What Are the WHO Criteria for Osteoporosis?

The World Health Organization defines osteoporosis as a bone mineral density (BMD) T-score of −2.5 or lower, measured by dual-energy X-ray absorptiometry (DXA) at the hip or spine. That single number, established in 1994, was chosen because it produced a prevalence in postmenopausal white women that roughly matched their observed lifetime risk of fragility fracture. The threshold has become the global diagnostic standard, but applying it correctly turns out to be more complicated than a single cutoff suggests, especially once you move beyond the population it was originally calibrated for.

The Four WHO Categories

The 1994 WHO working group divided bone density into four diagnostic bands, all defined by how far a person’s BMD falls from the average peak bone mass of a healthy young adult woman. That distance is the T-score, expressed in standard deviations. A T-score of 0 means your density matches the young-adult average. A T-score of −1 means you are one standard deviation below it, and so on. The categories are:

Normal: T-score at or above −1.0
Osteopenia: T-score between −1.0 and −2.5 (low bone mass, but not yet osteoporosis)
Osteoporosis: T-score at or below −2.5
Severe osteoporosis: T-score at or below −2.5 with one or more fragility fractures already present

These categories remain in clinical use worldwide. DXA is still considered the gold standard for noninvasive diagnosis, and the thresholds have not been officially revised since they were introduced.¹

Why −2.5 and Not Some Other Number

The cutoff was not chosen because bone suddenly becomes fragile at that exact point. Fracture risk increases continuously as bone density drops; there is no cliff edge. The WHO panel picked −2.5 because, in white women aged 50 and older, that threshold produced an osteoporosis prevalence that approximately matched the lifetime risk of hip, spine, or wrist fracture observed in epidemiological data.² In other words, the number was a practical calibration exercise rather than a biological discovery. The fact that it was calibrated to one specific demographic group has consequences for everyone else, which is where much of the controversy lies.

Where on the Body the Measurement Happens

DXA scans can be performed at several skeletal sites, but for formal WHO-based diagnosis, the recommended locations are the lumbar spine, the femoral neck, and the total hip. A diagnosis of osteoporosis can be made if any of these sites reaches a T-score of −2.5 or below. In practice, the spine and hip often tell different stories. A study of nearly 3,500 people found that spine and hip T-scores agreed in only about half of cases. Minor discordance, where the two sites placed a person in adjacent categories, occurred in roughly 42% of cases, and about 4% had major discordance, meaning one site said normal while the other said osteoporosis.³

This discordance is not a flaw in the machine. Different bones lose density at different rates depending on age, weight, physical activity, and whether the bone is mostly trabecular (spongy interior bone, dominant in the spine) or cortical (the dense outer shell, dominant in the hip). It is common for someone’s spine to qualify as osteoporotic while their hip reads as osteopenia, or vice versa. Clinicians typically use the lowest T-score from the measured sites when making a diagnosis.

When DXA Results Are Misleading

DXA works by projecting an X-ray beam through a region of bone and calculating density from how much of the beam is absorbed. Because it compresses a three-dimensional structure into a flat, two-dimensional image, anything extra sitting in the beam path can inflate the reading. In older adults, this is a real problem. Degenerative changes in the spine, such as bone spurs, disc narrowing, and calcified blood vessels, can artificially raise the lumbar spine T-score, sometimes dramatically. A recent case report described a woman in her nineties whose lumbar T-score appeared exceptionally high, masking actual bone fragility, because of severe spinal degeneration and vascular calcification.⁴ Clinicians are advised to cross-check DXA results against clinical and imaging findings rather than accepting the number at face value.

Prior compression fractures in the spine can also distort results by compressing vertebrae and paradoxically increasing the apparent density per square centimeter. For these reasons, the hip measurement, which is less vulnerable to degenerative artifacts, becomes increasingly important in older patients.

Men and the T-Score Debate

The original WHO criteria were developed using data from postmenopausal women. How to apply them to men has been debated for decades. The core question is straightforward: should a man’s T-score be calculated by comparing his bone density to the young-adult female reference, or to a young-adult male reference? Men generally have larger, denser bones, so using a female reference means fewer men will cross the −2.5 threshold. Using a male reference identifies more men as osteoporotic.

The Endocrine Society’s clinical practice guideline for osteoporosis in men uses a male reference, recommending treatment for men whose BMD at the spine, femoral neck, or total hip is 2.5 standard deviations or more below the mean of normal young white males.⁵ The International Society for Clinical Densitometry (ISCD), however, endorsed a female reference database for T-score calculation in men as of its 2013 consensus conference.⁶ The disagreement is not trivial. A study of older men found that which reference database was used substantially shifted where fractures fell across diagnostic categories. Using male reference data, the vast majority of fractures, about 87%, occurred in men without an osteoporosis diagnosis. When female reference data were used instead, the pattern changed, but most fractures still arose in men classified as non-osteoporotic, with the burden shifting toward the “normal” BMD group.⁷ Neither reference database captures most of the men who go on to fracture, which is one reason that density-based diagnosis alone is considered insufficient for predicting fracture risk in men.

Premenopausal Women and Children

The WHO T-score criteria were explicitly not designed for premenopausal women, men under 50, or children and adolescents. For these groups, the ISCD recommends using Z-scores instead of T-scores.⁸ A Z-score compares your density to others of the same age and sex rather than to a peak-bone-mass reference. In premenopausal women, the ISCD recommends that a Z-score of −2.0 or below be interpreted as “below the expected range for age,” while anything above −2.0 is “within the expected range for age.” The labels “osteopenia” and “osteoporosis” based on T-scores alone should not be applied to premenopausal women, according to these guidelines.⁹

Not everyone agrees. The International Osteoporosis Foundation (IOF) takes a slightly different position, recommending T-scores for adults aged 20 to 50 and suggesting that a T-score below −2.5 can define osteoporosis in that age range, particularly in the presence of a known cause of bone loss or a low-trauma fracture.⁹ The practical takeaway is that a young woman or man under 50 whose scan report shows a low T-score should not automatically be labeled osteoporotic. Context matters: whether they have had a fragility fracture, whether an underlying condition such as a hormone disorder or medication side effect explains the low density, and whether the Z-score paints a different picture.

It is also worth knowing that T-scores and Z-scores can diverge more than you might expect in younger adults. A study of over 4,200 people aged 20 to 49 found that the two scores differed by a full standard deviation or more in up to 11% of cases, partly because there is no single standardized method for calculating Z-scores across different DXA machines.¹⁰

The Reference Database Problem Across Ethnicities

The WHO threshold assumes a Caucasian female reference population, but bone size and density vary across ethnic groups. When DXA machines calculate your T-score, they compare your measurement to whichever reference database the scanner is loaded with. If that database does not match your population, the resulting T-score can be significantly off. A study of over 3,100 Thai women found that using a Caucasian reference database produced an osteoporosis prevalence of about 64% in postmenopausal women, while a Thai-specific database put the number at roughly 41%. Misclassification rates ranged from about 11% to nearly 49% depending on which reference was used and which skeletal site was measured.¹¹

The problem runs in different directions for different populations. A study of south Indian postmenopausal women with fractures found that a Caucasian database identified osteoporosis in 72% of them at the femoral neck, but a Korean database identified only 45%.¹² Meanwhile, research in African Americans with rheumatoid arthritis showed the opposite pattern: using Caucasian reference data, about 5% were classified as osteoporotic, but switching to an African American reference database tripled that figure to 16%.¹³ The direction of the error depends on whether the reference population has higher or lower average bone density than the group being tested. Either way, clinicians working with diverse populations need to be aware of which reference database their machine is using.

Beyond Bone Density Alone

One persistent criticism of the WHO criteria is that a T-score captures bone quantity but says little about bone quality. You can have two people with the same T-score, one who fractures and one who does not, because their bone microarchitecture, geometry, and material properties differ. Several tools have emerged to address this gap.

The trabecular bone score (TBS) is derived from the same DXA scan image but analyzes the texture pattern of the spine rather than the density. A low TBS suggests degraded bone microarchitecture and predicts fractures partly independent of BMD. TBS can be used to adjust FRAX, the WHO’s fracture risk assessment tool, to improve risk stratification.¹⁴ In practice, this means a patient whose T-score hovers around −2.0, technically just osteopenia, might be reclassified as higher risk if their TBS is low. Early research showed that TBS could help distinguish people with fractures from those without, even when their BMD values were similar.¹⁵

Bone turnover markers, measured through blood or urine tests, offer another angle. These biochemical markers reflect how actively your skeleton is being broken down and rebuilt. They do not diagnose osteoporosis, but they can track whether treatment is working and help gauge how rapidly bone is being lost.¹⁶

Alternative Imaging Methods

DXA is not the only way to assess bone density, and in some clinical situations, other modalities offer advantages. Quantitative computed tomography (QCT) measures volumetric bone density in three dimensions rather than the two-dimensional projection that DXA provides. This means QCT can separate trabecular bone from cortical bone, and its results are not inflated by spinal arthritis or vascular calcification. The trade-off is a higher radiation dose and greater cost. In studies comparing the two methods, QCT consistently detects osteoporosis in more postmenopausal women than DXA does. One study found that lumbar QCT identified osteoporosis in about 46% of participants compared to about 20% by combined spinal and hip DXA.¹⁷ That does not necessarily mean QCT is “better” at diagnosis; the WHO criteria were calibrated to DXA, and using QCT with DXA-based thresholds tends to overdiagnose relative to the original framework.

Quantitative ultrasound (QUS), typically applied to the heel bone, uses sound waves instead of X-rays and is portable, inexpensive, and radiation-free. That makes it attractive as a screening tool, especially in settings where DXA machines are scarce. Its sensitivity for detecting abnormal bone density can be quite high: one study reported about 89% sensitivity with perfect specificity when DXA was used as the reference.¹⁸ However, QUS measures a different property of bone than DXA does, and its specificity for identifying osteoporosis defined strictly by DXA T-scores can be lower, meaning it may flag people who would not meet the WHO threshold on a formal scan.¹⁹ Current guidelines do not accept QUS as a substitute for DXA in making a formal WHO-based diagnosis, but it has value as a triage tool to identify people who should go on to get a DXA scan.

Diagnosis Versus Treatment Thresholds

An underappreciated point is that the WHO diagnostic threshold and the threshold at which treatment should begin are not the same thing. The T-score of −2.5 tells you whether someone meets the definition of osteoporosis. It does not automatically tell you whether that person should start medication. The International Osteoporosis Foundation and ESCEO published a position paper specifically to clarify this distinction, arguing that diagnostic and intervention thresholds should be considered separately.²⁰

In practice, treatment decisions increasingly rely on fracture risk calculators like FRAX, which combine BMD with clinical risk factors such as age, prior fractures, smoking, glucocorticoid use, and family history of hip fracture. Someone with a T-score of −2.0, technically only osteopenia, might still qualify for treatment if their overall fracture probability is high enough. Conversely, a person with a T-score of −2.6 and no other risk factors might not need immediate pharmacological intervention. The WHO criteria set the diagnostic label; clinical judgment and risk tools set the treatment plan.

Access to Diagnosis in Lower-Income Settings

Even decades after the WHO criteria were established, access to DXA scanning remains highly uneven. A 2025 position paper from the International Osteoporosis Foundation noted that many low- and middle-income countries still base treatment decisions on bone mineral density where osteoporosis services exist at all, despite limited availability of densitometry.²¹ In regions without DXA machines, the WHO criteria exist in theory but cannot be applied in practice. This creates a diagnostic gap where fractures serve as the first sign of osteoporosis rather than a preventable consequence of it. Alternative approaches, including FRAX calculated without BMD input and portable ultrasound screening, are being explored as ways to bridge that gap, but no global consensus has replaced the DXA-based standard.