University of Southern California Dissertations and Theses

Persistence of pollution-induced lung function deficits in early adulthood: evidence from the Children's Health Study

(USC Thesis Other)

Persistence of pollution-induced lung function deficits in early adulthood: evidence from the Children's Health Study

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content
PERSISTENCE OF POLLUTION-INDUCED LUNG FUNCTION DEFICITS IN
EARLY ADULTHOOD: EVIDENCE FROM THE CHILDREN’S HEALTH STUDY

by
Trevor A. Pickering

A Thesis Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
MASTER OF SCIENCE
(BIOSTATISTICS)

August 2010

Copyright 2010 Trevor A. Pickering
ii

Table of Contents

List of Tables iii

List of Figures iv

Abstract vi

Chapter 1: Introduction 1

Chapter 2: Methods 4
Study Recruitment, Participation, and Measurements 4
Statistical Methods 6

Chapter 3: Results 11
Baseline Growth Curve 11
Lung Function and Pollution 11

Chapter 4: Discussion 44
“Betas” Model 44
Correlation between Age 18 and APEX 44
Possible Study Bias 46
Other Caveats 49
Relation to Other Studies 50

Chapter 5: Conclusion 52

References 54

iii

List of Tables

Table 1: PFT frequencies by age and study cohort. 7

Table 2a: Mixed model regression of ΔFEV
1
on pollution for each age 22
category (all cohorts).

Table 2b: Mixed model regression of ΔFEV
1
on pollution for each age 23
category (C and D cohorts).

Table 3a: Mixed model regression of ΔMMEF on pollution for each age 24
category (all cohorts).

Table 3b: Mixed model regression of ΔMMEF on pollution for each age 25
category (C and D cohorts).

Table 4a: Regression of age- and pollutant-specific β on age (FEV
1,
all 39
cohorts).

Table 4b: Regression of age- and pollutant-specific β on age (FEV
1
, 39
C and D cohorts).

Table 5a: Regression of age- and pollutant-specific β on age (MMEF, all 40
cohorts).

Table 5b: Regression of age- and pollutant-specific β on age (MMEF, 40
C and D cohorts).

Table 6: Regression of (APEX-18) pollution exposures on pre-18 pollution 42
assignments.

Table 7a: Regression of the (ΔFEV
APEX
- ΔFEV
18
) difference on age 18, 42
APEX, and (APEX-18) pollution exposures.

Table 7b: Regression of the (ΔMMEF
APEX
- ΔMMEF
18
) difference on age 18, 43
APEX, and (APEX-18) pollution exposures.

Table 8: Demographics of APEX and non-APEX children at age 18 for 48
categorical and continuous variables.

iv

List of Figures

Figure 1a: FEV
1
growth curve (all cohorts). 12

Figure 1b: FEV
1
growth curve (C and D cohorts). 12

Figure 2a: MMEF growth curve (all cohorts). 13

Figure 2b: MMEF growth curve (C and D cohorts). 13

Figure 3a: Beta coefficients from the regression of ΔFEV
1
on pollution, 14
across age categories (all cohorts).

Figure 3b: Beta coefficients from the regression of ΔFEV
1
on pollution, 16
across age categories (C and D cohorts).

Figure 4a: Beta coefficients from the regression of ΔMMEF on pollution, 18
across age categories (all cohorts).

Figure 4b: Beta coefficients from the regression of ΔMMEF on pollution, 20
across age categories (C and D cohorts).

Figure 5a: FEV
1
residual at APEX measurement plotted against residual at 27
age 18 (all cohorts).

Figure 5b: FEV
1
residual at APEX measurement plotted against residual at 28
age 18 (C and D cohorts).

Figure 5c: FEV
1
residual at APEX measurement plotted against residual at 29
age 18 (all cohorts), for males and females.

Figure 5d: FEV
1
residual at APEX measurement plotted against residual at 30
age 18 (all cohorts), for low-pollution and high-pollution
communities.

Figure 5e: FEV
1
residual at APEX measurement plotted against residual at 31
age 18 (C and D cohorts), for males and females.

Figure 5f: FEV
1
residual at APEX measurement plotted against residual at 32
age 18 (C and D cohorts), for low-pollution and high-pollution
communities.
v

List of Figures (continued)

Figure 6a: MMEF residual at APEX measurement plotted against residual 33
at age 18 (all cohorts).

Figure 6b: MMEF residual at APEX measurement plotted against residual 34
at age 18 (C and D cohorts).

Figure 6c: MMEF residual at APEX measurement plotted against residual 35
at age 18 (all cohorts), for males and females.

Figure 6d: MMEF residual at APEX measurement plotted against residual 36
at age 18 (all cohorts), for low-pollution and high-pollution
communities.

Figure 6e: MMEF residual at APEX measurement plotted against residual 37
at age 18 (C and D cohorts), for males and females.

Figure 6f: MMEF residual at APEX measurement plotted against residual 38
at age 18 (C and D cohorts), for low-pollution and high-pollution
communities.

vi

Abstract
We performed a follow-up study to track lung function of participants in the
Children’s Health Study (CHS), a multi-year longitudinal analysis of lung function
development for 3,787 children in twelve Southern California communities. Subjects
were followed past 18 years of age in order to examine the relationship between ambient
pollution and two lung function indices: maximal midexpiratory flow (MMEF) and
forced expiratory volume in 1 second (FEV
1
). There was a significant negative
relationship between several pollutants (acid vapor, NO
2
, particles with diameter less
than 2.5 microns, particles with diameter less than 10 microns, and elemental carbon) and
lung function. This relationship grew stronger with each year increase in age, up to 18
years (P<0.05 for all pollutants). There was little evidence to suggest that this trend was
reversed after 18 years of age. It also appeared that those who moved from areas with
higher NO
2
pollution to lower pollution were able to recover lung function for FEV
1

(P=0.008) and MMEF (P=0.01). This effect was also seen with PM
10
for FEV
1
(P=0.03).
These findings confirm results from previous CHS studies for subjects up to 18 years of
age. They also provide new evidence that the deficits accumulated in childhood are not
fully reversed into early adulthood, and that changes in air pollution exposure after age
18 may affect early adulthood lung function.
1

Chapter 1: Introduction
Ambient air pollution continues to pose a problem to several million Southern
California residents. Studies have demonstrated that exposure to both localized and
regional ambient air pollution is associated with diminished pulmonary capacity in
children, reflecting an overall decrease in lung function growth throughout adolescence
[1, 6-8]. These findings have been confirmed by the Children’s Health Study (CHS) as
well as other cross-sectional and longitudinal studies [9-13]. However, little data exists
to document the persistence of poor lung function into adulthood for those children living
in polluted environments.
Researchers frequently use lung function measurements as an indicator of overall
lung health. Deficiencies in these tests are especially important to note since deficits in
lung function can indicate future risk of chronic respiratory disease, such as asthma [1].
Diminished lung function at an early age has also been shown to be related to morbidity
and mortality in adulthood [2-5]. Several measurements of lung function can be acquired
quickly and non-invasively by blowing air into a mouthpiece connected to a spirometer.
These measurements include forced expiratory volume (FEV
1
, the amount of air forcibly
expelled from the lungs in the first second of exhalation) and maximal mid-expiratory
flow (MMEF, the expiratory flow measured over the middle portion of a forced
exhalation).
Children are an especially important study population, as the effects of pollution
during lung development can have serious and potentially long-lasting effects. Normal
lung function in children increases linearly, along with height, until adolescence. During
this adolescent phase, which begins around age 10 years in females and age 12 years in
2

males, there is a rapid non-linear increase of lung function growth [6]. Exposure to air
pollution may begin a chain of reactions that could lead to retarded pulmonary growth,
and possibly permanent loss of lung function, even in environments that meet current
EPA air quality standards [7].
Several factors—most of them measurable—potentially affect overall pulmonary
function. Tobacco smoke, allergens, household and occupational risks, and indoor and
outdoor air pollution have been shown to have a detrimental effect on the lungs [8]. It is
especially important to monitor these risk factors in areas where hazards are high, as in
Southern California, as they put the adolescent population at a greater risk. Effects of
pollution in the Southern California region have been well-documented by the CHS, a
multi-year longitudinal analysis of the lung function development in children throughout
this area. Initial cross-sectional findings showed that females experienced losses in lung
function in response to increased pollutant levels [9]. After four years of follow-up,
significant deficits in pulmonary function in response to ambient pollution were observed
for both males and females in three cohorts of 3,035 total children [10]. This effect was
replicated in a later cohort of 1,678 children, also followed for four years [11], and a
cohort of 1,759 children followed for eight years [12]. In these studies it was additionally
shown that local pollution exposures, such as those caused by motor vehicle traffic on
roads and freeways, resulted in lung function deficits for those living near these pollution
sources [13]. It has also been shown that long-term exposure to pollution resulted in
measurable lung function deficits between ages 9 and 18 years—the start of early
adulthood, when lung growth is completed in girls and nearly completed in boys [14].
3

It has been hypothesized, although not directly tested, that deficits in lung
function development during childhood and adolescence persist into adulthood [14, 15].
As part of the CHS, new lung function data was obtained for various subjects post-
childhood (ages 19 to 30 years). This thesis will analyze observed deficits in lung
function growth attributable to air pollution exposure during childhood to determine
whether these trends persist into young adulthood.
4

Chapter 2: Methods

Study Recruitment, Participation, and Measurements
Study subjects consisted of the first four cohorts of children enrolled into the
CHS. Children were recruited from twelve middle-income communities in the Southern
California area in 1993 and 1996. Four study cohorts were created: children enrolled into
the study from grades 10, 7, and 4 in 1993, (cohorts A, B, and C, respectively) and an
additional cohort of grade 4 children enrolled into the study in 1996 (cohort D).
Upon study entry, a residential and medical history survey was conducted for
each child through a self-administered written questionnaire. The survey asked about
several exposure variables of interest, including evidence of recent water damage in the
home, travel routes to and from school, and the use of home air conditioning, heating, or
fireplaces.
Evaluations of lung function were obtained annually through high school
graduation. Maximal forced expiratory flow-volume maneuvers were performed using
one of six rolling-seal spirometers in order to test for several measures of lung function
including forced expiratory volume (FEV
1
) and maximal mid-expiratory flow (MMEF).
These two measures were chosen as they represent the health of large and small airways
in the lungs, respectively, and are consistent with outcomes analyzed in previous CHS
studies. Additional details of lung function testing have been previously reported [9].
Ambient air pollution data were obtained from regional monitoring stations
located in each of the twelve study communities. A broad range of pollutants were
measured, including a continuous sampling of nitrogen dioxide (NO
2
) and ozone (O
3
) to
obtain hourly averages, hourly readings of particulate matter of size 10 microns or more
5

(PM
10
), and two-week integrated measurements of particulate matter of size 2.5 microns
or more (PM
2.5
), elemental carbon (EC), and strong (hydrochloric and nitric) and weak
(acetic and formic) acid vapor (acid). Annual averages of the 24-hour PM
10
and NO
2

averages were computed, as well as annual averages of the 2-week PM
2.5
, EC, and acid
averages. For O
3
, the computed annual average was restricted to include only data
collected between 10:00 A.M. and 6:00 P.M. as it better captures exposure to the
pollutant, which fluctuates throughout the day. The summary measure of each pollutant
level used in the statistical analysis is the mean of the pollutant’s annual averages from
1994 to 2004, in each community. Additional details of pollution exposure assessment
have been reported in previous CHS publications [9].
Subjects in all four cohorts were followed up to age 18, with expiratory tests
performed annually. Between high school graduation at age 18 and about twelve years
after the conclusion of the first CHS (by which time the oldest subjects were about 30
years of age), a subset of subjects were contacted and participated in additional lung
function testing. This study, termed APEX (Achieved Peak Exhalation), was aimed at
determining the growth trajectory of lung function into young adulthood and identifying
whether identified pollution-related deficits in growth persisted beyond age 18. Subjects
were recruited according to a specified priority, with oldest cohorts and females being
chosen first, as they have already achieved their peak lung function growth. Due to the
logistical, financial, and time constraints on the project, the operational boundary of the
APEX study was limited to subjects living within 500 miles of Los Angeles at the time of
follow-up testing. This included subjects residing in their original childhood communities
as well as subjects who had moved from the original twelve CHS study areas.
6

The general protocol for APEX lung function testing remained largely unchanged
from the previous CHS testing. In order to accommodate testing of subjects at locations
throughout California, smaller and more portable spirometers (Morgan Scientific
ScreenStar) were used. These instruments were selected based upon careful evaluation of
commercial alternatives and in consideration of continuity of data collection, as CHS
lung function testing had been performed on an earlier generation of Morgan Scientific
spirometers. To correct for measurement differences between the two generations of
spirometers, a model was built to translate the lung function measurements obtained in
the APEX study to the scale of measurement in the CHS studies. This model was based
on data that included measurements of lung function on subjects measured by both
devices. The model was implemented to scale the APEX measurements used in this
analysis. At the time of APEX testing subjects were also asked to complete a written
survey seeking information about possible residential or occupational exposures for the
time period since high school. The number of lung function tests available for study use
from each of the participating cohorts is presented in Table 1.
Statistical Methods
Descriptive analyses were used to track lung function over time and estimate
average FEV
1
and MMEF performance curves for males and females in the study. To
account for non-linearity in the overall growth curves, a linear spline method was
utilized, estimating an intercept and regression slope between seven age knot points (two-
year intervals between ages 12 and 24 years, inclusively). This method was implemented
for the overall data and for each specific community. In addition to the overall curves, a
prediction model was created to determine each subject’s expected lung function values
7

Table 1. PFT frequencies by age and study cohort. Annual testing was performed at age 18
years and younger.
Males Females Total
Age A B C D A B C D All Cohorts C and D
10 0 0 816 987 0 0 912 1072 3787 3787
11 0 0 941 784 0 0 977 795 3497 3497
12 0 25 854 717 0 44 906 770 3316 3247
13 0 358 710 732 0 466 735 795 3796 2972
14 0 439 610 699 0 529 661 702 3640 2672
15 32 400 558 627 43 470 627 718 3475 2530
16 374 317 596 577 448 354 665 612 3943 2450
17 392 260 516 502 424 292 580 548 3514 2146
18 318 212 402 334 311 238 435 355 2605 1526
19 44 23 49 25 25 18 28 130 342 232
20 2 0 0 77 0 0 0 78 157 155
21 0 0 51 33 0 0 78 17 179 179
22 0 0 83 1 0 1 98 0 183 182
23 0 11 69 0 0 18 74 0 172 143
24 0 39 15 0 0 53 4 0 111 19
25 0 36 0 0 0 43 1 0 80 1
26+ + + + 118 39 0 0 168 40 0 0 365 0

8

at each age. Additional predictors in this model included respiratory illness on test date,
lifetime history of doctor-diagnosed asthma, field technician, exercise on test date, log-
height, body mass index (bmi), bmi
2
, and race and ethnicity. Prediction models for
MMEF and FEV
1
were created separately for males and females. Because of sparse
sample size at older ages, all subjects age 26 or greater were grouped into the same age
category.
The primary outcomes, termed ΔFEV
1
and ΔMMEF, were calculated by taking
the difference between the corresponding observed lung function measure and the
expected value for each pulmonary test, the latter computed according to the growth
curve prediction model. These differences were computed for two different scenarios. In
one, a prediction model (“all cohorts”) was created utilizing data from all four cohorts,
with predictions computed for all cohorts. In the other, a prediction model (“C and D”)
using data from only cohorts C and D was created, with predictions applied to these two
cohorts. The overall growth curves for the “all cohorts” model closely matched the
growth curves for the “C and D” model. In addition, subsequent analyses and regressions
provided similar parameter estimates for the “C and D” and “all cohorts” method.
Because the scale of pollution levels varied widely among pollutants (e.g., a ten-
fold range for elemental carbon and a four-fold range for PM
10
) each pollutant was scaled
by its range. Dividing pollution values by their respective range (the difference between
the maximum and minimum pollutant concentration value) and using this quantity in lung
function models provided estimates that were more interpretable across pollutants.
Therefore the reported pollutant effects are interpretable as the difference in lung function
comparing those exposed to the highest vs. lowest recorded pollutant concentration.
9

The lung function measures ΔFEV
1
and ΔMMEF were averaged by town and age
category. These lung function averages were then regressed on town mean pollution
level to produce an age- and pollutant-specific estimate (β) of the effect of pollution on
ΔFEV
1
or ΔMMEF. In addition to the two-step regression approach above, we fit a
single mixed model that included a random effects for town. Parameter estimates from
this model also reflected the relationship between lung function and pollutant at each age,
with negative estimates indicating an inverse relationship between pollution levels and
lung function at a specific age.
These β estimates were plotted as a function of age for each of the six studied
pollutants: PM
10
, NO
2
, O
3
, PM
2.5
, EC, and acid. Of particular interest was whether these
parameter estimates changed after age 18, which would suggest a modification of trends
established in childhood. This was analyzed by fitting a regression of these parameter
estimates on age, weighted by their inverse standard error squared. This model was
parameterized to allow separate trends in pollution effects before and after age 18 years.
In other words, the model was able to detect a trend in β up to age 18 years, and then
estimate a different trend after age 18 years.
In addition to the mixed effects model, lung function estimates at age 18 (ΔFEV
18

and ΔMMEF
18
) were directly compared to the APEX measurements (ΔFEV
APEX
and
ΔMMEF
APEX
) for each pollutant. This approach determined whether children with
deficits in lung function at age 18 continued to have deficits in adulthood, or if there was
some “catch up” growth. This relationship was examined overall, and by levels of each
pollutant to allow for differing trends based on increasing pollution. It was also
examined by gender and town-wide pollution levels. For the purposes of analysis, the 12
10

original CHS communities were grouped into six “lower” and six “higher” pollution
towns based on historical monitoring data collected in the course of CHS testing, with
Lake Elsinore, Mira Loma, Long Beach, Riverside, San Dimas, and Upland classified as
high-pollution towns. The difference between the APEX and age 18 lung measurements
were computed, and regressed against CHS pollution levels, to determine if pollution
affects the magnitude of any “catch up” growth. In order to examine the performance of
subjects who changed pollution exposures in the interim, this analysis was also
performed with the difference in pollution exposure between APEX test date and age 18
as the independent variable.
11

Chapter 3: Results
Baseline Growth Curves
The population-averaged growth curves produced findings similar to that from
previous published CHS studies [9-12]. Lung function increased sharply during
adolescence (age 10-14) for both males and females. In males, lung function growth
slowed after age 18, but appeared to still increase beyond age 20 (Figures 1a, 2a). Lung
function growth slowed and appeared to level off beyond age 20 in females. Growth
curves restricted to the C and D cohorts produced results similar to the growth curves for
all cohorts (Figures 1b, 2b). In each case, lung function estimates using FEV produced
more stable curves than MMEF.
Lung Function and Pollution
Age-specific twelve-point plots of community-averaged lung function versus
community-averaged pollution generally showed a negative association for each
pollutant. That is, subjects in higher-polluted areas tended to perform worse with regard
to lung function than those in lower-polluted areas. The exception to this was O
3
levels,
which generally showed a positive relationship. The regression relationship (
t pollu age tan ,
β )
between lung function and pollution appeared to grow stronger in magnitude with
increasing age, up to age 18 (Figures 3a-b, 4a-b, Tables 2a-b, 3a-b). Trends were less
apparent after age 18 because of the reduction in sample size, which led to larger standard
errors on beta estimates (Figure 3a-b). Age-specific betas for the “all cohort” model were
similar for those in the “C and D” model, up to age 23 (Figure 4a-b).
12

Figure 1a. FEV
1
growth curve (all cohorts).

Figure 1b. FEV
1
growth curve (C and D cohorts).

0
1000
2000
3000
4000
5000
10 12 14 16 18 20 22 24 26+
Age (years)
FEV
1
(ml)
Females
Males
0
1000
2000
3000
4000
5000
10 12 14 16 18 20 22 24 26+
Age (years)
FEV
1
(ml)
Females
Males
13

Figure 2a. MMEF growth curve (all cohorts).

Figure 2b. MMEF growth curve (C and D cohorts).

0
1000
2000
3000
4000
5000
6000
10 12 14 16 18 20 22 24 26+
Age (years)
MMEF
(ml/s)
Females
Males
0
1000
2000
3000
4000
5000
6000
10 12 14 16 18 20 22 24 26+
Age (years)
MMEF
(ml/s)
Females
Males
14

Figure 3a. Beta coefficients from the regression of ΔFEV
1
on pollution, across age categories (all cohorts).
Error bars represent one standard error.

O 3
-300
-100
100
300
10 12 14 16 18 20 22 24 26+
Age (years)
β β β β
P M 1 0
-300
-100
100
300
10 12 14 16 18 20 22 24 26+
Age (years)
β β β β
N O 2
-300
-100
100
300
10 12 14 16 18 20 22 24 26+
Age (years)
β β β β
15

Figure 3a (continued). Beta coefficients from the regression of ΔFEV
1
on pollution, across age categories
(all cohorts). Error bars represent one standard error.

P M 2 . 5
-300
-100
100
300
10 12 14 16 18 20 22 24 26+
Age (years)
β β β β
A c i d
-300
-100
100
300
10 12 14 16 18 20 22 24 26+
Age (years)
β β β β
E C
-300
-100
100
300
10 12 14 16 18 20 22 24 26+
Age (years)
β β β β
16

Figure 3b. Beta coefficients from the regression of ΔFEV
1
on pollution, across age categories (C and D
cohorts). Error bars represent one standard error.

O 3
-300
-100
100
300
10 12 14 16 18 20 22
Age (years)
β β β β
P M 1 0
-300
-100
100
300
10 12 14 16 18 20 22
Age (years)
β β β β
N O 2
-300
-100
100
300
10 12 14 16 18 20 22
Age (years)
β β β β
17

Figure 3b (continued). Beta coefficients from the regression of ΔFEV
1
on pollution, across age categories
(C and D cohorts). Error bars represent one standard error.

P M 2 . 5
-300
-100
100
300
10 12 14 16 18 20 22
Age (years)
β β β β
A c i d
-300
-100
100
300
10 12 14 16 18 20 22
Age (years)
β β β β
E C
-300
-100
100
300
10 12 14 16 18 20 22
Age (years)
β β β β
18

Figure 4a. Beta coefficients from the regression of ΔMMEF on pollution, across age categories (all
cohorts). Error bars represent one standard error.

O 3
-800
-400
0
400
800
10 12 14 16 18 20 22 24 26+
Age (years)
β β β β
P M 1 0
-800
-400
0
400
800
10 12 14 16 18 20 22 24 26+
Age (years)
β β β β
N O 2
-800
-400
0
400
800
10 12 14 16 18 20 22 24 26+
Age (years)
β β β β
19

Figure 4a (continued). Beta coefficients from the regression of ΔMMEF on pollution, across age categories
(all cohorts). Error bars represent one standard error.

P M 2 . 5
-800
-400
0
400
800
10 12 14 16 18 20 22 24 26+
Age (years)
β β β β
A c i d
-800
-400
0
400
800
10 12 14 16 18 20 22 24 26+
Age (y ears)
β β β β
E C
-800
-400
0
400
800
10 12 14 16 18 20 22 24 26+
Age (years)
β β β β
20

Figure 4b. Beta coefficients from the regression of ΔMMEF on pollution, across age categories (C and D
cohorts). Error bars represent one standard error.

O 3
-800
-400
0
400
800
10 12 14 16 18 20 22
Age (years)
β β β β
P M 1 0
-800
-400
0
400
800
10 12 14 16 18 20 22
Age (years)
β β β β
N O 2
-800
-400
0
400
800
10 12 14 16 18 20 22
Age (years)
β β β β
21

Figure 4b (continued). Beta coefficients from the regression of ΔMMEF on pollution, across age
categories (C and D cohorts). Error bars represent one standard error.

P M 2 . 5
-800
-400
0
400
800
10 12 14 16 18 20 22
Age (years)
β β β β
A c i d
-800
-400
0
400
800
10 12 14 16 18 20 22
Age (years)
β β β β
E C
-800
-400
0
400
800
10 12 14 16 18 20 22
Age (years)
β β β β
22

Table 2a. Mixed model regression of ΔFEV
1
on pollution for each age category (all cohorts)*.
O3 NO2 ACID PM 10 PM 2.5 EC
Age β β β β P β β β β P β β β β P β β β β P β β β β P β β β β P
10 37.9 0.16 -29.5 0.25 -15.7 0.55 -43.7 0.14 -31.9 0.20 -31.0 0.16
11 40.3 0.11 -29.7 0.22 -17.5 0.49 -38.0 0.19 -27.7 0.24 -28.1 0.19
12 37.5 0.23 -27.8 0.35 -18.5 0.54 -31.9 0.37 -21.5 0.46 -26.8 0.30
13 20.6 0.62 -35.5 0.34 -37.1 0.31 -58.3 0.17 -40.0 0.26 -40.1 0.21
14 0.5 0.99 -21.4 0.58 -35.8 0.33 -53.6 0.21 -34.6 0.34 -27.4 0.41
15 -4.5 0.94 -93.7 0.04 -110.4 0.01 -130.7 0.01 -93.9 0.03 -84.8 0.04
16 13.1 0.76 -58.0 0.11 -64.0 0.07 -101.9 0.01 -71.9 0.03 -58.7 0.06
17 32.4 0.42 -87.5 0.001 -75.8 0.01 -98.9 0.00 -77.8 0.007 -79.2 0.007
18 2.8 0.95 -63.7 0.12 -74.1 0.05 -102.6 0.02 -77.1 0.04 -55.9 0.12
19 -5.8 0.95 -106.4 0.18 -103.2 0.17 -220.8 0.03 -163.5 0.04 -118.9 0.10
20 130.6 0.35 33.7 0.79 46.5 0.71 -130.0 0.39 -53.7 0.66 -34.9 0.76
21 -341.1 0.02 121.5 0.30 -16.0 0.89 11.8 0.92 44.1 0.66 80.8 0.43
22 91.9 0.45 -148.2 0.19 -106.9 0.36 -69.3 0.55 -63.4 0.54 -123.8 0.21
23 19.9 0.89 -93.8 0.48 -124.4 0.34 -188.8 0.16 -138.6 0.24 -129.4 0.26
24 -82.7 0.61 -10.3 0.95 10.8 0.95 193.0 0.33 126.7 0.45 57.7 0.70
25 -108.2 0.55 155.8 0.37 171.2 0.33 298.6 0.07 217.9 0.15 213.8 0.14
26+ -17.2 0.87 69.5 0.52 27.5 0.80 9.2 0.94 35.6 0.73 59.5 0.52

* β estimates represent the change in ΔFEV
1
(ml) corresponding to a one-range increase in each pollutant.

23

Table 2b. Mixed model regression of ΔFEV
1
on pollution for each age category (C and D cohorts)*.
O3 NO2 ACID PM 10 PM 2.5 EC
Age** β β β β P β β β β P β β β β P β β β β P β β β β P β β β β P
10 37.6 0.16 -30.1 0.24 -16.5 0.53 -44.2 0.13 -32.4 0.19 -31.0 0.16
11 39.8 0.11 -29.3 0.23 -17.3 0.49 -38.5 0.18 -28.2 0.23 -27.6 0.20
12 38.4 0.25 -25.2 0.43 -16.9 0.60 -32.9 0.38 -21.0 0.50 -25.2 0.37
13 32.8 0.45 -42.2 0.28 -38.5 0.32 -60.0 0.18 -38.2 0.31 -44.3 0.19
14 17.5 0.67 -37.0 0.31 -40.3 0.26 -56.0 0.19 -34.7 0.34 -40.0 0.21
15 10.8 0.87 -100.1 0.05 -110.4 0.02 -136.8 0.02 -92.0 0.07 -93.9 0.03
16 -7.2 0.87 -44.0 0.25 -61.5 0.08 -96.8 0.01 -63.4 0.07 -46.4 0.16
17 9.0 0.85 -92.9 0.006 -92.1 0.006 -113.3 0.005 -85.8 0.01 -85.4 0.003
18 -19.1 0.68 -73.0 0.06 -92.3 0.02 -95.6 0.05 -78.5 0.04 -59.4 0.09
19 -27.2 0.80 -130.8 0.18 -112.9 0.22 -190.6 0.11 -154.7 0.10 -142.1 0.11
20 111.3 0.45 44.3 0.73 41.3 0.75 -134.5 0.38 -55.3 0.66 -26.5 0.82
21 -334.9 0.02 135.0 0.26 -1.3 0.99 18.6 0.87 52.8 0.61 89.5 0.38
22 93.0 0.45 -142.7 0.21 -95.9 0.41 -56.2 0.63 -56.4 0.58 -112.7 0.26
23 -11.1 0.94 -75.5 0.60 -123.2 0.35 -182.2 0.17 -124.3 0.29 -123.3 0.28

* β estimates represent the change in ΔFEV
1
(ml) corresponding to a one-range increase in each pollutant.
** Data for age categories 24 and 25 were collected but not presented, as these age categories were
represented by only 10 and 1 towns, respectively.

24

Table 3a. Mixed model regression of ΔMMEF on pollution for each age category (all cohorts)*.
O3 NO2 ACID PM 10 PM 2.5 EC
Age β β β β P β β β β P β β β β P β β β β P β β β β P β β β β P
10 101.1 0.07 -42.9 0.46 -11.5 0.84 -68.9 0.31 -49.3 0.38 -44.3 0.38
11 62.9 0.21 -18.7 0.71 -2.3 0.96 -55.3 0.33 -37.9 0.42 -27.0 0.53
12 47.7 0.44 -44.9 0.44 -43.4 0.45 -101.7 0.11 -75.5 0.15 -52.0 0.30
13 27.6 0.72 -42.9 0.55 -50.8 0.47 -100.1 0.21 -69.5 0.30 -53.9 0.38
14 64.7 0.35 -30.3 0.65 -31.2 0.64 -97.3 0.19 -57.5 0.36 -41.1 0.48
15 42.0 0.69 -124.0 0.17 -135.0 0.12 -210.5 0.03 -141.5 0.09 -115.4 0.14
16 65.0 0.50 -103.0 0.24 -97.9 0.26 -173.0 0.08 -131.1 0.10 -84.0 0.28
17 40.1 0.65 -133.4 0.06 -120.1 0.09 -171.9 0.04 -139.2 0.04 -113.4 0.07
18 30.9 0.78 -94.8 0.34 -112.2 0.25 -211.7 0.05 -142.0 0.13 -93.0 0.29
19 100.4 0.56 -137.7 0.39 -69.3 0.66 -163.4 0.43 -206.9 0.20 -108.0 0.46
20 409.5 0.28 196.5 0.56 347.0 0.30 -88.5 0.83 -24.0 0.94 96.0 0.76
21 -826.5 0.004 355.7 0.16 18.2 0.95 -33.3 0.91 28.8 0.91 215.5 0.37
22 -139.9 0.59 -4.7 0.99 -136.3 0.57 -136.7 0.60 -34.4 0.88 -36.9 0.87
23 -56.0 0.84 -17.2 0.94 -38.9 0.87 -67.2 0.78 -70.2 0.74 -37.4 0.86
24 -7.4 0.98 -298.3 0.44 -244.4 0.52 -307.2 0.48 -302.1 0.41 -261.9 0.41
25 -318.9 0.32 201.2 0.52 120.2 0.70 221.2 0.46 170.1 0.53 167.5 0.52
26+ -135.0 0.47 224.7 0.23 110.8 0.56 245.4 0.25 254.3 0.15 253.0 0.11

* β estimates represent the change in ΔMMEF (ml/s) corresponding to a one-range increase in each
pollutant.
25

Table 3b. Mixed model regression of ΔMMEF on pollution for each age category (C and D cohorts)*.
O3 NO2 ACID PM 10 PM 2.5 EC
Age** β β β β P β β β β P β β β β P β β β β P β β β β P β β β β P
10 100.0 0.07 -41.9 0.47 -12.1 0.84 -70.1 0.29 -49.4 0.37 -42.7 0.40
11 60.0 0.24 -15.5 0.76 -0.6 0.99 -54.4 0.34 -37.2 0.43 -23.8 0.59
12 48.5 0.47 -39.8 0.52 -41.0 0.50 -100.3 0.15 -72.9 0.20 -48.3 0.37
13 72.1 0.40 -59.8 0.46 -47.0 0.56 -86.1 0.36 -60.8 0.43 -60.0 0.39
14 109.2 0.09 -61.2 0.34 -37.9 0.56 -104.8 0.15 -67.1 0.28 -63.1 0.26
15 68.5 0.51 -121.2 0.18 -124.3 0.17 -195.7 0.06 -126.3 0.15 -111.8 0.16
16 43.5 0.62 -43.9 0.59 -52.2 0.51 -120.2 0.19 -90.1 0.23 -35.5 0.63
17 68.5 0.41 -159.6 0.02 -125.4 0.08 -163.2 0.05 -146.1 0.03 -129.2 0.03
18 17.6 0.88 -91.1 0.38 -97.4 0.34 -133.1 0.29 -104.6 0.30 -66.6 0.48
19 202.9 0.38 -116.7 0.58 22.8 0.91 13.2 0.96 -108.6 0.60 -54.7 0.78
20 315.5 0.41 240.5 0.46 338.8 0.30 -94.9 0.81 -18.5 0.95 134.6 0.66
21 -800.7 0.005 359.1 0.13 37.8 0.89 -14.3 0.96 40.5 0.87 219.2 0.34
22 -139.0 0.56 -5.9 0.98 -122.3 0.59 -111.2 0.63 -36.3 0.86 -29.5 0.88
23 -99.3 0.71 8.1 0.97 -51.7 0.83 -93.4 0.70 -78.3 0.72 -28.5 0.89

* β estimates represent the change in ΔMMEF (ml/s) corresponding to a one-range increase in each
pollutant.
** Data for age categories 24 and 25 were collected but not presented, as these age categories were
represented by only 10 and 1 towns, respectively.
26

When these
t pollu age tan ,
β were regressed against age, the corresponding slope was
generally negative for those less than 18 years of age (Tables 4a-b, 5a-b). This supported
results showing an increasing magnitude of pollutant effect up to age 18. Allowing
flexibility beyond age 18 showed that, for the C and D cohort model, there appeared to be
no change in the
t pollu age tan ,
β after age 18. In the full-cohort model, the
t pollu age tan ,
β appear
to reverse the trend established before age 18.
A total of 1,071 subjects in all cohorts and 670 subjects in the C and D cohorts
had both a lung function test at age 18 and at least one during the APEX study period. A
regression of APEX lung function measurement against age 18 lung function
measurement suggested that an individual’s lung function estimate at age 18 was a good
indicator of lung function estimate at APEX test date (P<0.0001 for FEV and MMEF;
Figures 5a-b, 6a-b). That is, individuals with a pulmonary deficit at age 18 tended to
retain this deficit at APEX test date, and vice versa. There did not appear to be any
indication that this effect was different for those with a lung function deficit compared to
those performing better than expected. Results did not appear to change after
stratification by gender or town pollution levels (Figures 5c-f, 6c-f). It did appear, in the
C and D cohorts, that FEV
1
at age 18 was a better predictor of FEV
1
at APEX test date
for females, compared to males (Figure 5e).
When the difference between the APEX measurement and the age 18
measurements were plotted and regressed as a function of levels of each pollutant, there
was generally no statistically significant association. There were only two significant
relationships for FEV—increasing levels of NO
2
appeared to also increase the
27

Figure 5a. FEV
1
residual at APEX measurement plotted against residual at age 18 (all cohorts).
y = 0.75x + 4.81
R
2
= 0.49
-2500
-2000
-1500
-1000
-500
0
500
1000
1500
2000
2500
-2500 -1500 -500 500 1500 2500
A ge 18
A PE X

28

Figure 5b. FEV
1
residual at APEX measurement plotted against residual at age 18 (C and D cohorts).
y = 0.73x + 18.91
R
2
= 0.51
-2500
-2000
-1500
-1000
-500
0
500
1000
1500
2000
2500
-2500 -1500 -500 500 1500 2500
A ge 18
A PE X

29

Figure 5c. FEV
1
residual at APEX measurement plotted against residual at age 18 (all cohorts), for males
(top) and females (bottom).
y = 0.75x + 20.02
R
2
= 0.49
-2500
-2000
-1500
-1000
-500
0
500
1000
1500
2000
2500
-2500 -1500 -500 500 1500 2500
A ge 18
A PE X

y = 0.76x - 7.96
R
2
= 0.50
-2500
-2000
-1500
-1000
-500
0
500
1000
1500
2000
2500
-2500 -1500 -500 500 1500 2500
A ge 18
A PE X
30

Figure 5d. FEV
1
residual at APEX measurement plotted against residual at age 18 (all cohorts), for low-
pollution (top) and high-pollution (bottom) communities.
y = 0.75x - 8.02
R
2
= 0.49
-2500
-2000
-1500
-1000
-500
0
500
1000
1500
2000
2500
-2500 -1500 -500 500 1500 2500
A ge 18
A PE X

y = 0.75x + 15.84
R
2
= 0.49
-2500
-2000
-1500
-1000
-500
0
500
1000
1500
2000
-2500 -1500 -500 500 1500 2500
A ge 18
A PE X

31

Figure 5e.
FEV1
residual at APEX measurement plotted against residual at age (C and D cohorts), for males
(top) and females (bottom).
y = 0.68x + 39.71
R
2
= 0.46
-2500
-2000
-1500
-1000
-500
0
500
1000
1500
2000
2500
-2500 -1500 -500 500 1500 2500
A ge 18
A PE X

y = 0.81x - 1.03
R
2
= 0.58
-2500
-2000
-1500
-1000
-500
0
500
1000
1500
2000
2500
-2500 -1500 -500 500 1500 2500
A ge 18
A PE X
32

Figure 5f. FEV
1
residual at APEX measurement plotted against residual at age 18 (C and D cohorts), for
low-pollution (top) and high-pollution (bottom) communities.
y = 0.74x + 9.33
R
2
= 0.52
-2500
-2000
-1500
-1000
-500
0
500
1000
1500
2000
2500
-2500 -1500 -500 500 1500 2500
A ge 18
A PE X

y = 0.74x + 26.38
R
2
= 0.50
-2500
-2000
-1500
-1000
-500
0
500
1000
1500
2000
-2500 -1500 -500 500 1500 2500
A ge 18
A PE X
33

Figure 6a. MMEF residual at APEX measurement plotted against residual at age 18 (all cohorts).
y = 0.69x + 40.30
R
2
= 0.54
-4500
-3500
-2500
-1500
-500
500
1500
2500
3500
4500
-4500 -3500 -2500 -1500 -500 500 1500 2500 3500 4500
A ge 18
A PE X

34

Figure 6b. MMEF residual at APEX measurement plotted against residual at age 18 (C and D cohorts).
y = 0.69x + 63.89
R
2
= 0.54
-4500
-3500
-2500
-1500
-500
500
1500
2500
3500
4500
-4500 -3500 -2500 -1500 -500 500 1500 2500 3500 4500
A ge 18
A PE X

35

Figure 6c. MMEF residual at APEX measurement plotted against residual at age 18 (all cohorts), for males
(top) and females (bottom).
y = 0.76x + 87.66
R
2
= 0.56
-4500
-3500
-2500
-1500
-500
500
1500
2500
3500
4500
-4500 -3500 -2500 -1500 -500 500 1500 2500 3500 4500
A ge 18
A PE X

y = 0.60x + 12.69
R
2
= 0.54
-4500
-3500
-2500
-1500
-500
500
1500
2500
3500
4500
-4500 -3500 -2500 -1500 -500 500 1500 2500 3500 4500
A ge 18
A PE X
36

Figure 6d. MMEF residual at APEX measurement plotted against residual at age 18 (all cohorts), for low-
pollution (top) and high-pollution (bottom) communities.
y = 0.65x + 0.002
R
2
= 0.51
-4500
-3500
-2500
-1500
-500
500
1500
2500
3500
4500
-4500 -3500 -2500 -1500 -500 500 1500 2500 3500 4500
A ge 18
A PE X

y = 0.75x + 15.84
R
2
= 0.49
-2500
-2000
-1500
-1000
-500
0
500
1000
1500
2000
-2500 -1500 -500 500 1500 2500
A ge 18
A PE X

37

Figure 6e. MMEF residual at APEX measurement plotted against residual at age 18 (C and D cohorts), for
males (top) and females (bottom).
y = 0.75x + 106.7
R
2
= 0.53
-4500
-3500
-2500
-1500
-500
500
1500
2500
3500
4500
-4500 -3500 -2500 -1500 -500 500 1500 2500 3500 4500
A ge 18
A PE X

y = 0.64x + 36.8
R
2
= 0.57
-4500
-3500
-2500
-1500
-500
500
1500
2500
3500
4500
-4500 -3500 -2500 -1500 -500 500 1500 2500 3500 4500
A ge 18
A PE X
38

Figure 6f. MMEF residual at APEX measurement plotted against residual at age 18 (C and D cohorts), for
low-pollution (top) and high-pollution (bottom) communities.
y = 0.65x + 16.42
R
2
= 0.50
-4500
-3500
-2500
-1500
-500
500
1500
2500
3500
4500
-4500 -3500 -2500 -1500 -500 500 1500 2500 3500 4500
A ge 18
A PE X

y = 0.74x + 99.69
R
2
= 0.59
-4500
-3500
-2500
-1500
-500
500
1500
2500
3500
4500
-4500 -3500 -2500 -1500 -500 500 1500 2500 3500 4500
A ge 18
A PE X

39

Table 4a. Regression of age- and pollutant-specific β on age (FEV
1
, all cohorts)*.
O3 NO2 ACID PM 10 PM 2.5 EC
Age Est. P Est. P Est. P Est. P Est. P Est. P
Intercept 82.8 0.10 53.4 0.16 86.1 0.02 82.5 0.12 64.5 0.09 46.4 0.19
Age
≤18
-4.3 0.24 -7.5 0.01 -9.6 0.001 -11.2 0.008 -8.5 0.006 -6.9 0.01
Age
>18
-6.7 0.42 16.9 0.03 14.2 0.03 23.1 0.02 18.2 0.01 16.5 0.02

*Coefficients reflect β values. For example, each year increase in age (up to age 18) corresponds to a 4.3
unit decrease in β, for O3.
Age
≤18
: Coefficient for a one-unit increase in age, years 18 and younger.
Age
>18
: Coefficient for a one-unit increase in age, years 19 and older.

Table 4b. Regression of age- and pollutant-specific β on age (FEV
1
, C and D cohorts)*.
O3 NO2 ACID PM 10 PM 2.5 EC
Age Est. P Est. P Est. P Est. P Est. P Est. P
Intercept 116.0 0.04 47.7 0.26 97.8 0.007 70.1 0.09 54.7 0.10 38.0 0.28
Age
≤18
-7.1 0.09 -7.1 0.04 -10.7 0.001 -10.3 0.004 -7.6 0.006 -6.2 0.03
Age
>18
-4.2 0.80 10.7 0.45 7.0 0.48 6.2 0.58 6.4 0.50 2.2 0.85

*Coefficients reflect β values. For example, each year increase in age (up to age 18) corresponds to a 7.1
unit decrease in β, for O3.
Age
≤18
: Coefficient for a one-unit increase in age, years 18 and younger.
Age
>18
: Coefficient for a one-unit increase in age, years 19 and older.
40

Table 5a. Regression of age- and pollutant-specific β on age (MMEF, all cohorts)*.
O3 NO2 ACID PM 10 PM 2.5 EC
Age Est. P Est. P Est. P Est. P Est. P Est. P
Intercept 139.8 0.21 103.4 0.17 149.4 0.02 147.1 0.02 121.8 0.03 73.0 0.19
Age
≤18
-6.2 0.46 -12.3 0.04 -15.2 0.003 -19.8 <0.001 -15.5 0.001 -10.1 0.02
Age
>18
-30.7 0.09 39.4 0.004 25.3 0.02 45.9 <0.001 38.9 <0.001 35.6 <0.001

*Coefficients reflect β values. For example, each year increase in age (up to age 18) corresponds to a 6.2
unit decrease in β, for O3.
Age
≤18
: Coefficient for a one-unit increase in age, years 18 and younger.
Age
>18
: Coefficient for a one-unit increase in age, years 19 and older.

Table 5b. Regression of age- and pollutant-specific β on age (MMEF, C and D cohorts)*.
O3 NO2 ACID PM 10 PM 2.5 EC
Age Est. P Est. P Est. P Est. P Est. P Est. P
Intercept 108.2 0.36 104.0 0.20 123.5 0.08 61.1 0.23 75.2 0.05 52.8 0.35
Age
≤18
-2.8 0.76 -12.4 0.05 -12.7 0.02 -12.3 0.01 -11.3 <0.001 -8.3 0.06
Age
>18
-65.4 0.09 54.5 0.03 22.2 0.26 20.8 0.13 22.4 0.04 32.0 0.07

*Coefficients reflect β values. For example, each year increase in age (up to age 18) corresponds to a 2.8
unit decrease in β, for O3.
Age
≤18
: Coefficient for a one-unit increase in age, years 18 and younger.
Age
>18
: Coefficient for a one-unit increase in age, years 19 and older.
41

discrepancy between APEX measurement and age 18 measurement, suggesting those in
neighborhoods with higher values of NO
2
perform better than expected at APEX date
compared to age 18, compared to those with lower pollution values (P=0.01 for all
cohorts, P=0.02 for C and D cohorts; Table 7a). The effect was significant but in the
opposite direction for O
3
. The regression was significant for each pollutant for MMEF
for all cohorts, but the significance was not replicated for the C and D cohort subset
(Table 7b).
The differences between the APEX measurement and age 18 measurement were
also regressed against the differences between pollution exposure levels at APEX test
date and age 18 (that is, pollution increases or decreases between APEX and age 18).
This is because those who grew up in polluted areas were likely to have lower pollution
exposures if they moved, and vice versa (Table 6). This method showed no significant
relationships between pollution exposure changes and lung function changes when
examining only the C and D cohorts. However, it appears as though individuals who
experience an increase in NO
2
and PM
10
levels have lower expected FEV
1
at APEX test
date compared to age 18 (P=0.008 and P=0.03, respectively; Table 7a). This same trend
is seen for NO
2
when examining MMEF (P=0.01; Table 7b).
42

Table 6. Regression of (APEX-18) pollution exposures on pre-18 pollution assignments.
Pollutant Est. P

O3 -0.37 <0.001
NO2 -0.39 <0.001
PM10 -0.29 <0.001
PM2.5 -0.29 <0.001

Table 7a. Regression of the (ΔFEV
APEX
-ΔFEV
18
) difference on age 18, APEX and (APEX-18) pollution
exposures.

All Cohorts C and D Cohorts
Pollutant Est. P Est. P

Age 18 (CHS) Pollution
O3 -96.2 0.009 -70.3 0.13
NO2 84.6 0.01 96.1 0.02
Acid 46.9 0.16 59.1 0.14
PM 10 34.3 0.37 5.4 0.90
PM 2.5 46.8 0.15 28.4 0.46
EC 57.8 0.05 56.0 0.12
APEX Pollution
O3 -3.7 0.02 -2.3 0.20
NO2 0.3 0.87 0.6 0.76
PM10 -0.7 0.50 -0.8 0.47
PM2.5 -0.6 0.78 -0.2 0.94
(APEX-18) Pollution
O3 44.9 0.50 49.3 0.44
NO2 -167.5 0.008 -95.9 0.21
PM10 -176.7 0.03 -51.8 0.61
PM2.5 -115.5 0.07 12.5 0.87

43

Table 7b. Regression of the (ΔMMEF
APEX
-ΔMMEF
18
) difference on age 18, APEX and (APEX-18)
pollution exposures.

All Cohorts C and D Cohorts
Pollutant Est. P Est. P

Age 18 (CHS) Pollution
O3 -165.0 0.02 -56.9 0.55
NO2 174.6 0.009 160.3 0.06
Acid 138.2 0.04 155.0 0.06
PM 10 176.8 0.02 119.4 0.20
PM 2.5 154.2 0.01 111.6 0.15
EC 155.4 0.008 134.3 0.07
APEX Pollution
O3 -2.9 0.31 0.6 0.87
NO2 4.4 0.12 3.1 0.38
PM10 4.0 0.03 2.4 0.26
PM2.5 7.7 0.04 5.3 0.23
(APEX-18) Pollution
O3 326.9 0.009 112.4 0.51
NO2 -303.9 0.01 -118.2 0.43
PM10 -157.1 0.30 62.3 0.75
PM2.5 -160.9 0.18 56.3 0.70

44

Chapter 4: Discussion
Previous analyses from the CHS focused on data collected up to age 18 years.
Those studies showed that deficits in lung function associated with long-term air
pollution exposures persist through age 18 [12]. This study focused on the subsequent
lung function measurements after age 18 and their relationship to both adolescent
pollution exposure and exposure in early adulthood. The following section discusses our
results.
“Betas” Model
The regression of age and pollution-specific β on age generally showed a
decreasing trend in β for all pollutants except O
3
, demonstrating that the relationship
between lung function deficits and ambient pollution levels grows stronger every year up
to age 18. After age 18 the regression through the β coefficients was positive for most
pollutants in the “all cohorts” analysis (possibly indicating some reversal in previously
established trends), but was not significant for the “C and D” analysis. Thus it seems that
pollution was associated with decreases in lung function through age 18, and after age 18
the C and D cohorts did not show significant evidence that these decrements were
reversed.
Correlation between Age 18 and APEX
While not utilizing as much of the data as the β regression method, comparing the
age 18 measurement directly to APEX measurement gives a clearer picture of the direct
association between late childhood and early adulthood lung function. Overall, and
dichotomized by pollution levels, lung function at age 18 was a good predictor of lung
function at APEX test date. For both FEV
1
and MMEF the slope of the regression on
45

APEX date vs. age 18 was negative (around 0.75), which may suggest regression to the
mean as the variation of lung function measurements generally decreases at APEX test
date compared to age 18. Still, this provides additional evidence that deficits in lung
function at age 18 do not completely recover as teens age into their 20s.
With the exception of NO
2
, the difference between ΔFEV
APEX
and ΔFEV
18
was
not significantly related to air pollution in the original CHS communities. Almost all
associations, however, were in the positive direction, which could be evidence of some
catch up growth. For NO
2
, the difference in residuals was significantly greater for larger
values of pollution. This may suggest that children in communities with high values of
NO
2
are able to recover some deficiencies in FEV
1
by early adulthood. These results
were similar for MMEF, but the difference between ΔMMEF
APEX
and ΔMMEF
18
was
significantly related to air pollution for all pollutants in the full cohort. The effect
observed for MMEF (but not for FEV
1
) supports a possible hypothesis of partial lung
recovery, as improvement in small airway function (reflected in MMEF) is seen, with
less improvement observable in large airways (as measured by FEV
1
). However, the
change in pollution exposure after age 18 was inversely correlated with initial CHS
pollution assignment (Table 6). That is, those in heavily polluted areas generally moved
to less polluted areas, and vice versa.
Additional analyses revealed that patterns of lung function were significantly
associated with these pollution exposure changes (Table 7a-b). Therefore, lung function
trajectories during early adulthood showed associations with pollution that were
consistent with earlier observed associations during the adolescent growth period,
especially for the “all cohorts” method. The parameter estimates in the “C and D”
46

method were generally smaller and less significant, probably because those in the
younger cohorts did not have as much time for their lungs to react to any change in
pollution exposure.
When analyzing the change in expected lung function between APEX and age 18
test date as a function of the change in pollution exposure, those who moved to areas with
less NO
2
had partially improved test results for FEV
1
and MMEF, and vice versa. This
may suggest that exposure to NO
2
causes temporary damage that is more immediately
experienced and recovered, while other pollutants act on a deeper level and may retard
lung growth. It could be argued that the data also suggests that moving to an area with
higher O
3
is associated with an improvement in lung function. However, further analysis
(not shown) indicates that the change in O
3
level is highly inversely correlated with the
change in all other pollutant levels (P<0.001). When O
3
and NO
2
were put together into a
multivariable regression, the change in O
3
level was not significantly related to change in
lung function, and the change in NO
2
level was marginally significant (P=0.07). This
supports the explanation that the effects of O
3
are primarily seen due to its strong inverse
correlation with all other pollutants.
Possible Study Bias
These results may be biased for several reasons. First, age groups studied beyond
age 18 did not reflect all CHS study cohorts; the older age categories studied (24-30
years) were dominated by Cohorts A and B, while younger age categories (20-23 years)
were largely composed of cohorts C and D. Because there is a longer lapse in time
between age 18 and APEX test date, those in the A and B cohorts are more likely to
change their exposures through relocation, occupation, and lifestyle than those in cohorts
47

C and D. The finding that no FEV
1
catch-up growth was demonstrated in cohorts C and
D may hold more validity as they have had less time to experience misclassification of
exposures.
Second, selection bias may have been an issue in the presented analyses. This is
because only a subset of the many thousands of CHS subjects could be tested in early
adulthood, as part of the current (APEX) examination. While subjects included in the
APEX study are similar to non-APEX subjects at age 18 for most factors, it does appear
that those included in the APEX study have a different ethnic makeup (Table 8, P=0.02)
and higher value of ΔMMEF (P=0.03). Since the APEX subjects had higher values of
ΔMMEF, it may appear that these children had lung improvement in early adulthood
compared to the overall CHS sample at age 18. Still, this would only be applicable to the
“betas” model and does not affect the findings from the correlation of APEX and age 18
test dates.
Third, large, random variation after age 18 might be driving some of the results
from the “betas” model regression. For example, if the betas were randomly scattered
around 0 after age 18, and there had been a strongly established negative trend, we might
expect to see a positive regression slope after age 18. In this situation there would be a
positive post-18 slope, but it would not necessarily reflect an improvement in lung
function.
Perhaps most importantly, it is difficult to determine an accurate representation of
pollution exposure levels for each individual, as pollution was averaged over the study
period. Thus, there was generally no accounting for communities with air that becomes
progressively dirtier or cleaner. Additionally, there was no way to ensure that children
48

Non-APEX APEX P*
Gender
Female 50% 54% 0.15
Male 50% 46%
Race
Asian 6% 6% 0.07
Black 5% 3%
No Response 2% 1%
Mixed race 7% 7%
Other 15% 19%
White 64% 64%
Ethnicity
No Response 2% 1% 0.02
Hispanic 26% 31%
Non-Hispanic 72% 68%
Diagnosed Asthma
No 85% 87% 0.40
Yes 15% 13%
Smoking in House
No 83% 84% 0.48
Yes 17% 16%
Standardized BMI -0.08 (0.11) 0.23 (0.19) 0.15
Δ FEV
1
18.04 (10.91) -9.72 (16.27) 0.16
Δ MMEF 28.88 (36.47) 124.13 (23.00) 0.03
* P-value from a chi-square (categorical) or t (continuous) test
Table 8. Demographics of APEX and non-APEX children at
age 18 for categorical (percentages) and continuous (means
with standard error) variables.
49

who grew up in one community stayed in that particular community—therefore, the
pollution exposures did not necessarily remain constant throughout the length of the
study. Data do exist that documents pollution exposure history, and future work should
focus on using this comprehensive record to develop lifetime pollution exposure indices.
These data were partially utilized in this study to examine individuals that
changed their pollution exposures. It did not appear that expected lung function was
affected by changing pollution exposure, with the exception of NO
2
for MMEF, and NO
2

and PM
10
for FEV
1
. Since this effect was observed, it may explain the improvements for
those with high values of NO
2
(Table 7a-b). However, the effects seen for the other
pollutants do not appear to be an artifact of individuals in less polluted communities
moving to more polluted ones, and vice versa; these improvements may be attributable to
a delay in lung growth.
Other Caveats
This study may be limited by the sampling design and protocol. As previously
mentioned, only a fraction of CHS subjects were measured in the APEX study, and the
APEX study was essentially cross-sectional in design. Thus while there were 2,605
cross-sectional measurements for 18-year-olds, only 342 19-year olds and 157 20-year
olds were tested. The APEX participants were also measured on different spirometers,
with different calibrations from the initial studies. A regression approach was developed
within the CHS to adjust for these imbalances and predict corrected values of lung
function measurements, but error may have, nonetheless, been introduced into the
dataset.
50

Relation to Other Studies
Other studies have confirmed that pollution and other detrimental ambient
exposures during childhood can have long-lasting effects that reach into early adulthood.
For example, a longitudinal study showed that infants (2-3 months old) with low maximal
expiratory flows also had low values of FEV at age 22 [16]. In 1988 a random sample of
U.S. residents showed that the effects of smoking on lung growth were distinct and
irreversible with regard to FVC and FEV measurements [14].
It has been hypothesized that the reason for pollution-induced lung function
deficits is the retardation of small airway development in the lungs. While the effects of
many hazardous ambient exposures have clear effects on lung function (for example,
lung damage caused by smoking, or occupational dust exposure leading to reduced lung
capacity [18]), leading to actual damage in the lung airways, the complete pathway that
leads to pollution-induced lung function impairment remains unclear.
If lung function deficits are due to the retardation of airway growth in childhood,
then it is especially important to further study the nature of lung development. This
significance is enhanced by findings that lung function measurements in early adulthood
have been shown to be one of the strongest predictors of chronic pulmonary disease [19].
However, it has been shown that lung function growth rates are responsive to a change in
ambient pollution levels in children—and those that moved from areas of high pollution
to an area of low pollution were actually able to improve their rate of lung function
growth (based on FEV

measurements) [20]. This study also suggests that a change in air
quality has a similar impact on lung function trajectories in young adults. These findings
51

highlight the importance of continuing strategies to improve general air quality for the
health of both children and young adults.
52

Chapter 5: Conclusion
We studied the persistence of previously observed lung function effects
(associated with long-term exposure to ambient air pollution) into early adulthood among
a group of subjects who were initially studied annually from age 10 years until 18 years.
A prediction model was developed and proved to be useful in understanding observed
changes in lung function indices (MMEF and FEV
1
) through age 30 years. The
developed model drew upon the strength of long-term, longitudinal data collection in a
cohort of thousands of children. The spline model accounted for the non-linearity of lung
function trajectory changes in adolescence. By including several possible covariates in
the prediction model, and by stratifying by sex, we were able to reduce potential
confounding. The model also allowed for sex-specific measurements of FEV
1
and
MMEF after 18 years of age. Future studies could take into account other variables after
age 18, such as household exposures, pollution levels for movers, occupational
exposures, smoking, and other dynamic exposures. Additional follow-up in the future
might help to determine whether those with persistent deficits in lung function from
adolescence to adulthood experience clinically important outcomes as their lung function
declines with age.
This work provides evidence that overall lung function deficits accumulated in
childhood are not fully reversed in early adulthood. The data also suggest that changes in
air pollution exposure after age 18 years may affect early adulthood lung function. Based
on current data, it is not clear if effects on lung function growth continue to occur after
age 18, or if there is some degree of reversal of earlier deficits, if and as they occurred.
53

Future research and a more comprehensive set of post-childhood exposure and test data
are needed to clarify this key question for public health and regulatory policies.
54

References
[20] Avol, E.L., et al., Respiratory effects of relocating to areas of differing air
pollution levels. Am. J. Respir. Crit. Care Med., 2001. 164(11): p. 2067-2072.

[14] Dockery, D., et al., Effects of inhalable particles on respiratory health of children.
Am Rev Respir Dis, 1989. 139(3): p. 587-594.

[5] Friedman, G., A. Klatsky, and A. Siegelaub, Lung function and risk of myocardial
infarction and sudden cardiac death. N Engl J Med, 1976. 294(20): p. 1071-1075.

[10] Gauderman, W.J., et al., Association between air pollution and lung function
growth in Southern California children. American Journal of Respiratory and
Critical Care Medicine, 2000. 162(4): p. 1383-1390.

[11] Gauderman, W.J., et al., Association between air pollution and lung function
growth in Southern California children: results from a second cohort. American
Journal of Respiratory and Critical Care Medicine, 2002. 166(1): p. 76-84.

[12] Gauderman, W.J., et al., The effect of air pollution on lung development from 10 to
18 years of age. New England Journal of Medicine, 2004. p. 1057-1067.

[13] Gauderman, W.J., et al., Effect of exposure to traffic on lung development from 10
to 18 years of age: a cohort study. The Lancet, 2007. 369(9561): p. 571-577.

[18] Green, D.A., et al., Mineral dust exposure in young Indian adults: an effect on lung
growth? Occup Environ Med, 2008. 65(5): p. 306-310.

[8] Gilmour, M., et al., How exposure to environmental tobacco smoke, outdoor air
pollutants, and increased pollen burdens influences the incidence of asthma.
Environmental Health Perspectives, 2006. 114(4): p. 627.

[19] Hancox, R., et al., Systemic inflammation and lung function in young adults.
Thorax, 2007. 62(12): p. 1064.

[2] Hole, D.J., et al., Impaired lung function and mortality risk in men and women:
findings from the Renfrew and Paisley prospective population study. BMJ, 1996.
313(7059): p. 711-715.

[1] Islam, T., et al., Relationship between air pollution, lung function and asthma in
adolescents. Thorax, 2007. 62(11): p. 957-963.

[7] Jedrychowski, W., E. Flak, and E. Mroz, The adverse effect of low levels of ambient
air pollutants on lung function growth in preadolescent children. Environmental
Health Perspectives, 1999. 107(8): p. 669-674.17.
55

[17] Mauad, T., et al., Chronic exposure to ambient levels of urban particles afects
mouse lung development. Am. J. Respir. Crit. Care Med., 2008. 178(7): p. 721-728.

[9] Peters, J., et al., A study of twelve Southern California communities with differing
levels and types of air pollution. American Journal of Respiratory and Critical Care
Medicine, 1999. 159(3): p. 760-767.

[4] Schroeder, E.B., et al., Lung function and incident coronary heart disease: the
atherosclerosis risk in communities study. Am. J. Epidemiol., 2003. 158(12): p.
1171-1181.

[3] Schunemann, H.J., et al., Pulmonary function is a long-term predictor of mortality
in the general population. Chest, 2000. 118(3): p. 656-664.

[15] Siebelink, J., The relationship between respiratory illness in childhood and chronic
air-flow obstruction in adulthood. Am Rev Respir Dis, 1983. 128(5): p. 959.

[16] Stern, D., et al., Poor airway function in early infancy and lung function by age 22
years: a non-selective longitudinal cohort study. The Lancet, 2007. 370(9589): p.
758-764.

[6] Wang, X., et al., Pulmonary function between 6 and 18 years of age. Pediatric
Pulmonology, 1993. 15(2): p. 75-88.

Abstract (if available)

Abstract We performed a follow-up study to track lung function of participants in the Children’s Health Study (CHS), a multi-year longitudinal analysis of lung function development for 3,787 children in twelve Southern California communities. Subjects were followed past 18 years of age in order to examine the relationship between ambient pollution and two lung function indices: maximal midexpiratory flow (MMEF) and forced expiratory volume in 1 second (FEV1). There was a significant negative relationship between several pollutants (acid vapor, NO2, particles with diameter less than 2.5 microns, particles with diameter less than 10 microns, and elemental carbon) and lung function. This relationship grew stronger with each year increase in age, up to 18 years (P<0.05 for all pollutants). There was little evidence to suggest that this trend was reversed after 18 years of age. It also appeared that those who moved from areas with higher NO2 pollution to lower pollution were able to recover lung function for FEV1 (P=0.008) and MMEF (P=0.01). This effect was also seen with PM10 for FEV1 (P=0.03). These findings confirm results from previous CHS studies for subjects up to 18 years of age. They also provide new evidence that the deficits accumulated in childhood are not fully reversed into early adulthood, and that changes in air pollution exposure after age 18 may affect early adulthood lung function.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Spatial modeling of non-tailpipe emissions and its association with children's lung function

PDF

Disparities in exposure to traffic-related pollution sources by self-identified and ancestral Hispanic descent in participants of the USC Children’s Health Study

PDF

Long term effects of air pollution on lung function in the European community respiratory health survey

PDF

Association of traffic-related pollution and stress on childhood lung function

PDF

Airway inflammation and respiratory health in the Southern California children's health study

PDF

Genetic variation in inducible nitric oxide synthase promoter, residential traffic related air pollution and exhaled nitric oxide in children

PDF

Native American ancestry among Hispanic Whites is associated with higher risk of childhood obesity: a longitudinal analysis of Children’s Health Study data

PDF

Prenatal air pollution exposure, newborn DNA methylation, and childhood respiratory health

PDF

The impact of the environment on childhood allergic rhinitis: findings from the Children’s Health Study

PDF

Linking air pollution to integrative gene and metabolites networks in young adult with asthma

PDF

Ancestral/Ethnic variation in the epidemiology and genetic predisposition of early-onset hematologic cancers

PDF

A novel risk-based treatment strategy evaluated in pediatric head and neck non-rhabdomyosarcoma soft tissue sarcomas (NRSTS) patients: a survival analysis from the Children's Oncology Group study...

PDF

Chronic eye disease epidemiology in the multiethnic ophthalmology cohorts of California study

Asset Metadata

Creator Pickering, Trevor A. (author)

Core Title Persistence of pollution-induced lung function deficits in early adulthood: evidence from the Children's Health Study

School Keck School of Medicine

Degree Master of Science

Degree Program Biostatistics

Publication Date 08/03/2010

Defense Date 06/28/2010

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Adolescent,Children's Health Study,lung function,OAI-PMH Harvest,Pollution

Place Name California (states)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Gauderman, James W. (committee chair), Avol, Edward (committee member), Berhane, Kiros (committee member)

Creator Email tpickeri@usc.edu,trevor.pickering@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m3255

Unique identifier UC1103966

Identifier etd-Pickering-3925 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-366604 (legacy record id),usctheses-m3255 (legacy record id)

Legacy Identifier etd-Pickering-3925.pdf

Dmrecord 366604

Document Type Thesis

Rights Pickering, Trevor A.

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu