The news media, by definition, report new findings, not old ones. Their purpose is not to tell the world that everything is as they thought it was yesterday, but to tell us something new. And, given the frenetic competition within the media, the more unexpected the finding, the more coverage it gets. Thus, the assertion by Pons and Fleischmann, the University of Utah chemists who announced a few years ago that they had produced nuclear fusion at room temperature (cold fusion)-a highly implausible claim-was big news. On the other hand, a confirmatory finding of a scientifically plausible effect is unlikely to make the news. Furthermore, a study that finds no effect will receive almost no attention from the media, even when it is important. One of the consequences of the emphasis on both newness and positive effects is that a study that erroneously finds an effect gets more attention than ten studies that didn't. For example, about 10 years ago a study was published purporting to find an association between drinking coffee and cancer of the pancreas. It received an enormous amount of publicity; the authors were on television network news, and the newspapers and newsmagazines gave the study prominent coverage. Later, when it became apparent that the finding was almost certainly spurious, this news was relegated to the back pages.
Reporters are reluctant to downplay their stories by adding caveats. When a study finds a weak risk, reporters are not likely to emphasize how small the risk is and therefore how likely it is to be spurious. If they do, these qualifications are usually buried near the end of the story. Caveats are simply not a winning feature of news reports. Thus, weak risks, such as moderate obesity or eating red meat, may receive the same emphasis as strong risks, such as cigarette smoking or heavy drinking, and the public may react equally strongly to both. Furthermore, media reports are also likely to frame risks in the most impressive way. For example, a recent study, called the GUSTO trial, compared two anti-clotting agents in the treatment of heart attacks-one agent was t-PA, the other streptokinase. At the end of the trial, 6.3 percent of the patients receiving t-PA had died, compared with 7.3 percent of the patients receiving streptokinase. This one percentage point improvement was hardly a great difference. In fact, stated another way, the study showed that the chances of surviving a heart attack increased from 92.7 percent with streptokinase to 93.7 percent with t-PA-pretty good odds with either drug. Equally accurate would be the statement that t-PA was associated with a 14 percent reduced mortality. Somehow, framed this way, the finding sounds much more impressive. And that's exactly how the media tended to frame it. Even when reporters caution their readers not to embrace too enthusiastically news of a weak effect from a single study, the headline writers may have other ideas. In the Boston Globe, the t-PA finding was reported under the headline "Anti-Clotting Therapy Found to Spare Lives." While technically true, the effect of t-PA was meaningful only when applied to a large population. For an individual patient, the drug offered only a trivial increase in the odds of survival.
All of this is not to say that the media are not doing their job. They are. It's just that the job is not what we might think it is. The job of reporters is to tell the public what happened, and to do so in as engaging a way as possible. In medical reporting, this means telling the audience, as dramatically as possible, what researchers did and what they concluded. Reporters do not usually include in the scope of their job providing a context-that is, analyzing the strengths and weaknesses of a whole body of scientific evidence. Nor does the job include coming up with a reasoned conclusion for the public, although good reporters often do this. Instead, most reporters simply inform the public, study by study, of whatever research is most newsworthy that day-that is, most startling.
But startling research is more likely to be wrong than confirmatory research. Solid conclusions are usually reached bit by bit. The more studies done of a particular question, the more accurate they are likely to be, as the errors of earlier studies are avoided. News of the entire sequence is unlikely to make it to the media, at least not toward its end, when it is most reliable. To be sure, the media often present very good feature stories about medical subjects. In these, reporters analyze at some length what is known about the field, what questions remain to be answered, and what the implications are. These longer, more analytic stories often appear in newsmagazines, in the health sections of large newspapers, or in the Sunday feature pages. They emphasize a body of research, not just a single study. But the lion's share of media coverage of medical research is news of a single, dramatic study, and this is where the problems lie.
One frequent justification for the media making much of weak risk factors is that they are important for the public health, if not for individuals. An individual may not see much difference between his or her chances of surviving a heart attack with t-PA or streptokinase, but for the public health, it matters. Since there are nearly 1.5 million heart attacks a year in this country, a one-percentage-point improvement in the chances of surviving translates into 15,000 lives saved. Another example: A trial using cholestyramine to lower serum cholesterol about 9 percent in middle-aged men with high cholesterol levels reduced their seven-year risk of heart attacks or sudden death from 8.6 to 7.0 percent. Although such a reduction may not seem like much to an individual, particularly since it requires taking a drug with side effects, when spread over the estimated 1 million to 2 million Americans with similar cholesterol levels, it could account for up to 32,000 fewer heart attacks a year. The public health perspective has been gaining steadily in importance in the past several years, as policymakers have grown increasingly concerned about costs. Anything that reduces health-care expenditures, even if it doesn't reduce risks much for individuals, is important. But this switch in emphasis has occurred nearly subliminally. Reporters are rarely explicit about it. They may not always be aware of it themselves. Health risks continue to be reported as though they were meaningful for the individual, and, of course, people assume that the risks would not be in the news unless they were. But even well-established risk factors may have little importance for individuals. For example, the 10-year risk of death from cardiovascular disease is 4.9 percent in middle-aged men with high cholesterol levels. This difference in risk of about three percentage points may not be enough to induce an otherwise healthy man to try to lower his cholesterol level. Yet many American men have been led to believe that high cholesterol is a death sentence, and low cholesterol means they will have a long life.
[One Globe columnist's] lament is a little different from the problems of distinguishing weak from strong risk factors, sensational from solid studies, and news of importance to individuals from news of importance to the public health. [Ellen Goodman] was complaining that studies were inconsistent and often contradictory. Let's look at these complaints more closely. It is true that one study may find, say, that postmenopausal estrogen is associated with breast cancer and the next study may find that it isn't. Such inconsistency is common in medical research. It is particularly common in epidemiology, because these studies are so difficult to do. Instead of being a cause for lament, the phenomenon ought to suggest more caution in accepting the results of any one study. Caution is what inconsistency teaches scientists, and there is no reason why the public-and reporters and columnists-cannot draw the same lesson. Furthermore, the inconsistency may be more apparent than real. If a risk factor is very weak, but statistically significant, the media report this as showing an effect. Another study may have found exactly the same relative risk, but because the study was smaller, the difference was not statistically significant. The media would report the second study as showing no effect. The possibility that there was really no difference between the two study results would simply not be reported.
Goodman was also complaining about another problem. Some studies show that something is good for you in one way and bad in another. As she said, "estrogen may protect against heart disease and give you a better shot at breast cancer. If you run a lot, your bones may get brittle but your heart will stay strong. If you drink wine, you could wreck your liver but lower your bad cholesterol." There is no answer for this lament. Nature simply did not set out to make things uniformly good or uniformly bad for us. They are what they are. But the situation underscores another important problem in reporting health news. Media stories about a research study that focused on jogging and osteoporosis (Goodman's "brittle bones") too often reported the results in isolation. In reality, almost any food or activity that affects health in one way also affects it in others. It is a disservice to the public not to try to put stories about single research studies in their larger context. People feel whipsawed by science, when they are really being whipsawed by the media.
As the recent epidemiologic studies of breast implants and the diseases they are said to cause have been reported in the scientific literature, how have the media responded? By and large, very well. Since the studies have been so consistent, there has been little necessity to deal with contradictions. The limitations in the strength of the findings caused by the size of the studies have been appropriately mentioned, as were the flaws in the one study that found a link. Failure to find a difference between women with and without implants does not mean that there is no difference. The smaller the study, the more likely that a real difference will be missed. This fact is usually expressed by a "confidence interval," that is, a range of possible risks that is compatible with the evidence. In the Mayo Clinic study, for example, the relative risk for the diseases studied in women with breast implants was 1.0. This meant that compared with women without breast implants, those with implants were no more or less likely to develop the diseases in question. But because there were only 749 women with breast implants in the study, it was 95 percent possible that the relative risk was anywhere from 0.5 to 3.0. The closer to the extremes of the confidence interval, the more improbable, but it was still quite possible that breast implants were associated with as much as a threefold increase in these diseases. The press made this clear. Larger studies will have a narrower confidence interval. Therefore, the more large studies can find no association between implants and disease, the smaller any real risk must be not to have been detected. On the other hand, if there is a risk that has been missed, larger studies are more likely to find it.
What if it turns out that there is a small risk? A quick answer would be that no risk is acceptable. In this view, if there is any risk, no matter how small, implants ought to remain banned. After all, we should accept no unnecessary threats to health. Unfortunately, the real world cannot work that way. We accept unnecessary risks to health every day. Whenever we drive our cars or take an antibiotic or eat peanut butter, we take risks. (Antibiotics, after all, can have serious side effects, peanuts contain a mold that increases the risk of liver cancer, and automobile accidents are far more important causes of death and injury than most of the health hazards that occupy our attention.) Paradoxically, then, unnecessary risks may be necessary. The important question is the size of each risk and the costs, not just in dollars, of avoiding it. If a risk is so small that it is nearly impossible to detect, then perhaps it doesn't matter, just as it might not matter to an individual whether he or she receives t-PA or streptokinase after a heart attack. Maybe we should stop worrying and include whatever risk there is from breast implants with the multitude of small (and not so small) risks we accept willingly every day. On the other hand, we might be unwilling to accept certain small risks, but quite ready to accept others. The crucial point is that we owe it to ourselves to decide these matters explicitly. Unfortunately, Americans are not very good at evaluating risks of health hazards and deciding which ones we are willing to take. Oddly enough, many of us are also reluctant to rely on science to tell us about health hazards. Increasingly, we look to other sources, and some of these sources are strange indeed.
#
Marcia Angell, M.D., is executive editor of The New England Journal of Medicine. Reprinted with permission from her book, Science on Trial, published by W.W. Norton. Copyright (c) 1996 by Marcia Angell, M.D.