The kappa coefficient for the test-retest was 0.81, meaning high agreement between mothers’ answers and reliability of the instrument used. Table 1 shows the demographic data of the groups analyzed (JDM, monocyclic, polycyclic/chronic and control groups).

Smoking, occupational exposure and prematurity were associated with JDM in the univariate analysis of logistic regression models; no association was found between child’s secondhand smoking and JDM (Table 2).

JDM patients were more frequently exposed to factories or quarries at less than 500 m from the residential address compared to the control group (48.5% vs. 13.7%, p = <0.001). The same was found for the presence of factories/quarries at a distance < 500 m from work address during pregnancy (35.7% vs. 10%, p = 0.02), the address of daycare/school (20.8% vs. 6.45%, p = 0.04) and of residential address after birth (46.66% vs. 10.48%, p < 0.001). In the univariate analysis, the presence of emitting sources of pollutants farther than 200 m from work address and/or home during pregnancy was associated with JDM (OR = 2.36; IC95% 1.04–5.33; p = 0.04). Distances less than 100 m and between 100–200 m were not associated [(OR = 1.33; CI95% 0.50–3.55, p = 0.57) and (OR = 1.35; CI95% 0.53–3.41; p = 0.53)]. But distance farther than 200 m from the sources of pollutants and the address of daycare/school showed a negative association with JDM (OR = 0.24; CI 95% 0.05–0.74; p = 0.013).

There was no difference between the two groups in terms of weight at birth (88% vs. 89% of patients were born weighing between 2500 g and 4000 g, p = 0.73, respectively) and weight gain during pregnancy (50% vs. 48.39% of the mothers showed ideal weight gain, p = 0.49, respectively).

Exposure to O3, NO2 and SO2 during pregnancy, PM10 and SO2 in the first quarter, NO2, SO2 and CO in the third quarter, showed association with JDM with a p < 0.20 in the univariate analysis. In the multivariate analysis just between the pollutants during pregnancy, O3 exposition (second tertile 77.84–84.94 μg/m3, OR = 0.15, IC95% 0.04–0.54, p = 0.004) and PM10 (second tertile 38.69–49.41 μg/m3, OR = 3.95, IC95% 1.2–13.02, p = 0.024) persisted significantly associated with JDM.

Exposure to O3, PM10, SO2 and CO in the fifth year was associated with JDM in the univariate analysis and exposure to O3 was associated in the multivariate analysis (third tertile > 86.28 μg/m3, OR = 5.43; IC95% 1.29–22.7; p = 0.02).

Table 3 presents the multilevel logistic regression performed to identify possible risk factors for JDM diagnosis during gestation using all variables in the univariate analysis that showed p < 0.20 and the tropospheric pollutants that showed p < 0.05 in the multivariate analysis. In this analysis, maternal exposure to some occupational pollutants during pregnancy was found to be significantly associated with JDM. Table 4 presents the final multilevel analysis with the post-birth variables associated with JDM. In this analysis, exposure to O3 was found as a risk factor for JDM in the fifth year of life and sources of pollutants farther than 200 m from daycare/schools was a protective factor.

Regarding maternal smoking during pregnancy, there was no difference between the two groups (6.6% vs. 17.6%, p = 0.60, respectively). There was also no difference when evaluating the fetal smoking (40% vs. 59%, p = 0.48), as well as the child’s exposure to secondhand smoking after birth (37% vs. 40%, p = 1.00). In the univariate analysis in the logistic regression models, there was no association between maternal smoking (OR = 0.33; CI95% 0.03–3.31; p = 0.37) or secondhand smoking during pregnancy (OR = 0.33; CI95% 0.11–1.92; p = 0.29). There was no difference when assessing maternal occupational exposure during pregnancy (25% vs.17%, p = 1.00). Regarding the presence of sources emitting inhalable pollutants close to home (address during pregnancy and after birth), close to the mother’s work address during pregnancy and close to daycare/school, there was no statistical difference between the two groups (p > 0.05). In the univariate analysis, we found no association in the assessment of the presence of factories (OR = 2.86, CI95% 0.67–12.11, p = 0.15) and gas stations (OR = 0.28, CI95% 0.06–1.20, p = 0.09) near the address of the pregnancy. Prematurity was not different between the two groups (27% vs. 5%, p = 0.15, respectively) and was not associated with the disease course in the univariate analysis (OR = 0.153, CI95% 0.01–1.55; p = 0.19). Weight at birth (92.30% of vs. 83.55% of patients were born weighing between 2500 g–4000 g, p = 0.59) and weight gain during pregnancy (54.54% vs. 42.85% of mothers had an ideal weight gain, p = 1.00) showed no difference.

The pollutants that showed p < 0.20 in the univariate analysis were O3, PM10, NO2 during pregnancy, NO2 and CO in the first trimester, PM10, NO2, in the second trimester and O3 and NO2 in the third trimester. Only exposure to NO2 in the third trimester was associated with the monocyclic course in the multivariate analysis of pollutants.

Table 5 presents the final multivariate analysis performed to identify possible risk factors for the monocyclic course, including gestational and fetal factors. In this analysis, no variable was associated with disease course.

Exposure to O3 and CO (second tertile 1.32–2.44ppm, OR = 7.87, IC95% 1.10–56.12; p = 0.04) two years before diagnosis, PM10, NO2 one year before the diagnosis, PM10 and NO2 (second tertile 68.73–96.75µg/m3, OR = 0.12; IC95% 0.01–0.97, p = 0.047) one year after diagnosis and NO2 and CO two years after diagnosis was associated with monocyclic course in univariate analysis, but none remained association in multivariate analysis with disease course (p > 0.05).

Mean age at diagnosis was comparable between patients with refractory and non-refractory disease (6.3 ± 2.8 years vs. 5.3 ± 2.1 years, p = 0.37), as well as mean age at diagnosis and the frequency of female subjects (50% vs. 58.8%, p = 0.74). Mean age at symptom onset was 4.6 (±1.95) years and median of 4.5 years in the group of non-refractory patients and 5.6 (±2.7, p = 0.20) years and median of 5.3 years in refractory patients (p = 0.20).

Regarding maternal smoking during pregnancy, there was no difference between the two groups (11.7% vs. 12.5%, p = 1.00). There was also no difference when assessing maternal secondhand smoking (56.2% vs. 41.2%, p = 0.49), as well as in the analysis of child exposure to secondhand smoking (30.7% vs. 48.8%, p = 0.70). In the univariate analysis, maternal smoking (OR = 0.93; CI95% 0.11–7.55; p = 0.95), maternal secondhand smoking during pregnancy (OR = 0.54; CI95% 0.14–2.17; p = 0.39) and the child’s exposure to secondhand smoking (OR = 1.69; 95% CI95% 0.35–8.22; p = 0.52) were not associated to refractoriness and were not used in the multivariate evaluation (p > 0.20).

No difference was found regarding maternal work during pregnancy between the two groups (16.6% vs. 33.3%, p = 0.60). In the univariate analysis, maternal occupational exposure (OR = 0.87; CI95% 0.13–11.50; p = 0.87) was not associated with refractoriness.

The presence of factories/quarries and gas station at less than 500 m from the home during pregnancy and after birth, from the maternal work address during pregnancy, and from the daycare/school were not different between the two groups (p > 0.05). The univariate analysis of the presence of gas stations at less than 500 m from the mother’s work address during pregnancy showed OR = 0.07; CI95% 0.005–1.06, p = 0.055, the other variables showed p > 0.20. Prematurity (23.5% vs. 5.8%, p = 0.33, respectively), weight at birth between 2500 g–4000 g (86.61% vs. 91.67%, p = 1.00, respectively), ideal maternal weight gain during pregnancy (50% vs. 50%, p = 1.00, respectively) showed no difference between refractory and non-refractory group. In the univariate analysis, prematurity was not associated with refractoriness to treatment (OR = 4.92; CI95% 0.49–49.6, p = 0.18).

The tropospheric pollutants that showed p > 0.20 in univariate analysis were O3, PM10, NO2 during pregnancy, PM10 in the first trimester, O3, PM10, NO2, CO in the second trimester and PM10 and SO2 in the third trimester. None of them were associated with refractoriness in the multivariate analysis.

The variables involved in fetal exposure that were associated with refractoriness (p < 0.20) in the univariate analysis were the presence of gas stations less than 500 m from the mother’s work address during pregnancy and prematurity. However, in the multivariate analysis, there was no association with refractoriness to treatment (p = 1.00).

Further multivariate analysis was not performed as there was no association between tropospheric pollutants and refractoriness to treatment in the univariate analysis and the only variable associated with postnatal exposure that could be included in the analysis was prematurity (p = 0.18).