Association of Outdoor and Indoor Nitrogen Dioxide with Pulmonary Function in Schoolchildren

The effects of outdoor and indoor nitrogen dioxide (NO2) exposure on pulmonary function were studied in a cohort of children attending eight elementary schools in Chiba Prefecture. A three-year series of annual pulmonary function tests was conducted from 1989 through 1991. NO2 concentration in the living room of each child's home was measured in both the heating period and the non-heating period. Children were classified into four groups according to household annual average NO2 concentration. The indoor NO2 concentration varied with the area of residence and type of heating appliance used. Analyses using log-linear models, including interactions among these environmental factors, were performed to evaluate the association of NO2 exposure with pulmonary function in schoolchildren. Interregional comparison showed that pulmonary function values, adjusted for height and age, were lower among boys living in urban areas, where air pollution levels are high, compared with boys in rural areas. In log-linear models that considered the effect of indoor NO2 concentration, boys in urban areas showed significantly depressed values of FVC and FEV0.75. Indoor air pollution was not definitely associated with pulmonary function among boys, after adjustment for the area of residence. Among girls, high indoor NO2 concentrations were associated with low pulmonary function values, while no significant relationship between area of residence and pulmonary function was shown. Girls in the over-40-ppb group showed significantly depressed values of FVC, FEV0.75 and V25 in the second testing. In the third testing, V25 was the only parameter that was significantly low. These results suggest that pulmonary function is associated with area of residence among boys and with indoor NO2 concentration among girls. However, this study could not reveal the long-term effect of indoor air pollution on pulmonary function, since this association became weaker by the third testing. J Epidemiol, 1994; 4 : 137-146.

The effects of outdoor and indoor nitrogen dioxide (NO2) exposure on pulmonary function were studied in a cohort of children attending eight elementary schools in Chiba Prefecture. A three-year series of annual pulmonary function tests was conducted from 1989 through 1991. NO2 concentration in the living room of each child's home was measured in both the heating period and the non-heating period. Children were classified into four groups according to household annual average NO2 concentration. The indoor NO2 concentration varied with the area of residence and type of heating appliance used. Analyses using log-linear models, including interactions among these environmental factors, were performed to evaluate the association of NO2 exposure with pulmonary function in schoolchildren. Interregional comparison showed that pulmonary function values, adjusted for height and age, were lower among boys living in urban areas, where air pollution levels are high, compared with boys in rural areas. In log-linear models that considered the effect of indoor NO2 concentration, boys in urban areas showed significantly depressed values of FVC and FEV0 .75. Indoor air pollution was not definitely associated with pulmonary function among boys, after adjustment for the area of residence. Among girls, high indoor NO2 concentrations were associated with low pulmonary function values, while no significant relationship between area of residence and pulmonary function was shown. Girls in the over-40-ppb group showed significantly depressed values of FVC, FEV0.75 and V25 in the second testing. In the third testing, V25 was the only parameter that was significantly low. These results suggest that pulmonary function is associated with area of residence among boys and with indoor NO2 concentration among girls. However, this study could not reveal the long-term effect of indoor air pollution on pulmonary function, since this association became weaker by the third testing. J Epidemiol, 1994; 4 : 137-146.
nitrogen dioxide, outdoor air pollution, indoor air pollution, schoolchildren, pulmonary function tests As a result of growing vehicular traffic in recent years, air pollution in urban areas has not been improved substantially1), and anxiety has been aroused over the potential impacts of automobile exhaust on the health of inhabitants.
At present, major air pollutants in urban areas include nitrogen dioxide (NO,) and suspended particulate matter1). The former is also generated by indoor combustion apparatus, and indoor concentrations may exceed the environmental air quality standard in some cases2'3). Therefore, in evaluating the effects of outdoor air pollution on human health, the effects of indoor air pollution from sources such as unvented heating appliances and smoking should be studied in parallel").
As an epidemiological method of studying the effects of This report presents an interregional comparison of the results of pulmonary function tests conducted in 1990 and 1991. We also investigated the association of estimated pulmonary function levels with various environmental factors, including indoor NO, concentration.

Subjects
The study subjects were 1,081 children from eight elementary schools in Chiba Prefecture, the locations of which are shown in Figure 1. The children were in fourth grade as of October 1989.
A total of five schools in Chiba City (A, B), Funabashi City (C, D), and Kashiwa City (E) are located in urban areas ; all these school districts contain main regional highways. The school districts containing the two schools in Chiba City are adjacent to one other, as are the school districts containing the two subject schools in Funabashi City. Three schools in Sawara City (F), Ichihara City (G), and Tateyama City (H) are all located in rural areas where ambient air pollution is at low levels.
The annual average NO, levels in 1991, as measured at the General Air Pollution Monitoring Stations located in close proximity to these schools, were 0.025-0.028 ppm in urban areas and 0.007-0.010 ppm in rural areas15)

Pulmonary function tests
Pulmonary function tests were performed from October to January of each year for three years, from 1989 (fourth grade) to 1991 (sixth grade). For measurements of pulmo- Figure 1. Location of study schools in Chiba Prefecture.
A-E schools (closed circles) and F-H schools (open circles) are located in urban areas and in rural areas, respectively. nary function, an electrospirometer (DISCOM-21, CHEST Inc., Tokyo) was used. Prior to each day of testing, the spirometer was calibrated using a 3-liter syringe. Testing was carried out by three trained technicians.
On the day of testing, a questionnaire was administered concerning the presence or absence of cough, rhinorrhea, sputum, and fever on the day before and day of testing. The viability of tests on children with such symptoms was left to the judgment of doctors. Each subject was given an adequate explanation on the method of testing, and then submitted to repeated testing, with short pauses interposed between, until at least three but not more than eight reproducible forced expiratory maneuvers were obtained. The height of each subject was measured prior to testing. For all children, testing was conducted in the standing position and with a noseclip.
Only forced expirations that met the criteria recommended by the American Thoracic Society 16) were accepted. For children from whom reproducible forced expiratory maneuvers were obtained, the flow-volume curve used for the analysis was one in which the sum of forced vital capacity (FVC) and forced expiratory volume in one second (FEV,,0) was the maximum17). Pulmonary function parameters employed for the analysis were FVC, forced expiratory volume in 0.75 second (FEVO _75), and maximal expiratory flow rate at 25% of FVC (V,5). Values obtained were expressed in terms of the percentage of measured value to predicted value.
The prediction equations for the standard pulmonary function values were derived on the basis of sex, using data set obtained from tests conducted in 1988 on fourth-sixth grade children in the same three rural schools as those in this study. The apparatus and procedure for measurements are the same as the present study. Only data from children who met the selection criteria for normal individuals (444 boys and 381 girls) was included. Height and age (converted to months of age) were used as variables in the prediction equations according to our earlier study18)

Indoor NO, concentration measurements
Indoor NO, concentrations in each chill's home were measured twice : once during the heating period (January 1991), and once during the non-heating period (June 1991). Measurements were made in the living room using a filter badge NO, (Toyo Roshi Inc., Tokyo), and 24-hour average concentrations of NO2 were expressed in ppb19). Details of the measurement procedure are described elsewhere 14).
For homes in which valid measurements (i.e., covering an interval of between 22 and 26 hours) were carried out in the both periods, the geometrical mean of the two seasonal NO, concentrations was calculated according to the method proposed by Neas et al20). This geometrical mean was then employed as the annual average NO, concentration for each home.

Data analysis
In autumn 1988 (for F and H schools) or autumn 1989 (for the other schools), a standard respiratory symptom questionnaire, which was a modified Japanese version of ATS-DLD-78-C21), was administered to all subjects. The questionnaire was filled out by either of the parents, and replies submitted by households yielding valid indoor NO, measurements were analyzed. The following subjects were excluded from further analyses : those with less than 3 years of residence in the current area when this questionnaire survey was conducted, and those with bronchial asthma-like or wheezing symptoms in the 2 years prior to this time.
Pulmonary function values from the second testing (during fifth grade) and the third testing (during sixth grade) were analyzed, because indoor NO, measurements were carried out when the subjects were in the fifth grade (January 1991) and in the sixth grade (June 1991). Values obtained for children from urban and rural areas were compared using the Student's t test. Average indoor NO, concentrations were classified into the following four groups : 0-19 ppb, 20-29 ppb, 30-39 ppb, and 40 ppb or over. Pulmonary function values were evaluated by trend test using regression for four categories of indoor NO, concentrations.
Analyses using a log-linear model") were conducted in order to evaluate the effects of multiple factors affecting pulmonary function. Pulmonary function values were arranged in ascending order and classified into three groups : a low-level group, representing the sub-25th percentile ; an intermediate group, of the 25th to 75th percentiles ; and a high-level group, of the 76th percentile and up. Then the pulmonary function was investigated for associations with area of subject's residence, type of home heating appliance used, and indoor NO, concentration.
An appropriate model was chosen according to Bishop et a123), and odds ratios were estimated using the method described by Yanagawa24). The analyses were conducted using the SPSS-X package programs25). Table 1 provides the breakdown of subjects. For a total of 972 children, comprising 484 boys and 488 girls, reliable measurements of indoor NO, concentration were carried out during the heating and non-heating periods, and completed ATS-DLD-78-C questionnaires were obtained. Of these children, those with less than 3 years of residence (157 children) and those with bronchial asthma-like or wheezing symptoms over the past 2 years (56 children) were excluded from analysis. The final sample size was 759 children, and consisted of 382 boys and 377 girls.

Number of subjects
Children who were absent on the day of testing or who produced unacceptable testing results were subsequently excluded from the analysis : such subjects were 75 children at the second testing, and 63 children at the third testing. After the removal of these outliers, 684 and 696 children were submitted to analysis at the second testing and third testing, respectively. There were no differences between rejected and adopted children in terms of respiratory symptoms ; residential-housing structure ; type of heating appliances ; indoor NO, concentration ; and other factors.
Indoor NO, concentration Table 2 shows results for the two measurements of indoor NO, concentration, for a sample of 972 children. On the whole, the mean of indoor NO, concentrations during the heating period was 34.5 ppb in homes using Table 1. Number of subjects by testing and sex.
* Subjects who were absent on the day of testing or who produced unacceptable testing results according to the criteria of the American Thoracic Society. vented heating appliances, and 63.7 ppb in homes using unvented heaters.
The mean of indoor NO, concentrations was lower during the non-heating period, at 17.4 ppb. In all cities surveyed, indoor NO, concentrations were higher during the heating period than during the nonheating period, and were particularly higher in homes using unvented heaters. Indoor NO, concentrations during the non-heating period were higher in urban areas than in rural areas. The household annual average of two seasonal indoor NO, measurements was 22.6 ppb in homes using vented heating appliances, and 31.7 ppb in those using vented heaters. The annual indoor NO, averages were higher in urban areas than in rural areas regardless of type of heating appliance used.
Pulmonary function tests Table 3 shows the comparisons of pulmonary function values between children from urban and rural areas. In both the second and third pulmonary function testings, both boys and girls from urban areas showed lower pulmonary function values compared with those from rural areas. All differences between urban and rural boys were significant, except for V25 values obtained at the second testing (hereafter, 2nd V25). Significant differences in girls were seen in only 2nd FVC and FEV0 ,,S ; however, none were seen in any parameters from the third testing.  Table 4. In both the second and third testings, boys from high indoor  Analyses by log-linear models Using the log-linear model, the effects on pulmonary function levels of area of residence (A), type of home heating appliance used (H), and indoor NO, concentration (N) were analyzed. Some possible log-linear models are shown in Table 6. In the abbreviated bracket notation, we describe a hierarchical log-linear model including the highest interaction ; e.g., [ Values are mean±SD. * p value by trend test for categories of indoor nitrogen dioxide concentrations . NS, not significant. Similar results were obtained about the other pulmonary function parameters, indicating that the fits of models were satisfactory when the interactive effects of the area of residence and indoor NO, concentration on the pulmonary function were included. Using these models, the odds ratios for depressed pulmonary function levels were estimated in terms of area of residence and indoor NO, concentration.
As shown in Table 7, the estimated odds ratios of boys living in urban areas to those in rural areas were higher than 1 for all parameters except for 2nd V25. The area of residence was significantly associated with 2nd FEV0.75 (odds ratio (OR)= 1.55), 3rd FVC (OR= 1.63) and 3rd FEVu ,75 (OR= 1.90). In both the second and third testings, girls showed odds ratios of less than 1 for FVC and FEV0.75, and higher than I for V25. None of these were found to be significant, however.
Estimates of odds ratios for depressed pulmonary function levels by indoor NO, concentration are shown in Table 8. Boys showed odds ratios of over 1 in most cases.
However, only the 2nd V25 in the over-40-ppb group (OR=2.21) was significant, and no definite tendencies were noted in relation to indoor NO, concentration. In girls, most odds ratios were higher than 1, and significantly high odds ratios were obtained for the following parameters : 2nd FVC in the 20-29 ppb group (OR= 2.13) ; and 2nd FVC (OR= 1.80), FEV0 .75 (OR= 1.96), V25 (OR=2.08), and 3rd V25 (OR= 1.89) in the over-40-ppb group. All odds ratios except for 2nd FVC were highest in the over-40-ppb group.

DISCUSSION
The pulmonary function test has been widely utilized, along with the prevalence of respiratory symptoms, as an epidemiological method of studying the effects of ambient Table 6. Goodness-of-fit of some log-linear models for the associations of pulmonary function with various factors.
GI : the likelihood ratio chi-squared statistic, A : area of residence, H : type of heating appliance used, N : indoor nitrogen dioxide concentration, P, : FEV0 .75 for boys in the second testing, P, : FVC for girls in the second testing. a The abbreviated bracket notation means a log-linear model including the interaction among the variables . # The adopted model , * p<0.05.  Table 8. Estimates of odds ratios (ORs) and 95% confidence intervals (95% CIs) for depressed pulmonary function level by the annual averaee indoor nitrogen dioxide concentration.

By the model [AHN] [AP] [NP].
A : area of residence, H : type of heating appliance used, N : indoor NO, concentration, P : pulmonary function levels.
Odds ratios relftive to children in homes with indoor NO, concentrations of 0-19 ppb.
healthy children after adjusting for asthma and wheezing.
In the present study as well, when children with asthma and wheezing were excluded, a past history of respiratory disease before 2 years of age was not associated with depressed pulmonary function. Among the environmental factors, NO, is a representative ambient air pollutant, and causes lung injury at high concentrations"). Schoolchildren living in areas with high ambient NO, levels have been reported to show low pulmonary function values7. 35) Since NO, is also generated by the use of combustion apparatus in homes, indoor air pollution has been taken up as an environmental factor 2,3.3e) In the United States and European studies4-6,10-13,)the effects of cooking gas have been investigated as a potentially major source of indoor air pollution. In Japan, however, the use of unvented heating appliances has been frequently assessed as a potential source of indoor air pollution, since gas is normally used for cooking in almost all homes 14,39).
Berkey et al. 13) and Speizer et al.40) reported that children from homes where gas was used for cooking had slightly lower FVC and FEV, .0 values than those from homes using electricity, whereas Dodge") and Vedal et a136) reported that the use of cooking gas was not associated with pulmonary function in children. In terms of unvented heating appliances, Nitta et al.") similarly reported that pulmonary function in children showed no difference with type of home heating appliance used. As the above reports, indoor air pollution has not shown a consistent association with pulmonary function. The present study, similarly, showed no definite relationship between type of home heating appliance used and pulmonary function values in children.
As for passive smoking, many studies have reported that maternal smoking habits decrease pulmonary function levels in children 10,13,26> On the contrary, Dodge 12) and Lebowitz et al.") found no relationship between parental smoking habits and pulmonary function among children. In our study, maternal smoking was likely to decrease V25, values, but conversely to increase FVC values among girls. This result was in substantial agreement with the results of Nitta et al35) and Vedal et a.36).
In order to evaluate the effects on health of outdoor and indoor air pollution in combination, direct measurements of personal exposure to pollutants should be desirable 3). Some reports have discussed the association between personal NO2 exposure and individual health 39,12,43). However, few of these reports dealt with a sufficiently large number of subjects. Since an epidemiological study on a large number of subjects involves difficulties in measuring personal exposure levels, indoor NO2 concentrations, which are assumed to contribute largely to overall personal exposure 44>, have been measured in most cases5, 12 suggested that girls would be more likely to subject to increased exposure as they spend more time in the kitchen.
In the present study, girls showed a stronger association between indoor NO, concentration and pulmonary function than did boys. This finding appears to correspond to the findings of Hasselblad et al.10>. Girls may be more affected by indoor air pollution. The association was the strongest in V25, which reflects peripheral airway function. By the third testing, however, the number of significant parameters has decreased among girls, which failed to substantiate the long-term effect of indoor NO2 concentrations. Boys showed different pulmonary function values with varying levels of ambient air pollution, presenting an association between area of residence and pulmonary function. After controlling for indoor NO2 concentration, the association remained significant in FVC and FEV0.,5.
Thus, the adverse effects of outdoor and indoor NO2 may be mediated through different mechanism. The complex mixtures of pollutants present in outdoor and indoor air are clearly distinct under most circumstances. In the urban areas where ambient NO2 is at high levels, the presence of other pollutants, e.g., suspended particulate matter and ozone, with toxicity for the respiratory health would be anticipated.
In the present study, we carried out indoor NO2 measurements on only one day each during the heating-and non-heating periods, and did not survey the children's time activity patterns during testing. Therefore, the effects of outdoor and indoor air pollution on respiratory health should be studied further, including researches on methods of evaluating these effects.