The INMA (INfancia y Medio Ambiente; Environment and Childhood) Project is a network of birth cohorts in Spain aiming to study the impact of environment on pregnancy outcomes and child growth and development (Guxens et al. 2011
). Our study used data from four population-based birth cohorts that are part of the INMA project. These four cohorts—Asturias, Gipuzkoa, Sabadell, and Valencia—are located across eastern and northern parts of Spain (). The data for these four cohorts were collected prospectively during 2003–2008 using a common protocol and included a wide range of maternal and fetal characteristics (e.g., objective measures of gestational age by ultrasound examination), biological samples, and environmental measurements (e.g., air pollution) (Guxens et al. 2011
). Pregnant women who fulfilled the inclusion criteria [age ≥ 16 years, singleton pregnancy, no use of assisted reproductive techniques, intention to deliver at the reference hospital, and ability to speak and understand Spanish or a local language (e.g., Catalan or Euskara)] were recruited during the first trimester of pregnancy at primary health care centers or public hospitals. They were then followed throughout the pregnancy and their infants were followed from birth until 2 years of age. Additional information on the cohorts and data collection has been published elsewhere (Guxens et al. 2011
INMA birth cohorts and biogeographic regions across the Iberian Peninsula. Source: Mapa de series de vegetación de España, Spanish Ministry of Agriculture, Food and Environment (1987).
All participants gave written informed consent before enrollment in the cohorts. Each cohort obtained ethical approval from the ethical committee in its corresponding region.
The Iberian Peninsula encompasses two biogeographic regions with distinct climates and vegetation patterns () (Alcaraz-Segura et al. 2009
). The Eurosiberian region covers a narrow ridge across the northern part of the peninsula and is characterized by a humid climate with high water availability year-round, relatively cold winters, and maximum vegetation during summer months (Alcaraz-Segura et al. 2009
). The rest of the peninsula is considered a Mediterranean region, characterized by a dry climate with hot and dry summers, mild and rainy winters, and maximum vegetation between autumn and spring (Alcaraz-Segura et al. 2009
Of the four INMA cohorts included in our study, two (Asturias and Gipuzkoa) were located in the Eurosiberian region and two (Sabadell and Valencia) in the Mediterranean region (). To achieve maximum exposure contrast, we obtained data for surrounding greenness during the maximum vegetation period of the year for the corresponding biogeographic region of each cohort. We therefore abstracted surrounding greenness for Asturias and Gipuzkoa participants during the summer season and for Sabadell and Valencia participants during autumn to spring.
To determine the surrounding greenness, we used the Normalized Difference Vegetation Index (NDVI) derived from the Landsat 4–5 Thematic Mapper (TM) data at 30 m × 30 m resolution (Dadvand et al. 2012a
), which was obtained from the Global Visualization Viewer of the U.S. Geological Survey (2011). NDVI is an indicator of greenness based on land surface reflectance of visible (red) and near-infrared parts of spectrum (Weier and Herring 2011
). Its values range between –1 and 1, with higher numbers indicating more greenness. The Landsat TM data were acquired for year 2007, the most relevant year to the data collection periods of the cohorts (2003–2008), on days during the greenest months for each cohort when clear-sky (cloud-free) satellite data were available, specifically, 29 June for Asturias, 30 May for Gipuzkoa, 26 January for Sabadell, and 9 February for Valencia ().
NDVI maps of Asturias (June 29th), Gipuzkoa (May 30th), Sabadell (January 26th), and Valencia (February 9th) during 2007. Source: U.S. Geological Survey (2011).
For each participant, surrounding greenness was abstracted as the average of NDVI in buffers of 100 m, 250 m, and 500 m around her place of residence, which was geocoded according to the address at time of delivery (Dadvand et al. 2012a
; Donovan et al. 2011
We used separate linear mixed models with adjustment for potential confounders and a random cohort effect to estimate the change in birth weight (grams), head circumference (millimeters), and gestational age at delivery (days) associated with a 1-interquartile range (IQR) increase in surrounding greenness. Random intercepts were were used to adjust for potential confounding by unmeasured cohort characteristics (Chu et al. 2011
). The IQR was derived from the pooled distribution of all cohorts.
All analyses were adjusted for maternal age (continuous), ethnicity (white/other), socioeconomic status [Clasificación Nacional de Ocupaciones (CNO-94; three categories) (Domingo-Salvany et al. 2000
)], education level (none or primary/secondary/university), smoking (yes/no), alcohol consumption (yes/no), parity (0/1/≥ 2), infant sex (male/female), and season of conception (spring/summer/autumn/winter) (Dadvand et al. 2012a
). For birth weight, the analyses were also adjusted for gestational age at delivery, maternal pregestational body mass index (BMI), weight gain during pregnancy, and paternal BMI. Analyses of the head circumference were further adjusted for gestational age at delivery, maternal height, and paternal BMI.
Stratification of analyses according to socioeconomic status. There is some evidence that health benefits of green exposure depend on socioeconomic status, with people from lower socioeconomic groups benefiting more from green spaces, especially spaces near their place of residence (Dadvand et al. 2012a
; De Vries et al. 2003
; Lee and Maheswaran 2011
; Maas 2008
; Maas et al. 2009b
; Marmot 2010
). We therefore stratified analyses according to maternal education level [as an indicator of socioeconomic status (De Vries et al. 2003
; Maas et al. 2009b
)] to explore variation across socioeconomic strata. For these analyses, we removed the indicator of maternal socioeconomic status from the models.
Stratification of analyses according to the biogeographic region. We compared the associations between the two biogeographic regions (each encompassing two birth cohorts) by stratifying analyses (using NDVI average in 100-m buffer around maternal residential address) according to biogeographic region. Associations were expressed for a 1-IQR increase in surrounding greenness as defined for all cohorts combined (i.e., the same exposure contrasts used for the main analyses).
Evaluation of the interrelationship between air pollution, surrounding greenness, and pregnancy outcomes. Maternal exposure to nitrogen dioxide (NO2
) during the entire pregnancy was estimated using cohort-specific temporally adjusted land use regression (LUR) models that were previously shown to predict 51–75% of the variation in NO2
levels at different sampling points (Estarlich et al. 2011
). We repeated the main analyses by adding average maternal NO2
exposure levels during the entire pregnancy as a covariate to the models. This was done to explore the role of reduction in exposure to air pollution as an underlying mechanism for the association, if any, between surrounding greenness and pregnancy outcomes.
Season of data acquisition for surrounding greenness. For our analyses, we abstracted surrounding greenness using data from the greenest months for each biogeographic region. To investigate the robustness of our findings to this seasonal selection, we obtained the Landsat TM maps for all four birth cohorts during August 2003, one of the driest summers in Iberian Peninsula in recent years (U.S. Geological Survey 2011). Analyses were repeated using this alternative NDVI measure of surrounding greenness.
All births versus term births. We limited our analyses of birth weight and head circumference to those participants with term births (gestational age at delivery ≥ 37 weeks) to evaluate the robustness of our findings to the exclusion of preterm births.