P-649 Introduction: INMA-Valencia is a prospective cohort study included in the INMA network seeking to evauate the impat of environment and diet on foetal and early chilhood health. One of its aims is to asses exposure to outdoor nitrogen dioxide (NO2) for the 800 women followed during their pregnancy. Here we present the results mapping the NO2 levels over the study area. Methods: Our cohort is spread out over a great area covering a wide variety of environmental conditions. As part of sampling design, the area was divided in 4 zones, clearly different regarding to population and industrialization criteria: urban (U), metropolitan (M), sub-urban (SU) and rural (R). Each of U, M and S-U areas were grided with a fixed distance beween knots (0.5, 1, 3 km respectively). In the rural zone, one knot in each town center was placed as all of our women in R are living within the towns. 93 sites integrated the entire grid. Outdoor concentrations of NO2 were monitored during 7 days by passive samplers once a trimester (April, June and November 2004 and February 2005), intended to capture seasonal patterns. In each survey, NO2 data from the grid were interpolated using Universal Kriging with a linear drift. Maps were plotted using ArcView 3.2. Home addresses of women were geocoded using public Geopraphic Information Systems. We used cross-validation to compare kriging estimates with the value of the nearby sampler of the grid, and the inverse distance weighted average of neighobour samplers. Results: Median outdoor NO2 kriging estimates were: 40.5 (P10-P90: 23.5, 48.9), 27.9 (16.1, 35.3), 50.2 (33.1, 67.9), 42.5 (19.26, 48.16) μg/m3 in April, June, November and February surveys, respectively. By zone were: 47.2 (P10-P90: 32.5, 57.8), 44.6 (28.9, 59.3), 24.3 (16.7, 37.6) and 14.9 (8.9, 21.1) μg/m3 in U, M, S-U and R, respectively. Cross-validation showed prediction errors for kriging interpolation about of 35.8% of the range in true concentrations, for the neighobour samplers average were about 38.6% and for the nearby sampler about 44%. These results are consistent along sampling surveys. These results are not consistent by zone: In the rural area, nearby samplers provided the best estimates in all surveys. Graphs showed the impact of one motorway crossing the entire area. Discussion and Conclusions: Outdoor NO2 varied spatially within the area of study. We were able to assign levels of NO2 exposure to the women of our cohort: Kriging estimates were used except in the rural area, where the sampler sited in the same town was used. This is a first layer of exposure to be weigthed for covariates controlling the effect of nearby sources. This project was funded by G03/176; ISCIII-FEDER, FIS-FEDER: 03/1615, PI041509.
Read full abstract