Results

PREVALENCE DATA

Prevalence questionnaires were returned for 22 968 primary (83% response, 51% boys) and 27 826 secondary school pupils (86% response, 51% boys). The prevalence of wheeze in the past year was 15.1% in primary and 19.0% in secondary school children, and prevalence of currently active diagnosed asthma was 8.8% in primary children and 8.3% in secondary school children.

TRAFFIC FLOW DATA

Schools with the highest TAIs were located mainly within urban areas of Nottingham (fig 1). Primary school TAIs (given to the nearest million) ranged from 0 to 117 million vehicle metres/day/km² with one clear outlier of 247 and one missing value. Secondary school values ranged from 8 to 84 million vehicle metres/day/km², with one outlying value of 153. These outliers were initially excluded from the linear regression analyses of schools, but analyses were repeated with the outliers included. The tertiles used to define the exposure categories are given in the keys to figure 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Map of south Nottinghamshire showing (A) primary schools, and (B) secondary schools by traffic activity index (TAI).

ECOLOGICAL ANALYSIS OF SCHOOL PREVALENCE AND TAI

Individual primary school prevalences of current wheeze varied significantly from 3.1% to 21.5% (p<0.001) and secondary school prevalences from 11.9% to 26.7% (p<0.001). There was no association between TAI and primary school prevalence of wheeze (weighted regression coefficient β=−0.01, p=0.93, fig 2 A) and a non-significant negative relation for secondary schools (β=−0.18, p=0.26, fig 2 B). The findings for asthma were similar, with school prevalence estimates wide ranging (1.5%–17.9%, p=0.003 for primary schools and 4.4%–13.0%, p=0.004 for secondary schools), but showing no significant positive relation with TAI (β=0.10, p=0.28 for primary schools and β=−0.26, p=0.09 for secondary schools). Inclusion of the outlying schools in the analyses resulted in coefficients near zero for both primary (β=−0.08 and −0.02 for wheeze and asthma respectively) and secondary schools (β=−0.02 and 0.04 respectively).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Prevalence of wheeze in past year for (A) primary schools and (B) secondary schools by traffic activity index (TAI).

LOGISTIC REGRESSION ANALYSES OF INDIVIDUAL CHILDREN

The logistic regression analysis of wheeze status for individual primary school children resulted in no significant relation between wheeze in the past year and TAI, fitted as a continuous variable (p=0.82, table 1). However, when TAI was fitted as a factor to look for a non-linear effect, significantly higher odds ratios (ORs) of wheeze were found for medium and high TAI than for low (age and sex adjusted ORs 1.11 (95% CI 1.02 to 1.22) and 1.13 (1.03 to 1.24) table 1). A similar relation was found between asthma and TAI (OR of 1.16 (1.02 to 1.31) for medium TAI, and 1.16 (1.02 to 1.32) for high TAI, relative to low TAI). The relation between TAI and recent cough, examined in the random subsample of survey respondents (n=4570), was slightly stronger and also significant with ORs of 1.21 (1.02 to 1.44) for medium TAI and 1.22 (1.02 to 1.45) for high TAI.

The case-control set of 2648 primary school cases and 3928 controls (78% response) was used to control for social class and the major confounders in this data set, which were preterm birth (born ⩾3 weeks early), maternal age, and breast-feeding. After adjustment, there remained no significant effect of TAI, expressed as a continuous variable, on wheeze, asthma, or cough. Adjusted ORs for medium and high tertiles of exposure remained increased at 1.08 and 1.11 relative to low TAI, although in this smaller sample, the significance of the effect was reduced. Control for these confounding factors in the random subsample resulted in ORs of 1.02 (0.78 to 1.34) and 1.06 (0.80 to 1.41) for asthma, and 1.11 (0.93 to 1.34) and 1.15 (0.95 to 1.39) for recent cough. However, the ORs for asthma adjusted for age and sex only were lower in this subsample, and adjustment for confounders had only a small modifying effect.

There was no significant linear association between TAI and self reported wheeze in the past year in the secondary school children, or in the case-control sample of 5936 secondary pupils (58% response), before or after adjustment for age, sex, social class, and maternal smoking in the first 5 years of life, the risk factor in this data set (table 2). No significant association was found for TAI analysed as a factor, and if anything, the ORs for medium and high exposure were slightly reduced (table 2). No significant relations between asthma or cough and TAI were found.

Among both the primary and secondary school children with recent wheeze, the risk of severe wheeze was positively associated with TAI (adjusted ORs of 1.03 and 1.02 per 10 unit increase in TAI), but the linear associations did not reach significance (table 3). Inspection of the ORs for the tertiles showed non-significant increases in risk in those exposed to medium and high traffic activity compared with low activity.

LONGITUDINAL DATA

Of the 1330 children now attending secondary schools in the county who reported a history of wheeze in our 1988 survey, we were able to obtain data on 883 (66%) and assign a TAI to 765 children currently attending study schools. Persistence of wheeze was reported in 46% of this sample, and was found to have a weak linear relation with TAI (age and sex adjusted OR 1.05 (1.00 to 1.11) per increase in 10 million vehicle metres/day/km², p=0.06, table 4). Adjustment for social class, parental smoking, or early life factors made little change to the relation.

Discussion

It is important to establish whether exposure to traffic pollution has a significant effect on respiratory health in children. Previous studies of this question have been of low power, or have tended to use subjective self reported measures of traffic exposure. We have conducted a study of a large sample of school children, with an objective measure of traffic density as a marker of exposure to the real life pollution mix arising from vehicle exhausts. Our ecological analysis at a school level showed substantial variation in the prevalence of wheeze between both primary and secondary schools in our study, which, although likely to be due in part to differences in diagnostic recognition and labelling, suggests that local environmental exposures have a major influence on the prevalence of asthma. However, our findings also suggest that very little of this variation between schools was attributable to differences in traffic activity in the locality of the schools. Although this does not rule out any effect of traffic activity on the risk of wheeze, it does suggest that other factors are likely to be more important. As the use of ecological analyses has certain problems, particularly the difficulty of dealing with confounding at the level of the individual child as opposed to the level of the population unit, we also carried out analyses at an individual level to determine whether local traffic activity has an effect on a child's risk of wheeze with control for potential confounders. We found no evidence of a linear dose-response relation between TAI and wheeze, asthma, and cough. We did, however, identify evidence of a non-linear relation between exposure to traffic activity and occurrence of wheeze in primary school children, whereby children in areas of TAI in the middle and highest tertiles had a greater risk of wheeze than those in the lowest tertile, but with no difference in risk between the middle and highest tertiles. Similar effects were found for asthma and cough. This investigation of non-linear effects of TAI was a secondary analysis in this study and in view of the non-significant continuous relation, these findings may not be genuine. Nevertheless, if valid, these findings would have important implications for future studies, and we therefore include these results here.

To allow comparisons of our prevalence data with those from previous surveys in the same area, we needed to use the same questions to elicit information on symptoms as in our previous studies.11 12In fact these were very similar to the internationally validated questions used in the international study of asthma and allergies in childhood (ISAAC) studies.17 To measure exposure, we have concentrated on traffic activity in the vicinity of the child's school as children spend about one third of their daylight hours at school, a time when pollution levels are highest. But as a child's total exposure is also influenced by home and other exposures not measured here, some degree of error in assigning traffic exposure to school populations is inevitable. Further misclassification may have arisen from the estimation procedure used to assign flows to some roads, or from unmeasured influences from adjacent gridsquares. However traffic flows were measured blind to school prevalence and are therefore likely to be unbiased and any misclassification error would therefore tend to have resulted in underestimates of the true size of effects. This could explain the weaker associations among the secondary school children, who tend to live further away from their school than primary school children and are therefore subject to greater misclassification. We explored the possibility of confounding by factors including social class, preterm birth, maternal age, low birth weight, older siblings, parental smoking, and breast feeding, and found these to have only a small influence on each of the relations found. As data on these early life factors were collected retrospectively, there may be a problem with poor recall, which if less common among those with respiratory problems, could have resulted in bias.

The lack of a significant effect of TAI fitted as a continuous variable on risk of wheeze in primary school children is inconsistent with the significant plateau effect identified for TAI fitted as tertiles, and therefore places doubt on the stability of the results at the level of an individual child. Furthermore, the significance of the non-linear effect is based on SEs that are likely to be underestimated as a result of applying exposure from an area level to an individual child. Therefore the plateau effect found may simply be a chance result, or due to unmeasured confounders, and should be interpreted with caution. Even if a true effect, reductions in traffic activity would have to be extreme to have any public health impact and such reductions are unlikely to be achievable.

The few previous studies to look at the relation between symptoms of asthma and objective measures of traffic flow in preadolescent children have reported significant positive associations between reporting of respiratory symptoms and traffic counts, although none has investigated the shape of the exposure-response curve for symptoms.7 8One possible explanation for the plateau effect found in this study in primary school children is that traffic flows in the Nottingham area are generally high and may have reached a level of exposure for most of the population in which the exposure-response curve is relatively flat. The size of the effects identified are fairly small and amount to about a 10% increase in risk of asthma symptoms at levels of exposure above 22 million vehicle metres/day/km², although this may be underestimated by misclassification. It is also possible that TAI, which is a measure of exposure for a 1 km² area, may not reflect the most relevant exposure of an individual child. There is evidence to suggest that more localised exposure could be important as substantial contrasts in concentrations of oxides of nitrogen, carbon monoxide, and particulates have been reported to exist across distances of <100 m.18 Most previous studies looking at fairly localised measures of exposure to traffic have reported positive effects on respiratory health.8 10 19-23 However, many of these used self reported measures of exposure, which are likely to be biased, and one large study of prescribing rates for asthma showed no effect.24 Therefore there is a need to further explore the weak effect of exposure to traffic found in this study with a more precise measure of the localised exposure.

Our additional studies of severity and persistence of wheeze suggest a possible dose-related effect of traffic activity on severity among symptomatic school children, which supports previous reports of the effects of pollution on asthmatic children,25-27 and a similar effect on the persistence of wheeze into adolescence. However, these analyses were subject to relatively low power, and the effects found did not reach significance. Furthermore, in the longitudinal study we were only able to look at traffic activity near the child's current school, which is not a very accurate estimate of exposure for those children who moved during the follow up period. When the analysis was repeated on just those who had lived in their current house for at least 7 years (n=546, 71%), the OR for high TAI was reduced to 1.05. Some loss to follow up also occurred, which may have resulted in selection bias. Therefore, the associations found in these subsets need to be explored in larger studies.

To conclude, these findings suggest that the level of traffic in the 1 km² area surrounding a child's school is not an important determinant of the prevalence of wheeze in a school. Our study suggests that if there is any relation between traffic density and disease risk, it is that primary school children living in the lowest tertile of exposure in Nottingham (<22 million vehicle metres/day/km²) experience slightly lower risk of wheeze and cough symptoms than children exposed to higher levels. The fact that the effect of traffic activity in a child's local area on asthma symptoms seems to be only small, may be due to insufficiently large contrasts in pollution levels at this scale across our study area, in which case the effects of more localised exposure at home and school need to be explored in future studies.