A Bit of an Altercation with an Anti-smoking Vaper

I am at a bit of a loss what to talk about tonight, so perhaps a bit of ephemera is in order.

A commenter pointed me to:


The commenter told me that Masters had used the Doll and Hill ‘Hospital Study’ as a/the seminal paper on smoking and lung cancer. (Seminal means ‘seed-like’, in the sense that it started off the plethora of smoking-related papers which have avalanched over the past several decades) I pulled Masters up a bit on the grounds that the ‘Hospital Study’ was a minor affair and that the ‘Doctors Study’ was much more appropriate. Nowadays, we would call the ‘Hospital Study’ a ‘proof of concept’ study, which served the purpose of justifying the expense and effort involved in the ‘Doctors Study’. It was never actually stated to be so, but it is really rather obvious. The Hospital Study was planned in 1947, and the Doctors Study began in 1951. It is obvious that the two studies were related. It is impossible to believe that planning for the Doctors Study was not already under way either in concert with the planning of, or during the process of, the Hospital Study.

The Hospital Study  was not the ‘seminal’ study – studies in pre-war Germany preceded the Hospital Study and Doll was familiar with them.


You may have noticed that I referred to the person involved as ‘Masters’. That is his surname. In my first comment there, I called him ‘Mr Masters’, being polite. He was upset by that, demanding that I call him Doctor Masters because of his doctorate in maths. Nothing on his site says that he has this doctorate. When I pointed out that I could not know about his doctorate, he said that he was miffed because my comment seemed to portray him as an ignoramus. I suppose that even ‘doctors of mathematics’ can be touchy.

In his post, Dr Masters had said that tobacco kills half its users. I told him off about accepting Tobacco Control Industry slogans (perhaps that’s why he was miffed). Perhaps he was also miffed when I said that, mathematically, given that there is an average age at death, that 50% or deaths must be ‘premature’, and 50% must be ‘postmature’. (Only one person in the UK, or even the world, occupies the exact moment of average)

The Doctors Study did not prove that smoking causes lung cancer. It merely revealed that more smokers die before the theoretical average than do non-smokers. That is all. Lots of smokers die after the theoretical average, but not as many as non-smokers. Give or take a bit, all non-smokers die from the same conditions (diseases?) as smokers.

To give him his due, Dr Masters debated. I would have liked to continue the debate, but there was no point. I drew his attention to the McTear Case, but he would not accept that The Tobacco Control Industry failed to produce evidence that smoking causes lung cancer. He obfuscated. He says that he did not, but by quoting stuff that the Judge said about awareness of warnings on cig packets etc, and saying that this implied that the Judge knew that smoking causes lung cancer, he did indeed obfuscate. the whole point of the McTear Case was that Tobacco Control had the perfect opportunity to ‘prove’ (on the balance of probabilities) that smoking causes lung cancer. They could not produce ANY evidence that it was so. They relied upon the ‘authority’ of the WHO and similar, which means relying upon epidemiology.

To me, reliance upon epidemiology is much the same thing as guessing that malaria is caused by ‘miamas’ from swamps.

Finally, I have no trust in Doll at all. He was a communist and probably totally anti-capital, and therefore anti privately-owned big business. I cannot help but feel that people like him ‘captured’ the WHO at its inception.

I do not quite know what to make of Doll’s definition of ‘a smoker’. In the Hospital Study, he defined a smoker as a person who smoked at least one cigarette (or equivalent in pipe or cigar) per day for a whole year. As far as I know, that was the standard applied in the Doctors Study also. But suppose that a person smoked, say, five cigs per day for only one year? He would then be counted as a smoker, and, subsequently, an ex-smoker. Thus, taking into consideration the prevalence of smoking in the 1950s, 1960s, 1970s, almost everyone would have started off as ‘a smoker’. Thus, anyone who got lung cancer, and had smoked a little for a full year, would have succumbed to lung cancer as ‘a smoker’, no matter how little the smoking or how short the term, provided that the term was at least a year.

It isn’t always easy to get your brain around the figures, bearing in mind the definitions. Suppose that an individual doctor, at the start of the study, identified himself as ‘smoking five cigs per day’ and had been doing so for two years. That particular doctor was known by name and address. He slots into the category of ‘smoker’, albeit ‘light’ smoker. If he stops smoking, in the Doctors Study, he DOES NOT move into a category ‘ex-light smoker’ – he moves into the category ‘ex-smoker’. Thus, almost all persons who get lung cancer will have been smokers at some time, even for a brief period. Non-smokers will have been an elite. What other ‘bad habits’ did that Elite avoid? Where did they live? To what extent were they exposed to smog and war?

I have statistics (which I cannot be bothered digging out) which show that smoking has decrease in prevalence over the decades since about 1960 from about 65% of males to some 25% of males. I have statistics which show that lung cancer deaths have reduced over that same period by around 25%. I would need to look them up again. Of course, surely, over that period of time, earlier diagnosis and better treatments will have reduced mortality. But the simple fact is that the reduction in lung cancer deaths has not kept pace with the reduction in smoking.In the extreme, there are almost as many lung cancer deaths NOW as there were when smoking was at it height, in the period since WW1 and through WW2.


I must dig out the stats again. But nothing is simple. The question of immigration arises – what are the ages of immigrants and does it matter? I suppose that the naked figures are the first thing that need to be revealed. I think that Tobacco Control is batting on a very sticky wicket and it know it, which is why it is pushing so hard for legislation NOW. God only knows why Cameron et al succumb. It really is weird that they say, “Oh. Go on then. It isn’t important. Ban smoking in cars. Introduce PP. These are not important things, so why not? Who cares? They will not influence the general election”

But they WILL influence the GE. Everyone is sick to death of grinning puppets. But there is a long way to go before the Bullington Club id unseated.

This was supposed to be a short post! Forgive typos for tonight.






20 Responses to “A Bit of an Altercation with an Anti-smoking Vaper”

  1. Anthony Masters Says:

    I wish to correct a few inaccuracies about your description of my person. I would also like to begin by saying how much I enjoyed our polite and civil conversation.

    Firstly, the title of your post says I am an ‘anti-smoking vaper’. I am not an anti-smoker in the legislative sense, as I would hope the title of my blog would illuminate. It is not an activity I would wish to undertake, but neither do I believe in the use of force to punish those who wish to smoke. Restaurants, pubs and workplaces can come to their own arrangements. Also, I have never used e-cigarettes.

    Secondly, you write: “Nothing on his website says he has this doctorate.” The first sentence about my About page reads as follows: “I am a marketing analyst with a Mathematics PhD, and I am a passionate believer in liberty.”

    There are also posts on the blog about my PhD.

    Third, I sincerely apologise if I was “touchy”.

    The start of your first comment on my page states:
    “Oh Dear, Mr Masters.
    It is a matter of fact that, shortly after WW2, there was a continuous and fairly substantial increase in the import of oranges into the UK. At the same, there was a continuous and fairly substantial increase in divorces. Did the increase in the import of oranges cause the increase in divorces, or did the increase in divorces cause the increase in the import of oranges?
    That is a good example of why correlations do not constitute causation.”

    Again, I apologise if I overreacted, but the idea that some correlations are spurious, and so correlations do not necessarily imply causation, is well-known. However, ‘correlation is not necessarily causation’ does not mean that data analysis ends. There are many methods to investigate a data set.

    • junican Says:

      May I call you Anthony? I shall presume so.
      Anthony, I did observe the title of your blog ‘In defence of liberties’, but I did not ‘check’ your ‘About’ page. Perhaps I should have, but it would not have made a lot of difference, other than, perhaps, a slight change in mode of address. Whatever, but I most certainly would not have made a point of addressing you as ‘Doctor’ Masters, in just the same way that I would not have expected you to address me as ‘His Most Eminent Junican’.
      So let us forget all the stuff about titles.
      Fisher described certain factors which were important to ensure that correlations were, in fact, causal. He said that simply repeating the same investigation adds nothing to ‘causal’ argument. Thus, it does not matter if you repeat the same study with the same methodology several times in several countries – you will get the same results because you are doing the same things. For example, in the West, it is not unlikely that the increase in the import of oranges coincided with an increase in divorces everywhere in Europe. I do not imply that the smoking studies were incorrect merely for that reason. It is simply worth noting.

      I accept that there are methods of analysing data sets, but problems arise when the analyses become reliant upon smaller and smaller numbers of people in each subset.

      What is of the GREATEST importance is that epidemiology cannot. in itself, decide CAUSATION.

      • Anthony Masters Says:

        Yes, I think too many words have been typed about titles.

        Let’s assume that we see a persistent correlation between orange importation and divorce levels within the whole of Europe. This would lend credence to the idea that the relationship is not spurious: I would posit that population growth drives both variables.

        It is important to analyse plausible confounding factors when a correlation is observed. Again, the distinction between correlation and causation is not synonymous with the assertion all correlations are spurious. It is true that margins of error do widen as smaller data sets are considered: https://anthonymasters.wordpress.com/2013/07/11/statistics-and-lampposts-i-polling-and-confidence-intervals/

        “What is of the GREATEST importance is that epidemiology cannot. in itself, decide CAUSATION.” Data analysis of the epidemiological kind can strongly suggest causation, which has to be considered when a person wishes to decide “on the balance of probabilities”. For instance, if it were proposed that performing action X causes disease Y, then we would expect the segmentation by proclivity in action X to yield higher incidence rates of disease Y.

        Moreover, epidemiology can stimulate clinical and laboratory research, which would then bolster a causal hypothesis.

  2. garyk30 Says:

    Smoking rates and Lung cancer incidence

    It is acceoted that adult smoking rates in the USA fell from about 44% to about 20% during the time period of 1965 to 2010.

    Lung Cancer incidence rates have not followed suit.

    Table 8: Respiratory Cancers – Age-Adjusted Incidence Rates 1 by Site, Race and Sex, 1973-2006 per 100,000

    Lung & Bronchus
    1973………… 49.0

    1990………… 68.2

    2006………… 60.0

  3. garyk30 Says:

    ” tobacco kills half its users”

    Hmmmm, only half of smokers will die from the diseases ’caused’ by smoking?

    Or, half of all smokers deaths are caused by smoking? That would mean that only half of smokers deaths from lung cancer would be due to smoking and that half of smokers deaths from shark bite are actually caused by smoking?

    The dear Doctor should be so kind as to provide a mathamatical reason for his statement.

    • Anthony Masters Says:

      It is a simple matter of attribution. If you accept the premise that there are risk factors for developing lung cancer (as the above article explicitly does), then you can attribute those deaths to their risk factors.

      Not everyone died in the Great Smog of 1952, but that smog was certainly a cause of death: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1241789/

      More clearly, not all people die from being hit by a car, but vehicular manslaugter may be considered a cause of death.

      • garyk30 Says:

        “vehicular manslaugter may be considered a cause of death”

        “vehicular manslaugter” is a legal term that would cause a death certificate to be invalid if that term was listed as one of the underlying causes of a persons death.

        You seem to be stating that ‘half will die from their smoking’ implies that only half of smokers deaths from lung cancer will be due to smoking.

        The ‘half will die’ claim comes from here.
        Mortality in relation to smoking: 50 years’ observations on male British doctors
        Richard Doll, Richard Peto, Jillian Boreham, Isabelle Sutherland


        “What is already known on this topic:
        About half of all persistent cigarette smokers are killed by
        their habit—a quarter while still in middle age (35-69 years).”

        This seems to be justified by the chart on page 3 that shows that smokers have about double the overall
        mortality rate as never-smokers.

        That is 35.4 deaths per 1,000 people per year vs 19.38 for never-smokers.
        The difference of 16 deaths per year is 45%(about half) of 35.4.

        Obviously, the 50% would stupid if applied to car crashes, shark attacks, or causes of death other than from the diseases said to be ’caused’ by smoking.

        When the deaths from the diseases ’caused’ by smoking are considered:
        never-smokers = 84% of all deaths
        current smokers = 85% of all deaths

        Thus, both groups have the same probability of dying from the diseases ’caused’ by smoking and the same probability of dying from other causes.

        Doll’s group were either deficient in math or highly biased against smoking.

      • Anthony Masters Says:

        Apologies. It took an Excel file to realise what you are doing.

        In Table 1: for smokers, the proposed connected diseases account for an age-adjusted mortality rate 30.21, out of 35.35, that is, 85.5%.

        For non-smokers, the equivalent figures are 16.20 and 19.21, respectively, meaning 83.8% of all deaths.

        The reason that this difference does not appear overtly dramatic is two-fold: an increased mortality rate will affect both the numerator and the denominator. As an example, I produced a third column that used the smoking mortality rate for the supposed connected diseases and the non-smoking rate for everything else. This calculation yields an overall mortality rate of 33.35, with ‘associated’ deaths equal to 30.21, meaning the percentage is 90.6%.

        Secondly, the percentages are similar because smokers suffer higher mortality rates for other causes of death. It is difficult to assimilate such a finding with the idea that smoking has no effect on mortality.

        I may have been quite flippant in my earlier comment: for that, I apologise. Attribution for deaths will necessarily depend on the model underlying the calculation. Each disease has a base rate of manifestation, which may be heightened by the exposure to risk factors.

      • Tony Says:

        I’m afraid there’s a mathematical flaw in the concept of attribution of disease based on ‘risk factors’.
        I wrote about it here: http://www.f2c.org.uk/blog/2013/07/14/the-alice-in-wonderland-world-of-risk-factor-epidemiology/

      • Anthony Masters Says:

        Thanks for the comment.

        Modelling is an iterative process. This is not a “flaw in the concept” of disease attribution, but merely a recognition that such modelling will be adjusted if new factors are discovered.

        An analysis of known factors would first recognise their respective prevalence within the population, their relative risk factors, as well as interactions between these variables. The resulting model would seek to explain their variations observed within the population, before being tested, modified, re-tested, and so on.

        It would be better such figures were published as confidence intervals, given other risk factors may be identified in the future.

      • Tony Says:

        Thanks for your reply.

        Unfortunately, as the number of factors is potentially infinite these iterations would go on forever. What’s more there is no reason to suppose that the earliest discovered risk factors are the most prominent. Nor are interactions between them easy to determine.

        Hence any attempt at attribution, or at least attempting to put a figure on it, is doomed to failure. And that is even using the multi-risk factor equations which are almost never invoked by Epidemiologists.

        Some Epidemiologists do recognise this and accordingly, avoid attribution based on risk factors.

      • Anthony Masters Says:

        I am not a epidemiologist.

        Attribution is possible, and is undertaken in other settings, such as business intelligence and digital analytics.

        There are not “potentially infinite” factors, and factors with no predictive power, or those explained by existing variables, would be removed. The process of validation (against a fixed number of differing, randomly selected subsets of the original data set) can be finished. However, models do require retesting as times goes on.

        There is a distinction between the human limitations of knowledge and attribution modelling being “doomed to failure”.

      • Tony Says:

        Thanks for that interesting link on marketing.

        I accept that such methods are indeed useful in marketing. You’ll have a relatively small number of factors under your control and it is only those ones that are of interest for that reason. So it may well be reasonable/helpful to attribute X rise in sales to Y, at least to inform future campaigns.

        However I think attribution of disease by ‘risk factor’ is a different matter. It is quite wrong for an Epidemiologist to state that X number of deaths is attributable to ‘risk factor’ Y. The proportion cannot be known from ‘risk factor’ figures and there can be no implication of causality which is how they interpret their figures. I also still maintain that the potential number of factors is to all intents and purposes infinite.

      • Anthony Masters Says:

        Attribution calculations are always limited by what is currently known.

        Your two criticisms of these calculations are mutually contradictory: either epidemiologists should restrict their pool of risk factors to those with an experimentally-justified causal link to the disease, or they should widen their pools to include everything in the known universe. These two criticisms cannot be satiated simultaneously.

    • Tony Says:

      You say “either epidemiologists should restrict their pool of risk factors to those with an experimentally-justified causal link to the disease… “

      The trouble is, they don’t have any such risk factors for most of their studies.

      Taking smoking as an example, there is no experimentally-justified causal link to e.g. lung cancer. All animal experiments have failed as well as all human intervention trials such as MRFIT. And this is after over 60 years of trying. I’m talking here about active smoking rather than so called ‘passive smoking’. Yes, there are experimentally verified carcinogens in tobacco smoke. However to achieve dangerous exposure, an active smoker would need to smoke around 100,000 cigarettes a day.

      Even if they did have some experimentally verified factors, they could hardly exclude everything else.

      Hence their pool of possible risk factors does include everything in the universe.

  4. garyk30 Says:

    The rate keeps growing.

    Tobacco smoking and all-cause mortality in a large Australian cohort study: findings from a mature epidemic with current low smoking prevalence
    Compared to never-smokers, the adjusted RR (95% CI) of mortality was 2.96 in current smokers..

    Cause of death information was not available at the time of analysis.


    In Australia, up to two-thirds of deaths in current smokers can be attributed to smoking.

    Note: That is 67%, they say that a RR of about 3 means a 67% probabability of ‘causation’.
    2 is 67% of 3.

    Lung cancer caused by SHS exposure has a RR of 1.25.
    0.25 is only 20% of 1.25.

    So, the Aussie study has just shown that if a never-smoker exposed to SHS gets lung cancer, there is only a 20% probabliity that lung cancer was ’caused’ by the SHS exposure.

  5. A Rejoinder to the BSC | In Defence of Liberty Says:

    […] and developing lung cancer, a commenter wrote on the Bolton Smokers’ Club blog about me: A Bit of an Altercation with an Anti-Smoking Vaper. This article fulminates with fallacies and falsities, some of which are about me […]

  6. junican Says:

    This might interest you, Anthony:


    The headline is:

    Coffee intake linked to reduced risk of MS

    The result:

    From the US study, the team also found that participants who did not drink coffee in the year prior to symptom onset were approximately 1.5 times more likely to develop MS, compared with those who consumed at least four cups of coffee a day.

    As it happens, my wife has MS, so I have an interest.

    I’m going to write a post about it.

  7. junican Says:

    One last thing (I hope!) on this theme for tonight. The Charlatans do not give a toss about Health. THEY MAKE MONEY. In vast quantities. Science still exists, but is useful for industry. But there is also pseudo-science, which is what Government uses to enslave the people.

Comments are closed.

%d bloggers like this: