Image via Wikipedia
According to an article entitled "Junk Science" (National Review, Vol. 46, October 24, 1994), here's the story:
"Investigators .... were perplexed at finding no discussion in the professional public-health literature of this ubiquitous [one in a million] rule, which is the make-or-break test for new pharmaceutical products. Eventually they traced it back to the cranberryless Thanksgiving of 1959. That classic scare about possible pesticide residues led the then Secretary of Health, Education and Welfare, Arthur S. Flemming, to advise people against serving cranberies with their turkey. In response to the ensuing uproar, Nathan Mantel, a biostatistician at the National Cancer Institute, was asked by superiors to suggest guidelines for future warnings. In his paper he said that safety was a matter of chance and that "for purposes of discussion we'll assume, that anything that poses a one-in-a-hundred-million risk of developing cancer is safe. When asked recently by the authors of the EPA Watch study how he came up with this figure, Dr. Mantel said. "We just pulled it out of a hat." But the Food and Drug Administration used Dr. Mantel's number when it issued a 1973 directive euphoniously entitled: "Compounds Used in Food-Processing Animals: Procedures for Determining Acceptability of Assay Methods Used for Assuring the Absence of Residues in Edible Products of Such Animals."
The proposed rule floated around inside the FDA until 1977, and when it was made final it came out as one in a million - some unnamed bureaucrat had knocked two zeroes off Nathan Mantel's number. At that, this first rule said explicitly that one in a million was intended merely as a screening level: if the initial tests indicated more than a one-in-a-million chance of causing cancer, then there should be more study of how the risks and benefits stacked up.
But regulatory creep quickly established one in a million as a rule by which agencies would oppose or reject all manner of products - all on the basis of a researcher's "I pulled it out of a hat" and an unknown bureaucrat's decision to knock off two zeroes."