Friday 11 April 2014

The missing plain packs data

When the Chantler report on plain packaging was published last week, one piece of evidence was conspicuous by its absence. At first I assumed that the empirical research on teen smoking rates from the University of Zurich had been published too late to be included, but I was wrong. It transpires that Chantler's team not only had access to the study, but had spoken personally to its authors, Dr Ashok Kaul and Dr Michael Wolf.

Chantler had every reason to be interested in this research. It is the only study to date that addresses the question upon which all else hinges - does plain packaging help reduce the smoking rate amongst minors?

Kaul and Wolf had access to monthly smoking figures amongst 14-17 year olds that went all the way from 2001 to the end of December 2013, thirteen months after plain packaging was introduced in Australia. The data come from a large survey that has been used by anti-smoking researchers in the past and both authors are experienced professional statisticians. They concluded that there is no evidence from the data that plain packaging had any effect on smoking prevalence amongst this crucial age group. There was no increase in the gradual long term decline of smoking in this age group...

And when the long term decline was accounted for, plain packaging was shown to have had no discernible impact whatsoever...

This is very significant information, so why was it ignored by Chantler? It cannot be because it hasn't yet been peer-reviewed because Chantler looked at plenty of evidence that hasn't been peer-reviewed. It cannot be because it was commissioned by Philip Morris because Chantler looked at plenty of industry-funded research. In any case, if Chantler had any doubts about the research he could have looked at the raw data himself.

And we know that Chantler was aware of the research. We know this because Kaul and Wolf have now put online a 27 page transcription of the meeting he had with Chantler's team (the man himself didn't attend). It leaves no doubt that the study was carefully and patiently explained to Department of Health officials when they met the authors in London on 20 March.

At this meeting, Dr Kaul described that the methodology they used allowed maximum leeway for finding some effect from plain packaging. Alas, there was none...

"Even if you are not talking about statistical significance, I mean, sort of hoping for an effect, even a small one, you would at least expect something, a little, tiny effect, 12 months after in comparison to 12 months before but we don't find that, we find the contrary."

At the end of an extensive conversation, DoH official Christoper Cox thanked the pair for travelling to London and said of the presentation: "that is incredibly helpful and thoroughly interesting". So why was there not a trace of it in the final report? Why was the only empirical, real world evidence about underage smoking rates after plain packaging excluded?

Readers can draw their own conclusion.

(You can read the transcript here (PDF))


gh500 said...

We have reviewed the data presented in Ashok Kaul and Michael Wolf's paper4 and conclude that in view of the short time span since the measure was introduced, the variability in the measure, and the small sample size, this is neither an unexpected nor a meaningful conclusion.

Christopher Snowdon said...

"short time span since the measure was introduced" yes, it's about whether there has been a short term effect.

"the variability in the measure" there is plenty of variability but also a clear trend, see graph above

"the small sample size" actually quite a large sample size; large enough to show the trend both before and after plain packs came in and much larger than most pro-plain pack studies

gh500 said...

"short time span since the measure was introduced" yes, it's about whether there has been a short term effect.

The short time span doesn't allow for extrapolation regarding even short term effect when there's variability.

,"the small sample size" actually quite a large sample size; large enough to show the trend both before and after plain packs came in and much larger than most pro-plain pack studies
Who are you fooling about a large sample size wit a variability between 3% and 13%!

Christopher Snowdon said...

" Its sample size is twice the size of the two major official data sets, namely about 50.000 interviews per year. (NDSHS and AHS rely on about 25,000 interviews per year and are both cross-sectional data sets themselves.) It is available on a monthly basis, allowing researchers to trace month-to-month changes rather than developments in increments of several years. Also, the data have already been used to analyze smoking behavior in Australia by renowned tobacco control researchers [see e.g. Siahpush, M., Wakefield, M. A., Spittal, M. J., Durkin, S. J., & Scollo, M. M. 2009. “Taxation reduces social disparities in adult smoking prevalence.” American journal of preventive medicine, 36(4), 285–

BrianB said...

I found the transcript to be very enlightening.

Here you have a couple of professional guys (Kaul & Wolf) who admit they have tried to be 'helpful' "We did the best to find an effect but we couldn't", and including their use of 90% confidence intervals "so we have chosen the confidence level to be 90 percent and that's the lowest confidence level anybody would ever apply in empirical work. More typical is 95 percent and 95 percent would make the intervals wider. So we have chosen the smallest, reasonable
confidence level 90 percent"

Despite this, they cannot find an effect, and have the moral and scientific integrity to stick by their conclusions, even though there are a number of points where the 'Chantler' bods (Collis & Cox) are clearly trying to lead them towards a desired conclusion.

I would also add that the statistical analysis, simple though it is (and admitted so by Kaul & Wolf) went over the heads of Collis & Cox. There were odd signs that the latter are familiar with the statistical tricks played by the likes of Gilmore et al, eg "One or two of the ideas that spring to mind is you could look at just putting in dummy variables for the change", yet they didn't seem even to have a basic grasp of the role of confidence intervals.

All in all this discussion should have quite clearly led to the packaging issue being, at least kicked into the long grass for a few years, if not abandoned altogether. The fact that the only "compelling case" (Ellison, J.), is that Cyril C's opinions is that it is worth invoking the Precautionary Principle - although it ought to be argued that the PP should be invoked in the other (ie do nothing) direction, at least from a commercial and economic perspective.

What this ultimately boils down to is that the mere opinion of one consultant, albeit a known Tory bag man, is sufficiently compelling to introduce draconian legislation, yet the combined weight of opinion of 426,000 UK citizens is worth less that this one man's.

That is nothing less than corruption. It is a Government effcitively declaring war on its people, and it is time we took the war to them!

To finish on a quieter note: Kaul & Wolf clearly know a lot more about tobacco usage trends than the bastard tobacco 'control' industry, so one paragraph stood out as potentially very important for further research. It was this:

"To put it differently, we see this line in all the OECD countries, it looks very similar across all countries and in some countries we had heavy anti-smoking measures, in other countries we didn't, but we see essentially the same line in all countries. Even if you look at the quantitative data, you find a minus 0.4 percentage point effect per year. That is also very similar, had the same effect in Germany, I think in the UK must be very similar as well. I don't remember the exact numbers but I had a look at all the OECD countries and that looks very similar across the countries over the last, let's say, one and a half decades."

What this says to me is that, from their review of published statistics, anti-smoking measures in the 'developed' world have had no effect for the past 15 years.

We need to find out more, and build upon this if we are going to be in a position to destroy self-annointed tobacco 'control'.

Michael J. McFadden said...

Truly amazing. Not even significant at the 90% level and yet the Antis *STILL* try to claim there's some sort of evidence?

They simply make this stuff up, don't they?

(As if we didn't already know that, right?)


nisakiman said...

The short time span doesn't allow for extrapolation regarding even short term effect when there's variability.

But you can bet your bottom dollar that if the results had shown a trend towards lower take-up rates, the 'short time span' would be found to be quite acceptable and the report would have been included in the final summation. Indeed, it would have formed the central plank of a declaration that PP is wildly successful in Aus, and would have been cited as incontrovertible evidence that PP works. Such is the calumny and deceit of Tobacco Control.

Unknown said...

Reposted this article to my Facebook timeline with thanks.

nisakiman said...

Black market tobacco 'booming' in Australia: KPMG study

Of course, Tobacco Control are saying that because the KPMG research was funded by 'Big Tobacco' then it is to be dismissed out of hand.

"But anti-smoking campaigners, including the Cancer Council of Victoria, have slammed the report as ''self-serving rubbish''."

"The release of the second tobacco industry-commissioned report into illicit tobacco in less than six months smacks of desperation."

I would say their reaction smacks of desperation, myself.