Abstract
Background
Acute aortic dissection (AAD) is a fatal cardiovascular emergency whose environmental triggers remain poorly understood. Although particulate air pollution has been implicated, the role of gaseous pollutants, such as carbon monoxide (CO), remains unclear. This study was performed to determine whether short‐term exposure to ambient CO and other air pollutants triggers the onset of AAD.
Methods
We conducted a time‐stratified case‐crossover study of 798 patients with emergency department–confirmed AAD in an industrial metropolitan area from 2018 to 2021. Individual exposures to fine particulate matter (aerodynamic diameter ≤2.5 μm), particulate matter (aerodynamic diameter ≤10 μm), CO, nitrogen dioxide, ozone, sulfur dioxide, temperature, humidity, and wind speed were assigned from 24 monitoring stations. Conditional logistic regression models evaluated associations across lag 0 to 3 days, adjusting for meteorological factors. Effects were assessed using mean and interquartile range increments.
Results
No air pollutants were associated with AAD on the event day. However, ambient CO exhibited a strong delayed effect: at a 2‐day lag, each 1‐ppm increase in CO was associated with a 2.67‐fold increase in AAD risk (95% CI, 1.207–5.900), and each interquartile range increase of 0.19 ppm increased risk by 113% (95% CI, 1.075–4.239). The other pollutants showed no significant associations at any lag. Apparent temperature effects in mean‐based models were not confirmed by interquartile range–based analysis, indicating no robust thermal trigger.
Conclusions
Ambient CO is a potent delayed trigger of acute aortic dissection, likely acting through subacute inflammatory and oxidative vascular injury rather than immediate hypoxia. These findings suggest targeted public health alerts and emission control for CO in industrialized cities.
Keywords: air pollution, aortic dissection, carbon monoxide, emergency medical services system, time‐stratified case‐crossover
Subject Categories: Cardiovascular Disease, Epidemiology
Nonstandard Abbreviations and Acronyms
- AAD
acute aortic dissection
- CO
carbon monoxide
- NO2
nitrogen dioxide
- PM10
particulate matter (aerodynamic diameter ≤10 μm)
- PM2.5
fine particulate matter (aerodynamic diameter ≤2.5 μm)
Clinical Perspectives.
What Is New?
A significant delayed effect observed at a 2‐day lag after higher carbon monoxide exposure is associated to acute aortic dissection with a 2.67‐fold increase in risk (95% CI, 1.207–5.900).
What Are the Clinical Implications?
Findings highlight carbon monoxide as a key trigger and policy target in tropic industrialized regions.
Acute aortic dissection (AAD) is one of the most catastrophic emergencies in cardiovascular medicine, with mortality increasing by 1% to 2% per hour during the first 48 hours after symptom onset. Despite advances in imaging and surgical management, early recognition remains challenging because many patients present with nonspecific symptoms, and known risk factors, such as hypertension, connective tissue disorders, and aortic aneurysm, do not fully explain the timing of acute onset. 1 Traditionally, AAD etiology has centered on intrinsic risk factors, such as chronic hypertension, heritable connective tissue disorders, and preexisting aortic abnormalities. More recently, research has expanded to include environmental triggers that may precipitate acute cardiovascular events. 2 , 3 Increasing evidence suggests that transient external triggers may precipitate dissection in vulnerable individuals, yet these triggers remain poorly defined.
Ambient air pollution has emerged as an important, modifiable risk factor for cardiovascular diseases, such as myocardial infarction, stroke, and heart failure. 4 Increasing evidence also implicates air pollutants in the onset of AAD, with several studies reporting a strong association between short‐term fine particulate matter (fine particulate matter [aerodynamic diameter ≤2.5 μm] [PM2.5] and particulate matter [aerodynamic diameter ≤10 μm] [PM10]) exposure and higher hospitalization rates for AAD. 5 These findings support the concept that acute environmental stressors may destabilize the aortic wall and precipitate dissection.
Although the association between PM2.5 and AAD is increasingly acknowledged, the specific role of carbon monoxide (CO) remains under investigation. Large‐scale cohort studies have found that elevated CO, along with PM10, nitrogen dioxide (NO2), and sulfur dioxide, is independently associated with increased long‐term mortality in patients with established AAD, 6 suggesting that CO may serve not only as a trigger for dissection onset but also as a prognostic factor. The proposed mechanism involves CO‐induced hypoxia and inflammation that weaken the aortic media, predisposing it to tearing or propagation of dissection.
Prior studies have also reported substantial geographic heterogeneity in pollutant–AAD associations. In cold northern cities, particulate matter and nitrogen dioxide appear to dominate risk, 7 , 8 whereas other regions demonstrate weaker or inconsistent particulate effects. 9 These findings suggest that local emission sources and atmospheric composition determine which pollutants exert the greatest cardiovascular toxicity. Industrial port cities are characterized by elevated CO emissions from heavy industry, shipping, and combustion, yet their cardiovascular consequences remain underexplored.
Accordingly, we conducted a time‐stratified case‐crossover study of all confirmed AAD by emergency departments in a heavily industrialized metropolitan area between 2018 and 2021. We evaluated short‐term exposure to major ambient air pollutants and meteorological variables across multiple lag periods to test the hypothesis that CO acts as a delayed environmental trigger of AAD.
METHODS
Patients
The study population included all patients who presented to designated emergency hospitals in Kaohsiung and were diagnosed with AAD between 2018 and 2021. Kaohsiung, the largest metropolis (population >2.7 million) in southern Taiwan, was chosen because of its substantial air pollution burden stemming from heavy industry and dense population. Eligible cases encompassed patients who arrived at the emergency department independently, via the Emergency Medical Services system, or through interhospital transfers.
For each patient, data collected included age, sex, residential address, hospital arrival time, mode of arrival, and duration of emergency department treatment. All deidentified data were obtained from the Emergency Medical Management System. The study protocol was reviewed and approved by the Institutional Review Board at Kaohsiung Chang Gung Memorial Hospital (Institutional Review Board No.: 202300351B0) in accordance with the ethical standards of the 1964 Declaration of Helsinki and its subsequent amendments. Because of no any individual identification, informed consent forms were waived.
Ambient Air Pollution Data
Ambient air pollution data were obtained from the network of air quality monitoring stations operated by the Taiwan Ministry of Environment. These stations provided hourly measurements of particulate matter with aerodynamic diameters ≤10 μm (PM10) and ≤2.5 μm (PM2.5), as well as CO, ozone, NO2, and sulfur dioxide. To estimate individual‐level exposure, pollutant concentrations were assigned to each participant's residential address using data from the nearest monitoring station (totally 24 stations in the Kaohsiung metropolis). The monitoring network provides extensive spatial coverage, with at least 1 station in every district of the Kaohsiung metropolitan area. Further details about the network are publicly available online (https://airtw.moenv.gov.tw/). The Taiwan Air Quality Monitoring Network is operated by the Taiwan Ministry of Environment. The network undergoes rigorous, continuous automatic quality assurance and quality control procedures strictly in accordance with the US Environmental Protection Agency guidelines (https://airtw.moenv.gov.tw/ ENG/Information/QualityAssurance/QAIntro.aspx). In our study, the overall proportion of missing or invalid hourly data (eg, equipment calibration or malfunction marked by the US Environmental Protection Agency) was extremely low at 1.77%. To ensure data integrity, a valid daily average required at least 18 hours (75%) of valid hourly measurements. Days not meeting this threshold (accounting for only 1.78% of the total study days) were treated as missing and excluded from the specific single‐pollutant models to prevent information bias. Our time‐stratified case‐crossover analysis used 24‐hour daily averages (calculated from 00:00 to 23:59) to construct the lag exposures. Specifically, lag 0 represents the 24‐hour daily average on the calendar day of the AAD event, lag 1 represents the daily average on the preceding calendar day, and so forth. In addition to the Taiwan Air Quality Monitoring Network data, anonymous patient data could apply to Emergency Medical Management System, Kaohsiung City, for uses via an application proceeding.
Statistical Analysis
To examine whether short‐term exposure to air pollutants increases the risk of AAD, we used a time‐stratified case‐crossover design. In this approach, each individual serves as his/her own control, comparing pollutant levels on the event day (case day or lag 0) with those on matched pollutant exposures from the preceding 3 days (lags 1–3), which were analyzed to assess potential delayed effects. This typically yielded 3 to 4 control days per case.
Conditional logistic regression models were used to estimate odds ratios (ORs) and 95% CIs for the association between pollutant exposure and AAD risk. Models were adjusted for ambient temperature and relative humidity. ORs were calculated for both per‐unit increases and interquartile range (IQR) increments in pollutant concentrations, including PM10, PM2.5, and major gaseous pollutants. Statistical significance was defined as a 2‐sided P<0.05. All analyses were conducted using SAS software, version 9.4, with an α level set at 0.05.
Because conditional logistic regression models do not provide direct collinearity diagnostics, and the potential for high intercorrelation among air pollutants, Pearson correlation coefficients were also examined. To minimize collinearity‐related bias, highly correlated pollutants (|ρ|>0.7) were not simultaneously included in multipollutant models.
RESULTS
Table 1 summarizes the demographic, clinical, and selected environmental characteristics of the 798 patients with AAD identified between 2018 and 2021. The study population exhibited a marked male predominance (68.3%), with a mean age of 67.2 (SD=13.9) years at presentation. For the mode of arrival to the emergency department, self‐transport was most common (50.1%), followed by ambulance transport (35.5%) and interhospital transfer (14.4%). Baseline environmental data at the time of AAD onset are also shown. Overall, 34.7% of events occurred on “warm days,” defined as days with a mean ambient temperature >25.9 °C.
Table 1.
Demographic Characteristics of Patients With AAD in Kaohsiung Metropolis Between 2018 and 2021
| Characteristic | All | 2018 | 2019 | 2020 | 2021 |
|---|---|---|---|---|---|
| No. | 798 | 196 | 228 | 193 | 181 |
| Male sex, n (%) | 544 (68.2) | 134 (68.4) | 151 (66.2) | 135 (69.9) | 124 (68.5) |
| Age, mean±SD, y | 67.2±13.9 | 68±14.2 | 67.4±13.5 | 66.5±14.4 | 66.8±13.7 |
| Aged ≥65 y, n (%) | 432 (54.1) | 104 (53.1) | 123 (53.9) | 103 (53.4) | 102 (56.4) |
| Arrival method, n (%) | |||||
| Self‐transport | 399 (50.1) | 100 (51) | 113 (49.6) | 100 (51.8) | 86 (47.5) |
| Ambulance | 283 (35.5) | 70 (35.7) | 81 (35.5) | 62 (32.1) | 70 (38.7) |
| Hospital transfer | 116 (14.4) | 26 (13.3) | 34 (14.9) | 31 (16.1) | 25 (13.8) |
| Onset on warm days (>25.9 °C), n (%) | 277 (34.7) | 69 (35.2) | 84 (36.8) | 54 (28.0) | 70 (38.7) |
AAD indicates acute aortic dissection.
Table 2 summarizes the meteorological conditions and air pollutant concentrations in southern Taiwan from 2018 to 2021. Among particulate pollutants, the mean PM2.5 concentration was 22.53 μg/m3 (SD=12.43 μg/m3; range, 3.72–86.25 μg/m3), whereas PM10 averaged 46.57 μg/m3 (SD=22.49 μg/m3; range, 8.04–141 μg/m3). Both pollutants exhibited substantial temporal variability, as reflected by wide IQRs of 19.42 for PM2.5 and 35.1 for PM10. For gaseous pollutants, the mean CO concentration was 0.40 (SD=0.14) ppm, NO2 averaged 3.03 (SD=1.54) ppb, ozone reached a relatively high mean of 30.1 (SD=11.55) ppb, and sulfur dioxide maintained a mean level of 3.0 (SD=0.8) ppb. Meteorological parameters showed an average ambient temperature of 25.9 (SD=3.8) °C with an IQR of 5.8 °C. Relative humidity averaged 74.84% (SD=7.04%; range, 41.4%–95.84%; IQR=7.87%), and the mean wind speed was 1.84 (SD=0.4) m/s.
Table 2.
Summary Statistics for Meteorological Factors and Air Pollution From 24 Monitor Stations in Kaohsiung Metropolis, Southern Taiwan, 2018 to 2021
| Variable | Minimum | Maximum | Mean±SD | 25th Percentile | 50th Percentile | 75th Percentile | IQR |
|---|---|---|---|---|---|---|---|
| PM2.5, μg/m3 | 3.72 | 86.25 | 22.53±12.43 | 11.0 | 23.07 | 30.42 | 19.42 |
| PM10, μg/m3 | 8.04 | 141.0 | 46.57±22.49 | 27.26 | 46.98 | 62.35 | 35.1 |
| CO, ppm | 0.13 | 0.88 | 0.4±0.14 | 0.3 | 0.4 | 0.49 | 0.19 |
| NO2, ppb | 0.8 | 16.92 | 3.03±1.54 | 2.06 | 2.67 | 3.62 | 1.57 |
| Ozone, ppb | 5.25 | 68.25 | 30.1±11.55 | 20.71 | 29.5 | 38.41 | 17.7 |
| SO2, ppb | 1.11 | 8.47 | 3±0.8 | 2.43 | 2.92 | 3.45 | 1.02 |
| Temperature, ° C | 11.05 | 31.8 | 25.9±3.8 | 23.27 | 26.63 | 29.06 | 5.8 |
| Humidity, % | 41.4 | 95.84 | 74.84±7.04 | 70.79 | 74.46 | 78.66 | 7.87 |
| Wind speed, m/s | 0.69 | 4.16 | 1.84±0.40 | 1.6 | 1.78 | 2 | 0.4 |
CO indicates carbon monoxide; IQR, interquartile range; NO2, nitrogen dioxide; PM2.5, fine particulate matter (aerodynamic diameter ≤2.5 μm); PM10, particulate matter (aerodynamic diameter ≤10 μm); and SO2, sulfur dioxide.
A Spearman rank correlation analysis was performed to assess the interrelationships among selected air pollutants and meteorological variables using daily average data during study period in Table 3. CO showed a strong positive correlation with NO2 (ρ=0.8719) and PM2.5 (ρ=0.8164). Likewise, a robust positive correlation was observed between PM2.5 and PM10 (ρ=0.8441). Ozone demonstrated a distinct pattern that was strongly and positively correlated with ambient temperature (ρ=0.5847), consistent with the role of solar radiation and heat in driving photochemical ozone production. In contrast, ozone exhibited a negative correlation with NO2 (ρ=−0.3703). Ambient temperature was inversely correlated with NO2 (ρ=−0.5238). Relative humidity was positively correlated with NO2 (ρ=0.4202) but was negatively correlated with both temperature and ozone.
Table 3.
Spearman Correlation Coefficients for Air Pollutants and Weather Conditions in Southern Taiwan, 2018 to 2021
| Variable | Ozone | PM2.5 | PM10 | NO2 | CO | Temperature | RH |
|---|---|---|---|---|---|---|---|
| Ozone | 1 | 0.1264 | 0.2801 | −0.3703 | −0.1983 | 0.5847 | −0.3243 |
| PM2.5 | 1 | 0.8441 | 0.7621 | 0.8164 | −0.2523 | 0.2031 | |
| PM10 | 1 | 0.7001 | 0.7818 | −0.0764 | 0.0519 | ||
| NO2 | 1 | 0.8719 | −0.5238 | 0.4202 | |||
| CO | 1 | −0.3533 | 0.315 | ||||
| Temperature | 1 | −0.4449 | |||||
| RH | 1 |
CO indicates carbon monoxide; NO2, nitrogen dioxide; PM2.5, fine particulate matter (aerodynamic diameter ≤2.5 μm); PM10, particulate matter (aerodynamic diameter ≤10 μm); and RH, relative humidity.
Because CO and NO2 are both gaseous pollutants with high correlation (ρ>0.8), in the next step, we did not put CO and NO2 simultaneously in multigaseous pollutant models because of conditional logistic regression analyses without direct collinearity diagnostics. We tested signal pollutant in the next tables.
Table 4 presents the results of the preliminary time‐stratified case‐crossover analysis assessing the association between short‐term exposure to individual ambient air pollutants and meteorological variables with the risk of AAD. The analysis evaluated exposure effects across multiple lag periods, from the day of the event (lag 0) to 3 days prior (lag 3), with each environmental factor examined in a separate single‐pollutant model. Overall, the results indicate no statistically significant associations between any of the examined environmental variables and the onset of AAD in this initial analysis.
Table 4.
Unadjusted ORs for the Association Between Lagged Environmental Factors and Emergency Visits for AAD by Conditional Logistic Regressions
| Variable | P value | OR | 95% CI Lower | 95% CI Upper | |
|---|---|---|---|---|---|
| Lag 0 | PM2.5 | 0.5448 | 0.997 | 0.989 | 1.006 |
| Lag 0 | PM10 | 0.9810 | 1.000 | 0.995 | 1.005 |
| Lag 0 | Ozone | 0.6515 | 0.999 | 0.994 | 1.003 |
| Lag 0 | NO2 | 0.6409 | 0.996 | 0.980 | 1.012 |
| Lag 0 | CO | 0.0753 | 0.581 | 0.320 | 1.057 |
| Lag 0 | Temperature | 0.1901 | 0.976 | 0.941 | 1.012 |
| Lag 0 | Humidity | 0.4555 | 0.997 | 0.989 | 1.005 |
| Lag 1 | PM2.5 | 0.7753 | 1.001 | 0.992 | 1.011 |
| Lag 1 | PM10 | 0.3270 | 0.997 | 0.992 | 1.003 |
| Lag 1 | Ozone | 0.2646 | 1.003 | 0.998 | 1.007 |
| Lag 1 | NO2 | 0.1380 | 0.988 | 0.971 | 1.004 |
| Lag 1 | CO | 0.3410 | 0.747 | 0.409 | 1.362 |
| Lag 1 | Temperature | 0.3576 | 1.018 | 0.980 | 1.056 |
| Lag 1 | Humidity | 0.5576 | 1.002 | 0.994 | 1.011 |
| Lag 2 | PM2.5 | 0.4962 | 0.997 | 0.988 | 1.006 |
| Lag 2 | PM10 | 0.4194 | 0.998 | 0.992 | 1.003 |
| Lag 2 | Ozone | 0.1304 | 1.003 | 0.999 | 1.008 |
| Lag 2 | NO2 | 0.4414 | 1.007 | 0.990 | 1.023 |
| Lag 2 | CO | 0.1375 | 1.614 | 0.858 | 3.036 |
| Lag 2 | Temperature | 0.6053 | 1.010 | 0.973 | 1.047 |
| Lag 2 | Humidity | 0.7590 | 0.999 | 0.990 | 1.007 |
| Lag 3 | PM2.5 | 0.7681 | 0.999 | 0.990 | 1.008 |
| Lag 3 | PM10 | 0.9748 | 1.000 | 0.994 | 1.006 |
| Lag 3 | Ozone | 0.1324 | 0.997 | 0.992 | 1.001 |
| Lag 3 | NO2 | 0.4538 | 1.006 | 0.990 | 1.024 |
| Lag 3 | CO | 0.1085 | 1.756 | 0.883 | 3.494 |
| Lag 3 | Temperature | 0.4524 | 0.986 | 0.950 | 1.023 |
| Lag 3 | Humidity | 0.3470 | 0.996 | 0.988 | 1.004 |
AAD indicates acute aortic dissection; CO, carbon monoxide; NO2, nitrogen dioxide; OR, odds ratio; PM2.5, fine particulate matter (aerodynamic diameter ≤2.5 μm); and PM10, particulate matter (aerodynamic diameter ≤10 μm).
Based on the time‐stratified case‐crossover design, conditional logistic regression was used to evaluate the association between short‐term environmental exposures and the risk of AAD across multiple lag periods (Table 5). The analysis found no statistically significant associations between any measured air pollutants and AAD occurrence on the event day (lag 0). A delayed effect, however, was identified for CO at a 2‐day lag (lag 2). Exposure to CO was significantly associated with an increased risk of AAD, with an OR of 2.669 (95% CI, 1.207–5.900).
Table 5.
Association Between Environmental Factors (Mean) and AAD Based on Conditional Logistic Regression Analysis
| Variable | P value | Odds ratio | 95% CI Lower | 95% CI Upper | |
|---|---|---|---|---|---|
| Lag 0 | CO | 0.7531 | 0.888 | 0.424 | 1.859 |
| PM2.5 | 0.6955 | 0.997 | 0.985 | 1.010 | |
| PM10 | 0.8527 | 1.001 | 0.994 | 1.008 | |
| Temperature | 0.0554 | 1.026 | 0.999 | 1.054 | |
| Humidity | 0.1759 | 1.006 | 0.997 | 1.014 | |
| Wind speed | 0.1563 | 1.093 | 0.967 | 1.236 | |
| Lag 1 | CO | 0.7161 | 1.135 | 0.574 | 2.246 |
| PM2.5 | 0.9717 | 1.000 | 0.988 | 1.013 | |
| PM10 | 0.9139 | 1.000 | 0.993 | 1.008 | |
| Temperature | 0.0156* | 1.033 | 1.006 | 1.061 | |
| Humidity | 0.6253 | 1.002 | 0.994 | 1.011 | |
| Wind speed | 0.3910 | 1.055 | 0.934 | 1.190 | |
| Lag 2 | CO | 0.0153* | 2.669 | 1.207 | 5.900 |
| PM2.5 | 0.1947 | 0.992 | 0.979 | 1.004 | |
| PM10 | 0.4701 | 0.997 | 0.990 | 1.004 | |
| Temperature | 0.0293* | 1.030 | 1.003 | 1.057 | |
| Humidity | 0.6381 | 0.998 | 0.990 | 1.006 | |
| Wind speed | 0.3984 | 0.953 | 0.852 | 1.066 | |
| Lag 3 | CO | 0.3517 | 1.460 | 0.658 | 3.240 |
| PM2.5 | 0.2837 | 0.993 | 0.980 | 1.006 | |
| PM10 | 0.8519 | 0.999 | 0.992 | 1.007 | |
| Temperature | 0.0251* | 1.031 | 1.004 | 1.058 | |
| Humidity | 0.5151 | 0.997 | 0.989 | 1.005 | |
| Wind speed | 0.7301 | 0.981 | 0.879 | 1.095 |
AAD indicates acute aortic dissection; CO, carbon monoxide; PM2.5, fine particulate matter (aerodynamic diameter ≤2.5 μm); and PM10, particulate matter (aerodynamic diameter ≤10 μm).
P<0.05.
Ambient temperature also demonstrated a consistent and statistically significant positive association with AAD risk at lag 1 (OR=1.033 [95% CI, 1.006–1.061]), lag 2 (OR=1.030 [95% CI, 1.003–1.057]), and lag 3 (OR=1.031 [95% CI, 1.004–1.058]), suggesting that higher temperatures during the 3 days preceding the event were linked to an increased likelihood of AAD onset.
The other gaseous pollutants were not significant to AAD. Although NO2 was highly correlative to CO, however, it was also nonsignificant to AAD. The result was located in Table S1.
To evaluate the effect of exposure variability, a separate conditional logistic regression analysis was performed using IQR increments for each environmental factor in Table 6. Consistent with the mean‐based model, this analysis identified a statistically significant association only for CO at a 2‐day lag (lag 2). Specifically, each IQR increase in ambient CO concentration was associated with 113% higher odds of AAD (OR=2.134 [95% CI, 1.075–4.239]).
Table 6.
Association Between Environmental Factors (IQR) and AAD Based on Conditional Logistic Regression Analysis
| Variable | P value | Odds ratio | 95% CI Lower | 95% CI Upper | |
|---|---|---|---|---|---|
| Lag 0 | CO | 0.1088 | 0.595 | 0.316 | 1.122 |
| PM2.5 | 0.4718 | 0.996 | 0.985 | 1.007 | |
| PM10 | 0.4717 | 1.002 | 0.996 | 1.009 | |
| Temperature | 0.3494 | 0.975 | 0.926 | 1.028 | |
| Humidity | 0.7586 | 1.002 | 0.990 | 1.014 | |
| Wind speed | 0.9167 | 1.007 | 0.891 | 1.137 | |
| Lag 1 | CO | 0.4712 | 0.792 | 0.420 | 1.494 |
| PM2.5 | 0.3565 | 1.005 | 0.994 | 1.017 | |
| PM10 | 0.2721 | 0.996 | 0.990 | 1.003 | |
| Temperature | 0.4712 | 1.019 | 0.968 | 1.072 | |
| Humidity | 0.9987 | 1.000 | 0.989 | 1.011 | |
| Wind speed | 0.9885 | 0.999 | 0.884 | 1.129 | |
| Lag 2 | CO | 0.0303* | 2.134 | 1.075 | 4.239 |
| PM2.5 | 0.4751 | 0.996 | 0.985 | 1.007 | |
| PM10 | 0.5782 | 0.998 | 0.991 | 1.005 | |
| Temperature | 0.4518 | 1.020 | 0.969 | 1.074 | |
| Humidity | 0.5708 | 0.997 | 0.985 | 1.008 | |
| Wind speed | 0.9771 | 0.998 | 0.882 | 1.130 | |
| Lag 3 | CO | 0.0877 | 1.869 | 0.912 | 3.830 |
| PM2.5 | 0.6256 | 0.997 | 0.986 | 1.008 | |
| PM10 | 0.9402 | 1.000 | 0.993 | 1.007 | |
| Temperature | 0.6863 | 0.989 | 0.939 | 1.042 | |
| Humidity | 0.7469 | 0.998 | 0.987 | 1.010 | |
| Wind speed | 0.8451 | 1.013 | 0.892 | 1.149 |
AAD indicates acute aortic dissection; CO, carbon monoxide; IQR, interquartile range; PM2.5, fine particulate matter (aerodynamic diameter ≤2.5 μm); and PM10, particulate matter (aerodynamic diameter ≤10 μm).
P<0.05.
DISCUSSION
This study provides novel evidence that short‐term ambient CO exposure is a potent delayed trigger of AAD. A robust association was observed 48 hours after exposure, with >2‐fold increase in AAD risk in both mean‐based and IQR analyses. In contrast, particulate matter (PM2.₅ and PM₁₀) and other gaseous pollutants showed no association with AAD at any lag. These findings extend prior work on air pollution and aortic disease, and identify CO as a previously underrecognized cardiovascular hazard. 5 , 7 , 9 These discrepancies from prior literature warrant careful and nuanced interpretation.
The 2‐day lag observed for CO provides important mechanistic insight. Because carboxyhemoglobin has a short half‐life under ambient air conditions, direct hypoxia is unlikely to explain an effect occurring 48 hours later. 10 Instead, CO induces vascular inflammation, oxidative stress, and endothelial dysfunction, processes that develop over hours to days and weaken the aortic media. 11 , 12 , 13 Neutrophil activation and reactive oxygen species generation can degrade extracellular matrix and impair vascular integrity, providing a biologically plausible pathway linking CO exposure to delayed aortic wall failure. 12 , 13 Building on these findings, future research must bridge this epidemiologic observation with direct mechanistic evidence. Clinically, this requires prospective cohort studies tracking vulnerable populations (eg, those with known aneurysms) that integrate high‐resolution personal exposure monitoring for CO with serial blood sampling. Tracking biomarkers of inflammation and oxidative stress (eg, hs‐CRP [high‐sensitivity C‐reactive protein], interleukin‐6, and myeloperoxidase) during and immediately following high‐pollution episodes will help quantify the human vascular response. Furthermore, mechanistic validation studies using established animal models of aortic dissection are crucial. By exposing susceptible models to controlled ambient CO levels, researchers can perform chronological histologic evaluations to map the temporal sequence of neutrophil infiltration, local reactive oxygen species generation, and matrix metalloproteinase–driven extracellular matrix degradation in the aortic wall over a 48‐hour period. 11 This translational approach will provide the necessary biological evidence to definitively support the subacute inflammatory pathways suggested by our epidemiologic associations.
Our findings contrast with studies from other regions that identified particulate matter as the dominant AAD trigger. For example, Shanghai and Harbin studies reported associations with PM2.5, PM10, NO2, and O₃, 5 , 8 whereas others observed weaker or inconsistent effects. 7 , 9 These discrepancies support a regional pollutant profile model, in which the dominant local emission sources determine which pollutants exert the greatest cardiovascular toxicity. Kaohsiung is heavily influenced by industrial combustion and shipping emissions, which produce high CO concentrations relative to particulate‐dominated traffic environments. This mechanistic pattern aligns closely with the observed temporal lag.
In contrast, cities with pollution dominated strictly by traffic exhaust or biomass combustion may experience different toxicological effects driven by distinct particulate or NO2 compositions. Furthermore, regional climatic conditions likely dictate the prominence of these triggers. In temperate or cold regions, abrupt temperature drops cause acute sympathetic activation and hypertensive spikes, frequently acting as the dominant mechanical trigger for AAD. In Kaohsiung's tropical climate, the absence of extreme cold thermal stress likely allows the subacute, 48‐hour inflammatory and oxidative impact of CO to emerge clearly without being overshadowed by temperature‐induced hemodynamic surges. Therefore, our findings highlight a specific regional phenotype of environmental AAD risk. To definitively test this “Regional Pollutant Profile” hypothesis, future research must involve large‐scale, multicity collaborations. By harmonizing case‐crossover methods across a matrix of cities with diverse dominant emission sources and varying climates, researchers can systematically map how the hierarchy of environmental triggers for AAD shifts globally.
The study period (2018–2021) encompassed the onset of the COVID‐19 pandemic in Taiwan beginning in 2020. However, the data do not indicate any significant confounding effect. Annual AAD case numbers in 2020 (n=193) and 2021 (n=181) were comparable to prepandemic levels in 2018 (n=196), suggesting no major disruption in disease incidence or emergency care–seeking behavior for this critical condition. Although global studies have documented pandemic‐related alterations in air quality, any such changes were insufficient to obscure the strong, consistent CO–AAD association observed across the 4‐year study period.
Temperature showed inconsistent associations in mean‐based models but not in IQR‐based analyses, indicating statistical instability rather than a robust physiological effect. 14 , 15 This contrasts with colder regions where temperature drops strongly trigger AAD via acute blood pressure elevation and vasoconstriction. 16 , 17 In tropical climates, temperature variation is smaller and unlikely to dominate risk. This inconsistency suggests that the apparent temperature effect is likely a statistical artifact or a reflection of Kaohsiung city's limited temperature variability.
This study also possesses several methodological strengths. The time‐stratified case‐crossover design inherently controls for time‐invariant confounders, such as genetic predisposition, smoking, and chronic comorbidities, by allowing each patient to serve as his/her own control. The focus on a specific, severe, and clinically well‐defined end point (emergency presentation for AAD) enhances diagnostic precision and minimizes outcome misclassification. Furthermore, the systematic evaluation of multiple lag periods was instrumental in revealing the delayed CO effect, a key contribution of this research.
Nevertheless, several limitations must be acknowledged. Exposure estimates were derived from fixed‐site monitoring stations, which do not account for individual mobility, indoor–outdoor activity, or microenvironmental factors (eg, in‐vehicle exposure), potentially introducing nondifferential exposure misclassification that likely biases result toward the null, implying that the true CO effect may be even stronger. Moreover, the absence of detailed clinical data (eg, hypertension severity, medication use, or aortic abnormalities) limited subgroup analyses. Uses of time‐stratified case‐crossover design to control some confounding effect would overcome the limitation of data source lacking of granular, reliable medical history. On the other hand, the intercorrelation of the gaseous air pollutants lacked testing of multiple pollutant effects, because of conditional logistic regression without direct collinearity diagnosis. The use of single‐pollutant models may also not fully capture the interactive effects of complex pollutant mixtures. Finally, the study's focus on 1 industrialized metropolitan area with a tropical climate limits generalizability to other settings.
Despite these limitations, this study holds important implications for clinical practice and public health. The observed 2‐day lag between elevated ambient CO and increased AAD risk is clinically actionable. In heavy industrial and tropic regions, clinicians, particularly those in emergency medicine and cardiology, should maintain heightened suspicion for AAD in patients presenting with characteristic symptoms (sudden tearing chest or back pain, syncope, or pulse deficits) within 2 days after public alerts for high CO levels. Operationally, this can be achieved by integrating local environmental data into emergency department information systems. For emergency physicians, this environmental context can serve as a vital adjunct to clinical decision rules like the aortic dissection detection risk score, lowering the threshold for immediate bedside echocardiography (point‐of‐care ultrasound) and fast‐tracking definitive diagnostic imaging (computed tomography angiography). Prompt diagnostic imaging may reduce the high early mortality of AAD, which can reach 1% to 2% per hour within the first 48 hours. 18
From a public health perspective, these findings highlight the need to broaden air quality management beyond PM2.5 to include gaseous pollutants such as CO. Health advisories during high‐pollution episodes should explicitly warn of CO's acute cardiovascular risks, especially among vulnerable populations, including elderly individuals and individuals with hypertension, connective tissue disorders, or preexisting aortic disease. Regulatory strategies aimed at reducing CO emissions from both industrial and vehicular sources could serve as a primary prevention measure against fatal cardiovascular events. 19
Interpretation
Clinically, the 48‐hour delayed CO effect has direct implications. AAD mortality increases by 1% to 2% per hour in the first 2 days, and heightened diagnostic vigilance following high‐CO pollution episodes may improve early detection. From a public health perspective, these findings argue for expanding air‐quality advisories beyond particulate matter to include carbon monoxide as a cardiovascular risk indicator.
Sources of Funding
The study was supported by funds from the Taiwan National Science and Technology Council, NSTC 114‐2314‐B‐037‐060, and Kaohsiung Medical University Hospital (KMUH114‐4R85). This work was also supported partially by the Research Center for Precision Environmental Medicine, Kaohsiung Medical University, Kaohsiung, Taiwan, from the Featured Areas Research Center Program within the framework of the Higher Education Sprout Project by the Ministry of Education in Taiwan and by a Kaohsiung Medical University Research Center Grant (KMU‐TC114A01).
Disclosures
All authors have no conflicts of interest.
Supporting information
Table S1
Data S1
Acknowledgments
This is a short text to acknowledge the contributions of our colleagues, institutions, and Kaohsiung Emergency Medical Management System that aided the efforts of the authors.
Conceptualization, C.‐Y.C. and H.‐Y.C.; methods, F.‐W.L. and H.‐Y.C.; validation, F.‐W.L. and H.‐Y.C.; formal analysis, C.‐Y.C.; investigation, C.‐Y.C., F.‐W.L., and H.‐Y.C.; resources, H.‐Y.C.; data curation, C.‐Y.C. and H.‐Y.C.; writing (original draft preparation), C.‐Y.C.; writing (review and editing), F.‐W.L. and H.‐Y.C.; supervision, F.‐W.L. and H.‐Y.C.; project administration, H.‐Y.C.; funding acquisition, H.‐Y.C. All authors have read and agreed to the published version of the manuscript.
Supplemental Material is available at https://www.ahajournals.org/doi/suppl/10.1161/JAHA.126.049936
For Disclosures and Sources of Funding, see page 8.
This article was sent to William W. Aitken, MD, Associate Editor, for review by expert referees, editorial decision, and final disposition.
References
- 1. Nienaber CA, Clough RE. Management of acute aortic dissection. Lancet. 2015;385:800–811. doi: 10.1016/s0140-6736(14)61005-9 [DOI] [PubMed] [Google Scholar]
- 2. Reed MJ. Diagnosis and management of acute aortic dissection in the emergency department. Br J Hosp Med (Lond). 2024;85:1–9. doi: 10.12968/hmed.2023.0366 [DOI] [PubMed] [Google Scholar]
- 3. Rylski B, Schilling O, Czerny M. Acute aortic dissection: evidence, uncertainties, and future therapies. Eur Heart J. 2023;44:813–821. doi: 10.1093/eurheartj/ehac757 [DOI] [PubMed] [Google Scholar]
- 4. Brook RD, Rajagopalan S, Pope CA 3rd, Brook JR, Bhatnagar A, Diez‐Roux AV, Holguin F, Hong Y, Luepker RV, Mittleman MA, et al. Particulate matter air pollution and cardiovascular disease: an update to the scientific statement from the American Heart Association. Circulation. 2010;121:2331–2378. doi: 10.1161/CIR.0b013e3181dbece1 [DOI] [PubMed] [Google Scholar]
- 5. Chen J, Lv M, Yao W, Chen R, Lai H, Tong C, Fu W, Zhang W, Wang C. Association between fine particulate matter air pollution and acute aortic dissections: a time‐series study in Shanghai, China. Chemosphere. 2020;243:125357. doi: 10.1016/j.chemosphere.2019.125357 [DOI] [PubMed] [Google Scholar]
- 6. Lin M, Yan J, Tang J, Han S, Guo P, Wu S, Tao L, Xiao H, Chen Y, Tan X. Air pollutants and mortality risk in patients with aortic dissection: evidence from a clinical cohort, single‐cell sequencing, and proteomics. Environ Sci Technol. 2024;58:6509–6518. doi: 10.1021/acs.est.4c00534 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Xie N, Zou L, Ye L. The effect of meteorological conditions and air pollution on the occurrence of type a and B acute aortic dissections. Int J Biometeorol. 2018;62:1607–1613. doi: 10.1007/s00484-018-1560-0 [DOI] [PubMed] [Google Scholar]
- 8. Zhang H, Yin L, Zhang Y, Qiu Z, Peng S, Wang Z, Sun B, Ding J, Liu J, Du K, et al. Short‐term effects of air pollution and weather changes on the occurrence of acute aortic dissection in a cold region. Front Public Health. 2023;11:1172532. doi: 10.3389/fpubh.2023.1172532 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Wang Q, Huang W, Kou B. Examining the relationships between air pollutants and the incidence of acute aortic dissection with electronic medical data in a moderately polluted area of Northwest China. Inquiry. 2021;58:469580211065691. doi: 10.1177/00469580211065691 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Allred EN, Bleecker ER, Chaitman BR, Dahms TE, Gottlieb SO, Hackney JD, Pagano M, Selvester RH, Walden SM, Warren J. Short‐term effects of carbon monoxide exposure on the exercise performance of subjects with coronary artery disease. N Engl J Med. 1989;321:1426–1432. doi: 10.1056/nejm198911233212102 [DOI] [PubMed] [Google Scholar]
- 11. Chu LM, Shaefi S, Byrne JD, Alves de Souza RW, Otterbein LE. Carbon monoxide and a change of heart. Redox Biol. 2021;48:102183. doi: 10.1016/j.redox.2021.102183 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Thom SR, Bhopale VM, Han ST, Clark JM, Hardy KR. Intravascular neutrophil activation due to carbon monoxide poisoning. Am J Respir Crit Care Med. 2006;174:1239–1248. doi: 10.1164/rccm.200604-557OC [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Akyol S, Erdogan S, Idiz N, Celik S, Kaya M, Ucar F, Dane S, Akyol O. The role of reactive oxygen species and oxidative stress in carbon monoxide toxicity: an in‐depth analysis. Redox Rep. 2014;19:180–189. doi: 10.1179/1351000214y.0000000094 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Oudin Åström D, Bjursten H, Oudin A, Nozohoor S, Ahmad K, Tang M, Bjurbom M, Hansson EC, Jeppsson A, Holdflod Møller CJ, et al. Temperature effects on incidence of surgery for acute type a aortic dissection in the Nordics. Glob Health Action. 2022;15:2139340. doi: 10.1080/16549716.2022.2139340 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Henzi D, Platzmann A, Brtek J, Holubec T, Emmert MY, Vogt P, Mestres CA, Reser D. Increasing atmospheric temperature implicates increasing risk for acute type a dissection in hypertensive patients. J Thorac Dis. 2021;13:5799–5806. doi: 10.21037/jtd-21-824 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Kurz SD, Mahlke H, Graw K, Prasse P, Falk V, Knosalla C, Matzarakis A. Patterns in acute aortic dissection and a connection to meteorological conditions in Germany. PLoS One. 2024;19:e0296794. doi: 10.1371/journal.pone.0296794 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Zhang Q, Peng L, Hu J, Li H, Jiang Y, Fang W, Yan H, Chen J, Wang W, Xiang D, et al. Low temperature and temperature decline increase acute aortic dissection risk and burden: a nationwide case crossover analysis at hourly level among 40,270 patients. Lancet Reg Health West Pac. 2022;28:100562. doi: 10.1016/j.lanwpc.2022.100562 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Wang FX, Yuan QR, Cui J, Ke XQ, Wang FF, Dong XY, Wang T, Hahaerman A, Tong YF, Xu ZX, et al. The relationships of air pollution and meteorological factors with acute aortic dissection in Urumqi, China: a distributed lag nonlinear model analysis. J Multidiscip Healthc. 2025;18:3331–3343. doi: 10.2147/jmdh.S509228 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Afzal M, Agarwal S, Elshaikh RH, Babker AMA, Choudhary RK, Prabhakar PK, Zahir F, Sah AK. Carbon monoxide poisoning: diagnosis, prognostic factors, treatment strategies, and future perspectives. Diagnostics (Basel). 2025;15:581. doi: 10.3390/diagnostics15050581 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1
Data S1