When the Spanish flu reached the United States in the summer of 1918, it seemed to confine itself to military camps. But when it arrived in Philadelphia in September, it struck with a vengeance.

By the time officials there grasped the threat of the virus, it was too late. The disease was rampaging through the population, partly because the city had allowed large public gatherings, including a citywide parade in support of a World War I loan drive, to go on as planned. In four months, more than 12,000 Philadelphians died, an excess death rate of 719 people for every 100,000 inhabitants.

The story was quite different in St. Louis. Two weeks before Philadelphia officials began to react, doctors in St. Louis persuaded the city to require that influenza cases be registered with the health department. And two days after the first civilian cases, police officers helped the department enforce a shutdown of schools, churches and other gathering places. Infected people were quarantined in their homes.

Excess deaths in St. Louis were 347 per 100,000 people, less than half the rate in Philadelphia. Early action appeared to have saved thousands of lives.

Scientists are still studying the 1918 pandemic, the deadliest of the 20th century, looking for lessons for future outbreaks — including the possibility that H5N1, the avian influenza virus, could mutate into a form spread easily from human to human. This month, researchers published two new studies in The Proceedings of the National Academy of Sciences comparing public-health responses in cities like St. Louis and Philadelphia.

Using mathematical models, they reported that such large differences in death rates could be explained by the ways the cities carried out prevention measures, especially in their timing. Cities that instituted quarantine, school closings, bans on public gatherings and other such procedures early in the epidemic had peak death rates 30 percent to 50 percent lower than those that did not.

“It had been received wisdom that these interventions didn’t work,” said Dr. Richard Hatchett, the lead author of one of the studies, “because they looked at the variability between cities and concluded that there was some other factor than the interventions that caused the differing outcomes.

“That we were able to go back and ask the right questions,” Dr. Hatchett said, “is a function of a lot of modeling work that we did previously.”

Dr. Hatchett, who is a researcher at the National Institutes of Health, said the findings might hold lessons for the 21st century. “When multiple interventions were introduced early, they were very effective in 1918,” he said, “and that certainly offers hope that they would be similarly useful in an epidemic today if we didn’t have an effective vaccine.”

A two-week difference in response times, according to the researchers, is long enough for the number of people infected in an influenza epidemic to double three to five times.

Dr. Martin Cetron, director for global migration and quarantine at the federal Centers for Disease Control and Prevention in Atlanta, found reason for optimism in the study results.

“The thing I find encouraging about the Hatchett paper,” he said, “is that when you look back to 1918, you find that those who used nonpharmaceutical measures effectively were able to mitigate the impact of the severe pandemics, and this is consistent with some of the 21st-century simulation models.”

The second study, in the same issue of The Proceedings, suggests that in one sense preventive measures can be too effective. In an influenza epidemic, a certain number of people survive the illness and are immune to reinfection. As these numbers increase, the epidemic fades.

But an effective prevention program without a vaccine can leave enough people uninfected and still susceptible to the virus to start the epidemic again as soon as the controls are lifted. This is what happened in St. Louis. On Nov. 14, 1918 — in high spirits three days after the armistice that ended the war, and with influenza cases declining — the city reopened schools and businesses. Two weeks later, the second wave of the epidemic struck, this time with children making up 30 percent to 40 percent of the infections. Controls were immediately reinstituted.

The study examined the course of the epidemic in 23 cities: San Francisco, St. Louis, Milwaukee and Kansas City, Mo., had the most effective prevention programs, and time was of the essence. If restrictions were introduced too late or lifted too early, success rates declined substantially.

Neil Ferguson, a co-author of the second study and a professor of epidemiology at Imperial College London, explained in a telephone interview that the most successful interventions were in communities where the political and health authorities broadly agreed on what needed to be done and got significant cooperation from the public.

The key, Dr. Ferguson said, is to tune an imperfect intervention perfectly so that a single peak of minimal size is the result. Although no cities succeeded in doing this, those that got closest, like St. Louis, carried out early interventions before the first peak, and then reinstituted them when transmission rates began to rise again.

What these results mean for a future epidemic is not clear. “If avian flu became a pandemic tomorrow,” Dr. Ferguson said, “we would start a crash program to make a vaccine.”

But he added that rigid preventive measures like quarantines, mandated mask wearing and widespread business closings would still need to be put in place.

“What our study shows,” he continued, “is that interventions even without a vaccine can be effective in blocking transmission. What’s much less certain is whether society is prepared to bear the costs of implementing such intrusive and costly measures for the months that would be required to manufacture a vaccine.”