Tobacco Documents Online

Papers Environmental tobacco smoke exposure and ischaemic heart disease: an evaluation of the evidence M R Law, J K Morris, NJ Wald Abstract Objectives: To estimate the risk of ischaemic heart disease caused by exposure to environmental tobacco smoke and to explain why the assodated excess risk is almost half that of smoking 20 dgarettes per day when the exposure is only about i% that of smoking. Design: Meta-analysis of all I9 acceptable published studies of risk of ischaemic heart disease in li~elong non-smokers who live with a smoker and in those who live with a non-smoker, five large prospective studies of smoking and ischaemic heart disease, and studies of platelet aggregation and studies of diet according to exposure to tobacco smoke. Result.s: The relative risk of ischaemic heart disease associated with exposure'to en,Aronmental tobacco smoke was 1.30 (950/0 confidence interval 1.22 to 1.38) at age 65. At the same age the estimated relative risk associated with smoking one cigarette per day was similar (1.39 (1.18 to 1.64)), while for 20 per day it was 1.78 (1.31 to 2.44). Two separate analyses indicated that non-smokers who live with smokers eat a diet that places them at a 60/0 higher risk ofischaemic heart disease, so the direct effect of environmental tobacco smoke is to increase risk by 23°/0 (14°/0 to 33%), since 1.30/1.06 = 1.23. Platelet aggregation provides a plausible and quantitatively consistent mechanism for the low dose effect. The increase in platelet aggregation produced experimentally by exposure to environmental tobacco smoke would be expected to have acute effects increasing the risk of ischaemic heart disease by 34%. Conclusion: Breathing other people's smoke is an important and avoidable cause of ischaemic heart disease, increasing a person's risk by a quarter. Introduction Epidemiological studies have shown that the risk of ischaemic heart disease is about 30%0 greater in non-smokers who live with smokers than in those who do no~t'~ It seems implausible that the effect of environmental exposure to tobacco smoke should be so large when the excess risk associated with smoking 20 cigarettes per day is only about 80% at age 65 (the average age of ischaemic heart disease events in the studies).~'= Environmental exposure to tobacco smoke is only about l°~ that of smoking~'~; the risk is nearly half. In this paper we examine the possible explanations for this surprisingly large association. Methods We carried out five sets of analyses using published dam. Firstly, we conducted a meta-analysis of the stud- ies of exposure to environmental tobacco smoke (or passive smoking) and ischaemic heart disease.'~s" We identified relevant studies through Medline (MESH terms: smoking, tobacco smoke poRution), by scanning the reference lists of each study and of review articles, and by discussion with colleagues..~l the studies used spouse's smoking as an objective measure of exposure to environmental tobacco smoke (non-smokers who live with smokers have greater exposure both inside and outside the home~ ~z). We" extracted data on non-Fatal infarction or death ~rom ischaemic heart dis- ease in never smokers according to whether their spouses currently smoked or had never smoked, excluding data on ex-smoker spouses where possible. We calculated the average of the relative risk estimates, adjusted for age and sex, of the studies, each w~ighted by the inverse of its variance (as there was no heteroge- neity). For reasons given below in the discussion, we excluded two studies reported together)s ~ Secondly, to determine the risk of ischaemic heart disease associated with smoking at low doses, we analysed the dose-response relation between smoking and ischaemic heart disease from five cohort studies of men recruited during the 1950s (selected because of their large size).~q~ We analysed the five studies separately. In each the smokers had been divided into three or four categories according to the number of cigarettes smoked. We fitted logistic regression lines, in 10 year age groups, of the risk of ischaemic heart disease (relative to non-smokers) on the adjusted aver- age number of cigarettes smoked per day in each smoking category. The number of cigarettes was adjusted as described previously,3" using data on biochemical markers of tobacco smoke intake to allow for the fact that heavier smokers inhale less from each cigarette on average. From each regression line we determined the risk of ischaemic heart disease associated with smoking one dgarette per day by linear extrapolation. We then calculated the average of the five estimates, weighted (since there was no heteroge- neity) by the inverse of variance. S~ ~o~d/~ a~ p 980 Deparma~nt of En~en~ ~d M~ Worn ~dmm of ~d~e M~ St B~oMm~ ~d ~e London S~l of M~ ~ndon MR~w, r~ J K Mo~, NJ W~ Co~nd~ce m: Dr ~w 1997;315:973-80 VOLUME :~ 15 I~ Ot.-TG,3ER ,: '.~7 97:] - ~

Thirdly, we used the same cohort studies to determine how much of the excess risk of ischaemic heart disease is reversible many years after stopping smoking, as an indirect estimate of the extent of confounding. Fourthly, because people exposed to environmen- tal tobacco smoke eat less fruit and vegetables and this is associated with an increased risk of ischaemic heart disease, so confounding will arise,~' ~' we analysed pub- lished data to estimate the extent of confounding. We used published estimates of the increase in risk of ischaemic heart disease associated with a decrease of one standard deviation (SD) in consumption of fruit, vegetables, and antioxidant vitamins as markers of fruit and vegetable consumption.~ We identified studies that measured dietary intake of these nutrients (using weighed dietary inventory or 94 hour recall or food frequent, questionnaires with ~tuantitative estimates of portion size)~2"59 in smokers and non-smokers and in non-smokers who lived with smokers and did not (additional MeSH terms: fruit, vegetables, carotene, nutrition surveys, diet surveys, diet records, food habits). We calculated the differences as a proportion of the SD in each study. Thus we estimated the excess risk of ischaemic 'heart disease am'ibutable to differences in consumption of the nutrients in each stud): @e also examined the relation between smoking and other risk factors for ischaemic heart diseas~ Finally, because of the proposal that platelet aggre- gation may account for the large effect of exposure to environmental tobacco smoke on risk of ischaemic heart disease,~ we analysed published data (additional MeSH term: platelet aggregation). We fitted a logistic regression line to the data on risk of ischaemic heart <lisease according to platelet aggregation~° and estimated the increase in risk of ischaemic heart disease for a 1 SD increase in pl'atelet aggregation. We determined the effects of smoking and of exposure to environmental tobacco smoke on platelet aggregation (expressed in SDs) from published experimental studies6t~ and calculated the average increase, weight- ing by the number of subjects in each study. From these data we estimated the immediate increase in risk of ischaemic heart disease attributable to smoking and exposure to environmental tobacco smoke. We did not use cross sectional studies of platelet aggregation in smokers and non-smokers because they are insensi- tiven; the effects of smoking are short term6r and may 0.5 Studies Fig 1 Relative risk estimates (with 95% confidence intervals), adjusted for age and sex, from nine prospective studies (solid circles) and 10 case-control studies (open circles) comparing ischaemic heart disease in lifelong non-smokers whose spouse currently smoked with those whose spouse had never smoked (16 published studies (from left to right'~'=) and three with results cited by others from abstracts or theses= 3~) not be apparent in smokers who had not smoked for a few hours before blood was collected. Results Risk of ischaemic heart disease at low exposure to tobacco smoke Figure I shows the results of the I9 studies as the risk of ischaemic heart disease in never smokers whose spouses currently smoked relative to the risk in those whose spouses had never smoked (detail of design of the studies has been summarised previously~ ~).~*~ There were 6600 ischaemic heart disease events in total. There was no significant heterogeneity (X~ and figure 1 shows that the estimates from the individual studies axe consistent with each other. The summary estimate of relative risk was 1.30 (95% confi- dence interval 1.22 tol.38; P<0.00I), similar to the estimates from earlier meta-analyses with fewer studies?"5 Summary estimates were similar in women and men in cohort studies (in which almost all ischae- mic heart disease events were deaths) and case-conu:ol studies (in which most events were non-fatal infarcts) and with or without inclusion of three unpublished studies. It has been proposed that publication bias accounts for the association?~ The number of unpublished stud- Amed©zn Cancer Society 9 state 0 <10 10-19 ~20 British doctors 0 1-14 15-24 ~25 US veteranc American Cancer Society 25 state Pooling project o 1-9 10-20 21-39 ~,~0 0 '1-9 10-19 )~o 0 10 20 ~21 No of cigarettes smoked per day Fig 2 Estimates (with 95% confidence limits) from five studiess'~z of the risk of ischaemic heart disease in current cigarette smokers according to number of cigarettes smoked relative to never smokers (age standardised to age 65). Scale on the horizontal axis is linear with respect not to the number of cigarettes smoked but to the corresponding concentrations of biochemical markers= 974 BMJ VOLUMES15 18OCTOBER 1997

Papers ies would need to be implausibly large. As a simple approximation, eight of the 19 individual studies (fig 1) were significandy positive (a probability for each of less than 1 in 40 if there were no association). The total number of studies needed to generate these by chance would be more than 320 (8 x 60), of which only 19 were published. Exclusion from the analysis of all studies recording fewer than 100 events (as selective publication is more likely to affect small studies) made no difference to the relative risk estimate. nisk of smoking one ci~rnU per any Figure 2 shows the dose-response relation in each of the five studies of smoking and ischaemic heart disease (age standardised to age 65). The risk of ischaemic heart disease increases continuously with daily cigarette consumption, but in all five studies linear extrapolation of the regression line back to zero dose (no cigarettes per day) does not yield the expected relative risk estimate of 1.0. The estimates from the five individual studies of the risk at low dose ale consistent with each other, and the weighted average relative risk at one dgarette per day (defining risk in non-smokers as 1.0) was 1.39 (1.18 to 1.64; P<0.001). At 20 cigarettes per day this risk was 1.78. Table 1 shows the relative risk of ischaemic heart disease associated with smoking one cigarette per day accordkqg to age, firom the five cohort studies. The esti- mates decline with increasing age at death (P < 0.001). The average age of ischaernic heart disease events in the studies of exposure to environmental tobacco smoke was about 65 years and the estimate of 1.39 is dose to the estimate of 1.30 from these studies. With a linear dose-response relation the expected excess risk from smoking one cigarette per day is 4% (1/201h the excess risk of 78% from smoking 20 dgarettes per day); with environmental exposure it is 0.8% (1% of 78%). A high risk at low dose is therefore seen in the studies of smoking (59% excess risk v 40/0 expected) and in stud- ies of environmental exposure (30% v 0.8%). How much of the association is due to confounding by diet? Direct estimate The diet of smokers and of non-smokers who live with them differs from that of non-smokers who live with non-smokers. This dietary difference, rather than the exposure itself, may account for the high risk ofischae- mic heart disease,e' **' The most pronounced difference is a lower consumption of fruit and vegeta- bles,~54, st ~3 ~ ~7 which contain nutrients that may protect against ischaemic heart disease, including folic Table 1 Estimated dsk (95% confidence interval) of ischaemic heart disease relative to that in unexposed never smokers, from environmental exposure to tobacco smoke (see fig 1) and from actively smoking one cigarette per day (see fig 2) Age at death F, ovlmnmental Active smoking (yeats) exposure I cigarette per day 20 cigarettes par day 45 ~ 1.93 (0.99 to 3.78) 4A6 (1.21 to 18.42) 55 ~ 1.64 (0.95 tO 2.81) 3.07 (1.06 tO 8,88) 65 1.30 (1.22 tO 1.38) 1.39 (1.18 to 1.64) 1.78 (1.31 to 2.44) 75 -- 1.15 (0.83 to 1,60) 1.34 (0.72 to 2.51) acid,TM potassium, and linoleic acicU' These were not measured as potential confounding factors in studies of passive smoking and ischaemic heart disease. Fruit, vegetables, carotenes, vitamin C, and vitamin E ale highly correlated with each other, and the regression analyses for each nutrient were not adjusted for these correlations, so each nutrient will serve as a marker for the risk of ischaemic heart disease associated with fi'uit and vegetable consumption in general, even though I~ carotene and vitamin E do not themselves reduce mortality from ischaernic heart disease.'* 7t n Table 2 shows summary estimates from cohort studies of the relative risk of ischaemic heart disease associated with a difference of 1 SD in consumption of all fi-uit, all veg- etables, carotenes, vitamin C, and vitamin E as markers of fruit and vegetable consumption. Almost an studies showed a lower consumption of these nutrients in active smokers than non-smokers, but the differences were heterogeneous across studies (which is not surprising since fruit and vegetable consumption varies in different communities and different seasons, so the difference between smokers and non-smokers is likely to vary). Table 2 shows, from all studies reporting on each nutrient, the median dif- ference and the largest difference. The estimates of the excess risk ofischaemic heart disease corresponding to the median estimates of the dietary difference in no case exceed 3°/0 (suggesting that diet explains only a 3% excess risk of ischaemic heart disease in smokers); the largest estimates (based on the largest estimate of the dietary difference from any study and the upper confidence limits of the estimates of the association between the nutrients and ischaemic heart disease) do not cxceed 9%. The difference between smokers and non-smokers in plasma low density lipoprotein cholesterol is small--an estimated 0.07 mmolA,'~ corresponding to an excess risk of ischaemic heart disease (at age 65) of about 3%.TM The over-all excess risk of ischaemic heart disease attributable to dietary differences in smokers (fruit and vegetables and serum cholesterol) is thus about 6%. Blood pressure is no greater in smokers than Table 2 Estimates of increased risk of ischaemic heart disease in smokers relative to non-smokers, attributable to lower consumption of fruit and vegetables Madrer of Relative risk of ischaemic consumption of frail heart disease for decrease in and vegetables consumption of I $04~ Difference In consumption (smoker~ minus non-smoksr~) Oifference (prupod|on of I SOl Estlmat| of relative risk Of Ischaemio heart d~ssase No of studies Median" Largest Median LargesSe All fruit 1.16 (1.02 tO 1.31) 5~s~st =st --0.22 -0.33 1.03 1.09 All vegetabtes 1.23 (1.08 to 1.40) 6~*~s* ~ -.-O.t2 -0.25 1.03 Log Carotenes 1.06 (1.03 to 1.11) 9*t'~ --0.20 -0.34 1.01 1.04 Vitamin O 1.05 (1.00 to 1.09) 134a'~s -0.24 -0.49 1.01 1.04 V~'min E 1,05 (1.02 to 1.10) 6~z'~a -0.12 -0,27 1.01 1.03 *Median differences correspond to consumption lower by 10.15% in smokers than non-smokers. 1"Based on upper confidence limit of relative risk estimate and largest difference between smokers and non-smokers. .... : 975

tapers Table 3 Risk of death from ischaemic heart disease in men who had stopped smoking for 20 or more years, relative to that in men who had never smoked No of deaths from tschaemic heart disease Observed/expected Study Obselved Expected" (95% CI) US V'eteransn 2418 2303 1.05 (1.01 to 1.09) American Cancer Socie~/, 150 141 1.06 (0.90 to 1.25) 25 statese British doctorsT~ 200 174 ' 1.15 (1.00 to 1.32) All studies 2768 2618 1.06 (1.02 to 1,10) "From rates in never smokers, age adjusted. "i'15 or more years. non-smokers (and body mass index is less); these and other risk factors for ischaemic heart disease have a negligible effect on the relation between smoking and ischaemic heart disease.~5 For exposure to environmental tobacco smoke the median difference in fruit and vegetable consumption between non-smokers who do and do not live with smokers, based on all available studies (2-3 for each marker), was 0.10 SD for fruit,~5~ 0.06 for vegeta- bles," ~ 0.26 for carotenes,~" ~ and 0.18 for vitamin C.~ ~ ~ These are generally smaller than the median estimates for smokers (table 2). The corresponding estimates of the excess risk of ischaemic heart disease are I-2%. The differences in serum cholesterol'9 =' ~ ~ blood pressure)~ :~ 2s and body mass index~ 0-~ ~ ~ ~ in never smokers according to whether or not the spouse smoked were imperceptible (as expected since the dif- ferences between active smokers and non-smokers are so smaU). Combining the differences showed no significant difference, with upper confidence limits inconsistent with a serum cholesterol more than 2% higher, blood pressure more than 40/0 higher, and body mass index more than 6% higher. Estimates of the relative risk of ischaemic heart disease unadjusted and adjusted for blood pressure, serum cholesterol, body mass index, and a measure of sodal class (published in six epidemiological studies:~'~ ~ ~ ~o_) were similar: the weighted average of the adjusted estimates, 1.57 (1.00 to 2.13), was no Iower than that of the unadjusted esti- mates, 1.47 (1.00 to 2.19). Indirect estimate Three of the five smoking cohort studies cited above followed the men for 20 or more years. Almost all the excess risk reversed (table 3); the residual excess risk was 6% (2% to 10%). Thi~ sets an upper limit to any effect of confounding that is similar to our direct estimate of confounding. Table 4 Estimates of the extent of confounding and of the cause and effect relation in the associations of passive and active smoking with ischaemic heart disease at age 65 Relative flsk of [schaemic head disease (95% Cl) Active smoking Nature of association Passive smoking 1 cigarette per day 20 cigarettes per day Overall (from figs 1 and 2)° 1.30 (1.22 to 1.38) 1.39 (1.18 to 1,64) 1,78 (1,31 to 2A4) Irreversible: confounding" 1,06 (1.02 to 1,10) 1.06 (1.02 to 1.10) 1.06 (1.02 to 1.10) Revarsih~e: cause and effect 1.231. (1.14 to 1.33) 1,311. (1.11 to 1.55) 1.691" (1.23 to 2.33) • Estimates of the overall association in passive and active smokers, and of the extent of confounding, all apply to an average age at death of 65 years. 1"1,30/1.06=1.23; 1.39/1.06=1.31; 1.78/1.06=1.68. Confidence intervals take those of the "overall" and ~confounding" relative risk estimates into account; variances (in logarithms) were added. People who gave up smoking in recent years also changed their diet2"~ ~ We based our analysis on older studies, in which the former smokers would have given up before 1955, when dietary change was not widely advocated on health grounds (indeed, consumption of saturated fat increased in the United States between 1945 and 1955~). Significant dietary change in these studies is unlikely but even the recent data on dietary change after stopping smoking° ~" would increase the estimate of 6% to no more than 12%, similar to the direct estimate of the largest effect. " " Size of the causal assodation Confounding due to dietary differences accounts for a relative risk estimated as 1.06 in smokers. Of the over- all relative risk of ischaemic heart disease assodated with environmental exposure to tobacco smoke of 1.30, the estimated relative risk for the causal relation is therefore 1.2~ (1.~0/1.06). It will in fact be a little higher since dietary confounding is less than in smob ers` Even when based on the largest estimate of confounding in smokers it is 1.16 (1.:30/1.1,'2). For smoking one cigarette per day the overall relative risk is 1.39 and the esgimate Of the causal relation is 1.:31 (1.39/1.06). Table 4 summarises the estimates. Mechanism of effect: platelet aggregation Figure ~ shows the risk of ischaemic heart disease in a cohort of 2398 men (162 of whom had had a myocar- dial infarction) divided into five groups according to ranked measures of platelet aggregation (from the Caerphilly collaborative heart disease study~. The association is linear. The estimate from the logisdc regression line fitted to the data was that an increase in platelet aggregation of 1 SD (from ;my point on the distribution) is associated with a relative risk of ischae- mic heart disease of 1.]3 (1.19 to 1.48; P ~ 0.001). Table 5 summarises the experimental studies on smoking and platelet aggregation (from three different research groups).~* The effects of smoking and em,ironmental exposure are similar. The effect of smoking was similar in non-smokers who smoked on the one occasion (for experimental purposesfz~ and in habitual smokers abstinent for 8-10 hours,u'~ From the relation beBveen platelet aggregation and ischae- mic heart disease estimated above, the immediate increase in the risk of ischaemic heart disease amn~butTM able to effects on platelet aggregation is estimated to be 4:3% for smoking and :34% for environmental exposure (table 5). These estimates of the immediate effect are, as expected, a little higher than those of the long term effect of intermittent exposure over the day in table 4. The actions of smoking on other factors that increase risk of ischaemic heart disease are likely to increase continuously with dose such that the effect of exposure to environmental tobacco smoke is imper- ceptible. Thus plasma fibrinogen concentration is higher in active smokers but not detectably greater in non-smokers who live with smokers than in those who do not (the 950/0 confidence limits from combining three studies'~ ~ 7.~ were 4% lower and 40/0 higher). The effects on high density lipoprotein cholesterolv ~ and carboxyhaemog]obin~'~ ~ are also imperceptible. The more gradual increase in the excess risk from :39% smoking one cigarette per day to 780/0 at '2_0 per day (at 976 BMJ VOLUMES15 ISOCTOBER 1997

Papers P/ate/ot a#gre#atlon (SOs from the mean) Fig 3 Risk (95% confidence intsfval) of ischaemic heart disease relative to lowest group according to measurements of plate[et aggregation (induced by ADP), expressed as SDs from the mean, with fitted regression line (data from Elwood et al~°) age 65) is atm'outable to these and other non-platelet factors. Discussion Evidence for a large effect f~om a small exposm-e Environmental exposure to tobacco smoke is assod- ated with an excess risk of ischaemic heart disease of 30% and is estimated to cause an excess risk of 23% (95% confidence interval 14% to 330/0), increasing the risk of death fi'om ischaemic heart disease between the ages of 60 and 69 in British men who do not smoke from about 5% to 6%. So large an effect fi'om a relatively small exposure, though unlikely on first impression, is supported by a great deal of evidence. The association cannot be explained by bias in the studies of environmental exposure to tobacco smoke. Publication bias can be rejected, as discussed above. Misdassification bias arises because some people who claim never to have smoked are former or current smokers; they are at greater risk ofischaemic heart dis- ease and more likely" to have spouses who smoke. This bias has been found to be of minor importance in studies of lung cancer~; it will be negli~ole in studies of ischaernic hcar~ disease because the relative risk of ischaernic hcar~ disease in smokers is so much smaller than that of lung cancer (about 2 compared to 20). The cohor~ studies of smokers also show a substan- tial risk at low dose (table I). Studies directly measuring the e,xtent of atheroma- tous disease in arteries have confirmed a similar effect in smokers and in non-smokers exposc,d to environ- mental tobacco smoke, and a smaller effect in unexposed non-smoker~~' ~ The association cannot be explained by confound- ing. If all the excess risk of ischaemic hcar~ disease in non-smokers who live with smokers compared with those who live with non-smokers were to be explained by their dietary differences, about half the excess risk in smokers would also have to be attn'butable to differences in die~ The excess risk largely reverses on stopping smoking, indicating that this is not the case. The estimate of the extent of confounding as the excess risk that is not reversed many years after stopping smoking (6%) corroborates the estimate of the excess risk to be expected fi'om the dietary differences. The effect of tobacco smoke on platelet aggrega- tion provides a plausible mechanism for the low dose effect The immediate effect of a single environmental exposure is to increase risk by an estimated 34% (table 5). This will be an underestimate because of regression dilution bias (which could not be a/lowed for because repeat meastaes were not available), but will also tend to be an overestimate because it does not rdlect typical intermittent exposure to environmental tobacco smoke over the day, which will be somewhat less. One hour after a single exposure the effect on platelet aggregation is attenuated by about half," so the likely effect of intermittent exposure throughout the day ' could be consistent with our estimate of a 23% increase in risL Simple corroboration is provided by the obser- vation that aspirin abolishes the effect of tobacco smoke on platelet aggregation~ ~ and reduces the risk of ischaemic heart disease by about 25%.m The experi- mental evidence of a similar effect of smoking and environmental exposure on platelet aggregation is corroborated by studies showing that the generation of thromboxane A~ fi-om arachidonic acid (which leads to platelet aggregation) was also similar?= A small dose of an agonist seems to have a maximal effect on platelet aggregation." A mete-analysis of the studies of occdpadonal exposure to environmental tobacco smoke exposure indicated a disproportionately large effect: relative risk 1~6 (1.08 to 1.71).~ The occupational studies lack the susceptibility to confounding because non-smokers who work with smokers will not share their diet- In animal experiments, eight studies on four spedes (involving exposure to the smoke from simultaneous combustion ofbetwden one and 10 dga- rettes, generally for 4-6 hours per day over 6-16 weeks) all showed pronouuced vascular toxidty of the Table 5 Results of experiments of exposure to tobacco smoke and platelet aggregation and the associated immediate increase in risk of ischaemic heart disease Change i.n platelet aggregation ratio alter =xpo~mf Associated ralatlve dzk of ischznmic hall1 F.xpOlOre AIIso|O|I llo of standard deviations dlsll+Zl (9S%C|l~ Unexposed control period (non..smokem, n=lO~t) 0 0 1.00 Environmental tobacco smoke(20 minutes, n=lOm) 0.09" 1.03 1.34 (1.19 to 1.50)" Active smoking: smoldng one~ or two'~'~" cigarettes 0.11" 1.25 (awrage of 6 studies, n,,158) 1.4~ (1.24 to 1.63)" "P~O.OQ1. 1"Absolute ct~ange is an estimate of [~e propo~an of all ¢ircu{aling platelets t/~at were incorporated into a(jgre~ates as a result of ~e exposure; platelet aggregation ratios were on average 0.83 before and 0.72 after active smoking. $From lt~e assodation between platetet aggregation and ischaernic heart disease shown in figure 3. BMJ VOLUME31~ 18OLWOBER t~,)97 977

exposure.~ The size of the resulting infarct after experimental occlusion of a coronary artery was 50-1000/0 greater~ and arterial atheromatous disease was about twice as extensive"~ in exposed animals than in unexposed control~ Evidence against a large effect at low dose Pipe and dgar smokers have a low risk of ischaemic heart disease (the risk reladve to non-smokers was esti- mated as 1.13 for pipe smokers, 1.30 for dgar smokers, and 1.75 for dgarette smokers~). Pipe and dgar smob ers tend not to inhale when smoking and because of characteristics of the smoke can absorb nicotine through the mouth,9s but they must inhale their own environmental tobacco smoke in the same way that non-smokers inhale other people's smoke. Their low risk might at first sight weigh against the view that inhaling environmental tobacco smoke poses a material risk ofischaemic heart disease. However, men who smoke only pipes or dgars smoke much less frequently than cigarette smokers. In a study of 21 520 m.en, 9618 current pipe or cigar smokers smoked on average three times a day whereas 4184 dgarette smokers smoked on average 90 tLmes a day.~ The dif- ference in this frequency, in view of the short term nature of the effect on platelet aggregation (half life less than an hour~, may weal reconcile the observa- tions of passive smokers and the risk of ischaemic heart disease among pipe smoker~ Exz/aded stud/zs A separate analysis of one of the studies of environmental tobacco smoke exposure and ischaemic heart disease in the set of 19 studies (fig 1), and of two data sets not published elsewhere (from the US National Center for Health Statistics and the American Cancer Society) has been published by Layard and LeVois, consultants to the tobacco industry.~s They "reported a combined reladve risk estimate from the three studies of 1.00, with a narrow 950/o confidence interval (0.97 to 1.04).~ This negative result is statistically inconsistent with the estimate of 1~0 (1.99 to 1~8) from the above analysis of 19 studies (P<0.001). The difference is too great for the two groups of studies to be combined as separate valid esti- mates; one must be flawed. We took the estimate from the 19 studies as valid and rejected that of Layard and LeVois, since there is no reason to reject an analysis based on 19 independent studies in favour of one from a single group with a vested interest. Direct evidence supports this decision. Firstly, in one of the three studies the relative risk estimate (men and women combined) of Layard and LeVois of 0.98 (0.90 to1.07)s was inconsistent with that from an inde- pendent analysis of the same data commissioned by the American Cancer Society (the owners of the dam) of 1.91 (1.06 to 1~8)?I Secondly, even in the absence of a causal effect, the combined estimate of 1.00 is incon- sistent with any confounding from the dietary differences. Thirdiy, the result is inconsistent with the data on low dose active smoking, the evidence on platelet aggregadon, the animal studies and the other evidence summarised above • Analysis of 19 epidemiological studies shows that people who have never smoked have an estimated 30% greater risk of ischaemic heart disease if they live with a smoker (P < 0.001) • This is surprisingly large--almost half the risk of smoking 90 dgarettes per day even though the exposure is only 1% of that of a smoker • The excess risk from smoking one dgarette per day is 39%--similar to the risk in a non-smoker living with a smoker • The effect is mainly explained by a non-linear dose-response relation between expsoure to tobacco smoke and risk of heart disease • Detailed analysis shows no significant bias; dietary confounding can account for an excess risk of only 6%, so revising the excess risk from 30% to 93% Conclusions We believe that there is no satisfactory alternative interpretation of the evidence reviewed here than that environmental exposure to tobacco smoke causes an increase in risk of ischaemic heart disease of the order of 950/0. In proportionate terms this is of similar magnitude to the effects of exposure to environmental tobacco smoke on lung cancer,~ but the number of excess deaths from heart disease will be far greater because heart disease is so much more common than lung cancer ,in hon-smokers. Reversal of the effect would reduce the risk of ischaemic heart disease by about as much as taking aspirin, or by what many peo- ple could achieve through dietary change. The effect of environmental tobacco smoke is not trivial as is often thought. It is a serious environmental hazard, and one that is easily avoided. The evidence on ischaemic heart disease warrants further action in preventing smoking in public buildings and enclosed working environ- ments. The hazard in the home requires greater public education so that smokers recognise the risk to which they expose members of their family. It is also important that clinicians advise that families of patients with known coronary artery disease do not smoke in their presence_ We thank Richard Doll Martin Jarvis, Stanton GIantz, and Judson Wells for their comments on earlier drafts. Funding: The Department of Health (England) supported this work. although the views are our own. Conflict of interes~ None. Glantz SA, Pannley WW. Passive smoking and heart disease: epideminingy, physiology and biochemistry. C/reu/a/mt 1991;83:1-1R "6b.lls AJ. Passive smoking as a cause of heart diseas~J Am C,~ Can//~ 1994",~4:546-54. Glantz SA, Parmley WW. Passive smoking and heart disea_ce: mechanisms and risLJAMA 1995'~73:1047-53. SteeMand IL Passive smoking and the risk of heart ~ JAMA 1992",267.'94-9. Kritz I'L Schmid P, Sinzinger FL Passive smoking and cardiowa_q:tdar risk. Arc~ In~rn Med 1995;155:1942-8. Hammond EC., Gaffmkd L Coronary heart disease, su'oke, and aortic aneurysm.An:h ~ Hm/th 1969;19:167-89. Hammond EC. Smokingin relation to the death rates of one million men and women. In: Haemzel W, ed. E "pulemiologar~ ~ ~ th~ study of mnc, tr aad oam- rhron~ d/.u~ Bethesda, MD: US Depar~nent of Health, Education and ~dfare, 1966:127-204. (National Cancer Imtitu~," Mono- graph No 19.) Kahn HA.The Dorn stody of smoking and mortality among US vete~ms: r~port on eight and one-half years of obse~ation.in: Haenszei W, ed. Ep/- 978 BMJ VOLUME315 18OCTOBER 1997

~th~ MD: US Dep~em of H~, ~u~on ~d W~, 1966: L- 125. (Na~on~ ~cer Imfim~ Mono~ph No I9.) 9 H~mond E~ Horn ~ Smo~ng ~d d~ ~t~--~ on fo~-fo~ mon~ of~ollow~p o~ 187 783 m~J~ Z958;16~:1294-30~ 10 Doll ~ Hid of B~h d~o~ BA~ 19~;i:I3~-1410. 11 Doll ~ Hill .~ Mo~i~ of B~h d~m~ ~ ~lafion to smo~ o~- ~er Imfimte Mono~h No 19.) 12 ~ng ~ol~t~ stooling h~i~ ~lafive ~ight ~d ECG ~no~i~ to ~dd~ of ~jor co~ ~ena: fi~ ~ of ~e P~l~g pmj~J ~ Db 1978~ 1:20 I-3~. lung ~. BMJ I~7~ 15:980~. 14 ~d NJ, 1984~:~1. 16 ~ G, Conze ~ ~ck~ ~ Adl~f~ E L'p~ of ~ba~ ~oke 1~2;70~52~7. 17 ~j~ D~ A~ M~ Clan P, ~b~nJ, M~e ~ Do~d ~on ~ h~y yo~g ad~ N~JM~ 1996~34:15~. 18 H~ Z ~ive smo~ NZ M~J l~0fi~4. of ~i~ ~o~g on ~c h~ ~e mo~ oE aommo~ A~J ~ 1985;121:~50. 20 ~ PN, C~,~d~n 198~7-105. 21 Sv~ ~ Ku~ ~ M~ ~, ~e~ ~ ofp~i~ smok- ~g ~ ~e m~fiple ~k ~mr ~¢~fion ~ ~J E~I 1987;126:7~-9~ ~ h~ ~ Hole D], ~p~ h~ I989~99:~-7. 2~ S~ DP, Co~ G~ H~g K], Sho~ D~ D~ ~m ~ ~ ~ non-~mo~ who ~v~ 1989;79:163-7. smo~g ~d 20-y~ ~o~ ~ mo~ ~ong nommok- ~g ~v~ E~ Co.w, G~ AmJ ~l~ H~ 1990~0~01. ~6 Dob~n AJ, ~d~ ~ H~ ~, ~oyd D~L ~ive smo~g ~d ~e ~k o~h~ a~ or ~ d~ McdfA~¢ 1991;1~:79~-7. ~7 ~ V~ ~ D~o B, F~osi MG, To~o~ ~ ~ive smo~g ~ 28 ~ a ~k ~tor for n~ ~e~ BA~ 19~08~80~. 29 T~-P~ ~ B~ ~ ~ M, ~v~e ~ ~ive ~ok- ~g by ~ ~ ~d ~ co~e ~d ~e p~e of ~p~o~ ~ H~ 1995;49:1~9~3. • e ~k ofh~ a~ ~J ~L 1995:2~:715-9. ~d ~m~ h~ d~e ~ ~e ~ ~c~ S~e~ ~-D et ~ A C~i~ 1997;95:~7~9. ~ C~ J, ~ ~ ~ive [~s~I.E~ H~J 1996;17{~s~ suppl)~09. mo~ fo~owba~ s~'~. ~ T~ P~ 1995~ 1:180-~. smo~n~ h~ 1995;21:18~9L ~7 W~d ~, ~ie ~ V~i~on of sm~ on l~g ~ ~ non-smok~ $8 ~w M~ Mo~]~ ~n ~cer. BrJ ~ 199T;75:1690-3. 39 ~ L Col~ G~ ~o~d~g, m~t ~r, ~d publi~fion 1996;I~09-15, 1~6~48:1663~. ~ I ~w ~ Mo~J~ By how mu~ d~ ~it ~d v~le ~n ~uce ~e ~k of ~emic h~t d~e. E~f CI~ N~ t 997 (~ p~). ~t~on of ~83-96. 43 CadeJE. Margeits BM. Re/atiomhip between diet and smok~g-.-i= the diet of smoker~ dilferenOJ E~aidemiol C~ Hea/~h 1991;45:270-9 44 Bolton Smith C. Wtmdward M. Brown CA, Ttmatall-Pedoe I-L Nu~ient intake by duration of ex-smoking in the Scottish heart health stud)'. BrJ 45 Zondervan KT, Ock~ MC, Sink HA, Seldell JC. Do dietary and supplementary inmlu~ of antioxidants dilfer with smoking stores? Ep/dem/o~ 1996;25:70-9. 4~ j~rvinen, [rmckt P. Sepl~lnen. Reuna~en A, Heli6vaara M, Maatela, et aL Antioxidant vitamins in the di~ relationships with other person,d char- acteristics in HntancLj'F~ C.~mum~ Hca/~ 1994;48".549-54. 47 Margetts BM.j'ackson A~ Interactions between people's diet and their - smokJng habits: the die~mT and nu~'idonal survey of Britbh adula. 273./1. 1993",~07:138 I-4. 48 FehilyAM.Phillips IGM, YarnellJWG.Diet~smolCmg, sodal clax~andbody mass index ia the Caerphilly heart disease study. Am J 1984;40:8'27-33. 49 Lark~ FA, Basiotb PP, Riddick HA, Sykes KE, Pao EM. Dietary ofwomea sraokcrs and non-smokc-mj'Am D/~ As~¢ 1990;90:230-37. 50 Midgette AS, BaronJA. Rohan TE. Do dgaxette smokcr~ have diets that incre~e their risl~ of coronary hem't disease and cancer? ,t~J 1993; 137".52 I-9. 31 McPhillipsjrB, Eaton CB, Gan~ KM, Derby CA, La~awr TM, McKenneyJ'L, e~ al. Dietary differences in smoker~ and nonsmokers from ~uthca~em New Engi,~d communitie~ JAm Dia Aao¢ 1994;9,P2.87-92. .52 Suhar AF, Hadan LC, Ma~on ME. Food and nuU~em intake dllfe~mce~ be~veen smoke~ and nou-~moke~ in the US. Am J Pub//~ HmP.h 1990;80:I~23-9. 53 Emmon~ KM, Thompson B, Feng Z, Heber~ jR, Heime~dinger.~, Linnan L. Die~ry inta~ and exposure to e~vironmenml tobacco ~moke in worksim popula6or~ EurJ Clln Nutr 1995;49".536-45. ~t Kelth RE, D~kelIJA. F.~ec~ of chronic cigareue smoking on vitamin C szatu& lung ftmtion, and re~ting and excrete car~ova.tcuJar in humar~ Nutr//on ~ ~ 1980;21:907-12. 55 Triable DL, Giuliano LJ, Forunann SP, Reduced plasma a=corblc acid concemration~ in noramoke~ .regularly expo~,d to eavironme~ ~oba~o ~moke.,~1. C.~ ,Vu~" 1993",58".886-90. 56 Nuuen$ MC, Romon M. Ruidave~.]B, Arveiler D, Ducimefiexe P, Lecerf j'M, et aL Relatiomhlp between smoking and diet: the MONIC&-France proje~J/ntn'n Md 1992;231"349-56. 57 Thornton A, Lee R Frye. Difern~es between ~raoke~ ex-~moker~ sire smoke~ and no~smoke~,J C//n E/n~,m ~/1994;47:1143-62. 58 Le Mazr.hand L. Wilkerts LR, Hankin JH, HaJey NJ. Dietary pauerr~ of female nommokers with and without exposure to environmental wbacco smoke. ~nc~r Cm~" ~n,'m11991",2:11-~. 59 Sidne~ S, Caan BJ, Friedman GD. Die~ry intake ofcaromne in nommok- er~ with and withou~ pa~ive smoking a~ home. Am J 1989;129:1305-9. ~0 EJw~d P, Recta S, Sharp DS, B¢~Ick AD, O'Bden J'R, 1"amdl j~VG. I.~chaemic hear~ ~ and plamle~ aggreg~on: the Caevi~billy collabo- r=five heart dhea~ sm~.. C/nu/~u~n 199 I;83'2,8-44. 61 Davit ~, Shein3n L, Wamnabe IS, Arnold J. l~._~ive smoking alfec~ endothe|ium and pla~e~e~. An~/n~rn Ma/1989;149"38~.9. 69 Davis JW, Davis RF. Prevention of cig-areff.e smoking-induced platele~ aggregate formation by aspirin. An~ In/~rn Ma/1981;141:206-7. 63 Dav~ JW, Shehon L, Eigenberg DA, Hignite CE, ~ IS. F.~ec~ of ~oba~:o and non-~obacco cigare~e smoking on e~dothellum and platele~. ~ P/mrmawl T/m" 1985;37~29-3~. 64 Davi~ ~, Sh~dmn L, Harunan CR, Eigenberg DA, Rutfinge~ HA. Smoking-induced changes in e~dothelium and platele~ are not affected by' hydroxye~hylrumsides. BrJ Expt,6men~ Padm11986;67:765.71. 65 Davb JW, Dav~ RF, Hassanein KM. In he~Jthy habitual smoker~ acetyba/icylic acid abolishes the efl'~ts of r.obacco smoke on the platele~ aggrega~ m~. C~n Ma/AuocJ 1982:126:637-9. 66 Davb.~, Haru'nan CP,. Lew/~/-ID, Shelwn L, E[genberg DA, KM, e~ aL Cigarev, e smoking.induced enhancemem of pla~elet function: lack of prevention by a.~pirin lumen widl coronary artery dhe~tse.J Lab C//n Ma/1983;I05:479-83. 67 3chndd~ KG, Ra~mu.~en j~. Acute platele~ activation induced by smok- ing. T/mTmb Haemmt 1984;51:279-82. 68 Blache D, Bouthiltier D, Davignon J. Acute influlence of smoking on platelet behaviour, endothe~ium and plasma llpid~ and normalization by ~ph~n. At&,ro.Mcm, d~ 1992.'93:179-88, 69 Fit2C_,~d GA, Oams JA. Nov, ak j. Cigare~m smoking and hemosta~ic func~on.Am Heart.[ [988.' [ 13:267-71, 70 Boushey CJ, Ik,~'~ford SAA. Omefm GS, MomLsky AG. A quantitative assessment of plasma homocysteine as a risk factor for ",~a..~:ular J&~4A 1995;.'2"74:1049-37. 71 AJpha-tocopherol Beta-<arocene Cancer Prevention Study Group. The effect of vitamin E and beta carotene on the incidence of lung cancer and other cance~'~ ia ma/e smoker~ ,VF.ngtJMed 1994~30:1029-35. 72 Hennekem CH. BuringjE. MamonJE. 3tampfer ~ Romer B, Cook NIL et al. Lack of effect of Iong.term supplementation with beta carotene on the incidence of malignant neoplasms and cardiovascular diseas~ N J M~ 199~4:~ 145-9. 73 Craig WY, P'~omaki GE. HaddowarE. Clgareue smoking and serum lipid and lipoproteia concenwarlom: an an~/ysis of published da~ ~989;298:784-8. 74 Law ~ Wakt NJ', Thompson SG. By how much and how qu/dd~" z~luction in serum cholesteval concentration lower risk of heart disea.~ ~AfJ 1994~08.367-75. 75 Friedman GD. Dales LG, U~ HK. Mortaiky in mlddle-aged smoke~ and nommoke~. NF_,~J M~ 1979",300:21~-7. 76 Stephen AM, Wa/d N'J.Trends in individual consumption of dietary fac in the ~nited States, 19~0-1984. AmJ Clin Nua" 1990;3~:437-69. Papers

vapers 77 Rogot E. MurrayJL Srnokingand catue~ ofdeath among US veterans: 16 yea~ of obs~rvatlon. Public Hm/th Rep 1980;95:913-22. 78 Doll R. Peto R. Mortality, in reladon to smoking: 20 years' observations on male British doctot-~.BAfJ 1976:2:1525-36. 79 Iso H. Shimamoto T. Sato S. Koike K. Iida M, Komachi Y. Passive smoking ~md pL~m~ fibrinogen cancen~fiom. AmJ Epidtmi~ 1996; 144: 80 Howard O, Burke, GL Szkio M, Teil GS, Ec.kfeldt J, Ew-~ns G, et aL Active ~nd p,'~.~ive smoking are a~x'iated with incre~ed e~rodd ~ thickne~.~ Arch Inttr~ Mtd 1994;154:1277-82. 81 Amiplatelet Tri~li~tf Collabo~tinn. Collaborative o~'iew of domi~ed ~ of amiplate]et therapy. L Predation of death, myoeard~l inf~xion, and smoke by prolonged antiplatelet therapy in va~ou~ categoriet of patienm BMJ 1994~08:$1-10fx 82 Sdamid P. Karanika~ G. Kritz H, Pirich C. Stamatopoolos Y, Pe*kar B, et ai Pa~five smoking and platetet thromboxane- Thromb R~ 1996"81:451-60. 8~ Di Minno G, Silver MJ, Murphy S. Monitoring the envy of new platelet~ into the drculation meter ingestion of a~pirin. B/nod 198S:61:1081o5. 84 Welh AJ. Post-hearing comment re OSHA hearing: raeta-anal~i~ of rela- tive rLsk of heart disease from ETS exposure in the workplace- '#a~hing. ton, DC: US Department of Labor, 1995. (Docket H-122, Docket olrme Room N-2625.) 85 Prentice RC, Carroll R, Scanlon PJ, Thom~JX. Recent expo~re m do- r~te smoke incre-~e~ myocardi~ inf~ra size- J Am Co// Card~ 1989;13:124A. 86 Zhu B, Sun Y, Sieve~ RE, Glamz SA, Parraley WW, Wolfe CL Expom~ to environmen~.l tobacco smoke increa~ myoe-~dial infara size in C/feud/on 1994; 89:1282-90. 87 Zhu B, Sun ¥, Sim, er~ RE, Shuman JL, Glamz SA, Chatterjee K, et aL L-arginine decrea~t infara size in ratt exposed to environmental tobacco ,moke. Am Ht~tJ 1996:132.'9 I- 100. 88 Penn A. Snyder CA. Inhalation of ~ide~eam dgarette smoke aexdetate~ deveinpment ofarterio~cierodc pizquex Cimdatirm 1993;88:1820-5. 89 Penn A, Chen L-C, Snyder CA. Inhalation of steady-state sidesn'e~n smoke from one cigare~m promo~e~ anexio~¢lerodc plaque deveIopmenu ~l) Zhu ~ Sun Y-P, $ieve~ KE, Iaenbe~g WM, Glantz SA, Parmley WW. P~tdve *moking increa.te~ experimental atheroscierosi~ in cholesterol- fed riobitx fAm Coil Cardio11993;21".225-32. 91 Sun Y-P, Zhu B-Oo Sievert RE, Glamz SA, Parmiey XCW. Metoprolo| doe~ not attenuate athere~cierosis in lipid-fed rabbitt exposed to environmen- tal tobacco smoke. Cirmlamm 1994:89:2260-§. 92 Robertt KA. Re2ai AA.Pinkerton KE.RudedgejC. Effect of environmen- tal tobacco smoke on LDL accumulation in the artery walL Circu~ulon 1996;94:2248-53. 93 X~t~ld NJ, Idle M0 Boreham J. Serum cotinine levels in pipe smokef~: evl- dence against nlcodne a~ came of coronary heart ~ Lanea 1981',2:775-7. 94 Wald NH, Watt HC. Prospective study of effect of ~itching from cigarette* to pip~ or cigart on mortafity from thee smoking related di~ ~ BMJ 1997",3 (Aa~p~d 22 Se~ 1997) The accumulated evidence on lung cancer and environmental tobacco smoke A K Hackshaw, M R Law, NJ Wald a~d p 973 Department of Medicine, WoL~on Imdmte of Prev~dve Medicine, St Bartholomew% and Roy'ad London School of'Mediclne and Dem~try, London EC1M 6BQ A K Hack3haw, M R La~4 NJ Wald, Correspondence to: Profe~or Wald njwa/d@mdx qmw.a~uk BMJ 19~7".315.'980-88 Abstract Objective: To estimate the Hsk of lung cancer in Lifelong non-smokers exposed to environmental tobacco smoka Desigm Analysis of $7 published epidemiological studies of the risk of lung cancer (469-6 cases) in non-smokers who did and did not live with a smoker. The risk estimate was compared with that from linear extrapolation of the risk in smokers using seven studies of biochemical markers of tobacco smoke Main outcome measure: Reladve risk of lung cancer in lifelong non-smokers according to whether the spouse currendy smoked or had never smoked. Results: The excess risk of lung cancer was 24% (95% confidence interval 15% to $6°/o) in non-smokers who lived with a smoker (P < 0.001). Adjustment for the effects of bias (positive and negative) and dietary confounding had little overall effect; the adjusted excess risk was 26% (7% to 47O/o). The dose-response reladon of the risk of lung cancer with both the nnmber of dgarettes smoked by the spouse and the duration of exposure was significant. The excess risk derived by linear extrapolation from that in smokers was 190/% similar to the direct estimate of 26%. Condusiom The epidemiological and biochemical evidence on exposure to environmental tobacco smoke, with the supporting evidence of tobacco specific carcinogens in the blood and urine of non-smokers exposed to environmental tobacco smoke, provides compelling confirmation that breathing other people's tobacco smoke is a cause of lung cancer. Introduction Ten years ago scientific committees and national organisa~ons concluded that exposure to environmen- tal tobacco smoke (also called passive smoking) is a cause of lung cancer.~'~ Substantial additional evidence has since been published, and we report a new analysis. The additional data permit a more predse estimate of the sizc of the association, with a further assessment of whether it is cause and effect by seeking a dose- response relation and examining whether sources of bias and confounding could account for the assod- ation. We also compared the direct estimate of risk from epidemiological studies with that fi'om a low dose linear extrapolation of the risk in smokers using biochemical markers of exposure to tobacco smok~ As before~~ the estimate of effect was the relative risk of lung cancer in lifelong non-smokers according to whether the spouse currently smoked or had never smoked. Spousal exposure is the best a-v-ailable measure: it is well defined and has been validated using biochemical markers.~ It reflects exposure in general because non-smokers who live with smokers tend to be more exposed to tobacco smoke fi-om other sources, because they are more likely to mix socially with smok- ers) Workplace exposure varies considerably and is difficult to measure_ Me~o~ Direct estimate of risk of lung cancer from epidemiological studies Studies of environmental tobacco smoke and lung cancer were identified from Medline, the dtations in each study, and consultation with colleagues. We 980 BM~ VOLUME315 18 OCTOBER 1997