Available via license: CC BY 4.0
Content may be subject to copyright.
Association between exposure to particulate matter
and heart rate variability in vulnerable and
susceptible individuals: Application of the Bayesian
kernel machine regression model
Yong Whi Jeong
Yonsei University
Hayon Michelle Choi
Harvard T.H. Chan School of Public Health
Youhyun Park
Yonsei University
Yongjin Lee
Yonsei University College of Medicine
Ji Ye Jung
Yonsei University College of Medicine
Dae Ryong Kang
Wonju College of Medicine, Yonsei University
Article
Keywords:
Posted Date: September 27th, 2024
DOI: https://doi.org/10.21203/rs.3.rs-4983192/v1
License: This work is licensed under a Creative Commons Attribution 4.0 International License.
Read Full License
Additional Declarations: No competing interests reported.
Association between exposure to particulate matter and heart rate variability in vulnerable and
susceptible individuals: Application of the Bayesian kernel machine regression model
Yong Whi Jeong1, Hayon Michelle Choi2, Youhyun Park1, Yongjin Lee3, Ji Ye Jung4*, Dae Ryong Kang5*
1Department of Medical Informatics and Biostatistics, Graduate School, Yonsei University, Wonju,
Korea;
2Environmental and Occupational Medicine and Epidemiology Program, Harvard T.H. Chan School of
Public Health, Boston;
3Institute for Environmental Research, Yonsei University College of Medicine, Seoul, Korea;
4Division of Pulmonary and Critical Care Medicine, Department of Internal Medicine, Severance
Hospital, Yonsei University College of Medicine, Seoul, Korea;
5Department of Precision Medicine, Wonju College of Medicine, Yonsei University, Wonju, Korea
Address for correspondence:
*Ji Ye Jung,
Division of Pulmonary and Critical Care Medicine, Department of Internal Medicine, Severance
Hospital, Yonsei University College of Medicine,
50-1 Yonsei-Ro, Seodaemun-gu, Seoul 03722, Korea
E-mail: STOPYES@yuhs.ac
*Dae Ryong Kang,
Department of Precision Medicine, Wonju College of Medicine, Yonsei University,
20 Ilsan-ro, Wonju 26426, Korea.
E-mail: dr.kang@yonsei.ac.kr
Abstract
Particulate matter (PM) has various health effects, and individuals are simultaneously exposed to these
factors. Vulnerable and susceptible individuals are more sensitive to environmental factors than
nonvulnerable individuals. Exposure to PM causes cardiovascular diseases. Heart rate variability (HRV)
is a biomarker that may be used to identify cardiovascular diseases, and sensitive monitoring of HRV
is required. Most previous studies have evaluated exposure using environmental pollution monitoring
devices located in various districts. There is a lack of research exploring the relationship between
environmental pollutant exposure in personal living spaces and HRV using both indoor and outdoor
measurement devices. This study aimed to investigate the association between exposure to PM and
HRV using a model capable of multi-substance analysis in short-term exposures, in vulnerable and
susceptible individuals, including patients with environmental disease (patients with arrhythmia,
chronic airway disease, and stroke patients) and vulnerable populations (residents of an industrial
complex area, the elderly). We measured PM1.0, PM2.5, PM10, and digital biomarkers in 97 participants.
We evaluated the impact of short-term PM exposure on 24-h HRV over five days by measuring indoor
and outdoor exposure using personalized monitoring equipment and ECG monitoring via wearable
devices. The PM was calculated as a daily cumulative value and divided into days with high and low
cumulative concentrations. The association between exposure to single particulate and complex
mixtures and HRV was compared using multiple linear regression and Bayesian kernel machine
regression (BKMR). This study found that HRV showed a negative trend with increased PM exposure
on days with high cumulative PM concentrations, with statistically significant associations observed
between higher PM concentrations and decreased HRV on days with high exposure. The subgroup
analysis revealed that patients with chronic airway disease and residents of industrial complex areas
exhibited stronger negative correlations between exposure to PM and HRV. These associations were
more pronounced with complex exposure to PM1.0, PM2.5, and PM10. In short-term exposure, it was
confirmed that exposure to single and complex PM is negatively associated with HRV, and this
relationship varies depending on the sensitive characteristics of individuals. Integrating indoor and
outdoor personalized exposure assessments with 24-hour ECG monitoring has reinforced our
understanding of the complex interactions between PM and health. Our findings indicate that even
'acceptable' PM levels can harm HRV, suggesting that current thresholds may not adequately protect
sensitive individuals. This highlights the need for more stringent, particle size-specific standards for at-
risk groups.
INTRODUCTION
Environmental pollutants are known to have adverse effects on health, and air pollution, such as
particulate matter (PM), is a serious global issue due to its negative impact on human health. Numerous
studies have emphasized the effects of PM on mortality and morbidity related to cardiovascular diseases
(CVD) 1,2,3,4. Short- and long-term exposure to PM can have various effects on cardiovascular health,
with greater risks observed in vulnerable and susceptible populations, such as patients with
cardiovascular diseases, chronic obstructive pulmonary disease (COPD), and elderly individuals 1,3,5.
The proposed mechanisms to explain the association between PM inhalation and CVD include
oxidative stress, systemic inflammation 6,7, insulin resistance 8, epigenetic modifications 9, and
alterations in cardiac autonomic function and the autonomic nervous system 1,8. Notably, the impact on
the autonomic nervous system forms the basis of the link between PM and CVD, with heart rate
variability (HRV) serving as a useful non-invasive measure for assessing the autonomic regulation of
heart rhythm. HRV can be evaluated over short (typically 5–15 min) or long-term (24-h) periods 10.
Reduced HRV is a significant prognostic tool for various CVD, and long-term HRV assessments are
more accurate indicators for evaluating cardiovascular conditions than short-term HRV 10,11,12,13.
Epidemiological evidence suggests that short-term PM exposure is associated with a reduction in most
long-term HRV indices, particularly in sensitive groups such as the elderly 14, patients with
cardiovascular diseases 15,16,17, COPD patients 18, those with occupational exposure 19, or individuals
with hypertension or diabetes 5, 20.
There is consistent evidence that the association between the reduction in HRV and exposure to PM
depends on the particle size and subject population. In the general healthy population, occupational
exposure, and cardiovascular patients, exposure to PM1.0 has been linked to a reduction in HRV 21, 22. In
cardiovascular patients, patients with COPD, and the elderly, exposure to PM2.5, has been associated
with HRV reduction 15,23,24, while in patients with COPD, PM10 exposure has been linked to decreased
HRV 18. The type of particles and the magnitude of their association with HRV reduction may vary
depending on the subject population and specific pollutants involved.
Most previous studies have evaluated exposure using environmental pollution monitoring devices
located in various districts. There is a lack of research exploring the relationship between environmental
pollutant exposure in personal living spaces and HRV using both indoor and outdoor measurement
devices. Unlike outdoor air, indoor air quality can be altered at an individual level 25; therefore, both
indoor and outdoor air should be considered when assessing PM exposure 26. Additionally, because
individuals are not exposed to environmental pollutants in isolation but rather to a mixture of various
substances, an analysis of complex exposure is also necessary. Some complex exposure analyses have
utilized combinations of PM components 27, 28; however, the measurements did not fully reflect the
overall personal living environment and were not evaluated according to particle size.
Our study addresses these gaps by employing a comprehensive approach that simultaneously
examines indoor and outdoor air quality, while also considering the mixed exposure to different PM
sizes (PM1.0, PM2.5, and PM10) in real-time. Furthermore, the application of a Bayesian kernel machine
regression model (BKMR) 29 allows us to more accurately model the nonlinear and potentially
synergistic effects of mixed-substance exposures on HRV, particularly in vulnerable populations. This
novel approach not only fills a critical void in existing research but also provides a more nuanced
understanding of the health risks posed by particulate matter in diverse living environments. Therefore,
this study specifically aimed to investigate the relationship between short-term exposure and 24-h real-
time HRV measurements based on PM (PM1.0, PM2.5, and PM10) in patients with environmental diseases
(patients with arrhythmia, chronic airway disease, and stroke) and vulnerable populations (residents of
an industrial complex area, the elderly), using equipment capable of measuring both indoor and outdoor
environments. The study will take into account individual sensitive characteristics (medical history, age,
gender, disease) and personal living spaces and will examine both single-substance and mixture-
substance exposures.
METHODS
Study participants
In this study, a living labs was established to assess personalized exposure in patients with
environmental diseases and vulnerable populations. Patients with environmental diseases include those
with circulatory diseases, respiratory diseases, and neurological diseases, whereas vulnerable
populations include the elderly and industrial area residents. The circulatory disease group comprised
patients with atrial fibrillation or implanted cardiac devices (N=20). The respiratory disease group
included patients with COPD, asthma, or bronchiectasis (N=20). The neurological disease group
included 20 patients with acute stroke patients (N=20). The elderly group in the vulnerable population
includes individuals aged 60 and over who do not have severe chronic diseases or cognitive impairment
(N=20). Residents of industrial complexes were those living in industrial regions without severe chronic
diseases or neurological conditions (N=20). A total of 100 individuals were recruited from sensitive
and vulnerable living laboratories. However, three participants were excluded from the analysis due to
issues with device compliance, resulting in the analysis of data from 97 participants. This study was
approved by the Clinical Trial Review Committee of the Yonsei University Wonju Severance Christian
Hospital (approval number: CR321068). To use the data from the recruited participants, approvals were
obtained from the Institutional Review Board of Severance Hospital (4-2021-0550, 4-2021-0852),
Kyung Hee University Hospital (KNUH 2021-07-074-001), Gachon University Gil Medical Center
(GIRB-2021-351), and Gyeongsang National University (GIRB-A21-Y-0053).
Data collection
Demographic and clinical data of the participants, including age, sex, body mass index, smoking status,
and medical history, were obtained at the time of participant registration. Each participant visited their
respective hospital once a year over four years. During each visit, the participants were assessed for
personal health examination information, hospital visit history, and other relevant data. A personalized
exposure assessment was conducted to measure the exposure levels of environmental pollutants and
digital biomarkers in various microenvironments where the participants spent time tailored to their daily
living patterns. The total measurement period was five days. The research participants met in person on
the first day of the measurement, and the necessary measuring equipment was provided. Participants
carried the equipment during their daily activities. To ensure consistency in the measurements, the
research team visited the participants daily to check the equipment status. A daily activity log was
provided to the participants at the initial meeting to record their daily living patterns, and they were
instructed to fill it out each day. The quality of the daily activity logs was reviewed during the daily
visits. If the quality was found to be low, participants were interviewed about their activities from the
previous day to make necessary corrections. Because physical activity intensity can act as a moderating
variable affecting the level of PM exposure, an accelerometer was provided to measure the metabolic
equivalent of task (MET) with the participants' consent, excluding periods of showering and sleeping.
Digital biomarkers were measured using ECG monitoring during the same period as environmental
pollutants were measured, and the participants wore the device continuously unless a device
malfunction occurred. The researchers continuously monitored the environmental pollutant data and
physiological signals, and if data were not received, they managed the situation by contacting the
participants by phone. The living labs for residents of the industrial complex area was conducted from
December 2021 to February 2022; for arrhythmia patients, from June to August 2022; for chronic
airway disease patients, from October to November 2022; for the elderly, in June 2023; and for stroke
patients, in July 2023.
Assessment of exposures
Environmental pollutant data were measured using a lightweight portable device, AIR HEART-P21
(manufactured by ZINIDE, South Korea), which allows for real-time measurements every minute using
a light scattering method. The device measured PM1.0 (aerodynamic diameters <1 μm), PM2.5
(aerodynamic diameter ≥1 μm and <2.5 μm), PM10 (aerodynamic diameter ≥2.5 μm and <10 μm),
TVOC (total volatile organic compounds), temperature, and humidity. The data collected over the five
days, based on individual daily activity patterns, were converted into daily cumulative values for
comparison with 24-h HRV associated with short-term exposure. Accordingly, the days with the highest
and lowest cumulative PM concentrations were identified. Although data were collected over five days
for each individual, the start and end times of the measurements varied. Consequently, the first and last
days, when the measurement durations were shorter, fell under the days with the lowest PM
concentrations. To address this, the values from the first and last days were excluded and only data from
the three middle days were used to differentiate between days with high and low cumulative PM
concentrations.
ECG monitoring
Physiological signals were measured using the HiCardi SmartPatch (SW 1.101), a two-electrode patch
manufactured by Mezoo Co. Ltd (RM.808 200, Gieopdosi-ro, Jijeong-myeon, Wonju-si, Gangwon-do,
Republic of Korea) 30. The HiCardi device records 15,000 data points per minute at a frequency of 250
Hz. However, segments in which measurements were not possible due to poor contact or out-of-range
Bluetooth communication were excluded. Among the various methods for defining heart rate variability
(HRV), we excluded frequency-domain segments that could not be synchronized with the PM data. We
evaluated HRV using the standard deviation of normal-to-normal intervals (SDNN ), standard deviation
of the average NN intervals for each 5-minute segment of a 24-h HRV recording (SDANN), mean of
the standard deviations of all NN intervals for each 5-minute segment of a 24-h HRV recording
(SDNNI), and root mean square of successive RR interval differences (RMSSD) 31. As the cumulative
concentration of PM was evaluated daily, HRV was calculated over 24 h, and three preprocessing steps
were required to derive HRV from the ECG signals. First, outliers corresponding to extreme values due
to device compliance or measurement errors were removed. Second, noise in the signal, which would
obscure the identification of the R-peaks necessary for calculating the HRV, was removed using a
bandpass filter [32]. Third, missing values are replaced using interpolation methods 33.
Statistical analysis
Categorical variables are described as numbers and percentages. Continuous variables are summarized
as means and standard deviations or medians with minimum and maximum values. The independent t-
test or Wilcoxon rank-sum test was used to compare the baseline characteristics of PM based on the
cumulative concentration and HRV. To evaluate the association between environmental pollutants (daily
average PM1.0, PM2.5, and PM10) and HRV based on the days with the highest and lowest cumulative
PM concentrations, a multiple linear regression model was employed, adjusting for variables including
age, sex, BMI, respiration rate, smoking, alcohol consumption, METs, hypertension, diabetes mellitus,
TVOC, humidity, and temperature. To assess the association between complex exposure to PM (daily
average PM1.0, PM2.5, and PM10) of different particle sizes and HRV, the BKMR model was used to
determine the impact of environmental pollutants on the days with the highest and lowest cumulative
levels. BKMR is a flexible statistical method that allows the assessment of individual and joint effects
of exposure mixtures. It also accounts for the exposure-response relationships for each component of
the mixture, while identifying potential interactions using kernel functions 29. We used two BKMR
models on different days to evaluate the PM effects of PM. First, the models were constructed based on
days with high and low cumulative PM concentrations. Second, a model was developed to categorize
the evaluation based on environmental diseases and vulnerable populations. The combined effects of
the three mixtures (daily average PM1.0, PM2.5, and PM10) were examined in all models, and each model
was adjusted using the same variables as the regression models mentioned above. The software used
for the analyses was SAS (version 9.4; SAS, Cary, NC, USA) and R 4.0.3 (Institute for Statistics and
Mathematics, Vienna, Austria; http://cran. rproject. org).
RESULTS
Table 1 presents the basic descriptive statistics for each labs. The total number of study participants was
97, including 19 patients with arrhythmia, 20 with chronic airway disease, 19 with stroke in the
environmental disease group, 19 residents in an industrial complex area, and 19 elderly individuals in
the vulnerable group. There were more males than females, and the average age was 65.52 (12.04) years,
with the elderly group having the highest average age of 71.42 years. The average concentrations of
PM1.0, PM2.5, and PM10 were 9.89 (9.78), 10.46 (10.36), and 10.53 (10.44), respectively, with industrial
area residents having the highest average concentrations of PM1.0, PM2.5, and PM10. The participants'
24-h SDNN was 138.96 (63.06), SDANN was 98.25 (39.26), SDNNI was 68.01 (41.11), and RMSSD
was 77.37 (66.07). Among these, 24-h SDNN, SDANN, and SDNNI were the lowest in patients with
arrhythmia, whereas RMSSD was the lowest in patients with chronic airway diseases.
The concentrations of environmental pollutants and 24-h HRV were compared based on the days with
the highest and lowest cumulative PM concentrations (Table 2). The concentrations of PM1.0, PM2.5,
and PM10 were significantly higher on days with high cumulative concentrations. Although there was
no statistically significant difference in 24-h HRV between the days with high and low cumulative
concentrations, the values were lower on the days with high cumulative concentrations; SDNN was
137.60 (57.98), SDANN was 97.34 (37.31), SDNNI was 66.67 (39.14), and RMSSD was 76.85 (63.52),
compared to the days with low cumulative concentrations.
Figure 1 shows the results of the regression analysis stratified by days with high and low cumulative
concentrations of each environmental pollutant using age, sex, BMI, respiration rate, smoking, alcohol
consumption, METs, hypertension, diabetes mellitus, TVOC, humidity, and temperature as adjustment
variables. On both the high and low cumulative concentration days, each PM substance showed a
negative trend with 24-h HRV. Notably, on the days with high cumulative concentrations, PM1.0 PM1.0
(𝛽=-1.34, 95% CI=-2.45, -0.23), PM2.5 (𝛽=-1.28, 95% CI=-2.33, -0.23), PM10 (𝛽=-1.28, 95% CI=-2.32,
-0.24) exhibited a significant negative correlation with SDNN.
The overall effects exerted by the three PMs (PM1.0, PM2.5, and PM10) were estimated using the
BKMR method and stratified by days with high and low PM cumulative concentrations (Figure 2). This
regression model was adjusted for covariates including age, sex, BMI, respiration rate, smoking, alcohol
consumption, METs, hypertension, diabetes mellitus, TVOC, humidity, and temperature on both the
high cumulative concentration day (Figure 2 (A)) and the low cumulative concentration day (Figure 2
(B)). HRV (SDNN, SDANN, SDNNI, RMSSD) showed a negative trend as the degree of complex
exposure increased. On days with high cumulative concentrations, SDNN and the three PMs showed a
significantly negative association from the 70th to the 90th percentile compared with the 50th percentile
of exposure levels. Significance was observed at the 90th percentile when the dependent variables were
SDANN and SDNNI. Furthermore, when examining the exposure-response relationship on days with
high and low cumulative concentrations, it was confirmed that an increase in the exposure levels of the
three PMs was associated with a decrease in SDNN, SDANN, SDNNI, and RMSSD, demonstrating a
negative trend (Supplementary Figure 1).
Table 3 presents the results of the subgroup analysis based on the environmental disease group
(patients with arrhythmia, chronic airway disease, and stroke) and vulnerable group (residents of an
industrial complex area and the elderly). On days with high cumulative concentrations, regression
analysis was performed on the HRV for each environmental pollutant within each living labs, using age,
sex, BMI, respiration rate, smoking, alcohol consumption, METs, hypertension, diabetes mellitus,
TVOC, humidity, and temperature as adjustment variables. In both the environmental disease and
vulnerable groups, each substance generally showed a negative correlation with HRV. In the
environmental disease group, chronic airway disease showed a negative correlation with all HRV
indices for PM1.0, PM2.5, and PM10. In the vulnerable group, a negative association was observed
between PM1.0, PM2.5, and PM10 when the outcome was SDNN; residents of an industrial complex area
showed a significant negative correlation when SDNN was the outcome variable. On days with low
cumulative concentrations, exposure to PM substances generally demonstrated a negative trend with
24-h HRV, this was not statistically significant (Supplementary Table 2).
A subgroup analysis was conducted on the complex exposure assessment for each living labs (Figure
3). On days with high cumulative concentrations, the three PMs and SDNN showed a negative
correlation at the 90th percentile of exposure compared with when all exposures were fixed at the 50th
percentile in both the environmental disease and vulnerable groups. Additionally, in the environmental
disease group, particularly among patients with chronic airway disease, SDNN significantly decreased
when exposed to 3PMs at the 70th, 80th, and 90th percentiles compared to when all exposures were
fixed at the 50th percentile. Similarly, in the vulnerable group, particularly among residents of an
industrial complex area, SDNN significantly decreased when exposed to 3PMs at the 70th, 80th, and
90th percentiles. On days with low cumulative concentrations, complex exposure generally showed a
negative trend with 24-h HRV; this was not statistically significant (Supplementary Figure 2).
DISCUSSION
Vulnerable and susceptible individuals were recruited to evaluate the association between PM exposure
and HRV. During the five-day measurement period, we differentiated between days with high and low
cumulative concentrations to compare the associations between single-substance and complex-
substance exposures. On days with low cumulative concentrations of PM, we observed a negative trend
without statistical significance; however, on days with high cumulative concentrations, we found a
significant negative association. In the overall study population, both single- and complex-substance
exposure to PM1.0, PM2.5, and PM10 showed a negative association with SDNN, consistent with previous
studies that found that increased exposure to PM was associated with decreased HRV 18,19. In the
subgroup analysis by living labs, different associations were observed depending on the group. In the
environmental disease group, PM exposure was associated with a decrease in SDNN among patients
with chronic airway disease 18, whereas in the vulnerable group, residents of an industrial complex area
showed an association between increased PM exposure and decreased SDNN 19. However, unlike
previous studies, we did not find an association between reduced HRV and PM exposure in patients
with arrhythmia, stroke, or the elderly 14,15,16,17. This discrepancy may be due to differences in population
characteristics, study design, or environmental conditions, which warrants further investigation.
In this study, we highlight the importance of PM exposure in susceptible and vulnerable populations
by building on previously explored aspects. First, most previous studies evaluated exposure using
equipment located in urban areas, which cannot accurately reflect individual-specific information. We
used equipment capable of measuring environmental pollutants, both indoors and outdoors, to assess
personalized exposure levels in sensitive and vulnerable populations. At the same time, ECG signals
were measured in real-time using devices that could monitor participants during the same period when
environmental pollutants were being measured. This allowed for an accurate assessment of exposure
levels based on the participants’ residential areas, indoor and outdoor environments, and lifestyle habits
25. Beyond PM measurement, our equipment also recorded TVOC, temperature, and humidity, while
the ECG monitoring devices provided data on respiration rate and metabolic equivalents (METs). This
comprehensive data collection enabled us to control for individual characteristics as adjustment
variables, thereby enhancing the precision and relevance of our findings.
Second, after categorizing the exposure levels to environmental pollutants based on daily cumulative
concentrations, we analyzed the data by comparing days with high and low cumulative concentrations.
Interestingly, we found that the exposure levels on both high and low concentration days were still
within the 'good' or 'moderate' categories as defined by the WHO 34, the U.S. AQI 35, and the Korean
Ministry of Environment 36 for daily PM exposure (see Supplementary Table 1). Nevertheless, our
findings clearly demonstrate that even within these 'acceptable' levels, increased exposure to PM
significantly reduces heart rate variability in sensitive and vulnerable populations. Furthermore, we
observed a stronger association with decreased HRV as the PM particle size decreased 18, 37. These
findings underscore the necessity for more stringent, particle size-specific air quality standards for
vulnerable individuals, compared to those for the general population.
Third, we assessed the impact of exposure to complex mixtures of PM, in addition to single
substances, on HRV. This analysis revealed that the combination of PM particles of varying sizes had a
more pronounced impact on HRV than single-substance exposures alone. In the overall study population,
only SDNN was associated with single substance exposure; however, in the case of complex substance
exposure, significant associations were observed not only with SDNN but also with SDANN and
SDNNI. Notably, a negative association with SDNN was observed in both the overall environmental
disease and vulnerable groups, with statistically significant associations generally observed at the 90th
percentile of exposure. Although the daily average PM levels for both indoor and outdoor environments
were within the globally recommended 'good' and 'moderate' categories 34,35,36, our results suggest that
individuals with sensitive characteristics may be more strongly affected by decreased HRV when
exposed to a complex mixture of particles of varying sizes. These findings were made possible by the
application of BKMR, a sophisticated statistical approach that allows for the modeling of high-
dimensional and complex exposure-response relationships 29. BKMR enabled us to simultaneously
assess the effects of multiple pollutants, accounting for potential interactions between different PM sizes.
This approach provided a more nuanced understanding of how combinations of pollutants, rather than
individual substances alone, contribute to reductions in HRV. By capturing the non-linear and
potentially synergistic effects of mixed-substance exposures, BKMR enhanced the accuracy and
interpretability of our results, particularly in identifying the heightened vulnerability of certain
populations to complex pollutant mixtures.
Fourth, we conducted analyses by separating the participants into groups based on specific diseases
and vulnerable group characteristics. This approach demonstrated that the association between PM
exposure and HRV can vary depending on factors such as medical history, age, and region of residence.
Notably, patients with chronic airway disease and residents of an industrial complex area showed a
strong negative association with SDNN in both single and mixed PM substance exposures. In particular,
patients with chronic airway disease showed prominent associations across all HRV indices, indicating
a heightened sensitivity to PM exposure in this group. These two groups were measured during the
COVID-19 pandemic, when wearing masks was recommended outdoors, potentially resulting in lower
outdoor exposure levels, with most measurements being taken indoors (Supplementary Table 1). These
results may have influenced the observed associations, as reduced outdoor exposure could have altered
the overall exposure, emphasizing the importance of indoor air quality. While this study did not
specifically explore the mechanisms underlying the observed decrease in HRV due to PM exposure in
these groups, it is plausible that residents of industrial complex areas experience accumulated exposure
due to their living environment. Similarly, patients with chronic airway disease who suffer from a
progressive disease characterized by lung damage and inflammation due to particle and gas exposure
may experience heightened inflammation during respiration, making PM exposure more impactful 18.
These findings suggest that targeted interventions, especially those aimed at reducing indoor PM
exposure, may be necessary for individuals or groups with sensitive characteristics 26, 38.
Despite the advances reported in this study, several limitations must be addressed. First, the
participants were evaluated using electrocardiography (ECG) monitoring devices to assess HRV. Issues
such as device compliance or discomfort during use can lead to errors in data measurements 39. Although
we attempted to correct these errors through interpolation and outlier removal, the values might still be
overestimated compared with known HRV indices. Secondly, each living lab participant was evaluated
at different time points. Specifically, industrial residents, arrhythmia patients, and patients with
respiratory diseases were assessed during the COVID-19 pandemic. As mask-wearing was
recommended both indoors and outdoors during this period, the inhalation of environmental pollutants
may have been reduced 40. As a result, unlike previous studies, we only observed a trend of decreased
HRV with PM exposure in patients with arrhythmia, without statistically significant results 15,16,17. Third,
our evaluation of 24-h HRV based on daily exposure likely integrated various physiological responses
occurring under different conditions throughout the day. Although this averaging approach stratified the
data by days of highest and lowest cumulative exposure, the use of daily average values may have
obscured the precise effects of indoor and outdoor exposure from the exposure variable perspective, as
well as the differences in HRV between resting and active periods from the outcome variable perspective.
Future research should focus on disentangling the variables influencing HRV by separately analyzing
responses to indoor and outdoor PM exposures and examining HRV variations during activity and rest.
This would provide a clearer understanding of how environmental contexts and daily routines impact
HRV, enabling more targeted interventions. Additionally, exploring the biological mechanisms
underlying the significant associations found in vulnerable groups is crucial for understanding the long-
term health implications of PM exposure. Moreover, longitudinal studies incorporating repeated
measurements over extended periods are essential to move beyond mere associations and toward causal
inference. By capturing the chronic impacts of PM exposure on HRV over time, such studies would
enhance our understanding of the enduring effects of air pollution on cardiovascular health, ultimately
supporting the development of more effective public health policies and interventions.
Overall, we found that increased exposure to PM was associated with decreased HRV, and this
relationship varied depending on the sensitive characteristics of the individuals. The data with the
integration of indoor and outdoor personalized exposure assessments with 24-hour ECG monitoring has
enhanced our understanding of the complex interactions between PM and health. Furthermore, by
demonstrating that even 'acceptable' levels of PM exposure, as defined by global standards, can have
detrimental effects on HRV, our findings suggest that existing thresholds may not adequately protect
sensitive individuals. This underscores the need for more stringent, particle size-specific standards
tailored to these at-risk groups.
Author contributions
Y.W.J, and D.R.K. designed the study. Y.L. and J.Y.J. collected data. Y.W.J. and Y.P. conducted data
analyses. Y.W.J. and H.M.C. drafted the manuscript, with further development by J.Y.J. and D.R.K.,
J.Y.J. and D.R.K. supervised the study. All authors contributed to the revision of the manuscript and
approval of the final version.
Acknowledgments
This work was supported by the Korea Environment Industry & Technology Institute(KEITI) through
the Digital Infrastructure Building Project for Monitoring, Surveying, and Evaluating Environmental
Health funded by the Korea Ministry of Environment(MOE) (RS-2021-KE001338). Also, we would
like to thank Editage (www.editage.co.kr) for English language editing.
Competing interests
All authors declare no financial or non-financial competing interests.
Data availability
All data generated or analyzed during this study are included in this published article [and its
supplementary information files].
REFERENCES
1 Brook, R. D. et al. Particulate matter air pollution and cardiovascular disease: an update to the
scientific statement from the American Heart Association. Circulation 121, 2331-2378 (2010).
2 Gold, D. R. & Mittleman, M. A. New insights into pollution and the cardiovascular system:
2010 to 2012. Circulation 127, 1903-1913 (2013).
3 Pope III, C. A. et al. Cardiovascular mortality and long-term exposure to particulate air
pollution: epidemiological evidence of general pathophysiological pathways of disease. Circulation 109,
71-77 (2004).
4 Taneepanichskul, N. et al. Short-term effects of particulate matter exposure on daily mortality
in Thailand: a case-crossover study. Air Quality, Atmosphere & Health 11, 639-647 (2018).
5 Pieters, N. et al. An epidemiological appraisal of the association between heart rate variability
and particulate air pollution: a meta-analysis. Heart 98, 1127-1135 (2012).
6 Araujo, J. A. Particulate air pollution, systemic oxidative stress, inflammation, and
atherosclerosis. Air Quality, Atmosphere & Health 4, 79-93 (2011).
7 Brook, R. D. et al. Air pollution and cardiovascular disease: a statement for healthcare
professionals from the Expert Panel on Population and Prevention Science of the American Heart
Association. Circulation 109, 2655-2671 (2004).
8 Rajagopalan, S. & Brook, R. D. Air pollution and type 2 diabetes: mechanistic insights.
Diabetes 61, 3037-3045 (2012).
9 Peng, C. et al. Particulate air pollution and fasting blood glucose in nondiabetic individuals:
associations and epigenetic mediation in the normative aging study, 2000–2011. Environmental health
perspectives 124, 1715-1721 (2016).
10 Electrophysiology, T. F. o. t. E. S. o. C. t. N. A. S. o. P. Heart rate variability: standards of
measurement, physiological interpretation, and clinical use. Circulation 93, 1043-1065 (1996).
11 Kleiger, R. E., Miller, J. P., Bigger Jr, J. T. & Moss, A. J. Decreased heart rate variability and
its association with increased mortality after acute myocardial infarction. The American journal of
cardiology 59, 256-262 (1987).
12 Bigger Jr, J. T. et al. Frequency domain measures of heart period variability and mortality after
myocardial infarction. Circulation 85, 164-171 (1992).
13 Tsuji, H. et al. Impact of reduced heart rate variability on risk for cardiac events: the
Framingham Heart Study. Circulation 94, 2850-2855 (1996).
14 Mordukhovich, I. et al. Exposure to sub-chronic and long-term particulate air pollution and
heart rate variability in an elderly cohort: the Normative Aging Study. Environmental Health 14, 1-10
(2015).
15 Zanobetti, A. et al. Reduction in heart rate variability with traffic and air pollution in patients
with coronary artery disease. Environmental health perspectives 118, 324-330 (2010).
16 Huang, W. et al. Air pollution and autonomic and vascular dysfunction in patients with
cardiovascular disease: interactions of systemic inflammation, overweight, and gender. American
journal of epidemiology 176, 117-126 (2012).
17 Breitner, S. et al. Ambient and controlled exposures to particulate air pollution and acute
changes in heart rate variability and repolarization. Scientific reports 9, 1946 (2019).
18 Pan, L. et al. The short-term effects of indoor size-fractioned particulate matter and black
carbon on cardiac autonomic function in COPD patients. Environment International 112, 261-268
(2018).
19 Magari, S. R. et al. Association of heart rate variability with occupational and environmental
exposure to particulate air pollution. Circulation 104, 986-991 (2001).
20 Park, S. K., O’Neill, M. S., Vokonas, P. S., Sparrow, D. & Schwartz, J. Effects of air pollution
on heart rate variability: the VA normative aging study. Environmental health perspectives 113, 304-
309 (2005).
21 Baig, N. A. et al. Association between traffic emissions mixed with resuspended dust and heart
rate variability among healthy adults in Delhi. Air Quality, Atmosphere & Health 13, 371-378 (2020).
22 Chuang, K.-J., Chan, C.-C., Chen, N.-T., Su, T.-C. & Lin, L.-Y. Effects of particle size fractions
on reducing heart rate variability in cardiac and hypertensive patients. Environmental health
perspectives 113, 1693-1697 (2005).
23 Creason, J. et al. Particulate matter and heart rate variability among elderly retirees: the
Baltimore 1998 PM study. Journal of Exposure Science & Environmental Epidemiology 11, 116-122
(2001).
24 Gold, D. R. et al. Ambient pollution and heart rate variability. Circulation 101, 1267-1273
(2000).
25 Klepeis, N. E. et al. The National Human Activity Pattern Survey (NHAPS): a resource for
assessing exposure to environmental pollutants. Journal of exposure science & environmental
epidemiology 11, 231-252 (2001).
26 Hansel, N. N. et al. Randomized clinical trial of air cleaners to improve indoor air quality and
chronic obstructive pulmonary disease health: results of the CLEAN AIR study. American journal of
respiratory and critical care medicine 205, 421-430 (2022).
27 Hu, J. et al. Fine particulate matter constituents and heart rate variability: A panel study in
Shanghai, China. Science of The Total Environment 747, 141199 (2020).
28 Peralta, A. A., Schwartz, J., Gold, D. R., Coull, B. & Koutrakis, P. Associations between PM2.
5 metal components and QT interval length in the Normative Aging Study. Environmental research 195,
110827 (2021).
29 Bobb, J. F. et al. Bayesian kernel machine regression for estimating the health effects of multi-
pollutant mixtures. Biostatistics 16, 493-508 (2015).
30 Park, Y. et al. Development and Validation of a Real-Time Service Model for Noise Removal
and Arrhythmia Classification Using Electrocardiogram Signals. Sensors 24, 5222 (2024).
31 Stein, P. K., Bosner, M. S., Kleiger, R. E. & Conger, B. M. Heart rate variability: a measure of
cardiac autonomic tone. American heart journal 127, 1376-1381 (1994).
32 Fischer, S. E., Wickline, S. A. & Lorenz, C. H. Novel real‐time R‐wave detection algorithm
based on the vectorcardiogram for accurate gated magnetic resonance acquisitions. Magnetic
Resonance in Medicine: An Official Journal of the International Society for Magnetic Resonance in
Medicine 42, 361-370 (1999).
33 Ishaque, S., Khan, N. & Krishnan, S. Trends in heart-rate variability signal analysis. Frontiers
in Digital Health 3, 639444 (2021).
34 Organization, W. H. Air quality guidelines: global update 2005: particulate matter, ozone,
nitrogen dioxide, and sulfur dioxide. (World Health Organization, 2006).
35 Plaia, A. & Ruggieri, M. Air quality indices: a review. Reviews in Environmental Science and
Bio/Technology 10, 165-179 (2011).
36 Environment, M. o. Annual Report of Ambient Air Quality in Korea (Ministry of Environment
Seoul, 2010).
37 Huang, C. et al. Particulate matter air pollution and reduced heart rate variability: How the
associations vary by particle size in Shanghai, China. Ecotoxicology and Environmental Safety 208,
111726 (2021).
38 Kim, H. et al. The impact of life behavior and environment on particulate matter in chronic
obstructive pulmonary disease. Environmental Research 198, 111265 (2021).
39 Kwon, S. et al. Validation of adhesive single-lead ECG device compared with holter
monitoring among non-atrial fibrillation patients. Sensors 21, 3122 (2021).
40 Langrish, J. P. et al. Beneficial cardiovascular effects of reducing exposure to particulate air
pollution with a simple facemask. Particle and fibre toxicology 6, 1-9 (2009).
Table 1. Baseline characteristics of study population.
Living labs for environmental diseases
Living labs for vulnerable groups
Total
Patients with Arrhythmia
Patients with
Chronic Airway Disease
Patients with Stroke
Residents of Industrial
Complex Area
Elderly
(N = 97)
(N = 19)
(N = 20)
(N = 20)
(N = 19)
(N=19)
Sex (male)
59 (60.82)
13 (68.42)
17 (85.00)
13 (65.00)
7 (36.84)
9 (47.97)
Age (year)
65.52 (12.04)
65.42 (9.72)
59.55 (16.38)
62.15 (12.40)
69.53 (10.47)
71.42 (4.25)
BMI (kg/m2)
24.50 (3.67)
25.65 (2.61)
23.78 (2.56)
24.58 (2.72)
25.63 (3.31)
22.91 (5.76)
Smoking (yes)
11 (11.34)
2 (10.53)
2 (10.00)
1 (5.00)
4 (21.05)
2 (10.53)
Alcohol consumption (yes)
46 (47.42)
8 (42.11)
11 (55.00)
7 (35.00)
11 (57.89)
9 (47.37)
METs
160.81 (100.18)
208.03 (135.87)
162.47 (106.51)
122.55 (101.45)
155.34 (66.57)
157.59 (61.37)
Hypertension (yes)
52 (53.61)
12 (63.16)
9 (45.00)
16 (80.00)
9 (47.37)
6 (31.58)
Diabetes mellitus (yes)
18 (18.56)
4 (21.05)
0 (0.00)
7 (35.00)
3 (15.79)
4 (21.05)
PM1.0 (μg/𝑚3)
Mean (SD)
9.89 (9.78)
10.45 (11.46)
7.95 (4.55)
10.19 (13.11)
13.62 (11.49)
7.33 (2.73)
Median (q1, q3)
6.96 (4.68, 10.79)
7.37 (3.23, 12.30)
6.60 (4.67, 9.96)
6.73 (5.27, 9.74)
9.78 (6.13, 15.49)
6.72 (5.39, 8.68)
PM2.5 (μg/𝑚3)
Mean (SD)
10.46 (10.36)
10.94 (12.11)
8.44 (4.80)
10.77 (13.79)
14.51 (12.28)
7.72 (2.84)
Median (q1, q3)
7.36 (4.98, 11.35)
7.50 (3.31, 13.00)
7.04 (4.92, 10.48)
7.07 (5.75, 10.25)
10.34 (6.44, 16.45)
7.21 (5.67, 8.86)
PM10 (μg/𝑚3)
Mean (SD)
10.53 (10.44)
11.03 (12.11)
8.48 (4.83)
10.80 (13.86)
14.68 (12.51)
7.75 (2.86)
Median (q1, q3)
7.40 (5.00, 11.35)
7.64 (3.33, 13.03)
7.06 (4.94, 10.49)
7.07 (5.76, 10.27)
10.39 (6.45, 16.62)
7.27 (5.68, 8.92)
TVOC (μg/𝑚3)
Mean (SD)
340.92 (340.40)
258.73 (168.78)
361.02 (281.28)
407.70 (467.68)
375.10 (429.19)
297.47 (249.85)
Median (q1, q3)
238.28 (130.10, 424.59)
230.09 (142.47, 311.09)
308.74 (149.74, 443.92)
222.38 (127.13, 512.48)
206.11 (112.55, 491.56)
233.86 (112.48, 381.27)
CO2 (ppm)
Mean (SD)
650.19 (272.35)
554.08 (220.67)
671.55 (187.44)
595.58 (298.00)
837.49 (326.78)
592.95 (219.79)
Median (q1, q3)
615.94 (448.91, 777.40)
495.21 (407.30, 694.08)
624.54 (541.39, 764.96)
472.40 (390.76, 720.93)
758.29 (653.14, 850.40)
521.15 (409.82, 739.18)
Temperature (°C)
Mean (SD)
26.80 (3.47)
29.57 (3.84)
26.23 (2.00)
26.71 (2.76)
23.49 (3.18)
28.05 (2.15)
Median (q1, q3)
26.99 (24.67, 29.08)
29.71 (27.99, 30.99)
26.34 (24.60, 28.02)
26.59 (24.49, 28.68)
24.28 (20.80, 25.51)
28.31 (26.73, 29.86)
Humidity (g/𝑚3)
Mean (SD)
40.83 (13.25)
50.23 (13.25)
38.14 (8.93)
43.32 (11.05)
26.60 (6.28)
45.86 (12.18)
Median (q1, q3)
38.89 (30.89, 49.94)
47.94 (45.07, 54.32)
35.60 (31.30)
43.38 (34.72, 51.14)
28.31 (21.45, 31.22)
45.75 (34.62, 53.35)
Respiration Rate
13.97 (3.01)
14.88 (3.62)
14.11 (1.31)
13.74 (0.92)
12.88 (1.11)
14.23 (5.29)
Heart rate
75.39 (9.94)
76.44 (12.96)
78.01 (8.48)
77.45 (8.95)
73.97 (9.88)
70.84 (7.29)
Heart rate variability
SDNN
Mean (SD)
138.96 (63.06)
118.16 (52.99)
122.78 (50.29)
146.49 (60.35)
144.57 (77.11)
163.24 (63.71)
Median (q1, q3)
129.80 (91.68, 171.69)
108.88 (80.13, 157.20)
111.62 (83.90, 154.50)
137.14 (107.87, 174.51)
139.00 (82.96, 192.99)
152.66 (122.06, 200.06)
SDANN
Mean (SD)
98.25 (39.36)
85.98 (38.82)
89.66 (32.33)
95.84 (33.70)
108.63 (54.19)
111.72 (28.29)
Median (q1, q3)
94.17 (72.32, 119.61)
85.37 (59.34, 103.34)
90.74 (67.95, 113.94)
94.97 (68.88, 120.50)
95.38 (75.11, 134.31)
108.31 (89.21, 132.54)
SDNNI
Mean (SD)
68.01 (41.11)
59.96 (45.23)
61.79 (38.07)
77.92 (54.42)
72.23 (31.47)
67.94 (29.93)
Median (q1, q3)
57.16 (38.43, 84.91)
38.70 (29.36, 83.79)
46.98 (37.19, 76.93)
59.73 (42.92, 99.65)
63.89 (53.89, 83.56)
61.18 (44.40, 93.84)
RMSSD
Mean (SD)
77.37 (66.07)
73.14 (66.05)
63.33 (61.03)
87.62 (82.46)
78.62 (59.55)
84.33 (58.08)
Median (q1, q3)
48.63 (26.16, 118.20)
41.93 (21.58, 113.91)
34.66 (21.12, 91.94)
42.14 (26.31, 151.10)
62.27 (28.82, 103.34)
77.29 (29.28, 122.35)
Data are presented by n (%) or means (sd). SBP, systolic blood pressure; DBP, diastolic blood pressure; METs, metabolic equivalent of task; TVOC, total volatile organic compounds; PM,
particulate matter; SDNN, standard deviation of NN intervals; SDANN, standard deviation of the average NN intervals for each 5 min segment of a 24 h HRV recording; SDNNI, mean of the
standard deviations of all the NN intervals for each 5 min segment of a 24 h HRV recording; RMSSD, root mean square of successive RR interval differences.
Table 2. Baseline characteristics of PM1.0 PM2.5, PM10, and HRV according to the cumulative concentration
Highest
cumulative concentration day
Lowest
cumulative concentration day
P-value
PM1.0
daily cumulative
15,879.10 (16,277.67)
9,509.20 (9,457.47)
0.001
daily average
11.33 (11.25)
8.45 (7.86)
0.040
PM2.5
daily cumulative
16,820.90 (17,249.52)
10,024.00 (9,990.31)
0.001
daily average
12.00 (11.92)
8.91 (8.29)
0.037
PM10
daily cumulative
16,944.00 (17,386.07)
10,089.60 (10,107.86)
0.001
daily average
12.09 (12.01)
8.96 (8.36)
0.037
TVOC
351.88 (315.41)
329.96 (364.99)
0.655
CO2
659.78 (282.25)
640.60 (263.19)
0.625
Temperature
26.89 (3.59)
26.71 (3.36)
0.717
Humidity
41.09 (13.01)
40.57 (13.54)
0.786
HRV
SDNN
137.60 (57.98)
140.30 (68.05)
0.773
SDANN
97.34 (37.31)
99.16 (41.48)
0.748
SDNNI
66.67 (39.14)
69.34 (43.14)
0.653
RMSSD
76.85 (63.52)
77.88 (68.85)
0.913
Data are presented by mean (sd). P-value was calculated by Mann-Whitney U test or independent t-test. PM,
particulate matter; TVOC, total volatile organic compounds; SDNN, standard deviation of NN intervals;
SDANN, standard deviation of the average NN intervals for each 5 min segment of a 24 h HRV recording;
SDNNI, mean of the standard deviations of all the NN intervals for each 5 min segment of a 24 h HRV recording;
RMSSD, root mean square of successive RR interval differences.
Table 3. Association between PM1.0 PM2.5, PM10, and HRV according to the living labs on days with highest cumulative concentration.
Living labs for environmental diseases
Living labs for vulnerable groups
Total
Patients with Arrhythmia
Patients with
Chronic Airway Disease
Patients with Stroke
Total
Residents of Industrial
Complex Area
Elderly
𝛽 (95% CI)
𝛽 (95% CI)
𝛽 (95% CI)
𝛽 (95% CI)
𝛽 (95% CI)
𝛽 (95% CI)
𝛽 (95% CI)
Outcome: SDNN
PM1.0
-0.62 (-1.87, 0.62)
0.23 (-3.36, 3.83)
-7.32 (-13.06, -1.58)
-0.51 (-2.85, 1.83)
-2.92 (-5.47, -0.36)
-3.50 (-6.74, -0.26)
-3.50 (-26.62, 20.62)
PM2.5
-0.60 (-1.78, 0.59)
0.21 (-3.21, 3.62)
-7.12 (-12.75, -1.50)
-0.48 (-2.72, 1.75)
-2.72 (-5.12, -0.33)
-3.27 (-6.27, -0.28)
-3.49 (-26.19, 19.20)
PM10
-0.59 (-1.77, 0.59)
0.22 (-3.18, 3.62)
-7.11 (-12.73, -1.49)
-0.48 (-2.70, 1.73)
-2.69 (-5.04, 0.33)
-3.18 (-6.13, -0.24)
-3.41 (-25.97, 19.15)
Outcome: SDANN
PM1.0
-0.18 (-1.04, 0.68)
-0.11 (-2.40, 2.18)
-4.60 (-9.08, -0.12)
-0.08 (-2.36, 2.20)
-1.33 (-2.84, 0.17)
-2.65 (-5.62, 0.32)
-1.50 (-12.54, 9.54)
PM2.5
-0.17 (-0.99, 0.64)
-0.11 (-2.29, 2.06)
-4.57 (-8.90, -0.24)
-0.08 (-2.25, 2.09)
-1.25 (-2.66, 0.16)
-2.47 (-5.23, 0.29)
-1.38 (-11.98, 9.31)
PM10
-0.17 (-0.99, 0.64)
-0.10 (-2.28, 2.06)
-4.55 (-8.88, -0.22)
-0.08 (-2.23, 2.07)
-1.23 (-2.61, 0.16)
-2.38 (-5.10, 0.33)
-1.34 (-11.96, 9.19)
Outcome: SDNNI
PM1.0
-0.71 (-1.82, 0.41)
0.01 (-3.18, 3.20)
-3.01 (-5.98, -0.05)
-1.09 (-2.84, 0.65)
-0.43 (-1.66, 0.79)
-0.78 (-2.44, 0.88)
-3.43 (-15.01, 8.15)
PM2.5
-0.69 (-1.75, 0.37)
-0.01 (-3.02, 3.02)
-2.97 (-5.85, -0.08)
-1.04 (-2.70, 0.61)
-0.39 (-1.54, 0.75)
-0.72 (-2.26, 0.83)
-3.40 (14.54, 7.73)
PM10
-0.68 (-1.73, 0.38)
0.01 (-3.01, 3.02)
-2.96 (-5.84, -0.08)
-1.04 (-2.68, 0.61)
-0.39 (-1.51, 0.74)
-0.69 (-2.20, 0.83)
-3.34 (-14.41, 7.73)
Outcome: RMSSD
PM1.0
-0.70 (-2.39, 0.99)
1.03 (-3.44, 5.51)
-6.24 (-11.70, -0.78)
-1.66 (-4.18, 0.85)
-0.78 (-3.26, 1.71)
-1.87 (-5.23, 1.49)
-6.48 (-31.42, 18.46)
PM2.5
-0.69 (-2.29, 0.92)
0.97 (-3.28, 5.22)
-6.09 (-11.42, -0.76)
-1.58 (-3.98, 0.82)
-0.70 (-3.03, 1.63)
-1.72 (-4.85, 1.41)
-6.40 (-30.40, 17.60)
PM10
-0.67 (-2.27, 0.93)
0.98 (-3.26, 5.22)
-6.08 (-11.40, -0.75)
-1.57 (-3.95, 0.81)
-0.69 (-2.98, 1.60)
-1.66 (-4.73, 1.41)
-6.29 (-30.16, 17.57)
Linear regressions were adjusted for age, sex, BMI, respiration rate, smoking, alcohol consumption, METs, hypertension, diabetes mellitus, TVOC, humidity, and temperature. Boldface means
the significant results. 𝛽, regression coefficient; CI, confidence interval; PM, particulate matter; SDNN, standard deviation of NN intervals; SDANN, standard deviation of the average NN
intervals for each 5 min segment of a 24 h HRV recording; SDNNI, mean of the standard deviations of all the NN intervals for each 5 min segment of a 24 h HRV recording; RMSSD, root
mean square of successive RR interval differences.
Figure 1. Association between PM1.0 PM2.5, PM10, and HRV.
Figure 2. The overall effects PM1.0 PM2.5, PM10, and HRV were estimated using the Bayesian Kernel
Machine Regression method; overall effect of the mixture (95% CI), defined as the difference in the
response when all of the exposures are fixed at a specific quantile (ranging from 0.10 to 0.90), as
compared to when all of the exposures are fixed at their median value. (A) The day with the highest
cumulative concentration of particulate matter; (B) The day with the lowest cumulative concentration
of particulate matter
Figure 3. The overall effects PM1.0 PM2.5, PM10, and SDNN were estimated using the Bayesian Kernel Machine Regression method according to the living
labs on day with highest cumulative concentration; overall effect of the mixture (95% CI), defined as the difference in the response when all of the exposures
are fixed at a specific quantile (ranging from 0.10 to 0.90), as compared to when all of the exposures are fixed at their median value.
Supplementary Table1. Baseline characteristics of environmental hazardous substances and HRV according to the cumulative concentration day and living labs.
Living labs for environmental diseases
Living labs for vulnerable groups
Patients with Arrhythmia
Patients with
Chronic Airway Disease
Patients with Stroke
Residents of Industrial
Complex Area
Elderly
Highest day
Lowest day
Highest day
Lowest day
Highest day
Lowest day
Highest day
Lowest day
Highest day
Lowest day
Measurement period
22.06~22.08
22.10~22.11
23.07~23.07
21.12~22.02
23.06~23.06
*Measurement place
(day)
Indoor
78.61%
80.22%
72.36%
72.49%
78.61%
78.91%
78.40%
83.39%
82.78%
80.95%
Transportation
5.19%
4.61%
5.89%
6.28%
3.76%
4.71%
3.30%
2.64%
3.19%
3.26%
Workplace
5.41%
3.96%
7.76%
6.64%
11.82%
9.87%
3.51%
0.44%
2.34%
5.18%
Outdoor
6.75%
5.45%
9.53%
9.71%
4.43%
3.41%
8.68%
6.16%
5.45%
3.57%
Etc
4.04%
5.76%
4.47%
4.87%
1.90%
3.10%
6.10%
7.37%
6.25%
7.04%
PM1.0
12.09
(14.49)
8.80 (7.37)
9.50 (5.24)
6.39 (3.15)
12.09
(15.26)
8.28 (10.61)
15.64
(11.97)
11.60
(10.93)
7.38 (2.46)
7.29 (3.05)
PM2.5
12.69
(15.30)
9.19 (7.80)
10.11 (5.50)
6.77 (3.33)
12.79
(16.05)
8.74 (11.15)
16.75
(12.87)
12.27
(11.57)
7.74 (2.51)
7.70 (3.21)
PM10
12.77
(15.29)
9.29 (7.83)
10.17 (5.55)
6.79 (3.33)
12.84
(16.15)
8.76 (11.19)
16.96
(13.13)
12.39
(11.76)
7.77 (2.52)
7.72 (3.22)
TVOC
260.39
(161.09)
257.08
(180.55)
399.17
(306.73)
322.88
(255.45)
402.97
(403.39)
412.41
(534.97)
355.05
(364.18)
395.15
(495.15)
336.65
(294.04)
258.29
(196.52)
Temperature
29.70 (3.91)
29.44 (3.87)
26.49 (1.89)
25.98 (2.13)
26.87 (2.79)
26.54 (2.81)
23.15 (3.57)
23.84 (2.80)
28.28 (1.85)
27.81 (2.44)
Humidity
50.40
(13.29)
50.06
(13.57)
38.74 (8.49)
37.53 (9.53)
42.66
(10.53)
43.97
(11.77)
26.84 (6.53)
26.36 (6.20)
46.83
(11.88)
44.89
(12.73)
HRV
SDNN
121.92
(53.05)
114.41
(54.11)
123.30
(49.37)
122.26
(52.47)
136.29
(50.32)
156.68
(68.74)
133.36
(63.38)
155.79
(89.11)
174.21
(62.67)
152.27
(64.51)
SDANN
86.29
(32.86)
85.66
(44.92)
88.92
(33.51)
90.39
(31.95)
98.09
(35.54)
93.59
(32.51)
97.90
(50.47)
119.37
(56.97)
115.90
(26.59)
107.55
(30.02)
SDNNI
63.15
(51.70)
56.78
(38.87)
57.93
(31.91)
65.65
(43.87)
70.55
(49.48)
85.28
(59.30)
68.96
(26.82)
75.49
(35.97)
73.03
(30.95)
62.84
(28.79)
RMSSD
79.73
(73.65)
66.55
(58.75)
61.09
(56.71)
65.57
(66.49)
75.74
(71.59)
99.49
(92.38)
77.02
(54.74)
80.21
(65.47)
91.54
(61.51)
77.12
(55.14)
* The values for each measurement location for each living lab are expressed as %, and represent the % for the measurement place during 24 hours.
PM, particulate matter; TVOC, total volatile organic compounds; SDNN, standard deviation of NN intervals; SDANN, standard dev iation of the average NN intervals for each 5 min
segment of a 24 h HRV recording; SDNNI, mean of the standard deviations of all the NN intervals for each 5 min segment of a 24 h HRV recording; RMSSD, root mean square of successive
RR interval differences.
Supplementary Table 2. Association between environmental hazardous substances and HRV according to the living labs on days with lowest cumulative concentration.
Living labs for environmental diseases
Living labs for vulnerable groups
Total
Patients with Arrhythmia
Patients with
Chronic Airway Disease
Patients with Stroke
Total
Residents of Industrial
Complex Area
Elderly
𝛽 (95% CI)
𝛽 (95% CI)
𝛽 (95% CI)
𝛽 (95% CI)
𝛽 (95% CI)
𝛽 (95% CI)
𝛽 (95% CI)
Outcome: SDNN
PM1.0
-0.46 (-2.84, 1.92)
-2.26 (-8.59, 4.07)
-3.27 (-17.27, 10.73)
0.92 (-1.58, 3.41)
-2.83 (-5.87, 0.21)
-2.62 (-6.70, 1.45)
-4.09 (-18.15, 9.98)
PM2.5
-0.46 (-2.73, 1.78)
-2.13 (-8.08, 3.82)
-3.23 (-16.74, 10.29)
0.87 (-1.50, 3.24)
-2.71 (-5.59, 0.16)
-2.49 (-6.33, 1.36)
-4.50 (-17.37, 8.37)
PM10
-0.45 (-2.70, 1.80)
-2.09 (-8.01, 3.82)
-3.28 (-16.86, 10.30)
0.86 (-1.50, 3.23)
-2.67 (-5.51, 0.16)
-2.45 (-6.25, 1.36)
-4.49 (-17.29, 8.31)
Outcome: SDANN
PM1.0
-0.55 (-1.93, 0.83)
-2.32 (-7.16, 2.51)
-5.25 (-11.28, 0.78)
0.16 (-1.36, 1.68)
-1.51 (-3.38, 0.35)
-1.97 (-4.05, 0.11)
-0.44 (-7.55, 7.)
PM2.5
-0.54 (-1.85, 0.77)
-2.19 (-6.74, 2.35)
-5.05 (-10.88, 0.79)
0.15 (-1.29, 1.59)
-1.46 (-3.23, 0.30)
-1.87 (-3.83, 0.10)
-0.43 (-7.54, 6.69)
PM10
-0.54 (-1.85, 0.76)
-2.17 (-6.68, 2.34)
-5.09 (-10.95, 0.78)
0.15 (-1.29, 1.58)
-1.44 (-3.18, 0.30)
-1.84 (-3.78, 0.10)
-0.42 (-7.50, 6.66)
Outcome: SDNNI
PM1.0
-0.78 (-2.61, 1.05)
-0.79 (-4.66, 3.07)
-0.63 (-14.07, 12.81)
0.08 (-2.07, 2.22)
0.28 (-1.09, 1.65)
0.53 (-1.56, 2.62)
-2.19 (-8.55, 4.18)
PM2.5
-0.76 (-2.49, 0.98)
-0.75 (-4.38, 2.89)
-0.85 (-13.82, 12.12)
0.07 (-1.96, 2.11)
0.25 (-1.05, 1.55)
0.50 (-1.47, 2.48)
-2.22 (-8.07, 3.62)
PM10
-0.72 (-2.45, 1.00)
-0.73 (-4.33, 2.88)
-0.91 (-13.95, 12.13)
0.08 (-1.95, 2.10)
0.25 (-1.03, 1.53)
0.49 (-1.46, 2.45)
-2.22 (-8.03, 3.60)
Outcome: RMSSD
PM1.0
-0.73 (-3.49, 2.03)
-0.21 (-6.34, 5.92)
-6.44 (-25.65, 12.76)
0.23 (-3.10, 3.55)
-0.02 (-2.51, 2.51)
0.64 (-3.62, 4.90)
-5.99 (-17.47, 5.50)
PM2.5
-0.73 (-3.34, 1.89)
-0.21 (-5.97, 5.56)
-5.98 (-24.58, 12.61)
0.22 (-2.94, 3.38)
-0.02 (-2.39, 2.36)
0.61 (-3.42, 4.63)
-5.74 (-16.31, 4.82)
PM10
-0.68 (-3.29, 1.93)
-0.18 (-5.91, 5.54)
-5.97 (-24.68, 12.73)
0.22 (-2.93, 3.37)
-0.02 (-2.36, 2.33)
0.60 (-3.37, 4.58)
-5.70 (-16.22, 4.81)
Linear regressions were adjusted for age, sex, BMI, respiration rate, smoking, alcohol consumption, METs, hypertension, diabetes mellitus, TVOC, humidity, and
temperature. 𝛽, regression coefficient; CI, confidence interval; PM, particulate matter; SDNN, standard deviation of NN intervals; SDANN, standard deviation of the average NN intervals
for each 5 min segment of a 24 h HRV recording; SDNNI, mean of the standard deviations of all the NN intervals for each 5 min segment of a 24 h HRV recording; RMSSD, root mean square
of successive RR interval differences.
Supplementary Figure 1. The exposure-response relationship for the associations between PM1.0
PM2.5, PM10, and SDNN in the day with the highest cumulative concentration of particulate matter,
was obtained from hierarchical Bayesian Kernel Machine Regression models. (A) The day with the
highest cumulative concentration of particulate matter; (B) The day with the lowest cumulative
concentration of particulate matter
Supplementary Figure 2. The overall effects PM1.0 PM2.5, PM10, and SDNN were estimated using
the Bayesian Kernel Machine Regression method according to the living labs on day with highest
cumulative concentration; overall effect of the mixture (95% CI), defined as the difference in the
response when all of the exposures are fixed at a specific quantile (ranging from 0.10 to 0.90), as
compared to when all of the exposures are fixed at their median value
