We are developing the social individualist meta-context for the future. From the very serious to the extremely frivolous... lets see what is on the mind of the Samizdata people.

Samizdata, derived from Samizdat /n. - a system of clandestine publication of banned literature in the USSR [Russ.,= self-publishing house]

Honest science or propaganda?

Bernie Greene wonders just how scientific is the science behind the smoking debate?

Epidemiology began with a fellow called John Snow investigating to find the cause of a cholera epidemic in London in the 19th Century. He had the idea that it might be coming from contamination in a well. So he took a map showing the locations of wells and plotted the incidence of the disease on the map. Sure enough they were mostly in close proximity to one particular well. He had the well put out of service and there were no more new cases of cholera. That is a simple story of logic and surveying intelligently applied to test a theory.

It is very unfortunate that it was so simple to solve. He might then have left a better example for his followers.

What if he had found that the 166 1 total cholera cases were scattered all over the map pretty evenly but that they all had pink carnations on their coats? One hundred thousand people wore pink carnations and 99,874 did not get cholera.

What does he do now? Well if he were a tobacco investigator he would petition the government to do something about pink carnations. But let’s say he is a brighter boy.

He decides to go and interview the cholera cases in more depth. He asks them all kinds of questions about themselves and about pink carnations – how many years have they been wearing them? How old were they when they started? etc. etc. He gathers all the data and looks for similarities. He gets a few things and tabulates them. He then interviews other people who don’t have cholera but who do have pink carnations. He asks all the same questions again and tries to find something about pink carnations that is consistent among those who don’t have cholera and is absent from those who do, or vice versa.

He finds one item in common. The cholera sufferers all work in the City though they live all over London. There are lots of carnation wearers in the other group who also work in the City and don’t have cholera. So he knows that working in the City isn’t the cause either so he correctly calls it another correlation just like pink carnations. But he now has two correlating pieces of data and a line of enquiry worth following.

He now goes back to the cholera group and asks a single new question. “What can you tell me about working in the City and wearing pink carnations?” They tell him it is the thing to do. They tell him all the ladies like it. They tell him their bosses like it. They tell him they buy them from a vendor right next to St Paul’s Cathedral. etc. He notes their responses and tabulates them again. He asks the same thing of the non cholera group that also works in the City. They give him lots of very similar answers but on where they buy them from the answers vary and none of them are anywhere near St Paul’s Cathedral. So he now has a third correlation and this one is far more promising as it is entirely absent from the non cholera group. He knows it isn’t the cause itself because lots of people who go to St Pauls don’t also get cholera but in combination with pink carnations he knows it is a significant correlation.

He gets the vendor to close down for a few weeks to see if it makes a difference to the number of new cholera cases. There are no new cases after 10 weeks where formally there had been several new cases each week. This is progress he thinks. He could stop right here, have the vendor permanently closed down and that might be enough to prevent further cases. But as he still doesn’t understand how pink carnations and a vendor near St Pauls can be the cause of a cholera epidemic he continues to investigate.

He sets up a watching station opposite the vendor and sits there for many hours watching him ply his trade. He notices one thing that seems to offer a really plausible cause covering all the factors he has so far noted. A woman empties a huge chamber pot into a hole immediately behind the boxes of pink carnations. The scientist walks over and sees that some of the former contents of the woman’s chamber pot have spilt into the boxes holding the carnations. He finds they contain the feces of several cholera victims the woman tends to.

It wasn’t as easy a case to crack but it does illustrate what a scientist dedicated to truth and who won’t compromise with logic would do.

In the case of smoking research I would say that back when the Hill/Doll study of 1956 was done an honest and worthwhile conclusion would have been:

We now know that most cases of lung cancer occur in smokers of a certain amount of tobacco after a long period of time in a very small number of cases relative to the number of smokers who smoke that amount for that long. Tobacco smoking isn’t the cause of lung cancers but it is connected in some way that we don’t yet understand. We can predict that many more smokers will get lung cancer than will non smokers. Something like 24 times as many. But we must also say again that there are far more smokers than there will ever be cases of lung cancer. This in itself makes it obvious that something else is involved.

Further because we can predict that a very small fraction of all smokers will account for the majority of lung cancer victims, and that because we don’t know why this is so, we must responsibly inform smokers of this increase in risk. We must also inform smokers that the risk, whilst small compared to all smokers, is of a grave disease that is likely to be very painful and also fatal. Obviously smokers themselves have to decide if they want to cut down or quit smoking altogether because, as a scientist, our job is to research and discover threats to life but that it is the individual’s responsibility to act on that information as he
sees fit.

We now have something new to discover. Most cases of lung cancer occur in a small number of smokers, but if smoking were the cause of those cases then what is preventing it in so many more cases? If smoking is not the cause of lung cancer then what factor is present in those smokers who do get it but which is not present in those who don’t?

That to me would have been a logical, decent and honest approach. It may not have been so easy to understand as “Smoking causes lung cancer” but it would have been a whole lot more honest and the consequences of that could have changed subsequent history in many dramatic ways, and might have led some curious researchers into the field who would have known that there were still very significant questions to get answered, and they may have been capable of getting those answers by now.

Instead they told the government that tobacco was causing lung cancer and demanded that something be done about tobacco! Wrong target and thoroughly irresponsible as scientists in my opinion. That was a defining moment in the history of tobacco and it has been in decline ever since. Unfortunately so has epidemiology.

1 Don’t know what the actual number was. This was to make it closer to what Hill/Doll found with regard to lung cancer and smokers. 166 out of 100,000 after decades of 25 grams per day.

Competing interests: I once worked for a tobacconist shop. I am also a
pipe smoker. I’m not a scientist and I could be wrong. I would
appreciate being shown where with logical rather than ad hominem

Bernie Greene © 2003 . All Rights Reserved

21 comments to Honest science or propaganda?

  • The Wobbly Guy

    The worst that can be said about smoking is that it simply increases the risks of getting lung cancer(estimates from 4 to 8 times more probable; 24 times is an overestimate), and nothing more than that.

    The choice is up to the individual to decide if the risk of getting lung cancer balances the enjoyment derived from smoking. Deng Xiaoping was a heavy chain smoker, yet he lived to a ripe old age. Interesting, eh?

    When they say something else is involved, I can think of any number of factors that they’ll not have considered. Genetic propensity to cancer, genetic immunity levels, state of health, physical condition, diet, living conditions, stress, job type(it’s not good if you’re a smoker and working in a nuclear power plant).

    Simplistic arguments that state that ‘A causes B’ are often the most flawed on closer analysis. HIV/AIDS comes to mind…

    *rolls eyes*

  • toolkien

    What worries me more is that the dishonest scientist can get away so easily with press release/sound bite statements that are taken by the masses as is and who never question the information. For a dishonest statement to have efficacy it takes two to tango. Unfortunately socialist education makes for a servile public in which easily digestable tidbits go down smoothly and provide for the right jab of government intervention in combination with the left jab of the media.

  • Ed Poinsett

    It would be interesting to see the correlation between long time smoker risk (166 in 100,000) and second hand smoke risk. I would expect that risk to be almost negligible inspite of the current frenzy. The simple proof of this is in THE LIFE INSURANCE INDUSTRY. If there were any actuarial significance, life insurers would not issue “non-smoker” policies for any spouse or family member of a smoker.

  • Julian Morrison

    What about other stuff like emphysema? The useful statistic would be less “how dangerous is smoking for lung cancer?” and more “how dangerous is smoking, at all?”

  • Joel

    There is simply no way to suggest tobacco smoke is only statistically associated with health problems.

    You have left out the fact that tobacco smoke contains numerous carcinogens. The association, therefore, between smoking and cancer is not simply statisitcal.

    Smoking is also highly associated with other lung disease (emphysema) and with cardiovascular disease. If you want a shock, just look at the lungs of a severe emphysemic. They look like an old, ragged sponge with large irregular holes, instead of the normal, very fine spongy appearance of healthy lungs. ONLY smokers get this. I can mail you such a picture if you want.

    People who smoke pipes get things like lip and kidney cancer. You really don’t want to have a large part of your face taken off.


  • Joel says that “smoke contains numerous carcinogens”, but those carcinogens themselves are only linked to cancer by the statistics he condemns.

  • Doug Collins

    “If smoking is not the cause of lung cancer then what factor is present in those smokers who do get it but which is not present in those who don’t?”

    Another potential question is: “What factor is present in those who don’t get cancer?”

    That strikes me as the more interesting question.

  • Oh, scientists get away with all sorts of stuff all the time. You’ll be surprised at how much BS they can get away with when the government AND media is backing them.

  • Michael Last

    1) What statistics can tell you: Is there an association. I can tell you that if you smoke, you are a lot more likely to get lung cancer, emphysema,…,liver disease

    2) Association is not causation. The liver effects of smoking are well explained by if you are a smoker, you are more likely to drink

    3) We know there are carcinogens in tobacco. Take mice. Divide them into two groups. Expose one to tobacco smoke. Don’t expose the other group. In all other ways raise them identically. Best done with mice that are genetically identical. It has been observed (numerous times) that the group exposed to tobacco smoke has a significantly higher (yes, they use statistics) rate of cancer than the other group. There is no other difference than the tobacco exposure. Hence we know that there are carcinogens in tobacco. The question is, are there compounds which are carcinogenic to people?

    4) Why do some people get cancer and others not? Easy answer, though I don’t know if its right, is chance. Smoking increases your chance of getting cancer. It doesn’t mean you will get cancer, just you are more likely to.

    Then there are the detailed mortality studies on smokers vs. non-smokers, a rather thorough literature if you are interested.

    Read the studies before you criticize them – there are flaws with the initial studies (Berkson has a pretty good paper on some of them), but most of those have been investigated since.

    Let me put it another way (flamebait coming) – there is more evidence that smoking causes cancer than there is that private industry is more effecient than governments. There have been controlled experiments, which are missing in the field of economics….

  • Walter E. Wallis

    When you tell the researcher what you want to provr, it takes more charactor than most have to avoid the temptation to draw the curve, then plot the points.
    I suspect that if more attention had gone to curing cancer than finding new culprits, cancer deaths today would be as rare as polio paralysis.
    That is what is wrong with drawing the curve first.

  • A Specialist

    As an epidemiologist, I hardly know where to begin with this post. I encourage the author to read an introductory book on epidemiology before dismissing the field; I suggest Epidemiology in Medicine by Charles Hennekens. If he did read it he might not be a skeptic about smoking and lung cancer. To wit- there are several criteria that denote causality. Among them are temporality (exposure preceded disease), dose-response (more exposure = more disease), strength of effect (bigger numbers means that biases are less likely to have produced them) and biological plausibility (lung exposed to floating carcinogens… hmmmmm). Smoking fulfills all of these criteria.

    The author also appears not to know that causes come in different types. They may be necessary but not sufficient to cause disease. In this type of causation it is not surprising that only a fraction of exposed people develop disease.

    To read the author you would think that all research on lung cancer stopped in the 50’s, when in fact plenty of research continues to this day that explains, among other things, genetic susceptibility to developing lung cancer among smokers. For instance, N-acetyltransferase is an enzyme that clears tobacco carcinogens from the body. People with the slow-acting form (slow acetylators) are at a higher risk for lung cancer than people with the fast-acting form (fast acetylators). So it’s not as if there is no curiosity on the part of scientists to find out what else, besides smoking, affects risk of lung cancer. There’s an enormous literature on the subject if anyone cares to look on Medline. Incidentally, I would like to imagine the author coming up with a concise way to tell the public that it’s alright to smoke if they have the correct genetic polymorphisms for thirty different susceptibility genes.

    Lastly, the author criticised unnamed scientists for demanding something be done about tobacco. I truly doubt the author did anything like go back to the historical record to determine who advocated what law and when. Nevertheless, I’d agree that most of my fellow scientists instinctively fall back on statist solutions to modify disease risk. But let’s be clear- that does not invalidate the science. I’m sure the author can recognize an ad hominem argument when he sees one.

    Finally, here is a brief list of eliminable exposures and the diseases they caused as described by the dread Epidemiology:

    sleeping baby on the stomach – SIDS
    Hepatitis C – hepatocellular carcinoma
    diesterstilberol – vaginal cancer
    Rely tampons – toxic shock syndrome
    L-tryptophan – eosinophilia myalgia sydrome
    HIV – AIDS

    On behalf of my field, I accept your thanks.

  • Guy Herbert

    Even though he undermines his point a bit with the reference to SIDS–which is a long way from understanding/proof–I have to agree with Logan Spector. The post is muddled. Irritatingly so.

    There are problems with epidemiology, and its applications, but those are down to more than one cause (!).

    The fact is statistical reasoning is hard and even professionals in the field have to be careful not to be mislead by intuition. Sometimes predictions require assumptions about causes that may be mistaken, and the assumption dependency gets lost in discussion. The demand for prediction is usually at odds with the strength of the available models. And epidemiologists, just like all other investigators, are subject to confirmation bias: they’ll more readily follow up and elaborate existing models than scrap them and start with a blank sheet.

    However the trouble really start when the results (or sometimes the just the tools) get into the hands of innumerate doctors, politicians, bureaucrats, and media. Instead of asking, how certain? what else could it be? and, most important, how much? They take “statistical significance” (usually based on conventional confidence intervals, which can be pretty iffy) as demonstrating certainty–when it suits. And it usually suits only when there’s a moralistic rather than a scientific/economic case for doing something.

    Toolkin writes: “Unfortunately socialist education makes for a servile public in which easily digestable tidbits go down smoothly and provide for the right jab of government intervention in combination with the left jab of the media.”

    But it isn’t as conspiratorial as that. Logical reasoning, especially statistical reasoning, is hard. Most people, even well-educated professionals, never get it. And it is a peculiar psychological fact that nearly everyone reasons far worse presented with an abstract problem to analyse than a concrete example of the same question. (A disaster in the criminal courts, given increasing reliance on “scientific” evidence.)

    The media usually don’t understand what they are telling the public (including politicians), and the public understands that less. Positive relationships will be reported, negative not. One’s news, one’s not. Worse, most will assume that something can be done; that doing something will be better than doing nothing; that we know what to do to be effective; that if something effective can be done, it must be, regardless of other consequence; and that the state should do it. Public policy is subject to a string of implicit conditions (any of which may be false or break down in practice), and usually based on a misunderstanding of the evidence.

    So it isn’t reasonable to blame the epidemiologists for what’s done with their work. And they’d be better at it if subjected to proper critical analysis, rather than a choice of panic or flat disbelief.

  • Sorry, but epidemiology can only suggest, and is not the definitive method for determining cause and effect. It narrows down the list of possible factors, that is all.

    Only a controlled reproducible scientific experiment can properly determine cause and effect. That’s what I believe.

  • Guy Herbert

    The Wobbly Guy is right too–up to a point. Can we allow that Logan Spector was merely indulging in rhetoric when he wrote “…and the diseases they caused as described…” and not confounding aetiology and epidemiology, syndrome and disease?

    You need to understand mechanism to separate causation from association (which was the good point in the original post). But if you always need a controlled experiment, as opposed to a testable hypothesis, there are a number of sound sciences in big trouble.

  • toolkien

    So it isn’t reasonable to blame the epidemiologists for what’s done with their work. And they’d be better at it if subjected to proper critical analysis, rather than a choice of panic or flat disbelief.

    The thread begins with the notion of ‘dishonest’ scientists and I was working from that premise. The scientific community has self-interested motives for skewing data, the most basic reason that comes to mind is funding. I am of the opinion that so much of ‘science’ as it is defined today, is ensconced on college campuses, with a relatively cloistered culture, dependent on the ‘right’ sort of conclusions be drawn or funding will be diverted to those who will. Perhaps not all situations of falsified research rise to the level of the cold fusion blinkering a few years back but it is very prevelent IMO. Unfortunately scientific statements carry sacerdotal weight and remain unexamined (or perhaps unexaminable) by politicians, journalists, and the lay consumer much too often, which itself isn’t the unfortunate part of the situation, it is that the Force of the State is co-opted by certain forces based on intentionally faulty data.

  • A Specialist

    I’m glad to see that my comments generated some discussion. Some of my responses:

    To Guy Herbert:

    Actually, there has been a very clear decline in SIDS incidence since the institution of the “Back to Sleep” campaign in the U.S. (http://www.keepkidshealthy.com/welcome/safety/back_to_sleep.html). Just as John Snow did not know about Vibrio cholerae but knew to turn off the Broad Street pump, so do we know to turn babies on their backs without knowing precisely why it works.

    To the Wobbly Guy:

    “Only a controlled reproducible scientific experiment can properly determine cause and effect. That’s what I believe.”

    It’s unethical to intentionally expose people to substances thought to be harmful. That is why epidemiologists must rely on observational data. As I said before, there are criteria that indicate causality; together with lab and other data, they can be very convincing even absent controlled experiments. Wobbly Guy made a fine epistemological point, but a useless one in the real world.

    To Toolkien:

    Please don’t take the disparagement of thousands of people so lightly.

  • toolkien

    To Toolkien:

    Please don’t take the disparagement of thousands of people so lightly.

    I will if what they say doesn’t translate into poor public policy sacrificing billions in economic productivity and when transfers of my property are not required to fuel their research, and cozy relationships between the federal bureacracy and researchers comes to an end (I’m not enamored of business insinuating itself into the State, but I find it alarming how the transfers to researchers (based on traditionally left-leaning college campuses) gets a free pass). I have a high regard for science based on the scientific method, but have little regard for its alchemic cousin that passes for science today.

  • I can’t find the link offhand, but I saw a story linked from google news recently about lung cancer risks and genetics. The scientists in question had found a marker gene that correlates with getting lung cancer in smokers. Smokers with the gene were something like 126 times more likely than smokers without it to get lung cancer.

    Somehow I suspect the nanny statists to not like a cheap test for the gene, since if you’re negative, you can smoke and only probably get emphysema and friends, but not cancer. As a smoker, I’d personally really like to know if I’m risking The Big One or not!

  • Bernie Greene

    To all who read and made comments – Thank-you.

    To Logan Spector:

    If I offended with the generality – “That was a defining moment in the history of tobacco and it has been in decline ever since. Unfortunately so has epidemiology.” – then I apologise. In looking at it again I think it was an unfair statement as a generality.

    It is obvious that epidemiology can be used to find out useful things that can solve problems. The SIDS issue is a good example.

    The piece was not intended as a defense of tobacco smoking nor a summary of all that is known about it, but as a way of showing there is often a lot more to epidemiological studies than we get to hear about via the media, as filtered through the agendas of those with vested interests.

    It is often reported as an implied certainty that the subject of a study IS CAUSED by one of the many villains of the day, yet all that is presented in the way of data is a “relative risk” that can be made to seem very alarming and gives excellent grounds for the government to step in and “do something”. I wanted to show how a relative risk can be calculated and that it’s significance cannot he judged without having all the statistics to hand that were used to calculate it.

    It was never meant to be a definitive piece on smoking and lung cancer. Of course there have been later studies, though the particular one I had in mind (Hill/Doll) is still highly regarded and, to my knowledge, has not been invalidated. The point was that at a very early stage in tobacco research, and with a number of very significant questions unanswered, these researchers decided that the government should do something about tobacco smoking. Their research to that point is not in question. Their conclusions and subsequent actions are. I’m not certain of Hill’s involvement but Richard Doll went on to play a big part in the formation of ASH.

    On your point about causes coming in different types I deliberately bypassed discussing such things as I think they can be misleading. Rather than labeling one thing necessary and another sufficient I think it more precise to make a complete statement that includes all factors necessary to be sufficient, and call that whole statement a cause. I think it is easier to understand that way.

    I am also aware of Hill’s criteria (but could argue about some of those too :-))

    The problem with declaring smoking as a cause of lung cancer is that it is an incomplete statement. It doesn’t explain why so many smokers don’t get lung cancer. If it were found that a certain gene in lung cancer victims is in some way different than in people who don’t get lung cancer and put together with smoking triggers the disease then we could give a complete statement on what casues lung cancer in smokers. There must be a predisposition of some kind. Smoking is not sufficient to do it alone.

    To The Wobbly Guy:

    Your estimate of 4 to 8 times relative risk, as opposed to 24, is probably a bit low in my opinion but this is only a relative risk against non smokers.

    The risk should relate to how much one smokes, the dosage over a period of time, and it is this that adds some credence to the idea of smoking being a cause of lung cancer. Having said that, have you or anyone else ever seen any kind of “Tips for smokers concerned about their health” that gives any kind of table that might be used for assesing your risk? Such a thing could be produced and would be incredibly helpful in improving the health and loweing the likelihood of nasty diseases for smokers. It would include a number of other factors that may be involved with contracting various diseases. There is a concept called “pack/years” which could be used that way. A pack/year is one pack of cigarettes per day per year, or a pack/year could be 2 packs per day for six months etc. The more pack/years you accumulate the greater your risk.

    The science could be a lot more useful to smokers if they knew anything about it. But all we usually get is exaggerated claims in health warnings and higher taxes. Most of the studies into smoking are used by groups like ASH whose agenda is to stop us smoking and certainly not to give us any information that might help us minimise our health risks. It must be said that there are apparent risks in smoking. Some of the diseases that are possible can be very painful and even fatal. When I talk about minimising risks I mean lowering the number of chances of something nasty happening. It could still happen and would be just as nasty, but it is less likely.

    To Doug Collins:

    Yes of course you are right. It could be the presence or absence of something. There was a recent report of a study in genetics about longevity recently that found such a thing.

    Living Longer and Larger: It’s in the Size of Cholesterol-Carrying Molecules
    Source: New York Times, 2003-10-15

    Here is an interesting and telling quote from the researcher, “If you have this gene,
    you can smoke and you can be fat and you can not exercise. This sounds to me terrible”

    Well it is terrible English but why would it be terrible that you could do those things with impunity if you had this gene?

    There may be other factors apart from genes but it will be an interesting thing to watch what is done with such research. One obvious line to follow would be the possibility of developing some kind of drug that would compensate for the lack of that gene. If such a thing were developed and it were proven, maybe after many decades, that you could smoke with total impunity I wonder what folks like ASH would say about it.

    To Michael Last:

    On the experiments with mice. There have been lots using tobacco smoke on mice, rats, dogs and other creatures. None of them, to my knowledge, came up with much evidence of any kind of damage when just using smoke, even at very highly concentrated levels and constantly for days on end. However when they shaved off the fur from, I believe it was mice, and painted tobacco tars on the bare skin they were able to bring about the desired result of cancerous growths. So they can make a claim about tobacco that is true but if they told us the whole story as to how they arrived at it we would laugh, after being sick with disgust at what they put the animals through to get it.

  • Inspire 28

    Tell us about the epidemiology on 2nd hand smoke. I just saw a commercial touting the potential of 2nd hand smoke to cause all the same things that primary smoke caused. This from the California health boys. In engineering terms, they are full of shit and they know it, but they got a whole bunch of money to make commercials.

  • Old Patriot

    The smoking discussion is a bad one to try to use to show the “let’s get the solution down pat, then go work on the data” mentality of some groups (especially government-funded groups). Smoking and health is much an individual thing, and cannot be directly connected to a single vector, like the well used in the opening analogy. A much better case can be made on the new dragon to be slain, global warming.

    The problem in either case is that those who grasp at such villany hate to give up, and will do whatever is necessary to force their opinion, regardless of the data presented. The even bigger problem is premature exposure of data, when the number of factors is large and the known criteria limited. Each new day brings more and more information on how climate on Earth is influenced, and how it responds to that influence. New information indicates the sun is a greater player in climate variability than previously thought, that water vapor plays a significantly greater role not only as a greenhouse gas, but also as a buffer, and that even cosmic events affect our weather. Still, the people who have chosen to put their efforts behind reducing CO2 production have failed to change their ways, just as some smokers refuse to be swayed by statistics or other information. People believe what they want to believe, and continue to believe it until forced to change their opinion due to something that happens to them PERSONALLY – like watching beachfront property shrink, or they get cancer.