Rapid Responses to:

PAPERS: J B Copas and J Q Shi Reanalysis of epidemiological evidence on lung cancer and passive smoking BMJ 2000; 320: 417-418 [Abstract] [Full text]

Rapid Responses: Submit a response to this article

Rapid Responses published:

Passive Smoking: Wrong end of the telescope?

Norbert Hirschhorn   (12 February 2000)

Nothing new here

Stanton A Glantz   (12 February 2000)

No unpublished studies on passive smoking in Hong Kong

T H Lam   (14 February 2000)

Passive smoking risks: over- or underestimated?

T L P Watts   (15 February 2000)

Challenge to Professor T. H. Lam

Martha Perske   (15 February 2000)

Turning over the wrong stone

Kenneth C Johnson, James Repace   (17 February 2000)

Mooting the "Publication Bias" Issue

Leroy J Pletten   (24 February 2000)

Unfair scales for a visual test of publication bias

Christopher Cates   (4 March 2000)

Over-reliance on p values

Adam Jacobs   (25 March 2000)

Few unpublished studies on passive smoking

Lisa Bero   (6 April 2000)

Passive Smoking: Wrong end of the telescope? 12 February 2000
Norbert Hirschhorn, Consultant American University of Beirut Send response to journal: Re: Passive Smoking: Wrong end of the telescope?	In continuing to do meta-analyses on 37 published epidemiological studies on passive smoking and lung cancer by introducing some putative number of unpublished ones, Copas and Shi (BMJ 12 February) missed the essential biological point: As Hackshaw, et al. showed so well (1), passive smoking increases the risk of lung cancer linearly in a dose- related fashion, up to relative risks of 3 or greater. This makes eminent sense given the toxic chemical exposure, and is consistent with the finding that after a few hours of heavy exposure to sidestream smoke, non- smokers excrete nitrosamine metabolites (2). The essential biological questions are whether there is ever any "threshold" effect for creating DNA adducts in any one exposure; and whether multiple exposures inevitably result in some cancer risk, which when extrapolated to a large population with regular exposures amounts to a threat to the public's health. To put it more colourfully, what if "mad cow disease" were found to be associated with (and have a plausible biological mechanism for)lung cancer? The French would ban all British cigarettes! Norbert Hirschhorn, MD (1) Hackshaw AK, Law MR, Wald NJ. The accumulated evidence on liung cancer and environmental tobacco smoke. BMJ 1997; 315:980-8. (2) Hecht SS, Carmella SG, Murphy SE, Akerkar S, Brunnemann KD, Hoffman D. A tobacco-specific lung carcinogen in the urine of men exposed to cigarette smoke. New Engl. J. Med 1993;329:1543-6.
Nothing new here 12 February 2000
Stanton A Glantz, Professor of Medicine University of California San Francisco Send response to journal: Re: Nothing new here	My reaction to this paper is a big yawn. What these guys say is that they think that there is evidence of "publication bias" against small studies that reach the "negative" conclusion that secondhand smoke causes lung cancer. This is nothing new, nor is the analysis they present (based ons omething called a funnel plot). What they say is that they agree that a meta-analysis of the published studies on passive smoking and lung cancer shows a statistically significant increase in risk of 1.24. The compute that if only 60% of the studies that have ever been done were published and that the remaining 40% of studies that were done but never published --- and that no one has ever heard of -- were all negative, then the elevation in risk would only be 1.11, and not statistically significant. Of course, there is no evidence that these studies were ever done. In fact, our investigation (Bero et al, "Publication bias and public health policy on environmental tobacco smoke," JAMA 1994;272:133-136) suggests that there is NOT publication bias. In any event, the BMJ paper also points out that if only 70% of the studies were published, and all the unpublished tudies showed no elevation in risk, then the pooled risk wouldbe 1.13 and statistically significant p=.052. So, you could argue that they proved that, while failure to publish negative studies would lower the true risk of lung cancer associated with passive smoking, under any reasonable guess at how much "unpublished" research there was, there would still be an elevation of risk. Of course all this begs the question because, despite the fact that many people have tried to find these unpublished studies, no one has been able to find them. You can also be sure that the tobacco industry would make sure we knew about them. In other words, what this paper says is that if a bunch of people did studies that found no effect of passive smoking and lunc cancer and found no increase in risk, and we suddenly knew about these papers, then our estimate of how much the risk was elevated would be smaller. But the risk would still be elevated. The real killer from secondhand smoke is heart disease, not lung cancer. Heart disease kills about 10 times more people than lung cancer. And not even the tobacco industry has contested the evidence on asthma. So... what's the big deal?
No unpublished studies on passive smoking in Hong Kong 14 February 2000
T H Lam, Professor Department of Community Medicine, The University of Hong Kong Send response to journal: Re: No unpublished studies on passive smoking in Hong Kong	Dear Sir The meta-analysis of Hackshaw et al and of Copas and Shi included 37 studies, four of which came from Hong Kong. Of these, one was negative and this had the smallest number of subjects (Chan et al 1982: 84 cases and 139 controls). One was positive but was not statistically significant (Koo et al 1987: 86 cases and 136 controls), and the other two were positive and significant (Lam 1985: 60 cases and 144 controls; Lam et al 1987: 199 cases and 335 controls). Hong Kong is a small place and there have been few researchers involved in this research question. To the best of my knowledge, there are no negative studies in Hong Kong which have not been published and hence, there is no evidence that there is publication bias as far as Hong Kong is concerned. In the US EPA 1992 Report, the pooled relative risk (95% confidence interval), from the 4 Hong Kong studies with correction for smoker misclassification and weighting by the reciprocal of the variances of log odds ratio of the studies, was 1.48(1.21, 1.81) (p<0.01). When it is not known whether there are unpublished studies, publication bias should be considered as a possibility. However, when it is known that there are no unpublished negative studies, as in the case in Hong Kong, the pooled relative risk should stand.
Passive smoking risks: over- or underestimated? 15 February 2000
T L P Watts Send response to journal: Re: Passive smoking risks: over- or underestimated?	I was most interested in the paper of Copas and Shi(1) on possible overestimation of risks associated with passive smoking. I have long been concerned that all passive smoking studies actually underestimate true risks for the following reasons: 1. Few non-smokers are not frequently exposed to tobacco smoke in their daily activities. I live in a non-smoking house, travel on non- smoking trains, and work in non-smoking buildings. Yet I am exposed to tobacco smoke perhaps 15-20 minutes of each day, for instance when waiting for trains to arrive, entering and leaving my workplace, walking near smokers in towns and when going shopping. It may be impossible to find true negative control subjects for passive smoking studies. 2. The position in my childhood was far worse, with frequent exposure on public transport and many other places; this will have affected many control subjects in past studies. 3. Some subjected to passive smoking will undoubtedly become nicotine addicts, perhaps as children, and therefore become smokers themselves. I know of no good estimate of this risk and the subsequent damage. 4. As well as the serious risk of addiction, it is also unlikely for the above reasons that any study using realistic controls has been able to estimate the absolute effect of exposure versus total non-exposure to tobacco smoke. The effects of passive smoking therefore may be more serious than any studies have shown so far, publication bias notwithstanding. It is also likely that effects of smoking on smokers have been underestimated. I have no competing interests. (Dr) TLP Watts Senior Lecturer and Consultant in Periodontology, GKT Dental Institute, Floor 21, Guy's Tower, London Bridge, SE1 9RT. 1. Copas JB, Shi JQ. Reanalysis of epidemiological evidence on lung cancer and passive smoking BMJ 2000; 320: 417-418.
Challenge to Professor T. H. Lam 15 February 2000
Martha Perske, Private citizen Professional illustrator (see amazon.com) Send response to journal: Re: Challenge to Professor T. H. Lam	It seems that Professor T. H. Lam has omitted relevant information in his February 14 response to Copas and Shi. For example, Professor Lam states that the pooled relative risk (1.48) from the 4 Hong Kong studies on environmental tobacco smoke that were presented in the 1992 US EPA report should stand. What Professor Lam does not mention is that the US EPA, itself, raised questions about the reliability of most of those 4 studies. Listed below are direct quotes from the EPA report, pertaining to the Hong Kong studies. CHAN AND FUNG (1985): "Overall, methodological shortcomings hamper the interpretation of this study's results . . . the limitations of the original study, which was not designed to assess passive smoking, limit this study's value for assessing ETS exposure and lung cancer." US EPA 1992 report, pp. A-34,35. LAM ET AL. (1987): "Although this study was carefully conducted in most respects, the disregard for potential confounding effects leaves the authors' conclusion uncertain." US EPA 1992 report, p. A-98. LAM (1985): "Potential confounders (including age) have not been controlled for, however, so attribution of the elevated odds ratio to ETS exposure is uncertain." US EPA 1992 report, p. A-102. Also, Professor Lam is wrong when he says that a 95% confidence interval was used in the pooled relative risk of the 4 Hong Kong studies. On the contrary, the EPA did not use a standard 95% confidence interval as claimed by Lam. The report clearly states that a 90% confidence interval was used. US EPA 1992 report, Table 5-9, p. 5-28. Professor Lam further failed to mention that the US EPA 1992 report which he gives credance to was ruled null and void in a court of law. "A federal judge has ruled that the Environmental Protection Agency wrongly declared secondhand smoke a dangerous carcinogen..." (Washington Post, July 19, 1998, p. A1, "Secondhand Smoke Finding Struck Down") Martha Perske 159 Hollow Tree Ridge Road Darien, CT 06820 U.S.A.
Turning over the wrong stone 17 February 2000
Kenneth C Johnson, Senior Epidemiologist Cancer Bureau, Health Protection Branch, Health Canada, James Repace Send response to journal: Re: Turning over the wrong stone	Copas and Shi 1 assert that after allowing for publication bias the apparent average excess risk of lung cancer from passive smoking2 would drop from 24% to 15%. Despite the lack of data on which to base such a conclusion3, we are asked to believe this solely on the basis of a statistical inference that such data must be presumed to be hiding under a stone. However, they are turning over the wrong stone. More important is the underestimation of risk which occurs when studies assess exposure based solely on whether non-smokers either lived or did not live with a smoker, 2 when other exposure exists. Where other exposure is common (e.g., in childhood, in social situations, or in the workplace), lung cancer risks may be seriously underestimated. Spouses of non-smokers exposed in other circumstances will be misclassified as non exposed, contaminating the referent group, and attenuating the risk estimate. For example, Hackshaw et al.2 estimate that the odds ratio would have been 1.42 ( 1.21- 1.66) if spousal exposure alone was compared to those truly unexposed. By comparison, in a recent meta-analysis of risk associated with workplace exposure, Wells 4 found an estimated relative risk of 1.39 (95% confidence interval 1.15-1.68) for the five studies meeting basic study quality standards. Repace and Lowrey found that when both workplace exposure and an unexposed referent group were taken into account in the American Cancer Society study of passive smoking and lung cancer, a population relative risk of 1.2 increased to 1.7. 5 In fact, Repace and Lowrey modeled the risk of workplace exposure, estimating the average relative risk at 2.0 for U.S. office workers in the 1980's.5 This result is consistent with a value reported by Reynolds et al. for women with 30 or more years of workplace exposure, i.e. at ages at which lung cancer mortality begins to become significant.5 Of course, all of these analyses focus on average risk. Repace et al. 5 estimated that individuals at the 95th percentile (e.g., those experiencing high smoker density and low air exchange) have exposure -- and risk -- as much as four times as high as those at the median. This result is commensurate with observations of dose (ref 5) and risk (ref 2). In sum, turning over stones may indeed alter the estimated risk -- however, turning over the right stone suggests that in the original meta- analysis, the actual passive smoking-lung cancer risk is underestimated -- not overestimated. Kenneth C. Johnson, PhD, Senior Epidemiologist Environmental Risk Assessment and Case Surveillance Division, Cancer Bureau, Laboratory Centre for Disease Control Health Protection Branch Health Canada Ottawa, Ontario, Canada K1S 2Z7 Email and and James Repace, MSc., Health Physicist REPACE ASSOCIATES, Inc. Secondhand Smoke Consultants 101 Felicia Lane Bowie, MD 20720, U.S.A. Email: WEBSITE: 1. Copas JB, Shi JQ. Reanalysis of epidemiological evidence on lung cancer and passive smoking. BMJ 2000; 320: 417-418. 2. Hackshaw AK, Law MR, Wald NJ. The accumulated evidence on lung cancer and environmental tobacco smoke. BMJ 1997; 315: 980-988. 3. Bero LA, Glantz SA, Rennie D. Publication bias and public health policy on environmental tobacco smoke. JAMA 1994; 272: 133-136. 4. Wells AJ. Lung cancer and passive smoking at work. Am J Public Health 1998; 88: 1025-1029. 5. Repace JL, Jinot J, Bayard S, Emmons K, and Hammond SK. Air nicotine and saliva cotinine as indicators of passive smoking exposure and risk. Risk Analysis 1998; 18: 71-83. We have no competeing interests.
Mooting the "Publication Bias" Issue 24 February 2000
Leroy J Pletten, Medicine & Law History Substance Abuse Issues Counselor & Lecturer The Crime Prevention Group, Sterling Heights, Michigan Send response to journal: Re: Mooting the "Publication Bias" Issue	The above writers evidencing under-reporting of toxic tobacco smoke (TTS) risks are correct. There is no publication bias over-estimating it. On the contrary, any one study inherently underestimates it, not reciting anew the total number of casualties from all TTS effects—a "holocaust," the term assigned at the time of the then British "annual death toll of some 27,500." 1 Avoidance of under-reporting TTS risks necessarily requires consideration of the paucity of genuine nonsmokers. "It has rightly been observed that, if a 'non-smoker' is strictly construed as one who has never had any contact with tobacco-smoke, non-smokers in any . . . urban society are virtually non-existent. Not only, in such societies, is practically everyone exposed to passive inhalation of tobacco smoke, but a very considerable number of 'non-smokers' have once tried . . . smoking before renouncing the practice." 2 There is no publication bias in a legal sense against negative-result TTS studies, if any, which I agree with Dr. Glantz do not exist. But assuming them arguendo, editors rejecting them would not be committing "publication bias." Instead they would be demonstrating a prudent respect for the rule of law, requiring not publishing words that could foreseeably aid and abet and be accessory to foreseeable wrongful deaths. 3 Dr. Hirschhorn raised the issue of banning cigarettes, a concept that can lead to mooting the "publication bias" issue. As the TTS hazard was already well and sufficiently established in the nineteenth century, jurisdictions such as Tennessee and Michigan did bans of the type he implies. 4 Cigarette manufacture and sales bans were adopted pursuant to the common law ban on polluting others and their property, pursuant to the right to, e.g., "fresh and pure air." That right had been developing since at least 1306 with known judicial recognition (case law of record) since at least the early and mid 1700's. 5 Cigarette manufacture and sales bans preclude TTS exposure, i.e., institutionalize the common law right to "fresh and pure air." In consideration of smokers' addiction (e.g., impaired impulse and ethical controls), such institutionalizing actions are the only effective means of attaining that right for all people (born and unborn), at all times everywhere (homes, workplaces, travel facilities, shopping areas, etc.) as per above writers' references to TTS exposures in such locations, among others. Conclusion: No over-reporting "publication bias" occurs now, only under-reporting; and, when (if) pursuant to law, TTS-caused deaths cease (are at zero level), even the issue will foreseeably cease and become moot. 1 Royal College of Physicians of London, Smoking and Health Now (London: Pitman Med and Sci Pub Co, 1971), 9. 2 Larson PS, Haag HB, Silvette, H. Measurement of Tobacco Smoking. Med Times 1960; 88; 417-429. 3 Paladin Enterprises, Inc v Rice, 128 F3d 233 (CA 4, 1997) cert den 523 US 1074; 118 S Ct 1515; 140 L Ed 2d 668 (1998) (civil case); Julius Streicher, The Nurnberg Trial, 6 FRD 69, 161-163 (1946) (criminal case) 4 Pletten LJ. Alternative Models for Controlling Smoking among Adolescents. Am J Pub Health 1997; 87: 869-870 5 Tenant v Goldwin, 2 Ld Raym 1089; 1 Salk 360; 6 Mod 311 (1706); Rex v White and Ward, 1 Burr 333 (KB, 1757); Hay v Cohoes County, 2 NY 159; 51 Am Dec 279 (1849); Ross v Butler, 19 NJ Eq 294; 97 Am Dec 654 (1868); Washburn v Gilman, 64 Me 163; 18 Am Rep 246 (1873); State v Heidenhain, 42 La Ann 483; 7 So 621; 21 Am St Rep 388 (1890); McMorran v Fitzgerald, 106 Mich 649; 64 NW 569; 58 Am St Rep 511 (Mich, 1895); Shimp v New Jersey Bell Telephone Co, 145 N J Super 516; 368 A2d 408 (1976)
Unfair scales for a visual test of publication bias 4 March 2000
Christopher Cates, General Practitioner Manor View Practice, Bushey Health Centre, London Road, Bushey Send response to journal: Re: Unfair scales for a visual test of publication bias	EDITOR - Funnel plots can be useful to detect publication and related bias. The funnel plot in Copas and Shi's review of the epidemiological studies of passive smoking and lung cancer is, however, biased. In the absence of publication bias the plot can be assumed to be symmetrical only if relative risks are plotted on a logarithmic scale. The scale used by Copas and Shi is not logarithmic and will give the visual impression of publication bias even when there is none. Studies indicating that exposure to passive smoking increases the risk of lung cancer will spread out on the graph because the relative risk may range from 1.0 to infinity; in contrast studies showing a reduction in risk will be compressed in the range of 1.0 to zero. Visual interpretation of the data is therefore not possible using the scale presented.
Over-reliance on p values 25 March 2000
Adam Jacobs, Director Dianthus Medical Limited Send response to journal: Re: Over-reliance on p values	I am not completely convinced by Copas and Shi's assertion that 'at least some publication bias is needed to explain the trend we found'. The evidence for this statement seems to be that the correlation between study size and study result was significant at p = 0.012. This, while conventionally defined as significant, is not so small that it could not arise by chance. Reliance on this p value becomes especially dubious if this significant association was discovered before deciding to re-analyse the studies on passive smoking and lung cancer. I wonder if the authors decided to re-analyse this particular meta-analysis and then found a significant association, or whether they looked at many meta-analyses and chose to re-analyse the one with the smallest p value?
Few unpublished studies on passive smoking 6 April 2000
Lisa Bero, Associate Professor University of California San Francisco Send response to journal: Re: Few unpublished studies on passive smoking	Any effect of publication bias on meta-analyses of studies examining the association of passive smoking and lung cancer will occur only if unpublished negative studies exist. As Copas and Shi(1) point out, we conducted a study showing that about 20% of peer-reviewed articles examining the health effects of passive smoking had statistically nonsignificant results, suggesting that negative data on passive smoking are published (2). In a more recent study, we conducted in-depth interviews with 78 investigators conducting research on passive smoking (including tobacco industry funded researchers) (3). This study allowed us to determine the extent of unpublished results on the health effects of passive smoking. Of the 50 studies conducted in humans for which data analysis had been completed, 11 were unpublished, 2 of which had statistically nonsignificant results. Although the number of unpublished negative studies was small, there was a publication delay for passive smoking studies with nonsignificant results. 1. Copas, JB, Shi, JQ. Reanalysis of epidemiological evidence on lung cancer and passive smoking. BMJ 2000; 320: 417-418. 2. Bero, L, Glantz, S, Rennie, D. Publication bias and public health policy on passive smoking. JAMA 1994; 272: 133-136. 3. Misakian, AL, Bero, L. Publication bias and research on passive smoking: Comparison of published and unpublished studies. JAMA 1998; 280: 250-253.