Seth Berkley, president of the International AIDS Vaccine Initiative (IAVI), said there were many obstacles and few drugs in the pipeline, after the best hope failed in trials earlier this year.

"The earliest date to get a vaccine and a license is probably 2009," said Berkley. "But that's based on current timetables," he added, referring to the time it takes for drugs to go through the necessary testing phases.

One or two potential vaccines would begin trials in 2004 or 2005 and the trials would take four or five years, he said.

"How much has our world done to try to end this epidemic? The answer is virtually nothing. The only way is through a vaccine," Berkley said. The role of IAVI, a non-profit organization founded in 1996, is to make that happen.

Scientists were stumped by the virus, as unlike most illnesses it leaves no survivors. It also varies from region to region, presenting a problem for companies wanting to run tests.

"We don't know if we need a single vaccine or a cocktail of vaccines or whatever," said Wayne Koff, IAVI's senior vice-president for research and development.

Drug companies have also been slow to take up the challenge, since they sense no potential for money-spinners in the developing world where AIDS is most prevalent.

"The bigger companies are waiting," said Koff, although he added the U.S. firm Merck & Co. and Franco-German Aventis SA were developing one of about six drugs considered by IAVI to be the best hopes for a vaccine.

He said the first attempt to develop a vaccine, the AIDSVAX drug developed by U.S. biotechnology firm VaxGen Inc, failed in February after three years of trials costing about $150 million.

VaxGen's tests were in Europe and North America, close to laboratories for analyzing samples taken from the thousands of people involved in its trials. Firms face bigger bills if they test drugs in AIDS hotspots such as Africa.

Even a successful vaccine may be only 30 or 40 percent effective, enough to slow the epidemic and make drugs firms sit up and take notice.

"We think the next major advance is to find something that works a little bit so the rest of the field can build on it," Koff said.