Under gene therapy, a vehicle, usually a disabled virus, is used to transport a healthy gene around the body to replace a defective one which causes illness.
But the technique, still experimental, has not been risk free.
A young American teenage volunteer has died and the cancer has emerged among children treated for a rare immune deficiency disorder.
Gleicher wanted to see if he could introduce genes into an embryo, the idea ultimately being to reverse a single-gene defect in a lab dish and then reimplant the embryo in the uterus.
He extracted young cells called blastomeres from 16 human male cells and added one, two or three of these cells to each of 21 donated three-day-old female embryos.
He used male cells because the "Y" chromosome, which only males have, was easier to track in the developing embryos.
Twelve out of the 21 embryos "reached what we considered to be perfectly normal blastocysts," a cluster of cells that develops five to six days after fertilisation, Gleicher said on Tuesday at the annual conference of the European Society of Human Reproduction and Embryology (ESHRE).
However, four of the female embryos stopped growing and five developed abnormally.
Flaws in technique
An embryo of mixed sex or mixed species is called a chimera and to make one is illegal in many countries, although in others there are no specific laws.
The merged embryos were not intended to develop into a child, and were destroyed after a few days. But fellow experts have criticised the work as scientifically wrong and a breach of ethical standards.
Alan Trounson, a leading Australian fertility scientist, said Gleicher's work seemed "completely flawed."
There would be no time to test whether the correct version of the genes had been incorporated widely in the embryo before the embryo had to be transferred into the womb of the future mother, thus exposing the child to unknown health risks.
Another expert, Lyn Fraser, who chairs ESHRE's scientific committee, said she found it "hard to accept what they have done at all."
In any case, even if the healthy genes were incorporated across the embryos, the faulty genes would "all still be there," Fraser argued.
Gleicher, from the Center for Human Reproduction in New York and Chicago, said the study was a proof-of-principle to see whether embryonic gene repair is possible, and was not at all meant to intended as a "feasible or ethically acceptable" application.
But the British weekly New Scientist reported on Thursday he had already filed a patent for the concept.