Only decisive action within days of the first human to human cases of avian flu could prevent the triggering of a pandemic that would kill millions, scientists said today.

If the H5N1 virus currently circulating among birds in Asia evolves the ability to pass easily from person to person, health authorities will have just three weeks to contain it with drugs before it becomes a global threat, according to two sophisticted computer models.

Should this opportunity be missed, the result could be a pandemic that infects half the world’s population and kills even more than the 20 to 50 million who died in the "Spanish flu" of 1918-19.

Countries worldwide, including Britain, would be powerless to protect themselves against the virus, though judicious use of drugs and vaccination might reduce the death toll.

The findings, from two independent international research teams, show the urgency of building up a "mobile stockpile" of three million courses of the antiviral drug Tamiflu, that can be deployed anywhere in the world within three days. The World Health Organisation has just 120,000.

The models suggest that swift preventive use of the drug to treat people who come into contact with sick patients could contain an outbreak before it turns into a pandemic. They also point to the importance of thorough disease surveillance in the Asian nations in which a human H5N1 outbreak is most likely. If the first cases and clusters are missed, vital time will be lost and containment efforts could become futile.

"Control of a human outbreak is potentially possible but only when it is in its early stages," said Neil Ferguson, Professor of Mathematical Biology at Imperial College, London, who led the first study. "Once it is beyond this, once it reaches the UK, there is no chance of stopping it, only of mitigating its impact. We need to act quickly, detect cases quickly and treat people quickly. The challenges are great but the potential benefits are saving millions of lives.

"This is our only option for making a big difference to a pandemic outcome. If we let it spread we’ll at best be able to prevent the deaths of perhaps half those infected in the UK."

Ira Longini, Professor of Biostatistics at Emory University in Atlanta and the leader of the second team, said: "If detection and containment starts within three weeks or so, we have a good chance. Once the response takes more than a month there is a very poor chance of containment."

Both teams said the World Health Organisation (WHO) should buy three million doses of Tamiflu, which could be sent within three days to contain an outbreak anywhere in the world. WHO is currently negotiating over a large stockpile with the manufacturers, Roche, which is understood to be considering donating the drugs.

While there is an experimental vaccine against H5N1, a pandemic strain is likely to be genetically different from the present one, making it less effective. Tamiflu is considered the best option, as it can be used prophylactically to prevent infection.

H5N1 flu has caused 109 confirmed human infections and 55 deaths in Thailand, Vietnam, China and Indonesia, and the disease has also been found in birds in Russia and Kazakhstan. While the existing strain is virulent in humans, it is not readily passed between people - a key step in the development of a pandemic.

Professor Ferguson’s study, published in the journal Nature, examined the likely impact on Thailand of an outbreak of H5N1 flu. Detailed demographic data about the country’s population, such as numbers of households, schools and businesses and travel patterns, were used to make predictions as accurate as possible.

In the model, a single person in a rural village was infected with the mutant strain and began to pass it on to others. When nothing was done to contain this outbreak, the disease spread rapidly throughout the country, reaching the capital, Bangkok, within two months and then spreading abroad. A year after the first case, about 50 per cent of Thailand’s 85 million inhabitants had been infected. The model did not estimate deaths as it is impossible to predict how lethal a transmissible strain will be, but even if it is ten times less deadly than the present virus it would kill millions.

Professor Ferguson then considered what would happen if Tamiflu were given rapidly to everybody within a 5km or 10km radius of an infected person, and measures were taken to reduce contact by closing schools and workplaces. These approaches will contain an outbreak, but only if Tamiflu is given swiftly, preferably within 48 hours of a case being diagnosed. The prophylactic treatment campaign must begin before more than 30 to 40 people are infected, and 90 per cent must actually take the drugs they are given.

Professor Longini’s study, published in Science, analysed a smaller population of about 500,000 people, but reached similar conclusions. Antivirals contained the outbreak, so long as comprehensive action was taken within 21 days of the first case and 80 per cent of the potentially vulnerable population took the drugs.

Both models suggest that antivirals will be effective only if the virus’s replacement number - the number of new people infected by each sick patient - is no higher than 1.8. While there is no way of being certain about the replacement number of a new strain, both Professor Ferguson and Professor Longini said flu pandemics do not generally exceed this figure.

Peter Aldhous, who edited a recent Nature special edition on avian flu, said the twin studies create fresh pressure on international health authorities to step up preparations for a pandemic.

"History would be a harsh judge if we fail because the international community failed to provide the resources required or the political will to make these measures feasible," he said.

Explore World News

• World Agenda

• Iraq News

• US & Americas News

• Europe News

• Middle East News

• Asia News

• Africa News

• Ireland News

Times Recommends

• The Japanese Prime Minister is doomed: or is he?
• Brighton company 'involved in cyber attack'
• The art of survival: it's all in the mind