Asthma is a common chronic lung disease in which the airways become inflamed and produce increased mucus, making air flow in and out of the lungs difficult. The prevalence of asthma is higher among children 0–17 years of age compared to adults 18 years of age or older [1, 2]. Asthma is typically diagnosed starting at age 5; however symptoms may appear at younger ages. Children under the age of 5 may be more likely to have asthma if they have airborne allergies, wheeze without a cold, and if the parents have been diagnosed with asthma [2]. In addition to family history, children who had previously acquired respiratory syncytial virus (RSV), the virus primarily detected among infants diagnosed with viral bronchiolitis, were more likely to develop wheeze or asthma throughout both childhood and adulthood [3–5].

Several environmental exposures have suggested to increase the risk of asthma exacerbation, including both indoor and outdoor air quality [6–11]. Ambient air pollutants, including fine particulate matter (PM_2.5), have been shown to play a role in adverse respiratory outcomes among children with asthma [12]. PM_2.5 represents a combination of small particles and liquid droplets from vehicular exhaust, coal and wood burning, and industrial activities.

Measurements of residential distance to major roads and traffic density have been used as indirect estimates of traffic-related air pollutants due to the difficulty of measuring the mixture of traffic-related pollutants [13, 14]. Adverse respiratory outcomes have also been associated with exposure to areas of high traffic density or near major roadways [15–19]. Concentrations of vehicular pollutants have shown to decline to background levels after 150–300 m (m), while asthma prevalence has been reported to be higher among children who live closer to major roads [16, 17, 20]. The effects of air pollution in both warm and cool seasons have been associated with pediatric asthma attacks [21–24].

Although adverse health effects of ambient PM_2.5 exposure have been shown across all populations, infants and young children are sensitive subgroups. The small size of PM_2.5 makes the particles easy to inhale and penetrate deep into the airways [25]. In children, the immune and respiratory systems are not fully developed and their time spent outdoors is greater than adults. Young children also breathe more air per body weight than adults which in turn makes them more susceptible to ambient air pollutant exposures [26].

Notwithstanding some studies associating air pollution exposure and childhood asthma development and exacerbation, it is unclear whether lower levels of PM_2.5 trigger the same responses [27–29]. In addition, many large population-based studies have not studied effect modification between asthma exacerbations and short-term changes in PM_2.5 among potentially susceptible subgroups due to the lack of data. The aims of the study were to investigate the associations between early-life short-term PM_2.5 exposure and the risk of asthma or wheeze exacerbation, and to determine if there was any effect modification. To do this, we employed a case-crossover study design with short-term exposure lags using data on clinical encounters among all children 0–9 years of age in Massachusetts (MA) obtained from the innovative pregnancy to Early Life Longitudinal (PELL_ cohort record linkage system.

Approximately 10% of our study sample had missing PM_2.5 concentrations due to possible cloud or snow cover when the satellite based model was unable to provide estimates for a given day. For lag 0, the IQR for PM_2.5 was 5.9 μg/m³. The mean PM_2.5 concentration was 8.8 μg/m³ and the median concentration was 7.8 μg/m³. Similar distributions were observed for lags 1 and 2.

Per IQR increase in PM_2.5, non-significant increases in the odds of having an asthma clinical encounter were observed for preterm children for lag 1 (OR: 1.03; 95% CI: 0.97, 1.09) and lag 2 (OR: 1.02; 95% CI: 0.97, 1.08) compared to full-term children (lag 1 OR: 0.99; 95% CI: 0.97, 1.02; lag 2 OR: 0.98; 95% CI: 0.96, 1.00). The odds of an asthma clinical encounter were significantly higher (p < 0.05) among children who experienced a clinical encounter in the cold season (OR: 1.02; 95% CI: 0.99, 1.05) compared to the warm season (OR: 0.97, 95% CI: 0.94, 0.99) per IQR increase in PM_2.5 for lag 0 with similar results for lag 1. Among children whose insurance type was not paid through the government a protective effect was observed for lag 2 (OR: 0.97; 95% CI: 0.94, 0.99) compared to those who had government paid insurance (OR: 1.00; 95% CI: 0.98, 1.04) per IQR increase in PM_2.5. Similar results were observed for lags 0 and 1, although not significant (Table 3).

Results for the analysis of children 5–9 years of age are presented in Additional file 1: Tables S1-S3. Characteristics among the 6195 children and mothers in the 5–9 years of age group were consistent with those 0–9 years of age (Additional file 1: Table S1). Again, asthma clinical encounters were not associated with PM_2.5 exposures for lags 0, 1, and 2 regardless of whether only primary and secondary diagnosis fields were used (Additional file 1: Table S2). We observed negative associations for lag 0 when stratifying on gestational age, maternal race, residential distance to roadway, median income, breastfeeding initiation at hospital, and child’s gender. Per IQR increase in PM_2.5, preterm children (< 37 weeks) had significantly lower odds of having an asthma clinical encounter (OR: 0.81; 95% CI: 0.71, 0.92) compared to those were born full-term (> = 37 weeks) (OR: 0.98; 95% CI: 0.94, 1.03). Children who were born to mothers whose race was other than white had significantly decreased odds of having an asthma clinical encounter (OR: 0.91; 95% CI: 0.84, 0.98) per IQR increase in PM_2.5 compared to those born to white mothers (OR: 0.99; 95% CI: 0.93, 1.05). Among children who lived within 150 m from a roadway we did not observe any association per IQR increase in PM_2.5 (OR: 1.00; 95% CI: 0.94, 1.06), whereas a protective effect was observed among those who lived greater than 150 m from a roadway (OR: 0.92; 95% CI: 0.86, 0.98). Similar results were observed across lags 1 and 2, although not significant. Per IQR increase in PM_2.5, those with an annual household median income between $20,000–$70,000 had decreased odds of an asthma clinical encounter (OR: 0.92; 95% CI: 0.86, 0.98) compared to those whose median income was greater than $70,000 (OR: 1.01; 95% CI: 0.94, 1.09) and those who were less than $20,000 (OR: 1.02; 95% CI: 0.87, 1.19). Odds of having a clinical encounter per IQR increase in PM_2.5 were lower for female children (OR: 0.92; 95% CI: 0.85, 0.99) compared to male children (OR: 0.98; 95% CI: 0.93, 1.04). Lastly, children of mothers who initiated breastfeeding at the hospital following delivery were observed to have a protective effect against asthma clinical encounters per IQR increase in PM_2.5 (OR: 0.93; 95% CI: 0.88, 0.98) compared to those who did not initiate breast feeding (OR: 1.03; 95% CI: 0.95, 1.12) (Additional file 1: Table S3).

Several studies have suggested increased asthma exacerbation with short-term PM_2.5 exposure [25] but little has been reported on the subgroups of children with asthmatic symptoms that may be most susceptible to the effects of PM_2.5. Further, a broad range of PM_2.5 concentrations in the various study locations have been reported [27]. In our study area, we report lower levels of PM_2.5 with a median value of 7.8 μg/m³, however it has been suggested that asthma exacerbations may begin to increase at PM_2.5 levels as low as 4.00–7.06 μg/m³ [38]. Our study identified sensitive subgroups that may be more vulnerable to low levels of PM_2.5-induced asthma exacerbation resulting in a clinical encounter. LBW has been associated with a range of health effects in children including respiratory infections, lowered lung function and asthma [39–41]. LBW children have also displayed increased asthma symptoms related to air pollutants [42, 43]. When stratifying on birthweight, we saw increased odds of having an asthma clinical encounter across all lags, and statistically significant differences in the effect of PM_2.5 on asthma clinical encounters for LBW children compared to normal birthweight. This demonstrates that birthweight may play a role in immune development and subsequent asthma development.

In addition to LBW, the effect of PM_2.5 on asthma clinical encounters was significantly different in the cold season compared to the warm season (p < 0.05). Previous studies have identified the effects of air pollution in both warm and cool seasons to be associated with pediatric asthma attacks [21–24].

In this study we did not find smoking during pregnancy to be a significant effect modifier; however, we did observe increased odds ratios among women who smoked during pregnancy compared to those who did not across lags 1 and 2. Several studies have shown that maternal smoking during pregnancy is a modifiable risk factor for asthma [44]. The lack of effect modification by smoking in our analyses may be due to the relatively small number of women who reported smoking during pregnancy (n = 3922) and the under-reporting of smoking by pregnant women, causing misclassification of that variable.

Higher rates of asthma have been reported along major roadways [17] and exposure to traffic-related air pollutants have been associated with an increased risk of asthma hospitalizations [45, 46]. We found non-significant and inconsistent results within our data among children who lived within 150 m of roadways. In our analysis with children 5 and older, we found non-significant increases in odds ratios among children who lived within 150 m of a major roadway compared to those who lived greater than 150 m for lag 2 per IQR increase in PM_2.5, and a statistically significant protective effect among those who lived greater than 150 m per IQR increase in PM_2.5 for lag 0 (Additional file 1: Table S3). In our study, we specifically determined distance to major roadway from the household residence. However, studies have identified near roadway exposures at school may play a role in asthma outcomes as well [47].

Because asthma is best diagnosed at age 5 or older, we performed a secondary analysis restricting our cases those who had a clinical encounter at 5 years of age and older. We chose not to use these cases as our primary analysis because our case definition for asthma or wheeze included children who had the most severe symptoms leading them to seek care in the emergency department, and the majority of our cases were under the age of 5. Among those children who had a clinical encounter at age 5 or older, the association between PM_2.5 exposure and having an asthma clinical encounter was significantly different among preterm children compared to the full-term children for lag 0 (p < 0.05). We observed significantly decreased odds of asthma clinical encounters among pre-term children (OR: 0.81; 95% CI: 0.71, 0.92) compared to full-term children (OR: 0.98; 95% CI: 0.94, 1.03) per IQR increase in PM_2.5 (Additional file 1: Table S3). Preterm children may have other health issues that are typically treated more often in outpatient clinical settings, and therefore may have their asthma well-controlled and are seen less often in the emergency department. In addition, older children may have already been diagnosed with asthma in an outpatient clinical setting, or may have been seen in the emergency department for asthma or wheeze prior to age 5.

Our study tested several susceptible subgroups, and although we did not formally test for multiple comparisons, our significant findings exceeded the 5% type 1 error rate, or false positive rate, that we would have expected under the null hypothesis of no associations. Of the 42 tests between effect modifiers and PM_2.5 exposure among children 0–9 years of age, there were 5 (11.9%) significant interactions found (Table 3). In the analysis with children 5–9 years of age, there was only 1 significant interaction found between PM_2.5 and gestational age and therefore was within the 5% type 1 error rate and may have been an expected finding due to chance (Additional file 1: Table S3).

Our study design allowed our cases to serve as self-matched controls which provided several strengths. We did not need to control for modifiable asthma risk factors such as secondhand smoke. We were able to adjust for short-term meteorological changes that affect PM_2.5 concentrations. Another major strength of this study was our ability to use satellite-based measurements to obtain estimates of PM_2.5 at a 4-km resolution instead of relying on data from a relatively small number of fixed monitoring stations. Using short-term lags was reasonable since we would expect parents to seek medical attention for poorly controlled asthma or exacerbation of asthma or wheeze within the same day, or a day or two after symptoms begin. Another strength of this study was that the PELL data afforded both the large sample size and the record linkage necessary to exclude children who moved between their time of birth and clinical encounter. Exposure was assigned based on residential address so using the street-level birth address and matching that to the zip code at the time of their clinical encounter reduced exposure misclassification. Demographic characteristics among the subgroup that moved (including race, income, and maternal education) were consistent with the characteristics of those that did not move.

Although one advantage of using a case-crossover design is that there is no confounding by time invariant factors, one limitation is that the study design doesn’t allow for independent estimation of the effects for time invariant characteristics. Another limitation of this study was that our cases included only children who had more severe or poorly controlled asthma with severe enough symptoms to seek clinical care, and did not include children who are seen in outpatient clinical settings. Our study strictly focused on PM_2.5 exposure, however there has been accumulating literature focusing on traffic-related air pollutants (TRAP) as a whole instead of the independent effects of individual air pollutants [48]. Although we used distance to major roadway measurements to account for traffic-related air pollutants, further investigation of other air pollution components may be warranted.

Our study identified susceptible subgroups of children that may have increased risk for asthma clinical encounters resulting from lower-level PM_2.5 exposure. We found that LBW children had increased odds of an asthma or wheeze clinical encounter across all lags. Birthweight was a significant effect modifier at α = 0.05. We also found the season of clinical encounter to be a significant effect modifier for lags 0 and 1. This study suggests LBW children may be more susceptible to the effects of PM_2.5. We did not find strong evidence of PM_2.5 exposure and increased risk of asthma or wheeze clinical encounters among other subgroups.