Just 4% of Americans definitely believe – as President Donald Trump does – that taking Tylenol during pregnancy increases the risk of a child developing autism, a new KFF poll shows.

The poll comes following last month’s warning by Trump and the White House’s top health official, U.S. Secretary of Health and Human Services Robert F. Kennedy Jr., that taking acetaminophen, which is the active ingredient in Tylenol, can increase the risk of autism.

Though Trump and Kennedy blame Tylenol for putting children at risk for autism, science does not back up their claims. Their message of “don’t take Tylenol” has been rebuked by physicians, medical researchers and public health officials around the world.

As one example, the European Medicines Agency emphasized that its recommendations had not changed. It considers acetaminophen a safe way to address pain and fever in pregnancy, which are conditions that “can be dangerous to developing babies if left untreated,” according to reports. And in the U.S., scores of medical groups have also affirmed the drug’s safety with the American College of Obstetricians and Gynecologists warning that “the conditions people use acetaminophen to treat during pregnancy are far more dangerous than any theoretical risks.”

Thus, the KFF poll is good news for supporters of science. Very few, or just 4% of adults say the claim about a “causal relationship is definitely true,” KFF said in its poll analysis. Meanwhile, about one-third of the public, 35%, say the claim of a link between Tylenol and autism is “definitely false.”

But Trump’s claim has caused confusion. The KFF poll analysis says “most people are uncertain what to believe, with equal shares saying it is “probably true”and ‘probably false’ (30% in each case).”