Published ahead of print on November 17, 2005, doi:10.1164/rccm.200507-1123OC
© 2006 American Thoracic Society doi: 10.1164/rccm.200507-1123OC
Air Pollution and Markers of Inflammation and Coagulation in Patients with Coronary Heart DiseaseGSF–National Research Center for Environment and Health, Institute of Epidemiology, and Focus-Network Aerosols and Health, Neuherberg; IBE Department of Epidemiology, Ludwig-Maximilians-University of Munich, Munich; Department of Internal Medicine II, Cardiology, University of Ulm Medical Center, Ulm, Germany; and Department of Medicine and Dentistry, Vascular Medicine, and Pulmonary and Critical Care Unit, Rochester School of Medicine and Dentistry, Rochester, New York Correspondence and requests for reprints should be addressed to Regina Rückerl, M.Sc., GSF–National Research Center for Environment and Health, Institute of Epidemiology, Ingolstaedter Landstrasse 1, 85764 Neuherberg, Germany. E-mail: rueckerl{at}gsf.de
Rationale: Ambient air pollution has been shown to be associated with cardiovascular morbidity and mortality. Objectives: A prospective panel study was conducted to study the early physiologic reactions characterized by blood biomarkers of inflammation, endothelial dysfunction, and coagulation in response to daily changes in air pollution in Erfurt, Germany. Methods: Blood parameters were repeatedly measured in 57 male patients with coronary heart disease during the winter of 2000/2001. Fixed-effects linear and logistic regression models were applied, adjusting for trend, weekday, and meteorologic parameters. Measurements: Hourly data on ultrafine particles (UFPs; number concentration of particles from 0.01 to 0.1 µm), mass concentration of particles less than 10 (PM10) and 2.5 µm in diameter, elemental and organic carbon, gaseous pollutants, and meteorologic data were collected at central monitoring sites. Main Results: Increased levels of C-reactive protein above the 90th percentile were observed for an increase in air pollution concentrations of one interquartile range. The effect was strongest for accumulation mode particles, with a delay of 2 d (odds ratio [OR], 3.2; confidence interval [CI], 1.7, 6.0). Results were consistent for UFPs and PM10, which also showed a 2-d delayed response (OR, 2.3; CI, 1.3, 3.8; and OR, 2.2; CI, 1.2, 3.8, respectively). However, not all of the blood markers of endothelial dysfunction and coagulation increased consistently in association with air pollutants. Conclusion: These results suggest that inflammation as well as parts of the coagulation pathway may contribute to the association between particulate air pollution and coronary events.
Key Words: acute-phase reaction • air pollution • blood coagulation • cardiovascular diseases • C-reactive protein Increasing evidence suggests that ambient air pollution may adversely affect the cardiovascular system. It has been shown that ambient air pollution leads to increased cardiovascular mortality (1–6), and studies found associations between ambient air pollution and hospital admissions for various cardiovascular diseases, including congestive heart failure (7–9). Also, an increased risk for acute myocardial infarction (MI) (10) and cardiorespiratory symptoms (11) has been reported in association with particulate air pollution. The exact mechanisms linking the inhalation of ambient air particles to an acute exacerbation of cardiovascular disease are not completely understood (12). Seaton and coworkers (13) hypothesized that inhaled particles would lead to alveolar inflammation, which increases the level of blood coagulability, thus leading to an increased risk of ischemic events in susceptible individuals. DeMeo and colleagues (14) found reduced oxygen saturation in association with particulate matter of less than 2.5 µm in diameter (PM2.5). Pope and colleagues (15), who linked long-term exposure to particulate air pollution to various causes of mortality, found a strong and robust association between PM2.5 and cardiovascular disease mortality. They concluded that exposure to particulate air pollution and cardiopulmonary mortality risk is linked by accelerated pulmonary and systemic inflammation. Moreover, Peters and coworkers (16) demonstrated increased levels of plasma viscosity during an air pollution episode in central Europe, compared with less polluted days. Increased plasma concentrations of C-reactive protein (CRP), the classic acute-phase protein, were also shown during the 1985 air pollution episode (17). There is a strong link between inflammation and coronary heart disease (CHD) because factors involved in inflammation and infection seem to play a proatherogenic role and inflammation has been identified as a potent risk factor for acute coronary syndrome. Systemic inflammation could result in destabilization or even rupture of vulnerable atheromatous plaques, leading to acute ischemic events. Most of the cited studies have been conducted in the general population or in elderly healthy subjects. This study looks at a susceptible subgroup to provide insight into the ways in which air pollution might precipitate death in persons with underlying heart disease, based on the hypothesis that particulate air pollution can alter cardiovascular function. Repeated measurements of markers of an early inflammatory response, cell recruitment and coagulation, were compared with concurrent levels of air pollution. Our primary hypothesis was that CRP, a well-known marker for inflammation, would increase in association with a rise in levels of air pollution. Moreover, we analyzed various other markers of inflammation (serum amyloid A [SAA]), cell adhesion (E-selectin, von Willebrand factor antigen [vWf], intercellular adhesion molecule 1 [ICAM-1]), and coagulation (fibrinogen, factor VII [FVII], prothrombin fragment 1+2, D-dimer) on a more explorative basis hypothesizing that the levels of these blood markers would also go up in association with higher levels of air pollution, as seen in Figure 1. Some results have been previously presented in form of an abstract (18).
Study Design As part of the University of Rochester Particulate Matter Center, a prospective panel study was conducted between October 15, 2000, and April 27, 2001, in Erfurt, Germany. The panel consisted of male patients with CHD who were scheduled for 12 subsequent clinical visits. Each clinical visit included a short interview and the withdrawal of a blood sample. At the first visit, a baseline questionnaire was administered regarding health status, pulmonary and cardiac symptoms, medication intake, and smoking history. Sixty-one nonsmoking men, aged 50 yr or older, with physician-diagnosed CHD were recruited through a local cardiologist. Patients with pacemakers, recent (< 3 mo ago) MI, bypass surgery, or balloon dilatation were not included because the inflammatory processes involved in such a procedure might not yet have subsided. Persons with type 1 diabetes or on anticoagulation therapy (except for antiplatelet agents) were also not included. A written, informed consent was obtained from all subjects. The study protocol was approved by the German Ethics Committee of the "Bayerische Landesärztekammer" in Munich, Germany. All methods used in the study were conducted according to standard operating procedures and were tested beforehand in a 2-wk pilot study.
Air Pollution Monitoring Continuous ultrafine particle (UFP) counts (0.01–0.1 µm), accumulation mode particle (AP) counts (0.1–1.0 µm), and fine-particle mass (PM2.5) were measured with the mobile aerosol spectrometer (MAS). The MAS, described previously (24, 25), consists of two different, commercially available instruments covering different size ranges. Particles in the size range from 0.01 to 0.5 µm were measured using a differential mobility particle sizer (TSI, Aachen, Germany). Particles in the size range from 0.1 to 2.5 µm were classified by an optical laser aerosol spectrometer (PMS, Leonberg, Germany). PM10 (particulate matter < 10 µm in diameter) data were collected by the tapered element oscillating microbalance method (TEOM 1400a; Rupprecht and Patashnik, Albany, NY) and continuous data on elemental (EC) and organic carbon (OC) were measured with an ambient carbon monitor (ACM 5400; Rupprecht and Patashnick). Data on meteorologic variables for this period as well as concentrations of gaseous air pollutants were collected from existing networks. Missing values of the ambient UFPs between January 20 and February 13 were imputed by a linear regression model based on parallel measurements with a condensation particle counter and a scanning mobility particle sizer. The R squares for the regression model was 0.96. Also, between December 2000 and May 2001, approximately 15% of the PM2.5 measurements by MAS were lost. These missing values were replaced by corrected data based on parallel measurements with TEOM-PM10 and Harvard Impactor-PM2.5. (Air Diagnostic and Engineering, Inc., Naples, ME). For each person and visit, the individual 24-h average of pollutants immediately preceding the clinical visit (lag 0) up to Day 5 (lag 1–4) and 5-d running means before the examination were calculated if more than two-thirds of the hourly measurements were available for this period.
Clinical Measurements At each visit, ethylenediaminetetraacetic acid and citrate plasma samples were drawn (Becton Dickinson, Franklin Lakes, NJ). Samples were centrifuged and aliquots were immediately stored at –20°C until analysis. CRP (high-sensitivity assay), SAA, and fibrinogen were analyzed by immunonephelometry (Dade Behring, Marburg, Germany). ICAM-1, E-selectin (R&D Systems, Wiesbaden, Germany), and prothrombin fragment 1+2 (Dade Behring) were measured by means of a commercial ELISA. D-dimer and vWf were analyzed using an immunoturbidimetric method and FVII by clotting time measurement (Diagnostica Stago, Asnieres-sur-Seine, France).
Study Subjects
Statistical Analyses Prothrombin fragment 1+2, FVII, SAA, CRP, and E-selectin were log-transformed before analysis because their residuals remained skewed after multivariate modeling. Model building was done for each blood parameter separately without an air pollution variable. To explore the shape of the association between confounders and blood markers, nonparametric smooth functions on the basis of locally weighted least squares were applied for all confounders. Model fit was rated on the basis of the Akaike information criterion (AIC). In the final model, nonparametric smooth functions were replaced by appropriate polynomials (degree 2 or 3) or natural splines based on lowest AIC. After the model fit was completed, dose–response functions of the confounders were checked visually and in case implausible shapes were observed, degrees of freedom were decreased. Each pollutant was added separately to the final model. Data were analyzed using the statistical package SAS version 8.2 (SAS Institute, Inc., Cary, NC) and S-Plus version 6.0 (Mathsoft Engineering and Education, Inc., Cambridge, MA). Logistic regression models were used to determine whether the effect was limited to the upper tail of the distribution. Confounder adjustment was done in the same way as described for the linear regression models; however, more parsimonious models were used. Sensitivity analyses were done to explore the influences of the different confounder models.
Patient Characteristics Patient characteristics are summarized in Table 1. The study population comprised 57 nonsmoking men, aged 51 to 76 yr. Approximately 84% of them were retired. Except for one person, all patients had stopped smoking at least 1 yr before the examinations.
Air Pollutants The distributions of the 24-h average concentrations of the particulate and gaseous pollutants as well as meteorologic data are given in Table 2.
PM10, PM2.5, and AP were highly correlated (r = 0.90–0.91), whereas UFPs were only moderately correlated with PM10 and PM2.5 (r = 0.57 and 0.41, respectively). PM2.5 showed a moderate negative correlation with air temperature (r = –0.5; Figure 2). EC and OC showed high correlation (r = 0.96) and were also highly correlated with all other particle fractions (r = 0.63–0.90). Also, CO and NO2 were highly correlated (r = 0.82), whereas the correlation for UFPs with NO2 was slightly lower than with CO (r = 0.75 and 0.82, respectively).
Blood Parameters Levels of blood parameters are summarized in Table 3. Parameters of the acute-phase response, SAA and CRP, were correlated (r = 0.53), as were the adhesion molecules ICAM-1 and E-selectin (r = 0.53). However, no significant correlation was seen between markers of an acute-phase response and adhesion molecules (r = 0.08 to 0.31). SAA and CRP also showed a moderate correlation with fibrinogen (r = 0.44 and 0.34, respectively).
Regression Results Results for the regression of different blood markers are summarized in Table 4 (logistic regression) and Table 5 (linear regression). Effect estimates are presented together with 95% confidence intervals (95% CI) based on an increase in air pollution concentration from the first to the third quartile (interquartile range).
Inflammation and adhesion. For CRP, the odds of observing concentrations above the 90th percentile were consistently increased in association with PM10 and UFPs (Figure 3) as well as AP, NO2, and CO for lag 2. The highest odds ratio (OR) was seen with AP, whereas EC and OC showed no significant results.
The OR for observing high ICAM-1 levels increased, especially for lag 1 and 2. This pattern was seen for PM10 (Figure 3), AP, EC and OC, and CO. For ICAM-1, a decrease with lag 0 was also found for most pollutants. Results for SAA indicate an increase in association with particulate air pollution (e.g., with UFP concentrations); however, results are not as strong and consistent as for CRP (Figure 3). Linear regression analyses looking at the continuous distribution did not reveal significant results for CRP, ICAM-1, and SAA. Also, E-selectin did not show any association with ambient air pollution (Figure 3). Linear regression analyses of vWf (Table 5) revealed statistically significant positive associations for most pollutants with lag 0 and for the 5-d average exposure (Figure 3). For PM2.5 and AP, the effect was limited to the 5-d average exposure. Associations for the 1-d lag were found to be even stronger than for lag 0; however, this was not consistent throughout all pollutants.
Blood coagulation. Logistic regression results for FVII were in agreement with the results of the linear regression. For prothrombin fragment 1+2, the logistic regression revealed constant increases of the OR, with lag 4 showing a consistent pattern in all measured pollutants (data not shown). Fibrinogen only revealed very few significant effects, which might be due to chance. Analyses of D-dimer revealed a null result in linear as well as in logistic models (Figure 3).
Sensitivity analyses. For CRP, adding temperature, relative humidity, and air pressure resulted in higher AIC values. In these models, the results for the two-lagged effect of UFPs and AP were confirmed; however, these had generally wider confidence intervals. For prothrombin fragment 1+2, the AIC was reduced by adding air pressure to the model. However, estimates were up to twofold higher and results for the AP destabilized. Therefore, the more conservative and more stable model was used. Throughout all models, stable results were found for lag 2 with the UFPs and with lag 4 for AP, PM10, PM2.5, EC, and OC. The results for ICAM-1 also remained stable throughout all models. Moreover, we conducted sensitivity analyses comparing the results for those patients who were on lipid-lowering drugs, primarily statins, with those who were not. Results for the linear regression show that the effects were mainly driven by the patients who were not on lipid-lowering medication. The effects were larger than the overall effects but had wide confidence intervals due to reduced power. Stratified analyses for CRP showed stronger effects in the patients taking statins. We compared the results of a random-effects model with those of the fixed-effects model for the linear regression, showing consistent effect estimates (FVII, AP, 5-d average exposure: OR, –4.3; 95% CI, –8.1, –0.5). Some associations were found to have a nonlinear exposure response function as marked in Tables 4 and 5. Nonlinearity weakens the evidence for a strong influence of these pollutants; however, for the CRP, all associations were linear (Table 4).
Summary Our findings suggest increases in CRP and ICAM-1 in association with ambient air particles. For these markers, the effects were limited to the higher values of the parameters, showing an increase in the odds of observing high levels of the respective parameters with elevated levels of air pollution. CRP rose with a delay of 2 d for all measured pollutants except for EC and OC. For ICAM-1, a 1- and 2-d delayed increase was associated with most pollutants. Mean concentrations of vWf were shifted toward higher values revealing the strongest effect for the 5-d average exposure to almost all pollutants. Moreover, a rise in prothrombin fragment 1+2 in association with all pollutants was seen, which was consistent for lag 4. For FVII, clear and consistent negative associations were observed. UFPs, NO2, and CO were measured simultaneously and therefore allow a rough estimate on how well the two gases can be used as surrogates for the exposure to combustion-derived particles. Our results show a high correlation among these three pollutants, and also the estimates for the blood markers show comparable results, especially in logistic regressions. Similar conclusions have been drawn by Cyrys and colleagues (26), who found UFPs, NO2, and CO to be strongly correlated and to reflect motor vehicle traffic.
Possible Mechanisms
Inflammatory Pathway
Endothelial Dysfunction In addition, vWf may serve as a marker of endothelial dysfunction. In healthy mice, increased vWf expression on hepatic endothelium was detected after application of UFPs (41). vWf reflects endothelial cell release and probably vascular reactivity. Vascular reactivity could be secondary to inflammation, and because vWf can mediate platelet adhesion to damaged endothelium, this could be a predictor of coronary events (34, 42).
Coagulation Pathway FVII, one of the key enzymes of the extrinsic system of the coagulation cascade, is activated by tissue factor. Complexes of tissue factor with factor VIIa are central to the activation of factor X and to the formation of thrombin, which mediates the conversion of fibrinogen to fibrin (43). Results for FVII in the literature are inconsistent (28, 44). In our study, FVII activity decreased significantly in association with most pollutants. Regarding fibrinogen levels, we did not find any consistent results in association with air pollution. Controlled human exposure studies (45, 46) as well as epidemiologic studies (47, 48) demonstrated positive associations between fibrinogen or plasma viscosity and air pollution. However, decreases in fibrinogen levels also have been reported and the significance of these results is unknown (28, 41). Prothrombin fragment 1+2 is cleaved from prothrombin when it is activated to thrombin by factor Xa, thereby representing a marker of activation of the coagulation pathway (34, 49, 50). Our data indicate an increased concentration in association with ambient air pollutants. This significant increase indicates that an early step of blood coagulation has been activated. However, this activation was not associated with increased formation of fibrin, as would be detected by elevated D-dimer levels. The elevated levels of prothrombin fragment 1+2 are an important finding that shows that air pollution not only induces inflammation but also coagulation. The large number of blood markers measured in this study revealed inconsistencies that were already observed in previous studies (13). One possible explanation is that various particle fractions or components differ in their effects. CRP, for example, did not show any association with EC and OC, whereas other blood markers showed quite strong effects. Moreover, diverse time patterns in the reaction to air pollution, due to the differing biological mechanisms, are conceivable and were also seen in the data. While the results for inflammation and air pollution seem consistent, inhomogeneity exists in terms of coagulation markers. Our data strongly indicate that the pathway that links airway injury from air pollution and coronary events may include increased expression of adhesion molecules and a proinflammatory response. Furthermore, the coagulation system (prothrombin fragment 1+2) is activated, although not sufficiently enough to cause increased fibrin formation, as would have been reflected by elevated D-dimer levels. We do not have an explanation for the decreased concentrations of fibrinogen and FVII, but, whatever the cause, they may act to protect against clinical events secondary to coronary thrombosis. Our study represents measurements of background levels of blood markers and does not reflect changes that might relate to acute clinical events.
Strengths and Limitations We used logistic regression analyses in addition to linear regression because Peters and colleagues (16) had suggested that effects of air pollution may be seen on a small number of subjects with high values of a particular marker with less effect on the mean level. Also, some parameters showed a few extreme outliers, which strongly influenced the regression results. Although the fixed-effect models were adjusted for individual time-invariant factors, by design no adjustment for time-dependent individual-level variables was possible. A variety of pollutants were used for the analyses, because different pollutants may point toward different properties of the aerosol, and also represent different sources of air pollution. However, by testing multiple blood parameters and a set of air pollutants, the possibility that some effects might have occurred by chance cannot be excluded. Because the air pollution parameters are closely correlated, we considered especially consistent patterns in the data as actual effects. Moreover, thorough confounder adjustment for meteorologic variables was done to rule out the possibility that the detected associations resulted from meteorologic influences or seasonal differences. Only one central measurement site was used for the collection of ambient air pollution. However, the spatial representativeness of this site has been analyzed in detail previously by Cyrys and colleagues (51), who measured sulfate and PM10 levels simultaneously at three additional monitoring sites in the Erfurt area. The relatively high intersite correlation between the monitoring stations (0.69–0.98) indicates that regional episodes of sulfates and PM10 in Erfurt can be identified using one fixed monitoring site and that our site is generally representative for the urban background level of air pollution within Erfurt. Erfurt is a small city with one air mass confined by a mountain ridge on three sides and high rises on the fourth side. Because most of the participants of our panel were already retired, we assume that they spent the greater part of their day within the vicinity of their residence within the city of Erfurt. Many studies have demonstrated that individual exposures to PM are poorly correlated spatially with ambient concentrations (52). Some longitudinal exposure assessment studies of PM and specific PM components with repeated measures have found higher correlations between personal exposures and ambient concentrations. Janssen and coworkers (53) showed, for example, that ambient, indoor, and personal concentrations of PM2.5 were highly correlated in two European cities. However, the correlation many epidemiologists are interested in is not that between total personal exposure and outdoor concentrations but the correlation between that component of personal exposure that can be attributed to outdoor particles and the outdoor concentrations. Ebelt and colleagues (54) demonstrated that ambient concentrations and the contribution of ambient particles to personal PM exposure were highly correlated, with a Pearson correlation coefficient of 0.81 for PM2.5, of 0.71 for PM10 and of 0.73 for the coarse fraction (PM10 – PM2.5). Moreover, they show that ambient concentrations and exposure to nonambient PM2.5 are independent, which is an important assumption in epidemiologic studies that use ambient concentration as a surrogate for personal exposure. They conclude that their results give support to the use of ambient monitoring data in time series analyses. Cyrys and coworkers (55), who compared the relationship of indoor and outdoor levels of fine-particulate mass, particle number concentrations, and black smoke, concluded that ambient concentrations of PM2.5 and black smoke can be used as good approximations of indoor concentrations. A limitation to the study is that the examined panel consisted of male patients only, with a history of CHD, who were all taking cardiac medication. Therefore, they represent a highly selected group and the study results might not be generalizable to other population groups, such as females with CHD or healthy subjects. A differentiation between chronic and acute effects of higher levels of blood markers in the patients is not possible with this study design. Because of the short observation time, it is not clear whether these changes can lead to an onset or exacerbation of the disease. We observed short-term changes in various blood parameters; however, the implications for patients remain speculative. On the other hand, changes in blood markers due to air pollution have recently been observed not only in patients with CHD but also in young and healthy persons (56).
Conclusions
The authors thank Dr. O. Manuwald and his team and the German Weather Service (DWD). The study is funded through the U.S. Environmental Protection Agency STAR Center grant R-827354 and the Focus-Network of Aerosols and Health, GSF. The Focus-Network of Aerosols and Health coordinates and focuses all GSF research on health effects and the characterization of aerosols. It comprises research projects of the Institutes of Ecological Chemistry, Epidemiology, Inhalation Biology, Radiation Protection, and Toxicology at GSF.
Supported by the U.S. Environmental Protection Agency STAR center grant R-827354 and the Focus-Network of Aerosols and Health, GSF. Originally Published in Press as DOI: 10.1164/rccm.200507-1123OC on November 17, 2005 Conflict of Interest Statement: None of the authors have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. Received in original form July 21, 2005; accepted in final form November 11, 2005
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
