A lot of ink and pixels were spilled this week on a study that found "that maternal influenza infection was associated with a twofold increased risk of infantile autism," and "prolonged episodes of fever caused a threefold increased risk of infantile autism."
I can hear some readers howling already: The study did not find that fever "caused" an increased risk of autism, they're shouting. This is an association, not a finding of cause-and-effect. But "caused" comes directly from the "results" section of the abstract. To confuse things even further, the researchers report the following in the "conclusions" portion of the abstract:
Our results do not suggest that mild infections, febrile episodes, or use of antibiotics during pregnancy are strong risk factors for ASD/infantile autism. The results may be due to multiple testing; the few positive findings are potential chance findings.
So the discerning reporter has to ask: If these are potentially chance findings, why write about them at all? What is the chance they are chance findings? What is the chance they mean anything at all?
Admittedly, this is a difficult study to ignore. Autism is a big news story. But to their credit, many reporters who covered it were careful to note the limitations. Michelle Healy of USA Today began this way: "In a study that's already being greeted with notes of caution, Danish researchers report..." I like that. Put the caveat before the findings when they are as shaky as these are. She goes on to mention that the value of the study might be one piece of the puzzle in figuring out the causes of autism. She also says that current recommendations for treating pregnant women with the flu or fever should not change, according to the Centers for Disease Control and Prevention.
Maggie Fox at NBC News says the findings suggest "that in at least some cases, something is going on with a mother's immune system during pregnancy that affects the child's developing brain." And that's about all they suggest. Ryan Jaslow at CBS News takes a step too far, leading with "Moms-to-be who get sick during pregnancy may be more likely to have a child with autism, according to a new study." He quickly adds that mothers shouldn't be concerned, but after that lede, you'd expect them to be concerned.
And Bonnie Rochman at Time magazine goes way beyond the study's conclusions with this lede: "Expectant moms may have one more reason to get a flu shot." That certainly makes the connection far stronger than what was actually found. Nobody is saying flu during pregnancy causes autism, and nobody is saying that a flu vaccine will prevent it. The headline on the story is even worse: "Pregnant Moms' Flu Linked to Higher Risk of Autism Among Children." No. Remember the bit about how these are potentially chance findings? Why isn't that in the hed?
My favorite of all the stories I read was a commentary by the biologist and science writer Emily Willingham, whose headline at Forbes asked, "How much should you worry?" The finding "looks like big news," she begins. "But let's take a closer look." She walks readers through the study, pointing out all of its limitations, which go beyond what I've mentioned here. She does say pregnant women who are worried can get a flu shot, which might be a good idea for other reasons, but not something to do on the basis of this study, as I said above.
And she ends with a surprise personal observation that stopped my breath for a moment:
You can count me among the “flu-during pregnancy” population. I spiked a four-day 105 fever because of clinically proven influenza in my first trimester of pregnancy with our second son. But he’s not the son who’s autistic.
You might dismiss that as an anecdote that doesn't prove or disprove anything about the relationship between flu and autism. But it does remind us that it's a bad idea to be careless with studies, like this one, that will be read by people who are in the situation it describes. Imagine reading a draft of this story to a reader who is pregnant and just got over the flu. That might lead writers to temper their conclusions and emphasize the caveats.