With rapid global urbanization, air and noise pollution in subway systems (also called metro systems) and the potential cardiovascular hazards they pose have become a global public health issue. However, the joint effect of multiple pollutants, as well as the key pollutants that play a dominant role in cardiovascular health, remain unclear. A randomized crossover study with respirator and/or headphone interventions was conducted among healthy young adults from March 11 to May 28, 2017 in the Beijing subway. This study found that co-exposure to size-fractioned particulate matter (PM), black carbon (BC), and noise was strongly associated with changes in heart rate variability (HRV) indices. BC and noise may be the two dominant pollutants causing the overall impact. Analysis based on data from intervention phases suggested that the impact of BC or noise might be attenuated or even reversed when reducing the level of another pollutant (noise or BC). The findings were helpful in guiding the formulation and development of prevention and control strategies for key traffic-related pollutants that endanger the cardiovascular health of commuters.
Overall, 40 healthy young adults were recruited for this randomized crossover study from March 11 to May 28, 2017, in Beijing. Details of the study implementation could be found in previous publications (1-2). In brief, participants commuted for about 4 hours between 9∶00–13∶00 in the subway during 4 different periods with/without intervention (wearing respirator and/or headphone). The personal real-time levels of PM1 (aerodynamic diameters <1 μm), PM1–2.5 (aerodynamic diameter ≥1 μm and <2.5 μm), PM2.5–10 (aerodynamic diameter ≥2.5 μm and <10 μm), BC, and noise were measured. Simultaneously, HRV parameters were obtained using a 12-channel ambulatory ECG monitor. Total power (TP), very-low-frequency power (VLF), low-frequency power (LF), high-frequency power (HF), LF/HF and standard deviation of normal-to-normal intervals (SDNN) were included in this study. Bayesian kernel machine regression (BKMR) was used to further examine the joint health effect of multiple pollutants in the subway cabin. This model allows nonlinear relationships and potential interactions and has been used widely to assess the overall effect of mixed pollutants (3). Posterior inclusion probability (PIP) was estimated through variable selection to assess the relative importance of exposure variables. PIP values range from 0 to 1 and a larger PIP value means a higher importance. All exposure and outcome variables were standardized. The covariates included gender, age, body mass index (BMI), temperature, relative humidity and CO2 level. The number of subjects was used as a random effect term to control the correlation between repeated measurements. The lag effects from 5 min to 2 hours were examined, and the most significant effect was reported at 30 min lag. BKMR analysis was conducted using R software (version 4.0.3; R project for Statistical Computing) with the “bkmr” package. The study protocol was approved by the Institutional Review Board of Peking University Health Science Center (IRB number: 00001052-16066) and informed consent was obtained from each participant.
In total, 39 participants completed this study, of whom 18 (46.2%) were females. Their mean age and BMI were 21.2 years and 21.6 kg/m2, respectively. The mean levels of PM1, PM1–2.5, PM2.5–10, BC, and noise were 34.1, 51.6, 145.2, 9.5 μg/m3, and 75.9 dBA, respectively. Increased levels of co-exposure to these pollutants were significantly associated with decreased TP, VLF, SDNN, and increased LF/HF (Figure 1). According to the results of PIPs in Table 1, BC had a higher PIP than other pollutants in most HRV indices, ranging from 0.69 to 1.00. Specifically, BC had the highest PIP value in TP (1.00), VLF (1.00), LF (1.00), and SDNN (0.96), indicating the largest contribution to the overall effect. Noise was the second most important because it had a high PIP in LF (1.00), HF (0.96), and LF/HF (0.72).Figure 1. The joint effect estimates and 95% confidence interval (CI) of multiple air pollutants (PM1, PM1–2.5, PM2.5–10, BC, and noise) in subway cabin on HRV parameters of study participants in Beijing, 2017 Note: The plot compared each HRV index when all exposures were at a particular quantile to when all were at the median (reference). Abbreviations: TP=total power; VLF=very low frequency power; LF=low frequency power; HF=high frequency power; SDNN=standard deviation of normal-to-normal intervals; HRV=heart rate variability.
Variable TP VLF LF HF LF/HF SDNN PM1 0.51 0.45 1.00 0.80 0.61 0.26 PM1-2.5 0.48 0.43 0.76 0.70 0.73 0.27 PM2.5-10 0.59 0.63 1.00 0.75 0.76 0.62 Black carbon 1.00 1.00 1.00 0.90 0.69 0.96 Noise 0.60 0.53 1.00 0.96 0.72 0.21 Abbreviations: PM=particulate matter; TP=total power; VLF=very low frequency power; LF=low frequency power; HF=high frequency power; SDNN=standard deviation of normal-to-normal intervals.
* PIP is a measure of the importance of exposure variables, and a larger value means a higher importance.
Table 1. Posterior inclusion probabilities (PIPs) from basyesian kernel machine regression model for heart rate variability parameters of healthy young adults in Beijing, 2017*.
To clarify the effectiveness and necessity of BC and noise prevention and control in the subway system, the exposure-response relationships of BC and noise with HRV indices were plotted based on the data from both no-intervention and headphone/respirator intervention phases. As shown in Figure 2, a weaker effect of BC on VLF, LF, HF, and SDNN was observed in the headphone intervention phase (low noise level) than the no-intervention one (high noise level), and the effect on LF/HF was even slightly reversed from a positive to negative association. For the effect of noise on all HRV indices, the relationships significantly differed in the respirator intervention phase (low BC concentration), and positive associations between noise and TP, VLF, LF, HF, and SDNN were found during this phase.Figure 2. The exposure-response relationship of black carbon (BC) and noise in subway cabin with heart rate variability (HRV) indices of participants in Beijing, 2017. (A) Comparison of the effects of BC on HRV indices between no intervention (high-noise level) and headphone intervention (low-noise level) phase based on basyesian kernel machine regression analysis; (B) Comparison of the effects of noise on HRV indices between no intervention (high-BC concentration) and headphone intervention (low-BC concentration) phase based on LOESS. Abbreviations: BC=black carbon; HRV=heart rate variability; TP=total power; VLF=very low frequency power; LF=low frequency power; HF=high frequency power; SDNN=standard deviation of normal-to-normal intervals; LOESS=locally weighted regression.