pediatr 2002;141:204–210. 29. Van Der Sluis IM, Heuvel-Eibrink MM, Hahlen K, et al. Bone mineral density, body composition, and height in long-term survivors of actute lymphoblastic leukemia in childhood. Med Pediatr Onco 2000;135:415–420. 30. Hesseling PB, Hough SF, Nel ED, et al. Bone mineral density in long-term survivors of childhood cancer. Int J Cancer Suppl 1998; 11:44–47. 31. Atkinson SA, Halton JM, Bradley C, et al. Bone and mineral abnormalities in childhood acute lymphoblastic leukemia: Influence of disease, drugs and nutrition. Int J Cancer Suppl 1998; 11:35–39. 32. Cooper C, Dennison EM, Leufkens HGM, et al. Epidemiology of childhood fractures in Britain: A study using the general practice research database. J Bone Miner Res 2004;19:1976–1981.