The association between air pollution and the daily hospital visits for atrial fibrillation recorded by ECG: a case-crossover study

AF, coding I48. × 01 in 10th edition of International Classification of Diseases (ICD-10), was recorded by ECG. Generally, the ECG changes of typical AF are characterized by irregularly irregular R–R intervals, absence of distinct repeating P waves and irregular atrial activation (ESC guidelines definition). Patients’ data were eligible if they were testified with ECG indicating AF from Jan 1st 2015 to Dec 31st 2018 in Shanghai Sixth People’s Hospital. This case-crossover study involving human participants was in accordance with the ethical standards of 1964 Helsinki Declaration and its later amendments or comparable ethical standards. The study was also approved by the ethics committee of Shanghai Sixth People’s Hospital. The study consisted of 4933 male and 5392 female patients. The study did not include minors (the youngest person with AF was 19 years old). Total of 10,863 AF cases recorded by ECG were obtained and 538 cases repeated by ECG in 48-h were excluded. Demographic data included name, gender, age, and we collected information of ECG (ST-T: ST-T changes; FVR: fast ventricular rate; SVR: slow ventricular rate; CB: conduction block; NSVT: nonpersistent supraventricular tachycardia; VPB: ventricular premature beat). All data were sorted in a date-based ascending order and were organized as gender and age group distribution. A single-centered database of AF with demographic data and data of ECG was then established.

The daily averaged concentrations in Shanghai during the study period for fine particle mass and gaseous air pollutant, including PM_2.5, PM₁₀, O₃, SO₂, NO₂ and CO, were obtained from Environmental Monitoring Center in Shanghai. To quantitatively describe short-term air quality conditions and trends in the study area, air quality index (AQI) was introduced, ever since AQI indicators were implanted in China in 2012. The referencing standards for AQI grading calculation are “Environmental Air Quality Standards” (GB3095-2012) and “AQI Technical Regulations (Trial)” (HJ633-2012). The AQI is calculated as follows:

AQI = max (IAQI1, IAQI2, IAQI3, …, IAQIn). In this formula, IAQIn refers to the air quality sub-index and n refers to each pollutant. The values of AQI in different intervals represent a specific air quality grade (Additional file 1: Table S1). When AQI is greater than 50, the pollutant with the highest IAQIn is defined as the primary pollutant; if the pollutant of the highest IAQIn is a tie between two or among more, they are then listed as primary pollutants. The monthly and quarterly mean air pollutant concentration was calculated based on daily mean concentration in Shanghai.

Chi-squared tests and two-sample t-tests were applied to compare the demographic distribution of the two groups when appropriate. The Pearson correlation coefficient was used to analyze the association between AF and each air pollutant. A total of 10,325 AF in Shanghai Sixth People’s Hospital, from 2015 to 2018 was used to fit the model after adjusting for meteorological variables under the case-crossover analysis with the same year–month time stratification. Since each case serves as his or her own control, individual-level confounding factors that remain constant over a short period of time (e.g., age, race, gender, socioeconomic status) are controlled. The date of hospital visits served as the case day (lag 0). The association between lag 0–3 air pollutant concentrations and daily hospital visits for AF recorded by ECG was presented as an odds ratio (OR) with a 95% CI, and was conducted mathematically with a conditional logistic regression model using R (version 4.1.1; R Development Core Team, Vienna, Austria) with the season package. In order to make sure the relationship of NO₂ and hospital visits for AF, conditional Logistic regression (R software "season" package) was used to analyze the influence of different lag pollutants on the number of daily AF, including PM_2.5, PM₁₀, O₃, SO₂, NO₂ and CO and adjusted for weather factors (mean atmospheric pressure, mean temperature, mean relative humidity, mean wind speed). The parameters of the “lag” are defined in Additional file 2: Table S2. The statistical tests were two-sided, and the effects of p < 0.05 or p < 0.01 were considered statistically significant.

The AF recorded by ECG and the corresponding daily air pollutants are shown in Fig. 1. Air pollutants repeated regularly, but the concentration of air pollutants decreased year by year from 2015 to 2018. The concentrations of PM_2.5, PM₁₀, SO₂, NO₂ and CO were highest in winter (data not shown). The summary of basic demographic statistics of the study population and air pollution is shown in Table1. The mean age of patients was 76.49 ± 11.46 years old with 83.00% participants over 65. The mean age of patients of different genders was 74.41 ± 12.18 years old in males and 78.39 ± 10.40 years old in females, respectively. The occurrence of AF was different in terms of gender and age, and most of the AF occurrence was in patients over 65 years old (data not shown). Comparison of monthly and quarterly hospital visits for AF recorded by ECG in this study could be seen in Additional file 3: Fig. S1A, B. The monthly hospital visits for AF recorded by ECG reached its peak in January (2015 and 2016) and February (2018). However, the peak time in 2017 was seen in April. Besides, we did observe a rebound of the number of AF in spring in any year. The monthly average quality of NO₂ ranged from 30 to 80 µg/m³ (91.7%). The monthly average quality of SO₂ ranged from 5 to 20 µg/m³ (91.7%). The monthly average quality of PM₁₀ ranged from 30 to 65 µg/m³ (95.8%). The monthly average quality of PM_2.5 ranged from 25 to 80 µg/m³ (91.7%). The monthly average quality of CO ranged from 12 to 27 µg/m³ (93.8%). The monthly average quality of O₃ ranged from 25 to 90 mg/m³ (91.7%). Most of the air pollutants reached their highest concentrations in the first and fourth quarters annually, negatively devastating air quality in the same period. On the other hand, O₃ reached its highest concentration in the summer season, probably due to higher temperature.

The air pollution and meteorological measurements are shown in Additional file 4: Table S3. Daily NO₂ was positively correlated with CO (Pearson correlation coefficient, r = 0.703) and PM_2.5 (r = 0.680). NO₂ was also positively correlated with SO₂ (r = 0.607) and PM₁₀ (r = 0.264). On the contrary, O₃ was negatively correlated with NO₂ (r = − 0.201) and the weather factors were also correlated with the air pollutant measurements.

For the monthly level, O₃ showed a negative correlation with AF-onset daily visits (r = − 0.38, p < 0.05) and a positive correlation was seen between daily AF recorded by ECG and each of the rest of the detected air pollutants (Fig. 2a–f) (PM_2.5 r = 0.45 p < 0.05, PM₁₀ r = 0.50 p < 0.05, O₃ r = − 0.38 p < 0.05; SO₂ r = 0.46 p < 0.05, NO₂ r = 0.54 p < 0.05, and CO r = 0.38 p < 0.01). Since NO₂ showed the highest coefficiency on lag effects, we further tested if the influence was different in gender and age subgroups (Fig. 3a, b). While the slope of the regression curve in female patients (k = 0.02635, p < 0.01) was slightly sharper than that in male patients (k = 0.02162, p < 0.01), the hysteresis effect still remained in both genders. However, when it came to the age group fitting test, the hysteresis effect could only be seen in patients over 65 years old (k = 0.04732, p < 0.01). All results above were much more ambiguous when comparing weekly mean values of corresponding IAQI and AF. None of the absolute value of each Pearson correlation coefficient was above 0.20 (not shown).

To elucidate whether the air pollution could induce AF with lag effects, we analyzed the association between each air pollutant and the number of AF patients with ECG on day lag 0–3, respectively (Fig. 4). In order to make sure the relationship of NO₂ and hospital visits for AF recorded by ECG, conditional logistic regression (R software "season" package) was used to analyze the influence of different lag pollutants on the number of daily AF recorded by ECG, including PM_2.5, PM₁₀, O₃, SO₂, NO₂ and CO. And adjusted for weather factors (mean atmospheric pressure, mean temperature, mean relative humidity, mean wind speed) OR (95% CI). The PM_2.5, PM₁₀ and SO₂ have no relationship with daily visits of AF recorded by ECG. But NO₂ increases the risk of daily visits of AF recorded by ECG and the maximum OR at lag 0, and the OR value was 1.0381 (95% CI 1.0135–1.0634). The O₃ reduces the risk of daily visits of AF recorded by ECG and the maximum RR at lag 2, and the OR value was 0.9869 (95% CI 0.9791–0.9948) and adjusted for weather factors, The CO reduces the risk of daily visits of AF recorded by ECG at lag 0, and the OR value was 0.9972 (95% CI 0.9949–0.9995) (Fig. 4a–f).

The most important finding of this study was that significant increase of daily hospital visits for AF patients with positive ECG was associated with increments in NO₂ concentration at lag 0 in Shanghai. Also, this effect was stronger within female patients and patients over 65.

Nowadays, air quality is one of the most popular issues to human life [13]. It has been reported that air pollution, low temperature, air humidity and air pressure changes could threaten certain population who are more vulnerable to cardiovascular diseases [1‐4, 6, 14‐19]. During this research, we proved that ambient environmental changes resulted in gender-related arrhythmia. We found that people over 65 years old had a higher daily hospital visits for AF recorded by ECG, which was consistent with previous studies [11]. In this research, we found that NO₂ enhanced daily hospital visits for AF through the correlation between ECG records and meteorological pollution data. Though PM₁₀ showed similar hysteresis effect, as it increased the risk of daily hospital visits of AF recorded by ECG, its devastating function was much weaker than NO₂ and might need further detection [6]. Interestingly, the hysteresis effect of PM_2.5 was unexpectedly weaker than that of PM₁₀. O₃ and CO needs more evidence to show their capabilities in inducing or inhibiting air pollution-related arrhythmia, because our results were opposite to the previous findings by Rich et al. [8], who demonstrated a positive relationship between O₃ and CO concentrations and AF.

NO₂ has a variety of both natural (volcano eruption, forest fire, etc.) and anthropogenic sources, derived from fossil fuel combustion (mainly originated from traffic and industrial consumption) [20]. Various studies have indirectly suggested short-term NO₂ exposure would lead to cardiovascular mortalities, myocardial infarction, heart failure, and arrhythmias. Hoek et al. showed a significant correlation between air pollution and the daily mortality caused by arrhythmias in Netherlands [21‐25]. Strong effect of NO₂ inducing arrhythmias, including AF and increasing heart failure, has been described in Western Europe [26]. In Brazil, Santos et al. carried out a time-series study and reported the significant influence of concurrent-day NO₂ concentration on the number of arrhythmia emergency room visits [27]. Zhao et al. also concluded that NO₂ had larger and more robust effects on outpatient visits for arrhythmia than other pollutants [19]. A previous study led by Monrad M. et al. claimed that though long-term residential traffic-related air pollution could be associated with higher risk of AF, no clear regarding effect modification was seen between exposure to nitrogen oxides(NO_x) and AF in the aspects of gender, hypertension or other risk factors [7]. However, in our study, gender and age were clear modifications for short-term NO₂ exposure when exploring its hysteresis effect. Possible explanation for this difference could lie in later univariate analysis against population cognition towards typical arrhythmias and personal hospital visits behaviors. Hence, the mechanisms through which air pollutants cause AF need further research.

To our knowledge, the biological patterns of how air pollutants induce arrhythmia were still not fully understood. Most researchers believed that autonomic nerve system, inflammatory reaction, and direct damage of the gaseous factors might play an important role in the occurrence of AF mutually. Some hypothesized that exposure to PMs firstly caused acute imbalance of autonomic modulation, characterized by increased sympathetic and decreased parasympathetic nerve function, which led to accelerated conduction of myocardial cells. Certain bypass of susceptible population was then activated and consequently AF occurred. In other studies, controversial function changes of the autonomic nervous system were reported when exposing to air pollutants, suggesting that people with different susceptibility background react to air pollutant exposure variously. For instance, exposure to PMs is more likely to induce AF in people with cardiorespiratory disease [28‐30]. Other pathways through which air pollutants lead to AF may include the direct adverse effect against cardiac electrophysiological pathways as well as decreased oxygen-carrying capacity of the blood [31, 32]. The latter may result in the constriction of vessels, especially coronary arteries, and ascend blood pressure at the same time or sequentially. Afterwards, damaged vessel endothelium malfunctions and oxidative stress injury occur, leading to arrhythmia [33]. Also, we observed a hysteretic effect on the AF occurrence after exposure to higher NO₂. Thus, we believe that a potential and chronic mechanism exists if acute exposure to NO₂ is confirmed to be associated with later AF by further research. Seaton et al. speculated that air pollutants, through activating a coagulation mechanism and elevating circulating inflammatory mediator emission, might result in a secondary pulmonary inflammation and a delayed but more severe cytokine storm [34]. The hysteretic consequence of this pathway was probably arrhythmia.

This study does have several limitations. Since our hospital is the only center to perform this study in Shanghai, the conclusion of this study could represent the prevalence pattern of the eastern coastal area of China still needs more verification. Although case-crossover study was used, we excluded patients with AF who follow-up regularly, so we may have missed some patients of acute-onset AF who refused ECG.