In a few short months, the weather will turn crisp, the holiday season will draw near, and the coronavirus may embark on its third consecutive winter of death and devastation.
That prospect has federal regulators and their scientific advisors engaged in a high-stakes guessing game.
The question: How should the COVID-19 vaccine change?
Certainly, the circumstances have changed. The coronavirus strains responsible for 97% of infections today — BA.4, BA.5 and BA.2.12.1 — didn't exist in 2021, let alone in 2020. Yet all of the vaccines currently available in the U.S. are designed to recognize the version that left China in January 2020.
The shots have done an admirable job. Researchers credit them with saving 1.9 million U.S. lives in their first year of availability, and they continue to provide solid protection against severe illness and death from COVID-19. The ubiquitous omicron subvariants, however, have several mutations on their crucial spike proteins that make them less recognizable to an immune system primed to fight the 2 1/2-year-old virus.
The result: A real-world study found that the protection from three doses of mRNA vaccine is half as strong against omicron compared to the Delta variant that preceded it. Lab studies also have shown that exposure to omicron prompts the vaccinated immune system to pump out far fewer antibodies.
On top of that, the immunity induced by vaccines and initial rounds of boosters has waned. The Biden administration has placed an order for 105 million doses for a fall booster campaign, with an option to buy millions more.
Determining which recipe should go into those doses to induce optimal protection — the vaccine's "composition" — is an exercise in educated guesswork.
Their challenge comes in multiple parts. First, experts need to predict which coronavirus variant — or variants — will dominate the U.S. landscape in late 2022 and into 2023.
Then they need to forecast whether altered versions of the mainstay vaccines will be effective, and whether they pose new challenges regarding safety, cost or timely distribution.
Getting a timely and accurate fix on all this "is truly a challenge, and it is science at its hardest," Dr. Peter Marks, the U.S. Food and Drug Administration's vaccine chief, told the agency's independent advisors last month.
If the FDA waits too long to act on the scant available data, Americans may have subpar protection when coronavirus surges again, Marks warned. But speed involves risk. If manufacturers roll out reformulated vaccines without first testing them in full clinical trials, the FDA will need to trust that existing surveillance systems can quickly detect any new safety problems, he said.
"We're being asked, essentially, to have a crystal ball," fumed Dr. Arnold Monto, who chairs the FDA's vaccine advisory committee.
And if health officials make the wrong choice, a pandemic-weary nation could lose confidence in a vaccine that we're likely going to need for years to come.
In many ways, the task of updating COVID-19 vaccines is a turbo-charged version of a dilemma vaccine experts have faced for decades with the influenza vaccine.
Each year, several strains of the wily flu virus circle the globe and infect new populations. In the process, the mix of circulating strains changes, as does their genetic makeup.
So in early fall and early spring, the World Health Organization gathers an international group of vaccine experts to assess which mix of viruses is most likely to circulate in the next six months. Their findings dictate the composition of the shots offered for the flu season ahead.
Sometimes their predictions are accurate, but not always. Between 2001 and 2010, annual flu vaccines protected against circulating influenza B strains just 50% of the time. In the 2014-15 flu season, a wrong guess about influenza A viruses led to 758,000 flu hospitalizations among older Americans and 148 flu deaths in children.
The preferred COVID-19 vaccines in the U.S. — Pfizer and BioNTech's Comirnaty and Moderna's Spikevax — have a key advantage over traditional flu vaccines: Their mRNA technology allows the composition to be altered with comparatively blinding speed. Millions of targeted doses of both vaccines will be available by September or October.
To stretch a booster shot's protection across the widest possible expanse of variants, it would be ideal to know which of the omicron subvariants is most genetically distinct from the original coronavirus strain. Pairing the two in a single shot would give the vaccine "breadth," making it more likely to offer protection against a wide range of strains that continue to circulate.
It's easy to assume that because they cropped up most recently, BA.4 and BA.5 (which share the same spike protein) have the least in common with the ancestral strain.
But the fledgling field of genetic epidemiology isn't so sure. Dr. Kanta Subarrao of the University of Melbourne, who researches the immune system's response to emerging viral diseases, said the BA.1 subvariant of omicron is actually the bigger outlier.
If it were up to her to decide which version of the coronavirus to target with a fall booster, "I would choose BA.1," Subarrao said at the recent meeting of FDA advisers.
The agency didn't agree. At the end of June, the FDA asked vaccine manufacturers to produce "bivalent" doses that combine the original vaccine with one designed to recognize BA.4 and BA.5. Who will be advised to get it has yet to be determined.
Dr. Paul Offit, a virologist and immunologist at the University of Pennsylvania, thinks both approaches are wrong. In his view, the original vaccine and boosters are doing a fine job of preventing serious illnesses and death, and the potential upside of targeting omicron is too uncertain to justify the risks of releasing a shot that hasn't been subjected to a full clinical trial.
Sure, a bivalent vaccine booster might spur the immune system to generate more antibodies than a regular booster, but that doesn't necessarily mean recipients would be better off, Offit said in an interview.
"If they'd given me a choice of choice of voting 'no' or 'hell no,' I'd have voted 'hell no,'" he said.
To Offit and most other vaccine experts, the ideal response to a shape-shifting virus would be to develop a universal vaccine capable of neutralizing any and all variants that may emerge.
It is an approach that has been the holy grail of flu research, and despite years of work, it remains maddeningly elusive.
To withstand a virus' evolution, some scientists have tried to design vaccines that target a component that doesn't change — say, a protein that performs some housekeeping task that's crucial to its survival. Others have looked to animals such as llamas and alpacas, who produce very tiny antibodies capable of adapting to viruses when they change.
More recently, a team led by scientists at Caltech developed a multipronged vaccine that contains pieces of eight betacoronaviruses, including the one that causes COVID-19. When given to mice and monkeys, it trained their immune systems to recognize a wide range of viruses by showing them, in effect, a miniature family portrait.
The next step is a Phase 1 clinical trial in humans, where many promising universal vaccine candidates have foundered.
"It's not for want of money. It's not for want of desire or effort," Offit said. "It's just hard to make universal vaccines work."