What FRAX Does and Does Not Tell You

FRAX is a fracture-probability model, not a bone-strength test. It combines a limited set of clinical inputs with country-specific fracture and mortality data, with or without femoral-neck bone mineral density, to estimate two outcomes over 10 years: hip fracture and major osteoporotic fracture.

The output can help organize prevention and testing decisions, particularly when a bone-density T-score alone does not capture age and clinical risk, though its precision is conditional on correct inputs and fit to the relevant population. The number does not include every determinant of fracture, and a treatment threshold displayed in a guideline is not a biological boundary.

The two probabilities FRAX reports#

The first output is the 10-year probability of hip fracture. The second is the 10-year probability of a major osteoporotic fracture, commonly abbreviated MOF. In standard FRAX, MOF means a clinical vertebral, hip, forearm, or proximal-humerus fracture.

“Clinical vertebral” matters. Vertebral deformities found only on imaging are not counted in the output outcome in the same way, although a prior morphometric vertebral fracture can be relevant as a history input. Pelvic, rib, tibial, and other fragility fractures are not all included in the MOF endpoint.

A probability of 12% does not mean a fracture is 12% complete or that the model knows which person will fracture. It means that among people represented by the model with the entered features, approximately 12 in 100 would be expected to have the defined outcome within 10 years if the model is well calibrated to that setting and period.

Probability is different from a simple annual incidence multiplied by 10. FRAX accounts for the competing chance of death, because a person cannot experience a future fracture after death, so a high underlying fracture hazard may still produce a lower cumulative probability when mortality is also high.

What information goes into the model#

The standard web tool asks for age, sex, weight, height, and country or region. It also asks about previous fracture, parental hip fracture, current smoking, long-term oral glucocorticoid use under the model's definition, rheumatoid arthritis, selected causes of secondary osteoporosis, and alcohol intake at or above the stated level.

Most clinical factors are entered as yes or no. This makes the tool practical but compresses important detail: one prior wrist fracture and several recent vertebral fractures both turn the prior-fracture field to yes, even though risk may differ substantially. Smoking intensity, alcohol pattern, and medication dose are not represented continuously in standard FRAX.

Age limits and input definitions should be read on the selected calculator, and the original tool is generally used from age 40 through 90; values outside its modeled range are handled by boundary assumptions rather than by direct evidence for every age. The country model must match the intended population as closely as feasible.

Missing information is not a neutral category in the web tool. The official FAQ states that unanswered clinical factors are treated as no. A false no can bias the result downward, so do not treat a probability as complete when the history behind it is uncertain.

FRAX with and without BMD#

FRAX can be calculated without bone mineral density, which can support an initial risk assessment or help identify who may benefit from dual-energy X-ray absorptiometry, depending on the guideline and setting. Adding a valid femoral-neck measurement can reclassify risk.

When BMD is used, the standard field is femoral-neck BMD or its appropriate T-score using the specified reference. ISCD and IOF positions advise against substituting lumbar-spine or total-hip values. These sites have different distributions and were not interchangeable model inputs.

The distinction matters when the lumbar-spine and femoral-neck T-scores disagree. If the lumbar spine is much lower, standard FRAX may underestimate vertebral and overall skeletal concern. If it is much higher, the estimate may move in the opposite direction. Some approaches and FRAXplus adjustments can examine discordance, but the original number should be labeled clearly.

A T-score is a comparison with a young-adult reference. FRAX is a multivariable probability. Two people with the same T-score can have different probabilities because of age, prior fracture, parental history, smoking, body size, medications, rheumatoid arthritis, and the country model.

Why country selection changes the result#

FRAX models are calibrated to fracture and mortality patterns for specific countries or populations. The same clinical profile can generate different probabilities across models because the background hazards differ.

This does not mean nationality causes a fixed risk. Country models summarize population data and health conditions available during development. Migration, ancestry, access to care, secular trends, and within-country diversity complicate the match. The official FAQ notes uncertainty about which model to use for some migrants.

The United States model includes options intended to reflect differences among population groups, but categories cannot capture every person's genetic, social, and environmental context. Selecting a category is a model operation, not a biological identity statement. Calibration can also drift as fracture prevention, longevity, diagnostic practices, and population health change. The model version, country selection, date, and inputs should be documented whenever a number informs a decision.

What the 10-year horizon adds and hides#

A 10-year horizon creates a common frame for comparing risk and guideline thresholds. It is long enough to capture many preventable fractures and has been widely incorporated into policy.

It can also hide short-term urgency. Fracture risk is especially high soon after some fractures. Standard FRAX records prior fracture but does not fully encode how recent it was. If you fractured a hip last month, you may need assessment that should not wait for a calculator to cross a threshold.

At the other end, competing illness and life expectancy affect whether a 10-year preventive benefit is achievable. FRAX includes population mortality in the probability calculation but cannot understand your full prognosis, goals, or treatment burden. The horizon should also not be converted casually into a one-year number by dividing by 10, because risk is not necessarily constant across years and the competing-event calculation is nonlinear.

Important factors standard FRAX can underrepresent#

Falls are a major immediate cause of fractures, but fall count is not a direct input in standard FRAX. The development cohorts measured falls inconsistently. A history of frequent or injurious falls can therefore require separate assessment rather than being assumed to be captured fully.

Prior fracture is binary. Number, site, severity, and recency can add information. Multiple fractures and severe or recent vertebral fractures may imply higher risk than the standard input conveys. FRAXplus offers adjustment calculations for some factors, but adjusted results need their method and version stated.

Type 2 diabetes is not a direct standard input even though fracture risk can be elevated at a given BMD. Chronic kidney disease, frailty, sarcopenia, some medicines, balance, vision, home hazards, and physical function are also not fully described by the core fields.

The glucocorticoid field summarizes a specified pattern as yes or no. Dose, duration, route, and intermittent high-dose courses can change skeletal effects. Published adjustments exist, but they remain approximations and should not be mixed invisibly into an “official” FRAX number.

FRAX does not diagnose osteoporosis#

Osteoporosis can be diagnosed through accepted densitometric criteria in applicable populations or through particular fragility-fracture histories, depending on the guideline. FRAX is not required to confirm a diagnosis when those criteria are already met.

Conversely, a high FRAX probability identifies predicted fracture risk, not a pathology result. The causes could include low BMD, age, prior fracture, medicines, inflammatory disease, falls, or several contributors. Evaluation for secondary causes and fall risk may change management even when the probability is unchanged.

The 2025 USPSTF recommendation concerns screening, not treatment of an individual, and it recommends screening women aged 65 years or older and postmenopausal women younger than 65 years who are at increased risk after clinical assessment. It found insufficient evidence to determine the balance for population screening in men; those statements apply to defined asymptomatic populations and do not override evaluation after a fragility fracture or in secondary osteoporosis.

A treatment threshold is not part of the calculator#

FRAX reports probability. Guidelines decide how that estimate contributes to intervention. Thresholds reflect evidence, treatment effects, harms, costs, health-system priorities, and sometimes age or prior fracture.

The Bone Health and Osteoporosis Foundation clinician guide uses, among other criteria for the United States, 10-year thresholds of at least 3% for hip fracture or at least 20% for MOF in certain postmenopausal women and men aged 50 years or older with low bone mass. Those numbers are not universal FRAX cutoffs for every country, age, diagnosis, or patient.

The United Kingdom's NOGG framework uses age-dependent assessment and intervention thresholds and distinguishes high from very high risk. Other systems use different policies. Copying a threshold without its target population and guideline context can misclassify decisions.

Crossing a line by one decimal place should not be treated as a sudden biological change. Input error and measurement variation can shift the estimate. The full decision includes prior hip or vertebral fracture, BMD, falls, secondary causes, treatment options, expected benefits and harms, preferences, and feasibility.

FRAX is not a treatment monitor#

The official FAQ and ISCD positions state that FRAX is not sufficiently sensitive for monitoring treatment response or a treat-to-target strategy. Age increases automatically over time, while the model does not represent treatment effect as a dynamic input. A later probability can rise even when therapy has reduced risk relative to no therapy.

Recalculation may sometimes inform a new treatment decision after prior therapy, but it should not be interpreted as a direct measure of efficacy. Follow-up can instead consider new fractures, adherence, interval BMD where appropriate, height change, symptoms, fall history, laboratory findings, and the treatment-specific evidence. Comparing two outputs also requires the same country model, device data, BMD reference, and input definitions, since silent software updates can make an apparent change partly methodological.

How to read a FRAX report critically#

Confirm the model country, the date, and the age, sex, height, weight, and every yes-or-no factor you were asked for. Check whether the fracture history includes vertebral imaging findings and whether parental history is known, and identify whether BMD was omitted or entered from the femoral neck with the correct device and reference.

Read both outputs and their endpoint definitions. Ask whether recent or multiple fractures, frequent falls, diabetes, large spine-hip discordance, high-dose glucocorticoid patterns, frailty, or another secondary cause could make the core estimate incomplete.

Then identify which guideline is being applied to you. A probability without a decision framework is not a recommendation. A threshold without the exact target population is not transferable policy. The cleanest documentation separates the raw FRAX result, any adjustment, other clinical findings, and the reasoning used.

FRAX is most useful when its limited scope is respected. It converts several established predictors into an interpretable absolute probability. It becomes misleading when it is handed to you as destiny, diagnosis, universal treatment rule, or proof that unmodeled risks do not matter.

Sources#

1. Official FRAX calculation tool
2. Official FRAX frequently asked questions
3. USPSTF 2025 recommendation on osteoporosis screening
4. Bone Health and Osteoporosis Foundation clinician's guide
5. ISCD and IOF official positions on FRAX
6. NOGG 2024 clinical guideline for osteoporosis