Long-Term Exposure to Traffic-Related Air Pollution Associated with Blood Pressure and Self-Reported Hypertension in a Danish Cohort

Background: Short-term exposure to air pollution has been associated with changes in blood pressure (BP) and emergency department visits for hypertension, but little is known about the effects of long-term exposure to traffic-related air pollution on BP and hypertension. Objectives: We studied whether long-term exposure to air pollution is associated with BP and hypertension. Methods: In 1993–1997, 57,053 participants 50–64 years of age were enrolled in a population-based cohort study. Systolic and diastolic BP (SBP and DBP, respectively) were measured at enrollment. Self-reported incident hypertension during a mean follow-up of 5.3 years was assessed by questionnaire. We used a validated dispersion model to estimate residential long-term nitrogen oxides (NOx), a marker of traffic-related air pollution, for the 1- and 5-year periods prior to enrollment and before a diagnosis of hypertension. We conducted a cross-sectional analysis of associations between air pollution and BP at enrollment with linear regression, adjusting for traffic noise, measured short-term NOx, temperature, relative humidity, and potential lifestyle confounders (n = 44,436). We analyzed incident hypertension with Cox regression, adjusting for traffic noise and potential confounders. Results: A doubling of NOx exposure during 1- and 5-year periods preceding enrollment was associated with 0.53-mmHg decreases [95% confidence interval (CI): –0.88, –0.19 mmHg] and 0.50-mmHg decreases (95% CI: –0.84, –0.16 mmHg) in SBP, respectively. Long-term exposure also was associated with a lower prevalence of baseline self-reported hypertension (per doubling of 5-year mean NOx: odds ratio = 0.96; 95% CI: 0.91, 1.00), whereas long-term NOx exposure was not associated with incident self-reported hypertension during follow-up. Conclusions: Long-term exposure to traffic-related air pollution was associated with a slightly lower prevalence of BP at baseline, but was not associated with incident hypertension.

volume 120 | number 3 | March 2012 • Environmental Health Perspectives Research Exposure to particulate matter (PM) air pollution has been associated with myocardial infarction and stroke (Brook et al. 2010). The mechanisms believed to be involved include alteration of the autonomic function of the heart, vascular reactivity, induction of sys temic inflammation, and endothelial dys function (Brook et al. 2010), which in turn may affect blood pressure (BP) and the risk of hypertension. It has therefore been hypothe sized that high levels of air pollution may increase BP and the risk of hypertension.
Studies investigating associations between air pollution and BP have focused mainly on shortterm effects, with some studies report ing small increases in systolic and diastolic BP (SBP and DBP, respectively) (Brook et al. 2009;Chuang et al. 2010;de Paula et al. 2005;Dvonch et al. 2009;Urch et al. 2005;Zanobetti et al. 2004) and others reporting no association (Madsen and Nafstad 2006) or even inverse associations (Brauer et al. 2001;Harrabi et al. 2006;IbaldMulli et al. 2004). Two studies have investigated associations between longerterm air pollution exposure and BP (Auchincloss et al. 2008;Chuang et al. 2011). In a crosssectional study of approxi mately 5,000 persons 45-84 years of age, Auchincloss et al. (2008) reported that the 30day mean of PM 2.5 (PM ≤ 2.5 μm in aero dynamic diameter) was positively associated with SBP, whereas the association with DBP was weaker and statistically insignificant. In a crosssectional study of 1,023 elderly persons, Chuang et al. (2011) reported that systolic and DBP both were highly correlated with yearly mean levels of several pollutants.
Little is known about the effects of air pollution on hypertension. Shortterm expo sure to air pollution has been positively asso ciated with emergency department visits for hypertension (Guo et al. 2010a(Guo et al. , 2010b, and a recent study reported a significant positive association between estimated annual expo sure to residential PM 2.5 and the prevalence of selfreported hypertension (Johnson and Parker 2009). No studies have investigated the effects of longterm air pollution on the incidence of hypertension.
Residential exposure to air pollution and road traffic noise are positively correlated (de Kluizenaar et al. 2007; Sørensen et al. 2011). Because exposure to traffic noise has been associated with changes in BP and hypertension (Babisch 2006;de Kluizenaar et al. 2007), road traffic noise is a potentially important confounder in air pollution studies.
In this study we tested the hypothesis that longterm exposure to trafficrelated air pollu tion increases systolic and DBP and the preva lence and risk of hypertension, independent of shortterm exposure to air pollution and exposure to road traffic noise.

Study population.
The study was based on the Diet, Cancer and Health cohort (Tjønneland et al. 2007). In total, 57,053 of 160,725 resi dents of Copenhagen or Aarhus 50-64 years of age without a history of cancer (excluding non melanoma skin cancer) were enrolled into the original cohort between 1993 and 1997. An invitation to participate in a followup survey was mailed with a followup question naire to the 54,379 living cohort members still residing in Denmark in 2000-2002. The response rate was 83.3%, corresponding to 45,271 participants. All study participants pro vided written informed consent. The study was conducted in accordance with the Declaration of Helsinki (World Medical Association 2008) and approved by local ethical committees.
At baseline enrollment into the original cohort study, each participant completed a self administered questionnaire that included ques tions on lifestyle habits, health status, whether they suffered or had ever suffered from hyper tension, and whether they received or had ever received medication for hypertension.
Exposure assessment. Using the Danish AirGIS modeling system, we modeled nitrogen oxides (NO x ), nitrogen dioxide (NO 2 ), and nitrogen oxide (NO) concentrations in the air at each address at which the cohort members lived from 5 years prior to baseline until fol lowup was completed in 2000-2002. AirGIS calculates air pollution at a location as the sum of local air pollution from street traffic [cal culated with the Operational Street Pollution Model from input data on traffic (intensity and type), emission factors, street and building geometry, and meteorology (Berkowicz 2000;Kakosimos et al. 2011)]; urban background [from a simplified area source dispersion model that takes into account urban vehicle emis sion density, city dimensions (transport dis tance), and building height (Berkowicz et al. 2008)]; and regional background estimated from trends at rural monitoring stations and national vehicle emissions. Input data have been described elsewhere (RaaschouNielsen et al. 2010). The AirGIS system has been vali dated in several studies, and the correlation (r) between modeled and measured halfyear mean NO 2 concentrations at 204 positions in the greater Copenhagen area was 0.90 (Ketzel et al. in press;RaaschouNielsen et al. 2000). The AirGIS system calculates air pollution hour by hour, which was summarized as the yearly aver age concentration at each residential address.
We used NO x as a measure of exposure to air pollution from traffic because measured NO x correlates strongly with other traffic related pollutants in Danish streets: r = 0.93 for total particle number concentration (10-700 nm) and r = 0.70 for PM 10 (PM ≤ 10 μm in aerodynamic diameter) (Hertel et al. 2001;Ketzel et al. 2003). If NO x , NO 2 , and NO could not be calculated because of failed geocoding, we imputed the concentra tion calculated at the preceding or subsequent residential address of the cohort member as previously described (RaaschouNielsen et al. 2011). We then calculated 1year and 5year timeweighted average NO x , NO 2 , and NO concentrations before baseline enrollment (crosssectional study), and 1year and 5year timeweighted averages before a new diagno sis of hypertension or the end of followup (followup study).
Based on the enrollment address and the geographical information system (GIS) road network, we generated two additional traffic variables: a dichotomous indicator for the presence or absence of a street with a traffic density > 10,000 vehicles per day within 50 or 100 m of the residence, and the total number of kilometers driven by vehicles within 200 m of the residence each day (the product of street length and traffic density for all streets within a 200m radius).
We used hourly measurements at a urban background monitoring station (20 m above ground; chemiluminiscence NO/NO x model 200A; Teledyne Advanced Pollution Instrumentation, San Diego, CA, USA) to esti mate 3day average exposures to NO x , NO 2 , and NO (on the day of the BP measurement and the 2 preceding days) among partici pants enrolled by the Copenhagen center. The monitoring station was located in the center of Copenhagen, with a median residential dis tance from the monitoring station of 5.5 km (5th-95th percentile, 1.5-14.2 km). We used hourly measures of temperature and relative humidity from three locations (Copenhagen, Aalborg, and Odense) to estimate 3day aver ages for all participants. Previous studies on air pollution and BP have found different lags and cumulative exposures to be important (Auchincloss et al. 2008;Dvonch et al. 2009;Harrabi et al. 2006;Zanobetti et al. 2004). We calculated a 3day mean because this has been suggested to be related to BP Zanobetti et al. 2004).
BP measurement. At baseline enrollment, trained staff members measured brachial artery BP tomated TAKEDA UA 751 or UA743 using automated oscillometric sphygmoma nometers (model UA 751 or UA743; Takeda Pharmaceutical Co. Ltd., Osaka, Japan). The measurement was conducted with the sub ject in the supine position after a minimum of 5 min rest and at least 30 min after tobacco smoking and intake of food, tea, or coffee. If SBP was ≥ 160 mmHg more, or if DBP was ≥ 95 mmHg, the measurement was repeated after an interval of at least 3 min, and the lower of the two measurements was used. We excluded from the present analysis all partici pants who indicated on the enrollment ques tionnaire that they were taking or had ever taken medication for hypertension. Height and weight were measured at baseline according to standardized protocols.
Incidence of hypertension. Information on hypertension was assessed by questionnaire at enrollment and in the followup survey. Specifically, at enrollment participants were asked whether they had ever been hyperten sive or were taking or had ever taken hyper tension medication, and in the followup survey they were asked whether they had ever been diagnosed with hypertension by a medical doctor or were taking or had ever taken hypertension medication. In both the crosssectional study on hypertension and the followup study, we excluded all participants with hypertension at or prior to enrollment and participants with missing or contradictory answers to the hypertension questions.
Statistical methods. Cross-sectional analysis of BP and hypertension. We used general linear models to estimate associations between residential exposure to longterm NO x , NO 2 , and NO (1 and 5year aver ages prior to baseline) and systolic and DBP measured at baseline (among participants who did not report use of medications to treat hypertension), and logistic regression mod els to estimate associations between 1 and 5year average NO x , NO 2 , and NO concen trations and the prevalence of selfreported hypertension at baseline (PROC GLM and PROC GENMOD in SAS, version 9.1; SAS Institute Inc., Cary, NC, USA). Exposures were modeled as categorical variables (with cut points based on quartiles) and as continu ous variables after logarithmic transformation (log 2 ) to satisfy the assumption of linearity, which we evaluated using linear spline models with boundaries at deciles of exposure for the analytic cohort (BP) or cases (hypertension) (Greenland 1995). In addition, we estimated associations of BP and prevalent hyperten sion with shortterm NO x , NO 2 , and NO exposures averaged over 3 days (the day of BP measurement and the previous 2 days, log 2 transformed and categorical) among Copenhagen residents, and associations with the presence or absence of a major road within 50 m of the baseline residence and traffic den sity within 200 m of the baseline residence (log 2 transformed or categorical) among all participants.
We adjusted analyses for potential con founders: age (continuous), sex, calendar year, center of enrollment (Copenhagen or Aarhus), area [Copenhagen city, Aarhus city, or Copenhagen or Aarhus surroundings (defined as residence within 7-25 km of either city center)], length of school attendance (< 8, 8-10, > 10 years), body mass index (BMI; kilograms per meter squared, linear), smok ing status (never, former, current), alcohol intake (yes/no; grams per day among drink ers, linear), intake of fruit and vegetables (lin ear splines with a knot at 350 g/day), sport during leisure time (yes/no; hours per week among active, continuous), road traffic noise (L den ; decibels; residential exposure at enroll ment), season (winter, spring, summer, and autumn), mean relative humidity (continu ous), and ambient temperature during 3 days (the day of BP measurement and the 2 preced ing days). Temperature showed a weak inverse association with BP ≤ 11.5°C and a steep inverse association at temperatures > 11.5°C. Therefore, temperature was modeled using linear splines with a knot at 11.5°C. In addi tion we adjusted for the socioeconomic status (SES) of the participants' municipality (or dis trict for Copenhagen residents) classified as low, medium, or high based on information on average education, work market affiliation, and income at the time of enrollment. Analysis of associations with shortterm NO x , NO 2 , and NO concentrations were also adjusted by volume 120 | number 3 | March 2012 • Environmental Health Perspectives the 1year mean concentration of NO x , NO 2 , or NO, respectively, in the previous year.
In a secondary analysis restricted to Copenhagen residents (n = 21,507), we adjusted associations between longterm NO x by measured ambient NO x concentrations averaged over the day of the BP measurement and the previous 3 days. In addition, we con ducted sensitivity analyses restricted to par ticipants with normal BP (SBP ≤ 140 and/or DBP ≤ 90) or participants with SBP < 160 and/or DBP < 100.
In exploratory analyses, we tested for interactions between modeled longterm exposure to NO x and sex, education, smok ing, temperature, area, SES, and history of cardiovascular disease by introducing inter action terms into the model.
Graphical presentation of the functional form of association between NO x and SBP adjusted for the potential confounders was estimated with the OLS function in Design Library[R statistical software, version 2.9.0 (http://www.rproject.org/). Follow-up for hypertension. We ana lyzed data based on Cox proportional hazards model with age as the underlying time metric (Thiebaut and Benichou 2004). We used left truncation at age of enrollment, so that subjects were considered at risk from enrollment into the cohort, and right censoring at age of event (selfreported hypertension) or age at followup survey, whichever came first. We stratified all analyses by sex and calendar year. Exposure to longterm air pollution was modeled using timedependent variables of timeweighted average NO x , NO 2 , and NO concentrations at each year of age during followup (one row of data for each year of age that a participant contributed to followup).
We calculated incidence rate ratios (IRRs) for hypertension in association with 1 and 5year mean NO x , NO 2 , and NO concen trations at the time of diagnosis compared with 1 and 5year mean NO x , NO 2 , and NO concentrations for all cohort members at risk at that point in time. IRRs for the two traffic proxies (major road and traffic load) were cal culated using enrollment addresses. Analyses were adjusted for baseline information on smoking status, length of school attendance, alcohol intake, intake of fruit and vegetables, BMI, sport during leisure time, SES, area, and traffic noise. We interpreted a pvalue < 0.05 as statistically significant.

BP and baseline hypertension.
Of 57,053 par ticipants, we excluded 571 who had been diag nosed with cancer before baseline, but because of delay in the Danish Cancer Registry, were erroneously included; 2,737 with incomplete residential address information; 63 without BP measurement; and 2,961 with missing information on covariates leaving 50,721 par ticipants for the baseline hypertension anal yses. Of these, 6,285 received hypertension medicine at and/or before enrollment, leaving 44,436 participants for the BP analyses. Table 1 shows the distribution of base line characteristics in the study population. Longterm exposure to NO x and traffic load at the address at enrollment was corre lated, with a Spearman rank coefficient (r S ) of 0.95 between the 1 and 5year mean NO x (p < 0.0001) and 0.51 between traf fic load and 1year NO x mean (p < 0.0001). Modeled exposure to NO x , NO 2 , and NO at the enrollment address was highly correlated: 0.98 between NO x and NO, 0.97 between NO x and NO 2 , and 0.92 between NO 2 and NO (1year data; p < 0.0001). Short and longterm exposure to NO x were not cor related. There was a significant correlation between longterm exposure to NO x and L den at enrollment (0.69 and 0.67 for the 1 and 5year period preceding enrollment; p < 0.0001). The distributions of systolic and DBP were slightly rightskewed. However, similar results were observed for untransformed and log transformed values, and regression estimates for the untransformed data are presented.
Longterm exposure to NO x was inversely associated with BP ( Figure 1, Table 2). Although significant, the estimated changes were rather small. Categorical analyses showed a monotonically inverse dose-response rela tionship between the 1 and 5year NO x means and SBP, whereas this was not appar ent for the DBP. Corresponding estimates for longterm exposures to NO 2 and NO were generally consistent with those shown for NO x [see Supplemental Material, Adjustment for road traffic noise changed the estimates slightly (data not shown); for example, a doubling in 1year NO x was associ ated with a -0.39 mmHg change in SBP before adjustment for traffic noise and a -0.53 mmHg change in SBP after adjustment. Further adjustment by shortterm NO x concentra tions (among Copenhagen participants only) had little effect on estimates (data not shown). For example, the estimated changes in SBP per doubling of 1 and 5year NO x exposures were -0.50 mmHg [95% confidence interval (CI): -0.93, -0.07 mmHg] and -0.51 mmHg (95% CI: -0.94, -0.08 mmHg), respectively, after adjusting for shortterm NO x . When restricted to the 23,982 participants who had normal BP at baseline, the estimated changes in SBP per doubling of 1 and 5year NO x exposures were -0.24 mmHg (95% CI: -0.50, 0.01 mmHg) and -0.20 mmHg (95% CI: -0.45, 0.05 mmHg), respectively. When restricted to the 38,565 participants who had SBP < 160 and/or DBP < 100, the correspond ing estimates were -0.27 mmHg (95% CI: -0.55, 0.01 mmHg) and -0.22 mmHg (95% CI: -0.50, 0.05 mmHg), respectively.
Inverse associations between exposure and BP were also estimated for shortterm NO x and for the two traffic proxies ( Table 2). The categorical analyses for the 3day NO x mean and traffic load showed no clear dose-response relationship in relation to BP.
Sex, temperature, and a diagnosis of car diovascular disease appeared to modify the association between NO x and SBP (Table 3). The inverse relationship between exposure  . The presence of a major road within 50 or 100 m of the residence seemed to be associated with a lower prevalence of hyper tension, whereas there was no evidence of an association with traffic load.
Follow-up for hypertension. Of the 45,271 persons that filled out the followup question naire, we excluded 7,110 with hypertension at or prior to enrollment, 1,841 participants with missing or contradictory answers to the hypertension questions, 2,897 with incom plete residential address information, and 148 with missing information on covariates, leaving a study base of 33,275 participants with an average followup period of 5.3 years. Among these, 3,195 participants reported that they had been diagnosed with hypertension within the followup period. Table 1 shows the distribution of base line characteristics in the study population. The distribution of baseline characteristics among the 33,275 participants followed up for hypertension was very similar to the distri butions in the baseline study cohort.
We found no clear associations between exposure to trafficrelated air pollution and risk for selfreported hypertension in the sub set of participants who responded at the fol lowup survey (evaluated as IRRs; Table 4). In analyses of NO x , point estimates were slightly elevated, but CIs included the null. Estimates did not demonstrate monotonic doseresponse relations with increasing quartiles of exposure. Participants who lived within 50 m of a major road had a 13% higher risk for hypertension (95% CI: 0.97%, 1.32%). Exclusion of participants with a history of myocardial infarction (n = 335) or stroke (n = 206) at baseline resulted in only minor changes in estimated associations (data not shown). Also with regard to exposure to long term NO 2 and NO, no clear associations were found between exposure and risk for hyper tension [see Supplemental Material,

Discussion
Longterm exposure to trafficrelated air pol lution was inversely associated with systolic and DBP and the prevalence of selfreported hypertension in a crosssectional design, whereas longterm exposure to trafficrelated air pollution was not associated with the risk of selfreported hypertension during approxi mately 5 years of followup.
Strengths and limitations. Strengths included the large study population, with detailed information on potential confound ers. Furthermore, access to residential address histories improved estimation of longterm air pollution. In addition, we adjusted for exposure to road traffic noise, which poten tially is associated with trafficrelated air pol lution (de Kluizenaar et al. 2007;Sørensen et al. 2011) and has been associated with BP and hypertension (Babisch 2006). However, we cannot rule out residual confounding, for example, by individual SES or intake of sodium or potassium.
Although the dispersion models used to estimate longterm exposures to air pollution in the present study have been successfully validated and applied Berkowicz et al. 2008;Ketzel et al. in press;RaaschouNielsen et al. 2000), such estimates are inevitably associated with some degree of uncertainty, which would result in exposure misclassification. However, such misclassifica tion should be non differential with respect to BP and hypertension.
A limitation of the study of measured BP at baseline part of this study is the crosssectional design. Although we have adjusted for many possible confounders, associations should be confirmed using a longitudinal design with repeated measures. Results of previous studies of air pollution and BP measured at different points in time have been inconsistent, with some studies reporting positive associations (de Paula et al. 2005;Dvonch et al. 2009;Zanobetti et al. 2004), and others reporting inverse associations (Brauer et al. 2001;Ebelt et al. 2005;IbaldMulli et al. 2004) without clear relations between the results observed and the design of the study.
The measurement of systolic and DBP in our study was standardized but did not fol low standard clinical protocols for diagnos ing hypertension, which require several measurements of BP. We repeated measure ments only if SBP was ≥ 160 mmHg or DBP was ≥ 95 mmHg, and used the lower of the two measurements, which may have resulted in a systematic bias toward lower values in participants with higher BPand could have biased the BP estimate toward an inverse asso ciation. When we restricted the sample to par ticipants with normal BP values, who were less likely to have had repeated BP measurements, inverse associations were less pronounced but still evident between longterm NO x and BP. However, it is not possible to determine whether or how much differences observed after restriction reflect a reduction in misclas sification of the BP measurements versus selec tion bias caused by limiting the analysis to potentially less susceptible participants. Our prospective study of hypertension also has some limitations. First, information on hypertension was selfreported, and the actual number of hypertensive participants is probably underestimated. Therefore, a number of participants who were actually hyperten sive at baseline were falsely included as non hypertensive. Such misclassification may have led to a systematic bias; for example, if under reporting was most prominent in lowSES groups, who are often exposed to the highest levels of air pollution, risk estimates may have been biased downward. Also, all participants had their BP taken at baseline, and it is very likely that those with a high BP measurement would subsequently have been examined fur ther by their physician. Therefore, many with undiagnosed hypertension at enrollment will potentially be diagnosed immediately after enrollment. However, exclusion of the cases diagnosed within the first year (27%) from the analyses did not change the estimates markedly, indicating that this did not result in a systematic bias.
Systolic and DBP. Our finding of a weak inverse association between air pollution and BP was robust across different model speci fications. Furthermore, both longterm and shortterm air pollution exposures, as well as two proxy measures of traffic exposure, were inversely associated with BP.
Two studies of longerterm exposure to air pollution and BP have reported positive asso ciations (Auchincloss et al. 2008;Chuang et al. 2011). Results of previous studies of short term exposures and BP have been inconsistent (Brauer et al. 2001;Brook et al. 2009;Chuang et al. 2010;de Paula et al. 2005;Dvonch et al. 2009;Harrabi et al. 2006;IbaldMulli et al. 2004;Urch et al. 2005;Zanobetti et al. 2004). IbaldMulli et al. (2004) suggested that a pos sible mechanism for a decrease in BP caused by exposure to air pollution could be a shift in sympatho vagal balance due to an increase in vagal tone. Another explanation could relate to the effect of NO as a potent vasodilator that diffuses freely across membranes. NO is pres ent in exhaust from vehicles and is converted to NO 2 through reaction with ozone. Because ozone is generated from oxygen reacting with sunlight, NO is usually present in lowest con centrations during summer. NO, NO 2 , and their sum, NO x , are highly correlated (Hertel et al. 2001;Ketzel et al. 2003), and it is there fore extremely difficult to disentangle effects of the three exposures. A closer look at the results reported by Auchincloss et al. (2008) indicates that the positive association between longterm air pollution exposure and BP was evident only in the warmer season (> 10°C); whereas at tem peratures < 10°C, the association tended to be negative although not statistically significantly. Similarly, we found that the inverse association between air pollution and BP was present only at temperatures < 15°C. The concentration of NO x is rather constant during a year, but dur ing summer the contribution of NO is reduced because of higher ozone concentrations. This and other factors that could influence NO concentrations, such as geography and season, might also explain differences in results among studies of different populations.
Our results suggest that among patients with a previous diagnosis of cardiovascular disease, longterm exposure to NO x might be positively associated with BP, indicating that these patients might be a susceptible group. This analysis was, however, based on relatively few patients with cardiovascular disease.
Hypertension. We found inconsistent associations of longterm air pollution with hypertension. In the crosssectional analysis of selfreported hypertension at baseline, we saw a small inverse association, consistent with the results of the BP analysis. However, expo sure was not inversely associated with incident selfreported hypertension, and results could indicate a slight positive association, although estimates did not indicate a monotonic dose dependent relationship. Studies using vali dated hypertension as outcome are necessary to disentangle possible sources of bias.
To our knowledge, this is the first study to estimate effects of longterm air pollution on the incidence of hypertension. A few pre vious studies have reported that shortterm air pollution was associated with emergency department visits for hypertension (Guo et al. 2010a(Guo et al. , 2010b and that longterm air pollu tion was positively associated with prevalence of selfreported hypertension (Johnson and Parker 2009). Exposure to air pollution has been associated with increased inflammation and oxidative stress, as well as endothelial dys function (Brook et al. 2010;Hoffmann et al. 2009), which may contribute to the develop ment and progression of atherosclerosis and risk of hypertension. Because most studies have focused on PM 2.5 and not NO and NO 2 , direct comparisons are difficult. In contrast to wellknown vasodilatory effects of NO, PM mixtures are extremely variable and may have very different physiological effects depending on the predominant constituents.

Conclusions
Longterm exposure to trafficrelated air pol lution was associated with a slightly lower BP but was not consistently associated with self reported hypertension.