An increased risk for coronary heart disease among nonsmokers related to environmental tobacco smoke (ETS) exposure has been reported in both cohort 1–7 and case-control 8–13 studies, but important aspects of the association are still poorly understood. Several mechanisms by which passive smoking may increase the risk for coronary heart disease have been suggested, such as increased platelet aggregation, increased blood carbon monoxide levels, nicotine-induced hypertension, and accelerated atherosclerosis due to exposure to polycyclic aromatic hydrocarbons. 14–17 Only a few of the epidemiologic studies on ETS have investigated myocardial infarction (MI) specifically. 1,8,10,11,13 Although current exposure has shown greater effect estimates than previous exposure in several studies on ETS and MI, no trend analysis regarding risk reduction related to years since last exposure to ETS has been reported. Previous studies have adjusted for age and gender in the analysis, and additional corrections for other risk factors, including body mass index (BMI), hypertension, diabetes mellitus, and family history of coronary heart disease, have been made in some studies. 18,19 Nevertheless, diet and psychosocial factors have usually not been taken into account, leaving open the possibility of residual confounding.
Although both spousal and occupational ETS exposure have been investigated, the effect of combined exposure from different sources of ETS has been studied only sporadically. 1,13 Generally, studies of ETS and MI have examined average amount, duration, and current vs previous exposure. In studies of ETS and lung cancer, different dose metrics have been used in an effort to describe a relevant long-term biological exposure in various settings, including cumulative amount of exposure (“pack-years”) and weighted duration, for example, “hour-years” (1 hour per day for 1 year) of exposure. 20 Only one study has investigated the association between MI and cumulative ETS exposure using pack-years, and none has used hour-years as exposure metric. 6 Moreover, little is known regarding a possible interaction between ETS exposure and other risk factors for MI.
We investigated the relation between ETS exposure and MI among male and female never-smokers in a large Swedish population-based case-control study (SHEEP–Stockholm Heart Epidemiology Program). In particular, we aimed to study the effect of combined exposure in residential and occupational settings, intensity of exposure, duration of exposure, time since last exposure, and possible interaction with other risk factors.
Subjects and Methods
The SHEEP study has been described in detail elsewhere. 21 In brief, the study comprised all nonfatal and fatal first events of MI among Swedish citizens age 45–70 years who resided in Stockholm county during 1992–1993 (1992–1994 for women), and population controls from the corresponding study base. The upper age limit was set at 65 years during the first 10 months of subject recruitment and 70 years for the rest of the enrollment period. Cases were identified from the coronary and intensive care units at the departments of internal medicine at all of the emergency hospitals in Stockholm County, the Hospital Discharge Register for the county, or death certificates from the National Cause of Death Register at Statistics Sweden. The diagnostic criteria for MI used to determine case inclusion were those applied by the Swedish Association of Cardiologists. 22 They require at least two of three conditions to be met regarding certain symptoms, specified blood enzyme changes (creatinine kinase and lactic dehydrogenase), or specified electrocardiogram changes. In addition, myocardial necrosis detected at autopsy that could be related to the time of disease onset was also included. One control per case, matched on gender, age, and hospital catchment area, was randomly selected from the study base within 2 days of the inclusion of a case. All controls were initially checked for previous MI and were alive when recruited, regardless of the vital status of the corresponding case. In total, the SHEEP study included 2,246 cases and 3,206 controls. The questionnaire response rate among cases was 72% for women and 81% for men, and the corresponding figures among controls were 70% and 75%. The subjects responded to the same extent in different age groups and were equally inclined to participate from the different catchment areas. In the present study only nonfatal never-smoking cases (N = 334) and never-smoking controls (N = 677) were included because of the potential inaccuracy in exposure information from relatives of fatal cases, especially regarding ETS exposure at work.
All study subjects received a postal questionnaire covering a large set of potential risk factors for MI, including physical and psychosocial work environment, social factors, different lifestyle factors, dietary intake, and ETS exposure. A supplementary telephone interview was conducted to fill in missing data. A special health examination was also carried out on hospitalized cases and their controls to collect data on various biological parameters related to cardiovascular disease. The biological variables for analysis were primarily based on data from the health examination, but some (for example, BMI) were complemented with questionnaire data for subjects not participating in the clinical testing.
Socioeconomic status was determined from questions regarding professional background according to a Swedish socioeconomic index classification of occupations. 23 A three-level variable was constructed comprising blue-collar workers (manual workers, for example, unskilled or skilled employees in service or goods production) and white-collar workers of either lower level (assistant nonmanual employees with or without subordinates) or intermediate to upper level (for example, professionals and executives). The most recent working period before inclusion in the study was considered for the socioeconomic classification, and we ignored periods of undefined employment (for example, self-employed farmers and former self-employed disability pensioners with no record of company size, number of employees, or agricultural capacity). Overweight was defined as BMI above 28 kg/m2, which corresponds roughly to the 75th percentile for all controls (27.5 kg/m2 for men and 28.1 kg/m2 for women), calculated from height and weight measurements from the health examination (91.4%) or from values reported in the questionnaire. Job strain was measured in accordance with the Swedish version of the demand-control concept based on the Karasek-Theorell questionnaire. 24 We calculated the ratio of the sum of the scores of the five questions on psychological demands and the sum of the scores of the six questions on decision latitude for the last 5 years before inclusion in the study. Subjects with a ratio above the 75th percentile among all controls in the SHEEP study (men and women combined) were classified as being exposed to job strain. 21 Hypertensive subjects were defined as those fulfilling at least one of the following criteria: receiving antihypertensive drug therapy when included in the study, having a history of regular antihypertensive drug therapy during the last 5 years, or having a systolic blood pressure ≥170 mmHg or a diastolic blood pressure ≥95 mmHg and no history of antihypertensive drug therapy during the last 5 years. An extensive food frequency section of the questionnaire provided data on consumption of fiber, saturated fat, and polyunsaturated fat. The dietary variables, expressed in grams, were calorie adjusted with the energy contribution from alcohol excluded, for men and women separately, using the residual method. 25,26 To gain statistical stability, the food frequency answers from the whole study population were used (N = 3,980), excluding those with extreme nutrient calorie intakes (men <600 and >4,000 kcal, women <600 and >3,500 kcal). The dietary variables were dichotomized at the 75th percentile according to the distribution among all never-smoking control subjects. Diabetes mellitus was defined either from the questionnaire, as those who stated they were controlling diabetes with insulin, drug treatment, or diet control at the time of inclusion in the study, or from the health examination, as those with a fasting blood glucose level above 6.7 mmol/liter. Family history of coronary heart disease was defined as having at least one biological parent or sibling who had suffered from or died of coronary heart disease before the age of 65 years.
Subjects who had never smoked regularly for at least 1 year were considered never-smokers. For ETS, spousal exposure was defined as having lived with a smoking spouse or cohabitant. The number of cigarillos and cigars smoked by a spouse was converted to cigarette-equivalents by multiplying by 4. Intensity of spousal ETS exposure was expressed as the average number of cigarette-equivalents per day, and total duration of exposure from spouse or work was given in years. In addition, cumulative exposure from the spouse was calculated as the number of pack-years (1 pack-year = 365 packs or the equivalent of 1 pack per day for 1 year). Cumulative time-weighted duration of exposure at work was calculated as the number of hour-years (1 hour-year = 365 hours or the equivalent of 1 hour per day for 1 year). When duration of exposure from spouse and work were combined, the same year was counted only once to avoid overlap of exposure periods from both sources.
Odds ratios (ORs), as estimators of relative risk, and 95% confidence intervals (95% CIs) were computed using unconditional logistic regression, adjusted for the matching variables age (continuous), gender, and catchment area (ten categories). The covariates included in the regression model were BMI (dichotomous), socioeconomic status (three levels), job strain (dichotomous), hypertension (dichotomous), diet (dichotomous), and diabetes mellitus (dichotomous). In addition, we used conditional logistic regression conditioning on the matching variables and adjusting for covariates. Categorization of ETS exposure variables was based on cutpoints according to the distribution among exposed controls (25th, 50th, and 75th percentiles). Interaction between ETS and other risk factors was defined as departure from additivity, expressed as the proportion of cases among those with joint exposure that is attributable to their interaction. 27 Attributable proportion due to interaction adjusted for confounders was calculated using an SAS program. 28 All other statistical analyses were performed with STATA 6.0.
The distribution of cases and controls according to age, gender, hypertension, overweight, job strain, diabetes mellitus, socioeconomic status, and selected dietary factors is shown in Table 1. Hypertension and overweight were slightly more common among cases, whereas a relatively greater proportion of controls were exposed to job strain. Diabetes mellitus was considerably more common among cases than controls.
Ever having lived with a smoking spouse was associated with an OR for nonfatal MI of 1.53 (95% CI = 0.95–2.44) for women and 0.96 (95% CI = 0.64–1.44) for men (Table 2). Among the women, current spousal exposure appeared particularly important (OR = 2.59, 95% CI = 1.27–5.29). For work place exposure, there was no major difference in risk estimates between the genders, and primarily current exposure seemed to be associated with an elevated risk. Exposure to ETS from parents during childhood showed no consistent relation to an increased MI risk (not shown in table).
Table 3 shows that subjects exposed to an average of more than 20 cigarettes per day from their spouse had an OR for MI of 1.58 (95% CI = 0.97–2.56). Living with a heavy smoker appeared particularly important for women (OR = 2.13, 95% CI = 1.05–4.28, not shown in table). The risk of MI was slightly higher for a longer duration of residential ETS exposure, with an OR of 1.25 (95% CI = 0.77–2.02) for those exposed for 33 years or more. A similar pattern was seen regarding duration of work place exposure. For cumulative measures, pack-years of spousal exposure displayed an increase with a higher exposure level, with an OR of 1.33 (95% CI = 0.81–2.20) in the highest category. The highest category of cumulative time-weighted duration of exposure at work (hour-years) showed an increased risk estimate for MI of 1.48 (95% CI = 0.99–2.22). For both of these cumulative exposure variables, the estimates in the highest category were similar for men and women, providing support for combined analysis of both genders when using refined exposure variables.
Table 4 illustrates the main analyses focusing on total ETS exposure from the two major sources. The risk of MI consistently decreased for each category of years since last exposure from either spouse or work (OR for trend over categories = 0.98, 95% CI = 0.97–1.00). For duration of exposure from spouse and work, there was an elevated OR for those exposed for at least 24 years. A trend analysis based on mean values of categories resulted in an OR of 1.15 (95% CI = 1.05–1.26) for an increase of 10 years of exposure. Weighted duration (hour-years) seemed to follow an exposure-response pattern, and the OR for a linear increase of 10 hour-years (based on mean values of categories) was 1.04 (95% CI = 1.01–1.06). An alternative categorization of hour-years based on whether the subject belonged to the first, second, third, or fourth quartile of either exposure source also showed a high risk estimate in the top category of exposure. In addition, we also included exposure during childhood in some analyses, in an attempt to better describe total lifetime exposure. This information did not result in stronger associations, however, which suggests that more recent exposure was of greater importance for the MI risk. The results of conditional regression analyses were similar to those of the unconditional analyses but tended to yield slightly higher ORs, especially in the top category of the variables in Table 4; for example, the OR was 1.75 (95% CI = 1.09–2.83) for more than 90 hour-years of ETS exposure (with 268 observations dropped because of incomplete matching sets, mainly as a result of small catchment areas and the exclusion of smoking individuals). According to the combined exposure measures in Table 4, the attributable proportion of MI among never-smokers from exposure to work place or spousal ETS would be 12–13%.
Most variables describing spousal ETS exposure showed higher effect estimates for women than for men as well as for younger subjects. For the highest category of cumulative exposure, in which we expect least misclassification, estimates were similar in the two genders, however. Neither gender nor age seemed to modify the effect from work place exposure. Analysis of possible interaction between ETS exposure and other risk factors for MI indicated synergistic interaction only for diabetes mellitus and family history of coronary heart disease. The proportion of cases due to interaction among those with both cumulative ETS exposure from spouse and work above the 75th percentile and another risk factor was estimated to be 52% for diabetics and 25% for those with a family history of coronary heart disease. None of the other risk factors indicated interaction with ETS.
Our findings support the hypothesis that passive smoking is associated with an increased risk of nonfatal MI. We found increased risks for those living with a smoking spouse or working in a location where smoking occurs, as also reported in previous studies. 1,2,8,10,11,13 The never-ever exposure variable commonly used in ETS studies fail to take some biologically relevant aspects of the association into account, such as intensity, duration, and time since last exposure. In addition, single-source models analyzing parental, spousal, and work place exposure separately ignore the importance of considering total exposure to the pollutant in different environments. Our study shows stronger associations between ETS and MI with more refined exposure variables, such as combined and cumulative measures, suggesting that simple dichotomous variables in present-day populations may incorporate substantial misclassification. 29 The variables used displayed dose-response relations between exposure and disease.
The average daily number of cigarettes smoked by a subject’s spouse appeared to be of special importance for the risk of MI, which is in accordance with the previous studies analyzing intensity of ETS exposure and MI risk. 8,11,13 The risk for MI also seemed to be higher with increasing number of years exposed to ETS from the spouse or work, whereas other studies have not found consistently higher MI risk with increasing duration of ETS exposure. 1,8,11 Nevertheless, others have reported a higher risk for current than previous exposure, which was also demonstrated in our study, both for spousal and work place exposure. Current exposure also implies longer duration on average than previous exposure. In addition, our data suggest a trend of risk reduction from recent to past periods of exposure. In studies of active smoking and coronary heart disease, the risk decreases with smoking cessation, and it is plausible that this could be the case also for exposure to ETS. 17
For chronic diseases, for example, heart disease and cancer, concurrent modeling of duration and intensity or use of long-term time-weighted average exposures are more suitable as a biologically relevant exposure metric if the effect is due to cumulative damage, whereas exposures during the final months or years, or present exposure, may be more relevant for more acute health effects. MI can be viewed both as a long-term progression and an acute event, and there is evidence to suggest that ETS exposure, as well as active smoking, may have both types of effects. Because of the difficulty in modeling long-term average exposure and duration simultaneously, cumulative measures are an alternative. Cumulative time-weighted duration (hour-years) of ETS exposure at work showed the strongest association between exposure and disease in our study. This measure may be a better cumulative measure with regard to exposure in the work place than spousal exposure, whereas pack-years might better describe the exposure pattern from the spouse. Work place environments polluted with tobacco smoke most often include the smoke from more than one person at a time, which is not the case for spousal ETS exposure. In addition, the environment in a work place where people smoke indoors is likely to be replenished with tobacco smoke repeatedly, resulting in a more even and prolonged exposure pattern compared with a residence where a spouse is smoking from time to time. Average nicotine concentrations in work places with no smoking restrictions are commonly greater than concentrations in the homes of smokers. 30
To calculate the combined measure of hour-years demonstrated in Table 4, we used a conversion factor from pack-years of exposure calculated from a Swedish study of ETS and lung cancer 20 to compute hour-years of spousal exposure, because number of hours of residential exposure was not directly available from questionnaire data. These were combined with hour-years of exposure from work to obtain total cumulative dose from these two major sources of ETS exposure. 31 Although this calculation might have resulted in some misclassification of combined hour-years, it would probably affect cases and controls similarly. Thus, because the sources are different and the variables difficult to combine, our estimates for total cumulative exposure may be more heavily weighted for spousal than for work place exposure, or vice versa. Therefore, an alternative way to combine the cumulative dose measures from spouse and work is presented in the last section of Table 4, which disregards the specific level of the two possibly different exposure sources.
Analysis of potential interaction suggested that among never-smokers as many as half of all MI cases among diabetics exposed to ETS might be due to interaction between the two factors, and every fourth MI case among the ETS-exposed individuals with a family history of coronary heart disease. The data, however, are too sparse to allow for any conclusive results. Nevertheless, our findings imply that certain subgroups in the population, such as diabetics or those with a family history of coronary heart disease, could be more susceptible to develop MI from ETS exposure.
The study population comprised all never-smoking first-event MI cases and controls from a large case-control study. The strengths of this study are the good quality of diagnosis and the high reliability of case identification, 32,33 as well as low probability of important systematic error from selection or recall bias. 21 Possible sources of bias to be considered primarily include nonparticipation and exposure misclassification. For example, bias would be introduced if those who did not respond to the questionnaire differed in exposure reporting from those responding, provided that this nonresponse was related to disease status. Nevertheless, a major difference needs to be present to have an important effect on the results, especially in view of the high participation rates in the SHEEP study. According to an environmental health survey in 1992, the prevalence of ETS exposure was 25% in Stockholm County among never-smokers between 45 and 65 years of age. 34 This figure is consistent with the data in our study for older subjects, among whom 21% of the controls and 23% of the cases reported daily ETS exposure at home or work between 1992 and 1994. Another possible validity problem is recall bias. We excluded fatal cases from the study, because of the probability of poor quality and possible bias in exposure information from their relatives. Validation by next-of-kin interviews in a study of ETS and lung cancer in Stockholm, 1989–1995, found good agreement between the study subjects and their spouses or other next-of-kin on reported smoking habits of the subjects’ spouses. 35 The sensitivity for smoking status was 99%, and there was also a good correlation for cumulative amount, similar for cases and controls. Because the cases answered the questionnaire after their infarction, bias in the reporting of other variables cannot be excluded, but is unlikely to explain the excess risks.
It has been argued that the association between ETS and MI might be explained by differences in dietary habits between smoking and nonsmoking families, which have previously been reported. 36 To account for this possible confounding, we adjusted for dietary intake of fat and fiber in the analyses, which has not been done in other studies. Such adjustment did not affect the results. Other potentially important confounders, for example, dietary cholesterol and blood lipids, were also considered and had no or very little effect on the results. In particular, possible confounding by smoking was avoided by focusing on never-smokers. Previous data have suggested that possible bias from misclassification of never-smoking status does not lead to any important validity problem even in studies of ETS and lung cancer, in which the much higher smoking-related risks make this a greater concern. 35 Data from population validation studies of reported smoking have shown about 5% misclassified ever-smokers in the never-smoking category, who were mainly light or long-term ex-smokers, and even lower misclassification rates have been found in case-control studies, in which 1.2% of the “never-smokers” were reported by their next-of-kin to be former regular smokers. 35,37
Some concerns have been raised regarding the discrepancy between the high risk for MI associated with ETS in relation to the small concentration of passively inhaled cigarette smoke compared with active smoking. 19 One possible explanation for this might be that studies on active smokers usually include passive smokers in the reference category, whereas the reference category in ETS studies consists of those who have reported never being exposed to ETS. 1 Other explanations might be that active smokers have adapted themselves to many of the toxins in cigarette smoke or that secondhand smoke differs quantitatively from the smoke inhaled by smokers. 14 In the SHEEP study, the risk of active smoking appeared somewhat higher than previously reported for corresponding age spans. 21 Our data suggested that the proportion of MI cases among never-smokers that might be attributable to passive smoking at home or work was 12–13%.
In conclusion, the results from our study confirm a risk of nonfatal MI from exposure to ETS. Intensity of spousal exposure, combined exposure from spouse and work, and time since last exposure appear to be of special importance for the risk of MI from ETS.
The SHEEP Study Group comprised the following: Institute of Environmental Medicine, Department of Public Health Sciences, Units of Social Medicine and Occupational Health, and Department of Medical Epidemiology, Karolinska Institute, Stockholm; Department of Occupational Health, National Institute for Working Life, Stockholm; National Institute for Psychosocial Factors and Health, Stockholm; Departments of Environmental Medicine, Epidemiology, Occupational Health, and Social Medicine, Stockholm County Council; Departments of Medicine at Danderyd, Huddinge, Löwenströmska, Nacka, Norrtälje, Sabbatsberg, St Görans, Södersjukhuset, and Södertälje Hospitals; and Departments of Cardiovascular Medicine and Clinical Chemistry, Karolinska Hospital, Stockholm, Sweden.
1. Kawachi I, Colditz GA, Speizer FE, Manson JE, Stampfer MJ, Willet WC, Hennekens CH. A prospective study of passive smoking and coronary heart disease. Circulation 1997; 95: 2374–2379.
2. Svedsen KH, Kuller LH, Martin MJ, Ockene JK. Effects of passive smoking in the Multiple Risk Factor Intervention Trial. Am J Epidemiol 1987; 126: 783–795.
3. Sandler DP, Comstock GW, Helsing KJ, Shore DL. Deaths from all causes in non-smokers who lived with smokers. Am J Public Health 1989; 79: 163–167.
4. Hole DJ, Gillis CR, Chopra C, Hawthorne VM. Passive smoking and cardiorespiratory health in a general population in the west of Scotland. BMJ 1989; 299: 423–427.
5. Humble C, Croft J, Gerber A, Casper M, Hames CG, Tyroler HA. Passive smoking and 20-year cardiovascular disease mortality among non-smoking wives, Evans County, Georgia. Am J Public Health 1990; 80: 599–601.
6. Steenland K, Thun M, Lally C, Heath C Jr. Environmental tobacco smoke
and coronary heart disease in the American Cancer Society CPS-II cohort. Circulation 1996; 94: 622–628.
7. Garland C, Barrett-Connor E, Suarez L, Criqui MH, Wingard DL. Effects of passive smoking on ischemic heart disease mortality in nonsmokers: a prospective study. Am J Epidemiol 1985; 121: 645–650.
8. Muscat JE, Wynder EL. Exposure to environmental tobacco smoke
and the risk of heart attack. Int J Epidemiol 1995; 24: 715–719.
9. Lee PN, Chamberlain J, Alderson MR. Relationship of passive smoking to risk of lung cancer and other smoking-associated diseases. Br J Cancer 1986; 54: 97–105.
10. Dobson AJ, Alexander HM, Heller RF, Lloyd DM. Passive smoking and the risk of heart attack or coronary death. Med J Aust 1991; 154: 793–797.
11. La Vecchia C, D’Avanzo B, Franzosi MG, Tognoni G. Passive smoking and the risk of acute myocardial infarction
. Lancet 1993; 341: 505–506.
12. He Y, Lam TH, Li LS, Li LS, Du RY, Jia GL, Huang JY, Zheng JS. Passive smoking at work as a risk factor for coronary heart disease in Chinese women who have never smoked. BMJ 1994; 308: 380–384.
13. Ciruzzi M, Pramparo P, Esteban O, Rozlosnik J, Tartaglione J, Abecasis B, Cesar J, De Rosa J, Paterno C, Schargrodsky H. Case-control study of passive smoking at home and risk of acute myocardial infarction
. J Am Coll Cardiol 1998; 31: 797–803.
14. Glantz SA, Parmley WW. Passive smoking and heart disease: mechanisms and risk. JAMA 1995; 237: 1047–1053.
15. Glantz SA, Parmley WW. Passive smoking and heart disease: epidemiology, physiology, and biochemistry. Circulation 1991; 83: 1–12.
16. Taylor AE, Johnson DC, Kazemi H. Environmental tobacco smoke
and cardiovascular disease. Circulation 1992; 86: 699–702.
17. Surgeon General. The Health Benefits of Smoking Cessation. Rockville, MD: U.S. Department of Health and Human Services, 1990.
18. He J, Vupputuri S, Allen K, Prerost M, Hughes J, Whelton P. Passive smoking and the risk of coronary heart disease: a meta-analysis of the epidemiologic studies. N Engl J Med 1999; 340: 920–926.
19. Law MR, Morris JK, Wald NJ. Environmental tobacco smoke
exposure and ischaemic heart disease: an evaluation of the evidence. BMJ 1997; 315: 973–980.
20. Nyberg F, Agrenius V, Svartengren K, Svensson C, Pershagen G. Environmental tobacco smoke
and lung cancer in nonsmokers: does time since exposure play a role? Epidemiology 1998; 9: 301–308.
21. Reuterwall C, Hallqvist J, Ahlbom A, de Faire U, Diderichsen F, Hogstedt C, Pershagen G, Theorell T, Wiman B, Wolk A. Higher relative, but lower absolute risk of myocardial infarction
in women than in men: analysis of some major risk factors in the SHEEP study. J Int Med 1999; 246: 161–174.
22. Cardiac Intensive Care (in Swedish). Behandlingsprogram för Danderyds sjukhus, Ersta sjukhus, Huddinge sjukhus, Karolinska sjukhuset, Löwenströmska sjukhuset, Nacka sjukhus, Norrtälje sjukhus, Sabbatsbergs sjukhus, S. t Görans sjukhus, Södersjukhuset, Södertälje sjukhus. Stockholm: Stockholm County Council; 1990.
23. Statistics Sweden. Swedish Socioeconomic Classification (in Swedish). Reports on Statistical Co-ordination. Stockholm: Statistics Sweden, 1982; 4.
24. Theorell T, Perski A, Åkerstedt T. Changes in job strain in relation to changes in psychological state: a longitudinal study. Scand J Work Environ Health 1988; 14: 189–196.
25. Willett W, Stampfer MJ. Total energy intake: Implications for epidemiologic analyses. Am J Epidemiol 1986; 124: 17–27.
26. McGee D, Reed D, Yano K. The results of logistic analyses when the variables are highly correlated: an empirical example using diet and CHD incidence. J Chron Dis 1984; 37: 713–719.
27. Rothman KJ, Greenland S. Modern Epidemiology. Boston: Little, Brown, 1986; 313–315.
28. Lundberg M, Fredlund P, Hallqvist J, Diderichsen F. A SAS program calculating three measures of interaction with confidence intervals. Epidemiology 1996; 7: 655–656.
29. Nyberg F, Pershagen G. Accumulated evidence on lung cancer and environmental tobacco smoke
(Letter). BMJ 1998; 317: 347–348.
30. Hammond KS. Exposure of U.S. workers to environmental tobacco smoke
. Environ Health Perspect 1999; 107 (suppl 2): 329–340.
31. Nyberg F. Environmental and genetic factors in lung cancer: epidemiological and biomolecular studies focusing on never-smokers (Doctoral thesis). Stockholm: Karolinska Institutet, 1998.
32. Hammar N, Nerbrand C, Ahlmark G, Tibblin G, Tsipogianni A, Johansson S, Wilhelmsen L, Jacobsson S, Hansen O. Identification of cases of myocardial infarction
: hospital discharge register data and mortality compared to myocardial infarction
community registers. Int J Epidemiol 1991; 20: 114–120.
33. Alfredsson L, Hodell A, Spetz CL, Åkesson LO, Hammar N, Kahan T, Ysberg AS. Assessment of the Quality of Diagnoses of Acute Myocardial Infarction
in Three Swedish Counties (in Swedish). Stockholm: Swedish Board of Welfare, Project Report No. 1997 84-8.
34. Department of Environmental Health. Miljöhälsorapport 1994: On Associations between Environment and Health in Stockholm County (in Swedish). Stockholm: Stockholm County Council, 1994.
35. Nyberg F, Agudo A, Boffetta P, Fortes C, González C, Pershagen G. A European validation study of smoking exposure in nonsmoking lung cancer cases and controls. Cancer Causes Control 1998; 9: 173–182.
36. Osler M. The food intake of smokers and nonsmokers: the role of partner’s smoking behavior. Prev Med 1998; 27: 438–443.
37. Nyberg F, Isaksson I, Harris JR, Pershagen G. Misclassification of smoking status and lung cancer risk from environmental tobacco smoke
in never-smokers. Epidemiology 1997; 8: 304–309.