A Case-Crossover Analysis of Particulate Matter Air Pollution and Out-of-Hospital Primary Cardiac Arrest Author(s): Drew Levy, Lianne

Sheppard, Harvey Checkoway, Joel Kaufman, Thomas Lumley, Jane Koenig, David Siscovick Reviewed work(s): Source: Epidemiology, Vol. 12, No. 2 (Mar., 2001), pp. 193-199 Published by: Lippincott Williams & Wilkins Stable URL: http://www.jstor.org/stable/3703622 . Accessed: 01/11/2011 15:56
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Lippincott Williams & Wilkins is collaborating with JSTOR to digitize, preserve and extend access to Epidemiology.

http://www.jstor.org

Case^Crossover and Levy,1 Lianne

Analysis Out-of^Hospital Sheppard,2,3 Jane Harvey Koenig,3

of

Particulate Primary Cardiac Kaufman,3,4

Matter

Air Arrest

Pollution Drew

Checkoway,1,3 and David

Joel

Thomas

Lumley,2

Siscovick1,4

Numerous epidemiologic studies have reported increases in the daily incidence of cardiovascular mortality and morbidity as? sociated with increases in daily levels of particulate matter air pollution. We studied the association between the incidence of primary cardiac arrest and two daily measures of particulate matter using a case-crossover study of 362 cases of out-ofhospital cardiac arrest. All cases were attended by paramedics and had no history of clinically recognized heart disease or life-threatening comorbidities. We compared particulate mat? ter levels at index times with particulate matter levels from referent days matched on day of week within strata defined by month and year. The estimated relative risk at a lag of 1 day for an interquartile range (IQR) change in nephelometry (0.51 X

10'1 km'1) was 0.893 (95% CI = 0.779-1.024). The estimated relative risk at a lag of 1 day for an IQR change in PM10 (19.3 was 0.868 (95% CI = 0.744-1.012). Other lag periods jiLgm'3) gave similar results. We did not find evidence of confounding by carbon monoxide or sulfur dioxide. Analysis of effect mod? ification by individual-level variables did not reveal any sus? ceptible subgroups. These findings do no support an association between particulate matter and increased risk of primary car? diac arrest among persons without clinically recognized heart disease. The null results of this study may result from several factors, including the highly selected nature of this case series and the relatively low particulate matter levels in the Seattle metropolitan area. (Epidemiology 2001;12:193-199)

Keywords: case-crossover studies, air pollution, particulate matter, sudden cardiac arrest

evidence that in? There is accumulating epidemiologic creases in ambient air pollutants are associated with increases in non-accidental daily mortality. Most evi? dence comes from ecologic time-series studies in North America and Western Europe.1'7 Overall, daily variation in levels of ambient particulate matter (PM) air pollu? has found in urban environments tion as commonly been associated with daily total mortality increases of 0.5% per 10 /ig/m3 increase in particulate matter less than 10 jLtm in diameter (PM10).8 For cardiovascular mortality, summary estimates of 1.4% per 10 /xg/m3 have been reported.1,3,9 Corroborative evidence has been ob-

of Fromdepartmentof Epidemiology, 3Department 2Department Biostatistics, of of of Environmental Health,4Department Medicine, University Washington, WA 98195-7232. Seattle, of Box 357232, Department to: Addresscorrespondence LianneSheppard, of Seattle,WA 98195-7232. Biostatistics, University Washington, This research fundedin part by the Health EffectsInstituteResearch was Protection 97-2-2 andin partby the UnitedStatesEnvironmental Agreement Research Center R827355forthe EPANorthwest agreement Agencythrough and forParticulate Pollution Health. Air and the does reflect viewsof eitherfunding Thisresearch not necessarily agency endorsement shouldbe inferred. no official Submitted accepted August1, 2000. April25, 2000;finalversion & Inc. ? Williams Wilkins, Copyright 2001by Lippincott

of cardiovascular tained from analyses hospital admissions.10'13 Despite the apparent consistency of the epidemiologic findings for PM effects on daily cardiovascular disease mortality, there remain important unresolved issues that limit causal interpretation. These include design limita? of disease outcomes analyzed in tions, the non-specificity most time-series studies, uncertainties regarding the most toxic components of PM, and the frequently noted and potential problems of exposure misclassification, and factors, co-pollutants, confounding by climatic other putative disease risk factors. The ecologic timeseries design, where in exposures and health outcomes are investigated at the population level, rather than at the individual level, is used in the majority of existing studies. In addition to the utilization of ambient pollu? tion measures as a proxy for personal exposure, the ecologic design requires additional fundamental assump? tions before relative risk estimates can have individuallevel interpretation.14 Furthermore, the absence of personal-level risk factor data also limits the ability to assess effect modification by other risk factors. Characteriza? tion of effect modification, which can assist in the iden? tification of susceptible subgroups in the population, becomes especially important in epidemiologic studies of relative risks PM because of the small population-wide that are usually observed. Few studies have been able to evaluate specific forms of cardiovascular effects attribut? able to PM because of the challenges of sensitive and

193

194

Levy et al

Epidemiology

March 2001, Vol. 12 No. 2

outcome surveillance and specific population-based ascertainment. Cardiovascular disease accounts for a sizeable propor? tion of daily mortality and is thus an important area of focus for PM research. Possible effects to the cardiovas? cular system from transient changes in PM levels should be most readily detected in studies of acute-onset events. Schwartz15 found a disproportionate increase in sudden death on high air pollution days, indicating cardiac arrest may be an important component of the PMmortality association. Cardiac arrest refers to the abrupt, unexpected loss of cardiac function due to a life-threatening arrythia, and is usually fatal. Death from primary cardiac arrest accounts for about 10% of mortality in US adults.16 The proximate cause of fatal cardiac arrest is typically ventricular tacharrhythmia, which may be provoked by a series of triggering events acting on the for particulate myocardium.17 No biologic mechanism(s) air pollution effects on cardiac arrest have been elucidated, although some candidate models, such as electri? cal disburbances and inflammatory reactions, have been proposed.18'22 Ideally the exposure measures used in a cardiovascular health effects study will represent the biologically active area several PM exposure. In the Seattle metropolitan measurements are routinely collected. These include mea? gravimetric measures of PM10 and nephelometry sures of fine PM. Measures of light scattering from an are highly correlated with the integrating nephelometer of particles between 0.1 and 1.4 jutm mass concentration in diameter.23,24 Since the submicrometer fraction con? tains a large portion of the respirable particles and these have a composition that may affect health, the light scattering measurement should provide a useful exposure metric for health studies. We hypothesize that if the fine fraction of PM is the biologically active component, then the relative risks observed for light scattering will be stronger than that for PM10. This study assessed the association between PM and incidence of primary cardiac arrest, also known as sud? den cardiac death, in a well-defined existing case-series of primary cardiac arrests among persons without histo? ries of clinically detected cardiovascular disease, and who were free of other life-threatening conditions. We anticipated that the specificity of the case identification process for this group may facilitate inference about the mechanisms of possible underlying pathophysiologic PM. Our secondary objectives were to contrast associa? tions for two measures of PM and to evaluate potential effect modifiers.

1994- To be eligible for inclusion, cases had a sudden pulseless condition in the absence of a non-cardiac con? dition. We reviewed emergency medical service reports, death certificates, and (when available) medical examiner and autopsy reports to confirm the absence of noncardiac causes. We excluded cases if they had a history of clinically recognized heart disease (including angina, myocardial infarction, coronary artery bypass surgery, or angioplasty, congestive heart failure, arrhythmias, carvalvular, or congenital disease), or lifediomyopathy, comorbidities (cancer, or end-stage lung, threatening liver, or renal disease). The series was further restricted to 25- to 75-year-old married King County residents whose spouses participated in an in-person interview (83% of those eligible) to ensure uniform data collection for survivors and non-survivors of primary cardiac arrest. Air Pollution Data

Methods Study Subjects We used data from a population-based case-control study of out-of-hospital primary cardiac arrest conducted by Siscovick and colleagues.25 The 362 cases were a subset of all cases of paramedic-attended out-of-hospital primary cardiac arrest in Seattle and suburban King from October 3, 1988 to July 25, County, Washington

Air pollution data were obtained from the Puget Sound Clean Air Agency (PSCAA) for the study period (October 3, 1988 - July 25, 1994). Data for daily average temperature at Seattle-Tacoma airport were obtained from the National Oceanic and Atmospheric Administration. The primary exposure metric used in this study was 24-hour average particulate matter measured by nephe? lometry (reported as bsp in units of km'1, and referred to as the light scattering extinction We also coefficient). had gravimetric measures of PM10. Both were obtained from three King County monitoring sites: Duwamish, Lake Forest Park and Kent (Figure 1). We used the mean of daily average values from the three sites as our expo? sure measurement. Where data were missing for a par? ticular monitoring station on a given day, the values from the remaining monitors were used to compute the average. Missing data prevented us from using PM2 5 in measures, however, are a good analyses. Nephelometry data correlate surrogate for PM2 5 since nephelometry well with gravimetric particle measurements in the 0.11.4 aerodynamic diameter range23,24 and we have found nephelometry data to be highly correlated with PM2 5 in the greater Seattle metropolitan area.26 We obtained daily average sulfur dioxide (SOz) mea? surements from a monitor co-located with the PM mon? itors situated at the urban industrial Duwamish site. For carbon monoxide (CO), we combined daily averages from four street canyon locations into a single daily mean value. Ozone was measured only in the summer months and therefore we did not consider ozone in these analyses. We summarize distributions of all available air pollutant measurements and temperature in Table 2 and give correlations among all variables we consider in health effects analyses in Table 3. Further descriptive detail of these data can be found in Sheppard et al.26 Statistical Analysis

The case-crossover study design was proposed by Maclure27 to study the effects of transient, intermittent exposures on the subsequent risk of rare acute-onset

Epidemiology

March 2001, Vol. 12 No. 2

PM AIR POLLUTION AND CARDIAC ARREST

195

of Libraries GIS Mapcreatedby University Washington Datacopyright CHyofSeattle, 1909 .

FIGURE

1.

Map of air pollution monitors and case residences.

196

Levy et al

Epidemiology

March 2001, Vol. 12 No. 2

TABLE

1.

Characteristics of Cases

events in close temporal proximity to exposure. This design can be regarded as a special type of case-control study in which each case serves as his or her own referent. It requires exposure data only for cases. Expo? sures for each case during an "at risk" (index) period before the event are compared with the distribution of exposure during a referent period. The referent expo? sures should be representative of the expected distribu? tion of exposure for follow-up times that do not result in a case. Selection of the index period (usually a single day in air pollution studies) follows similar logic to conven? tional air pollution time-series studies - the index day may be the day of the event or some previous day allowing for a lag between exposure and manifestation of the event. The choice of referent days in air pollution research poses greater methodologic challenges because of the need to minimize multiple competing biases ow? in the air pollution time ing to lack of stationarity series.28'32 These include biases from long-term time in exposures, trends, seasonal patterns, autocorrelation work stemming and day-of-week effects. Methodologic from this study31,32 suggests that time should be stratified prior to analysis (eg, into separate months) and referents selected to be all days falling on the same day of the week within the same stratum as the index day. We performed conditional logistic regression to ob? tain estimates of relative risks and 95% confidence in? tervals associated with interquartile range (IQR) expo? sures from nephelometry (0.51 X IO'1 km'1 bsp) and PM10 (19.3 /xg/m3). We conducted separate analyses for lags of 0 through 5 days since the induction period for an association between PM and out-of-hospital sudden carTABLE 2. Distributions

diac arrest is unknown. We then selected a single lag for models of PM with either CO or S02. multipollutant We examined effect modification by considering cate? of selected variables, specifically age, current cig? gories arette smoke exposure, aspirin use, consumption of al? coholic beverages, long-chain N-3 polyunsaturated fatty acid consumption, physical activity, and a composite of risk factors quantifying risk for coronary heart disease. We classified current cigarette smoke exposure as cur? rent smokers and subjects exposed to passive smoking for one or more hours in an average week. Aspirin use pertained to subjects taking the equivalent of two or more aspirin tablets per week. Alcoholic beverage con? sumers imbibed one or more alcoholic beverages per day. We classified subjects with consumption above the me? dian of a Fish Intake Scale25 along with those who reported taking fish oil supplements as exposed to longchain N-3 polyunsaturated fatty acid. We classified cases as physically active if their average total kilocalories of recreational physical activity expended per week was greater than or equal to one.33 For indications of coro? nary heart disease risk, we included treatment for diabe? tes mellitus, high blood total cholesterol, or hyperten? sion, or family history of early myocardial infarction or sudden death (any parent or sibling experiencing myo? cardial infarction or sudden death before age 56 for males and before age 66 for females). We also assessed the role of time by examining effect modification by season and time categories (before or after the midpoint of the study period).

Results We assumed a linear exposure-effect model and esti? mated interquartile range relative risks and 95% confi? dence intervals for PM from light scattering for lags of 0 to 5 days. Effect estimates range from 0.89 (0 day lag; 95% CI = 0.78-1.02) to 1.01 (3 day lag; 95% CI = We did not observe any pattern over the set 0.90-1.12). of the lags suggestive of a relation between PM and the incidence of out-of-hospital primary cardiac arrest. The corresponding findings for PM10 display a generally sim? ilar pattern of no strong or consistent association with cardiac arrest. We selected lag 1 for subsequent analyses because, absent any compelling evidence for any other lag, it is the most proximal to the exposure among non-zero lags. At lag 0 we could not rule out that the time of the cardiac arrest may have preceded most of the

of Daily Means of Air Pollution Variables and Temperature

Epidemiology

March 2001, Vol. 12 No. 2

PM AIR POLLUTION AND CARDIAC ARREST

197

TABLE

3.

Pearson Correlation Coefficients Nephelometer 1 PM10

among Air Pollutants SOz CO

Variable Nephelometer PM10 so2 CO Temperature

for a nonlinear exposure. We assessed the potential exposure effect at lag 1 by estimating effects by quartile and by using regression spline smoothers with one and three knots. None of these models was an improvement over the linear exposure-effect model. Table 4 gives relative risk estimates for single- and models at lag 1. The two PM effect multi-pollutant estimates are essentially the same. There is no indication of association between either CO or S02 and out-ofhospital primary cardiac arrest. The multi-pollutant models do not give evidence of confounding of the PM effect by any of the pollutant variables. Table 5 summarizes the results of analyses for PM measured by nephelometry stratified by potential effect modifiers. We did not observe any apparent susceptible subgroups for the cardiovascular risk factors examined, in estimates although we did observe heterogeneity across seasons. The estimate for autumn was greater than one while those for the remaining seasons were less than one. If PM is indeed associated with primary cardiac arrest, we would expect at least as large relative risks to occur in winter owing to the prevalence of residential wood burning and the potentially greater toxicity of these particles. This expectation is not consistent with the results. We repeated the analyses for the remaining lags, for PM10, and for other categorizations of the effect modifiers. These results were not materially different from those reported in Table 5. Discussion The possibility of detrimental health effects from par? ticulate matter air pollution remains controversial. A large number of studies have found associations between PM and both morbidity and mortality outcomes. Most studies have linked daily variation in ambient PM con-

TABLE 4. Interquartile Range Relative Risk Estimates for PM from Light Scattering, PM10, CO, and SOz at lag 1

centrations to daily variation in total and cause-specific mortality, hospital Temperature admissions, or emergency department visits. For instance, Schwartz and Mor? ris11 observed an association between PM10 and hospital admissions for isch? emic heart disease, congestive heart failure and dysrhythmias in Detroit, Burnett et al.10 reported Michigan. that fine sulfate particles were associated with heart disease admissions in Ontario, Canada. Schwartz12 found associations between PM10 and cardiovascular disease admissions in Tucson, Arizona. He also observed that daily variation in PM10 was associated with hospital admissions for heart disease in eight US counties from various parts of the country, although the relation was weakest for Seattle among the eight metropolitan areas studied.13 His summary relative risk estimate was 1.025 (95% confidence limits = 1.018, 1.032) for a 25 jug/m3 increase in PM10. He posited that approximately 5% of hospital admissions for heart disease may be attributable to air pollution. The mechanisms of PM action on the cardiovascular system are uncertain. Pope et a!34 tested the hypothesis that exposure to PM acutely reduces blood oxygenation. While they did not observe an effect on blood oxygen? ation, they did find an association with slightly increased heart rate. The biological significance of this finding remains unclear. Liao et al35 reported transient reduc? tions in heart rate variability associated with increases in exposure to PM, suggesting a reduction in parasympathetic control of heart rate among elderly persons. Other hypotheses concerning effects of PM on the cardiovas? cular system have suggested ultrafine PM may induce alveolar inflammation, with the release of mediators capable of increasing blood coagubility, possibly increas? ing the risk for ischemic events.36 Peters et al31 found increases in plasma viscosity to be associated with an air in total sus? pollution episode marked by elevations and other pollutants. Their data, how? pended particles from ever, did not distinguish between contributions ultrafine particles, other air pollutants, or other unmea? sured confounding factors. Our results do not support an association between PM and out-of-hospital primary cardiac arrest. The patterns of results were nearly identical for PM measured by and gravimetrically (PM10). Furthermore, nephelometry we did not find evidence of effect modification by known personal risk factors for sudden cardiac arrest. Time-series studies of PM and cardiovascular disease are in the range of 1.4-4.2% increases in daily cardiovas? cular mortality for a 10 jLtg/m3 increase in PM10.13,38 Because we have a refined endpoint, however, we ex? pected a priori that the effect of PM on out-of-hospital primary cardiac arrest would show stronger associations. Instead, our results suggest there is no association be? tween PM and out-of-hospital primary cardiac arrest. The null findings were similar across the range of 0 to 5 day lags, although there were fluctuations of the effect estimates. While it is tempting to select the lag most

198

Levy et al

Epidemiology

March 2001, Vol. 12 No. 2

TABLE 5. Interquartile Range Relative Risks for PM from Light Scattering at a Lag of 1 Day, Stratified by Potential Effect Modifiers

* ETS,environmental tobacco smoke. of for t DM, history treatment diabetes of for mellitus; Chol, history treatment high cholesterol; Htn, of for of infarction sudden or history treatment hypertension; family FHx, death. history early myocardial

might be that the mechanisms of PMrelated cardiovascular toxicity do not involve short-term triggers that culminate in cardiac arrest. The most likely however, relates to the explanation, select group of study subjects highly comprising our case series. They were free of major comorbidity and any his? tory of clinical evidence of coronary artery disease.25 In contrast, Peters et al42 studied patients implanted with cardioverter defibrillators to assess whether potentially life-threatening arrhythmias are associated with partic? ulate air pollution episodes. They found an increase in N02 was associ? ated with increased tachycardia and ventricular fibrillation 2 days later, and that the most susceptible patients (those with repeated events) were es? ar? pecially at risk of experiencing rhythmia after increases in PM2 5 and NOz. Together these two studies sug? gest that air pollution effects on the risk of potentially life-threatening ar? rhythmia are more plausible in suscep? tible individuals, eg, people with a his? tory of severe cardiovascular disease. References

consistent with our prior hypothesis of a positive effect, the plausibility of selected positive associations must be evaluated in the context of all the effects estimated. Otherwise, we introduce model selection bias into our reported results.39 For instance, even though the point estimate of the relative risk for PM10 at lag 4 suggests an 8% increase in risk is possible, the lag 1 effect is even greater in absolute value, but for an inverse association. Furthermore, lag 1 represents the most plausible induc? tion period for PM health effects on cardiac arrest a priori. There are several plausible explanations for the ab? sence of an observed effect of PM on risk of primary cardiac arrest in this case series. Our study relied on daily city-wide exposure measurements. Seattle is topographically diverse and has localized PM sources from wood burning, particularly in the winter. While we found location effects on PM levels that varied with atmo? in a small exposure substudy, a respheric conditions fined analysis accounting for measurement error did not masked an suggest bias due to exposure misclassification association in this study.40 Furthermore, it may be pos? sible that exposures in Seattle are overall of the wrong or too low to cause an effect. Supporting composition evidence comes from a time-series analysis from the same location and general time period. That study is consistent with these case-crossover results; it showed no elevated risks related to PM for cardiovascular and isch? emic heart disease mortality.41 Another explanation

1. PopeCA, Schwartz Ransom MR.Dailymortality PM10 and in J, pollution Utah valley.ArchEnvHealth1992;47:211-217. 2. Dockery DW,PopeCA, XuX, Spengler Ware FayME,Ferris Jr, BG JD, JH, FE. between pollution mortality six U.S. air and in Speizer An association cites.N EnglJ Med1993;329:1753-1759. 3. Schwartz Air pollutionand dailymortality Birmingham, in Alabama. J. 1993;137:1136-1147. AmJ Epidemiol 4. PopeCA, BatesDV, Raizenne Healtheffectsof particulate pollu? ME. air tion:Timeforreassessment? 1996;7:225-230. Epidemiology 5. Verhoeff HoekG, Schwartz vanWijnen Airpollution daily AP, and J, JH. in mortality Amsterdam. 1996;7:225-230. Epidemiology 6. Zmirou Schwartz Saex M, Zanobetti Wojtyniak Touloumi D, J, A, B, G, SpixC, Poncede LeonA, Le Moullec Bacharova Schouten Ponka Y, L, J, K. of A, Katsouyanni Time-series and analysis airpollution cause-specific 1998;9:495-503. mortality. Epidemiology D. 7. Fairley Dailymortality airpollution SantaClara and in Califor? county, nia: 1989-1996.Environ HealthPersp1999:107:637-641. 8. Dominici Samet XuJ,Zeger Combining SL. F, evidence airpollution on JM, anddailymortality the twentylargest cities:A hierarchical from US mod? A ellingapproach.RoyalStat Soc Series 2000;163:263-302. J 9. Schwartz Dockery DW. Particulate pollution dailymortality air and J, in Ohio.Am J Epidemiol Steubenville, 1992;135:12-19. 10. Burnett DalesR, Krewski VincetR, DannT, Brook Associations RT, D, JF. between ambient sulfate admissions Ontario and to for particulate hospitals cardiac respiratory and diseases. J Epidemiol Am 1995;142:15-22. 11. Schwartz Morris Air pollution hospital R. and admissions cardiovas? for J, culardisease Detroit, in Am Michigan. J Epidemiol 1995;142:23-25. 12. Schwartz Airpollution hospital and admissions cardiovascular for J. disease in Tucson. 1997;8:371-377. Epidemiology 13. Schwartz Airpollution hospital and admissions heartdisease eight for in J. U.S. counties. 1999;10:17-22. Epidemiology L. 14. Sheppard Ecologic AH, study design.In:Shaarawi Piegorsch WW,eds. of Environmetrics. York: New Encyclopedia JohnWiley& Sons,in press. 15. Schwartz Whatarepeopledyingof on highpollution Environ J. Res days? 1994;64:26-35. 16. Kannel WB.Sudden death: Lessons subsets population from in studies. Am J Coli Cardiol B:141B-149B. 1985;suppl 17. Goldstein Bayes-de-Luna Guindo-SoldevilaSudden S, A, Cardiac Death. J. NY: Armonk, Furura 1994. Publishing,

Epidemiology

March 2001, Vol. 12 No. 2

PM AIR POLLUTION AND CARDIAC ARREST

199

of I. 18. Weisensee Bereiter-Hahn Schoeppe LoW'FriedrichEffects W, D, J, Int cardiac on of myocytes. J Immunocytokines the contractility cultured 1993;15:581-587. pharmacol as BD. basis 19. Verrier Nearing Electrophysiologic forT wavealternans an RL, fibrillation.Cardiovasc to indexof vulnerability ventricular J Electrophysiol 1994;5:445-461. 20. Killingsworth Lessandrini Murthy GGK,Catalano Paulauski CR, F, PJ, JD, and chemokine Godieski Inflammation, expression, deathin monocroJJ. Inhalation Toxicol fuel rats taline-treated following oil fly ash inhalation. 1997;9:541-565. 21. Clarke P, GGK,Sioutas Paulauski C, RW,Catalano Koutrakis Murthy J, PJ, alters inhalation Coull B, Ferguson Godieski Urbanair particulate S, JJ. in model inflammation a rodent and function induces pulmonary pulmonary Toxicol1999;11:637-656. of chronic bronchitis. Inhalation 22. ClarkeRW, CatalanoP, Coull B, Koutrakis Murthy GGK,Rice T, P, urban particles air in Godieski Age-related JJ. responses ratsto concentrated Toxicol2000;12:73-84. (CAPs).Inhalation 23. Ruby DL, E, AP, MG,RoodMJ, JG. Waggoner Robinson Blumethal Watson of and measurement scattering light coefficient Integrating nephelometer of In: fine particle concentration. Lodge et al, eds.Methods Air Sam? JP, Ml: 1989;450-457. Publishers, Chelsea, Lewis plingandAnalysis. and between 24- Thomas Gebhart Correlations A, J. gravimetric lightscattering AtmosEnv 1994;28:935-938. aerosols. for photometry atmospheric 25. SiscovickDS, Raghunathan King I, WeinmannS, Wicklund KG, TE, intake LH, P, V, J, Albirght Bovbjerg Arbogast SmithH, Kushi etal. Dietary N-3 levelsof long-chain polyunsaturated acidsand andcell membrane fatty cardiac arrest. AMA 1995;274:1363-1367. the riskof primary J of air 26. Sheppard LevyD, Norris Larson Koenig Effects ambient TV, L, G, JQ. admissions Seattle, in asthma on Washington, pollution nonelderly hospital 1987-1994.Epidemiology 1999;10:23-30. transient A M. 27. Maclure The case-crossover design: methodfor studying on effects the riskof acuteevents.Am J Epidemiol 1991;133:144-153. in and trends case-crossover caseand S. 28. Greenland Confounding exposure time-control 1996;7:231-239. designs. Epidemiology with time case-crossover 29. Navidi W. Bidirectional designsfor exposures trends. Biometrics 1998;54:596-605.

variation time trendin and 30. Bateson Schwartz Controlforseasonal TF, J. of studies acuteeffects environmental of case-crossover exposures. Epidemi? ology1999;10:539-544. for 31. Lumley LevyD. Biasin the case-crossover T, design: implications studies Environmetrics 11:689-704. of airpollution. 2000; selec? H. 32. LevyD, Lumley Sheppard Kaufman Checkoway Referent L, T, J, of tion in case-crossover analyses acute health effectsof air pollution. 2001;12:186-192. Epidemiology 33. Lemaitre Siscovick Raghunathan Weinman Arbogast Lin TE, S, P, RN, DS, arrest. D. Leisure-time cardiac Arch and activity the riskof primary physical Int Med1999;159:365-372. 34. Pope CA, Dockery DW, Kanner Villagas GM, Schwartz Oxygen RE, J. a air saturation, rateandparticulate pollution: dailytimeseries panel. pulse Am J RespCritCareMed1999;158:365-372. R. 35. Liao D, Creason Shy C, WilliamsR, WattsR, Zweidinger Daily J, controlin of and autonomic variation particulate pollution poorcardiac air Environ the elderly. HealthPerspect 1999;107:521-525. air 36. SeatonA, MacNeeW, Donaldson GoddenD. Particulate pollution K, Lancet1995;345:176-178. andacutehealtheffects. W. 37. Peters Doring Wichmann Koenig Increased A, HE, A, plasma viscosity Lancet1999;349:1582an a during airpollution episode: linkto mortality?. 1587. air effects particulate pollution. of 38. Dockery PopeCA. Acuterespiratory DW, Ann Rev Public Health1994;15:107-132. selection L. seasonal and 39. Lumley Sheppard Assessing T, confounding model control and anal? biasin airpollution using epidemiology positive negative 2000;11:705-717. yses.Environmetrics H. for 40. Sheppard LevyD, Checkoway Correcting the effectsof location L, in on and atmospheric conditions airpollution exposures a case-crossover (in J study. ExpAnalEnviron Epidemiol press). and Matter Pollution Air 41. LevyD. A Case-Crossover of Particulate Study PhD Cardiac Arrest KingCounty, in Washington. Primary Out-of-Hospital of Seattle,WA, 1999. Thesis,University Washington, 42. Peters LiuE,Verrier Schwartz GoldDR,Mittleman Baliff Oh M, A, RL, J, J, of and DW. Air pollution incidence K, JA, Allen G, Monahan Dockery cardiac 2000;11:11-17. Epidemiology arrhythmia.