A comparison of population air pollution exposure estimation techniques with personal exposure estimates in a pregnant cohort.

There is increasing evidence of the harmful effects for mother and fetus of maternal exposure to air pollutants. Most studies use large retrospective birth outcome datasets and make a best estimate of personal exposure (PE) during pregnancy periods. We compared estimates of personal NOx and NO2 exposure of pregnant women in the North West of England with exposure estimates derived using different modelling techniques. A cohort of 85 pregnant women was recruited from Manchester and Blackpool. Participants completed a time-activity log and questionnaire at 13-22 weeks gestation and were provided with personal Ogawa samplers to measure their NOx/NO2 exposure. PE was compared to monthly averages, the nearest stationary monitor to the participants' home, weighted average of the closest monitor to home and work location, proximity to major roads, as well as to background modelled concentrations (DEFRA), inverse distance weighting (IDW), ordinary kriging (OK), and a land use regression model with and without temporal adjustment. PE was most strongly correlated with monthly adjusted DEFRA (NO2r = 0.61, NOxr = 0.60), OK and IDW (NO2r = 0.60; NOxr = 0.62) concentrations. Correlations were stronger in Blackpool than in Manchester. Where there is evidence for high temporal variability in exposure, methods of exposure estimation which focus solely on spatial methods should be adjusted temporally, with an improvement in estimation expected to be better with increased temporal variability.