Tuesday 23 July 2013

That plain packaging study

A few days ago, anti-smoking sociologist Simon Chapman sent out some tweets suggesting that there was compelling evidence for plain packaging on its way and he called on the British government to specify what kind of evidence would satisfy them.

I don't claim to speak for the government but I expect that the least they want to see is: (1) a sharp decline in smoking prevalence, particularly underage smoking prevalence; (2) no increase in illicit cigarette sale and production; (3) a successful and inexpensive settlement of the various intellectual property disputes/lawsuits.

The study Chapman referred to was published yesterday. It was a phone survey conducted by one of his mates and it didn't remotely address any of these issues. Its main finding—reflected in headlines such as Plain cigarette packs 'encourage smokers to quit'—was that people who smoked out of plain packs were more inclined to think about giving up smoking. According to the survey, 57.1% of those smoking from branded packs were "seriously considering quitting in the next 6 months", whereas this rose to 68.8% for those smoking from plain packs.

These numbers are pretty feeble. The British government is contemplating introducing plain packs on the basis that it might deter underage nonsmokers from taking up the habit. At best, this new study suggests plain packaging might make a handful of adult smokers say that they are thinking about giving up smoking. It is doubtful whether any government will cross the Rubicon of plain packaging on the basis that it might make an extra ten per cent of adult consumers express a vague aspiration.

The researchers did not bother to chase up these people six months later to see if they actually did quit. As with all previous policy-based evidence on plain packs, stated preferences are assumed to align with revealed preferences. They don't.

It wouldn't be surprising if there was a spike in smokers expressing an interest in quitting, although this is more likely to be due to the new warnings than the elimination of branding (the authors acknowledge that they can't distinguish the effect of the warnings from the effect of the plain pack). It is often the case that new warning labels catch smokers' eyes, but it is also well known that smokers soon get used to them and quickly ignore them. Indeed, there is evidence that the more graphic the warning, the less effect it has. That may be why graphic warnings "have not had a discernible impact on smoking prevalence".

I suspect that the government is more interested in whether people actually quit (or never start) than in whether Australia's vastly expensive experiment briefly makes a few smokers express an aspiration. In fact, the study's findings are even flimsier than they first appear for the following reason.

Unlike previous studies on the topic, this research has the benefit of surveying people in a real life situation. It was conducted in November and early December of last year when the supply of branded packs was dwindling or non-existent (it was illegal to sell branded packs from December 1st). As a consequence, some people were smoking from branded packs and others (the majority) were smoking from plain packs.

I was reminded of an e-mail I received from an Australian reader in late October. He sent me some photos of the new plain packs—I published some of them in this post—including a picture of the large collection he had stockpiled for his own consumption (below).

I'm sure my correspondent was not alone in stocking up on conventional packs in the weeks before the ban came in. It should be quite obvious that anyone who buys a six month supply of cigarettes is not intending to give up smoking and is unlikely to express such a desire in a survey.

The point is that many of the people who were smoking from branded packs in November and December were inherently more committed to smoking than those who weren't. Getting hold of branded packs was increasingly difficult in the last few weeks before the ban (which is why the majority were on plain packs by then). Some effort was required to obtain them. Those who were smoking from branded packs when they were interviewed in early December can only have been doing so because they had a stockpile.

This is no mere hunch. The difference in character between the plain and branded pack purchasers is shown in Table 2 of the study. Of those who had never tried to quit, only 54% were buying plain packs, but of those who had tried to quit at least once in their lifetime, 76% were buying plain packs. This is actually the strongest result of the lot (with an odds ratio of 2.61) and it clearly suggests reverse causation, ie. it isn't that branded packs made people less likely to express an interest in quitting, but that people who were less interested in quitting were seeking out branded packs. The authors acknowledge this point in the text and when they adjust their results to control for this confounding factor, most of the associations they report disappear (ie. they fail to achieve statistical significance).

Ultimately, we are interested in what people do, not what they say. Stated preferences are never more meaningless than in the field of smoking where around 97% unassisted quit attempts end in failure and there is strong social pressure to express anti-smoking sentiments. The weak findings of this study do not merit the news coverage they have received, nor the spin put on them by campaigners, as Bernard Keane notes at Crikey...

[Health Minister Tanya] Plibersek’s media release statement that the study shows plain packaging is “working to put people off smoking” appeared at least a little disingenuous, though it depends on what you mean by “putting people off”. By the time some in the media had finished with it, we had “packs helping smokers kick the habit” (the ABC) and “plain cigarette packaging works: study” (the supposedly more rigorous Conversation). Perhaps the reports might have better if journalists had the actual report to work from, rather than a press release from one side of the debate?   

The idea that this phone survey might be instructive, let alone decisive, is absurd. Incredibly, the author of this study was so keen to influence UK policy that she flew from Australia to England in April to share the findings of this study with the government. No complaints from the public health lobby about Australians lobbying politicians on that occasion! In the end, she failed to meet any MPs and had to settle for briefing tobacco controllers at the Department of Health, but it is interesting to note how blurred the distinction between researcher and campaigner is in tobacco control.

When the government says it wants evidence, it means it wants to see what happened. How difficult can this be for people to understand? A natural experiment is underway in Australia which the rest of the world can study. Let's see what the smoking prevalence and cigarette sales figures show. We don't have long to wait. The desperate rush of anti-smoking campaigners to 'prove' the policy works before the real data surfaces suggests that they know it hasn't.


Vova said...

I tried to leave a comment on the BMJ tobacco control blog - pointing out that the study contained no data on health outcomes nor smoking initiation (which plain packs were supposed to be about) yet they still headlined their piece 'Australian study shows cigarette plain packaging works'. Then a comment about EBM and 'public health' 'advocacy'.
Stone me - it didn't get past moderation.

PJH said...

One thing is puzzling me, and it's in regard to this:

`According to the survey, 57.1% of those smoking from branded packs were "seriously considering quitting in the next 6 months",`

Where on earth did they find a cohort where 57% of the subjects 'seriously'[1] wanted to quit to begin with? That seems abnormally high to me.

Anyone any idea of the 'normal' percentage of smokers who 'seriously'[1] want to quit and if it bears any resemblance to the subjects selected for this piece of research?

[1] As opposed to including the 'I supposed I would want to quit if I could be bothered' which seem to appear in the high 60-70% the CDC in the States seem to quote.

Unknown said...

Somebody didn't like this )


Vova said...

This reminds me of the panic to rush out Stockwell's analysis of Canadian (not) minimum pricing (absence of) effect on alcohol problems earlier this year. Here we have a study of what is basically irrelevant information masquerading as showing plain packaging 'works'. It shows what suckers Govt and politicians are for confirmation bias. If people like Stockwell and Chapman really believe these studies mean anything like they say they do then they are suckers too. On the other hand, if they realise the work is basically irrelevant crap and use it to play to Govt stupidity they pose a serious problem for the rest of us, in that they pull the policy strings without ever being submitted to public scrutiny/approval.

CricketIsLikeFishMan said...

"The desperate rush of anti-smoking campaigners to 'prove' the policy works before the real data surfaces suggests that they know it hasn't"

One could as easily say that the desperate rush of tobacco companies to prevent these initiatives suggests that they know they would work.

Christopher Snowdon said...

The reason why tobacco companies strongly oppose the idea are obvious, but not in the way you think. Even Simon Chapman once accidentally told the truth about this...


Any industry would oppose having their intellectual property destroyed and I would be right behind them.