The relationship between entered height, weight, sex, and changes in patient thickness on Trabecular Bone Score using a Hologic Horizon Dual Energy X-ray Bone Densitometer - A cadaveric spine phantom study

Abstract

Purpose/Aims

To explore the sensitivity and precision of Trabecular Bone Score (TBS) calculations to changes in the entered weight, height, sex, and scan mode on a Hologic Horizon-A densitometer, using a radiographic cadaveric spine torso phantom.

Rationale/Background

Currently, TBS adjusts the score using BMI of the patient, using calibration points derived from a TBS Calibration Phantom that has segments with differing attenuation over the targets within it. If patient thickness is different than predicted by BMI, or if data entry error occurs this can impact both TBS, and FRAX, and TBS-adjusted FRAX. Sex selected alters TBS, and this can have implications for those undergoing gender transition.

Methods

A cadaveric bone torso phantom (Radiology Support Devices, Inc, Long Beach, CA, USA) was scanned on a Hologic Horizon A, using Apex Version 5.6.1.2 Rev 009 software, five times each, using array and fast-array spine scan modes, and again with the addition of one or two 4mm thick acrylic plates to simulate changes in soft-tissue thickness. All scans were done without repositioning of the phantom during sets. (FIGURE 1) The default auto-analysis was accepted after verification of intervertebral line placements for the first scan in each series of five, and the “Auto-Compare” analysis of the remaining scans to reduce operator effects on results. For the fast array scans using two acrylic plates, the software applied “Auto low-density” analysis algorithm. Upon calculating the average Effective Epoxy Thickness (TH) values of the phantom without absorbers, an index height and weight was determined using the average height and weight of a convenience sample of patients with matching TH values, taken from the scanner database. TBS values were then calculated (TBS iNsight version 3.1.2) after varying the height in 1 inch increments or the weight in 5 pound increments across the BMI range permitted for TBS (BMI range 15-37 kg/m2). The results at each height/weight were recorded for both sexes.

Results

Precision error, as the SD of the five scans in each scan mode, for TBS, BMD, and TH were smaller for array scans than fast-array of the phantom. But this was reversed for BMD and TH with 4 and 8cm of absorber. TBS SD, however was consistently poorer in fast-array at all three phantom thicknesses. There was a slight increase in BMD with additional attenuation, but no significant differences between BMD in array or fast array at each phantom thickness. TBS scores decreased with additional absorber while BMD increased slightly with additional absorber. (TABLE 1) When altering BMI whether by weight or height, TBS was proportional to BMI based on data entry. At all values of BMI, males have are higher TBS than females but the slopes remain similar until a BMI of around 27 kg/m2, where the slope for males increases compared females. Fast-array values are consistently higher than array values for both men and women.

There is no difference whether BMI was altered using height or weight. (FIGURES 2, 3)

Implications

Data entry errors regarding weight and height will affect calculated TBS scores. As the thickness of our phantom was increased, measured TBS decreased. This suggests that using Hologic EET values may allow for a more accurate adjustment of TBS due to decreased image resolution, contrast, and statistical noise than estimates based on BMI only. In addition, different scan modes produced different TBS values at the same height, Weight, BMI values. This could have an impact when following patients over time or at different facilities if different scan modes are used.