Attack of the killer frying pans
There’s a headline in the Herald: Heart disease linked to non-stick cookware: study. There seems to have been some loss in translation for both the article and the headline.
The journal press release says
Exposure to perfluorooctanoic acid (PFOA), a manmade chemical used in the manufacture of some common household products, appears to be associated with cardiovascular disease and peripheral arterial disease in a study of 1,216 individuals
That is, PFOA is used to make non-stick cookware, but cooking with non-stick cookware isn’t an especially important source of it, and the article doesn’t say it is.
Unlike the Teflon in non-stick coatings, which is as inert as a very inert thing, PFOA is quite chemically interesting. There are traces of PFOA in all sorts of things, and it accumulated in the environment, which is why it’s interesting to public health researchers. If you want to read more, Wikipedia is a bit alarmist about the health evidence but reasonably informative.
The story goes on to say
The study reviewed the levels of the chemical in 1216 people with heart problems.
which isn’t true. The study used a few years of data from the wonderful NHANES health surveys in the US, which is a random sample of the US population. A subset of the NHANES participants had PFOA levels measured in their blood, and that’s where the number 1216 comes from. Probably about 50 or so had heart disease, though as far as I can see the article doesn’t actually say. The researchers compared the PFOA levels in the people with and without heart disease, and then did the same thing for the related ‘peripheral arterial disease’.
Importantly, the heart disease was not measured by a doctor, participants were asked “Has a doctor ever told you that you had coronary heart disease”. This was at any time in the past, probably years before the PFOA was measured, and the “Has a doctor ever told you..” questions have a much higher error rate than you would expect.
The research is fine as far as it goes, and the researchers admit that what they need is a longitudinal study where PFOA is measured in healthy people who are followed up to see if they become sick. On a small scale this could be done with NHANES, since the data have been linked to Medicare records precisely to allow follow-up studies, though you have to go to a CDC data center to use the linked data set.
From a statistical point of view it’s strange that the researchers just used four years of NHANES data, from two non-adjacent two-year periods. The study keeps going year after year, and they are now measuring PFOA and related compounds on greater numbers of people. As a service to the StatsChat readership, I just spent 15 minutes downloading and analysing the 2007-2008 data, which has PFOA measurements on 2100 people, 80 of whom reported coronary heart disease. I didn’t do as thorough a job of ruling out other risk factors (such as cholesterol or high blood pressure), but it’s still interesting to note that there is absolutely no sign of an association between coronary heart disease and PFOA levels in the 2007-2008 data.
Thomas Lumley (@tslumley) is Professor of Biostatistics at the University of Auckland. His research interests include semiparametric models, survey sampling, statistical computing, foundations of statistics, and whatever methodological problems his medical collaborators come up with. He also blogs at Biased and Inefficient See all posts by Thomas Lumley »