Approval has been given for a study where some cardiac arrest patients will be given a dummy drug, rather than adrenaline to kick-start their heart.
There are concerns that adrenaline, used for more than 50 years as a key part of resuscitation, can do more harm than good.
But the study will mean some patients across England and Wales being given a placebo without their consent.
Experts say the research throws up a number of ethical challenges.
More than 50,000 people suffer cardiac arrests outside of hospitals each year in the UK.
During arrests their hearts stop beating properly, preventing blood from being pumped around the body. Most people do not survive - currently only 8% leave hospital alive.
Paramedics follow a protocol for anyone found without a sufficient heartbeat - this can involve chest compressions, shocking the heart with a defibrillator and giving a shot of adrenaline.
But over the past decade some scientists have suggested adrenaline could do more harm than good.
There is some evidence it may damage the brain by reducing blood supply to the head and could diminish the chances of survival.
The Warwick University study will involve 8,000 people in London, Wales, the West Midlands, and the South Coast and north-east of England who have a cardiac arrest.
Half the patients will receive the dummy medicine and half will get adrenaline - but as they will be unconscious there will be no consent gathered as the drug is administered.
Paramedics will also not attempt to get consent from relatives or passers-by as the researchers emphasise that time is critical during resuscitation.
But they say local residents will be given information on the trial and details about how to opt out.
Depending on local arrangements, this could involve ringing their ambulance service and being put on a computerised register - but details are yet to be finalised.
'No active consent'
After the arrest, relatives will not be told whether their family members received the drug or the dummy version.
Dr Daniel Sokol, a medical ethicist and barrister who is not involved in the study, said: "A challenging aspect of this research is the lack of informed consent regarding a life-and-death intervention.
"The question then is: 'Is this research sufficiently potentially beneficial that we can forego proper consent?'"
ANALYSIS: Ethics of life-and-death research
There will always be a drive to update conventional medical practice as science advances. And at the heart of modern trials is the idea of informed consent - all participants must agree to take part and be told about the risks.
But gaining consent for studies looking at life-and-death situations is difficult. People who are unconscious - either after a cardiac arrest or in intensive care, for example, will not be able to give consent while a study is taking place.
This raises questions about how to ethically advance treatment for cardiac arrest - which has extremely poor survival rates. The results of the trial could save many lives. But the researchers feel the only way to get these answers involves testing on a group of patients who cannot be part of the decision-making process.
And scientists will have grappled with the question of whether to inform relatives about the trial - whether the patient survives or not.
Researchers stress they have had lengthy discussions with the Oxford research ethics committee that granted approval for the study.
Prof Peter Weissberg, of the British Heart Foundation, said: "It is important to remember we don't actually know whether this routine use of adrenaline is a safe and effective practice.
"The concern is it could be doing patients more harm than good. The only way to answer this crucially important question is to do a well-designed clinical trial
"It is always difficult to conduct a trial in situations where people are too ill to give their consent. But there are well established ethical guidelines for undertaking such studies.
"What is unacceptable is to continue giving a treatment that could be doing more harm than good."
The study will start in the autumn.