More journalists with reading disabilities

« previous post | next post »

Or maybe this is about press-release writers who don't express themselves clearly. According to "Chemical in Tap Water Linked to Food Allergies", Drugwatch 12/7/2012 (emphasis added, here and throughout):

A new study in the Annals of Allergy, Asthma and Immunology links high levels of dichlorophenols–chemicals used for chlorinating tap water–to a higher risk of food allergies. According to the study, people with higher levels of these chemicals in their urine have a greater risk of developing food allergies.

Out of the 2,211 people with high levels of dichlorophenols who participated in the study, 411 had food allergies and 1,016 had an environmental allergy, according to researchers.

Or "Pesticide Suspected in Rising Food-Allergy Cases", Voice of America 12/7/2012:

The researchers identified more than 2,500 individuals with measurable levels of dichlorophenols — chemicals found in pesticides and chlorinated water — in their urine. The study team, led by allergist and immunologist Elina Jerschow, narrowed their sample down to 2,200. Out of this group, Jerschow says, 411 of the subjects had some sort of food allergy.

Or "Dichlorophenols found in pesticides, tap water may trigger food allergies" Healio 12/7/2012:

Researchers learned that 2,548 participants (aged 6 years and older) of 10,438 people in the study had measurable dichlorophenols (greater than 0.14 mcg/L) in urine and allergen-specific serum IgE testing.

Or again, "Tap Water Chemical Dichlorophenol Linked With Food Allergies: Study", Huffington Post 12/3/2012:

The findings are based on 10,348 people who participated in the U.S. National Health and Nutrition Examination Survey between 2005 and 2006. Of those people, 2,548 had detectable levels of dichlorophenols in their urine, and of those people, 2,211 were actually used for the study.

Of those 2,211 people, 411 had food allergies, while 1,016 of them had some sort of environmental allergy, researchers found.

And in the Daily Mail, "Chlorine in tap water linked to increase in number of people developing food allergies", 12/3/2012:

In a study of 2,211 American adults with the chemical in their urine, 411 were found to have a food allergy, while 1,016 had an environmental allergy.

The study in question is Elina Jerschow et al., "Dichlorophenol-containing pesticides and allergies: results from the US National Health and Nutrition Examination Survey 2005-2006", Annals of Allergy, Asthma and Immunology 109 (2012). If you glance at that paper, you'll quickly learn that 2,548 was the total number of people whose urinary dichlorophenol level was checked, NOT the number with "detectable levels of dichlorophenols in their urine"; and that 2,211 was the total number of people with both dichlorophenol measurements (zero or otherwise) and also other covariates needed for the study, NOT "the 2,211 people with high levels of dichlorophenols".

It's true that of these 2,211 people, 411 had food allergies and 1016 had environmental allergies (according to a blood test with a significant false-positive rate) — but these are the baseline rates for all the participants in the study, and tell us nothing whatever about the effects of dichlorophenols. The actual number of study participants with high levels of dichlorophenols was 225 with one high urine dichlorophenol (> 75th percentile level of any of the six isomers), and 431 with two high urine dichlorophenols. These did actually have somewhat elevated rates of food sensitivities:

How in the world did the people who wrote those articles get these simple facts so completely wrong? Well, for a start, they probably didn't read the published study, but rather just relied on the press release, "Dichlorophenol-Containing Pesticides Linked to Food Allergies, Study Finds; Chemical Also Used to Chlorinate Tap Water". And the press release says:

Among 10,348 participants in a US National Health and Nutrition Examination Survey 2005-2006, 2,548 had dichlorophenols measured in their urine and 2,211 were included into the study. Food allergy was found in 411 of these participants, while 1,016 had an environmental allergy.

Apparently the writers of the quoted articles interpreted "had dichlorophenols measured in their urine" to mean "had measurable levels of dichlorophenols found in their urine" (though that reading makes the sentence false), rather than "had the level of dichlorophenols in their urine checked" (a reading that makes the sentence true).

I should mention that the discussion section of the published study appeals to the "worms and germs are good for you" theory as a possible explanation for the association that they found:

Studies in children have found that healthy commensal flora influence mucosal immune tolerance. These flora are thought to be responsible for the transition from the TH2-predominant profile in early life to a more balanced phenotype in nonallergic individuals. Pentachlorophenol and its metabolites have a potent antibacterial effect. Contamination with pentachlorophenol metabolite 2,4-dichlorophenol has been found in produce, such as cocoa-powder and fruit juices. Therefore, drinking of chlorinated water and ingestion of crops treated with dichlorophenol-containing pesticides may alter the microbial spectrum to which humans are naturally exposed. […]

A daily exposure to dichlorophenols through water or food may have an ongoing effect on the bacterial diversity. There are indications that the intestinal microflora of nonallergic children differs from the microflora of allergic children. In addition, altered intestinal microbiota is associated with an increase of CD4-mediated inflammation in mice. Whether changes in intestinal microflora have a similar effect on allergic sensitization in adults is not known.


  1. amandachen said,

    December 16, 2012 @ 4:12 pm

    "produce, such as cocoa-powder and fruit juices" or "produce such as cocoa-powder and fruit juices"?

  2. Briefly | Stats Chat said,

    December 16, 2012 @ 5:07 pm

    […] Language Log examines the inability of journalists around the world to get the basic numbers right in reporting a study on water chlorination and allergies (a story that the NZ media seem to have had the good sense not to pick up). […]

  3. Denise Szymczak said,

    December 16, 2012 @ 5:31 pm

    FYI – "disability" does not mean "incompetent". This seems like the kind of blog where I shouldn't have to explain this.

  4. D.O. said,

    December 16, 2012 @ 8:41 pm

    How's that possible that there are more people with two isomers at high level than with one? Is there some funny correlation + multimodal distribution + some strange redefinition of percentiles?

    [(myl) Beats me. Maybe there's a biochemist who can help us out?]

  5. Ethan said,

    December 17, 2012 @ 12:49 am

    The categories are "exactly one DCP level is high" and "exactly two DCP levels are high". The first category is not a subset of the second.

    [(myl) Right, but it's odd (for instance) that there were apparently not enough people with high levels of three isomers to consider them a separate category.]

  6. Rohan F said,

    December 17, 2012 @ 2:00 am

    @Denise Szymczak, the very first definition for "disability" in the full Oxford English Dictionary is "[l]ack of ability (to discharge any office or function); inability, incapacity; weakness" and gives citations up to the year 2000 to prove it.

  7. Jeroen Mostert said,

    December 17, 2012 @ 2:47 am

    The press release formulation is horrendous. Of course people would interpret "Among 10,348 participants in a US National Health and Nutrition Examination Survey 2005-2006, 2,548 had dichlorophenols measured in their urine" as meaning "2,548 had measurable levels of dichlorophenols in their urine", because the alternate reading only makes sense if you'd think it was obvious that although 10,348 people participated, only 2,548 were actually measured (which is what happened, but the press release doesn't say why). Leaving out the salient fact of how many participants actually had increased levels of dichlorophenols means there is nothing to contradict the incorrect reading. It would be very hard for anyone to draw the proper conclusions unless they went to the trouble of checking the source material, but nobody has time for that these days, especially not journalists. News is news, after all — anyone who needs reliable news probably shouldn't count on journalists to get it for them beyond the dim outlines.

    In this case, I'd place the blame squarely on the press release writer's inability to write decent summaries. "Making sure journalists who will never read the source material don't grossly misinform the public" is not an unreasonable burden to place on the shoulders of press release writers. Or even "making sure journalists, who will never read the source material, …"

  8. J.W. Brewer said,

    December 17, 2012 @ 4:11 pm

    Humorously enough, the underlying research was supported in part by grant money from the "National Center for Advancing Translational Sciences." Whatever "translational" means in this context presumably doesn't have to do with translating research results into accurately-written ordinary-English press releases for journalistic consumption. Is the "correct" reading even possible? If a report on drug testing said that of 10,000 employees at such-and-such company, 2,500 had cannabinoids measured in their urine would anyone read that to mean that 7,500 employees were never tested in the first place and thus the 2,500 includes some (possibly quite significant number) who proved, upon testing, to have no measurable cannabinoids in their urine?

  9. Coby Lubliner said,

    December 17, 2012 @ 5:46 pm

    Flora as a plural (of *florum?) is new to me.

    [(myl) "Flora and fauna" for "plants and animals" is pretty common phrase in vaguely biological settings, and things like "intestinal flora" are also commonly encountered.]

  10. Ethan said,

    December 17, 2012 @ 7:28 pm

    "Translational" is a recent NIH buzzword that is supposed to indicate that the results from whatever it is attached to will be delivered from academia to private industry for commercialization. As with many such buzzwords, it gets attached to many things where this reading makes little sense.

    It is often the case that only some of the possible measurements are available for any given participant in a large study. This is particularly true when the study is large by virtue of being a combination of several smaller studies, which may not originally have been interested in the same set of scorable attributes. So having measured DCP levels for only 1/4 of the participants is believable, although of course it is not very satisfactory from the point of view of the specific claim in the press release.

    [(myl) According to the paper, the authors took uninary dichlorophenol levels for a randomly-selected subsample of the adults in the overall studied population. They don't say why they didn't test everyone, but presumably they felt that this subsample would be large enough, and wanted to avoid the unnecessary expense of more testing.]

    @Cory: "flora" is the usual collective term for multiple species of bacteria occupying a shared niche, as in the phrase "intestinal flora" that myl mentions. I don't know why "flora", since bacteria are not plants.

  11. Dan Hemmens said,

    December 18, 2012 @ 7:31 am

    In this case, I'd place the blame squarely on the press release writer's inability to write decent summaries.

    On the contrary, I'd place the blame squarely only the press release writer's ability to write summaries that do *exactly what they are supposed to do*.

    The press release as written is magnificently constructed. It manages to present a misleading, exaggerated picture of the study (which makes the study look far more significant than it actually was) without actually including any false information whatsoever.

  12. Coby Lubliner said,

    December 18, 2012 @ 1:19 pm

    Mark and Ethan: I know what flora and fauna mean. It's just that I (like every dictionary I have consulted) know them as singular nouns (whose plural is given as either floras/faunas or florae/faunae), unlike, say, bacteria. I just wonder if some undereducated scientists, when they see a noun ending in -a, assume it to be plural. I also know that some plural words ending in -a have become singular in English, e.g. agenda and media. This is the first time I have seen the reverse.

    Incidentally, the UC Berkeley Recreational Sports Facility, where I work out, recently got some new exercise machines, some of which are marked "bicep" and "tricep".

  13. Jerry Friedman said,

    December 18, 2012 @ 1:52 pm

    @Coby Lubliner: "Bicep" and "tricep" are universal in my limited experience of weight rooms and writing on fitness, and common in literary use.

    "Right biceps" is still more common in books than "right bicep" but not by much, according to this Google ngram result.

  14. Sven said,

    December 21, 2012 @ 12:00 pm

    Cory: It happened to the word pease, which became peas with the back-formed singular pea. But for me, too, this is the first time I have seen "these flora" and I wonder if it arose as an overcorrection by someone who has been told too many times that data is plural.

RSS feed for comments on this post