Monday 9 December 2013

New study finds no link between secondhand smoke and lung cancer

This news hasn't got any mainstream media coverage, despite being reported in the Journal of the National Cancer Institute and being a large, prospective study with 76,000 participants and 901 lung cancer cases, including 152 who never smoked.

I wonder why?


A large prospective cohort study of more than 76,000 women confirmed a strong association between cigarette smoking and lung cancer but found no link between the disease and secondhand smoke.

Awkward.

Investigators from Stanford and other research centers looked at data from the Women’s Health Initiative Observational Study (WHI-OS). Among 93,676 women aged 50–79 years at enrollment, the study had complete smoking and covariate data (including passive smoking exposure in childhood, adult home, and work) for 76,304 participants. Of those, 901 developed lung cancer over 10.5 mean years of follow-up.

The incidence of lung cancer was 13 times higher in current smokers and four times higher in former smokers than in never-smokers, and the relationship for both current and former smokers depended on level of exposure. However, among women who had never smoked, exposure to passive smoking overall, and to most categories of passive smoking, did not statistically significantly increase lung cancer risk. The only category of exposure that showed a trend toward increased risk was living in the same house with a smoker for 30 years or more. In that group, the hazard ratio for developing lung cancer was 1.61, but the confidence interval included 1.00, making the finding of only borderline statistical significance.

There's no such thing as borderline statistical significance. It's either significant or its not, and if it includes 1.0, it's not. It will be interesting to see the full results when they are published, but judging by the Journal of the National Cancer Institute report, it appears that this major cohort study did not find any statistically meaningful association between secondhand smoke and lung cancer in the home, in the workplace or in childhood. Even if we disregard statistical significance (and why would we?), it seems that only decades of chronic exposure in the home might, at most, have an extremely modest effect on lung cancer risk. There is nothing to indicate the slightest risk to those who work in or patronise the bars and restaurants which have been forced to ban smoking.

This should not be particularly surprising. Very few passive smoking/lung cancer studies are published these days compared to the glut of the 1980s and 1990s, but the handful that have appeared in recent years continue to support the null hypothesis. For all the campaigners' talk of "overwhelming evidence", the link between secondhand smoke and lung cancer has always been very shaky. It tends to be the smaller, case-control studies which find the associations while the larger, cohort studies do not (and, as the JNCI report notes, case-control studies "can suffer from recall bias: People who develop a disease that might be related to passive smoking are more likely to recall being exposed to passive smoking.")

Taken as a whole, the epidemiological evidence is a mess of conflicting results and weak, non-significant findings that go in both directions. The World Health Organisation's attempt to settle the controversy with a pan-European study of its own, ended in failure in 1998. Attempts to generate significant relationships with meta-analysis have relied on cherry-picking and questionable statistical practice.

The only real surprise is the knowledge that large, prospective studies are still being carried out in a field that has already served its political purpose of justifying smoking bans. And it's interesting to see how sanguine the researchers and others are about these findings. It's almost as if they never believed the hypothesis in the first place and are just happy to have smoking bans.

[Gerard Silvestri, MD, of the Medical University of South Carolina, a member of NCI’s PDQ Screening and Prevention Editorial Board] finds some reassurance in the passive-smoking findings. “We can never predict who is going to develop lung cancer,” he said. “There are other modifiers. But you can say, with regard to passive smoke, it’s only the heaviest exposure that produces the risk. We kind of knew that before, but it’s a little stronger here.”

“We’ve gotten smoking out of bars and restaurants on the basis of the fact that you and I and other nonsmokers don’t want to die,” said Silvestri. “The reality is, we probably won’t.”

Take that in for a moment. Contrast that little admission with the quackery of Stanton Glantz who has spent the last few days implying that one minute of secondhand smoke 'exposure' could be lethal. Contrast it with the BMA's assertion that "there is overwhelming evidence, built up over decades, that passive smoking causes lung cancer" (a quote made in response to yet another study that showed no such relationship.)

According to Dr Jyoti Patel, of Northwestern University School of Medicine, the real reason for smoking bans isn't saving nonsmokers from lung cancer, but denormalising smoking. Who woulda thunk it?

“The strongest reason to avoid passive cigarette smoke is to change societal behavior: to not live in a society where smoking is a norm. It’s very reassuring that passive smoke in the childhood home doesn’t increase the risk of lung cancer [in this study],” said Patel. “But it doesn’t decrease the need for us to have strong antismoking measures.”

And, as the person who presented these findings says...

“The fact that passive smoking may not be strongly associated with lung cancer points to a need to find other risk factors for the disease [in nonsmokers],” said Ange Wang, the Stanford University medical student who presented the study at the June 2013 meeting of the American Society of Clinical Oncology in Chicago.

Well, yeah. Maybe that's what we should have been doing all along.

14 comments:

Anonymous said...

I'll dig it out tomorrow Chris the WHO/Boffetta 1998 study had incomplete data from Sweden and Portugal and RR of 1.16 becomes 1.05.

What causes LC in non smokers? The same as what causes it in smokers genetic mutations.

Non smokers it is generally the mutation of the EGFR gene and what causes that is unknown.

"The K-Ras-gene is often mutated in tumours from smokers, but seldom in tumours from non-smokers; whereas the EGFR-gene is mutated in tumours from non-smokers, and not in smokers."

"INTERPRETATION:
Lung cancer in never-smokers should probably be regarded as a different disease-entity than smoking-induced lung cancer. This could impact prognosis as well as treatment."

http://www.ncbi.nlm.nih.gov/pubmed/19844277

Anonymous said...

Hi Chris must get to bed but the blog post and the COMMENTS is where it all is.

http://daveatherton.wordpress.com/2011/06/11/more-on-the-whoboffetta-paper-hat-tip-klaus-k/

Anonymous said...

Quite odd to revert to the confidence interval when looking at significance which usually is expressed when I suppose log odds are reported, against a F- or a t-Distribution. Hence, confidence intervals give some hint, but usually they do not tell too much about significance, least of all "borderline significance". Never heard of it, must be some psycho-therapeutic adaptation of statistics. The entire result to me looked very much like: We tried very hard to find some relationship between passive smoking and lung cancer, but we failed entirely, even on the borderline...

Jonathan Bagley said...

Interesting comment by Wang. In wealthy areas of the USA, the proportion of lung cancer victims who never smoked is substantial and as non-smoking lung cancer is a fairly common cancer with low survival rates, minds have become concentrated on finding the real causes and a cure, rather than persecuting smokers via the passive smoking issue.

Vova said...

Confidence intervals are produced using the same distribution as produces p-values. So if a 2-sided 95% interval contains the null hypothesis value then the (2-sided) p-value is >0.05. It's quite common & in fact often insisted on by journal editors to quote confidence intervals rather than or as well as p-values.
More to the point, if epidemiology was like other applications of statistical science (it isn't and in fact the epidemiological 'literature' has papers arguing it shouldn't be) then the evidence would be considered clear that passive smoking does not cause lung cancer.

Unknown said...

Yet another a major study (without the fingerprints of the tobacco industry) that will undoubtedly be ignored by those who stand to gain from the (not so) hidden agenda. Also makes a total mockery of the 3rd hand shite and need for outdoor bans.

Vova said...

I love this
“We’ve gotten smoking out of bars and restaurants on the basis of the fact that you and I and other nonsmokers don’t want to die,” said Silvestri. “The reality is, we probably won’t.”

The reality is they certainly will.

Fredrik Eich said...

Well according to Enstrom,

"between 1914 and 1968 among persons who never smoked cigarettes. For white males the relative increase for ages 35–84 years has been about 15-fold; the relative increase for ages 65–84 years has been about 30-fold. For white females the relative increase for ages 35–84 years has been about sevenfold."

enemy guest said...

Here's another keeper Chris...

Overdiagnosis in Low-Dose Computed Tomography Screening for Lung Cancer

http://archinte.jamanetwork.com/article.aspx?articleid=1785197

enemy guest said...

Lung cancer ‘harmless’ in one of five cases, and may not require treatment: U.S. study

http://life.nationalpost.com/2013/12/10/lung-cancer-harmless-in-one-of-five-cases-and-may-not-require-treatment-u-s-study/?utm_source=dlvr.it&utm_medium=twitter

Jonathan Bagley said...

IMO, JD is correct about confidence intervals. Substitute "brain cancer for "lung Cancer" and nearby mobile phone use" for "passive smoking" and you will find a more reasoned reaction to the results of epidemiological studies - even from epidemiologists. Like guns and knives, epidemiology isn't the problem - it's those who use it for political purposes.

Steve Kelly said...

Lifestyle epidemiology, properly used (as often it is not), has the power to suggest a clear (although imprecise) link between chronic smoking and risk of lung cancer. If there even could be some such thing as a 20% (odds ratio 1.2) increase in the risk of lung cancer amongst never-smokers from a single common lifestyle factor -- an idea which in itself beggars belief given the diversity of individuals' constitutions and also of the innumerable outside factors we are variously exposed to -- lifestyle epidemiology would be comically insufficient to demonstrating such a thing.

The efforts to demonstrate ETS risk statistically are nothing more than a farce. How many are “killed by ETS”? How many angels can dance upon the head of a pin? Doctor Gio Gori's characterization of "the ETS fraud" is apt indeed.

Doctor Gori's research on virtually safe cigarettes at the US National Cancer Institute was stifled by abolitionists decades ago. To have curtailed such a vital project was nothing short of a vicious crime. The anti-smoking crusade of recent decades has been all-in-all a vicious crime.

Those never-smoker lung cancer / ETS studies made between 1981-2011 which are generally considered procedurally acceptable are listed in a PDF document at www.olivernorvell.com. There is no "peril" from ETS. The real motivation behind the anti-smoking movement’s ETS scares has been what the campaigners like to call "denormalization": frankly stated, vilification, and social division.

Lifestyle epidemiology derives directly from statistical procedures developed by men such as Francis Galton and his protégés Karl Pearson and Ronald Fisher: the original eugenicists who likewise sought social division as a social good.

Moderating our habits, generally, is wise. Introducing virtually safe cigarettes, decades ago, would have been a positive godsend. But safer cigarettes, and all attempts at a reasoned approach to tobacco policy, have been assiduously blocked by empowered abolitionists for years upon years to the present day. Contemporary anti-smoking is an evil, and a wicked shame, which our era will bear ineradicably into the future.

Vova said...

>Steve Kelly

What do you mean they 'sought social division as a social good'? Whatever, as far as Fisher is concerned at the end of his career he was widely attacked for arguing against smoking as a cause of lung cancer To quote his biographers Yates and Mather, "It has been suggested that the fact that Fisher was employed as consultant by the tobacco firms in this controversy casts doubt on the value of his arguments. This is to misjudge the man. He was not above accepting financial reward for his labours, but the reason for his interest was undoubtedly his dislike and mistrust of puritanical tendencies of all kinds; and perhaps also the personal solace he had always found in tobacco.

Steve Kelly said...

What I meant about the old eugenicists is that they saw humanity as being divided into two groups, the fit and the unfit, and saw elimination of the unfit as a social good. Fisher was critical of 1950s tobacco research and I don't fault him for that. I fault him for the elitism and racism he displayed throughout his life. I also dislike that statistical practices he developed have been used to bad purposes both during and after his lifetime.