INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Increases in respiratory and cardiovascular morbidity and mortality that accompanied episodes of extreme air pollution earlier in this century demonstrated that air pollution can adversely affect human health (1). More recent investigations of health effects of particulate air pollution include daily time-series studies that observed changes in daily mortality associated with short-term changes in particulate air pollution at relatively low concentrations. These pollution-mortality studies have been conducted in more than 30 cities in the United States, Europe, and other parts of the world (1). When cause of death was evaluated, particulate air pollution was associated with both respiratory and cardiovascular mortality, but generally not with mortality due to other causes. Recent prospective cohort studies observed that long-term exposure to fine particulate air pollution was associated with an increased risk of mortality, primarily cardiopulmonary mortality (4, 5). Epidemiological investigations have also observed that particulate pollution is associated with respiratory and cardiovascular hospital admissions (6).

Several reviews of the epidemiological evidence linking particle air pollution with cardiopulmonary mortality and morbidity suggest that the estimated effects have been reasonably consistent and coherent (1). It has been hypothesized that particulate air pollution (especially ultrafine particles) may provoke alveolar inflammation, resulting in the release of potentially harmful cytokines (3). Nevertheless, the underlying mechanisms or pathways of particle-induced mortality and morbidity still remain largely unknown. A general pathophysiological pathway linking particulate air pollution and cardiopulmonary death might be pollution-induced lung damage (potentially including oxidative lung damage and inflammation) declines in lung function, respiratory distress, and hypoxia. Recent studies have reported evidence of particulate-related pulmonary inflammation (10) and small transient decreases in lung function measured daily in panels of children and asthmatic patients (11). Studies of pollution-related hypoxia or elevated pulse rate after exposures to particulate air pollution have not been reported.

The primary objective of this study was to evaluate potential associations of daily measures of respirable particulate air pollution with acute changes in oxygen saturation of the blood (measured by pulse oximetry, Sp_O2) in panels of elderly adults. A secondary objective was to evaluate potential associations between pollution exposure and changes in pulse rate. The performance of the pulse oximeters and the ability to observe daily changes in oxygen saturation and pulse rate associated with changes in barometric pressure were also evaluated.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Period and Area

The data collection period for this study was from November 18, 1995 through March 15, 1996. This study was conducted in Utah Valley in central Utah (Figure 1). During low-level temperature inversion episodes common to winter months, particulate concentrations become elevated when local emissions become trapped in a stagnant air mass near the valley floor. Sources of particulate air pollution included an integrated steel mill, motor vehicle emissions, and wood smoke. The study area is approximately 1,400 m above sea level. In elderly people at this altitude, normal partial pressure of oxygen in arterial blood is 55 to 67 mm Hg (16). This may provide a better opportunity to test for acute effects on oxygen saturation as declines in partial pressure of oxygen (PO₂) will cause greater relative declines in Sp_O2 than would be seen at lower altitudes owing to the shape of the oxyhemoglobin dissociation curve.

View larger version (63K): [in this window] [in a new window]   Figure 1.   Study area, Utah Valley.

Subject Selection and Recruitment

Two panels were selected for the study. The first panel was a "private home panel" which included 50 retired Brigham Young University faculty, staff, and spouses plus two additional participants who met the eligibility requirements. Participation was limited to elderly persons who met the following eligibility requirements: (1) lived with spouse; (2) were willing to participate; (3) lived near air pollution monitors for PM₁₀ (particles with an aerodynamic diameter  a nominal 10 µm) (Figure 1); (4) had homes without special air filtration systems; (5) were nonsmokers and had smoked less than 10 packs of cigarettes in their lifetime; and (6) had no serious medical condition that would preclude their participation, including mental illness, being currently under oxygen therapy, or having had heart failure in the last 6 mo. The number of home sites equaled 26 with 52 participants in the private home panel.

The second panel was a "retirement home panel" consisting of residents of a retirement home, located east of the steel mill and the Orem PM₁₀ monitor (Figure 1). All residents of this home were nonsmokers. They were retired elderly individuals who were relatively independent and not in need of regular medical/nursing care. Residents lived in one large building as couples or individually in separate units that included bedrooms, a bathroom, a kitchen/dining area, and a living room. Although they could eat in their own units, most of the residents ate together in a common dining area. Thirty-eight of these residents were also selected as participants. A much more detailed description of the study is provided in a report to the Health Effects Institute (17).

Materials and Methods

Research protocol and consent forms were approved by institutional review boards for human subjects at Brigham Young University, University of Utah School of Medicine, and Harvard School of Public Health. Each home received a Nellcor N-20P pulse oximeter with finger sensor and printer (Nellcor Inc., Hayward, CA). This pulse oximeter measures Sp_O2 based on the transmission of light in two wavelengths through a vascular bed, the index finger. The pulsatile flow of blood through the tissue bed modulates the light reaching the detector. Some measurement error can be produced by dyshemoglobins, such as carboxyhemoglobin, or conditions that reduce the amplitude of the arterial pulsations, such as hypertension or ischemia. Pulse rate is also measured. The pulse oximeters printed out both Sp_O2 and pulse rate. Participants also received a customized bound folder with health diaries, envelopes for oximeter printouts, copies of consent forms, and necessary information and instructions. Participants measured their Sp_O2 and pulse rate twice daily (unless out of town), shortly after rising in the morning and before breakfast, and before retiring to bed. Also, each night before retiring to bed they completed their health diaries indicating the presence of various symptoms.

When measuring Sp_O2 and pulse rate, participants sampled in the continuous operating mode for approximately 3 min. At the end of the sample, average pulse rate and Sp_O2 over the entire interval were printed along with date and time. After writing their name or an identification number on the printout, the printouts were placed in the envelope attached to their diary. Before starting each test, participants were instructed to remain sitting at rest for at least 1 min and to remain still throughout the test. They were also instructed to use the left index finger to keep uniformity and to avoid the presence of bright light sources. Some participants had to remove fingernail polish. Approximately 7 d after the initial visit, research assistants made follow-up visits to interview the participants, evaluate compliance, test the performance of the oximeter, identify any problems, and collect the printouts from the oximeters. Follow-up visits for the same purposes were conducted every 2 wk throughout the study period.

Similar protocols were used for measuring Sp_O2 levels and pulse rates of participants at the retirement home. The primary differences were that these participants only took measurements once each day between 5:00 and 7:00 P.M., and they were always assisted by a research assistant assigned to the retirement home. The research assistant would set up three pulse oximeters in a room adjacent to the dining area and participants would stop in before or after their evening meal. The subjects would be seated and have the 3-min measurement taken. The oximeter printout would then be labeled and filed by the research assistant. Symptom diaries were not completed for the retirement home panel.

Quality Control

The Nellcor N-20P pulse oximeter determines pulse rate and functional Sp_O2 (oxygenated hemoglobin expressed as a percentage of the hemoglobin that is capable of transporting oxygen). It is automatically calibrated each time it is turned on or whenever a new sensor is connected. The oximeter sets sensor-specific calibration coefficients by reading a calibration resistor in the sensor. Each pulse oximeter was tested for accuracy once each month with an SRC-2 pulse oximeter tester (Nellcor Inc.). Also, during follow-up visits a control pulse oximeter was brought to each home and a 3-min test was run on each participant using the control oximeter on the right hand's index finger, and the participant's oximeter on the left hand's index finger. Both oximeter readings were printed and compared for consistency. After collection, data were sorted, checked, and photocopied, and entered as computer files. Computer data files were proofread independently twice and checked by computer for consistency and congruity with known parameters.

Pollution and Weather Data

Air pollution monitoring was conducted by the Utah State Department of Health. Daily PM₁₀ data were collected in accordance with the U.S. Environmental Protection Agency's reference method (18) at three sites in the valleyLindon, Provo, and Orem monitoring sites (Figure 1). Measures of PM₁₀ were available for 85%, 90%, and 94% of the days at the Lindon, Orem, and Provo monitoring sites, respectively. Monitoring of carbon monoxide (CO) was conducted at five sites in the valley, including daily 24-h average and 8-h high CO levels. Limited monitoring indicated very low concentrations of sulfur dioxide (SO₂), nitrogen dioxide (NO₂), and wintertime ozone (O₃).

Weather data were obtained from a weather station at Brigham Young University (Figure 1). High and low temperatures, high and low relative humidity, and barometric pressure at 5:00 P.M. local time were collected for each day in the study period. Barometric pressure readings were reported by the weather station in inches of mercury adjusted to sea level. For this analysis, barometric pressure values were back converted to millimeters of mercury (mm Hg) indicating actual barometric pressure at the valley floor (i.e., not adjusted to sea level values).

Statistical Methods

Key variables were plotted together over time. Pearson correlation coefficients between Sp_O2, pulse rate, pollution variables, and weather variables were calculated to evaluate simple pairwise correlations. PM₁₀ levels at the three different monitoring sites were similar and highly correlated over the study period. Therefore, in most of the analytic plots and statistical analyses, PM₁₀ levels averaged across available data from the three sites were used except for 2 d with no monitored data. On these 2 d PM₁₀ was estimated by simple linear extrapolation.

Fixed-effects autoregressive multiple regression models were estimated. Values of Sp_O2 and pulse rate were regressed on indicator variables for each participant and the time-dependent covariates including pollution, barometric pressure, and other weather variables. Temperature and relative humidity were divided into quintile ranges, and indicator variables for quintiles of temperature and relative humidity were included in the models. To evaluate potential lagged relationships, various lagged moving average models were estimated. Because statistically significant (p < 0.05) autocorrelation was observed, first-order autoregressive models were estimated using the maximum likelihood estimation method (19). Models were also estimated after stratifying the data by health status, sex, and age.

The mean pulse rate across the full study period was calculated for each participant. Binary variables were created to indicate days when pulse rate was elevated by more than 5 or 10 beats per minute (beats/ min) greater than the study period mean for the participant. These binary high pulse rate event variables and symptom variables were analyzed by estimating fixed effects logistic regression models with the same control variables as described earlier (20).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Compliance and Oximeter Performance

Table 1 provides summary information about the participants in the study and oximetry data collected. The participants were elderly and the majority suffered from heart or lung disease, or both. Although daily symptoms data were not collected for the retirement home participants, daily health diary data collected from the home panel reflected chronic disease with little daily variability in symptoms. Based on logistic regressions using these limited daily symptoms data, statistically significant positive associations between symptoms and PM₁₀ were not consistently observed (17). The Sp_O2, and pulse rate measures for the private home panel were 88% complete. Measures for the retirement home panel were 74% complete. There was never an occasion during the study period when there was evidence that any of the oximeters failed to operate properly. When coreadings of Sp_O2 and pulse rate with the study oximeters and the control oximeter were compared, the readings were identical or only one digit different 93 and 99% of the time, respectively. Never were the Sp_O2 or pulse rate control comparisons different by more than 3% or 3 beats/ min, respectively. Control tests using the Nellcor SRC-2 pulse oximeter tester indicated that, based on manufacturer's specifications, none of the oximeters ever failed the quality control tests for either Sp_O2 or pulse rate.

Figure 2 presents plots of daily concentrations of PM₁₀ at each of the three monitors and CO concentrations at the South Orem monitor over the study period. Also presented in this figure are barometric pressure and the mean nighttime/ evening Sp_O2 readings for participants in both panels. As can be seen in Figure 2, the PM₁₀ levels at the three monitoring sites were nearly the same and were highly correlated. Pearson correlation coefficients (r) between PM₁₀ levels at the three sites equaled 0.92 to 0.96. PM₁₀ was less strongly correlated with carbon monoxide levels (r = 0.61 to 0.76) and barometric pressure (r = 0.38 to 0.48).

View larger version (42K): [in this window] [in a new window]   Figure 2.   PM10, CO, barometric pressure, and mean SpO2 values plotted over days in the study period.

Associations with Oxygen Saturation

An association between mean Sp_O2 and barometric pressure was observed (Figure 2). The most robust result for Sp_O2 from the various regression models was the positive highly statistically significant relationship between Sp_O2 and barometric pressure (Table 2). This association was consistent across models and was not affected by the inclusion of other covariates in the models. The estimated effect of barometric pressure (25 mm Hg) on Sp_O2, depending on other covariates included in the models, ranged from 0.59% (SE = 0.08) to 0.69% (SE = 0.07). Also, no lag structure was observed for barometric pressure. The estimated effect of barometric pressure on Sp_O2 appeared to be due to concurrent day's barometric pressure.

Consistent, statistically significant associations between Sp_O2 and PM₁₀ were not observed. Concurrent day PM₁₀ levels were often positively associated with Sp_O2. Measured Sp_O2 was generally negatively associated with previous day and 5-d lagged moving average PM₁₀, but the estimated effect was small (Table 2) and generally not statistically significant. For some of the strata, most notably male participants who were 80 yr old or older, the negative association between Sp_O2 and PM₁₀ was statistically significant. Estimated effects of PM₁₀ or barometric pressure on Sp_O2 were not sensitive to the inclusion of weather variables in the regression models.

Including CO in the regression models resulted in slightly larger estimated effects of PM₁₀ on Sp_O2 but the coefficients remained small and statistically insignificant. Coefficients on CO were positive, but small (approximately  0.02) and statistically insignificant (p > 0.10). Exposure to CO should falsely raise Sp_O2 slightly, as the oximeter cannot read carboxyhemoglobin and ignores it and, in addition, CO shifts the oxyhemoglobin dissociation curve to the left which increases Sp_O2 for a given PO₂.

Associations with Pulse Rate

Figure 3 presents daily mean pulse rate for all the participants and the proportion of the participants with pulse rates more than 10 beats/min above their study-period mean pulse rate. These are both plotted with the previous day's PM₁₀ concentrations across the days in the study period. A careful examination of Figure 3 indicates that there is much variability in pulse rate but also suggests that a small subtle positive association between pulse rate and PM₁₀ concentrations may exist. Regression analysis confirms the existence of a small but statistically significant association between pulse rate and PM₁₀ exposure. Table 2 presents the results of various regression models for pulse rate where PM₁₀ is included in the models as concurrent day PM₁₀, previous day PM₁₀, and different lagged moving averages of PM₁₀. Table 3 presents estimated odds ratios for pulse rate being elevated by 5 or 10 beats/min.

View larger version (46K): [in this window] [in a new window]   Figure 3.   Daily mean pulse rate for all participants and the proportion of participants with pulse rates more than 10 beats/min above their study-period mean plotted with the previous day's PM10 levels across the day in the study period.

A consistent result from the regression models was the negative statistically significant relationship between pulse rate and barometric pressure. A 25 mm Hg increase in barometric pressure was associated with an average decrease in pulse rate equal to approximately 1.5 beats/min. The estimated effect of barometric pressure on pulse rate was not highly affected by the inclusion of pollution variables in the models. The estimated effect of barometric pressure (25 mm Hg) without any pollution variable was 1.27 (SE = 0.47). Also, no clear lag structure was observed between barometric pressure and pulse rate. Barometric pressure's estimated effect on pulse rate appeared to be due to the concurrent day's barometric pressure.

Positive associations between pulse rate and PM₁₀ were observed. Daily pulse rate was associated with exposure to particulate pollution on the previous 1 to 5 d. A 100 µg/m³ increase in previous day PM₁₀ was associated with an average increase of 0.78 beats/min. Also, the odds of having pulse rate elevated by 5 or 10 beats/min was associated with exposure to particulate air pollution levels on the previous 1 to 5 d. A 100 µg/m³ increase in previous day PM₁₀ was associated with an increase of 29% and 95% in the odds of experiencing an elevated pulse rate of 5 or 10 beats/min, respectively (Table 3). As with the analysis for Sp_O2, controlling for weather variables in the regression models had little impact on the estimated effects of PM₁₀ or barometric pressure on pulse rate. Including concurrent-day CO in the regression models resulted in mostly unchanged or somewhat larger estimated effects of PM₁₀ on pulse rate with negative but statistically insignificant coefficients on CO. When lagged CO levels were included in the models with PM₁₀, the coefficients for both CO and PM₁₀ were less stable, reflecting the collinearity between these two pollutants. Also, the estimated PM₁₀ effect was generally strengthened with adjustment for barometric pressure.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Over the last several years, various studies specific to Utah Valley have evaluated the health effects of air pollution (21). As with the present study, previous studies used central site fixed monitors to measure ambient air pollution. Changes in ambient particulate air pollution levels at least partly reflect differences in personal exposures over time. Pollution levels at PM₁₀ monitoring sites are highly correlated, suggesting that fine particulate air pollution is dispersed across the study area. Although the elderly participants spend most of their time in their homes, outdoor/in-home differences in particulate concentrations were reduced by including only homes without any smokers (including the retirement home) and homes without special air filtration systems.

This study evaluated declines in blood oxygenation or increases in pulse rate as potential pathways for previously observed cardiovascular mortality associated with particulate air pollution. Within the context of this basic objective, this study provides information regarding at least four questions of interest, as follows:

1. How usable and reliable are the data from pulse oximetry in a study of this type? The performance and usability of the pulse oximeters were excellent. The oximeters were easy to use and well received by the participants. Printouts provided a ready way to check for compliance and collect hard copies of results. Quality control comparisons with control readings and test readings suggested that Sp_O2 and pulse rate measures were consistent and reproducible.
2. Is the study design capable of observing small changes in Sp_O2 such as those associated with small daily changes in barometric pressure? Yes. The range of barometric pressure that occurred during the study period was only 26 mm Hg (635 to 661). Based on the estimated regression coefficients for barometric pressure, the mean increase in Sp_O2 that would be expected for an increase in barometric pressure equal to 26 mm Hg would be less than 1% (about 0.6%). This is approximately what would be expected at this point on the oxygen-hemoglobin dissociation (saturation) curve and given a comparable relative change in PO₂ in the blood (approximately 2 mm Hg or about 0.12 volume percent [vol %] of oxygen).   Although the direction and size of the observed association between barometric pressure and Sp_O2 are reasonable, care must be taken not to overinterpret the results. Most oximeters, including Nellcor, read somewhat high at the upper end of the oxygen dissociation curve (22, 23). Also, the differences in Sp_O2 across this range of barometric pressure are much smaller than the differences in Sp_O2 across individuals in the study. However, in this study by looking at daily means across an entire cohort of participants, or by estimating fixed effects regression models, individual differences could be averaged out or controlled for in the analysis. These results suggest that barometric pressure might be considered as a relevant weather variable in daily time series studies of morbidity and mortality and air pollution. However, the relevance of the observed association between barometric pressure and Sp_O2 with respect to daily time-series studies of cardiopulmonary health requires further study.
3. Are there observable declines in Sp_O2 after exposure to differing daily concentrations of particulate air pollution? Statistically significant negative associations with Sp_O2 and contemporaneous PM₁₀ were not consistently observed. Small negative associations between Sp_O2 and 1- to 6-d lagged PM₁₀ levels were observed, but for analysis that included all participants, observed associations between Sp_O2 and even lagged PM₁₀ were not statistically significant. It was observed, in stratified analysis, that there were statistically significant associations between Sp_O2 and lagged PM₁₀ levels for males participants 80 yr old or older, suggesting that the effects may be larger in older, less healthy individuals. Baseline testing of cardiac or pulmonary function may have provided more guidance as to the most sensitive subjects.
4. Are there observable increases in pulse rate after exposure to differing daily concentrations of particulate air pollution? Yes. Daily pulse rate, as well as the odds of having a pulse rate elevated by more than 5 or 10 beats/min, were associated with exposure to particulate air pollution on the previous 1 to 5 d. The association between pulse rate and PM₁₀ was small compared with the overall variability in pulse rate. In this study, however, pulse rate was obtained as integrated pulse rate over approximately 3 min using a pulse oximeter for 90 participants measured daily for 119 d. The use of aggregated data across the entire cohort of participants and the use of fixed effects regression models provided the opportunity to peer through the expected variability and observe and evaluate the subtle associations between pulse rate and particulate air pollution.

The importance of pulse has been emphasized in medicine for many centuries (24). Recent studies have observed that pulse rate is associated with hypertension, coronary heart disease, and mortality (25). Autonomic nervous system activated acutely elevated heart rate and blood viscosity along with decreased heart rate variability and reduced ventricular fibrillation threshold may also increase the likelihood of sudden cardiac death (32), but a role for fine particulate exposure is not documented. A recent study has observed increases in pulse rates of rats following exposure to elevated levels of particulate air pollution (33). Also in dogs, electrocardiogram (ECG) abnormalities were observed after exposure to fine particles (34). Although we are unaware of previously published studies that have evaluated potential associations between particulate air pollution and pulse rate in humans, air pollution was linked to elevated levels of plasma viscosity in both men and women living in Augsburg, Germany during an air pollution episode (35). The researchers speculated that "altered blood rheology due to inflammatory processes in the lung which induce an acute phase reaction might therefore be part of the pathological mechanisms linking air pollution to mortality." Within the context of treating plasma viscosity as an indicator for impaired blood flow properties, these results may be relevant to the present study. The results of this study suggest a lag structure of 1 to 5 d, consistent with the hypothesis that inflammatory responses to particulate pollution are involved (3) and that it takes a day or so before the inflammatory response and cytokine release are apparent.

In conclusion, this study provides little evidence of pollution-related hypoxia. The observed associations between elevated pulse rate and previous exposure to particulate air pollution are the most intriguing findings of this study. Nevertheless, the medical or biological relevance of these observed increases in pulse rate after exposure to particulate air pollution remains unclear. If there is a nonspurious association between particulate air pollution and elevated pulse rate, it is almost certainly only a part of substantially complex pathological mechanisms linking air pollution to cardiopulmonary mortality.