Thursday, 18 November 2010

On a clear day you can still see the evidence

Because of the significance of secondhand smoke (SHS) to the anti-smoking movement in the last thirty years, part of my research when writing Velvet Glove, Iron Fist was to read all the studies relating to SHS and lung cancer. Not just the numerous reports, meta-analyses and summaries which informed the political debate, but the studies themselves which showed the real evidence. Taken together, they provide a far more (for want of a stronger word) equivocal view of the science than is routinely presented by campaigners. In fact, it's difficult to imagine so many conflicting studies showing weak or negative associations being cited as "overwhelming" evidence in a less politicised field of epidemiology.

My annotated list of these studies can be downloaded here and most of the full papers can be read online, so I'll leave the interested reader to make up their own mind, but one question that has been raised is why so many of them are from the '80s and '90s and why are so few of them from the last five years?

The answer, quite simply, is that with SHS popularly recognised as lethal, there is little incentive to carry out further research. Researchers follow research grants and today—with private homes now the focus of anti-smoking campaigns—the big money is to be made in thirdhand smoke, childhood exposure and maternal smoking research.

Studies that produce epidemiological evidence for SHS exposure to adults in normal settings are now few and far between. They tend to include SHS as one possible hazard amongst many and/or receive minimal press attention.

Amongst the first category we might include the very obscure Neuberger study of 2006 ('Risk Factors for Lung Cancer in Iowa Women: Implications for Prevention', Cancer Detection and Prevention) which found that passive smokers were 73% less likely to develop lung cancer than the unexposed group—a finding that the researchers understandably chose not to dwell on.

Amongst the latter category, we could include last year's Tse study ('Environmental Tobacco Smoke and Lung Cancer Among Chinese Nonsmoking Males: Might Adenocarcinoma Be the Culprit?', American Journal of Epidemiology) which found no statistically significant increase in lung cancer risk for passive smokers in either the home or the workplace. As with Neuberger, this 'unhelpful' finding was brushed over in the text of the study by authors, and while they gamely tried to find a stronger (but still nonsignificant) association with adenocarcinoma (which is the type of lung cancer least associated with smoking), this was not a study that received a big press release.

And now—well-spotted by Dick Puddlecote, since it received no press attention—comes another null study. This one (D. Brenner, 'Lung cancer risk in never-smokers: a population-based case-control study of epidemiologic risk factors', BMC Cancer, 2010) found precisely zero increased risk of lung cancer from either childhood or adult exposure to SHS (1.0 95% CI: 0.6-1.8, and 1.0 95% CI: 0.5-2.0 respectively) and no statistically significant increase for workplace exposure (1.2 95% CI: 0.7-2.0). It did, however, find a near-trebling of risk for exposure to paints and solvents (2.8 95% CI: 1.6-5.0) and for exposure to smoke-soot and exhaust (2.8 95% CI: 1.4-5.3). The authors conclude:


Our results support the concept that exposure to exhaust fumes and or soot/smoke (from non-tobacco sources) is a source of carcinogenic exposure.


This study is—or should be—of particular interest since its sample group is unusually large, comprising some 445 lung cancer cases. This puts it high in the rankings of large, well-conducted studies (size being very important when it comes to accurately quantifying risk).

Instead, this study—like most others that fail to support the passive smoking theory—has gone largely unnoticed because, as we all know, 'the debate is over'.

119 comments:

JJ said...

Chris – The seeds of the anti-tobacco lobby destruction already lay in the facts that you so cogently point out. There are quite simply no epidemiological studies whatsoever that give results that the antis so badly desire …no matter how adroitly massaged.
Besides which, how does anyone distinguish between SHS and carbon monoxide poisoning from vehicle emissions, industrial pollution from home and abroad…and of course pesticides, fungicides, herbicides and insecticides which are routinely sprayed on farmers crops worldwide as well as gardens where pest control is necessary.

All of this swirls around in the atmosphere and we breathe it all in…but the antis desperately try to make a case about a few wisps of smoke (second hand smoke) emanating from someone’s cigarette, cigar or pipe which immediately dissipates into the surrounding atmosphere and is 90% water vapour, so any residual evidence simply does not exist for any case to be made.

Just recently a programme on the BBC outlined two types of breast cancer which have been pinned down to rogue genes…I wonder as the genome is further explored and subsequently understood – whether we shall see a particular type of lung cancer normally associated with smoking – being attributed to genetic malfunctions.

Interestingly, I’ve also noticed when items about lung cancer are aired smoking was not mentioned on a few occasions when previously it had. Why? Could an understanding that perhaps smoking is not such an obvious culprit after all...is making its presence felt?

Anonymous said...

Debating the economic, political and public health implications of tobacco and/or anti-tobacco is one half of what has always made the topic of tobacco use fascinating to me.

The other half is being a living witness to the cultural and sociological transformation taking place, as well as (the tragic, in my view) consequent psychological transformation taking place in people I've known as I've followed these issues over the years. The very same elders in my family who made a cumulus cloud of smoke over the kitchen table a fixture of my childhood now banish me outdoors to smoke with an accompanying lecture, for instance.

While I'm disgusted by it all, I'm simultaneously fascinated to watch the unfolding of the very things I learned of in the preventative history lessons of my teenage years.

Perhaps I'm a die hard, but I think all of the battles over secondhand smoke, in reality, amount to little more than the equivalent of making a worldwide, trillion dollar issue of the debate you might have with your spouse regarding keeping the window open or the fan on at night.

For the risk of being insensitive (and going on too long), I want to wish all of you and your families well in the struggles you have been having with the worldwide bird flu epidemic. I know that you have likely lost someone to the bird flu. After all, the media and experts warned us all with either front page stories in the newspapers, or leading stories on the TV news, that we, or someone we know, would be dead of bird flu by now.

If you are not dead, it's likely only because of the many billions of taxpayer dollars the government paid to provide us all with a vaccine, and you have received that vaccine shot. Who received the billions to produce that vaccine that all of us lined up for and are sure that we've received? Remember getting that life saving vaccine?

I assume it was Big Pharma, but I might be wrong. Maybe it was some humble, independent, bird flu warrior of a university research scientist. Funny that no one knows, considering what a big deal the bird flu was.

http://articles.cnn.com/2005-11-01/health/us.flu.plan_1_pandemic-strain-vaccine-makers-flu-pandemic?_s=PM:HEALTH

Anonymous said...

Debating the economic, political and public health implications of tobacco and/or anti-tobacco is one half of what has always made the topic of tobacco use fascinating to me.

The other half is being a living witness to the cultural and sociological transformation taking place, as well as (the tragic, in my view) consequent psychological transformation taking place in people I've known as I've followed these issues over the years. The very same elders in my family who made a cumulus cloud of smoke over the kitchen table a fixture of my childhood now banish me outdoors to smoke with an accompanying lecture, for instance.

While I'm disgusted by it all, I'm simultaneously fascinated to watch the unfolding of the very things I learned of in the preventative history lessons of my teenage years.

Perhaps I'm a die hard, but I think all of the battles over secondhand smoke, in reality, amount to little more than the equivalent of making a worldwide, trillion dollar issue of the debate you might have with your spouse regarding keeping the window open or the fan on at night.

Well, what can one do?

To anyone reading this, best wishes to you and yours in coping with the "bird flu". I know it's having an impact on you. The media told me it definitely would five years ago or so.

http://articles.cnn.com/2005-11-01/health/us.flu.plan_1_pandemic-strain-vaccine-makers-flu-pandemic?_s=PM:HEALTH

Anonymous said...

Sorry. I thought I lost the first one due to length, so I edited and double posted.

Rollo Tommasi said...

I see you’re still cherry-picking your evidence Chris! Why focus on adulthood-only exposure or childhood-only exposure for never smokers only, when the study also shows results for exposure in childhood and/or adulthood – and gives these results for both the total population and never smokers? Oh, it’s because these results show a higher relative risk (1.2, 95% CI 0.7-1.9 & 1.1, 95% CI 0.6-1.9 for total population and never-smokers respectively) than your carefully selected figures do.

And why ignore the killer finding that SHS exposure from both home and work increases the relative risk of lung cancer even more highly - to 1.4 (95% CI 0.9-2.2) for the total population and 1.2 (95% CI 0.7-2.1) for never-smokers?

If you take this study’s full results, instead of cherry-picking one or two of the sub-findings, it is clear that it actually SUPPORTS the established figures about passive smoking increasing the risk of lung cancer.

Your dubious claim that this study “fail(s) to support the passive smoking theory” relies on that old lie that if a finding is not statistically significant in itself, then there is no significant risk.

Rab said...

Rollo: Since the CI in all the results you cite encompass a RR of 1.0 they can all be regarded as statistically insignificant.
I would hardly count that as a "killer finding".

Rab said...

Rollo

"Your dubious claim that this study “fail(s) to support the passive smoking theory” relies on that old lie that if a finding is not statistically significant in itself, then there is no significant risk."

Harldy a lie, actually more of a definition issue. A significant risk would not "by definition" be statistically insignificant when you calculate it.
A significant risk would show an unequivical correlation. Besides, even if the figure was bang on a 0.2 rise on the baseline hardly qualifies as *significant* especially when that translates into a figure of 1 in 50,000 per annum.
I mean honestly Rollo, if you bought 5 lottery tickets per week would you count yourself as having a significant chance of winning the lottery?

Snowdon said...

Rollo,
Written at 9.01 am? It's nice to know this is the first blog you check when you get to work. And it's comforting to know you're still quietly lurking and waiting for a chance to debate SHS, but better to stay silent and be thought a fool than to speak up and remove all doubt.

The never-smokers are the relevant group to study, since the total population sample includes smokers. Nearly all passive smoking studies focus on the effect on non-smokers for obvious reasons. And the authors focus on them in this study, concluding:

Cases and controls did not vary significantly in the total hours exposed to ETS during childhood or adulthood at home (data not shown). Among never smokers in our population, we observed no association between either exposure to ETS at home or at the workplace and lung cancer risk (Table 2).

It's probably why the study is called 'Lung cancer in never-smokers', eh? But even if we are to focus on the ever-smokers...

"In general, the effect estimates for ETS exposure were similar between the total population and only among never smokers."

ie. there is, and let's repeat that, "no association".

To AVOID cherry-picking, my list of passive smoking studies all looked at the exactly the same thing: domestic exposure to adults (Why? Because there are more studies of that than anything else). The RR of domestic exposure to adults here is 1.0 (0.5-2.0). If you think that a combined childhood/adult figure of 1.1 (0.6-1.9) is a 'killer figure' I can only assume you are on heavy medication. A 10% increase in lung cancer risk to never-smokers is not significant in either a conventional or a statistical sense. And statistical significance is not a "lie", it is the most basic test in epidemiology, as even Stanton once admitted:

http://velvetgloveironfist.blogspot.com/2010/10/stanton-glantz-then-and-now.html

Magnetic said...

“but better to stay silent and be thought a fool than to speak up and remove all doubt.”

Wonderful advice, Chris. However, concerning the intentional fool, Rollo, all doubt has long been removed.

Too much damage has already been done by the antismoking bigots posing as “promoters of health” to entertain an incompetent sub-amateur like Rollo.

Anonymous said...

Rollo, no matter whether arising from a single study or a large meta analysis, a relative risk of 1.1 or 1.2 would not be given much weight by conventional statisticians, regardless of whether the confidence interval did not include 1.0, i.e. significant in the statistical sense. This is because of systematic bias which is not removed by conducting more or larger studies. Systematic bias caused by hidden and residual confounding and misclassification errors, to mention just two out of many problems. It would be helpful if you revealed your identity. I don't know the level at which I should pitch my comments. I am interested to hear the views of professional statisticians on what level of relative risk should be taken seriously in a meta analysis of case-control studies.
Jonathan Bagley.

Pogo said...

@anon, 12:10... In answer to your closing question:-

In epidemiologic research, [increases in risk of less than 100 percent] are considered small and are usually difficult to interpret. Such increases may be due to chance, statistical bias, or the effects of confounding factors that are sometimes not evident .[Source: National Cancer Institute, Press Release, October 26, 1994.]

"As a general rule of thumb, we are looking for a relative risk of 3 or more before accepting a paper for publication." - Marcia Angell, editor of the New England Journal of Medicine"

"My basic rule is if the relative risk isn't at least 3 or 4, forget it." - Robert Temple, director of drug evaluation at the Food and Drug Administration.


Quoted in Prof Brignell's "Numberwatch" summary of Relative Risk.

Gary K said...

"Save the children."

OT; but, only a little bit.
Let's look at a 15 year old kid that starts smoking and their chances of lung cancer death(LCD) or cardiovascular disease(CVD) death over a lifetime(60+ years) of smoking.

First some facts.

The median age of LCD is about 72 and smokers and non-smokers get lung cancer at the same age and if a smoker gets lung cancer there is about a 90% probability it was caused by factors other than their smoking.

The median age for CVD death is about 77 and if a smoker gets a CVD there is a 99% probability it was caused by factors other than their smoking.

The US Centers for Disease Control(CDC) says that there are NO smoking related disease deaths below the age of 35.

There are very,very few LCD's or CVD deaths below the age of 45.

Our 15 year old child will only be a child for 3 years, the age of 18 is considered being an adult.

A 15 year old can smoke for 20 years and have ZERO chance of dying from a smoking 'caused' disease.

A 15 year old can smoke for 30 years and have a very,very small chance of a LCD or CVD death.

If our 15 year old is going to die from lung cancer, there is only a 50% chance that LCD will occur during their first 57 years of smoking and if they do get lung cancer, there is about a 90% probability that cancer was 'caused' by something other than smoking.

If our 15 year old is going to die from CVD, there is only a 50% chance that CVD death will occur during their first 62 years of smoking and there is a 99% probability that CVD was caused by something other than their smoking.


If anyone wishes sources and data, I will be happy to post them.

Gary K said...

Most anti-smokers seem to think that a positive RR is proof of a disease or death occurring.

Since most diseases(all smoking diseases) are multi-factorial in causation, an RR is a 'probability ratio' between factors and NOT a proof of 'causation'.

SHS exposure and never-smoker lung cancer has an RR=1.20.

1 is the probability of other factors being the 'cause' and .2 is the probability of SHS being the 'cause'.

Since 1 is 83% of 1.2, this tells us that a never-smoker exposed to SHS has an 83% probability of some other factor being the 'cause' of their lung cancer.

Anti-smokers will claim that if 10/10,000 never-smokers(not exposed to SHS) get lung cancer and 12/10,000 never-smokers(exposed to SHS) get lung cancer it proves that those 2 lung cancers were 'caused' by SHS exposure.

All this shows is an 83% probability that those 2 lung cancers were NOT 'caused' by SHS!!

An RR=2, a 100% increase, still gives a 50% probability of some other factor being the cause of a disease or death.

Rollo Tommasi said...

What a curious post, Chris. I see you offer no explanation for the cherry-picked sub-findings you chose to refer to in your article, ignoring the major findings of the study in the process. Yes, more studies look into never-smokers and domestic exposure; but that is not to say that other forms of exposure and risks for ever-smokers should be disregarded. And it certainly doesn’t explain why you quoted sub-findings in the first place.

Then you consciously misquote the study, by focusing on one sentence in the middle of the report about a lack of association, but then ignoring the fact that the report explains that “ETS exposure was not found to significantly increase risk among never smokers in this study, however, several potential explanations are possible”. It then focuses on two main limitations in the study which may “lead to a bias toward to null”.

Then you go and consciously misquote my own words! At what point did I state that “statistical significance is a….lie”? My point was that the conclusion to be derived from a statistically non-significant result is NOT that there is no additional risk. It is that the results cannot determine in statistical terms whether there is an additional risk or not. Want some evidence for this?

Non-significance does not mean ‘no effect’. Small studies will often report non-significance even when there are important, real effects which a large study would have detected.” (http://www.medicine.ox.ac.uk/bandolier/painres/download/whatis/What_are_Conf_Inter.pdf).

The fact is the results of this study are broadly consistent with previous results about the risk of lung cancer from exposure to SHS. And if you want a proper pooled RR for never-smoking women exposed to passive smoking from spouses, which is more scientifically worthy than your list, try this - 1.27 (95% CI 1.17–1.37) (from http://ije.oxfordjournals.org/content/36/5/1048.full#B51).

Snowdon said...

"I see you offer no explanation for the cherry-picked sub-findings you chose to refer to in your article"

Uh, yes I did. I think I explained it very clearly.

"..ignoring the major findings of the study in the process."

The major finding was that "we observed no association between either exposure to ETS at home or at the workplace and lung cancer risk." Is that clear enough for you.

Etc. Etc.

Life really is too short to keep repeating myself like this. Do any other readers feel I haven't already answered Rollo's questions?

Gary K said...

" try this - 1.27 (95% CI 1.17–1.37)"

Ahhh Rollo,
You do realize that by insisting on the .27 being 'real', you are insisting on the 1 also being 'real'.

Thus, out of 13 cancers, 10 are NOT caused by SHS.

You are claiming that out of 13 cancers 3 are caused by SHS and 10 are not!!

Rollo Tommasi said...

And life's too short to keep pursuing Chris Snowdon on his dodging.

Shame. If he were a serious commentator, he'd be open to a proper informed discussion.

JJ said...
This comment has been removed by the author.
Snowdon said...

Gary K,

I'm afraid that's not how it works. You can't make those calculations from the relative risk alone - it depends on how many other risk factors there are and what their values are.

I am curious to know what your source for saying that "if a smoker gets lung cancer there is about a 90% probability it was caused by factors other than their smoking" however.

Gary K said...

RR's

A Relative Risk increase ALWAYS has to be in reference to some previous amount that would have happened without the presence of the factor being considered.

Thus Rollo's 27% increase is in addition to some amount that already occured.

Anti-smokers have gotten away with claiming the total amount has been caused by smoking.

For instance the 3,400 never-smoker lung cancer deaths 'caused' by SHS and a RR = 1.2%.

To get the 20% increase they first must have 2,833 deaths that would normally occur and then 567 extra deaths due to SHS for the 3,400 deaths.

The BIG LIE comes when they claim all 3,400 deaths to be due to SHS when 83% were not.

The 567 extra deaths are doubtful since each of them has an 83% probability of being 'caused' by some other factors.

Of the about 50,000 never-smoker total deaths(USA) claimed to be due to SHS, at least 40,000 are 100% due to other factors and each of the 10,000 others have at least an 80% probability of other factor causation.

Gary K said...

" a 90% probability "

The factor that is never adjusted for is that not all ex-smokers have the same risk of lung cancer/deaths.

http://profiles.nlm.nih.gov/NN/B/B/D/B/_/nnbbdb.pdf
1990 Surgeon General's Report
Lung Cancer

Page 110
In the British Physicians Study, U.S. Veterans Study, and ACS CPS-II, former smokers
who had been abstinent for 15 years or more showed an 80- to 90-percent reduction in
risk compared with current smokers.

http://stopsmoking.uchicago.edu/benefits.html
Chicago STOP Smoking Research Project
Benefits of Quitting Smoking
Health benefits from quitting smoking:
lung cancer death rate decreases by half in 5 years, and is similar to that of nonsmokers after 10 years

http://www.lungusa.org/finding-cures/our-research/trend-reports/Tobacco-Trend-Report.pdf

Using data from the site above(table 3), we see that there were 46 million current smokers in 1990 and 2008.
Using data from table 15, we can figure out that there were 44 million x-smokers in 1990 and 48 million x-smokers in 2008.

This shows that there are 44 million x-smokers with a lung cancer risk similar to never-smokers and 4 million x-smokers with a lung cancer risk similar to current smokers.

44 million is 92% of 48 million and 92% of the ex-smokers' lung cancers and LCD's must be shown in the grouping with never-smokers.

http://www.cdc.gov/mmwr/preview/mmwrhtml/mm5644a2.htm

Table 2:

lung cancers( never-smokers) = 17.9%
157,000 lung cancer deaths per year x 17.9% = 28,103 lung cancer deaths(LCD's)

lung cancers(ex-smokers) = 61.2% = 96,084 LCD's

lung cancers(current smokers) = 20.9% = 32,813 LCD's

After adjusting for the fact that 92% of the ex-smokers have the same risk of lung cancer as never-smokers, we have 180 million non-smokers having 116,500 LCD's.
This is a rate of 6.5/10,000.

Current smokers = 32,813 LCD's per 46 million smokers.
This is a rate of 7.1/10,000.

If they had never smoked, current smokers would still have 6.5/10,000 LCD's.

6.5 is 91% of 7.1.

Gary K said...

A couple of other points.

The number of years smokers have smoked has no bearing on the age at which they are diagnosed with lung cancer;because:

Smokers and never smokers are diagnosed with lung cancer at about the same age!!

http://seer.cancer.gov/statfacts/html/lungb.html
US Mortality:From 2001-2005, the median age at death for cancer of the lung and bronchus was 72 years of age.

http://jco.ascopubs.org/cgi/content/full/25/5/472
RESULTS
Although never smokers were slightly older at lung cancer diagnosis than current smokers in two population-based cohorts (MEC and NHEFS), this difference was not observed in the majority of cohorts evaluated (NHS, HPFS, CTS, and U/OLCR; Table 2).

DaveA said...

@Rollo



Here is the OR for smoking and cervical cancer "OR for current smoking habit 20+ per day = 2.57; 95% Cl = 1.49-4.45). " So Rollo should we ban smoking to cure CC? No of course, because they now know what the sole cause of CC is HPV 16 and 18 aka genital warts and is contracted sexually. The confounder and the correlation here is that HPV is associated with promiscuous people and promiscuous people are more likely to smoke.



Also the Brenner paper with its relative strong response to "..find a near-trebling of risk for exposure to paints and solvents (2.8 95% CI: 1.6-5.0) and for exposure to smoke-soot and exhaust (2.8 95% CI: 1.4-5.3), " might explain the small elevated risks if at all in some studies. Smoking is often a sign of relative poverty, if you are working class you are twice as likely to smoke. Hence it seems logical that working class people are exposed to SHS more than middle class people, but it is not the SHS that induces LC but the fact that maybe a painter and decorator, or car mechanic are exposed to solvents and soot and hence have a higher incidence of LC. Sorry Rollo back to the drawing board.



http://www.ncbi.nlm.nih.gov/sites/entrez?Db=PubMed&Cmd=ShowDetailView&TermToSearch=11076670

Gary K said...

CardioVascularDisease(CVD)/Deaths

The 15 year stat also applies to CVD's.


here:
http://www.cancer.org/Healthy/StayAwayfromTobacco/GuidetoQuittingSmoking/guide-to-quitting-smoking-benefits



The American Cancer Society says that 15 years after quitting: The risk of coronary heart disease or stroke is the same as a non-smoker's(never smoker-GK).

The CDC table data allows us to show that there is NO increased risk for CVD due to being a current smoker.

Ex-smokers will have higher rates of lung cancer and CVD because these are diseases that are found more commonly among the elderly.

About 33% of ex-smokers are over the age of 65, compared to about 8% of current smokers and about 12% of the never-smokers.

Rollo Tommasi said...

Gary K – Chris is actually right on this one. Numbers of deaths attributable to a particular risk don’t just depend on relative risks. They also depend on the numbers of people exposed to that risk and the total number of deaths from the cause to which the risk relates.

Jamrozik’s 2005 study showed that around 1,400 of the 32,000 lung cancer deaths in the UK each year are attributable to passive smoking. So too are around 5,200 of the 112,000 heart disease deaths. All the calculations Gary K talks about were made in order to reach these 6,600 deaths attributable to passive smoking.

I trust Gary K agrees with me that 6,600 – or even a number considerably lower than that - is an appallingly high number of deaths from exposure to secondhand smoke.

Rollo Tommasi said...

DaveA – Sorry mate. You know I have nothing against you, but I find your latest post bixarre..

What possible relevance does the number of cervical cancer deaths from promiscuous sex have to this debate on passive smoking? How many people who happen to be in the vicinity of playful couples are exposed to secondhand HPV?

As for your remarks on the Brenner paper, that is nothing more than conjecture on your part. There is no evidence to suggest that lung cancer deaths attributed to SHS are really due to paints, solvents or soot. Furthermore, how do you explain the c. 5,200 heart disease deaths each year attributable to passive smoking? There is no suggestion that these other factors increase the risk of heart disease.

Ann W. said...

Rollo, are you referring to this study?

BMJ 2005; 330 : 812 doi: 10.1136/bmj.38370.496632.8F (Published 1 March 2005)

* Paper

Estimate of deaths attributable to passive smoking among UK adults: database analysis

1. Konrad Jamrozik

Anonymous said...

To emphasise DaveA's point about working class exposure, take a look at table 3 of the Brenner paper:

54 never smokers were exposed to paints etc out of a total of 466 in the control group, whereas 152 were exposed out of the total control population of 948 (which includes the never smokers). So out of 482 current or former smokers, 98 were exposed. (948-466=482 and 152-54=98).

So smokers and former smokers were nearly twice as likely to be exposed to paints etc compared with the never smokers. This relation holds for all the pollutants measured, perhaps because they were more likely to be manual workers.

It is also odd that the LC OR for (non-tobacco) pollutant exposure for smokers is much lower than for never smokers. But that might be to do with adjustments made by the authors.
The control group exposure figures mentioned above are not adjusted.

The ORs for solvents, welding and other non-tobacco pollutants are statistically significant and well above 2.0 so should be taken as seriously suggestive. Of course correlation does not imply causation even there.

Tony

Rollo Tommasi said...

Ann - Yes.

Klaus K said...

New Study acquits passive smoking

This is a Google translation of my artice about the Brenner study, which also features an interview with Geoffrey Kabat.

He now points to the fact, that since scientists may have found, that lung cancer in smokers and never-smokers are two different diseases, it is getting harder to argue, that SHS can be a cause of Lung Cancer.

The translation is not perfect, but I hope you understand.

Snowdon said...

Please do read Klaus's article, folks!

Ann W. said...

Chris, was my last post removed or was there a malfunction..........

Pogo said...

@Rollo Tommasi...

Jamrozik’s 2005 study showed that around 1,400 of the 32,000 lung cancer deaths in the UK each year are attributable to passive smoking.

Thanks Rollo, interesting cite. I'm assuming that you never got past the abstract.

Jamrozik arrives at these figures by a bit of guesswork and by torturing the results found in Woodward & Laugesen, "How many deaths are caused by second hand cigarette smoke?" (Tobacco Control 2001;10:383–388)...

Woodward & Laugesen is a metadata study based upon a trawl through the literature 'using key words “tobacco smoke pollution”, “second hand smoke”, “environmental tobacco smoke”, and “passive smoking”.' and although it managed to find and cite the highly dubious "US Nurses health Study" (which incidentally had a statistically-insignificant result in this case) it failed to find the Boffetta et al report "Multicenter case-control study of exposure to environmental tobacco smoke and lung cancer in Europe." - one of the largest ever done, under the auspices of the World Health Organisation. I suppose that we should be relieved that, unlike Jamrozik, they didn't cite Stanton Glanz!

That said, the Relative Risk figures quoted in W&L never reached a level of significance - eg Lung Cancer, RR 1.24 (95% CI 1.13 1.36).

So, to sum up... Jamrozik's paper was a metadata reports based upon metatdata reports that didn't produce any significant findings.

If that's the best you can do Rollo... :-(

Snowdon said...

Ann,

I haven't removed anything and never do.

Chris

DaveA said...

Here is a paper from Oxford Statistian Peter Lee on 80 SHS studies. I also must do some more research into misclassification as Lee says the avaerage is 1%-4% averaging out at 2%. This would reduce assuming an RR of 10-20 of active smoking to 0.85-1.05.

http://www.pnlee.co.uk/documents/refs/lee2010B.pdf

Ann W. said...

Rollo, one of the problems with studies that estimate death attributions is that they are based on assumptions. If one or more of these assumptions are wrong, then all the calculations are also wrong.

An example of this is in the very detailed TABLE 5 Projected age-specific rates (per 100,000) of smoking-attributable deaths by age, sex and province, Canada, 2000
in the study "Projected Smoking-attributable Mortality in Canada, 1991-2000" by Larry F Ellison, Yang Mao and Laurie Gibbons, Cancer Division, Bureau of Chronic Disease Epidemiology, Laboratory Centre for Disease Control, Health Canada, Tunney's Pasture, Postal locator: 0601E2, Ottawa, Ontario K1A 0L2
http://www.phac-aspc.gc.ca/publicat/cdic-mcc/16-2/c_e.html
Ann - Although we have taken no time to conduct a detailed evaluation, we judge that the discrepancies between the estimated values and what you have observed, likely resulted from a projection assumption violation or a combination of the violations, in particular the mortality rate reduction in major chronic diseases and the smoking prevalence rate decrease
observed in males. It worth noting that the projections are done for each province and than aggregated to the national total. In some provinces, small numbers could cause large data variation and increase the likelihood that the projected values will differ from the true values. Please feel free to e-mail me or call me for questions and clarification. If it is more convenience, you may provide me you phone number, so I can call you and discuss the issue more. - Dr. Yang Mao.

So of course the question now is "what did I observe?"

when I test the figures to actual deaths, they came out so grossly overstated that we would have to bring back the non smoking dead to balance out the smoking-attributable deaths for each category under age 85.

Rollo Tommasi said...

Pogo, Ann – Thanks for your interesting comments.

But Pogo – While you may complain about the lung cancer results in the Woodward & Laugesen, the fact is that their results are replicated in other studies.
> The SCOTH report refers to an additional lung cancer risk from passive smoking of 24%.
> The US Surgeon General’s report refers to a 20-30% additional risk.
> A 2007 meta-analysis (http://ije.oxfordjournals.org/content/36/5/1048.full#B51) refers to a 27% additional risk.

And ALL of these reports take account of the Boffetta study which you argue was ignored by Woodward & Laugesen. So in what ways are Jamrozik’s conclusions exaggerated?

And Ann – I see nothing from your remarks to provide evidence that Jamrozik’s reasoning was flawed. I also have to ask – by what percentage would Jamrozik’s estimate of 6,600 SHS-related deaths each year have to be exaggerated before the number of deaths became insignificant in your mind? To me, even if you could claim the actual number of deaths was considerably lower than 6,600, it would still be a significant and worrying figure.

Ann W. said...

Rollo, I think that you are forgetting that all those numbers are a mathematical and statistical exercise. Another words, No persons were harmed or died in the making of these figures.

**note: we did not look at a literal body count but on well-documented economic theories and assumptions,**
http://www.ccsa.ca/Eng/Priorities/Research/CostStudy/Pages/default.aspx

I weep for real people not for theories or assumptions.

Ann W. said...

Konrad was diagnosed with advanced sarcoma in September 2009 and lost his battle with this disease in March 2010.
http://www.news.uwa.edu.au/201004272430/school-population-health/memorial-celebration-konrad-jamroziks-contributions-public-hea

Holy cow, He was 54 years old. I wonder what type of sarcoma he died from............

Risk Factors for Classical Kaposi’s Sarcoma
http://jnci.oxfordjournals.org/content/94/22/1712.full.pdf

The strongest association was a reduced risk of KS with cigarette smoking (OR = 0.25, 95% CI = 0.14 to 0.45). Cigarette smoking intensity and duration could be evaluated for men, among whom the risk for KS was inversely related to the amount of cumulative smoking (Ptrend<.001).

KS risk decreased approximately 20% (OR = 0.81, 95% CI = 0.74 to 0.89) for each 10 pack-years reported, and it was decreased sevenfold (OR = 0.14, 95% CI = 0.07 to 0.30) with more than 40 pack-years. In multivariable analysis, a decreased KS risk was associated with smoking (OR = 0.23, 95% CI = 0.12 to 0.44);
but an increased KS risk was associated with
topical corticosteroid use (OR = 2.73, 95% CI = 1.35 to 5.51),
infrequent bathing (OR = 1.85, 95% CI = 1.04 to 3.33),
and a history of asthma (OR = 2.18,95% CI = 0.95 to 4.97)
or of allergy among men (OR = 2.59, 95% CI = 1.15 to 5.83) but not among women (OR = 0.09, 95% CI = 0.003 to 2.76).
KS was not related to other exposures or illnesses examined.

Conclusion: Risk for classical KS was approximately fourfold lower in cigarette smokers, a result that requires confirmation by other studies. Identification of how smoking affects KS risk may lead to a better understanding of the pathogenesis of this malignancy and
interventions for its prevention. [J Natl Cancer Inst 2002; 94:1712–20]

Rollo Tommasi said...

Ann W: Do you not realise that your latest 2 posts directly contradict each other?

You complain about the results of epidemiological evidence when it suits your purposes. But you’re happy, not just to accept but also to push, epidemiological evidence when you think it suits your purposes. What a hypocritical line you are pushing!!

Using your approach to try to deny the actual, probable or even possible deaths caused by passive smoking is also utterly callous. How many lives are you prepared to risk just to spare some smokers the slight inconvenience of not lighting up in enclosed public places?

Pogo said...

@Rollo Tommassi, 20 November 2010 00:39:-

Thanks for coming back with more information...

SCOTH, a metastudy, on page 5 states - In most studies considered individually the observed odds ratios failed to reach statistical significance. They were nevertheless comfortably within the confidence limits of the pooled odds from the 1997 meta-analysis presented to SCOTH...

That's not correct. You cannot claim that the sum total of a collection of insignificant studies somehow becomes significant. That said, SCOTH still only manages to claim an RR of 1.24 - which is in itself not significant.

Taylor, Najafi and Dobson, yet another metastudy, looked at many of the same individual reports and, not unsurprisingly, came to similar conclusions. Pooled RR of 1.27 - insignificant.

It's interesting to note that, as you say, the above studies include the findings from the Boffetta et al report that I quoted previously. However, in SCOTH it's mentioned as being part of a group of studies wherein "The authors of these publications have consistently reported an increased risk of lung cancer associated with exposure to other people’s smoke" and in the Taylor study it's referenced as supporting the claim that "Our findings are also consistent with results from exposure to ETS in workplace in both men and women.", which is in fact completely contra the conclusions reached in Boffetta

There seem to have been so many "Surgeon-General's" reports that it's difficult to know which one to reference... However, using your quoted "20-30% additional risk". this is still insignificant.

"Insignificant"... Why do I say this? First, may I draw your attention to the quotes I included in my 18 November 2010 14:13 response to Jonathan Bagley...?

In epidemiologic research, [increases in risk of less than 100 percent] are considered small and are usually difficult to interpret. Such increases may be due to chance, statistical bias, or the effects of confounding factors that are sometimes not evident .[Source: National Cancer Institute, Press Release, October 26, 1994.]

"As a general rule of thumb, we are looking for a relative risk of 3 or more before accepting a paper for publication." - Marcia Angell, editor of the New England Journal of Medicine"

"My basic rule is if the relative risk isn't at least 3 or 4, forget it." - Robert Temple, director of drug evaluation at the Food and Drug Administration.


Any professional statistician will confirm the above. Under virtually no circumstance will an RR of less than 2.0 (or greater than 0.5 in the case of an inverse risk) be considered significant. It might be acceptable in the case of a very closely controlled double-blind, single factor study, but observational studies, as are all the above, need to show an RR of at least 2.0 to achieve significance and, if the New England Journal of Medicine or the Food and Drug Administration are to be believed, an RR exceeding 3.0 is required.

There has not been a single study of which I'm aware that has shown any statistical significance of the effects of SHS. I'm not disputing that smoking is not good for the active smoker, but there has been too much public policy on the subject of SHS based on far too little "evidence" and an awful lot of very bad "science" - which, as a non-smoking, retired physicist I find far more objectionable than the smell of tobacco smoke.

Pogo said...
This comment has been removed by the author.
Belinda said...

There is no 'actual' that I can see, there may be 'probable' or 'possible' (more likely improbable and highly unlikely going on the figures). I see no contradiction: there are good and bad uses of epidemiology, and bad uses are made consistently when trying to make the case for passive smoking being dangerous.

As for talking of your opponents in the debate as callous: well, don't be silly. There is nothing callous in speculating about hypothetical and not very probable deaths, (especially when the smoking ban policy does not stop people smoking anyway, just restricts them to their homes). To paraphrase: I know I talk a lot of rubbish, but isn't the bare possibility that I am right enough for you? Isn't the bare possibility that lives will be lost if I don't get my way enough for you? (There is however callousness in encouraging the eviction of elderly residents who smoke, which is happening in USA on the basis that secondary smoke floats between different apartments.)

Rollo Tommasi said...

Good grief, Belinda. You frequently resort to the tactic of describing deaths attributable to passive smoking as “statistical” or "hypothetical" and not “real”. If you genuinely had grounds for that view, you would be able to argue how passive smoking was safe. You’d be able to show how the relative risks identified in studies like the IARC monograph, SCOTH report and US Surgeon General’s report were all wrong. And you’d be able to argue why Jamrozik’s work in translating relative risks into actual deaths was fundamentally flawed. But you never do, because you can’t.

You simply use the argument of “hypothetical” deaths as a device to allow you to pretend that passive smoking is safe. That approach is as cynical as it is callous.

Rollo Tommasi said...

Pogo - I have tried to respond to your post several times now. But every time I do, my post soon disappears. Don't know what's happening.

Rollo Tommasi said...

Pogo – Part 1 of my response.

I’m afraid you’ve been tricked by the lies of pro-smoking groups. The truth is that epidemiological science expects that a relative risk needs to be 2 or more ONLY if you are trying to reach a conclusive finding from the results of a SINGLE study. It does not apply to relative risks established from a large body of evidence, as is the case with passive smoking, in which case smaller relative risks ARE relevant. And that approach is quite right. Otherwise epidemiologists would have to conclude that either something presents a risk of double or more, or else there is no additional risk at all – which is clearly nonsense.

Rollo Tommasi said...

Let’s look at your quotes, Pogo, to further prove my point. The Robert Temple quote relates to a discussion about the relative risk required before a "single epidemiologic study is persuasive by itself" (http://www.nasw.org/awards/1996/96Taubesarticle.htm). The deception by pro-smoking groups is even more blatant in the way they misquote Marcia Angell in the same article, because they deliberately and cynically ignore the rest of her quote when she states "particularly if it is biologically implausible or if it's a brand-new finding”.

To further prove my point, while Ms Angell was executive editor at the New England Journal of Medicine, it published a report which showed that " nonsmokers exposed to environmental smoke had a relative risk of coronary heart disease of 1.25 (95 percent confidence interval, 1.17 to 1.32) as compared with nonsmokers not exposed to smoke " and concluded that "Given the high prevalence of cigarette smoking, the public health consequences of passive smoking with regard to coronary heart disease may be important"? Surely if your claim about relative risk was right, she wouldn’t have accepted a study with those findings for publication (http://content.nejm.org/cgi/content/full/340/12/920).

Ann W. said...

Actually Rollo, I think my post supports my stance. Do you really think that I believe that smoking would protect someone from Classical Kaposi’s Sarcoma any more then increased bathing? Give your head a good shake.

Rollo says "utterly callous".

Tell me what in life does not come with a risk? Or are you one of those that decides what risks are okay and what aren't?

Risk and harm are not the same thing.

You place someone else at risk ever time you do something but that does not mean that any harm has occurred.

If I wept for every risk I placed someone at, my tears would never stop........

(I too am having problems with my posts appearing)

Belinda said...

Rollo, I am sorry, if name-calling is all you can do to convince me that deaths described in hypothetical and statistical studies (based on assumptions) are real, all I can say is that your powers of argument are far less persuasive than the many people who are arguing against you.

Belinda said...

Also you haven't explained how results with a confidence interval spanning 1 (and including RR < 1) can have actual significance.

Belinda said...

Surely a large body of studies with findings that span RR 1 or very close to RR 1 confirm very low significance of the elevated risk – not, as Rollo appears to believe, higher significance of the elevated risk?

Rollo Tommasi said...

Ann: Firstly, you did not explain the point you were trying to make with reference to that Classical Kaposi’s Sarcoma study. Even now I’m not fully sure. After all, that reported risk relates to active smoking, not passive smoking. If your point is that even higher relative risks may not mean there’s a serious risk, then yes I would agree with you. Epidemiology is not simply a numbers game. And it certainly does not usually turn on the results of a single study.

Rollo Tommasi said...

Ann, I certainly agree that we cannot and should not try to control every potential risk out there. But we can and should take reasonable steps to reduce risks, particularly if the risks are either high or potentially very serious. How do we decide that? We need to look at a number of factors:

How great is the increased relative risk for the individual? In the case of passive smoking, the additional risk to a specified individual is relatively modest.

How many people are exposed to this additional risk? In the case of passive smoking, high numbers of people have been exposed, although levels of exposure have been lowering in recent years as more smokers either quit or change their smoking habits.

Potential risk of what? In the case of passive smoking, it couldn’t be any more serious. Death – particularly from lung cancer or heart disease. Also certain other potential risks, including respiratory diseases and difficulties, and deaths from stroke and SIDS.

How many people succumb to this result? Substantial numbers. Around 130,000 people in the UK will die each year from either heart disease or lung cancer.

The typical pro-smoking argument is to focus on the first of these points alone – how much extra risk to an individual. However, you have to ask yourself all 4 questions. Doing so helps to explain why Jamrozik calculated that around 6,600 of UK deaths each year from lung cancer or heart disease are attributable to passive smoking.

Can we not agree that 6,600 deaths each year is an unacceptably huge number? And can we not also agree that simply asking smokers not to smoke when they are in enclosed public spaces is a reasonable step to reduce these risks, which does not place a major burden on most smokers?

Rollo Tommasi said...

Belinda – I was not “name-calling”. I was attacking your arguments, not you personally. And I think it is highly hypocritical for you to accuse me of name-calling, when in a recent exchange you referred to me as “the mass debater”!

And, true to form and as I highlighted in my last post to you, you still don't substantiate your claim that there are no "real" deaths attributable to passive smoking, becuase you have nothing to substantiate that bogus claim with.

Gary K said...

"Can we not agree that 6,600 deaths each year is an unacceptably huge number?"

The 6,600 deaths is an imaginary number and has two HUGE problems:

1) There are NO deaths that have actually been proven to be caused by SHS exposure. Out of 6,600 you should be able to dig up 5 names of REAL people.

2) The 6,600 deaths include those that would have died had there been NO SHS exposure and that number may well be 6,600.

Rollo Tommasi said...

As for Belinda’s second post, I have not tried to argue that individual results straddling 1.0 can be statistically significant. My point has been to crush the myth that a statistically insignificant result means “no risk”. In fact it means “it is not clear from these results whether there is an additional risk or not”.

When you have scores of studies, most of which are weighted above 1.0, that collective evidence suggests it is likely that there may be additional risk.

In the case of passive smoking, meta-analyses have been used to provide more clarity. Their findings? That passive smoking does increase the risk of lung cancer and heart disease – through statistically significant results .

e.g.
Lung cancer
Taylor et al (2007): 1.27 (95% CI 1.17–1.37)
SCOTH (2004): 1.24 (95% CI 1.13-1.36)

Heart disease
US Surgeon General (2006): 1.27 (95% CI 1.19–1.36)
He et al (1999): 1.25 (95% CI 1.17- 1.32)

As for Belinda's last post, I simply refer her to my last post to Ann. Belinda just looks at one question, and ignores the other 3 questions I list.

Rollo Tommasi said...

Gary:

1. That is a fatuous argument. I challenge you to find any respected scientific or medical expert who would make that point. If you took that line, then we would not consider there to be any risk of lung cancer from vehicle pollution or radon gas.

2. Wrong. See my post of 18 November 2010 23:04.

Gary K said...

Rollo,

1)We are not talking about vehicle pollution or radon. Engstrom & Kabat's study as well as others have shown NO risk of death from lung cancer due to SHS exposure.

2)Risk is NOT the same as probability.

A 20% increased risk(RR 1.2) means that the factor being considered has ONLY a 17% probability of being causative(1 is 83% of 1.2).

When applied to large numbers of people it is total fantasy to state that an RR1.2 implies that 20% of the people have a 100% probability of that factor being causative when each and everyone of them has an 83% probability of that factor NOT being causative.

Rollo Tommasi said...

Gary K: I’m afraid to say you’re both talking nonsense and completely ignoring points I made to you earlier in this thread, which you chose not to respond to at the time.

1. I see you are unable to find a respected scientific or medial expert who’ll back you up on your earlier claim. Radon gas and vehicle pollution are relevant, because your earlier post tried to nullify any scientific understanding based on epidemiological studies of populations, instead of death certificates of individuals. Our awareness of the risks of radon gas and pollution – like SHS - in relation to lung cancer are based on epidemiological evidence.

2. You continue this bizarre argument that most people who die of lung cancer or heart disease do not do so due to passive smoking. Who’s suggesting otherwise? The 6,600 deaths attributable to passive smoking, based on best available evidence, may be a small proportion of the 130,000 lung cancer and heart disease deaths overall in the UK each year. But 6,600 is still a huge number of deaths – all of which are preventable over time if people take reasonable steps to avoid subjecting other people to SHS exposure.

If you think we should not concern ourselves with these 6,600 deaths, then you truly are heartless.

Rollo Tommasi said...

Gary - Also meant to say: Enstrom & Kabat did not show "no risk of death from lung cancer due to SHS exposure."

Ann W. said...

Rollo said "The 6,600 deaths attributable to passive smoking, based on best available evidence."

Are you stating that Konrad's guessimations are the best available evidence? Have you even read his paper?

Konrad makes so many assumptions and invents so many figures to come up the 6,600 figure that one has to take a huge leap of faith that everything he states is factual and that he has not made any errors (see my comments from 19 November 2010 13:38 re:"Projected Smoking-attributable Mortality in Canada, 1991-2000")

Ann W. said...

"Ann: Firstly, you did not explain the point you were trying to make with reference to that Classical Kaposi’s Sarcoma study. Even now I’m not fully sure."

Rollo, my point was and still is how silly and meaningless these studies can be, unless you have an agenda.

I really wasn't expecting to see that smoking was going to show a protective effect or that infrequent bathing was a risk for KC when I googled Sarcoma.

Did you know that Sarcoma is rare in adults, accounting for only 1% of all adult cancers.

Now your comment "After all, that reported risk relates to active smoking, not passive smoking" does raise my curiousity.

Aren't people who smoke the most exposed to passive smoke?

Gary K said...

"epidemiological evidence"= a lot of junk

HAH, here is how that it gets perverted.

Say the death rate from lung cancer for never-smokers NOT exposed to SHS is 4/10,000.

Now we will say a study is done that shows the lung cancer death rate for never-smokers exposed to SHS is 5/10,000.

You and your 'experts' will claim that all 5 of the exposed peoples deaths were due to SHS; when in fact, 4 of them would have died anyway!

4 lung cancer deaths that are claimed to be due to SHS would have an 100% probability of being 'caused' by factors other than SHS and 1 would have an 80% pprobability.

Only the very corrupt would claim that a 20% probability is adequate to PROVE causality.

At least 80% of your 6,600 SHS 'caused' deaths have a 100% probability of being caused by other factors and the rest have at least an 80% probability of NOT being caused by SHS.

Rollo Tommasi said...

Ann - I have read Jamrozik's paper.

If you are so critical of it, you're welcome to share your reasoning.

Sorry, but arguing that Jamrozik must be flawed because you think another study is flawed does not cut it.

You say "Did you know that Sarcoma is rare in adults, accounting for only 1% of all adult cancers." The answer is yes. Which is why I say you don't look only at the RRs in epidemiological studies. You have to take other factors into account. Passive smoking is a significant public health hazard because it increases the risk of common causes of death (lung cancer and especially heart disease).

Rollo Tommasi said...

Gary - Your last post is unsubstantiated homespun rubbish.

As I have already pointed out, deaths from other factors were already filtered out before the 6,600 figures of deaths attributable to SHS was reached.

Ann W. said...

Rollo, you didn't address

"After all, that reported risk relates to active smoking, not passive smoking" does raise my curiousity.

Aren't people who smoke the most exposed to passive smoke?

Belinda said...

Rollo (MD) says: 'When you have scores of studies, most of which are weighted above 1.0, that collective evidence suggests it is likely that there may be additional risk.'

So if you want a positive policy outcome just pay for lots of studies reaching insignificant results and spin a story about how it makes them significant.

I can't see anything in Jamrozik's study accounting for other factors than secondary smoke. All I can see is guesses about almost everything in sight, including the level of passive smoke exposure at 'home' or at 'work', with absolutely no indication of how intense the exposure was either at work or at home, how big the rooms were, how good the air conditioning, how prolonged or heavy the exposure was. There is no certainty about anything in the study, yet it is used as the basis of conclusion and speculation.

Ann W. said...

"I have read Jamrozik's paper. If you are so critical of it, you're welcome to share your reasoning."

My criticism is that Jamrozik has used far too many assumptions and estimates that require a leap of faith that he has been honest, made no errors, his assumptions and estimates are correct and he is not working towards an agenda.

In order to use a percentage based population attribution like he has, one has to assume that everyone and everything is homogeneous
• As there are no equivalent estimates available for passive exposure at work, I applied the same figures for that setting
• I obtained counts of deaths from lung cancer (International Classification of Diseases, 10th revision, codes C33, C34), ischaemic heart disease (I20-I25), and stroke (I60-I69) in England and Wales during 2003 from the Office for National Statistics27 according to age ( < 65 or ≥ 65). I adjusted the figures upwards by 12.9% and 12.4%, respectively,8 to obtain UK estimates.
• I derived estimates of the cause specific numbers of deaths for both sexes combined for people aged < 65 and ≥ 65 and for each site of exposure (home or work), using the formula {[p.(RR − 1)]/[1+p.(RR − 1)]} for population attributable proportion (where p = prevalence of passive smoking at home and RR = relative risk) and applying the resulting fraction to the relevant total numbers of deaths from a specific cause.
• I assumed that all employees in the hospitality industry were exposed to the same amount of passive smoking at home as the rest of the population.
• The proportion of staff for whom hospitality work represents their chief lifetime occupation is not known but I estimated it at 20%.
• Though the data relating to passive smoking and cerebrovascular disease remain limited, I have accommodated this potential objection by recalculating the figures on the assumption that the relative risks for stroke and ischaemic heart disease are the same.
• In partitioning deaths in the whole population to those occurring within and outside the workforce, no allowance has been made for a healthy worker effect. Arguably, both of these factors might contribute to an overestimation of deaths caused
by passive smoking. On the other hand, I have omitted severe morbidity from vascular disease in individuals who might have not developed these conditions had they been able to avoid being exposed to tobacco smoke.

Michael J. McFadden said...

Why hello again Rollo. It's been a while.

You wrote, "The truth is that epidemiological science expects that a relative risk needs to be 2 or more ONLY if you are trying to reach a conclusive finding from the results of a SINGLE study. It does not apply to relative risks established from a large body of evidence, as is the case with passive smoking, in which case smaller relative risks ARE relevant."

You'll note the part I bolded Rollo. I'm sure you'll be nice enough to cite three or four examples of such "proven" risks that are widely accepted out there in the scientific world along with the body counts that have been computed for them.

After all, this is the "normal" way you say science is conducted, true?

Michael J. McFadden

Rollo Tommasi said...

AnnW (1st post): There are 2 issues. The risks of active smoking are likely to include elements which someone subjected to passive smoking would face (breathing in smoke from the burning end of the cigarette), and other elements from what they inhale through the other end.

But there is also a difference in that risks of active smoking are based on the smoker’s own actions and behaviours. Risks from passive smoking are based on the actions and behaviours of other people which impact on those around them.

Rollo Tommasi said...

AnnW (2nd post): I’ve looked through your examples. It seems to me you are questioning his model just because it includes some assumptions (which are inevitable in any epidemiological study). More than this, however, you simply assume that he must have been working to some kind of agenda, even though you offer no evidence to suggest that his model would artificially inflate the actual number of deaths. Turning to your examples:

• As there are no equivalent estimates available for passive exposure at work, I applied the same figures for that setting - that only applied to Jamrozik’s calculations of stroke deaths, which are not part of the 6,600 deaths figure I’ve quoted
• I obtained counts of deaths from lung cancer (International Classification of Diseases, 10th revision, codes C33, C34), ischaemic heart disease (I20-I25), and stroke (I60-I69) in England and Wales during 2003 from the Office for National Statistics27 according to age ( < 65 or ≥ 65). I adjusted the figures upwards by 12.9% and 12.4%, respectively,8 to obtain UK estimates. - referring to official statistics to get total numbers of deaths? Whatever is wrong with that?! His calculation for determining UK figures from English stats is also reasonable
• I derived estimates of the cause specific numbers of deaths for both sexes combined for people aged < 65 and ≥ 65 and for each site of exposure (home or work), using the formula {[p.(RR − 1)]/[1+p.(RR − 1)]} for population attributable proportion (where p = prevalence of passive smoking at home and RR = relative risk) and applying the resulting fraction to the relevant total numbers of deaths from a specific cause. - This has no bearing on the 6,600 figure – only what proportion of that figure are aged < 65 and ≥ 65. Besides, it seems a reasonable calculation to make.

Rollo Tommasi said...

• I assumed that all employees in the hospitality industry were exposed to the same amount of passive smoking at home as the rest of the population. - Again, this has no bearing on the 6,600 figure – the hospitality employee figures were separately calculated and then treated as a sub-set of the 6,600 total. Besides, it seems a reasonable calculation to make.
• The proportion of staff for whom hospitality work represents their chief lifetime occupation is not known but I estimated it at 20%.- - Once again, this has no bearing on the 6,600 figure – the hospitality employee figures were separately calculated and then treated as a sub-set of the 6,600 total.
• Though the data relating to passive smoking and cerebrovascular disease remain limited, I have accommodated this potential objection by recalculating the figures on the assumption that the relative risks for stroke and ischaemic heart disease are the same. setting - As your first bullet, that only applied to Jamrozik’s calculations of stroke deaths, which are not part of the 6,600 deaths figure I’ve quoted
• In partitioning deaths in the whole population to those occurring within and outside the workforce, no allowance has been made for a healthy worker effect. Arguably, both of these factors might contribute to an overestimation of deaths caused by passive smoking. On the other hand, I have omitted severe morbidity from vascular disease in individuals who might have not developed these conditions had they been able to avoid being exposed to tobacco smoke. - Once again, I don’t know what your problem is. Highlighting potential confounding factors is part and parcel of any epidemiological study. Any inaccuracies might as well under-estimate actual death figures and over-estimate them. Besides, there is absolutely nothing to suggest that these factors would be so important that they would reduce the huge 6,600 figure to an insignificantly small total number of deaths.

Rollo Tommasi said...

So now Belinda, having been called for the hypocrisy of her name-calling allegation, can’t resist a little bit of name-calling of her own again. You couldn’t make it up!

If passive smoking constituted no risk, you would expect to find as many studies concluding with RRs <1.0 as >1.0. And you certainly wouldn’t find statistically significant findings of additional risk from meta-analyses.

Jamrozik’s figures for lung cancer and heart disease deaths are based on respected relative risk figures, which reflect typical levels of exposure to passive smoking amongst those so exposed. There was no need for Jamrozik to separately assess levels of exposure for individuals.

Belinda, let me return to a question I keep asking you and you never reply to. What is your point in questioning the accuracy of Jamrozik’s figures? How much lower do you think is 6,600 deaths figure would have to be before you would consider it an insignificant number not worthy of concern? And how would you show that any inaccuracies in his figures were large enough to mean that the real figure was low enough not to warrant concern?

Rollo Tommasi said...

Hello again Michael. Nice to have you ask questions of me.

I really don’t know why you’re asking me to prove that an RR of 2 or more is not required where there is a large body of evidence. After all, the only arguments I’ve seen that an RR of 2 or more is always required are based on lies and misquotes. So, if you believe that I am wrong and that an RR of 2 or more is always required, please show your evidence.

But I’m also not going to duck your question. This article gives a couple of examples (re intravenous streptokinase and beta-blockers) where actions have been demonstrated over time to be clinically significant even where the RR was neither >2 or <0.5: http://www.bmj.com/content/315/7119/1371.full?ijkey=56e76871c0af23ff41bc08edb84a947f344402a0&keytype2=tf_ipsecsha

Pogo said...

@Rollo... 20 November 2010 14:02

Please don't make assumptions and by so doing insult my intelligence and integrity. I'm as much influenced by "the lies of pro-smoking groups" as I am by the lies of the anti-smoking lobby, ie not at all. I call it as I see it, based upon a mixture of common sense and a working life spent in "hard" science. I realise that when all is said and done, the "tobacco debate" left the world of science and descended into the nether regions of politics many years ago, but, as I've said before, I'm a lifelong non-smoker so I don't have a dog in this race, merely a concern that "science" not be manipulated and misstated in order to forward an agenda.

I'm afraid that your statements about RRs of less than 2 in epidemiology appear to be completely at odds with academic standards. Whilst checking I ran across the following:- "Breslow and Day, 'Statistical Studies in Cancer Research' Vol: 1 'The analysis of case-control studies', IARC (ie W.H.O.) Scientific Publications 32 (Lyon 1980)" - which appears to be the standard text work on the interpretation of statistics in cancer research - and on page 36 states "relative risks of less than 2.0 may readily reflect some unperceived bias or confounding factor, those over 5.0 are unlikely to do so".

As I said earlier, although RRs of less than 2.0 might be taken into account in a double-blind trial they are essentially meaningless in the case of observational studies based upon meta-analyses or data dredges. Combining the results of a group of non-significant studies does not magically make the result significant. Full Stop.

Michael J. McFadden said...

Thank you for responding Rollo, and you did indeed provide some examples outside of ETS exposure where the cumulative method was used to support small RRs. In the cases cited in that article however the scientists were dealing with cases of fairly well-defined clinical trials where there was no pressure to find results tending in one direction or the other. In the case of ETS studies however there are at least three clear pressures brought to bear on researchers:
'
1) Publication pressure: I believe most would feel that they are more likely to have their research accepted for publication if their results point to ETS being evil rather than good -- a pressure unlikely to be felt around such non-politically-sensitive issues as beta blockers and streptokinase.
'
2) Ideological pressure: Few researchers doing research on those latter issues would have any strong bias of moral desire toward the outcome, while I think it's fair to say that MOST researchers on ETS/LC would have the desire that their results support efforts to reduce the "disease and death caused by smoking."
'
3) Financial pressure: If the researchers found one way or the other for the latter issues it would be unlikely to influence their chances much in terms of getting future grants. Producing a study saying that ETS exposure has a good effect on people however could make getting future grants somewhat more problematic.
'
All three of those forces would work toward pushing the results in a particular direction, something that cannot be said about such things as beta blocker and streptokinase research. And before you accuse me of impugning the good names of any researcher, let me point out that there are many researchers who conducted those studies and it would only take a few biased results to move the final result into nonsignificant territory. And those few wouldn't even have to be due to overt fraud: remember why there's an importance attached to double-blind experimentation. Even the most innocently introduced bias is still bias.
'
When you add that together with the point clearly made by other posters here about the difference between simple statistical significance and true causality then the idea of assigning some imaginary number of deaths becomes quite questionable, particularly when you yourself indicate that even "When you have scores of studies, most of which are weighted above 1.0, that collective evidence SUGGESTS it is LIKELY that there MAY BE be additional risk." (emphasis added)
'
I'd agree there's a "suggestion" out there Rollo, but that "suggestion" comes quite largely from studies not only afflicted with the problems above but also from studies largely based on situations where the smoke concentrations often likely ranged up to 200% or more of what you'd commonly find today.
'
- MJM

Michael J. McFadden said...

btw Chris, don't let Rollo make you feel too bad when he says "you’re still cherry-picking your evidence Chris!" The last time he stopped in to visit me on a forum he led off with this:
'
"I see you’re at it again. Trying to paint a picture about the science by referring a couple of completely cherry-picked, unrepresentative and irrelevant anecdotes. ... And you also don’t mention how you just cherry-pick one or two studies in order each your distorted conclusion"
'
I then repeated a question I had asked him in the past:
'
"Name three studies that you are prepared to defend and where the data is publicly accessible for verification in some form where any level of significant risk of harm to health from the levels of smoke that would commonly be encountered in bars and restaurants today has been shown to be harmful to the general public. I'm quite happy to let YOU do the "cherry picking" from the mountain of studies you believe support your position and then defend them."
'
and, not unexpectedly, he disappeared.
'
- MJM

Ann W. said...

“Besides, it seems a reasonable calculation to make.”
What can I say Rollo, if you are comfortable with the paper and all of Jamrozik’s assumptions and estimates and I am not, then there really is no point in continuing our discussion.
But I am curious where you came up with the 6,600 deaths attributed to passive smoking. I must have missed that number in Jamrozik’s paper.
“Passive exposure to tobacco smoke at work may cause more than 600 deaths each year in the UK, including over 50 in people employed in the hospitality industry”
“Exposure at home may account for 2700 deaths in those aged 20-64 and 8000 in those aged ≥ 65”

Doesn’t that equal 10,700?

Ann W. said...

In Canada the most current figures for deaths attributed to passive smoking is calculated at 831.
(Lung cancer – 252 and Ischaemic heart disease – 579)
15-29 yrs Lung cancer – 0 and Ischaemic heart disease – 1 = 1
30-44 yrs Lung cancer – 4 and Ischaemic heart disease – 8 = 12
45-59 yrs Lung cancer – 55 and Ischaemic heart disease – 73 = 128
60-69 yrs Lung cancer – 78 and Ischaemic heart disease – 97 = 175
70-79 yrs Lung cancer – 83 and Ischaemic heart disease – 169 = 252
80+ yrs Lung cancer – 34 and Ischaemic heart disease – 231 = 265
Table I-S-1. Smoking-attributable fractions, mean age at death, and number of deaths attributable to tobacco by sex, age and disease category in Canada, 2002

And as you can see this is a far more detailed and complex calculations then what Jamrozik offers. And because the data used is available it is easier to see if they have made any mistakes in their assumptions and estimates, which there is.
If they can’t even get something as simple as their adding to work, how can one rely on anything else they present?
As the boomer generation, one of history’s largest cohorts, enter into their senior years, the number of deaths will natural increase proportionate to the size of that population and by using death attribution calculations to risk, it would make it appear as if an epidemic has occurred.

Rollo Tommasi said...

How strange Michael. I show you how epidemiology accepts relative risks of <2 and >0.5 in other circumstances. What’s your response? To claim that results from passive smoking studies might possibly be that way because of bias and pressures upon researchers. Any evidence or examples for this view? Absolutely not. Only a plea that readers should trust the “unbiased” views of Michael J McFadden (Author of “Dissecting Anti-Smokers’ Brains”). That would be an absolutely comical stance, were it not for the fact that you are challenging measures designed to save lives on the basis of such unsubstantiated hogwash.

I also note that you still are unable to provide any evidence to back up claims that a relative risk of 2.0 or more is always required.

As for your final comments, I don’t deny that the average level of exposure to SHS is declining, due to a combination of reduced smoking levels, more considerate practices by smokers and of course the smoking laws themselves. But that does not mean there is no need for the laws to remain, in the same way as very few people would advocate the reintroduction of asbestos into construction materials because fewer people are dying of asbestosis nowadays.

Rollo Tommasi said...

By the way, MJM, I’m delighted you’ve brought up the subject of that previous discussion (from the “Bonjour Paris” website). I actually tried to reply on 3 or 4 occasions, but unfortunately they were not added to the thread. You’ll be delighted to know that I actually saved my response, so I could copy and paste it into the thread. And here it is on the following message……..

Rollo Tommasi said...

Bonsoir Michael. Thank you for responding. As before, I am more than happy to take you up on your challenge.

You want 3 publicly accessible reports? I’ve already given you two: The IARC report of the 2004 and the US Surgeon General’s report of 2006. You can add to that this 2007 assessment of available evidence on the risks of lung cancer from passive smoking: http://ije.oxfordjournals.org/cgi/reprint/dym158v1

These reports are thorough. They take account of most published studies available at the time of analysis, and not just a small number of cherry-picked examples. Their analysis is robust. And they take account of exposure to second-hand smoke in a range of locations (as you know, second-hand smoke does not become either any more or any less toxic because it is in a pub rather than in a private dwelling).

If you want to challenge either these reports or the underpinning studies on which they are based, I am more than happy to consider your critique.

PS I note how you avoided answering my point about how the Helena study has had virtually or absolutely no influence in decisions to introduce smoking laws. The Helena study is one of a small number of recent studies which look at SHORT-term risks to health from passive smoking. The smoking laws have been introduced on the back of long-standing, established and thorough evidence showing LONG-term risks to health from passive smoking.

It is quite deceitful of you to try to confuse the two.


[RT response to MJM in Bonjour Paris]

Rollo Tommasi said...

Ann - I remain content with Jamrozik's conclusions. i note you have no response to the replies I gave to your earlier points. It seems to me you have no grounds on which to criticise Jamrozik's work, let alone argue that his results are so extraordinarily inaccurate that the true number of deaths is actually insignificantly low.

I have consistently quoted the 6,600 deaths figure, which is a combination of the total number of deaths Jamrozik estimates from lung cancer and heart disease. The 10,700 figure also includes an estimated number of deaths from strokes attributable to passive smoking. While there is good reason to believe that passive smoking can contribute to strokes, the evidence until now has been less strong than for lung cancer and heart disease. Personally, I prefer to exclude the stroke figures - 6,600 deaths from lung cancer and heart disease is already a horrifically high number without adding in stroke figures too.

Michael J. McFadden said...

Did I provide evidence or examples of bias in the large epidemiological studies? Of course not: there's obviously no way to get the resources to investigate the basic data for these studies and I haven't examined enough of them in enough detail to comment more generally. What I *did* provide was clear indication of motivation, and a clear indication that the motivation would be far more powerful than the simple motivation of money which has recently been noted all by itself to corrupt 15% or more of drug research that is conducted. If those estimates are correct, the additional motivations in the ETS field would likely place the bad research up around 20% or even higher.
'
Do I ask readers to trust my "unbiased" views? Not at all. Which is why I laid out the reasoning behind them. The readers can judge for themselves whether it's likely that my reasoning indicates the likely existence of enough bias to have a significant influence on your meta-analytical results. I believe it does.
'
And you know perfectly well that the measures are not "designed to save lives," at least not nonsmokers' lives. The measures are designed to reduce smoking, which is why no reasonable alternatives such as comfortable smoking lounges that employees don't have to enter and separate exhaust are considered. And don't get silly about employees being poisoned by "third hand smoke" while cleaning Rollo -- that should be beneath even you.
'
You then make it sound like I've claimed RR >= 2.0 is "always required." I believe you know perfectly well that I've never claimed that and that is why you worded it that way -- insinuations are always better than direct charges when you don't have the back up, eh?
'
Finally you claim to have "taken up my challenge" by providing the names of several "reports." Of course you did NOT take up my challenge, since the entirety of it clearly read exactly as follows:
'
====

Name three studies that you are prepared to defend and where the data is publicly accessible for verification in some form where any level of significant risk of harm to health from the levels of smoke that would commonly be encountered in bars and restaurants today has been shown to be harmful to the general public. I'm quite happy to let YOU do the "cherry picking" from the mountain of studies you believe support your position and then defend them.


Remember though: we're talking about actual scientific studies: not reports, statements, opinions, fact-sheets, suppositions, reviews, etc etc. No one can mount a reasonable argument against a report that summarizes hundreds of studies except by showing that the individual studies themselves are made out of sand.

====
'
So unless you were particularly visually handicapped that day, you would have noted the statement I have bolded here for extra emphasis. You most certainly did not answer my challenge. You also noted that "as before" you were happy to take me up on the challenge. Did you make the same error the previous time? Odd that I don't remember -- care to cite that previous time Rollo?
'
- MJM

Ann W. said...

"Ann - I remain content with Jamrozik's conclusions. i note you have no response to the replies I gave to your earlier points"

No Rollo, you only address what you wanted to address. no comment on this?
"In order to use a percentage based population attribution like he has, one has to assume that everyone and everything is homogeneous"

and then you only stating "it seems a reasonable calculation to make." without qualifying why you felt that way.

We could be discussing if God existed and you could have written the same statement but that takes a leap of faith in both situations and you have not shown why I should have any faith in someone that has an agenda, be it in religion or smoking attributed deaths. And Jamrozik does have an agenda.

Sorry Rollo, there are far too many cause of IHD in seniors that someone can state with any certainly, by using a simply mathematical exercise, that those deaths were caused by passive smoking.

Ann W. said...

The calculations presented in this paper were commissioned by
SmokeFree London, a collaboration of all 33 local borough councils in
London concerned with extension of smoke-free policies in that city. I
received no payment or any other consideration for conducting them.
These data were originally presented at a conference on passive smoking in
the hospitality industry organised by the Tobacco Advisory Group of the
Royal College of Physicians in May 2004, and are published here with permission
of SmokeFree London.
Contributors: KJ is the sole author.
Funding: None.
Competing interests: None declared.
Ethical approval: Not required.

Rollo Tommasi said...

And, without going off at a tangent, separate smoking areas have not been demonstrated to adequately protect staff and punters.

You’re a bit miffed that I should expect you to point to evidence showing that a RR of 2.0 or more is always required. I believe I am entitled to, as you asked me to point to contrary evidence (which I did), even though nobody has been able to point to valid evidence that I might be wrong.

As for the Bonjour Paris challenge, the studies I referred to ARE “actual scientific studies” – they were undertaken in accordance with proper professional scientific principles.

I actually agree that “No one can mount a reasonable argument against a report that summarizes hundreds of studies except by showing that the individual studies themselves are made out of sand.” But if the pro-smokers’ concerns about the validity of SHS research were sound enough, then those essential criticisms should already be available. So where are they?

I’ve never suggested that any one or two individual studies are enough to demonstrate that passive smoking is harmful. That comes from the evidence collectively. So I’ll leave the cherry-picking to you and Chris.

By the way, after this time and without checking back, I can’t remember what our previous debate had been.

Rollo Tommasi said...

AnnW: Jamrozik's figures are estimates and relate to a particular point in time. Nobody is suggesting an annual number of deaths of exactly 6,600 from lung cancer/heart disease. And certainly changes in the age profile of the population, as well as changes in the level of exposure to SHS, can affect actual death rates over time. But Jamrozik's work gives a very good sense of the scale of deaths attributable to passive smoking.

That estiamte can be calculated in a range of ways. Jamrozik did so one way; the study you referred to in another. Each has its advantages and disadvantages.

But I keep coming back to this point, which you continue to ignore: you are simply not providing any evidence that Jamrozik's results are so extraordinarily inaccurate that the true number of deaths is actually insignificantly low.

Pogo said...

Rollo...

"Breslow and Day, ..."...?

No comment???

Rollo Tommasi said...

Pogo - I did reply. But I've been having ongonig problems getting posts to stick and I didn't notice mine to you disappear.

Here's a hasty response afresh. Bottom line - I believe you're misinterpreting B&L. They did not set a hard and fast rule that all studies must show a RR of 2 or more. They actually said studies should be interpreted using 3 criteria:

1. Strength of evidence - a RR of <2 may give rise to possible bias.

2. Consistency of evidence. Results which are consistent with previous studies are more likely to be accurate.

3. Biological plausibility of evidence.

Passive smoking studies may not be strong for (1), but they are strong for (2) and (3). There are scores of studies showing similar levels of additional risk. And there is biological plausibility, in that evidence shows a dose-relationship between exposure and risk of harm, and the similarity to known harm from active smoking.

Rollo Tommasi said...

Pogo: As for your other points....

I stand by my statement about the lies of pro-smoking groups. If you had read the quotes about RR from original sources, it would have been clear to you that Angell and Temple were not stating that a RR of >2 is always required. Instead, you followed lies used by pro-smoking groups.

And you are just inventing your own scientific theory when you claim "Combining the results of a group of non-significant studies does not magically make the result significant". Meta-analyses are a proper and respected way of finding overarching results from a collection of studies, as this article points out: http://www.bmj.com/content/315/7119/1371.full?ijkey=56e76871c0af23ff41bc08edb84a947f344402a0&keytype2=tf_ipsecsha

Meta-analyses consistently show that passive smoking increases the risks of heart disease and lung cancer.

Ann W. said...

I have ignored nothing Rollo.

To address your responds that I have not provide any evidence that Jamrozik's results are so extraordinarily inaccurate (nice wording), is insulting.

Jamrozik's paper is a hypothetical conclusion using mathematical calculations based on assumptions and estimates.

The onus is not on I to provide it wrong, it is on you to prove it factual.

"Passive smoking produces elevated risks for CHD that are similar to low-level active smoking (e.g., about one cigarette per day). Glantz and Parmley have suggested that the reason active smoking does not produce higher (i.e., linear) dose-response results is because the effects of cigarette smoke on the heart may reach a saturation point, making a monotonic dose-response effect unlikely. "
http://www.phac-aspc.gc.ca/publicat/cdic-mcc/23-1/b_e.html

Michael J. McFadden said...

Rollo, as you well know, separate smoking areas as I defined them -- a separate room with exhaust ventilation -- render the nonsmoking areas perfectly safe in terms of any real world consideration. The level of exposure would be on the order of 1/1,000th as much as having all the smokers crowded into the same room and using ventilation levels similar to those used in the 1950s through 80s when the base exposures for the US EPA meta-analysis were based.
'
The US EPA figures indicate roughly one extra lung cancer for every 40,000 worker-years of exposure (based on a base nonsmoker LC rate of about four in a thousand with a 19% increase). If my 1/1,000th estimate is correct then we'd see one extra lung cancer for every forty MILLION worker-years of exposure in the nonsmoking room. If my estimate is off by an entire order of magnitude we'd still see only one per every four million worker-years of exposure. And again, even that is only true if we accept the EPA estimate as correct.
'
I was miffed that you implied I had made the claim about 2.0 being always required, not that you asked if I had sources for it. I never claimed it and therefore have no sources for it. I would however agree with the epidemiologists who feel it is a good rule of thumb due to the inherent difficulties of controlling for variables in population based long-term epidemiological studies.
'
As for the challenge, don't pretend to be thick headed Rollo: I very specifically asked for "studies" and even specifically defined such as NOT including generalized "reports." So you clearly did NOT respond adequately when you chose to cite three "reports" instead of what I asked for. A report is not a study, and you know the difference perfectly well.
'
- MJM

Klaus K said...

Even if a significant result is obtained by pooling all 80 published studies on ETS & lung cancer in a meta-analysis, there is still no proof of causation. Statistician Peter Lee is tracking all the studies for the tobacco industry - his paper on ETS and lung cancer:

http://www.pnlee.co.uk/documents/refs/lee2010B.pdf

The pooled risk 1.22 - CI (1.13-1.31) is a very small risk, as Brenner 2010 reminds us: The RR of lung cancer from exposure to various workplace hazards is 10 times higher in the Brenner study.

Furthermore as Peter Lee notes, 68 of the 80 ETS-studies (85%) showed no significant risk. So we have a 85% majority of studies that do not show any association between ETS and lung cancer. Despite this fact the anti-smoking lobby is claiming that not only association is proven, but also causation. That is a leap into fantasy.

Four of the five biggest studies showed no significant risk. The biggest study even showed a significantly negative risk. This means, that the many biased small studies are forcing the pooled CI above 1.0. Furthermore most of the studies suffer from biases like misclassification, many uncontrolled confounding factors (20% of the studies did not adjust for age) and evidence of considerable publication bias.

A big doubt also arises when asking: what is measured? Is it possible at all to measure smoke in peoples lives by asking them? People are different. What is actually measured is how much smoke people remember being exposed to.

Considering these many biases in the studies and the small reported risks, it is more than reasonable to say, that no risk is proven. In Peter Lee's words: Taken as a whole, the epidemiology does not support the claim that ETS causes lung cancer in non-smokers.

Estimating number of deaths from the RR's in these studies must be considered a joke.

Rollo Tommasi said...

AnnW: Jamrozik’s work was not “hypothetical”. It was a practical translation of the established additional risk of death as a result of passive smoking into actual numbers of deaths in the UK each year.

I still don’t understand the point you are trying to make. Are you trying to suggest that the actual number of deaths attributable to SHS is less than Jamrozik estimated? If so, how much less? Or are you trying to deny that any deaths at all are attributable to SHS? It seems to me you are trying to cast any doubt you can on his work, in a desperate effort to pretend that passive smoking might not cause any deaths.

Jamrozik’s work has been published and is referred to by scientific professionals. I am not aware of any substantive evidence-based challenges that have been made to the thrust of his findings (although it is recognised that the actual numbers of deaths may be a bit higher or lower than he estimated). In the absence of specific and substantive challenge, Jamrozik’s work needs no further justification.

And I am not sure about the point you are trying to make with the study you refer to at the end.

Rollo Tommasi said...

Michael: Separate smoking rooms are not the answer. That’s not my view. That’s the view of publicans themselves, who can’t agree on what they want. It’s also not what Freedom2Choose is campaigning for. Separate smoking rooms are not an option for most of the small community pubs which have suffered most in recent years from a number of factors, owing in large part to physical and cost constraints.

As for the RR claim, I don’t know why you were miffed. You asked me to give evidence that a RR of 2.0 or more is not always required. Isn’t it reasonable to assume that you’d only make that demand if you believe that it is?!

As for the challenge, I am not pretending to be thick-headed. I’m being perfectly serious. I don’t accept your premise that you can cherry-pick a few studies for scrutiny, because I have never suggested that any studies would be conclusive on their own. My views are based on the holistic evidence. But you are cherry-picking what you are allowing me to cherry-pick! As I said before, if the pro-smokers’ concerns about the validity of SHS research were sound enough, then those essential criticisms should already be available. So where are they?

Rollo Tommasi said...

KlausK: I disagree.

You insist on trying to suggest that a non-statistically significant result means there is “no association”. It doesn’t. It means that an association has not been demonstrated in statistically significant terms. But that is very different from evidence that there is no association at all. Meta-analyses of these studies demonstrate that there the additional risks of lung cancer and heart disease from SHS ARE significant.

I’m curious about your claim that “the biggest study even showed a significantly negative risk”. I thought the biggest lung cancer study was by Fontham et al (653 lung cancer cases), which found a statistically significant positive risk.

Your point about recollection is valid to an extent, in relation to case-control studies. However, it is less of a problem re cohort studies. And the results of cohort studies into passive smoking are very similar to those of case-control studies, indicatibng that recollection of exposure does not materially affect results (see Taylor et al (2007)).

I’ve been quoting the combined wisdom of the US Surgeon General, the IARC and SCOTH, supported by most medical professional bodies worldwide. You are quoting Peter Lee. Why should I believe his word over theirs?

Gary K said...

Rollo says:
"I’ve been quoting the combined wisdom of the US Surgeon General, the IARC and SCOTH, supported by most medical professional bodies worldwide."

Indeed and all of them will quote studies to prove their points and ALL of them will state that smokers have a risk for lung cancer that is 23 times that of never-smokers.

The US CDC states that never-smokers account for 18% of the 157,000 yearly lung cancer deaths or 28,260 cancer deaths.

Thus, ALL of those 'experts' are stating that smokers have 649,980 lung cancer deaths every year.

That is ONLY 4.14 times the actual number of lung cancer deaths that actually occur!!!!

The 'experts' and their studies are full of CRAP!!!!

Ann W. said...

"AnnW: Jamrozik’s work was not “hypothetical”. It was a practical translation of the established additional risk of death as a result of passive smoking into actual numbers of deaths in the UK each year.”

Rollo since you chose to ignore my comment "In order to use a percentage based population attribution like he has, one has to assume that everyone and everything is homogeneous"

Maybe you will understand this then

“An ecological fallacy (or ecological inference fallacy) is an error in the interpretation of statistical data in an ecological study, whereby inferences about the nature of specific individuals are based solely upon aggregate statistics collected for the group to which those individuals belong. This fallacy assumes that individual members of a group have the average characteristics of the group at large. Stereotypes are one form of ecological fallacy, which assumes that groups are homogeneous. For example, if a particular group of people are measured to have a lower average IQ than the general population, it is an error to assume that all members of that group have a lower IQ than the general population. For any given individual from that group, there is no way to know if that person has a lower than average IQ, average IQ, or above average IQ compared to the general population.”
http://en.wikipedia.org/wiki/Ecological_fallacy

Rollo Tommasi said...

GaryK: Your calculations are well dodgy!

For a start, they assume that there are equal numbers of smokers and non-smokers.

Also, from a quick look, I see the rish of lung cancer in men only is given as 22 or 23 times higher for smokers than never-smokers. For women the figure is around 12 times.

Those are just a couple of factors which destroy your duff stats.

Rollo Tommasi said...

AnnW: I did not ignore your comment. I said "That estiamte can be calculated in a range of ways. Jamrozik did so one way; the study you referred to in another. Each has its advantages and disadvantages."

Your quote is completely irrelevant, as it deeals with identifying characteristics of individuals, whereas Jamrozik was applying results to large population groups. The relative risks Jamrozik used already average levels of exposure amongst people exposed to SHS.

Rollo Tommasi said...

And Ann, let me remind you again of the question you always ignore. Why do you keep raising this issue?

In all of your desperate efforts to taint Jamrozik, are you really trying to deny there are significant numbers of deaths attributable to passive smoking?

Fredrik Eich said...

"In all of your desperate efforts to taint Jamrozik, are you really trying to deny there are significant numbers of deaths attributable to passive smoking?"
Rollo, there is proof that people attribute deaths to passive smoking but one would have thought that the burden of proof is with those people to prove deaths are caused by passive smoking. I could attribute lung cancer deaths to annual rain fall but asking someone else if they deny lung cancer is attributable to rain fall does not mean it is auto-magically proved that rain causes lung cancer. Which it (rain) definitely does not.

Ann W. said...

Rollo, I am not and have never tried to taint Jamrozik. All I have ever said is that you have to have faith that everything he claims is factual and that he has not made any mistakes in his assumptions or estimates.

I have shown, in two studies, that errors can and are made in those areas, which you have chosen to ignore.

You have every right to believe what you want but you DO NOT have the right to bully, insult or force me or anyone else to have the same faith.

Pogo said...
This comment has been removed by the author.
Pogo said...

@Rollo, 21 November 2010 20:02:

Meta-analyses are a proper and respected way of finding overarching results from a collection of studies, as this article points out: http://www.bmj.com/content/315/7119/1371.full?ijkey=56e76871c0af23ff41bc08edb84a947f344402a0&keytype2=tf_ipsecsha

Indeed that is true - for single-component, studied trials such as those described in the report you cite. As I have repeatedly said, RRs of <2 can be taken as statistically significant if the type of study merits it ie double-blind or similar. However, NONE of the individual ETS studies are of this type, they are either observational or data-trawls, which, by statistical convention, require far higher RRs to be considered significant

Meta-analyses consistently show that passive smoking increases the risks of heart disease and lung cancer.

No. Meta-analyses of statistically-insignificant ETS studies consistently produce statistically-insignificant RRs.

You seem incapable of understanding the differences between controlled and observational trials, and as I, unlike you, am not a paid shill I have no intention of wasting any further time on the subject and will quietly withdraw in order to allow you righteously to wallow in your ignorance.

Michael J. McFadden said...

Rollo, separate smoking rooms are most assuredly the answer for larger pubs that have the room for them. Your earlier comment was quite different than concern about what the publicans wanted however: You said, quite clearly, "separate smoking areas have not been demonstrated to adequately protect staff and punters." and you are wrong to the extent that they HAVE been shown to reduce exposures by a hundredfold or more -- which any rational person would take as "adequate protection." Dodging back to the practicalities of what publicans and F2C might *ideally* prefer doesn't work, and I'm sure you were well aware of that.

Re RR of 2 or more: I asked you for some examples to the contrary as I was not aware of such. You provided them and I accepted them with a certain degree of criticism. Read my original post for that criticism, which was mild.

As for the challenge, I submit to anyone reading this that you were, and are, indeed being thickheaded since I asked you VERY specifically for something and VERY specifically outlined what I excluded in my request. You REFUSED to provide the information I asked for and VERY specifically provided the information I had excluded.... and then went on to claim you had answered the challenge. Other tnan in the sense of admitting you could not provide what was asked for you most certainly did NOT answer the challenge.


You accuse the opposition of cherry picking and I responded in the most powerful way possible: by inviting YOU to cherry pick yourself; and even with that offer you are unable to support your position. Return to my original statement if you wish and see if you can provide what was asked for unless you want to make a commitment that you stand by those reports IN THEIR ENIRETY. I.E. you stand ready to defend each and every point/study in those reports or else admit they do not adequately support your contentions. Do you truly feel that confident in their perfection that you will lay yourself on that line?

- MJM

Anonymous said...

Men and women lift-a-holics are always searching for the best site to encounter. Numerous have never thought being online with others who hold extensive knowledge about which drugs and proteins to accept in order to expand with results. This is because weightlifting is a sport, and with any sport you go where the fanatics are to discover about it! To discover more about Female Bodybuilding

To push the statement a level further, you might be able to go and take advice from a individual trainer about precise techniques but their knowledge will be only be so vast and voluminous. In order to learn you must go where other like you talk in a group to talk about about one thing. This is because a lone bad raise or protein can break everything you have worked so long for already. Even more so, there are few groups
and places you can go in order to conversate about steroids and other drugs used in lifting.

But yet if weightlifting and increasing muscle ins't what you are looking for, many Weight Loss Programs are around solely for individuals to discover how to get rid of those extra five pounds, and with the culture of today it seems everyone is getting larger and fatter. So no matter who you are in the end, obliging to a fit diet and the right exercises can play all the difference!

Gary K said...

Rollo says:"Meta-analyses consistently show that passive smoking increases the risks of heart disease and lung cancer."

Perhaps; but, they give NO useful information about 'actual risks'.

A 25% increase over a very little risk is very,very darn little risk.

In the USA there are about 136 million never-smokers and the claim is that they have 3,400 lung cancer deaths due to SHS/ETS exposure.

Since the smokers are of all ages and the deaths are of all ages, we can arrive at a life-time risk of lung cancer death due to SHS/ETS exposure.

3,400 deaths per 136 million never-smokers is 1 lung cancer death per 40,000 people exposed.

1 divided by 40,000 is 0.000025.

A never-smoker has a .0025 of a 1% lifetime risk of dying from lung cancer due to SHS/ETS exposure.

That is 2.5/1,000th of a 1% risk.

Put another way,since 39,999 out of 40,000 will NOT die, there is a 99.9975% probability that a never-smoker exposed to SHS/ETS will NOT die from a lung cancer due to that exposure!!!

In the USA lung cancer deaths due to SHS/ETS are only 7% of the total 'claimed' deaths. Rollo says there are 6,600 deaths in the UK.

7% is ONLY 449 lung cancer deaths out of the 10's of millions of people involved!!

Anonymous said...

Ann W'
Your paper on Kaposi's Sarcoma is a fascinating find!
I can't find any details as to what sarcoma Jamrozik died of but according to the ONS, Kaposi's is very rare - only about 15 UK cases in 2005.
Interestingly, a google search for "Konrad Jamrozik" brings up a freedom2choose blogspot article as the first hit. A Yahoo search bring up a Velvet Glove Iron-Fist article as the second hit.

Quick note on RR's: I think that any RR below about 2.0 is bound to be suspect. I recently knocked up a programme that runs multiple epidemiological studies using random data. i.e. data where it is known that no correlation exists. With this, I can generate statistically significant meta-analyses with RRs of up to around 25% every single time using only a small level of publication bias. By contrast, generating a RR of 2.0 or more would require me to ignore well over 95% of these random studies - much less likely to happen in the real world.
What's more, even a single study is often made up from several blocks of data and so can be suspect for the same reason.
Of course, as MJM has pointed out, publication bias is just one of man potential ways to bias a study.
Tony

Michael J. McFadden said...

Rollo, one of the differences between us is that I believe I am quite aware of my bias and try to correct for it (though it may not always show on the battlefield.) I think you are biased and are NOT as aware of it... and that that is one of your weaknesses.

Let me give an example. You will say something like "Meta-analyses consistently show that passive smoking increases the risks of heart disease and lung cancer." and believe you have said something true.

You have not.

Can you explain why?

- MJM

Rollo Tommasi said...

Okay guys. 7 responses back at me since I was on this morning. Happy to reply but I've better things to do and can't respond to all 7.

I'll give a quick response to some just now. Taylor et al (2007) answers in itself MJM's question in his latest post, by listing previous meta-analyses on lung cancer. It also addresses Pogo's point - because it shows that meta-analyses of cohort, population-based case control and other case control studies ALL show a statistically significant increased risk of lung cancer.

GaryK's calculations show no sign of improving. First of all, he treats ANNUAL risks to non-smokers as LIFETIME risks. Then he fails to take into account the additional risk of heart disease from passive smoking.

I have to say all of you are making bold claims, based on nothing more than your say-so (other than Fredrik's reference to his own blog - sorry Fredrik, but your cute comparison would easily fail Bradford Hill's criteria of causation).

Otherwise, if you guys want to choose one of the above posts for me to respond to, then I'm happy to do so.

Anonymous said...

From Dave Atherton

Rollo you also forget about misclassification on smokers as non smokers. People are far more likely to fib about their smoking than the other way. Peter Lee suggests typically it could be 2.5% and with an RR of 10x for active smnoking and LC an RR of 1.25 would then become 1.00. In practical terms the 2nd URL is one done on Cork bar workers and 4% fibbed.

http://jnci.oxfordjournals.org/content/91/6/560.full.pdf


http://www.springerlink.com/content/v15664h75x2k8u6l/

Michael J. McFadden said...

So you want US to choose one post out of 7 because 7 is too many. But meanwhile you want US to provide you with full refutations, item by item of several THOUSAND studies listed in the Reports you cite as your backup.

Meanwhile, I sincerely doubt Taylor addresses my last question Rollo. I think instead it will reinforce the fact that you are sadly unaware of your bias. But I will look, nonetheless.

- MJM

Michael J. McFadden said...

OK, I've examined Taylor. Rollo, I will admit that you surprised me. It seems likely that you not only understood the point I was getting at but managed to put your finger on a study that addresses it fairly well. I don't agree with the conclusion but at least it was addressed.

Now about those 7 compared to thousands....

:?
MJM

Fredrik Eich said...

"sorry Fredrik, but your cute comparison would easily fail Bradford Hill's criteria of causation" - That it would Rollo, that it would.

Gary K said...

Rollo says:"GaryK's calculations show no sign of improving. First of all, he treats ANNUAL risks to non-smokers as LIFETIME risks."

The risk MUST be a lifetime risk; if not, it would be cumulative and eventually you would have several thousand never-smokers with a 10% risk for the present year and the rest of the exposed never-smokers would have ZERO risk

Michael J. McFadden said...

So Rollo, have you completely given up on even trying to answer seven simple misguided posters here while still expecting us to be able to post detailed analyses of tens of thousands of studies for you?

- MJM

Michael J. McFadden said...

Well, since it's now been over a year, I guess it *is* safe to say that Rollo gave up.

- MJM