Long-term exposure to air pollution is considered a major public health concern and has been related to overall mortality and various diseases such as respiratory and cardiovascular disease. Due to the spatial variability of air pollution concentrations, assessment of individual exposure to air pollution requires spatial datasets at high resolution. Combining detailed air pollution maps with personal mobility and activity patterns allows for an improved exposure assessment. We present high-resolution datasets for the Netherlands providing average ambient air pollution concentration values for the year 2009 for NO 2 , NO x , PM 2.5 , PM 2.5absorbance and PM 10. The raster datasets on 5×5 m grid cover the entire Netherlands and were calculated using the land use regression models originating from the European Study of Cohorts for Air Pollution Effects (ESCAPE) project. Additional datasets with nationwide and regional measurements were used to evaluate the generated concentration maps. The presented datasets allow for spatial aggregations on different scales, nationwide individual exposure assessment, and the integration of activity patterns in the exposure estimation of individuals.

Introduction: Previous analysis from the large European multicentre ESCAPE study showed an association of ambient particulate matter < 2.5 mu m (PM2.5) air pollution exposure at residence with the incidence of gastric cancer. It is unclear which components of PM are most relevant for gastric and also upper aerodigestive tract (UADT) cancer and some of them may not be strongly correlated with PM mass. We evaluated the association between long-term exposure to elemental components of PM2.5 and PM10 and gastric and UADT cancer incidence in European adults. Methods: Baseline addresses of individuals were geocoded and exposure was assessed by land-use regression models for copper (Cu), iron (Fe) and zinc (Zn) representing non-tailpipe traffic emissions; sulphur (S) indicating long-range transport; nickel (Ni) and vanadium (V) for mixed oil-burning and industry; silicon (Si) for crustal material and potassium (K) for biomass burning. Cox regression models with adjustment for potential confounders were used for cohort-specific analyses. Combined estimates were determined with random effects meta-analyses. Results: Ten cohorts in six countries contributed data on 227,044 individuals with an average follow-up of 14.9 years with 633 incident cases of gastric cancer and 763 of UADT cancer. The combined hazard ratio (HR) for an increase of 200 ng/m(3) of PM2.5_S was 1.92 (95%-confidence interval (95%-CI) 1.13; 3.27) for gastric cancer, with no indication of heterogeneity between cohorts (I-2= 0%), and 1.63 (95%-CI 0.88; 3.01) for PM2.5_Zn (I-2= 70%). For the other elements in PM2.5 and all elements in PM10 including PM10_S, non-significant HRs between 0.78 and 1.21 with mostly wide CIs were seen. No association was found between any of the elements and UADT cancer. The HR for PM2.5_S and gastric cancer was robust to adjustment for additional factors, including diet, and restriction to study participants with stable addresses over follow-up resulted in slightly higher effect estimates with a decrease in precision. In a two-pollutant model, the effect estimate for total PM2.5 decreased whereas that for PM2.5_S was robust. Conclusion: This large multicentre cohort study shows a robust association between gastric cancer and long-term exposure to PM2.5 S but not PM10 S, suggesting that S in PM2.5 or correlated air pollutants may contribute to the risk of gastric cancer.

Background: The South Durban (SD) area of Durban, South Africa, has a history of air pollution issues due to the juxtaposition of low- income communitieswith industrial areas. This study usedmeasurements of oxides of nitrogen (NOx) to develop a land use regression (LUR) model to explain the spatial variation of air pollution concentrations in this area. Methods: Ambient NOx was measured over two two-week sampling periods at 32 sites using Ogawa badges. Following the ESCAPE approach, an annual adjusted average was calculated for these results and regressed against pre-selected geographic predictor variables in a multivariate regression model. The LUR model was then applied to predict the NOx exposure of a sample of pregnant women living in South Durban. Results: Measured NOx levels ranged from 22.3-50.9 mu g/m(3) with a median of 36 mu g/m(3). The model developed accounts for 73% of the variance in ambient NOx measurements using three input variables ( length of minor roads within a 1000 m radius, length of major roads within a 300 m radius, and area of open space within a 1000 m radius). Model cross validation yielded a R-2 of 0.59. Subsequent participant exposure estimates indicated exposure to ambient NOx ranged from 19.9-53.2 mu g/m(3), with a mean of 39 mu g/m(3). Discussion and Conclusion: This is the first study to develop a land use regression model that predicts ambient concentrations of NOx in a South African context. The findings of this study indicate that the participants in the South Durban are exposed to high levels of NOx that can be attributed mainly to traffic. (C) 2017 Elsevier B.V. All rights reserved.

Background: Exposure to air pollution during pregnancy may increase attention-deficit/hyperactivity disorder (ADHD) symptoms in children, but findings have been inconsistent. We aimed to study this association in a collaborative study of eight European population-based birth/child cohorts, including 29,127 mother–child pairs. Methods: Air pollution concentrations (nitrogen dioxide [NO2] and particulate matter [PM]) were estimated at the birth address by land-use regression models based on monitoring campaigns performed between 2008 and 2011. We extrapolated concentrations back in time to exact pregnancy periods. Teachers or parents assessed ADHD symptoms at 3–10 years of age. We classified children as having ADHD symptoms within the borderline/clinical range and within the clinical range using validated cutoffs. We combined all adjusted area-specific effect estimates using random-effects meta-analysis and multiple imputations and applied inverse probability-weighting methods to correct for loss to follow-up. Results: We classified a total of 2,801 children as having ADHD symptoms within the borderline/clinical range, and 1,590 within the clinical range. Exposure to air pollution during pregnancy was not associated with a higher odds of ADHD symptoms within the borderline/clinical range (e.g., adjusted odds ratio [OR] for ADHD symptoms of 0.95, 95% confidence interval [CI] = 0.89, 1.01 per 10 µg/m3 increase in NO2 and 0.98, 95% CI = 0.80, 1.19 per 5 µg/m3 increase in PM2.5). We observed similar associations for ADHD within the clinical range. Conclusions: There was no evidence for an increase in risk of ADHD symptoms with increasing prenatal air pollution levels in children aged 3–10 years. See video abstract at, http://links.lww.com/EDE/B379.

Background: Ambient air pollution contains low concentrations of carcinogens implicated in the etiology of urinary bladder cancer (BC). Little is known about whether exposure to air pollution influences BC in the general population. Objective: To evaluate the association between long-term exposure to ambient air pollution and BC incidence. Design, setting and participants: We obtained data from 15 population-based cohorts enrolled between 1985 and 2005 in eight European countries (N = 303 431; mean follow-up 14.1 yr). We estimated exposure to nitrogen oxides (NO2 and NOx), particulate matter (PM) with diameter <10 mu m (PM10), <2.5 mu m (PM2.5). between 2.5 and 10 mu m (PM2.5-10). PM2.5 absorbance (soot), elemental constituents of PM, organic carbon, and traffic density at baseline home addresses using standardized land-use regression models from the European Study of Cohorts for Air Pollution Effects project. Outcome measurements and statistical analysis: We used Cox proportional-hazards models with adjustment for potential confounders for cohort-specific analyses and meta-analyses to estimate summary hazard ratios (HRS) for BC incidence. Results and limitations: During follow-up, 943 incident BC cases were diagnosed. In the meta-analysis, none of the exposures were associated with BC risk. The summary HRs associated with a 10-mu g/m(3) increase in NO2 and 51-mu g/m(3) increase in PM2.5 were 0.98 (95% confidence interval [CI] 0.89-1.08) and 0.86 (95% CI 0.63-1.18), respectively. Limitations include the lack of information about lifetime exposure. Conclusions: There was no evidence of an association between exposure to outdoor air pollution levels at place of residence and risk of BC. Patient summary: We assessed the link between outdoor air pollution at place of residence and bladder cancer using the largest study population to date and extensive assessment of exposure and comprehensive data on personal risk factors such as smoking. We found no association between the levels of outdoor air pollution at place of residence and bladder cancer risk. (C) 2016 European Association of Urology. Published by Elsevier B.V. All rights reserved.

Air pollution has been classified as carcinogenic to humans. However, to date little is known about the relevance for cancers of the stomach and upper aerodigestive tract (UADT). We investigated the association of long‐term exposure to ambient air pollution with incidence of gastric and UADT cancer in 11 European cohorts. Air pollution exposure was assigned by land‐use regression models for particulate matter (PM) below 10 µm (PM 10 ), below 2.5 µm (PM 2.5 ), between 2.5 and 10 µm (PM coarse ), PM 2.5 absorbance and nitrogen oxides (NO 2 and NO X ) as well as approximated by traffic indicators. Cox regression models with adjustment for potential confounders were used for cohort‐specific analyses. Combined estimates were determined with random effects meta‐analyses. During average follow‐up of 14.1 years of 305,551 individuals, 744 incident cases of gastric cancer and 933 of UADT cancer occurred. The hazard ratio for an increase of 5 µg/m 3 of PM 2.5 was 1.38 (95% CI 0.99; 1.92) for gastric and 1.05 (95% CI 0.62; 1.77) for UADT cancers. No associations were found for any of the other exposures considered. Adjustment for additional confounders and restriction to study participants with stable addresses did not influence markedly the effect estimate for PM 2.5 and gastric cancer. Higher estimated risks of gastric cancer associated with PM 2.5 was found in men (HR 1.98 [1.30; 3.01]) as compared to women (HR 0.85 [0.5; 1.45]). This large multicentre cohort study shows an association between long‐term exposure to PM 2.5 and gastric cancer, but not UADT cancers, suggesting that air pollution may contribute to gastric cancer risk.

BACKGROUND:Cohorts based on administrative data have size advantages over individual cohorts in investigating air pollution risks, but often lack in-depth information on individual risk factors related to lifestyle. If there is a correlation between lifestyle and air pollution, omitted lifestyle variables may result in biased air pollution risk estimates. Correlations between lifestyle and air pollution can be induced by socio-economic status affecting both lifestyle and air pollution exposure.OBJECTIVES:Our overall aim was to assess potential confounding by missing lifestyle factors on air pollution mortality risk estimates. The first aim was to assess associations between long-term exposure to several air pollutants and lifestyle factors. The second aim was to assess whether these associations were sensitive to adjustment for individual and area-level socioeconomic status (SES), and whether they differed between subgroups of the population. Using the obtained air pollution-lifestyle associations and indirect adjustment methods, our third aim was to investigate the potential bias due to missing lifestyle information on air pollution mortality risk estimates in administrative cohorts.METHODS:We used a recent Dutch national health survey of 387,195 adults to investigate the associations of PM10, PM2.5, PM2.5-10, PM2.5 absorbance, OPDTT, OPESR and NO2 annual average concentrations at the residential address from land use regression models with individual smoking habits, alcohol consumption, physical activity and body mass index. We assessed the associations with and without adjustment for neighborhood and individual SES characteristics typically available in administrative data cohorts. We illustrated the effect of including lifestyle information on the air pollution mortality risk estimates in administrative cohort studies using a published indirect adjustment method.RESULTS:Current smoking and alcohol consumption were generally positively associated with air pollution. Physical activity and overweight were negatively associated with air pollution. The effect estimates were small (mostly <5% of the air pollutant standard deviations). Direction and magnitude of the associations depended on the pollutant, use of continuous vs. categorical scale of the lifestyle variable, and level of adjustment for individual and area-level SES. Associations further differed between subgroups (age, sex) in the population. Despite the small associations between air pollution and smoking intensity, indirect adjustment resulted in considerable changes of air pollution risk estimates for cardiovascular and especially lung cancer mortality.CONCLUSIONS:Individual lifestyle-related risk factors were weakly associated with long-term exposure to air pollution in the Netherlands. Indirect adjustment for missing lifestyle factors in administrative data cohort studies may substantially affect air pollution mortality risk estimates.

Several studies have indicated weakly increased risk for kidney cancer among occupational groups exposed to gasoline vapors, engine exhaust, polycyclic aromatic hydrocarbons and other air pollutants, although not consistently. It was the aim to investigate possible associations between outdoor air pollution at the residence and the incidence of kidney parenchyma cancer in the general population. We used data from 14 European cohorts from the ESCAPE study. We geocoded and assessed air pollution concentrations at baseline addresses by land‐use regression models for particulate matter (PM10, PM2.5, PMcoarse, PM2.5 absorbance (soot)) and nitrogen oxides (NO2, NOx), and collected data on traffic. We used Cox regression models with adjustment for potential confounders for cohort‐specific analyses and random effects models for meta‐analyses to calculate summary hazard ratios (HRs). The 289,002 cohort members contributed 4,111,908 person‐years at risk. During follow‐up (mean 14.2 years) 697 incident cancers of the kidney parenchyma were diagnosed. The meta‐analyses showed higher HRs in association with higher PM concentration, e.g. HR = 1.57 (95%CI: 0.81–3.01) per 5 μg/m3 PM2.5 and HR = 1.36 (95%CI: 0.84–2.19) per 10−5m−1 PM2.5 absorbance, albeit never statistically significant. The HRs in association with nitrogen oxides and traffic density on the nearest street were slightly above one. Sensitivity analyses among participants who did not change residence during follow‐up showed stronger associations, but none were statistically significant. Our study provides suggestive evidence that exposure to outdoor PM at the residence may be associated with higher risk for kidney parenchyma cancer; the results should be interpreted cautiously as associations may be due to chance.

We thank Drs. Ha and Mendola for their thoughtful comments (1) on our study of associations of meteorological conditions and air pollution levels with preterm birth risk (2). Over the last decade, the spatial resolution of exposure models for atmospheric pollution has improved. Ha and Mendola mention, among other models, dispersion models as possibly more relevant alternatives to the land-use regression (LUR) approach used in the European Study of Cohorts for Air Pollution Effects (ESCAPE), in which our study was embedded. It should be noted that, although not explicitly considered by LUR models, meteorological conditions are still indirectly taken into account through their influence on local air pollution levels, which constitute an entry parameter of our seasonalized LUR model (2, 3). A detailed comparison of the yearly estimates of our LUR model and of dispersion models at home addresses, made in the European Study of Cohorts for Air Pollution Effects, showed a median (Pearson’s r) correlation of 0.75 for nitrogen dioxide and 0.29 for fine particulate matter, with stronger agreement in areas where dispersion models were more predictive of the local measurements done to define LUR models (4). These models should not be opposed, and estimates from dispersion models and other models can actually be fed into LUR models to increase their predictive ability (5). Further improving the spatial resolution of models below the 10-m to 100-m values of typical current models is unlikely to significantly improve accuracy in exposure estimates as long as the time-space activity of pregnant women is not considered (6). In a small-scale study of pregnant women carrying Global Positioning System devices in a mid-size French city, incorporating time-space activity in an exposure estimate based on an (outdoor) dispersion model produced little change in exposure estimates for fine particulate matter (a pollutant with limited spatial variability within urban areas) and somewhat larger but still limited changes for nitrogen dioxide exposure estimates (7). More generally, it has been shown that bias in dose-response functions due to ignoring time-space activity is likely to increase as spatial resolution of exposure models becomes finer (6). Taking indoor air pollution levels into account seemed to have a greater impact on exposure estimates (7), which is coherent with the limited correlation reported between personal and outdoor exposures outside the context of pregnancy (8). However, dosimeters cannot easily be carried for more than a few weeks during a pregnancy, thus offering a better consideration of indoor levels at the cost of a decreased ability to test numerous exposure windows during the pregnancy. Modeling indoor infiltration of outdoor pollutants could be a way to better take indoor levels into account without decreasing the temporal resolution of exposure estimates (9). Considering larger areas is an option for increasing the sample size. As Ha and Mendola rightly pointed out (1), there is no consensus as to the best way to correct for bias possibly resulting from the consideration of large study areas (10). The main concern here relates to confounding bias. In the context of a birth cohort, Pedersen et al. (10) showed that as the area considered around the city centers was extended, thus increasing sample size, the heterogeneity of the population increased in terms of disease risk factors, thus increasing the potential for bias in a situation where all confounders cannot be perfectly measured; this is but an illustration of the well-known bias-variance tradeoff. We chose to adjust for the study area using a randomeffect covariate (2). In the case of associations with firsttrimester atmospheric pressure, not adjusting for study center at all did not yield an increased point estimate (odds ratio per 5-mBar increase = 1.04, compared with 1.06 after controlling for study center with a random-effect variable) (2); taking the study center into account might indeed lead to overadjustment, but it can reduce confounding bias due to variations between locations in the risk factors for preterm

BACKGROUND:The evidence from observational epidemiological studies of a link between long-term air pollution exposure and diabetes prevalence and incidence is currently mixed. Some studies found the strongest associations of diabetes with fine particles, other studies with nitrogen dioxide and some studies found no associations.OBJECTIVES:Our aim was to investigate associations between long-term exposure to multiple air pollutants and diabetes prevalence in a large national survey in the Netherlands.METHODS:We performed a cross-sectional analysis using the 2012 Dutch national health survey to investigate the associations between the 2009 annual average concentrations of multiple air pollutants (PM10, PM2.5, PM10-2.5, PM2.5 absorbance, OPDTT, OPESR and NO2) and diabetes prevalence, among 289,703 adults. Air pollution exposure was assessed by land use regression models. Diabetes was defined based on a combined measure of self-reported physician diagnosis and medication prescription from an external database. Using logistic regression, we adjusted for potential confounders, including neighborhood- and individual socio-economic status and lifestyle-related risk factors such as smoking habits, alcohol consumption, physical activity and BMI.RESULTS:After adjustment for potential confounders, all pollutants (except PM2.5) were associated with diabetes prevalence. In two-pollutant models, NO2 and OPDTT remained associated with increased diabetes prevalence. For NO2 and OPDTT, single-pollutant ORs per interquartile range were 1.07 (95% CI: 1.05, 1.09) and 1.08 (95% CI: 1.05, 1.10), respectively. Stratified analysis showed no consistent effect modification by any of the included known diabetes risk factors.CONCLUSIONS:Long-term residential air pollution exposure was associated with diabetes prevalence in a large health survey in the Netherlands, strengthening the evidence of air pollution being an important diabetes risk factor. Most consistent associations were observed for NO2 and oxidative potential of PM2.5 measured by the DTT assay. The finding of an association with the oxidative potential of fine particles but not with PM2.5, suggests that particle composition may be important for a potential effect on diabetes.

Oxidative potential (OP) has been suggested as a health-relevant measure of air pollution. Little information is available about OP spatial variation and the possibility to model its spatial variability. Our aim was to measure the spatial variation of OP within and between 10 European study areas. The second aim was to develop land use regression (LUR) models to explain the measured spatial variation.OP was determined with the dithiothreitol (DTT) assay in ten European study areas. DTT of PM2.5 was measured at 16-40 sites per study area, divided over street, urban and regional background sites. Three two-week samples were taken per site in a one-year period in three different seasons. We developed study-area specific LUR models and a LUR model for all study areas combined to explain the spatial variation of OP.Significant contrasts between study areas in OP were found. OP DTT levels were highest in southern Europe. DTT levels at street sites were on average 1.10 times higher than at urban background locations.In 5 of the 10 study areas LUR models could be developed with a median R-2 of 33%. A combined study area model explained 30% of the measured spatial variability. Overall, LUR models did not explain spatial variation well, possibly due to low levels of OP DTT and a lack of specific predictor variables. (C) 2016 Elsevier Ltd. All rights reserved.

Background: Tobacco smoking increases the risk of liver cancer, but little is known about possible risks associated with ambient air pollution. Methods: We evaluated the association between exposur...

Atmospheric pollutants and meteorological conditions are suspected to be causes of preterm birth. We aimed to characterize their possible association with the risk of preterm birth (defined as birth occurring before 37 completed gestational weeks). We pooled individual data from 13 birth cohorts in 11 European countries (71,493 births from the period 1994-2011, European Study of Cohorts for Air Pollution Effects (ESCAPE)). City-specific meteorological data from routine monitors were averaged over time windows spanning from 1 week to the whole pregnancy. Atmospheric pollution measurements (nitrogen oxides and particulate matter) were combined with data from permanent monitors and land-use data into seasonally adjusted land-use regression models. Preterm birth risks associated with air pollution and meteorological factors were estimated using adjusted discrete-time Cox models. The frequency of preterm birth was 5.0%. Preterm birth risk tended to increase with first-trimester average atmospheric pressure (odds ratio per 5-mbar increase = 1.06, 95% confidence interval: 1.01, 1.11), which could not be distinguished from altitude. There was also some evidence of an increase in preterm birth risk with first-trimester average temperature in the -5°C to 15°C range, with a plateau afterwards (spline coding, P = 0.08). No evidence of adverse association with atmospheric pollutants was observed. Our study lends support for an increase in preterm birth risk with atmospheric pressure.

Introduction: An association between air pollution and breast cancer risk has been suggested but epidemiological evidence is sparse and inconclusive. Aim: We examined association between long-term ...

Background: The available evidence of air pollution exposure on fertility is scarce. In a previous analysis, we have shown that PM2.5 was associated with a higher risk of having use Assisted Reproductive Technology (ART), now we wanted to identify if any PM element was associated with ART. Objectives: To assess the association between PM2.5 elements and the use of Assisted ART. Methods: This study was based on a cohort of pregnant women (N = 8391) that were followed in a major university hospital in Barcelona during 2001-2005. Socio demographic variables and home addresses of the women were recorded. The medical record also included the question about the use of ART to achieve the pregnancy. Annual average exposure to copper, iron, potassium, nickel, sulfur, silicon, vanadium and zinc, each respectively derived from particles with aerodynamic diameters ≤ 10 μm (PM10) and 2.5 μm (PM2.5) were estimated using standardized land use regression models developed within within the framework of the TRANSPHORM European project and assigned to enrolment address. Associations between air pollution and ART use was assessed with logistic regression and were adjusted by age, smoking and socio-economical status. Results: The mean age of the women was 30 yeas (SD 6). 18% smoked during pregnancy. 110 women (mean age 34 (5)) reported the use of ART, 55% used IVF and 30% artificial insemination. We found a negative association in the crude analyses between PM2.5K and Ni with ART (OR=0.64 95%CI(0.43-0.96) for K and 0.78 (0.62-0.99) for Ni) but the association disappeared after adjustment (0.75 (0.48-1.16) for K and 0.87 (0.67-1.13) for Ni). No associations were observed for the other PM elements. Conclusions: None of the elements included in this analysis were clearly associated with ART use. The suggested negative associations found for the PM2.5 Ni and K needs further investigation and may be due to high negative correlations with other pollutants not measured in this study.

[背景]已有研究表明,在死亡率和颗粒物空气污染长期暴露之间存在关联,而估计颗粒物元素成分对死亡率影响的队列研究很少. [目的]研究自然原因死亡率和长期暴露于颗粒物元素成分之间的关联. [方法]遵循严格的标准化方案,并使用19个欧洲队列研究中死亡率和混杂因素的数据.描述住宅暴露于优先选择的8种颗粒物(PM)成分的特征.使用土地利用回归模型估计粒级≤2.5 μm(PM2.5)和≤10 μm(PM10)的颗粒物中铜、铁、钾、镍、硫、硅、钒和锌的年平均浓度,并在meta分析之后,按照统一的标准,使用Cox比例风险模型对队列特异的死亡率和空气污染之间的关联进行统计分析. [结果]研究包括291 816名参与者,其中25466在随访期间(平均随访时间14.3年)死于自然原因.几乎所有元素的风险比都呈阳性,并且PM2.5硫的风险比具有统计学显著性(每200 ng/m3为1.14,95％CI:1.06～1.23).在一个双污染物模型中,校正PM2.5总质量之后,死亡率与PM2.5硫之间的关联保持稳定,而与PM2.5总质量的关联性降低. [结论]长期暴露于PM2.5硫与自然原因引起的死亡率有关联.这个关联在校正了其他污染物和PM2.5之后依然稳定.

Industrialization has been linked to the etiology of inflammatory bowel disease (IBD).

AimsWe investigated whether traffic-related air pollution and noise are associated with incident hypertension in European cohorts.Methods and resultsWe included seven cohorts of the European study of cohorts for air pollution effects (ESCAPE). We modelled concentrations of particulate matter with aerodynamic diameter ≤2.5 µm (PM2.5), ≤10 µm (PM10), >2.5, and ≤10 µm (PMcoarse), soot (PM2.5 absorbance), and nitrogen oxides at the addresses of participants with land use regression. Residential exposure to traffic noise was modelled at the facade according to the EU Directive 2002/49/EC. We assessed hypertension as (i) self-reported and (ii) measured (systolic BP ≥ 140 mmHg or diastolic BP ≥ 90 mmHg or intake of BP lowering medication (BPLM). We used Poisson regression with robust variance estimation to analyse associations of traffic-related exposures with incidence of hypertension, controlling for relevant confounders, and combined the results from individual studies with random-effects meta-analysis. Among 41 072 participants free of self-reported hypertension at baseline, 6207 (15.1%) incident cases occurred within 5-9 years of follow-up. Incidence of self-reported hypertension was positively associated with PM2.5 (relative risk (RR) 1.22 [95%-confidence interval (CI):1.08; 1.37] per 5 µg/m³) and PM2.5 absorbance (RR 1.13 [95% CI:1.02; 1.24] per 10 - 5m - 1). These estimates decreased slightly upon adjustment for road traffic noise. Road traffic noise was weakly positively associated with the incidence of self-reported hypertension. Among 10 896 participants at risk, 3549 new cases of measured hypertension occurred. We found no clear associations with measured hypertension.ConclusionLong-term residential exposures to air pollution and noise are associated with increased incidence of self-reported hypertension.

BACKGROUND:Leave-one-out cross-validation that fails to account for variable selection does not properly reflect prediction accuracy when the number of training sites is small. The impact on health effect estimates has rarely been studied. The objective of this study was to develop an improved validation procedure for land-use regression models with variable selection and investigate health effect estimates in relation to land-use regression model performance.METHODS:We randomly generated 10 training and test sets for nitrogen dioxide and particulate matter. For each training set, we developed models and evaluated them using a cross-holdout validation approach. Cross-holdout validation develops new models for each evaluation compared with refitting the model without variable selection, as in standard leave-one-out cross-validation. We also implemented holdout validation, which evaluates model predictions using independent test sets. We evaluated the relationship between cross-holdout validation and holdout validation R and estimates of the association between air pollution and forced vital capacity in the Dutch birth cohort.RESULTS:Cross-holdout validation Rs were generally identical to holdout validation Rs, but were notably smaller than leave-one-out cross-validation Rs. Decreases in forced vital capacity in relation to air pollution exposure were larger for land-use regression models that had larger holdout validation and cross-holdout validation Rs rather than leave-one-out cross-validation R.CONCLUSION:Cross-holdout validation accurately reflects predictive ability of land-use regression models and is a useful validation approach for small datasets. Land-use regression predictive ability in terms of holdout validation and cross-holdout validation rather than leave-one-out cross-validation was associated with the magnitude of health effect estimates in a case study.

Introduction. Land use regression modelling has emerged as a promising technique in exposure assessment. Its use in developing countries is limited. This study reports the development of a model in a city in a developing country, with regions of high industrialisation (SD), compared to lower levels of industrial development (ND). In addition the SD region has a history of air pollution concerns, coastal proximity & basin like topography. This study aimed to characterise exposure in these two regions for oxides of nitrogen (NOx) as part of a birth cohort, the Mother and Child in the Environment (MACE). Methods. NOx were measured over two two-week sampling periods at 40 sites (SD=32; ND=8), with Ogawa badges. Using the ESCAPE approach, two models were developed, a combined version (SD+ND) and a SD model only to test the variation across the two regions. The model was applied to the addresses of pregnant women to determine their individual exposure. Results. The significant predictor variables in the combined model were length of major road within 300m, area of open space within 100m & 1000-100m with an R2 of 0.71. The significant variables in the reduced model included length of major road within 300m, minor road within 1000m, and area of urban formal land use within 200m & open space within 1000m, with an R2 of 0.69. Sensitivity analyses for the SD model showed an increase in model predictability (R2=0.73) by omitting the urban formal land use variable. Conclusions. The findings further suggest that the observed NOx levels are mainly attributed to road traffic, due the significant associations observed between the road length variables and measured NOx levels. The shortcomings identified in model development for the SD, include the lack of industrial variables & classification as minor / major as well as emissions inventory, meteorological variables such as wind frequency/direction & coastal impacts, which are a premise for future research in this area.