The dose you can't see and the supply chain you never think about

Two invisible costs

Every previous topic on this site has dealt in things you can count on a bill: machines, radiologists, dollars, wait times. This one is about the two costs of imaging that never itemise. The first is radiation, a genuine but easily-overstated risk that the field manages through the twin principles of justification (should this scan happen?) and optimisation (at the lowest dose that answers the question). The second is supply-chain fragility, which is almost never discussed until a factory in Shanghai closes or a Dutch reactor springs a leak — at which point it becomes the only thing radiology departments can talk about. Both costs share a property: they are invisible right up until the moment they are catastrophic.

The dose ladder is the literacy that matters

The single most useful thing a patient or clinician can carry is the shape of the dose ladder. Ultrasound and MRI sit at zero — no ionizing radiation at all. A chest X-ray is a reference unit of about a tenth of a millisievert, roughly ten days of natural background. From there it climbs steeply: a head CT is around twenty chest X-rays, an abdomen-pelvis CT perhaps a hundred, a multiphase study more still. The striking fact buried in the data is that dose is not fixed by exam type — the same CT study varied thirteenfold between the highest- and lowest-dose institutions in Smith-Bindman's work, which means most dose variation is a choice about protocol and equipment, not an inevitability of physics. That is also the good news: it is the part the field can optimise.

The 103,000 number, handled honestly

The 2025 JAMA Internal Medicine projection that a single year of US CT scanning will cause about 103,000 future cancers is the most important — and most contested — statistic in this topic, and it deserves to be presented with its dispute intact rather than as a clean headline. It is a model, not a body count: it scales organ doses from a large sample up to 93 million scans and applies risk coefficients from the BEIR VII framework, which were largely derived from populations exposed to far higher doses than a CT delivers. The American College of Radiology's objection — that no published study has directly linked CT scans to cancer, and that the linear no-threshold extrapolation may overstate low-dose risk — is a legitimate scientific position, not industry defensiveness. The honest reading is that the per-exam risk is small (on the order of 0.1% against a lifetime cancer baseline near 50%), the population-level number is large because the scan count is enormous, and the true value is genuinely uncertain. All three things are true at once.

Concentration is the real supply risk

The through-line of imaging's supply crises is concentration. Half the US contrast market depended on one GE plant; when Shanghai locked down in 2022, the second-largest supplier had no spare capacity and everyone else held under 5% between them, so a factory closure became a national rationing event within days — worsened by just-in-time inventories holding roughly a week's supply. The isotope chain is structurally identical: forty thousand US scans a day depend on technetium-99m eluted from molybdenum-99 that cannot be stockpiled because it decays in hours, produced in a half-dozen reactors mostly older than fifty years, two of which went dark simultaneously in 2024. Helium, cooling every superconducting MRI magnet on the planet, is non-renewable and geographically concentrated enough to have swung 50–100% in price. The uncomfortable lesson the contrast crisis taught — that a large Australian network could run on 10% of its normal contrast when forced to — is double-edged: it shows real resilience is possible, and it hints at how much routine use was never strictly necessary in the first place.