The drug, sodium valproate, marketed under the brands Epilim and Orlept, is used to prevent seizures, the idea being that the foetus could be harmed by lack of oxygen or by a fall if the mother has a convulsion.

The researchers studied 249 children aged between six to 16 who had been born to mothers with epilepsy.

Forty-one had been exposed to sodium valproate; 52 to another anti-seizure drug, carbamazepine; 21 to phenytoin; 49 to a mixture of drugs; 80 had not been exposed to any drug.

The children were given a standard intelligence test, based on nine verbal and socialising skills. Six out of the 243 could not take part in the test.

Children whose mothers had taken valproate alone had an IQ "in the low-average range," the study found.

Their verbal IQ, determined by an index called the Wechsler scale, was on average seven points lower than would have been expected, and they were likelier to take longer to acquire skills than the other children.

In addition, a score of 69 or below was more than three times likelier in children exposed to valproate alone compared to children whose mothers had not taken any anti-epileptic drug.

The researchers say the results are a source of worry, given that valproate has been in use for more than 40 years.

The drug has previously been implicated with a rise in neural-tube defects -- a key setback in foetal development.

They also say, though, that the results have to be interpreted cautiously. There could be statistical bias because only 40 percent of the 547 mothers originally approached actually responded, they say.

And they also record that a lower verbal IQ was also likelier if the mother had had frequent seizures, and valproate is effective in preventing some kinds of fits.

The research is led by David Chadwick, a professor at the Walton Centre for Neurology and Neurosurgery in Liverpool.

It is published in the Journal of Neurology, Neurosurgery and Psychiatry, part of the British Medical Association's publishing stable.

In a commentary also published in that journal, British neurologist Simon Shorvon described the findings as "deeply worrying ... potentially a major clinical concern."

"Further exploration of this vital area is now urgently needed," he said.

According to figures cited in the study, epilepsy affects about five in every 1,000 people in Britain.

One third of them are women of reproductive age; one in every 200 women attending antenatal clinics receive anti-epileptic drugs.