Tobacco Documents Online

5 Heart Disease Table5.1' Heart disca.sc risk in never smokers in relation to spouse/household smoking 187 Study Sex Cases Relative risk (95 Ya limits) Factors' adjusted for U E Butler F 20 60 1.05 (0;65-1.70): Age Garland F 2 17 3.51 (0:80-15.3) None'- He F 9 25 1.50 See text Hirayama F 118 376 1. fS (0.94-1.42) - Age of wife Hole Gillis et a1. F 2 19 3.56 (0:83-15.4): None3 M 18 14 1.30 (0:64.2.64) None3 Hole eral. M+F 30 54 2.01,(1.21-3.35) Age. sex.,class. BP: Huasble lI F 27' 49 1.59y0.99-2.57) cho11 BbS7 Age, sex. BP. chol. Ltt F 22 55 0:97 (0!56-1.69), BMI None!i M 26 15 1.34 (0:64-2:80); None' Martin F -23- 2.6 None= Palmer F -?- 1.2 (signifinnce Not known unknown) Sandier Il F 437 551i 1.19 (1.04-1.36) Age. schooling. M 248 122 1.31 (1.05-1.64) housing: maritall Svendsen M 8 5 2.23 (0.72-6.92) sutus Age. BP; chol...-t. drinks. education U I- Unexposcd; E - exposed. ' BP - blood aressure; chol - eholesterol;, BMU- body, mass index: w•t - wcight.. ~ See text_ s Adjustment for age had l'tttle effect, and adjusted 95% Iimits could not be calculated. 'Numbers of cases are approximate, based on age-adjusted rates. has not been usedL It would be incredibly unstable (estimated 95% limits, about 0.2-5(90)„sinee it is probably impossible to adjust prop- erly for multiple factors with so few deaths. Mantel [pen. commun.] has also said that 14.9 was an error, the appropriate value being log, (14.9) or 2:71! 2 in-_ ~'

198 Environmental Tobacco Smoke and Mortality Secondly, unadjusted relative risks have been used for the individual sex results from the Hole study (published in Gillis et al [29]), since age standarditation had very little effect and relative risks could only be properly calculated for the unadjusted data. (v) The relative risk given in the He study [102], a case-control study in which there was matching omagc, race, place of residence and occupa- tion, is adjusted for previous and family history of hypertension, family history of coronan• heart disease, amount of exercise, history of drinking, and hypercholesterolaemia. Confidence limits could not be calculated, but the relative risk was stated to be significanttp < 0:01), but this seems not to be true (see section 2'.7:1). (vi) The results presented in Table 5.1 for the H'irayama study are for ischaemic heart disease. Hirayama [20] notes that no significant rela- tionship was seen between spouse smoking and risk of 'other heart disease' (undefined), base& on 68& deaths, or risk of hypertensive heart disease, based on 226 deaths. Relative risk estimates were not provided for these two disease categories. Table 5.I provides 4 independent relative risk estimates for men and 10 for women; 13 of which.are greater than 1, with 4 of them significant: both the male and female estimates for the Sandltr IC study, and the female estimates for the He an& Manin studies. Hole also showed a significantl} increased relative risk when results for the sexes were combined. Although an overall estimate of relative risk (as for example calculated by Wells [7]) is probably of little meaning given the extreme variability in study designs and populations involved, the data in this table - considered without regard to study design and a variety of other methodological problems - indicate a weak association between exposure to ETS from the spouse or in the household and' risk of heart disease. Five of the l l studies provided some information on risk of coronary, beam disease and extent of exposure to ETS from the spouse or in the household (Table 5.2): The Hirayama and He studies both showed evi- dence of a dose-response relationship; with a significant (p < 0.05) trend and elevation in risk for women whose husbands smoked 20 or more cig- arettes a day. The Hole and Svendsen studies also showed the highest riskk in~the highest exposure group„though here numbers of deaths were smalli and' the trend statistic was not significant. In the large Sandler 11 study there was no evidence of a relationship of risk to ETS exposure in either sex, risk of heart disease being similar in those classified as exposed to light or heavy ETS exposure.

5 Heart Disease 189 Table 5.2 Heart disea.cc risk in never, smokers in relation to extent of spousehrousehold smoking Study Sex Exposure Hirayattu F Husband: never smoker ex-smoker or current 1-19 cigs/day current 20+ eiYs/day He F Husband: never smoked' during marriagc smoked 1-20 cigs/day smoked 21+ eigs/day Hole F Household smokmg none low•, high (cohabitant 15+ cigslday) Sandier !1 Household exposure (tcare)`. F 0(none) 1-5 (light) 6+ (heavy) M 0(none) I-S (light) 6+ (heavy) Svendsen M- Wife:: did not smoke smoked 1-19 cigs/day smoked 20+ cigslday Cases Relative Factors risk adjusted for 118 1.00 Age of wife 240 1.08 136 1.30- ` Cases and 9 1.00 eontrols 12 230 matched for 13 6.86 age, race, + occupation, cesidena 3 1.00 Age 14 2.09 16 4.12 437 1.00 Age 252 1.20 schooling, 299 1.27 housing 248 1.00 quality, 56 1.39 marital status 66 1.24 8 1.00 0 None 1 0.90 4 3.21 Information on hean disease in relation to other indices of ETS expo- sure is fairly sparse. In the Lee study, subjects were classified on a score rangin,g frorn 0 to 12 according to whether they considered they were exposed not at all (0), a little (1), average (2), or a lot (3-12), separately for at home, at work, during travel; and during leisure. No significant relation- ships were seen,,relative risk estimates for scores 0-1, 2-4 and 5-12 being 1, 0.43 and 0.43 in males (based on 30 deaths), and 1, 0.59 and 0.9 1 in females (based on 36 deaths): ~a~~Sg :1.22 0 G_'! ~! ~'F

190 Table 5.3 Environmental Tobacco Smoke and Mortality Effect, of adjiestment for various factors on estimated rel'ative risk of heart disease in never smokers according to spouse/household smoking Study Sex Adjustment factors Relative risk (959b limits) Butler, F None 1.36 (0.82-215) Age 1.05 (0.65-1.70) He F Matching factors (age, race, residence, occupation) onl) 3.52 (11.43-8.65). Matching factors.,also previous and family history of hypertension, familv history of CHD. exercise, drinking, hypercholesterolacmia .50 Hirayama F None 1.00 (0!8'1-1!.23) Age of wife 1.15 (0;94-1'.42) Age of husband! 1.15 (0;93-1.41) Age and occupation of husband 1.16 (0.94-1.43) Holt F None 3.56 (0:83-1 S:4) Gillis et aL Age 3.25 M None 1.30 (0.64-2.64) ! Age 1.29 Hole et all M+F Age 1.75 (1.1i0-2:83) Age, sex, social class. BP„chol, BMI 2.01 (1.21-135) Huntblt If F Age 1.34 (0.84-2.21) Age, BP, ehot! Bh4I 1.59 (0.99-2.57) Lee F None 0.97 (0.56-1.69) Age, marital'status 0.93 M None 1.34 (0.64-•`2.80) Age, marital status 1.24. Sandler 11 F None 0;66 (0:59-0.75) Age, housing quality, schooling, marital status 1.19 (1.04-1.36) M None 1I.17 (0.95-1.46) Age„housing quality; schooling, marital status 1.31 (1.05-1.64) Svendscn M None 2:12 (0.69-6.46), Age, BP, chol, wt, diinks/week, education 2:23 (0.72-6.92)1 ~...', BMI - Body mass index; BP - blood pressure; ehol- eholesterolrCH'U - eoronary heart disese; wt - weight. ~~. l~' "-

© 5 Heart Disease ® 191' In the Svendsen study, a relative risk estimate, adjusted for age andd wife's smoking status, of 2.6 (p - 0.23; 95% limits, 0,5-12.79 was derived when men whose co-workers smoked: were compared with men whose co+ workers did not. This result, and that in Table 5.1, was for coronary death. Svendsen also provided results for the end-point fatal or non-fatal coro- nary event. Here, relative risk estimates were derived for four categories: (a) neither: wife nor co-worker smoked, f.0 (base); (b) co-Worker smoked but not wife, 1.U; (c) wife smoked but not co-worker,-11.2: (d),both wife and co-worker smoked, 1.7. No result was significant. In the studti by Butler which, in the AHSMOG eohort„related heart disease risk to the number of years lived& and the number of years worked with a smokcr, some 'suggestion' or 'in- dication' of an effect was reported in both sexes, but no, detailed results were reported. No other study provided information on other indices of exposure. In the study by He, relative risks of I, 1.83, 3.07 and' 5.49 were reported in relation to 0, I-10,, 1 1-20: and' 21+ years of ETS exposure, and relative risks, of I', 1.54, 2:30„5.07 and 12.67 were reported in relation,to 0; 1-199, 200-399, 400-599 and 600+ cigarette-years of smoking by the husband. Both, trends were statistically signif cant (p < 0.01). t,eaving aside the Garland study for reasons noted in~ subsection (iv) above, 7 studies provided some information on the extent to which adjust- ment for various risk factors affecte6the estimates of heart disease risk in relation to spouselhousehold smoking. The results are summarized in. Tablc 5.3. Two main conclusions can be drawn from this tabl'e. First. that in some studies age adjustment made a substantial difference to the relative risk. This effect, which would depend on the design of the study and on the frequency of smoking by age and sex in the eountn concerned, is evident in the Hirayama study and is also probably a contributor to the lgrge asso- ciation reported in the Sandler II study. The second main conclusion is that there is no clear effect from addi- tional adjustment- for the classical coronan• risk factors. Thus, while the Hole, Humble 11 - and perhaps the Svendsen - studies showed some increase in relative risk after adjustment, the He study showed a substan, tial' decrease. Some of the prospective studies cast more light on the possibility of confounding by various risk factors, since they present data comparing exposed and non-exposed women at the start of the stud'\. The results are 2023,512224

192 Environmental Tobacco Smoke and Morsality. Table 5.4 Comparison of heart disease risk facton in tTSexposed and non-FLS-exposed subjeets Study Sex Risk factor Spouselbousehold Signifi« exposure nnce no yes Sandier 1'J F School gradc(1,2+) (%) 33:8 35.4 p < 0.1 Housing index (&-10), (%) 83,3' 81.1 p < 0.01 M School'.grade (12+) (%) 38.0 43.8 p < 0.01 Housing index ($-I0), (96) 83.8~ 80.4 p < 0.05 Garland F Years of marriage Mean 36.01 34.2 p < 0:1 Systolic blood pressure Mean 140.1 138_2 NS Obesity index Mean 3.50 3.43 NS Plasma cholesterol Mean 225.7 226.7 NS Svendsen M Diastolic blood pressurt Mean 103.1 I'.03:3 NS Systolic blood pressure Mean 150.8 152.3 NS Serum cholesterol Mean 264.4 266.0 NS HDL eholesterol! Mean 42:7 43.4 NS LDL cholesterol Mean 167.1 166,5 NS Weight (Ibs) Mean 190.4 194.6 p < 0.05 ~ Drinkslweek Mean 7.6 9.7 p < 0.01 Education (years) Mean 14.2 13.8 p - 0.05 lncome (S 000), Mean 22.3 22.1 NS HDL - High d'ensity lipoprotein; LDL - low density lipoprotein. summarized in Table 5.4. Significant differences were seen in respect of weight (ETS-exposed heavier), drinks per week (ETS~exposed drink more), housing index (ETS-exposed~worse);,and years of ed'ucation (ETS-exposed more in Sandler 11, less in Svendsen), but is not clear that these differences were large enough to cause substantial bias. The Chinese case-control study of He also reported differences in blood fat and apolipoprotein levels according to ETS exposure, but did not, attempt to adjust for these in the analysis. One risk factor which might be relevant„but which was not investi- gated, was the number of cohabitants. In the Sandier 11 and Hole studies the index of ETS exposure seemed b.+ its very construction to be correlated with the number of cohabitants. In particular, the Sandler 11: study would ®

5 Heart Disease 1'93 Table 5,5 Effect of other variables on estimated relative risk of hean, disease in never smokers aaoording to spnuse smoking Study Sex VariableJlevel Relative nisk' (95Y% limits) Factors adjusted for Hirayana F Age of wife: None 40-49 0.87 (0.47-1.62). 50-59 L03 (0.72-1.46) 60+ 1.30 :(0.9$- L.73) Age of husband: Nbne 40-49 1,.34' (0,74-2.42) 50-59 1.25 (0.81-1.92) 6U+ 1,07 (0.83-1.40); Occupation: agricultural'worker I 1.32 (0.99-1.74) Age of husband other 0.99 (0:72-1.35) Humble Il F Blacks 1.78 (0.86-3.71) Age, BP; chol,, High social status Whites 1.97 (0.72-5.34) BMI Low social status Whites 0.79 (0.32-1.96) BMI - Body mass index; BP - blood pressure; chol - cholesterol: I For wives whose husband smoked cornpared to t,hose whose husband did not smoke. have included all people living on their own in the non-exposed group, and both studies were very likely to include people living in homes with many occupants in the exposed group. Since household size may correlate with many facets of disease, it seems to be a statistical error not to adjust for it in analysis. Two studies provided some information on variation in relative risk according to the level of some risk factors. Results are summarized in Table 5.5, and show how the association with ETS exposure varies by'age and occupation in the Hirayama study and by race and social, status in the Humble 11 study. Although the association with spouse smoking is evident only in agricultural workers in the Hirayama study and only in Blacks and high social status Whites in the Humble 11 study, there is in fact no significant heterogeneity between the relative risk estimates in either study. ~a2;~~1~126

194 Environmental Tobaeao Smoke and Morulily 5.4 Discussion and Conciusions. Although lung cancer is a rare cause of death in those who have never smoked, heart disease is not, and it is much easier, to conduct an ade-' quately large study for heart disease than for lung cancer. Yet there are far more studies of ETS and lung cancer than of ETS and heart disease. Ut is striking that so many of the latter studies are based on very smallinumbers of deaths or cases and/or have not been properly reported in the litera- ture. Only 2 studies are sufficiently large to pick up a moderate increase in risk as statistically significant„and neither is satisfactory: Both lack data on classical heart: disease risk factors, such as blood pressure„choltsterol and body mass index, and both have a number of problems that have been referred to in detail earlier. Certainly, neither is a straightforward prospec- tive study conducted acevrding to acceptable methodology, with collection of risk factor data at intervals and essentially complete follow-up of deaths. Apart from the generdlly unimpressive nature of the studies that have been oonducted, the other circumstance that stands out is that 13 of the 14 sex-specific estimates of relative risk of heart disease for spouse or house- hold exposure imTable 5.1 show a positive (though for the most part not statistically significant) association. !n considering this fact, a number of points have to be taken into account: (i) Active smokers have an increased risk of heart disease, as is clear from numerous epidemiologicalistud'ies. However, the relative risk is much lower than it is for lung cancer. For example, the 1989 US Surgeon General's report [711j cites results from the latest American Cancer Society prospective study showing that, compared with, never smok- ers, current smokers have relative risks of heart disease of 1.94 in malesand 1.78 in females„as compared to relative risks of lung cancer of 22.36 in males and 11'.94 in females. Vapour phase components of cigarette smoke have been implicated in the aetiology of heart disease (rather than particulate phase compo- nents for lung cancer (1j), and the relative exposure of ETS-cxposed non-smokers as compared with active smokers is substantially higher for vapour phase than for particulate phase components, but the low relative risks for heart disease for active smoking suongly suggest that if ETS does increase risk of heart disease this increase would be quite modest. The six estimates of over 1.5 in Table 5.1 seem difficult to

5 Hurt Disease 1955 rtooncile with the dosimetry, especially bearing in mind that active smokers have very substantialiETS exposure. (ii) Heart disease is certainly multifactorial, and many of the risk factors have not been taken into account in many of the studies. Confounding is therefore a possibility, particularly in the 2 studies (Hole,, Sandler, Il), where the index of ETS exposure used was likely to be extremely, strongly correlated with household size. Whilt the evidence discussed' in Tables 5.3 and 5,4 does not clearly derttonstr;Lte important con- founding by blood pressure„cholesterol or body mass index it is rather limited and somewhat inconsistent. More evidence is clearly needed on this important potential source of bias. (iii) Bias due to misclassil••ication of active smokiiig status is likely to occur, but since the increase in risk in relation to active smoking is relatively so much less for heart disease than for lung cancer, the extent of the bias will be that much smaller. Since the bias is proportional to the excess risk (see section 3.4.9), its magnitude will be only 5-1096 of that illustrated in typical situations for lung cancer. (iv) Publication bias is one major source of bias that can certainly not be excluded as relevant. There are two major reasons for believing this may be an important issue. First, there is a strong tendenc)• in Table 5.1 for the large relative risk estimates to be based on very small, studies. From, the 13 sex-specific estimates, the rank correlation is highly significantly (p < 0.05) negative. Who would bother to try to publish a paper showing no association based on very few deaths? Sec,ondly,,certain studies that couldipublish findings have not done so. Of particular importance is the fact that the first American Cancer Society study of over a million men and' women, which published results for ETS and lung cancer in 1981, has never published results for ETS and heart disease. 11 is very likely that no association was found. If this were so; it would have a very large effect on the results of any meta-analysis (or con- sequent estimate of heart disease deaths 'due to ETS'). Mainly because of the problems caused by the strong likelihood' of severe publication bias, it cannot be concluded from the existing evidence that ETS is associated with heart disease. The present author understands that the American Cancer Society intends to publish within the next year ~ or so findings related to ETS based on its second large prospective study: lt ~ is hoped that results from its first prospective study will also be released. r~ Until there is such evidence, and hopefully also evidence from other stud- W ies involving substantial numbers of deaths from heart disease with good ~ ~ ~ ~ ~

I 196 Environtnental Tobacco Smoke and Mortality control of confounding and with evidence on ETS exposure from sources other than the spouse or in the home, it is certainly premature to comc to any conclusions. Note Added in Proof Since this section was completed. Dobson et al. (166J, reported results from an Australian easecontrol' study of myocardial infarction and suddcn death.,Among non-smokers there was no positive relationship of risk to ETS exposure at work in either sex. Nor was therc a positive relationship of ETS exposure at home in males. In contrast, in femalbs a signifi- nnt positive relationship of risk was rtported to ETS exposure at home. In this study, data on smoking habits were collected by completely difierent methods for ea.ses and oontroas, the potential of bias being underlined by the wide variation in smoking fre- Quenay reported in controls according to how and where the data were collected. In addition, virtually no relevant confounding variables were taken into aeeount,.