More needs to be done to monitor safety during drug trials on children, experts say.
The study looked at over 700 drug trials
An study of 739 trials found just 2% had set up independent safety monitoring committees with the others relying on internal monitoring.
The Nottingham University team said it was "unacceptable" as independent oversight was essential, the Acta Paediatrica journal reported.
But industry officials said there were already strict rules in place.
The researchers looked at trials funded by the pharmaceutical industry, government bodies and charities.
All drug trials monitor safety, mostly through their own internal processes which have to be reported to drug regulators within 15 days in the EU. Similar procedures are followed elsewhere in the world.
The team found adverse drug reactions, including bleeding, high blood pressure and seizures, were reported in just over a third of trials.
But trials with independent monitoring committees were more likely to be terminated early because of problems.
In total, six trials were ended - all of which had the independent committees.
The researchers said they would not expect the findings to be any different for trials involving adults.
Lead researcher Dr Helen Sammons, a professor in child health, said: "We feel that the small number of studies that reported having safety monitoring committees was unacceptable.
"It is invaluable to have an independent monitor who can swiftly question any adverse drug reactions or differences in illness and death rates between groups taking part in the clinical trials.
"Parents also need to be made aware of the risks of adverse drug reactions when a child takes any medicine so that they can make informed decisions that balance those risks against the possible benefits the drug may provide their child."
But the Association of the British Pharmaceutical Industry, which represents drug firms, said the procedures currently used were adequate.
A spokesman said: "Safety is already strictly governed. The independent committees are really only need in large scale trials conducted across a number of sites."