ARRT Bone Densitometry common mistakes
ARRT Bone Densitometry · Understand the traps and how to avoid them.
On every spine scan, accepting the software's default L1–L4 labels without confirming that the lowest rib-bearing vertebra is T12.
Auto-labeling follows expected anatomy and can shift by one level with transitional vertebrae or when the scan starts too high or low; a level shift changes which vertebrae are averaged and which fracture or artifact is excluded.
What to do instead
Count ribs on the localizer or spine image, identify T12 as the lowest rib-bearing vertebra, relabel L1–L4 manually, and document any level change.
Scanning the proximal femur with the patient's foot in neutral or external rotation.
External rotation foreshortens the femoral neck and can falsely elevate femoral neck BMD, while distorting the greater trochanter and making region-of-interest placement inconsistent.
What to do instead
Use a positioning device, internally rotate the foot 15–25° or per manufacturer instructions, and verify on the image that the lesser trochanter is barely visible and the femoral neck is elongated.
Keeping a vertebra with a compression fracture, laminectomy, spinal hardware, or dense degenerative change in the L1–L4 average because the software included it.
These features alter attenuation and make that vertebra's BMD unrepresentative; including them can falsely elevate or lower the average and change diagnosis or monitoring decisions.
What to do instead
Inspect each vertebra, exclude fractured, artifact-laden, or surgically altered vertebrae and those with focal structural change, ensure at least two evaluable vertebrae remain, and document the exclusions.
Reporting a T-score and WHO category to a 32-year-old premenopausal woman or a 45-year-old man.
The T-score compares the patient with a young-adult mean; in premenopausal women, men under 50, and children the age-matched Z-score is the appropriate comparison, and WHO criteria do not apply.
What to do instead
Use the Z-score for these groups, report 'below expected range for age' when Z is –2.0 or lower, and reserve T-score/WHO categories for postmenopausal women and men 50 and older.
Calculating the least significant change as 1.96 times the precision CV.
LSC at 95% confidence must account for error in both measurements and uses 1.96 × √2 ≈ 2.77; using 1.96 underestimates LSC and causes random variation to be called real change.
What to do instead
Compute LSC = 2.77 × precision (CV%) for 95% confidence, and know that 90% confidence uses about 2.33 × precision.
Initialing the daily phantom QC log without plotting the value or checking it against control limits and trends.
A single pass/fail check misses gradual drift or a step shift, and an uninvestigated shift produces a systematic BMD bias that affects every patient.
What to do instead
Plot each phantom BMD on a control chart, apply manufacturer and program limits, investigate trends or shifts by repeating the scan and checking phantom positioning, escalate for service when indicated, and document corrective action.
Interpreting DXA BMD as a true volumetric density in g/cm³ and comparing values across different manufacturers.
DXA produces areal BMD in g/cm², which depends on bone size and on edge-detection and calibration differences; values from different manufacturers are not interchangeable.
What to do instead
Report and interpret BMD as g/cm², recognize the areal limitation for small or large bones, and use cross-calibration or same-machine comparison for longitudinal care.
Declaring a real BMD change on follow-up by comparing scans from a different machine or software version without cross-calibration.
Different machines, software versions, and reference databases introduce systematic offsets larger than the LSC, so apparent change may be technique-related rather than biological.
What to do instead
Compare same-machine and same-software scans whenever possible; if not possible, use cross-calibration or standardization and account for it before applying the LSC.
Skipping the pre-scan interview for recent iodinated contrast, barium, or radionuclide studies, or for pregnancy status.
Contrast media and radionuclides alter attenuation and can invalidate BMD measurement, and pregnancy status affects policy and patient counseling even at low dose.
What to do instead
Use a standardized questionnaire, postpone DXA per institutional and program timing guidance after contrast, barium, or nuclear medicine studies, and document pregnancy screening.
Using Ward's triangle or the greater trochanter as the diagnostic hip site, or applying the wrong reference database to the femoral neck.
Only the lumbar spine, total hip, femoral neck, and 33% radius in specified cases are diagnostic; Ward's triangle and trochanter are not. Femoral neck T-scores should use the NHANES III young-adult reference.
What to do instead
Base diagnosis on lumbar spine, total hip, and femoral neck; use NHANES III for femoral neck T-scores and the manufacturer reference for spine; reserve 33% radius for when hip and spine are unavailable or invalid.