SUPPLEMENTARY SUBMISSION FROM ASH

Early reports from Ireland are encouraging.

31st May 2004 - the Office of Tobacco Control in Ireland published its first report on compliance for one month after the smoke-free law came in (covering the period 29th March when the ban was introduced to 30th April 2004). The report comprises of data from three sources: the National Tobacco Control Inspection programme, the smoke-free workplace compliance line and market research on public attitudes and behaviours.

The report found that 97% of premises inspected under the smoke-free workplace legislation were compliant with the law (i.e. no one smoking and no evidence of smoking in contravention of the law) and indicated that levels of visits to pubs and restaurants remained constant, with one in five smokers choosing not to smoke at all when out socialising.

Prior to the introduction of the smoke free workplace law, 91% of the population stated they would be either more likely or just as likely to visit a restaurant to eat. Since the law was introduced, this figure is 92%.

The rate of smokers visiting pubs has remained steady at 74% since the legislation was introduced. The frequency of non-smokers visiting pubs has increased from 67% to 70%.

The full six page report is available on their website www.otc.ie under Publications.

Progress on smoke-free public places is being made elsewhere in Europe.

Tuesday 1 June 2004 - legislation in Norway to introduce smoke free public places is implemented.

May 12, 2004 - the Swedish parliament votes to ban smoking in bars and restaurants, starting on June 1, 2005.

Smoke free New York - one year review shows success.

The Smoke-Free Air Act took effect on March 30th 2003. On May 12, 2004 the New York City Department of Health and Mental Hygiene (DOHMH) announced an 11% decline in the number of smokers in New York City over the previous year - the fastest drop in smoking rates ever recorded nationally. This drop represented 100,000 fewer New Yorkers smoking in 2003 compared with 2002. Those who continued to smoke were also smoking less. The DOHMH attributed the fall in smoking rates to its program of tobacco control, including the ban on smoking in public places.

Concerns had been expressed about the potential economic impacts on business of a ban. Data from the DOHMH one year review showed that:

• Business tax receipts in restaurants and bars were up 8.7%
• Employment in restaurants and bars had increased by 10, 600 jobs (about 2,800 seasonally - adjusted jobs) since the law's enactment
• 97% or restaurants and bars were smoke-free
• New Yorkers overwhelmingly supported the law
• Air quality in bars and restaurants had improved dramatically
• Levels of cotinine, a by product of the body's metabolising tobacco smoke, decreased by 85% in non-smoking workers in bars and restaurants
• 150,000 fewer New Yorkers were exposed to second-hand smoke at work

SUPPLEMENTARY SUBMISSION FROM TOBACCO MANUFACTURERS’ ASSOCIATION (Part 1)

Introduction

On 8 th June 2004, the Tobacco Manufacturers’ Association (TMA) gave oral evidence to the Health Committee on the Prohibition of Smoking in Regulated Areas (Scotland) Bill. During the course of those proceedings the TMA undertook to provide the Committee with certain further information, hence this supplementary written evidence.

The whole debate about smoking in work and public places revolves around and is founded on the assertion that ETS is harmful to the health of the non-smoker. In particular, the Committee asked for further information: on the epidemiological studies which have been undertaken concerning ETS; about the balance of the findings of those studies; and effectively why the TMA did not believe that they justified or supported the popular perception that ETS causes serious diseases in non-smokers. Additionally, the TMA offered to provide further information on legal cases brought against employers.

In order to provide a comprehensive answer to those questions, and to enable the Committee to reach its own conclusions on the available evidence on an informed basis, it is not sufficient simply to list the ETS studies that have been published. The studies need to be put into a proper context, their design and terminology explained and a guide provided as to how their findings should be interpreted.

A chronology

In the US Surgeon General’s reports of 1972 and 1975, initial speculations were raised about the possible consequences of exposure to environmental tobacco smoke (ETS). The US Surgeon General’s 1979 report noted several adverse outcomes that appeared to have an association with ETS; but also that there was only a limited amount of systematic information available regarding the health effects of ETS. The Surgeon General’s 1982 report raised the concern that ETS might cause lung cancer. Following that report a number of epidemiological investigations were published which claimed to show a relationship between ETS and lung cancer.

Then in 1986, the US Surgeon General’s report, as well as reviews by the National Research Council and National Academy of Science (sponsored by the US Environmental Protection Agency (EPA)), concluded that ETS caused lung cancer and claimed an increase in risk of 30%, with the latter two reviews also associating ETS exposure with adverse respiratory outcomes in young children.

However, a review published in 1986 by the International Agency for Research on Cancer (IARC) of the World Health Organisation came to different conclusions. It did not produce estimates of risk but concluded that available studies:

“had to contend with substantial difficulties in determination of passive exposure to tobacco smoke and to other possible risk factors. The resulting errors could arguably have artefactually depressed or raised estimates of risk, and, as a consequence, each is compatible either with an increase or with an absence of risk.”¹

Nonetheless, in June 1989, the US EPA issued a public notice that stated categorically that ETS “is a known cause of lung cancer”. However, the EPA did not provide an analysis of the data on which it had based its conclusion. It was pressed to do so but did not produce its analysis and risk assessment until 1992² . This took the form of a review of selected published studies. It was subjected to devastating criticism, not least by members of the US Congressional Research Service appearing before a Committee of the US Senate, who said:

“The EPA study analysed and summarised 30 studies of passive smoking lung cancer effects. Critics have questioned how a passive smoking effect can be discerned from a group of 30 studies of which 6 found a statistically significant (but small) effect, 24 found no statistically significant effect, and 6 of the 24 found a passive smoking effect opposite to the expected relationship.”

“… our evaluation was that the statistical evidence does not appear to support a conclusion that there are substantial health effects of passive smoking.”³

The report was later also challenged in the courts⁴ where the EPA was found to have knowingly, wilfully and aggressively disseminated false information with far reaching regulatory implications in the US and worldwide. Judge Osteen found that the EPA had :

“changed its methodology to find a statistically significant association . . .In conducting the ETS Risk Assessment, EPA disregarded information and made findings on selective information; did not disseminate significant epidemiologic information; deviated from its Risk Assessment Guidelines; failed to disclose important findings and reasoning; and left significant questions without answers … Gathering all relevant information, researching, and disseminating findings were subordinate to EPA’s demonstrating ETS a Group A carcinogen.”

Yet to this day, despite that judgement which vacated (annulled) the report after ‘forensic’ investigation of the EPA’s review and process, the report is used as a ‘gold standard’ by the authorities. It is the ultimate foundation of the estimates made by UK authorities of UK deaths resulting from exposure to ETS. The report and its methods have subsequently been used as a model for other reports by the Californian EPA⁵ , the National Health & Medical Research Council of Australia⁶ , and the UK’s Scientific Committee on Tobacco and Health (SCOTH)⁷ . In 1998, the US National Toxicology Program accepted the EPA 1992 report and its twin from California as the basis for listing ETS as a known human carcinogen.

At the time the EPA prepared its 1992 report, there were only around 30 published studies seeking to determine lung cancer risks associated with exposure to ETS. There have now been well over 100 studies and reviews that have been published; a great many more are thought to have been undertaken but not been published.

The significance of publicationand publication bias

Whilst, therefore, the total number of studies and reviews that have been undertaken is likely to be very much larger, only those that have been published form part of the accepted compendium of information on ETS. This means that every party has access to the same information upon which they may make their own judgements. Unpublished studies are not concealed or used; publication is the determining factor. Such differences of opinion as do exist about ETS studies and reviews arise out of the critical examination and analysis to which they may then be subjected, and the interpretations and judgements which may then be made as to their data and findings.

Given this significance of publication, it is well recognised that what epidemiologists term ‘publication bias’ may arise:

“Publication bias occurs in two quite separate forms. Studies with positive results are more likely to be submitted for publication and more likely to be accepted; and significant findings, such as an association with a particular occupation or exposure, are often given prominence by the authors, particularly in case-control studies [explained at paragraph 21 et seq.], while other exposures that were analysed but were not significant may not be mentioned at all. Both types of bias tend systematically to exaggerate associations in the published literature.”⁸

“Quite different conclusions might be drawn from a review of all published and unpublished studies.”⁹

“The presence of even a modest degree of publication bias can lead to a substantial increase in the estimated risk.”¹⁰

“The result is a biased understanding of the differences and similarities in the disease patterns of populations and an exaggerated view of the importance of associations between risk factors and disease outcomes.”¹¹

Publication bias is well recognised as existing particularly when a consensus develops among the ‘experts’ themselves – albeit that consensus opinion may not be correct . Once a large number of people believe something, it can be difficult and costly to argue to the contrary. For example, academics and researchers who then go against the grain can find it difficult to achieve publication of their opinions and research, or struggle to find posts or research funds.

An illustration of the reception that can be given to the publication of views which do not conform to the accepted wisdom – and which thereby illustrate the strong force that publication bias represents was provided by the reaction to the publication by the British Medical Journal in May 2003 of a major new ETS study¹² , in respect of which the BMJ carried the front-page headline, “Passive smoking may not kill”. This prospective study measured the relationship between ETS, as estimated by smoking in spouses, and long-term mortality from tobacco related disease and was conducted on over 100,000 Californian adults between 1960 and 1998. The conclusions of the study stated:

“The results do not support a causal relation between environmental tobacco smoke and tobacco related mortality although they do not rule out a small effect. The association between exposure to environmental tobacco smoke and coronary heart disease and lung cancer may be considerably weaker than generally believed.”

The publication of the study by the BMJ gave rise to a violent storm of criticism from the medical community. In responding, the editor of the BMJ was minded to comment -

“Of course the study we published has flaws – all papers do – but it also has considerable strengths: long follow-up, large sample size, and more complete follow up than many such studies. It’s too easy to dismiss studies like this as fatally flawed with the implication that the study means nothing . . . I found it disturbing that so many people and organisations referred to flaws in the study without specifying what they were. Indeed, this debate was much more remarkable for its passion than its precision.”

“We must be interested in whether passive smoking kills, and the question has not been definitively answered. It’s a hard question, and our methods are inadequate.”¹³

The heterogeneity of studies and reviews

Whilst it is now common for the statistical findings of ETS epidemiological studies to be expressed in a common manner – in terms of a reported estimated relative risk [explained at paragraph 32 et seq] there is no accepted common study design and “few epidemiological studies satisfy the stringent methodological criteria that should ideally be applied.”¹⁴ . Thus individual studies and reviews exhibit wide variations in design, methodology, data collection, country, population and study size. It is therefore not surprising that findings show little consistency. This makes interpretations and comparisons both difficult and contentious. This is particularly so as even where a positive association between ETS and a disease has been reported, it has been of a very low order of risk, It has been of a magnitude that might easily be accounted for by bias or confounding [explained at paragraphs 24 and 28 respectively], or by inadequate adjustment in the study of such bias and confounding. It has also been of a magnitude well below that normally regarded as being significant and appropriate as a guide for public policy.

Meta-analysis

Given the g reat variability of individual studies, in undertaking collective reviews of studies, a weight of evidence approach is frequently used. This involves considering the quality of individual studies, discarding some and including others in making an overall judgement. Inevitably, this approach involves a great many subjective judgements about the available studies.

Meta-analysis involves the quantitative synthesis of the results of separate studies, to provide a summary of the pooled results. However, for this to be a valid approach, the studies need to be similar and comparable in design and many other respects, otherwise the result is no better than mixing apples with oranges. Such inappropriate mixes may result from pooling studies of widely varying design and methodology; studies from different countries and populations in respect of which there may be significant and varying confounding variables; studies undertaken in significantly different time frames; and from the selective inclusion of studies based on the researcher’s impressions of study quality.

For example, almost all of the ETS studies that have been undertaken have been of populations outside the UK, particularly in the United States and Asia. They are very different populations to the UK in a great many respects. They have been undertaken over a time period since 1981 and there is a marked difference in the findings between those studies conducted before and after 1989. The difficulties of extrapolating data on one population and applying it to another on the basis that one group of people is broadly equivalent to another has been vividly illustrated by the extrapolation of risk scoring methods for coronary heart disease derived from the US Framingham heart study¹⁵ and its application to the UK. The Framingham study played a key role in quantifying risks such as smoking and high cholesterol. The UK researchers compared the Framingham results with the British regional heart study¹⁶ . They found that using Framingham, there was an over-estimation of the risk of non-fatal coronary events of 57%, and also that 84% of British heart deaths occurred in the 93% of men classified as low risk by Framingham criteria. The fact is that substantial variations in coronary heart disease are found between different regions and different ethnic groups, socio-economic status and family history of coronary heart disease.

Nonetheless, in recent years, meta-analysis has been increasingly used to combine evidence from epidemiological ETS studies of quite different design. This can result in a combined relative risk estimate that has narrow confidence limits [explained at paragraph 35 et seq]; it may appear to be precise, but can in fact be an inaccurate estimate of the true association, if any.

Understanding and interpreting the results of ETS epidemiological studies

“In experimental animal research and in some situations in clinical medicine, for example testing the efficacy of a new drug, it is possible to carry out clinical ‘experiments’ comparing groups receiving different treatments. However, in epidemiological research requiring large populations for the evaluation of potentially harmful exposures, alternative approaches are needed. For example, to ‘prove’ that ETS causes cancer or heart disease would require the conduct of long term experiments (randomised controlled trials) involving hundreds of thousands of individuals half of whom would be randomly assigned to long term ETS exposure and the other half assigned to non exposure. But because it is not ethical to expose human subjects to a potentially harmful substance (in this case ETS), the only research approaches possible are those based on observational studies of non-smokers. Either disease rates in individuals exposed to ETS at home or at work are compared with rates in individuals not so exposed (cohort study); or past ETS exposures are compared in cases (those with the disease in question e.g. heart disease or lung cancer), and in those without these conditions (controls) (case control study). There is no certainty in either type of study that the two groups being compared are similar with respect to other relevant variables. Thus there is the possibility that any differences observed between the groups could be due to factors other than the ETS exposure. If such factors also affect the risk of disease, they are referred to as confounding variables. The consequence is that part or all of the observed association between ETS and the disease may be spurious.”¹⁷

A ‘cohort’ study follows a population group through a lengthy time period. It tracks the disease incidence in the cohort, and can assess possible lifestyle factors and calculate their relationship to the disease incidence. Cohort studies are larger and lengthier than case control studies, and hence are more costly. However, they are thought to be somewhat more reliable than case control studies, especially when multiple risk factors are involved.

However, the vast majority of the investigations that have been undertaken into ETS have been case-control studies . These have typically compared the incidence of certain diseases in non-smokers living with smoking spouses, as compared with non-smokers living with non-smokers. For chronic diseases, such investigations need to assess exposure over a period of thirty to forty years. This is usually achieved through questionnaires - obviously relying on the personal recollections of people - of the intensity and duration of exposure to ETS over a lifetime. The uncertainty involved in this form of data collection makes such epidemiology a relatively imprecise tool.

Bias

In statistical terminology, ‘bias’ relates to deviations from the facts arising from such factors as flaws in study design, data collection or analysis. ETS studies are particularly susceptible to many forms of bias. Aside from the comparative unreliability of individuals’ memories – known by epidemiologists as recall bias - questionnaires are often administered not to the actual members of the populations being studied, but to surviving family members, so increasing recall unreliability and introducing or aggravating other possible sources of bias.

Smokers tend to marry smokers and non-smokers non-smokers and a proportion of people are known not to tell the full facts about their present or past smoking habits. Together, these facts are recognised to give rise to substantial misclassification bias.

Also there cannot be certainty about the precise cause of death, given both the difficulty of establishing that fact and also that “inaccuracies in the registered cause of death are recognised, especially with multiple causes”¹⁸ . In any event, death certificates do not record what caused the illness stated on the death certificate.

Publication bias is also possible – that is the likelihood that studies are published only if they produce positive results or results which conform to the accepted wisdom.

Confounding

Studies are also subject to confounding – distortion because there may be an association of disease with factors other than ETS, such as diet, alcohol consumption, socio-economic circumstances, the level of exercise, the history of disease in the family, that happens to correlate with being in a household with a smoker. While some ETS studies have attempted to collect information on some confounding factors, there has generally been an inconsistency and inadequacy of approach. Yet confounding is a most important consideration in ETS studies. Diseases in smokers that have been associated with smoking are well recognised to be multi-factorial. For example, cardiovascular disease has been associated with over 300 different factors.

There are methodological and statistical techniques to adjust for likely confounding and biases, but again they are not applied uniformly in each individual study, nor are they anything other than devices that may not reflect the true situation, and are themselves subject to limitations.

In reality, therefore, ETS epidemiological studies are statistical exercises, the measurements of which have limited credibility in terms of accuracy. That is not to say that they are irrelevant but it is to put them into a proper context. Epidemiology is “a crude and inexact science”¹⁹ ; and “…until we know exactly how cancer is caused and how some factors are able to modify the effects of others, the need to observe imaginatively what happens to various different categories of people will remain.”²⁰

In other words, epidemiological findings are not incontrovertible, objective conclusions; the judgements made about epidemiological data which indicates a low level of risk, are inevitably subjective. And in the case of ETS, “the judgement as to whether the links observed are causal or not remains difficult.”²¹

continued...

¹IARC, 1986: p.308

² Respiratory health effects of passive smoking: lung cancer and other disorders, EPA, Washington DC, 1992

³ Oral statement of Dr Jane Gravelle & Dr Dennis Zimmerman of the Congressional Research service, the Library of Congress, Washington DC, May 11 1994

⁴ Flue-cured Tobacco Stablization Corporation et al v United States Environmental Protection Agency and Carol Browner, District Court for the Middle District of North Carolina before District Judge Osteen, Order and Judgement, 17 July 1998

⁵ Californian EPA 1997

⁶ NHMRC 1998

⁷ SCOTH 1998

⁸ Peto, J, Meta-analysis of epidemiological studies of carcinogenesis, Mechanisms of Carcinogenesis in Risk Identification, ed Vainio H et al, IARC, 1992

⁹The Lancet, April 23, 2004 on the research commissioned by the National Institute for Clinical Excellence into the prescribing of anti-depressants drugs to children; and The Independent, April 23 2004

¹⁰ Copas J, Shi J, BMJ 2000;320: 417-418

¹¹ Bhopal, R.S, Professor of Public Health, University of Edinburgh, Concepts of Epidemiology, p 91, OUP 2002

¹² Enstrom J E & Kabat G C, Environmental tobacco Smoke and tobacco related mortality in a prospective study of Californians 1960-1998, BMJ 2003;326: 1057-1061

¹³ Richard Smith, editor BMJ, BMJ 2003;327:505

¹⁴ Peto, J ( Institute of Cancer Research), Meta-analysis of epidemiological studies of carcinogenesis, in Mechanisms of Carcinogenesis in Risk Identification, p572, IARC 1992

¹⁵ Dawber T R et al. An approach to longitudinal studies in a community: The Framingham Study, Ann. NY Acad. Sci. 1996; 107:539-556

¹⁶ Brindle P et al, Predictive accuracy of the Framingham coronary risk score in British men: prospective cohort study, BMJ 2003; 327:1267-1270

¹⁷ Report on the health effects of environmental tobacco smoke in the workplace. Commissioned by the Health and Safety Authority of Ireland and the Office of Tobacco Control from an independent scientific group, January 2004.

¹⁸ Derek Wanless, Securing Good Health for the Whole Population, Final Report, HM Treasury, February 2004, 5.49

¹⁹ Dr Charles Hennekens, Harvard School of Public Health, New York Times, 1995

²⁰ Doll R & Peto R, The causes of cancer: Quantitative estimates of avoidable risks of concern in the US today. Journal of the National Cancer Institute 1981:66, 5-6: 1191-1308

²¹ Report on the health effects of environmental tobacco smoke in the workplace. Commissioned by the Health and Safety Authority of Ireland and the Office of Tobacco Control from an independent scientific group, January 2003.