Content uploaded by Irene A Kreis
Author content
All content in this area was uploaded by Irene A Kreis
Content may be subject to copyright.
Available via license: CC BY-NC 4.0
Content may be subject to copyright.
Copyright© 2010, IRANIAN JOURNAL OF ALLERGY, ASTHMA AND IMMUN OLOGY. All rights reserved. 117
ORGINAL ARTICLE
Iran J Allergy Asthma Immunol
June 2010; 9(2): 117-126.
Air Pollution Effects on Peak Expiratory Flow Rate in Children
Narges Bagheri Lankarani1, Irene Kreis2 and David A. Griffiths3
1 Royan Research Institute, Tehran, Iran
2 CEDIR, University of Wollongong, Wollongong, Australia
3 School of Mathematics and Applied Statistics, University of Wollongong, Wollongong, Australia
Received: 30 November 2009; Received in revised form: 11 March 2010; Accepted: 11 April 2010
ABSTRACT
Airway mucus hypersecretion Health effects caused by air pollutants may range from
subtle biochemical or physiological signs, such as mildly reduced lung function, to difficult
breathing, wheezing, coughing and exacerbation of existing respiratory conditions such as
asthma. The aim of this study was measuring the adverse health effects of air pollution on
lung function of primary school students.
The lung function of students was measured daily for seven weeks in two elementary
schools in District 12 of Tehran, after obtaining permission from the two principals and
signed parents’ consent forms. Twenty four hourly air pollution levels were used as potential
predictors of lung function. The principal analysis conducted was a logistic regression on a
subset of the data using a case-crossover design.
The outcomes data consisted of the results of lung function tests for 356 female and 206
male students over the six-week period. Using the difference between mean (87) and
maximum (125) concentration of moving average of NO in this period to judge the size of
the effect, such an increase in NO is predicted to lead to an increase in the probability of
poor lung function (OR=20) based on population-based predicted value.
This study has shown strong and consistent associations between children’s poor lung
function and outdoor air pollutants in District 12 of Tehran for some pollutants. The strong
association found in this study was an increase in seven-day moving average of NO using
both definitions.
Key words: Air pollution; Asthma; Children; Poor lung function prevalence
INTRODUCTION
Epidemiological studies have demonstrated a clear
Corresponding Author: Narges Bagheri Lankarani, PhD;
Royan Research Institute, Tehran, Iran. PO Box: 198196-4697,
Tel: (+98 21) 2356 2652, Fax: (+98 21) 2230 6481,
E-mail: narges.lankarani@gmail.com
association between air pollution as it occurs in various
places around the world and lung function.1-9 Acute
effects of urban air pollution on respiratory health of
children were reported in many researches.10
Although a fairly large number of studies have
explored the respiratory impacts of air pollution,
because of its unique location and the children as
N. Bagheri Lankarani, et al.
118/ IRANIAN JOURNAL OF ALLERGY, ASTHMA AND IMMUNOLOGY Vol. 9, No. 2, June 2010
susceptible population to air pollution we studied the
association between air pollutants and respiratory
symptoms of primary school children in Tehran for the
first time. The question is whether air pollution, as it
occurs in Tehran, will show similar association with
respiratory health of primary school children.
This article addresses the association between air
pollution levels and the poor lung function.
MATERIALS AND METHODS
Data Collection
Lung function data were collected on students in two
schools. Over seven weeks, daily measurements of peak
expiratory flow rate (PEFR) were obtained using a mini-
Wright flow meter. The recorded lung function (PEFR)
for the first week was removed from the database, as
this was treated as a learning period.
The recorded measurements for the remaining six-
week period were analysed. During the period, six
measurements per week were taken on each child, there
being no school on Fridays.
Data Analysis
Case-crossover analysis of the data used worst lung
function definitions.11 The presence of poor lung
function is more complex than often presented. A Case
date would be any date which lung function measured
less than 50% the predicted value or the personal best
blow. Thus the case definition for case-crossover
analysis uses two alternative definitions of poor lung
function for the analyses.
Two definitions of poor lung function case were
used. Using a definition of poor lung function based on
PEFR less than 50% of predicted value, the case date
for each student identified as having poor lung function
was the date of worst lung function.
The second definition of poor lung function used the
best PEFR that each student produced during the six-
week data collection period was identified. Subjects
were deemed eligible to be a case if any observed PEFR
was below 50% of the student’s best PEFR (best blow).
Each such student was identified as a case on the
day of his or her worst lung function. The method of
defining a case and case date were otherwise the same
as for the first definition of case.
The case date is defined as the date of worst lung
function for each person according to either definition
one or two. The control dates are two weeks before and
after each case date. Therefore, by definition, a student
is his or her own control on a day of better lung
function. However, that day’s lung function may
measure (by either definition).
Statistical Model
Conditional logistic regression was used to analyse
case-crossover data, with the response variable taking
the value 1 for a case and 0 for a control. Variables used
as putative predictors in the regression model were daily
average of air pollution based on the teaching shift,
seven-day moving averages of air pollutants, daily
temperature and squared temperature (allowing the
model to incorporate a non-linear temperature effect.
The daily averages based on teaching shift are
calculated for a 24 hour window which differs between
teaching shifts, since the morning shift runs from 09:00
to 13:00 and the afternoon shift runs from 13:00 to
17:00 (times to be verified).
RESULTS
The air pollution and lung function data are
summarized in Table 1-6 and Figure 1-8. The lung
function data consisted of the results of 4,088 lung
function tests on 356 girls and 3,112 tests on 206 boys.
The pollution data are a temporal subset of the values
described.12
Although there were significant data integrity
problems for air pollution over the study period of over
two years, the case-crossover analysis only needed data
on case and control dates (and, for the calculation of one
week moving averages, the immediately preceding
seven days.
As indicated in Table 1 and Table 2, the air pollution
data for Fatemi look insufficient to analyse. However,
since case-crossover analysis just needs the case dates,
air pollution data for Fatemi were used in this study.
In addition, there were no SO2 data for either station
during the seven week period of lung function data
collection, so this was necessarily removed from the list
of potential predictors.
Lung Function Based on Predicted Value
The descriptive data of worst lung function based on
the predicted value (case definition one) are
summarized in Table 3. Pollutants exposures used were
the average of the current and seven-day moving
average based on the teaching shift.
Air Pollution Effects on Peak Expiratory Flow Rate in Children
Vol. 9, No. 2, June 2010 IRANIAN JOURNAL OF ALLERGY, ASTHMA AND IMMUNOLOGY /119
Table 1. Summary statistics for air polluion data
Variables
N
(Days)
N Miss
(Days)
Minimum
(µ
µµ
µg/m3)
Median
(µ
µµ
µg/m3)
Mean
(µ
µµ
µg/m3)
Maximum
(µ
µµ
µg/m3)
Std Dev
Fatemi
SO
2
- 43 - - - - -
PM
10
24 19 35 86 90 165 39.3
NO 24 19 58 92 97 160 24.6
NO2 24 19 91 133 142 224 39.0
NOX 24 19 86 132 137 229 37.6
O
3
10 33 4 8 7 14 3.2
CO 24 19 5703 9219 10167 16828 2874.9
Bazaar
SO2 - 43 - - - - -
PM10 42 1 30 84 324 5000 1046.4
NO 42 1 14 68 76 207 36.2
NO2 32 11 24 30 33 48 7.0
NO
X
42 1 10 66 73 178 29.9
O3 27 16 28 31 37 117 22.4
CO 42 1 3078 10589 9466 20172 4109.2
Table 2. Summary of distribution of coeff icients for
significant air pollutants at Fatemi station
MA PM10 MA NO MA CO
100% (Max) 4 9.2 78.8
75% (Q3) 0.8 1 4.2
50% (Median) 0.3 0.2 -6.9
25% (Q1) -0.2 -0.6 -18.6
0% (Min) -4.7 -8.6 -80.9
Estimates of the increase in prevalence are shown
using the hazard ratio. As indicated in Table 3, there
were 70 cases of worst lung function using predicted
value and 140 control observations which is 2 controls
per case. The average lung function was 101 L/min. A
matched conditional logistic regression was carried out
to investigate the relationship between an outcome and a
set of prognostic factors in matched case-control
studies. This
analysis used the PHREG procedure in SAS, with a
conditional logistic model and a stratum for each
matched set.
The results of case-crossover analysis are presented
in Table 4. Stepwise backward elimination was used to
choose the final model in which the seven-day moving
average of PM10 and NO from Fatemi were the pollution
variables significantly associated with poor lung
function (see Table 4).
In addition, using the difference between mean (81)
and maximum (133) concentration of moving average
of PM10 in this period to judge the size of the effect,
such an increase in PM10 is predicted to lead to a
decrease of poor lung function rate of 0.1. Using the
difference between mean (87) and maximum (121)
concentration of moving average of NO in this period to
judge the size of the effect, such an increase in NO is
predicted to lead to an increase in the probability of
poor lung function (OR = 19).
Lung Function Based on best Blow
The descriptive data of worst lung function based on
best blow (case definition two) are presented in Table 5.
As indicated in Table 5, there were 166 cases of worst
lung function using personal best blow and 332 control
observations (two weeks before and after case dates).
The average lung function was 125 L/min.
Conditional logistic regression using the PHREG
procedure in SAS was performed, and used a
conditional logistic model with a stratum for each
matched set.
N. Bagheri Lankarani, et al.
120/ IRANIAN JOURNAL OF ALLERGY, ASTHMA AND IMMUNOLOGY Vol. 9, No. 2, June 2010
Table 3. Summary statistics for lung function data based on case definition 1 and corresponding air pollution
data over six weeks
Variable N N Miss Minimum Median Mean Maximum Std Dev
Lung function 210 0 60 100 101 140 20.8
Height 210 0 113 129 131 151 9.7
Age 210 0 6 8 8 11 1.3
Fatemi
SO
2
0 210 - - - - -
PM
10
102 108 36 72 76 146 49.1
NO 102 108 66 90 91 160 26.6
NO
2
102 108 91 125 130 224 41.2
NO
X
102 108 86 116 126 229 39.9
O
3
54 156 4 6 7 14 3.3
CO 102 108 5 7 8 13 2.3
MA PM
10
comb 158 52 39 81 81 133 18.2
MA NO comb 158 52 23 88 87 121 13.8
Bazaar
SO
2
0 210 - - - - -
PM
10
198 12 30 78 81 157 30.3
NO 204 6 14 64 69 161 28.0
NO
2
168 42 25 30 33 48 7.3
NO
X
204 6 10 63 67 137 23.4
O
3
138 72 28 31 38 117 24.1
CO 204 6 2 9 8 16 3.3
-Not available
The results of case crossover analysis are presented
in Table 6. All risk factors such as daily temperature,
squared temperature, school, teaching shift, their
interactions, seven-day moving average of daily air
pollution PM10 from Fatemi, NO, NO2, O3 and CO
comb at both stations (see Table 6) were included in the
initial model. Stepwise backward elimination was used
to choose the final model in which the seven-day
moving average of PM10, NO and CO at Fatemi station
were the pollution variables significantly associated
with poor lung function. From Fatemi station, using the
difference between mean (83) and maximum (140)
concentration of moving average of PM10 in this period
to judge the size of the effect, such an increase in PM10
is predicted to lead to a decrease in the probability of
airway obstruction (OR = 0.1).
Table 4. Conditional logistic regression for lung function based on predicted value
Variable
Parameter
Estimate
Standard
Error
Chi Square
Pr>Chi-
Square
Hazard
(95%CI)
ratio
Lower
upper
Daily Temperature (T) 0.34
0.19
3.33
0.07
1.4
1.0
2.0
T2 -0.02
0.01
2.13
0.14
1.0
1.0
1.0
Shift 0.24
0.38
0.41
0.52
1.3
0.6
2.7
MAPM
10
comb(F)* -0.05
0.02
4.94
0.03
0.9
0.9
1.0
MA NO comb(F)* 0.08
0.03
5.94
0.01
1.1
1.0
1.2
* Statistically significant p = 0.05
Air Pollution Effects on Peak Expiratory Flow Rate in Children
Vol. 9, No. 2, June 2010 IRANIAN JOURNAL OF ALLERGY, ASTHMA AND IMMUNOLOGY /121
Table 5. Descriptive data of lung function based on definition 2 and air pollution
Variable N N Miss Minimum Median Mean Maximum Std Dev
Lung function (BB)
Fatemi
498 0 60 120 125 220 31.0
SO
2
0 498 - - - - -
PM
10
260 238 24 89 91 165 39.1
NO 260 238 43 92 94 160 25.3
NO
2
260
238
74
133
140
224
38.5
NO
X
260 238 62 132 134 229 37.4
O
3
156 342 4 8 8 17 3.7
CO 260 238 4 8 8 13 2.2
MA PM
10
comb 375 123 19 84 83 140 20.6
MA NO comb 375 123 17 87 88 125 16.5
MACOcomb(ppm) 375 123 2 8 8 11 1.4
Bazaar
SO
2
0 48 - - - - -
PM
10
443 55 30 87 91 184 35.2
NO 449 49 6 67 75 207 31.8
NO
2
361 137 25 30 33 48 6.6
NO
X
449 49 5 66 72 178 26.2
O
3
296 202 28 31 37 145 23.0
CO 469 29 2 9 8 16 3.2
- Not available
Using the difference between mean (88) and
maximum (125) concentration of moving average of
NO in this period to judge the size of the effect, such an
increase in NO is predicted to lead to an increase in the
probability of airway obstruction (OR = 80). Using the
difference between mean 8 and maximum 11
concentration of moving average of CO in this period to
judge the size of the effect, such an increase in CO is
predicted to lead to a decrease in the probability of
airway obstruction (OR = 0.1).
Table 6. Conditional logistic regression for lung function at Fatemi
Variable
Parameter
Estimate
Standard
Error
Chi-Square
Pr>Chi-Square
Hazard
(95%CI)
ratio
Lower
upper
Daily Temperature (T)* 0.41
0.13
9.48
0.0021
1.5
1.2 2.0
T2* -0.02
0.01
6.22
0.0126
1.0
1.0 1.0
Shift 0.20
0.28
0.54
0.4606
1.2
0.7 2.1
Gender -0.29
0.41
0.50
0.478
0.8
0.3 1.7
Shift × Gender
0.15
0.62
0.05
0.8156
1.2
0.3
3.9
MA PM
10
comb(F)* -0.04
0.01
6.99
0.0082
1.0
0.9 1.0
MA NO comb(F)* 0.12
0.04
11.09
0.0009
1.1
1.1 1.2
MA CO comb(ppm)(F)* -1.02
0.36
7.89
0.005
0.4
0.2 0.7
*Statistically significant p = 0.05 MA: seven-day moving average
N. Bagheri Lankarani, et al.
122/ IRANIAN JOURNAL OF ALLERGY, ASTHMA AND IMMUNOLOGY Vol. 9, No. 2, June 2010
Figure 1 Daily lung function of all students during a six-week period
Figure 2. Daily PM10 levels over six-week, units in µg/m3 for the morning teaching shift
Figure 3. Daily PM10 levels over six-week, units in µg/m3 for the afternoon teaching shift
0
100
200
300
400
500
600
9/11/02 16/11/02 23/11/02 30/11/02 7/12/02 14/12/02 21/12/02
L/min
Max
0
1000
2000
3000
4000
5000
9/11/02 16/11/02 23/11/02 30/11/02 7/12/02 14/12/02 21/12/02
Fatemi Bazaar
0
100 0
200 0
300 0
400 0
500 0
9/11 /02 1 6/11 /02 2 3/1 1/02 3 0/11/02 7/12 /02 1 4/12 /02 2 1/12/02
Fatemi Bazaar
Air Pollution Effects on Peak Expiratory Flow Rate in Children
Vol. 9, No. 2, June 2010 IRANIAN JOURNAL OF ALLERGY, ASTHMA AND IMMUNOLOGY /123
Figure 4. Daily NO levels over six-week, units in µ
µµ
µg/m3 for the morning teaching shift
Figure 5. Daily NO levels over six-week, units in µ
µµ
µg/m3 for the afternoon teaching shift
Figure 6. Daily CO levels over six-week, units in µ
µµ
µg/m3 for the morning teaching shift
0
200
400
600
9/11/02 16/11/02 23/11/02 30/11/02 7/12/02 14/12/02 21/12/02
Fatemi Bazaar
0
200
400
600
9/11/02 16/11/02 23/11/02 30/11/02 7/12/02 14/12/02 21/12/02
Fatemi Bazaar
0
10000
20000
9/11/02 16/11/02 23/11/02 30/11/02 7/12/02 14/12/02 21/12/02
Fatemi Bazaar
N. Bagheri Lankarani, et al.
124/ IRANIAN JOURNAL OF ALLERGY, ASTHMA AND IMMUNOLOGY Vol. 9, No. 2, June 2010
Figure 7. Daily CO levels over six-week, units in µ
µµ
µg/m3 for the afternoon teaching shift
Figure 8. Daily temperature levels over six-week, units in °C
DISCUSSION
This study examined the association between air
pollutants and poor lung function in elementary school
children in District 12 of Tehran over six-week. In total,
562 students from two schools participated in the lung
function study. For the case-crossover analysis used
here, the number of cases was rather less than the
number of students. The final analysis also omitted
those cases for which the pollution data were missing on
the relevant days. That is 158 and 375 observation were
used from 210 and 498 observation read using students'
predicted values and best blow respectively.
This study has shown a statistically significant
relationship between outdoor air pollution and child’s
poor lung function. In this study, associations of seven-
day moving averages of PM10 , NO and CO with poor
lung function were found. However, the relationship
between the rest of air pollutants and child's poor lung
function has not shown up may be because of their
many missing values. While the frequency of poor lung
function was high in December, this study showed the
effect of NO concentration could increase the number of
days with poor lung function. In other hand, higher
concentration of NO is associated with higher rate poor
lung function.
0
10000
20000
9/11 /02 16/11/02 23/11 /02 30/1 1/0 2 7/12/0 2 14/1 2/0 2 21 /12/02
Fatemi Bazaar
0
10
20
30
9/11 /02 16/11 /02 23/11 /02 30/11/0 2 7/12/02 1 4/1 2/02 2 1/12/02
Air Pollution Effects on Peak Expiratory Flow Rate in Children
Vol. 9, No. 2, June 2010 IRANIAN JOURNAL OF ALLERGY, ASTHMA AND IMMUNOLOGY /125
Nitric oxide (NO) is the most common form of
nitrogen directly emitted into the atmosphere.1 In
ambient outdoor air, nitric oxide (NO), which is emitted
by motor vehicles, combines with oxygen in the
atmosphere under the action of sunlight, producing
nitrogen dioxide (NO2) a major air pollutant and other
NOX. The previous studies showed nitric oxide does not
significantly affect human health. On the other hand,
elevated levels of NO2 cause damage to the mechanism
that protect the human respiratory tract and can increase
a person’s susceptibility to respiratory infections.13,14
In particular populations living near busy roads, NO2
is of particular concern. At levels currently observed in
Europe, exposure to NO2 may decrease lung function15
and increase the risk of respiratory problems,
particularly in children.16 Short-term exposure to peak
levels can increase respiratory allergic reactions.
Overall, the study had good response rates with a
response of 72%. Therefore, it can be considered
representative for the busier areas of Tehran at least. To
assure consistency in the measurements of lung
function, the researcher used competition between
students to get their best PEFR. The students were
blinded to the hypothesis investigated in this study. One
reason was that they could not check out the air
pollution level every day as it was not available
everywhere. It was presented every day on an electronic
screen on Fatemi station only at the time were the data
were collected and it was available online but the
number of students who had access to internet was
limited. Another reason was that the population did not
express much concern about the level of air pollution. A
possible confounder that was not controlled for in the
analysis was the use of asthma medication. However,
only 10 students in that sample used this medication.
The effect of this confounder would be to limit our
ability to detect an effect, as it would potentially mask
the effect of air pollution on lung function. Therefore, it
is expected that selection and respondents bias have not
substantially influenced the results of this study.
As a form of case-control was used in the analysis, a
selection bias might be of concern in the selection of the
controls. To prevent this, a symmetric bidirectional
method was used; this was described as providing
adequate control.17
Thus, this study has shown strong and consistent
associations between children’s poor lung function and
outdoor air pollutants in District 12 of Tehran for some
pollutants. The strong association found in this study
was an increase in seven-day moving average of NO
using both definitions. These impacts also appeared to
be distinct from any temperature effects. PM10 and CO
are not consistent with the literature. To answer the
study question whether there is an association between
air pollution levels and poor lung function, the study
showed there is association between lung function
changes or airway obstruction with air pollution levels.
These data indicate that in circumstances in which NO
levels are chronically elevated, the levels of exposure to
NO in the previous seven-day can influence the level of
lung function in children. Interestingly, other study on
absenteeism also showed the positive association
between the concentration of NO and respiratory related
absenteeism from school.12 This study adds the effect
and positive association of NO on lung function.
ACKNOWLEDGEMENTS
This research was supported in part by funding from
the University of Wollongong. We thank Ms Vicki
Kendrick for assistance in using SAS.
REFERENCES
1. Brunekreef B, Janssen NAH, De Hartog J, Harssema H,
Knape M, VanVliet P. Air Pollution from Truck Traffic
and Lung Function in Children Living Near Motor Ways.
Epidemiology1997; 8:298-303.
2. Gauderman WJ, McConnell R, Gilliland F, London S,
Thomas D, Avol E, et al. Association between Air
Pollution and Lung Function Growth in Southern
California Children. Am J Respir Crit Care Med 2000;
162(4 pt 1):1383-90.
3. Peled R, Bibi H, Pope CA, 3rd, Nir P, Shiachi R, Scharff S.
Differences in lung function among school children in
communities in Israel. Arch Environ Health 2001;
56(1):89-95.
4. Gauderman WJ. USC Study Confirms Air Pollution Linked
to Slowed Lung Function Growth in Children.2002 01/07/.
5. Neuberger M, Moshammer H, Kundi M. Declining
Ambient Air Pollution and Lung Function Improvement in
Austrian Children. Atmos Env 2002; 36(11):1733-7.
6. Frye C, Hoelscher B, Cyrys J, Wjst M, Wichmann HE,
Heinrich J. Association of lung function with declining
ambient air pollution. Environ Health Perspect 2003;
111(3):383-7.
7. Frye C, Hoelscher B, Cyrys J, Wjst M, Wichmann HE,
Heinrich J. Association of Lung Function with Declining
N. Bagheri Lankarani, et al.
126/ IRANIAN JOURNAL OF ALLERGY, ASTHMA AND IMMUNOLOGY Vol. 9, No. 2, June 2010
Ambient Air Pollution. Environmental Health
Perspectives2003; 111(3):383-7.
8. Neuberger M, Schimek MG, Horak J, Friedrich,
Moshammer H, Kundi M, et al. Acute Effects of
Particulate Matter on Respiratory Diseases, Symptoms, and
Functions: Epidemiological Results of the Austrian Project
on Health Effects of Particulate Matter (AUPHEP). Atmos
Env 2004; 38(24):3971-81.
9. Jedrychowski W, Maugeri U, Jedrychowska-Bianchi I,
Flak E. Effect of Indoor Air Quality in the Postnatal Period
on Lung Function in Pre-Adolescent Children: A
Retrospective Cohort Study in Poland. Public Health 2005;
119(6):535-41.
10. Van der Zee SC, Hoek G, Boezen HM, Schouten P,
Wijnen JHv, Brunekreef B. Acute Effects of Urban Air
Pollution on Respiratory Health of Children with and
Without Chronic Respiratory Symptoms. Occup Environ
Med 1999; 56(12):802-13.
11. Schwartz J. How Sensitive is the Association between
Ozone and Daily Deaths to Control for Temperature? Am J
Respir Crit Care Med 2005; 171(6):627-31.
12. Bagheri Lankarani N. Adverse Health Effects of Air
Pollution on Primary School Children in Tehran [PhD
thesis]. Wollongong: University of Wollongong; 2006.
13. Gilliland FD, Berhane K, Rappaport EB, Thomas DC,
Avol E, Gauderman WJ, et al. The Effects of Ambient Air
Pollution on School Absenteeism due to Respiratory
Illnesses. Epidemiol 2001; 12(1):43-54.
14. Hwang J-S, Chen Y-J, Wang J-D, Lai Y-M, Yang C-Y,
Chan C-C. Subject-Domain Approach to the Study of Air
Pollution Effects on Schoolchildren's Illness Absence. Am
J Epidemiol2000; 152(1):67-74.
15. Peters JM, Avol EL, Navidi W, London SJ, Gauderman
WJ, Lurmann F, et al. A Study of Twelve Southern
California Communities with Differing Levels and Types
of Air Pollution. II. Effects on Pulmonary Function.
American Journal of Respiratory and Critical Care
Medicine1999; 159(3):768-75.
16. Neuberger M, Moshammer H. [Suspended particulates and
lung health]. Wien Klin Wochenschr 2004; 116(Suppl 1):8-
12.
17. Bateson T, Schwartz J. Selection Bias and Confounding in
Case-Crossover Analyses of Environmental Time-Series
Data. Epidemiology 2001; 12(6):654-61.