7.2 Trabecular Bone Score (TBS) Principles & Clinical Utility

Interpret image texture alongside BMD

Trabecular bone score (TBS) is a unitless index derived from the grayscale texture of a compatible lumbar spine dual-energy X-ray absorptiometry (DXA) image. It can be calculated from an existing suitable acquisition without an extra X-ray exposure. Areal bone mineral density (BMD) measures mineral content per projected area. TBS captures a different property of the projection that is associated with aspects of microarchitecture and fracture risk. Two patients with similar areal BMD can therefore have different TBS values and different clinical risk profiles.

The texture calculation uses the variation in pixel intensity at different separations. In a simple experimental semivariogram, if $I(x_i)$ is the gray level at pixel $x_i$ and there are $N(h)$ pairs separated by a lag $h$, one estimate is:

This equation illustrates the statistical idea behind the adapted variogram used by the software. Nearby pixels with similar values produce smaller squared differences at that lag. The algorithm uses the relationship between variation and distance to derive a score. It does not reconstruct the three-dimensional network, count individual trabeculae, or determine an individual's mechanical failure load.

Verify eligibility and software version

The International Society for Clinical Densitometry (ISCD) states that TBS is appropriate in adults aged 40 or older, regardless of sex, race or ethnicity, and prior or current osteoporosis treatment, within the applicable technical limits. It is most likely to change management when an individual is close to a pharmacologic intervention threshold. Use the instructions for the installed product, DXA system and software version.

For example, a BMI-based report may be unavailable because the patient is outside that installation’s BMI range. That finding does not establish whether a different installation can report the case. Verify the actual version and its requirements, including the recorded tissue thickness if the software uses it.

A software version can change the tissue correction and thus the reported value. A generic BMI limit or fixed set of texture categories should not be carried across versions. Even when the installed software reports descriptive bands, an isolated score is not a diagnosis of osteoporosis or a treatment threshold. Use the result in the context of the person's scan quality, BMD and fracture risk.

Use the full L1–L4 region and withhold severe artifact

The ISCD adult position uses L1–L4 without exclusions for TBS and TBS-adjusted FRAX, including when there are moderate degenerative changes or chronic lumbar compression fractures. If severe structural or pathological artifact is present, such as vertebra plana, laminectomy, hardware or a metastatic lesion, the position recommends withholding the TBS report. A score obtained by discarding the severely affected level is not the workaround specified in the adult position.

A lumbar spine BMD analysis has a different objective. Use the appropriate BMD exclusion criteria for invalid vertebrae. For example, if a local degenerative change renders one level unsuitable for BMD but the full image remains eligible under the moderate-change TBS rule, the reported vertebral sets can differ. If metal hardware causes severe distortion, withhold TBS even when enough other levels remain for a reportable BMD.

Worked reporting decision

A 67-year-old patient has three evaluable vertebral bodies for BMD. L3 contains substantial spinal fusion hardware that distorts the TBS image. The other three levels may still support a spine BMD result if they meet the BMD criteria. The clinician should not report a three-level TBS or use it for an adjusted FRAX estimate. If the risk estimate is clinically useful, use an appropriate calculation without TBS and explain the limitation. Even when TBS is reportable, routine monitoring and reporting of TBS change is not recommended in the 2023 Adult Positions.

ISCD Adult Official Positions · FDA TBS iNsight V4 summary · Pothuaud and colleagues’ texture study