Central sensitization

Why does pain sometimes persist for years after the injury that caused it has fully healed?

Here is a situation that ought to be impossible. A wrist is broken, set, and healed. The X-ray is clean, the swelling is gone, the surgeon is satisfied — and the person still reports pain, two years on, sometimes spreading past the wrist into the forearm, sometimes triggered by a shirt sleeve.

The tempting reading is that something must still be wrong at the wrist and we have not found it. But suppose we look for it and there is genuinely nothing. What then? We are forced to a harder question: what if pain was never a readout of tissue damage in the first place — and if it is not, what is it a readout of?

Start with the machinery, because the answer lives in a specific place. Damage in tissue is detected by nociceptors, whose thin fibres carry signals into the dorsal horn of the spinal cord. There the signal is handed to a second neuron, and only then does anything travel to the brain. Notice that the handover is a synapse, and a synapse is not a wire. It has a gain, and gain is a thing that can be turned.

So ask: does that gain ever change? It does, and the first version is fast and unmysterious. Deliver identical noxious volleys to the same fibre at a sufficient rate, and the spinal neuron's response grows with each one rather than staying flat. This is called wind-up, and its mechanism is known: the NMDA receptor sits at that synapse plugged by a magnesium ion, and the plug is only expelled when the membrane is already depolarised. So one stimulus does nothing much, but a train of them holds the cell depolarised long enough to unblock the receptor, and the next volley now lands on a much more responsive cell.

That explains amplification during the input. It does not explain amplification after. For that we need changes that outlast the stimulus, and several are documented: phosphorylation and trafficking of glutamate receptors so more of them sit in the membrane; loss of inhibitory tone, since the dorsal horn is normally held down by GABA and glycine signalling; microglial activation releasing BDNF, which shifts the chloride gradient in dorsal horn neurons so that inhibition weakens; and facilitation descending from the brainstem, where cells that normally damp incoming traffic switch to amplifying it.

Now the crucial question — how would we ever know this is central rather than peripheral? Ask what each account predicts. If the amplification lives in the injured tissue, then numbing that tissue should abolish it, and nothing outside the injured territory should hurt. If it lives in the cord, neither holds.

The experiment is routine in volunteers. Inject capsaicin into a patch of skin. Within minutes, an area of uninjured skin surrounding it becomes painful to pressure, and light brushing — normally carried by large touch fibres that signal no pain at all — is reported as painful. That surrounding zone is not injured. Anaesthetising the injected patch does not abolish it. Something downstream of the skin has changed what a touch signal means.

What would refute this. If secondary hyperalgesia vanished the moment the original site was anaesthetised, the amplification would be peripheral and the account collapses. If increased sensitivity never extended past the injured territory, we would need no central story at all. The spread into skin the injury never touched carries the most weight, because no account confined to the periphery can reach there.

Where I am simplifying, and where the field is unsettled. Everything above is well established in animal preparations and in short-lived human models. The step to a person in pain for years is weaker. Human measurement is indirect — questionnaires, quantitative sensory testing, temporal summation — and these proxies do not map cleanly onto the spinal mechanisms shown in rodents. "Nociplastic pain", the recent label for pain attributed to altered nociceptive processing without identifiable tissue damage, is a clinical category still being argued over. And plenty of persistent pain is not this: it is ongoing disease not yet found. Distinguishing those cases is a clinical judgement requiring examination of an actual person, and belongs to a clinician, not to a mechanism.

Two corrections are worth making explicit. First, this is not "the pain is in your head" in the dismissive sense. The amplification is a physical change in nervous tissue, and the pain it produces is as real as any other. What has changed is the relationship between damage and pain, not the reality of the pain.

Second, sensitization is not the opposite of habituation. A neighbouring piece asks why repeated noise fades while repeated pain sharpens, and answers it in terms of how a nervous system responds to a stimulus that is still arriving. This piece asks the further question: what makes the amplified state outlive the stimulus entirely. And it is a different question again from opioid-induced hyperalgesia, where the same spinal machinery is driven by a drug rather than by an injury — there the raised baseline is produced by the treatment, here by the original insult.

• Why some people's cords sensitize after a routine injury and others' do not, and how little of that is currently predictable.
• How conditioned pain modulation — pain suppressing other pain — is used to probe descending control in humans, and how noisy that measure is.

Every term the collection defines is gathered in the glossary.