You are on page 1of 5

Ambient Air Pollution and Oxygen Saturation

Dawn L. DeMeo, Antonella Zanobetti, Augusto A. Litonjua, Brent A. Coull, Joel Schwartz, and Diane R. Gold
Channing Laboratory, Department of Medicine, and Department of Pulmonary and Critical Care Medicine, Brigham and Womens Hospital, Harvard Medical School; and Exposure, Epidemiology, and Risk Program, Department of Environmental Health, Harvard School of Public Health, Boston, Massachusetts

We investigated the association between fine particulate air pollution and oxygen saturation as measured with a peripheral oxygen saturation monitor during a 12-week repeated-measures study of 28 older Boston residents. Oxygen saturation and air pollution particulates with a mean diameter less than or equal to 2.5 m were measured continuously during a protocol of rest, standing, exercise, postexercise rest, and 20 cycles of slow, paced breathing. In fixedeffect models, mean pollution concentration was associated with reduced oxygen saturation during the baseline rest period (6 hours: mean, 0.173%; 95% confidence interval [CI], 0.345 to 0.001), postexercise (6 hours: mean, 0.173%; 95% CI, 0.332 to 0.014), with a trend toward decrease during postexercise paced breathing (6 hours: mean, 0.142%; 95% CI, 0.292 to 0.007) but not during exercise. Participants taking -blockers had a greater pollutionrelated decrease in oxygen saturation at rest (6 hours: mean, 0.769%; 95% CI, 1.210 to 0.327) (interaction for particulates with a mean diameter less than or equal to 2.5 m by -blocker, p 0.0005) than did those not taking -blockers (p 0.25). The reduction in oxygen saturation associated with air pollution may result from subtle particulate-related pulmonary vascular and/or inflammatory changes. Further investigation may contribute to our understanding of the mechanisms through which particulates may increase respiratory and cardiac morbidity among vulnerable populations. Keywords: air pollution; epidemiology; respiratory physiology

individuals. Results from a similar analysis have been reported previously in the form of an abstract (6).

Recruitment of Participants
To examine associations of ambient pollution with outcomes including changes in oxygen saturation, heart rate variability, and electrocardiographic segment level, volunteers were recruited from a housing community in the city of Boston, Massachusetts between May and June of 1999, and if eligible were evaluated between July and August of 1999. A screening questionnaire provided information concerning demographics, self-reports of doctor-diagnosed pulmonary and cardiac disease, medication use, diet, and smoking behaviors. Exclusion criteria included unstable angina, arrhythmias (such as atrial utter or brillation), and the presence of a pacemaker. All participants were required to be able to ambulate on level ground.

Research Protocol
Human subject approval was obtained from the Institutional Review Board at Brigham and Womens Hospital (Boston, MA) and informed consent was obtained from all participants. If participants met inclusion criteria and agreed to participation, they were given a time and day of the week when weekly testing would be done, with the goal of 12 weeks of repeated-measures testing. Individual testing was performed Monday through Friday, between 9:00 a.m. and 2:00 p.m. Weekly questionnaires were completed concerning cardiac and pulmonary symptoms, selfreport of symptoms and physician-diagnosed disease, doctor and hospital visits, and medication use. The experimental protocol involved 5 minutes of rest in the seated position, 5 minutes of standing, 5 minutes of outdoor exercise (involving a 5-minute walk with a climb up a slight incline), 5 minutes of recovery during a postexercise rest period, and 5 minutes of slow, paced breathing. Paced breathing consisted of 20 cycles during which the participant was asked to breathe in for 5 seconds and then out for 5 seconds, as coached by a research assistant. Continuous oxygen saturation and pulse monitoring were performed during all parts of the protocol, using an oximeter (Nellcor, Pleasanton, CA). These measures, which represent instantaneous measures of oxygen saturation and pulse rate, were printed at 30-second intervals for all parts of the protocol on oximetergenerated data strips; these data strips included the protocol start and stop times and the date of testing. A total of 18,497 data points for oxygen saturation were analyzed.

Epidemiologic studies have demonstrated that particulate air pollution is associated with an increase in respiratory and cardiovascular morbidity and mortality (13). Pollution-associated hypoxia has been proposed as a mechanism through which pollution might increase the risk for acute coronary or other cardiovascular events (4, 5), but scant evidence of pollution-related hypoxia has been demonstrated. One repeated-measures study in Utah of elderly participants suggested no consistent effect of particulates with a mean aerodynamic diameter less than or equal to 10 m (PM10) on oxygen (O2) saturation. On stratifying the analysis by age, signicant negative association between O2 saturation and lagged pollution measures were observed in male participants 80 years of age and older (4). Our study investigated the association of longitudinal repeated measures of particulate air pollution and changes in oxygen saturation during a rest and exercise protocol in older urban dwelling individuals, adjusting for the inuence of individual comorbidities and medication use as potentially relevant factors inuencing oxygen saturation outcomes. We hypothesize that particulate air pollution has measurable effects on oxygen saturation in a panel of healthy, older

Exposure Monitoring
Hourly pollution data were collected in 1999 from June until the end of August. Within 1 km of the study site, continuous measurements were made of ne particles with a mean diameter of 2.5 m or less (PM2.5), using a tapered element oscillating microbalance (TEOM; Rupprecht & Patashnick, Albany, NY); carbon monoxide was measured with a model 48C gas analyzer (Thermo Electron, Waltham, MA). Ozone, nitrogen dioxide, sulfur dioxide, temperature, and relative humidity measures were obtained from the Massachusetts Department of Environmental Protection local monitoring site, 5 km away from the study site. Hourly meteorologic measures, such as mean temperature ( C), dew point temperature ( C), and barometric pressure (inches of mercury), were obtained from the National Weather Service at Logan Airport in East Boston, and were extracted from climatic records (EarthInfo, Inc., Boulder, CO).

(Received in original form February 25, 2004; accepted in final form May 11, 2004) Supported by HL07427 and K08 HL072918-01 (D.L.M.); also supported by EPA 826780-01-0 and NIH grant 1P01ES 09825-01. Correspondence and requests for reprints should be addressed to Dawn L. DeMeo, M.D., M.P.H., Channing Laboratory, 181 Longwood Avenue, Boston, MA 02115. E-mail:
Am J Respir Crit Care Med Vol 170. pp 383387, 2004 Originally Published in Press as DOI: 10.1164/rccm.200402-244OC on May 13, 2004 Internet address:


Characteristic Mean number of visits (median) Mean age, yr (range) Sex, n (%) Male Female Race/ethnicity, n (%) Black White Other Mean pack-years of cigarette smoking for smokers only (SD)* Number of ever-smokers* Number of never-smokers Physician diagnosis of: COPD, n (%) Asthma, n (%) Angina, n (%) Heart attack, n (%) Heart failure, n (%) Hypertension, n (%) Medication, n (%) Calcium channel blocker -Blocker ACE inhibitor Diuretic Self-rated health, n (%) Excellent Very good Good Fair Number (n 28)

Quality Assurance
The oximeter device has been manufactured to automatically perform calibration when switched on. Every 30-second percent values for oxygen saturation and absolute values for heart rate were printed out directly from the oximeter device, and double-entered into the data set. The data were checked for zero values suggesting that the oximeter lost the signal required to register accurate measures. As part of the cleaning process of these data, the zero value and the point above and below were excluded. Aberrant single-point changes of 5% or greater in oxygen saturation were excluded (e.g., we excluded a 30-second reading of 54% if the previous and subsequent 30-second readings were 100%).

11 (12) 73.3 (60.489.2) 7 (25.0) 21 (75.0) 8 (29) 19 (68) 1 (4) 36.6 (39.7) 13 11 2 1 3 3 2 11 3 5 8 7 7 6 11 4 (7) (4) (11) (11) (7) (41) (11) (18) (29) (25) (25) (22) (39) (14)

Statistical Analysis
In separate models for each portion of the protocol, we examined the association of pollution with the median oxygen saturation, treated as a continuous variable. Primary analyses were performed in Splus using generalized linear models. All models examining oxygen saturation as an outcome included date and time and smooth functions for temperature. Natural spline smoothing functions were used when predictor variables had previously been demonstrated to have nonlinear relationships with health outcomes. Single pollution models were evaluated, as were models that included carbon monoxide as a nonlinear predictor. All estimates are presented for the effect of PM2.5 on oxygen saturation as an outcome. These results are presented as a decrease in oxygen saturation for an interquartile increase in particulate air pollution, calculated by multiplication of the value for PM2.5 from the regression model by the interquartile range of air pollution during the specic lag or mean time frame. In the rst phase of analysis we used xed-effect modeling with an indicator variable for each participant in the study. With this approach individual participants served as their own controls. To examine the association of pollution with oxygen saturation over the entire study period, we also performed a nonstratied repeated-measure analysis using SAS (SAS Institute, Cary, NC), in which oxygen saturation at each portion of the protocol was considered a continuous outcome and part of the protocol (rst rest, standing, exercise, postexercise rest, and paced breathing) was a covariate in the analyses. We followed with a second phase of analysis using random effects models, to investigate the interaction between covariates and PM2.5. This part of the analysis was to obtain specic effect estimates for measured covariates and to examine interactions between time-varying or time-invariant covariates and PM2.5, with a focus on the interaction between air pollution and medication use. These models included a random intercept and a random slope to explain the heterogeneity by subject and the heterogeneity in response to air pollution, after accounting for differences in medication use.

Definition of abbreviations: ACE angiotensin-converting enzyme; COPD chronic obstructive pulmonary disease. * Includes one current smoker; smoking information missing for four participants.

Characteristics of the 28 individuals who completed more than 1 visit are demonstrated in Table 1. The majority of participants were elderly white females; all participants had normal oxygen saturations at rest. Approximately 10% reported a history of angina, heart attack, or heart failure. The mean pack-years of cigarette smokers was 36.6 pack-years, among those classied as ever having smoked. Fewer than 10% had a diagnosed lung disease, although about 40% of the participants had a history of hypertension. All individuals taking -blockers (n 5) were taking at least one other antihypertensive agent (three combined diuretic with -blocker; one angiotensin-converting enzyme inhibitor with blocker; and one angiotensin-converting enzyme, -blocker, and calcium blocker combination). For rst rest, postexercise rest, and paced breathing, but not for the exercise portion of the protocol, small, often statistically signicant inverse associations were detected with individual hourly lags of PM2.5 concentrations measured between 2 and 7 hours before exercise, using xed-effect models. For the rst rest, this effect was most pronounced at 6 hours, demonstrating a 0.172% (95% condence interval [CI]: 0.313 to 0.031) decre-

ment in oxygen saturation for each 11.45- g/m3 increase (1 interquartile range) in PM2.5. The trend for the variation in oxygen saturation during the rst rest by the individual lags for air pollution is presented in Figure 1. For rst rest, postexercise rest, and paced breathing, the mean 6-, 12-, 24-, and 48-hour PM2.5 concentrations also tended to be associated with a small reduction in oxygen saturation. However, during exercise no signicant reduction in oxygen saturation was observed. In the postexercise rest period that followed exercise, the effect for the 6-hour mean returned to the level observed during the rst rest period (Table 2). When oxygen saturation was considered for the overall protocol (and not stratied by part of the experimental protocol), there was a signicant inverse association of increasing PM2.5 on oxygen saturation (Table 2). The effect of PM2.5 on oxygen saturation was modied by the use of -blocker medications (p value for PM2.5 by -blocker interaction, less than 0.0005). Those taking -blockers had the most signicant decrease in oxygen saturation at rest. For each increase in PM2.5 of 13.42 g/m3 (1 interquartile range), there was a signicant decrease in 6-hour mean oxygen saturation of 0.769% (95% CI: 1.21 to 0.327); the decrease was not signicant in those individuals not taking -blockers (mean, 0.062%; 95% CI: 0.248 to 0.123; p value for interaction, greater than 0.25). This nding was not observed during other parts of the protocol. Compared with those not taking -blocker medications, the ve individuals taking -blocker medications had higher mean resting supine and exercise systolic blood pressures (132 vs. 125 mm Hg during resting supine measurement and 155 vs. 139 mm Hg during exercise) and overall mean lower

DeMeo, Zanobetti, Litonjua, et al.: Air Pollution and Oxygen Saturation

385 Figure 1. Variation in oxygen saturation during the first rest period versus individual hourly lag measurements for particulates with a mean diameter less than or equal to 2.5 m (PM2.5). On the y axis, EFFECT represents the effect estimate for the oxygen saturation outcome as calculated by multiplying the for air pollution from the multivariate regression models by the interquartile range for air pollution at the specific time lag. Hourly time lags are presented on the x axis. Of note, Hour 1 is Lag 0 for air pollution, Hour 7 is Lag 6 for air pollution, and so on. Confidence intervals (CI) are represented by the length of the line extending through the point estimate, with the bounds of the CI denoted by a dash.

heart rates during the protocol (65 vs. 80 beats per minute). The effect of PM2.5 on oxygen saturation was not modied by age, sex, smoking status or smoking history, self-report of hypertension, or other self-reported diagnoses.

In a repeated-measures panel study of older individuals studied during the summer of 1999, we found a small but statistically signicant inverse association of ambient PM2.5 level and oxygen saturation. Older individuals (65 years of age or older) are potentially more susceptible to the health effects of air pollution (7) and represent an important vulnerable group to include in studies of pollution-related health effects. Past studies have demonstrated the associations of particulate air pollution with cardiac

and respiratory morbidity and mortality, but epidemiologic evidence of particulate-related decreases in oxygen saturation as an explanation for these associations has been scant. Pope and colleagues have demonstrated a statistically signicant negative association between PM10 and oxygen saturation among elderly, male individuals who were at least 80 years of age. They concluded that the effects of particulates on hypoxia might be most relevant in older and sicker individuals (4). In this study, Pope and colleagues (4) suggested that a study design that relies on the pulse oximeter for readings is robust. Past studies have investigated clinical characteristics that may designate an increased susceptibility to air pollution, such as chronic obstructive pulmonary disease (8), coronary artery disease (9), arrhythmias (10), and congestive heart failure (9). These

Protocol Rest Hour 6 12 24 48 6 12 24 48 6 12 24 48 6 12 24 48 6 12 24 IQR ( g/m3) 13.42 10.81 10.26 10.57 13.42 10.81 10.26 10.57 13.42 10.81 10.26 10.57 13.42 10.81 10.26 10.57 13.42 10.81 10.26 Percent Change 0.173 0.160 0.169 0.153 0.005 0.014 0.001 0.011 0.173 0.128 0.113 0.157 0.142 0.139 0.121 0.082 0.131 0.120 0.112 95% CI 0.345 0.308 0.316 0.304 0.215 0.196 0.180 0.196 0.332 0.266 0.250 0.295 0.292 0.269 0.248 0.211 0.247 0.221 0.212 to to to to to to to to to to to to to to to to to to to 0.001 0.012 0.022 0.002 0.205 0.168 0.182 0.174 0.014 0.010 0.023 0.019 0.007 0.010 0.007 0.047 0.015 0.020 0.013


Postexercise rest

Paced breathing

Summary over protocol

Definition of abbreviations: CI confidence interval; IQR interquartile range. * Baseline. Pollution data represent hourly means at 6, 12, 24, and 48 hours.



studies have been performed in cohorts of frail individuals. De Leon and colleagues investigated the association between PM10 and mortality in New York City between 1985 and 1994. In participants older than 75 years of age, they observed that individuals with respiratory disease had a higher mortality from circulatory deaths, leading to the conclusion that older individuals may be more at risk for deleterious effects of air pollution in the setting of respiratory disease (11). Our cohort represents a population of older, independent individuals, most of whom rate their health as good, very good, or excellent. However, the ndings of effect modication in this cohort by -blocker use but not by other cardiac or pulmonary factors may dene a true subpopulation at risk. All the individuals taking -blockers were administered more than one antihypertensive medication and demonstrated higher systolic blood pressure than the remainder of the cohort, possibly representing a group with more difcultto-control blood pressure. Such a group may be more susceptible to pollution-related vascular effects of inammation, leading to subtle changes in oxygenation. In our participants, the magnitude of the decline in oxygen saturation is small, potentially representing a subtle local airway and alveolar inammatory effect. Conversely, these effects might be greater and more clinically relevant in a more debilitated population on -blockers. The negative association between oxygen saturation and PM2.5 is not apparent during exercise and suggests we are dealing with a relatively healthy group, or may be due to the noise associated with peripheral oxygen saturation measurements during rapid movement. Strengths of this study include the repeated measures for oxygen saturation both during the experimental protocol as well as through longitudinal follow-up during 12 summer weeks, with varying levels of particulate exposures. There was little to no intercurrent illness and thus each subject acted effectively as his or her own control during the study. Potential limitations include the lack of pulmonary function data, which would identify the presence of occult pulmonary disease that may contribute to the nding of decreased oxygen saturation. We do not have information about environmental tobacco smoke exposure in the home (but the majority of the participants lived alone and were nonsmokers), nor was direct monitoring of particulates in the home performed as part of this study. As well, we do not have measures of arterial oxygen levels, but we believe that the investigation of changes in oximetry provided a reliable and reproducible assessment for this epidemiologic investigation. Nine individuals completed only one visit and were excluded from the analysis. However, comparison of those who completed only one visit with those who completed more that one visit did not reveal any systematic differences between the two groups. Pollution-associated airway and alveolar endothelial inammation may result in impairment of oxygen diffusion and subtle decrements in oxygen saturation levels due to edema formation. These subtle oxygen saturation changes may directly inuence cardiovascular responses, or, more likely, the process involved in producing these small decreases in oxygen saturation may lead to subsequent adverse cardiovascular physiology. The local inammatory process associated with small changes in oxygen saturation may lead to systemic inammation capable of inuencing cardiovascular outcomes. In another study we found that elevated ambient particle levels predicted increased exhaled nitric oxide, a marker of pulmonary inammation (12). Seaton and colleagues suggested that particulate air pollution might inuence alveolar inammation. They also suggested that particles lead to inammatory mediator release, potentially exacerbating lung disease and hypercoagulability (3). Pulmonary inammatory effects from particulate air pollution have been observed in bronchoalveolar lavage and mucosal biopsy speci-

mens in healthy young people exposed to diesel exhaust, with evidence of upregulation of cytokines (such as interleukin-8). Increased expression of endothelial adhesion molecules and increased neutrophils, mast cells, and lymphocyte burden in the lung have also been observed in response to diesel exhaust exposure (13, 14). Human investigations into the systemic inammatory effects of particulate air pollution have demonstrated increased C-reactive protein (15), plasma brinogen (16), and plasma viscosity (17) in association with air pollution. The relative decrease in oxygen saturation may also be a marker of oxidative stress that may result in neural feedback to the heart, or may alter cardiovascular physiology through changes in local pulmonary and cardiac vascular inammation and blood coagulability. Particulate pollution-related pulmonary inammation and cytokine cascades may inuence autonomic responses via stimulation of pulmonary vagal receptors. In multiple epidemiologic investigations, air pollution has been associated with increased heart rate (4, 18) and decreased heart rate variability (1921). Recent studies on air pollution and heart rate have suggested that autonomic imbalances, as evidenced by increases in heart rate and decreases in heart rate variability, may specically contribute to the increased mortality (1921). In conclusion, we demonstrate a statistically signicant effect of ambient particulate air pollution on decreased oxygen saturation at rest in a population of free-living older individuals, with a more signicant interaction in the individuals taking -blockers. The biological and/or clinical relevance of this magnitude of effect in terms of hypoxia may be small, but we hypothesize that this nding reects subtle effects of the inammatory cascade, which results from exposure to ambient particulate pollution. Further investigation is needed to characterize susceptibility characteristics and to elucidate the functional/mechanistic signicance of these ndings.
Conflict of Interest Statement : D.L.D. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; A.Z. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; A.A.L. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; B.A.C. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; J.S. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript; D.R.G. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript.

1. Dockery DW, Pope CA III, Xu X, Spengler JD, Ware JH, Fay ME, Ferris BG Jr, Speizer FE. An association between air pollution and mortality in six U.S. cities. N Engl J Med 1993;329:17531759. 2. Samet JM, Dominici F, Curriero FC, Coursac I, Zeger SL. Fine particulate air pollution and mortality in 20 U.S. cities, 19871994. N Engl J Med 2000;343:17421749. 3. Seaton A, MacNee W, Donaldson K, Godden D. Particulate air pollution and acute health effects. Lancet 1995;345:176178. 4. Pope CA, Dockery DW, Kanner RE, Villegas GM, Schwartz J. Oxygen saturation, pulse rate, and particulate air pollution: a daily time-series panel study. Am J Respir Crit Care Med 1999;159:365372. 5. Schwartz J, Morris R. Air pollution and hospital admissions for cardiovascular disease in Detroit, Michigan. Am J Epidemiol 1995;142:2335. 6. DeMeo DL, Schwartz J, Jacobsen M, Gold D. Longitudinal analysis of ambient pollution and oxygen saturation in a cohort of elderly individuals [abstract]. Am J Respir Crit Care Med 2002;165:A306. 7. Aga E, Samoli E, Touloumi G, Anderson HR, Cadum E, Forsberg B, Goodman P, Goren A, Kotesovec F, Kriz B, et al. Short-term effects of ambient particles on mortality in the elderly: results from 28 cities in the APHEA2 project. Eur Respir J Suppl 2003;40:28s33s. 8. Zanobetti A, Schwartz J, Gold D. Are there sensitive subgroups for the effects of airborne particles? Environ Health Perspect 2000;108:841 845.

DeMeo, Zanobetti, Litonjua, et al.: Air Pollution and Oxygen Saturation

9. Goldberg MS, Bailar JC III, Burnett RT, Brook JR, Tamblyn R, Bonvalot Y, Ernst P, Flegel KM, Singh RK, Valois MF. Identifying subgroups of the general population that may be susceptible to short-term increases in particulate air pollution: a time-series study in Montreal, Quebec. Res Rep Health Eff Inst 2000;97:7113; discussion 115120. 10. Kwon HJ, Cho SH, Nyberg F, Pershagen G. Effects of ambient air pollution on daily mortality in a cohort of patients with congestive heart failure. Epidemiology 2001;12:413419. 11. De Leon SF, Thurston GD, Ito K. Contribution of respiratory disease to nonrespiratory mortality associations with air pollution. Am J Respir Crit Care Med 2003;167:11171123. 12. Adamkiewicz G, Ebelt S, Syring M, Slater J, Speizer FE, Schwartz J, Suh H, Gold DR. Association between air pollution exposure and exhaled nitric oxide in an elderly population. Thorax 2004;59:204209. 13. Salvi SS, Nordenhall C, Blomberg A, Rudell B, Pourazar J, Kelly FJ, Wilson S, Sandstrom T, Holgate ST, Frew AJ. Acute exposure to diesel exhaust increases IL-8 and GRO- production in healthy human airways. Am J Respir Crit Care Med 2000;161:550557. 14. Salvi S, Blomberg A, Rudell B, Kelly F, Sandstrom T, Holgate ST, Frew A. Acute inammatory 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;159:702709.

15. Peters A, Frohlich M, Doring A, Immervoll T, Wichmann HE, Hutchinson WL, Pepys MB, Koenig W. Particulate air pollution is associated with an acute phase response in men: results from the MONICAAugsburg Study. Eur Heart J 2001;22:11981204. 16. Schwartz J. Air pollution and blood markers of cardiovascular risk. Environ Health Perspect 2001;109:405409. 17. Peters A, Doring A, Wichmann HE, Koenig W. Increased plasma viscosity during an air pollution episode: a link to mortality? Lancet 1997; 349:15821587. 18. Peters A, Perz S, Doring A, Stieber J, Koenig W, Wichmann HE. Increases in heart rate during an air pollution episode. Am J Epidemiol 1999;150:10941098. 19. Pope CA III, Verrier RL, Lovett EG, Larson AC, Raizenne ME, Kanner RE, Schwartz J, Villegas GM, Gold DR, Dockery DW. Heart rate variability associated with particulate air pollution. Am Heart J 1999; 138:890899. 20. Gold DR, Litonjua A, Schwartz J, Lovett E, Larson A, Nearing B, Allen G, Verrier M, Cherry R, Verrier R. Ambient pollution and heart rate variability. Circulation 2000;101:12671273. 21. 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;107:521525.