smoked. Estimates of ETS exposure, based on cotinine measure- ments, suggest that involuntary smokers absorb about 0.5 to 1 percent of the nicotine that active smokers absorb (Jarvis et al. 1984, Haley and Hoffmann 1965; Wald et al. 1984; Russell et al. 1966). Dose-Response Ret!ationships and Threshold for Risk -response relationships for active smoking can provide in- sights into the expected magnitude of disease resulting from the exposure of nonsmokers to ETS. These data are reviewed to de&mine whether disease can be expected in association with E'I'S. Data from cohort and cas+control studies demonstrate dose- response relationships for lung cancer, which extend to the lowest levels of reported active smoking. The dose-response relationship of active smoking with lung cancer risk has been described by several investigators in several different date sets (Whittemore and Altshu- ler 1976; Doll and Pet0 1978; Pathak et al. 1986). Although the mathematical forms of these models vary, none have included a threshold level of active smoking that must be passed for lung cancer tc develop. The dose-response relationship for active smoking and lung cancer has been used to project the lung cancer risk for nonsmokers (Vutuc 1964). Such projections yield risk estimates of 1.03 to 1.36 for exposures, considered to be reasonable estimates of involuntary smoking exposures, i.e., 0.1 t.c 1.0 cigarettes per day. The reference population for these risk estimates is the risk for nonsmokers as a group, including those with higher and those with lower exposures to environmental tobacco smoke. In contrast, the reference population for the risk estimates in studies of involuntary smoking is the lung cancer risk in only that group of nonsmokers who have lower exposure to EITS. Comparisons of lung cancer risk estimates from active smoking studies with those from involuntary smoking studies require reference to the same exposure group for proper mterpreta- tion. In general, the lung cancer experience .of all nonsmokers (i.e., those with higher and lower involuntary smoking exposure com- bined) has been used to establish the reference rate of lung cancer occurrence (i.e., set as a risk of 1) in studies of active smoking. The use of all nonsmokers as the reference group averages the lower risks of nonsmokers with less ETS exposure with the higher risks of those with more ETS exposure. Thus, with the relative risk for the entire group of nonsmokers set to unity, the relative risk for nonsmokers with lower exposure is below 1 and that for the group with higher exposure is above 1. As a consequence, relative risk estimates from studies of involuntary exposure cannot be directly compared with risk estimates extrapolated from active smoking, unh c(qm%on to a single level of exposure is possible. Failure to 26 consider the differences between the reference populations explains the apparent discrepancy noted by Vutuc. Consider, for example, the mortality study reported by Hirayama (1981a). In this study, the relative risk of lung cancer for nonsmoking wives of smoking husbands (current and former) compared with nonsmoking wives of nonsmoking husbands (as calculated from Figure 1 in Hirayama 1981a) was 1.78. If the relative risk for nonsmoking wives of nonsmoking husbands were expressed in relation to the combined group of nonsmoking women, then a value of 0.63 is obtained, while with a similar calculation, that for nonsmoking wives of smoking husbands (both current and former), yields a value of 1.12. Thus, when the appropriate comparison is made, the risk estimates developed by extrapolation of the active smoking data (1.03 to 1.36) closely approximate those actually found in a study of lung cancer risk due to involuntary smoking. Dose-response relationships between active smoking and the level of lung function, the rate of decline of lung function in adult life, and the development of chronic airflow obstruction are well established (US DHHS 1984). Different measures of dose have provided the strongest correlation with functional decline in different studies. Pack-years, a cumulative dose measure, was the strongest predictor of the level of forced expiratory volume in 1 second (FEVI) in the Tucson epidemiologic study (Burrows, Knudson, Cline et al. 1977). Duration of smoking and the amount smoked were found to be the best predictors in male subjects in a study of three U.S. communities (Reck et al. 1981), and pack-years was the best predictor in female subjects. In both of these studies, however, the estimated dose accounted for only about 15 percent of the variation of age- and height-adjusted FEW1 levels. The relatively low predictive capability of cigarette smoking variables in these studies most likely reflects a lack of information on the dete rminants of individual susceptibility to tobacco smoke. Further, exposure variables obtained by question- naire, such as the number of cigarettes smoked daily, may only roughly approximate the dose delivered to target sites in the respiratory tract. Many factors, such as puff volume, lung volume at which inhalation starts, and airways geometry will influence the smoke dose and its distribution within the lungs. Extrapolation from the results of these studies to the pulmonary effects of exposure to ETS is, therefore, likely to be inaccurate. Another approach for assessing lowdose exposures is to consider the information available from studies involving children and teenagers who have recently taken up smoking. Even with brief smoking experience, cross-set tional studies of active cigarette smok- ing by children and adolescents have demonstrated an increased frequency of respiratory symptoms (Rawbone et al. 1978; Rush 1974; Bewley et al. 1973; Seely et al. 1971) and small but statistically 27 significant reductions in lung function (Seely et al. 1971; Peters and Ferris 1967; Lim 1973; Walter et al. 1974; &&house 1975; Woolcock et aL 1984). Longitudinal studies involving children and adolescents have demonstrated that a physiologic impairment attributable to smoking may be found in some children by age 14 and may be present after only 1 year of smoking 10 or more cigarettes per week in children with previously normal airways (woolcock et al. 1934), and that relatively small amounts of cigarette use may lead to significant effects on FEVl and on the growth of lung function in adolescents (Figure 1) (`l'ager et al. 1935). When considering the risk of lowdose exposures for the develop ment of chronic respiratory disease, the existence of a spectrum of risk and a distribution of dose within the population should be taken into consideration. The characteristics of the part of the population most susceptible to involuntary smoke exposure is still being clarified. Evidence is accumulating that airways hyperrespon- siveness, atopy, childhood respiratory illness, and occupational exposures may all influence response to ETS. Current understanding of lung injury suggests that individuals with one or more of these characteristics that place them at the most sensitive end of the susceptibility curve may be the most likely to develop symptoms or functional changes es a result of ETS exposure. Dose of ETS also varies in the population, and the coincidence of high dose and increased susceptibility may convey a particularly high risk. Fur- thermore, ET3 exposure may damage lungs that are also affected by other insults. Pathophysiologic Cbsiderations Cancer Carcinogenesis refers to the process by which a normal cell is transformed into a malignant cell with uncontrolled replication. Carcinogenesis has been conceptualixed as a multistage process involving a sequence of alterations in cellular DNA that terminate with the development of a malignant cell. Agents acting early in this sequence are referred tc as initiators; those actii later are referred to as promoters. Compounds with both initiating activity and promoting activity have been identified in tobacco smoke. Carcinogenesis reflects DNA damage; although some repair may take place, biological models have not suggested that there is a threshold of damage that must be exceeded. Rather, carcinogenesis has been considered to involve a series of changes, each occurring at a rate dependent on the dose of a damaging agent. Higher doses increase the probability that the entire sequence will be completed, but lower doses may also lead to mahgnancy. o 0 10,ooo 20300 30,ooo 4o.wo 50,ooo 60,ooo 70,Doo 30,ooo Number of cigamtes consumed 150 d o ? ? o o ? ? ? o o ? o o o o o o o o s + 25 0 lO,ooa 2a.ooo 30,ooo 40,ow 5o.ooo ewoo 70,ooo 3o.ooo Number of cigarettes consumed FIGURE l.-Relationship between levels of predicted for FEW, (A) and FEFSWIJ (B) at examination 8 and cumulative number of cigareti smoked during examinations 4 through 8 NOTEzMenandwome.nmmbined@J-44). SOURCE: Tagor et al. (1936). 29 with the PiZZ or other phenotypes, are modest particulate exposures likely to increase the risk for disease to an appreciable extent. The development of acute and chronic airzoay disease or symptoms of cough, phlegm production, and wheeze may require a considerably smaller exposure than changes in the lung parenchyma, and it is not unreasonable to hypothesize that these symptoms may be related to repeated and continuous exposure to EYES in the susceptible individu- al. Strong evidence that lowdose active smoking causes increased rates of respiratory symptoms and functional impairment comes from the studies of children and adolescents discussed earlier (Woolcock et al. 1984; Tager et al. 1985). Because of the length of exposure, it is likely that these reflect airway rather than parenchy- mal effects. Another pathophysiological mechanism by which exposure to EX'S may increase an individual's risk for the development of chronic airflow obstruction is through respiratory viral infections. Mounting evidence indicates that the very young child (under 2 years of age) exposed to ETS is at increased risk for lower respiratory tract viral infections (Harlap and Davies 1974; Colley 1974; Colley et al. 1974; Leeder et al. 1976a; Fergusson et al. 1981; Dutau et al. 1979; Pedreira et al. 1985). There is also increasing, though still inconclusive, epidemiologic evidence that respiratory viral infections in early life may be associated with an accelerated decline in F'EVl and, therefore, an increased risk for the development of chronic airflow obstruction in adult life in smokers (Burrows, Knudson, Lebowitz 1977; Samet et al. 1983). By increasing the occurrence of viral infections of the lower respiratory tract in early life, exposure to ETS in childhood may have an appreciable, but indirect, effect on the risk for the development of chronic airflow obstruction in adult life. The structural basis for this increased susceptibility has not yet been elucidated, however. Furthermore, the child whose parents smoke is also more likely to take up smoking than is the child of nonsmoking parents. Thus, the child made susceptible to the effecta of active smoking by prior PITS exposure is also more likely to become an active smoker. The possibility that exposure to constituents of tobacco smoke in utero may exert a prenatal effect must also be considered. This exposure is clearly not the same as ETS exposure, since the lungs of the fetus are not being exposed to ETS; rather, the developing fetal lung is exposed to compounds absorbed by the mother and delivered to the fetus transplacentally. Evidence of an in utero effect in pregnant rats has been reported by Collins and coworkers (1985). These investigators reported that pregnant rats exposed to smoke from day 5 to day 20 of gestation, in comparison with control rats, showed reduced lung volume at term and saccules that were reduced in number and increased in size as a result of the reduced formation 31 Lung Disease The noncarcinogenic pathophysiologic effects of active smoking on the respiratory tract can be separated into (1) effects on the airways and (2) effects on the lung parenchyma. In the airways, the structural changes include inflammation in the small airways and mucous gland hypertrophy and hyperplasia. In the parenchyma, the main structural change is alveolar wall destruction. Both the airways and the parenchymal changes are caused by active smoking, but the interrelationships of these changes are not clear. They may be independent pathophysiologic pi, linked only by their joint association with tobacco smoking. As discussed earlier, there is evidence showing an approximately linear d-response relationship between F'EWl level and amount smoked; however, the d-response relationships have not been as well described for the underlying pathophysiologic changes in the airways or in the lung parenchyma. Host factors and other environ- mental factors presumably interact with active smoking to affect an individual's risk for the development of disease. In this regard, present evidence would suggest that only 10 to 15 percent of smokers develop clinically significant airflow obstruction, although parenchy- mal and airways changes can be demonstrated in a substantially higher percentage at autopsy (US DHHS 1984). Extrapolation from the evidence on active smoking to the likely effect of exposure to environmental tobacco smoke on the airways and parenchyma suggests that pathophysiologic effeds on both the airways and the lung parenchyma might be expected. Because the dose of smoke components from ETS exposure is small in comparison with the dose from active smoking, the extent of lung injury would most likely also be much smaller than that found in active smokers. Small changes in the lung may be below the threshold for detection on pulmonary function testing. If clinically significant chronic airflow obstruction occurs in nonsmokers exposed to EYES, the risk is likely to be concentrated among those individuals highly susceptible to the airway or parenchymal effects of cigarette smoke. This susceptible group may include individuals with bronchial hyperre sponsiveness and with other, as yet unidentified, genetic and familial risk factors. Identifying the risk factors for susceptibility to the airway and parenchymal effects of both mainstream smoke and EL'S is an important priority. The dose of environmental tobacco smoke received by the nonsmoker is unlikely, by itself, to he sufficient to cause a clinically significant degree of purmchymul disease (em- physema) unless an individual is at the extreme end of the susceptibility distribution. Any particulate load is likely to increase the elaatase burden in the lungs by causing an influx of neutrophils. However, only in the individual with very inadequate lung defenses, specificaIly severe deficiency of protease inhibitor (pi) associated 30 of saccule partitions. These hypoplastic lungs showed an internal surface area that was decreased. Whether this study in rats has any relevance to humans is not yet clear, but this issue deserves further investigation. Whether continued exposure to EX'S during childhood, while the lung is remodeling and growing, affects the process of growth and remodeling is not yet clear. In general, rapidly dividing cells and immature organs are more susceptible to the effects of enviromnen- tal toxins than are cells undergoing a normal rate of division and mature organs. Apart from the evidence, cited above, linking lower respiratory tract viral infections in very early life to an accelerated decline of F'EVl in adult life, there is no information yet to link the rate of growth of lung function during childhood to the rate of decline of hmg function in adult life hecause the nw longitudi- nal studies have not heen done. More information is needed to describe the relationship of exposure to ETS at various times during childhood to the maximal level of lung function achieved at full lung growth. Mefiod~logical Considerations in Epidemiologic Studies Measumnent of Expure h mamsing the health effects of EX'S exposure, as with other enknmental pollutant8, accurate assessment of exposure is critical for obtaining estimates of this agent's effects. Both random and systematic misclassification of the exposures of subjects in an investigation are of concern. Random misclassification refers to errors that occur at random; the consequence of such random misclassification is to bias toward fmding no effect. Systematic misclassification refers to nonrandom errors in exposure assessment; the consequence maybe to bias toward a greater or lesser effect than is actually present. Biased answers in response to a questionnaire may introduce systematic misclsssification. Some misclassification occurs in most observational (nonexperi- mental) epidemiological studies, and is inherent in all epidemiologi- cal studies of ETS. Tobacco smoking is ubiquitous in nearly all environments; few people escape being exposed to EX'S. Thus, the exposure variables for ETS in epidemiological studies do not separate nonexposed subjects from exposed subjects, rather, they identify groups with more or less exposure, or with a qualitative or semiquantitative gradient of exposure. In assessing exposure to ETS, the information should cover the biologically appropriate time period for the health effkct of interest and be collected in a form that permita the construction of biologically appropriate exposure measures. However, the collection of a full lifetime history of IZTS exposure, as in a study of malignancy, may not he feasible, and the accuracy of the informa- 32 tion may he limited. In evaluating the effects of ETS exposure, cumulative exposure, duration of exposure, and intensity of exposure may each influence the magnitude of effects, as may the timing of exposure in relation to age and level of development. Because of the difficulties inherent in assessing exposures through questio maims, increased emphasis has been placed on meamuing exposure through the use of molecular or biochemical markers. With available markers, this approach is limited to providing an indica- tion of recent (within 48 hours) exposure, which may not necessarily correlate with past exposure. A marker has not yet been devised for total integrated dose. Nevertheless, biological markers provide another method for classification of current exposure, and a stan- dard for validating questionnaires. The strengths and weaknesses of the existing methods of measur- ing exposure are further discussed below. Atmospheric Markers A number of different markers of atmospheric contamination by tobacco combustion products can be feasibly measured. Ideally, the atmospheric levels of the air contaminant or class of contaminants that are implicated in producing the adverse health effects would be measured. A variety of contaminants have been measured as indicators of ETS, but no single measure can adequately index all of its myriad components. Further, some contaminanta are produced by sources of environmental contamination other than tobacco smoke. Nicotine is ahsorbed only from tobacco and tobacco combustion products. Some of the pollutants that have heen measured include (1) carbon monoxide, (2) respirahle suspended particulates CRSP), (3) nicotine, (4) a number of aromatic hydrocarbons, such as benzene, toluene, benxopyrene, and phenols, and (5) acrolein. Some of these are in the vapor phase and some in the particulate phase. Some, such as nicotine, may exist in one phase (particulate) in MS and in the other (gas) phase in SS. Until more is learned about the contaminants and their physical state in ETS, the results of monitoring for a particular ETS component will be difficult to relate to ita diseasecausing potential. At a practical level, the technology for measuring nicotine levels and RSP levels is available and accurate. Personal Monitoring Both active and passive personal monitors can be used to measure an individual's total exposure to an air contaminant at the breathing xone. Active personal monitoring systems .employ pumps to concen- trate the air contaminants on a collection medium for laboratory analysis or to deliver the air to a continuous monitor. Passive 33 personal monitoring systems use diffusion and permeation to concentrate gases on a collection medium for laboratory analysis. Personal monitoring should provide a more accurate estimate of the dose of a contaminant than area mOnitoring, because the actual air in the breathing zone is sampled and the subject's time-activity pattern is inherently considered. As with area monitoring, the results for a particular component of ETS may not adequately characterize exposure to other components responsible for a particular disease or effect. Respirable suspended particulates can be measured with accuracy and give a reasonably accurate measurement of current exposure. Questionnaires me que&ionnaire has heen the most frequently used means of estimating exposures for epidemiological investigations. Question- naires typic&y have obtained information about the smoking habits of parents, spouses, or other family members and often about exposure outside the home. From this information, the subject is classified as exposed or not exposed to Errs, and the extent of exposure may be estimated. The questionnaire approach for exposure estimation has several potential limitations. First, the information obtained cannot exhaus- tively cover lifetime exposure to ETS; therefore, a completely accurate reconstruction of integrated dose over the years cannot be achieved. Second, in evaluating El% exposure in the home, the usual daily smoking of the smokers has often been used as a measure of exposure intensity at home. This assumption may not be correct, since smoking does not occur only in the home. For example, a one- pack-a-day smoker may smoke only five cigarettes a day in the home environment and smoke the rest at work or elsewhere outside the home. Third, quantitation of exposure in the workplace is inherently Sfficult because of changes in jobs and the varying exposure in any particular workplace. Despite these shortcomings, the information obtained by question- taires does discriminate between more exposed and less exposed ubjects. The evidence validating the questionnaire method is trongest for domestic exposure. In several studies, levels of cotinine m body fluids have varied with reported exposure to tobacco smoke at home (Greenberg et al. 1984; Wald and Ritchie 1984; Matsukura et al. 1984; Jarvis et al. 1984). In fact, residence with a smoker may identify a population that is more tolerant of ETS, and therefore more likely to be exposed outside the home. Evidence in support of this speculation is provided by a study of urinary cotinine levels in nonsmoking men in the United Kingdom (`Wald and Ritchie 1984). In this study, the men married to women who smoked reported a 34 greater duration of exposure outside the home than men married to women who did not smoke. Until accurate and inexpensive exposure markers are available for cumulative ETS exposure, the questionnaire approach will remain the simplest means of obtaining exposure information. It is, there fore, important to consider the misclassification that can be intro duced by using this indirect measure of exposure, In studies of the effect of ETS exposure, two types of misclassification are of concern: misclassification of current or former smokers as never smokers and misclassification of the extent of ETS exposure. Because active smoking has a greater effect on the lungs than exposure to ETS? the inclusion of active smokers within a larger group of nonsmokers may lead to the fmding of a significant effect on lung function, which is actually attributable to active smoking rather than to involuntary smoking. Misclassification of undeclared active smoking is a particularly important source of error in studies involving teenagers. Misclassification of smoking status is also of concern in casecontrol studies of the association between exposure to M`S and lung cancer. Information about smoking habits for these studies often comes from interviews with a surviving spouse or surrogate, who may have been a close family member, neighbor, or friend, or from a review of medical records. The smoking habits of the subject may he incorrectly reported. Classification of individuals who are current or former smokers as never smokers would lead to a spurious increase in the relative risk for lung cancer in nonsmokers exposed to ETS, because the smoking habits of spouses tend to be correlated. The extent of this bias in the case-control studies is uncertain. The proportion of people reported as never smokers, but who in fact did smoke in the past, is unknown. The proportion of current smokers who report themselves as nonsmokers can be estimated from studies using markers to validate questionnaires. Using biochemical markers of tobacco smoke ahsorption, the propor- tion would appear to he about 0.5 to 3 percent, depending on the population studied and the questionnaire used (Wald et al. 1981; Saloojee et al. 1982). Misclassification of the extent of ETS exposure can also occur, and may reduce the observed risk if a nonsmoking spouse of a smoker is not exposed to smoke at home. Friedman and colleagues (19831, reporting on a survey of 38,000 subjects, noted that 47 percent of nonsmoking women married to smokers reported that they were not exposed to tobacco smoke at home. Measurements of Absorption The difficulties inherent in estimating exposure and dose have provided the impetus for the development of biological markers for exposure to both MS and ETS. The marker that at present holds the 36 highest promise is cdhine, the mqjor metabolite of nicotine. Cotinine may he measured in saliva, blood, or urine. Numerous studies have demonstrated that there is good correlation between these measures of cotinine and the estimated exposure to tobacco smoke under laboratory conditions (Russell and Feyerabend 1975; HofEnann et al. 1984) and under conditions of daily life (Russell and Feyerabend 1975; Feyerabend et al. 1982; Foliart et al. 1983; Wald et al. 1984; Wald and Ritchie 1984; Jarvis et al. 1984; Matsukura et al. 1984; Greenberg et al. 1984). Cotinine is probably the best marker for tobacco smoke intake because it is highly sensitive and specitlc for t&acc~ smoke and because it can be detected both in active smokers and in individuals exposed to EX'S. Further details about cotinine and other markem are to be found in Chapter 4. Pohntiully Gmfounding Variables In any epidemiological study, the confounding factors must be considered and their effects controlled. Confounding refers to the bii effect of a factor that independently influences the risk for the disease of concern and is also associated with the exposure under evaluation. Confounding is of particular concern when the effects of the exposure of interest are expe&d to be small. The potential confounding variables depend on the health outcome of interest. For lung cancer, occupational exposures, diet, and exposure to other combustion products are of concern. For acute and chronic pulmonary effects, potential confounders include airways hyperresponsiveness, other indoor air pollutants, outdoor air pollu- tion, respiratory tract infections, occupational exposure, and socio economic status, which may potentially influence disease risk through its environmental correlau23. While this list is extensive, it may not be inclusive; in any single investigation it may not be possible to meesure and control all potentially confounding vti- able& In general, the evidence on active smoking in combination with the dosimetrp of involuntary smoking leads to the conclusion that the effecta of ETS on a population will be substantially less than the effecta of active smoking. The effects of E!R3 on infants and young children are an important exception. The association of E'IS with an adverse effect in an individual study may reflect bias, chance, or a causal relationship. Statistical signiticance testing is used to quantitate the role of chance; by convention, a p (probability) value less than 0.05 is deemed statisti- cally significant. A p value less than 0.05 means that the observed results would occur by chance less than 5 times out of 100, if there is 36 truly no association between ETS and the effect. The choice of 0.05 is arbitrary, and as the significance level declines, the probability that the observation could have occurred by chance lessens. For effects of small magnitude, as may he anticipated for some consequences of exposure to ETS, a large study population may be necessary to demonstrate statistical significance. The absence of statistical significance for an association may refled an inadequate sample sixe and is not always indicative of the absence of an association. In this regard, reports describing the absence of effects of ETS should provide the calculations needed to demonstrate the study's statistical power (ability to detect effects of the magnitude expected) or a confidence interval for the estimate of effect. An additional statistical issue is the directionality of statistical significance testing. Either one-sided or two-sided tests may he used, in the fmt, only effects in one direction are considered a possibility, whereas twosided tests consider the possibility of effects in opposing directions, i.e., increase or decrease of risk Given the strength of the evidence on active smoking and disease risk, one-sided testing in the direction of an adverse effect seems appropriate for most potential consequences of ETS. However, one-sided tests have not been performed in all investigations of ETS; the use of two-sided tests makes these studies conservative, as statistical significance will less often be attained. Respiratory System Effects of Involuntary Cigarette Smoke Exposure This section reviews the evidence on involuntary smoking and the adverse physiologic effects, respiratory symptoms, and respiratory diseases in nonsmoking adults and children. Health effects related to fetal exposure in utero from active smoking by the mother are not discussed. Lung growth and development may he influenced by in utero exposure, and the effects of such exposures have not been separated from those of exposure after birth. More complete treatments of this issue have heen published (US DHEW 1979; US DHHS 1980, Abel 1980; Weinberger and Weiss 1981). This section begins with a review of the data on infants and children who are exposed primarily through parental smoking. The health effects examin ed are increased respiratory illnesses, of both the upper and the lower respiratory tracts, increased chronic respiratory symptoms and illnesses, and alterations in lung growth and development. Studies of adults, whose exposures to environmen- tal tobacco smoke occur in a variety of settings, are examined with regard to symptoms and changes in measures of lung function. The potential for J3TS to produce bronchoconstriction in asthmatic and nonasthmatic subjects is also examined. 37 InfantsandChildren Acute Respimtory Illness Longitudinal Studies A number of studies, based on a variety of different designs, have examined the effects of involuntary smoking on the acute respira- tory illness experience of children (Table 1). Several different end points have been ev&&ed in these investigations: hospitalization for bronchitis or pneumonia as 888e88ed by hospital records (Harlap and Davies 1974; Rantakallio 1978); questionnaire assessment of hospitalization for bronchitis or pneumonia or of doctor's visits (Colley 1971; Leeder et al. 1976a) or both G'ergusson et al. 1981; Fergusson and Horwood 1985); questionnaire assessment of reapira- tory illness within the last year (Cameron et al. 1969; Schenker et al. 1983; Ware et al. 1984); chest illness before age 2 (Schenker et al. 1983); hospitalization for respiratory syncytial virus 0 infection (Sims et al. 1978; pullan and Hey 1982); physiciandiagnosed bronchitis, tracheitis, or laryngitis (Pedreira et al. 1985); and tonsillectomy as an indication of recurrent respiratory infection (Said et al. 1978). These diverse end points range from illnesses associated with a specific etiologic agent, e.g., RSV bronchiolitis, to clinician&agnosed syndromes, e.g., bronchitis of undetermined etiology. The possibility of reporting bias must be considered for the studies that have used questionnaires to measure iUness experience. In most of these studies, parents, usually the mother, have responded for the child and reported on the child's illness experience. Some investiga- tors have suggested that mothers with respiratory symptoms are more likely to report symptoms for their children and that stratifica- tion of subjects by the symptom status of their parents removes this element of recall bias (Lebowitz and Burrows 1976). Removal of symptomatic parents, however, may result in overcorrection for recall bias because cigarette smoking is associated with symptoms in the adult. This analytical strategy would not be expected to adjust for biased parental recall of early life events. Additionally, in all studies in which potential reporting bias was examined, control for parents' status reduced, but did not eliminate, associations of involuntary smoking with health -outcomes (Colley et al. 1974; Leeder et al. 1976a,,b; Schenker et al. 1983; Ware et al. 198.4). Further, the consistency of these studies, in spite of differing study populations and methods, weighs against bias as the sole explanation for the effect of involuntary smoke exposure. Harlap and Davies (1974) studied 10,672 births in Israel between 1965 ad 1968 and observed that infants, whose mothers, at a prenatal visit, reported that they smoked, had a 27.5 percent greater hospital admission rate for pneumonia and bronchitis than children 38 TABLE 1.-E&y childhood reepiratory illnew and involuntary c@arette smoking 0 l-10 11-m m+ C-Y' 2$4x birth& 1963-1966, fJt&hdm bItmdLiti/ 7.6 10.4 11.1 153 (1971) prhemo&tIrrtyeuoflifo 10.3 16.1 14.6 292 Fa=1.78 for a?m pumt maker RR=280 fbr two puwlt mllokam numberaEm&rwx l&d-rhdr TABLE ~-continued Study Saidetal. w78) 8ohenkeretalal. mfw Eht&b fihP Dlneu~ rata per 100 ckllnumb 3,920 children, ajpd lo-!20, -w ondhr 28.2 41.4 60.9 ~&~notobu France edem, geaiemlly before pmultmmkiDghmldt~rspac age6,Milldi&orof~ueult timediredb-~ mepimtoly tact infedion r wP== lo+ Y=- 4,071 children, eged 6-14, che6tiuaembeforeage2 6.7 7.9 11.6 Trend#forbothsirmrhMt united stata ChdUblW>SdQ%3iDptlCt 8.8 11.8 13.6 Ye= Parent- NOMIlWkGI chrmlt- camerolJ et al. w60) Leederetal US?Bp, b) am8 et al. (1978) 168 children, agmi M; Reepimtory iuners, rertricted 1.33 1.4 lllneumportednot~ parenta t&phone activity sad/or medical notclearhoymportbgbdult quationnaire, united stata conmllhllwhlutyeu ml&dtoohud -ludy 2J4!3 infant& born l96s- RR - 2.0 for Wants with two Not provided Puew mlponu bin unlikely, 1986, EngLand amokhg per&a el&cte-Torinfmltod -P-P-e- Rp&wIul~~not invutigatod Lcmgiw Btlldy 36 children, hcapitakad, Bonierline eipifhlt in- in Not pmided R8vblvndlioli~36 Noh6niCant&ctfor matelnalmlloking,6lwtye8rof colltro~ E&end ?? ???o?*?? o ? life emountemokedorabrfw RR=266 IYLm~ofMthncul~ c---M* TABLE l.-Continued lZd&dliO (1Qw Pullan and Hey ww l&t21 children of amoking sienifieant incream in Not `& FlUBpdiV8fOllOUUpddda motbera hapMid for rupiratory vidt4hoapi~dmtba l&322 children of ilhemduringfimt6yeamoflih uPbrsas&- n-king mothem RRz1.74 alldingwal~ FWmd Iaaei-w 190 children hooPitaliced, 8ignis~teffe!ctofmatemal Not prwidd c---M-b B8vinfectioQfh%tywIrof (Rk1.96) and ~&msJ (RR-1.62) life; 111 nonhonPi~ smoking at time of rtudo; -hb FwNd Eignaant maternal lrmobhg effectduringiintyenroflife (RR=l.M) F&&a et al. ww Nommokex hoker -- 1,144 infanta in prdintric 8ignibntincreer,inrupiratory Brolrchitis 71 109 l%dia~nothlimkdb pi-act& united 8tatu ibUDESunonp~O~ 21 40 eqmuwmaffactmfa childrcm 4 7 LV" tia of nonsmoking mothers. In addition, they demonstrated a dose- response relationship between the amount of maternal smoking and the number of hospital admissions for these conditions. The infants were classified by the mothers' prenatal smoking behavior and not by the mothers' smoking behavior during the first year of the child's life. Maternal smoking habits would probably have remained relatively stable across the short observation period. British. investigators (Colley et al. 1974) followed children born between 1963 and 1985 in London and also observed an increased frequency of bronchitis and pneumonia during the first year of life in the children of parents who smoked. This difference did not persist at 2 to 5 years of age. Tbis effect was independent of the parents' personal reports of winter morning phlegm and increased with the amount of smoking by parents. The annual incidence of bronchitis and pneumonia during the first year of life also increased with a greater number of siblings. This variable was not controlled in the original analysis, however, Leeder and colleagues (1976b) subse quently reported that, in this same cohort, a dose-response relation- ship with parental smoking persisted for bronchitis and pneumonia in the first year of life, after control for parental respiratory symptoms, the sex of the child, the number of siblings, and a history of respiratory illness in the siblings. Fergusson and colleagues (1981) studied 1,265 New Zealand children from birth to age 3. They demonstrated an increase in bronchitis and pneumonia and in lower respiratory illness during the first 2 years of life in children whose mothers smoked compared with child.ren whose mothers did not smoke. Correction for maternal age, family size, and socioeconomic status did not affect the relationship between the amount of maternal smoking and the rate of respiratory illness. The effect of maternal smoking declined with increasing age of the child. In a second report (Fergusson and Horwood 1985) the followup was extended to include the first 6 years of life. The results conf%rmed the initial fmdings. Maternal, but not paternal, smoking was associated with a statistically significant increase in lower respiratory illnesses during the first 2 years of life. However, after age 2 there was no signif?cant effect of maternal smoking on respiratory illness occur- rence. Rmbkdlio (1978) followed more than 3,696 children during the first 5 years of life; half of the children had mothers who smoked cigarettes during pregnancy and half did not. The children of mothers who smoked had a 70 percent greater chance of hmpi&- tion for a respiratory illness than the children of nonsmoking mothers. Pedreira and associates (1985) prospsctively studied 1,144 infants and their families in the greater Washington, DC., area. Mate& 42 smoking was associated with an excess frequency of acute bronchitis, tracheitis, and lary&tis, as diagnosed by the pediatricians caring for these families. Episodes of croup, pneumonia, and bronchiolitis were not increased by maternal smoking. A family history of chronic respiratory symptoms ~88 also associated with excess respiratory illness. Ware and coworkers (1984) studied more than 10,009 children in six American cities. Maternal cigarette smoking was associated with increased parental reporting of a doctordiagnosed respiratory illness before the age of 2 years and of an acute respiratory illness within the past year. The prevalence of positive questionnaire responses increased consistently with the current daily cigarette consumption of the mother; the d-response relationships were unchanged by adjustment for maternal symptoms and educational status. Cross-Sectional Studies Schenker and coworkers (1983) studied 4,071 children between the ages of 5 and 14 years in a cross-sectional study in Pennsylvania. Both chest illness in the past year and severe chest illness before age 2 were more frequently reported in nonsmoking children of parents who smoked. These investigators found that symptom and illness rates were higher in children of parents with respiratory symptoms. However, a significant effect of maternal smoking on these illness variables remained after adjustment for the parents' own respira- tory symptom history. In a study of 1,355 children between 6 and 12 years of age in the Iowa public schools, Ekwo and coworkers (1983) found that the presence in the home of at least one parent who smoked was significantly associated with reported hospitalization of the child for a respiratory illness during the first 2 years of life. As in other studies, the effect was stronger for maternal smoking than for paternal smoking. Case-Control Studies In England, Sims and colleagues (1978) examined 35 children at 8 years of age who had been hospitalized during infancy for RSV bronchiolitis and compared them with 35 control children of similar age. Maternal smoking was associated with a relative risk of 2.65 for hospitalization due to bronchiolitis. The sample size was small, and this effect of maternal smoking was not statistically significant. Pullan and Hey (1982) studied children who had been hospitalized with documented RSV infection in infancy. They found significantly greater smoking by their mothers at the time of the infection, compared with children hospitalized for other illnesses, including respiratory disease for which RSV infection was not documented. At 43 age 10, the children previously ill with RSV infection had an excess reported occurrence of wheeze and asthma and had lower levels of pulmonary function in comparison with the controls. The research- ers could not determine whether the RSV infection had caused persistent damage that affected the maturation of the lung or whether these children were already more susceptible to severe RSV infection because of pulmonary problems that antedated the RN infection. In summary, the results of these studies show excess acute respiratory illness in the children of parents who smoke, particularly in children under 2 years of age. This pattern is evident in studies conducted with different methodologies and in different locales. The increased risk of hospitalization for severe bronchitis or pneumonia associated with parental smoking ranges from 20 to 40 percent during the first year of life. Young children appear to represent a more susceptible population for the adverse effects of involuntary smoking than older children or adults. The timeactivity patterns of infants, which generally place them in proximity to their mothers, may lead to particularly high exposures to environmental tobacco smoke if the mother smokes. Acute respiratory illnesses during childhood may have long-term effects on lung growth and development, and might increase the susceptibility of the lung to the effects of active smoking and to the development of chronic obstructive lung disease (Samet et al. 1983; US DHHS 1984). Cough, Phlegm, and Wheezing A number of crossse&onal studies from different countries (Table 2) have shown a positive association between parental cigarette smoking and the prevalence of chronic cough and chronic phlegm ,in children; some studies have shown a relationship for persistent wheeze. However, not all studies have shown a positive relationship for all symptoms. The results of some of these studies may have been confounded by the child's own smoking habits (Colley et al. 1974, Bland et al. 1978; Kasuga et al. 1979). The association with parental smoking was not statistically &i&ant for all symptoms in all studies (Lebowits and Burrows 1976; Schilling et al. 1977; Schenker et al. 1983). However, the majority of studies showed an increase in symptom prevalence with an increase in the number of smoking parents in the home. A recent report (Charlton 1984) provides crosssectional data on parent-reported cough for 15,000 children, 8 to 19 years of age, in northern England. Chronic cough in the children was related to their age and to their own cigarette smoking status. However, with control of these factors by stratification, the number of parental smokers in the home was positively associated with the occurrence of chronic 44 TABLE Z.-Chronic respiratory symptoma in children in relation to involuntary smoke exposure Bates per 100 by numb8r of amoking parents Study SUhjfXbJ Beapiitory eymptonu or illness 0 1 2 c!ammellte Wley et al. (1974) 2,426 children, agfd 614, Chronic cough; quentionnaire blend completed by parent 16.6 17.1 22.2 Trend eignifican~ reportiog bina -pcmible result of parent symptoma or ecth smoking in children, ldikdYtOC!IpL3blflIU&CtOf trend Bland et al. W378) 3.106 children, awl 12-13, did not admit-te-ever emoking cigarettes., England Couah durina day or at nkht - . Morning oough 16.4 19.0 29.6 1.6 2.8 2.9 Croewectional study children's self-reported symphme end emoking hietory a&&d aimul~ual~ morning end daytime-&e&- different dimeen, could be difference in expoeure (expomre more likely awnke than deep) Crowaxtionel study, adju&d for cbild'e own smoking habita Wek et al. 8E4 children, aged 6-9, wm United States Chronic wugh end phlegm 1.7 2.1 3.4 Per&tent wheeze 1.8 6.8 11.8 Tread not rign&ant Tend significant Cmm-wtional study, edjueted for parental symptoms and child% ~WII SmOkiIlg Charkon aw 16,ooO children, eged S-19 yeara, E&and 40.0 46.0 66.0 TABLE 2.4htinued Study Subject.3 Wee per 100 by number of smoking parent-g Respiratory 6ymptoms or illness 0 1 2 Comment8 Dodge (1982) 628 children, grades 3-4, twwparent households; parent questionnaire response, United States Any wheeze 27.6 27.9 40.0 Phlegm 6.4 10.9 12.0 Cwxh 14.6 23.0 27.8 Au trends tdgnlfiwn~ some aFect might relate to po.renM Wm@moa but no trend influence !ik.ely &os-e&onal study Schenker et al. u9w Lehowitz and Burrowx (1976) SchiIllng et al. (1977) Kaeuga et al. (1979) 4,071 children, aged 6-14, United States 1,625 children, <16 yenm old, United States 816 children, age 7+, United States 1,937 children, aged 6-11, Jaw Chronic cough chronic phlegm Persistent wheeze Pereietent cough Pen&tent phlegm whesre couph, phMm, wheeze wheeze. asthma Trend not significant not austed for parental eymptome, although parental symptom effect dyr-ed Croeee&ional study Hiiher rates in symptomatic parent .houeebol&, trends persisted for asymptomatic households; no adjustment for child`e own smoking Cros+e&ional etudy Specific data not provided -ional study Adjust.4 for distance of home from main tratl-ic, highway cmee-sectional study Ekwo et al. (1983) 1,366 children, eged 6-12, United States Coughs with adds Wheeeing apart from c&3 6.2 1.0 8.3 4.1 4.8 4.0 1.2 7.7 6.4 Never Parent smoker . smoker 3.7 7.2 10.0 12.8 23.4 94.1 No swcant effect Increased prevalence in heavy smoker (>21 &/day) family less clear effect in light smoker (~21 c&z/day) family Odds ratios: 1.4 for smoker father, I.6 for smoker mother 2 if only Bnlpker mother Gasstmeuse measnrd not contded for; no con&tent doe+ ZEZtional study cough. The mother's smoking had a greater effect than the father's smoking. Burchfiel and colleagues (1986) have conducted a longitudinal study of 3,482 subjects from Tecumseh, Michigan. Subjects were initially between the ages of birth and 10 years and were followed up by questionnaire and examination 15 years after entry into the study. Age-specific incidence rates were calculated for a number of chronic respiratory symptoms, including cough, phlegm, wheeze, and bronchitis. Incidence rates for all symptoms were higher for children with two parental smokers when compared with children of non- smokers. Adjustment for potential confounding variables, including age, parental education, family size, and personal smoking, did not explain these results. British researchers &eeder et al. 1976b) studying a birth cohort over a 5-year period demonstrated an increased incidence of Nheez- ing among nonasthmatic children with two parents who smoked in comparison with children whose parents did not smoke, one parent who smoked, or parents whose smoking changed during the study (Leeder et al. 1976a). However, when this association was examined by logistic regression with control for. other factors, parental smoking was not a significant predictor of wheeze or of asthma. McConnochie and Roghmann (1984) performed a retrospective cohort study to examin e the influence of mild bronchitis in early childhood on wheezing symptoms 8 years later when the subjects had reached a mean age of 8.3 years. Involuntary smoking was a significant predictor of current wheezing (odds ratio 1.9). In a related study (McConnochie and Roghmann 1985) with these same children, involuntary smoking did not affect lower respiratory tract illness experience. In a study of 650 children aged 5 to 10 years (Weiss et al. 19801, a significant trend in the reported prevalence of chronic wheezing with current parental smoking was found; the rates were 1.9 percent, 6.9 percent, and 11.8 percent for children with zero, one, and two parents who smoked, respectively. Although the data given are for all households, when the analysis was restricted to those households where neither parent reported symptoms, the results were identical. The stability of the fmdings with this restriction suggests that reporting bias introduced by parental symptoms W&B not responsible for the observed results. Schenker and coworkers (1983) e xamined the influence of parental smoking and symptoms on the reporting of chronic respiratory symptoms of cough, phlegm, and persistent wheezing in children. These investigators found that the mothers were more likely than the fathers and symptomatic mothers were more likely than asymptomatic mothers to report these symptoms in their children. 47 Parental smoking had no significant effects on chronic respiratory symptoms. L&,,~z m,j BU~OWB (1976) assessed the effects of household mea=' smoking on respiratory 8ymptoms in 6% `hwm Chihhn younger b 15 y- of age. children from homes with current smoke= m higher symptom rates than those from homes with ex- smokers ad ee never smokers. However, the effect of household Bmow w w88 atistically significant only for persistent cough. h a general population s~u&, &hilling and ~lleagues (1977) reported no mhtion between wheeze and involuntary smoking. ware md ~i&,es (1984) enrolled 10,106 children between 6 and g Y- of we from six U.S. cities in a prospective study. The pmdene of persistent cough and persistent wheeze, measured at the second mtion, was higher in children whose parents smoked. `JJg effect was greater for maternal smoking than for paternal smoking. Symptom prevalence rates increased linearily tith the number of cigarettes smoked daily by the mother. In a multiple logistic model, the effect of maternal smoking persisted after adjustment for reported ilhess in the parents. Dodge (1982), studying third and fourth grade children in Arizona, found that symptoms, including wheeze, were related to both the presence of symptoms in the parents and the number of smokers in the household. In summary, children whose parents smoke had a 30 to 80 percent excess prevalence of chronic cough or phlegm compared with children of nonsmoking parents. For wheezing, the increase in risk varied from none to over sixfold among the studies reviewed. Many studies showed an exposurerelated increase in the percentage of children with reported chronic symptoms as the number of parental smokers in the home increased. Misclassification as nonsmokers of children who are actively smoking could bias the results of these studies. Adolescent 8mokers may be reluctant to accurately report their smoking habits, and more objective measures of exposure may not help to distinguish active experimentation with cigarettes from mvohmtary exposure to smoke Cl'ager 1986). Although miscla&fica- tion of children who are actively smoking as nonsmokers must be considered, many studies showing a positive association between P-nM smoldng amI symptoms in children, including children at 4P bdke f%#.ficant experimentation with cigarettes is prevalent. h addition, many 8tudies (Bland et al. 1978; Weiss et al. 1980; Chm-hn 19% Schenker et al. 1983, Dodge 1982; Burchfiel et al. 19%) found significant effects of parental smoking after ~~ide~g active smoking by the children. chronic resPbtorY symptoms represent an immediate health h&n for the child. However, the long-term simlcance of chrodc r@PhtorY sYmPbmS for the health of the child is unclear. &et 48 available data are cross-sectional, and followup studies of chroticdy symptomatic children are necessaq to determine the long-term health consequences of chronic respiratory symptoms. In recent YW, the effect of parental cigare&e smokjng on pulmonary function in children has been examined in crossgedional studies (Table 3) and a few longitudinal studies. The crmonal studies have demonstrated lower values on tests of pulmonary function (FEV75z, LEVI, FEFLJMS, and flows at low lung volumes) in children of mothers who smoked compared with cmdren of non- smoking mothers. The longitudinal studies (Table 4) have confirmed the cross-sectional results and provide some insight into the imp&+ tions of the cross-sectional data. Dose-response relationships have been found in both cross-se&on- al and longitudinal studies Stager et al. 1979; Weiss et al. 1980; Ware et al. 1964; Berkey et al. 1986); the level of function decreases with an increasing number of smokers in the home. As would be anticipated from the mother's greater contact time with the child, maternal smoking tends to have a greater impact than pa&Id smoking. Younger children seem to experience greater effects than older children (Tager et al. 1979; Weiss et al. 1960), and in older children the effects of personal smoking may be additive with those of involuntary smoking (Tager et al. 1979,1985). As noted by Tager (1986), the effect of maternal smoking on lung function may vary with the child's sex. Some studies have reported greater effects on flows at lower lung volumes in girls than in boys (Burchfiel et al. 1986; Tashkin et al. 19% Yamell and St. Leger 1979; Veda.l et al. 1964). Flows at higher lung volumes seem more affected in boys (Burchfiel et al. 1986, Yarnell and St. Leger 1979; F&-key et al. 1986; Tad&in et al. 1964). Whether these sex effects represent differences in exposure, differences in susceptibility to environmental cigarette smoke, or differences in growth and devel- opment is unclear. Tager and colleagues (1983) followed 1,156 children for 7 years to determine the effect of maternal smoking on the growth of PRO- nary function in children (Figure 2). After correcting for previous level of FEVI, age, height, personal cigarette smoking, and correla- tion between mother's and child's pulmonary function level, mater- nal smoking was associated with a reduced annual increase in F'EVI and FEFs75, using two separate methods of analysis. If the effect of ma&d smoking is maintained to 20 years of age, then a 3 to 5 percent reduction of FEV, and FEFz+75 due to maternal Smoking would be projected. The validity of this PrOjeCtion remains to be &&h&ed. Because few mothers changed their smoking habits, the 49 TABLE 3.-~uhonary ~UC&~UI in CII.WEW exposed to involuntary smoking Study Subjecta SchiIling et al. (1977) 816 children, aged 7-17, Comwcticut and South ceroh united statm FEV, 88 percent pITdictA No e!Tect of parental amokin No control for db&ip ab Or comlntion of eihliug pllmoaur function; for dludren who nevw enlow vwu60 elgdamtly lean in cbikhn with!Jmokhglnotbm T-r et al. om w&u et al. a9w Vddetd. (lgw Lebowlte end BIllTowS (ls7s) 444 children, e&fed 619, 3ht Bcmfml, Massaehusetb, united state &5ochudren,aged~,Eeet Boston, Memlechueetb, united state6 r,oca children, aged MS, united stake 371 houoehol&, complete hietnriee of parent Emokiug and pulmonery flmctlon of children, age >6, Tucson, Ak-hna. united ststa MMEF in etadad deviation units MMEF in etandard deviation unite m,, Fvc, vmuKJl vcdm VUdO FBV,, Fvc, o,, vmu7s derivedfromM?dEPVcune8, ea etaderd deviation units signll~t effect of parental Ml2Oki43 slgnifimt effect of parental ~Okhg FVC poaitivcly emocietedl flown negatively omocieti Noeffectofpnmbdemoking c!mtmued for oitahip aim and cadetion of sibling pulmowy fll&iOll controlled for nibobip aim and -htiOlOf&UBgpllmoovy function FlowBdaab-raporuewlth emountarnokedbymother chggahioa: may bs d dlfferenm in hdonr leveb of rlspa_mcomprsdwithmors no*rly climebe L8bmitz et al. (1982) 339 childreb Tuaon, Arimm, united stat83 FEV,, I acorn No effect of pan&al wnokhg Higher lweb of puhoaaq flulctlon for 4Thikhn of tmokiqparenbtbanforna- emokeexpwalchudren