Published ahead of print on February 16, 2006, doi:10.1164/rccm.200510-1552OC
© 2006 American Thoracic Society doi: 10.1164/rccm.200510-1552OC
Parental Smoking and Lung Function in ChildrenAn International StudyInstitute Environmental Health, ZPH, Medical University, Vienna, Austria; Institute for Risk Assessment Sciences, Utrecht University, Utrecht, The Netherlands; Department of Environmental Health, Harvard School of Public Health, Boston, Massachusetts; Environmental Health Unit, National Center of Public Health Protection, Sofia, Bulgaria; GSF–National Research Center for Environment and Health, Institute of Epidemiology, Neuherberg, Germany; Department of Health Statistics and Informatics, Regional Authority of Public Health, Banska Bystrica, Slovakia; Public and Environmental Health Research Unit, London School of Hygiene and Tropical Medicine, London, United Kingdom; National Institute of Environmental Health, "József Fodor" National Center for Public Health, Budapest, Hungary; Institute of Public Health, Ostrava, Czech Republic; and Department of Epidemiology, Institute of Occupational Medicine and Environmental Health, Sosnowiec, Poland Correspondence and requests for reprints should be addressed to Prof. Manfred Neuberger, M.D., Kinderspitalgasse 15, A-1095 Wien, Austria. E-mail: manfred.neuberger{at}univie.ac.at
Rationale: Both prenatal and postnatal passive smoking have been linked with respiratory symptoms and asthma in childhood. Their differential contributions to lung function growth in the general children's population are less clear. Objective: To study the relative impact of pre- and postnatal exposure on respiratory functions of primary school children in a wide range of geographic settings, we analyzed flow and volume data of more than 20,000 children (aged 6–12 yr) from nine countries in Europe and North America. Methods: Exposure information had been obtained by comparable questionnaires, and spirometry followed a protocol of the American Thoracic Society/European Respiratory Society. Linear and logistic regressions were used, controlling for individual risk factors and study area. Heterogeneity between study-specific results and mean effects were estimated using meta-analytic tools. Main Results: Smoking during pregnancy was associated with decreases in lung function parameters between –1% (FEV1) and –6% maximal expiratory flow at 25% of vital capacity left (MEF25). A 4% lower maximal midexpiratory flow (MMEF) corresponded to a 40% increase in the risk of poor lung function (MMEF < 75% of expected). Associations with current passive smoking were weaker though still measurable, with effects ranging from –0.5% (FEV1) to –2% maximal expiratory flow (MEF50). Conclusions: Considering the high number of children exposed to maternal smoking in utero and the even higher number exposed to passive smoking after birth, this risk factor for reduced lung function growth remains a serious pediatric and public health issue.
Key Words: child • passive smoking • pregnancy • spirometry • tobacco
Several studies have reported that prenatal and/or postnatal passive exposure to smoking adversely affects the lung function of children. A review summarized much of the earlier literature (1). More recent articles on environmental tobacco smoke (ETS) and lung function in children were either of a general nature (2, 3) or concentrated on specific subgroups of children with presumed higher vulnerability (e.g.,
Details of substudies (12–16) and their selection (17) have been given before. Some studies (North America, Austria, Germany, and the Netherlands) were designed to have all participating children undergo lung function testing, although in the Central European Study on Air Pollution and Respiratory Health (CESAR; Hungary, Poland, Slovakia, and Czech Republic), only the older children were recruited for lung function testing. Thus, the eligible population tested for spirometric outcomes comprised 22,712 schoolchildren (Table 1).
Lung function testing was performed according to the protocol of American Thoracic Society (18), except for minimum exhalation time of 6 s (not feasible for children) and except for the Dutch study, which followed the protocol of the European Respiratory Society (19). In addition to continuous flow and volume variables, "poor lung function" was analyzed (16, 20), specified by FVC (< 85% of predicted by age, sex, height, and weight from all study participants of the same country), FEV1 (< 85% predicted), low PEF (< 75% predicted), or maximal midexpiratory flow (MMEF; < 75% predicted). These cut points were chosen because they were anticipated to define the group of subjects with the lowest 5% lung function. This assumption was fairly correct for FVC, FEV1, and PEF, whereas MMEF exhibited a larger variability. Prediction formulas were developed for each country separately, according to the same regression model. The natural logarithms of lung function variables were regressed on the logarithms of age and weight, and an interaction between sex and the logarithm of height (16, 21–23). With this model, the exponentials of the effect estimates of exposure can easily be interpreted as percentage differences. Questions on household smoking and smoking during pregnancy had been asked by comparable questionnaires (17), except for "smoking during the first 2 yr of the child's life," which was unavailable for Austria and Germany. For Germany, "smoking in the first year of life" was available, and was used instead. Thus, three different exposure variables were considered: in utero, early postnatal, and current exposure. Pooled study results are presented: (1) the "basic model," adjusted for age, weight, height, and sex, for all children; (2) the "partially adjusted model," additionally adjusted for the potential confounders described in the next paragraph; (3) the "fully adjusted model," additionally controlled for the other smoking variable (smoking during pregnancy controlled by current ETS exposure and both current and first year's exposure controlled by smoking during pregnancy); in order to study possible interactions, selected analyses were performed stratified by sex, age group, doctor-diagnosed asthma, and (for postnatal exposure) by smoking during pregnancy. We controlled for seasonal trends (by adding a dummy for the four seasons) and potential confounders: recent respiratory infections, current medication, maximal parental education, household crowding, unventilated gas/oil/kerosene heater, mold, birth order, and "ever had a pet." We also controlled for technicians and spirometers where applicable. Children with a nationality or native language different from each country's main population had been excluded a priori from the CESAR datasets (12) and were therefore also excluded from the other studies (< 10% of children) in the adjusted models. Sensitivity analyses assessed the impact of unavailability of covariates, by comparing (in studies that did have the variables) results with and without adjustment.
Using linear regression (or logistic regression for the binary outcomes), study-specific effects were estimated by Stata software (version 8; StataCorp, College Station, TX). Finally, possible geographic clustering of outcomes was accounted for by including an "area" term as a random effect (xtreg and xtlogit commands). Study-specific estimates and their standard errors were entered into a meta-analysis, from which Forest plots, a mean estimate, and a measure and Cochran
Participation rates were high in most substudies (Table 1). Of 35,390 subjects participating in the studies with lung function testing, we analyzed the data of 22,712 children with valid lung function data and information on smoking exposure (Table 1). Because passive smoking was not of primary interest in the original studies, the percentage of missing values was rather high in some of the studies. Especially for "ETS exposure in the first 2 yr of life," this percentage ranged from 0.3% in North America to 20.5% in Hungary. In the Central European Studies, by design, only the older children participated in lung function testing. So, for example, in Poland, spirometry was performed in 1,472 children, whereas the questionnaire information was available for 1,862 children. Among children who underwent spirometry, acceptable curves were obtained for 640 children (43.5%). Some of them had missing exposure data, leaving 626 children eligible for analysis. For the pooled analysis, either the "clean" datasets (only including eligible children) or the total datasets (from Poland, Hungary, Austria, and Germany) were available. In these four total datasets, there was no significant difference in any passive smoking variable between those children with and without valid lung function data. The Austrian data stemmed from routine school health survey and thus exhibited a high participation rate, but concerns were raised about the accuracy of the lung function results due to the possibly poor cooperation of some (e.g., handicapped) children in a total population of elementary school children. Therefore, a variable indicating the cooperativeness of the child was added. This was assigned by the technician who performed the lung function testing and who was unaware of the passive smoking status of the child. Children rated as cooperating poorly were then excluded from the analysis. Because of poor cooperation and of invalid lung function values, approximately 20% of the Austrian children were not eligible for the study. Among the children with poor cooperation, the number of those with missing information on passive smoking exposure (31%) was significantly higher than in the other children (18%). Approximately half of the children were recruited from the American 24-City study. The age range of the children differed among studies with older children (9–12 yr) in the United States and the CESAR countries, and in younger elementary school children (mostly 6–10 yr) in Austria. In the CESAR countries, the lung function tests were only performed in spring; in the other countries, these were performed throughout all seasons.
Smoking during Pregnancy In the basic model, "smoking during pregnancy" was associated with a significant decrease in all lung function parameters except FVC. For the other lung function parameters, the point estimates for the percentage difference ranged from –1.1% for FEV1 to –5.4% for MEF25 in the meta-analyses, with no indication of heterogeneity between studies except for MEF25 (Table 2). After controlling for possible confounders (partially adjusted model), the point estimates were reduced by 10 to 20%, but remained significant (Table 2, Figure 1). Adjusting for "current ETS exposure" (fully adjusted model) had practically no influence on the estimates (Table 2). Despite little evidence of heterogeneity, the more conservative random-effect estimates are reported for the other models.
Logistic regression in the basic model showed 31% (PEF) to 40% (MMEF) increased risk of poor lung function for children exposed to maternal smoking in utero. In the partially adjusted model, the estimates were lower (15–34%) but mostly remained significant (Table 3).
ETS Exposure after Birth ETS exposure after birth—in the first 2 yr and currently—showed a high concordance in all countries except Germany (Table 1). The associations of these two indices of postnatal exposure with lung function parameters were very similar and significant (Table 4), except for FVC. Effect estimates were smaller than those from "smoking during pregnancy" and ranged from –2.5% (MEF25) to –0.4% (FEV1) for "current passive smoking in the household" and from –2.6% (MEF25) to –0.5% (FEV1) for "passive smoking in the first 2 yr." Although controlling for confounders decreased the already small effects, they remained significant on PEF, MMEF, MEF25, and MEF75, and there was little indication of heterogeneity across countries (Figures 2 and 3).
In the basic model, significant effects on poor lung function were seen for PEF and MMEF. The increases in poor PEF and MMEF were stronger for ETS exposure in the first years of life (24 and 27%) than for current exposure (23 and 20%). These estimates were reduced in the partly adjusted model (17 and 23% for the exposure in the first years and 13 and 16% for current exposure; the 13% increase in low PEF with current exposure was no longer significant, p = 0.078). When controlling for smoking during pregnancy, effects of current ETS exposure were reduced to –0.1 (FEV1) and –0.7% (MMEF). Results were no longer significant. Only for PEF (–0.5%) was the criterion for significance nearly met (p = 0.054). Effect estimates for exposure in the first 2 yr were in a similar range (–0.3% for FEV1 and –0.7% for PEF and MEF50). The results were not significant (p = 0.08 for PEF, and up to 0.5 for the other parameters) and displayed some evidence of heterogeneity between countries. In the fixed-effects model (data not shown), the risk estimates were somewhat higher (up to –1% for MEF50), and reached significance for MEF50 and FEV1.
Subgroup Analysis ETS exposure in the first 2 yr was associated with smaller lung function values, independently of smoking status during pregnancy. Smoking during pregnancy did not have significantly different effects in the age groups 6 to 8, 9 and 10, and 11 and 12 yr. The effect modification of sex was studied by stratifying by sex and by introducing an interaction term. Both methods indicated no influence of sex on the effects of smoking during pregnancy except for PEF where the effect of smoking during pregnancy was significantly stronger (p = 0.008) in boys by approximately 1%. Current ETS exposure tended to affect boys more, but this difference did not reach significance. Within the subgroup of children with doctor-diagnosed asthma, no clear association was seen between current ETS exposure and pulmonary function (data not shown). But smoking during pregnancy had an apparently stronger effect on the children without asthma (in the basic model): in children with asthma, differences ranged from –0.7% (FEV1) to –4.2% (MMEF), and in the nonasthmatic children from –1.1% (FEV1) to –5.9% (MEF25).
Sensitivity Analysis The Austrian dataset was the second largest with generally somewhat smaller effect estimates and with the youngest group of children. Effect estimates for smoking during pregnancy on exclusion of the Austrian data were a little stronger for the flow variables. The exclusion of the Austrian dataset did not change the results for current ETS exposure. Inclusion or exclusion of the confounders missing from the Austrian dataset ("crowding" and "unventilated heaters") also did not change any findings. Without Austrian and German data (with and without overcrowding, unventilated heaters, and birth order), the results were quite similar to those obtained when only excluding the Austrian data.
In recent years, considerable attention has been focused on the harmful effect of involuntary "secondhand" or "passive" smoking, confirming ETS as a serious and substantial public health problem, with particular impact on respiratory health of children (25, 26). The PATY project offered the unique opportunity to study effects of in utero exposure to maternal smoking and postnatal ETS exposure on children's lung function in a variety of environmental settings in Europe and North America, differing in confounding factors. Despite geographic, social, and environmental differences, the results showed a high consistency across all substudies, with nearly no indication of heterogeneity. The effect estimates found were similar to those from other studies: Cook and colleagues (1) performed a meta-analysis of up to 21 studies, but they did not consider smoking during pregnancy and current ETS exposure of children separately. Separate analyses were provided by Li and colleagues (7) (also discerning between boys and girls) and Gilliland and colleagues (3) (also discerning by the number of smokers in the household and by past and current exposure, and giving combined estimates for smoking during pregnancy and current ETS exposure). Smoking during pregnancy was more influential and effects were stronger on flow parameters than on FEV1. No effects were seen on FVC, which is also in line with previous findings. ETS exposure was not of primary interest in any of the PATY studies, but was asked about as a possible confounder in air pollution studies. This may have reduced the recall bias. In contrast to studies on respiratory symptoms, recall bias was not an issue for pulmonary function as the outcome variable. Selection bias was an issue at two stages:
The PATY study (on air pollution effects) focused on children because they are usually not exposed to many confounding exposures like occupational hazards or active smoking. In the older children, some active smoking cannot be ruled out entirely and cannot be assessed correctly by parental questionnaires. The smoking habits of the family tend to influence the smoking behavior of the adolescent (27). Therefore, in the subset of the older children, those with ETS exposure at home might tend to try cigarettes earlier and more frequently. If this is the case, the effect of ETS exposure on lung function would appear to be stronger in older children. Because this was not seen in this study, the influence of active smoking of the (older) children was not strong. Authors of other studies have stressed the point that the effects found were rather small. In this study, this is true for the population's mean lung function decrement, but not for the percentage of children with a "poor" lung function. The definition of poor lung function was somewhat arbitrary but yielded case frequencies not uncommon in clinical practice. The choice of another definition would not have changed the results substantially. We found that a 4% decrease in MMEF (from "smoking during pregnancy" in the basic model) corresponds to a 40% higher risk of having clinically poor lung function. Not only do changes in the mean predict changes in the number of cases with extreme values (28) but small decreases of the mean in the general population of healthy schoolchildren are associated with a relevant increase in the number of children with clinically poor lung function. The effects of smoking during pregnancy as well as of current ETS exposure became smaller after controlling for possible confounders. But factors like crowding or socioeconomic status (measured by the parents' education) might also influence the strength of the ETS exposure, whereas factors like "recent respiratory infections" might already have been influenced by ETS. Therefore, controlling for confounders could have led to some overadjustment, which should be kept in mind when reading the adjusted results. Effects of current ETS exposure became smaller and generally statistically nonsignificant in the fully adjusted model. The same finding was reported by Gilliland and colleagues (3). Maybe some of the current ETS exposure effects were taken over by the stronger exposure indicator of smoking during pregnancy, which could also be an indicator for heavy current smoking (of the mother/in the household). In the small subgroup of children exposed to maternal smoking during pregnancy, current ETS exposure was associated with higher lung function values. This was not the case for smoking exposure in the first years of life. Therefore, continued smoking in the household led to improvements in lung function (which does not seem likely) or the parents quit smoking after the first years of life of their child if their smoking was associated with respiratory problems of the child. In such a case, the respiratory problems are the cause and the change in smoking behavior is the outcome ("healthy child's smoking parents effect" analogous to the "healthy worker effect"). In cross-sectional studies, such a change in smoking behavior in reaction to a child's health problem biases the effects of current ETS exposure to the null. Because of the small number of mothers who smoke during pregnancy but not afterwards, the effect of prenatal smoking only is difficult to estimate and could be overestimated in comparison to postnatal smoking if mothers who smoke during pregnancy smoke more cigarettes than others (26). In general, however, we expect an underestimation of risk in this cross-sectional study due to exposure misclassification (29) and the behavioral changes discussed above. School-age children stay out of home more often than very young children. Therefore, smoking behavior in the household is no longer a very good indicator of the child's true exposure, leading to misclassification and, in case of higher exposures outside the home (30), to an underestimation of current ETS exposure (as compared with early life or even fetal exposure). Unfortunately, no bronchodilatation was performed in children with poor lung function. Such measures are not feasible when investigating large groups of (mostly) healthy children. The amount of reversibility of the bronchial obstruction would have indicated the actual contribution of acute versus persistent effects. The influence of current ETS exposure remained significant for the peak flow, whereas effects of smoking in pregnancy were strongest on the mid- and endexpiratory flows. Early (pre- and postnatal) ETS exposure might exert its effects especially on the small airways (31), whereas current exposure (at school age) seems to also lead to (reversible?) obstruction of the large airways. There was some evidence for a greater effect in boys, but this was not strong. The same finding was reported by Cook and colleagues (1) from their meta-analysis. The effect of current ETS exposure was not stronger in children with doctor-diagnosed asthma. The effects of smoking during pregnancy (in the basic model) were actually stronger in nonasthmatic children. This could be due to underdiagnosis of asthma in children from low socioeconomic status where ETS exposure tends to be higher. In contrast Li and coworkers (7) reported stronger effects of in utero exposures in children with asthma. For current and past ETS exposure, they found influences of asthma divergent by sex (increased susceptibility in boys with asthma and nonasthmatic girls), whereas increased susceptibility of nonasthmatic boys in this study reached significance for effects of prenatal exposure on PEF. No clear trend with age was evident. This is interesting for "smoking during pregnancy" for which a dilution of the effect with increasing age would have been expected. Other studies also found effects of in utero exposure to maternal smoking on children of different age (31), ranging from birth (32) until the age of 16 yr (2) and adulthood (9, 10, 33, 34), with little evidence of a change in the magnitude of the effects. Together, this suggests a lasting, possibly permanent effect of in utero exposure (35).
Conclusions Poor lung function in childhood has lasting effects because it predicts a worse prognosis of asthma in adulthood (34). Approximately 60% of all children were exposed to ETS in the first years of life, 50% at the time of questionnaire, and 20% in utero. The population's average decrements found were small, but the risk of a clinically poor lung function was substantial. Considering the high number of exposed children, this indicates that both ETS exposure and smoking during pregnancy remain a severe public health problem.
The authors thank all scientists and technical staff involved in the planning and conduction of the studies, and the children, parents and teachers for their cooperation.
The PATY study was supported by the EU Fifth Framework Quality of Life Program (proposal no. QLRT-2001-02544). U.G. was supported by a research fellowship within the postdoctoral program of the German Academic Exchange Service (DAAD). Originally Published in Press as DOI: 10.1164/rccm.200510-1552OC on February 16, 2006 Conflict of Interest Statement: None of the authors has a financial relationship with a commercial entity that has an interest in the subject of this manuscript. Received in original form October 3, 2005; accepted in final form February 16, 2006
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
1-antitrypsin deficiency [
2 test of between-study heterogeneity were obtained. Study-specific estimates are assumed to follow a random distribution with a common mean. The estimation of this mean and its confidence interval takes into account both the variation in study-specific estimates and the uncertainty (due to sampling variability) related to each study-specific estimate (
