Statistical analyses
To study the association between fetal growth and PM10, we applied robust linear regression models for the continuous out- come birth weight and logistic models for the binary outcome SGA. Since we expected the effects of PM10 exposure and covari- ates on birth outcomes to depend on gestational age, analyses were performed within the following three groups: extremely preterm ( o 32 weeks), moderately preterm (32–36 weeks), and term ( 4 36 weeks) births.