Tobacco Documents Online

AMERICAN Journal of Epidemiology Fwnw +h Al¢RIG1N )O(JLWAL OF Hy!GgNB O 1987 by The Jomns Hopkina Univenity Schooi of Hygiene and Public H.a/tb' roR VOL. 126 NOVEMBER 1987 NO. 5 Original Contributions Y ! EFFECTS OF PASSIVE SMOKING IN THE MULTIPLE RISK FACTOR !, INTERVENTION TRIAL KENNETH H'. SVENDSEN,' LEWIS H. KULLER,' MICHAEI: J. MARTN' Avn JilIDITH K. OCKENE' Svendsen, K. H. (Coordinating Centers for Biometric Research, U. of Minnesota, Minneapolia, MN 55414), L H. Kuller, M. J. Martin, and J: K. Ockens. Effects of passive smoking in the Multiple Risk Factor Intervention Trial. Am J Epfdernlol 1987;126s763-95.. The Multiple Risk Factor Intervention Trial (MRF1T), conducted in 1973-1982, provided a unique opportunity to study the effect of passive smoking on men whose wives smoke. MRFiT participants who nsported at entry tfW they had never smoked tobacco products were classified according to the smoking status of their wiwes. Men with wives who smoked had similar mean levels of serurn thiocyanate (54.3 vs. 53.9 µmol/liter, p = 0.83) but higher mean levels of expired urbonmonoxide (7.7 vs. 7.1 ppm, p = 0:001). Lower levels of pulmonary function (by maximum forced expiratory volume in one second) were also observed in these men 1 (3,493.1 vs., 3,591.9 ml, p = 0.04). The relative risks, for men whose wives smoked compared with men whose wives did not smoke, for the endpoirtls coronary heart disepse death; fatal'or nonfatal coronary heart disease event, and deathfrorn any cause were 2.11 (p = 0.19, 95% ' confidence Interval (CI) 0.69- 6.46); 1.48 (p = 0.13, 9594 Cl 0.89-2.47), and 1.96 (p = 0.08, 95% Cf 0.93-4.11), respectively.lMhen smokers who quit prior to entry were included in the analyses, the relative risks, for men whose wives smoked compared with men whose wives did not smoke, for the above endopoints were 1:45 (p = 0.25, 95% C10.77-2.73), 1.19 (p = 0.29, 95% CI 0:85-1.65), and 1.72 (p = 0.01, 95% CI 1.12-264), respectively. These relative risk estimates did not change approckbfy after adjusting for ofher baseline risk factors. The results suggest that passive expo- sure to cigarette smoke may have a deleterious Impact on the health of non- smokers and that nonsmokers may be at an incrsased risk of death 1Maqh passive exposun to cigar.tte smoke. eoronary disease; ',tiobacco smoke poQuUon x: Passive smoking is defined as exposure ing from another person's tobacco smoke. of an individual to the air poilution result- The products of tobacco smoke are divided Received for publication September 3; 198fi; and'in one eecond; MRFIT, Multiple FWk Factor Interven- final form January 21, 1987: tion Trial: Abbreviationa: FEV;, forced eryiratory volume in ' Coordinsting Centers for Biometric Research, 783

I 784 SVENDSEN ET AL into two components. Those directly ex- haled by the smoker are called mainstream smoke, while those from the lit end of the cigarette, cigar, or pipe which are dis- charged into the environment are referred to as sidestream smoke. The composition of sidestream smoke (1) differs substan- tially from that of mainstream smoke, de- pending upon the different temperatures at which the substances burn and' the avail- able oxygen supply. Particulates, for ex- ample, are about 10 times greater in main- stream~ smoke than in sidestream smoke. After inhalation, sidestream smoke prob- ably reaches the more distant alveolar spaces in the lung (2). Sidestream smoke also contains much more free nicotine in the gas phase, generates more carbon mon- oxide (1)„ and contains mu& higher con- centrations of the reduced products of ni- trogen including several highly carcino- genic substances (3). Most environmental tobacco smoke is from sidestream smoke, and only a very small amount is from ex- haled mainstream smoke. Environmental exposures to tobacco smoke depend on the number of smokers im the area and the amount they smoke, the size of the area, and the ventilation rate. It is now an accepted fact that cigarette smokers have an increased risk of many diseases. In recent years, there has been~ a growing concern that nonsmokers exposed to environmental tobacco smoke may also be at increased risk of certain diseases, especially cancer, chronic obstructive pul- monary disease, and, possibly, heart' dis- ease. Friedman et al. (4) reported that 63.3 per cent of adults were ezposed' to passive Di.vion of Biometry ;, School of Public Health. Uni- venity of Minnesota. Minneapolis, MN. ' Graduate School! of Public Health, University of Pittsburgh, Pitt.burgh. PA. ' Clinical Epidemiology Program, San FYanciaco General Hospital, San Francieco, CA. ' Department of Preventive and Behavioml Medi- eine, University of Massachusetts Medical' Center, worceeter. MA. Reprint requests to Kenneth H; Svendben, Coor- dinating Centers for Biometnc Researeh, Suite 508, 2829 University Avenue S.E., Minneapolis, MN'55414. smoking for at leastone hour per week. A higher percentage was exposed away from home, usually at work. Repace and Lowrey (5) have estimated that the exposure to environmental tobacco smoke of the non- smoking adult population was about 1.43 mg of tar per day. A cigarette smoker, on the other hand, can be expected to inhale about' 420 mg of tar per day (14 mg of tar per cigarette for an average of 30 cigarettes per day). Thus, the dose from passive smok- ing is much less than the dose from ciga- rette smoking. Studies on passive smoking reported to date have depended on self-report.e&histo~ ries of environmental tobacco smoke ex- posure. A workshop on the respiratory ef- fects of enviro~ ental tobacco smoke in 1983 sponsored'by the Division of'~ Lung Diseases at the National Heart, Lung, and Blood Institute (6)! noted that'~ lack of objec- tive measures of dose or exposure, con- founding variables, methods of statistical analysis,, and quantificat'ion~ of other vari~ ables were major concerns in the evaluation of current and future studies. Participants in the Multiple Risk Fact~)r Intervention Trial (MRFIT) (7)! offered an unusual opportunity to study the effect of environmentali tobacco smoke on, men„ es- peciallyy inithe home. Objective measures of cigarette smoking behavior, as wellias other critical risk factors for cardiovascular and other diseases, were carefully monitored in a large population followed for an average of seven years. Fortuitously, at entry into the study, prior to randomization, a de- tailed smoking history was obtained for each of the participants subsequently ran- domized. This history included not only their own smoking history but also that of their wives, family members, and cowork- ers. This trial, to our knowledge, is the first longitudinal study that was able to objec- tively define the participants' smoking sta- tns and possible exposure to environmental tobacco smoke. The study design was also unique because the index subjects were men who did not smoke and who were at high risk of heart disease, and the exposure in- ~

I I I EFFECTS OF PASSIVE SMOKING dex was the smoking behavior of their wives.. MATERIALS AND ME'I'HDDS' The Multiple Risk Factor Intervention Trial was a primary prevention trial de- signed to test the effect of a multifactor intervention program on mortality from coronary heart disease. The design of the MRFIT has been de- scribed (7). Briefly; men aged 35-57 years were recruited in 18' US cities. They were screened to select those in the upper 10-15 per cent of' a risk score distribution deriveti from Framingham data, based on serum cholesterol concentration, cigarette smok- ing, and diastolic blood pressure. Those free of overt' coronary heart disease by history and resting electrocardiogram who con- sented to participate were randomized to either the special intervention or usual care groups. After randomization~ special inter- vention men participated in an intensive intervention program aimed at lowering blood cholesterol' by nutritional means, eliminating cigarette smoking through ed- ucation and behavior modification tech- niques, and reducing the diastolic blood pressure of those who were hypertensive primarily by using a stepped-care drug reg- imen. Usual care participants were referred to their customary source of medical care with information on their risk factor status but with no~adviee as to intervention. Both special intervention an& usual care partic- ipants were seen annually over six to eight years for risk factor measurement and a medical ezamination. A detailed smoking history was obtained' from all participants during screening and'at each annual'visit. This paper focuses on the effects of pas- sive smoking on participants who reported that they did not smoke cigarettes, pipes, cigars, or cigarillos prior to randomizationn into the trial. Most analyses are restricted to men who had never smoked cigarettes. Endpoint results are shown for never smok- ers and~ all nonsmokers at entry; non- smokers included never smokers and ez- smokers who quit prior to entry into~ the 785 study. Data on the smoking habits of the participants' wives were collected at base- line for participants who smoked and those who di& not smoke. The smoking status of the wife is used as an index of passive smoking exposure for the men who di& not smoke. Only a limited' amount of informa- tion was collected about exposure to to- bacco smoke on the job. Participants were asked the smoking status of their cowork- ers. The results of all'~ analyses presented are for the special intervention and usual care groups combined. Separate analyses for each study group yielded similar results. Measurernents of smoking exposure Serum thiocyanate was measure&duringg screening and at each annual visit. In the planning stages of the MRFIT, it was rec- ognized that special intervention partici- pants who were repeatedly urged to stop or reduce smoking cigarettes might be more likely to misreport their cigarette smoking status than usual care participants. Serum thiocyanate is elevated in smokers because of the cyanide present in tobacco smoke which is metabolized to thiocyanate. The half-life of serum thiocyanate is approzi- mately 14 days, reflecting long-term expo- sure to cigarette smoke. At the th'srd~ and' sixth annual examina- tions, expired air carbon monoxide was measured, using an ecolyzer (series 2000, Energetics Science, Inc., Elmsford, NY), which permitted~ a visual meter reading on a 0-104 parts per million (ppm) scale. The levels of expired air carbon monoxide are directly related to carbozyhemoglobia in the bloodi The half-life of elevated carboxy- hemoglobin levels after exposure to envi- ronmental carbon monoxide is only two to four hours; thus, its measurement reflects only very recent exposures. Other factors, especially any incomplete combustion of carbon-containing substances, can increase environmental carbon monoxide levels and blood carbozyhemogiobin levels. Pulmonary function testing was con- ducted at screening and at each annual ezamination using a 10-L Stead Wells

I 786 svErDsEh Er A.L water-filled spirometer (Warren E. Collins, Inc., Braintree, MA) i The forced expiratory volume in one second (FEV1) is defined as the voltime of gas exhaled over an interval of one second, with ezpiration as rapid and as complete as possible. The selection of tracings for analysis was based on carefuli quality control standards defined prior to the current' analyses. The maz.imum of three to five measurements meeting quality standards (maximum FEVI), adjusted for age and height, is used to quantify pulmo- nary function in this paper. The quslity control procedures and measurement tech- niques are described in detail elsewhere (8). Endpoints Classification of cause of death was. per- formed by a committee of three cardiolo- gists who were unaware of treatment as- signment (special intervent'ion f usual care) or passive smoking status. They used hos- pitalrecords, physicians' reports, nert-of- kin interviews, death certificates, and au- topsy reports, when available. Coronary heart disease deaths were subclassified as 1) documented myocardial infarction; 2)) sudden death within 60 minutes, or be- tween one and 24 hours of symptom onset, without documented myocardial infarction; 3)' congestive heart failure due to coronary heart disease; or 4) death associated with surgery for coronary heart disease. Resultss are aiso presented for the endpoint fatal' or nonfatal coronary heart disease event. Thiss endpoint includes coronary heart disease death, serial change from baseline on a resting electrocardiogram, and/or docu- mented evidence of myocardial infarction from a review of hospital records by a panel of physicians (9). Statistical methods Differences in baseline characteristics and changes in risk factor levels from base- line to the sixth annual examination for men who did not smoke who had wives who smoked'versus men who did not smoke who had wives who were also nonsmokers were tested for statistical significance using the Student's t test (two-sided) or the 2 x 2 chi-square tesL. For comparison of mea- sures of smoking exposure between the two groups, mean levels of thiocyanate and the maximum FEV, were calculated for base- line and the average of baseline and all foll'ow-up visits. The latter results in im~ proved precision but smaller sample size. The maximum FEV, means were adjusted for age and height by analysis of covariance. Mean levels of expired air carbon monozide were calculated for year 3 and the average of years 3 and 6. Differences in the means between the two groups for thiocyanate and expired air carbon monoxide were assessed by the Student's t test. Differences in the adjusted means for maximum FEV, were assessed by analysis of covariance. Tests for a dose effect of smoking exposure were performed using regression models with number of cigarettes smoked per day re- ported by wife as an independent variable. Relative risk estimates, for men whose wives smoked compared with men whose wives did not smoke, for the endpoints death from any cause, coronary heart d+.s- ease death, and fatal or nonfatal coronr: i, heart disease event were calculated using the Cox proportionali hszards model (10)) with Breslow's approximation (11)~ Results are shown both unadjusted and adjusted for age,, baseline blood pressure, cholesterol, weight, education (as a measure of socio- economic status), and drinks per week. RESULTS Sample size There were 1,4001 of 12,866 randomized' participants who reported that they ha& never smoked cigarettes, pipes, cigars, or cigarillos at entry into the MRFIT. Of these never smokers, 1,245 were married; 286 to women who smoked and 959 to women who ~ did not smoke (table 1). Q Comparabiiity of neuer smokers by smoking N status of~ui(e ~ Baseline characteristics of these 1,245 }.i men by smoking status of wife are sum- ~.i FV

EFFECTS OF PkS51'VE SMOKING 787 marized in table 2: The two groups of men are similar with respict to age, blood pres- sure, and cholesterol. The average weight for men with wives who smoked was 4.2 pounds greater than that of men whose wives did not smoke (p <0.01)'. Men whose wives smoked consumed an average of 2.1 more alcoholic drinks per week (p < 0:01) and had' 0.5 years less formal' education TAecs 1 than men with wives who did not smoke (p Frequency d;.,rrtbuuion of smoking status at encry- < 0.05). Income was similar between the Multiple Risk Factor Intervention Trial, 1973-/982 groups. Table 3 shows risk factor changes Smokers' 9,244 71.8 Ea-smokers 2,222 17:3 Never smokers 1,400 10.9 Not married 155 1.2 Wife a nonsmoker 959 7.5 Wife a smoker 286 2.2 Total 12:866 100.0 • Includes smokers of cigarettes, pipes, cigars, or cignrillos. and the percentage of men prescribed anti- hypertensive medications at the sixth an- nual examination by smoking status of wife. There were no statistically significant differences between the two groups. Comparisons of smoking exposure Mean serum thiocyanate levels at base- line and the average of baseline and all annual' follow-up visits are shown in table TAat.l: 2' Mean values of selected variables at entry /or 1.245 men urho reported never smoking ci8arrertes„pipes, cigars, or cignrillos, by smoking status of mi(e at entry: Multiple Risk Factor Intervention Trial,' 1973-1982 Smking status of wife 95% Smoker (n - 286) Nonsmoker (n - 959) Differsooe' ooebdence intsrval Age (years) 1 47:4 47.5 -0.2 -1.0-0.6 Diastolic blood preeaue (mmHg) 103.3 103,1 0.2 -0.4-0.9 Systolic blood pressure (mmHg) 152.3 ' 150~8 1.5 -0.4-3.4 Serum cholesterol (mg/dl)' 266.0 264.4 1.6 -2.3-0.5 High density lipoprotein cholesterol (mg/dl) 43.4 42.7 0.7 -0.7-2.0 Low density lipoprouin cholesterol (iag/dl) 166.5 167.1 -0.6 -5.0-3.9 Weight (lbs)' 194.6 190.4 4.2 0.6-7:8 Drinks/week (n) 9.7 7:6 2.11 0.8-3.3 Education (years) 13.8 14.2' -0.5 -0.9-0.0 Income (1,0005) 22.1 22:3 -0.1 -1.4-1.2 ' Difference may not agree because of rounding. TAata 3 Mean change in selected uariables (sixth annual minus baseline uanunation) for men who reported never smoking cigaretter, pipes, cignrs, or cigariLds, by smoking stahu of wife at entry: Multiple Risk Factor Intertxntion Tri4 1973-1982 Smoking status of wife 96% Smoker (n -266) Nonsmoker (n - 889) Difference oonfideaa intervalI Diastolic blood pressure (mmHg)', -10.1 -9.9 -0 3 -1.7-1.1 Systolic blood pressure (mmHg) -12.6 -13.6 1.1 -1.1-0.2 Plasma cbolesterol (mg/dl). -11.4 -11.0 -0!4 -4.7-3.9 High density lipoprotein cholesterol!(mg/dl) -1.4 -0.7 -0!7 -1.9-0.5 Low density lipoprotein cholesterol (mg/dl) -10.8 -10.4' -0!4 -4.4-3.7 Weight (lba) -2.2 -2.5 0!3 -1.4-2.0 Drinks/week (n) -2.7 -2.1 -0:6 -1.7-0.4 On antihypertensive medication (96) 66.5 62.5 4.0 -2.7-10.6

798 SVENDSEN ET Ai.. 4 by smoking status of wife. The mean thiocyanate levels are similar for the two groups, both at' baseline and averaged over aIli visits.. El pired air carbon monoxide was mea- sured at the third and sixth annual exami- nations. The average expired air carbon monoxide at the third annual examination for men whose wives smoked was 7.7 ppm compared with 7.1 ppm for men whose wives did not smoke (table 5). The differ- ence, 0.6, is statistically significant (1p = 0.001), as is the test for linear trend (p = 0.03). Similar results were obtained when the averages of the third and sizth annual' carbon monoxide measurements were com- bined. Men with wives who smoked had signif- icantly lower levels of puImonary function at baseline as measured by the maximum FEV, (table 6). The mean maximum FEVy is 3,493.1 ml for men whose wives smoked versus 3,591.9 for men whose wives did not smoke, a difference of about 100 ml. Similar results were obtained when averaging over all visits, although the difference between the two groups was not statistically signif- icant (p = 0.16). Endpoint results for never smokers Table 7 gives the event rates by smoking status of wife and table 8 shows the relative risk estimates (for men who did not smoke whose wives smoked compared' with those whose wives did not smoke) for the end- points death from any cause, coronary heart disease death, and fatal or nonfatal coronary heart disease event. TAB[s 4 Mean leveLs of rhiocyanate (Kmol/liter) ar baseline and average over all'virits for men who reported never snurking cigarerre.s, pipea, cigan, or eigarillos, by .xrwking statw of wife at entry. Mukiple Risk Factor fnteruention 7}iaf, 1973-1982 Ba.eline Ayerage over all visits Smok'in` .utus of wife - - n Mean n A;rr Nonsmoker 878 Smoker 264 1-19 cigerettee/day 125 220 cigarettes/day 139 53.9 704 51.G 54.3 212 52.3 54.0 102 51.6 54.6 110 52.9 Smoker/nonsmoker difference 0.4 (-3.7; 4.6)' 0.7 (-2.7, 4.0)', p value for linear trend 0.99 0.55 • 95% canfidence limits. TAat.c 5 men who Meon e~ire d air mrbon monazidt (pprn) at the thind anrwa! visir and averuge over all visiLr for reported'never smoking cigarettes, pipes, ci8ar:, or cigariUvs.,by smoking status of wife at entry` Multipk Risk Factor Intervention T}ia 1973-1982 Smkin~ ta:r o(wife T6ied annual .vit n Mun Nonsmoker 828 71 Smoker 244 7:7 1-19 cigarettes/day 112 7:7 220 ciQaretru/day 132 7:8 Smoker/nonsmoker difference 0.6 (0:2. 1.0)' p value for linear trend 0.03 ' 95:'u confidence limits. Avrrate wer al] ivi.itm n Mean 760 6.7 228 7:1 ~ 106 7.1 122 7:2 ~ 0.5 (0.2, 0:7)) <0.01 ~ W ~ T

EFFECTS OF PASSIVE SMOKING 789 3 8 ~ I I TAeix 6 Mean mazimum FE V, ,(m1) adjusted /or age and height at bcseline and average ouer a!l uiuiti for men who reported never smoking cigaretiei, p+pes, cigars, or cigarillos, by smoking status o( wi/e at entry: Multiple Risk Factor 1'ntervention Tria4 1973-1982 Smoking wtw of wife Baseline Average over alI visi4 n Mean n M.an Nonsmoker 514 3,591.9 257 3,491.3 Smoker 162 3,493.1 81 3,403.3 1-19 cigarettes/day 66 3,412.1 • 31 3,263.3 t20 cigarettes/day 96 3,548 8 50 3,489.0 Smoker/nonsmoter difference -98.9 (-192:4„-5.4)' -87.8 (-210.7; 35.2) p value for linean trend 0.52' 0.99' ' 95% confidence limits. TABLE 7 Number of deaths /rom, any rasse and from coronary heart disease and'jatal or nonfatal coronary heart disease events for men who reported'neuer smoking cigarettes, pipes, cigars, or cigarillos, by'smoking status of wi(t at entry:,Multip/e Risk Factor Intervention Tr~ 1973-1982 Smoking eutw No. Death from Coronary 'heart. of wife of men any uuse dia..ue death Fatal or nonfital coronary beart dise..e event Nonsmoker 959 19 (2.83)' 8'(1.19) 48'(7,28) Smoker 286 11 (5.55) 5(2.52) 21 (10.81) 1-19 cigarettes/day 133 3(3.21) 1(1.07) 8(8.70) z20 cigarettre/dny 153 8(7.65) 4 (3.82) 13(12.71) p value for linear trendY 0:08 0:04 0.20 ' Rates per 1,000 person-years. t From Cox proportional hazards regression using number of cigarettee smoked per day by.vife as a covariste. TAaLE 8 Relative risk estimates, , u i/e mho smoked compared with wife who did not smoke, and their 95 per crru eonjidence uuervaLs for men who reported never smoking cigarettes, pcpes„cigara, or cigari!los: Multiple Risk Factor Intervention Trial. 1973-1982 Endpomt Relatuve risk p value 95% rnnfidence interval Death from any cause Unadjusted 1.96 0.08 0.93-4.11 Adjustcd' 1.94 0.08 0.91-4.09 Coronary heart disease death Unadjusted 2.11 0.19 . 0.69-~&46 Adjusted 2.23 0.17 0.72-6.92 Fatal or nonfatal coronary heart disease eventi Unadjusted 1.48 0.13 0.89-2.47 Adjusted 1.61 0.07 0.9fr-2.71 ' Adjusted~ by Cox prnportionali haaards regression for age, baseline blood pressure, cholesterol, weigbt, diinks per week, and education. As of February 28, 1982„after an average married to nonsmokers (2.8 per 1,000 per- of seven years of follow-up, 11 of 286 men son-years). There is some suggestion of a married'to smokers had died (5.6 per 1,000 dose effect for the endpoint death from any person-years)! compared with 19 of 959 men cause, with 3:2 deaths per 1,000 person-

I 790 SVENDSEN ET AL years in the category wife smokes 1-19 cigarettes per day and 7.7 deaths per 1,000 person-years in the category wife smokes 20 or more cigarettes per day, although the test for a linear trend was not significant (p- 0.08). The numbers are small for the endpoint coronary heart disease deat'h, but they fol- low the same pat+tern as those for the end- point death from any cause. The coronary heart disease death rate is 2.5 per 1,000 person-years for those whose wives smoked compared' with 1.2 for those whose wives did not smoke. The test for a linear trend was significant (p = 0.04). Among men~ with wives who smoked; there were 10.8' fatal or nonfatal coronary heart disease event' endpoints per1,000 per- son-years versus 7.3 per 1,000 person-years for those whose wives did' not smoke. The event rate is higher for those whose wives smoked 20 or more cigarettes per day com- pared with those whose wives smoked 1-19 cigarettes per day, although the test for linear trend for the endpoint fatal or non- fatal coronary heart disease was noU signif~ icant. The relative risk estimates, for men whose wives smoked compared with men whose wives did not smoke, for the end- points death from any cause, coronary heart disease death, and fatal or nonfatal coronary heart disease event are 1.96 (1p = 0.08, 95 per cent confidence interval! (CI) 0.93-4.11), 2.11 (p = 0.19, 95 per cent C1 0.69-6.46), and 1.48 (p 6 0~:13,,95 per cent CI 0.89-2.47), respectively. These relative risks did not change appreciably after ad- justing for other baseline risk factors. Endpoint results for all nonsmokers. Table 9 presents unadjusted and adjusted relative risk estimates, for men whose wives smoked compared with men whose wives did not smoke, for the endpoints death from any cause, coronary heart disease death, and fatal or nonfatal coronary heart disease event for all nonsmokers at entry; non- smokers included' never smokers and e:- smokers who quit prior to entry into the study. For the endpoint death from any cause, the relative risk estimate is 1.72,, which differs significantly from 1.0 (p = 0.01, 95 per cent CI 1.12-2.64). For the endpoints coronary heart disease death and fatal or nonfatal coronary heart disease event, the relative risk estimates are 1.45 (p = 0:25; 95 per cent CI 0.77-2.73) and 1.19 (p = 0.29, 95 per cent CII 0.85-1.65), respectively. As with the analysis restricted to never smokers, adjusting for baseline risk factors did not change the relative risk estimates. Endpoint results by amoking exposure on the job Only a limited amount of information was collected'about exposure tb tobaceo '1'ASLE 9 Relatiue risk ertimates, mi/i who amoked'eompared with urite who did not smo/re, and their 95 per cent mnjidenett interuda for nonsmokera': Mukiple Risk Factor Intertrention 7}ial, 1973-1982 Endpoint Relative risk Death from any nuse Unadjueted 1.72 Adjusudfi 1.79 Coronary heart disese death Unadjusted 1.45 Adjusted~ 1.59 FataJ or nonfatal coronary heart disease event Unadjusted 1.19 Adjusted 1.32 p value 95% aonSdence intsrva! 0.01 1.12-2.64 <0.01 1.17-2.76 0.25 0:77-2.73 0.15 0.84-3.02 0.29 0.85-1.65 0.10 0.95-1.84 • Includes both never smokers and e:•smokers who quit prior to entry into the trial, t Adjusted by Cox proportional hazards regression for age, baseline blood preasure, drinks per week, educationj and past smoking history. cholesterol, weight, r

EFFF.G 715 OF PASSIVE SMOICING 791 I smoke on the job. The participants were asked the smoking status of most of their coworkers. Of 1,237 never smokers, 906 (73.2 per cent) reported that most cowork- ers were smokers, and 331 (26.8 per cent) reported that most coworkers were non- smokers. The relative risk for the endpoint death from any cause, for men whose co- workers smoked compared with men whose coworkers did not smoke, adjusted' for age and wife's smoking status is 1.2 (p = 0.63, 95 per cent CI 0.5-1.8). For the endpoint coronary heart disease death, the relative risk is 2.6 (p = 0.23; CI 0.5-12.7), and' for fatal or nonfatal coronary heart disease event, the relative risk is 1A (p = U6; CI 0.8-2.5). Because of the small number of deaths, the joint impact of a spouse who smoked an&coworkers who smoked was estimated only for the endpoint fatal or nonfatal cor- onary heart disease event. The risks for the categories wife and coworkers who smoked, wife who smoked and coworkers who did not smoke, and coworkers who smoked and wife who did not smoke relative to the category wife and coworkers who did not smoke are 1.7 (p = 0.14, 95 per cent CI 0.8- 3.6), 1.2 (p = 0.75, 95 per cent CI 0.4-3.7):, and 1.0 (p = 0.99, 95 per cent CI 0.5-1.9):, respectively; DlscusstoPt To our knowledge, this is the first longi- tudinal study of' the relation between pas- sive smoking and total and coronary heart disease mortality that has included mea- sures of other major risk factors, objective monitoring of smoking behavior in a well defined population at risk, and a careful unbiase& ascertainment and evaluation of causes of death. Our findings, which sup- port the hypothesis that passive smoking is associated with an increase in morbidity and mortality among nonsmokers, are dis- cussed below. Thiocyanate levels did not vary by envi- ronmental tobacco exposure. This finding is similar to that reported by Friedman et al. (4). In other studies, conducted~ in smok- ing chambers, a direct dose-response rela- tion between exposure to tobacco and the cotinine levels in saliva, urine, and blood was found (12).. Jarvis et al. (13) also found a positive correlation between urinary co- tinine levels and'self-reported ezposures to sidestream cigarette smoke. Similar find- ings using urinary cotinine were noted by Mat'sukura et al. (14) i and Wald et ali (15). In these studies, the differences in bio- chemical levels by environmental exposure were small compared with the differences between smokers and nonsmokers. For ex- ample, Wald et al. reported that the median urinary cotinine levels were 1,645 ng/m]!in cigarette smokers, 6 ng/m11 in nonsmokers exposed to environmental tobacco smoke,, and approximately 2 ng/ml in nonsmokers not so exposed. The increase in ezpired' air carbon mon- ozide resulting from passive smoking is rel- atively small even if statistically significant and in and of itself is of relatively little biologic significance. The increase probably reflects exposure to environmental tobacco smoke (16). The half-life of expired air carbon monoxide is somewhat short, around four hours. The men may have been exposed to their wife's tobacco smoke at home prior to going to the clinic for their annual examination or while traveling by car to the clinic. The differences in ezpired' air carbon monoxide or blood carboxyhe- moglobin levels may have been substan- tially greater immediately after exposure to environmental tobacco smoke. The differ- ences presented here also may be conser- vative because of the fact' that the smoking status of the participant's wife was avail- able only at baseline. By the time carbon monoxide was measured, some wives who were smokers may have quit, while others who were nonsmokers may have restarted. This type of misclassification would tend to decrease any observed difference in car- bon monozide. The health effects of exposure to low doses of carbon monoxide are not known~ at present. Earlier studies have reported' that individuals with cardiovascular disease

I t 792 SVENDSEN Er AL. (17, 18) have an adverse response to rela- tively low doses of environmental carbon monoxide. There has been controversy con- cerning these findings (19, 20), however, and the studies are currently being repeated in different laboratories. It is possible that transient elevations of carbon monoxide due to environmental tobacco smoke in high-risk individuals may be associated with an increased risk of heart attacks and perhaps cardiovascular deaths. The major- ity of sudden and unezpect,ed deaths in the community occur at home (21). The acute precipitant of many of these heart attacks is unknown but could relate to certain in- door air pollutants. Occupational studies (20) ~ of exposure to carbon monoxide and risk of heart attack have been equivoca] in their results, as have community studies of the relation between ambient carbon mon- oxide and coronary heart disease mortality (22). . There have been a few studies of pulmo- nary function and' exposure to passive smoking among adults (23-28). Three stud- ies in the United States (23), France (24),, and Holland (25) have demonstrated de- creased pulmonary function among pas- sively exposed individuals, with usuallyy about a 1OO- ml difference in FEVI, betweenn the passively exposed compared with the nonexposed nonsmokers. A study in Ha- gerstown, Marylan& (26), noted that 5 per cent of nonsmoking men not passively ex- posed and 7.1 per cent of those passively exposed had FEV, less than 80 per cent predicted (relative risk of 1.4). The relative risk was not statistically significantly dif- ferent from one. Forty families were iden- tified in a study of three communities in the United States in which the mother was a smoker and the father a nonsmoker (27). There was a statistically significant de- crease in the mean residual FEVI for the fathers married to women who smoked compared with those married to women who did not smoke. The effect was, how- ever, substantially reduced when tha ez- smoking men were excluded. A recent re- port from the Federal Republic of Germany (28) also failed to demonstrate any effect of passive environmental tobacco smoke on pulmonary, function among a rel atively young occupationai; cohort. There was also no apparent effect from direct cigarette smoking on~either the forced vital capacity or FEVy. Lebowitz et al. (29), in several studies in Arizona, have been unable to demonstrate any effect of environmental tobacco smoke on pulmonary function among adults who do not smoke. The approximate 100-m1 differences in the FEV', at baseline as noted' in table 6 are consistent with those of several of the other larger studies previously discussed (23-25). It is unlikely that the relatively small dif- ferences in pulmonary function in our study can contribute substantially to chronic ob- structive pulmonary disease or disability. It is possible, however, that there is a subset ofind'ividuals in whom a hypersensitivity to environmentali tobacco smoke causes further progression of pulmonary disease and disability: The excess total' and coronary heart dis- ease mortality and morbidity amoni, MRFIT men who were exposed to environ• mental tobacco smoke is further evidence of a potential! serious health risk for a large segment of the nonsmoking population. Inn the MRFIT study, 23 per cent, of the men who did not smoke were exposed at home to the environment;al tobacco smoke of their wives (table 1). As noted, a study by Friedman et al. (4) has suggested that up to two thirds of nonsmokers are exposed to environmental tobacco smoke. At present, the number of cancer deaths in this study is too small to allow any evaluation of the relation between environmental tobacco smoke and specific cancer and other causes of death. Other studies have evaluated the relation between environmental tobacco smoke and ~' lung or other cancers. 1*learly al1 the cancer ~' studies have been case-control studies (30- N' 36). The cases have usually beea lung or W other cancers and the controle either hos- ~ pital patients, community residents, or ~ friends of the cases. Practically all the stud- ~ ~ ~ r