A T-Score That Improved by Moving House
Ms. R, 66, had her baseline DXA two years ago at an academic center on a Hologic Discovery: femoral neck T-score -2.4, and she started alendronate. She has moved and now attends your clinic, which uses a GE Lunar Prodigy. Today her femoral neck T-score is -1.9. She is delighted and says she wants to stop her alendronate because 'it worked.' Her weight is stable, she has taken the drug regularly, and both studies were acquired and analyzed by experienced technologists. Your facility's in-vivo precision for the femoral neck is 1.9% (RMS-CV), giving a least significant change of 5.3%. She has brought her old report but not the images.
1. Is the apparent improvement real? Show the reasoning, not just the conclusion.
Read the worked answer
You cannot call this an improvement, for three separate reasons that all point the same way:
- Different manufacturers. Hologic and Lunar use different calibration and reference databases, and the systematic offset is site-specific and often 5-10% at the hip. A half-T-score 'gain' can be produced entirely by changing machines.
- Magnitude versus precision. A 0.5 T-score at the femoral neck is roughly half a young-adult SD, on the order of 5% of BMD. Your LSC is 5.3%. So even if both scans had been done on your Lunar by the same technologist, a 5% change would sit right at the edge of significance, not comfortably beyond it.
- Added variance. Two machines, two technologists, and two analysis sessions widen the confidence interval around the difference considerably beyond a single-site LSC.
The honest answer: the change is uninterpretable, not reassuring.
2. What would you need in order to compare these two studies properly?
Read the worked answer
- Cross-calibration. The standard is an in-vivo crossover: a set of subjects (or a phantom, for a device-level correction) scanned on both systems, from which a correction factor or at least an offset estimate is derived. Without it there is no defensible arithmetic that turns a Hologic T-score into a Lunar T-score.
- The original images/DICOM, not just the report. Verify which hip, which region was reported, the level labeling, and whether an exclusion or a misdrawn region explains part of the gap.
- A common segment. If the old study analyzed a different femoral neck region or included artifact, the comparison is invalid before device issues even arise.
- Your own precision data, used prospectively: the LSC should be calculated at your site for your technologists, on your machine, for the site you are monitoring.
If cross-calibration is unavailable, the practical substitute is to establish today's value as a new baseline and compare forward on the Lunar only.
3. How do you counsel her today, and what do you put in the report?
Read the worked answer
- Do not endorse stopping alendronate. Nothing in this comparison shows the drug worked or that she can safely stop; stopping a bisphosphonate is a physician decision weighing fracture history, risk, and time on therapy.
- Explain the machine issue in plain language: 'The two scanners use different rulers. A small difference between them does not mean your bone changed.' Note that her actual value could have gone up, stayed the same, or gone down, and this comparison cannot distinguish those.
- Report language: give both values, state that the prior study used a different manufacturer with no cross-calibration available, state that the observed difference does not exceed the LSC for the site, and recommend that the current value serve as the baseline for future monitoring on this device.
- Plan forward: repeat in 1-2 years on the same scanner, same site, same technologist where possible; if her physician needs a formal comparison, arrange a cross-calibration study. Send the report to the prescriber with the caveat stated, not buried.