The Doctors Study 1951 – 2001
02/09/2012
The first part of this post is concerned with the ‘Preliminary Report’ which was published in the BMJ in 1954. It concerned the first 29 months of the study.
Introduction
In 1950 Doll and Hill completed the Hospital Study (see sidebar). [For brevity, I shall refer only to Doll from now on] That study was planned in 1947 and the results were published in the British Medical Journal in 1950. The Hospital Study was a ‘retrospective’ study, which means that it ‘looked backward’. That is, people with lung cancer were asked about their smoking history in the past and the results were then analysed. The problem with that type of study, in connection with smoking and lung cancer (because of the long lead time before the effects show themselves), is that people find it difficult to remember with any accuracy how much they smoked, say, ten years before. About that time, Doll and Hill decided to do a ‘prospective’ study, which means ‘looking forward’. In that type of study, people are asked what they smoke now and they are followed for some years to see how smoking might affect their health and possibly cause their deaths. This study is called ‘The British Doctors Study’. It continued for fifty years, although Doll said that it was not originally intended to keep it going for so long.
Towards the end of 1951, a questionnaire was sent to every doctor on the medical register, which amounted to some 60,000 doctors. The British Medical Association sent the letters out for him. Some 40,000 responses were received. Doll was a little concerned about the missing 20,000 since it might cast doubts upon whether or not this study could be considered to be representative of all doctors, but he discounted that idea on the grounds that his purpose was to compare the incidence of disease between smokers and non-smokers and that the numbers were sufficient for that purpose. Among the 40,000 were some 6,000 women. He decided to confine himself to the male population of doctors for simplicity. After excluding the women and a few responses which were incomplete, he finished up with about 34,000 male doctors.
The questionnaire was intentionally very simple. Doll said that he deliberately kept it simple to encourage as many responses as possible. Simply put, the questionnaire asked:
a) Are you a non-smoker? (Doll allowed for a small amount of smoking provided that it did not amount to more than ‘one cigarette per day for as much as one year‘).
b) If you used to smoke, when did you stop and how much were you smoking at that time?
c) If you are a smoker, do you smoke a pipe, a mixture of pipe and cigarettes or just cigarettes? Also, to say how much tobacco and/or cigarettes you smoke daily.
Once Doll had received as many responses as he could reasonably expect (which took about six weeks), he then started to enquire as to the fate of the doctors. He was looking for deaths and causes of deaths. He had arranged with the Registrar of Deaths to let him know whenever one of the doctors on his list died and the cause of that death. Since his principal concern was with lung cancer, he made further enquiries from various consultants as to the certainly of the diagnosis.
In 1954, he published the first of four reports in the BMJ (British Medical Journal). I am going to give the URLs of all four reports:
Published 1954: Period – first 29 months
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2085438/?page=2
Published 1964: Period – 10 years 1951 to 1961. (Part 1 – Part 2 one week later)
http://www.bmj.com/highwire/filestream/238745/field_highwire_article_pdf/0/1399.full.pdf
Published 1964: Period – 10 years 1951 to 1961. (Part 2).
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1814697/
Published 1976: Period – 20 years 1951 to 1971.
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1690096/?tool=pubmed
Published 1994: Period – 40 years 1951 – 1991.
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2541142/?tool=pubmed
Published 2004: Final report. Period 50 years 1951 to 2001.
Click to access bmj32801519.pdf
The first report that I read was the last one, as most people will have. I then read the 1954 report. Before summarising that report, I decided to read all of them. I’m glad that I did, because the later reports shed some light upon where Doll was heading – probably, from the beginning of the whole project. You see, further questionnaires were sent out asking for more information. For example, inhaling was introduced (it was in the original Hospital Study, but was not included in the Doctors Study until later). Also, questions were asked about cigar smoking. Further, although in the 1954 report he referred to other diseases, as the number of deaths increased, it was possible to include more detail about other causes of death and produce data showing that smoking ‘causes’ almost every terminal illness known to man. As time passed and more and more data became available, Doll became more and more hysterical. The mantras and slogans which we now see repeated over and over began to appear, such as, “Smoking causes half of smoker deaths”. We shall pick them out as we plough our way through the dense thicket of statistics. As regards the statistics, fear not, since I have no intention of reproducing them all!
—————-
THE 1954 REPORT.
During the 29 months from mid 1952 (when all the responses had been received and the monitoring systems set up), a number of deaths occurred. Doll was therefore able to make this preliminary analysis.
He begins with a list of nine other studies on the subject of smoking habits between 1950 and 1953. These were all ‘retrospective (looking backward) studies. He claims that enough retro studies had been done, and it is time for a ‘prospective’ (looking forward) study. He explains how he recruited the ‘cohort’ (all the doctors who had responded) and the arrangements for collecting deaths information.
[Can I say at this point that I am going to avoid a lot of the verbiage and concentrate on the Tables, because they contain the important information, but I shall link them. I shall not include all the information in the tables.
I shall now go to the past tense.
He claimed that, unlike the previous Hospital Study, there was no need to go into a person’s smoking history in detail and that the experience gained from the Hospital Study showed that the most recent smoking activity was sufficient to present a true picture.
Remember that Doll had excluded women and had 34,000 male doctors in his ‘population‘. [I should explain here that, in epidemiology, a ‘population’ is the group of people who you are studying. It can be large or small. For example, if one was studying Members of Parliament, the ‘population’ would be the 650 MP s] He decided to exclude the under 35 year-olds in this first study because they rarely suffer lung cancer. There were about 10,000 doctors under 35 years old and so the ‘population’ which Doll was interested in (until the under 35s became old enough to be involved) was reduced to 24,000. Of course, that is a lot of people and sufficient for the purpose for the time being.
Table 1.
Shows the age distribution of the doctors in ten-year groups (35 to 44, etc) and the amount of tobacco smoked. For example:
[NS = non-smoker. g p/d = grams per day. (NB. 1 cig = 1g. 1 ounce of pipe tobacco = 4g.]
Age………NS….1-14 g p/d…. 15-24 g p/d……25 g p/d + ……Total.
35 – 44.….1457.…..2864.……….2888.…………1716.……….8925.
Etc.
————————————————————————-
Totals of
Columns….3093.….8431.……….7662.…………5203.……..24389.
What we should select from these figures and bear in mind is that smokers outweigh non-smokers by about 21,000 to 3,000. Of course, Doll takes this into account when he calculates death rates, but we should keep it in mind anyway.
Having broken the doctors down into the above age groups and quantity of tobacco groups, he then broke down the crude quantities of tobacco into types of tobacco smoked in Table 2. Here is the first line and the totals:
Age………Pipes……….Mix pipe and cig……Cigs…………Total.
35 – 44.…..1001.…………..1240.………….5227.…………8925.
Etc.
————————————————————————
Totals of
Columns….3558.………….4504.………….13234.………21296.
So we now have:
a) Numbers of non-smokers and smokers in age groups.
b) Smokers broken down by amounts of tobacco smoked.
c) Smokers broken down by method of smoking.
In what follows, Doll will use these breakdowns to analyse death rates by age, amount and method.
During the first 29 months, out of the total 24,389 NS (non-smokers) and S (smokers), 789 men died. Of these, 35 were lung cancers (being, very roughly, 5%). Doll described in Table 3 how he certified, as best he could, that these 35 deaths were definitely lung cancer. No need to reproduce it.
——————–
We are now getting to the nitty-gritty of the report – the proportions of deaths from various causes according to whether people were non-smokers or smokers. Also, the proportions for smokers of different amounts of tobacco. But there is a problem. Clearly, the older that people are, the more likely they are to die. How can you compare a death rate from cancer in old people with a death rate among young people? For example, suppose that there were two groups of people, 10,000 for each group. Let’s call them Group A and Group B. Suppose that you were told that the number of cancers in Group A, during the course of a year, was 2, and the number in Group B was 1,000. Might you not say that there must be something seriously wrong with the people in Group B? But what would you say if I then told you that Group A is taken from school children and Group B is from pensioners? Do you the point?
A very long time ago, a method was found to avoid this problem. It is called “Age Standardisation”. It involves some messy equations and calculations, but I can give you an example to give you the idea.
Suppose that you have a group of 1,000 people aged between 45 and 49, of which 12 died from cancer. Suppose that you have another group of 1,000 aged between 65 and 69, of which 14 died from cancer. Is there a way by which you can directly compare the ‘significance’ of cancer in the two groups? One way (and it really is simply to enable a comparison of some sorts) is to take the age distribution in the general population of 45 – 49 year olds and 65 – 69 year olds. Now suppose that, out of 1,000 people generally, there tended to be 80 in the first age group and 60 in the second (the total of all the individual numbers in each age group will add up to 1000, of course). You can ‘standardise’ the death rate in each group by multiplying the actual number of deaths by the ratio of people in that age group as a proportion of the total number (in this case, 1,000 people altogether when all the numbers in each age group are added up).
So, we get this calculation:
Group
45 – 49.…….12 deaths times 80 per 1000:
12 x 80/1000 = 96.
Group
65 – 69.…….14 deaths times 60 per 1000:
14 x 60/1000 = 84.
Purely and only for comparison purposes, you can invent a nonexistent number of deaths for each group. In this case, 96 for the 45 -49 group and 84 for the 65 – 69 group. So you can say that despite the actual deaths in the 45 – 49 group (12) being less than the number of actual deaths in the 65 – 69 group (14), the rate of deaths is greater in the first group than it is in the second! You could even calculate the percentage rate of excess deaths at around 12%!
I have very much simplified the method purely for illustration purposes, but do you get the idea? There is nothing wrong with the idea of ‘age standardisation‘. Insurance companies, for example, use it all the time.
It is important to grasp the general idea for when we look at Doll’s figures for death rates in what follows.
——————
Doll then started to apply ‘age standardisation’ to his age groups of deaths in the doctors. But it gets even more complicated because he also has to allow for the distribution of deaths among his non-smoking and smoking categories and also differentiate them by amount of tobacco smoked!
I propose to skip all the complicated mathematical stuff and accept that all the calculations are correct. It is unthinkable that Doll would have tried ‘to get away with it’ as regards fiddling the maths since any half-decent statistician would have spotted it immediately!
At this point, Doll produced a Table (Table 4) comparing lung cancer death rates with death rates from other causes. Remember that these death rates are artificial figures which are of value only for comparison purposes. I shall reproduce only three of the disease categories (out of the six) so that we can see what Doll was referring to:
Disease……Actual deaths……NS…….1g p/d…….15 g p/d ….25g p/d.
Lung canc…….36.……………….0.………0.48.……….0.67.……….1.74.
Heart probs….235.……………3.89.…….3.91.……..4.71.……….5.15.
Other canc……92.……………..2.32.…….1.41.……..1.50.………..1.91.
Apart from the ‘Actual Deaths’ column, the numbers are “deaths per thousand” (age standardised, percent standardised, etc). All that they are good for is to compare one with another.
What we can see immediately is a big change in the rate of deaths when comparing non-smokers with heavy smokers as regards lung cancer, especially from light smokers to heavy smokers. Heavy smokers have three times the death rate of light smokers. But we should note that the death rates are only a half a death per thousand people for light smokersand one and a half deaths per thousand people for heavy smokers in the lung cancer category.
As regards ‘heart probs’ (coronary thrombosis), we do not see a big increase, but we see a steady increase as people smoke more. On the other hand, we see that non-smokers have a bigger death rate for ‘other cancers’. Can I remind you again that the age at death of any individual has been taken care of by the ‘age standardisation’ process so that so there is no point in asking, “Ah, but how old were these people when they died?” As a sort of ‘back-up’, he then took the age group 55 – 64 in which 13 deaths from lung cancer occurred. He opined that, had smoking not be involved in these deaths, they ought to have been spread out over the groups of non-smokers and smokers in accordance with ratio of non-smokers to smokers. In that age group, the NS group was (about) 9%, the light smokers (1 g) were 34%, the moderate smokers (15 g) were 32% and the heavy smokers (25 g) were 25%, so he reckoned that, if it were not for smoking, the 13 deaths in that age group should have been spread out as follows:
Age group……NS………1 g p/d……..15 g p/d……….25 g p/d.
55 – 64.…………4.………….14.…………11.……………7.
(Those figures approx).
Whereas the actual distribution of deaths was:
…………………..0.………….12.…………11.…………..13.
He drew attention to the doubling of the heavy smokers and said that the numbers were statistically significant (but he said nothing about the lack of deaths in the non-smoking group!)
For simplicity of visualisation, he then produced a block chart showing that, even for other diseases which caused deaths, heavy smoking was always a factor, even if only a small one (except ’other cancers’).
He next briefly examined the difference between pipe smokers and cigarette smokers. He observed that the number of pipe smokers was less than should have occurred had the deaths been distributed according to how the doctors who had died had described their smoking habits. That is, statistically, comparatively more cigarette smokers had died (as compared with what should have happened, in theory) and comparatively less pipe smokers. Having said that, the numbers involved were so small that they did not attain statistical significance.
He observes here that none of the other five disease groups (heart problems, breathing problems, circulatory problems, other cancers and other diseases) show any significant difference between pipe smokers, mixed pipe and cigarette only smokers.
He then mentioned the suggestion that physicians might take more care to diagnose heavy smoker deaths correctly than non or light smokers, but discounts the suggestion because it would affect death rates in other categories, of which there is no sign.
————-
Doll concluded with a short discussion which contains some surmises about the possibility of some doctors ’under-reporting’ their smoking habits. He gave a couple of instances – one being a doctor who had reported that he smoked cigarettes, but who had previously smoked also a pipe; another was of a doctor who had reported that he smoked ’moderately‘, but the doctor who confirmed the diagnosis of lung cancer said that he was one of the heaviest smokers that he had ever known.
Finally, Doll briefly summarised. He said that the figures for lung cancer clearly indicated a connection with smoking – the heavier the smoking, the clearer the connection. There was also a less pronounced connection with coronary thrombosis, but little connection with other diseases – so far.
——————————–
That concludes my summary of Doll and Hill’s Preliminary Report based upon the first 29 months after the start of the Doctors Study. I hope that it is reasonably easy to read and understand. I do not propose to comment at this time, but I can make the following observations:
a) There was no mention of ’addiction’ – smoking is described as a ’habit’.
b) There were no slogans or mantras – no ’smoking causes half of smokers to die prematurely‘, etc.
c) There was an admission that, at that time, there was no apparent connection between smoking and any other disease (other than a tendency for coronary thrombosis deaths to occur a little more frequently in heavier smoker).
d) After only 29 months, the number of deaths was a very small fraction of the total of 21,000 male doctors over 35 years old. Such figures, after such a short period of time and in such small numbers for each disease category, can be greatly distorted by state of health of only a few doctors who responded to the survey. For example, a doctor who knew that he had lung cancer and was a smoker and believed that smoking had caused his lung cancer, might well exaggerate his level of smoking. Doll does not mention that possibility, but he does talk about doctors who might be very ill with one of the diseases which he mentions and do not complete the questionnaire, thus causing an under-estimation of deaths from the causes in the early days of the study.
And so we move on to the 10 year report which covers the period from 1951 to 1961. After 10 years, Doll and Hill had vastly more data to work with – and it shows!
THE ANALYSIS OF THE TEN YEAR REPORT WILL BE ADDED HERE ONCE COMPLETED.
Post of 10th Sept 2012: The Doctors Study.
I am currently studying the Doll and Hill Doctors Study 10 year report. For the uninitiated, Doll started the Doctors Study in late 1951 when he (and his colleagues) sent out 60,000 questionnaires to all the doctors in the UK to ask them about their smoking habits. He received about 40,000 replies. After that, he checked to see what they died from and drew statistical conclusions about the dangers of smoking, based upon the number of doctors who died from lung cancer especially, as compared with non-smoking doctors. He produced a preliminary report which was published in the BMJ in 1954. At that time, only a small number of doctors had died, but he was still able to produce some figures which seemed to indicate that smoking caused lung cancer.
In 1964, he published a second report in the bmj. By this time, after 10 years, there were many more deaths among the doctors who responded at the beginning. He used the causes of death to deduce an awful lot of statistics about smoking as a killer – not only regarding lung cancer, but also other ‘diseases’. After the 10 year report came the 20 year report and then the 40 year report, and, finally, the 50 year report.
The 10 year report is hard work. What I am trying to do is to render it comprehensible to us ignoramuses and, of course, to shorten it. That is, to pick out only the important things which apply to the situation as it exists now, and, in particular, to enable us to understand how the persecution of smokers came about.
This may take some time.
Post of 17th Sept 2012: The 1961 Reports in the BMJ of The Doctors Study THE 10 YEAR REPORT
Having reviewed the preliminary report of 1954, I shall now review the 1964 report of the results of the first ten years of the study (1951 – 1961).
The first thing that I need to point out is that there is no point in trying to compare this report to the 1954 report. The reason is that the 1954 report excluded those doctors under 35. That was because very, very few under 35s caught the disease of Lung Cancer. (I use the phrase ‘caught the disease of’ for lack of a better phrase!) So, in effect, we must treat this as the first report for all intents and purposes.
What we can do, to start with, is state again how many doctors responded to Doll’s questionnaire of October 1951. There were 40,637 responses altogether (out of a total of some 60,000 doctors in the UK). 34,445 were men and 6,192 were women. Because the number of women were comparatively small, Doll did almost all his work on the male doctors with only a brief reference to the situation as regards women at the end. Since this analysis is intended to be as brief as possible, I shall ignore any reference to the women – we gain nothing by including them except complications and even more tables.
May I remind readers that, in epidemiology, a ‘population’ refers to the group of people under consideration. For example, the 650 Mps at Westminster would be regarded as ‘the population’, if the study was about Mps, even though there are only 650 of them.
We start, therefore, with a population of 34,445 male doctors. We are going to see how they fared over the 10 year period from 1951 to 1961. We are going to see how many died and what they died from, and we shall see how many deaths for each disease were smokers and how many were non-smokers, but we shall have to take into consideration that there were many more smokers than non-smokers. The easy way to do this is to express deaths as percentages. For example, in the non-smoking group, 10 deaths out of 200 (if there were 200 non-smokers) is 5% and, in the smoking group, 120 deaths out of 1000 (if there were 1000 smokers) is 12%. Thus, in that example, it is clear that smokers died at more than twice the rate of non-smokers. But the report is far, far more complicated than that. But never fear! It is not my intention to even try to enter into detail. I shall pick out some detail here and there to illustrate Doll’s statistical analysis of death rates. If we do not simplify, then we shall become lost among the masses of data.
—————————–
As we begin to examine the 10 year report, let us never forget that Doll had a closed group. No new entrants into the group were accepted (or indeed acceptable) into the group. This means that, year after year, the whole group diminished. Remember also that, for simplicity, Doll considered only the male doctors. So we can start with these facts:
Number of doctors: 34,445 (males only).
Deaths at 1961: 4,597.
We should also remind ourselves of the breakdown of the doctors into groups. I cannot find a full breakdown as at 1951 because Doll ignored doctors who were less than 35 years old in his first report in 1954. The best I can do is this calculation of the breakdown in 1951 from 1958 figures:
Non-S….Ex-S…..Pipe….Cigs&Ors…..Cigs……Total.
5439.…..4812.….3613.….3803.……..13542.…31208.
(Some 2500 had died and a number had been ‘lost’, which accounts for the difference between the total at 1958 (31208) and the total at 1951 (34445)).
Bear those figures in mind.
———————-
Between Nov 1957 and Oct 1958, Doll sent a second questionnaire to the male doctors. He discovered how many had quit, how many had reduced or increased their smoking, how many former quitters had started again. He also received information about cigar smoking and mixed smoking, and asked about inhaling. He produced Tables 1, 2 and 3 showing these variations. No need to re-produce them since he decided not to incorporate the changes in the figures for the time being since the variations were mostly minor changes. He decided to use the initial distribution of non-smokers, light smokers, moderate smokers and heavy smokers that he obtained in 1951. He then started into the nitty-gritty.
His chief consideration was the comparison of death rates among these categories:
a) Smokers and life-long non-smokers.
b) Cig smokers and pipe smokers.
c) People who had given up smoking before Nov 1951 and continuing smokers.
d) The different amounts of tobacco smoked.
He then gave a list of percentages of death rates, but it is easier to go straight to the Tables which serve the same purpose than to extract that list. Table 4 gives the death rate figures for lung cancer.
[May I once again warn readers that he used standardised death rates. Standardisation of death rates is a legitimate statistical formula which enables comparisons to be made of death rates among different age groups. The important thing to remember is the number stated are not actual deaths. They are only useful for comparing death rates.]
Table 4.
Death rates per thousand per an from Lung cancer.
(The actual number of deaths in each group is in brackets)
Age…………NonS………….Cigs p/d.
………………………………….1 -14.……..15 – 24.……..25 +.
35-44.……0.05 (1)……….0.07 (1)….0.00 (0)……0.11 (1).
45-54.……0.00 (0)……….0.31 (3)….0.62 (9)……0.75 (8).
55-64.……0.00 (0)……….0.48 (3)….2.31 (20)…..3.88 (26).
65-74.……0.00 (0)……….2.69 (9)….5.16 (17)…..6.48 (14).
75 +……..1.11 (2)………..2.68 (6)….7.27 (8)……16.33 (8).
—————————————————————-.
All ages….0,07 (3)………..0.57 (22)…1.39 (54)….2.27 (57).
(NB. I have no intention of reproducing every table! I have reproduced the whole of this table only because it gives us a flavour of the tables and there are lessons to be learnt from it)
First we notice that he still hasn’t brought in doctors under 35. This may be because there are hardly any deaths from LC at all below that age. However, I am not sure whether or not it is because the 25 year olds were now 35 year olds. The text does not make clear what dates the age groupings refer to. Having said that, it is hard to think that the age groups can be other than as at 1961 (that is, the 25 – 34 group at 1951 are now the 35 – 44 group at 1961), otherwise, how can he say that X number have died in that group over the last 10 years? It really is very confusing.
Secondly, note the significance of standardisation. Look at the two figures which I have made bold. There were more deaths in the 65-74 group (14) than in the 75 + group (8), but, allowing for the number of people in that older group and the number who were ‘heavy‘ smokers, the death rate was much higher.
Doll himself said at this point that the numbers were too small to draw any general conclusions, but opined that the number in the higher smoking groups were indicative.
He next produced a chart showing death rates per thousand per an which shows that the death rate increases for cigarette smokers the more cigs that are smoked. The figures go from a half a death (per thousand remember!) for a smoker of 5 cigs per day to 3 for a smoker of 45 cigs per day.
—————–
I have had a think about what I can expect from continuing to try to get to grips with this report. To be honest, I see no advantageous point. Doll ET AL had gained the use of a computer to process the data. “We are grateful also to Miss Margaret Devine, who programmed the more complex data for analysis by the London University Unit’s computer Mercury” I wonder how sophisticated that computer was? That is not to say that it was badly programmed – it is more a question of whether or not Doll Et Al knew what they were doing. I say this because nowhere, either in the 1954 report or in the 1961 report, do they give a breakdown of the original figures for non-smokers, quitters, current smokers, and their ages, etc. Those figures are absent. For that reason, it is not possible to know just how many 70 year old or 29 year old doctors there were in the study. I dare say that one could work these figures out, but, to be honest, I see no point in wasting time doing so (I have tried!).
I have wondered to what extent Doll was ’playing with’ the computer. Shit in shit out. Did anyone ever examine the answers to the original questionnaires? I have seen no evidence that it was so.
————————
So I will simply state the gross subjects which Doll covered:
a) Lung cancer in smokers as compared with non-smokers and a comparison of death rates dependent upon the amount of tobacco smoked.
b) A comparison of smokers and non-smokers as regards other diseases and cancers, some of which seemed to indicate correlation.
c) Death rates depending upon age, amount smoked, type of tobacco smoked, changes in smoking habits, etc.
d) The significance of where people lived.
e) The significance of inhaling.
f) Women.
I propose to waste no more time on the 10 year study. It is (imperfectly) computer produced gobble-de-gook, based upon very small numbers in every category and every age group.
I shall move on to the 20 year report and see if that makes any more sense.
UPDATE 2.15 am 17th Sept 2012.
I have just had a quick look at the 1971 report. Guess what? The first tables give a full breakdown of the numbers as at 1951, which is just what I was complaining about above. Perhaps I was not the only one to find the absence of those figures disturbing.
I am much too tired to continue further with that report tonight. I hope that the 1971 report contains less computer gobble-de-gook.
Post of the 19th Sept 2012: The Doctors Study
I had to abandon the 10 year report because it did not give a breakdown of number of smokers and non-smokers in the different age groups under consideration. But I suspect that statisticians complained to Doll about the missing info because, in the 1971 report (20 years later), the original info about how many doctors there were in each age group (eg. 20 -24, 25 – 29, etc) appeared.
Given that Doll had all the information that he needed to say what the incidence of death was in the different age groups, you would think that he would have stated these figures. But he did not. He did not say that, in the age group 35 – 39, for example, X number of doctors died.
It really is very difficult to find simple, basic figures. I want to know how many doctors in the age range 35 -39 died, and from what ‘disease’ they died, after 20 years had elapsed (when they were 55 – 59). That simple info is not given. Only ‘death rates per hundred thousand (age adjusted)’ is given.
By rooting around in the tables, I have made discoveries, but these discoveries have thrown up anomalies. For example, one table seems to give a figure of deaths in a specific age group as, say, 100 – but another table seems to say that the number was 130. I’ve spent several hours this evening puzzling out the discrepancies. I MUST BE WRONG! Doll could not get such simple things wrong.
I just want to be able to say that, in the age range 35 to 39, out of X number of doctors, Y number died leaving Z number still alive. But Doll’s numbers do not tell me that and do not permit me to calculate it. Well….. maybe they do, but the figures do not add up.
I’ll try again tomorrow…………
Post of 20th Sept: A Preliminary Comment on the 1971 Report of the Doctors Study
Maybe these comments will give readers an idea of why I am struggling with the Doll and Hill Doctors Study. The Judge’s decision in the McTear Case was some 600 pages long, but I found it reasonably easy to condense in a fair manner. This report by Doll is very difficult, not least because the figures given make it difficult to track the calculations. There is a specific difficulty which is:
How can you ‘age standardise’ effects which occurred maybe 30 years after the cause of the events, if, indeed, the supposed cause was the actual cause? Here is a quote from the 1971 report:
“These ratios suggest that between half and a third of all cigarette smokers will die because of their smoking, if the excess death rates are caused by smoking” [From the abstract of the 1971 report]
That statement is a typical propagandist, cynical, ASH-type statement. “…if the excess death rates are caused by smoking” Isn’t the whole point of the study to prove on the balance of probabilities that smoking is the cause of the excess deaths?
Therefore, I want to find out, if I can, what the reason for doubt is. Why does Doll not simply say that his figures prove (on the balance of probabilities) that smoking causes the excess deaths?
So far, after a struggle, I have evinced only the following definite information in macro terms:
Breakdown by age of the (male) respondents to the first questionnaire (Oct 1951):
Ages………..Total Number…………percent NON-smokers.
20 – 24……………886………………………..43%.
25 – 29…………..4375………………………. 30.
30 – 34………….4855………………………..21.
35 – 39…………..5086………………………..17.
40 – 44………….3802………………………..15.
45 – 49………….3538………………………..13.
50 – 54………….3577…………………………11.
55 – 59………….2177…………………………10.
60 – 64…………1893………………………….9.
65 – 69…………1477…………………………..8.
70 – 74…………1211…………………………..9.
75 – 79………….905………………………….11.
80 – 84…………451………………………….13.
85 +……………..177………………………….16.
Total………..34440.
There are certain things which spring to mind on seeing that list:
a) The total number of doctors in the UK at that time was some 60,000, only 40,000 of which responded (of which 34,440 were male – the subjects of Doll’s study). Why did some 20,000 not reply? Doll did not follow them up. But we can see that there is a drastic drop in replies from the age 55 +. Would it be reasonable to say that the non-respondents were getting on a bit and nearing retirement (or actually retired) and simply treated Doll’s questionnaire as just another nuisance? How many other ‘damn fool questionnaires’ were they receiving? It also strikes me as odd that Doll did not identify the non-repondents since he had access to the Medical Register. Surely that register would have the non-responding doctors’ ages? (Frankly, I suspect that he did, but did not wish to publish the results)
b) Taking into consideration the effects of a), do we not notice the rapid reduction in responses after the age of 54? What was happening to doctors over the age of 54? Where did they go to? And why did Doll bother to record events for doctors who had retired since they were not actually doctors any more?
c) Another point also comes to mind. Doctors must surely have been aware of the initial symptoms of lung cancer (much more so than the general population). How many were caught in the early stages and ‘cured’? Doll was not a doctor – he was an ‘ignorant’ epidemiologist (he described Fisher, the Father of Statistics, as ‘an ignorant geneticist’). He would have known nothing about the detail.
————————
I have the initial age groupings. I now want to know what happened to these people. Starting with the oldest (the 80 + group in 1951), what happened to them? Doll had access to their death certificates. What did they die from? I happen to know, from National Statistics, that pneumonia polishes off vast numbers of the over 80s. It hardly touches younger age groups. At this point, it hardly matters whether they smoked or not. What did they die from? Answers come there none.
————————
The whole situation is very troubling. If National Government policies are to be decided by ‘scientific’ knowledge, then the knowledge must be better than guesstimates. The judgement in the McTear case was about 600 pages which I reduced in summary to about 60 pages. Doll’s article in the BMJ was 12 pages. Should national policy be based upon such minuscule information?
———————-
I am trying to figure out if it is possible to backtrack from the ‘per 100,000’ figures to the actual deaths in any specific age group. It ought not to be necessary. What complicates matters is the practice of ‘standardising to the UK population’. How could we know the actual figures if the numbers have been ‘adjusted’? The least that Doll could have done is state the actual numbers of deaths in each age range.
———————
I fear that I am wasting my time. I fear that Doll’s reports were deliberately engineered to support the Eugenicist agenda. He was, after all, a Rockefeller student. That is not to say that he lied. I am sure that everything that he published was true, but it was ‘cherry-picked’ to suit the agenda.
——————–
It is hard to believe just how powerful Doll became. Read the McTear Case and see that he was chairman of the IARC committee, which produced, in a roundabout sort of way, the Framework Convention on Tobacco Control. Very few people were involved in the PRINCIPLES included in that document (which were, for all intents and purposed, simply aimed at destroying Big Tobacco). Doll was chairman of that committee and involved in all the others. I can’t help but feel that there was a personal vendetta between Doll and Big Tobacco. Perhaps Big Tobacco rubbished his studies. Whatever, to be fair, he has conquered all, at this time. But he could not have done so without the support of the billions of dollars from ‘the eugenistic foundations which paid for the NGOs which were set up BEFORE the UN passed the Millennium Goals agenda.
—————-
Enough. Maybe, ‘for the record’, I’ll try again to summarise Doll’s reports without comment. I’ll try, but it is very difficult because of the fact that the reports are agenda driven. I’ll try to tell it as Doll tells it and, if I can, point to the cherry-picking as I go along. But it is difficult – it may take some time. But what’s the hurry?
I do not trust ANYTHING that Tobacco Control says.
Post of 21st Sept 2012: The Doctors Study (Finding Figures)
The Doctors Study (Finding Figures)
Gosh! This is hard work!
One would not believe how hard it is to extract, from the Doctors Study, simple numbers of how many smoked and how many did not, how many died in each age group and what they died from. It is like wading through a swamp. The problem is that Doll did not provide the basic numbers. He transcribed them all into ‘deaths per hundred thousand (age standardised to the whole population of the UK)’. The problem is: “How can anyone check the calculations without knowing the actual detailed figures of deaths in each age group?” There is a reference to the original info being held at some university/hospital records department, but I’m damned if I am going to try to find that info – far too much work for very little return.
But I have made some progress. I have the breakdown of age groups of doctors involved in the study (five-year groups from 20 years old), and the percentages of smokers and non-smokers. I have had to estimate some of the figures because Doll uses 10 year groupings some of the time. But I am getting there.
It is all very complicated because I have to work backwards from Doll’s ‘per hundred thousand (age standardised) per annum’ to actual deaths in each age group. Why did he not start his 5 year, 10 year, 20 year reports with the simple facts about deaths in each age group? In other words, he complicates things which should be simple and simplifies things which should be complicated.
He goes to great efforts to identify precisely that, when a person died from lung cancer, the diagnosis was correct. But he is quite happy to lump together people who smoked one cigarette per day with people who smoked fourteen cigarettes per day. Fourteen cigarettes per day is fourteen times more dangerous than one cigarette per day! In other words, he is very precise about what interests him, but very imprecise about boring things (like precise quantities of tobacco smoked) which do not interest him.
That must be a major fault in the study.
I think that it is just about possible to translate the ‘per hundred thousand per annum’ figures into actual numbers of deaths, but do not count on it. A first stab on one particular number produced a 100% error. But that was just a preliminary stab at a calculation. My method of calculation may have been crap.
I shall persevere until I find it impossible to get at the facts. I find it really odd that it takes a little old man like me to try to make sense out of this blather when famous statisticians and budding famous statisticians are silent. It really is frightening.
23rd Sept: The Doctors Study (1971 Report) – Ischaemic Heart Disease
Ischaemic Heart Disease describes the condition where the blood supply to the heart (from itself in a roundabout sort of way) is insufficient to supply the heart muscles with ‘energy’. When the condition becomes bad, the heart stops and the person dies. Almost 10% of deaths last year were caused by this condition. You can see why this condition was one of Doll’s targets for ‘smoking related diseases’. I have been struggling to find actual death figures for lung cancer which relate both to the age groups (in five-year bands: 20 – 24, 25 – 29, etc) and the amount of tobacco smoked. But I found some figures for Ischaemic Heart Disease which can just about be applied to both age groups and smoking history. The trouble is that he starts from the age of 45 (there were very few deaths below that age for this condition), also, he groups the deaths in 10 year increments and not 5 year increments.
I have had no alternative but to take liberties with precision by splitting the 10 year number into two parts and assigning one part to one group and the other part to the other group. For example, if Doll said that the number of deaths in the group 40 – 49 was 400, I have had to simply break the 400 into 2 parts and put 40 – 44 as 200 and 45 – 49 as 200. I could have used the national death figures as a pattern to split the numbers more accurately, but it would have been difficult and would have achieved very little, since there is not a lot of difference between numbers in the high risk groups (from 45 to about 65). I mean, in the example I have given above, the numbers would have been something like 180 and 220. Also, I have treated all the non-smoking and smoking quantities numbers in the same way so that the ratios have the same relationship, give or take a bit. There again, as I have mentioned before, Doll lumps together smokers of 1 cig per day with smokers of 14 cigs per day, so there is a big built-in inaccuracy already. Anyway, here we go……
———————–
[In what follows, I shall give only examples of the calculations. I shall put the full lists at the end]
First, I created a spreadsheet with the numbers in each age group:
20 – 24………….25 – 29……………..30 – 34……………….etc.
886………………4375…………………4855…………………etc.
Doll gave the percentage of NON-smokers in each group:
43%……………….31%……………………23%…………….. etc.
So I could split the totals by non- S (non-smokers) and S (smokers):
Non S.
381……………etc.
S
505…………etc.
——
886…………
—————————————
At this point I have the split of smokers and non-smokers by age group. Bu there is a problem. Doll gave figures for Isch Heart Disease only for non-S and Cigs smokers, so I had to remove people who smoked only pipes or a mixture of pipe and cigarette from the smoker numbers. Fortunately, in the 5 year report (1954), Doll gave a breakdown of smokers by method of smoking (pipe etc), but they were in 10 year age groups! So I had to split those groups in much the same way as I split deaths numbers. After a lot of messing about, I managed to extract the cig smokers and arrived at this point:
………………………..20 – 24…………..25 – 29……………..30 – 34………..etc.
Non S…………………382………………..1312………………..865……………etc.
Cig only S……………405……………….2063……………….2904………….etc.
But there was another problem. I needed to split the cig smokers into groups by amount smoked. In the 5 year report, Doll gave these figures, but in grams of tobacco! All I could do was split the cig smokers pro rata, which I did. So I now have:
…………………………..20 – 24…………..25 – 29………….etc.
Non S…………………..381……………….1312……………..etc
1 -14 (cigs)……………152…………………773……………..etc.
15 – 24…………………152…………………773……………..etc.
25 +…………………….101………………..517………………etc.
The reason that many of the cigs numbers are the same is that I could only use simple ratios like 2:3:2 or 4:3:2. I doesn’t matter since, at the end, variations will only mean differences of tenths of a percentage point or so (as you will see), and the divisions into ‘1 – 14′, ’15 – 24′, ’25 +’ are themselves artificial.
—————————-
So I now have the number of doctors by 5 year age bands and the number of non-smokers and number of cig smokers, with the cig smokers split into numbers of cigs smoked daily. Now I need to allocate Isch Heart disease deaths to each group. Doll provided the actual death figures, in 10 year bands (starting at 45 years old) in Table V of the 1971 report. I had to split these figures into 5 year bands also, but the numbers were quite small and no harm done by doing so. For example, in the 45 -54 group, Non-S deaths were 32, 1 – 14 p/d were 38, 15 – 24 were 90 and 25 + were 69. I slightly adjusted those figures to favour the older groups, since there is an age progression of deaths. Here is an example of deaths:
…………………………..40 – 44…………..45 – 49………………50 – 54…………..etc.
Non S………………………..2………………..15……………………17…………………etc.
1 – 14………………………..6…………………17……………………17.
15 – 24……………………..12………………..45…………………..45.
25 +………………………..18………………..34…………………..35.
That is most of the statistical numbers and allocations done. Now we move the important part – the ratios.
——————————
The above death numbers are not much use in themselves. We need populations so that we can say ‘2 deaths out of x number of non-smokers’, etc. Below is an example:
40 – 44…………45 – 49…………etc.
2/570……………15/460 (15 deaths out of 460 doctors in the age group 45 – 49)……….
6/870…………..17/594………..
12/870…………45/892……….
18/583…………34/594……….
And now we can convert these ratios into percentages:
……………..40 – 44…………45 – 49…………etc.
NonS0…….0.35%…………..3.27%…………
1 – 14……….0.69%………….2.86%………….
15 – 24……..1.38%……………5.04%…………
25 +…………3.09%…………..5.72%…………
How now to interpret these percentages?
If you were cherry picking data, you would select the 40 – 44 column and say,”Wow! Look at the difference between the non-smokers and heavy smokers! Flipping heck! The heavy smokers are TEN TIMES more likely to peg out from a heart attack!” You wouldn’t pick the 45 – 49 column which shows less than double the risk, would you? Let’s look at the 65 – 69 column:
65 -69.
NonS0…….33.9%.
1 – 14……….20.25%.
15 – 24……..25%.
25 +…………40%.
Well, well! Light and moderate smokers are dying at a lower rate than non-smokers and heavy smokers are only dying at a slightly higher rate! What’s the problem?
Below are all the figures for the ratios and percentages. A couple of notes:
a) When you look at the ratios, pay attention to the numbers involve. Ask yourself – How many doctors died in that age group and how many stayed alive?
b) The same for the percentages.
c) When the reach the end of the ratios, EVERYBODY IS DEAD. (At the end of the ratios, more people seem to have died than were alive! That’s because the numbers were becoming so small that minor calculation errors have a big effect)
d) Note especially that, as the groups of doctors get older and older, more and more of them die from heart disease, regardless of smoking habits.
Age groups of male doctors:
|
20 – 24
|
25 – 29
|
30 – 34
|
35 – 39
|
40 – 44
|
45 – 49
|
50 – 54
|
55 – 59
|
60 – 64
|
65 – 69
|
70 – 74
|
75 – 79
|
80 – 84
|
85 +
|
Non-smokers.
|
381
|
1312
|
1019
|
865
|
570
|
460
|
393
|
218
|
170
|
118
|
109
|
100
|
63
|
28
|
Cigs 1 – 14 per day.
|
152
|
773
|
988
|
1089
|
870
|
594
|
594
|
531
|
333
|
321
|
280
|
167
|
170
|
43
|
cigs 15 – 24 p/d.
|
152
|
773
|
988
|
1089
|
870
|
892
|
892
|
354
|
333
|
200
|
200
|
70
|
72
|
8
|
cigs 25 + p/d.
|
101
|
517
|
660
|
726
|
583
|
594
|
595
|
180
|
334
|
100
|
100
|
23
|
25
|
2
|
ANYONE LOOKING AT THIS JUST AFTER 10.30PM 22ND SEP, THERE’S MORE TO COME. I HAD TO CHECK HOW THE COPYING LOOKS!
Deaths from Ischaemic heart disease:
|
20 – 24
|
25 – 29
|
30 – 34
|
35 – 39
|
40 – 44
|
45 – 49
|
50 – 54
|
55 – 59
|
60 – 64
|
65 – 69
|
70 – 74
|
75 – 79
|
80 – 84
|
85 +
|
Non – smokers.
|
1
|
2
|
15
|
17
|
37
|
42
|
40
|
43
|
32
|
30
|
30
|
1 – 14
|
2
|
4
|
6
|
17
|
17
|
45
|
46
|
65
|
69
|
38
|
38
|
37
|
15 – 24
|
3
|
7
|
12
|
45
|
45
|
61
|
62
|
50
|
51
|
17
|
17
|
16
|
25 +
|
2
|
6
|
18
|
34
|
35
|
62
|
63
|
40
|
41
|
9
|
9
|
9
|
Ratios:
|
20 – 24
|
25 – 29
|
30 – 34
|
35 – 39
|
40 – 44
|
45 – 49
|
50 – 54
|
55 – 59
|
60 – 64
|
65 – 69
|
70 – 74
|
75 – 79
|
80 – 84
|
85 +
|
Non-smokers.
|
0/1019
|
1/865
|
2/570
|
15/460
|
17/393
|
37/218
|
42/170
|
40/118
|
43/109
|
32/100
|
30/63
|
28\28
|
1 – 14.
|
2/988
|
4/1089
|
6/870
|
17/594
|
17/594
|
45/531
|
46/333
|
65/321
|
69/280
|
38/167
|
38/170
|
37\43
|
15 – 24.
|
3/988
|
7/1089
|
12/870
|
45/892
|
45/892
|
61/354
|
62/333
|
50/200
|
51/200
|
17/70
|
17/72
|
8\8
|
25 +.
|
2/660
|
6/726
|
18/583
|
34/594
|
35/595
|
62/180
|
63/334
|
40/100
|
41/100
|
9\23
|
9\25
|
2\2
|
Percentages of deaths to numbers of doctors in each age group:
|
20 – 24
|
25 – 29
|
30 – 34
|
35 – 39
|
40 – 44
|
45 – 49
|
50 – 54
|
55 – 59
|
60 – 64
|
65 – 69
|
70 – 74
|
75 – 79
|
80 – 84
|
85 +
|
non-smokers:
|
?
|
?
|
0%
|
0.11%
|
0.35%
|
3.27%
|
4.33%
|
16.97%
|
24.70%
|
33.90%
|
39.45%
|
32%
|
47.62%
|
100%
|
1 14.
|
?
|
?
|
0.20%
|
0.37%
|
0.69%
|
2.86%
|
2.86%
|
8.47%
|
13.81%
|
20.25%
|
24.64%
|
23%
|
22.35%
|
86.05%
|
15 – 24.
|
?
|
?
|
0.30%
|
0.64%
|
1.38%
|
5.04%
|
5.04%
|
17.23%
|
18.62%
|
25%
|
25.50%
|
24.29%
|
23.61%
|
100%
|
25 +.
|
?
|
?
|
0.30%
|
0.83%
|
3.09%
|
5.72%
|
5.88%
|
34.44%
|
18.86%
|
40%
|
41%
|
39.13%
|
36%
|
100%
|
——————————–
Comments.
I have not done any ‘age standardisation’ to the above figures. I don’t want or need some sort of summary which produces overall percentages for all doctors. I don’t much like this ‘age standardisation’ trick. I get an impression that age standardisation hides the fact that, although ‘heavy’ smokers die proportionally in greater numbers in each age group, the number of deaths is small until you get to the age of 55 or so. But the number of deaths in non-smokers seems to me to be surprisingly high – thirty seven out of two hundred and eighteen in the age group 55 -59. Hang on……I’ve just had a think about that. I was forgetting that the number of deaths is the number occurring over a period of twenty years! The doctors in the age group 55 – 59 would have become progressively older as time passed. Thus, a 55 year old in 1951 would have been 75 years old in 1971. This also true of the other age groups. For example, there were 594 ‘heavy’-smoking doctors in the age group 44 – 49 at the beginning of the period in 1951. There were 34 deaths from Isch Heart Disease in that group, but those deaths accumulated over a period of 20 years! By 1971, the oldest persons in that group would have been 69, and 560 of those doctors (heavy smokers, remember) in that group were still alive. My, how easy is it to make things look worse that they are! UPDATE 10 am 23rd: ERROR! It does not mean that 560 were still alive. It means only that they had not died from Isch Heart Disease – they may have died from something else. I shall come back to that later.
Another reason that I don’t like age standardisation is connected to the claims by Doll and Tobacco Control in general is that smoking only kills after decades have elapsed. I can’t really think it through, but it just feels wrong.
———————————
There is only one more thing that I wish to mention about the 1971 report which is ‘smoking related diseases’. He identified those diseases which he thought were smoking related. Here is a list:
Closely related.
Cancers – lung, oesophagus, other respiratory.
Tuberculosis, bronchitis, emphysema,.
Pulmonary heart disease, aortic aneurism.
Hernia (through coughing).
Ischaemic heart disease.
Other associated.
Cancer – rectum, pancreas, bladder.
Pneumonia.
Myocardial degeneration, hypertension, arteriosclerosis.
Cerebral thrombosis.
Liver sclerosis (alcoholism), Peptic ulcer, suicide, poisoning.
Possibly protective.
Parkinsons.
Unknown.
Everything else.
And so we see that, even as long ago as 1971, the Health Zealots and Tobacco Control Advocates had started to blame almost all the important causes of death on smoking.
————————-
He also examines inhaling and the good effects of stopping smoking. He also, via his age standardisation process, starts to apply his study results to the whole population of the UK.
———————–
I am done with the 1971 report, but it may be necessary to come back to it. The reason is that it may be the last report which was reasonably objective. By the time of the 1991 (the forty-year report) Tobacco Control was in full swing, and so anything in Dolls reports becomes even more suspect.
I hope that this summary has been reasonably easy to read. I know that some people’s minds go blank when they see statistical tables! All I can suggest is that you simply allow your eyes to dwell on them. In particular, as regards the ratios and percentages quoted, remember that the deaths have accumulated over twenty years, but the number of doctors relates to the original number in 1951, and that very few of them died over the twenty year period in middle age.
24th Sept:The 20 year Report on the Doctors Study – some further
I have just read through my last post about the Doctors Study. I would not blame anyone for finding it gobbledygook. The problem is that the way that Doll presents the facts is also gobbledygook! I am sure that, to him, it all made perfect sense, but non-statisticians and non-epidemiologists would struggle. But I had to describe things in the way that I did so that I could add my cogitations to the summary in the sidebar. In retrospect, it is clear that Doll ought to have begun his 20 year report with a detailed statement of:
a) The number of doctors in each 5 year age group at the beginning of the study (and clearly stated to be so).
b) The number of doctors who died, in each age group, and what was the cause of death.
c) The proportions of those deaths by cause and a ‘running total’ of the diminution of the numbers in each age group.
d) The proportions of the deaths as related to smoking habits, in greater detail than ‘1 to 14’, for example (not ‘age standardised’).
e) Age standardisation at the end purely for international comparison purposes.
It is clear to me that the above requirements are ‘sine qua non’s.
But we must make the best of what we have available.
I have been able, very roughly, to sort out the lung cancer deaths in the 20 year report. I cannot go into detail tonight because there is still work to be done. Suffice to say that, as stated by Doll, the number of lung cancer deaths during the whole of the period of twenty years from 1951 to 1971, was 441 (male doctors) out of some 24,000 male doctors. That is FACT. To what extent they smoked is open to debate, but that FACT is not. Nor is the number of doctors in each age group open to debate.
Those numbers are also FACTS.
More and more ‘sleight of hand’ in these Doll reports is coming to light.
Post of 25th Sept: The Doctors Study (1971 Report): Lung Cancer
I did not think that enough information was to be found in the 1971 report to enable me to find the facts that I needed to get to grips with the detail of how many doctors in each age group died from lung cancer during the 20 years from the beginning of the study in 1951 until the 1971 report. The figures which Doll gave were per hundred thousand, age adjusted’, which weren’t much use for tracking back to the actual number of deaths in each age group. VERY, VERY, VERY IRRITATING! Why did he not start with a chart of the basic deaths information? He had it all.
All I had to work with was:
a) I had the number of doctors in each 5 year age band starting from the group 20 to 24 and ending at 85 +.
b) I had the percentage of those doctors who were non-smokers in 1951 for each 5 year band.
c) I had the total number of deaths from lung cancer, being 441 (out of a total of doctors of some 34,400 – that’s 1.28%. We should also note that some 28,500 were smokers at the beginning of the study in 1951). To put the following observations in context, we can see that throughout the whole of the twenty years elapsed time from the beginning of the study until 1971, only 441 doctors died from lung cancer. Some 28, 000 of the smokers did not. Also, during that period, some 10,000 of the doctors died, of which only 441 were lung cancer deaths. So, in terms of major causes of deaths, LC did not really figure, either for smokers or non-smokers.
What I did NOT have was:
a) A breakdown of the LC (lung cancer) deaths by age bands.
b) A breakdown of non-smoker LC deaths by age bands.
c) A breakdown of the LC deaths among smokers by amount smoked (1 -14 per day, 15 – 24 per day, 25 + per day).
I seemed to be totally stymied. But I had a couple of brainwaves. As regards a), and in view of the fact that I had little other alternative, I thought about the LC deaths by 5 year age groups in the ONS (Office of National Statistics) statistics for the whole population of England and Wales. It occurred to me that the breakdown of LC deaths in the national statistics would probably suffice to give me ratios for deaths of doctors in 1971 in age groups. Rough thought the application of these ratios might be, it is unlikely that they would be very much different. And so I checked the 2010 statistics and found the following ratios:
2010 LC deaths by 5 year age groups (male).
Ages…20-24…25-29…30-34…35-39…40-44…45-49…50-54…
%………0.02…..0.02…..0.04…..0.21……1.25…..3.00….5.96….
Ages .55-59…60-64…65-69…70-74…75-79…80-84…85+.
%…….11.32…..14.79…..17.42….18.64…15.48….9.66…..8.
What do we immediately see? Well, well! We see a gradual increase in LC deaths among the population at large by age groups!
So, while bearing in mind variations will have occurred between 1971 and 2010, it is not unreasonable to apply these percentages to the 1971 number of LC deaths. And so we get the following calculated number of LC deaths by age group in 1971:
[I have rounded the figures to the nearest full death]
Ages…20-24…25-29…30-34…35-39…40-44…45-49…50-54…55-59…
Deaths..0………0…………0………1…………5…………13……..26……..50…..
Ages….60-64…65-69…70-74…75-79…80-84…85+.
Deaths…65……….73………83…….68……..43……..4.
Remember that those deaths are accumulated figures over a period of twenty years! For example, even in the highest death rate group, the 70-74 group, the actual number of deaths, on average, was only 4 per year. Oh…and those deaths include both smokers and non-smokers (even though almost no non-smokers seemed to have died from LC during the 20 years….
——————-
So I have solved the problem of a) above [the LC deaths by age band]. What can I do about problems b) [non-smoker deaths by age bands] and c) [smoker deaths by amount smoked]?
Another little brainwave reminded me that there was a breakdown of both of those figures in the 10 year report (Table 4). I looked it up and found that, indeed, that table had numbers of non-smoker deaths and smoker deaths broken down into age groups and smoking amounts. These were the figures:
Age………..NS………….1-14…………..15-24………….25+.
35-44………1…………….1…………………0………………1.
45-54………0……………3…………………9………………8.
55-64………0…………..3…………………20……………26.
65-74………0…………..9…………………17…………….14.
75+…………2…………..6…………………..8……………..8.
So I thought, “Why not apply the ratios above to my calculated age band numbers?” (for lack of any better way). So I did. But there is a problem – all those zeros in the NS column. However, by reference to the ‘per 100,000 per an’ figures in the 1971 tables, I found that approx x deaths from LC had occurred in non-smokers by 1971. The numbers are very samll and not worth calculating, so I left them as zero. It doesn’t matter.
———————–
Having found reasonable ways to get around the lack of basic info in the Doll 20 year report, I was able to produce these very approximate figures:
The ratio of LC deaths to the number of doctors in each age band along with amount of tobacco smoked:
Ages..20-24..25-29..30-34…35-39…40-44…45-49…50-54…55-59…
NS…….?…………..?……0\1019..0\865..1\570..0\460..0\393..0\218..
60-64…65-69…70-74…75-79…80-84…85+.
1\170…..0\118….0\109..6\100..4\63…..0\28.
Ages..20-24..25-29..30-34…35-39…40-44…45-49…50-54…55-59…
1-14……………………….0\988…0\1089..0\870..2\594..4\594…3\531..
60-64…65-69…70-74…75-79…80-84…85+.
4\333…17\321…19\280..16\167..11\170..1\43.
Ages..20-24..25-29..30-34…35-39…40-44…45-49…50-54…55-59…
15-24…………………….0\988..0\1089..2\870..6\892..12\892..20\354..
60-64…65-69…70-74…75-79…80-84…85+.
27\333..33\200..35\300..23\70..14\72..1\8.
Ages..20-24..25-29..30-34…35-39…40-44…45-49…50-54…55-59…
25+………………..0\660…1\726…..2\583..5\594…10\595..27\180..
60-64…65-69…70-74…75-79…80-84…85+.
34\334..27\100…29\100..23\23..14\25..1\2.
The percentages of deaths to numbers in each age band.
Ages..20-24..25-29..30-34…35-39…40-44…45-49…50-54…55-59…
NS.%….……………………0……..0…………0.18……..0……….0……..0…..
60-64…65-69…70-74…75-79…80-84…85+.
0…………0………….0………6……..6.35……0.
Ages..20-24..25-29..30-34…35-39…40-44…45-49…50-54…55-59…
1\14%……………………….0………0……..0………0.34…….0.67…..0.56……
60-64…65-69…70-74…75-79…80-84…85+.
1.2……5.3…….6.79….9.58……6.47…..2.33.
Ages..20-24..25-29..30-34…35-39…40-44…45-49…50-54…55-59…
15\24%……………………0………0…….0.23…….0.67……1.35……5.65…
60-64…65-69…70-74…75-79…80-84…85+.
8.11……16.50…17.50.…32.86…19…12.50.
Ages..20-24..25-29..30-34…35-39…40-44…45-49…50-54…55-59…
25+%……………………….0…….0.14……0.34……0.84……1.68……15…
60-64…65-69…70-74…75-79…80-84…85+.
10.18…..27…….29……100…….56……50.
—————–
I have been determined to complete these calculations, despite the fact that they are very, very inaccurate; and despite the fact that hardly anyone will read them. My sole objective has been to show how very small numbers can be manipulated and extrapolated and applied to big populations. For example, the percentages above show that, even at the ‘busiest’ death times, there were never two deaths per annum from LC in each 5 year age group. In other words, in the worst LC deaths 5 year age group, over the period of twenty years, only one-and-a-bit persons died from lung cancer in any given year. In other words, throughout the UK, less than two doctors in any five year age group died from lung cancer – and there are only fourteen age groups.
It is very late and I need some sleep. I admit that this post is messy. What I have been trying to reveal is that the risk of lung cancer is small, regardless of your smoking habits. Sure, non-smokers tend to survive through middle age better, but they all die from something or other before they attain the age of one hundred years.
—————-
We come back to the original complaint that I had with the Doll study and the statistics in the 1971 report. that complaint was that he did not provide a basic Table showing the actual deaths from all causes (grouped as appropriately) so that people could form judgements. His ‘age standardised’ rates only went to ‘hide the decline’. And what is ‘the decline’ in this scenario? It is the effects of the aging of the human body! It is as simple as that! It is not about specific ‘diseases’. The specific cause of death for any otherwise healthy individual (other than accident) MUST be some fault in that person’s physical make-up. Thus, it is quite possible that a minority of people have a ‘genetic’ fault which smoking aggravates. There again, there might be lots of factors which might ‘aggravate’ the genetic fault.
But I am getting carried away and falling for the Eugenisist trick in that I am accepting that all human beings have the same basic template. But do they? Obviously, most of us will be, basically, ‘healthy’ when we are born and will continue to be ‘healthy’ for many years. But an awful lot of people are not so privilaged. Many will have faults in their hearts, or lungs, or livers, or kidneys, or brains, or skin, or blood, or red blood cells, or white blood cells, or avioli or folicles or ….. whatever.
When the Government tries to make ‘one size fit all’, it errs. There is no ‘one size fits all’. My wife suffers from Multiple Scelosis. She has done so for some thirty years. Fortunately, it is not such that it developed rapidly. She is pretty stable and has been for many years. We go often to Mallorca on holiday. We cope. But she is one example of the extremes of ‘one size fits all’ as Governments see things, and act and waste enormous amounts of our tax moneys.
I know that Government has been trying to change the ‘one size fits all’, especially since my daughter became a teacher of infants. (She chucked-up a good job in insurance to do so). The Gov introduced the idea that ‘every child matters’. Who could possibly disagree? But, worthy as that slogan is, how does it differ from expectations fifty years ago? Why does Gov assume that ‘every child’ DID NOT MATTER fifty years ago?
I have digressed.
It is my opinion that DOLL ET AL (as they were at the time) were Holy Zealots of the worst kind. Even in 1971, the Zealots were setting up the demonisation of tobacco smoke and the dehumanisation of people who enjoy tobacco. What aggravated the situation was that DOLL ET AL deliberately applied the damaging effects of tobacco smoking on the vulnerable to the whole population. Thus, the enjoyment of tobacco is a perfectly simple personal decision which requires no imput from government at all, other than the requirement that ‘smoking establishments’ display a sign. Non-smoking places should also display a sign.
A strange post tonight. I do not know what to make of it….
Post of 26th Sept: A Final Thought on the 1971 Doctors Study Report
Doll said that the Hospital Study (see sidebar) was started in 1947. He published his paper on that study in 1950 in the British Medical Journal. I read somewhere that he was so disappointed by the lack of interest in that study that he decided to begin the Doctors Study. But that cannot be true. The Doctors Study involved writing to 60,000 doctors in October of 1950. That exercise would have taken some considerable organising. It must have been under consideration for much longer than a couple of months. I strongly suspect that both of these studies were conceived more or less together at a much earlier date. Doll was a ‘Rockefeller Student’, which means that he was substantially subsidised financially by the Rockefeller Foundation. That organisation was very much a eugenistic organisation and would certainly have funded only those who were ‘believers’. I strongly suspect that the anti-tobacco campaign was initiated by the remnants of the Prohibition Movement, shortly after the collapse of prohibition in 1933. The whole thing stinks of a very small number of people with massive financial muscle. I have no doubt that lessons were learnt from the collapse of prohibition, and that the idea of ‘salami slicing’ was dreamt up around that time. Doll’s report on the Hospital Study was published in September 1950 and the questionnaires for the Doctors Study were sent out commencing in October of 1950. It seems clear to me that the publication of the Hospital Study (for doctors to read) was preparatory for the commencement of the distribution of questionnaires.
The ‘small number of people’ includes Sir George Godber. Godber became Deputy Chief Medical Officer in 1950, and then Chief Medical Officer in 1960 (until 1973). (He lived to be 100, dying only in February 2009) He, of course, was the author of the ‘Godber blueprint’, which set out the steps to be taken to salami slice the introduction of Tobacco Control, along with the need to emphasis second-hand smoke harm and ‘for the children’. See how the dates tie in? There was an American counterpart to Godber, but I forget his name.
For the above reasons, I strongly distrust anything that these people produced. As we know, Fisher (the ‘father of statistics’) was very critical of the Doll studies and other studies about smoking (it is surprising how many such studies appeared around the same period of time – 1950). He said that they all covered the same ground and used the same methods, so that it was not surprising that they came to the same conclusions.
One of the oddities of the Doctors Study (as reported in 1971 – after 20 years) was the sheer lack of lung cancers in non-smokers. According to Doll, there were only 10 LC deaths among some 5,500 non-smoking doctors over the period of 20 years between 1951 and 1971, given that diesel fumes are now recognised as carcinogenic (among other possible influences).
I do not trust a word these people say. While I don’t suppose that there was actual deliberate falsification, the possibility of mis-diagnosis of the cause of death among non-smokers must have been a significant factor. Remember that Doll stated that he checked that the lung cancer deaths were indeed primary lung cancers, but he did not check that other causes of death were accurate. He accepted what the death certificate said in those cases. One wonders how many non-smoking doctors had lung cancer (which would have killed them) but died from heart attacks and strokes and things before the lung cancer became evident? Also, I have been rather concerned about the arbitrary division of cigarette smokers into ‘1-14′ per day, ’15-24′ p/d and ’25 +’ p/d’. I know that Doll had figures for 1-4,5-9,10-14, etc because he used them in one Table (to do with inhaling). Also, on one of his graphs, he used ’45 cigs per day’ as a marker.
I am also very suspicious of Doll’s use of ‘standardisation’ of death rates. It smells a bit of ‘hide the decline’. By ‘hide the decline’, I mean hide the effects of declining health as people get older. Almost nobody dies from heart disease or lung disease before the age of 45. Thereafter, there is a steady increase, year after year by age, in the number of deaths from these causes. Age standardisation hides these effects by smoothing.
One last thing. Where did the idea of a several decades delay in the effect of smoking come from? How was that decided? It seems to me that this delay is a rationalisation of a circular argument. Nobody dies from heart disease or lung cancer before the age of 45, even if they have been smoking since the age of 19, THEREFORE the ill effects of smoking must be delayed for several decades. But there is a serious problem with that argument. As I understand it, lung cancer (particularly squamous cell carcinoma of the bronchi) appears quite suddenly and is rapidly fatal (within a few months of diagnosis). Remember the McTear case? Mr McTear was able to go on holiday to Malta, despite having squamous cell carcinoma, and then died a couple of weeks after returning from Malta. It makes no sense to say that he started to get squamous cell carcinoma when he first started smoking, and that it got steadily worse over the next several decades, when we know that there is only a short interval of time between the first appearance of the cancer and death.
I do not believe a word they say. I have mentioned before the fact that my wife has multiple sclerosis. She was first diagnosed around 1981 – 30 years ago. (Please do not think of this as unbearable! One learns to cope with not being able to walk etc!) We have been going on holiday to Majorca two, three and sometimes four times a year. We have never had any problems because of her condition until the last couple of years. Only in the last couple of years has a tendency for pressure sores and infections started to appear. But she is now 71 and has MS. Are the pressure sores and the infections caused by the MS? Of course not! They are not ’caused’ at all. They come from bacteria which have invaded this damaged body which is aging. Pneumonia (which Doll tried to attribute to smoking) polishes off vast numbers of over-eighties. But pneumonia is not itself a ‘disease’. It is more like drowning, in that the lungs produce fluid when trying to remove particles which are breathed into the lungs, such as fungus spores, pollen, dust, etc. The fluid blocks the avioli and stops the transfer of oxygen.
I do not believe a word they say. Even if it were true that smoking exacerbates existing conditions, the result of stopping people smoking will simply prolong life a bit and result in more cases of Alzheimer’s disease etc. Thus, in due course, there will be no gain for the NHS. In fact, Public Health will become an even greater liability.
We should also note that, in the 1971 report, Doll definitely claims that smoking CAUSES lung cancer, although he edges around that claim. He also claims that smoking CAUSES heart attacks and all the most common causes of death. The propaganda blitz was in preparation…..
I do not believe a word they say.
I shall now move on to the 40 year report. I am glad that I have seriously looked at the previous reports because I am now aware, in advance, of some of the trickery. It might take a few days……..
The Doctors Study 40 Year Report (Part 1 of 3) : 5th Nov 2012
Many readers will be aware that I have been following through the Doctors Study. We started with the 27 month report about the progress of the study, and have looked at the following reports at 5 years, 10 years and 20 years. We are now considering the 40 year report.
In my ‘summaries’ of the reports prior to the 40 year one, I have kept them ‘general’ in character. That is, I have paid little attention to Doll et al’s structure of the reports. In my description of the 40 year report, I have (so far) stuck to the construction of the report in its parts and divisions. The reason that I have been quite particular about the way I have dealt with this report is that it has seemed to me that, as regards the very important McTear Versus Imperial Tobacco Company lawsuit (which concluded in 2005), it would have been the 1991 results of the study which would have been important. Doll et al produced a final report in 2001, but that report was, essentially, just a summary, since almost all the doctors were dead by 2001. Results between 1991 and 2001 added nothing new to speak of, since, after 50 years, even the very youngest doctors at the beginning (apart from a few who were under 25 at the beginning) would have been well over 75 years old by 2001.
The 40 year report is very dense – statistic after statistic after statistic. It is hard to envisage any ordinary doctor, who might have read the British Medical Journal, getting beyond the first page, unless he was especially interested. Of course, by 1991, there may well have been lots of Holy Zealots who were interested. Even so, unless those persons were really zealous, I quite imagine them abandoning the thing after a few paragraphs and accepting Doll’s word for it. I see no way that the likes of Arnott, Duggan, Pell, etc could ever be bothered to read it, never mind understand it.
——————-
I decided to split this examination into three parts. This is the first part. It deals with the TEXT of the report. Part two will deal with the TABLES and GRAPHS. Part three will be COMMENTS. What I am hoping for in the comments is an explanation as to why the Expert Witnesses in the McTear Case did not bring their studies into evidence. Even the general observation that Judges do not much care for epidemiological evidence in specific, individual cases does not excuse the exclusion by the Expert Witnesses. The point is that there was no actual physical evidence that tobacco smoke caused McTear’s cancer. Note the difference between smoking and smoke.
——————
Let us make a start….
In 1951, a study was started called ‘The Doctors Study. It was to be a ‘prospective’ study – meaning that people would be asked about their smoking habits and followed for a period of time to see how their health was affected by smoking. The forty-year report on the results of this study was written by Doll, Peto and three others. I have tried to summarise it briefly, but with reasonable accuracy, regarding what is significant from my point of view. My point of view derives from the question: “Why did Tobacco Control not introduce the Doctors Study (and the US studies if necessary) as evidence that smoking causes lung cancer in the McTear versus Imperial Tobacco court case?” [A summary of that court case can be accessed via the sidebar] TC must have known that some attempt at proof would be required – even though the Judge’s decision needed to be based only ‘on the balance of probabilities’ (since this was a civil case) rather than the more stringent requirement of ‘beyond reasonable doubt’ (which applies in criminal cases). In fact, all the more reason to produce the epidemiological evidence. TC failed to produce any evidence other than ‘because we say so’.
We will quickly pass through THE TEXT of the report in the first instance, and then look at the TABLES and GRAPHS in Part 2.
NB. Remarks in square brackets are mine.
[Please excuse the mixture of tenses]
———————–
MORTALITY IN RELATION TO SMOKING: 40 YEARS OBSERVATIONS ON MALE BRITISH DOCTORS
Abstract
The objective of the Doctors Study was to assess the hazards associated with long-term use of tobacco.
Over 34,000 British male doctors took part. 10,000 died during the first 20 years of the study and another 10,000 died in the second 20 years, leaving about 14,000 survivors in 1991.
Doll claims that positive associations with tobacco were observed regarding these illnesses: Various Cancers, COPD (chronic obstructive pulmonary disease), Vascular Conditions, and Peptic Ulcer. Possible associations also with Cirrhosis, Poisoning and Suicide. Some protection seemed possible re Parkinson’s Disease. The excess mortality was chiefly from disease caused by smoking.
The effect of stopping smoking has great benefits. The sooner the better.
Doll asserts that smoking kills half of smokers.
Introduction
Early results from the study showed a strong connection between smoking and Lung Cancer in particular and also with other causes of death. 48 causes of death are considered. Other studies which had been summarised by the College of Physicians, the US Surgeon General and the International Agency for Research on Cancer [IARC – which was the creation of Doll et al!] confirmed the findings. What is important about the Doctors Study is the length of time of the study. Lung Cancer is higher in Britain than other places and smoking is very prevalent in Britain. The hazards of long-term smoking in Britain are likely to be very informative.
Methods
34,500 male doctors had replied to a questionnaire about their smoking habits out of about 55,000 male doctors. Some 6,000 women also took part but were not included in this study. The first questionnaire was quite simple, asking only if the respondent was a) a non-smoker (meaning never got into the habit of smoking – (Any doctor who never smoked the equivalent of one cigarette per day for as long as one year was counted as a non-smoker) b) Smokers were asked to estimate how much they smoked at the time of the questionnaire . c) They were asked if they’re smoking was cigarettes only, pipe only, or a mixture of the two. Cigar smoking was not included at that time, but was introduced later. During the course of the 40 years, 5 additional questionnaires were sent out in 1957, 1966, 1972, 1978 and 1990 and from these, changes in smoking habits were obtained. Over 90% of questionnaires were returned.
Follow up
After the first 20 years (1971), the ‘vital status’ (alive or dead) of 99.7% of the doctors who responded in 1951 was known. 10, 074 had died. A few people had been excluded as a result of being struck off, for example. By 1991, a further 10,449 had died. Again, a few were excluded for various reasons. Deaths were monitored in various, overlapping ways.
Causes of Death
Death certificates were the main source. In the period to 1971, any deaths from lung cancer were confirmed by additional enquiries, but after that, death certs were accepted since very few LC causes were changed in the first 20 years.
Statistical Methods
Doctors were allocated to groups:
Non-smokers, cigs only, pipe or cigar only, cigs and others.
Some ‘standardisation’ was used [but, as far as I can see, it was confined to ‘equalising’ within age groups. For example, in the 10 year age group 40 – 49, there may have been a lot more doctors approaching 49 than just over 40. In that case, small, statistical adjustments were made to allow for the discrepancies. I do not think that these adjustments were important].
The life-threatening effects of smoking generally take many years to appear.
Results
Changes in smoking habits.
Partly because smokers had been dying off prematurely, and partly because many quit, few were still smoking in 1991. Only 18% still smoked of which only 7% were cigs. [There is quite a lot about percentages still smoking here which I have not produced]. Doll said that many may have quit because of his questionnaires and the information which he provided to doctors of results to date! He remarked that, among those who were 45 in 1951, smoking was down to 2%, [but we must bear in mind that the 45 year olds in 1951 would have been 85 in 1991].
Mortality by smoking habit and cause
The study shows that smokers of pipes and cigars were hardly affected, although they fared worse than non-smokers. For simplicity, mortality rates are given in these categories: smokers of cigarettes only, other current smokers and other former smokers. What the Tables show was then described. Rates for 48 causes of death were shown, some being grouped together due to shortages of numbers. Rates for seventeen forms of cancer were given. The forms of cancer which the International Agency for Research on Cancer (IARC) said were caused by smoking and were few in number, were grouped together. There next follows some detail about which forms of cancer were most particularly associated with smoking and which were least. A separate Table described other diseases. COPD (Chronic Obstructive Pulmonary disease) was almost as bad a lung cancer; Pneumonia had a weak connection. Asthma deaths in smokers and former smokers were double those of non-smokers. (8.3 per 100,000 compared with 3.7) Pulmonary heart disease was most closely related, but ischemic heart disease was only related a little. But there were lots and lots of deaths, so the less serious association was more important in absolute terms. In relation to Parkinson’s disease, there seemed to be a negative association, which could be because sufferers could not hold a cigarette. Alzheimer’s also showed negativity but there were very few deaths.
Mortality by smoking habit, by age.
No deaths under 25 years of age and only 67 up to 34. The only useful date came from those over 35. Over the 40 years, smoking harm was twice as bad for continuing smokers as for non-smokers through middle age and early old age. After that, things evened out somewhat.
Actuarial survival by smoking habit.
This section consists mostly of graphs which compare survival rates among non-smokers, light, moderate and heavy smokers.
Change in mortality and survival over time
Many changes occurred over the 40 year period which could have modified the effect of smoking. Doll decided to split the 40 years into two parts – the 20 year periods 1951 – 71 and 1971 – 91. In the second half, mortality among smokers was worse than in the first half. There is a graph showing this. Also, improvements in the treatment of diseases seem not to have helped smokers. Further graphs show the effect of quitting – the earlier the better, but quitting even as a 60 year-old still has an effect.
In the next paragraphs, Doll described how greater longevity helped non-smokers much more than smokers in connection with a number of diseases.
DISCUSSION
Causation, confounding and chance
Doll said that this study, along with others reviewed by US Surgeon General and IARC, showed that smoking is ‘causal in character’ regarding mortality among smokers. Smoking may act in concert with other factors like alcohol, diet, blood pressure and others, including possibly infections. The association with alcohol was confirmed when a questionnaire was sent out in 1978 and completed by 12,300 doctors. It showed that non-smokers drink less units of alcohol than smokers do – from 2% in non-smokers up to 20% in heavy smokers. [He makes a strange conjecture here. He says that, if small quantities of alcohol are beneficial to the heart, then it is possible that heavy smokers who drink moderately will receive some protection against the effects of smoking on the heart. He suggests that there could have been an even greater adverse heart effect from smoking than was observed. He gives ratios of what might have been the situation had the drinkers not been drinkers as well as smokers. With reference to Ischemic Heart Disease deaths, the ratio of smokers to non-smokers is 1.58:1. He conjectures that, had the drinking smokers not also been drinkers, then the ratio would have been 1.76:1]
In the next paragraph, he discusses rectal and colorectal cancers There was evidence of association with colorectal adenomas, but there was an induction period of several decades. There seemed to be a strong association with peptic ulcer and association with suicide, gunshot wounds, etc, but they were excluded.
Temporal trend in excess mortality in smokers
A new feature discovered by the Doctors Study is that excess mortality in smokers was greater in the second half of the study (1971 – 91) than in the first half (1951 – 71). In the period 51 – 71, excess mortality in smokers was about twice that of non-smokers during middle age. In the period 71 – 91, it was about treble. [He makes a rather strange statement here. He says “If, as is likely, most of the difference in mortality between smokers and non-smokers is actually caused by smoking, then a threefold excess would imply that about two-thirds of the deaths in middle age among smokers were caused by tobacco” Condensed, does that sentence not mean, “If smoking causes death, then smoking causes death”?]
He continues with a general discussion of various factors:
a) Median survival of non-smokers in the first period was about 5 years. In the second period it was 8 years. The average seems to be 6 and a half, but, in fact, it was 7 and a half because of the greater excess of smoker deaths in the second period.
b) Advances in the treatment of many diseases helped non-smokers. Environmental standards (such as coal-smoke pollution), diet and other things, had improved. Lung cancer deaths of non-smokers stayed much the same as time passed.
c) Aids had little effect on the generations studied.
The difference in the two halves reflects the fact that only non-smokers benefited from improved treatments. [There follows some conjecture about the effects of low tar cigarettes, but Doll reckons that low tar would not have made much difference] He talks about the maturing of ‘the epidemic’ of deaths from smoking in the second half so that all the excess deaths in that period are the result of the maturing of the epidemic. This conjecture was supported by massive cohort studies by the American Cancer Society.
In the developing world, we can expect a similar epidemic of deaths.
He concludes with the statement:
“The substantial differences between the effects of tobacco during the first and second halves of the present study underline the need for a worldwide network of prospective studies that can monitor the increases, and eventual decreases, in this great epidemic throughout the world”
———————
Shorn of most of the rhetoric, it is not particularly impressive, is it? So in the first half, smokers tended to die 5 years before non-smokers and in the second half, 8 years before non-smokers. So what, if most of the deaths were in very old age anyway?
But let us wait until we have looked at the tables and the graphs. Hopefully, I should have Part 2 organised within a week.
The Doctors Study: 1991 Report: Part 2: The Tables and Graphs: of 2012
ALTHOUGH THIS ARTICLE ‘STANDS ALONE’, READERS MAY LIKE TO READ PART 1 FIRST. IT IS DATED 5TH NOV “THE FORTY YEAR DOCTORS STUDY REPORT (PART 1 OF 3)”.
It may be as well to think about how Doll might have organised his records. He sent out 60,000 questionnaires. How was he to organise the returned, completed q’s?
We must bear in mind that, as far as I can see, Doll did not have access to a computer at that time.
There are various ways. My own inclination would be to transfer the info onto some sort of ‘card index’ (bearing in mind that the durability of the record over 20 years plus is important). The cards would have to be of a suitable size (say, 10” x 5”) to allow for unexpected info in the future. The cards would bear the name, address, phone number and date of birth of the doctor. As each questionnaire came in, I would have completed a card for that person. Doll said that he only asked how much doctors were smoking at the time of the questionnaire, but he must also have asked how long they had been smoking, otherwise he would not have been able to say that the average age of starting was about nineteen and half (among other things). He also asked people who used to smoke, when they stopped and how much they were smoking at the time.
So, on the cards, I would have also recorded:
1. If a person was a non-smoker and always had been (apart from the possibility of a little experimentation).
2. If an ever-smoker (a person who stopped), amount smoked and date quit.
3. If a continuing smoker,
(i) Date of starting smoking.
(ii) Type of smoking (cigarette, pipe, mixture).
(iii) Amount smoked.
Thus, I would have a base, durable record for each doctor in the study.
———————-
But that is only the start. How best to organise the index cards? In the first instance, I would merely have kept the cards in strict alphabetical order – that is all. I would then have waited for death reports to come in. (Arrangements had been made for a form to be completed by the registrars and sent to Doll when a doctor died). When a report came in, I would have extracted the appropriate card and written the date of death and cause on the card. Bearing in mind that deaths occurred only at a rate of some 45 per month (probably more in winter than in summer), I would probably had stored the death reports in a box file as they came in and dealt with them monthly – staff and time and such.
When a death report had been dealt with, alphabetical order of the cards was no longer required. As cards were noted with date of death and cause, they could simply go into a box ‘awaiting further action’. Perhaps when about a hundred cards had accumulated in the ‘dead’ box, I would have actioned them further. How?
————————-
I now have a little problem. On the one hand, I want to analyse them by ‘smoking status’; on the other hand, I want to analyse them by ‘cause of death’. I also want to analyse them by ‘age’. How to do so? This is where it becomes very tricky.
Let us imagine how it might be done.
In the first place, he has a chart/table. Something like this:
…………..CAUSE OF DEATH.
Age……….LC…………Heart………..Stroke……….etc
35 – 44……..2……………..4……………..1……………..
45 – 54……..4.……………10…………….3……………..
55 – 64……..8…………….25…………….6……………..
You get the idea.
Next, he takes the group “Age and LC” (the bold figures in the chart above) and allocates them by smoking:
………………………………..Non-S……..Cigs only………..Pipe only………..Mixture.
35 – 44 with LC…………….0………………..2………………….0…………………..0.
45 – 54 with LC…………….0………………..3………………….0…………………..1.
55 – 64 with LC……………1………………..6…………………..1…………………..0.
And so on.
Now he can take the next step – analysing the Cig Smokers:
………………………………………………..1 – 14…………15 – 24…………25 +.
Cig smoker and 35-44 with LC……….0………………..1………………1.
Cig smoker and 45-54 with LC……….0………………..1………………2.
Cig smoker and 55-64 with LC……….1…………………1……………..4.
And so on.
And what would be the easiest way to create these tables/charts? Well, WALL CHARTS of course!
That is how I imagine Doll et al doing the work. Not difficult, once set up. As far as I know, Doll did not have access to a computer in the early stages, but he did by the ten year report as at 1961.
Why have I gone to the trouble of working the above out and describing it? The answer is that I believe that it will help us to understand the TABLES AND GRAPHS which will follow.
————————-
THE TABLES
[NB. I have rounded numbers for ease of demonstration]
We must remember that Doll had 40 years to decide how best to use his statistics to his advantage. As a Eugenicist/Prohibitionist, he would naturally favour that information which attacked tobacco.
Table 1 seems to have little purpose other than as a form of propaganda. It takes THE SURVIVORS of the original cohort of 34,000 doctors after forty years, and describes their smoking status. The survivors numbered 10,800. Of these, 2,600 were originally (in 1951) non-smokers. A few (some 300) had started smoking, so that, of the surviving non-smokers, 2,300 were still non-smokers. The Table contrasts the non-smoker situation with smokers. I shall take only cigarette smokers as an example. Of the original number of cig smokers, 4,400 had survived. Of these, 3,400 had stopped smoking and become non-smokers. The remaining 1,000 continued to smoke. Among those, some 500 were cig smokers and some 500 were pipe smokers. Only a few (47) were mixed smokers.
Of the FORMER smokers in 1951 (1,500 of whom were still alive in 1991), 1,400 were still former smokers. Only 100 had relapsed.
In Table 1, Doll also indicates the percentages of smokers and non-smokers among survivors. Non-smokers now numbered 25%; former smokers (in 1951, remember) were 13%; cig smokers were 41%; the rest (21%) were pipes and mixtures. It might be as well to remind ourselves of the original distribution in 1951:
Non-smokers……………………………………5439.
Ex smokers……………………………………….4812.
Pipe and/or cigar………………………………3613.
Cig and other……………………………………3802.
Cig only………………………………………….13542.
……………………………………………….
Total……………………………………………..31208.
Roughly , some 17% were non-smokers; 15% were ex-smokers; 68% were smokers. What Doll is saying is that, of the original cohort, the SURVIVORS were 22% were non-smokers; 60% were ex-smokers; and the smokers were only 62%. Thus, smokers had been dying off more rapidly than non-smokers.
Table 2 is not significant. It merely illustrates Doll’s statement that:
“Those who did smoke cigarettes reported smoking much the same amount at each age at the beginning of both the first and second halves of the study” Table 2 does not seem to have much significance.
Table 3 is much lengthier. It is headed up: “Mortality by smoking habits from neoplastic [cancer] diseases”. There are 15 specifically names sites with a breakdown of numbers by smoking habit; there is an additional one named ‘other sites’ and another named ‘site unknown’. This table shows “Annual mortality per 100,000 (male doctors in the study)” It would be natural for readers who have not studied these things to say, “Erm…there was not 100,000 people in the study altogether. How can Doll talk about X per 100,000 who were smokers of, say, pipes when there were only a few hundred pipe smokers?”
The answer, I believe, is that, for international comparison purposes, it was decided some time ago to state percentages in the form of ‘per 100,000’ rather than ‘per 1,000’ or ‘per 100’ (which, of course, is what ‘percent’ means). Let’s put it this way. 500 per 100,000 is only 5 per 1,000, which is only 0.5%. Not very impressive. Stated as 500 (per 100,000) makes the figures sound much more impressive, don’t ya think? Whether that was a deliberate intention or not, I do not know. Even more impressive is the translation of these numbers into proportions of the population as a whole. For example, with a population of 60,000,000 people in the UK, 500 per 100,000 becomes 30,000 people!
It is worth thinking for a moment about this. Let us suppose that, throughout the UK, x people go to the doctor with flu symptoms. Of these, a few (who are very, very old or very, very young) die. Some epidemiologist in the DoH picks up these figures. Disregarding the ages, he multiplies up the incidence to whole population proportions. He computes that x hundreds of thousands will contract flu, and x thousands will die. He reports to the Minster, who reports the PM who asks, “What can be done?” The zealots say, “Order a million doses of flu vaccine and spend hundreds of thousands of pounds on advertising. Force doctors to contact their ‘vulnerable’ customers and get them to exercise persuasion, and make it a condition of their contracts. The money spent on the vaccines does not matter because the SAVINGS in terms of NHS costs more than makes up for it 100 times.
But it does not work. People refuse the vaccinations. The epidemic does not occur. Egg hits faces, but is quickly wiped off. Millions of pounds poured down the drain for nothing, and nobody either knows or cares. Those millions of pounds could have been put to good use elsewhere, but nobody cares. That is GOVERNMENT and POLITICS. And that is why our political system stinks.
————————–
But back to business.
The first item in Table 3 is ‘deaths from cancer of upper respiratory track’, which includes several separate sites such as the larynx. The incidence of these causes of death was 1 per 100,000 per annum in non-smokers.
Making sense of these figures can be really tricky. Take the above as an example. The death rate for non-smokers from upper respiratory cancers was 1 per 100,000 non-smoking doctors per annum. The period involved is 40 years, which implies that there were 40 deaths over the whole period per 100,000. At the beginning of the study, there were 5,400 non-smokers. Multiplying the 40 per 100,000 by 5,400 equals 2.16. So the actual number of non-smokers who died from those causes was 2. (Don’t bother about the 0.16 – there is bound to be some rounding involved!) I did the same calculation for ex-smokers, pipe and mixed smokers and cigarette smokers. I calculated, using the ‘per 100,000’ figures that a total of about 230 doctors had died from those causes. But Doll says that there were only 98 in total. There must therefore be some factor which is not clear. Perhaps the ‘per 100,000’ relates only to those who had died and/or he was moving people who had stopped smoking around. Impossible to say.
The main point about Table 3 is, of course, that the figures show smokers dying at a far greater rate than non-smokers from most of these cancers. I shall pick out a few as examples. I am not going to quote all the variations of smoking. I have picked out only non-smokers, average smokers, heavy smokers and pipe smokers for simplicity.
HIGH VARIATION
DEATHS PER 100,000 PER ANNUM
………………Non-S…..all cig smokers….heavy cig smokers……pipes etc.
Upper resp……..1……………24…………………48……………….15.
Lung…………….14………….209……………….355………………112.
Oesophagus……4……………30………………..45………………..23.
Pancreas……….16…………..35………………..49………………..24.
Apart from lung cancer (with 893 deaths – out of around 20,000 in total), the numbers are quite small where smoking damage is said to be important. Pancreas is nearest with 205 deaths. Stomach cancer is quite high at 277, but the figures for non-smokers and smokers with stomach cancer are not that far apart (smokers are a somewhat short of double – 26 NS compared with 43 S).
Here are a few which do not present significant hazard via smoking. I have picked the ones which have significant numbers of deaths:
LOW VARIATION
DEATHS PER 100,000 PER ANNUM
………………Non-S…..all cig smokers….heavy cig smokers……pipes etc.
Colon………….36…………….46…………………..62……………………..47.
Prostate………68……………..67………………….84…………………….64.
Other sites…..39…………….40…………………..45…………………….31.
The numbers for those three were respectively 437, 568 and 344.
Perhaps we should put these numbers into some sort of perspective. For example, as regards lung cancer, the total number of deaths (893) has accumulated over forty years. Which means that the average toll per annum was 893 divided by 40, which is approximately 25 per annum – even among the heaviest of smokers. We should also bear in mind that a ‘heavy smoker’ is one who smokes 25 or more per day. There is no upper limit. So, for all we know, the LC deaths in heavy smokers could all be of people who smoked 50 a day and drank a bottle of whiskey a day and were obese.
Table 4 is the same as Table 3 except that it covers respiratory and vascular diseases.
Respiratory Diseases.
Only 5 are named. Of these, only COPD (chronic obstructive pulmonary disease) has significant numbers of deaths (542). Pneumonia is included with 864 deaths, but ONS stats show that pneumonia is rarely fatal except in extreme old age. I suppose that Doll had to include it to keep is figures straight. Here are the COPD figures:
DEATHS PER 100,000 PER ANNUM
………………Non-S…..all cig smokers….heavy cig smokers……pipes etc.
COPD…………10…………….127………………………225…………………51.
‘Other respiratory diseases’ (216 deaths) do not seem to be particularly significant:
ORD…………..19……………..30…………………………33…………………18.
Vascular Diseases.
This section concerns everything concerned with heart and circulation matters. It includes the biggest killer of all – Ischaemic Heart Disease (6438 deaths). Other big numbers are also outlined.
DEATHS PER 100,000 PER ANNUM
………………………..Non-S…..all cig smokers….heavy cig smokers…pipes etc.
Ischaemic (6438)….572……………892……………………1025……………653.
Myocard (841)………..61…………….125……………………..173……………..85.
Cereb throm (956)…..93……………122……………………..143…………….106.
Cereb haem 607)…….59……………..81……………………….92………………58.
Other cereb (1025)….94……………164……………………..188……………..103.
In the last block, we see that, statistically speaking, heart diseases do not have very significant differences. Certainly, the heaviest smokers were almost double the non-smokers, but, again, not anywhere near that of lung cancer. It certainly seems that other factors (like alcohol and obesity) could be big considerations here.
Table 5 concerns all other causes of death. The numbers are not large and contain causes which may have serious confounders. For example, Cirrhosis of the Liver is much more likely to be associated with alcohol. I propose to skip Table 5.
Table 6 is headed “Total Mortality by smoking habits and age at death”
Doll uses it to show that continuing cig smokers die twice as fast as life-long non-smokers in ‘middle age and early old age’.
The Table lists age groups in ten year periods starting at 35 (up to 85 +). An oddity here is that he reverts to ‘per 1000 per annum’ rather than the ‘per 100,000 which he has been using everywhere else.
UPDATE 14TH NOV. I have reason to believe that the reversion to ‘per 1,000’ is a misprint, and that it should read ‘per 100,000’.
[NS = Non-smoker.
ExS = Former smoker.
S = All Current smokers.
HS = Heavy smoker.]
DEATHS PER 1,000 PER YEAR (Should be per 100,000).
Ages……………NS………..Ex-S…………….S………….HS.
45-54……………4.0…………4.9……………..8.1………..10.8.
55-64……………9.5………..13.4…………….20.3……….26.0.
65-74…………..23.7……….31.6…………….47.0……….60.7.
There is also another peculiarity. Notice how the ratio of non-smoker to smoker is almost exactly double, regardless of the age group – 4.0:8.1, 9.5:20.3, 23.7:47.0. Ought not the heavy smokers be drawing away from the non-smokers as time passes? Why do they not? We saw this self-same trend in the graphs which I published a few days ago. (Of which more shortly).
He does not say whether or not he is moving people from one group to another as he goes along. For example, he could be moving people who stopped being heavy smokers into the ex-smoker group. He says somewhere (can’t remember where this instant) that smoking status was based upon the answers to the latest questionnaire, which suggests that people were sometimes moved from their original status in 1951. I can’t be sure.
There is a Table 7, but that comes after THE GRAPHS SECTION, so we shall leave that until the end.
—————————–
THE GRAPHS
[ALL GRAPHS CAN BE CLICKED TO ENLARGE]
Doll describes the graphs as ‘figures’. I prefer the word graphs because I am going to translate the details onto graph paper so that I can keep the detail as accurate as possible and use similar scales for both the X and the Y axes. I have already done the first two graphs which show that smokers die younger than non-smokers (Graph 1) and that heavy smokers die before moderate smokers who die before light smokers who die before non-smokers (Graph 2). But let us first look at the way in which Doll drew Graph 1. It seems to suggest a fairly drawn out series of events. There is nothing wrong with the way in which Doll drew the graph, but it is a bit misleading. If you look carefully, you can see that the Y axis (the vertical line on the left) is squashed up when compared with the X axis (the bottom, horizontal line). It isn’t easy to spot. It is because the ‘breaks’ are in 20% sections on the vertical axis whereas they are in 15 year sections on the horizontal axis.
Now look at that graph when the percentage axis and the ages axis are drawn to the same scale [NB. I HAVE CONVERTED THE ABOVE GRAPH INTO ‘DEAD’ RATHER THAN ‘STILL ALIVE’]:
Do you see the point? No deaths until age about 40 and then smoker and non-smoker deaths begin to diverge a little (about 2% more smoker deaths). Only around age 65, does the divergence become sizeable. But it is also easier to see that everything happens really in old age and that, smokers and non-smokers die at about the same rate (the red and green lines are just about parallel). In fact, towards the end, they start to converge when non-smokers hit the age 80 (the green line begins to approach more closely to the red line). If continuous smoking is so bad, when compared with not smoking at all, shouldn’t the red line and the green line continue to diverge?
Graph 2 is similar to graph 1 except that there are four lines – heavy smokers, moderate smokers, light smokers and non-smokers.
Graph 2:
Somewhat more messy than Graph 1, but again we see a small divergence around 50ish with heavy smokers starting to die earlier than non-smokers, but, again, we see the lines running parallel and converging towards the end. Again, I ask why is it that so many heavy smokers survive to be so old?
[By the way, I must admit that there are probably a couple of miscalculations on the above graph. That lurch to the left on the first line looks very suspicious. Not to worry – the overall effect is correct]
Graph 3.
As I said, I reversed graph 1 and 2 to show ‘dead’ rather than ‘alive’. Graph 3 shows how people who gave up smoking before 35 fared as compared with continuing smokers and non-smokers. It shows that people who gave up smoking by thirty five, suffered no ill effects as compared with non-smokers. I shall not ‘reverse’ it, so this graph shows ‘alive’ rather than ‘dead’. [I’ll try to be more accurate!]
So first, let’s look at Doll’s original:
[The faint gridlines are mine]
Again we note what seems to be a gradual process of people becoming less than alive.
Here is my version of tht graph with the scale of percentages and ages equally spaced:
You can easily see that nothing much happens until the age of 40. Only about then do smokers start to die in small numbers with non-smokers dying at a lower rate. This difference continues to increase until about the age of 60, when things ‘settle down’, and smokers and non-smokers start to die at the same rate, albeit smokers dying earlier than non-smokers. What is also odd, as I have said before, is that smokers stop dying off so readily at about the age 80. Their death rate slows down in comparison with non-smokers).
But has anyone noticed the really, really odd thing about this graph? Look at the yellow line (people who smoked but stopped at or before age 35). The yellow line shows that they actually outlive non-smokers! At first, I thought that it was simply because Doll could not overlay the non-smoker and ex-smoker lines, but look back at the Doll version. Note that the dotted line starts inside the non-smoker line, crosses it at about the age 65-70 and stays outside it for the rest of the time until about 95, when it comes back inside again. I do not recall Doll making any mention of this revelation in the text. I wonder why? People who smoke from the age of 19 but stop by 35 OUTLIVE non-smokers!
Graph 4
There are four separate graphs in this figure. They describe how much ‘extra’ old age people get if they stop smoking from the age of 45 upwards. I do not propose to redraw them as I did with Graphs 1-3. I am sure that readers will be able to imagine the graphs elongated in the vertical direction:
Do not confuse the first graph with graph 3. Remember that Graph 3 concerned people who stopped smoking before age 35.
I see no point in commenting on Graph 4.
Graph 5
This figure purports to show that the excess of smoker death rates in the second half of the study (1951-1971) were greater than in the first half (1971-1991). I reproduce this graph without comment:
All I’ll say is what the above means. The bottom line represents smokers who were, say, 45-54 in 1951. The top line represents people who became 45-54 in 1971. That is, smokers who were 45, say, in 1951 died off less rapidly in subsequent years than smokers who were 25 in 1951 and became 45 in 1971. In the text, Doll suggested various possible reasons for this, principally, as a result of non-smokers gaining advantages from better treatment of diseases.
Graph 6
Again, this concerns the different death rates in the two halves of the study:
We are not particularly concerned about it, except to note that careful examination shows that the widening of the gap is caused by non-smokers living longer.
———————————
We now go back to the tables. As I said earlier, Table 7 follows the graphs. The reason is that Table 7 gives the detail of Graph 6. You can see that, in total, smokers died from the listed disease at much the same rate in both halves. Non-smokers died from these diseases at a lower rate in the second half. Again, I’ll reproduce it for reference only:
There is a further Table which is given as an Appendix. It is a big one. I’ll reproduce it for readers who wish to study it:
[Click to enlarge!]
If you read this Table, remember that the numbers are PER 100,000.
————————————
That concludes Part 2 of my summary of the Doll 1991 Report on the results of the Doctors Study. The final part is going to be my comments. However, I would be grateful for any comments readers might wish to make.
WHY DID TOBACCO CONTROL NOT PRODUCE THIS STUDY IN THE McTEAR V IMPERIAL TOBACCO LAWSUIT?
And so I ask again – Why Did Tobacco Control not Produce Epidemiological Evidence in the McTear Case? 17th Nov 2012
Actually, the answer is quite simple. It lies in the second requirement stated by Judge Nimmo. TC would have to prove, on the balance of probabilities, that smoking caused Mr McTear’s lung cancer.
Let’s have a look at what we now know.
Mr McTear was born in 1944. He was diagnosed with LC in 1992 at the age of 48. He died in 1993 at the age of 49. We can look at Doll’s graphs and figures to see what they tell us about the incidence of LC in persons of or about that age.
Let’s look again at the graph showing deaths in doctors up to the age of 50:
What does the graph show us about the deaths of doctors up to the age of 50? Enlarge the graph by clicking it and look at the situation at the age of 50. At that age, 2% of non-smokers had died and 4% of smokers. There were about 10,000 smokers in the study who were under 50 at the start of the study. 4% of 10,000 is 400. But remember that this 400, was from all causes! How many of that 400 died from lung cancer?
That information is not easy to find out from Doll’s tables. I tried to get some sort of ratio from the ONS (Office of National Statistics). The total of deaths in people between the ages if 25 and 49 in 2010 in England and Wales was about 16,ooo. Of those, some 650 were lung cancer deaths. The ratio then is 650 to 16,000, which is about 4%. If that ratio is applied to the total of deaths from all causes among smokers up to the age of 50 in Doll’s Doctors Study, the number of deaths from lung cancer among doctors up to the age of fifty would have been about 16.
There is another way to discover that figure, which is to take whatever figures that we can find in Doll’s tables and work from them. I shall reproduce again the Table which formed the Appendix to the 40 year report:
[CLICK TO ENLARGE]
Don’t be alarmed! We are only looking at the first column (lung cancer) and the first three age groups in that column (35-39, 40-44 and 45-49). What do we see? Among current smokers, the table says that there were 42 deaths in those age groups, but that is 42 per 100,000. Per hundred thousand what? We are not certain, but if it refers to smokers in those age groups, then we know that there were about 7,300 cigarettes only smokers in those age groups at the beginning. A quick calculation indicates that there were only 3 actual deaths from LC in those age groups.
So we have two widely differing figures – from ONS stats ratio, we get 16. From Doll’s table, we get 3. But it does not matter! All we are looking for is a rough idea.
So we can now summarise:
1. There were some 10,000 smokers in the age range 25-49 at the start of the study.
2. By the age of 49, as little as 10 or so died from lung cancer.
How therefore could Tobacco Control bring forward to the court a suggestion that this meagre number could in any way support the contention that the death of Mr McTear at the age of 49 from lung cancer was caused by smoking? If anything, Doll’s figures show the opposite.
QED.
But we can go further. Any attempt to prove that smoking causes lung cancer would have met the same sort of problem. Tobacco Control have a major problem which is that they state that lung cancer from smoking takes decades to appear. But, surely, that is a circular argument? Upon what facts does that claim depend? It can only be that so few lung cancer deaths appear until old age. So that argument is based upon an assumption and not upon facts. The assumption is that smoking causes lung cancer! Put it this way – Only if it is true that smoking causes lung cancer can it be held that it takes decades to appear.
There is also another major problem for TC. That is – How does the risk crystallise into lung cancer in any particular person? Without an explanation for that, TC could never uphold causation.
The Doctors Study: The 50 Year Report: 7th Dec 2012
First, a reminder of the URL for the 50 Year Report:
http://www.bmj.com/highwire/filestream/400720/field_highwire_article_pdf/0/bmj.38142.554479.AE
Despite this final report being several pages long, there is not a lot to it. It is more in the nature of a summary. This is hardly surprising since almost all the original cohort of doctors were dead. When the study began in 1951, the youngest persons were about 25 years old (although a very few were younger). By 2001, the youngest people still alive would have been about 75 years old. Even so, some 6,000 of the original 34,000 were still with us. It is clear that not much could be added to the study’s findings by enumerating the mortality of the 5,000 or so very old men who died between 1991 and 2001. In fact, Doll identifies the objective of the Report as being:
“To compare the hazards of cigarette smoking in men who formed their habits at different periods, and the extent of the reduction in risk when cigarette smoking is stopped at different ages.”
The words: “who formed their habit at different periods” tell us that the Report is concerned with the differences between periods. In this case, the periods are those doctors born before 1900, and the decades thereafter up to 1920-30 (being the final decade of birth for the doctors being studied). A large part of the Report is concerned with two groups as such – those born before 1900 and those born after 1900. Also, it describes the effects of stopping smoking in some detail.
An awful lot of the text is exhortation, and so I propose to go straight to the tables. There are six, along with four figures comprising of ten graphs.
Table 1
We must bear in mind the totals involved. There were 34,400 men to start with. Over the years, some doctors were lost trace of for various reasons – some were disqualified, some said that they did not want to continue, others went abroad. They totalled some 3,000. Which left some 31,500 in 2001. Of those, 25,300 had died and 5,900 remained. Table 1 gives the causes of those 25,300 deaths in 11 broad catagories. I’ll list them rapidly:
Cause……………………………….Number of deaths.
Lung cancer…………………………..1052.
Cancers mouth, larynx, etc……….340.
Other cancers………………………..3893.
COPD…………………………………….640.
Other respiratory…………………..1701.
Ischaemic Heart…………………….7628.
Cerobrovascular…………………….3307.
Other vascualar…………………….3052.
Other medical………………………2565.
External causes………………………891.
Unknown causes……………………277.
——————————————.
Total………………………………..25,346.
I have listed them so that readers can judge for themselves the relative significance of each individual cause as a proportion of the total. Clearly, for example, lung cancer accounted for 1 in 25. You might think that very odd, in view of the fuss that has been made about the smoking and lung cancer and the fact that 83% of the d0ctors were smokers at the beginning of the study. Even odder is the prevalence (or lack of) of COPD (Chronic Obstructive Pulmonar Disease) with a ratio of 1 in 75. It is more than obvious that the vast majority died from heart failure, stroke, a variety of cancers and ‘other medical’.
But Doll’s main point is to draw attention to the two main points – a) smokers died earlier than non-smokers and, b) that as regards lung cancer and COPD in particular, the ratios are hugely bad for cigarette smokers. For example, for every 1 (per thousand per year) non-smoker lung cancer death there were 15 current cig smoker deaths and, especially, 25 heavy cig smoker deaths. As regards COPD, those ratios are 1 to 14 to 24. (Remember that these ratios ‘per thousand per year’ are not real numbers. They are a statistical way of comparing non-smoker and smoker deaths on equal terms).
The actual ratios of these deaths quoted by Doll are:
Per thousand per year.
Cause……………………..Non-S……………….Current S……………………Heavy S.
Lung Cancer……………..0.17……………………2.49…………………………..4.17.
COPD………………………0.11…………………….1.56…………………………..2.61.
On the other hand, the ratio of the biggest cause of death (Ischaemic Heart Disease) is:
Isch Heart Disease…….6.19…………………..10.01………………………..11.11.
Which, as can clearly be seen, translates into 6 to 10 to 11 per thousand per year. It is not worth translating the ratios into single units (well…go on then…..1 to 1.617 to 1.795).
One thing worth observing is that Doll wants ‘his cake and ha’penny’ regarding these stats. On the one hand, he wants to say that the huge differences in risk appertaining to those causes which applied to only a few doctors were very important, but, on the other hand, the risks where the difference in risk was very small where also very important. He justifies this on the grounds that heart attacks affected lots of people and so the smaller difference actually caused more premature deaths. Well, yes Mr Doll, but then you must surely admit that a cause of only a very few real deaths is of very little importance…….surely?
I do not propose to go further with these figures. You can look at the figures yourself. I would, however, like to point out that, the bigger the numbers of deaths involved, the lower the ratios! I’m not sure how significant that is, but it ought to be reasonably significant, but I can’t think why at the moment!
Table 2
Table 2 lists some confounders, especially alcohol intake. Taking the first item in the table, ‘vascular [heart failure] risk factors’, the figures go like this:
Alcohol units per week
Current smokers………………………………………………..19.0.
Ex smokers less than 10 years……………………………..18.1.
Ex smokers more than 10 years…………………………..14.8.
Non smokers………………………………………………………8.3.
Somehow or other, by some convoluted logic, Doll deigned to suggest that the fact that smokers drink more protects them to some extent from heart attacks and that they should suffer heart attacks at an earlier age than they do! Other risk factors mentioned were blood pressure and body mass index which were, on average, similar for all. (Is similar on the average of any use at all?)
The second item in Table 2 enumerated why some people quit smoking. Heart problems and breathing problems were the main things. Doctors had also been asked about whether they had any heart problems, ‘shortage of breath’ problems or ‘bringing up phlegm in winter’ problems. Frankly, I do not see that they prove anything at all, so I will move on.
[Incidentally, I have just been having a bit of fun with BMI. I am about 5’6″ and about 10 stone. My BMI is 22.9. A healthy BMI for me is between 18 and 25. I once met Alex Murphy. Hardly anyone, if anyone at all, who reads this, will know the name. Alex Murphy was scrum half (the guy who puts the rugby ball into the scrum) for St Helens (England!) Rugby League Team. HE WAS THE BEST EVER! Over, say, 15 metres, he accelerated so fast that he could break through the opposition line like a knife through butter. I once met him fleetingly in a pub. I am skinny. He was about my height, but built like a brick shithouse – neck like a bull, shoulders like an ox, legs like tree truncks. If he was less than 15 stone, I’ll eat my hat. According to the BMI index, his BMI was 35.9. “Your BMI is 35.9, which means you are in the very overweight or obese category. Losing weight would make a big difference to your health – you might want to speak to your GP about the best way to do this” So much for BMI]
Table 3
This table is very small. It compares the risk of death for people born before 1900 with that of people born after 1900 (up to 1930, of course – the latest DoB for doctors in the study). It shows simply that the first group (those born before 1900) had a lower level of risk of death because they smoked less than those born after 1900. Just as a taste, here are a couple of figures:
Relative Mortality Risks (non-smokers are base 1)
Century of birth………………non S………………Current S total……………..Heavy S.
Before 1900……………………….1………………………….1.46………………………..1.83.
After 1900…………………………1………………………….2.19………………………..2.61.
Thus, as compared with non-smokers, cigarette smokers born after 1900 had a greater risk of dying early than those born before 1900. Doll conjectured that this was because the habit of smoking cigarettes was only inculcated in WW1, as regards those born before 1900, and therefore did them less damage. On the other hand, those born after 1900 were smoking cigs as a matter of course between the wars and were further encouraged to smoke cigs during WW2. Those bloody generals! They were putting their troops in serious harms way by giving them free cigs, the bastards! It would have been much better for their health to have forced them to do without such dangerous comforts as cigarettes.
Enough of this idiocy. Let us move on.
Table 4
Prior to Table 4, there are some graphs illustrating the effects of the ratios indicate in Table 3. We can come back to them later.
I don’t think that there is any point in even contemplating this table. It compares trends in doctors who attained the ages of 70 to 89. Sure, it shows that non-smokers in that age-range lived longer. Wow! So what?
Table 5
Gives death rates (per thousand per year) in 10 year age groups. The purpose of this table is to show that giving up smoking prolonged ex-smokers lives. I’ll give one example:
Age group 65 – 74 AT DATE OF DEATH.
Death rate for non smokers………………………………18.6 (per 1000 per an).
D R for quitters at age 35-44…………………………… 22.7.
D R for quitters at age 45-54…………………………….31.7.
D R for quitters at age 55-64…………………………….36.4.
DR for continuing smokers…………………..50.7.
In this table, you can compare death rates either with non smokers or with continuing smokers. I think that the figures speak for themselves – that is, that non smokers and worried quitters (taking into consideration the fact that smokers tended to drink much more than non smokers) looked after their bodies better than smokers.
That concludes the Tables.
———————-
THE GRAPHS
I’m going to examine the graphs quite quickly because I am getting fed up, and, in any case, we saw much the same thing in the forty year report. So let’s get going.
Figure 1
Figure 1 contains two graphs. Readers may remember that I mentioned above that non-smoking doctors lived longer. These two graphs illustrate that. I have decided, on this occasion, to redraw the graphs so that the scales are the same. In the report, the percentage Y axis is squashed while the ages X axis is stretched. Here is what graph 1 looks like when the ‘percentage alive’ Y axis is drawn to the same scale as the ‘ages’ X axis:
[CLICK THE GRAPH TO ENLARGE]
The graph in the Report gives the impression that difference between smokers and non-smokers is much bigger because the graph is elongated to the right. Thus, what is in reality a very abrupt process (as illustrated above) is shown as though it is a slow and gradual process.
It purports to show how badly smokers who were aged 60 fared as compared with non-smokers aged 60. It is true that smokers died a few years earlier than non-smokers, but if you take the whole of life rather than concentrating on the period of life over sixty, you can see that non-smokers die at the same rate, even though they be dying at an older age. This clear from the fact that the lines are almost parallel.
That graph is for those born before 1900. The second graph is for those born between 1900 and 1930:
[CLICK THE GRAPH TO ENLARGE]
I haven’t redrawn the graph. I have simply added the non-smoker line (born after 1900) to the graph. The smoker line for those born after 1900 is not worth adding because it varies only slightly.
Doll said, words to the effect, that “Excess mortality among smokers, as compared with non-smokers, increased in the second half of the study”. Before I became familiar with ‘the tricks of the trade’, I would have understood that to mean that smokers were dying even earlier than before. But that is not true. What was happening is that non-smokers were dying later and later. and not smokers dying earlier and earlier. In fact, although non-smokers were gaining a great deal of extra years in old age, smokers were in fact gaining a little also. At the age of 70, smokers still alive went from 68% to 71%; at the age of 80, smokers still alive went from 31% to 32%. Not much, but not “Excess mortality” in GENERAL terms, that is, as though one was comparing smoker deaths in the SECOND period with smoker deaths in the FIRST period. Even smokers were living longer.
Did Doll deliberately intend to give false impression of what had been happening, both by elongating the graphs and describing smoker deaths as “increasing excess nortality”? We don’t know and never will.
Figure 2
Figure 2 is much the same as Figure 1, except that it breaks the figures down into individual decades (of date of birth) from 1900 to 1909, 1910 to 1919, and 1920 to 1930 (1930 being the latest possible date of birth for people in the study). Again it does not show smokers dying earlier and earlier, but it does show non-smokers living long and longer in very old age.
Figure 3
Has only one graph. Again, it is elongated the figures show that, at the age of 50, 97% of non-smoking doctors were still alive while 94% of smokikng doctors were still alive. At 60, 91% of NS were still alive compared with 81% of NS. At 70, the figures were 81% and 58%. At 80, they were 59% and 26%.
Jolly good show!
Figure 4
Contains four graphs, purporting to show that people who give up smoking live longer. The earlier, the better. In fact, the first graph seems to indicate that people who started smoking at, say, 19, actually live longer that non-smokers!
I have no doubt that it is all true. The only problem is that the quitters were self-selected. That is, was it the case that the quitters were ‘worried-well’ who really ought not to have been smoking at all? Were they not genetically suitable for smoking anyway?
There are many, many imponderables. How was it that so many doctors, over the perion of the Study, were able to stop smoking without the aid of patches, gum, tablets, etc? Surely, since they were smoking untipped, full-strength Capstain and Senior Service, they would have been hopelessly addicted?
————————
Following the completion of this Summary of the Doctors Study, a few comments are in order. But not yet. We have to think a bit For example, Bradford Hill was THE prime instigator of the Study. But, for some odd reason, he retired at the age of 64/65. It is unlikely that he was ill since he survived until the age of about 90. In fact, in 1965 he produced a paper in which he defined the requirements for epidemiological studies to have a semblance of ‘causation’. What motivated him to produce that paper? Why did he retire? Was he pushed?
Bolton Smokers Club 
Leave a Comment