Bone remodelling

Why does bone thicken exactly where it is worked and melt away where it is not?

We speak of the skeleton as a frame — the scaffolding inside us, laid down in childhood and thereafter simply there, like the joists of a house. But a tennis professional's racquet arm carries visibly more bone than the other, built by nothing but use. An astronaut loses bone in months. A leg in a cast comes out thinner. Frames do not do this. So what kind of thing is a bone, if it can be reshaped by how you spend your afternoons?

Start with a demand no static frame could meet. Bone is loaded in ways that change over a lifetime — a growing child, a pregnancy, a heavy job, a fracture that alters how a limb is used. A structure poured once cannot be right for all of them. And there is a second problem: bone accumulates microscopic cracks under repeated loading, as any material does. A frame that cannot replace fatigued material eventually fails. What would a structure need in order to solve both?

It would need to be removable. Not repaired at the surface, but taken out and put back. And that is what actually happens. Two cell types work the bone surface: osteoclasts, which dissolve mineral and digest the matrix beneath them, and osteoblasts, which lay down fresh matrix that then mineralises. They do not work at random. They work in teams, at discrete sites, in a fixed order — a patch is excavated over a few weeks and refilled over several months, then falls quiet. Thousands of such sites are open in you at any moment. The adult skeleton is turned over continuously; you are not carrying the bone you had a decade ago.

Now the question that matters: what tells a team where to dig? Notice that neither cell can sense load — both sit on the surface, active only briefly. But there is a third cell we have not mentioned. When an osteoblast finishes, some become buried in the matrix they secreted, and there they persist as osteocytes: the most numerous bone cells by far, wired to each other and to the surface by fine processes running through tiny channels. Load a bone and it flexes minutely; flexing drives fluid through those channels past the osteocytes. That is the sensor — strain converted into a flow the buried cells can feel.

Follow the loop. Osteocytes under habitual strain signal to restrain resorption. Osteocytes deprived of strain — or killed outright by a microcrack tearing through their processes — signal the opposite, and a resorption team is recruited to that spot. So the excavation goes where loading is absent or where damage has occurred, and refilling follows the excavation. No plan is needed. The structure is continuously dismantled everywhere, and retained only where it is being used.

This is what Wolff described in 1892 when he observed that the internal struts of the femur run along the lines of principal stress. He read it as a mathematical law of architecture. We would now say the architecture is the residue of a feedback loop — and Harold Frost's mechanostat framing puts it well: there are strain thresholds, below which bone is removed and above which it is added, with a wide band in between where turnover simply maintains what is there.

The correction worth making is to the word "law". Wolff's law is a real observation dressed as more than it is: trabecular alignment does follow stress, but the original mathematical formulation does not hold, and the modern account is a cellular feedback loop with thresholds, not a rule of geometry.

The loop's properties explain the clinical picture. It responds to change, not to absolute magnitude — which is why habitual walking builds little, while a new and unusual load builds a lot, and why the effective stimulus turns out to be high strain applied in short bursts with rest between, rather than long steady effort. It is also local: loading the legs does little for the wrist.

It is not symmetric, either. Removing load acts fast, because resorption is quick and formation is slow. Astronauts lose bone from the hip and spine at roughly 1 to 1.5% per month without countermeasures — a rate that would take years to accumulate by ageing — and bed rest does the same. Rebuilding takes far longer than losing, and after a long unloading the recovery may be incomplete, particularly in trabecular architecture: once a strut has been fully resorbed there is no surface left for osteoblasts to work on, so the shape does not simply come back. This asymmetry, not the loss itself, is why prolonged immobility is taken so seriously.

• Why oestrogen withdrawal at menopause accelerates the loop, and what that implies about treatment timing.
• How a fracture heals, and why the callus is remodelled away once the bone is loaded again.

Every term the collection defines is gathered in the glossary.