This study used data from the first Portuguese National Health Examination Survey (INSEF), collected between February and December 2015. This survey was described in more detail by Nunes et al.11 The present analysis was restricted to the subsample of INSEF participants from mainland Portugal (n=3467) who consented to have their data linked, provided a residence zip code number, were living within a 30-km radius of an air quality monitoring station with available PM10 concentration values and had blood pressure values measured in INSEF (n=2272). The INSEF survey received approval from the Ethics Committee of the Portuguese National Health Institute Doutor Ricardo Jorge, the National Data Protection Authority (Authorization no. 9348/2010) and from the regional Ethics Committees.

Health data collection in INSEF was performed by trained health professionals, according to the European Health Examination Survey (EHES) procedures.12 Blood pressure values were measured according to the previous published study of Rodrigues and colleagues.4

Sociodemographic (age, sex, educational level and occupation), lifestyle (smoking, excessive alcohol consumption, sedentary and unhealthy diet) and health status variables (diagnosed-dyslipidemia, diagnosed diabetes, lipid-lowering medication usage and diabetes medication usage) were obtained by self-report through the interview. Smoker's definition includes current daily and occasional smokers and excessive alcohol consumption was defined as reporting three or more days/week of consumption of at least one of the following alcoholic beverages: wine, beer, brandy/spirits, Port wine/martini/liqueur, whiskey/gin/vodka. Unhealthy diet was defined as no consumption of fruit and vegetables at least once a day and sedentary was defined as reporting reading, watching television or other sedentary activities as the best description of leisure time activities during the last 12 months.

Hourly validated PM10 values were obtained from the QualAr database, available online at the Portuguese Environment Agency (APA) website (https://qualar.apambiente.pt/) and the individual allocated PM10 concentrations were obtained as previously reported by Gaio et al.13 Briefly, for each individual the allocated three-day average PM10 concentration was the weighted average of three-day averaged PM10 concentrations of all stations within a 30 km radius from each participant's residence. This average was weighted by the inverse of the squared distance between the residence and the air quality monitoring stations.

The individually allocated three-day average temperatures were also obtained using data from the National Oceanic and Atmospheric Administration database (www.ncdc.noaa.gov) and we assumed the three-day average value of the closest temperature monitoring station as being representative of the individual exposure.

The statistical analysis was performed using the R statistical package version 3.6.3.14 All estimates were weighted to account for different selection probabilities resulting from the complex sample design of INSEF 2015 and to match the geographic region, age group and sex population distribution of 2015. T-test and the Wilcoxon test were used to access differences of quantitative variables according to their adherence to the normal distribution or not. Proportions were compared using Pearson's chi-squared test. The significance level for all analysis was set at 5%.

Regression coefficients of effect (β) of PM10 on DBP and SBP with the corresponding 95% confidence intervals (CIs) were obtained by generalized linear regression models analyses for each 1 μg/m3 increment of PM10. We used the svyglm function from the “survey” R package to run each Gaussian family model with a link log function (family=gaussian (link=“log”)). Then, percent change per 10 μg/m3 increment of PM10 with corresponding 95% CIs were calculated by using the formula 100×[exp (β)−1]*10. Additionally, to obtain estimates comparable with previous studies, we also run the models using the (family=gaussian (link=“identity”)) to obtain results in mmHg increase per each 10 μg/m3 increment of PM10 and these results are presented as supplementary data.

First, an unadjusted exposure-outcome model was fitted for each outcome. Then, a second model confounder-adjusted for sex (Male/Female), age group (25–49 years; 50–74 years), educational level (Low education/Medium education/High education), occupation (White-collar occupation/Blue-collar occupation), smoking (smoker/no smoker), excessive alcohol consumption (Yes/No), sedentary (Yes/No), unhealthy diet (Yes/No) and individual allocated three-day average temperature (continuous) was performed for each outcome. Additionally, as the objective was to assess only the short-term effects of PM10 exposure, an adjustment for the long-term effect of PM10 was also included by adding the variable one-previous year average of PM10 concentrations.

To assess the sensitivity of our analysis to the 30 km radius criteria, we also fit the models considering only participants living within a 20 km radius of an air quality monitoring station with available PM10 measurements. To evaluate the sensitivity of our analysis regarding the exposure assessment method, we also fit the models considering the modeled PM10 concentrations obtained by the application of a numerical air quality model – the WRF-CAMx modeling system that has been extensively applied over Portugal region with satisfactory skills, as previously described.13

Finally, we also repeat the analysis after excluding the participants taking anti-hypertensive medication and also when considering only more vulnerable groups, namely restricting the analysis to the hypertensive participants (defined as measured SBP ≥140 mmHg or measured DBP ≥90 mmHg or self-report of taking antihypertensive medication) or participants with abdominal obesity (assessed by using the formula: waist to height ratio >0.5).

The general characteristics ((Table S1) analyzed were similar among included and excluded participants. Among the 2272 participants in our study, 52.8% were females, 53.0% aged between 25 and 49 years old, 57.8% had low education level and 63.6% had a white-collar occupation. Most participants reported to be non-smokers (78.5%), non-excessive alcohol consumers (64.2%), to have a healthy diet (73.6%) and to be not sedentary (55.0%). The prevalence of hypertension was 36.4% and 25.1% of the participants reported to take antihypertensive medication. The mean values of DBP and SBP were 74.4 mmHg and 125.1 mmHg, respectively and the Individual allocated three-day average temperature was 16.1°C. Differences between males and females were found regarding level of education, occupation, smoking, alcohol consumption, unhealthy diet, hypertension prevalence and blood pressure values (Table 1).

The individual allocated three-day average PM10 concentration values ranged between 3.31 μg/m3 and 64.35 μg/m3 (median=17.19 μg/m3, IQR=12.40–24.27 μg/m3) and differences between males and females were not found (Figure 1). The distribution of the three-day average PM10 concentration values is above the daily limit value given by the EU Ambient Air Quality Directives (not to be exceed on more than 35/year) and also by the World Health Organization air quality guidelines (not to be exceed more than three days/year).

Figure 1.

Boxplot of the PM10 distribution according to sex. The dashed lines represent the daily limit value given by the EU Ambient Air Quality Directives (not to be exceed on more than 35 days/year) and also by the World Health Organization air quality guidelines (not exceeded more than 3 days/year).

(0.12MB).

There was no association between individual allocated three-day average PM10 concentration and DBP or SBP values (0.42% DBP increase per each 10 μg/m3 PM10 increment, 95% CI: −0.85%; 1.70% and 0.47% SBP increase per each 10 μg/m3 PM10 increment, 95% CI: −0.86%; 1.79%). Correspondent estimates in mmHg increase per each 10 μg/m3 increment of PM10 are presented as supplementary data, in Table S9 (0.33 mmHg DBP increase per each 10 μg/m3 PM10 increment, 95% CI: −0.62%; 1.27% and 0.57 mmHg SBP increase per each 10 μg/m3 PM10 increment, 95% CI: −1.10%; 2.54%). Similar results were obtained in the stratified analysis, when considering only males or females (Table 2).

Regarding the sensitivity analysis, when we restricted our sample to the participants living within a 20 km radius of an air quality monitoring station with available PM10 values, similar results were found when considering the total sample and also when considering both males or females. Per each 10 μg/m3 PM10 increment, there was a 0.51% (95% CI: −1.35; 2.37) increase in the DBP values and a 0.75% (95% CI: −1.02; 2.53) increase in the SBP values, being these results not statistically significant (Table S2). When we considered different exposure periods of time (2 previous days or 5 previous days instead of three previous days) similar results were found (Tables S3 and S4). Additionally, when we considered the individual allocated PM10 concentrations obtained by the air quality modeling system (WRF-CAMx), no associations were neither found (Table S5). Finally, when we restricted our analysis to participants not taking hypertensive medication, or when we considered only more susceptible individuals like those obese or diagnosed with hypertension, results remain similar (Tables S6–S8).

In this study, no statistically significant association was found between PM10 exposure and both DBP and SBP values. However, the point estimates obtained were similar or even higher than those from the last published meta-analysis about the global association between ambient air pollution and blood pressure6 (DBP: 0.33 versus 0.15 mmHg increase per each 10 μg/m3 PM10 increment; SBP: 0.57 versus 0.21 mmHg increase per each 10 μg/m3 PM10 increment). These results were consistently maintained by the sensitivity analysis performed, especially when considering a more restrictive criteria to the exposure assessment (participants living within a 20 km radius of an air quality monitoring station with available PM10 values), when considering the two or five previous days instead of three previous days as the exposure period of time, when considering an individual allocated PM10 concentrations obtained from an air quality model instead of the direct measurements from the air monitoring stations or when considering only hypertensive, participants with abdominal obesity or participants not taking anti-hypertensive medication.

According to a recent systematic review and meta-analysis performed on the association between ambient air pollution and blood pressure, that pools studies published before May 2017,6 evidence already supported the association between short-term PM10 exposure and DBP values (0.15 mmHg, 95% CI: 0.01; 0.29, per each 10 μg/m3 increase of PM10), which our study was unable to confirm. However, the point estimate obtained in our study was higher than the one reported by the cited meta-analysis (0.33 versus 0.15 mmHg increase per each 10 μg/m3 PM10 increment). Additionally, the same meta-analysis did not support the association between PM10 exposure and SBP values (0.21 mmHg, 95% CI: −0.01; 0.43, per each 10 μg/m3 increase of PM10), which is in concordance with our results.

It is important to mention that the polled estimates of the meta-analysis previously mentioned6 were obtained using data from studies mostly from America and Asia regions where the levels of PM10 are higher in comparison to the majority of the European countries. When we consider only the pooling of the estimates from studies performed in European countries,15–18 we found a −0.01 mmHg DBP decrease (95% CI: −0.14; 0.13) per each 10 μg/m3 increase of PM10. Accordingly, when considering only European countries, the published meta-analysis also did not support the association between short-term exposure to PM10 and diastolic (DBP) or systolic (SBP) blood pressure values, which is corroborated by our study.

After the publication of the previously cited meta-analysis,6 additional studies were published on this issue but the association between short-term PM10 exposure and blood pressure values remains controversial.19,20 The majority of the studies, mostly performed in China, provide evidence of a positive association between PM10 and blood pressure values19,21 but there are some other studies that did not support this association.20 For instance, in a panel study performed in Italy and Sweden, with a range of PM10 values comparable to those from Portugal, no associations were found between PM10 levels and either systolic (SBP) or diastolic (DBP) blood pressure but short-term exposure to PM10 results in reductions in carotid elasticity among elderly population which means that PM exposure could be associated with other important CV outcomes.20

This study suggests that exposure to PM10 concentrations in the levels observed in Portugal, similar to other European countries, has little or no effect on the blood pressure values, however it could have impact in other important cardiovascular outcomes. In fact, in a recent published study performed in the Portuguese population,13 a significant association between short-term PM10 exposure (3-day average PM10 concentration) and white blood cells (WBC) among females was found (a 2.76% WBC increase, 95% CI: 0.65–4.87, per each 10 μg/m3 PM10 increment) suggesting that PM10 exposure could impact on other outcomes with potential to mediate a cardiovascular event. Additionally, it is also important to mention that studies reporting significant results are more likely to be published and the possibility of a publication bias on this issue cannot be rejected.22,23

The point estimates obtained in this study were not statistically significant, which will be probably due to the lack of statistical power to estimate effects of this magnitude with a sample size of just over two thousand participants. However, even if the point estimates had statistical significance, we considered that obtained estimates would represent a small effect of the exposure to PM10 concentrations on blood pressure, taking into account the daily mean values of PM10 concentrations observed in the Portuguese context, in 2015 (three-day average PM10: 18.7 μg/m3).

Our study has some limitations. A first one is related to the exposure assessment method, due to the high distance considered between participants’ residence and the air quality monitoring stations (30 km) in the inclusion criteria. The number and the spatial distribution of air quality monitoring stations in the Portuguese mainland does not allow us to apply a more restrictive distance due to a substantial sample reduction that could compromise the power of the estimates and result in selection bias. However, our results remain similar when we apply more restrictive criteria (including only participants living within a 20 km radius of an air quality monitoring station with available PM10 measurements) or when we used data from a numerical model.

A second limitation of our study is related to the possibility of an erroneous assumption on the temporal relationship between exposure to PM10 and the BP changes. In fact, the considered time window of exposure, that was the three previous days in our study, varied a lot in the previously published studies, ranging from days to months and even years. Despite the time window of PM exposure that is causally linked to BP levels is not well established, according to the perspective of Newman and his collaborators,10 it will be on the order of a few days given the rapid occurrence of the physiological responses whereby PM influences on BP. Consequently, we assumed that the average three previous days’ PM10 concentrations were adequate to represent the individual exposure of each participant. Additionally, we also tested the associations when considering two and five previous days instead of the three previous days and results remain the same.

Finally, even considering the adjustment for several potential confounders, there is still a possibility that the observed estimates could have been affected by residual confounding due to other unmeasured confounders. In addition, effect estimates presented in this study were based on a single-pollutant model and considering only PM10, due the scarce available air quality monitoring data for the remaining air pollutants, but there are important interactions between the atmospheric pollutants, namely the potential additive effects of multiple pollutants, that should be considered in future studies.24,25