It's time to write up the recipe for next year's flu vaccine — and the nation's influenza experts are aiming for better protection than this year's shot wound up offering.

The flu vaccine must be reformulated every year to keep up with the fast-evolving influenza virus, and this year the government made a rare wrong bet on which strains would cause the most disease.

The flu season got off to a slow start, but it rocketed in mid-January because of some new strains that are sickening even people who got vaccinated. It seems the vaccine is a good match for only about 40 percent of the virus now spreading in the U.S., according to the Centers for Disease Control and Prevention.

On Thursday, the Food and Drug Administration brought together flu specialists for the annual rite of predicting what strains are most likely to strike next winter.

On the agenda: A complete overhaul, brewing next year's vaccine to protect against three strains not in this year's inoculation but that are circulating the globe now. They include a strain called Brisbane/10 that's responsible for much of the current U.S. misery — one first spotted in Australia late last February, too late for scientists to include in this year's vaccine recipe even if they had predicted it would gain steam.

"We didn't have enough information to know that this was going to be a serious player," said CDC flu director Dr. Nancy Cox. In fact, "we thought we were going to have a pretty mild season until a few weeks ago."

What virus strains to include in the vaccine is a decision made months in advance to give manufacturers time to brew more than 100 million doses before the next fall, a time crunch that makes keeping a step ahead of influenza difficult.

"Influenza viruses are changing all the time. We're trying to have a window which allows us enough time to manufacture a vaccine. We can't just turn on a dime," said Dr. Norman Baylor, FDA's chief of vaccine review.

Still, the nation has a pretty good track record: 16 of the last 19 flu seasons had well-matched vaccines.

The last time the vaccine didn't provide enough protection was in the 2003-04 season, when the Fujian flu emerged in Asia too late to be included in the vaccine formula. Even then, when researchers analyzed who got sick, the less-than-perfect vaccine worked about 52 percent of the time for healthy adults. Usually, it's about 70 percent to 90 percent effective.

Indeed, the CDC says it's still a good idea for the vulnerable to get vaccinated this year. Because Brisbane/10 is a relative of a strain in the vaccine, the shot should offer some protection, perhaps meaning a milder case, Cox said.

Flu viruses come in different strains that constantly mutate, until one that few people have immunity against emerges and is able to spread widely. Each year's vaccine contains protection against two varieties of the harsher Type A flu — subtypes known as H1N1 and H3N2 — and one from the more benign Type B family.

This year, two vaccine components turned out not to be a good match. That troublesome Brisbane/10 strain is different enough from the vaccine's H3N2 version, named Wisconsin, that it now accounts for most of the nation's laboratory-confirmed flu. A different Type B strain is causing illness, too.

It's too early to tell if this winter's flu will be more deadly than usual. Every year the flu infects up to 20 percent of the population, hospitalizes 200,000 people and kills 36,000.

Last week, the World Health Organization made its own recommendation for the recipe for the Northern Hemisphere's flu vaccine for next year, three strains that weren't in this year's shot: H3N2/Brisbane/10; another new Type A strain called H1N1/Brisbane/59; and Type B/Florida.

The U.S. always follows the health organization's lead.

But Thursday's meeting allows the FDA's scientific advisers to scrutinize the data before the U.S. officially accepts that recipe and manufacturers start the laborious process of growing virus in chicken eggs to brew into vaccine.