The Risks Of Risk Reporting

The Risks Of Risk Reporting

January 31, 1999

By now, most readers will have heard news that a team of scientific experts empanelled by U.S. District Court Judge Sam Pointer has essentially vindicated silicon gel-filled breast implants of charges that they cause a raft of connective tissue and immune system diseases. That mainstream reporters have been so willing to accept this vindication has been widely hailed as a positive note in science journalism. But while the episode should be viewed positively, it raises an important question about the overall accuracy of science reporting.

In November, the National Science Panel concluded that there is no evidence that breast implants "cause or exacerbate" a range of illnesses including lupus, rheumatoid arthritis, sclerodoma, or other connective tissue or immune system diseases. What is surprising (indeed encouraging), is that this scientific consensus has been accepted so completely, and reported so broadly by journalists and editorialists alike.

Unfortunately, this reporting doesn’t necessarily indicate a general increase in scientific literacy among health reporters. Take, as an example, the reporting on another recent court decision concerning the science of secondhand smoke. Last July, U.S. District Court Judge William Osteen struck from the official record an Environmental Protection Agency (EPA) study in which the agency called secondhand smoke "a known human carcinogen." The reporting on that decision focused largely on procedural issues, including Judge Osteen’s opinion that the "EPA publicly committed to a conclusion before the research had begun" and "adjusted established procedure and scientific norms to validate its conclusion."

One very important bit of information has been overlooked in nearly all of the reporting on secondhand smoke, however. Even if the EPA had not committed such statistical manipulation, its published findings still would not prove secondhand smoke to be a carcinogen. Like much other research published before and after it, the EPA study found that secondhand smoke increased the likelihood (or relative risk) that non-smokers would develop lung cancer by roughly 19 percent (a relative risk of 1.19). Epidemiologists, however, generally view increases in relative risk of less than 100 percent (or 2.00) to be essentially meaningless.

When risk factors are lower than this, statisticians believe it is quite difficult to separate their findings from the statistical noise caused by various confounding factors. Therefore, a small finding is usually the equivalent of no finding at all, a point that, judging from published accounts, seemed to be lost on most reporters.

What does all this have to do with breast implants? The answer is unlikely to be obvious, but tucked comfortably inside the National Science Panel report is the following information. The best epidemiological evidence suggests that breast implants increase the relative risk of rheumatoid arthritis, and other autoimmune and rheumatic conditions by 15 percent; dermatomyositis and polymyositis by 52 percent; definite connective tissue diseases by 14 percent (statistically significant at the 95 percent confidence level); and Sjögren’s Syndrome, a chronic autoimmune disease, by 47 percent (statistically significant). National Science Panel members agreed (rightly so) that such low levels of increased risk are very weak associations and do not constitute compelling evidence that breast implants caused these diseases. In other words, the media got it right this time, but may not have realized why.

That the mainstream media ended up on the right side of this issue is helpful. That the mainstream press still ignores important concepts like "relative risk," however, bodes poorly for the coverage of future controversies.

Gregory Conko (conko@cei.org) is a Policy Analyst at CEI.