The PROFID investigators pooled individual participant data from 32,214 patients with a left ventricular ejection fraction of 35% or less after myocardial infarction: 7,477 with a primary-prevention implantable defibrillator and 24,737 without. They estimated propensity scores on age, sex, ejection fraction, renal function and diabetes, applied overlap weighting, and then split the cohort by year of inclusion into 1995–2004, 2005–2014 and 2015–2020. There were 12,097 deaths over a mean 43.7 months.
The association between having an ICD and surviving weakened steadily. The hazard ratio was 0.54 in the earliest period, 0.67 in the middle one, and 0.89 in the most recent, where the confidence interval ran from 0.73 to 1.07 and the weighted cumulative mortality curves barely separated. The test for interaction across periods was significant at P below 0.001, so this is not noise between subgroups.
The honest reading is that background therapy has moved. Early revascularisation, quadruple heart failure therapy and better secondary prevention have lowered the arrhythmic death rate the ICD was competing against. What this study cannot do is tell you to stop implanting, because it is observational: the patients who received a device in 2018 differ from those who did not in ways propensity weighting on five variables will not fully capture, and the most recent period has the fewest events and the widest interval.
- The finding is about the size of the benefit, not its direction — it has not reversed
- Observational and propensity-weighted, so residual confounding by indication remains plausible
- The most recent period carries the fewest patients and much the widest confidence interval
- Nothing here changes secondary-prevention ICD indications, which are a separate question
- Worth raising in device clinic when a patient asks what an ICD will actually buy them
The statistics, in plain English
A hazard ratio of 0.89 with a confidence interval from 0.73 to 1.07 means the data are compatible with a 27% reduction in death, with no effect at all, and with a 7% increase. When an interval crosses 1.0 the study has not shown a benefit — but it has also not shown the absence of one, and with the fewest patients in this period that width is expected. The significant interaction across periods is the more robust finding: whatever the true effect now, it is not the same effect as in 1995.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for cardiology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free