An Analysis: the Doll and Hill study – ‘Smoking and Carcenoma of the Lung’ (1950) (The Hospital Study)

Today, 16th May 2012, I have copied the below analysis from the original post of 8th May 2012. I have put an entry into the sidebar referring to it to assist any person who wishes to know what the original study contained without too much detail.

The comments are not carried forward – they remain on the original post. At the end of this analysis, I have made some notes about the apparent faults in the Doll and Hill study, most of which originated from the comments to the original post.

UPDATE 10th May.

I forgot to put the URL in for the study itself:

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2038856/pdf/brmedj03566-0003.pdf

UPDATE 1.15AM 9TH May.

Those who read this post earlier will have observed that the tables were in a mess. After some hours of struggle, I have got them into a reasonable state. It is not possible to get them exactly right. This is because of the different formats used by WordPress and MS Word or Open Office. The corrections should make the post much more readable. I was tired when I published the post. I need to have another look at some aspects.

The Doll and Hill study has been taken apart again and again and I have seen a variety of remarks about it. But I have never before actually ‘studied the study’. Over the last few days, I have corrected that situation. I found the actual study on the net and gave it my (very imperfect, but ‘what the hell’) attention.

Below is my synopsis of the study – take it or leave it. I have tried my best to be non-judgemental and to describe the study as I have seen it. I have also tried to simplify it so that only the main points appear. For example, the female aspects of the study are essentially the same as the male aspects. Some of the Tables are complicated by separating males from females. That just makes the Tables more difficult to read and to understand. And so, I have simplified by eliminating the female figures. There were not many females involved anyway, as compared with the males.

So here we go:

In September 1950, Doll and Hill published an article in the Brit Med Journal entitled Smoking and Cancer of the Lung.

[Henceforth, I shall refer only to ‘Doll’ for brevity]

Doll starts by referring to the “phenomenal” increase in lung cancer deaths in the early parts of the 1900s. In the quarter century 1922 – 1947, such deaths rose from 612 to 9287. He discounts the increase in the population. He quotes Stocks (1947) which gave an increase per 100, 000 from 1 male person’s death to 10 male persons’ deaths, between the 1920s and the 1930s (approx), allowing for changes in age distributions.

He then discusses the possibility that better diagnosis is responsible for the apparent increase. He says that the increase was as evident in country districts as in London (where the better diagnostic facilities were located). He also draws attention to similar evidence from large cities elsewhere and locations in Wales. He says:

The large and continued increase in the recorded deaths

even within the last five years, both in the national figures

and in those from teaching hospitals, also makes it hard to

believe that improved diagnosis is entirely responsible. In

short, there is sufficient reason to reject that factor as the

whole explanation, although no one would deny that it

may well have been contributory. As a corollary, it is

right and proper to seek for other causes.

We note, therefore, that he does accept some possible uncertainty due to diagnosis improvements.

He then moves on to ‘possible causes of the increase’. He says that there have been two mooted – atmospheric pollution (car fumes, tarmac, etc) and smoking tobacco. He does not discuss the pollution issue, but quotes a number of small-scale studies which implicate tobacco, both here and abroad. He quotes a large-scale study from the USA (Wynder and Graham, 1950) which showed that only 1.3% of lung cancer patients were non-smokers, while 14.6% of general patients (not LC) were non-smokers. Also, 51.2% of LCs smoked more that 20 per day compared with only 19.1% of other patients. This is taken to be a strong indicator that smoking is responsible for lung cancers.

Doll then moves on to:

THE PRESENT INVESTIGATION.

I propose to quote the whole of the first paragraph:

The present investigation was planned in 1947, to be

carried out on a sufficiently large scale to determine

whether patients with carcinoma of the lung differed

materially from other persons in respect of their smoking

habits or in some other way which might be related to the

atmospheric pollution theory. Patients with carcinoma of

the stomach, colon, or rectum were also incorporated in

the inquiry, as one of the contrasting groups, and special

attention was therefore given at the same time to factors

which might bear upon the aetiology [cause] of these forms of

malignant disease. A separate report will be made upon

these inquiries. The present study is confined to the question

of smoking in relation to carcinoma of the lung.

He states later the number of patients with other cancers, but, after that, does not mention them again in this study.

Twenty London hospitals took part. They were asked to notify the investigators of incoming cancer patients. Not every incoming patient was accounted for, but Doll says that there was no bias in this since the hospital staff knew of the purpose only in broad outline. There was no reason for them to pick and choose.

Incoming patients, suspected of having lung cancer (hereafter described as LC), were interviewed by a researcher. At about the same time, an equal number of non-LC patients were interviewed. These non-LC patients were selected to be in the same age group and of the same sex as the LC patients so that the two groups of people (LC and non-LC) were directly comparable. I shall say here that, later in the report, Doll excludes bias on the part of the researchers on the grounds that they only had a sketchy idea of the detailed purposes of the report.

The research took about 18 months to complete. Altogether, there were 2370 admissions of cancer suspects, but not all were interviewed for various reasons. In the first place, Doll had decided to exclude patients of 75 or older on the grounds that their recollections [of their smoking history?] might be suspect. Other reasons were: too ill, discharged, dead, deaf, and others. In short, the number of suspected cancer patients (ALL cancers) who were interviewed finished up at 1732.

All these variations in types of cancer etc become rather confusing. It is better not to get involved in the   ‘whys and wherefores’ of questioning bowel cancer suspects, for example. Doll said that he had in mind a separate study about the general cancer situation. What we must be aware of is that, of the 1732 persons interviewed, only the lung cancer patients are important in this study.

I am now going to show Table 1. Bear in mind what I have said above about only the lung cancers being important, otherwise, you will be confused by the table:

Disease………………..Confirmed….Other Diagnosis…….Total.

Cancer, lung………………489……………220……………….709.

Cancer, Stomach………….178……………..28……………….206.

Cancer, Colon, rectum……412…………….19………………..431.

Cancer, other……………..xxx…………….xxx………………..81.

Non- cancer (control)…….xxx…………….xxx………………709.

Others…………………….xxx……………xxx……………….335.

Excluded………………….xxx……………xxx…………………4.

———————————————————————-.

Total…………………………………………………………..2475.

We have already observed that Doll had said that ‘other cancers’ would be the subject of a separate investigation. I have shown the full table simply to illustrate how Doll arrived at the total number of interviews – being 2475. ONLY THE ‘CANCER, LUNG’ PATIENTS AND THE ‘NON-CANCER CONTROL’ PATIENTS MATTER IN THIS REPORT. The bold figures simply draw the reader’s attention.That is, only the 709 LC patients and the 709 non-LC patients figure in the report from now on.

It would be reasonable to ask how Doll managed to get his non-LC patients to exactly equal his LC patients! That sort of question pops into ones mind, when confronted with tables of this nature. It is reasonable and natural that it should do so. The answer is that the total number of cancer suspects (2370) is not associated with the number of interviews (1732), or the number of cases (in the above table: 2475). Essentially, these various totals exist simply to approximate the idea that 2000 people represent the whole population of the country statistically (in my view). So how did Doll arrive at exactly equal numbers? Very simple. Only the number of LC patients is an absolute number. The interviewers, having interviewed a LC patient, then found a non-LC patient of the same sex and age and asked them to partake in the study. That is, LC patients were paired with non-LC patients of the same age and gender. Apparently, no one ever refused to partake.

I felt that it was important to explain Table 1 in some detail because it is the only Table which I propose to detail in full. For us amateurs, such tables can be very confusing, and rightly so. The more constituent parts there are in a table, the more difficult it is to comprehend. So let us exclude those parts which serve no purpose as far as we are concerned.

Have we achieved any clear understanding so far? I think so. I think that we have seen that Doll has acquired 709 completed questionnaires from LC patients and an equal number of completed questionnaires from non-LC patients, all equalised in respect of age and gender. In this regard, I must now publish Table 2, which shows the age distribution of people interviewed:

Age……………..LungCancer…………………Non-LC controls.

25 – 29…………………….2…………………………………..2.

30 – 34…………………….6…………………………………..6.

35 – 39……………………18…………………………………18.

40 – 44……………………36…………………………………36.

45 – 49……………………87…………………………………87.

50 – 54……………………130 ………………………………130.

55 – 59……………………145……………………………….145.

60 – 64…………………..109………………………………..109.

65 – 69………………….  88…………………………………89.

70 – 74……………………28…………………………………27.

Also in Table 2 are social class distributions and places of residence distributions. Both are roughly similar (I have omitted the detail).

One might reasonably ask how such close proximity was obtained. My own view is that far more interviews occurred in respect of non-LC patients as compared with LC patients, and that the age/sex comparisons were ‘discovered’ at a later date. I do not mean that there was anything underhand – it was simply the easiest way to do it.

At this point, we can say that Doll has achieved his ‘starting position’. He has 709 LC patients’ questionnaires and 709 non-LC patients’ questionnaires, all equally balanced by age and sex.

We can now move on to other aspects.

—————————————-.

ASSESSMENT OF SMOKING HABITS.

The assessment of the relation between tobacco-smoking

and disease is complicated by the fact that smoking habits

change.

Right – so what to do about the calculations? Oh dear! Now things become rather complicated:

In this investigation, therefore, the

patients were closely questioned and asked (a) if they had

smoked at any period of their lives; (b) the ages at which

they had started and stopped; (c) the amount they were in

the habit of smoking before the onset of the illness which

had brought them into hospital; (d) the main changes in

their smoking history and the maximum they had ever been

in the habit of smoking; (e) the varying proportions smoked

in pipes and cigarettes; and (f) whether or not they

inhaled.

Oh dear! It cannot be denied that the above sentence has introduced six varying confounders, all of which can affect results to a greater or lesser extent, both individually and collectively. Doll glosses over this problem by introducing a paragraph about what constitutes ‘a smoker’. He defines a smoker as a person who smokes one cig per day for a year:

A smoker was therefore defined in this inquiry

as a person who had smoked as much as one cigarette a

day for as long as one year, and any less consistent amount

was ignored.

But the definition of ‘a smoker’ has little to do with the six confounders which Doll himself has listed above. ‘One cigarette per day for a year’ is about as meaningful as our cat catching one mouse per week. What is the effect of ‘Marcus’ (the cat) catching one mouse per week, on average, on the health or population of mices within one square mile of his home territory?

————————————–.

In my opinion, it is not difficult to see how this study is beginning to fall apart. But that is only my own opinion. Too many variables are appearing. Doll tries to defuse this anomaly by reference to certain extra interviews. That is, re-interviewing some people to check that their answers to the questions are not significantly different on re-examination. Here is Table 3:

On second thoughts it is not worth the trouble. It just shows that Doll asked 60 interviewees the same questions again, and that a few of them changed their stories just a little bit, but not significantly. The Table merely shows that some people changed their minds about precisely how many cigs they smoked per day. Only 60 people were involved. This little study was sufficient to confirm that people gave genuine answers, according to Doll. I doubt it, but it is not important.

————————————–.

WE COME NOW TO THE ABSOLUTE NITTY-GRITTY.

Excuse my language, but it seems to me that the following Table is the be-all and end-all of this study:

Table 4:

Disease………………Smokers………………Non-Smokers.

Lung Cancer…………..647………………………….2.

Non-LC……………….622…………………………27.

That is the evidence that smoking causes lung cancer. Only 2 of the 709 lung cancer  were non-smokers, while 27 of the 709 non-lung cancer patients were non-smokers. That, essentially, is it.

EDIT 10TH MAY 2012.

647 + 2 = 649 (622 + 27 = 649): and not 709! Answer? We have excluded the females!

But Doll goes on to break the figures down by sex, age and number of cigarettes smoked. For simplicity, I propose to ignore the female numbers because there were few of them in comparison with the males – 120 compared to 1298. We lose nothing by doing so.

We can now move on to Table 2 (females excluded). This Table merely demonstrates that there were equal numbers of LC patients and non-LC patients.

Ages…………………..LC……………………..Non-LC.

25 –  29…………………2……………………………2.

30 – 34…………………6…………………………….6.

35 – 39…………………18………………………….18.

40 – 44…………………36………………………….36.

45 – 49…………………87………………………….87.

50 – 54……………….130………………………….130.

55 – 59………………..145………………………..145.

60 – 64………………..109………………………..109.

65 – 69………………  88…………………………..89.

70 – 74………………..28…………………………..27.

Also in Table 2 are social class and place of residence. The numbers break down reasonably evenly between LC and non-LC people, so I have left the detail out.

It can be seen that there is a ‘bulge’ in the numbers in the middle age area. Doll explains that later.

Assessment of smoking habits.

Doll explains the difficulty caused by the fact that people change their smoking habits from time to time. This problem was overcome by close questioning. Here are the questions:

(a) if they had smoked at any period of their lives; (b) the ages at which

they had started and stopped; (c) the amount they were in

the habit of smoking before the onset of the illness which

had brought them into hospital; (d) the main changes in

their smoking history and the maximum they had ever been

in the habit of smoking; (e) the varying proportions smoked

in pipes and cigarettes; and (f) whether or not they inhaled.

At this point, Doll defines what he means by ‘a smoker’:

A smoker was therefore defined in this inquiry

as a person who had smoked as much as one cigarette a

day for as long as one year, and any less consistent amount

was ignored.

He raises the point that people might have difficulty in remembering accurately. To check for accuracy, he re-questions 60 people after six months. He finds that only a very small number of people changed their account, and only in a small way. He gives details of the re-questioning in Table 3. I propose not to reproduce Table 3. It tells us little.

Smokers and non-smokers.

I have already shown Table 4, but I shall reproduce it here:

Table 4:

Disease………………Smokers………………Non-Smokers.

Lung Cancer…………..647………………………….2.

Non-LC……………….622…………………………27.

As I said, I have left out the female element for simplicity.

Amount of smoking.

As I said earlier, Doll now moves on to break down the figures by number of cigarettes smoked. In the first instance, he asks about number of cigs smoked immediately prior to entering hospital.

Table 5 gives the figures:

Disease……….Cigs smoked.

……………..1 – 4……5 – 14……15 – 24……25 – 49……50+.

LC………….33……..250………..196……….133………32.

……………(5%)……(38%)……..(30.3)……..(21%)…….(5%).

Non-LC…….55…….293………..190…………71……….13.

…………….(8.8%)…(47.1%)……(30.8%)……(11.4%)….(2.1%).

Observe the bold numbers. Doll’s point here is that the Non-LC patients smoked considerably less cigs than the LC patients. For example, in the Table, at the lower end, more non-LCs smoked less cigs than did the LCs. At the other end of the scale,  133 LC patients smoked more than 25 cigs per day, whereas only 71 non-LC patients smoked more than 25 cigs per day. (By the way, he converted pipe tobacco into the equivalent of cigs by a simple formula)

He produces two graphs to illustrate (one for men and one for women). No need to reproduce them.

The next Table concerns maximum amount of tobacco per day ever consumed regularly. Here is Table 6:

Disease……………….Cigs Smoked.

…………..1 – 4…….5 – 14……15 – 24…….25 – 49……..50+.

LC………..24……….208………196………..174………..45.

……………(3.7%)….(32%)……(30.3%)……(26.9%)……(7%).

Non-LC…..38………242………201………….118……….23.

…………..(6.1%)……(38.3%)…..(32.3%)…….(18%)…….(3.7%).

Again, I have emboldened the figures which Doll considers to be the important ones – those which illustrate the difference in tobacco consumed by LCs and non-LCs. The important thing is to note that, at the left of the Table, the ‘light smoking’ end, more non-LCs smoke only a little as compared with LCs. At the right side of the Table, the ‘heavy smoking end’, the opposite is true – LCs smoke more than non-LCs.

Table 7 goes a step further. Doll estimates the total amount of tobacco consumed in the lifetimes.

Here is Table 7:

Disease…………….Total cigarettes  consumed EVER.

……………..365…….50,000……150,000……250,000…….500,000.

LC………….19………145………183…………225…………..75.

……………..(2.9%)….(22.4%)….(28.3%)……(34.8%)………(11.6%)

Non-LC…….36……..190………182…………179………….35.

……………..(5.8%)….(30.5%)….(29.3%)……(28.9%)……..(5.6%).

Again Doll shows that the LCs smoked more heavily than the non-LCs.

Smoking history.

Table 5 showed figures related to most recent amount smoked. Doll says that, since Tables 6 and 7 (total numbers of cigs smoked per day, and total consumed during lifetime) closely relate to Table 5, then he is in order to simply use  most recent amount smoked as sufficient measure during the rest of the report. This is because he is now going to examine:

Comparisons of the age at which patients began to

smoke, the number of years they have smoked, and the

number of years they have given up smoking are shown in

Table VIII.

Table 8 has three distinct sections. To make it easier to read, I have split Table 8 into separate sections.

Table 8 (a): (NB. Both male and female patients are lumped together in this table)

Age of starting smoking.

Age………………………LC………………………Non-LC.

Under 20……………….541 (78.6%)…………488 (75.1%).

20 – 29…………………118 (17.2%)…………129 (19.8%).

30 – 39………………….17 (4.2%)…………….22 (5.1%).

40 +…………………….12………………………11.

(The % for 30s and 40s includes both).

Again, I have highlighted the important figures, which are that LCs started smoking earlier than non-LCs. (Well..the under twenties did – the over twenties didn’t).

Table 8 (b):

Number of years smoking.

Years smoking……….LC……………………….Non-LC.

1 – 9…………………..14 (5.1%)……………….18 (7.7%).

10 – 19……………….21………………………….32.

(The % lumps both groups together).

20 – 39……………..351 (51%)……………….338 (52%).

40 +………………..302 (43.9%)……………..212 (40.3%).

The longer people have smoked, the more likely to have LC.

Table 8 (c):

Years stopped smoking.

Years stopped…………..LC………………………Non-LC.

0……………………….649 (94.3%)…………..590 (90.8%).

1 – 9………………….  30 (4.4%)……………….37 (5.7%).

10 – 19………………….4 (1.3%)……………….14 (3.5%).

20 + …………………….5……………………….9.

(The 10s and 20s are lumped together as a %)

Doll reckoned that Tables (a) and (b) are not very significant, but that Table (c) is significant. Here’s what he says:

It will be seen that the lung-carcinoma patients showed

a slight tendency to start smoking earlier in life, to continue

longer, and to be less inclined to stop, but the differentiation

is certainly not sharp and the difference is

technically significant only with respect to length of time stopped.

Perhaps the significant bit is the percentage of the nine 10s and 20s LCs who had stopped over 20 years before as compared with the twenty three non-LCs.  He does not say.

Cigarettes and Pipes.

Doll mentions the difficulties of working out the quantities of tobacco smoked by pipe smokers. The figures suggest that pipe smokers do not suffer from lung cancer to the same degree as cig smokers. He suggests that the reason for this is that pipe smokers smoke less tobacco. He suggests that they could be classed as ‘light smokers’. He gives little factual info re pipe smoking and there are no tables.

Inhaling.

The great statistician, Fisher, was not very happy about the way in which Doll brushed aside the indications given by the replies to his questions about inhaling. Remember that all the patients who smoked were asked whether they inhale or not, so we are not talking about some small percentage. Thus, any findings about inhaling are just as important as any others, if not more so. Here are the figures quoted in the study about the percentage of inhalers in the two groups:

(Not a table in the report!)

Patients……………..inhaled………………did not inhale.

LC (688)……………..61.6%……………………38.4%.

Non-LC (650)……….67.2%……………………32.8%.

Bearing in mind what I said about ALL the patients (both LC and non-LC) being involved, the difference in the percentages cannot just be lightly brushed aside, as Doll attempted to do. In this case, I shall quote the whole paragraph about inhaling:

Inhaling.

Another difference between smokers is that some inhale

and others do not. All patients who smoked were asked

whether or not they inhaled, and the answers given by

the lung-carcinoma and non-cancer control patients were

as follows: of the 688 lung-carcinoma patients who smoked

(men and women) 61.6% said they inhaled and 38.4% said

they did not; the corresponding figures for the 650 patients

with other diseases were 67.2% inhalers and 32.8% non-inhalers.

It would appear that lung-carcinoma patients

inhale slightly less often than other patients.

However, the difference is not

large, and if the lung-carcinoma patients are compared

with all the other patients interviewed, and the necessary

allowance is made for sex and age, the difference becomes

insignificant.

Fisher did not agree that the difference was insignificant. In fact, he said that inhaling seemed to be protective to a certain extent. That may be so, but what bothers me is Doll’s apparent attempt to brush the matter aside. Does that imply a certain bias?

It is interesting to note that the question about inhaling was dropped from the much larger Doctors Study which followed.

The statistical facts upon which Doll bases his case that smoking causes lung cancer are now concluded. In the next section of the paper, Doll addresses various ‘objections’ which might be raised against his results, such as bias of one sort or another.

—————————————

Interpretation of the results.

In this section, Doll deals with various potential confounders and biases. Here is Doll’s list:

  1. Could the LC patients in the study have not been truly representative of LC patients generally?
  2. Were the non-LC patients truly comparable?
  3. Could the LC patients have exaggerated their smoking habits?
  4. Was there bias among the researchers?

He produces Tables 9 to 14 which purport to show that there was no bias. I do not propose to reproduce them. Suffice to make the following comments:

As regards 1:

A number of patients came from areas outside London proper. This was to be expected since the better hospitals for treating cancer were in London. But Doll had figures for District hospitals (‘local feeder’ hospitals, one might say). Table 9 gives these figures. These figures show that LCs smoke more than non-LCs, even though there are only a few people involved, in much the same proportions.

As regards 2:

Doll produced Table 10. This table shows the proportions of number of cigarettes smoked by people suffering from conditions other than LC, such as heart disease.  He says that the proportions are OK.

As regards 3:

I haven’t seen this question actually addressed.

As regards 4:

Doll goes to some trouble to show that his researchers were not biased. Tables 11, 12 and 13 refer.

Table 11 is concerned with the question of whether or not the researchers chose light smokers. He shows that hospitals referred lots of light smokers, but researchers chose heavier smokers.

Tables 12 and 13 are tricky. They show that researchers truly recorded what patients said. Thus, in the cases where researchers interviewed people thought to have lung cancer and subsequently found not to have lung cancer, the researchers truly recorded that their smoking was ‘light’.

By the above explanations and tables, Doll showed that there was no bias.

———————————–

We now come to the final section – The Discussion.

The first para says that bias can be discounted. Doll concludes that there is a real association between smoking and LC. An interesting observation, however, is this:

Further, the comparison of the smoking

habits of patients in different disease groups, shown in

Table X, revealed no association between smoking and

other respiratory diseases or between smoking and cancer

of the other sites (mainly stomach and large bowel). The

association therefore seems to be specific to carcinoma of

the lung.

Now is that not interesting? The archbishop (or even pope?) of tobacco control said that his study showed that smoking only affects lung cancer and nothing else!

There are some tricky  arguments in ‘The Discussion’.  How about this one:

If it be assumed that the patients without carcinoma of

the lung who lived in Greater London at the time of their

are typical of the inhabitants of Greater London

with regard to their smoking habits, then the number of

people in London smoking different amounts of tobacco

can be estimated.

What!!!! Sick people, many of them aged and damaged, can be regarded as typical of the whole population of London, young and old? Is this guy out of his mind?

Frankly, I have had enough. What he says in his ‘Discussion’ is neither here nor there – it is just his opinion. There is in fact a lot of speculation in the ‘Discussion’. I have found that difficult to read and understand. At the moment, the mixture of statistical inferences and speculation has defeated me. But I shall come back to it tomorrow and figure it out.

For now, I shall ‘publish and be damned’.

APPARENT ERRORS IN THE DOLL AND HILL STUDY.

Taken mostly from comments. These are very briefly put. People should refer to the original comments on the posts for more detail. The posts in question are dated between xxx and xxx.

1. Doll gave the impression in his study that he was not personally involved in gathering data, but he admitted later to being involved in certain re-interviews.

2. London was not representative of the population of the UK at the time. There was a significant excess of bronchitis, pneumonia and tuberculosis there at that time.

3. The influence of gassing with mustard gas of soldiers in WW1.

4. The apparent ‘stupendous’ increase in lung cancers (hereafter referred to as LCs) must have had some of its origin in much improved diagnosis.

5. Doll made no attempt to compare ‘life experiences’ such as occupation, war and forces service, genetic and ailments history, etc.

6. Lack of representation in LC cases of older people.

7. Inhaling protective?

8. Exclusion of over 75s.

9. LC a ‘modern’ disease? Smoking had been going on for 300 years. Why the apparent sudden increase in LCs? Diesel fumes? Heavy industry? Air pollution?

10. A misclassification (of lung cancers and non-lung cancers which mimic lung cancer) error of only 1% would render Doll’s figures statistically insignificant.

11. Doll said that the increase in LCs was ‘stupendous’, but the reality was that the occurrence of LC was ‘stupendously’ small as a proportion of smokers.

12. Contrary to what Doll claimed, his figures show that not enough persons in LC wards from older age groups and from heavier smoking patients. The highest number of patients were in the middle-age group and in the 15 – 24 cigs-per-day group. Numbers of patients should have been greater in the ‘higher smoking for longer’ groups than was seen.

13. This is a bit complex. 99% of LC patients were smokers and 96% of non-LC patients were also smokers. But the percentages depend upon Doll’s definition of ‘a smoker’ as ‘one who smoked at least one cigarette per day for as long as one year’. Doll would defend this definition on the grounds that he used the same definition for both groups – LCs and non-LCs. Thus, he was comparing like with like. Did he therefore exaggerate the number of smokers by setting the definition so low? The answer is indirect. It does not matter since people who tried smoking for a little while would appear in the ‘given up smoking for X years’ table! The fact is that hardly anyone at all had stopped smoking, therefore, the definition was irrelevant – people either smoked or they did not. Very few people appeared in the 1 -4 cigs per day group in the table. I had understood that only some 60% of the population smoked around that time – 1950. But how many of the male population smoked? I have read that it was about 95%! ALL THE MORE REASON THEREFORE TO BELIEVE THAT THERE OUGHT TO HAVE BEEN VASTLY MORE LUNG CANCER PATIENTS THAN THERE WERE, IF SMOKING CAUSED LUNG CANCER.

Comment on this page is disabled. It is intended just to be a record. If the reader wishes to comment, by all means do so. Whatever subject I may be talking about in any given post, if you wish to comment on this analysis, do so. Merely say that your comment is ‘about the hospital study analysis’. No problem at all.

Advertisements

%d bloggers like this: