Beyond the T-Score: A More Honest Fracture-Risk Model
The answer to this series’ central question is precise: a T-score alone cannot define fracture risk because it describes bone density at a measured site, while a fracture is an outcome produced by density together with age, bone quality, falling, fracture history, clinical conditions, medications and the limits of the evidence used to estimate risk. The useful judgment is not to abandon the number, but to read it as one part of a decision pathway whose purpose is preventing clinically important fractures.
The number is precise about density, not destiny
A T-score is a standardized comparison of bone mineral density with the average density of a reference population. In the definition described by the US Preventive Services Task Force, a value of −2.5 or lower at the hip or lumbar spine meets the densitometric definition of osteoporosis. The classification is important: it gives clinicians a shared language, supports diagnosis and appears in the entry criteria of many treatment trials.
But the T-score does not directly measure the event that a person fears. A fragility fracture is a fracture from a fall from standing height or less that would not fracture most healthy people. Major osteoporotic fractures include the hip, spine, wrist and shoulder. They can lead to distress, another fracture, loss of independence and difficulty with daily activities. After a hip fracture, the USPSTF reports, only 40% to 60% of people recover their prefracture mobility and ability to perform daily activities. Fracture prevention is therefore the outcome against which a measurement and a treatment pathway should ultimately be judged.
Central DXA is valuable because it measures BMD at clinically relevant sites, uses a low dose of radiation and correlates with bone strength and clinical fracture outcomes. The location is part of the information. A femoral-neck result and a lumbar-spine result are not interchangeable readings of one abstract whole-person risk. The USPSTF notes that risk at a specific site is best predicted when density is measured at that site. The scan is therefore a well-established measurement, but it remains a measurement of density at a location.
The gap between density and fracture becomes visible in three related facts. First, age is a stronger determinant of fracture than BMD alone. Older adults have much higher fracture rates than younger adults with the same density because risk also rises with declining bone quality and an increasing likelihood of falling. Second, the USPSTF notes that most fragility fractures occur in people without densitometric osteoporosis, meaning with T-scores above −2.5. Third, BMD alone has useful but variable predictive performance: across the cohorts reviewed by the USPSTF, AUCs ranged from 0.60 to 0.80 for major osteoporotic fracture and from 0.64 to 0.86 for hip fracture.
Those facts are not contradictory. They describe different layers of the problem. A low T-score can identify a group with clinically important low density. A higher-than-threshold T-score does not place someone outside the possibility of fracture. An AUC can show that BMD tends to rank risk better than chance in a studied population without telling us the exact probability attached to one person’s result. The statistical signal is real; the claim that can safely be made from it is narrower than “this person will” or “this person will not” fracture.
Age, falling and bone quality also do not act as interchangeable checkboxes. Age changes the background rate of fracture and accompanies changes that the BMD value does not report. Falling connects the bone to the force that can produce a fracture, while the scan reports density rather than how often a person falls. A history of fracture or a medical condition can further change the clinical question. The measurement remains the same kind of measurement, but its decision meaning changes as the surrounding conditions change.
Imagine two people with the same femoral-neck T-score. One is younger and has no reported falls. The other is older and has recurrent falls. The density comparison is identical, but the context in which a fracture might occur is not. Add a prior fragility fracture or a medication associated with bone loss, and the question may no longer be ordinary screening at all. These examples do not justify assigning either person a personal percentage from a T-score. They show why the number cannot erase information it never contained.
The same principle explains why T-score remains indispensable even after its limits are accepted. Removing it would discard the established density classification and the measurement used to determine eligibility in nearly all trials of bone-conserving pharmacotherapies.