3. Results
3.1. Descriptive statistics
There were no missing data in this study. Table 1 provides the sum- mary statistics of these participants. Sixty percent of the subjects were hyperpietics and 40% were under regular medication of antihyperten- sive during our study period. The results from the self-administered questionnaires showed that they did not smoke and were not exposed to environmental tobacco smoke at home or the workplace. They also did not drink alcohol, participate in strenuous physical activity, or leave the urban areas 3 days before the scheduled body examination.
Table 2 provides the summary statistics of BP measurements. In total, we obtained 210 measurements of BP throughout the 6 follow- ups. On average, the range of variation in SBP for the same participant was25mmHg,17mmHginDBPand22mmHginPP.SBPwasstrongly correlated with PP (r = 0.82), and moderately correlated with DBP (r = 0.34). The correlations between PP and DBP were relatively weak (r = −0.26).
According to our study design, we performed 24 body examinations (6 times for 4 subgroups). Table 2 provides summary statistics of the 24-h averages of air pollutant concentrations and weather variables prior to each follow-up. PNCs decreased greatly with increasing diameter, and were dominated by particles b 0.40 μm which accounted for 86% of the total PNCs. There were 1615 particles b 1 μm per 1 cm3 air and only 5 for particles of 1.0–10 μm; therefore, we excluded PNC for size 1.0– 10 μm. Alternatively, we used PMC for 1.0–10 μm in this analysis. The av- erage PMC ≤ 2.5 μm was 60 μg/m3 during the follow-ups, much higher than the range of 10–20 μg/m3 commonly reported in North America and Western Europe (Dominici et al., 2006). The average temperature and relative humidity were 22 °C and 72%, respectively, reflecting the sub- tropical climate in Shanghai.
Generally, size-fractionated PNCs were strongly correlated with each other and with mass concentrations of PM2.5 (Pearson r = 0.8– 0.9), and moderately correlated with PM2.5–10 (r = 0.4–0.6); PNCs were moderately correlated with gaseous pollutants and weakly corre- lated with weather conditions (r = 0.1–0.2).
3.2. Regression results
In single-pollutant models, we estimated the associations between BP and size-fractionated PM over multiple single lags from 0 to 72 h (see Fig. 1). Overall, particles of all the fractions other than 2.5–10 μm are significantly associated with SBP and PP using at least one lag. How- ever, there was not any statistically significant associations between DBP and PM in all of the size ranges using all lag structures.