This article was downloaded by: [New York University] On: 05 May 2015, At: 05:53 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Archives of Environmental Health: An International Journal Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/vzeh20
Air Pollution and the Duration of Acute Respiratory Symptoms Joel Schwartz
a
a
U.S. Environmental Protection Agency , Washington, DC, USA Published online: 03 Aug 2010.
To cite this article: Joel Schwartz (1992) Air Pollution and the Duration of Acute Respiratory Symptoms, Archives of Environmental Health: An International Journal, 47:2, 116-122, DOI: 10.1080/00039896.1992.10118764 To link to this article: http://dx.doi.org/10.1080/00039896.1992.10118764
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/termsand-conditions
Air Pollution and the Duration of Acute Respiratory Symptoms
Downloaded by [New York University] at 05:53 05 May 2015
JOEL KHWARTZ U.S. Environmental Protection Agency Washington, DC
ABSTRACT. The possibility that air pollution exposure can extend the duration of respiratory symptoms was examined in a diary study of student nurses. This diary study has already shown associations between air pollution and incidence rates of respiratory symptoms. After individual risk factors and temperature were controlled for, photochemical oxidants were significantly @I < .oOOl) associated with the duration of episodes of coughing, phlegm, and sore throat. Some heterogeneity of response to oxidants was seen; there was l i i l e effect on asthmatics, but the impact increased as family income increased. Plots of the mean duration of symptoms, by quintiles of oxidants, for which the other covariates were contdled, showed strong signs of a dose-response relationship for coughing and phlegm and moderate signs of a monotonic doseresponse relationship for sore throat. The relationships continued for concentrations below the current ambient standard for ozone. Chest tightness or discomfort was significantly associated with sulfur dioxide @I = .016), but the effect seemed mainly restricted to asthmatics. However, evidence for a d d e p e n d e n t increase was weak.
of these symptoms means that the number of days on DESPITE NUMEROUS STUDIES of acute responses to which a subject was eligible to remain symptomatic air pollution,lB the impact of such exposures is not fully understood. Daily diaries of respiratory symptoms have were much lower than the number of days on which a subject could become symptomatic. This gives such proved to be a valuable tool in this regard because the populations studied serve as their own control, and the studies less power. Recently, air pollution has been associated with the duration of respiratory symptoms in recall error of annual questionnaires is avoided. These children;" this prompted an examination of symptom studies raise several methodological issues, including duration in a diary study in which an association be(a) the possibility of serial correlation, in which obsertween symptom incidence and air pollution has recentvations that occur close together in time tend to be ly been reported." more alike than observations that occur far apart in time; and (b) heterogeneity of individual responses. These issues have been discussed in detail else~here.~ Methods and materials Some studies have examined incidence rates, rather than prevalence rates, for respiratory symptoms; this A population of individuals who were first-year nurshas been done, in part, to reduce the serial correlation ing students in Los Angeles was recruited for a study of risk factors for acute illness. Smoking histories and the of the data," and particularly because the degree of presence of asthma, hay fever, chronic sinusitis, and serial correlation may differ between incidence and allergies to house dust and pollen were determined for persistence of symptoms. Other studies have treated all subjects. Only students who were residents of the prevalence rate^.^.'','^ nursing school were eligible for inclusion in the study. To date, duration of symptom episodes has received Daily diaries of acute symptoms were distributed and less attention, partially because the low incidence rate 116
Archives of Environmental Health
Downloaded by [New York University] at 05:53 05 May 2015
collected each Monday for 3 y. Further details of the study have been published previously.'"'' The respiratory symptoms examined were episodes of cough, sore throat, phlegm or sputum, and chest tightness or discomfort. Daily measurements of air pollution were obtained from a monitoring site located within 2.5 miles of the student nurses' dormitories and hospitals. Only the maximum 1-h concentrations of photochemical oxidants, nitrogen dioxide, sulfur dioxide, and carbon dioxide were available. The minimum and maximum daily temperatures for each day were obtained from the National Oceanic and Atmospheric Administration (NOM) weather station, which was located about a mile from the air pollution monitor. Duration of an episode was defined as the number of consecutive days the symptom was reported by the individual. The distribution of duration was highly skewed for all outcomes, with a minimum of 1 d and a long tail. A logarithmic transformation approximately normalized the episode durations. The means and standard deviations for the natural logarithm of duration are shown in Table 1 for the first four symptoms. The distributions of air pollution, allergies, chronic conditions, smoking, and temperature have already been published.'o*'' Individual nurses had more than one episode of each symptom during the 3 y of study. If there is heterogeneity of individual risk across subjects, the duration of different episodes for the same subject will be more alike than between subjects. This has been dealt with as a random subject effect. The inferences of classic regression analyses depend on the validity of the assump tions. In particular, heteroscedasticity, or other deviations from normality of the residuals, can result in biased t statistics for individual risk factors. Robust variance estimates provide hypothesis tests on the variables that are asymptotically unbiased, even if the variance is misspecified. The generalized estimating equations of Liang and Zeger'' were used to estimate the random subject effects and robust variance estimates. Because of their computational complexity, models were initially fit by use of ordinary least squares regression, and the Liang and Zeger methodology was used to refit the final models. Initial regressions considered covariates from the individual histories for each subject, including age, family income, allergies, the presence of asthma, hay fever, or sinusitis, and both current and cumulative smoking. Once the best subject covariate model had been fit, air fa& 1.-Mean (8and Standard Deviation 0) of the Logarithm of Symptom Duration
~Vmptom Cough Phlegm Sore throat Chest discomfort
No. of episodes 1 747 1 249 1998 845
MarcWAprill992 [Vd. 47 (NO.?)I
R
SD
0.84 0.85
0.94
0.73
0.77
0.60
0.75
0.96
pollution and weather terms were considered. Including a linear air pollution term in a model for the logarithm of symptom duration is equivalent to assuming an exponential increase in duration as exposure increases. Although such a relationship is possible, a less steep rate of increase is also quite plausible. Hence, linear terms for both air pollution concentration and temperature and their natural logarithms were considered. The pollution variable used was the average pollution level during the episode. The average levels for the week and the month before the episode began were considered as an alternative. Air pollutants were considered, one at a time. if more than one pollutant was a significant predictor of the duration of any symp tom, multiple pollutant models were considered. The dose-response nature of any significant association with air pollution was examined, i.e., the mean duration by quintile of pollutant, after adjustment by regression for the significant covariates, was computed and displayed in a plot. Similar plots were constructed for the smoking variables. Recent chamber studies of ozone exposure have shown that certain subjects have a considerably enhanced forced expiratory volume response to ozone, although these subjects were not distinguishable by either baseline decrements in lung function or by the presence of known risk fact~rs.'~ This suggests that factors that are not themselves main effects for respiratory outcomes may be effect modifiers for pollution exposure. This was tested; once basic models were estab lished, interactions between subject covariates and any significant pollution variable were examined-even for covariates that showed no main effect. Results The only subject covariates associated with duration of respiratory episodes were family income and smoking. As previously noted,'O current cigarettes per day was a better predictor of duration of phlegm and chest discomfort than pack years, but pack years of exposure was a better predictor of the duration of coughing episodes, Smoking was not associated with duration of sore throat episodes. Family income was a risk factor for phlegm. The maximum daily temperature on the day the episode began was a better predictor of symp tom duration than either its logarithmic transform or other averaging periods. In Figure 1, a-c, is shown the relationship between cigarette smoke and the duration of episodes of coughing, phlegm, and chest discomfort. When air pollution was considered, the mean pollution during the episode was always a better predictor than pollution during the previous 7 or 30 d, or on the day the episode began. The natural logarithm of pollution was also always a better predictor of symptom duration than the untransformed pollutant. In Table 2 are shown the final regression models for the logarithm of duration of episodes of coughing, phlegm, chest discomfort, and sore throat. Photochemical oxidants were a significant predictor of symptom duration for cough episodes. Either nitrogen dioxide or oxidants were significant in single-pollutant models for phlegm 117
6
I
I
l
a
-k m
5
-u
v
4 M
1
3 Ir
4
0
c
0 .* Y
n$
3
2
Downloaded by [New York University] at 05:53 05 May 2015
NONE
-
b
-A -8
-
M
u
a
2.0
L.
0
C 0
. I
CI
E
G
1.5
1.o
ISmoking Status
duration. The model with oxidants had greater explanatory power. When both pollutants were considered in a model simultaneously, nitrogen dioxide became insignificant and was dropped. Oxidants were significantly associated with duration of sore throat, and sulfur dioxide was significantly associated with the duration of chest discomfort. The impact of heterogeneity is illustrated in Table 3, in which the coefficient and robust t statistics of the pollution variables, estimated without the random effects model and without control for the individual covariates, are shown. The result with the random effects model are repeated for ease of comparison. The intraclass correlation coefficients in the random effects models were 20 for phlegm episodes, .09 for sore 118
=. lO/day
Fig. l a x . Duration of cough, phlegm, and chesi discomfort epC sodes, by categories of smoking, controlling, by regression, for the signifisant covariates in Table 2.
2.5
&
4
IO/day
Smoking Status
Smoking Category
3.0
Archives of Environmental Health: An International Journal Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/vzeh20
Air Pollution and the Duration of Acute Respiratory Symptoms Joel Schwartz
a
a
U.S. Environmental Protection Agency , Washington, DC, USA Published online: 03 Aug 2010.
To cite this article: Joel Schwartz (1992) Air Pollution and the Duration of Acute Respiratory Symptoms, Archives of Environmental Health: An International Journal, 47:2, 116-122, DOI: 10.1080/00039896.1992.10118764 To link to this article: http://dx.doi.org/10.1080/00039896.1992.10118764
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/termsand-conditions
Air Pollution and the Duration of Acute Respiratory Symptoms
Downloaded by [New York University] at 05:53 05 May 2015
JOEL KHWARTZ U.S. Environmental Protection Agency Washington, DC
ABSTRACT. The possibility that air pollution exposure can extend the duration of respiratory symptoms was examined in a diary study of student nurses. This diary study has already shown associations between air pollution and incidence rates of respiratory symptoms. After individual risk factors and temperature were controlled for, photochemical oxidants were significantly @I < .oOOl) associated with the duration of episodes of coughing, phlegm, and sore throat. Some heterogeneity of response to oxidants was seen; there was l i i l e effect on asthmatics, but the impact increased as family income increased. Plots of the mean duration of symptoms, by quintiles of oxidants, for which the other covariates were contdled, showed strong signs of a dose-response relationship for coughing and phlegm and moderate signs of a monotonic doseresponse relationship for sore throat. The relationships continued for concentrations below the current ambient standard for ozone. Chest tightness or discomfort was significantly associated with sulfur dioxide @I = .016), but the effect seemed mainly restricted to asthmatics. However, evidence for a d d e p e n d e n t increase was weak.
of these symptoms means that the number of days on DESPITE NUMEROUS STUDIES of acute responses to which a subject was eligible to remain symptomatic air pollution,lB the impact of such exposures is not fully understood. Daily diaries of respiratory symptoms have were much lower than the number of days on which a subject could become symptomatic. This gives such proved to be a valuable tool in this regard because the populations studied serve as their own control, and the studies less power. Recently, air pollution has been associated with the duration of respiratory symptoms in recall error of annual questionnaires is avoided. These children;" this prompted an examination of symptom studies raise several methodological issues, including duration in a diary study in which an association be(a) the possibility of serial correlation, in which obsertween symptom incidence and air pollution has recentvations that occur close together in time tend to be ly been reported." more alike than observations that occur far apart in time; and (b) heterogeneity of individual responses. These issues have been discussed in detail else~here.~ Methods and materials Some studies have examined incidence rates, rather than prevalence rates, for respiratory symptoms; this A population of individuals who were first-year nurshas been done, in part, to reduce the serial correlation ing students in Los Angeles was recruited for a study of risk factors for acute illness. Smoking histories and the of the data," and particularly because the degree of presence of asthma, hay fever, chronic sinusitis, and serial correlation may differ between incidence and allergies to house dust and pollen were determined for persistence of symptoms. Other studies have treated all subjects. Only students who were residents of the prevalence rate^.^.'','^ nursing school were eligible for inclusion in the study. To date, duration of symptom episodes has received Daily diaries of acute symptoms were distributed and less attention, partially because the low incidence rate 116
Archives of Environmental Health
Downloaded by [New York University] at 05:53 05 May 2015
collected each Monday for 3 y. Further details of the study have been published previously.'"'' The respiratory symptoms examined were episodes of cough, sore throat, phlegm or sputum, and chest tightness or discomfort. Daily measurements of air pollution were obtained from a monitoring site located within 2.5 miles of the student nurses' dormitories and hospitals. Only the maximum 1-h concentrations of photochemical oxidants, nitrogen dioxide, sulfur dioxide, and carbon dioxide were available. The minimum and maximum daily temperatures for each day were obtained from the National Oceanic and Atmospheric Administration (NOM) weather station, which was located about a mile from the air pollution monitor. Duration of an episode was defined as the number of consecutive days the symptom was reported by the individual. The distribution of duration was highly skewed for all outcomes, with a minimum of 1 d and a long tail. A logarithmic transformation approximately normalized the episode durations. The means and standard deviations for the natural logarithm of duration are shown in Table 1 for the first four symptoms. The distributions of air pollution, allergies, chronic conditions, smoking, and temperature have already been published.'o*'' Individual nurses had more than one episode of each symptom during the 3 y of study. If there is heterogeneity of individual risk across subjects, the duration of different episodes for the same subject will be more alike than between subjects. This has been dealt with as a random subject effect. The inferences of classic regression analyses depend on the validity of the assump tions. In particular, heteroscedasticity, or other deviations from normality of the residuals, can result in biased t statistics for individual risk factors. Robust variance estimates provide hypothesis tests on the variables that are asymptotically unbiased, even if the variance is misspecified. The generalized estimating equations of Liang and Zeger'' were used to estimate the random subject effects and robust variance estimates. Because of their computational complexity, models were initially fit by use of ordinary least squares regression, and the Liang and Zeger methodology was used to refit the final models. Initial regressions considered covariates from the individual histories for each subject, including age, family income, allergies, the presence of asthma, hay fever, or sinusitis, and both current and cumulative smoking. Once the best subject covariate model had been fit, air fa& 1.-Mean (8and Standard Deviation 0) of the Logarithm of Symptom Duration
~Vmptom Cough Phlegm Sore throat Chest discomfort
No. of episodes 1 747 1 249 1998 845
MarcWAprill992 [Vd. 47 (NO.?)I
R
SD
0.84 0.85
0.94
0.73
0.77
0.60
0.75
0.96
pollution and weather terms were considered. Including a linear air pollution term in a model for the logarithm of symptom duration is equivalent to assuming an exponential increase in duration as exposure increases. Although such a relationship is possible, a less steep rate of increase is also quite plausible. Hence, linear terms for both air pollution concentration and temperature and their natural logarithms were considered. The pollution variable used was the average pollution level during the episode. The average levels for the week and the month before the episode began were considered as an alternative. Air pollutants were considered, one at a time. if more than one pollutant was a significant predictor of the duration of any symp tom, multiple pollutant models were considered. The dose-response nature of any significant association with air pollution was examined, i.e., the mean duration by quintile of pollutant, after adjustment by regression for the significant covariates, was computed and displayed in a plot. Similar plots were constructed for the smoking variables. Recent chamber studies of ozone exposure have shown that certain subjects have a considerably enhanced forced expiratory volume response to ozone, although these subjects were not distinguishable by either baseline decrements in lung function or by the presence of known risk fact~rs.'~ This suggests that factors that are not themselves main effects for respiratory outcomes may be effect modifiers for pollution exposure. This was tested; once basic models were estab lished, interactions between subject covariates and any significant pollution variable were examined-even for covariates that showed no main effect. Results The only subject covariates associated with duration of respiratory episodes were family income and smoking. As previously noted,'O current cigarettes per day was a better predictor of duration of phlegm and chest discomfort than pack years, but pack years of exposure was a better predictor of the duration of coughing episodes, Smoking was not associated with duration of sore throat episodes. Family income was a risk factor for phlegm. The maximum daily temperature on the day the episode began was a better predictor of symp tom duration than either its logarithmic transform or other averaging periods. In Figure 1, a-c, is shown the relationship between cigarette smoke and the duration of episodes of coughing, phlegm, and chest discomfort. When air pollution was considered, the mean pollution during the episode was always a better predictor than pollution during the previous 7 or 30 d, or on the day the episode began. The natural logarithm of pollution was also always a better predictor of symptom duration than the untransformed pollutant. In Table 2 are shown the final regression models for the logarithm of duration of episodes of coughing, phlegm, chest discomfort, and sore throat. Photochemical oxidants were a significant predictor of symptom duration for cough episodes. Either nitrogen dioxide or oxidants were significant in single-pollutant models for phlegm 117
6
I
I
l
a
-k m
5
-u
v
4 M
1
3 Ir
4
0
c
0 .* Y
n$
3
2
Downloaded by [New York University] at 05:53 05 May 2015
NONE
-
b
-A -8
-
M
u
a
2.0
L.
0
C 0
. I
CI
E
G
1.5
1.o
ISmoking Status
duration. The model with oxidants had greater explanatory power. When both pollutants were considered in a model simultaneously, nitrogen dioxide became insignificant and was dropped. Oxidants were significantly associated with duration of sore throat, and sulfur dioxide was significantly associated with the duration of chest discomfort. The impact of heterogeneity is illustrated in Table 3, in which the coefficient and robust t statistics of the pollution variables, estimated without the random effects model and without control for the individual covariates, are shown. The result with the random effects model are repeated for ease of comparison. The intraclass correlation coefficients in the random effects models were 20 for phlegm episodes, .09 for sore 118
=. lO/day
Fig. l a x . Duration of cough, phlegm, and chesi discomfort epC sodes, by categories of smoking, controlling, by regression, for the signifisant covariates in Table 2.
2.5
&
4
IO/day
Smoking Status
Smoking Category
3.0
