Routine CT scans of 78,000 patients map how every vertebra and rib loses density with age

Bones get thinner with age, and thin bones break more easily. Doctors usually check bone density with a special X-ray scan of the hip and lower spine, but that scan is done mostly when someone is already suspected of having weak bones. So we know a lot about a few spots in the skeleton and much less about the rest, or about exactly when women and men start to lose bone.

There is a cheap way to look further. Millions of people get a CT scan of the chest or belly for some other reason, such as checking the lungs or the kidneys. Each of those scans also shows every bone in view, and the brightness of bone on a CT image, measured in Hounsfield units (HU), goes up with its density. Computer programs can now outline each bone automatically, so these scans can be turned into bone measurements at no extra cost and with no extra radiation.

What they did

The team used the Japan Medical Image Database, which collects anonymized scans from Japanese university hospitals. They took every CT scan of the torso done without contrast dye in 2024 at nine hospitals, keeping one scan per person. That left 78,620 patients aged 15 to 100, a little over half of them men, with an average age of 65.

A published program called TotalSegmentator outlined 25 vertebrae (from the neck down to the sacrum, the bone at the back of the pelvis) and 24 ribs in each scan. For each bone, the team recorded the average CT value of the whole bone, including its hard outer shell. They then drew age curves separately for men and women, and built reference charts that show what counts as low, typical or high for a person's age and sex, like the growth charts used for children. They also asked which bones could stand in for the first lumbar vertebra (L1), the bone most often used for this kind of check.

What they found

• Loss almost everywhere. In women, bone density declined with age at every one of the 49 sites; in men it did so at 46 of them. The three exceptions were neck vertebrae in men, where the change was close to zero.
• A sharp drop in women around 52. In women, the speed of loss peaked at a median age of 52 years, at every single site, all within a narrow range of 50.5 to 55.5 years. In men there was no such peak: loss was fastest at the youngest ages studied and slowed after that. Between 50 and 60, the typical site lost about 4.8 HU a year in women and about 1 HU a year in men.
• The sexes cross over. Younger women had denser bones than men of the same age, and older women had less dense bones. The curves crossed at a median age of 57.
• Larger loss overall in women. From 40 to 80, the typical site lost about a third of its CT value in women (34%) and about an eighth in men (13%).
• Neighbours can stand in, ribs cannot. The vertebra just below L1 (L2) matched it closely. Ribs did not: converting a rib value to an L1 value could be off by up to about 80 HU, which covers most of the range doctors care about.
• Hospitals differ. The same bone gave somewhat different values at different hospitals, even after taking age and sex into account. L1 varied least of all 49 sites.

What to keep in mind

• A snapshot, not a film. The authors point out that the design is cross-sectional: the curves compare different people of different ages, not the same people as they grow older.
• Patients, not healthy volunteers. Everyone in the study had a CT scan for a medical reason, so the charts describe hospital patients. Few people under 30 were included.
• Much is unknown. The database has no diagnoses, medicines, fractures or menopause status. The link between the drop at 52 and menopause is the authors' interpretation: they note that the median age at natural menopause in Japanese women is 52.1 years, but menopause was not recorded.
• Different measure, different cut-offs. Because the team measured the whole bone rather than only its spongy inside, the usual CT thresholds for weak bones don't apply to these numbers. Values also depend on the scanner and hospital.
• One country, not yet reviewed. All patients were Japanese, scans were not checked by hand, and the work is a preprint that has not yet been peer-reviewed.

Why it matters

Bone loss is one of the most familiar parts of aging, but this study shows its timing across the whole spine and rib cage at once, in tens of thousands of people. The striking result is how tightly the faster loss in women is clustered around the early fifties, at every bone measured. The authors say the reference charts and their computer code will be released with the published paper, so other hospitals could use the CT scans they already take to spot people whose bones look thin for their age, though the authors stress that values have to be calibrated for each measurement method and hospital.