Statistics, Vaccines, and Autism

Scientific American reports on statistical problems of debates on mammography and autism:

Misapprehension of statistics and scientific process has been even more apparent in the misunderstandings surrounding vaccines and the onset of autism.

Given the age at which children receive immunizations and that at which many cases of regressional autism manifest themselves (in which a seemingly normally developing child suddenly loses much of the ability to communicate as well as other acquired functions), "by chance alone" there will be a lot of children who regress at some point after getting their scheduled vaccines, Daniel Salmon, a vaccine safety specialist at the U.S. Department of Health and Human Services (HHS), said here on Tuesday.

As he pointed out, however, "temporality is insufficient to show causality." But underlying—and perhaps highlighted by—this "logical fallacy," he explained, is a frequent hang-up of science communication: the devil is in the details, and the details can be complicated (and not too catchy) to explain.

When former Jenny McCarthy, an advocate of the vaccine-autism link, goes on CNN's Larry King Live and says, "'Vaccines cause autism,' that's a very clear, simple message," Salmon noted.

Most respected scientific bodies, however, are not prone to such blanket statements. In a 2004 report essentially dismissing the assertion that vaccines cause autism, the Institute of Medicine (IOM) was notably more measured than McCarthy, concluding that "based on this body of evidence, the committee concludes that the evidence favors a rejection of a causal relationship between thimerosal-containing vaccines and autism…" The report also included appropriate caveats about the limitations of the available data, which many in the anti-vaccine community took as evidence that there was in fact evidence for a link.

Much of the rallying around vaccines (and/or their thimerosal additive) and autism has centered on powerful stories about children who developed regressional autism shortly after receiving immunization. But "the anecdote is not data—though it often seems that way" in the public debate about autism's causes, Salmon said. And likewise, correlations are not the same as causation. All kinds of outside factors, from recommended vaccines to the size of the internet, can be plotted to match the rising curve of autism rates, he noted, underscoring that "I'm not suggesting that the internet causes autism."

Despite the dearth of durable data showing that vaccines can induce autism, some 25 percent of parents in the U.S. still believe them to be a possible cause (a statistic which is "really quite remarkable," Salmon said with a bit of distress).

And the cautious nature of science is unlikely to be able to dispel belief in the link any time soon. "It's exceedingly difficult—and some in epidemiology would say impossible—to prove a negative," Salmon said.