She did everything right. Sixty-three, active, a little worried after her mother’s hip gave out, she asked her doctor for the scan. The result came back osteopenia — thinning, but not the disease. Not osteoporosis. “Keep doing what you’re doing,” she was told. “You’re not in the danger zone.”

Fourteen months later she caught her foot on a curb, put a hand down, and broke her wrist. A “fragility fracture” — the kind that isn’t supposed to happen to people who aren’t in the danger zone.

Her scan wasn’t wrong. It measured exactly what it measures. The problem is that what it measures is not the same thing as who breaks — and the distance between those two things is the most important, least discussed number in bone health.

The test we trust to catch fragile bones misses three out of four of the people who break them.

That is not a rhetorical flourish. It is roughly what the data say, and they have said it, consistently, for twenty years.

A Number That Shouldn’t Be Possible

Start with the landmark. In the early 2000s a study called NORA followed roughly two hundred thousand postmenopausal women and did something deceptively simple: it waited to see who fractured, then looked back at their bone density. The expectation was that the low-density women — the osteoporotic ones — would account for the fractures. They didn’t. The large majority of women who fractured had bone densityabove the osteoporosis line. Only a minority were “osteoporotic” at all.

A second major study, the Rotterdam cohort, found the same shape: most non-vertebral fractures arrived in people whose density was merely low-ish, or even normal. A 2019 analysis in a leading endocrinology journal put a blunt title on it — most fractures are not attributable to osteoporosis as we define it. And in 2024, an international working group reviewing best practices for the density scan itself reaffirmed the uncomfortable core of it: the majority of people who fracture will have a bone-density score better than the diagnostic threshold.

Line those findings up and they converge on the same figure. Somewhere around seven in ten to three in four fragility fractures occur in people the scan would not call osteoporotic. The screening test we built to find fragile bones flags, at best, a minority of the people who go on to break one.

How does a test survive that? Because it was never designed to do the job we assigned it.

The Line Drawn for Bookkeeping

In 1994, a World Health Organization study group needed a way to count osteoporosis — to say how many people had it, compare countries, and study treatments. To do that you need a definition, and a definition needs a line. They drew it at a T-score of −2.5: bone density two and a half standard deviations below that of a healthy young adult.

It was a reasonable place to draw a line for that purpose. But read the fine print the field has been trying to un-forget ever since: it was an epidemiological threshold, chosen to make the numbers work, not a biological cliff where bones suddenly become dangerous. Risk does not leap when you cross −2.5. It climbs smoothly the whole way down — and it is already meaningfully elevated in the wide territory the line leaves on the “safe” side.

−2.5 was never a biological cliff. It was a line drawn for bookkeeping.

So the word “osteopenia” — the in-between zone — does a quiet, dangerous thing. It sounds like reassurance. It is heard as you’re fine. But it describes tens of millions of people whose individual risk is real and whose collective share of fractures is enormous. Which brings us to the part that feels like a paradox and isn’t.

The Osteopenia Paradox

Here is the trap that fools intuition. A person with osteoporosis is, individually, at higher risk than a person with osteopenia. That’s true. And yet more total fractures come out of the osteopenia group — because there are so many more people in it.

Epidemiologists have a name for this: the prevention paradox. A large number of people at modest risk will generate more cases than a small number of people at high risk. Osteoporosis is the small, high-risk group. Osteopenia and “normal-ish” is the vast, moderate-risk crowd. Do the multiplication and the crowd wins — most of the fractures come from the people nobody flagged.

This is why a screening program built around a single threshold, no matter how carefully chosen, will always let most of the fractures slip through. It is aimed at the wrong end of the distribution. And it is why the interesting question was never “is your density below the line?” The interesting question is “what actually breaks a bone?” — and density is only one of the answers.

What the Number Can’t See

A bone-density scan measures one thing well: how much mineral is packed into a given area of bone. Call it quantity. But a fracture is not a quantity problem. A fracture is a collision — a moment when the force arriving at a bone exceeds the strength that bone can muster. Two whole categories of information decide that collision, and the density number contains neither.

The first is bone quality. Density tells you how much material is there; it says nothing about how well that material is built — the honeycomb architecture inside, the protein scaffold, the tiny struts that either brace one another or stand alone. Two bones with identical density can differ dramatically in strength depending on that internal engineering, and the scan sees straight past it.

The second is everything that isn’t bone at all. Most fragility fractures require a fall, and whether you fall — and whether you can catch yourself, and how much muscle and soft tissue cushion the landing — is governed by strength, balance, vision, medications, and nerve. A strong person who doesn’t hit the ground doesn’t fracture, whatever the scan says. A frail one who goes down hard might, even with respectable density.

Put those together and the density number starts to look like what it is: one input into a problem with many. Necessary, often useful — but nowhere close to sufficient. And there is one more input, louder than almost anything on the scan, that the scan will never show you.

The Loudest Signal Isn’t on the Chart

Ask what most sharply predicts a person’s next fragility fracture, and the answer is not their T-score. It is whether they have already had one.

A fragility fracture — a break from a fall at standing height, the kind that shouldn’t break a sound bone — roughly doubles the odds of the next one, and the risk is at its highest in the months and couple of years right after. Researchers call it imminent risk. It is one of the strongest, most actionable signals in the whole field.

The loudest predictor of your next fracture is your last one — and no density scan will tell you that.

And yet the wrist that breaks at sixty-three is so often filed as bad luck — splinted, healed, and forgotten — while the far bigger fracture it forecasts goes unspoken. The information was there. It just wasn’t on the scan, so nobody read it out loud.

The Fix Already Exists

None of this is an argument against the density scan. It is an argument against using it alone, as a single number with a single line — because the tools that fill the gap already exist and are, for the most part, neither expensive nor exotic.

A risk calculator such as FRAX reframes the question from “is your density below the line” to “what is your actual chance of fracturing in the next ten years,” folding in age, prior fracture, family history, steroids, smoking, low body weight — the factors density ignores.

A quality measure — the Trabecular Bone Score read off the same scan, or a radiation-free REMS Fragility Score — estimates the architecture the density number can’t see, and adds predictive power on top of it.

And the cheapest tool of all is a question: have you broken a bone as an adult from a minor fall? If the answer is yes, that person is high-risk today, whatever their T-score, and modern guidelines increasingly treat them accordingly — by risk, not by a line.

These are not rivals to the density scan. They are the rest of the picture it was always missing — layers, not replacements. The scan measures the bone. The fracture factor is everything else that decides whether it breaks.

Go back to the woman with the broken wrist. Nothing about her scan was false. What failed her was the story built around it — that a number better than −2.5 meant safety, that osteopenia meant “not yet,” that a wrist fracture at sixty-three was an accident rather than an announcement.

A different clinician, reading the same scan, would have seen a sixty-three-year-old with a family history and asked a fuller question — run the risk, added a quality read, and treated the wrist not as bad luck but as the warning it was. Same bone. Same number. A completely different outcome, decided entirely by how much of the picture someone was willing to look at.

The density scan tells you how much bone there is. Whether it breaks is a bigger question — and the good news buried in that harder truth is this: most of the fracture factor is made of things you can actually change.