Air pollution and development of asthma, allergy and infections in a birth cohort

Study population

The PIAMA prospective birth cohort study had an initial enrolment of 4,146 children recruited during the second trimester of pregnancy from a series of communities of variable size in the Netherlands 27, 28. The study protocol was approved by the institutional review board of each participating institute and informed written consent was obtained from all participants. At inception, the study included a group of 855 high-risk children who were randomly allocated to control and intervention groups in which allergen-free mattress covers were used. In addition, the study population included a natural history group of 3,291 children, including 472 high-risk children. High-risk children were identified before birth by obtaining information regarding self-reported asthma or reporting of physician-diagnosed allergy to house dust, house dust mites, pets or pollen in the expecting mothers. The current analysis included all children (natural history and intervention) who remained in the cohort at 4 yrs of age.

Exposure assessment

Details of exposure assessment have been reported elsewhere 29–31, 33. Briefly, air pollution concentrations at the home address at birth were calculated by a model combining air pollution measurements with a Geographic Information System (GIS). Fine particles (defined according to PM_2.5), NO₂ and soot (determined as the reflectance of the PM_2.5 filters) were measured at 40 sites designed to capture the maximum variability in pollution from traffic sources. At each location, measurements were conducted for four 2-week periods dispersed throughout 1 yr (March 1, 1999 to April 20, 2000; corresponding to a period when the cohort was on average 2–3 yrs old) and then adjusted for temporal trends to calculate long-term average concentrations. The adjustment consisted of: 1) calculation of an annual average for a single reference site where 14-day integrated samples were made for the entire study period; 2) calculation for the reference site of the difference between the annual average and the measurements for each 2-week sampling period at all other sites; 3) addition of the difference for each sampling period to the measured concentrations at each site during this period; and 4) calculation of an annual average from the adjusted values 31. Therefore, this approach to exposure assessment emphasises spatial contrasts in long-term average concentrations during the first 4 yrs of life, since it was hypothesised that it would have greater relevance for the development of chronic respiratory disease than short-term temporal variations in concentrations. At the same locations, GIS data were also collected regarding traffic in the vicinity of each monitoring location. Regression models were developed to relate the annual average concentrations measured at the 40 monitoring sites with the GIS variables; e.g. the number of high-traffic roads within a 250-m radius of a location, the presence of a major road within a distance of 50 m, the density of buildings within a 300-m radius and an indicator for the region of the country were used in the model to predict soot concentrations. Models with similar variables describing traffic intensity were developed for PM_2.5 and NO₂. These models explained 73, 81 and 85% of the variability in the annual average concentrations for PM_2.5, soot and NO₂, respectively 29–31. These models were applied to the same GIS variables measured at the home addresses of each cohort member in order to obtain individual long-term ambient air pollution exposure estimates. Exposure was assessed for the address at birth to reflect the present authors’ interest in the effects of early life exposures on respiratory health.

Questionnaire data

From questionnaires completed by the parents of cohort members, the same health outcomes as in the present authors’ previous analysis 26 were selected to describe symptom frequencies and physician diagnoses for indicators of asthma, allergic disease and infections (table 1⇓). Questionnaires were completed during pregnancy, at the time the child was 3 months old and at the time of the child's first, second, third and fourth birthday. Each questionnaire contained a core set of questions that were supplemented with additional questions as the cohort aged, i.e. additional variables describing wheeze and asthma development were added to the current analysis (table 2⇓). The online supplementary material lists symptoms, diagnoses, exact questions, definitions of early-onset transient, late-onset and persistent wheeze, and a definition of “possible/probable asthma” based upon reported symptoms and other conditions. The definition was developed in the PIAMA project according to guidelines of the Dutch General Practitioners Association, with the relevant questions identical to those of the International Study of Asthma and Allergies in Childhood 34. Analyses were conducted with the same potential confounding variables as in the present authors’ previous study (table 3⇓). The data for these variables were selected, from the earliest questionnaire available, to coincide with the estimated exposures for birth addresses. Details are presented in the legend of table 3⇓.

Sensitisation data

A blood sample was taken at 4 yrs of age in a total of 713 children who also had a home visit at 3 months of age and who completed the year 4 questionnaire. Consent for blood sampling was obtained from all parents. Blood sampling was offered to children participating in the intervention study and those children in the natural history part of the cohort who had had a home visit at 3 months of age 35. Total and specific immunoglobulin (Ig)E levels for house dust mite, cat, dog, Dactylium, birch pollen, Alternaria, egg and milk were measured. IgE was determined by a radioallergosorbent test-like method according to the standard operating procedure used at the Sanquin Laboratories (Amsterdam, The Netherlands). Positive reactions were defined as those with IgE levels >0.35 IU·mL⁻¹. Sensitisation was defined as elevated IgE to any of the measured allergens, any indoor allergens, any of the outdoor allergens or any of the food allergens.

Statistical analysis

The association between exposure and health outcomes was analysed by multiple logistic regression, with adjustment for confounding variables. Sensitivity analyses were conducted for the subset of the cohort who remained at the birth address at 4 yrs of age and those who remained at the same address since their first birthday. While the primary analysis focused on the relationship between health outcomes at 4 yrs of age and risk factors in the first year of life, a separate analysis was also conducted, including risk factors information (gas stove, pets, siblings and home dampness) from the questionnaire at 4 yrs. Odds ratios (OR) are presented for an interquartile range increase in air pollutant concentrations equivalent to 3.3 µg·m⁻³ for PM_2.5, 0.58×10⁻⁵ m⁻¹ for soot (corresponding to 0.84 μg·m⁻³ elemental carbon) 30 and 10.6 µg·m⁻³ for NO₂. A stratified analysis was also performed on the subset of the population with sensitisation measurements in order to assess whether the association between air pollution and symptoms was stronger in children who were sensitised.

Prevalence of health outcomes

At the time of the first questionnaire (3 months of age), 212 children had dropped out of the study (n = 3,934). At 1 yr the cohort size was 3,745, and was 3,730 at 2 yrs. At 4 yrs, data were collected from 3,538 children, corresponding to 85.3% retention. Of those children remaining in the study population at 4 yrs of age, 608 (17.2%) were in the intervention arm, representing a slightly lower proportion (19.3%) than at the initiation of the cohort. The main reasons for dropout were loss of interest, lack of available time and subjects moving and becoming untraceable. Although complete data was not available for any of the covariates, subjects who dropped out of the study (see online supplementary material) were more likely to have allergic mothers, parents with less than a medium level of professional education, mothers who smoked during pregnancy and were more likely to live in homes with a smoker, mold and an unvented gas heater. These subjects also were less likely to have been breastfed at 3 months of age, were born to younger mothers and were more likely to live in the western part of the Netherlands. Finally, those subjects that dropped out had higher estimated exposures to traffic-related air pollutants.

Health outcomes are presented separately for those variables defined by the questionnaire or sensitisation measurements conducted at 4 yrs of age (and referring to diagnoses and symptoms experienced in the fourth year of life; table 1⇑) and cumulative indicators of disease (table 2⇑) based on the experiences during the first 4 yrs of life. The highest prevalence of symptoms at 4 yrs of age was found for ear/nose/throat infections (table 1⇑). The prevalence of wheeze was about three times higher than doctor-diagnosed asthma.

Of the 3,538 subjects who completed the questionnaire at 4 yrs, 1,836 (52%) subjects resided in the same house since birth and 2,497 (71%) resided in the same house from 1–4 yrs of age. Prevalence of health outcomes were very similar for these subsets of the cohort (data not shown) relative to the full cohort. A slightly elevated prevalence for early wheeze was observed in the subset remaining at the birth address (23.6%) and those residing at the same address from age 1 yr (22.7%) relative to the full cohort (21.8%).

The prevalence of sensitisation to any allergen was high and primarily due to sensitisation to food allergens (28.1%), with relatively few children (8.2%) being sensitised to outdoor allergens. It is noteworthy that this prevalence cannot be compared with the very low prevalence of self-reported allergy in table 2⇑, since the sensitisation data were available for a subgroup of children in which children born to allergic mothers were overrepresented 35. As a result, the subpopulation with blood sampling differed from the full cohort in some respects. The online supplementary material presents prevalence of the health outcomes and potential confounding variables for this subpopulation. At 4 yrs of age, these children reported significantly higher prevalence of wheeze, physician-diagnosed asthma, cough, bronchitis, itchy rash and physician-diagnosed eczema in the previous 12 months than did the full cohort. Further, this subpopulation was more likely to be female, to have older siblings, to have been breastfed for ≥3 months and to have a lower prevalence of smoking during pregnancy, smoking in the home, pets in the home and fathers with less than a medium level of professional education. These differences suggest that the subset was likely to be at greater risk of developing allergic disease but were more aware of relevant risk factors and took steps to minimise such risks. The tendency for families with children at increased risk for the development of allergic diseases, based on maternal allergy status, to avoid known risk factors has been reported previously for this same cohort 27.

Prevalence of potential confounders

Table 3⇑ lists the distribution of potential confounders in the study population. Prevalence of smoking was 24%. Most households used gas for cooking (79%). Unvented gas water heaters were present in only 3% of the homes. Virtually all children were of Dutch ethnicity. There were no major differences in the prevalence of potential confounders between the full cohort and the subset who remained at the birth address.

Exposure to air pollutants

Exposures were successfully calculated for the home addresses of 3,532 of the 3,538 children (99.8%) and are summarised in table 4⇓. For six subjects, exposures were not calculated due to an inability to geo-code their home addresses or an error in the GIS calculation of one variable. The exposure estimates for the different pollutants were very highly correlated. The correlation between soot and NO₂ was 0.96, and between soot and PM_2.5, 0.97. The correlation between NO₂ and PM_2.5 was 0.93. There were only very minor differences in exposure (<2% for median or 90th percentile exposures) between the full cohort and those subsets who remained at the birth address. Given the similarities in exposure, prevalence of potential confounders and health outcomes between the full cohort and those who did not move, no sensitivity analyses of the associations between air pollution exposure and the main outcomes were conducted for these groups. Such sensitivity analyses were performed for the sensitisation relationships given the small size of the group with sensitisation measures.

Associations between air pollution exposure and health outcomes

The association between exposure to air pollution and development of respiratory symptoms and sensitisation is reported in tables 5⇓–⇓7⇓. Before adjustment for confounders, air pollution was associated with increased incidence of wheeze, dry cough at night, ear/nose/throat infections and flu/serious colds (table 5⇓), based on responses referring to the fourth year of life only. After adjustment for the full set of confounders, the associations between air pollution and increased risk of wheeze (e.g. OR (95% confidence interval (CI)) 1.18 (0.98–1.41) per interquartile range increase in soot), ear/nose/throat infections (e.g. 1.15 (1.01–1.31) for soot) and flu/serious colds (e.g. 1.18 (1.02–1.36) for soot) remained elevated, but the OR for dry cough at night was somewhat decreased (e.g. 1.13 (0.97–1.30) for soot). Sample sizes for the adjusted OR were reduced due to incomplete questionnaires for some of the confounding variables. When analyses were restricted to adjustment for early-life risk factors, results were similar with somewhat elevated OR for ear/nose/throat infections (e.g. 1.16 (1.03–1.31) for soot), flu/serious colds (e.g. 1.19 (1.04–1.37) for soot) and physician-diagnosed asthma (e.g. 1.30 (0.98–1.71) for soot; table 5⇓).

Sensitisation to any common allergen was positively associated with air pollution, both before and after adjustment, with substantially elevated ORs (e.g. adjusted OR (95% CI) 1.45 (1.11–1.91) for soot; table 6⇑). The association between sensitisation and air pollution was only present for food allergens. As mentioned previously, this subset was composed of children with higher familial risk of allergy but who lived in lower-risk home environments. Air pollution exposures for the subset did not differ from those of the full cohort.

Table 7⇑ presents the results of analyses in which information from different years was combined to produce cumulative measures of disease status. In the unadjusted analysis, significant associations were found between air pollution exposure and any reporting of wheeze (wheeze ever) and with probable asthma, while there were also indications of associations with physician-diagnosis of asthma and with persistent wheeze. In the adjusted analysis, the sample size was substantially reduced, because only those subjects who completed all questionnaires during the 4 yrs of follow-up were included (n = ∼2,500). After adjusting for potential confounders, the relationship with wheeze ever remained significant (e.g. OR (95% CI) 1.18 (1.04–1.34) for soot) and a significant association with physician-diagnosis of asthma was observed (e.g. OR (95% CI) 1.26 (1.02–1.56) for soot). The association with probable asthma observed in the unadjusted analysis disappeared, while a relationship with early-onset transient wheeze emerged. Associations with persistent wheeze remained borderline significant. No associations were found with doctor diagnosis of bronchitis, eczema or reported itchy rash.

Symptom ORs for nonsensitised and sensitised children were very similar. Since air pollution exposures were assessed only for the birth addresses, sensitivity analyses were also conducted by restricting to those subjects who remained at the birth address until 4 yrs of age. Adjusted ORs for sensitisation to food allergens were somewhat reduced, but were still greatly elevated for those who had never moved, and were similar to the full cohort for those who had not moved from 1 yr of age.