Air Pollution and Ischemic Stroke Among Young Adults
Background and Purpose—Studies have demonstrated consistent associations between cardiovascular illness and particulate matter (PM) <10 and <2.5 μm in diameter, but stroke received less attention. We hypothesized that air pollution, an inflammation progenitor, can be associated with stroke incidence in young patients in whom the usual risk factors for stroke are less prevalent. We aimed to evaluate the association between stroke incidence and exposure to PM <10 and <2.5 μm, in a desert area characterized by a wide range of PM.
Methods—We included all members of the largest health maintenance organization in Israel, who were admitted to a local hospital with stroke between 2005 and 2012. Exposure assessment was based on a hybrid model incorporating daily satellite remote sensing data at 1-km spatial resolution. We performed case-crossover analysis, stratified by personal characteristics and distance from main roads.
Results—We identified 4837 stroke cases (89.4% ischemic stroke). Interquartile range of PM <10 and <2.5 μm was 36.3 to 54.7 and 16.7 to 23.3 μg/m3, respectively. The subjects’ average age was 70 years; 53.4% were males. Associations between ischemic stroke and increases of interquartile range average concentrations of particulate matter <10 or <2.5 μm at the day of the event were observed among subjects <55 years (odds ratio [95% confidence interval], 1.11 [1.02–1.20] and 1.10 [1.00–1.21]). Stronger associations were observed in subjects living within 75 m from a main road (1.22 [1.03–1.43] and 1.26 [1.04–1.51]).
Conclusions—We observed higher risk for ischemic stroke associated with PM among young adults. This finding can be explained by the inflammatory mechanism, linking air pollution and stroke.
Epidemiological studies have demonstrated a consistent increased risk for cardiovascular morbidity and mortality after exposure to air pollution.1–4 Stroke, a common cause of death and severe chronic disability,5 has received less attention.6 Of the studies that examined the association between particulate matter <10 μm in diameter (PM10) or <2.5 μm in diameter (PM2.5) and stroke, several have found positive associations5,7–13 predominantly with incidence of the ischemic stroke.3,6,8,10,13 Other studies did not find associations between PM and stroke risk.14,15 A major limitation of most studies is the exposure assessment that was based mostly on monitoring stations, a method that could have potentially lead to some inaccuracy and biased results.16,17
We have recently presented a new method of assessing spatiotemporal resolved PM2.5 and PM10 exposures18,19 in the Negev (South Israel)— a semi-arid area, located in the global dust belt, which extends from West Africa to the Arabian Desert.20 Unlike many commonly used exposure models, our model makes use of satellite aerosol optical depth measurements, which allowed us to estimate daily, spatially resolved, PM10 and PM2.5 concentration with the inclusion of populations in rural areas not living nearby monitoring stations.
Stroke in older patients is mostly related to atherosclerosis.21 In the young, however, stroke is less common, and the cause is much more diverse and often unexplained.21 We hypothesize that PM exposure is positively associated with stroke, especially among young adults.
We included all Clalit Health Services members with available geocoded addresses in Southern Israel, admitted to Soroka University Medical Center (SUMC) with hemorrhagic (International Classification of Diseases Ninth Revision [ICD9] 430–432) or ischemic stroke (ICD9 432–435) between the years 2005 and 2012. SUMC is the only medical center providing acute neurological care in the area, serving a population of ≈1 million residents; therefore, all the patients with stroke are referred to SUMC. Clalit Health Service is the largest health care provider in the area, covering ≈70% of the population. To avoid within subject correlation among recurrent events, we used only the first stroke event, within the study period, for each subject.22
Clinical and Demographic Characteristics
Demographic, clinical, laboratory, and medication prescription data were available for all subjects. Clalit Health Service uses electronic medical records system with subjects identified by the national ID, and thus, the following data were fully available: age, sex, ethnicity, comorbidities, body mass index, smoking status, medications purchase, and socioeconomic status. Socioeconomic status was defined based on home addresses and according to the definitions of the Israel Central Bureau of Statistics assigning socioeconomic status level in a scale of 1 to 10, based on the population residing in the same statistical zone.23
Air Pollution and Meteorologic Data
Exposure assessment was based on a hybrid satellite-based model incorporating daily satellite remote sensing data at spatial resolution of 1×1 km.19 Briefly, we make use of a new algorithm developed by National Aeronautics and Space Administration –Multi-Angle Implementation to Atmospheric Correction,24 which provides aerosol optical depth data in a high resolution. Using mixed models, we regressed daily PM10 and PM2.5 mass concentration from the Ministry of Environmental Protection against aerosol optical depth, traditional land use regression temporal predictors (obtained from the Technion Center of Excellence in Exposure Science and Environmental Health air pollution monitoring database) and spatial predictors (obtained through the Israeli Central Bureau of Statistics and Survey Bureau mapping service). When aerosol optical depth was not available, we fitted a generalized additive model with a thin plate spline term of latitude and longitude to interpolate the estimates. Good model performance was achieved, with out-of-sample cross validation R2 of 0.79 and 0.72 for PM10 and PM2.5, respectively. Both model predictions had little bias, with cross-validated slopes (predicted versus observed) of 0.99. More in-depth details can be found in Kloog et al.19 The daily average concentrations of the pollutants, estimated throughout the studied region, were assigned for each patient based on proximity to his geocoded home address.
Daily data on air temperature and relative humidity for the study period were obtained from a central monitoring site located in the center of the largest city of the region.
Results are presented by mean±SD and interquartile range (IQR) continuous variables and as percentages for categorical data. Continuous variables were compared using independent t test or ANOVA. Mann–Whitney test or Kruskal Wallis test were used if parametric assumptions could not be satisfied. Pearson χ2 test was used to compare categorical variables.
Air Pollution Exposure
We examined the association between stroke and PM10 or PM2.5 using a case-crossover design25; in which we sample only cases and compared the subject’s exposure in the case day to that subject’s exposure in other time periods. Because each subject serves as their own control, all measured or unmeasured characteristics that do not vary over time are matched by design and therefore cannot confound the association with air pollution. To avoid confounding by seasonal patterns, we used Navidi bidirectional approach and matched control days on day of the week from the same month of the same year.26 Case period was defined as the hospitalization day. One- to 4-day legged exposures were tested as well. Associations with PM10 and PM2.5 were tested separately and were adjusted for daily average temperature and relative humidity. PM values higher than the 99th percentile were considered as outliers and were excluded from the analyses.
In a stratified analysis by age, sex, comorbidities, medications use, seasons, smoking, and distance from main roads, we assessed whether the effect was modified by specific diagnoses and personal characteristics. In the latter, patients were grouped by the quartiles of distance from main roads (in meters). We have a priori defined young adults as subjects <55 years.27 Others were grouped into 2 age groups (55–65 and >65 years).
We further explored the associations with PM in groups in which we observed significant associations with PM, using penalized splines to allow the detection of nonlinear associations with the pollutants. Odds ratio (OR) and 95% confidence intervals (CI) were then obtained from the regression predicted values and were calculated for IQR increases in selected concentrations across the range of the pollutants.
These data were analyzed using a conditional logistic regression (coxph, R statistical software, version 3.1.1). Results are presented as OR and 95% CI for an IQR increase of the pollutants daily average concentration.
We identified 4837 subjects admitted to SUMC with stroke between the years 2005 and 2012, of which 4325 were cases of ischemic stroke. Figure 1 shows the study region. Half of the subjects were males, and the mean age was 70 years. Diagnoses of ischemic stroke and history of cardiovascular diseases were less frequent in younger patients. The purchase of aspirin, lipid-modifying agents, and antidiabetic medications were similar in all groups (Table 1).
Pollution and Meteorology
The IQR of PM10 ranged between 36.3 and 54.7 μg/m3, reaching a maximal value of 235.2 μg/m3. The IQR of PM2.5 ranged between 16.7 and 23.3 μg/m3, reaching a maximal value of 75 mg/m3. Daily PM2.5 and PM10 concentrations were higher than the threshold, defined by the World Health Organization guidelines, in 19% and 35% of the days, respectively.28
The climate in the study region is relatively hot and dry, reaching a maximal temperature of 33.5°C in the summer (Table 2).
Short-Term Exposure Effect
Using a case-crossover design, we examined the association between stroke and short-term exposure to PM10 and PM2.5. Positive significant associations between ischemic stroke and PM10 or PM2.5 concentrations in the day of the event were observed among subjects <55 years (OR [95% CI], 1.11 [1.02–1.20] and 1.10 [1.00–1.21], respectively).
No associations were found with hemorrhagic stroke (1.05 [0.96–1.14] and 1.04 [0.93–1.15] for an IQR elevation in PM10 and PM2.5, respectively) or all available cases of ischemic stroke (Table 3).
In a stratified analysis by season, associations with PM10 and PM2.5 were observed only among subjects admitted in the spring season (1.08 [1.01–1.15] and 1.077 [0.99–1.15], respectively). Sex, comorbidities, socioeconomic status, smoking, ethnicity, and medications use did not modify the associations.
Exposure Range and Distance From Main Roads
Among subjects <55 years, we found stronger associations with ischemic stroke within the lower ranges of PM. For instance, an increase of IQR concentration of PM10 from 30 to 48 μg/m3 was associated with an OR of 1.16 (95% CI: 1.09–1.23), compared with an OR of 1.10 (95% CI: 1.01–1.20) associated with IQR elevation of PM10 between 160 and 178 μg/m3. Similar associations were observed with PM2.5 (Fig. 2).
Of the young subjects (<55 years), 25% lived <75 m from a main road, 50% lived 150 m from a main road, and 75% lived 250 m from a main road. When stratifying these cases by the quartiles of distance from the nearest main road, we found higher and significant risk for ischemic stroke only among subjects living within 75 m from a main road (OR [95% CI], 1.22 [1.03–1.43] and 1.26 [1.04–1.51] for IQR increase of PM10 and PM2.5, respectively). The risk estimates in all other distance groups were similar and were, therefore, combined into 1 group: no associations were observed among subjects living in further locations (OR [95% CI], 1.07 [0.97–1.18] and 1.04 [0.93–1.17] for IQR increase of PM10 and PM2.5, respectively; Table I in the online-only Data Supplement).
No associations were found with PM10 or PM2.5 concentrations 1 to 4 days before the hospitalization day.
In this large-scale case-crossover study, we found higher risk of ischemic stroke associated with daily average PM10 and PM2.5 concentration on the day of the event, among young subjects. The associations were more pronounced in the lower range of the pollutants, where particles most likely originate from traffic,29 and among subjects who reside in proximity to main roads.
Inflammation is a known casual factor for several central nervous system diseases.30 Among the environmental exposures that may induce inflammation and oxidative stress leading to central nervous system pathology, air pollution may be considered as the most prevalent source.31 The 2 possible pathways in which PM can affect neurocognitive function are (1) oxidative stress induced by respiratory intake of PM and the penetration of the inflammatory compounds into the blood–brain barrier32; (2) direct entry of PM into the central nervous system through the olfactory bulb.33
Stroke in the Young
Unlike other studies that found higher risks for air pollution–related stroke in older patients,5,13 we observed significant increased risk among subjects <55 years and found no associations in other age groups.
Similarly, Ruidavets et al34 found that acute myocardial infarction in patients with no vascular risk factors may be related to air pollution. Their suggested explanation for this finding was reactive increased vascular resistance in patients who are not chronically treated with statins and vasodilators (as the older population with chronic vascular risk factors).34 Another possible explanation is that atherosclerosis makes the vessels less reactive,35 which actually protects the vessels air pollution effects in older patients.
In younger adults, atherosclerosis remains an important risk factor, but cardiometabolic cause is more common.36 Among other components, the cardiometabolic syndrome can be associated with hypercoagulability and mediated by inflammatory response and interferences in the coagulation process, which may contribute to increased risk for cardiovascular events.37
Atrial fibrillation, a common cause for cerebral emboli38 is believed to be under diagnosed in patients with stroke of undetermined cause. Presentation of this common arrhythmia was found39 to be associated with exposure to air pollution in young adults. This yet unexplained association may contribute to the association between stroke and air pollutants in young adults.
Exposure to PM was associated with elevated levels of C-reactive protein, fibrinogen, white blood cells, and several inflammatory factors.4 Considering the underling mechanism in which air pollution may be related to ischemic stroke is based on inflammatory response, it is possible that air pollution–induced stroke is more pronounced in younger patients, where cardiometabolic-related stroke is more common.
Similar to previous studies, PM exposure in our study was associated with ischemic but not with hemorrhagic stroke.3,6,8,10,13 Oudin et al,8 for instance, found a 13% increase in hospital admissions for ischemic stroke associated with PM10 30 μg/m3 compared with PM10 <15 μg/m3; and Wellenius et al13 found an estimated OR of 1.11 (95% CI: 1.03–1.20) for ischemic stroke comparing the 25th and 75th percentiles of the daily average PM2.5 concentration.
Although we observed no association with PM in all available ischemic stroke cases, the associations observed among specific subgroups were higher than most association reported to date.3,5,7,13 Commonly used methods for exposure assessment in many of the studies rely on ground monitoring data and are therefore prone to exposure misclassification, which may dilute air pollution–associated health effects.16 Using our PM prediction models, we were able to include patients reside both in urban and in rural places and to obtain spatially resolved PM estimates, which allowed us to reduce this misclassification.
Traffic- Versus Natural-Related Air Pollution
Given the natural and anthropogenic sources of PM in the Negev area, the range of PM is wide, with different compositions that characterize the pollutants.40 Krasnov et al29 assessed the constant anthropogenic contribution of PM10 as an average value of 34 μg/m3. Values >71 μg/m3 were found to be associated with dust events.29 In its lower range, PM originates mostly from traffic emissions.41 Higher levels of PM, recorded during dust events, are mostly of natural origin and have different mineral and chemical compositions.29,40
Studies that evaluated the toxicity of PM by its source, size, and composition have found particles with industrial and traffic components to be highly toxic.42,43 Although we did not examine the speciation of PM, we did find stronger associations with PM within its lower range and among subjects who reside in proximity to main roads, emphasizing the role of PM composition in addition to its mass.
Our study had several limitations. First, we did not have information on the exact onset of stroke symptoms, which may have caused misclassification in the definition of a case day. However, because stroke events require urgent care and hospitalizations and because SUMC is the only available hospital in the region, we think that the misclassification will be minor. Second, we were not able to adjust the effect for traffic noise and gaseous air pollutants in our models and therefore cannot rule out the contribution of these factors to the higher stroke risks found among subjects residing in proximity to major roads.44 In addition, we did not have information about smoking history and therefore could not fully assess the modification by smoking status. Finally, the use of 1×1 km model could also be improved as higher resolution satellite data become available. As satellite remote sensing evolves and progresses, finer resolution data should enable us to assess more precise estimated daily individual exposure as they relate to different locations such as residence and work place for data sets where individual addresses are available.
We observed higher risk for ischemic stroke associated with PM in the day of the event, among young adults. This finding can be explained by the inflammatory mechanism, linking air pollution and stroke. Our findings suggest stronger effect of traffic pollution and not pollution of natural sources.
Sources of Funding
This research was supported by the Milgrom Foundation of Science (M.F.S.).
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.115.010992/-/DC1.
- Received July 26, 2015.
- Revision received September 30, 2015.
- Accepted October 1, 2015.
- © 2015 American Heart Association, Inc.
- Peters A,
- Dockery DW,
- Muller JE,
- Mittleman MA.
- Brook RD,
- Rajagopalan S,
- Pope CA III,
- Brook JR,
- Bhatnagar A,
- Diez-Roux AV,
- et al
- Hong YC,
- Lee JT,
- Kim H,
- Kwon HJ.
- Tsai SS,
- Goggins WB,
- Chiu HF,
- Yang CY.
- Wellenius GA,
- Schwartz J,
- Mittleman MA.
- Pope CA III,
- Burnett RT,
- Thurston GD,
- Thun MJ,
- Calle EE,
- Krewski D,
- et al
- 23.↵Central Bureau of Statistics. Characterization and Ranking of Local Authorities According to the Population’s Socio Economic Level in 2001. http://www.cbs.gov.il/hodaot2004/13_04_22.htm. Accessed May 25, 2015.
- Maclure M.
- You RX,
- McNeil JJ,
- O’Malley HM,
- Davis SM,
- Thrift AG,
- Donnan GA.
- 28.↵World Health Organization. Who Air Quality Guidelines for Particulate Matter, Ozone, Nitrogen Dioxide and Sulfur Dioxide. http://whqlibdoc.who.int/hq/2006/WHO_SDE_PHE_OEH_06.02_eng.pdf?ua=1. Accessed June 17, 2015.
- Genc S,
- Zadeoglulari Z,
- Fuss SH,
- Genc K.
- Ruidavets JB,
- Cournot M,
- Cassadou S,
- Giroux M,
- Meybeck M,
- Ferrières J.
- Wolf PA,
- Abbott RD,
- Kannel WB.
- Yitshak-Sade M,
- Vodonos A,
- Novack V,
- Friger M,
- Amit G,
- Katra I,
- et al
- Moeller DW.
- Sørensen M,
- Hvidberg M,
- Andersen ZJ,
- Nordsborg RB,
- Lillelund KG,
- Jakobsen J,
- et al