Cardiology is the specialty that taught medicine to test itself.

In August 1988, The Lancet published a trial that had enrolled 17,187 people arriving at 417 hospitals with a suspected heart attack. Some were given an hour of streptokinase, some a month of aspirin, some both, some neither. Five weeks later, 13.2% of those given neither had died of a vascular cause, against 8.0% of those given both. A tablet that cost almost nothing, swallowed in the first hours, had saved lives at a scale nobody had measured before.

Three years later the same field learned the opposite lesson. After a heart attack, extra heartbeats predicted sudden death, and drugs that suppressed those beats were prescribed on that logic. The Cardiac Arrhythmia Suppression Trial randomised 1,498 such patients to encainide, flecainide or placebo, and had to withdraw the two drugs early: 43 arrhythmic deaths on the drugs against 16 on placebo. The reasoning had been sound. The patients had died anyway.

And in November 1994, 4,444 Scandinavians with coronary disease and a raised cholesterol, randomised to simvastatin or placebo and followed for 5.4 years, settled an argument that had run for a generation: 256 deaths on placebo, 182 on the drug, a relative risk of 0.70. Whether lowering cholesterol was worth doing stopped being a matter of opinion that month.

These are three of the thousands of studies that shaped modern cardiology. The field changes its mind on evidence, publicly and by name, and it has done so more often and more quickly than almost any other branch of medicine: the beta-blocker, the stent, the anticoagulant that needs no monitoring, the diabetes drug that turned out to treat heart failure. Each time, the answer was in a paper years before it was in a textbook.

It is still happening. Recently, in August 2025, JAMA published a trial for people whose blood pressure stays high whatever they take. The trial enrolled 1,083 adults at 159 sites in 13 countries. Every one of them was already on two to five blood pressure drugs. The trial added lorundrostat, a pill that blocks the hormone aldosterone at its source, or placebo. After six weeks, systolic pressure had fallen by 16.9 mm Hg on the pill, against 7.9 mm Hg on placebo. The pill raised potassium and lowered sodium in some people, and fewer than 1% stopped it for those reasons. A blood pressure nothing else could move had come down. Practice moves the month the paper comes out, and the guidelines catch up within the year. It is a good culture. It has exactly one cost.