A Half-life of Burch
7. Passive Smoking
Sandler on passive smoking
Since Hirayama’s study in 1981, a flood of papers claiming to demonstrate the dangers of passive smoking had appeared. At first they mostly used the smoking habits of relatives of their subjects as a way of estimating passive exposure to cigarette smoke. Later, biochemical assays of saliva made more accurate estimates possible, and it became evident that most exposure occurs in the workplace. Burch commented on some of this work, with open sarcasm and ridicule, for instance two papers by Sandler, Wilcox and Everson which appeared at the end of 1985 and the start of 1986.
The authors claimed, by analysis of data from a hospital in North Carolina that passive smoking, measured by a count of smoking relatives, increased the risk of lung cancer. Burch pointed out that their reasoning contained a hidden assumption which implied that active smoking cancels out the effects of passive smoking and that passive smoking in childhoood would have to be positively beneficial. A fuller version of this section with the mathematics is here.
“I have to admit being partial to ‘A + P = 0’,” Burch wrote to Thornton.
Repace and Lowrey
J.L. Repace and A.H. Lowrey attempted to derive the number of deaths attributable to passive smoking as measured by various different techniques. The contents of their paper A quantitative estimate of nonsmokers’ lung cancer risk from passive smoking were ridiculed by Burch in a critique that can hardly be improved on. A summary of the paper in the form of quotations from Burch is here.
On 20 June 1986 The Times printed a Medical Briefing with the headline “Passive smoking no significant danger”. This raised the curtain on an affair which was to continue all year. Burch was only marginally involved, but the story is worth telling as it marks the point at which it became heresy to question the dangers of passive smoking. The Times report began
Over the past few years continuous passive smoking (inhaling other people’s smoke) has become widely accepted as dangerous but a major piece of research to be published in next month’s British Journal of Cancer has the hallmark of turning the received wisdom into one of the medical controversies of the year.
The researchers – from the Institute of Cancer Research in Surrey – conclude that passive smoking, for life-long non-smokers, carries no significant increase in risk of lung cancer, bronchitis, or heart disease – all commonly associated with smoking.
The principal author of the leaked paper, which duly appeared in July, was Peter Lee, and The Times had fairly reported its conclusions. A study of 47 lung cancer patients who had never smoked in their lives (or so they claimed) and their spouses, as against otherwise similar controls who did not have lung cancer and their spouses, showed no significant risk of lung cancer from spouse’s smoking: the relative risk of 1.33 was not statistically significant. In one respect the Times report was misleading. It stated that the study “involved over 12, 000 people” whereas this was the pool from which the 47 subjects and 96 controls were drawn. The Lee paper also contained a survey of previous studies, most of which proved to suffer from poor experimental design.
On 25 June, The Times found itself reporting the conclusion, but not the content, of a speech by Sir Richard Doll disputing Lee’s finding. On 5 July it printed a letter from Professor Robin Weiss, director of the Institute for Cancer Research, and Julian Peto. These two cited the relative risk of 1.33 as reason for “further restrictions on smoking in public places”. As for statistical significance “It need not imply that an effect is negligible… but that a particular survey included too few subjects to demonstrate an effect credibly.” (Yet again, a study with negative results allegedly lacked power.) Lee drafted a reply to Weiss and Peto, but if it was sent it was not published.
On 23 June, Brian Simpson of the Tobacco Advisory Council arranged for copies of the Times report to be distributed to every member of Parliament. In consequence, Michael Brown MP tabled an Early Day Motion (a motion which is debated but not put to a vote) noting its conclusions and calling for the current government funded campaign on passive smoking to be axed. Weiss wrote an angry letter to Brown claiming that his Institute had been misrepresented and “suggesting” that the motion be withdrawn. Brown replied, “I have no intention of agreeing to this impertinent request” since Lee’s actual paper substantiated the Times summary and government spending on the passive smoking campaign was in fact higher than the motion had noted.
Burch missed all this, being on holiday on the continent.
We had a most enjoyable time in Spain and France with only one day of really bad weather. We took advantage of it to visit two caves, both near to Altamira. My appreciation of Spanish cathedrals has been greatly enhanced and I was most impressed by the New Cathedral (15th Century) in Salamanca. In that city, alas, the nearside window of our car was shattered by thieves who stole the radio and miscellaneous contents.
Thornton, who had been following matters closely, brought him up to date on the story. “Who leaked the findings to the press,” Burch wondered. “Possibly a journalist or someone in Br. J. Cancer, not I think a Tobacco Industry [employee],” Thornton told him. (The leak, as he made plain to Simpson, had not strengthened the industry’s hand.)
Thornton in Canada
Thornton then departed for Canada, in part to meet Theodore Sterling on Burch’s recommendation. His report to BAT makes entertaining reading.
I had expected to meet T.D.S. at Simon Fraser University, Barnaby, Vancouver, but he had requested I meet him in downtown Vancouver, where, it subsequently transpired, he has an apartment and a business office.
It also transpired that T.D.S. was virtually immobile due to a yachting accident and the meeting took place in brilliant sun, on the top of the apartment with T.D.S. resplendent on a sun-bed and distinctly resembling, in several ways, one’s popular conception of an ancient sea-captain, possibly Long John Silver. I was offered a drink, asked for coffee, but given a very large bottle of whisky and told to pour my own as the elderly secretary “didn’t know what to do with it” (quote)
T.D.S is interested in low risk epidemiology and he said that he liked to do three kinds of research:
(b) to refute (i.e. bad papers on passive smoking), and
(c) to anticipate (i.e. other people’s problems).
He tended to assume he was a rather important person, and seemed disappointed to hear I was only a chemist and not medically qualified. I took T.D.S. to be a radiation physicist.
I was given a paper, in the press, criticizing Repace and Lowry’s calculations on the effects of passive smoking and lung cancer, and another relating to lung cancer incidence in the U.S.A.
We discussed low-risk epidemiology, T.D.S. agreeing with Burch that much of the work was of low quality. T.D.S. said environmental chemicals must be responsible for some cancers and that Burch could only be partly right (genetic disposition for cancer).
T.D.S. said he could easily identify neutral epidemiologists who could identify poor work. T.D.S. said he believed that Sir Richard Doll must be financed by the Chemical Industry to make the sort of remarks he does. T.D.S. seemed to imply that industrial interests controlled most things, not surprising in view of the existence of Theodor Sterling Ltd., with his son Elia Sterling, as Research Director. Elia joined us for part of the discussions.
I suspect that the business interests of the Sterling family (which I had not known about before the trip) were not far away during our discussion. T.D.S.’s expertise in low risk epidemiology, passive smoking, radon levels etc. are probably available through Theodor Sterling Ltd. for an appropriate fee, but there was not the free discussion I had anticipated. In view of discussions with the Fraunhoffer Institute I decided to leave discussions with T.D.S. somewhat indeterminate.
T.D.S.’s knowledge of epidemiology, particularly in the U.S., could be useful, but some thought is needed as to how to apply this to the project with the Fraunhoffer Institute. Also, I would like to take another view on the scientific competence of Theodor Sterling Ltd. before proceeding further, since T.D.S. is clearly not in the same intellectual ball-park as Philip Burch, even if his professed views are more conventional, and his bank balance larger.
Thornton shared the story of his trip with Burch, who replied “I am sorry to-hear about Ted Sterling; I find his company very congenial although I disagree with so many of his environmental views.”
The events of June and July had made Lee enemies, as he was soon to discover. The September issue of the British Journal of Cancer led with a Guest Editorial, Passive Smoking, by Julian Peto and Richard Doll sternly denouncing the Times report, the Early Day Motion and Lee’s paper itself. These had led to “the unauthorised association of the Institute’s name with misleading propaganda”. Passive smoking “must be assumed” to cause lung cancer because “it is now generally accepted that a safe threshold is unlikely to exist for most carcinogens”. Though existing studies made questionable assumptions they came to the right conclusion, and therefore “the suggestion that the possibility of a cancer hazard should be added to the certainty of unpleasant pollution in the movement against unrestricted smoking in public places seems entirely reasonable”.
A significant passage in the editorial reopens the question of the dose-response relationship, linear or quadratic. Doll and Julian Peto are now prepared to entertain a possibly linear relation again, because a quadratic relationship extrapolated downwards to low doses of smoke predicted a negligible increase in the risk of lung cancer.
Lee protested, in a private letter to Julian Peto and a published letter to the editor of the British Journal of Cancer that the editorial misrepresented him. A reply by Peto and Doll addressed minor points but not the central one of poor experimental design. “The most surprising aspect of Mr Lee’s letter, however, is that it should have been addressed to us at all.” The Times and the Tobacco Advisory Council were to blame for any misrepresentation.
Lee had been working on passive smoking for years, and had a second study ready for publication. It concerned a study of 1775 British men and women comparing self-reported smoking rates with rates estimated from cotinine levels in saliva. 20 “non-smokers” with implausibly high levels of cotinine were mostly married to smokers, and statistical analysis suggested that the entire supposed risk of lung cancer from spouses’ smoking could be accounted for by “misreporting” – in plain language, lying.
Thornton had an advance copy which he passed to Burch, who greatly enjoyed it. “One of the interesting things about it is that the conclusions are valid whether the association between active smoking and lung cancer is causal, constitutional or a mixture of the two… Fig. 1 is priceless.” (Figure 1 graphically highlighted the discrepancy between self-reported smoking and the evidence of cotinine.)
Lee delivered the paper at a conference in Switzerland in September which was attended by Thornton and also Nicholas Wald who, as chance would have it, had been conducting passive smoking research on very similar lines to Lee. The two men held what Thornton described as a “lively exchange”.
Burch, writing to Thornton on the Guest Editorial, commented:
This is all very entertaining. Has the earlier rapport between Lee and Peto broken down? But why did the Guest Editorial feature J. Peto rather than R. Peto? Was this because of J. Peto’s affiliation? The resort to the Guest Editorial itself looked like an act of desperation on the part of the establishment and extreme indignation over the report in The Times. Sooner or later someone might even raise the question: Does association imply causation?
Will the Br. J. Cancer publish the letter by Lee et al. which, in any case, is in need of some polishing? If this is Chapter 2 then I can barely wait for Chapter 3.
With ill-concealed delight.
Burch was prescient about Lee’s prospects of publication. What happened next was summarised, somewhat telegraphically, by Thornton in a memo to Sheehy (PNL 2 is the second Lee paper, NW is Wald).
Discussions between Lee and Wald on possible joint paper. Agreed not possible but agreed to exchange drafts: Draft of PNL 2 sent to Wald by Lee.
ca November 1 1986 – Nothing heard from Wald for 4 weeks. Then Wald’s draft sent to Lee with note stating already submitted.
November 8 1988 NW paper appears in BMJ, some 7 days after submission thereby establishing a world record for quick publication and instant refereeing, not to mention type-setting. Leaked to press, but only the day before.
ca November 10 1986 Lee submits PNL 2 to BMJ for publication .
ca November 15 1986 – Lee informed that BMJ will not accept PNL 2, but Lee can write 400 words commenting on Wald or a precis of his study (hardly enough for both).
The Wald paper appeared as Does breathing other people’s cigarette smoke cause cancer? It is a reanalysis of existing work on passive smoking. Its thesis is that the only problem with them was small sample sizes, and that the samples combined reinforce the conclusion that passive smoking is harmful. What of Lee’s point about misclassification? A section of the paper “Discussion” argues at length that the effect disappears if you consider lies told by ex-smokers as well as by current smokers. A short paragraph explains that a factor of 3.3 has been “adapted” from an estimate of 3.94 in “P. N. Lee, unpublished findings”
Wald’s letter to Lee informing him of the new facts of life, and Lee’s reaction, can speak for themselves.
Wald to Lee, 3 November 1986:
As I indicated, the BMJ has decided to take our paper. They are trying to publish it before the publication of the NAS Report and have just delivered to us the proof which we have corrected. In doing so I pointed out that the 10-fold relative risk in your own analysis is linked to the 1.4% of misclassified current smokers rather than the 12% of all estimated misclassified smokers. This should avoid the potential misunderstanding.
I will ask Stephanie to ring you about the other matters.
With kind regards
Lee to Richard Smith, editor of the British Medical Journal, 5 November 1986:
Dear Dr. Smith,
I enclose 3 copies of my paper “Passive smoking and lung cancer. Association a cause of bias?” which I would like to submit for publication in the British Medical Journal. As I explained on the phone, this paper relates strongly to the paper by Prof. Nicholas Wald which you have recently accepted. Both papers discuss the likely magnitude of bias to the passive smoking/lung cancer association caused by a small proportion of smokers denying they smoke. The possibility of such a bias is one I had considered in an unpublished document in 1983 and had alluded to at a conference in 1984. I had always felt it was necessary to get good data in order to be able to quantify this bias and the enclosed paper describes the results of the studies designed to do so and the conclusions from them. These contrast strongly with Wald’s. I have had a number of discussions with Wald on the issue (though I had not seen his paper until earlier this week) and I am somewhat surprised that he makes no mention of the fact that the original idea of such a bias came from me. My 1983 document was sent for consideration to the National Academy of Science Committee on Passive Smoking at Wald’s request, and I imagine was the reason why Wald started work in misclassification.
We met somewhat over a month ago and at that time my understanding was we would exchange drafts of our papers for comment. I duly sent my paper to him a month ago and awaited his reply before submitting. His replay came a few days ago, indicating that he had already submitted because of pressure from the NAS.
I would obviously value highly the chance to get my own point of view heard on this important issue.
Peter N. Lee
Burch, a veteran of this kind of treatment, exchanged letters with Lee.
I hope that you will reply to Wald et al. in this week’s BMJ!
I am indeed responding to Wald et al in the BMJ. I had submitted, at virtually the same time as Wald, a paper which came to a completely different conclusion. Not totally unexpectedly, in view of medical journals’ general reluctance to publish papers contrary to mainstream opinion, it was rejected although they asked me to write a letter. This letter and the paper, which I am now submitting to the Lancet without a lot of hope, are enclosed.
The BMJ appears to have been guilty of outrageous discrimination in publishing the paper that most favours the anti-smoking policy of the BMA. Crusades in medicine are, I suppose, inevitable but they do not advance the cause of scientific objectivity. I have also despatched a letter to the BMJ (copy enclosed) but I fear it will be even less to their liking than yours.
The physicians’ trade union, the British Medical Association (BMA), is the publisher of the British Medical Journal (BMJ).
To Burch’s surprise, both his letter and Lee’s were published. “Did the editor of the BMJ suffer a twinge of conscience?” he wrote to Thornton.
All credit to the BMJ, given its own heavy bias, for publishing all three letters. (I do not suppose that Peter Lee feels as charitable.)
Three letters on the topic appeared in the British Medical Journal for 6 December 1986. P. Sherwood Burge and A. S. Robertson observed that the workplace accounted for far more passive exposure to smoke than did spouse’s smoking. Burch wrote that correlation is not causation and once more publicised Whitehall and MRFIT. (“I content myself with going for the jugular” he told Thornton.) Lee supplied a concise version of his evidence and conclusions.
Lee’s paper was rejected in turn by the Lancet and the British Journal of Cancer and finally appeared in Human Toxicology in 1987.