DXA Precision and Least Significant Change

A follow-up bone density result only means something if the change is bigger than the error of the measurement itself. Dual-energy X-ray absorptiometry (DXA) precision assessment measures that error at each skeletal site, and the Least Significant Change (LSC) converts it into the smallest bone mineral density (BMD) change a facility can call real with 95% confidence. Get the precision study and the LSC right, and serial monitoring becomes defensible; get them wrong, and you either chase noise or miss real bone loss.¹²

Introduction

Osteoporosis management lives on the follow-up scan. A single DXA gives a diagnosis against the World Health Organization T-score thresholds, but the decisions that change a patient's therapy — is the bone loss accelerating, is the drug working, has treatment stabilized the skeleton — all depend on comparing one scan to another over time.¹⁶

The problem is that the biological changes DXA is asked to detect are small, often on the order of a percent or two per year, and they are competing against the measurement error of the test. Reposition a patient a few millimeters, let the automated bone-edge detection place a region of interest slightly differently, and the reported BMD moves even when the skeleton has not. If a clinic reads every wiggle as real, it will change management on noise. If it ignores change until it is obvious, it will miss the window where intervention matters.¹³

Precision assessment is how a densitometry program separates signal from noise. It quantifies the reproducibility of the test in that facility, with that scanner, and with that technologist, and it feeds a single, decision-ready number: the Least Significant Change. This article explains what precision and the LSC are, how to run and compute a precision study correctly, why the root-mean-square method matters, what the International Society for Clinical Densitometry (ISCD) considers acceptable, and how the LSC is used to read serial scans. DRPS delivers this work through bone densitometry (DXA) physics testing and accreditation support across Florida, Maryland, Virginia, Washington DC, California, and Nevada.

Topic Explanation

What is precision, and what is the LSC?

Precision in densitometry is the reproducibility of a BMD measurement — how close repeated measurements of the same patient come to each other when nothing about the patient has changed. It is not the same as accuracy, which is how close the measured BMD is to the true mineral content. A scanner can be biased (inaccurate) yet still be highly reproducible (precise), and for monitoring change over time, it is precision, not accuracy, that governs whether a follow-up result is trustworthy.³

Precision is expressed as a precision error: the standard deviation (SD) of repeated measurements in absolute units (g/cm²), or the coefficient of variation (%CV) when the SD is normalized to the mean BMD and expressed as a percentage. A smaller precision error means a more reproducible test.³

The Least Significant Change (LSC) turns that precision error into a clinical threshold. It is the smallest difference between two scans that exceeds the measurement error at a defined confidence level. The ISCD defines the LSC as 2.77 times the precision error, which corresponds to 95% confidence for the comparison of two measurements.¹² A change equal to or greater than the LSC is reported as real; a change smaller than the LSC is statistically indistinguishable from noise and should not be called progression or response.¹

For related quality-control context on the equipment side of DXA, see our guide to bone densitometry (DEXA) QC.

Why the manufacturer's number will not do

Every DXA vendor publishes a precision figure. It is essentially useless for clinical monitoring, because precision depends on things the manufacturer cannot replicate: your patient population (body habitus, degenerative change, artifacts), your technologists' positioning consistency, and the specific behavior of your installed unit. The ISCD is explicit that manufacturer precision and LSC values cannot be applied to a clinical center; each facility must determine its own.² A center that borrows the vendor's optimistic number will set its LSC too low and start reading noise as real change.

Key Technical Principles

The precision study design

A precision study estimates the precision error by scanning real patients more than once, with complete repositioning between scans, so the measurement captures the full patient-plus-technologist-plus-scanner variability. The ISCD accepts two designs that deliver the same statistical power — 30 degrees of freedom:¹

• 30 patients, each scanned twice, or
• 15 patients, each scanned three times.

Patients get off the table and are repositioned between acquisitions, because a study that leaves the patient in place measures only scanner reproducibility and badly understates the real-world error. Degrees of freedom, not the raw number of subjects, drive the confidence of the estimate: for a design with $m$ subjects each scanned $n$ times, the degrees of freedom are $m(n-1)$, which is 30 for both accepted designs.¹³

Computing precision the right way: root-mean-square

The single most common error in a precision study is averaging standard deviations the wrong way. Precision error must be computed as the root-mean-square (RMS) average of the per-subject standard deviations, not the arithmetic mean. For $m$ subjects, where subject $i$ has a standard deviation ${\text{SD}}_i$ across their repeated scans, the RMS precision error is:

$${\text{SD}}_{\text{RMS}}=\sqrt{\frac{1}{m}\sum _{i=1}^m{\text{SD}}_i^2}$$

Expressed as a coefficient of variation, each subject's SD is normalized to that subject's mean BMD before the RMS average:

$$\%{\text{CV}}_{\text{RMS}}=\sqrt{\frac{1}{m}\sum _{i=1}^m{\left(\frac{{\text{SD}}_i}{{\overline{\text{BMD}}}_i}\times 100\right)}^2}$$

Using the arithmetic mean of the SDs instead of the RMS, or failing to account for degrees of freedom, underestimates the true imprecision — by up to about 25% for duplicate measurements with an arithmetic-mean error, and by larger amounts when degrees of freedom are mishandled.³ An underestimated precision error produces an LSC that is too small, which is exactly the failure mode that turns noise into false "progression."

From precision to the LSC

Once the RMS precision error is known for a skeletal site, the LSC follows directly. For a comparison of two measurements at 95% confidence:

$$\text{LSC}=2.77\times \text{precision error}$$

The 2.77 factor is not arbitrary. It is the product of the 95% two-sided normal quantile and the propagation of error across two independent measurements:

$$2.77=1.96\times \sqrt{2}$$

The $\sqrt{2}$ accounts for the fact that a comparison involves the error of the baseline scan and the error of the follow-up scan, and the 1.96 sets the two-sided 95% confidence level. The relationship is linear, so the LSC scales with the precision error. Worked directly from the ISCD framework: a site with a 1% precision error has an LSC of about 2.8%, and a site with a 2% precision error has an LSC of about 5.6%.¹

Converting to absolute units, which is how serial change should ultimately be judged, a lumbar spine with a mean BMD of 1.000 g/cm² and a precision error of 1.5% (an SD of 0.015 g/cm²) has:

$$\text{LSC}=2.77\times 0.015{\text{g/cm}}^2\approx 0.042{\text{g/cm}}^2$$

so only a change of about 0.042 g/cm² or more between scans at that site is statistically significant for that facility.

ISCD acceptable-precision limits

Precision is not just measured; it is judged against a standard. The ISCD publishes maximum acceptable LSC values that a technologist's precision must meet at each site. If a measured LSC exceeds the maximum, positioning and analysis technique should be reviewed and the study repeated.¹

The lumbar spine and total hip are the sites the ISCD recommends for monitoring, precisely because they combine biological responsiveness with achievable precision. The femoral neck's small region of interest makes it noisier, so it is generally not the site on which a monitoring decision is hung.¹

Clinical Impact

The LSC is the number that decides whether a follow-up DXA changes patient care. When a rheumatologist or endocrinologist compares this year's scan to the last, the correct question is not "did the number go down?" but "did it go down by at least the LSC?" Only then is the change beyond the measurement error and eligible to be called real bone loss or a real treatment response.¹⁴

This has direct consequences. A patient on anti-resorptive therapy whose spine BMD falls by 2% when the site LSC is 4% has not, statistically, lost bone — the apparent drop is within noise, and switching or escalating therapy on that basis is unjustified. Conversely, a rise of 4% at a site with a 3% LSC is a genuine response worth documenting. Reading serial scans without the LSC invites both errors: overreacting to noise and underreacting to real change.¹⁴

The ISCD also emphasizes that serial comparison should be based on absolute BMD in g/cm², not on T-scores or Z-scores, because the reference-population statistics behind T- and Z-scores add variability that is irrelevant to whether this patient's bone actually changed.¹ Percentages remain useful for communicating with clinicians and patients, but the underlying significance decision is made against the LSC.

Precision quality also compounds across a program. Because the LSC is set by the least precise contributor, a single technologist with poor positioning technique can inflate the effective LSC for the patients they scan, blunting the whole clinic's ability to detect change. This is why precision is assessed per technologist, and why continuity of technique matters when patients return for follow-up.

Practical Optimization Tips

Run the precision study on real, representative patients

Use patients who resemble the clinic's actual monitoring population, not just easy-to-position volunteers. A precision estimate built on slim, artifact-free subjects will look excellent and then fail to describe the harder scans the LSC is actually applied to.²³

Reposition fully between scans

Have the patient stand up and be repositioned between each acquisition. Leaving the patient on the table measures scanner-only reproducibility and produces an unrealistically small precision error — and therefore an LSC that is too permissive.³

Compute with RMS, not the arithmetic mean

Use a validated precision calculator (the ISCD provides one) or a correctly built spreadsheet that performs the root-mean-square averaging and respects degrees of freedom. Do not average the per-patient standard deviations arithmetically.³

Assess each technologist and each scanner separately

Precision belongs to the operator–machine pair. New technologist, new precision study. New scanner or major software upgrade, new precision study. Do not let one technologist's value stand in for the department, and do not carry an old LSC across a hardware change.¹²

Keep the follow-up scan comparable

Whenever possible, monitor on the same scanner, with the same acquisition mode, and with matched positioning and region-of-interest placement. If a patient must be moved to a different manufacturer's system, absolute BMD values are not interchangeable and cross-calibration is required before change can be interpreted.¹

Common pitfalls to avoid

• Using the manufacturer's precision or LSC. It does not describe your patients, staff, or machine.²
• Averaging SDs arithmetically. This understates imprecision and sets the LSC too low.³
• Not repositioning between scans. This measures the scanner, not the real test.³
• Comparing T-scores instead of absolute BMD. Reference-population variability is not part of the patient's change.¹
• Applying the spine LSC to a hip result (or vice versa). Precision and the LSC are site-specific.¹
• Never revisiting precision after a scanner swap, upgrade, or staffing change.¹²

Regulatory Considerations

DXA precision assessment sits inside the quality framework for a radiation-producing X-ray machine, not a radioactive-materials program. A DXA unit generates X-rays with an X-ray tube; it contains no byproduct material, so it falls outside U.S. Nuclear Regulatory Commission (NRC) jurisdiction. Instead, DXA is regulated as a diagnostic X-ray system under the U.S. Food and Drug Administration's electronic-product performance standards (21 CFR 1020.30) and by state radiation-control programs that register the machine and set inspection and quality requirements.⁹ Of the states DRPS serves, the machine is registered and inspected through each state's radiation-control program; the specifics vary by state, so confirm requirements with the authority having jurisdiction.

On the quality side, precision assessment and the LSC are professional-practice standards from the ISCD, reinforced by international guidance. The ISCD Official Positions require each facility to determine its own precision and calculate the LSC, and to repeat the study when a new DXA system is installed.¹² The IAEA's guidance on DXA for bone mineral density places precision and quality control within a broader technical-standards framework,⁸ and the ACR–AAPM technical standard for diagnostic medical physics performance monitoring of DXA equipment sets expectations for a Qualified Medical Physicist's evaluation of the system at installation and periodically thereafter.¹⁰ Accrediting bodies expect documented precision studies and site-specific LSC values as part of a defensible densitometry program.

Diagnostic classification itself rests on the WHO operational definition: a T-score of −2.5 or lower at the lumbar spine, femoral neck, total hip, or one-third radius indicates osteoporosis, a T-score between −1.0 and −2.5 indicates low bone mass (osteopenia), and a T-score of −1.0 or above is normal.⁶⁷ Precision and the LSC do not change the diagnostic threshold; they govern whether a change in BMD over time is real.

Frequently Asked Questions (FAQs)

What is the Least Significant Change (LSC) in DXA?

The Least Significant Change is the smallest change in bone mineral density between two DXA scans that is larger than the measurement error of the test, at a stated confidence level. The ISCD defines it as 2.77 times the facility's own precision error, which corresponds to 95% confidence for a comparison of two measurements. A change smaller than the LSC cannot be distinguished from noise.

How do you perform a DXA precision study?

A facility scans a representative group of patients more than once with repositioning between scans, then computes the root-mean-square standard deviation of the repeated measurements at each skeletal site. The ISCD accepts two designs that both yield 30 degrees of freedom: 30 patients scanned twice each, or 15 patients scanned three times each. Each technologist and each scanner should have its own precision value.

Why can't I use the manufacturer's precision value?

The manufacturer's quoted precision is measured on their own phantoms and patients by their own staff, so it does not reflect your patient mix, your technologists' positioning, or your specific scanner. Because the LSC depends directly on precision, using a manufacturer value can make real bone loss look like noise, or make noise look like a real change. Every center must measure its own precision.

What is an acceptable precision for DXA?

The ISCD publishes maximum acceptable Least Significant Change values that a technologist's precision must meet: 5.3% for the lumbar spine and 5.0% for the total hip. If a technologist's measured LSC at a site exceeds the ISCD maximum, the precision study should be repeated and positioning technique reviewed. The femoral neck is inherently less precise because the region of interest is small.

Should serial BMD be compared as a percentage or as an absolute value?

The ISCD recommends comparing serial bone mineral density using absolute values in grams per square centimeter rather than T-scores or Z-scores, and judging whether the change meets or exceeds the LSC. Percentages are easier to communicate to clinicians and patients, but the underlying decision about whether a change is real should be made against the LSC at that skeletal site.

How often should the precision study be repeated?

A precision study should be performed for each technologist after they are trained on DXA, and it should be repeated when a new DXA system is installed, when there is a major software or hardware change, or when a technologist's positioning technique may have drifted. The precision value and the LSC derived from it are specific to the operator and the machine.

Does precision assessment expose patients to meaningful extra radiation?

No. A DXA scan delivers a very small dose, and the ISCD has concluded that the patient-care benefit of knowing the true LSC outweighs the trivial additional exposure from a precision study. The scans used for a precision study are ordinary clinical acquisitions, and the resulting ability to distinguish real change from noise directly benefits the patients being monitored.

Key Takeaways

• Precision, not accuracy, governs monitoring. For serial BMD, what matters is how reproducible the test is, quantified as the precision error at each skeletal site.³
• The LSC is the decision threshold. It equals 2.77 × precision error (95% confidence for two measurements), where 2.77 = 1.96 × √2. A change below the LSC is not statistically real.¹
• Compute precision with RMS. Root-mean-square averaging of per-subject SDs, with correct degrees of freedom, is mandatory; arithmetic averaging understates imprecision and sets the LSC too low.³
• Use an accepted study design. 30 patients × 2 scans or 15 patients × 3 scans, with full repositioning, both give 30 degrees of freedom.¹
• Measure your own precision per technologist and scanner. Manufacturer values do not apply, and the LSC is set by the least precise contributor.²
• Meet the ISCD limits. Maximum acceptable LSC is 5.3% (lumbar spine) and 5.0% (total hip); compare serial change as absolute BMD in g/cm².¹

Conclusion

Precision assessment is the quiet backbone of clinical densitometry. It does not appear on the patient's report, but it decides what the report is allowed to say. Without a valid, facility-specific precision error and the LSC derived from it, a follow-up DXA is just two numbers with no rule for telling change from noise.

A defensible program measures precision on real patients, computes it with the root-mean-square method and correct degrees of freedom, sets a site-specific LSC for each technologist and scanner, checks it against the ISCD limits, and then reads serial scans against that threshold using absolute BMD. Done well, precision assessment is what lets a densitometry service make honest statements about bone loss and treatment response — and defend them.

How DRPS Can Help

Diagnostic Radiation Physics Services helps densitometry programs build and document the precision framework that makes serial DXA defensible. That includes designing and analyzing precision studies, computing site-specific LSC values, reviewing technologist positioning technique, evaluating scanner performance and calibration stability, and preparing the quality-control documentation accrediting bodies expect — delivered by board-certified medical physicists through our bone densitometry (DXA) physics testing and accreditation support services.

DRPS supports facilities across our service locations, including Florida, Maryland, Virginia, Washington DC, California, Nevada, New York, Pennsylvania, New Jersey, and Delaware.

A strong densitometry program is not just about producing a BMD number. It is about knowing which changes in that number are real.

Related Resources

• Bone densitometry (DEXA) QC
• Diagnostic reference levels
• ACR accreditation physics requirements
• Bone densitometry (DXA) physics testing
• Accreditation support
• Medical physicist consulting

References

1. Slart RHJA, Punda M, Ali DS, et al. Updated practice guideline for dual-energy X-ray absorptiometry (DXA). Eur J Nucl Med Mol Imaging. 2024;52(2):539-563. doi:10.1007/s00259-024-06912-6. doi.org
2. Baim S, Wilson CR, Lewiecki EM, Luckey MM, Downs RW, Lentle BC. Precision assessment and radiation safety for dual-energy X-ray absorptiometry: position paper of the International Society for Clinical Densitometry. J Clin Densitom. 2005;8(4):371-378. doi:10.1385/jcd:8:4:371. doi.org
3. Glüer CC, Blake G, Lu Y, Blunt BA, Jergas M, Genant HK. Accurate assessment of precision errors: how to measure the reproducibility of bone densitometry techniques. Osteoporos Int. 1995;5(4):262-270. doi:10.1007/BF01774016. doi.org
4. Shuhart CR, Cheung AM, Gill RK, Gani LU, Goel H, Szalat A. Executive Summary of the 2023 Adult Position Development Conference of the International Society for Clinical Densitometry: DXA Reporting, Follow-up BMD Testing and Trabecular Bone Score Application and Reporting. J Clin Densitom. 2023;27(1):101435. doi:10.1016/j.jocd.2023.101435. doi.org
5. Leslie WD, Moayyeri A, Sadatsafavi M, Wang L. A new approach for quantifying change and test precision in bone densitometry. J Clin Densitom. 2007;10(4):365-369. doi:10.1016/j.jocd.2007.08.002. doi.org
6. Kanis JA. Assessment of fracture risk and its application to screening for postmenopausal osteoporosis: synopsis of a WHO report. WHO Study Group. Osteoporos Int. 1994;4(6):368-381. doi:10.1007/BF01622200. doi.org
7. World Health Organization. Assessment of Fracture Risk and its Application to Screening for Postmenopausal Osteoporosis. WHO Technical Report Series 843. Geneva: WHO; 1994. who.int
8. International Atomic Energy Agency. Dual Energy X Ray Absorptiometry for Bone Mineral Density and Body Composition Assessment. IAEA Human Health Series No. 15. Vienna: IAEA; 2011. iaea.org
9. U.S. Food and Drug Administration. 21 CFR 1020.30: Diagnostic X-ray Systems and Their Major Components. ecfr.gov
10. American College of Radiology. ACR–AAPM Technical Standard for Diagnostic Medical Physics Performance Monitoring of Dual-Energy X-ray Absorptiometry (DXA) Equipment. acr.org