Wednesday 4 June 2014

Simple Simon says

Simon Chapman's latest confrontation with reality has attracted the attention of Redheadfullofsteam who has written a nice takedown, as ever. Simple Simon is continuing his war against e-cigarettes (which has already become a "hey, we still don't know so let's keep them banned in Australia" rearguard action against e-cigarettes) by resorting to numerology.

It would be genuinely interesting to know whether this guy is as stupid as he seems or if he merely assumes that the people he needs to persuade are as stupid as he hopes. His line of attack on Robert West's recent study in Addiction, which showed that smokers were around 50% more likely to quit using e-cigarettes than if they went cold turkey, begins by accurately describing the research:

It found 93 out of the 464 people who used e-cigarettes were successful (20%), while 194 out of the 1,922 people using nicotine replacement therapy made it (10.1%), and 535 out of 3,477 of the people trying to quit unassisted did so (15.4%).

Almost incredibly, he concludes with the following words:

...headlines about the English study should probably have said there were nearly six times as many smokers who quit without any assistance than vapers who did.

As used to Chapman's cretinism as I am, this still leaves me gobsmacked. Let me explain in words that even he will understand.

This study involved a self-selecting sample. Sorry, have I lost him already?

Let's try again.

This study looked at people who had chosen their own way of quitting. Nearly 3,500 smokers attempted to quit cold turkey. Only 464 tried e-cigarettes. What matters is what proportion of each group successfully quit. A significantly higher proportion of the e-cigarette group successfully quit. By looking at the percentages in each group, we can see which method was most successful.

The fact that there were more quitters—in absolute numbers—in the cold turkey group is a reflection of the fact that there were many more people in it to start with. That isn't the point. The number of people who attempt different methods can be changed over time and policy can help do that. The obvious conclusion from this study—to anyone who has eyes to see—is that if the number of people who chose to quit smoking by using e-cigarettes was as large as the group that currently uses the less efficacious method of cold turkey, there would be many more people successfully quitting. Growing that number requires encouraging people to use e-cigarettes rather than banning them, disparaging them and over-regulating them. In other words, it requires doing the opposite to what puritanical statists like Chapman instinctively want to do.

I hate to patronise you, dear reader, with this explanation, but it seems that there is at least one person on the planet who cannot tell the difference between relative numbers and absolute numbers.

Except that we know that Chapman does understand relative numbers. We know this because he has spent most of his long life championing the dubious epidemiology of passive smoking and the relative risks it reports. He knows better than anyone that if absolute risks were accurately reported in countries with relatively small populations then "journalists will be hardpressed to write anything other than 'Official: Passive smoking cleared - no lung cancer".

Nevertheless, the great sociologist scoffs at the e-cigarette study by focusing on the absolute numbers:

Let’s look at these numbers another way. In this large study, 80% of smokers trying to quit by vaping were still smoking compared with 84.6% of those [who] tried to quit on their own. That hardly looks like a champagne-popping difference deserving the accolades abounding in narratives about vaping.

What he's saying here is that most of the smokers failed to quit. 80%, 84.6%, what's the difference?

Consider using that logic with Chapman's beloved passive smoking theory. Here, if you wink one eye, ignore statistical significance and exclude the studies you don't like, you find a lung cancer risk from a lifetime's secondhand smoke exposure of about 1.20, which is to say 20%.

What's a nonsmokers' lifetime risk of getting lung cancer? It's about 1%.

What does an increased relative risk of 1.20 from passive smoking look like? It's about 1.2%.

So, to put it in Chapman's terms, 99% of the unexposed group don't get lung cancer and 98.8% of the exposed group don't get lung cancer. This small difference of 0.2% "hardly looks like a champagne-popping difference deserving the accolades abounding in narratives about" smoking bans, does it?

Same logic, same rules, but now the crank finds himself having to attack observational studies, rather than glorify them, and so the interpretation changes.

12 comments:

Anonymous said...

And this is what I dislike about these surveys. They only offer a quit or not-quit choice.

The reality of the vaping experience is that a vast number of vapers still smoke a bit, but much less than they did previously. For example, I now only smoke a few cigarettes each week, mainly with a beer. However, I would be classified as not-quit.

That's an entirely inappropriate classification of my transformation from regular, daily smoker to social smoker.

Anonymous said...

Hello Mr Snowdon

I know from the Catch debates on Frank's blog a few years back that you are a believer in the 'smoking causes lung cancer' theory. A theory which was born in the 50s on statistics showing a larger number of lung cancer among smokers compared to nonsmokers. If we look at the official statistics now, we can see that the decline in smoking prevalence parallels the decline in LC incidence for men, but not for women. LC incidence for women has been on the rise since the 70s and is now almost double while the smoking prevalence among women has almost halved.
http://www.cancerresearchuk.org/cancer-info/cancerstats/types/lung/incidence/uk-lung-cancer-incidence-statistics fig 1.2 and 1.3
for more details on smoking prevalence http://www.hscic.gov.uk/catalogue/PUB11454
what are your thoughts on this ? thank you

Jonathan Bagley said...

Anonymous, there is a time lag of around 30 years between smoking prevalence and lung cancer incidence and women started smoking in large numbers after men. For me, the most compelling evidence that smoking increases the risk of lung cancer comes from Sweden, where snus use among men has been popular for several decades. Consequently Sweden has the lowest male smoking prevalence in the developed world. It also has the lowest lung cancer incidence - around half that of other European countries. Swedish women do not tend to use snus as it is perceived as unfeminine and, I was told, turns your teeth black. Consequently, female smoking prevalence and lung cancer incidence are more typical of other similar countries.

Anonymous said...

Mr Bagley, that is what the official propaganda would say, but their own data contradicts them. If you look at the second link, you can see that in 1948 65% of men and 41% of women were smokers. In 1980 42% men and 36% women were smokers. So there isn't any 30yrs period during which the smoking prevalence of men went down while the prevalence for women doubled. Furthermore, how is this 30yrs time lag supposed to work ? If someone starts smoking around 15-20yrs of age, it's unlikely they would have LC by 45-50yrs as LC average onset age is 72. From 1980 the downward trend for both is very pronounced, so that in 2010 (last year of data) 20% men and 19% women were smokers. How can you explain that although the prevalence of smoking among women has been in decline since 1948 (and a steep one since 1980) the LC incidence has doubled ? (22.6 cases per 100k in 1975 vs 40.9 per 100k in 2011.) These confusing trends are true for USA as well. I don't have an explanation, I just observe that smoking doesn't seem to have anything to do with LC.

Jonathan Bagley said...

Anonymous. Your CRUK fig 1.3 illustrates my point, though maybe the time lag is longer than the thirty years I stated. Female smoking appears to have peeked around 1970, yet male smoking had then been falling for at least 20 years. I can't find the earlier figures you quote for women. If women's smoking has been in decline since 1948, you have a point, but all the graphs I've seen back up the lag theory, both for men and women. The discussion about cohorts beginning below fig 1.4 is informative.

Jonathan Bagley said...

This small difference of 0.2% "hardly looks like a champagne-popping difference deserving the accolades abounding in narratives about" smoking bans, does it?

Yes, to explain it, I suggest imagining sitting in smoke-filled bars for 4 hours a day for fifty years and then, at the age of 71, putting your hand into a box of 499 white balls and 1 black ball. If you pick the black ball, you get lung cancer. And that's with believing the 1.2 RR claim, which is almost certainly false.

Fredrik Eich said...

Jonathan,
Sweden is one of the reasons I dispute the 30-20 years time lag for smoking to cause lung cancer (and If I remember correctly Wynder thought it would have to be a 60 year time lag!). If you compare ciggie use in Sweden and the USA , it is clear that Swedens cigs cause lung cancer in the year of purchase , where as US cigs take about twenty years. I can believe that ciggies cause lung cancer straight away and I can believe that it takes decades but I don't believe both theories can be true simultaneously, unless they act synergistically with some other cause between 1950 and 1990 when both the US and Sweden had their (Male) epidemics. IN which case cigarettes are NOT a sufficient cause of lung cancer.

And that is not even talking about why Eastern Block countries saw their lung cancer epidemics drop in the 1990's just like almost every other country in the world despite continuing with high smoking prevalence. When I get the data I hope to show that Russian cigarettes cause lung cancer 20 years before the year of purchase , unlike Swedish ones which cause lung cancer in the year of purchase or American ones which cause lung cancer twenty years after purchase.

Jonathan Bagley said...

What you say is at first sight puzzling, Fredrik. Have you looked at countries where there was little or no smoking until recently? There must be several where very few women smoked. Did their female lung cancer rates increase during the fallout period?

nisakiman said...

Thailand is a country which traditionally has very low prevalence of smoking in females - it's a cultural thing - and I believe that male smoking prevalence is about 40% as opposed to 3% for females.

Lung cancer is defined as one of the major health problems in Thailand and has been the most common cause of death since 1999 [4]. Lung cancer in Thailand is the second most common cancer in males after liver cancer and the fourth in females after cervix, breast and liver cancers. There is a higher incidence rate of lung cancer in northern Thailand than other areas [5]. In 2005, there were 535 new cases of lung cancer diagnosed in 2005 (326 males, 209 females) (Figure 2 and ​and3)3) in Northern Thailand. This constituted 25.6% of all cancers in males and 14.9% of those in females.

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3097734/

Those LC figures don't really seem to bear much relationship to the M / F smoking rates.

Fredrik Eich said...

Jonathan,
There is a lot of volatility for female LC rates across the world. But here they are (Age adjusted) for US and Sweden and I have added Thailand as nisakiman bought it up.

Anonymous said...

here are the statistics for laryngeal cancer http://www.cancerresearchuk.org/cancer-info/cancerstats/types/larynx/incidence/uk-laryngeal-cancer-incidence-statistics
the numbers for both men and women are the same now as they were in 1975...i wonder why is that smoking which is supposed to to so much damage to the lungs, doesn't affect the larynx ? The smoke has to pass through it in order to get to the lungs...very strange

Junican said...

I have been looking for the actual West study. There are two closely similar studies. I found it.
What I was interested in was how West got his subjects. It is clear that Natstats 'Household Surveys" were the source. The importance is 'randomness'. Natstats might survey tens of thousands of people, and it is possible to select those parts of Natstats' surveys which deal with smoking and quitting. Thus, it is reasonable to assume that West's source was truly random. (IE, not derived from quit-line calls, for example)
The importance of the randomness is that it reduces the effect of 'pressures' which might induce a person to call a quit-line, for example.
What this means is that the numbers of 'quitters' using whatever method, "Just so happens" to be X in the case of 'cold turkey', Y in the case of e-cigs and Z in the case of patches and such. The important phrase is "Just so happens".
Quitting 'cold turkey' and using ecigs have this in common - they cost the taxpayer nothing. It is clearly to the advantage of the people that people should quit in those ways rather than by using prescription drugs and aids.