Diabetic complications

Why does mildly high blood sugar damage nerves and kidneys only decades later?

A person's blood sugar runs a little above the normal range, and they feel entirely well. Twenty-five years later their retina is bleeding, their kidneys leak protein, and their feet have gone numb.

Two things there deserve to be refused rather than accepted. Why so slow — what kind of poison waits a quarter of a century? And why those tissues, when blood carries glucose everywhere, and the muscle and fat that consume most of it are largely spared? A toxin that reaches everything and injures a short list is telling us something about the list.

Start with the chemistry, because it explains the clock. Glucose is not a poison in the way cyanide is; it blocks no enzyme. What it does is react slowly, without any enzyme's help, with amino groups on proteins — sticking to them, then over much longer periods maturing into stable cross-links between protein strands. The rate depends on how much glucose there is and how long it has to work, so mild excess multiplied by decades reaches the same place as gross excess multiplied by months.

We are not inventing this, because the reaction is already the basis of the standard test. Glycated haemoglobin is glucose stuck to the haemoglobin inside red cells. A red cell lives roughly 120 days (a recalled figure), and circulating cells are of all ages up to that, so the measure integrates glucose exposure over something like the preceding two to three months — a window derived from the cell's lifespan rather than chosen. The monitoring test is a specimen of the damage mechanism, read on a short-lived protein.

Which hands us the answer to the timing. Haemoglobin is destroyed every few months, so its modifications never accumulate. Now ask which proteins are not replaced on that schedule. Collagen in vessel walls and basement membranes, the crystallins of the lens, the structural proteins of nerve sheaths — turnover measured in years. Those are the molecules where early adducts have time to become permanent cross-links, and where each year's dose is added to every previous year's. Nothing has to accelerate for the disease to appear late. A constant, slow, poorly-reversible reaction on a long-lived molecule is enough.

Now the second puzzle: why that particular list of tissues? Ask what happens inside a cell when plasma glucose rises. Muscle and fat take glucose up through a transporter that only appears at the cell surface when insulin tells it to; when insulin action is absent or ineffective, those cells are actually taking up less, not more. But capillary endothelium, the kidney's mesangial cells, the retina's pericytes and the Schwann cells of peripheral nerve take glucose up down its concentration gradient, unregulated. They cannot decline delivery. Whatever is in the plasma is, near enough, in them.

That is the selectivity, and it is testable rather than decorative: the complications of diabetes are a roll-call of exactly the tissues whose glucose uptake is not insulin-gated. Inside those cells the excess spills out of ordinary glycolysis into side routes — the polyol pathway, the hexosamine pathway, activation of protein kinase C — alongside the glycation already described. A widely taught proposal unifies these as consequences of one upstream event, excess superoxide from overloaded mitochondria; I would call that a leading hypothesis rather than settled, since antioxidant trials have not delivered the protection it seems to predict.

One more piece is needed, or the timing still looks too abrupt. Damage that accumulates smoothly produces symptoms suddenly, because organs carry spare capacity: a kidney filters far more than survival requires, a nerve trunk has many fibres. The silent decades are the reserve being spent, and the symptom is the moment it runs out.

The load-bearing claim is narrow: injury is dose multiplied by time, acting on cells that cannot exclude glucose, and recorded in molecules that are not replaced. That predicts two things beyond what we have used to build it.

First, control should matter more the earlier it happens, because cross-links already formed do not undo. The long follow-ups of the major control trials support this: cohorts that had better control early retained an advantage over cohorts that caught up later, years after the two groups' glucose levels had converged. That "legacy" pattern is what an accumulation model predicts and what a model of moment-to-moment glucose toxicity does not. I have deliberately quoted none of the trials' effect sizes — the percentages vary with the complication counted, the era, and the endpoint definition.

Second, and this is where my account is weakest, it should apply across all the vascular damage of diabetes. It does not. Tight glucose control reduces the small-vessel complications — retina, kidney, nerve — much more convincingly than it reduces heart attacks and strokes. That dissociation is real, it is not fully explained, and it is the single observation that most limits the glycation story. If anything refutes the account as a complete one, that is it.

Two smaller corrections. Risk rises continuously with exposure rather than switching on at a cut-off, so the threshold defining diabetes is a clinical convention, not a biological edge. And exposure does not determine outcome — some people with long, poorly controlled hyperglycaemia never develop retinopathy, pointing to genetic modifiers we cannot yet name.

• Why large-vessel disease in diabetes responds so much less to glucose control than small-vessel disease does.
• Why some people with decades of high exposure never develop retinopathy, and what that genetic protection consists of.

Every term the collection defines is gathered in the glossary.