BMD Measurement and Precision: A Comparison of GE Lunar Prodigy and iDXA Densitometers

Abstract

This study assessed bone mineral density (BMD) comparability and precision using Lunar Prodigy and iDXA densitometers (GE Healthcare, Madison, WI) in adults. Additionally, the utility of supine forearm measurement with iDXA was investigated.

Lumbar spine and bilateral proximal femur measurements were obtained in routine clinical manner in 345 volunteers, 202 women and 143 men of mean age 52.5 (range: 20.1–91.6) yr. Seated and supine distal forearm scans were obtained in a subset (n = 50). Lumbar spine and proximal femur precision assessments were performed on each instrument following International Society for Clinical Densitometry recommendations in 30 postmenopausal women.

BMD at the L1–L4 spine, total proximal femur, and femoral neck was very highly correlated (r² ≥ 0.98) between densitometers, as was the one-third radius site (r² 5 0.96). Bland-Altman analyses demonstrated no clinically significant bias at all evaluated sites. BMD precision was similar between instruments at the L1–L4 spine, mean total proximal femur, and femoral neck. Finally, one-third radius BMD measurements in the supine vs seated position on the iDXA were highly correlated (r² = 0.96). In conclusion, there is excellent BMD correlation between iDXA and Prodigy densitometers. Similarly, BMD precision is comparable with these two instruments.

Introduction

Dual-energy X-ray absorptiometry (DXA) is the current gold standard for the clinical diagnosis of osteoporosis based on measurement of bone mineral density (BMD) (1). As DXA technology continues to evolve, new instruments and technology are introduced, making it necessary to document how these advances compare to prior densitometers (2). GE Healthcare has recently developed the Lunar intelligent DXA (iDXA), a fan-beam densitometer that uses slightly higher amounts of radiation and enhanced detector capabilities, the latter yielding improved spatial resolution (3,4).

Prior comparison of lumbar spine and proximal femur BMD has demonstrated good agreement between iDXA and Prodigy (3–6). Similar BMD precision for iDXA and Prodigy has been reported in abstract form by the manufacturer (6) but not independently validated. As such, this study compared BMD at routine clinical sites, the L1–L4 lumbar spine, proximal femur, and distal forearm, obtained using a Lunar Prodigy and Lunar iDXA densitometer. Additionally, a precision assessment was performed at the lumbar spine and proximal femur with both instruments, and forearm BMD measurement in the supine vs seated position using iDXA was compared.

Methods

Study Participants

Volunteers (n = 345, 143 men and 202 women) aged 20 yr and older consented to participate in this study, which was approved by the University of Wisconsin Health Sciences Human Subjects Committee. From this group, 30 women aged 60 yr and older also participated in a precision assessment. A separate subset of 50 volunteers (24 women and 26 men) also had right distal forearm measurements on both instruments. The latter group had forearm scans obtained while sitting and supine on the iDXA. Demographic data for all groups are provided in Table 1. Subjects with surgical hardware or other anatomical abnormalities precluding adequate acquisition for diagnostic interpretation were excluded from this study.

Table 1.

Subject Demographics

	Main study sample (n = 345)	Precision cohort (n = 30)	Radius cohort (n = 50)
Age (yr)	52.6 ± 18.5 (20.1–91.6)	69.6 ± 4.9 (61.8–78.9)	51.7 ± 17.0 (22.5–87.5)
Body mass index (kg/m2)	26.5 ± 5.5 (17.4–48.8)	26.1 ± 4.6 (18.1–33.4)	26.1 ± 5.7 (18.7–41.8)
L1–L4 BMD (g/cm2)	1.201 ± 0.195 (0.658–2.066)	1.116 ± 0.130 (0.816–1.388)	1.171 ± 0.167 (0.722–1.564)
Left total femur BMD (g/cm2)	1.012 ± 0.175 (0.592–1.603)	0.856 ± 0.103 (0.658–1.128)	0.977 ± 0.158 (0.619–1.322)
Left femoral neck BMD (g/cm2)	0.970 ± 0.182 (0.588–1.631)	0.818 ± 0.103 (0.610–1.012)	0.936 ± 0.179 (0.588–1.381)
One-third radius BMD (g/cm2)	N/A	N/A	0.903 ± 0.127 (0.538–1.170)

Data reported as mean ± standard deviation and (range); proximal femur = total femur and femoral neck.

Note: iDXA BMD data reported.

Abbr: BMD, bone mineral density; N/A, not applicable.

DXA Acquisition

All participants were scanned on a GE Healthcare Lunar Prodigy and iDXA in routine clinical manner per manufacturer recommendations (7). Lumbar spine (L1–L4) and bilateral proximal femoral scans were obtained in 345 volunteers. Right forearm scans were obtained with subjects sitting in a chair next to each instrument. The right forearm was imaged, rather than the nondominant, as at the time of data collection, supine positioning with iDXA was available only for the right forearm. To potentially enhance patient comfort, increase throughput, and offer technologists an alternative for better forearm positioning, iDXA allows this scan to be performed with a patient lying supine on the table. These iDXA supine forearm measurements were compared with those obtained when sitting. For the Prodigy, Encore software GE Healthcare, Madison, WI) version 9.2 was used for acquisition and 11.4 for analysis; with iDXA, Encore software version 9.3 was used to acquire scans with version 11.0 used for analyses.

The precision assessments were performed in routine clinical manner according to International Society for Clinical Densitometry (ISCD) recommendations (1). Specifically, a cohort of 30 women aged 60 and older (described in Table 1) had lumbar spine and proximal femur scans obtained twice on both instruments; they stood up from the table between each set of scans. All precision scans were obtained at the same study visit and were performed by the same ISCD-certified technologist (MC).

Statistical Analyses

Bone mineral content, area, and BMD from Prodigy and iDXA were compared using Deming regression assuming equal error variance and Bland-Altman analyses (Analyse-it, Leeds, UK). Two-tailed paired t-tests of mean differences were calculated with Minitab (State College, PA). The precision (percent coefficient of variation and least significant change with 95% confidence interval; 2-sided testing) was calculated using the ISCD precision calculator (www.iscd.org). Comparison of precision by instrument was determined by calculating each sample variance; group difference was then determined using the F-test (Microsoft Excel, Redmond, WA).

Results

Bone Mineral Density

There was excellent agreement of BMD measurements at the lumbar spine and proximal femur between the two densitometers. Specifically, at all measured sites, BMD was highly correlated, r² ≥ 0.98. Additionally, the BMD mean bias was≤0.007 g/cm² at all skeletal sites (Fig. 1 and Table 2). Statistically significant mean differences were observed at the lumbar spine (p < 0.05) and femoral neck (p < 0.001). Of note, although statistically significant, these differences are likely of no clinical consequence as the numerically greatest mean BMD difference (observed at the femoral neck) was only 0.007 g/cm². Similar to the spine and proximal femur comparisons, one-third radius BMD correlation between densitometers was very good, with an r² = 0.96. Mean one-third radius BMD bias was −0.021 g/cm². Finally, in this study, one-third radius BMD obtained supine or sitting demonstrated good correlation, r² = 0.96 with a mean bias of 0.007 g/cm² (Fig. 2, Table 1). Finally, the possibility that BMD comparability might differ between males and females was explored by performing ordinary linear regression with dummy variables for male slope and intercept differences. There were statistically significant differences in intercept and slope only at the left total femur. The estimated male BMD difference was small, from +0.009 to −0.013 g/cm² through the entire range.

Fig. 1.

(A, B) Lumbar spine BMD and left femoral neck BMD correlation between instruments. Lumbar spine BMD (A) is highly correlated with an r² of 0.98. Additionally, Bland-Altman analysis (inset) demonstrates a BMD bias of −0.003 g/cm² confirming instrument similarity. Left femoral neck BMD (B) is also highly correlated with an r² of 0.99. Similarly, Bland-Altman analysis (inset) demonstrates a femoral neck BMD bias of 0.007 g/cm² confirming instrument similarity. BMD, bone mineral density.

Table 2.

BMD Correlations Between iDXA and Prodigy by Site and Forearm Technique

	BMD agreement (r2)	Bias (g/cm2)	RMSE	Slope (95% CI)	Intercept (95% CI)	p Value
L1–L4	0.98	−0.003	0.024	0.987 (0.974, 1.001)	0.0126 (−0.004, 0.029)	0.04
Total femur
Left	0.99	−0.001	0.015	1.002 (0.993, 1.011)	−0.003 (−0.012, 0.006)	0.11
Right	0.99	< 0.001	0.012	0.992 (0.984, 1.000)	0.008 (0.000, 0.016)	0.92
Femoral neck
Left	0.99	−0.007	0.021	0.988 (0.975, 1.000)	0.005 (−0.007, 0.018)	< 0.001
Right	0.98	−0.007	0.020	1.005 (0.993, 1.017)	−0.011 (−0.023, 0.001)	< 0.001
One-third radius
Between instruments	0.96	−0.021	0.026	0.997 (0.936, 1.058)	−0.018 (−0.074, 0.001)	< 0.001
Between iDXA techniques	0.96	0.007	0.026	0.944 (0.887, 1.001)	0.057 (0.005, 0.109)	0.10

Abbr: BMD, bone mineral density; RMSE, root mean square error of the linear regression fit; 95% CI, 95% confidence interval; p-value = probability of paired t-test for mean difference = 0.

Fig. 2.

One-third radius BMD correlation between instruments and using 2 positioning techniques on the iDXA. One-third radius BMD measurement is highly correlated between iDXA and prodigy with an r² of 0.96 (A). Bland-Altman analysis (A, inset) confirms measurement similarity with a mean bias of −0.021 g/cm². Additionally, one-third radius BMD measurement using the recumbent and supine positioning techniques with iDXA are well correlated with an r² of 0.96 (B) and a mean bias of 0.007 g/cm² demonstrated on Bland-Altman analysis (B, inset). BMD, bone mineral density.

Precision

No clinically relevant differences in BMD precision were observed at the L1–L4 spine or proximal femur sites (Table 3).

Table 3.

Comparison of BMD Precision: Prodigy and iDXA at the Lumbar Spine and Proximal Femur

	Prodigy		iDXA		Difference
	LSC (g/cm2)	CV (%)	LSC (g/cm2)	CV (%)	p Value
L1–L4 BMD (g/cm2)	0.043	1.45	0.056	1.81	0.12
Femoral neck BMD (g/cm2)
Right	0.035	1.52	0.027	1.15	0.06
Left	0.028	1.16	0.026	1.15	0.46
Total femur BMD (g/cm2)
Right	0.019	0.76	0.021	0.92	0.14
Left	0.016	0.67	0.015	0.65	0.44

Abbr: BMD, bone mineral density; LSC, least significant change; CV, coefficient of variation.

Discussion

As densitometer hardware and software evolve, it is necessary to document comparability of new instrumentation with preceding scanners (2). A substantial difference between iDXA and Prodigy technology is use of slightly higher radiation dose with iDXA and enhanced X-ray detectors (8), thereby improving image quality (3,6). Potentially, one could postulate that these technological differences might lead to differences in BMD measurement. Moreover, precision might be improved as the automated analysis techniques might place edges or regions of interest more “truly” using iDXA; such speculation is clearly not supported by these results. This study demonstrates excellent BMD measurement agreement between the two instruments at the lumbar spine and proximal femur in adult Caucasian men and women, similar to other reports (3–6). This study also demonstrates very good comparability in BMD measurement between these instruments at the one-third radius site when using the standard sitting technique for acquisition.

This study also supports the utility of supine forearm measurement with iDXA. This feature has subsequently been incorporated into Prodigy software as well. Supine forearm measurement is of potential benefit as it can be challenging to obtain an ideal image when performing forearm acquisition using seated technique. The classical seated positioning approach requires a patient to place his/her arm on the densitometer table with the elbow bent at a 90° angle; this can sometimes be uncomfortable. Additionally, patients of short stature often do not have a long enough humerus to place the rest of their body away from the scanner arm. The combination of these factors can make it difficult to position patients appropriately or for them to hold an ideal position despite the short scan duration. This is especially difficult on the iDXA, as there is a greater distance between the edge of the densitometer table and the scan field. Availability of the supine forearm technique requires less manipulation of a patient and allows technologists more flexibility to obtain an optimal image. It should be noted that this study did not evaluate supine forearm precision or follow-up; therefore, the same technique (seated or supine) should be used each time a given patient is scanned.

Limitations of this study include performance in a research setting, and only one precision assessment by one technologist was conducted. As such, these results may not be generalized to every pair of instruments, any technologist, or every patient population. It is important to recognize that precision takes into account 3 variables, the reproducibility of the instrument measurement, technologist positioning, and patient population (9). Additionally, although it is likely that BMD comparability between arms in a given individual would be similar, it should be acknowledged that an additional limitation is that this study imaged the right forearm, rather than the nondominant forearm.

In conclusion, this study documented excellent BMD comparability between GE Healthcare Lunar iDXA and Prodigy densitometers at the clinically relevant sites of the lumbar spine, proximal femur, and one-third radius. The precision observed with both densitometers was superb. One-third radius BMD measurements obtained using the supine position with iDXA demonstrated good agreement with classical seated positioning.