I VDI Itcoorts 888 ASSOCIATION OF GERMAN ENG~NEERS COMMISSION ON AIR POLLUTION OF THE VDI AND DIN CARCINOGENIC SUBSTANCES IN THE ENVIRONMENT Oriein Measurement Risk Minimization Mannheim Collocuium. 23 to 25 Avril 1991 VDIVERLAG Ln rl-_i ~10 1--~ BATCo document for Legal Services : Health Canada 19 October 1999 2 Passive smokinp- - Evaluation of the cvidcmiolop-ical findings K.-H. Mckel. Bremen Summarv The biologically plausible hypothesis that environmental tobacco smoke causes lung cancer appears to be supported by the evidence available from epidemiological studies. Both case control studies and cohort studies are in agreement in indicating a trend towards increased relative risks which. in a meta-analysis based on a total of 14 studies. reveal a 35% higher risk of passive smoking due to the partner who is a smoker. The criticism expressed in these studies will be discussed and will be appraised in the overall context. In addition. the results of a study of the methods and initial results of a case control study will be reported. I . - Unroducdon Tobacco smoke consists of very many individual chemical substances. a large number of which are toxic or have been shown to be carcinogenic both in animals and in humans /3/. While the human carcinogenicity of active cigarette smoking is no longer seriously disputed. the health effects of so-called "passive smoking", more appropriately referred to as Environmental Tobacco Smoke (ETS), are the subject of vigorous scientific controversy. It is undisputed that many of the active carcinogens are present in the secondary smoke stream in very much higher concentrations than in the main smoke stream inhaled by the active smoker. ETS, the mixture inhaled by the passive smoker, is the sum of the two individual components secondarv-stream smoke and exhaled main-stream smoke of the smoker or smokers present in the room. Whether the individual carcinogenic substances of this mixture actually effectively reach the "target -object", the passive smoker, there is still controversy about gas phase and particle phase /7/. Animal experimenti considered in isolation have been unable in the past, and presumably will continue to be unable in the future, to provide convincing results, since -these experimenEs have unusually high pollutant concentrations compared with human situations. The aim of this article is to describe the possible contribution, to date and in the future, of epidemiology to the question of the carcinogenicity of passive smoking, against the background of the situations presented above. The methodological problems encountered and their relevance will be discussed in particular-, possible solutions for overcoming such methodological problems will be indicated and the initial results of an ongoing study will be presented. Although active smoke ~*otent carcinogen also for areas other than the lung (cf. /3/), we shall restrict ourselves to the effect of lung cancer. This is because of the eitremely high potency of active smoking in connection with pulmonary carcinoma on the one hand and the number of epidemiological investigations on the subject to date on the other hand. 00 CD (JI BATCo document for Legal Services : Health Canada 19 October 1999 3 2. Available epidemiological evidence In a review which app=ed recently, Saracci and Riboli /10/ present I I epidemiological case control studies and 3 cohort studies which are concerned with the relationship between lung cancer and exposure to passive smoking. Owing to the small numbers of cases in the individual studies (cf. Table 1), the confidence intervals for the relative risk of passive smoking are greater than one for none of the case control studies presented and only for one cohort study. I corresponding to statistically significant proof of a carcinogenic effect. However, in a synopsis of all studies, a so-called metaanalysis, Saracci and Riboli obtain an estimate for the relative risk of 1.35 with a 95% confidence interval from 1.2 to 1.52. As the smallest common denominator of the study, passive smoking is understood as exposure to an actively smoking partner. A relative risk of 1.35 means that a nonsmoker exposed to the smoking spouse has a 35% higher risk of contracting lung cancer than an unexposed nonsmoker. Although a number of further epidemiological studies on the stated problem have also appeared after 1987, they do not alter anything with regard to the evidence accumulated so far, owing to the relatively small number of cases, assuming that the fundamental validity of the studies is not in doubt. It is interesting to note that the relative risk of 1.27 determined from the case control studies presented does not differ to a relevant extent from that of the cohort studies (1.44), which contradicts the frequent presumption that people ~uffering-kom a disease (in this case from lung cancer) tend to attribute their condition to an exogenic factor, namely the partner's smoking. Nevertheless, the crux of epidemiological evidence presented to date is clear from the above formulations. Passive smoking is reduced to the presence of a smoking partner. However, it is known that people can also be exposed to tobacco smoke in places other than * in their private life. A much more serious problem, however, is that a carcinogenic effect can be convincingly proved, if at all, only in the so-called "complete nonsmokers', and one is evidently dependent on the information provided by the pdrsons questioned. This Sises the problem of "distortions', which we shall. consider Wow.. AW BATCo document for Legal Services : Health Canada 19 October 1999 4 Table 1. Epideitiolwicz.* studies on the relationship between lup(] cancer and passive smokinq Case control studies Author and year Covitry Lung cancer controls Relative 951 confi- cases risk dence Exposurell Exposure interval Yes/No Yes/No Chan ; rung, Ron; Kong 'w 34/550 66/73 0.75 0.45-1.31 1982 Correa et al., U.S.A. w 14/ 8 61/72 2.03 0.91-5.08 1983 11 2/ 6 26/154 2.29 .30-17.33 is Trichopoulos Greece w 38/24 81/109 2.11 1.17-3.78 et al., 1983 Buffler et al., U.S.A. V 3311 a 164/32 0.80 0.32-1.99 1984 Kabat & Wynder, U.S V 13/11 15/10 0.79 0.25-2.48 1994 . it S/ 7 S/ 7 ' 1.00 0.20-5.06 Garfinkel et al., U.S.A. w . 91/43 254/148 1.23 0.81-1.86 1985 . Akiba et al., '~M!*_ ' ' " if" 73/21 188/8-2 1.48 ' 0.87-2.52 3/16 91101 2.45 .45-13.45 Lee at al., England W 22/10 45/21 . 1.03 ' 0.41-2.58 8/ 7 14/16 1.30 0.37-4.54 Zoo at at., 90M. Kong W 51,135 6617D 1.54 0.89-2.67 1988. Pershaqen Sieden V 33/34 150/197 1.27 0.75-2.18 et al., 1986 .. . . . :.,. Humble et al., ~ U.Sj. * 1 15/ 5 * -91/71 2.16 0.84-5.52 1987 TOW 440/295 1291/1197 1.27 1.06-1.52 (according to Saracc, . Riboli (1989)) Exposures def ined as living together with a smoking partner CD 00 CD C) BATCo document for Legal Services : Health Canada 19 October 1999 5 Table I (continuedl. Epideiiiological studies on relationship between lunq cancer and oassive siLokinQ [cohort studies I I I jAuthor and year Country Lung cancer Controls Relative 951 confi- cases risk dence Exposurell Exposure interval Yes No Yes No lGarfinkel, 1981 U.S.A. W 88 65 127164 :49422 1.18 0.90-1.54 lGillis et al Scotland W 6 2 1388 521 1.00 0.20-4.91 11984 a 4 2 306 515 3.25 .60-17.65 1 lairavama Jaoan W 146 17 63287 21858 1.63 1.25-2.11 let ai 1984 1 7 57 1003 19222 2.25 1.04-4.86 1 lTotal 251 163 193148 91538 1.44 1.20-1-72 ImetaanalySis for case control 1.35 1.20-1.53 Istudies and cohort studies l(accordino to Saracci & Riboli 1(198911 Exposures defined as living together with a sitokinq Dartner Possible s urces: of distortions Various studies have shown that a number of people who regard themselves as complete nonsmokers have in fact actually smoked at some point in their lives. Since empirical findings indicate /61 that smokers marry smokers rather than nonsmokers, it follows that, among s - mokers. incorrectly classified as complete nonsmokers (with a substantially higher risk ofj~mg cancer than tme complete nonsmokers.). a hip-her proportion answer i'n the affirmative with regard to exposure to passive smoking. The effect of this so- called misclassification is shown as a hy o etical example in Figure 1. 'P th It assumed that 2.5 % of all self-reported complete nonsmokers are in actuality s ' mokers and Oat.661* ofall true complete nonsmokers but 82% of all alleged complete nonsmokers- answer- in the affirmative with regard to,exposure to passive smoiing, Tihese hypothetical data are based on the book by Lee /6/, whose work was supported by the Tobacco Advisory Council, a sub-organization of the British cigarette industry. If a rate of newly contracted' disease of -5 ' 100.000 is assumed for complete nonsrqokers. i-d es (Z) re2a I s of whe&r these are pedpie exposdto passi've smoking or unexposed people, the alleged relative risk is 1.26the risk for active smoking being 20 times higher (rate of newly contracted disease 100 : C;-> 100,000 for active smokers). This means that -With the assurped. prd~rs of rpagnitude and C) BATCo document for Legal Services : Health Canada 19 October 1999 6 without an actual effw of passive smoking, an apparent increase in risk by 25% is due to the assumed misclassifica[ion and the "aggregation behaviour" of smokers. It should be pointed out here that a basic rate of newly contracted disease other than 5 : 100,000 does not change the artificial finding mentioned. ~hus, in an epidemiological study. a relative risk of i 'es 1.26 would be endrely plausible on the basis of the source of i istortion; d cribed - if Lhis alone were valid - wiliout there being any risk of passive smoking in reality. Thus, Lee 16/, too, concludes that the higher risks of passive smoking observed in epidemiological studies to date are very probably due to such distortions. The fact that Wald et at. /I I/ obtained controversial results with slightly different assumptions is frustrating: a 30% increase in the risk of passive smoking may be entirely probable. The effect of the various parameters on the apparent increase in risk in the hypothetical example presented above are shown in Table 2. 0 0 U4 BATCo document for Legal Services: Health Canada 19 October 1999 0 0 CL 0 0 Ga PASSIVE, SMOKING 81100,000 LUNG CANCER CASES SURE POSITIVE EXPO TRUE N - 643 500 N - 32 175 NEVER SMOKERS N - 975.000 I'ASSIVr SMOKING, 5.100.0oo I"XI'DS111W. NFGATIVI LI fN(o CANrr-.It (-ASIj.S U - 331 Soo N 16 5 15 FICTITIOUS COHORT OF NEVER SMOKERS OWN INFORMA17ON N - I.ooo.ooo M PASSIVE SMOKING 10C)SIO (I. a 00 2.5% a2% XPOSURB POSITIVE LUNG CANCER CASES N - 20.500 20 5 EVER SMOKERS 0 N - 25.000 PASSIVE SMOKING1001100,000 EXPOSURE NECATIVE -~ LUNG CANCER Q N - 4 $00 N - 4 5 90 0 RELATIVE RISK FOR EXPOSURE TO PASSIVE SMOKING 32.17S 0 20.5 4 5 0 il43.300 6 2o.500 1.26 rilk 331.500 6 4.500 0 Cr (D Figure 1. Effect Of smokers as nonsmokers on the e0valumisclassification complete ation of the passive smoking risk by the -L 11LIGband (hypothetical exi-imple) w cc (D H,iSHOGS 7 66% Passive smokin(g 5:100,000 Lung cancer cases exposure positiveN 32.175 N = 643,500 True complete nonsmokers N = 975,000 P a s s i v e smoking 5:100,000 Lung cancer cases 97.5% e x p o s u r e negative N = 16.575 34% N 331,500 Fictitious cohort of "complete nonsmokers" according to own information N = 1,000,000 Passive smoking 100:100,000 Lungcancer cases 2.5% 82% exposure positiveN 20.5 N = 20,500 Never smoked N = 25,000 Passive smoking 100:100,000 Lung cancer cases 18% exposure negativeN 4.5 N = 4,600 32,175 20.5 16,575 - 4.5 Relative "risk" for exposure to passive smoking 1.20 643,500 20.300 331,500 - 4,500 Figure I - Effect of misclassification of smokers as complete nonsmokers on the evaluation of the passive smoking risk by the husband (hypothetical example) UT CD CD CD CD (il BATCo document for Legal Services : Health Canada 19 October 1999 8 Table 2. Apparent increase in risk for different assumptions about the misclassificati0fl rate 7r. relative risk of smokine RR and exoosure of' smokers to 1)assive smokinz a (complete nonsmokers exposed to passi-e smoking 68%) RR\ci 70% 80% 82% r = 2.5 % 2 + 0. 4 % 1.5% +1.7% 1 10 + 3.3 % +12.1% +14.0% 1 1 1 20 + 5. 8 5-c +22.6% +26.5% 1 7r = 5 % 2 + 0. 8 5-; + 2.9% + 3.3% 1 1 1 0 1 10 +5.5~r + ~2_ 1. 2 % +24.8% 1 1 .1 20 + 8.87c +36.2% +43.0% The remarkable feature here is the fact that the apparent increase in risk depends in a relevant manner on the assumed relative risk of active smoking. This counteracts the criticism of Lee and others: thus, it does not appear ver , y plausible that smokers misclassified as complete nonsmokers are grouped together with the very heavy smokers. for whom relative r-isks greater than 10 must be assumed. Instead. what is plausible is that they belong to the group consisting of those who smoke rarel.v or to a small extent and for whom we have to assume substantialiv lower risks on the basis of all epidemiological evidence. Another damper for critics who favour the above-mentioned example is the fact that the epidemiological studies since performed in connection with the risk of cardiovascular diseases also indicate relevant increases in risk (cf. /11). If the observed relationship with lung cancer were due only to the misclassification described. the estimated lisk of passive smoking would have to be an order of magniaide lower (with the same pathogenetic chain of effects) owing to the very much lower risk of active smoking for cardiovascular diseases (about 2:1 compared with 10: 1 for lung cancer), which is also evident from Table 2. Over and above this, it must not be forgotten that. apart from the source of distortions just described, there may be a number of other possible sources of distortions, some of which may have effects in the opposite direction (cf. Table 3). ui CD CD c'O (J-J CD Lr, BATCo document for Legal Services : Health Canada 19 October 1999 9 Table 3. Effects of further distortions on a derived relative risk (RR) of passive smoking I Cause Effect on RR" I "Non-exposed people" are in reality exposed (occurrence of ETS everywhere) Nondifferential misclassification of exposure Cases underestimate/controls exaggerate the exposure Cases exaggerate/controls underestimate the exposure t Lung cancer cases not involving smoking are not recognized as such 41 means that the relative risk is underestimated or overestimated. Whether and how the sources of distortions described affect epidemiological findings cannot be established in theory. In empirical terms, it appears that more knowledge can be gained only through an improvement of the epidemiological instrument. For this purpose, it is particularly important to arrive at a valid quantitative estimate of the exposure to passive smoking taking into account sources of exposure other than the smoking spouse. 4. Results of a study of methods Developing an instrument for a more valid estimate of the exposure was the task of an international study performed by the International Agency for Research on Cancer (IARC), U-I in which the Bremer Institut fu-r Pr5ventionsforschung und Sozialmedizin (BIPS) participated ~:D CZD as one of 13 study centres in 10 countries. In 1986, 100 female nonsmokers in Bremen were Ca CD BATCo document for Legal Services : Health Canada 19 October 1999 10 questioned about their exposure io passive smoking over the previous 7 days by means of a standardized questionnaire, and the excretion of cotinine (a specific metabolite of nicotine) in the urine was detcrmined. The results of both the national and the international study have now been published /8/, 191. Of the 100 women between 39 and 67 years who were questioned, 49 were in employment and 51 did not have gainful employment. Figure 2 shows that a relevant proportion of the exposure to passive smoking as a function of time was due to the smoking husband, this being based on the evidence of the women themselves. Women in employ- Women not in ment (N = 49) employment (N = 51) Husband Other people at home Other private occasions Other public occasions Workplace Figure 2. Proportion of exposure to passive smoking according to source of exposure (in %) However, it is also clearly evident that other sources of exposure, for example the workplace or other public occasions, make a relevant contribution to the exposure to passive smoking which was subject to the questionnaire. As shown in Table 4, there is very good correlation between the "objective" marker of exposure to passive smoking, the cotinine in the urine and the total exposure determined from the questionnaire- (-71 CD CD 00 CD Ln BATCo document for Legal Services: Health Canada 19 October 1999 11 Table 4. Correlation between relative cotinine level and exposures stated in questionnaire (Spearman's rank correlation coefficient). Weighted exposure" as a function of time, based on subject's information Without With" consideration Source of the concentration Total exposure 0.71 0.61*** Husband 0.59*** 0.59*** Other people 0.43*** 0.42*** at home Workplace 0.15 0*10 0 1 1 I Vehicles 0.26* 0.37*** Other occasions 0.17 0.22, Taking into account the half-life of cotinine (32 hours) Volume and amount -P < 5% *- P < 1% P < 0.1% Based on the individual sources of exposure, it is true in comparable terms. although the workplace does not reach the limit of statistical significance. In the international study /91, however, a statistically significant relationship between the subjects' own information and the codnine level was also found for this source of exposure. It therefore must be assumed that the work-place is a relevant source of exposure for passive smoking, even though the determination appears to involve considerable problems, owing to, among other things, the highly variable room sizes of the workplaces. Overall, however, these findings are likely to encourage the use in future studies of an improved instrument of investigation which leads to a more valid estimate of the exposure of nonsmokers to passive smoking. An instrument derived from the knowledge gained in this study of methods was developed in cooperation with the IARC and is currently being used as part of an international study. Initial results of the German part of the study which is intended to contribute towards clarifying workplace-related lung carcinogens are presented U-I below. 00 CD CD BATCo document for Legal Services: Health Canada 19 October 1999 12 5. initial results Of an ongoing study on (tie relationship between lung cancer and passiv.91 smokin In a case control study supported by the EIMFr (support code 01 14K 546/8, cf. /5/), the relationship between lung cancer and risks at the workplace is to be investigated in detail. This study is based on 1.000 lung cancer patients and the same number of controls from the Bremen and Frankfurt regions. All people are asked to answer detailed questions about their professional history and smoking habits. People who stated that they had never smoked regularly for a period of more than 6 months (complete nonsmokers and occasional smokers) were asked about their life-long exposure to ETS. Passive smoking in childhood, through the spouse, through other people in the house, in vehicles and on other occasions are considered as potential sources of exposure. In the middle of the study, relevant information is available for 33 cases and 115 controls who regard themselves as complete nonsmokers and occasional smokers. Without being able to go into details of the design, the initial results of this study will be presented below. Owing to the small number of cases, the results presented here cannot and should not be regarded as a conclusive finding but rather as a first interim report on a study which is attempting to solve the problems described in Section 2. In line with the review character of this work, it is not possible to assume the function of a basic epidemiological publication, so that the interested reader will inevitably miss important pieces of information. However, since German studies in this subject have not been available at all to date and the number of cases reached are in the region of the studies shown in Table 1, it appears justified to report initial results here. The group of complete nonsmokers and occasional smokers accounts for 2 % of all cases and 16% of all controls for men. For the women, the corresponding figures are 29% and 49%. If living with a partner who is a smoker is the sole factor considered in exposure to passive smoking. a relative risk of 2.46 is obtained (cf. Table 5). In order to overcome the difficulties already described, an alternative exposure index which also includes the other exposure sources was determined. The person is considered to have been exposed if, for any source of exposure, he or she specifies exposures in the upper quartil of the common distribution of all subjects questioned. The relative risk obtained from this definition of exposure is 2.64, with a statistical confidence interval of 1. 15 to 6.07 (cf. Table 6). LJ4 BATCo document for Legal Services : Health Canada 19 October 1999 Table 5. Initial resulLs of an epidcmiological case control study on lung cancer~'. Effect of exposure to passive smoking Cases Controls RR" 95% confidence interval Partner smokes (smoked)" W 17 25 yes 2.46 0.94-6.43 m 3 11 W 6 20 no 1.0 Reference m 6 59 Supported by funds from the BMFT, suppori code 01 HK 546 8 1 1 Adjusted for sex, smoking status (complete nonsmoker, occasional smoker) 3) No information for one case Table 6. Inibal results of an epidemiological case control study on lung cancer": Effect of exposure to passive smoking Cases Controls RR" 95% confidence interval Relevant exposure" for any source * 13 20 yes 2.64 1.15-6.07 * 8 26 no w 10 25 1.0 Reference * 2 44 33 115 1 Supported by funds from the BMFT, support code 01 HK 546 8 Adjusted for sex, smoking status (complete nonsmoker, occasional smoker) For any exposure source in the upper quartil of the distribution, exposure sources: childhood, spouse, home, workplace, on other occasions U-1 CD CD co CD CD BATCo document for Legal Services : Health Canada 19 October 1999 14 As is evident from Table 7, this definition of exposure also makes it possible to estimaie the simultaneous effect of several sources of exposure. Table 7. Relative risk as a function of the level of exposure to passive smoking RV P value 1. Relative exposure" Average 2.12 13.2% High 3.43 1.87. 11. Relevant exposure without spouse Present 2.53 4.4% Relevant exposure" I due to spouse Average 1.82 40.1% High 3.49 11.5% 111. Relevant exposure due to spouse Present 2.60 10.1% Relevant exposure" to other sources Average 1.83 27.4% High 3.67 Z.9% Adjusted for sex and smoking status (complete nonsmoker, occasional smoker) Present = for any source in the upper quartil of the distribution Average = Present but not high High = Above the 9M. percentile of the distribution 3) Additional exposure If the two sources "exposure to the spouse" and "other sources of exposure (= totality of all sources without spouse)' are considered, the risk of passive smoking due to one of the two sources can be estimated and the effect of the other sources in each case can be investigated in a dose-dependent manner- Risk gradients which show statistically significant risks for the highest level of the graded feature in each case are obtained. The results obtained to date appear to support the epidemiological evidence presented in Section 1 of this article without being able to provide final certainty, owing to the limited numbers of cases. It should be mentioned at this point that subgroup analyses (separately according to sex or smoking status (complete nonsmokers or occasional smokers)) indicate the same trend without reaching the statistical limit of significance. It would certainly be desirable to validate the results obtained from the questionnaires. However, there are at present no appropriate means for this U, I purpose. CD CD W CD Url BATCo document for Legal Services : Health Canada 19 October 1999 15 6. Discussion It should now be evident that the question of the health effects of passive smoking, in particular of carcinogenicity of the lungs, cannot at present be considered to have been finally clarified scientifically. Certainly, the arguments of the critics of the available epidemiological evidence /6/, /7/ must be taken seriously. However, the considerations in Section 3 should have made it clear that a biased concentration on selected sources of distortions is not suitable for dealing with the overall problem. However, studies of methods (cf. Section 4) indicate that a valid retrospective determination of ETS is possible (cf. M, /8/, /9/). Investigations based on this, such as the study described in Section 5, are capable of ensuring relevant progress in terms of knowledge. They can certainly become more informative through the supportive measure of validity studies. On the basis of knowledge to date, the relevant risk of passsive smoking cannot be ruled out in the overall evaluation. The importance of this statement with regard to health policy may be illustrated by the following consideration: in a representative cross-section of 25 to 69 year olds in the population (The German Cardiovascular Preventive Study (GCP), 1988) /2/, 12.9% of all those questioned regard themselves as complete nonsmokers and at the same time answered "yes" to the question on exposure to passive smoking due to the spouse or at the workplace, If, for the sake of simplicity, exposure to passive smoking is assumed to entail a 35% higher risk, this would mean that 4.3% of all cases of lung cancer are due to exposure to passive smoking from these two sources of exposure in this segment of the population. Based on the complete nonsmokers, this figure was 9.9%. This does not take into account the fact that a cross-sectional questionnaire does not include life-long exposure but deals with only the instantaneous exposure situation. Against this background and on the basis of the consideration that a reduction in passive smoking certainly has absolutely no adverse effects on health and that the danger of active smoking has been adequately established, everyone's efforts should be concentrated on reducing active and passive smoking, regardless of the search for further scientific evidence. The fact that such measures are supported by a large majority of the population is demonstrated by a representative investigation by the Bremer Institut fitir Mventionsfbrscl~ung und Sozialmedizin (BIPS) /4/ in Bremen and Nordrhein-Westfalen, according to which 62% of all those questioned agree with a general ban on tobacco advertising and 59% are in favour of making access to cigarette machines more difficult. 7- Reference Co (~D Lri BATCo document for Legal Services : Health Canada 19 October 1999