Late last winter, a committee of vaccine experts designing this season's flu shot considered their choices. They had two, and both seemed bad.

Should they stick with last year's formula, even though a new strain of the bug was ominously building strength? Or should they try to make a new vaccine and risk complications or delays that could result in a shortage or maybe even no vaccine at all?

In the end, the committee voted 17-1 to bring back last year's version, even though they feared they were telling millions of Americans to roll up their sleeves for shots that might not work very well.

Many of them probably agreed with Dr. Theodore Eickhoff of the University of Colorado, who said: "For the first time in many years of participating in these deliberations, I must add I am very uncomfortable with the recommendation."

What Eickhoff and the others dreaded is exactly what happened. That new strain of flu became the dominant variety, accounting for three-quarters of all cases as the disease got an unusually early start this fall.

About 83 million doses of vaccine were made, but no one really knows how much protection from illness it gives. It almost certainly will not be the usual 70 percent to 90 percent, and some experts fear it is below 50 percent.

"We agonized. We asked repeatedly, 'Is there another choice?' " remembered Dr. David Stephens, who chaired the panel and heads infectious diseases at Emory University. "The bottom line is, we weren't really given a choice."

Their experience shows the frustrating and often imprecise nature of humanity's labor to stay ahead of this perennial nuisance and sometime killer.

The flu virus mutates constantly. The Food and Drug Administration, with the help of its expert committee, must decide in late winter what varieties will be the biggest threats. Picking the best combination is a mixture of science, luck and seat-of-the-pants instinct.

"By the time you know what's the right strain, you can't do anything about it," said Dr. Michael Decker, head of scientific affairs at Aventis, one of the three U.S. vaccine makers.

The first inkling of something worrisome dawned on flu experts at the end of January. Just two weeks before committees were scheduled to meet at the World Health Organization in Geneva and the FDA in Rockville, Md., to settle on the makeup of this fall's vaccine, scientists who track the flu noticed a new strain was gathering mass.

The vaccine could theoretically protect against several strains of the virus, but because production is slow, the shot is limited to just three. Any of these flu bugs can make people very sick, but since it emerged in 1968 the one most likely to result in pneumonia or death is a type called H3N2.

Flu viruses are categorized according to the makeup of their two key proteins, hemagglutinin and neuraminidase, the "H" and "N" in their names. Changes in the virus' hemagglutinin is especially troublesome, since this is the protein the human body aims for when it makes antibodies to fight off the flu.

For five years, the vaccine had protected against an H3N2 strain called Panama. Now that virus had mutated. A version with two differences in its hemagglutinin was causing outbreaks in Asia and had also turned up in Europe and North America.

The FDA's committee met in February and heard the bad news: The current vaccine might not reliably keep people from catching this emerging strain, called Fujian.

Nobody knew if the new strain would die out or gain strength, but Dr. Roland Levandowski, the FDA's flu vaccine expert, warned that new flu variants sometimes spread rapidly.

The WHO had already postponed its decision on H3N2. The FDA committee did the same.

When the FDA committee met again in March, the situation was, in some ways, even worse. Ten percent to 20 percent of H3N2 viruses around the world were Fujian. But the Centers for Disease Control and Prevention was having trouble isolating a sample that could be the basis of a vaccine.

"This is a very urgent issue," CDC flu chief Nancy Cox told the committee. "We've been working on this very intensively for what seems like a very long time. We're very disappointed."

Still ahead were many other steps, as well. The Fujian strain's hemagglutinin and neuraminidase genes would have to be transferred into tame flu viruses that grow nicely in hens' eggs so vaccine makers could produce them in bulk. Even then, it would take weeks to know if the process would reliably generate the vast quantities needed.

"It became, Do we go with a vaccine we know will be partially effective?" remembered Eickhoff. "Or do we wait around and try to identify a possible candidate strain?"

When the vote came, only Peter Palese, head of microbiology at Mount Sinai School of Medicine in New York, chose to switch to the Fujian strain despite the unknowns. He worried that an ineffective formula would give the flu vaccine a bad name because many people might get sick.

The WHO made the same decision as the FDA. In hindsight, was it correct?

Decker recalled what happened in 2000. Delays resulting from a switch to a new strain, along with a virus that produced poorly, contributed to a vaccine shortage.

A last-minute change to Fujian this year "could easily have meant not only a severe shortage but also the wrong vaccine," he said. "Right now, people are saying, 'You idiot, why didn't you choose Fujian?' But what if Fujian had petered out?"