Glass onion

After yesterday's discussion of the "secondhand smoke in a car is x times more toxic than in the home", I've been looking again at the origins of this claim.

My first port of call was the ASH website, which has changed its fact sheet on the subject since I last visited in March. It's at the same URL because ASH have a slightly Orwellian tendency to delete old information as if it had never existed. Nevertheless, the old page said:

According to a report by the Ontario Medical Association, secondhand smoke levels in cars can be 23 times greater than in a house.

As I mentioned yesterday, the Ontario Medical Association's figure has been exposed as bogus. On their new page, ASH say:

In 2005, The State of California’s Air Resources Board (CARB) conducted a comprehensive review of studies which measured secondhand smoke particle concentrations in a variety of environments. The review found that in-car concentrations were up to 60 times greater than in a smoke-free home, and up to 27 times greater than in a smoker’s home.

And this time we have a reference: Proposed Identification of Environmental Tobacco Smoke as a Toxic Air Contaminant, a document prepared by the California Environmental Protection Agency (Cal-EPA). It is indeed thorough, and although it never uses the "27 times" figure, it does rely on a study that came close to doing so.

The study (Offermann et al., 2002) is one I've seen referenced before but had never found. Today, after much searching, I did. It's obscure because it seems never to have been published or peer-reviewed, but it was presented at the 9th International Conference on Indoor Air Quality and Climate (book early to avoid disappointment!) in Monterey in 2002. And in it, Offermann makes a nearly-familiar claim:

The calculated exposure for a five hour automobile trip with the windows closed/ventilation off and with a smoking rate of 2 cigarettes per hour is 25 times higher than the same exposure scenario in a residence.

Given the lack of ventilation and the small size of the smoking environment, this is certainly plausible. Offermann measured respirable suspended particles (RSP) in a moving car under three different conditions and got the following readings:

(1) Windows open/ventilation off: 92 µg/m3

(2) Windows closed/ventilation on: 693 µg/m3

(3) Windows closed/ventilation off: 1,195 µg/m3

(In experiments of this kind, the idea of drivers having the windows open and the ventilation on is apparently considered too far-fetched to be worth studying.)

It is worth noting that although Scenario 1 says 'windows', it is clear from the text that only the driver's window was open. The driver also held the cigarette in his right hand despite being in a left-hand drive vehicle. And RSP levels fell to nearly zero within 60 seconds of the cigarette being extinguished.

In any case, it is not Scenario 1 that Offermann focuses on; he prefers to dwell on the much more unlikely Scenario 3. This is what he bases his "25 times higher" figure on.

Offermann's unpublished study forms the basis of Cal-EPA's own figures. Offermann didn't measure RSP in the home, but Cal-EPA found some other studies to piece the picture together. Their summary is shown below (click to enlarge):

As you can see from the top line, average RSP concentrations in smokers' homes were found to be between 44-125.6µg/m3. Average RSP levels in nonsmokers' homes were between 20-87.8µg/m3. The reason these two sets of figures are so close together is that a smoker's home is not necessarily a home with a smoker smoking (and there are other sources of RSP other than cigarettes). Studies in rooms with people actually smoking found much higher RSP concentration of between 160-5,500µg/m3—much higher than anything recorded in Offermann's car (actually it was a minivan, but let's not get bogged down in the detail.)

As I said, Cal-EPA never use any of this to say that smoking in a car is x times worse than smoking in a home or a pub. But let's speculate for a moment and say that someone (from ASH?) looked at these figures, picked the lowest home reading (44µg/m3), compared it with the highest vehicle reading (1,195µg/m3) and did a quick calculation. 1,195 divided by 44 = 27. Bingo!

And let's say they picked the lowest nonsmoking home figure (20µg/m3) and did the same thing. 1,195 divided by 20 = 60. And so, once we unravel all the layers, I think we may have found the original source of what was written in The Guardian yesterday:

Second-hand smoke can be 27 times more toxic in a car than a smoker's home, it says in a report published today.

And, to be fair to the UK Faculty of Public Health, they did mention that this was a completely unventilated car:

In a closed car, levels of second-hand smoke can be extremely high – the concentration in cars can be up to 60 times higher than in a smoke-free home, and up to 27 times greater than in a smoker’s home.

It's cherry-picking, of course, and the words "up to" should always ring alarm bells when statistics are involved. Such calculations make no attempt to find the average readings under average (real life) conditions. We could use the same set of figures to make the claim that there is nearly as little secondhand smoke in a smoky car (92µg/m3) as in a completely smoke-free house (87.8µg/m3) and, even then, only while the cigarette is burning. After that, a smoky car is less smoky than a smokefree house! 

That would be a dodgy use of statistics as well, and—lest we forget—it all hinges on an unpublished study which neither ASH nor the Faculty of Public Health appear to have read (ASH only reference the Cal-EPA report, the Faculty didn't even do that.)

Nevertheless, I owe ASH and the Faculty of Public Health an apology. I accused them of falling for someone else's bogus figure. In fact, they came with their own.

The frustrating thing is that even after all this we are no closer to getting to the source of the "23 times" claim that conquered the world on the back of a brief report in Rocky Mountain News. Offermann's study came out in 2002 and the Cal-EPA report is from 2005. But Rocky Mountain News published its story back in 1998 and the trail has long-since gone cold. Some mysteries are destined never to be solved.

UPDATE: This is a comment from Carl V. Philips, e-mailed to me after he was unable to post in the comments section:

[This is an attempt to recreate a comment I wrote but somehow managed to lose rather than submit. I am not going to redo the calculations, so I will try to do it from memory, and also try to write my long comments offline from now on!]

First, I want to say that this is great research work, Chris. Thinking of the silly attempt to hush you up re: The Spirit Level Delusion, I have to say that if half the people publishing in health science journals were even half the scientist that you are, the field would be improved enormously.

As for the ventilation in cars study itself, if you are taking this further I think there must be some problems with the numbers. Such problems do not compare to the extremists lying about what the numbers mean, of course, but it is a separate problem.

I know this is not my field, but the statistics should be pretty straightforward, and they do not seem to add up. Comparing tests 1 and 2, the air exchange rate is very similar. The half-life for a bit of air (and thus a bit of smoke) is somewhere between half and 3/4 of a minute for both of them (assuming perfect mixing). That is, a particular smoke molecule has a 50% chance of being gone in that much time. It does not seem plausible that there would be such a difference in concentration then.

But the more important one is the difference between test 3 and 2. If 3 is really right, and it is the one that the antis want to say is right, then 2 and 1 do not seem plausible. The air exchange rate in 3 translates into a half life of about eight minutes. If someone smokes a cigarette over the course of four or five minutes then the vast majority of the smoke is still in the car at the end of the period.

This compares to the other scenarios when, by the time of the last puff, almost all the smoke from the first few puffs is gone and even half the smoke from the previous puff is gone. It does not make any sense, then, that the concentration is only reduced by half in 2 compared to 3. This suggests that the numbers for a ventilated car – i.e., the worst realistic-case scenario, since as you point out, no one turns the ventilation off – are exaggerated.

An interesting comparison is that the air exchange rate (according to what I looked up) for a room in a house is in the order of 10/hour (a pub is in the range of twice that). So, though the space is much larger, the smoke lingers a lot longer. Thus, it is not entirely clear to me that the exposure in a car is actually worse. Since in a realistic car situation, the smoker who is not trying to annoy others in the car at least cracks the window, holds the cigarette near there, and exhales toward the opening (harm reduction!), typical exposure will be even lower than the best case of the tests run. It may be that momentarily the concentration is higher, but drops lower than a room fairly rapidly. The question would be “is it worse to have a higher peak exposure or lower exposure for longer” – an interesting question, but one we will never know the answer to because the millions of dollars spend studying these matters are never directed at anything practical. (Actually, we might know. There is physiologic research on effects of smoke being done by industry, even though the “health” people are not interested in gaining useful knowledge about reducing effects, and so someone might find something relevant.)

Anyway, the bottom line is that the actual difference between smoking in a car with someone or a room of a house is not actually large and it is not even clear which is a greater exposure under realistic conditions.