- Email: [email protected]
Ecology of passive smoking by young infants
Ecology of passive smoking by young infants R o b e r t A. G r e e n b e r g , MD, MSPH, Karl El B a u m a n , PhD, L u c i n d a H. G l o v e r , MPH, V i c t o r J. S t r e c h e r , PhD, MPH, D a v i d G. K l e i n b a u m , PhD, N a n c y J. H a l e y , PhD, H e l e n C. S t e d m a n , BS, M a r y G l e n n Fowler, MD, a n d Frank AI L o d a , MD From the Department of Pediatrics,University of North Carolina School of Medicine, Chapel Hill; the Department of Health Behavior and Health Education an d the DePartment of Biostatistics, University of North Carolina School of Public Health, Chapel Hill The Frank Po[ter Graham Child Development C e n t e r University of North Carolina at Chapel Hill; and the Clinical Biochemistry Laboratory, Division of Nutrition and Endocrinology, American Health Foundation, Valhalla, New York This study provides a detailed description of passive smoking by 433 infants (mean a g e 18 days) enrolled from a representative population of healthy neonates in central North Carolina during 1986 and 1987. Sixty-four percent (276) lived in households with smokers or had contact with nonhousehold smokers. During the w e e k before data collection, two thirds (!84) of these 276 infants reportedly had t o b a c c o smoke produced in their presence. Seventyfive percent of smoking mothers smoked near their infants. The amount smoked by the mother near the infant correlated with the amount smoked near the infant by nonmaternal smokers. Cotinine, an indicator of smoke absorption, was found in the urine of 60% (258) of all study infants. The amount smoked in the infant's presence, as well as the amount smoked farther a w a y from the infant, especially by the mother, were the most significant correlates of the urine cotinine concentration. The results of thisstudy suggest that efforts to reduce passive smoking in young infants should emphasize the importance of the mother's smoking behavior, smoke produced anywhere in the home, and household social influences on smoking behavior near the infant. (J PEDIATR 1989;I14:774-80)
Numerous studies have found a direct association between parental smoking and lower respiratory illness in infants. 1 This association appears to be strongest during the first year of life. Vulnerability in early life is of particular concern because of the possible association of acute lower respiratory illnesses during infancy and pulmonary abnormalities in later life. 2 Evidence is also accumulating that
Supported by grant No. 28895 from The National Heart, Lung, and Blood Institute, National Institutes of Health. Submitted for publication Sept. 23, 1988; accepted Nov. 2, 1988. Reprint requests: Robert A. Greenberg, MD, MSPH, Division of Community Pediatrics, CB 7225, Medical School Wing C, University of North Carolina, Chapel Hill, NC 27599-7225.
774
other health problems might also be associated with passive smoking in childhood. 1,3 In response to these conc~ns, health care providers have been encouraged to counsel parents to stop smoking for their own and their children's benefit? Another approach is to reduce the opportunities for infants to absorb tobacco smoke despite persistent contact with people who smoke. Such efforts might be more effective if based on an understanding of the circumstances in which infant passive smoking occurs. We pursued a detailed investigation of those circumstances, including a description of the characteristics of the infants, the smokers, the places of exposure, the amount of smoke absorbed, and their interrelationships. We focused on the first weeks of life, a time when an infant is
Volume 114 Number 5
vulnerable to the harmful effects of tobacco smoke and when intervention and behavior change might be most effective, before smoking habits around the infant are firmly established and when a family's concern for protecting the infant's health may be greatest. METHODS Study population. The infants were enrolled in the study at birth, from April 1986 to May 1987. They are part of an intervention trial to reduce passive smoking, but all data in this study were collected before the intervention. To be eligible for enrollment, an infant had to have no significant postnatal problems, have a birth weight of at least 2000 gin, and reside in Alamance County or Chatham County in central North Carolina. Infants were recruited from three hospitals where approximately 80% of the births in those two counties occur. Although 1096 infants were eligible for the study, some were not included because (1) the mothers did not give informed consent (n = 394), or they gave informed consent but withdrew it before data collection (n = 68); (2) the infants were randomly selected not to participate until a later stage of data collection (n = 166); or (3) they did not produce the urine sample required for inclusion (n = 35). The remaining 433 infants constituted the study sample. Table I describes the study sample. Between the study sample and the remainder of the population that was eligible for enrollment, there were no significant differences (p >_ 0.05) in the sex of the infant, educational level of the head of households educational level of the mother, mother's age, and smoking behavior of the mother. The study population included relatively mor e black mothers: 35% of study mothers were black compared with 24% of mothers of the remainder of the eligible infants (p < 0,001). This difference remained significant after the variable of educational level of the head of household was controlled. The mean _+ SD number of people living in study households, including the infant, was 4.2 _+ 1.2. The most common composition of the households was the infant mother, father, and one or more siblings (38%), followed by the infant and just the mother and father (29%). Data collection procedures. Data collectors visited the mothers of all eligible infants in the hospital soon after birth to request enrollment in the study and to obtain informed consent and demographic data. Data collectors then visited the homes of enrolled infants when the infants were approximately 3 weeks of age (mean age _+ SD 18 _+ 6 days, range 8 to 51 days) and before the family's assignment in the intervention trial. Before beginning the interview, the data collector placed a urine collection bag on the infant and explained to the mother that the urine
Passive smoking by infants
Table
775
I. Description of study sample Study sample (N = 433)
Infants Female (%) Male (%) Mothers Race White (%) Nonwhite (%) Age Mean _+ SD (yr) Range Education Less than HSG (%) HSG only (%) More than HSG (%) Smoking behavior Smokers (%) Mean _+ SD (cigarettes/day) Range (cigarettes/day) Household heads Education Less than HSG (%) HSG only (%) More than HSG (%)
47.1 52.9
64.7 35.3 25.2 _+ 5.4 13 to 45 25.9 37.2 37.0 26.8 13.7 -+ 8.5 2-40
24.3 37.9 37.9
HSG, High schoolgraduate.
would be analyzed for tobacco smoke products to determine whether the baby had been exposed to smoke. The mother was then asked to describe the smoking habits of all household members and any recent exposure of the infant to tobacco smoke. The exposure data were gathered in a systematic manner by a series Of questions? For both "a typical day last week" and "during the last weekend," the mother identified all the places where the infant spent time, including rooms in the home, at other homes, in cars, and in public places. She then estimated the amount of time the infant spent in each place, the size of the place, and the amount of ventilation. For each place the mother reported all smokers who entered while the baby was there; whether they smoked, what, and how much; how close they were to the infant while smoking; and how much time the smokers spent in each place while the infant was there. Definitions of variables. An infant was considered at risk for exposure to tobacco smoke if the infant lived in the same household as a smoker or was in the same place at the same time as a n0nhousehold smoker, either in the house or elsewhere. A smoker was defined as anyone who smoked at least one cigarette, cigar, or pipeful per day. There were six Cigar smokers and one pipe smoker in the study. Exposure to tobacco smoke was said to occur when
776
Greenberg et al.
29%exposedto / smokefrom~ ( non-household smokers~
The Journal of Pediatrics May 1989
~ I
7%~ y
47%exposedto smokefrom mother Figure. Sources
%exposedto smokefrom household members otherthan mother
of smoke for 184 infants exposed to tobacco smoke. Because of rounding, percentages outside circles may not exactly equal the sum of their respective components within circles.
smoke was produced in the infant's presence, that is, in the same room or vehicle as the infant, affording the infant an opportunity to absorb the smoke. The amount of exposure was expressed as cigarettes per week smoked in the infant's presence. This was calculated by multiplying the number Of cigarettes smoked in the infant's presence during "a typical day" of the previous week by 5 and adding the number of cigarettes smoked in the infant's presence during the previous weekend. A cigar or piPeful was considered equivalent to one cigarette. S m o k e absorption was defined as sufficient airway contact with tobacco smoke to result in absorption of nicotine. It was measured-by the urinary concentration of cotinine, the major metabolite of nicotine. Cotinine in body fluids is considered a good indicator of tobacco smoke absorption. 1, 3 The urine samples were frozen and shipped to the American Health Foundation, in Valhalla, New York. Researchers there analyzed Specimens without knowing the exposure status of the infant. Cotinine concentration was quantitated by radioimmunoassay with a modification of the method originally described by Langone et al. 6 This protocol uses specific antisera produced by rabbits and has interassay and intraassay variations of 5%, with a sensitivity of 360 pg/ml (2102 pmol/L). Creatinine was measured by dry chemistry methods on a Kodak Ektachem 400 analyzer (Eastman Kodak Co., Rochester, N.Y.). The cotinine was divided by the creatinine concentration to adjust the data to reflect the concentration of the urine sample. 3 Statistical methods. Multiple linear regression models were used tO identify the correlates of the amount smoked in the infant's presence by each source of exposure to tobacco smoke and the correlates of the infant's urine cotinine concentration. A stepwise procedure, with back-
Laboratoryprocedures.
ward elimination of variables from a full model containing all appropriate candidate independent variables, was applied after adjustment for any collinearity. To remain in a model a variable had to have a partial F test p value <0.05. In the regression model involving the amount smoked in the infant's presence by the mother as the dependent variable, the total number of cigarettes she smoked per day was designated, before data analysis, to be kept in the model. This allowed us to study the effect of the other potential correlates (e.g., educational level of the head of household) after taking into account, or controlling for, the total number of cigarettes the mother smoked per day. Likewise, in the regression model involving the amount other household members smoked in the infant's presence as the dependent variable, the total number of cigarettes smoked daily by other household members was designated as a controlled variable. The significance of first-order interactions of the controlled variable with all candidate independent variables was also tested. The 21 infants of smoking, breast-feeding mothers were eliminated from all analyses that used cotinine as a continuous variable because of th e relatively large amount of nicotine and cotinine absorbed through breast-feeding, compared with the amount absorbed through the airways. 7 The effect on an infant's urine cotinine concentration of the size of the place in which the infant was exposed to tobacco smoke, the ventilation of that place, the time the infant spent there, and the distance between the smoker and the infant was summarized in an index that adjusted the number of cigarettes smoked in the presence of the infant during the previous week. First, the number of cigarettes each person smoked was weighted by the distance of the smoker from the infant. Then the weighted total number of cigarettes smoked in each place was further weighted by the size and ventilation of the place and the time the infant spent there? Distributions of categorical variables were compared by the Pearson chi-square ~statistic. The Cochran-MantelHaenszel chi-square method was used when a categorical variable was controlled. The two-sided Student t statistic was used to compare means of continuous variables for two groups. RESULTS
riskforexposuretotobaccosmoke.
Infants at A total of 276 infants, or 64% of the entire study population, were at risk for exposure to tobacco smoke; that is, the infant either lived in a household with a smoker or was in contact with a nonhousehold smoker. Of these infants at risk for exposure, 42% had a smoking mother, 70% lived in
Volume 114 Number 5
Passive smoking by infants
777
Table II. Final linear regression model* for amount of exposure of infant to smoke produced by mothert Variable
Intercept Total maternal smoking (cigarettes/day) Amourit of exposure to nonmaternal smoking (cigarettes/wk) Maternal age (yr) Place where infant spends maximum time Siblings
Unstandardized coefficient
V a r i a b l e values
Mean • SD 13.7 + 8.5 Range 2-40 Median 14 Range 0-241 Mean _+ SD 24.6 • 5.3 Range 14-38 Bedroom: 1 Other areas: 0 Yes: 1 No: 0
F
p
25.7 l 6 4.752
172.03
0.0001
0.340
22.43
0.0001
-1.939
9.68
0.0024
-18.694
9.35
0.0028
16.146
5.88
0.0170
*Model F = 57.56; df= 5, 110;p = 0.0001; R2 = 0.72. tFor 116 infants of smoking mothers, with control for the total number of cigarettes mother smoked per day.
households with smokers other than their mothers, and 26% had contact with nonhousehold smokers. Thirty-six percent of the infants at risk were in contact with more than one of those categories of smokers. A total of 239 infants, or 55% of the entire study population, lived in households with smokers. Of these households, 61% had one smoker, 33% had two smokers, and 6% had more than two. Infants exposed to tobacco smoke. Of the 276 infants at risk for exposure to smoke, 184 (67%) were reportedly exposed during the previous week. Of the exposed infants, 47% were exposed to smoke produced by their mothers (75% of smoking mothers smoked in their infant's presence), 66% were exposed to smoke from household members other than mothers, and 29% to smoke from nonhousehold smokers (Figure). Forty percent of the exposed infants had more than one category of smoker as their source of exposure. The mean total amount of exposure was 71 cigarettes per week per infant, with a range of 1 to 350 cigarettes per week. Mothers contributed, on the average, 33% of the amount to which an infant was exposed. All household members other than the mother contributed, on the average, 47% of the total amount of exposure, and nonhousehold smokers contributed 20%. The most important correlates of the amount of exposure of the infant to smoke from the mother were maternal smoking, measured by the total number of cigarettes smoked anywhere by the mother each day, and the number of cigarettes smoked in the infant's presence by people other than the mother (Table II). There was more exposure to tobacco smoke produced by the mother if the infant had a younger mother, spent most of the time outside the bedroom, and had siblings. Nonsignificant variables
included the educational level of the head of household, the sex of the infant, the mother's race, whether the infant was breast fed, the number of rooms in the house, and first-order interactions between maternal smoking and other variables. The significant correlates of the amount of the infant's exposure to tobacco smoke produced by nonmaternal household members were the amount of the infant's exposure to maternal smoking and the interaction of the mother's race with the total number of cigarettes nonmaternal household members smoked per day (Table III). An increase in the total number of cigarettes smoked by nonmaternal household members resulted in a greater increase in the amount smoked in the presence of nonwhite infants compared with white infants. The nonsignificant variables were the mother's age, sex of the infant, educational level of the head of household, number of rooms in the home, place where the infant spent maximum time, and first-order interactions between total smoking by nonmaternal household members and other variables, except race. The only significant variable in the regression model of the correlates of the amount of exposure to smoke produced by nonhousehold smokers was the amount of exposure to maternal smoking (p < 0.0001, model R 2 = 0.21). Nonsignificant variables were mother's age and race, infant's sex, educational level of the head of household, place where the infant spent the greatest time, and amount of exposure to smoking by nonmaternal household members. Absorption of tobacco smoke. O f the 433 study infants, 258 (60%) had measurable cotinine in their urine. The median concentration of those excreting cotinine, excluding the 21 infants of breast-feeding, smoking mothers, was
778
Greenberg et al.
The Journal of Pediatrics May 1989
T a b l e III. Final linear regression model* for amount of exposure of infant to smoke produced by nonmaternal household members']" Variable
Intercept Total nonmaternal household member smoking (cigarettes/day) Race of mother:l: Total nonmaternal household member smoking • race of mother Amount of exposure tO maternal smoking (cigarettes/wk)
V a r i a b l e values
Mean _+ SD 20.7 _+ 15.5 Range 1-120 Nonwhite: 1 White: 0
Mean + SD 25.5 + 50.6 Median 0 Range 0-280
Unstandardized coefficient
F
p
2.31
0.1303
1.035
7.08
0.0085
0.306
44.31
0.0001
19.224 0.253
-24.955
*Model F= 20.84; df= 4, 186;p = 0.0001; R2= 0.31. ~'For 191 infants in households with nonmaternal smokers, with statistical control for the total number of cigarettes they smoked per day. :~Thisvariable must be included in the model because it is involvedin the significant interaction variable.
121 n g / m g of creatinine (77.7 nmol/mmol), with a range of 6 to 2273 n g / m g (3.9 to 1459 nmol/mmol). The median and range of cotinine concentrations unadjusted by ereatinine were 9 n g / m l and 2 to 138 ng/ml (51 nmol/L and 11 to 784 nmol/L), respectively. The concentrations of cotinine in the urine of the infants in our study were similar to those in other studies of passive smoking by infants 79 and in studies of adult passive smokers} The significant correlates of the urine cotinine concentration were the amount of smoking the mother did in the infant's presence, the amount she smoked elsewhere, the amount of smoking other household members did in the infant's presence, the amount they smoked elsewhere, and the intensity of exposure to tobacco smoke (Table IV). The last variable was a measure of whether the place where the infant spent the most time coincided with the place where the infant received the maximum exposure. The amount of exposure to smoking by nonhousehold smokers was not a significant correlate. For exposed infants the Spearman correlation coefficient between the number of cigarettes smoked in the infant's presence during the previous week and the urine cotinine concentration was 0.54 (p = 0.0001). The correlation coefficient did not change when the number of cigarettes was adjusted for the size of the place in which the infant was exposed, the amount of ventilation of that place, the time the infant spent there, and the distance between each source of exposure and the infant. DISCUSSION A large proportion of our study infants (64%) was at risk for exposure to tobacco smoke. This percentage is within
the range for households with at least one smoker, reported in studies of older children9 The actual proportion at risk is probably greater than 64% because 37 (9%) of the study infants had cotinine in their urine yet reportedly lived in nonsmoking households and were not in contact with smokers during the previous week. The estimated half-life of elimination of cotinine from the urine of nonsmoking adults is 1 to 2 days, 3 which indicates that those 37 infants probably had contact with tobacco smoke during the previous week. An alternative source of cotinine would be the milk of breast-feeding mothers who used smokeless tobacco products. However, only three of our mothers reported using these products, and they did not breast-feed their infants. We considered an infant to be exposed to tobacco smoke when smoke was produced in the infant's presence (i.e., in the same room or vehicle). Mothers were a major source of tobacco smoke exposure, but other household members and nonhousehold smokers also contributed. The amounts smoked in the infant's pre~ence by maternal and nonmaternal smokers were associated. The amount smoked by the mother in the infant's presence was the most important correlate of the amount smoked in the infant's presence by nonmaternal household members and by nonhousehold smokers. Conversely, the amount smoked in the infant's presence by nonmaternal smokers made an important contribution to explaining the variation in the amount smoked by the mother in the infant's presence. These results are consistent with the thesis that social influences are important to smoking near the infant. The total number of cigarettes a mother smoked near her young infant was strongly related to the total amount
Volume 114 Number 5
Passive smoking by infants
779
T a b l e IV. Final linear regression model* for absorption of tobacco smoke (urine cotinine/creatinine ratio) for 411 infants Variable
Intercept Amount of exposure to maternal smoking (cigarettes/wk) Maternal smoking not in same place as infant (cigarettes/day) Amount of exposure to nonmaternal household member smoking (cigarettes/wk) Comparison of place where infant spends maximum time with place where maximum amount of exposure to smoke occurs Total nonmaternal household member smoking not in same place as infant (cigarettes/day)
V a r i a b l e values
Mean _+ SD 11.7 _+ 33.8 Median 0 Range 0-280 Mean _+ SD 1.5 _+ 3.9 Median 0 Range 0-30 Mean _+ SD 11.7 _+ 26.7 Median 0 Range 0-161 Same place: 1 Other situation: 0
Mean _+ SD 7.0 _+ 12.7 Median 0 Range 0-120
Unstandardized coefficient
F
p
50.168 1.541
23.76
0.0001
13.996
39.06
0.0001
1.227
10.58
0.0012
64.078
6.07
0.0142
1.375
3.99
0.0464
*Model F = 39.64; df= 5, 405;p = 0.0001; R2= 0.33. Not includedare 21 infants of breast-feeding, smokingmothers and one infant with no reported smoke exposure but with the highest urine cotinine/creatinineratio in the study (2273 ng/mg).
she smoked. This relationship may weaken as the infant grows older and they spend less time close together. Thirty-seven percent of the children who absorbed smoke were reportedly not exposed to smoke during the previous week. Although inaccurate maternal recall and reporting, discussed in more detail later in the article, could contribute to this result, other explanations arc that the infant was brought into places where smoking had previously taken~ place, Or that smoke produced elsewhere had dispersed into places Where the infant was. The correlates of smoke absorption accounted for only 33% of the variation in urine cotinine excretion. However, tobacco smoke exposure and absorption should not necessarily be more closely correlated, because they are conceptually distinct. Exposure in our study referred to tobacco smoke in the environment, and absorption referred to a nicotine metabolite excreted in urine a s a reflection of tobacco smoke absorbed. Nevertheless, tobacco smoke exposure and absorption are related, and there are factors that could have affected our measure of that relationship. Reconstruction by the mother of a "typical day" for her infant during the previous week and her recall of events during the previous weekend might have been inaccurate. A recognition that smoking near infants is harmful might have led to an underestimate of the amount of smoking by household members, particularly in the infant's presence. Although the infants were only several weeks of age, there were times during the day when some mothers were not with their infants, or even in the same house; knowledge of
smoke exposure during those times might have been particularly limited. Days that were not considered "typical" by the mothers might have been reflected in the urine cotinine concentration, particularly if those atypical days occurred close to the day the urine specimen was collected. The amount of smoke that reaches an exposed subject is said to be affected by the size of the space into which it disperses, the amount of air exchange in that space, and the distance of the subject from the source of smoke, l However, adjusting the number of cigarettes smoked in the presence of the infant by those factors did not strengthen the relationship between the number of cigarettes and the urine cotinine level. These results might be explained partially by the rapid dispersion of tobacco smoke in a defined space? Our need to rely on maternal recall, especially to estimate ventilation and distance of the infant from each smoker, prevents us from dismissing these factors as unimportant. Assessment of their influence on ~smoke absorption might be better under experimental conditions. Some variables that were not measured in this study could lead to variation in urine cotinine values. These include differences in cigarette nicotine content and individual smoking habits (such as puff depth and frequency and butt length) that could affect the nicotine content of environmental tobacco smoke. 3 The amount of time the infant was actually in contact with smoke, no matter where it was produced, also was not measured. Characteristics of
780
Greenberg et al.
the infants themselves, such as respiratory rate and depth at the time of exposure and differences in metabolism of nicotine, could lead to different concentrations of urine cotinine even though the amount of exposure to tobacco smoke was the same. 1 Despite the preceding limitations, the amount of variation in urine cotinine explained by our model (R 2 -- 0.33) is similar to that of Benowitz and Jacob, ~~who studied adult smokers under conditions in which nicotine intake could be accurately estimated. They found a correlation of 0.62 between nicotine intake and urine cotinine excretion, and they attributed variations in urine cotinine, despite similar nicotine exposure, to differences in individual metabolism. Cessation of smoking is the most obvious effective solution for reducing environmental tobacco smoke. However, it may not always be feasible. Woodward e t a l . 1~ found little success in encouraging and helping parents of newborn infants to stop smoking. A combination of promoting cessation for those who are interested and controlling tobacco smoke for those who are unwilling to stop may be the most realistic strategy available. However, the results from our study suggest that control measures would have to be stringent. Simply blowing smoke away from the infant, going into another room to smoke, or increasing the ventilation in a room will probably not prevent the infant from eventually absorbing tobacco smoke. If people who have contact with the infant must smoke, they might decrease the infant's exposure by smoking outdoors or in areas that do not contribute air to the place where the infant is or might be brought. Although these conclusions and others might reasonably follow from our results, final recommendations for an intervention program should be based on controlled trials. We are grateful to the pediatricians of Alamance County and Chatham County, North Carolina, for supporting our study of their patients. The cooperation of the administrative and nursing
The Journal of Pediatrics May 1989 staff of Alamance County Hospital, Alamance Memorial Hospital, and North Carolina Memorial Hospital is also greatly appreciated. We thank the staff of the Infant Health Study, whose work made this study possible.
REFERENCES 1. U.S. Department of Health and Human Services. The health consequences of involuntary smoking: a report of the Surgeon General. Rockville, Md.: Office of Smoking and Health, 1986; DHHS publication No. (CDC) 87-8398 (pp 38-49, 58-9, 139-42, 145-6, 186-98, 203-6, 211-2). 2. Samet JM, Tager IB, Speizer FE. The relationship between respiratory illness in childhood and chronic air-flow obstruction in adulthood. Am Rev Respir Dis 1983;127:508-23. 3. National Research Council. Environmental tobacco smoke: measuring exposures and assessing health effects. Washington, D.C.: National Academy Press, 1986:25-7, 137-46, 269-76. 4. Perry CL, Silvis GL. Smoking prevention: behavioral prescriptions for the pediatrician, tiediatries 1987;79:790-9. 5. Bauman KE, Greenberg RA, Strecher VJ, Haley NJ. A comparison of biochemical and interview measures of the exposure of infants to environmental tobacco smoke. Evaluation and the Health Professions (in press). 6. Langone J J, Gjlka HB, Van Vunakis H. Nicotine and its metabolites: radioimmunoassays for nicotine and cotinine. Biochemistry 1973;12:5025-30. 7. Schwartz-Bickenbach D, Schulte-Hobein B, Abt S, Plum C, Nau H. Smoking and passive smoking during pregnancy and early infancy: effects on birth weight, lactation period, and cotinine concentrations in mother's milk and infant's urine. Toxicol Lett 1987;35:73-81. 8. Luck W, Nau H. Nicotine and cotinine concentrations in serum and urine of infants exposed via passive smoking or milk from smoking mothers. J PEDIATR1985;107:816-20. 9. Greenberg RA, Haley N J, Etzel RA, Loda FA. Measuring the exposure of infants to tobacco smoke: nicotine and cotinine in urine and saliva. N Engl J Med 1984;310: 1075-8. 10. Benowitz NL, Jacob P. Daily intake of nicotine during cigarette smoking. Clin Pharmacol Ther 1984;35:499-504. 11. Woodward A, Owen N, Grgurinovich N, Griffith F, Linke H. Trial of an intervention to reduce passive smoking in infancy. Pediatr Pulmonol 1987;3:173-8.
Numerous studies have found a direct association between parental smoking and lower respiratory illness in infants. 1 This association appears to be strongest during the first year of life. Vulnerability in early life is of particular concern because of the possible association of acute lower respiratory illnesses during infancy and pulmonary abnormalities in later life. 2 Evidence is also accumulating that
Supported by grant No. 28895 from The National Heart, Lung, and Blood Institute, National Institutes of Health. Submitted for publication Sept. 23, 1988; accepted Nov. 2, 1988. Reprint requests: Robert A. Greenberg, MD, MSPH, Division of Community Pediatrics, CB 7225, Medical School Wing C, University of North Carolina, Chapel Hill, NC 27599-7225.
774
other health problems might also be associated with passive smoking in childhood. 1,3 In response to these conc~ns, health care providers have been encouraged to counsel parents to stop smoking for their own and their children's benefit? Another approach is to reduce the opportunities for infants to absorb tobacco smoke despite persistent contact with people who smoke. Such efforts might be more effective if based on an understanding of the circumstances in which infant passive smoking occurs. We pursued a detailed investigation of those circumstances, including a description of the characteristics of the infants, the smokers, the places of exposure, the amount of smoke absorbed, and their interrelationships. We focused on the first weeks of life, a time when an infant is
Volume 114 Number 5
vulnerable to the harmful effects of tobacco smoke and when intervention and behavior change might be most effective, before smoking habits around the infant are firmly established and when a family's concern for protecting the infant's health may be greatest. METHODS Study population. The infants were enrolled in the study at birth, from April 1986 to May 1987. They are part of an intervention trial to reduce passive smoking, but all data in this study were collected before the intervention. To be eligible for enrollment, an infant had to have no significant postnatal problems, have a birth weight of at least 2000 gin, and reside in Alamance County or Chatham County in central North Carolina. Infants were recruited from three hospitals where approximately 80% of the births in those two counties occur. Although 1096 infants were eligible for the study, some were not included because (1) the mothers did not give informed consent (n = 394), or they gave informed consent but withdrew it before data collection (n = 68); (2) the infants were randomly selected not to participate until a later stage of data collection (n = 166); or (3) they did not produce the urine sample required for inclusion (n = 35). The remaining 433 infants constituted the study sample. Table I describes the study sample. Between the study sample and the remainder of the population that was eligible for enrollment, there were no significant differences (p >_ 0.05) in the sex of the infant, educational level of the head of households educational level of the mother, mother's age, and smoking behavior of the mother. The study population included relatively mor e black mothers: 35% of study mothers were black compared with 24% of mothers of the remainder of the eligible infants (p < 0,001). This difference remained significant after the variable of educational level of the head of household was controlled. The mean _+ SD number of people living in study households, including the infant, was 4.2 _+ 1.2. The most common composition of the households was the infant mother, father, and one or more siblings (38%), followed by the infant and just the mother and father (29%). Data collection procedures. Data collectors visited the mothers of all eligible infants in the hospital soon after birth to request enrollment in the study and to obtain informed consent and demographic data. Data collectors then visited the homes of enrolled infants when the infants were approximately 3 weeks of age (mean age _+ SD 18 _+ 6 days, range 8 to 51 days) and before the family's assignment in the intervention trial. Before beginning the interview, the data collector placed a urine collection bag on the infant and explained to the mother that the urine
Passive smoking by infants
Table
775
I. Description of study sample Study sample (N = 433)
Infants Female (%) Male (%) Mothers Race White (%) Nonwhite (%) Age Mean _+ SD (yr) Range Education Less than HSG (%) HSG only (%) More than HSG (%) Smoking behavior Smokers (%) Mean _+ SD (cigarettes/day) Range (cigarettes/day) Household heads Education Less than HSG (%) HSG only (%) More than HSG (%)
47.1 52.9
64.7 35.3 25.2 _+ 5.4 13 to 45 25.9 37.2 37.0 26.8 13.7 -+ 8.5 2-40
24.3 37.9 37.9
HSG, High schoolgraduate.
would be analyzed for tobacco smoke products to determine whether the baby had been exposed to smoke. The mother was then asked to describe the smoking habits of all household members and any recent exposure of the infant to tobacco smoke. The exposure data were gathered in a systematic manner by a series Of questions? For both "a typical day last week" and "during the last weekend," the mother identified all the places where the infant spent time, including rooms in the home, at other homes, in cars, and in public places. She then estimated the amount of time the infant spent in each place, the size of the place, and the amount of ventilation. For each place the mother reported all smokers who entered while the baby was there; whether they smoked, what, and how much; how close they were to the infant while smoking; and how much time the smokers spent in each place while the infant was there. Definitions of variables. An infant was considered at risk for exposure to tobacco smoke if the infant lived in the same household as a smoker or was in the same place at the same time as a n0nhousehold smoker, either in the house or elsewhere. A smoker was defined as anyone who smoked at least one cigarette, cigar, or pipeful per day. There were six Cigar smokers and one pipe smoker in the study. Exposure to tobacco smoke was said to occur when
776
Greenberg et al.
29%exposedto / smokefrom~ ( non-household smokers~
The Journal of Pediatrics May 1989
~ I
7%~ y
47%exposedto smokefrom mother Figure. Sources
%exposedto smokefrom household members otherthan mother
of smoke for 184 infants exposed to tobacco smoke. Because of rounding, percentages outside circles may not exactly equal the sum of their respective components within circles.
smoke was produced in the infant's presence, that is, in the same room or vehicle as the infant, affording the infant an opportunity to absorb the smoke. The amount of exposure was expressed as cigarettes per week smoked in the infant's presence. This was calculated by multiplying the number Of cigarettes smoked in the infant's presence during "a typical day" of the previous week by 5 and adding the number of cigarettes smoked in the infant's presence during the previous weekend. A cigar or piPeful was considered equivalent to one cigarette. S m o k e absorption was defined as sufficient airway contact with tobacco smoke to result in absorption of nicotine. It was measured-by the urinary concentration of cotinine, the major metabolite of nicotine. Cotinine in body fluids is considered a good indicator of tobacco smoke absorption. 1, 3 The urine samples were frozen and shipped to the American Health Foundation, in Valhalla, New York. Researchers there analyzed Specimens without knowing the exposure status of the infant. Cotinine concentration was quantitated by radioimmunoassay with a modification of the method originally described by Langone et al. 6 This protocol uses specific antisera produced by rabbits and has interassay and intraassay variations of 5%, with a sensitivity of 360 pg/ml (2102 pmol/L). Creatinine was measured by dry chemistry methods on a Kodak Ektachem 400 analyzer (Eastman Kodak Co., Rochester, N.Y.). The cotinine was divided by the creatinine concentration to adjust the data to reflect the concentration of the urine sample. 3 Statistical methods. Multiple linear regression models were used tO identify the correlates of the amount smoked in the infant's presence by each source of exposure to tobacco smoke and the correlates of the infant's urine cotinine concentration. A stepwise procedure, with back-
Laboratoryprocedures.
ward elimination of variables from a full model containing all appropriate candidate independent variables, was applied after adjustment for any collinearity. To remain in a model a variable had to have a partial F test p value <0.05. In the regression model involving the amount smoked in the infant's presence by the mother as the dependent variable, the total number of cigarettes she smoked per day was designated, before data analysis, to be kept in the model. This allowed us to study the effect of the other potential correlates (e.g., educational level of the head of household) after taking into account, or controlling for, the total number of cigarettes the mother smoked per day. Likewise, in the regression model involving the amount other household members smoked in the infant's presence as the dependent variable, the total number of cigarettes smoked daily by other household members was designated as a controlled variable. The significance of first-order interactions of the controlled variable with all candidate independent variables was also tested. The 21 infants of smoking, breast-feeding mothers were eliminated from all analyses that used cotinine as a continuous variable because of th e relatively large amount of nicotine and cotinine absorbed through breast-feeding, compared with the amount absorbed through the airways. 7 The effect on an infant's urine cotinine concentration of the size of the place in which the infant was exposed to tobacco smoke, the ventilation of that place, the time the infant spent there, and the distance between the smoker and the infant was summarized in an index that adjusted the number of cigarettes smoked in the presence of the infant during the previous week. First, the number of cigarettes each person smoked was weighted by the distance of the smoker from the infant. Then the weighted total number of cigarettes smoked in each place was further weighted by the size and ventilation of the place and the time the infant spent there? Distributions of categorical variables were compared by the Pearson chi-square ~statistic. The Cochran-MantelHaenszel chi-square method was used when a categorical variable was controlled. The two-sided Student t statistic was used to compare means of continuous variables for two groups. RESULTS
riskforexposuretotobaccosmoke.
Infants at A total of 276 infants, or 64% of the entire study population, were at risk for exposure to tobacco smoke; that is, the infant either lived in a household with a smoker or was in contact with a nonhousehold smoker. Of these infants at risk for exposure, 42% had a smoking mother, 70% lived in
Volume 114 Number 5
Passive smoking by infants
777
Table II. Final linear regression model* for amount of exposure of infant to smoke produced by mothert Variable
Intercept Total maternal smoking (cigarettes/day) Amourit of exposure to nonmaternal smoking (cigarettes/wk) Maternal age (yr) Place where infant spends maximum time Siblings
Unstandardized coefficient
V a r i a b l e values
Mean • SD 13.7 + 8.5 Range 2-40 Median 14 Range 0-241 Mean _+ SD 24.6 • 5.3 Range 14-38 Bedroom: 1 Other areas: 0 Yes: 1 No: 0
F
p
25.7 l 6 4.752
172.03
0.0001
0.340
22.43
0.0001
-1.939
9.68
0.0024
-18.694
9.35
0.0028
16.146
5.88
0.0170
*Model F = 57.56; df= 5, 110;p = 0.0001; R2 = 0.72. tFor 116 infants of smoking mothers, with control for the total number of cigarettes mother smoked per day.
households with smokers other than their mothers, and 26% had contact with nonhousehold smokers. Thirty-six percent of the infants at risk were in contact with more than one of those categories of smokers. A total of 239 infants, or 55% of the entire study population, lived in households with smokers. Of these households, 61% had one smoker, 33% had two smokers, and 6% had more than two. Infants exposed to tobacco smoke. Of the 276 infants at risk for exposure to smoke, 184 (67%) were reportedly exposed during the previous week. Of the exposed infants, 47% were exposed to smoke produced by their mothers (75% of smoking mothers smoked in their infant's presence), 66% were exposed to smoke from household members other than mothers, and 29% to smoke from nonhousehold smokers (Figure). Forty percent of the exposed infants had more than one category of smoker as their source of exposure. The mean total amount of exposure was 71 cigarettes per week per infant, with a range of 1 to 350 cigarettes per week. Mothers contributed, on the average, 33% of the amount to which an infant was exposed. All household members other than the mother contributed, on the average, 47% of the total amount of exposure, and nonhousehold smokers contributed 20%. The most important correlates of the amount of exposure of the infant to smoke from the mother were maternal smoking, measured by the total number of cigarettes smoked anywhere by the mother each day, and the number of cigarettes smoked in the infant's presence by people other than the mother (Table II). There was more exposure to tobacco smoke produced by the mother if the infant had a younger mother, spent most of the time outside the bedroom, and had siblings. Nonsignificant variables
included the educational level of the head of household, the sex of the infant, the mother's race, whether the infant was breast fed, the number of rooms in the house, and first-order interactions between maternal smoking and other variables. The significant correlates of the amount of the infant's exposure to tobacco smoke produced by nonmaternal household members were the amount of the infant's exposure to maternal smoking and the interaction of the mother's race with the total number of cigarettes nonmaternal household members smoked per day (Table III). An increase in the total number of cigarettes smoked by nonmaternal household members resulted in a greater increase in the amount smoked in the presence of nonwhite infants compared with white infants. The nonsignificant variables were the mother's age, sex of the infant, educational level of the head of household, number of rooms in the home, place where the infant spent maximum time, and first-order interactions between total smoking by nonmaternal household members and other variables, except race. The only significant variable in the regression model of the correlates of the amount of exposure to smoke produced by nonhousehold smokers was the amount of exposure to maternal smoking (p < 0.0001, model R 2 = 0.21). Nonsignificant variables were mother's age and race, infant's sex, educational level of the head of household, place where the infant spent the greatest time, and amount of exposure to smoking by nonmaternal household members. Absorption of tobacco smoke. O f the 433 study infants, 258 (60%) had measurable cotinine in their urine. The median concentration of those excreting cotinine, excluding the 21 infants of breast-feeding, smoking mothers, was
778
Greenberg et al.
The Journal of Pediatrics May 1989
T a b l e III. Final linear regression model* for amount of exposure of infant to smoke produced by nonmaternal household members']" Variable
Intercept Total nonmaternal household member smoking (cigarettes/day) Race of mother:l: Total nonmaternal household member smoking • race of mother Amount of exposure tO maternal smoking (cigarettes/wk)
V a r i a b l e values
Mean _+ SD 20.7 _+ 15.5 Range 1-120 Nonwhite: 1 White: 0
Mean + SD 25.5 + 50.6 Median 0 Range 0-280
Unstandardized coefficient
F
p
2.31
0.1303
1.035
7.08
0.0085
0.306
44.31
0.0001
19.224 0.253
-24.955
*Model F= 20.84; df= 4, 186;p = 0.0001; R2= 0.31. ~'For 191 infants in households with nonmaternal smokers, with statistical control for the total number of cigarettes they smoked per day. :~Thisvariable must be included in the model because it is involvedin the significant interaction variable.
121 n g / m g of creatinine (77.7 nmol/mmol), with a range of 6 to 2273 n g / m g (3.9 to 1459 nmol/mmol). The median and range of cotinine concentrations unadjusted by ereatinine were 9 n g / m l and 2 to 138 ng/ml (51 nmol/L and 11 to 784 nmol/L), respectively. The concentrations of cotinine in the urine of the infants in our study were similar to those in other studies of passive smoking by infants 79 and in studies of adult passive smokers} The significant correlates of the urine cotinine concentration were the amount of smoking the mother did in the infant's presence, the amount she smoked elsewhere, the amount of smoking other household members did in the infant's presence, the amount they smoked elsewhere, and the intensity of exposure to tobacco smoke (Table IV). The last variable was a measure of whether the place where the infant spent the most time coincided with the place where the infant received the maximum exposure. The amount of exposure to smoking by nonhousehold smokers was not a significant correlate. For exposed infants the Spearman correlation coefficient between the number of cigarettes smoked in the infant's presence during the previous week and the urine cotinine concentration was 0.54 (p = 0.0001). The correlation coefficient did not change when the number of cigarettes was adjusted for the size of the place in which the infant was exposed, the amount of ventilation of that place, the time the infant spent there, and the distance between each source of exposure and the infant. DISCUSSION A large proportion of our study infants (64%) was at risk for exposure to tobacco smoke. This percentage is within
the range for households with at least one smoker, reported in studies of older children9 The actual proportion at risk is probably greater than 64% because 37 (9%) of the study infants had cotinine in their urine yet reportedly lived in nonsmoking households and were not in contact with smokers during the previous week. The estimated half-life of elimination of cotinine from the urine of nonsmoking adults is 1 to 2 days, 3 which indicates that those 37 infants probably had contact with tobacco smoke during the previous week. An alternative source of cotinine would be the milk of breast-feeding mothers who used smokeless tobacco products. However, only three of our mothers reported using these products, and they did not breast-feed their infants. We considered an infant to be exposed to tobacco smoke when smoke was produced in the infant's presence (i.e., in the same room or vehicle). Mothers were a major source of tobacco smoke exposure, but other household members and nonhousehold smokers also contributed. The amounts smoked in the infant's pre~ence by maternal and nonmaternal smokers were associated. The amount smoked by the mother in the infant's presence was the most important correlate of the amount smoked in the infant's presence by nonmaternal household members and by nonhousehold smokers. Conversely, the amount smoked in the infant's presence by nonmaternal smokers made an important contribution to explaining the variation in the amount smoked by the mother in the infant's presence. These results are consistent with the thesis that social influences are important to smoking near the infant. The total number of cigarettes a mother smoked near her young infant was strongly related to the total amount
Volume 114 Number 5
Passive smoking by infants
779
T a b l e IV. Final linear regression model* for absorption of tobacco smoke (urine cotinine/creatinine ratio) for 411 infants Variable
Intercept Amount of exposure to maternal smoking (cigarettes/wk) Maternal smoking not in same place as infant (cigarettes/day) Amount of exposure to nonmaternal household member smoking (cigarettes/wk) Comparison of place where infant spends maximum time with place where maximum amount of exposure to smoke occurs Total nonmaternal household member smoking not in same place as infant (cigarettes/day)
V a r i a b l e values
Mean _+ SD 11.7 _+ 33.8 Median 0 Range 0-280 Mean _+ SD 1.5 _+ 3.9 Median 0 Range 0-30 Mean _+ SD 11.7 _+ 26.7 Median 0 Range 0-161 Same place: 1 Other situation: 0
Mean _+ SD 7.0 _+ 12.7 Median 0 Range 0-120
Unstandardized coefficient
F
p
50.168 1.541
23.76
0.0001
13.996
39.06
0.0001
1.227
10.58
0.0012
64.078
6.07
0.0142
1.375
3.99
0.0464
*Model F = 39.64; df= 5, 405;p = 0.0001; R2= 0.33. Not includedare 21 infants of breast-feeding, smokingmothers and one infant with no reported smoke exposure but with the highest urine cotinine/creatinineratio in the study (2273 ng/mg).
she smoked. This relationship may weaken as the infant grows older and they spend less time close together. Thirty-seven percent of the children who absorbed smoke were reportedly not exposed to smoke during the previous week. Although inaccurate maternal recall and reporting, discussed in more detail later in the article, could contribute to this result, other explanations arc that the infant was brought into places where smoking had previously taken~ place, Or that smoke produced elsewhere had dispersed into places Where the infant was. The correlates of smoke absorption accounted for only 33% of the variation in urine cotinine excretion. However, tobacco smoke exposure and absorption should not necessarily be more closely correlated, because they are conceptually distinct. Exposure in our study referred to tobacco smoke in the environment, and absorption referred to a nicotine metabolite excreted in urine a s a reflection of tobacco smoke absorbed. Nevertheless, tobacco smoke exposure and absorption are related, and there are factors that could have affected our measure of that relationship. Reconstruction by the mother of a "typical day" for her infant during the previous week and her recall of events during the previous weekend might have been inaccurate. A recognition that smoking near infants is harmful might have led to an underestimate of the amount of smoking by household members, particularly in the infant's presence. Although the infants were only several weeks of age, there were times during the day when some mothers were not with their infants, or even in the same house; knowledge of
smoke exposure during those times might have been particularly limited. Days that were not considered "typical" by the mothers might have been reflected in the urine cotinine concentration, particularly if those atypical days occurred close to the day the urine specimen was collected. The amount of smoke that reaches an exposed subject is said to be affected by the size of the space into which it disperses, the amount of air exchange in that space, and the distance of the subject from the source of smoke, l However, adjusting the number of cigarettes smoked in the presence of the infant by those factors did not strengthen the relationship between the number of cigarettes and the urine cotinine level. These results might be explained partially by the rapid dispersion of tobacco smoke in a defined space? Our need to rely on maternal recall, especially to estimate ventilation and distance of the infant from each smoker, prevents us from dismissing these factors as unimportant. Assessment of their influence on ~smoke absorption might be better under experimental conditions. Some variables that were not measured in this study could lead to variation in urine cotinine values. These include differences in cigarette nicotine content and individual smoking habits (such as puff depth and frequency and butt length) that could affect the nicotine content of environmental tobacco smoke. 3 The amount of time the infant was actually in contact with smoke, no matter where it was produced, also was not measured. Characteristics of
780
Greenberg et al.
the infants themselves, such as respiratory rate and depth at the time of exposure and differences in metabolism of nicotine, could lead to different concentrations of urine cotinine even though the amount of exposure to tobacco smoke was the same. 1 Despite the preceding limitations, the amount of variation in urine cotinine explained by our model (R 2 -- 0.33) is similar to that of Benowitz and Jacob, ~~who studied adult smokers under conditions in which nicotine intake could be accurately estimated. They found a correlation of 0.62 between nicotine intake and urine cotinine excretion, and they attributed variations in urine cotinine, despite similar nicotine exposure, to differences in individual metabolism. Cessation of smoking is the most obvious effective solution for reducing environmental tobacco smoke. However, it may not always be feasible. Woodward e t a l . 1~ found little success in encouraging and helping parents of newborn infants to stop smoking. A combination of promoting cessation for those who are interested and controlling tobacco smoke for those who are unwilling to stop may be the most realistic strategy available. However, the results from our study suggest that control measures would have to be stringent. Simply blowing smoke away from the infant, going into another room to smoke, or increasing the ventilation in a room will probably not prevent the infant from eventually absorbing tobacco smoke. If people who have contact with the infant must smoke, they might decrease the infant's exposure by smoking outdoors or in areas that do not contribute air to the place where the infant is or might be brought. Although these conclusions and others might reasonably follow from our results, final recommendations for an intervention program should be based on controlled trials. We are grateful to the pediatricians of Alamance County and Chatham County, North Carolina, for supporting our study of their patients. The cooperation of the administrative and nursing
The Journal of Pediatrics May 1989 staff of Alamance County Hospital, Alamance Memorial Hospital, and North Carolina Memorial Hospital is also greatly appreciated. We thank the staff of the Infant Health Study, whose work made this study possible.
REFERENCES 1. U.S. Department of Health and Human Services. The health consequences of involuntary smoking: a report of the Surgeon General. Rockville, Md.: Office of Smoking and Health, 1986; DHHS publication No. (CDC) 87-8398 (pp 38-49, 58-9, 139-42, 145-6, 186-98, 203-6, 211-2). 2. Samet JM, Tager IB, Speizer FE. The relationship between respiratory illness in childhood and chronic air-flow obstruction in adulthood. Am Rev Respir Dis 1983;127:508-23. 3. National Research Council. Environmental tobacco smoke: measuring exposures and assessing health effects. Washington, D.C.: National Academy Press, 1986:25-7, 137-46, 269-76. 4. Perry CL, Silvis GL. Smoking prevention: behavioral prescriptions for the pediatrician, tiediatries 1987;79:790-9. 5. Bauman KE, Greenberg RA, Strecher VJ, Haley NJ. A comparison of biochemical and interview measures of the exposure of infants to environmental tobacco smoke. Evaluation and the Health Professions (in press). 6. Langone J J, Gjlka HB, Van Vunakis H. Nicotine and its metabolites: radioimmunoassays for nicotine and cotinine. Biochemistry 1973;12:5025-30. 7. Schwartz-Bickenbach D, Schulte-Hobein B, Abt S, Plum C, Nau H. Smoking and passive smoking during pregnancy and early infancy: effects on birth weight, lactation period, and cotinine concentrations in mother's milk and infant's urine. Toxicol Lett 1987;35:73-81. 8. Luck W, Nau H. Nicotine and cotinine concentrations in serum and urine of infants exposed via passive smoking or milk from smoking mothers. J PEDIATR1985;107:816-20. 9. Greenberg RA, Haley N J, Etzel RA, Loda FA. Measuring the exposure of infants to tobacco smoke: nicotine and cotinine in urine and saliva. N Engl J Med 1984;310: 1075-8. 10. Benowitz NL, Jacob P. Daily intake of nicotine during cigarette smoking. Clin Pharmacol Ther 1984;35:499-504. 11. Woodward A, Owen N, Grgurinovich N, Griffith F, Linke H. Trial of an intervention to reduce passive smoking in infancy. Pediatr Pulmonol 1987;3:173-8.