In dozens of studies across the world, particulate air pollution has been associated with increases in daily deaths and hospital admissions. 1,2 Other studies have shown that these associations are not confounded by gaseous copollutants, seasonal patterns, or weather. 3–8 However, it is unclear which populations are particularly susceptible to the adverse effects of PM10 (particulate matter with aerodynamic diameter less than 10 microns), and this issue has been identified as a key research need. 9 This question is linked to the question of potential mechanism, which is another key concern.
In recent studies, exposure to airborne particles has been associated with reduced heart rate variability, 10–12 increased c-reactive protein 13 concentrations, and higher peripheral white cell counts. 14,15 Diabetes is a chronic disease characterized by disturbances in all of these cardiovascular risk factors 16–18 and we hypothesized that diabetics might therefore be at increased risk of PM10-associated cardiovascular events.
We previously examined race, sex, and socioeconomic status as modifiers of the effect of air pollution on mortality, 19 and found little evidence of substantial modification. Similar results were reported by Samet and coworkers. 6 We have also found only weak patterns for hospital admissions. 5,20
Goldberg and coauthor found a higher risk of all-cause mortality in diabetics than in nondiabetics, associated with particulate air pollution in Montreal. 21 We have examined the question of effect modification by medical condition in Chicago, and we have previously reported 22 that diabetes was a modifier of hospital admissions for heart, but not lung, disease in persons 65 and older.
In this study, we enlarge the Chicago analysis to four cities in the United States; we examined the effect modification by concurrent diagnosis of diabetes overall and by age group. We applied a hierarchical model to verify that the effects of PM10 on persons with diabetes differ substantially from those on persons without diabetes. In the first stage, separate Poisson regressions were fit in each city and within each stratum of age and diabetes. In the second stage, the estimated effect size for PM10 within city and strata were regressed against the stratification variables, allowing for a random effect.
The health data were extracted from billing records of the Health Care Financing Administration (MEDICARE), which provides hospital coverage for all U.S. citizens age 65 and over. We analyzed data from Cook County (Chicago), Illinois, Wayne County (Detroit), Michigan, Allegheny County (Pittsburgh), PA, and King County (Seattle), WA, between 1988 and 1994. These cities were chosen because they are in the most populous counties in the U.S. with daily PM10 monitoring. Within each city, we computed daily counts of hospital admissions for cardiovascular disease (International Classification of Disease, 9th Revision [ICD-9], 390–429) as given by the discharge diagnoses. We stratified these counts into admissions with or without a notation of diabetes (ICD-9, 250) as a secondary, contributing factor for the admission. We also stratified by age, producing four daily counts of cardiovascular admissions: with diabetes, age 65–74; with diabetes, age 75+; without diabetes, age 65–74; and without diabetes, age 75+.
Weather data for the four cities were obtained from the EarthInfo CD-ROM (EarthInfo CD NCDC Surface Airways, EarthInfo Inc., Boulder, CO), and air pollution data were obtained from the U.S. Environmental Protection Agency's Aerometric Information Retrieval System (AIRS) network. 23 Where more than one PM10 monitor was operating in a city, we computed the daily mean using an algorithm that accounts for different means and variances among the monitors. 3
The associations between hospital admission counts and PM10 were investigated with a generalized additive robust Poisson regression model. 24 Such models have been fit previously in these cities to examine the association of PM10 with cardiovascular admissions in all persons age 65 and older. 3,6
In this analysis, those same models were fit to stratum-specific daily counts. The generalized additive model allows us to better model the nonlinear dependence of daily admissions on weather and season by using nonparametric smoothing. The covariates we examined were temperature, prior day's temperature, relative humidity, barometric pressure, and day of the week. To control for weather variables, we chose the smoothing parameter that minimized Akaike's Information Criterion. 25 To model seasonality, we chose the smoothing parameter that reduced the residuals to white noise. Where that was not possible, we incorporated short-term autoregressive parameters and chose the smoothing parameter that reduced the remaining residuals to white noise. Further details have been published previously. 3,6
The mean of PM10 on the day of the admission and the day before the admission was used as our exposure variable; this was treated linearly. Separate models were fit for each of the four subgroups in each of the four cities.
Second Stage of the Analysis
To examine effect modification by diabetes in the four cities we fit a meta-regression model of the form:EQUATION
Where βij is the estimated PM10 effect in city i and group j, Dj is the dummy for diabetes, and Aj is the dummy variable for age 75+. Then α tells us how much the effect of PM10 changes for a person with diabetes, γ indicates how much it changes for a person over 75, and η allows for an interaction between the two categories. β* is the effect of PM10 on persons without diabetes who are younger than 75 years old.
In this model we assume that:EQUATION
where βij and μj are defined above, Sij is the estimated variance-covariance matrix, and δ is the random variance-covariance matrix component, reflecting heterogeneity in response among the cities. δ ≈ 0 corresponds to the fixed effect estimates.
To estimate the model parameters, the method described by Berkey et al.26 is applied. More specifically, the iterative generalized least squares method to estimate model parameters is:EQUATION
where V = Ŝij + δ and θ = (β*,α,γ,η)T. The covariance matrix δ is estimated by maximum likelihood.
Table 1 shows the 25th, 50th, and 75th percentile values for the number of daily admissions for cardiovascular disease with and without a concurrent diagnosis of diabetes. Admissions without diabetes were around four times higher than admissions for those causes with diabetes, except in Seattle, where they were seven times higher.
Table 1 also shows the 25th, 50th, and 75th percentile values for the environmental variables in the four cities. The median values for PM10 are similar in all four cities.
Table 2 shows the results for the effect of PM10 in each city by strata of diabetes and age. These city-specific results show higher effects of PM10 in diabetics than in nondiabetics in all four cities, although the differences are modest in all cities.
Table 3 shows the results of the hierarchical modeling. We applied a meta-regression with indicator variables for diabetes and for older age; the interaction term of diabetes by older age was quite small.
Diabetes was a modifier of the effect of PM10 on cardiovascular hospital admissions, with percentage increases almost doubled compared with nondiabetics. Age was also a modifier of risk. There was a suggestion of a somewhat smaller difference in the effect of diabetes in the older age group than in the younger group.
As Chicago had more events than any of the other cities, and because we have previously reported results from Chicago, we repeated the meta-regression for the three new cities, excluding Chicago. The findings were quite similar to those attained by using all four cities.
We have confirmed our initial observation of effect modification by diabetes in a multicity hierarchical model. We found that diabetics have double the percentage increase in PM10-associated cardiovascular admissions compared with nondiabetics. We also found that persons age 75 and older had higher risk, and that the interaction between the two categories (diabetes and age) was less than additive. If the latter relation is confirmed in other studies, it may represent a survivor effect among diabetes, enhanced susceptibility among nondiabetics, or both. In this study, we assume that the relation between hospital admissions and particulate matter is linear. In a previous study, we tested for the linearity of the exposure response function in all persons age 65 and older 27 in 10 U.S. cities, and we found an almost linear relation with steeper slopes at lower doses.
The observation of higher risks among diabetics is interesting for several reasons. As noted in the introduction, recent studies have reported that airborne particles are associated with reduced heart rate variability, 10–12 increased plasma c-reactive protein, 13 and increased plasma fibrinogen 14 and white cell counts. 14,15 These are dimensions of cardiovascular vulnerability that are affected by diabetes as well, which makes the observation plausible. More interestingly, it raises the issue of whether other cardiovascular risk factors affected by diabetes are also vulnerable to airborne particles. For example, diabetics are at increased risk of vascular injury. Whether particulate exposure has similar effects on the association between diabetes and vascular disease may be a fruitful area of research. Some preliminary reports support this hypothesis. For example, Vincent and coworkers 28 have recently reported that exposure to urban particles increased endothelin I and endothelin III concentrations in plasma. Another recent study reported that rabbits exposed to urban particles showed, in addition to increases in white cell counts, increases in atherosclerotic lesions in coronary arteries. 29
Diabetics also show an upregulation of inflammatory activity. Recent reports have indicated that in addition to elevating white cell counts, particle exposure is associated with increased concentrations of c-reactive protein in plasma. 12 C-reactive protein is an acute phase marker of inflammation that is associated with risk of cardiovascular events such as myocardial infarction and death. 30
Salvi and coworkers 15 recently reported that human subjects exposed to diesel particles had elevated levels of vascular cell adhesion molecule-1 and intercellular adhesion molecule-1. The association of factors such as these with particle exposure would appear to be a fruitful area of research.
Our findings indicate that the interaction of particle exposure with diabetes deserves more attention. Even though the lack of detailed medical information on subjects is a limitation of this study, diabetes is a highly prevalent chronic illness and thus public health implications of this association are potentially large.
1. Pope CA, Dockery DW, Schwartz J. Review of epidemiologic evidence of health effects of particulate air pollution
. Inhal Toxicol 1995; 7: 1–18.
2. Schwartz J. Air pollution and daily mortality: a review and meta analysis. Environ Res 1994; 64: 36–52.
3. Zanobetti A, Schwartz J, Dockery DW. Airborne particles are a risk factor for hospital admissions
for heart and lung disease. Environ Health Perspect 2000; 108: 1071–1077.
4. Schwartz J. Assessing confounding, effect modification, and thresholds in the association between ambient particles and daily deaths. Environ Health Perspect 2000; 108: 563–568.
5. Schwartz J. Air pollution and hospital admissions
for heart disease
in eight U.S. counties. Epidemiology 1999; 10: 17–22.
6. Samet JM, Zeger SL, Dominici F, et al
. The National Morbidity, Mortality, and Air Pollution Study. Part II: morbidity, mortality, and air pollution in the United States. Res Rep Health Eff Inst 2000; 94 (pt 2):5–79.
7. Samet J, Zeger S, Kelsall J, Xu J, Kalkstein L. Does weather confound or modify the association of particulate air pollution
with mortality? An analysis of the Philadelphia data 1973–1980. Environ Res 1998; 77: 9–19.
8. Samet JM, Dominici F, Curriero FC, Coursac I, Zeger SL. Fine particulate air pollution
and mortality in 20 U.S. cities, 1987–1994. N Engl J Med 2000; 343: 1742–1749.
9. National Research Council. Research Priorities for Airborne Particulate Matter. Washington DC: National Academy Press, 1998.
10. Gold DR, Litonjua A, Schwartz J, et al
. Ambient pollution and heart rate variability. Circulation 2000; 101: 1267–1273.
11. Pope CA III, Verrier RL, Lovett EG, et al
. Heart rate variability associated with particulate air pollution
. Am Heart J 1999; 138: 890–899.
12. Liao D, Creason J, Shy C, Williams R, Watts R, Zweidinger R. Daily variation of particulate air pollution
and poor cardiac autonomic control in the elderly. Environ Health Perspect 1999; 107: 521–525.
13. Peters A, Frohlich M, Doring A, et al
. Particulate air pollution
is associated with an acute phase response in men: results from the MONICA-Augsburg study. Eur Heart J 2001; 22: 1198–204.
14. Schwartz J. Air pollution and blood markers of cardiovascular risk. Environ Health Perspect 2001; 109: 405–409.
15. Salvi S, Blomberg A, Rudell B, et al
. Acute inflammatory responses in the airways and peripheral blood after short-term exposure to diesel exhaust in healthy human volunteers. Am J Respir Crit Care Med 1999; 159: 702–709.
16. Singh JP, Larson MG, O'Donnell CJ, et al
. Association of hyperglycemia with reduced heart rate variability (The Framingham Heart Study). Am J Cardiol 2000; 86: 309–312.
17. Stee JJ, Silberhatz H, Tofler GH, et al
. Association of fibrinogen with cardiovascular risk factors and cardiovascular disease in the Framingham Offspring Study. Circulation 2000; 102: 1634–1638.
18. Gerritsen J, Dekker JM, TenVoorde BJ, et al
. Glucose tolerance and other determinants of cardiovascular autonomic function: the Hoorn Study. Diabetologia 2000; 43: 561–570.
19. Zanobetti A, Schwartz J. Race, gender, and social status as modifiers of the effects of PM10
on mortality. J Occup Environ Med 2000; 42: 469–474.
20. Zanobetti A, Schwartz J, Gold DR. Are the sensitive subgroups for the health effects of airborne particles? Environ Health Perspect 2000; 108: 841–845.
21. Goldberg MS, Burnett RT, Bailar JC III, et al
. The association between daily mortality and ambient air particle pollution in Montreal, Quebec. 2. Cause-specific mortality. Environ Res 2001; 86: 26–36.
22. Zanobetti A, Schwartz J. Are diabetics more susceptible to the health effects of airborne particles? Am J Respir Crit Care Med 2001; 164: 831–833.
23. Nehls GJ, Akland GG. Procedures for handling aerometric data. J Air Pollut Control Assoc 1973; 23: 180–184.
24. Hastie T, Tibshirani R. Generalized Additive Models. London: Chapman and Hall, 1990.
25. Akaike H. Information theory and an extension of the maximum likelihood principal. In: Petrov BN, Csaki F, eds. Second International Symposium on Information Theory. Budapest: Akademiai Kiado, 1973; 267–281.
26. Berkey CS, Hoaglin DC, Antczak-Bouckoms A, Mosteller F, Colditz GA. Meta-analysis of multiple outcomes by regression with random-effects. Stat Med 1998; 17: 2537–2550.
27. Schwartz J, Zanobetti A. A meta-smoothing approach to analyze the dose response relationship between particulate matter and hospital admissions
. Epidemiology 2000; 11: S74–177.
28. Vincent R, Kumarathasan P, Mukherjee B, et al
. Exposure to urban particles (PM2.5) causes elevations of the plasma vasopeptides endothelin (ET)-1 and ET-3 in humans. Am J Crit Care Med 2001; 165: A313.
29. Suwa T, van Eeden SF, Quinlan K, et al
. The effect of PM10 on coronary aterosclerosis. Am J Crit Care Med 2001; 165: A311.
30. Devlin RB, Johnson TA, Ghio A, et al
. Pulmonary and cardiac changes in pigs exposed to residual oil fly ash (ROFA). Am J Respir Crit Care Med 2001; 165: A311.