What's the fastest way for the government to get word of life-saving medical discoveries out to the doctors and patients who need to know?

Nobody is sure of the answer, but almost everyone seems to agree that it's not medical journals, judging from comments by participants at a daylong conference held last week at the National Institutes of Health. In fact, evidence from six recent cases discussed at the conference suggested that, at times, medical journals can slow the flow of information.

All six involved research discoveries with the potential for saving lives or preventing disastrous complications like blindness or paralysis. In five of the six, researchers eventually published their findings in the prestigious New England Journal of Medicine.

In several cases, before alerting the public, researchers and government officials first negotiated with the journal's editors to make sure that publicizing the discovery would not hurt the study's chances of later appearing in what is regarded as the premier U.S. medical journal. In others, scientists refused to release more than a summary of their findings because they believed that the journal's rules forbade giving the details before the article appeared.

"There seem to be two enormous centers of power in the dissemination of crucial information in American medicine," said panelist Robert Bazell, a NBC News science correspondent, toward the end of the meeting. "One is the New England Journal of Medicine and one is the NIH."

Arnold S. Relman, the New England Journal's recently retired editor-in-chief, immediately disputed that assessment.

"We're just an instrument," he said. "We reflect the judgment and will of the community of physicians who produce the information and use the information."

Relman defended his journal's rule prohibiting prior publicity about the studies it publishes, and said the policy is widely misunderstood. Known as the "Ingelfinger rule" after Relman's predecessor at the magazine, the policy states that the journal will not publish a study if its findings have already appeared elsewhere. However, researchers are not penalized for presenting their work at a scientific meeting, even if a journalist writes an article about the presentation.

Relman said the New England Journal's editors are willing to waive the rule and allow prior publicity if researchers and government officials can persuade them that a study's findings are so important that doctors and patients urgently need to know about them. That is what happened in five cases discussed at the conference.

In four cases -- three studies involving promising treatments for AIDS or breast cancer, and one that found that widely prescribed drugs were causing an increase in cardiac deaths -- the journal's editors promptly gave researchers permission to publicize the results.

But in a case last year involving a readily available drug, methylprednisolone, that had been found to reduce paralysis following spinal cord injuries, editors at the journal took four months to review the manuscript reporting the study, and then required changes before accepting it.

Michael B. Bracken, a Yale epidemiologist who headed the research team, said scientists who conducted the study were unwilling to publicize the results until it had been formally accepted, because they felt the findings should undergo the critical review by outside experts that the journal routinely provides.

Eight months elapsed between the discovery that the drug could help seriously injured patients and the news conference announcing the finding. Then, researchers refused to release full details of the study, saying the journal's copyright rules prohibited it.

Michael D. Walker, director of the division of stroke and trauma at the National Institute of Neurological Disorders and Stroke, said the brief summary of the study's findings released by his institute provoked a barrage of complaints from doctors, who charged that the announcement said "too little, too late" and did not reach the right audience.

Relman said editors at the journal would have moved faster if they had been alerted by the NIH. He added, to the surprise of many at the meeting, that "the NIH is perfectly free to write up its own version of the results any time it wants" in such urgent situations, as long as it does not claim that its version is the same as what will appear in the medical journal.

Paul Meier, a professor of statistics at the University of Chicago, said researchers have a legitimate interest in publishing their findings in a respected journal, but that this goal should not prevent the release of life-saving information.

"There comes a point when we should be ready to speak, and somehow the schedule of the printer gets in the way," he said.

Scientists and government officials at the meeting differed on the question of how much outside review of a study is needed before important findings can be considered conclusive enough to announce to doctors and patients.

Anthony S. Fauci, director of the National Institute of Allergy and Infectious Diseases, said that his institute's expert panels periodically review the findings of ongoing studies of AIDS treatments. He said once such a panel decides to stop a study because a treatment has been shown to help patients, as has happened several times in studies of drugs for AIDS, doctors should be alerted promptly.

"If it's unethical to continue the study, isn't it then unethical not to let the private doctors know, because they may be doing something contrary to the study," he asked.

Even when everyone agrees that a medical discovery should be announced at once, no one is sure how to reach the right audience. Doctors prefer to read a study before their patients learn about it from news media reports. Yet Relman noted that no medical journal is read by all doctors. And not all doctors belong to medical organizations, use computer networks, watch medical-education shows on cable television, or even read their mail carefully.

NIH officials said that each time they have released information on new findings, they have been criticized for failing to inform enough physicians. "We have not worked out, at least in my mind, an adequate mechanism for the dissemination of information," said Fauci.

Relman contended that deciding when and how to make such announcements are a government responsibility, and should not be delegated to journals, drug companies or medical organizations.

"There are very few times when a government action can have a direct impact on what doctors do in the management of disease," he said. "There ought to be a mechanism here at the NIH or at the Centers for Disease Control or at the Food and Drug Administration . . . for getting information of this kind out to practicing physicians within a few days."