To assemble its mRNA production network, Pfizer used its own money and didn't take any from the federal government. Executives said they didn't want to give agencies outside the FDA more leverage over the design of the trials.

Rival Moderna, which took funding from a division of the U.S. Department of Health and Human Services, suffered a three-week delay in completing its own mRNA trial after federal officials there asked the company to slow down enrollment to boost the racial and ethnic diversity of study subjects.

Starting in March, Mr. McDermott spent $500 million to buy and design equipment, more than double what he budgeted. Pfizer retrofitted an Andover, Mass., plant to make the mRNA, and assigned a St. Louis factory to make the raw materials for the shots. Numerous times it returned to suppliers or contractors to increase orders.

Instead of making mRNA in giant steel tanks, which would add months, Pfizer used disposable bags. Each fits at least 500,000 doses.

In April, the company began buying machines the size of a single-car garage that play a crucial role in production. These formulation machines deposit the mRNA vaccine in the fatty, or lipid, envelope that carries it to a cellular target. Without the safe passage, the vaccine breaks down during the journey.

Pfizer bought seven of the machines, each costing $200 million, destined for company plants in Puurs, Belgium, and Kalamazoo, Mich. It takes about a week to make a batch of doses, which then undergo two weeks of testing. The finished product is stored at subzero temperatures to stay stable and potent.

Mr. McDermott and his team took steps to maximize output. Companies typically put one vaccine dose in a single vial so a nurse doesn't need to draw out and measure how much to inject. Pfizer went with vials capable of holding multiple doses, because filling them was faster.

Demands for doses kept increasing, sometimes frustrating Mr. McDermott. He was a 30-year industry veteran who became part of Pfizer in 2009, and has years of experience making vaccines such as the pneumonia shot Prevnar. A hands-on engineer, he likes crawling under the machinery for a look.

Then on June 30, regulators in the U.S. and Europe said they wanted the late-stage trials to have far more subjects than initially anticipated -- 30,000 rather than 8,000.

That meant Pfizer and BioNTech would have to make triple the dose supplies it had planned -- within weeks. And Mr. McDermott's vaccine-research counterparts hadn't settled on a final candidate, which meant the company would have to make materials for two candidates.

"It was like, 'Oh, my God, we can't do this,' " Mr. McDermott said, recalling his reaction.

Within a day, he and one of his deputies, Kevin Nepveux, got a call from their Belgian plant. Engineers would speed up preparation of filling lines to make more trial doses. They also figured out how to squeeze another formulation machine into the facility, helping make commercial doses six months earlier.

It wouldn't be cheap. Mr. McDermott, working out of his New Jersey home office, rang up Pfizer Chief Financial Officer Frank D'Amelio. "I got an idea, and I need $10 million. Is it an issue?" Mr. McDermott said.

"Just do it," Mr. D'Amelio said.

Pfizer wasn't sure how long doses would last in a refrigerator before expiring, and the company didn't have time to conduct studies to figure it out. Manufacturing staff decided to play it safe and ship the supplies at subarctic temperatures.

In April, Pfizer started working with contractors to create a container to keep the shots cold. The suitcase-size box could hold nearly 5,000 doses. To test it, the company ferried containers on hundreds of trips around the world, visiting Dubai, Africa, and even the doorsteps of Pfizer employees.

Mr. Nepveux's grown son was visiting his parents' New Jersey home when a package arrived. "When it got delivered, he thought it was some kind of fancy beer cooler," Mr. Nepveux said.

Just before the July 4 weekend, Mr. McDermott's team showed off the container to colleagues. The team was excited because of the quick turnaround. Then came Mr. Bourla's reaction.

"What in the hell is CVS going to do with 5,000 doses?" he said. Pharmacies run by CVS Health Corp. and doctor's offices, he said, would need to receive fewer doses.

Mr. McDermott's team went back to the drawing board. Pfizer now plans to release a 25-vial, or 125-dose, package next year.

The decision to go ahead with the second vaccine candidate, made at the end of the month, was another shock. While colleagues on the call cheered the milestone, Mr. McDermott did some math in his head and groaned.

His team had anticipated the first candidate would be the choice. The second one required a larger dosage, which meant fewer shots could be packed in a vial.

By that point, Pfizer was targeting as many as 100 million doses for 2020. Mr. McDermott asked whether they could toggle back to 80 million. Mr. Bourla wouldn't relent. "Find a way," he said.

The late-stage trial, cleared to start on July 27, began quickly. Within hours, the first four healthy volunteers received injections at the University of Rochester Medical Center in upstate New York.

Evaluating whether the experimental shot protected against Covid-19 meant finding people for the trials in places where a study subject had a good chance of getting exposed. Then researchers could see if they developed symptoms or not, and whether they had taken the vaccine or placebo.

The company deployed mathematical models that crunched infection rates down to ZIP Codes, to identify hot spots.

Researchers hoped the seriousness of the disease would encourage people to sign up. Some 20,000 enrolled through August. By the end of the month, however, enrollment had veered off track.

There were two problems. People weren't signing up in target areas as fast as the companies expected. And the mathematical models weren't doing a good job at predicting virus hot spots more than a few weeks out.

The virus had shifted to other places by the time subjects had joined the pivotal study and taken both shots. Due to local efforts to curb the pandemic, infection rates were falling in places like New York and Washington state where researchers had been recruiting.

"It is like a puzzle, and the puzzle is changing," Dr. Jansen said.

Dr. Jansen convened researchers and computer scientists on Sept. 1 to review the virus's behavior and make changes to the trial.

Within days, researchers stopped enrolling patients at certain sites, including in New York City. They upped the recruitment goals in emerging hot spots such as Argentina. They also expanded the trial by 14,000 subjects to increase the odds of exposure.

Still, by early October, fewer subjects than expected had become sick. Pfizer scientists were perplexed. Perhaps study subjects were taking precautions to stay healthy, such as wearing masks and social distancing, the researchers theorized. Maybe the FDA-authorized tests Pfizer was using to confirm cases weren't accurate, Dr. Jansen said.

Pfizer stopped processing test results and asked the FDA to permit use of a different test. The FDA, meanwhile, wanted Pfizer to wait until 64 subjects became sick before taking its first peek, rather than just 32, as the company had planned.

By the time the two sides struck an agreement in early November, the issue was moot: 94 people had become sick, triggering a preliminary review that found the vaccine was more than 90% effective and looking safe.

Mr. Bourla learned the results Sunday afternoon. When he gave the findings to senior company officials on a call, Mr. Bourla said people jumped up from their chairs.

To get a head start building up supplies, in mid-August the company began producing doses at its Belgium factory that could go immediately into use should any country give clearance.

By the end of September, the company had set aside 1.5 million doses to be frozen and ready to ship.

Yet most of Pfizer's manufacturing network wouldn't be ready for making shots until October. Installing, testing and certifying the machines made supplies correctly took weeks, culminating in engineering runs in September.

Looking for workarounds, Mr. McDermott tweaked the production process in a way he figured would add 5 million more doses a week. But in the end, despite making enough mRNA for 100 million doses, Pfizer and its partners incurred a three-week delay securing enough raw materials to make the lipid nanoparticle that wraps around the vaccine. On Nov. 9, Pfizer disclosed it would make 50 million doses this year, half what Mr. Bourla wanted.

Still, the chief executive is looking on the bright side. "I always try to shoot for the stars, because I know that even if you miss it, you will land somewhere in the moon," he said. "But right now it looks like we landed, more or less, in the stars."

--Peter Loftus contributed to this article.

Write to Jared S. Hopkins at jared.hopkins@wsj.com

(END) Dow Jones Newswires

12-11-20 2149ET