When reading articles like this, I often find it useful to at least skim the paper that the article discusses. The article's opening line seems designed to dem

When reading articles like this, I often find it useful to at least skim the paper that the article discusses. The article's opening line seems designed to demonize: "Diacetyl, a flavoring chemical linked to cases of severe respiratory disease, was found in more than 75% of flavored electronic cigarettes and refill liquids tested by researchers at Harvard T.H. Chan School of Public Health." 75%! Tested by Harvard! This is huge! However, take a look at the actual study. Specifically, page 6, under Methods, e-Cigarette Selection http://ehp.niehs.nih.gov/wp-content/uploads/advpub/2015/12/ehp.1510185.acco.pdf http://ehp.niehs.nih.gov/wp-content/uploads/advpub/2015/12/e... "A convenience sample of 51 e-cigarette flavors was selected for use in this study. Electronic cigarette cartridges, liquids, and their associated devices and batteries were purchased online and in retail locations. We evaluated 51 flavors, including all available flavors from three large cigarette companies (Brands A, B, and C, with 2, 2, and 7 flavors, respectively); 5 flavors from a large independent e-cigarette company (Brand D); and 24 additional flavors from three ecigarette distributors (Brands E, F, and G, 10, 8, and 6 flavors, respectively) that we selected based on their potential appeal to children, teenagers and young adults (Table 2). In addition, we evaluated 11 e-liquid flavors that are inserted into a cartomizer (disposable cartridge and atomizer system) (Brands H and I, 6 and 5 flavors, respectively)" To break this down a bit.. Let's start with convenience sample. A convenience sample is a non-probability sampling technique. It's made up of subjects (in this case, nicotine juice) which are easy to reach. It is _not_ a representative sample of the products offered by the entire industry. You can't make probability statements about non-probability samples. The numbers are worthless in relation to the greater whole. EG, you can't say "Diacetyl was found in more than 75% of flavored electronic cigarettes." It would be like going to a Donald Trump campaign rally to poll public opinion because it's close to your house and then reporting that "75% of Americans believe Donald Trump should be the next President of the United States". Add a couple of words, however, and although your numbers are still worthless, you can use a technicality to slip past whatever fact-checking apparatus is in place that would call you out on a bold-faced lie and you can lend some additional credibility to your prejudicial pseudo-science. "Diacetyl was found in more than 75% of flavored electronic cigarettes tested by researchers at Harvard." The average person is not going to consider the added qualifying statement in the proper context. Technically, it lets the reporter off the hook for generalizing because they have now (importantly) stated that the number is only in relation to what was tested. Effectively, though, it lends a big name institution and all that it represents in support of the reporter's agenda. Most people will read that sentence as "Harvard reports that 75% of flavored electronic cigarettes contain diacetyl, which causes severe respiratory disease'. It does help the article's credibility that this 'research' was indeed performed at Harvard, was under grant from the NIH, and that the probabilistically irrelevant results were posted on a Harvard website in a fairly authoritative manner as a press-release. The writer of the article isn't credited, so this appears to come from the university itself. But anyway, back to the numbers they listed. Their sample size, of the many thousands available, was 51. And it came from... 3 cigarette companies, 1 e-cigarette company, and 3 e-cigarette distributors. They use only seven (convenient, non-representative) sources to paint the industry as a whole. If we ignore the blanketing insinuation of the industry as a whole, though, the number they came up with is still of very limited utility. One company contributed 2 datapoints, another contributed 10. Biased contributions make the overall '75%' figure irrelevant in describing even just those 7 companies comprising their 'research'. If a company uses diacetyl in one of their products, odds are very high that they will use it in their other products. That aside, since the selection of flavors tested are not representative of sales of individual products, there's no point in repeatedly testing the products of a single company once a positive result has been reached. I'm very disappointed and irritated that an institution like Harvard would lend its credibility to a piece like this. It leads me to wonder whether they are acting with an ulterior motivation or simply incompetently. In the end, their intent means little, though; their words will have impact to discourage the spread of e-cigarettes as an alternative or cessation aid for traditional cigarettes. Is that a good thing? I don't know. We need better research into this youthful industry that has a significant potential health impact. We rely on valid research by trustworthy institutions to inform our decision making process. There's an opportunity cost to choosing to perform the wrong research and to performing the right research incorrectly. While some good does come of this, in that specific companies are outed (sort of -- the paper does not list the company names) for their use of dangerous ingredients, the research does not address safety concerns around e-cigarette use in the scenario where a responsible vendor takes precautions by ensuring that known dangerous additives aren't added to their products. I know it's not very sexy to report, and might require some additional effort, but why not look at the impact of the additives that _all_ of the e-cigarette companies are using because they are assumed benign in the absence of evidence to the contrary?