Eighteen-Month Follow-Up Across Two Manufacturers
A 70-year-old woman began alendronate five years ago. Her baseline DXA at Clinic A (GE Lunar Prodigy) showed L1–L4 BMD 0.820 g/cm² (T −2.6) and femoral neck 0.640 g/cm² (T −2.5). She moved to another city 18 months ago and is scanned today for the first time at Clinic B, which uses a Hologic Horizon A. The new report lists L1–L4 BMD 0.760 g/cm² (T −3.1) and femoral neck 0.605 g/cm² (T −2.8). The physician's note reads "marked bone loss - consider switching therapy." A spine radiograph from Clinic A shows no vertebral fracture. The Clinic B technologist confirms that positioning, region-of-interest placement, and image quality were satisfactory and that the patient reports taking alendronate consistently. The patient asks whether she is "getting worse."
1. Can the difference between the two studies be interpreted as bone loss? Explain your reasoning.
Read the worked answer
No - the comparison cannot support that conclusion.
- Different manufacturers. GE Lunar and Hologic systems use different calibration, different edge-detection and region-of-interest algorithms, and different reference databases. The same patient scanned on both can differ by several tenths of a T-score unit on the same day, and absolute BMD differences of this size are entirely plausible.
- T-scores are not interchangeable across systems. The T-score depends on the reference database, not only on the measured BMD, so "T went from −2.6 to −3.1" mixes a machine difference with a reference-population difference.
- No LSC applies across systems. The facility's least significant change is derived from repeated measurements on one densitometer; it cannot be used to judge a cross-manufacturer comparison, and no cross-calibration was ever established for this patient.
- Clinical plausibility. A true 7% loss over 18 months in an adherent patient on alendronate, with no fracture on radiograph, is unlikely. The arrow points at the measurement, not the skeleton.
What is valid is the current scan's own classification on the current system: a T-score of −3.1 at L1–L4 meets WHO criteria for osteoporosis at that site.
2. What would a valid monitoring program look like, and what could be done to salvage this comparison?
Read the worked answer
Valid monitoring program: the same densitometer for every follow-up, ideally the same technologist, same software version, same acquisition mode, and consistent positioning and analysis; comparison in g/cm² against the facility's own LSC rather than in T-scores; and an interval long enough for change to exceed the LSC - commonly 1–2 years.
Salvaging this particular comparison:
- In vivo cross-calibration. Scan this patient on both systems within a short window (same day or within a few days) and derive a patient-specific offset. That offset, not the raw difference, is what could be compared.
- Phantom cross-calibration. A cross-calibration phantom scanned on both systems can estimate the systematic difference, but it does not capture patient-specific factors such as body composition and edge detection.
- Published cross-calibration equations exist for some manufacturer pairs, but they are population-level adjustments and should be labeled approximate rather than used to assert a precise percentage change.
- Make Clinic B the new baseline and follow from there on one system. Absent a cross-calibration, the most defensible statement is that the 18-month interval cannot be assessed.
Note also that forearm scanning is not a workaround for cross-manufacturer comparison; the same calibration problem applies.
3. What should appear in the report and what should you communicate to the physician?
Read the worked answer
In the report:
- The current BMD and T-score for each site, classified on the current system's reference database (L1–L4 T −3.1 = osteoporosis; femoral neck T −2.8 = osteoporosis).
- An explicit technical statement: "Baseline study performed on a different densitometer (GE Lunar Prodigy). Direct comparison of BMD values or T-scores between systems is not valid; no least significant change can be applied across systems."
- No phrase such as "marked bone loss" or "significant decline," because neither is supported.
To the physician:
- The classification on today's scan is valid; the comparison is not.
- Do not change therapy on the basis of a non-comparable scan. If monitoring is the goal, arrange a follow-up in 1–2 years on a single system, or arrange the cross-calibration scan first.
- Other inputs - adherence history, fracture history, clinical risk factors, and any secondary causes - should carry the weight that a spurious 7% number is currently carrying.
- If the patient asks whether she is getting worse, the honest answer is that today's scan cannot tell us that, and explain why; that is more reassuring and more accurate than either false alarm or false reassurance.