Dexrazoxane won on four measures, and matching built the group it was measured against

Four measures, fifty patients

The numbers all point one way. Bleeding volume inside the heart muscle was measured at 2.0 per cent of the left ventricle against 6.3 per cent; infarct size came out at 29.1 per cent against 43.8 per cent, and left ventricular ejection fraction at 39.8 per cent against 34.7 per cent. Haemorrhagic infarction appeared in 6 of the 25 treated patients and in 16 of the 25 comparators. Intravenous dexrazoxane, 250 milligrams, was given before the artery was reopened and again at 4, 8 and 12 hours, and no drug-related serious adverse events were reported.^[1]

Before reading that table it helps to read how the group was built. The published paper describes the work as a single-centre, non-randomised, placebo-controlled phase IIa study: the 25 patients who received the drug were compared with 25 patients matched from 78 who received placebo. There was no draw; the matching was a choice about which patients entered the comparison group. Matching balances the variables you know about and leaves the ones you do not.^[1]

Who chose the comparison group

A second result landed the same week, and it reads from the opposite direction. At 69 sites, 959 adults were assigned in a double-blind, placebo-controlled design to three arms, and over six months none of the three groups — cognitive dysfunction, autonomic dysfunction and exercise intolerance — showed an advantage over placebo. That empty result says more than a 25-patient group winning on four measures can say, because a draw separated its arms. What makes the difference is how the comparison group was settled, rather than the drug or the disease.^{[1], [2]}

I asked the same question from the other end a day ago: in the column of 30 August 2026 on the ablation trial where a sham arm with no catheter still lifted the quality-of-life score, the comparison group carried an effect of its own. There, the sham showed how much of an endpoint that measures what a patient feels it could account for on its own. Here, the matched comparison group leaves it unclear how much of the imaging measures it could account for. In both cases the argument is about the group placed alongside, rather than about the drug or the procedure.^{[1], [3]}

The distance from image to patient

All four endpoints measured are imaging findings: bleeding volume, infarct size, ejection fraction and how often haemorrhagic infarction appeared. Death, readmission and heart failure were not measured at this size. The announcement's headline says the treatment could prevent millions of heart failures and deaths a year. The distance between a difference on images in fifty patients and millions of deaths a year is the distance this study does not close.^[1]

The authors stop at the same place and call for a randomised multicentre trial. That does not make the result worthless: haemorrhagic infarction has no specific therapy today, and the difference in bleeding volume is a signal worth testing. What remains open is in which patients, at which centres and on which endpoint the signal holds. The job of a phase IIa study is to make the next trial possible, and answering the patient is a different job.^[1]