P .N . LEE STATISTICS AND COMPUTING LTD . From: Peter N . Lee 25 Cedar Road Consultant in Statistics end Sutto n Adviser in Epidemiology and Toxicology Surrey SM2 5DG TELEX 91714 3 FAX 01-643-6453 453 Telephone : 01 .642 8265 (4 lines ) VAT Reg. No. 3184017 78 PNL/DPM 19 March 198 7 Dr .R .E .Thornton British-American Tobacco Co .Ltd . Westminster Hous e 7 Millbank LONDON SWIP 3JE . 2 3 MAR X37 ) Dear Ray, I enclose my comments on the paper by Gori you sent me . As you will see, I think the paper is pretty valueless, partly as it is completely unbalanced, partly as the (many) wild claims made are not substantiated by detailed evidence . Best wishes . Yours sincerely , encl . Directors : P .N . Lee, MA(Oxon) M . Lee Registered in England No . 1688551 Registered Office : 228 Bishopsgate, London EC2M 400 . Source: http://industrydocuments.library.ucsf.edu/tobacco/docs/jfkp0203 -1- "On the positive side : a position paper" Some comments on the January 87 paper by Gio Cor i Author : P .N . Lee Date : 19 .3 .8 7 1 . General impression s While there are undoubtedly some beneficial effects of smoking, this paper seems to me "over the top" in its enthusiasm for the cause . There are 3 major flaws in the argument : (i) inclusion, in the list of diseases negatively related to smoking, for some of which the evidence does not justify this status . (ii) overestimation of the effect that taking effects of confounding variables might have in reducing the number of deaths related to smoking . (iii) lack of justification for the wild statement that "the net sum of longevity effects of smoking may actually turn out positive . " 2 . Diseases "prevented" by smokin g Let us start by looking at the diseases which Cori claimed to be prevented by smoking . (As an aside, "prevention" is surely the wrong word, implying to many that smokers do not get these diseases at all) . Source: http://industrydocuments.library.ucsf.edu/tobacco/docs/jfkp0203 -2- 2 .1 Breast cancer . Baron's excellent review of smoking and oestrogen-related disease in the American Journal of Epidemiology (1984, 119, 9-22, see also Review 245) cited evidence from 10 case-control and 8 prospective studies . Relative risks for smoker vs non-smoker, ranked in order, were as follows : 0 .65, 0 .69, 0 .71, 0 .77, 0 .80, 0 .81, 0 .83, 0 .85, 0 .86, 0 .88, 0 .92, 0 .99, 1, 1, 1, 1 .1, 1 .1, 1 .13, 1 .2 and 1 .39 (some studies providing more than one estimate) . While there are more relative risks less than 1 (12) than greater than 1 (5), the median reduction is only about 10% . Baron concluded that the evidence regarding smoking and breast cancer is inconclusive and my reading of the literature, including later published studies, tended to agree with him . In contrast to the 10% median reduction, Gori notes that "several studies show that smoking has a preventive effect of about 20-30% for breast cancer", apparently dismissing studies which found no protective effect only on the grounds that they were "earlier" . He also cites "newly published evidence" that "such an effect is more evident only for specific types of breast cancers, namely those poor in oestrogen receptors ." This is interesting and may be important, but unfortunately Gori does of directly reference the study, and I am not aware of it . (As another aside, it is totally unsatisfactory for a paper not t o attach references to the text, and only to give long lists at A O the end) . • Source: http://industrydocuments.library.ucsf.edu/tobacco/docs/jfkp0203 O W CO O to -3- 2 .2 Endometrial cancer . Baron cites data from 4 studies of endometrial cancer incidence, all showing a relative risk less than 1 (0 .40, 0 .69, 0 .79 and 0 .83) and 2 studies of death, showing no evidence of a reduction (1 and 1 .89) . Baron argues that though the studies of incidence are more likely to be relevant than the studies of cancer deaths, a number of doubts about the adequacy of the incidence studies means that more evidence is needed before a conclusion is reached . A subsequent case-control study by Lesko et al (Review 278) found a relative risk of 0 .7, but did not fully resolve all the unanswered questions regarding the possible role of smoking in protecting against endometrial cancer . Overall, the data are strongly suggestive that smoking may reduce endometrial cancer risk, but 50% is probably overoptimistic . 2 .3 Ulcerative colitis . I have recently reviewed the evidence here (paper by Cope et al to appear in Human Toxicology) . The evidence is very consistent that relative to never smokers current smokers have a reduced risk - I estimated from 16 independent data points a RR of 0 .25, i .e . a reduction of 75% . Cori's "over 75%" is not too far from the truth . Source: http://industrydocuments.library.ucsf.edu/tobacco/docs/jfkp0203 -4- 2 .4 Colon cancer . While ulcerative colitis can predispose to colon cancer, this only accounts for a very small proportion of colon cancer . I are totally unaware of the evidence supporting Cori's claim that smoking reduces risk of colon cancer by up to 50% . Indeed, I thought the epidemiological evidence was consistent with smoking having no reducing effect at all . Thus, the ACS million person study, the US Veterans Study and the British Doctors Study all find a relative risk for smoking in relation to colon cancer of fractionally over 1 . Elsewhere (TG 1986), I considered whether Garfinkel was sensible to choose colon cancer patients as non-smoking related disease controls in his case-control study of passive smoking and lung cancer, and concluded he was . I thus do not remotely believe Cori's claim here - where is the evidence ? 2 .5 Parkinson's Disease . That smoking is negatively related to Parkinson's disease is clear enough from numerous studies, which I am currently reviewing . It is important to note, though, that people with PD have normal life expectation so that the claim of additional deaths if smokers were to quit is dubious . The statement that smoking is responsible for the "prevention of neurological diseases such as Parkinson's" falsely implies a much more general conclusion than the evidence permits . 2 .6 Sarcoidosis . What is the evidence here? Source: http://industrydocuments.library.ucsf.edu/tobacco/docs/jfkp0203 -5- 3 . Confounding factor s Cardiovascular disease (CVD) is certainly multifactorial, Smoking is certainly associated with reduced body weight and reduced blood presure . Intervention trials on smoking and cardiovascular disease have certainly not produced a clear result . However, it is not made clear that taking account of the reduced body weight and blood pressure statistically would only tend to reinforce the association between smoking and CVD after all, smokers have an increased risk despite these favourable characteristics . To make a valid case for the smoking/CVD association being non-causal, or partly so, one needs to cite CVD risk factors more prevalent in smokers . I find it hard to believe the claimed effects of giving up smoking on obesity are so large as stated on average, nor that they would have this effect on premature deaths . After all, there is abundant evidence that ex-smokers have a reduced risk of CVD compared to continuing smokers . One just cannot make such statements without detailed supporting evidence . I do not propose to comment in detail on whether the number of deaths caused by asbestos, radon and other occupational exposures are anything like as large as claimed . Rather, I will make 2 points - firstly one must have the detailed supporting evidence, and secondly, a death related to these factors may also be attributable to smoking . If smoking multiplies risk by J~:CD 10 and asbestos by 5, a lung cancer in an asbestos worker who p W Co O Source: http://industrydocuments.library.ucsf.edu/tobacco/docs/jfkp0203 -6- smokes is 80% due to asbestos and 90% to smoking, by the standard calculations . The final statement of the section on lung cancer is ludicrous . Even given (a) that of the 110,000 new lung cancer cases a year, perhaps only between 35,000 and 65,000 a year can be attributed to smoking and (b) that there are a few thousand extra lung cancer cases in non-smokers that are not diagnosed, the proportion of lung cancers due to smoking would still be an extremely substantial one . How can one expect to persuade anti-smokers that it is "only a small fraction of lung cancers" that "could be linked to smoking . " 4 . Overall effec t The only justification for the wild statement that "the net sum of longevity effects of smoking may actually turn out positive" is a table before the references that is not mentioned at all in the text . Although no details are given as to how the figures were calculated, some major criticisms are obvious : (a) obesity is only rarely entered as the underlying cause of death . In no way are there 200,000 reported deaths with underlying cause obesity per year in the US . Many deaths to which obesity contributes are from GVD . The fact that Gori gives no corrected deaths from CVD due to smoking and -50,000 from obesity due to smoking, implies he Source: http://industrydocuments.library.ucsf.edu/tobacco/docs/jfkp0203 -7- believes that smoking is beneficial as regards CVD, no reason being given for the conflict with the epidemiology . (b) the colorectal contribution is unjustified as colon cancer is not negatively related to smoking . (c) these two contribute very largely to the overall negative effect . If one restricted attention to the diseases for which there is good evidence of a negative relation (Ulcerative colitis, endometrial cancer, Parkinson's Disease), there is clearly no question that the adverse effects for which there is good evidence of a positive effect - even excluding heart disease - clearly outweigh the relatively small benefits . Source: http://industrydocuments.library.ucsf.edu/tobacco/docs/jfkp0203