Most health systems now decide what to pay for by setting a treatment's benefit against its cost. In England, NICE values health gain in quality-adjusted life years and judges a treatment against a set cost for each year gained. Germany prices new drugs on their added benefit over existing care. The United States has no national threshold, and its law restricts the use of the quality-adjusted life year in Medicare's decisions. The same drug can be judged good value in one country and poor value in another, on the same evidence.
Almost every such judgement rests on a model: a simulation of what happens to patients over a lifetime, built from trials that ran for a year or two. The verdict depends on what the model assumes beyond the trial: how long the benefit lasts, how many people stay on the treatment, what they would otherwise have died of. Small changes in those assumptions can move a treatment from good value to poor. A model's track record, meaning how its forecasts compare with what later happened, matters as much as its conclusion.
The newest pressure comes from treatments for large populations. The GLP-1 drugs are cost-effective for an individual in many analyses, yet their total cost to a health system is large because so many people are eligible. Prevention poses the opposite problem: its savings arrive years later, and often in another budget. Value for one patient and affordability for a system are now separate questions, and the evidence speaks more clearly to the first.
We report what these evaluations conclude and how their models have fared. We do not make a judgement of value of our own.