Saturday 22 May 2010

A tale of two studies


The International Agency for Research on Cancer (IARC) is a well-respected body set up by the World Health Organisation. It has conducted many large epidemiological studies into possible carcinogens. Let's take two of them. We'll call them Product X and Product Y.

There were two major findings for Product X. They were:

Odds ratio: 1.40 (1.03-1.89)

Odds ratio: 1.15 (0.81-1.62)


There were also two major findings for Product Y. They were:

Odds ratio: 0.78 (0.64-0.96)

Odds ratio: 1.16 (0.93-1.44)


You will notice that each study found one small but significant finding and one small but non-significant finding. In the case of Product Y, however, that significant finding suggested a protective effect. 

None of these findings are particularly strong, but—if you had to pick—you would say that Product X was the most likely to be the real carcinogen, right? After all, both findings for Product X show a potential increased risk, and the largest of them is not only statistically significant but is more than twice as large as Product Y's.

But that's not how these findings were reported at all. The WHO issued a press release saying that there was no conclusive evidence that Product X caused cancer and blamed "biases and errors" for the study's findings. The WHO also issued a press release for Product Y, saying that it definitely did cause cancer and blamed weaknesses in the study for its failure to show this more clearly. 

Consequently, the BBC reported that Product X "does not appear to increase the risk" of getting cancer, but reported that Product Y represented "a definite, although small, risk" of getting cancer.

So why would the weakest associations be hyped up while the stronger associations were downplayed?

Product Y is passive smoking. Product X is a mobile phone. 

The World Health Organisation has not decided to wipe mobile phones off the face off the earth.



29 comments:

Leg-iron said...

Do you have 'X' and 'Y' the right way round?

Christopher Snowdon said...

Not at the end. Thanks! Didn't make much sense as it was.

manwiddicombe said...

Is it just me or do the links at the end both target the same page on the BBC?

Christopher Snowdon said...

It's good to know I have attentive readers! That, too, has now been corrected. That's what you get when you're rushing to get out in the sunshine!

Cheers

Leg-iron said...

Sunshine? our forecast was 'fine with occasional showers'. What we have is 'overcast with occasional monsoons'.

Won't do the pubs any good this weekend.

Christopher Snowdon said...

You need to get away from Scotland and get yourself down to Brighton, Leg-Iron. I'll get the drinks in.

Rollo Tommasi said...

Well Chris, I have to say you’ve really had to twist the evidence to make your case on this blog (even when the links are in the right place).

The main figures from this latest mobile phone study are actually OR 0.81(0.70-0.94) and OR 0.79 (0.68-0.91). Kind of bursts your bubble, doesn’t it? The figures you quoted relate to the 10% of user with highest exposure. As for next steps, Dr Christopher Wild, Director of IARC is quotes as saying: "An increased risk of brain cancer is not established from the data from Interphone. However, observations at the highest level of cumulative call time and the changing patterns of mobile phone use since the period studied by Interphone, particularly in young people, mean that further investigation of mobile phone use and brain cancer risk is merited."

To me that conclusion and that will to investigate mobile phone use further is an entirely responsible response to those findings.

As for the Boffetta study on passive smoking from 1998, the figures relate to ever exposure – not the most highly exposed decile. More to the point, the case for the dangers of passive smoking is hardly based on that study alone. Instead it contributes towards a much larger body of evidence. As the report concludes: “The risk from ever exposure to spousal ETS was consistent with the combined available evidence from European studies, but it was lower than some previous estimates—a result that could be explained by the large number of subjects whose exposure to ETS ended several years earlier.”

And as for the 0.78 child exposure figure, you choose to ignore that the Boffetta report discusses that too – pointing out that it is only one of several findings on child exposure and that the evidence overall is inconsistent.

If this blog what counts for rigorous scientific scrutiny in your world, Chris? Because it is riddled with holes.

Christopher Snowdon said...

As usual Rollo misses the point of the post, which is about how these studies are reported, both by the WHO and the media. I don't believe that mobile phones cause cancer, but you can bet your bottom dollar that if a study of passive smoking showed the same results, even if -as here - it was only for the most heavily exposed, it would be reported in a very different way. (I don't think many people seriously think that occasional mobile phone use increases cancer risk, so it's reasonable to look at the heaviest users. Unlike secondhand smoke, of course, for which there is 'no safe level of exposure').

There are biases in every study, and it's interesting to see how those biases were ignored or downplayed with the IARC passive smoking study but were hyped up, even by its authors, in the IARC mobile phone study. (The mobile phone study was part-funded by the telecoms industry. Imagine if the passive smoking study had been part-funded by the tobacco industry! You'd never hear the end of it, as Enstom & Kabat well know).

Most of the media have used the IARC report as proof that mobile phones are safe, or at least that the evidence is inconclusive. The Daily Telegraph reporter, however, uses the exact same data to say they are unsafe. Regardless of whether or not he is right, that article gives a taste at how the study would have been reported if it had about secondhand smoke.

Rollo Tommasi said...

Sorry to say this Chris, but you are talking absolute nonsense.

Since you spent the first half of the article discussing the actual results of the studies (NOT how they were reported, no matter how much you may try to backtrack now), let’s look at what the major findings of the two studies REALLY were.

Boffetta et al on passive smoking and lung cancer:

Adult exposure (ever exposure to spousal ETS): OR 1.16 (0.93–1.44).
Adult exposure (ever exposure to spousal ETS): OR 1.17 (0.94–1.45).
Childhood exposure: OR 0.78 (0.64–0.96).

Interphone Study Group on mobile phones for people ever having been a regular mobile phone user
Glioma: OR 0.81 (0.70-0.94)
Meningioma: OR 0.79 (0.68-0.91)

Those are very different findings, no matter how much you try to cherry-pick results especially from the mobile phones study in a weak effort to make them appear similar. Your excuse for not using like-with-like exposures for the two sets of groups (i.e. any exposure for passive smoking but only highest exposure for mobile phones) is pathetic. As you should know, like-with-like measurement is a feature of part of proper scientific method. And besides, whatever happened to your own claim that “the dose makes the poison”?

When you consider the actual key findings, and compare them with existing evidence on the subjects, the conclusions drawn in both studies are quite reasonable.

Right, let’s turn now to how the studies were reported. You claim errors and confounding issues were downplayed in the Boffetta study. Really? I’ve just reread the study report and the discussion about confounding factors and efforts taken to overcome them are set out in detail.

In the mobile phones study, the media release highlights errors and confounding with the highest decile for meningioma. That was the only decile group to show a positive result – one decile group out of 20. Results involving small subsets of study groups are invariably shakier than results for the whole group, because of the small sample size involved.

Finally, you try to blame IARC for the media’s reaction to the stories. As if IARC has editorial rights over what newspapers state!! If you’ve got a problem with how the studies have been reported, you should be directing your complaints to the media outlets involved, not IARC.

brankach said...

As I am aware, passive smoking studies have been done for spousal, workplace and childhood exposure. However, I think one could argue these are also sort of heavy exposures, not occasional ones.
Am I right Chris?

Christopher Snowdon said...

Rollo,

Your quibble seems to be over my use of the phrase 'main findings' to describe the RRs of the most heavily exposed. If I was trying to make out that mobile phones caused cancer you might have a point, but since I'm not, you don't. If it bothers you, substitute that for 'headline findings' or 'stand-out findings' or whatever, but don't pretend I've dug out some obscure part of the study. It's right there in the abstract and every major news source has focused on it. For example:

The BBC:

"But they also dismissed as problematic the finding that at the other extreme end of the spectrum those using the phone for the longest cumulative periods, more than 1,640 hours, appeared to have a higher risk, regardless of over what period of time this was spread. This was as much as 40% higher for glioma, and 15% higher for meningioma."

Associated Press:

The survey of almost 13,000 participants — the biggest ever of its kind — found up to 40 percent higher incidence of glioma, a cancerous brain tumor, among the top 10 percent of people who used their mobile phone most.

TIME:

"the 10% of people who used their phones most often and for the longest period of time — 30 minutes a day or more on average for at least 10 years — had a substantially higher risk of developing some form of brain cancer than those who didn't use a mobile phone at all."

And so on, and on and on. Perhaps there is a legitimate question about the findings for the heaviest users. Since most sensible people understand that there is a safe level of exposure for everything, maybe the heaviest users have crossed that line? I don't know. What interests me is how these statistically significant findings are explained away as being the result of bias and chance, when more feeble findings in the sphere of secondhand smoke are taken on trust.

To give one example, let's take recall bias.

The BBC:

"The report noted that people with brain tumours were more likely to overestimate the role of a potential risk factor"

Fair enough. It's very plausible that subjects who are already ill will exaggerate or misremember their exposure to something that they are told could have caused their disease. But if this is true of brain cancer and mobile phones it is certainly true of ETS and lung cancer, but the IARC never entertained that as a possibility and anyone who does is accused of working for the tobacco industry or worse.

It is the reaction to that particular finding for heavy users that interested me. Not the finding itself, but the reaction.

By the way, do you believe that using a mobile phone really reduces the risk of glioma by 19% and the risk of meningioma by 22%? Both are stronger associations than the IARC found for passive smoking and lung cancer, and they reach statistical significance. So what's it to be?

Rollo Tommasi said...

So Chris, if I get you right, you are judging the “major findings” (or whatever alternative term you might use) on what the media chooses to pick up. Not the key findings of the research. Not the headline points set out in IARC’s accompanying media releases. But on what the media picks up. And apparently IARC is responsible for however the media reacts to its research

As for recall bias, why are you quoting the BBC report so selectively? It gives reasons for possible recall bias, which are distinctive for the mobile phone study – but which you choose to ignore. This is what the article says:
“The report noted that people with brain tumours were more likely to overestimate the role of a potential risk factor, and that the disease interferes with memory and cognition, undermining the accuracy of the recollections of such extreme use.
Unlike lung cancer, where the risk rises the more cigarettes are smoked, this mobile phone data shows no increased risk until the very heaviest use begins. This was also seen as casting doubt on the reliability of the reports.”
Of course, cohort studies are a good way of reducing the possible effects of recall bias. As far as passive smoking and lung cancer are concerned, the results of cohort studies are remarkably similar to the results of case-control studies. So why is it that you suggest passive smoking studies are so tarnished by recall bias?

The last paragraph exemplifies two misconceptions you continue to hold throughout your arguments. These are, you insist on adopting a strictly formulaic, numbers-based approach to assessing studies; and you also insist on treating each study in isolation. The fact that most health bodies and scientists generally now agree that passive smoking is associated with lung cancer comes not from one or two studies, but a whole series of studies which have produced consistent findings.

Because there is now a large portfolio of broadly consistent evidence about passive smoking and lung cancer, scientists are prepared to recognise the potential dangers there even though the excess risk for an individual is by some standards relatively modest. If that parcel of corroborating studies is not there (as is the case with mobile phones – note the call from researchers for further research), scientists are reluctant to reach a definitive view.

And if you think that’s a handy device which allows scientists to rule that passive smoking is harmful but other factors aren’t, then you are wrong. There are plenty of examples of scientists and health bodies determining that the evidence is insufficient to link passive smoking to a particular condition. For example, IARC concluded in Monograph 83 that evidence overall is insufficient to associate passive smoking with breast cancer – even though there are some studies showing relative risks of 2 or 3 or more, the scientists rightly also take account of other studies which suggest no added risk. Similarly, SCOTH (2004) concluded that evidence available at the time was insufficient to conclude an association between passive smoking and strokes, despite the existence of one study which produced an odds ratio of 1.82 (1.34-2.49).

Christopher Snowdon said...

Answer the question.

Rollo Tommasi said...

If you want an explicit answer to the question in your last paragraph, which should already be clear from my last post anyway, here it is. The figures surprise me a bit, but I neither believe nor disbelieve that using a mobile phone really reduces the risk of glioma by 19% and the risk of meningioma by 22%. Without reading the full study report, I cannot state definitively if I believe the study was conducted well, but the media release provides promising signs that it was by expressing findings and conclusions in a balanced and reasonable tone. But more than anything, I would not read a too much from the results of a single study. It has to be looked at alongside other evidence (so it’s promising that IARC intend to produce a monograph on this issue).

I’m slightly surprised you expect me to answer every part of your comment directly, when you have not offered me the same courtesy. I see, for instance, in your last post that you chose to duck points I made that:

a. you had cherry-picked findings from the two studies. You argue that the media picked up on the findings you cited. True, but they also referred to the other key findings which you chose to ignore in order to make some kind of argument.
b. contrary to what you claimed, neither study downplayed the potential for bias or confounding factors;
c. you seem to be arguing that IARC should somehow be responsible for how media outlets choose to refer to its work.

At least now I know your ground rules, I can be confident that you will answer each of these points properly, as well as each of the other points I made in my last post.

Can't I?

Christopher Snowdon said...

a. Yes, I was using those specific figures to make an argument. Well spotted. Obviously, it's not the argument you wanted to have (ie. another generic slanging match about passive smoking), but it's my blog and I'll write about what interests me. I'll say it again: I'm not arguing that mobile phones cause cancer or saying that those are the overall odds ratios. I was looking at the way the IARC and the media looked at those particular well-publicised findings. Let's also bear in mind that the 'heaviest users' in the phone study were only using them for half an hour a day. That's pretty average these days, isn't it, so you can see why everyone except you considered that to be the most important finding in the study.

b. I didn't say the mobile phone study downplayed the biases. I said the opposite. And I said biases are 'ignored and downplayed' in passive smoking studies. Which they are. See below.

c. Rollo, in the media there are things called 'press releases'. Journalists don't read through the whole study and pick out what they think is the interesting bit. They rely on these 'press releases' to form the content and angle of their article. In this instance, it was WHO press release number 200, which said:

In the tenth [highest] decile of recalled cumulative call time, ≥1640 h, the OR was 1.40 (95% CI 1.03-1.89) for glioma, and 1.15 (95% CI 0.81-1.62) for meningioma; but there are implausible values of reported use in this group.

Not the word 'implausible' there? That's a subjective term which downplays the finding.

It also says:

Biases and errors limit the strength of the conclusions that can be drawn from these analyses and prevent a causal interpretation.

Then it lists several quotes from various people which were reprinted verbatim by the press. See how this works now? Do you see how the IARC influences how the study is reported? Now compare that to the infamous WHO press release 'Do not let me them fool you' which didn't mention any of the potential biases, didn't mention the childhood exposure result and blamed the lack of statistical significance on small sample size.
Even you, Rollo, must be able to spot a bit of spin here.

Rollo Tommasi said...

Well Chris. Thank you for demonstrating your hypocrisy for all to see.

In your response to point (a), you freely concede that you are selectively cherry-picking certain figures to make an argument (on the grounds that “it's my blog and I'll write about what interests me”). Yet at (c) you accuse IARC of “spin” because their press release didn’t cover every salient detail of the Boffetta study. So blatant cherry-picking is alright for you, but it’s not alright for IARC to omit any vaguely relevant detail?

I think the IARC press release about Boffetta is quite reasonable. The main purpose of both releases is to set out what the two studies add to existing understanding on their respective subjects. In the case of mobile phones, there is potential new learning from this study, but the results need to be treated with caution. That is not the same situation as with the Boffetta study, where the main point of learning is that the Boffetta study was consistent with previous studies, therefore reinforcing existing evidence that passive smoking is harmful.

Of course, the press release on the Boffetta study had to serve a second purpose – which was to reassure the world that the misinformation spread by Big Tobacco was nonsense. If only you would criticise those who pedalled the misinformation that IARC wanted to withhold the Boffetta study from publication.

As for (b), you try to evade the issue. You argued that the Boffetta study downplayed biases when, as I have said, it did no such thing.

Christopher Snowdon said...

I'm not going to keep going round in circles with you Rollo, and I'm not going to keep repeating myself. Last time you were on this blog I pointed out your habit of not listening to a word you're told, of constantly changing the subject and of twisting everything beyond all logic. You have done nothing to merit having your hand held through every nuance of the debate.

I've spent years reading about this issue and I assure you that you are bringing nothing new to the table. Why should I have to go over it all again for the benefit of an anonymous and aggressive commentator on the internet? Have you even read my book? If not, why don't go off and get a copy and when you come back, drop the attitude and try to argue rationally.

Even as a rhetorical contest it falls flat. For a debate to be worthwhile, the other person needs to engage with what is being said. You seem totally unable to do that. This is a psychological problem that you need to address and I know I'm not the first person to tell you this.

As a starting point, look at your last post. It's a classic example of how you argue. It simply doesn't make sense. Using an example isn't the same thing as cherry-picking, and cherry-picking isn't the same thing as spin. 'Big Tobacco' have got nothing to do with what we're talking about. You are confusing completely different issues, building straw men and going off on unrelated tangents. It's a collection on non-sequiteurs that don't have any relation to the original article, to my replies or to the other comments. It doesn't address the topic of the article and it doesn't respond to my last reply to you. It's just a bunch of opinions about some vaguely related topics.

If this is deliberate debating technique, stop it. If it's something you can't help, accept it as a problem and try to correct it in the future.

Rollo Tommasi said...

It didn’t take long Chris, did it? You try to defend your statement a couple of times, but you quickly give up and resort instead to attacks on me. Well done.

No matter how you try to criticise me and my technique (and actually, you are virtually the only person to say such things), the points I made remain.

Let’s recap on that the major findings from the 2 studies actually were:

Boffetta et al on passive smoking and lung cancer:

Adult exposure (ever exposure to spousal ETS): OR 1.16 (0.93–1.44) (consistent with previous evidence)
Adult exposure (ever exposure to spousal ETS): OR 1.17 (0.94–1.45) (consistent with previous evidence)
Childhood exposure: OR 0.78 (0.64–0.96) (added further inconsistency to an already heterogeneous body of evidence)

Interphone Study Group on mobile phones for people ever having been a regular mobile phone user

Glioma: OR 0.81 (0.70-0.94)
Meningioma: OR 0.79 (0.68-0.91) (some evidence of increased risk at very highest exposures, but the evidence for this subset is not robust)

In this piece, you have tried to level a charge of inconsistency on IARC. But in order to do so, you have had to create your own version of what the “major findings” from the two studies were in order to make IARC appear different from what a fair and objective person would believe. So you have deliberately downplayed some of the real major findings and raised the profile of other findings.

You have consistently ducked addressing the charge of falsely arguing that the Boffetta report downplayed biases in the results.

And, as I said earlier, your views are blinkered by a couple of enormous blind spots, based on misunderstandings of scientific practice. One is that you insist on adopting a strictly formulaic, numbers-based approach to assessing studies. The other is that you also insist on treating each study in isolation. This is latterly shown up in your apparent refusal to accept that it is reasonable for the press release on the Boffetta study to include different key points from the press release on the mobile phones study, because the results of Boffetta added to a large body of pre-existing work.

If you run a blog, you should accept that not all the comments you receive will be from pro-smoking lackeys. You should expect to be challenged from time to time. If you can’t deal with these challenges in a mature and responsible way, then that is your problem Chris.

Fredrik Eich said...

Rollo,
You have to understand that anti smokers can show a lack of maturity, take for example this tirade on the Daily Mail
"Smokers should do their nasty business in their own homes only. Public places should be out of bounds to these stinkers. These dirty people huddled outside offices, pubs and in cars are a STINKING nuisance. RolloTomasi, Bangkok, Thailand, 24/3/2010 4:21"
Anyhow, what do you think we should do with skeptic Sir Richard Peto after apparently saying at the Fifth World Conference on Smoking and Health that the chief relevance of passive smoking is "political rather than health."(ref), Should we send him off to an anti-smoking
political re-education camp?

Rollo Tommasi said...

Fredrik: It is up to individuals to account for their own comments – whether we are talking about Christopher Snowdon or that other Rollo Tomasi from Thailand (who is not me, by the way….).

I’m a bit disappointed that you should choose to use a 26 year old quote from Sir Richard Peto. Most of the evidence about passive smoking wasn’t available then, so it's not surprising that the quote, even if correct then, is outdated. If you want a more up-to-date statement of Peto’s views, try this: http://news.bbc.co.uk/1/hi/health/6244926.stm#Peto

“Cigarette smoke is the most important cause of cancer in the world and so the exposure of non-smokers to it is going to cause some risk of death”.

Fredrik Eich said...

Rollo,
Thanks for ID the clarification!

I just thought that it was worth pointing out that Sir Richard Peto has not my knowledge quantified risk for passive smoking and has indicated that any such risk is hard to measure (as per your conversation) He has not said "passive smoking kills X number of people a year which is more than car accidents and hoards of man eating sea otters combined" for example.

Klaus K said...

Tobacco industry consultants did make several comprehensive reviews of the Bofetta 1998 IARC study. You can read one of them here:

Analysis of IARC 1998 Study

The conclusion: Once you adjust for the relevant confounders, the point estimates of risk will not be 1.16 and 1.17, but 1.08 for spousal smoking - and 1.04 for workplace ETS. Neither are significant.

Everybody can do the same calculations directly from the figures in the published IARC-study.

It should also be noted, that the risk in three of the eight european centers was negative, i.e. France (0.72), Italy (0.73) and Western Germany (0.88) - while the risk in fx. Sweden was 2.20. All figures unadjusted for confounders.

Omitting Sweden and Portugal from the study results in zero risk for Europe.

The IARC-study is the far biggest study of ETS-risk in Europe, and I agree with Snowdon the results are inconclusive. In fact all ETS-studies of workplace risk in Europe are inconclusive.

Even if you pool all existing european workplace ETS-studies together - the Surgeon General 2006 did just that in their meta-analysis - the results are inconclusive: 1.13 (0.96-1.34) - Please remember, that this estimate contains the unadjusted figures from the IARC-study.

Surgeon General 2006, chapter 7 - see page 436

There is no other meaningful conclusion to this, than to say, the risks of lung cancer from ETS in Europe is inconclusive.

On the other hand: All the existing european childhood ETS-studies pooled together is conclusive: 0.81 (0.71-0.92) - i.e. protective against lung cancer.

Doesn't quite make the case stronger, does it?

Rollo Tommasi said...

Klaus offers an interesting analysis. Strange that it only appeared on Brown and Williamson’s website, though, and not fed into the professional discussion about the Boffetta study. As it happens, a tobacco industry consultant – Peter Lee – did present a similar critique of the study, which Boffetta duly addressed:

http://jnci.oxfordjournals.org/cgi/reprint/91/6/560

Klaus’ other comments are more bizarre. The Boffetta report clearly states the study controlled for most potential confounders, despite Klaus’ claims.

And what exactly is the logic for omitting the results for Sweden and Portugal? Omit the studies from France and Italy 1 results and you get a statistically significant positive result. Likewise, omitting France and Portugal 1 from the child exposure studies gives you a statistically non-significant reading.

Klaus then refers to the 2006 Surgeon General’s report. Yet bizarrely, he decides that the only results worth noting are results from Europe about workplace exposure. What about all eligible studies? And what about spousal exposure?? A meta-analysis of case control spousal studies shows a statistically significant risk of 1.21. And a meta-analysis of cohort studies (which overcome any risk of recall bias) show an even stronger statistically significant risk of 1.29.

Klaus K said...

Thank you for your comments, Mr. Tommasi.

It is true, that the authors of the IARC 1998 study controlled for some confounders, but if you read the IARC study closely you will discover, that the point estimates (1.16 and 1.17) are the unadjusted figures - i.e. without controlling.

The authors mention some potential confounders (not all of them) and show how much they bring down the point estimates one at a time - fx. they write:

"Restriction of the analysis to histologically verified cases had minor effects on the risk estimates: The OR for spousal or workplace exposure was 1.11 (95% CI 0.86–1.43)." This however, is not a "minor effect", but a huge effect, bringing the point estimate 33% closer to zero.

Controlling for other confounders, i.e. consumption of vegetables, lowered the point estimates to 1.14, and 1.15 when controlling for urban / rural residence. But several important confounders has not been controlled for, unfortunately - fx. intake of fat / saturated fat, as noted by Dr. Denson: Denson letter to the editor.

Please note the rather strange answer from the authors. Of course intake of fat will not explain the whole result. But controlling for this and the above mentioned confounders would in fact lower the point estimate to zero - as shown in this analysis.

I fail to find any significant result from any of the study centers in the IARC study. The authors state, that 4 of 12 centers are below zero. Two countries, Sweden and Portugal have very high point estimates (2.20 and 2.33, although still not significant).

My logic behind trying to omit these two countries is showing you, that without them the result would be exactly zero - it is in fact quite sensible to make this conclusion from the study:

Passive smoking is no risk for lung cancer in Europe - only maybe a weak insignificant risk in Sweden and Portugal.

This is apparent to anyone with a calculator using the figures directly from the study paper.

Regarding the European workplace studies in Surgeon General 2006: Yes. It is true. They show no sigificant risk, even all of them (seven) pooled together: 1.13 (0.96-1.34) - including the IARC study. Please go check yourself: Surgeon General 2006, chapter 7 - page 436.

You ask: What about all eligible studies? And what about spousal exposure??

Well, we live in Europe (Snowdon and I). It seems that passive smoking in workplaces is no risk for lung cancer here. There could be many reasons for that. At least I fail to see, why American or Asian workplace studies should determine, that we should ban passive smoking in all European workplaces, just because there is a tiny lung cancer risk abroad, that does not exist here.

Please also note that the authors could not find a risk from exposure in public places: "The range of estimates for ETS exposure in public indoor settings such as restaurants was 0.24–2.32, with an overall estimate of 1.03 (95% CI 0.82–1.29). Analyses by duration of exposure did not suggest any consistent pattern ..."

Spousal smoking has nothing to do with workplace bans in my view. Banning the spouses' smoke in the workplace will on the contrary lead to more smoke in the home. This was already warned against in 2006 by among others the House of Lords.

It seems like the smoking bans in Europe are standing on a very shaking house of cards. Don't you agree?

Rollo Tommasi said...

Thanks for your further comments Klaus. I don’t agree with your assessment of the Boffetta study when you claim the 1.16 and 1.17 figures are not adjusted for confounders. The report states at page 1446 that “we controlled for most potential confounders”.

You also refer to a report which would show that “controlling for [certain] confounders would in fact lower the point estimate to zero”. That’s the same article as you cited last time. As I mentioned, it only appears to have been published on the Brown & Williamson tobacco conmpany’s website and not fed into professional debates about the Boffetta study. Peter Lee did make similar claims, which Boffetta and others addressed.

I find your methods for cherry-picking some studies from the Boffetta report bizarre, if I may say so. I asked you before what reason you have for choosing to exclude the two studies you chose. You haven’t dealt with that point. As I mentioned, you could cherry-pick out a couple of other studies, and that would give you a statistically positive result from the remaining studies that passive smoking is linked to lung cancer.

It seems your logic appears to be that the studies show definitive results about risk each country studied and that a “non-statistically significant” result means the same as “no risk”. Both are wrong assumptions to make. You can argue that the results in themselves are not enough to allow a conclusion that passive smoking is harmful. But I do not see what grounds you have for claiming “Passive smoking is no risk for lung cancer in Europe”.

I also fundamentally disagree with your view that “Spousal smoking has nothing to do with workplace bans in my view.” If secondhand smoke is harmful, it is harmful in any enclosed setting. It doesn’t somehow become safer when it occurs in a workplace.

And there is no evidence for your claim that “Banning the spouses' smoke in the workplace will on the contrary lead to more smoke in the home”. People are taking steps to reduce exposure to secondhand smoke in their homes. Recent research has found that 69 per cent of respondents in 2009 said smoking was not allowed in their home, compared with 61 per cent in 2006.

So I say smoking restrictions are based on well-founded grounds.

Rollo Tommasi said...

My attempt at setting a hyperlink in the penultimate paragraph failed.....

Here is the page I wanted to link to:

http://www.statistics.gov.uk/pdfdir/smokenr0709.pdf

Michael J. McFadden said...

Just wandering in here, with no intention of getting involved in this debate beyond a single simple comment.

I've always thought it interesting that the authors themselves (not media or politicians) characterized the one truly significant finding of their study, the 22% "protective" effect of smoke exposure on children, as indicating "no association."

The WHO press release headlined the nonsignificant results while totally ignoring the significant one.

The fact that the raw information in the study was treated this way points strongly to a significant bias, both on the part of the authors and on the part of the WHO. Given the importance of bias at all stages in scientific studies in controversial areas, whatever the findings were they should be discounted to some unknown but possibly significant degree toward the "no effect" conclusion.

Michael J. McFadden

Michael J. McFadden said...

An addendum: the existence of such bias might well account for the seeming preference to isolate the confounding effects so as to minimize them rather than combining them - which might well have resulted in final results far below what ended up being published.

- MJM

Klaus K said...

Yes, Mr. Thommasi - some adjustment for confounding was done in the IARC 1998 study, but the published point estimates, 1.16 for spousal smoking and 1.17 for workplace smoking are the unadjusted figures. Let me show you:

Please take a look at the raw figures of the IARC study 1998: IARC: Technical Report Nr. 33, 1998

Go to page 122: "Spouse, ever exposed" and see the 5 columns: Starting from the top row, the first column on the left side is the published point estimate: 1.16 (0.93-1.44). The following columns are the different adjustments for confounders. Please refer to the bottom of the page to see precisely, what they are.

In column 3 you will find the adjustments for urban/rural residence and education - two very important confounders in lung cancer risks. Adjusting for these confounders brings the point estimate down to 1.11 (0.85-1.46).

No go to page 220: "Workplace, ever exposed" - the published, unadjusted point estimate, top left is 1.17 (0.94-1.45) but the adjusted figure in column 3 (adjusting for urban/rural residence & education) is only 1.08 (0.83-1.39).

Now if you go two rows down in column 3, you will find the adjusted figures without the centers with no dietry data, (excluding just 7 cases from the study) - this leaves us with a workplace risk at 1.02 (0.78-1.35).

As you can see, just these two confounder adjustments reduced the risks quite significantly against the zero. And it is very likely, that adjusting for dietry fat as Dr. Denson suggested would have reduced the adjusted spousal risk (1.11) to a figure close to zero too: Dr. Denson: Letter to the editor

What I mean to say is that I understand perfectly well, why the authors wrote in the article, that their results were weak. This is in fact a zero risk study - and I agree with Mr. McFaddens last comment: This would have been very clear, if the authours had combined the adjustments for confounding effects and published those figures.

One last thing, that has not been mentioned makes the results even weaker: In Table 5 of the article the combined (unadjusted) risk of either spousal or workplace exposure risk i shown: 1.14 (0.88-1.47)

It is a strange finding, that this risk goes up to 1.18 (0.88-1.59) two years after ending the exposure - and then further up to 1.20 (0.89-1.62) after 3 to 15 years of no exposure. How is this possible?

This is a strong indication of something else playing a part - some other factor than passive smoke is distorting the results heavily through all the figures.

It would have been fair if WHO and the authors had written honestly in the press release, what this study really shows: No evidence of any risk.