ML20106A357
| ML20106A357 | |
| Person / Time | |
|---|---|
| Site: | Harris  |
| Issue date: | 06/15/1984 |
| From: | Harrington J, Laird N, Speizer F, Spengler J, Wilson R HARVARD UNIV., CAMBRIDGE, MA |
| To: | |
| References | |
| OL, OL-S-3, S-3, NUDOCS 8408170192 | |
| Download: ML20106A357 (102) | |
Text
MMN MMM'EEM 4
.. ,.a c >.f, . j t*, g ,, 'r .f... ': . '
..-c.,, ,..s
.. 3 ., ,, -
- g. , . '
g...
..g.- ' ..; . r . .' e e s
g .., -- y .. . ., .
- =. I,$Y. h. ,' ..... ' . ,.,*>4 .
e N.. . :-. . *
- l- - I.
s c a..
m , * >rYdI_e ah !. *
-.b' ; , e,
- . ._ '%
- ,,,p.[.j.sg.n,. ,,J.,Q(: ,' b.' . ;.; j) ._ -.; .c,.,......p .. g . q.y., [ .j.. _ . y. .. -.
.. .Qi: . t , .. _
h .$s. .. b%.&+ k., "."%s , ,,s.
s'.'F + f. M ,: -,. ..}-%.,..C
-1( . .-
1 4.' ,,,
e .
. n#l Q:... : i j '. i.e * .,. '. ,4' . ~. M., 1g . * ..,.G;e' =-
. / g* . : , , . . , .
v.; mar . -
c - dW.
- *l"**)'.F.. - -
. .. .; . . y ' .- . . ', ... : .~. . . .
.. >=..~s..
t M
A .,, .a'.er.4 4, g. -
$PT.* ~
sa. a. .v. .. ,- - . ...,,y,,.
.
- q4g w - e -s .,.
f..')7.,
g sp, %
V; ; e' . . -) [. .
c.~ .j;J,y r U... - '
.v K , . .: _
4' ,. k .. 4 sp.c ,,,7..y*c .; I'.m. N
<N f.<ff
.vir,i.,.h~
m....'.W. 4 ,11.IM T'$. g-@p., sr .:-J.9, -.: .m$
\
n,.,n. ; ;;.3./
' %.y,' ;a. s .,. . , .
W 9. %,Q.q. W. --e . w ~. y.a.-
Sq . ' .
.... . .= . ....,t. . - + .
4 .
W &, : 1, , ' :
..A., gw.-g/,;c ~ , .. ..~.,;
/ ow '. ' .y*f .* f. , -. Z, . . ,.y,......;....... :< v;-n. yp a ,
- s..;' V.* ..?
- ,M.g f- .'. ..e , . h. . . .. . ,
.,i. 3 . , ..* . q .. , A. . T . ..,
4: .
2 .. .
kf. , .
M- .
k f . .
- W '
. g v. - . : v,- .s . ;
~
- s. s ,p ..
_ v.,
.,;.. .e-
.x. . . . . :.. ,.:..'., %.,:
- . , s_. , . . _
Y, hgm
~. ,. ,. ;- , - , , g%
. . . . . , . - " ._ ., ;. . y. . e. . , . . 3, %e. .
. . .* . . *3 ; J.- .
.. s - '9, n, . ; . . . . . .
4, yr ..
s .,
. . . . v,r .,s g, ; .. :
..r ..c.- .
.s- -,.,.
. :. .g.". 1. r ..
.. ... - ..o.,. /. .
~*% } ./ * , ,,
4
- p". .
4 }.
. ,? .' J. ,~ . ] ,, h .g . . . ,
., , ] ,. **
. , . ., y ,. , *** d t' - * * ' * * * * .; ,re r e a : e - . " , . - .. **e h k8t.,. f. .h*-h- ,,4,e -
-e - - . .f # - 1
- ,. A .
' 94 ,'. 4* .
.. E Y. $ .j ' ^ * .,
i I T , b.ISTYw '., 6.. jt., .
. (( ^- ' rs -
-*l- 4 ,
L * . p, p.. }.f* , f '. f,..
T
..3
-- e
.., .c , . ' ,
t, ,,.. .. .
,p .
g g x+ ~' [ ..g.-f $.
- .*s.= .....(.[8
,eg [ ' k,. .'{ f .. , ,. . . ~ .f,.*/ ','.'. '
' g .... . " , . ., x, . .#-f- .j ,,
( .f' , . -l.m 'a. ..
~ g1. ;% . . ..1... . . , : . . , . ,s . .a y . . .r . . -
,.7.-
v-
-- ,ar .r.
t M . ; ,* . .
v s.. .. .
- -2 .. , . . ;
c_s. j
/ .;, . ,. . , .
m t
}f%
j g, - ; y, ,,-) .14 .*.s..y.j.J.? ' l f:." ?; 7 .i 1.l;. O f . '.. %,. ,.. :
\ .- : :s ' . , ~ 4& z ~ - : :,-
.L
. :, 7*
.& . >.4 K,... . . i;V m i ? .. ' . *; .:*". . .,... - .
- . ; ; ,' . .,,,** ... . /> ..<
>.~ - .!.
5~ s . - . . ..1 +
4 . . y y sg .n. y.,,. . . . . , - " . - . .. ... c ..; .. a .y ~ ,
. : ., ! ......s.
A. n
- . 'r p . s .} ..w ... . . . ,. . g m . -
~ , . . 1 T .', , \':.} Yl *
- i l :
'. ~ / ,D , - jli- ' ~ - ' \. ',,L. .f'.f, ?: ;; ) r.... '., ; ;l" ' .b ' . .? . ,'l .'.
5l?lf; ,;, u*n.c / 'l.l' n.\ g .
_ y ; _ ;, . ,- l :
-~'j., _ .l' +
h Q WQ~ %r ;,e, . ',..as (.s . .:,
,a u4 .- . .
J,4.. &- -(a ,
e'
..,.y -
,. . . . (
..n . %
g
. ..u :. ; yx . .. .. : .... ; .y; '. . , . ;}.b. .; . ,;;_ a; *. .v :,, ; - .e ,m . . , . __
.an w Q
.?
s, av y c .- o,. m - -
,4 .f .~( '.,- -
.. .;,,..j,, j~
C. .. ,. . . . ..- , ,,.. 4 4- g. _ j ,y. gig y, .. % *ns'tf.4 % e'. f..2..g.;&
.4 7,
-[;[..f:s_:-l..}g..Q,;*.;.. . .
-2* g -
.,n'. ,. .; .' ; _ ;. _.
. 3
- y. _ ....;
m.
a . ; . ..
0,s,a%g y W p ..N .....
........r.
. vt,. .z.
. . , , ... . . . .. - . - ,. .; . , . .. ..,,....w,a,...,,. .
,h?$ & : Y: y~'.l_ .~ 'f l ~fffflI i
_. fAV2 1 .
y : ., , Y';E.f 5:f,&;$fh?Yb'&'; ;.;I
- hef -. ;} .Q ' [Q _ _lw,t.d'?j , s ': : f }g . A, i. J
[,s.%. ;
~
-Q g .... .... bf.t. y. - :,
- ll. . [. ,l ~ ,
' s / . .. ; ... -u ( ?.9 [-l j
,, ,, . 1 4' ..
.$9., % :e.9 . J. ht. n. .~n..., . .-
- y, .
Q !'& s ' *
. ~ ,. : .: .;.c. ;. -
.g:.. :
I.,~ .',y ..[ ..* * ', h~ ..y .%V ..h. . : R. W.y. l,.l,Tb: Q f.Q*.?
- ' '~
. .ll
~Ye* ' ,y -n.L';~ .. . yr L .-. . 4.
s ,.
~.
.. . . .p . -
t . . .
.s - ..-- . . .
e... .
., c
.s: .
3; y . .. ,.
f M ,
, . , ~ . . - : ., . ,
4, p,.f4 4,,}h! .
kf t Q
hh .' .*N kh *
- y, y,4.; . . n. . ,. ,,,p
- y gsl W.
2,dm . ;3,9y: 4
.t .
1 n,s ..; ..
M b .. .m.Nf8NNbhN '
A y y;,. m p g w ,a g.; y g. O m'
W"~ ~ ' .
.., y.
' ^
__3 _
f,gy, 3 e > - -e* -s .
?~* ,
~ ., 4 1 4 uw -
u ; 0
- p. .
.4 iv ,
I ush j,
4 ^ , 'f-
- 1 -
_'4.
- t h
s
- e. .
, \\\
\ .\
\
\\\\ys 7=
4 4 (1 3 P. .I4
& r! $
\>'g\\\\\
8"
?. b }r: %'6 i Et s
1
\;\ \ \
\'\
Q '
\\ '\ \ \ \
udDD O
-'n - - - . '
~~
f x
- c
- :
m c\
}[< ;., J ' u' -
j i
, \ M, pOC;GlI"MG & j siay;CLLRAPCU
'. \
(, s EECY M C /
/
.-j- '
ic "
ANALYSIS OF HEALTH EFFECTS RESULTING FROM -
POPULATION EXPOSURES TO AMBIENT PARTICULATE MATTER
(-
, HEALTH AND ENVIRONMENTAL EFFECTS DOCUMENT
.2 e 1983 Prepared by:
Harvard University Energy and Environmental Policy Center 140 Mt. Auburn St., Cambridge, MA 02138 1
Assessments performed under the direction of:
J.D. Spengler, Ph.D. - P.I.
Harvard School of Public Health J.J. Harrington, Ph.D. - Co-P.I.
Division of Applied Sciences, Harvard University and Harvard School of Public Health N.M. Laird, Ph.D. - Co-P.I.
, Harvard School of Public Health F. Speizer, M.D. - Co-P.I.
Harvard Medical School o@ R. Wilson, Ph.D. - P.I.
Physics Department, Harvard Unive.*sity o -Principal Authors of the reports i, H. Oskaynak, Ph.D. (Project Manager) G.D. Thurston, Sc.D.;
T.D. Tosteson, M.A.; C.M. Smith, M.S.; P.L. Kinney, M.S.: B. Beck, Ph.D.:
W. Skornik, M. A. ; S.D. Colome , Sc.D.s A. Schatz, M.S.
October 1983 Health and Environmental Risk Analysis Program U.S. Department of Energy '
Agreement No. De-AC02-81EV10731 o
CL .. n g . +.. . - . ji
_ ~- ~ .- . - - . . . - -_
+
.Q. .
i 7 ;..
k
, ACENONLEDGEMENTE
.This NEED, " Analysis of Nealth Effects Resulting from Population 3 Exposures to Ambient Particulate Matter",is based upon the contributions
.o' of the staff, consultants and co-principal investigators of the study-on Nealth Effects of Esposures to Airborne Particles which is administered by.the Energy and Environmental Policy Center at Narvard University,
'5- John F. Eennedy School of Government. The results of our study are also contained both in the annual progress reports previously submitted
- g to NERAF of DOE and in the three Appendices submitted along with this
- HIED . -
We appreciate and acknowledge the technical contributions of the 4
following advisors or consultants to the. project.
J.D. Brain, Sc.D. - Marvard School of Public Health R.B. Husar, Ph.D. - Washington University D.T. Nage, Ph.D. - U.S. Environmental Protection Agency-P. Matthews, M.S. - Stanford University
- P.J. Lloy, Ph.D. - Institute of Environmental Medicine, NYU M. Lippmann, Ph.D. - Institute of Enviromental Medicine, NYU M.B. Schenker, M.D. - University of California, Davis .
Y.S. Kim, Ph.D. - University of Texas J.H. Ware, Ph.D. - Harvard University
. g' We also appreciate the support and guidance provided to us by ,
Dr. Nathaniel F.-Barr and Dr. Paul Cho of DOE's Health and Environmental Risk Analysis Program. Technical discussions with several members of p government and private research labs (in particular: EPA, BNL, ORNL, ITRI, LIJfL, LEMR and IWG Corporation) are also gratefully acknowledged.
T Finally, we acknowledge the outstanding efforts of Ms. Cecilia J. Nicks -
g in the preparation of this document.
4 e
g -
9-
.e8 B
D 11
~
a
-C
- TABLE OF C0NTENTS -
i
'l
+ ,
,, Executive Summary . . . . . . . . . . . . . . . . . . . 1
. Introduction. . . . . . . . . . . . . . . . . . . 1 Epidemiologic Assessment. . . . . . . . . . . . . 3 Toxicologic Analysis. . . . . . . . . . . . . . . 9 i,
Conclusions and Research Recommendations. . . . . 12 h
- I. Introduction. . . . . . . . . . . . . . . . . . . 16 o II. Epidemiologic Assessments of Health Effects of Exposures to Ambient Particulate Matter. . . . 13
. II.1 Review of Evidence from observational Epidemiology: Morbidity Effects of Particle Pollution . . . . . . . . . . . . . 18 II.2 Fine Particle Pollution and
,,. Human Morbidity. . . . . . . . . . . . . . . 23 i ',
II.3 Investigations of the Mortality Effects of Air Pollution Using Time-Series Studies . 35 11.4 An Analysis of Cross-Sectional Mortality Data Incorporating Fine and Inhalable Particle Mass Indices. . . . . . . . . . . . 45 III. Toxic Effects of Airborne Particles . . . . . . . 51 III.1 Carcinogenic Effects of Particulate Matter: Review of Evidence from Bioassay Experiments. . . . . . . . . . . . 51 10 111.2.1 Non-Neoplastic Toxicity of -
, Particles and Their Significance to Human Health: Metals. . . . . . . . . 69 111.2.2 Non-Neoplastic Toxicity of .
~
Particles and Their Significance to Human Health: Acid Sulfates . . . . . 75 III.2.3 Non-Neoplastic Toxicity of Particles and Their Significance l e to Human Health Nitrate Aerosols. . . . 78
,q. IV. Princ!. pal Conclusions and Future Research Needs . 81 Principal Conclusions . . . . . . . . . . . . . 81 Future Research Needs . . . . . . . . . . . . . 85 Glossary of Commonly Used Abbreviations . . . . . . . . 88 1
References. . . . . . . . . . . . . . . . . . . . . . . 91 6
111
i lC 1 EXECUTIVE SUMMART
- Combustion-related energy systems represent a major source
- g* of airborne particles. An assessment of the current and future health impacts of energy systems requires an understanding of the nature, magnitude and uncertainty of potential health ef f ects from these particles. This generic Realth and Environmental Ef f ects Document (REED) on airborne particles presents results f rom the second year of an ongoing s tudy of human exposure and _
. c*, response to ambient particulate matter. As such, it also draws
, from information included in our 1982 REED. The 1983 REED emphasises our efforts over the past year and continues to develop estimatas of'the magnitude and range of potential health 3
effects from exposure to particles.
' Using published findings and re-analyses of data available from epidemiologic, toxicologic and aerometric s tudies, we have, in this report, attempted to evatuste effects of particles.
Where the available data permit, we have related health effect outcomes (which include morbidity and nortality) to concentra-tions of particulate matter according to sise, composition and C' source of aerosol.
Special attention is given to the uncertainty associated with estimates of ef f ec ts. Deficiencies are identified in the data bases required to make assessments, and recommendations are given for future studies and analytical ef forts which would serve L to better identify and quantify impacts.
INTRODUCTION A wide variety of particles exist in the ambient environ-ment. These particles have been f ound to vary in size, shape, O concentration and composition, both spatially and temporally. -
Through condensation, evaporation and chemical transformations gases may become particles and vice versa. Moreover, dif f erent
' forms and types of particles are known to have varying biological activity, so that a characterisation of airborne particles f or
, health studies is complicated (Section 111.1). In this work, we
,- have recognised these differences (see Appendices I and II), and have expended considerable ef f ort to characterizo aerosols in order to examine the influences of particle size and composition
, on the estimation of health effects.
'i
. There are many complications involved in applying e p id e m i-ologic or laboratory analyses of particle pollution to assess-ments of human health risk. Observational epidemiologic studies of community air pollution exposures have not generally been designed to provide dose-response relationships suitable for risk assessment (Section 11.1 and Appendix II). Therefore, the exist-
, ing data base is lim it ed. Most esposure asasurements for these studies were f or the total mass of suspended particles, or for the reflectance of light off bulk particle samples such as British Smoke (38). These qualities are incomplete characteri-astions of an aerosol and, more Laportantly, are probably not '
1
_. ..- ... . . - ~ . _ _ _ . - _ .
3l parameters of greatest interest for an understanding of health effects. Furthermore, the f requency and duration of pollutant measurements employed were usually too limited la scope to allow a separation of the chronic versus acute ef f ects of particles. F Additionally, the observational epidemiologic literature repre-seats exposure levels-that are generally above current and '
projected concentrations in the United States. As a !
result assessments of risk have relied heavily on the limited ~
data bases that are available for lower concentrations and/or entrapolated findings from the studies of , higher concentrations. 'l Thus, currently available literature on health effects from com- . !
avaity air pollution cannot alone be expected to fully disen-
.; tangle the separate actions of individual pollutant sources or i aerosol components f rom one another. In contrast, laboratory ,
j studies of humans, asiaals or cellular systems provide control of factors not possible in observational epidealology. Insights may be provided into mechanisms of action and relative potency f or '! ]
different particle sources and components. However, the control i of emperimental parameters briass with it new dif ficulties in projecting back to the more complicated free-living human envi-rossent. Thus, while historical epidemiologic and laboratory ,~
studies provide valuable sources of inf ormation on the health '>
effects of particle air pollution, each methodology also has )
inherent limitations. l Recossising the above-noted data analysis limitations, we ;
began to combine the evidence and understanding from individual ,
toxicologic, epidemiologic and exposure assessment data sets into H a staste coherent and consistent saatysis. In most cases, it was necessary to take aerometric data sets developed for other pur- ?
poses ( e.g . , the EPA's flae particle network data) and apply them [
to compatible epidemiologic and toxicologic data. Major efforts .)
undertaken this year to derive health potency f actors for air- .,
borne particulate matter included: -
(1) Improvement of particle esposure estimates. This was accomplished using recent fine particle data sets to develop general relationships with atmospheric visual range records, as well as with historic records of other particle metrics.
(2) Application of improved estimates of particle esposures to esisting cross-sectional mortality and sorbidity
- studies. The data examined included the 1960 and 1979 total nortall'ty data f o r U.S. Standard Metropolitas "
Statistical Areas (SMSA) and the 1979 National Realth Interview Survey (IIS) norbidity data for 12 SMSAs.
(3) Evaluation of carcinogenic and non-carcinogenic potential of airborne particles using in vitro and in zigg bioassay data. Available data were restructured "
and a comparative relative potency model applied to Ames bioassay data for a variety of particle types.
s.
'O A major objective of this REED has been to addre s s unc er-tainties and gaps in understanding when reporting epidemiologic, l toxicologic and exposure results, thereby providing an integrated Q f ramework usef ul f or both risk assessment and policy making.
Where po s s ible, quantitative risk estimates are derived. The source of these estimates and their associated uncertainties are described in the f o llo w ing sections. Throughout this effort we have also identified remaining gaps in knowledge and data base -
limitations. This has led to recommendations for future studies O and analyses that will serve to further define the nature, magni-tude and uncertainties of particle pollution health risks.
EPIDEMIOLOGIC ASSESSMENT o
In this REED, we hav e continued our analyses of the health effects of particulate matter in the general population. This work (presented in Section II) includes the review and re-analysis of previously reported data bases, a synthesis of pub-lished findings, and an original analysis of health data using new models or recent data. The objectives of this work are to:
(1) identify health outcomes associated with different levcis of k~
pollutants; and, (2) characterize the uncertainties of findings and the sensitivities of results to choice of data base and analysis methods.
Observational Enidemiolorvt Review RI. Morbidity Literature Existing epidemiologic literature on the morbidity effects of human exposure to particulate matter has been summarized herein. The outcomes considered are various measures of respira-tory health, including symptoms and pulmonary function. Uncer-tainty is introduced in the relationships between the measured outcomes and particulate pollution because of uncertain exposure estimates, confounding variables, and the high pollutant concen-trations typical of this literature. Upper and lower respiratory symptoms and reduced pulmonary function has been found in asso-ciaton with ' long-ter$ exposure to TSP-equivalent concentrations in excess of 180 ug/m (Ware et a l. , 1981). Short-term (24-hour) concentrations of particles have been associated with hospital l ad m is s ion s. A recent study indicates a relationship between
! total and respiratory emergency room admissions with particulate matter over a wide range of particle concentrations (Samet et
- al., 1981). The role of particles in that study was statis-tica11y significant, but it explained only a small proportion of Q the variance in emergency room visits. Particulate matter is not expected to have a greater role in emergency room admissions, but the size of its effect leaves open the possibility that other factors (not accounted for in the analysis) could be responsible for the observed relationship. Very few studies have addressed acut h 4
ug/ m$ measured (ealth ef fin ec24-hour ts of particle averagepollution betov levels TSP equivalent of 1000 concentra-tions). 'Those which have studied effects at lower concentrations (Martin, 1964 1 and Lawther et al., 1958 and 1970) show an increase in hospital admissions for cardiac and respiratory ,
111 ness as well as for bronchitis symptoms. Even in these 1
1 j 6 3
L --
n_i ._ __ _ m _.,..._-._.._.]
.i
.?.: i j studies, the average ambient concentrations'of TSP were high i
,'{ relative to current levels in the U.S. However, the data for
.; short- and long-tera exposures to particulate matter do not sus-gest existence of effect thresholds. For this reason, simple 7 linear coefficients have been derived relating particle concen-trations to respiratory inf ections, ho spitalisation, and symptoms of chronic bronchitis (see Table 1).
- While the above-noted extrapolation problem is a major con-l tribution to' uncertainty of morbidity risk estimates, other .
6 sources of uncertainty are also important. These can be cate- -
'l gorized into sampling and non-sampling errors. Sampling errors 4 refer to the lack of precision or the representativeness of a d sample result. This error occurs when a particular sample .
analysed happens, by chance, to be dissimilar f rom the general
, population in some way. Non-sampling errors, on the other hand, J includes biases in determining personal exposures from central site monitoring data; a lack of his torical chemical- or sise-selective measurements of particles; confounding factors reisted to personal habits such as active and passive smoking, alcohol
. consumption, etc.; and, cross-community dif f erence in particle source composition, fuel mix, type of housing, sad population 4y.
migration, among others. Such sampling and non-sampling errors are-inherent, not only to the morbidity effects discussed above, ;
but also to all of the other mo rb id it y and mortality investiga- ;
tions which follow, e
Morbidity An a l y s e s 1E.21 1]Lg, N a t io n a l H e a l t h I n t e rv i e w s u rv e y The 197 9 National Realth In t erv ie w Survey of the Na-tional Center for Realth Statistics provides a data base on
_ individual level health status for approximately 100,000 people.
The 1979 survey used a two-week reference period for the health
- questions and collected a variety of economic and demographic )l information. On one-third of the sample, the survey also !
included a detailed questionnaire on smoking habits for adults.
In this phase of our research, we used the RIS outcomes of total !
Restricted-Activity Days (RAD) and Work-Loss Days (WLD), relating them to particle exposures in dif f erent SMSAs. Unfortunately, y particle mass seasurements in the U.S. have typically been made e every sixth day in most communities. Since the RIS data base specifies the health ef f ects outcomes in terms of incidences over ,
a two-week period, it became essential to improve upon the . l available esposure metrics.
Using fine particle and visibility data collected in an initial set of 12 SMSAs, we were able to develop site-specific daily estimates of fine particle exposures, which could then be averaged over the two-week recall periods employed by the survey. -
'In Appendix 1 of our 1982 NEED, we developed the theoretical basis of this approach f or estimating fine particle exposures y using a erosol extinction coef ficient (from airport visibility i data). As indicated by numerous researchers (see, f or example, t Trijonis and Yuan (1978), U.S. epa (1979a) and Section 11.2 of ,
this EEED] under most circumstances, this method yields reliable 4
]
- 3. + . %1 -asai s k
TO Table 1 .
~q Sm ,o.,f Principal Results from the Assessment i
of Population Health Effects Resulting from Exposures to Airborne Particles * -
i .3
,V I
, masa er ,
Type et ' Type of sma&e Signeus assaammed pytasapa& asumuma
., , w - er one.e Wernehle augemuse./Rempamme
_a_- hm iata han emmtes. -.f an.al e cent ** = < = at r - '
muutedter assess ed e&ammed Ismer M es t (**) Comme / year per aementeseLasttatama (6asspoupremes seedy
- sempiamessy a designe lastead of hea&th and
! I % &adestiamo pm&&matan instahlmes me&ouseme of egmeuse I ** I*""" e mmede amammbstas N' " "8" a g SB 3 3 e to 113 sp(eeH "" "al .p"uuhaam ed Ise dame emessen&a=
sendta i Asume sempiamener W 1e3 3 e ne ges a g team acenduettag ens to " ^^.
, d&amano dampagnes and teampasse& saammse eatif&=
hmmes SW 8 i I. - g 3 g to y 4 8 mA*F I"m8==O'"AA"ssa"ge"'"
e I9" "IA "8 9 seg6ameusy 46amamos ammmme teena emme- Tap 3e a e i gamer some v6 site Andref a et amme m etey 79 0
- g.e1 ans y .as estimmes reemme them a dasset amasure gg9, gg Boye (msL (for ee/u rp of rp mese te eens .cmafenendaag offesas et pesema es e eksmate per 3-emah seter pm&&ueesee and passase -ma=g, meg faamew6e laattee&amel asemase
^
.aa-aaaad =' ames&a used &a the i survey and umst-4ame Beye (Idal FP G
- 8.004 15 3 ama&ye&e my me he hwa austed na amanyma (8er pasemme wie Sag as EIS daea *cedmetag utahame 4 -
- "*** sere F, assumas umsessammi per w ==== 'a= **= 'adaa=* 'a 'a= ==** =r -
i gy ,,g ,,g y he agmeenst *amamong daea guestamm &a est l edeseemed e&ffaanley la psamaaeny aneshang j does the 3-emah rema&& pertad use re empamuses
- mareauer taas emesse an&&y unreauer oss e.st a e.ept' assaan/my .aume ese muse &y meamend to eaammond eme.
l
'#8"" ' " to e.es a s. ass. per 14, pues tem
,,g, rammer g g,,,,,g than parese,,as
,g,,, gg , na,,me g,,g ,.9,museass
,.t com e. semey ameta&sey . eses. en ans puseas.
j eersese la eseamme&ag empemusam der the 1 pape& sedan lastag na uut by noaag anagne.
i sessesm man &eartag done
- amen &.emand us.
l ** 0b**
l Caume* Ammua& nerea&&ey PP 1.3 3 4.4 taaeha/ yens .amm&yese empneyed FF mean egasuso eme&*
amans haamd upon emmerd site essa opuse t
l 8"""E 8 IO e '""' antas amado n daag vastantaa mee _
i Amm&yede see se4' FF (e.g.. gammmes pe&&enament = app 1&amh&e to j typeant amn6ams amuses &
88n, M and l 3.4 8 1.4 asy mee appay to emesume assammaa *emumm&- t toy quase&ame hommme ambtems lams & ege*
eures and meta &&er
. a6aammer W ' -' - Ikasemme la t.&e e emmyd6an d S 8 14*'
- 8.4 " Issum &a en *6&attatamme ed human hea&eb duas and Demandese et annantem amtmaan asan (mp
, ,gg.e ,g,3
.__ - - ettem agemmes*Ade; f and See Commer g,,,g,,, ,,,,,
,, g ,,, ,,,,,,
gunny et tas se&anase passesy amen & and
( esamme Suommsy mmesa per eg4' the L&amarley ammusteam some ta se&an&ag i ,, ,,,, amm essa &yst 3 3 14*'
- 4.3 ,,,,,,,,g, esposure to sempanse *emmaus6am of Ameer =
l mynsk omgame esaten effesas evestatesme &a poesmey luy i (1* 98. 89*) -
# #""' ' " "'' '8"##8*
S E te*'
- 8.43 -
esamt i -' ed paratele amuseo g,, _
Ammedema&a& temene 4 3 14*'
- 4.03 eyyme la ensus of ease -
esgne=
PRC 4 3 18**
- 9.3
.? ! -i- i i.e*.1 r 011 1 3 10**
- g.! l 1
, Im m i , *The messenteesme surgemetag all ei the rich samfflesmes shuun &a thee tam &e ese em lasgo that the pues&h&14ef ed an ed9 mets (&.e., mese i riaal sammad alempe be : La emme&esag er es&ay them. L .. the appa sa=mi- et thmee emoffte&amee emummes amasumo ammeae and apost#4essten ed -- E of gus&mmeeses and ^^. The semene te aged to summe&t the appenprisee sese4em of tade W teostamme 13.1. 23.3. !!.3. II.4 amt 133.1) for foemer sha- and guademme. ,
- ttese ebene ese amm " and de amt iso & ado emup&ame ad 7 'at 8essese, se seppast thee, et the psumme taae, the sammensd sever (el el the estimens he emusedesed ha&f as lasgo as the seedmese tenumm&ous.
l ***9h8 'Tyne of empenseems ese ase ammd&demme teemrumas, g>.e seeker tan range of pmanahte posee estaanese of empumwe/rempamme omedfieassee.
l y 'emme estammes et emedfissang tance, numreur, altasumates ament fesumaneesmo have toen shamn to gave empsowe/vempamme emetttetees e to ? ,
samme haamme tamm ason.
I "LaPeased saad&demme teeerve&e. I I
- m. >
5 i'
- u, _--~~_ =
_z_w: ::
+ . . - ~ . . . . . . - - -
- 1
-t ,
3 1.
-. i . .
estimates - of fine particle . concentrations. Moreover, because it
- ~ ?
provides area-wide as opposed to local exposure .information, this technique is pref erred over estimating fine particle mass: .
i from hi-vol measurements of TSF sass (cf. Appendix I.2). 9 o
In our preliminary' analysis, two ' multivariate models (ordi-nary least squares and logistic regression) have been used to relate the fine particle exposure es timates to the health .out- -
- . - comes, while accounting for weather-related, demographic and available behavioral health variables. The continuing analysis 36
~
of these data is reported'in section II.2 of the text, and a -
4 summary o'f the preliminary results is provided in Table 1.
l According to these estimates, persons with chronic lim it a t io n s -
are expected to experience in a eve f eek 0.01 RADS and 0 to 0.004 WLDs per. ug/m exposure period 0 to of average fine particle
,,3
- mass concentration. Iowever, since further model refinements are in progress, these numbers must still be considered to be tentative.
Uncertainties are introduced in this morbidity analysis through the use of a particle pollution index and through errors g~ >
i in measurement of the health outcomes. To this date, we have
- considered a single measure of pollution based upon a derived relationship between particle mass concentrations and visibility.
,' ' The physical relationship between inverse of visual range and i Fine Particle (FF) mass measurements is consistently linear (see i Section II.2), but other pollutants have not been co'asidered that s might be related to fine particle and visibility levels. For instance, nitrogen dioxide is known to contribute to visibility l reduction, while osone coveries with sulf ates (a major FF compo-nent) at some locations. Also, since we do no know the exact day of interview, the pollution measures are not perfectly F matched with the periods considered by the retrospective health ~ ); '
laterviews. However, unlike past efforts to analyse such data,
- we have been able to obtain estimates of daily fine mass concen-l trations for the entire period, rather than occasional sampling j data (e.gi, every sixth day). As is always the case, confounding
! variables might include the effects of passive smoking or other unidentified factors related to persons 1 exposures. With these a
g;
{ caveats in mind, the major conclusion is that the models employed ;
l indicate an acute air pollution health effect, even when control-
[ ling f or all available f actors thought to relate to health out-comes. The outcome asasure most sensitive to the effects of air .
)! ~
pollution is restricted-activity days among people repo-ting chronic conditions including asthma, emphysema, chronic bron- ;
i chitis, heart and mental conditions, among others.
! Time-Series Mortality j In the absence of long-tera studies designed specifically to
' detect'the nortality effects of air pollution exposures, attempts ;
have been made to utilise available nortality and pollutant index data to search for a po s s ib le cause-effect relationship. Time- .
series analysis provides one means by which to test for such a ,
relationship.
6 se em a. ap
-wwww 4 w-me'e we ww-,-c'
g gw e diit % . me., m=we.*- ma a- =m- v' .--n =
q 10 4
i In Sec tion ' II.3, we have re-analyzed 14 years (1963 - 1976)
, - of data on mortality and air pollution f or New York City. Uti-
't .11 sing-a range of models designed to correct for assumed con-
- O f ounding due to low-f requency (seasonal) cycles and the ef f ects of temperature on mortality, we have obtained a range of coeffi-cients (see Table 1. and also Table II.7, in Section II .3 ) esti-mating the association between daily mortality and Coef ficient of Haze (COH), a crude surrogate of exposures to ambient particles. -
l The range of results provides a measure of the magnitude of l
uncertainties in risk coefficients due to uncertainties in l
?' modeling assumptions. The risk coefficient reported by Schimmel 1
(1978) is near to the lower end of our coefficient range. Super-imposed on our range of risk coefficients are the statistical e uncertainties in each es timate, as - quantified by its standard error. These range f rom 30 to 60 percent of the mortality risk
. estimates.
. Combining the modeling and statistical uncertainties yieldg a range of daily risk that extends f rom 0 to 0.03 deaths per 10 persons per day per unit COH. However, as we discuss in Saction II.3, this rang e is probably s till too narrow because: (1) pop-U ulation exposure misclassification occurs in utilizing pollution
- data from one fixed ambient monitoring site; (2) the exposure metric, COE, is imperfectly related to respirable particle mass concentration; and , (3) the range of models we fit may not have been diverse enough. For example, independent weather influences on mortality may-not have been fully considered. The first two issues tend to bias the estimated risk coefficients toward zero, while the third issue might result in positive bias.
Finally, the results quoted here (and summarized in Table 1) j_ apply directly only for the mix of sources and time pattern of L
- concentrations observed in' New York City between-1963 and 1976. ~
The general applicability of these results will not be certain until validation studies in several other cities with dif f erent pollution characteristics and weather patterns are undertaken.-
Cross-Sectional Mortality Cross-sectional analyses-of mortality during one time in t erva l (e.g., during a given year) provide another means by which to search for. a particle pollution mortality relationship.
. - Using recent (1979-1981) data on inhalable and fine particle mass, we have. conducted preliminary cross-sectional regressions C of total mortality on both socio-economic and particle pollutant variables. This analysis was performed to investigate the influence on cross-sectional analyses of particle measures
., such as IP and FP rather than TSP. Tab le II.11 in Sec tion II.4 compares regression results for the various measures of particle pollution employed in the analysis. The regressions repeat prior O analyses -of 1960 SMSA data ( see 1982 HEED,Section IV) incor-porating recent inf ormation on particle size relationships and 1979-1980 census, mortality, and fine particle pollution data. -Important socio-economic variables ( e.g. , age, percentage ,
l
~
~
7
=- - - -- - -
! _ _ _ _ _ _ . , l_ _.. . . - . . - - - - "
? .:. . .. -~.--. ..-. -:..,.--- .. . - -x - - --
1 I R l
I
- of non-white population, population density, etc.) are controlled l f or in these analyses, with the various particle size mass con-I centrations being sequentially replaced by one another as the particle pollution index'in the regressicas. These regressions indicate that, .as coarse particles are eliminated from the mes- O sure of exposure (in going from TSP to IP to FP), the general l tendency observed is a. reduction in standard errors, an increase in statistical significance and a stronger relationship between particle air. pollution and mortality. This is consis tent with _
i the hypothesis that the IP fraction (especially FP) includes the j component of . particulate matter responsible for health effects. D l If the coarse f raction acts as a biologically non-significant. -
random component on top of the IP. f raction, the coef ficient f or TSP is expected to. be biased downward (i.e., too low) in a manner l
consistent with the results of this work. .
t l If one wishes to apply a coefficient from the available 9 l cross-sectional mortality analyses to predict the health effects I
of particle pollution, then we would recommend particle risk c effi ie t (e.g., 1.3 the use of g per-0.6 dea ths / year /10 fine l sons per ug/m 3 FP, as shown in Table 1), rather than a TSP or sulfate coefficient. The problem as sociated with using a TSP s coefficient is that a major portion of that mass is not inhalable ~
by~ humans, while sulfates represent only a portion of a total mix of ambient fine particles. In the absence of FP data, the coef-ficient noted in Table 1 f or sulf ates may be applied with cau-tio=. However, if a sulf ate coef ficient were to be applied in cases in which sulfates are present in substantially less or .g greater than usual proportions ( e.g . , relative to organic par- '
ticles or trace metals), then the results would b e mis leading..
It can be expected that the use of the entire fine particle mass should be less sensitive to errors introduced by compositional i variation from case to case. Also, it should be kept in mind that the FP coef ficient is mos t representative f or an " average" urban aerosol composition and will, to some extent, be subject - '
to the biases noted for sulfates when applied to aerosols having a askeup very dif f erent from the mean compo sition. It may b e .
that this problem can be addressed through the development of coef ficients f or each of the numerous aerosol components (e.g.,
auto particles, soil particles, oil combustion particles, e tc.). y However, until aerosol component-specific coefficients are =
developed, the use of fine particle coef ficient (rather than a TSP or sulfate co e f f ic ien t) appears to provide the more accept-able alternative for risk. analysis at this time. .
i Although the use of a fine particle mortality coef ficient ~
should provide an improvement over previously used cross-sectional indices of particle air pollution, we mus t emphasize the large uncertainties surrounding any such damage coefficient.
Indeed, despite the f act that the coef ficient is s tatis tically greater than zero, uncertainties not considered by such analyses (e.g... errors in the measurement of the exposure variable) make ;
it possible that the mortality risk might in fact be zero. Such coef ficients have in the past been applied without adequate attention to their. actual applicability to the situation and the ,
I 8 -
-4&+-w-g.e -haiy g e 7 -'y W'r-GT* -T--wwww w-wmw--'gy*ee--tt +p-7m - y -yy++g-*gww
10
. uncertainties involved. We refer readers to the lim it a t io n s regarding the application of mortality risk coef ficents noted in Section II.4 o f this report, as well as in Section IV and the
, Executive Summary of the 1982 HEED.
10
, T0XICOLOGIC ANALYSIS The 1983 HEED toa*. city analyses have concentrated on the evaluation and ranking of carcinogenic potentials and non- -
neoplastic health risks of particles. These assessments have O been based upon a number of 1p. vitro and in,y_iza, bioassays. Such inf ormation may be useful in the ranking of risks which are not separable using epidemiologic tools to confirm epidemiologic observations, to extend analyses into areas where epidemiologic
.6 s t u ', i e s have no t been undertaken and to provide insight into biological mechanisms of damage. The uncertainties discussed i
? underscored the current lack of fundamen:a1 knowledge regarding the biological processes involved, and point toward the need for further research to clarify disease mechanisms.
1 Carcinorenic Effects o_[ P a r t ic l e s C The main objective in this area was to obtain estimates an1 ranks of the carcinogenic risks posed by various sources o' airborne particles, including those derived from coal, oil, wood, diesel, and leaded and unleaded g a s o lin e combustion. A compara-tive potency statistical model, based upon a number of short-term bioassays and human epidemiologic data, was developed (see Section III.! ) . Application of the model to data generated using the Ames Salmonella typhimurium bioassay provided estimates of increased relative risk of cancer and associated degrees of uncertainty. Ames bioassay data were used primarily because it was the only assay for which sufficient information was available in the published literature to allow f or comparative analyses between the particle types of interest to our group. _
In this analysis, estimates of human cancer . risk f rom ex trac tab le org anic s were presented as increments in rela tive risk per ug/m years. As shown in Table 1, estimates were
, generated for emissions from light-duty diesel vehicles, catalyst
- .~- spark engines, non-catalyst spark engines, woodstoves, residen-l tial heaters, fluidized bed coal combus tion, conventional coal l and o il c o mb u s tion. jhese estimates produced increments in relative risk per ug/m from approximately 10~3 extractgble l . organics - years ranging l to 10- . Due to various sources of
'-y uncertainty (predominately involving inter-test potency relation-
. ships), the 95 percent confidence intervals ranged over approxi-mately five orders of magnitude. Further, this range cannot be viewed as conservative, since certain non quantifiable or inade-quately quantified elements are not fully captured by the model.
For example, these include uncertaintie s in the epidemiologic
- , data to which the bioassays are linked in the statistical model, potential variations in potency due to different extraction pro-tocols, and potential non-linearities in dose-response functions.
However, given the magnitude of the range determined in this .
,:.= :.----. -.r -.-.:: L~ _ _.-_. . z. . - . - . - - . - . - . - - - . -
--- er .
. . . . . . . - - . . , . . . ~ ~ . . - - . - - - ~- - .-- - -- - - " - - - " ~
i
. 7-i analysis, we do not feel that the exclusion of these.further
. uncertainties are likely to be of significance. Because of the limitations inherent in these estimates, we conclude that it is premature to use data generated from short-term bioassays of g-complex combustion products f or anything more than f or general
! comparisons.
The extension of this analysis to comparisons of risks due to actual exposures experienced by the population as a whole is -
i difficult. Exposure data have generally been collected ' or .3 estimated in terms of particulate mass, while the risk estimates generated by this analysis are in terms of extractable matter. '
.i An approximate interconversion between these units based.upon extractable f raction (the percentage of particle mass which is extracted) for each particle type is possible. Such a conversion
- allows for general comparisons when' source-apportioned ambient 9 exposure data are available, but this process also introduces additional uncertainty into the analysis. Given the large uncer-tainties already evident, we have decided that to proceed with this aaslysis was not justified at this time.
I- Several points of significance have emerged from this inves- -?
tigation (see Section II I .1 ) . First, the potency values clearly contribute the largest degree of uncertainty in any analysis of
_ specific health effects, far outweighing uncertainty resulting from exposure estimates. This underscores the need for the development of tes t s ys tems which more clo s ely approach the is, y.iyg. human situation. Such tests must await a better under- !
, standing of the fundamental mechanisms involved. Second, due to
' the significant overlap of the 95 percent confidence intervals, one cannot currently make any fira distinctions between the
~ potencies of these particles. Therefore, from a policy perspec-tive, it would appear vise to concentrate on controlling those particle sources which contribute most heavily to soluble v organic exposures. In situations where the magnitudes of expo-i sures are similar, then central potency estimates, or more con-servative upper-bound risk estimates, may be used to guide policy decisions. Estimates of predicted numbers of lives lost do not a
'ppear warranted at this time due to large uncertainties involved and the obvious economic and social implications. ,
3 Bioassays Relevant 3,g Non-Neonlastic L.g.ng Disease l
l Evaluation of non-neoplastic pulmonary effects has focused -
on 13, vitro and 13.Vi?9. studies of a wide range of particles, particularly metals and sulfates. The effects of particles have k been evaluated using inf ormation f rom five categories of bio-assays: ( .' ) measurements of macrophage function 13, vitro using both aniasi and h u r. a n cells; (2) pulmonary function mes-surements;s3) inf ectivity models; (4) lung lavage fluid changes; and, (5) mucocilliary transport.
Our preliminary studies in dose-response relationships (dis-cussed in Section I I I .2.1 ) for in vitro tests of metals have determined, by linear regression analysis, the estimated concen- '
10 -
~
_.-L.. A <
! 'O
! trations required to produce an experimental outcome level (which is specific for each bioassay) of half the maximum effect (gC5.0 )
- The five metals investigated had relatiply stable )toxgetty
" +
.c r ag*k in g s f p+r d if f e r e n t bioa s says (i.e., Cd ) YO 3 Ni )
l Cr ) Mn ). Further, this order was consistent with ranked l
exposure limits f or occupational Threshold Limit Values (TLV).
l Using an 13, y,iy g, b io a s s ay that assesses increased su s c ep t ib ilit y to infection and extrapolating from experimental doses in mice to _
equivalent doses in man, it appears that the dose of some metals
.. from urban exposures in man reach levels that are within a factor
- of 10 smaller than those doses which enhance susceptibility to bac-terial infection in animals I'
l
. However, no reliable . scientific data exist which directly i demonstrate effects in humans due to chronic exposures to these
- metals at ambient levels. An extensive data base does exist
. which demons trates serious biological ef f ec ts in a variety of animal bioassays f o llo w ing short-term exposure to metal concen-trations above those found in ambient air. The importanca of the
'ef f ects and 'the similarities of the biological sys tems between animals and humans suggest that chronic exposure of humans to L these metals may result in significant, but as yet unquantified, health effects.
We have focused our toxicologic review of sulfates on mucocilliary clearance and respiratory mechanics. As discussed in Section I I I r2. 2 , these effects appear to be related to the strength of the acid and, therefore, most attention is given to the e f.f e c t s of sulfuric acid. Following repeated exposure to su lf uric acid, several animal systems have developed persistent alterations in mucocilliary clearance rates. Histological exami-nation of airways shows changes which account for the reduced clearance seen in airways of animals exposed to chemical irritants. -
Since the sequence of events and tissue histology following
. sulfuric acid exposure resembles those seen in humans who smoke cigarettes, an important ques tion is whether sulf uric acid and
. diminished mucociliary transport contribute to the development of e- bronchitis. At this time, however, the progression of clearance dysfunction in the pathogenesis of chronic bronchitis is not known.
e The potential impacts.of nitrates on humans and animal health has received relatively little attention. However, infor-nation on effects can be inferred from exposure studies for nitrogen dioxide, since the gas dissolves in the respiratory epithelium and f orms nitrates. For example, animal inhalation s tudies with NO2 expo sure show produc tion of f unctional decre-ments and anatomical changes in peripheral airways and air-spaces which are consistent with impairment of respiratory d ef en s e s y s t em s (s e e S ec tion III.2.3). Thus , while the effects of nitrate aerosols on respiratory disease have not been estab-lished, there is a body of toxicologic data which raises suf fi- '
cient concern to justify further investigation.
s 11
- m. _ .. . , - . . _ . _ _ _ . . . . _ - . _. ,. ..;_._.
i j.
CONCLUSIONS AND RESEARCR RECOMMENDATIONS l'
In the Executive Summary of the 1982 HEED'on airborne par-ticles, we pointed out the need f or the development and use of p!
more biologically plausible exposure indices (rather than the I typically available TSP or SO 4" measures) in assessments of epi- .)
demiologic data regarding health eff ects and air pollution. O u r- !
second E year's ef forts . were direc:ed toward improving our under- -'
. standing -
of the extent of health risks resulting from the Dl components of inhaled airborne particles. We were partially , l successful in completing this task. Therefore, the research '
summarized in this report provides estimates of mortality'and morbidity effects of exposures to ambient particles in terms of
, , various particle mass measures (see Table 1).
- There were several interesting findings from our epidemio-logic investigations of the morbidity ef f ects of exposures to '
airborne particles. A survey of the exis ting literature on the morbidity ef f ects of human exposures to particulate - matt er indi-cated that very few studies have addressed ac health' effects.
of particle pollution below levels of 1000 ug/m$te(measured in 24- 7' hour average TSP equivalent s). However, the available data f or short- and long-term exposures to particulate matter did not suggest the existence of effect thresholds. For this reason, simple linear coefficients were derived which related particle
. concentrations to morbidity effects. Also, an original epidemio- ,.
logic study using the National Health Interview Survey morbidity --
data and an'index of fine particle pollution (based upon airport visibility data) indicated a correlation between fine particle
- air pollution and human morbidity. This relationship persisted even when the analysis was controlled for inter-city and seasonal effects, but was evidenced only among persons reporting periodic ,
limitations due to chronic conditions. >!
Time-series analysis or historical Coefficient of Haze (COH) and mortality data collected in New York City (NYC) indicated-that COR was related to temporal variations in mortality.
However, until validation s tudies in several other cities with . . .
different pollution characteristics and weather patterns are l' undertaken, the results derived from this analysis apply directly
, only for the six of sources 'and time patterns of concentrations
, observed in NYC between 1963 and 1976.
Cross-sectional analyses of total mortality in Standard ~,
Metropolitan Statistical Areas (SMSA) across the U.S. indicated that the use of fine particle health risk coefficients provides
-an improvement over previously employed particle mass measures
- (esoecially TSP). Aside from the biological plausibility of such a finding, statistical analyses showed FP measures to be more consistently associated with mortality and morbidity health "
effect outcomes than either TSP or sulf ate. Unfortunately, as noted elsewhere in this report, the uncertainties surrounding even these new estimates remain so large that we are still com- '
pelled'to emphasize the need to consider the pos sibility of a 12
.-.,v--- , .w,...m,,,y_,,,,_...,m,_, ,,.,_-mm,_,,__m_,,_-____ _,__,.,,_,._,,,,__m.,__,m,,.m,-..~m_-.,_.______._e.,
. ~ . . . . _ _ __
- _ - . - - -- .- . . ~.
O
~! sero-risk ~ coefficient-(or no effects ~due to exposures to' airborne 3 particles at the levels studied). It was also found that, using
- . mean - exposures 'and expressing mortality risks in terms of similar a units, the NYC time-series analysis indicated less than half the mortality risks prdicted by our cross-sectional analysis. While preliminary in nature,. this f ind ing is consistent with the expec-i tation that the cross-sectional studies may capture more of the chronic-health effects of air pollution than would time-series _
^
. studies ( e .g . , Evans et al., 1983).
C An important source of uncertainty in the analyses discussed
, c above is contributed by the use of central-site pollution da ta, as opposed to more accurate estimates of personal exposures
-e (including separate indoor, outdoor and in-transit components).
Another key contributor to the uncertainties identified (see
- C Sections II.1 and II.4, and Appendices I and II) is in the spa-tix1 and temporal variability of the aerosol composition and its toxic components. Since siza-specific mass measures do not
-account f or these dif f erences, it is not yet pos sible to b e tter
! resolve the size of errors as sociated with the current estimates based upon particle mass. However, as these uncertainties in 4
..U exposure estimates are reduced, it is expected that the resultant biasing effect (thought to be downward) on the FP coef ficients will diminish and that the confidence interval of the estimates can be reduced. This will then allow further improvements in the statistica1 reliability of health ef f ect assessments derived from observational e p id em io lo g y.
During our bioassay investigations, we attempted to address some of the potential source composition problems mentioned
, above. We studied the non-neoplastic toxicities of various L metals and found their rankings to be consistent with the TLV rankings. Using a relative potency model, we also estimated the range of carcinogenic risks posed by different types of airborne -
particles such as those emitted f rom coal, oil, wood, diesel and gasoline combustion. The main conclusion arrived at after devel-oping estimates of incremental relative risk of lung cancer (based upon Ames data) was that, due to large uncertainties , it
, is not possible to reliably discriminate between the potencies of e" different types of particles. For this reason, the population risks associated with exposures to various types of ambient particles must, at this time, be ranked on the basis of quanti-
.- ties of extractable (soluble) organics emitted rather than indi-vidual estimates of potencies.
After evaluating the evidence derived from epidemiologic and toxicologic components of our research, we concluded that epi-demiologic risk coefficients (although quite crude and only appropriate for the development of bounding estimates) are per-haps the only useful tools readily available to air pollution risk analys ts today. Nevertheless, we s trongly believe that a proper application of these coefficients (preferably those based upon fine particle exposures) demands extreme caution and a specification of qualifications and uncertainties. ,
I 13 i.m--ear--*-.y-A--+mm. .- _ --
~~ . - . .
. ._____.. _ .- .__ .. _ _ _ _ . . .____, . .-._c.-...
t .T i
4 There are several avenues of research which may yield more useful assessments of the population health risks resulting from exposures to airborne particles.
For. future health studies,.we need combinations of both
. improved ~ exposure and health measures. For example, the use of ,
health surveys or census data would be greatly enhanced by the !
addition of measures that would at least partially account for cigarette smoke and indoor exposures. -We clearly need more- -
prospective health studies to enable characterization of chronic 3' and acute health effects. Also needed are better estimates of '
personal exposures to particles, including information on' indoor /
outdoor exposures by source and chemical composition. However, since new data s ets . will take a decade or two to develop, existing retrospective population health data sets should
. continue to be re-analyzed with.more representative exposure a j estimates. Especially for the study of morbidity, existing data bases should receive additional attention.. For this purpose,
-historic data sets might be re-analyzed using novel exposure estimates (as attemp ted in Section II.2 of this HEED using air-port visibility to estimate ambient fine p a r t ic 'A e concentra-tions). In addition, existing health surveys such as HIS and ^
NHANES (National Health and Nutrition Examination Surveys) should
!. be' expanded to better accommodate analysis for air pollution effects. Further, various exposure-averaging times should be
- examined in order to address ques tions regarding response times
! associated with observed biological ef f ects of air pollution.
Particular attention should also be paid to improving exposure estimation by incorporating available personal monitoring data and exposure modeling techniques. Finally,_ associations obtained l' betw'een air pollution exposures and different measures of human
! morbidity should be compared in terms of their significance and biological p lau s ib ilit y.
- 1. ..
k l For time-series analysis, we need an improved model of the
! ef f ects of confounding variables (such as temperature) on acute l mortality and morbidity. This also suggests the need to develop
- a physiologically based model of acute mortality in time-series studies.
O Based upon our results from cross-sectional mortality analyses (which indicate the importance of fine aerosol frac-tion), ambient measurement of-fine particle should be expanded.
Moreover, future work should explore the sources and composition
- of fine particles to examine their respective importance in the ,
interpretation of epidemiologic data bases. -
Future bionssay studies should address the question of bio-logical significance of assays with respect to human disease (bo th neoplas tic and non-neopla s tic). Basic research into the
, physio-chemical mechanisms of carcinogenesis may eventually pro-vide a model that, in conjunction with bioassays, will be quite useful to risk assessment.
em
~
14 -
. s 1
l
- ry '
Future research should also be directed toward th'e refine-ment of present bioassay technology, especially those utilizing the comparative potency approach (which necessitates the use of-
... multiple systems and compounds). For instance, approaches such
-as those used in the gPA's cancer bioassay studies may be improved upon by expanding the potency sets to include the auch larger group of known animal carcinogens. However, relationships between as says still need to b e. determined for a greater data _
base to account f or interlab and intersample variability.. The-s ,. ~ effects of differences in chemical co mp o s it io n on assay results
- also need further clarification. It seems probable that certain assays will be ef f ective as predictors of carcinogenic potency f or certain chemical clas s es. Finally, ques tions of bioavail-
, ability must also be addressed in assessing mutagenic potentisi of urban aerosols.
t
.G 1
4 i g 3
(
i 1
e k e
^
15
. - . ~ ~ .. , _ : --~~ - :. _ , ==..- ::- - . ,_ _. :: - _ . - - , _ _ . , _ - - . _ - , - - . =
e.~. . . , - - - w
_. % ,__ w. O mew.A, AEms
,j , .
.[,
O i ..
.y I. Introduction -
i This report is the second generic Health and Environmental
'I Effects Document. (REED) on airborne particles. . The 1983 BEED prodides results to date of continuing analyses of the nature, 3 magnitude and uncertainty of potential health impacts of airborne particles.
The scope of this REED ' includes those particles. which are most commonly associated with general types of fossil fuel energy technologies. . Particle types considered in the toxicity assess-
- ments included
- fine, inhalable and total suspended particles; -
L sulf ates; nitrates; various trace metals; particles f rom coal,
- oil and gasoline combustion; and, diesel particles. Characteri-
! zations of potential health ,ef f ects .resulting from exposures to .
airborne particles primarily have been developed from the ,
analysis.of toxicologic and epidemiologic data utilizing ambient '-
air pollution measurements. To a larg e extent, predictions of population exposures to particles were derived f rom the EPA's Storage and Retrieval of Aerometric Data System ( S A R O A D') and Inhalable Particles (IP) network data bases. Characterizacions
) of ambient particle concentrations and apportionments of typical ,
l sources of airborne particles have been discussed in various *'
-sections of this report (e.g., II.2, II.4 and III .1 ) as well as in Appendix I.
Section II. presents ep id em io log ic assessments of the health ef f ec ts of human - expo sures to particulate matter. In Section s II.1, w e included an updated review and a preliminary re-analysis l of previously reported morbidity data bases. Using the 1979 Health Interview Survey (HIS) data base of the National Center for Realth Statistics, in Section II.2 ve present investigations into the association between fine particle pollution and human morbidity. In Section II.3, we devo ted mo s t of our efforts to ,
l describing the results from our re-analysis of 14 years of daily -
mortality and air pollution data collected in New York City.
Section II.4 includes the as s es sments derived f rom our cross-sectional analysis component of our epidemiologic investigations.
This analysis searched for relationships between geographic dif-f erences in fine particle, to tal su spended particulate ma tter, y inhalable particle and sulfate concentrations in the U.S. and .
geographic dif f erences in mortality races. Further details on
~
epidemiologic investigations regarding mortality and morbidity effects of air pollution can be found in Appendix II. .
Analyses of the toxicity of airborne particles (presented in e.
Section III and Appendix III) concen.trated on the evaluation of neoplastic and non-neoplastic health risks of particles. In Section II I .1, we developed estimates of the carcinogenic risks posed by various sources of a irb o rn e particles, including those derived from coal, oil, wood, diesel, and leaded and unleaded gasoline combustion. In Section III 2.1, we eva lu a t ed bioassays 3 relevant to non-neoplastic lung djsease u s ing*inCr" y.ly.g studies of metals including Cd +, YO ~, Ni vit and ro and Mn p.
Our toxicologic review of acid sulfates3 ( p r e s e n, t e d in Section ,
16
.O III.2.2) focused on the effects of sulfuric acid exposu-res on
- mucociliary clearance and respiratory mechanics. Potential implications of these lung function changes to human chronic bronchitis is also included in Section III.2.2. Finally, in Section II I .2.3 , we reviewed the available literature on human and animal health effects of nitrates.
Our assessments in this HEED are concluded in Section IV with a summary of the principal conclusions and research needs -
indentified during our analysis of health risks resulting f rom exposures to ambient particles. The reader is referred to the three Appendices accompanying this report for details on specific points.
if r
C
- 1 i
( -
i l
l i e
- (
h 17
. . -.. - ...- - ~ ~
. - . . . - - - - - - . . . . . - . = . -.-- --. .. . - -- =
. l. -
-S j
II. Enidemiolonic Assessments d Health. Effects g Esnosures ig, Ambient Particulate Matter In Section IV of the 1982 REED on airborne particles, we provided epdemiologic assessments .regarding the nature and the extent-of association between air pollution and population health effects.. In this year's . HEED , .w e attempted to expand our epi-i
~demiologic studies on the morbidity and mortality effects of- -
particle air pollution in to order to: .(1) explore the likely ,
range of air pollution. effects on alternative health outcome and ~
exposure variables; and,(2) charac teriae the uncertainties and
- sensitivit-les of results to choice of data base and models used.
, The morbidity ef f ect s component of our investigations -
I included both an updated review of the epidemiologic literature 3 ~
on morbidity ef f ects of human exposures to ambient particulate matter and new analyses of available morbidity data. Our review this year encompassed several new studies, most of which were
, made available to us af ter the development of the 1982 HEED. We
, also conduc ted new analyses using morbidity data f rom the 197 9 i National Center for Health Statis tic s (NCES) Health Interview '
. Survey (HIS). The principal health outcome variables included in this preliminary analysis were to tal Res tricted-Ac tivity Days z.(RAD) and to tal Work-Los s Days (WLD). This work f ocus ed mainly on examining the extent of the realtionships between fine par-ticle pollution and human morbidity measured by either RADS or
~
WLDs. '
i l To estimate the magnitude of possible mortality risks associated with particle exposures, we concentrated our recent
. efforts on studies of daily time-series and annual cross-sectional nortality data bases. In particular, we re-examined the 1963-1976 New York City daily mortality and air pollution 1 J series to develop estimates of daily mortality risk coefficients. '
l Under a separate analysis, we also s tudied the implications of l using alternative exposure indices in cross-sectional studies of air pollution and total nortality. Using a methodology similar to the one employed in the 1982 HEED, we analyzed 1960 and 1980 cross-sectional data bases, along with information from the EP A'S .'
1979-1981 Inhalable Particle (IP) network data base.
- In the remainder of this s ec t io n , we discuss the results of our epidemiologic assessments regarding morbidity and nortality -
effects of air pollution. Further details on our epidemiologic l
studies can be found in Appendix II of this report. ~
II.1 Review g Evid enc e f.gg.a ob s erva t ional Inidemioloav:
Morbidity Effects d Particle Pollution Introduction
. .a This section updates the initial asses sment of morbidity consequences of particle exposure included in the 1982 HEED. We have also reviewed the morbidity analysis f rom an independent -
risk evaluation (Manuel et al.,1983) that wa s de s igned to d e t er-c 18 L
O 1
1 aine the benefits as s oc ia t ed with alternative standards for par-4 ticulate matter.
7, 4 .O In developing this REED, most o'f our efforts in quantitative analysis have been directed toward nortality as an outcome.
However, if ~ mortality is plausibly an outcome of human exposure to particulate air pollution, there should be a relationship between this f orm of air pollution and non-f atal disease out- ~'
.. comes. Unfortunately, ascertaining whether particulate pollution
. , at current U.S. concentrations is linked to morbidity is hampered by an extremely limited data base. We have identified several reasons for this limitation, including:
Morbidity outcomes are measured with less regularity and precision-when compared with the relatively
.L ), complete vital statistics available for mortality.
This severely limits the opportunity to conduct
- aggregate-level, cross-sectional investigations of morbidity outcomes.
, c.. . Many morbidity outcomes (e.g., lung f unction) require the investigator to make direct measurements on study 2 subjects.. Also, most
~
o b s e rva tional s tud ie s have involved relatively few subjects and, therefore, do not have the statistical power to detect slight changes in nonspecific disease outcomes. Unfortunately, this is exactly the type of effect we would hypothesize to
, result from particulate exposure at current levels.
, . The majority of studies were conducted at particulate-matter concentrations exceeding levels of concern for standard-setting and risk evaluations ( i.e. , years ago
- and/or at very polluted' sites). .
I . Most experimenters do not design their studies to 4 develope quantitative dose-response information.
l Therefore, many studies which suggest a health effect
! from particulate matter are not useful for quantitative
{-p interpretations.
Interpretation of the morbidity studies mus t be qualified similarly to all non-experimental epidemiologic s tudies. The individual epidemiologic studies of morbidity can demonstrate an
, association between par t icu la t e matter and ill health, but they U cannot prove the causality of that association. The morbidity studies, however, do have one' major design advantage over the aggregate-level studies that use vital statistics information:
the individual-level analysis of ten employed in these studies
- permits more careful control of personal f actors (e.g., smoking and occupation) which might otherwise act as confounding items in V an : aggregate-level study.
I r
,e 19 i
- .. _: - .-_ = .__.= = = . _._ -. - _ -
, _ _. ._.__ ._ _ - ~
4 .
f Re s u l t s 1,g,g,3 Review g,[_ Morbidity Literature q
, Studies of the morbidity effects of chronic exposure to -v i particulate matter have shown upper and . lower respiratory symptoms-(including bronchitis) and reduced pulmonary function to he associated with particle [in Total Suspended Particulate Matter (TSP) eqgivalents] concentration in excess of approxi-mately 180 ug/m (Ware et aL,1981). No evidence exists in these ~
data to-suggest an effect threshold. The observational studies ?(
on short-term particle exposure are more sparse, of i
these studies address TSP levels in excess of 1000 ug/m ang(most 24-hour-j averag e). Those which have studied effects at lower concentra-tions.(Martin, 1964; and Lawther et al, 1958 and 1970) suggest j increased (with TSP athospital 600 ug/m adyissions in association .for cardiac with 802 or respiratory at 400 ug/m ig)inessand Ol 4
worsgning of health status among500 ug/m{ tics bronch (with TSP at 350 ug/m in c.omb ina t ion wit h $02 at ). As is the case in j chronici exposures, these studies do not suggest an effect
- threshold.
- Initial Results 1gg,3 Assessment 2.1 Morbidity Effects "'
In a preliminary attempt to derive simple linear-coeffi-cients for morbidity, we selected several highly regarded studies that provided quantitative estimates for air pollution concentra-tion and morbidity outcomes. Only the study of hospital visits ,
i addresses a concentration range for particles that is relevant to -
current air pollution levels in the U.S. The other studies 4 address relatively high particulate levels, and.one must extrapo-late beyond the range of observed concentrations in these studies to conduct a useful risk assessment. This approach must obvi-ously be taken with caution as very little evidence exists to _.
indicate whether or not the relationships observed at higher ..
J-
! particle concentrations hold at lower concentrations. With these l qualifications, coef ficients were derived and are presented in
. Table II.1.
I The coefficients shown in Table II.1 are left in the
! original air pollutant form (i.e., Smoke using OECD calibration, ,
1 British Smoke (BS) or (tsp)], but nonetheless present a bio-logically con s is t en t impression with the coefficients typically
, derived for sortality. The right-hand c o lu m n of Table II.1 indicates the range over which health ef f ects were observed in
- each of the studies. Air pollution is only responsible for a
, portion of this effect, since the intercept for the health effect '~
is generally greater than sero. This column is included because i any extrapolation outside the effects range will produce an l uncertain error.
i I
In Table I I .1, we also present results from the Machtech -'
! analysis for their lin ea r functional form (Manuel et al . , 1983).
j In' Appendix I I .1, we discuss the two functional f orms used by j Machtech in their analysis of morbidity outcomes and describe our '
- reasons for selecting the linear f orms as mos t jus tifiable and parsimonious. The-additional studies relied upon by Machtech are 20
.,v---
} (
"r r, c
- l g, _
~
Q 4 ;
d :
! Table II.1 Assessment of Morbidity Response Rates f i 't j tstimated txposure- ,
Response Slope Pollution Range Pollutant Range of Total ,
j ThatEstimatejs Measured Health Effects l l Type of Health Response (cases per /pg/m year /g)0,000 Based Upon (pg/m )** in Study Observed in Study *** ;
I
- Childhood lower respiratory 60i* BS (97-301) BS 23-36% reporting infection S02 (123-275) symptoms l Bronchitis in Males 100-170f* Smoke (DECD 4-27% reporting
} (smokers and nonsmokers) calibration) symptoms i BS (90-170) i, . Bronchitis in Females 0-10i* 2-10% reporting h (smokers and nonsmokers) symptoms F
i
[ age
! Acute Respiratory Disease 0-24 yrs 540i* TSP (0-200) ISP - -
1 (Manuel et al., 1983)* 25-54 yrs 100t*
l M 55+ yrs 120t* .l i.
l Excess Emergency Room Visits ++ 131** 24.5 t 9.3 daily l for Respiratory Disease TSP (14-696) TSP visits Excess Total Emergency' Room 20t* S02 (4-369) 94.3 i 14.2 daily Visits ++ visits Excess Emergency Room Visits 8.21*
for Respiratory Disease ++
Standard errors can not be estimated from these studies in a straightforward manner. Errors from sampling considerations are under investigation lby us; however, larger errors will result from nonsampling considerations. In our opinion, the errors should be considered to be half as large as ,
estimates.
Based upon annual average concentrations with exception of emergency-room-visit study that was based upon 24-hour averaging period.
- This column represents the inclusive range over dich health effects were observed in each of the studies. A portion of these effects will be due to factors other than air pollution.
+ Based on data of Saric et al.,1981. Assuming, conservatively, that TSP level is 0 in clean city.
++ These coefficients are based upon regression results for which TSP explained only approximately 1 percent of the variance in hospital amergency room visits (Samet et al.,1981). The 13 and 20 rates can a be computed by assuming a population base of 20,000 for the hospital. The value of 8.2 is based on estimates of Manuel et al.,1983, assuming population base of 31,000 and compares with our coefficient of 13 for excess emergency room visits for respiratory disease.
t f .
-___.-.._.m- . . _
w . . _ ..___ _ . . _ . . . . . . _ .
q: . . .
- [ 7 consistent with the results obtained in ths preliminary-analysis
- - (Table II.1).
a g~
It should be mentioned that the range of coefficients shown in Tab le II.1 au s t logically include 0 because of the extrapo-la tion' b elo w conc entrations ob s erved in the s tudie s.. Further-more, in addition to .this major extrapolation problem, there are ~
other important c o n t r ib u t o r s to the uncertainties of the.mor-bidity risk estimates. These can be categorized into sampling O and non-sampling errors. .
. Sampling errors ref er to the lack of precision of a sample result. If a sample were to be collected without sampling error, '
then one could reproduce from the sample the.results which would have been obtained if the entire population had been included. a To the extent that this is not true, sampling error exists. Non-sampling errors include a variety of factors that influence the uncertainty of the estimated particle air pollution / morbidity relationships. They include: confounding factors ( e .g . , cigar-ette smoking, socio-economic status, occupational exposures, race, prior exposures and residence); collinearities with other 2 pollutants (e.g., particles and sulf ur dioxide); changing mea-sures of particle pollution that are not extirely comparable
[e.g., British Smoke or Coefficient of Haze (C0H) versus TSP];
oversimplifications in estimating personal exposures from data collected at fixed-site monitors; and, biases due to historical and.across-community differences in particle and source -
composition.
These and other caveats (see, for example,Section II.4 and Appendix II of this HEED; and,Section IV of the 1982 REgD) should always be considered when morbidity risk es timates are ,,
In particular, if these coefficients are applied to aero- "
used. -
sols having a dif f erent makeup than the ones they have been based upon, the results may be quite misleading. In conclusion, due to potential error contributions by the factors noted above, the s tandard errors of the estimates should be considered half as large as the estimates themselves. Since further efforts to quantify some of these errors are currently underway, thevalues .-
presented in this section should be considered as tentative.
l I
d O
em W
r
, . O II.2 T,,ig g Zgg gigl,g P o l l u t i o n 13j, 1333A N o r b i d i t y .
As described previously, published literature on observa-tional epidemiology does not provide information on the nature
, 'O and the extent of the relationship between human morbidity ef-facts and population exposures to respirable particles. The major problem in determining fine particle / morbidity relation-ships .has been the limitation of the available aerometric data
' bases characterizing ambient concentrations of respirable or fine -
particles in the U.S. However, for the purposes of air pollution O health ef f ects investigations, mos t of the health s tudies con-ducted in this country also lack the desired number of pertinent health, behavioral, socio-economic and demographic variables which have been used in the surveys administered to large popula-
. tions. ' Faced with these serious limitations, we elected the option of using a comprehensive national health data bas e, the v 1979 National Center for Health Statistics (NCES) Health Inter-view Survey (HIS).
In order to examine the possible effects of particle air pollution on the morbidity measures ob tained from the HIS, we L,
developed estimates of fine particle mass exposures. These esti-i mates were derived from established r e la t io n s hip s between fine mass, from the EPA's IP Network data, and relative humidity-corrected aerosol extinction coefficients, f rom airport visi-bility observations. In addition to the daily estimates of fine particle mass concentrations averaged over each two-week survey i
period, a number of other weather and socio-economic controlling variables were developed. ,
The essential features of the data bases used and the
- results from the preliminary analysis are summarized below.
Further discussion regarding this analysis can be found in
, , Appendix II.2.
1 -
Descrintion gi 3,][g National Health Interview Survey The health data utilized in this report were gathered during
! the 1979 National Health Interview Survey. This survey-is a y national multi-stage probability sample of approximately 100,000
- . individuals. From information gathered in interviews, the survey attempted to describe the social, demographic and economic aspects
) of i llness, disability and the use of medical services. The
- j. survey has been conducted continuously s ince 1957, and is cur-3 rently managed by the Census Bureau for the NCHS. The health e concepts employed by the NCES differ somewhat from those used in scientific and medical studies.
, According to the NCHS, morbidity is considered to be depar-
- cure f rom a state of physical or mental well-being resulting from disease or injury of which the affected individual is
, .,; aware. Illness, however, is considered only one form of evidence of the existence of a morbidity condition, since a morbidity condition may lead to other types of actions such as the t
j *, ^
4 23
. . _ , _ . - . , _ _ , _ , , , , [_.,.._,.._.,. _._,,_.__,____.,_,__.__,_.,..m_ _ . , _ _ - . . . .
^' '~ ' ~ " ^ ' "
_ : .i . _ . . . . . . . - _ _ _ _ . - _ . _
4 restricting of usual activities, bed disability, work loss, the seeking of medical advice or the taking of medicines.
l An acute condition according to the NCHS (1975) is defined , '
as a condition which had its onset during the two weeks imm ed i-ately preceding the interview week and which involved either medical attention or restricted activity during that two . week period. Acute condition groups include: . infective and parasitic diseases, respiratory conditions ( e.g. , influenza, pneumonia and -
bronchitis), digestive system conditions, injuries, and other 3 conditions such as diseases of the ear and skin, as well as ,
headaches. By contrast, a morbidity condition is considered to be chronic if (1) the condition is described by the respondent as having been firs t noticed more than three months bef ore the week of the in t e rv ie w , or (2) it is one of the co nd it ion s listed
- in Table 1 Appendix II. Chronic conditions include, among '
others: asthma, emphysema, chronic bronchitis, and heart and mental conditions.
The survey also collected a variety of e:onomic and demo-graphic information, as described in various NCHS publications.
The 1979 questionnaire included supplements on smoking habit s for ?
adults and residential mobility. The smoking supplement was given to one-third of the sample, but unfortunately was not given to all adults in a household, thereby precluding an assessment of passive smoking effects for most of the sample. (Children in homes with a single adult may be an exception.) -
Preliminary Analysis U s ina 11Z,1 H,I.,3, F il e s . EP A's J,1, N e t w o r k h J.,g,g,3, 3,p_d,, d H i s t o r i c Airnort Visibility Records.
The health outcome variables considered in our analysis were: ,
(1) total Restricted-Activity Days (RAD);
(2) total Work-Loss Days (WLD);
(3) acute respiratory condition incidence; and, (4) RADS due to acute respiratory c o nd it io n s .
The two principal disability terms that we have used as health ,
outcome variables were RAD and WLD.
A restricted-activity day according to the NCHS (1975) is a day on which a person c u t s' down on his usual activities for
- the whole of that day becaus e of an illnes s or an injury.
A work-loss day according to the NCHS (1975) is a day on which a person did not work at his job or busines s f or at least half of his normal workday because of a specific illness or injury. In the Health Interview Survey, the number of days lost from work is determined only f or persons 17 years of age and over who reported that at any time during the 2-week period '
covered by the interview they either worked at a job or had a business.
=
~
24
\
. - - =. . . . .
O-From the 1979 HIS files, we have constructed two data sets f or analys is. Table 2 in App endix II.2 provid es a list of the variables selected and a brief description of these variables.
Prior to the modeling of the four health variables, we
'O (1) attempted to determine those behavioral and demographic fac-tors having the greatest influence on the observed disability rates; and (2) obtained estimates of daily fine particle concen-trations representative of population exposures in the Standard l Metropolitan Statistical Areas (SMSA) studied. We briefly ~
describe below the approaches used in these p r e lim ina ry investi-
- p gations prior to analyzing fine particle / morbidity associations.
Demoeranhic ABA Behavioral Health Determinants' Due to the extemely skewed nature of the survey data on the s morbidity outcomes (which were mostly zero) in the analysis undertaken, the health outcome variables were treated as dichoto-mous ra ther than continuous variables. Thus, grea ter than or r
equal to one RADS or WLDs were all grouped together in the cate-gory of " effects" versus the category of "no effects". Further-more, due to computational and aerometric data base concerns, g only the 12 SMSAs (representing 3,431 individuals) identified in Figure II.1 were used in our initial analysis. The cities were chosen after consideration of the availability of air pollution
. data on fine and coarse particles. The correlations between the dichotomized morbidity variables and other demographic, socio-economic and behavioral factors were then analyzed to determine the factors having the greatest influence on the observed rates.
All of the health measures show a s trong seasonal pat tern, as illustrated in Figure 3 of Appendix II.2. Fall and winter quarters show the highest rates. Not surprisingly, estimated fine mass exposures also demonstrate seasonality (fine mass con-4 9 centrations are higher in the summer months than in other months .
of the year).
t As displayed in Figure I I .1, all rates show noticeable in ter-city variability. Los Angeles is consistently among the cities showing higher incidence rate of RAD and WLDs. WLD and U RADS appear to have a similar pattern, as do incidence due to t
acute respiratory condition and RAD due to respiratory conditions i (c.f. Appendix II.2).
l' One of the most potent predictors of RAD and WLD is the presence of existing limitation due to chronic co nd it io n s.
,' G As indicated in F ig u r e 8 Appendix II.2, a lmo s t half of all individuals reporting res tricted activity also have a limiting chronic condition. However, for the incidence of restricted ac tivity due to acute respiratory conditions, the limited and no n-lim it ed groups exhibit greater similarity and low incidence-rate (around 0.04).
l ,
Economic factors and age clearly play a role in restricted-activity and work-loss days. The analysis discussed in Appendix i II.2 indicates opposite trends for RAD and WLD over the three l
~
u 25
.,, ~'7.K - --- , _ , - J - - . - - - - _ , . . . , . - . . . . - ~ - , . - - - - , . - . , _ . ~ _ _ - _ _
. _ _ . ~ . . . . ,m___ - , - .
.. . . -. n.= ~ -
.4 s
Figure II.1 Procortion gf Sample Reportina M 314, M hy_ h a
4 Baltimore ***************************
Buffalo **********************
.y Dallas ***********************************************
Kansas City *******************************************
Los Angeles ********************************************************
Minneapolis ************************************** ,
New York **************************************************
Philadelphia *******************************************
Pittsburgh **********************************
San Francisco ************************************ 3 J St. Louis ****************************
Washingten, D.C. ***********************************
0.02 0.04 0.56 0.$8 0.k 0.i2 0.[4 0.[6 0.[4 Restricted-Activity Days
. Baltimore ********************************
Buffalo ******************************************** _ J Dallas **************************************************************
Kansas City ******************************
Los Angeles ****************************************************************
Minneapolis *******************************
New York *********************************************
Philadelphia *********************************************
Pittsburgh **************************************************
San Francisco ************************************************* ~'
St. Louis *********
Washington, D.C. **************
MOG M12 0.018 0.02% 0 03 0036 0042 0 044 0.054 0 06
)
Work-Loss Days
.Em 26
_-m , _ _ _ . , _ - - _ . _- -----
O' categories of income. This may occur as a result of the higher levels of unemployment in the lower income categories, and because RAD data was collected from both working and non-working
!; populations. Occupational status may affect rates of WLD.
V Individuals classified by census occupation categories above 400 j (blue-collar workers, farm workers and service workers) were i
noted to have a auch higher rate of WLD, perhaps due to injuries or job-related financial compensatory effects. In contrast, the 4
opposite trend, which was not immediately apparent, was shown for ~
, RADS (c.f. Appendix II.2). Age is also an important determinant.
As shown in Figure 12, Appendix II.2, the incidence of restricted activity increases with age while WLD rates decrease with age. .
We expected that smoking status would be an important factor for all of the health outcome rates. In fact, as shown in
' Figure 13, Appendix II.2, the apparent effects were rather small, particularly for the respiratory condition measures.
-Marit-al status and sex also appear to be influential f ac- I tors. Single adults , either divorced or never married, tend to have higher rates of health ef f ects than married adu'lts. Rates g also vary according to sex, as shown in Figure 15, Appendix I I .2. ;
In every case, women reported a greater incidence of health I effects than men.
In summary, this initial analysis showed that the health effects rates derived from the HIS data are strongly influenced by demographic factofs, as well as personal and seasonal factors.
Since these variables may also be correlated with the exposure variables of concern, in the subsequent analysis of the morbidity effects of air pollution utilizing the 1979 NCHS/HIS records, we
- explored the extent of the influence of these potential con-founding variables on the modeled effects.
! .C .
Daily Lig,3, Particle Exoosure Estimates D e r iv ed fJ,,g.3 A iro o r t Visibility Observations Since the Health Interview Survey data was reported for individuals during specific two-week periods, it was necessary to G develop mean Fine Particle (FP) exposure estimates for these specific periods. At the time of the HIS work, little or no FP aass data were being collected at these 12 SMSAs. Typically, for example, sulfate data collected by the National Aerometric Sur-veillance (NASN) stations have been limited to one-every-sixth-day measurements. However, by using FP and visibility data O. collected in these SMSAs during 1980 and 1981, combined with airport vis ib ilit y records during the same periods, it was possible to develop site-specific and daily estimate 2 of FP exposures which were then averaged for each two-week period.
This was so because there -is a roughly linear relationship between fine particle concentrations and invers e visual rang e, with visual range decreasing as particle mass concentrations increase [see, for example, U.S. EPA (1979a), Latimer et al.
(1978); Waggoner and Weiss (1980); and Trijonis and Yuan (1978)].
~
O V 27
~.~ .7 ------.L__.-._:._--_-.-. - - , _ . - - . _ . - - . - - -
__ 7_ ,_ __ _ , _ _ _ ____7 i : -
The FF-visibility relationship is founded upon the principle that fine particles absorb and scatter light, causing an extine-tion of light and loss of visual contrast (reduction in visual
. range, V,). The extinction coefficient can be estimated from -
visibility via the Koschmieder Formula:
5,xg (ka'I) = 3.91/V, (km)
A summary of the site-by-site FF/B relationships utilised in this analysis is shown in Table II.Y.t Details of their develop- *
- meat may be found in Appendix I.2. Applying these formulae (developed from 1980-1981 data) to.the visibility data recorded l during the 1979 RIS study allowed for the estimation of average -
FF mass exposures in each SMSA during the individual periods of
-' interest.
Estimation d Norbidity Effects d gg,Polluti m Weather and,Other Factors Given the nature of our exposure a ca su re. (i.e., airport ?
observations of visibility which integrate pollution levels over a wide dis tanc e), we decided to aggregate central and non-central city data. Furthermore, since the reported incidence levels for the acute respiratory conditions and the restricted activity due to acute respiratory conditions were relatively low, we decided _
to limit our final analysis to the study of the associations between fine particle exposures and RAD or WLD.
For our initial modeling, we used Ordinary Least Squares (OLS) to estimate the regression of the dichotonous variables on air pollution and demographic variables. This method yields .,
unbiased estimators of coef ficients of a linear model for the s incidence probabilities and their standard errors. However, the d is t ribu t io nal assumptions associated with ordinary p-values are not satisfied, and these must be viewed with more than usual skepticism.
Table II.3 shows the results for RAD and WLD. The same
- sites of demographic variables were included in each fit for purposes of comparison. Two coefficients for fine mass are estimated--one for the limited group (defined in terms of chronic limitation) and one.for the non-limit ed group. As expected from *
, our previous graphical analysis (see Figure 17, Appendix II.2),
the sign of the coefficient switches to negative for the non- '
limited group. The coefficients may be interpreted as the change in the probabilityofanoccurrenceofahealthofject in a two-week period. For instance, an increase of 10 ug/m average fine mass concentration would be expected to increase the probability of a restricted-activity day in a two-week period by 0.056. If "
s we assume that the probabilities in successive periods are inde-pendent, then we can use the laws of probability to calculate occurrence probabilities for periods greater than two weeks.
- Under the independence of events assumption, an increase of 28 I
-a '
l-
- v. r O .c n . r.
. e 0 > o- i i t l ".! t 1-l . 's-i
_j i !
Table II.2 i
Light Extinction-Particle Mass Relationships at 12 U.S. Cities ~1 nf-K18e stI** Sent - K+
- Kal"fl
- Kalac l** -
Total Mean leean FP Coefficient Coefficients .__
SADOAO Num6er heat . h as (ngl E Ee Kg E2 SMSA Site Code of Obs. (km*8) (p9/m )
8 (ke-sp3/m )
8 (ka-ag g,37,g g ,37 ,y
- 1. Ims An9eles, CA 054100103 104 0.333 2 .034 2t. 911.9 69.82 4.2 0.16 1.07 8.621.5 -2.6*2 2.9
- 2. San Francisco, CA 056300003 !!9 0.204 2 .016- 14.42 1.1 62.41 3.0 0. Den 1.03 10.410.9 2.9** 1.6
- 3. tsashin9 ton, D.C. 090020017 94 0.194:.008 26.12 3.6 138.0t8.0 0.15 1.02 2.31 0.6 -0.!* *0.8 swa
- 4. Baltimore HD 210120009 76 0.2312.016 24.32 1.5 90.526.7 0.18 i.05 5.411.7 -3.2* i 2.4
- 5. seinneapolis, see 242260051 3!! 0.228 2 .0 % 86.7t0.8 65.41 4.6 0.0'# t .08 14.21 2.7 -2.e* t 1.4
- 6. St. Rouis, too 260010001 47 0.2862.020 21.12 1.4 69.52 3.9 0.12 2.04 s.0 t 1.6 .. 0.l
- t 1.4
- 7. Kansas City,seo 262 3enoo2 75 0.24ae.015 19.121.1 6a.st 4.3 0.20 t.04 6.32 1.5 -2.0*1 1.5
- 8. Duffato, NT 3 5064dMMI Il4 0.2762.014 40.0t 2.0 124.0t 7.0 0.19 f .0 5 2.4 2 0.fm 0.6** 0.9
- 9. Icew York, NY 334680005 50 0.2632.Ol? 22.02 1.4 82.4t 3.7 0.0e t .01 S.221.3 'O.0*11.3
- 30. Pittsburgen PA 390lotMae>9 77 0.290t.019 21.0t1.5 71.2t 4.0 0.13 f.04 7.7t1.3 -0.4*11.2
!!. Philadel46ta 6 PA 397840036 67 0.2331 .085 24.291.3 93.92 3.4 0.07 1.03 7.02 0.8 1.6*2 1.6
- 12. Dallas, TI 451130050 97 0.235t.009 39.3 t0.9 78.32 3.5 0.15 t.02 4.22 0.9 0.781 0.7 i-
- ht statistically sign 6ficant at I. . 05
- enessessions 1.ase1 e4== a selectemt seeset of ent a re data set (see Assewtin 13.
+
. ;5 9
. . J. l. . . . ~ '~
^ ~
~
'{ :
,I i*)
Table II.3 1
Results from Ot.s nearessions for mestricted-Activity Days and Work-Ioss D&Ys Q Restricted-Activity Days Work-toss' Days Estimate Significance Estimate Significance (Standard Error) Invol (Standard Error) Imvel .'!
.; variables
- X 1000 (P-value) X 1.000 (P-value) ,
i 8
' Age -0.74 0.042 -0.95 0.0001 (0.36) (0.22) .
, Noon 0.001 0.79 0.25 0.57 7' temperature (0.72) (0.45)
Hours of 12.65 0.69 7.29 0.36 precipita- (12.9) (7.97) tion ,
^~
Fine mass: 5.59 0.002 1.78 0.11 with chronic (1.81) (1.12) condition Without -0.95 0.28 -0.76 0.17 ,
chronic (0.89) (0.55) condition *
. ~ *Model used in regression analysis also included categorical variables on the following: income, city, sex, city residence, employment, smo: ting status, V,.
quarter of the year, chronic conditions, marital status, health status, and ..
blue-collar occupation. These variables represented fixed effects that could
- f. -
not be estimated directly.
N = 3,431 R8 = 0.14 e
e.
O
./
+
m' no ea 30 0 9 g
.-- - - - - - - - _~ - - - _ . _. - _ - . _ - -
_ .~ _. ._ ,
f O .
1 us/m 3 of the fine particle mass concentration over a year's period But, will translate into an incrafse in the RAD probability by I
0.14. an increase of 20 ug/m (over the mean of the base-line level) of fine particle mass concentration may result in an additional.95 percent chance in experiencing one or more restricted-activity days during the year due to (fine particle)
! air pollution.
4 i
, In an additional analysis, we fit the model for restricted-
! U activity days using logistic regression.
In this model, the j occurrence probability is modeled with the logistic function, and
' the coefficients, therefore, may not be compared directly. In this analysis, we excluded cases with no reported chronic limits-tion. Table II.4 compares the coefficients (a linearized coeffi-cient is shown for the logistic regression). As can be seen, the coefficients and p-values agree fairly well.
Discussion 21. Results The preliminary r e s't i t s reported in this REED regarding the
'. use of the (1979) NCES/ BIS morbidity files and estimated fine part icle mas s, to our knowledge, are the first of such studies in open literature. We are, however, aware of the ongoing work (which is similar in character to ours) by F. Fortney and his colleagues at Resources for the Future. Since the results from their work are not yet available, we can only compare our g' results to those of 3. Ostro (1983 a,b) and Crocker et a l.
(1979). These investigations, hewever, have studied the morbidity ef f ects of air pollution using the 1976 NCES/EIS dad
._S A10AD (the S torag e Re trieval o f Aerome t ric Data s ys t em)-(TSF, 304, etc.) air pollution files. Overall, our results seem to be consistent with those of Ostro, although our exposure measure is c quite dif ferent (two-week average fine mass versus annual average ,
TSP or sulf ate). Assuming an average fine mass to TSP ratio of 0.4 as Ostro's.
(1982 REED), our estimates can be " scaled" to the same unitj With this approach, our estimate of 1.7 8 x 10~
incremental WLD probability estimate converts to 7.5 x 10~',
which is about half of Ostro's estimate (0.0013 - 0.0017) as c
presented in ostro (1983b). Giver. the fact that our es timated significance level is around 10 percent, this cannot be inter-preted as a different result, since Ostro's estimates will most likely be within the estimated confidence intervals.
'Another way of (at least, empirically) comparing our o findings is to translate our estimated coefficients into measures of elasticity (a method especially popular among economists).
Simply put, the elasticity indicates the percentage of change of a dependent health outcome variable which is associated with a one percent change in air pollution, and thus provides an index of relative effects. Fine mas s coefficients (for the limited a group) f rom Table II.3 were scaled by mean pollution over mean effects to develope Table II.5 presenting estimated air pollution / ef f ect elasticity coefficients. saploying work-loss days as an example. Table II.5 suggests that a one percent reduc- .
tion in FFconcentrations would reduce work-loss thedays of the 31
~. : .....~~ . . : ~ ~:. . . . . . . . . . - . ._L...~~~'
- ~ '
4
,i-Table II.4 Concarison of Logistic,Recression ggl, M ,e
. Coefficient and Model ,
Standard Error p-value Linear 0.0065 0.042 (0.0032)
,' Logistic 0.0072 0.042 (0.0035) ,
- Restricted-Activity Days excluding cases with no reported chronic limitation.
Table II.5 Estimated Elasticities for the Outcome Variables Mean Fine Mass Mean Estimated outcome ug/m 3' Outcome Elasticity Restricted- 18.75 0.149 0.70 ..
Activity Days 18.75 0.403* 0.26
- Work-Loss Days 18.75 0.046 0.73 18.75 0.072* 0.46
- Represents averages for the chronically limited group. ,
O s
9 4
32
iO.
limited group b'y half a percent. These results and elast cities are comparable to those given in Ostro-(1983a).
O Perhaps the most important difference, or an issue of further concern, is the negative sign of the sulfate damage coefficient reported by Ostro versus the positive fine mass coefficients obtained in our study. One obvious reason, of course, is the use of a single particle exposure measure in our -
work versus the multiple particle exposure measure employed by
'_ps Ostro. However, it should also be mentioned that in our analysis we were able to use daily estimates of fine particle mass to i predict average FP concentrations corresponding to the individual-specific two-week recall periods. In contrast, Ostro's annual average sulfate estimates were based on the
! , limited (every sixth day) EPA /NASN sulfate measurements. Never-theless, since sulfates and fine mass are typically highly cor-related, caution is advised in the use of our current estimates until joint regressions investigate collinear effects of various particle and gaseous pollutants have been completed.. Presently, our results tend to sugges t morbidity ef f ects of fine mass air i ,s pollution at the significance levels of p = 0.1 or less. The
, most significant effect (p = 0.002) is for the restricted-i activity day outcomes for the chronically limited group.
However, it must be stressed that, for all the non- (chronically) limited groups, the fine mass ef f ects were non-significant and i negative, suggesting that the association of morbidity outcomes
, , o r. l y for the sensitive population.
l Our results can be summarized in terms of possible ranges l (or 95 percent confidence levels) of incremental risk f or the L population subgroups with chronic limitations, as follows: '
.0 (1) for restricted-activity day outcome: .
l 0 - 0.01 per ug/m 3 FP per 2-week exposure period (2) for work-loss day outcome:
- 0 0 - 0.004 pe. ug/m 3 FP per 2-week exposure period l l* These ranges are based on the limited OLS and logistic model runs f i
reported in the text and should be considered tentative, since G further model refinements are presently being considered.
In summary, we are aware of the f ollowing f actors contri-buting to the overall uncertainty in our results. Only one measure of pollution has been considered. This is a problem .in two ways. Firs t, although we have found a strong relationship ,
L between fine mass and visibility, a similar relationship may also hold between estimated fine particle concentrations and other pollutants such as sulf ates or ozone. Second, even if we were using fine mas s directly, we would still have to consider the -
conf ounding ef f ects of other air pollutants as well as passive 33 w_:. a .= - _ _ . - - - - -
~
~
,. ._ . - _. ^
i smoking. The matching of health responses to air pollution and
,a meteorological observations are expected to introduce additional
+
uncertainties. Since we only _know the week of the interview, the ,
_. assigned exporure does not conform exactly to the two-week recall 't period. In addition, the adequacy of monitoring data as a measure of personal exposure is, as always, highly questionable.
Finally, the problems of missing data and ,the possibility of r epor t ing b l.a s e s r equ ir e f ur ther inve s t ig a t ion. So far in our analysis, we have classified the missing values in the unknown }, .
category.
9 9
r.%
J B
J 4
W D
i g,
).
4 O 34
__ m . . ._. .
- n. .. .~.-. . - ,
0-
.II.3 Inv e s t ie n t ion s g 1]Lg, Mo r t a li t y Effects' ~
g- M Pollution g,g.iB1 Time-Series Studies In this section, we present the~nain results of our con-
. l<*
tinuing exam ina t ion . . o f time-series epidemiologic methods. Our primary eaphasis was on analyses useful f or interpretation of results froa.recent time-series mortality studies (Schimmel, 1978; Masuadar, -Schimmel and Higgins,.1982; Mazundar and Sussaan, 1983).
~
Thess:recent studies have exam in ed the air pollution /
.aortality as sociation in _ New York City, London and Pittsburgh,
.'s .together representing the main body of results currently avail-able. .In order.to provide background.for the' analyses that follow, we first review some of the special f eatures and chal-
- , lenges of observational (as opposed to exp er im en t al) time-series sortality studies.
Introduction Using many years'of dVily observations, the time-series approach uses statistical methods to es timate the influence of :
daily air pollution on. daily mortality. There are, however, l several issues which preclude direct estimation of effects and/or 1 4
cloud. the interpreta tion of the results ob tained. One issue is l that of " temporal confounding" (i.e., the potential existence of.
variables, either measured or not, which are correlated in time with air pollution and exert influence on mortality independently of air pollution). Ignoring such variables in the analysis might lead-to over-estimates of pollution ef f ects. Temporal confounding has been thought ' of by mo s t investigators is falling into two !
categories or components: low frequency and.high frequency.
l
. -The lo w-f r equency component is-the shared seasonal cycle of daily mortality and air pollution. In the typical time-series
- C approach, this annual correlation is assumed to reflect con- --
founding by unmeasured variables,-and an attenyt is made to remove its influence prior to the analysis. This is of ten done by re-expr e s s ing ' the variables to' be analyzed as residuals from
. their 15-day moving averag e s. This approach has the ef f ect of "f ilt e r in's ou t" (r emov ing) c e r ta in lo w-f r e qu en c y co mponen t s in
- -L .the da ta.
After the variabics have been' filtered, the remaining "high g ,_ f requency" residuals are us ed in regres sion analysis. At this stage, consideration is given to which variables, other than 1
pollution, should be included as explanatory variables. Tempera-EC ture is one such varisble.It is known, for example, that death rates are elevated during summer heat waves. Other more moder-ate temperature excursions might also be influential. Of course, variations in temperature and other weather variables are closely linked to variations in air pollution concentrations. This is true both for se.2sonal-trends (Iow 3 frequency) and for day-to-day
- variations-(high f requency). Thus, weather is a potential con-l f ounder of the air pollution / mortality as sociation. Since the (
4.
Iow-f requency component s in the variables (including weather) have already been removed through filtering, what remains is just -
m l
35
. _ __. _ _ . .._._._______.._.______-.J._._____..____._--_. . _ . .,_ J
- i. L . :- - . .. . ~
. :- .. - -. . - - " ~ . .- . : 3::~ ~ :.7.C
.w the "high' frequency": conf ounding. Most. time-series investiga-tions have included weather variables (af ter appropriate fil-
. tering).in their regressions. For example, Schimmel (1978)'
i; included nine - functions of temperature as explanatory variables r in his regressions.
It.,is'important to recognize that these concepts-of temporal confounding are difficult to disentangle. Our concep tual model ,
of the system studied suggests-that the potential for confounding -
j exists, but tells us little about its exact nature.- W it hou t such 9 ~
- 1 information, the choice of a method used
- to control for temporal
.7! confounding introduces a certain amount of uncertainty. The
- results of the modeling process, quantitative estimates of-the
]7 influence of daily air pollution on mortality, carry with them
! this source of potential error. Clearly, too little control .
, could lead to over-estimates of pollution effects, while too much a control could lead to under-estimates. However, without a sound basis f or choice of model, it is impossible to know whether any
- particular choice leads to too little or too much control.
1 Another issue which clouds the interpretation of time-series results is the expression of coefficients for' particle air pollu- p tion effects in terms of Coefficient of Haze (COH) or British Smoke (BS), rather than, for example, Total Suspended Particulate l -(TSP) or Inhalable Particle (IP) concentrations. The latter set i of. unit s (or other particle-size-classified mass concentrations)
- vould be of more direct use to policy analysts and might also be
- more physiologically interpretable. However, soiling data are 1 i
almost always used in time-series s tudies because they are the
. only historic data generally available on a daily basis for
- _ extended periods in large cities. As stated in our 1982 HEED l .(Appendix II .3 ) , the relationship of filter soiling measures (COH
- , and BS) to measures of mass concentration varies over both loca-
- . tion and time. [ A recent study of one summer in Detroit (Wolf f 3
'~] et al, 1983) found that' COR was closely related to elemental ~
< . .: carbon content of the aerosol.}
. S e n s i t i v i e v A n a 1 y s i s ,g 1 J,gg Y_gI,h, ,Q,,i,,g,,y.
gg,_ Pollution / Mortality Results As previously mentioned, the lack of theoretical underpin- -
ning of the temporal confounding hypothesis leads to uncertainty in choosing methods for its control. In this section, we sun-marize results from our time-series investigations which tested .
.'J the sensitivities of results and uncertainties to a' range of ,
- plausible modeling choices. - 4 Descrintion gi p,,ggg a
The data used in the sensitivity analysis is a subset of the
- New York City data set ob tained f rom Dr. Herbert Schimmel (see ,
Schimmel, 1978). It consists of 14 years (1963-1976) of daily s measurements of sortality (the sum of heart, other circulatory, respiratory and cancer mortality), coefficient of haze, sulf ur dioxide (SO2) and temperature. The means and standard -
36
_m-eE-
..m ;
. l n
.A deviations of these variables are given in Table II.6. Through the middle of 197 5, the pollution data were collected at the East 121st Street "laborabory" s tation in Manhat tan. The last year and a half's data were collected at the Roosevelt Island station, L with correction factors applied to adjust for the generally lower concentrations at that site. This correction was made using an approach developed by Schimmel. .To further conform to Schimmel's
, preliminary data manipulations, the mortality and pollution data were initially " indexed" by dividing each daily value by a cen- ~
s tered 365-day moving average. The indexed variables have means
.f- of approximately one and lack long-term (cycles greater than one year) trends.
Sensitivity ig Alternative Filters l- ,
d' This anlysis consisted of estimating regression coefficients for COH and SO2 after all variables were preprocessed with one of several filters (see Table II.7); regression results with no filtering are pres ent ed firs t. The first set of three filters 4
consisted of taking residuals from 7, 15- or 21-day moving averages of the data. These three filters remove primarily low-
.g -frequency components from the data. Shorter moving averages cut more severely into the high-frequency components than do longer
- ones. (Inthe limit, taking residuals from a "one day" moving average r e m o v e s All v a r ia t io n from the d a t a.) The regression coefficients were found to increase with the length of the moving
- average, suggesting that the mortality / air pollution association i , increases as a w id e r band of high frequencies are allowed to
! remain in the da ta (s ee Appendix II.3). The next three filters were "id e a l" in the sense that they perform ed precise frequency cu ts. They removed all cycles in the data which f ell beyond the indicated period (1/ f requency) lengths, measured in days. For example, the ideal 2-4 filter removed all cycles with periods of 0 greater than 4 days. It should be noted that the shortest period -
that can be evaluated in data measured daily is two days. The three ideal filters focused on high-f requency bands of varying I widths. Qualitatively, this was the same as the three moving
!. averages, but we gained a clearer sense of which frequencies were removed from the data. The results were also in qualitative 3 agreement with pollution coefficients increasing as band width increases.
~
Overall, the regression coefficients for COH ranged from 1.2 to 5.4 daily deaths p er unit COH. The lower value was ob tained when only periods of 2 to 4 days were evaluated. The higher
( coefficient resulted when no filtering was applied. For'the six filters which isolate various high-frequency bands, the results ranged from 1.2 to 2.0 daily deaths per unit COH. Since these six filters represent a range of reasonable approaches to removing low-frequency confounding from the analysis, the range of coefficients derived from them is also expected to provide the s relevant sens itivity measure.
l l
\
\
37 i
- .. . . . . . - , . . _ -__ ; ,. ... {
t .. , _ . . _ _ - . . . . . _ _ _ .. ._-. i _. _ _ . _ . . _ _ __-..
. ..m__ u. -
- .4*
.r
-f Table II.6
! i$.
Means and Standard Deviations (SD) -
of Princiole Variables in Analysis O
s
'i .
Mean S.D.
., Mortality * -182 25.9 CCH 1.96 0.965-SO2 (PPM) 0.103 0.098 7 Temperature (CF) 54.5 17.5
/
- The sum of respiratory, heart disease, other circulatory and cancer mortality. See Schimmel 1978, p. 1073 for corresponding ICD codes. s t
'u
.h e
9 e'
"h S-'
N s
38
.. -~- .a Table II.7
, . Regressions Using Different Filters
[All regressions include 502 same-day tempera-ture, and day-of-week dummy variables. Coefficients represent number of daily deaths per unit COH.
(Mean CCH = 1.96)) _
s' s-
+
Statistical Filter
- Coefficient Error nene 5.40 0.46 7-day MA 1.40 0.46 15-day MA 1.76 0.46
+
21-day MA 1.97 0.46 Ideal 2-4 1.20 0.71 Ideal 2-7 1.21 0.56 Ideal 2-14 1.38 0.49
- see text for description of filters used.
MA stands for moving average.
~
l .C' i
O l
k' .
l l
me D
39
_. __ __ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __ _.. . . _ _ . - _ . _ . _ _ _ _ . . , _ . _ .? _ _ _ _ _.
...A J
L '
- I 3
Ana ly s i s o f Egy. . lgd, h Tim e-S e r ie s *
.; Results E1131 Different Tennerature Corrections In this set of analyses, variables representing various ?
-: functions of weather were introduced into the regressions. For
, : practical reasons, the weather vector was approximated by average daily temperature. Throughout the analyses, variables were initially filtered by taking residuals from their 15-day moving _
! average. The COH-regression coefficients presented in Table II.8 ',
- are f rom models with no temperature variables, with same-day -
temperature, with same-day temperature by month, with Schimmel's .
nine-temperature variables, and with Schissel's nine-temperature l variables by month (the latter results were taken from Schimmel, 1978). The regression coefficients decrease from 3.1 to 1.3 ,
3 (daily deaths per unit COH) as more elaborate temperature vari- ,
, ables are introduced. Such behavior is not too surprising since. "
i filtered C0H and filtered temperature are moderately correlated t
(r = .34). Current understanding of the independent physiologic effects of temperature on mortality, which might guide the choice i L of appropriate temperature specification, is unfortunately l , limited. However, it is worth noting that, for the range of -
3 temperature specification f rom "same day" to "nine by month," the f
COH coefficient ranges only f rom 1.8 to 1.3. Thus, the uncer-tainty in results due to this factor is not large.
Summary ggi Discussion q
For a reasonable range of preliminary data filters meant to -
conerti for low-f requency (seasonal) confounding, estimated coef-
! ficients from a model relating daily mortality to COH range from 1.2 to 2.0 deaths per day per unit COH. Similarly, a reasonable
, range of variations 'in the specification of temperature resulted
- in coefficients ranging f rom 1.3 to 1.8 deaths per unit COH. 9
, These results can be re-expressed as risk coefficients by -
dividing by the mean New York City population for the period I
- 1963-1976, es timated to be 9.8 million persons. This operation j :j - yields a range of risk coefficients from 0.01 to 0.02 deaths per
( day per unit COH per 100,000 persons. If the statistical uncer-tainties as quantified by the standard errors are added, then the l ;
i range of risk coefficients we obtained can be said to extend from -
!. just below sero to about 0.03 deaths per day per~ unit COH per j' 100,000 persons.
l -
, For purposes of comparison with mortality risk coefficients
- derived f rom cro s s-s ectional s tudie s , these daily risks es 2 be ,,
! converted into annual risks. If we assume daily probabilities are l'
l independent, we can then use the binomial expression to estimate an annual risk coefficient from the daily probabilities. However, for low-probability events, such as those we are considering, this is equivalent to simply multiplying the daily risk by 365.
- - Annual risk due 'to pollution at its mean is ob tained by multi- .'
I plying by the mean COH level f or the period of s tudy - ( i . e . , 1. 9 6 ). !
\
- l. It should be noted that the coefficient range given above is , ;
I probably too narrow f or characterizing our uncertainty of the n
a ,
40 l i
s
~.
.w -
L:
i Table II.8 1
Regressions Using Different Temperature Variables (All' variables.were first filtered using 15-day MA deviation filter. ' Regressions include 502 variable and day-of-week dunumer variables. Coefficients represent number of daily deaths per unit COH. (Mean CCH = 1.96) ] .
A Temperature Statistical variables coefficient Error .
none 3.09 0.44 ,
same day 1.76 0.45 same day, by month 1.88 0.44 Schinsnel's nine* 1.45 0.46 Schinunels ' nine , 1.26 0.42
'by month **
- Including: 1) Same-day tenperature
- 2) Previous-day temperature
- 3) ' Average of second and third lags
- 4) Average of fourth to sixth lags
- 5) Average of seventh to thirteenth lags
- 6) The square of the positive deviation of same-day temperature from its expected temperature C- (where expected tengerature = least squares '
fit of sine wave to 14 years of data)
- 7) Same as 6 for negative deviations
- 8) The square of the positive deviation .of the average of same-day"and previous-day from their expected value
,C - 9) Same as a for negative deviations
- Taken from Schinsnel,1978. No day-of-week dummy variables.
.4 o
D
.... a
~
41
7
. relationship between daily mortality and COR in New York City.
Several sources of potential error were not explicitly considered in our analysis, including: (1) population exposure misclassifi-cation resulting from utilisation of pollution data from one .
g fixed ambient monitoring site; (2) the imperfect relationship between the exposure metric, COE, and a more biologically optimal metric such as respirable particle mass concentration; and, (3) too little diversity in the range of models we fitted. Regarding
~
the latter point, we chose to analyse a range of models that, in our opinion, spanned the approximate extent of current modeling '
uncertainties. .
There are also more subtle reasons why our range of coeffi-cients does not include the full range of results reported in '
earlier analyses of this data set (see Schimmel and Greenburg, 1972; Schimmel and Muravski, 1976; and the 1982 HEED, Appendix 3, 7 pg. 89). One reason is that we haven't performed separate analyses of subsets of the data set (e.g., jackknif e analysis).
Early analyses of the New York City data were, by necessity,
^
performed on various early segments of the full 14 year data set i for which air pollution ef f ects tended to be larger than those in later periods. Further, a simplifying feature of our sensitivity 7
. analysis wa s the use of a design in which model variations of one
, type (f or example, filtering) were perf ormed while keeping the l .other model variable fixed. This reduced computation complexity and simplified presentation of results, but also understated the I
variability that would ha_ve b een ob tained had a f ull ma trix of ,.
model variations been tested. Lastly, we made no attempt to J
-evaluate the delay (or lag) structure of air po llu t io n-m o r ta lit y
- 4ssociations. We expect this to lead to an underestimate of
- total pollution effects. Schimmel (1978) found, for example, i .that total air pollution ef f ec ts, when summed over coef ficient i estimates for same-day and up to four-day lags of pollution were about 60 percent higher than ef f ects computed f rom a simple same- - -
day pollution coefficient.
- Finally, the results quoted herein apply direc tly only f or the six of sources and time pattern of concentrations observed in New York City between 1963 and 1976. The general applicability of these results will not be certain until validation studies in .
several other cities with dif f erent pollution characteristics and weather patterns are undertaken.
A I.3 g.1 g.f, Ih.3, T i m e- S e r i e s M e t h o d o l o n y Usina S imu l a t ion 2.1. LS.g. Ann a l e s p.gn Various investigators have conducted studies combining fil-tering and regression to investigate the relationship between daily mortality and air pollution (e.g., Schimmel, 1978; Mazundar. Schimmel and Eiggins, 1982). In the previous section, we investigated the sensitivity of regression results to filter variations and alternative regression specifications using New '
York City da ta. The purpose of this section was to inves tigate the behavior of analysis in cases where the "t ru e" results were known in advance. Based upon actual mortality, air pollution and
- 42
a\
l weather data from Los Angeles County, simulated mortality series * '
with pollution effects (as similar as possible to the actual i mortality) were generated. These pollution ef f ects were then "back calculated" using typical time-series methods (see l Schimmel, 1978). l The data set for the simulation study was based upon data from Los Angeles County for the period 1973 to 1977 and ,
including:
(1) Total daily mortality. l (2) Daily average particulate level in KM units from one
, station. EM is a reflectance measure of the particles' blackness. It is proportional to the quantity of elemental carbon in the atmosphere (Cass, 1983).
(3) Daily maximum temperature, minimum temperature and relative humidity.
(4) Indicator variables for day of week.
Simulated nortality data were generated (see Appendix II.3) based on assumed dose-response relationships f or air pollution effects. The first case simulated was that of no pollution effects. .In 500 simulations, the estimated linear regression 4
coefficient was not significantly different from aero, suggesting that the procedure gives a reasonably unbiased estimate of 5, the I
constant of proportionality relating KM to daily deaths. Next, various simulations were run with linear pollution effects.
I Again, the estimated results were reasonably unbiased with approximately the same variance as in the no-effects case. Using I
this inf ormation, we computed the probability of rejecting the hypothesis of no pollution effects f or dif f erent value s o f B. ~
Results are given in Table II.9.
Table M Probability, P, of rejecting the
- hypothesis of no pollution effects given various "true" effects B i (deaths per day per unit KM).
B P O.0 .05 l 0.5 .33 0.7 .51 1.4 .95 1.6 .99 These results may be more easily interpreted if we consider
" effects at the mean," i.e., the product of 8 and the mean pollu-j t ion lev e l, KM = 2.9. For example, an air pollution ef f ect that 43
.r.,,., .,_,,,_,-.----.,,.,n.,,,___,,,,...,,,___,_,_.n_.,, . , . . , , _ . - . , , . , . . . , . , ,,n__,_- m,..,.._.,,
. _ _ _ _ _ _ ~_ ~ . .. ._
. J 1Z CJ'_;.T '
'i . .
O would produce (an average) 0.7 I 2.9 = 2 deaths per day in Los Angeles could be detected 50 percent of the time. To be detected
- 95 percent of the time, the mean ef f ec t would have to be twice as large, or 4 deaths per day. O The 1975 population of Los Angeles County is estimated to have been 7.26 x 10 6 p e r s o n s.* Dividing through by this figure, the numbers in the above table can be converted to risk esti- '
mates. Thus, our simulation results indicate that, in a series of length such as 1826 days with a mean pollution level of 2.9 C
, KM, the lowest slope of a proportional dose-response curve .
could be detected 95 percent of the time would be 1.9 x whic) 10-deaths / person / day per KM.
The results given above provide information about the abil-icy of typical time-series regression methods to detect air i pollution / mortality effects. We don't consider these results to be directly applicable to the New York City analysis, because of sev ral unique features of the simula tion analys is. We note, however, that the risk coefficient that could be detected 95 per9ent of the time, w 10" x 2.9 KM = 5.5 x 10 hendeaths expressed at the meanis exposure (1.9tox ..
/ person / day), about three V' five times larger than those we estimated for New York. The exact interpretation of this result must await further analyses aimed at both replicating the simulation using New York City data and estimating ef f ects in the real Los Angeles data set. How-ever, if these results prove to be comparable, then it would -
appear that the air pollution health effects indicated by the New York time series are below or near the level at which mortality
-ef f ec t's might typically be detectable by such techniques.
.y .
ee.
.)
- 0btained by 11aear incorpglati n 1970 a d 1980 census ,
figures (7.04 x 10 6 and 7.48 x 10 , respectively).
e 44
[e-II.4 gg Analysis gg, Cross-Sectional Mortality 23,g,3, Inc o rn o r a t inn F, iga gad, Inha lab l e P ar t ic l e, gggg -
Indicas Introduction
- q, In prior cross-sectional analyses of the health effects of particle mas s and sulf ate air pollution, it has been f ound that those pollutants are contributors to mortality, even after socio-economic control variables are considered (e.g., see analyses by '
Evans et al, 1982). Such past work has employed Total Suspended J- Particulate Matter (TSF) and Total Suspended Particulate Sulfates (TSP 50 4 "), as these were the particle measures available for the 1960 and 1969 total mortality data sets considered. However, it is now thought that it is primarily the Inhalable Parti'c les (IP)
(d. < 15um) and (especially) Fine Particles (FF) (d < 2.5 u m ) ,
that may af f ect mortality. It is this subset of TS$ which can enter the body and is expected to have the greatest human health implications.
It is the purpose of this work to test the hypothesis that the air pollution health ef f ects estimate derived f rom cross-p sectional analyses will be found to be both more reliable ( i.e. ,
regression coefficient more significant) and of larger magnitude (i.e., greater mean effect), if the exposure estimates are improved. This hypothesis is based upon the statistical fact that if there is error in a predictor variable (x) in a regres-sion, then the coefficient for that x aust be biased downward 4 <
(e.g., s ee Snedecor and Cochran,196 7). Since TSP measures are, at best, indirect indicators of true human exposures to l particles, the past use of TSP indices can be expected to have l introduced error (and bias) into the -mortality modeling process.
I We wish to now cost whether or not employing s ize-f rac tionat ed
! particle mass concentrations (which should more accurately repre-O sent variations in human particle exposures) enhances the statis- .
tical reliability and the magnitude of estimated particle air pollution health effects.
Method
.C In this work, the Lave and Seskin 1960 cross-sectional mortality data set (previously analyzed by this study (Evans et al,1982) is first reexamined in light of recent date regarding tha relationships between TSF, TSP SO4", FF and IF (Trijonis, l** 1983). The TSP and sulfate data reported for each of 98 Standard Metropolitan Statistical Areas (SMSA) are employed in formulae L developed by Trijonis (1983) primarily from the IP aerosol sampling network (Watson et al.,1981). These formulas were derived on a regional basis by Trijonis (see Figure II.2), and allow an estimation of FF and IF mass concentrations from historic TSP mass and sulfate data (see Ta b l e II.10 ) . These estimates of each SMSA's mean FP and IP concentration can there-C fore be readily substituted into the previously developed 98 SMSA 1960 cross-sectional sortality regressions. This allows a pre-11minary test of the hypothesis that improved particle exposure On 45 e
E l]
a 1!
. Table II.10 ,
,j ~ Regional Equations for the Estimation of Q and IP, P Mass from TSP and Sulfate Data
- S i
REGION EQUATIONS .
1 Cal - San Francisco FP = 1.1 (SOI) + .20 (TSP - 1.4 SOE) O!
IP = 1.2 (SCI) +
Cal - Central Valley FP = 1.1 (SO") + .18 (TSP - 1.4 SOI)
IP = 1.2 (SO") + .49 (TSP - 1.4 SOI) .g Cal - Los Angeles FP = 1.1 (SO") + .23 (TSP - 1.4 SO*) i IP = 1.2 (SOI) + .64 (TSP - 1.4 SO")
Pacific Northwest FF = 1.1 (SO") + .15 (TSP - 1.4 SO") 2 IP = 1.2 (S0") + .52 (TSP - 1.4 SO")
Arid Southwest FP = 1.1 (50") + .17 (TSP - 1.4 SO")
IP = 1.2 (SO") + .56 (TSP - 1.4 SO") ?
North Central FP = 1.1 (SO") + .14 (TSP - 1.4 SO")
IP = 1.2 (SO") + .48 (TSP - 1.4 50")
Northeast FP = 1.1 (SOT) + . 2 3 (TSP - 1. 4 SOT) .
IP = 1.2 (SO") + .60 (TSP - 1.4 507)
Southeast FP = 1.1 '(50") + .18 (TSP - 1.4 SOI)
IP = 1.2 (SO") + .50 (TSP - 1.4 SO") O
- Listed equations are those derived via regression from 1979-1981 inhalable particle data for each region (Trijonis,1983) . ;
t 4
O 46 ;-
-l,
-4 estimates should increase the statistical reliability and' magni-tude of the estimated air pollution mortality effects.
3
\ _
pe wie =m n P -
r:
- oarm A\ No V5 y ser ==s s .'"'1"
- n i I se coine e f es , sg*
y Figure II.2 Sites and Regions Utilized t_o, o Develop Regional FP and IP Estimation Formulae. (S denotes suburban locations, N denotes nonurban locations, and sites not marked by a letter are metropolitan locations.
Source: Trijonis (1983)
As a further tes t of the above-s ta ted hypo thesis , t he 1960 mortality analyses were repeated using 1980 Census and 1979 mortality data (the most recent available) for some 30 SMSAs for C. which direct IP and FP measurements were made during 1979-1980. ..
Although only a subset of the c.iginal 98 SMSAs can be con-sidered, this test does provide an independent and direct means by which to check the 1960 results.
Results
.C Tab le II.11 p res ents an intercomparison of mortality regres-sions in which various particle measures are separataly added to
- a base model (described in detail in the 1982 HEED on Airborne Particles) containing the following socio-economic control vari-ables f or each SMSAs percentage of population over 6 5, popula-C tion median age, percentage of non-white population, population density, percentage of college educated, smoking index, and per-contage of poor in the population. For the 1960 data, both the TSP mass and TSP sulf ate variables approached statistical sis-nificance at the 95 percent confidence level (Models 1 and 2).
However, when the Trijonis formulae were applied to the 98 SMSAs (Models 2 and 3), the resulting estimates of IP and FP yielded a noticeable improvement in the significance of the particle mes-sures in predicting mortality, and a 50 percent rise in the mortality effect at the mean. -
47
. . L .
- u t
. t W
Table II.11 Intercomparison of Cross-Sectional Total Mortality Regressions for Alternative Particle Pollution Measures yj 1960 Data Analysis 1980 Data Analysisi (n = 98 SMSAs) (n = 30 SMSAs) t Ratio of Pollutant !
Ratio of Pollutant Mean Effect (bi Yi) f!
3eg. Coefficient Mean Effect (biYi) to.the Mean ,
Particle Pollution (bi) [ deaths / year Mean Value to the Mean Effect Effect Calculated Model variables Added 5 per 10 ple Coefficient of variable Calcul ted for for Sulfates ;.
No. (i) to the Base Model* Per pg/m ] p-value (Ki )[pg/m 8] Sulfates /p-value of bi **
g 1 Sulfates 2.6 .064 10 1.0 1.0 / p = .001 ;
2 TSP O.31 .052 121 1.4 0.05 / p = .88 3 IP 0.58** .029 70 1.5 0.8 / p = .06 4 FP 1.3** .029 31 1.5 1.3 / p = .002 t
, CBase Model includes an intercept, percent of population 2 65, median age, percent college educated, smoking index,
,and percent poor.in each SMSA (for details, see Appendix 3 of 1982 HEED) .
. TSP = Total Suspended Particulate Matter.
IP = Inhalable Particle Mass (da < 15 pm) .
~
f' FP = Fine Particle Mass (da < 2.5 pm) . ,i
tSee Appendix IIfor complete presentation of 30 SMSA Analysis of the 1980 data set. !
' i ,1 e
9 e
- t
' y o _ _
> a- a o '1
Table II.11 also contains summary information from the 1980 (30 SMSA) analysis.* It is interesting to note that the TSP
-0 coefficient is no longer a significant predictor of mortality.
that IP approaches significance, and that sulfate and FP remain highly significant. In terms of the mean mortality effect implied by the particle coefficient, it can be seen that FP retains its greater magnitude of effect (relative to sulfates),
while IP and (especially) TSP do not.
,C, Conclusions Based upon the 1960 and 1980 analyses, it is ' concluded that the use of a FP mass measure in c ro.s s -s e c t io n a l nortality analyses does indeed result in a more significant and larger magnitude health effect than previous measures employed in such analyses. This agrees with both physiological and statistical expectations.
The TSP measure would have appeared to have been a reason-able choice of exposure variable based upon the 1960 analysis alone, but the 1980 data indicated that this relationship was not consistent. For 1960, TSP was found to yield a larger and (slightly) more statistically significant effect than sulfates, but in the 1980 mortality regression TSP was not at all ,
significant. gased upon these conflicting results, the use of TSP for the estimation of particle health effects is not recom-
, sended.
The commonly employed TSP sulfste measure was found to be a consistently significant predictor in both the 1960 and 1980 analyses, indicating it to be a useful measure in the absence of FP asasurements. This is not surprising in that the conditions conducive to the formation of sulfates are also conducive to the .
formation of other fine particle constituents, such as secondary carbonaceous asterial and nitrate aerosols. Thus, sulfates are often highly correlated with fine particle mass concentrations.
However, it should be no ted that the FF mean ef f ect ranged f rom 30 to 50 percent higher than the sulfate mean effect. This p indicates that the use of sulfate as a surrogate for all consti-tuents of the FP sass, while useful, introduces additional error into the es timation of exposure, and thus may cause underestima-tion of actual nortality health ef f ects of particle air pollu-tion. As a result, it is recommended that, until a 98 SMSA 1980 analysis is completed, in TableII.11
.c (1.3 deaths / year per 10phepeople FP co eper f ficieng) ug/m s hown be employed in the assessment sulfate coefficient of the health (2.6 ef deaf ects of air t hs / yea r p pollugion, withper e r 10 people theug/m 19g0) being employed when estimates of FP impacts are unavailable.
a
- Complete details of the 30 SMSA methodology and results can be found in Appendix II.
b
.6
. . _ . - - _ _ - . .-__ - -. -_- -_ _ - ~ - _ . . _ _ _ - ----
.._____. 2 . C ' . ._~. n. .___
_ . a 1_ _ . _ .. -
4 Oc When applying the coef f icients noted above, consideration must be gives to- the composition of the aerosol sisture -involved.
For emaaple, if a sulf ate coef ficient were to be applied to a case in which sulf ates are present in substantially lesser or 7 greater than usual proportions (e.g., relative to organic l particles or trace metals), them the results would be misleadias.
, It can be espected that the use of the entire fine particle mass i should be less sensitive to errors introduced by compositional l variation from case to case. Also, it should be kept in mind l
that the FF coef ficient is mos t representative f or an " average" 3l
, urbaa aerosol composition and will, to some entent, he subject to .
j the biases noted for sulf ates when applied to aerosols having a I l ; askeup very dif ference from the mean composition. It may be that i this probles can be addressed through the development of coef ficients for each of the numerous aerosol composeats (e.g.,
i auto particles, soil particles, oil combustion particles, e t c.). 9 l Rowever, until aerosol component-specific coefficients are developed, the use of fine particle coef ficient (rather than a
- TSP or sulfate coef ficient) appears to provide the more accept-able alternative for risk analysis at this time.
1
[ Although the use of a fine particle mortality coef ficient O
! should provide sa improvement over previously used cross-i sectional indices of particle air pollution, we must emphasise l .the large uncertainties surrounding any such damage coefficient.
Indeed, despite the f act that the coef ficient is s ta tis tically
- greater than sero, uncertalaties not considered by such analyses
! (e.g., error s in the measurement of the esposure variable) make 1 i it possible that the sortality risk might la fact be sero. Such l coefficients have, in the pas t, been applied without adequate j attention to their actual applicability to the situation and the j uncertainties involved. We refer readers to the coefficient j limitations noted in Section IV and the Summary of the 1982 NEgD.
Improper application of cross-sectional coef ficients may lead to .
V erroneous conclusions regarding health risks.
- Frobably the most important conclusion to be drawn from this j analysis is that making refinements in the estimation of human
- exposures to air pollution does indeed enhance our ability to l guantify the health effects of these exposures. The use of FF is ,
- an improvement in that it represents the breathable mass f rac-l tion, which is also less subject to errors introduced by the i siting of specific monitors (because it is less specially
- variable). It is important to note, however, that further i improvements can and should be made in pollution esposure esti-mates. These include the estimation of variations in personal esposure (due to influences such as indoor air pollution), as
, well as in individual dosages, given similar esposures. It seems i clear that, as the estimates of esposure and dosage are refined,
, our ability to detect and to be confident in our estimates of the hussa health ef f ects of air pollution will improve.
^
! 50
C i III. Toxie Effects gi Airborne Particles .
I The 1982 REED on airborne particles summarized the litera-ture on animal and human studies relevant to the component-
,'O specific toxicities of airborne particles. Toxicity data on metals, sulfates, nitrates, natural dusts, diesel particles and B(a)P were reviewed. Our second year toxicity evaluations have concentrated on the evaluation and ranking of health risks based upon in vitro a nd in 111g b io a s s a y s . In particular, mutagen- -
icity, carcinogenicity, and animal bioas says on non-neoplastic
- i. particle toxicity ef f ects were analyzed. In addition to these assessments (those derived f rom bioassay and relative potency evaluations), we have extended our review of the toxicity of airborne acid sulfates and nitrates, which in our 1982 HEED were
. identified to be among the most significant factors in the inter-pretation of observational and epidemiologie data on population exposures to ambient particles.
In this section, carcinogenic and noncarcinogenic ef f ec ts and risks associated with particle exposures will be described.
III.1 Carcinosenie Effects gi Particulate Matter:
Review gi Evidence LLgg Bioassav Exneriments The main objective of our work in this area was the estima-tion and comparison of the carcinogenic potencies of various particle types (dies el, coal fly ash, wood s tove, oil, ga soline c engine emis s ions , e tc.). Another important goal of this compo-neat of our research was the identification of the sources of uncertainty involved and the estimation of their magnitude. As discussed further in the Appendices to this and the previous HEED on airborne particles, we have evaluated three broad categories of studies relevant to accomplishing this task. They are:
(1) Epidemiologic/ occupational s tudies (see Appendix II).
(2) Animal studies (see 1982 REED Appendix 2).
(3) Short-term bionssays (see Appendix III.1 and 1982 HEED g Appendix 2).
Strengths and weaknesses, with respect to the quantification and ranking of part icle potencias, were identified in each of
- three categories. In brief, they are as follows:
.C (1) Epidemiologic studies a) high degree of relevance; but, b) studies often lack estimates on many complex particles or components of airborne particles.
(2) Animal studies
,2 a) extrapolations between species are difficult and controversia1s b) again, definitive data lacking.
g &
51
. -- . _. _. _ _ _ _ _ _ _ ._r
' ' f ~~
~
~L .. ~::: Y
.6 .J i- '(3)-'short-term bioassays a) techniques for extrapolation are 'not yet estab-lished; however,
.h b) considerable data bases to support risk analytic .
studies are available. 0 3
Subsequent to our initial scoping efforts, we decided to concentrate on-bioassays, mainly because of the availability of
, data allowing for correlation _between in vitro and human carcino- _,
i' genicity'results. Looking further at our options, we focused on Ames bioassay data since this data base was the most extensive 7' e and among the most standardised. .
Activities in this area have been divided into three phases.
In' the firs t phase (contained in the 1982 HggD), the biological ,
validity and quantitative shortcomings of short-term bioas says ',
were reviewed. In light of the inadequacy of traditional surro- -
gates for carcinogenicity such as the B(a)P content of complex mixtures, a proposal was developed for utilizing potencies from short-term bioassays directly in our risk assessment of. airborne particles.
In the second phase, a number of projects have been under-taken in an effort to provide a human risk interpretation of the
, quantitative results of several b io as say systems. The progress of these is summarized below. (Further details may be f ound in Ap p end ix III.1)
- s In the third and final stage of the project, published
bioassay results pertinent to urban aerosols and combustion emis-sions are presented and interpreted using the results of the
-. previous phases.
Toxicity Ranking gi Particles h1 Usina Potencias y Estimated iggg Short-Tern Bionssays -
, Rationale apf Method I One approach which is useful in ongoing assessments of carcinogenic risk involves comparisons through a ranking scheme.
These rankings can be formulatd by weighting various types of -
evidence which relate to human carcinogenic potential. In the past, emphasis has focused on ranking compounds as to the likeli-hood of their being carcinogens. More recently, attention has ,
shifted toward attempts to go beyond mere determinations of 4
carcinogenic potential. Such efforts seek to estimate the rela- ,
tive carcinogenic potencies of these substances.
, Many inves tigations have approached this problem through comparisons of activities in different short-term bionssays and animal model systems (Albert et al.,1982;DuMouchel and Harris, 1983; Cuddihy et ' a l., 1981). For ~ complex mixtures, the relative concentrations of surrogste substances which exhibit carcinogenic acitivity, such as B(a)P, have also been used (Cuddihy et al., .
h 52 -.
.. L g ,, 1d . -
._. _.. ..._'h:T"'2 .-
1980). The limita tion s of this latter approach are discussed in more detail in the 1982 REED.
These comparisons can provide a rank ordering of pctancies within certain well-defined tests or between disparate tests. A consistent response within a system which is believed to closely s in ic . t h e in yiy g, s i t u a t io n in humans would provide the best evidence of human cancer risk. Although it seems reasonable to assume that short-term assays using whole animal systems (skin -
painting) or eukaryotic cell lines are more representative of j' humans than prokaryotic bacterial systems, the relationships between these sys tems are not clear. Thus, it does not appear advisable at this time to completely discount the results from any of these tests when attempting to rank the potencies of suspect carcinogens. Consistent responses, or rankings, across disparate sys tems should provide the analys t with an increased degree of confidence in any proposed ranks. For pollutants tested in only a f ew or even one tes t sys tem, confidence in the rank potencies cannot be great.
1 As a firs t' step, potency rankings provide comparative infor-nation while avoiding a direct quantitative calculation of the magnitudes of risk involved. Of course, quantitative estimates can also be used as the basis for a ranking scheme. However, the '
i numbers generated of ten assume a greater importance and degree of validity than is jus tified or intended. Still, rank es timates may prove useful in determining the value or urgency of research or control efforts among different pollution sources.(Holmberg, 1983). They may also be used to follow the ef f ec t s upon g eno-toxic emis s ions subsequent to various alterations in combustion l conditions or pollution control technology. Yet even here, an
! implicit assumption must be made that the respective tests are in i fact relevant as predictors of human health risk. Therefore, the C presentation and use of such data requires caution. Many factors .
. aust be considered, including: fundamental assumptions and non-quantifiable sources of uncertainty, which are of ten uns tated; i differences in the extent and magnitudes of exposure; compar-
[ ability of data; and, issues which are less scientific in nature ;
relating to questions of equity and dif f erential assumption of j f; risks or hypothetical risks.
l Sources gj, Uncertainty 12.d. P R.E.8. ].A.g.R. C o n s id e r a t io n s Comparisons of both a qualitative and quantitative nature are best made when the various samples are trea ted within the C same experiment (Claxton and Huisingh, 1979). Our attempts to extend a comparative potency analysis to additional particles has necessitated the use of results from many experiments and a number of separate investigations. Although consistency between and within laboratories can be demonstrated for a number of assays, many sources of uncertainty exist which have not yet been adequetely quantified. These include factors integral to the assays themselves,as well as other variations in sampling,
, extraction, etc. Variations in these factors is likely to be greater between laboratories than within, and for certain assays .
53 l
_ _ . . _ . _ . _ _ _ _ _. .m . _ _ _ _ _ .
i 0
as compared.to others which have a more standarized protocol.
3 ; Rowever, restricting an analysis to only those. studies where 4 multiple sources . have been compared within the same laboratory
- would drastically limit an already meager data set and, as impor-tantly, discount differences occurring within an emis s io n cate- ,"
gory. For this reason we have considered all data available, despite having been reported by a number of laboratories using varying analysis techniques.
3 -
Many'of the.following comparisons'are taken from data pre- .s seated ia Appendix III.1. Due to . inconsistencies in experimental methods and data presentation, extensive comparisons between dif f erent particle types has proven dif ficult. Throughout the i analysis certain assumptions have been necessitated. These are 1
explicitly stated when made. Data from specific research groups .
covering different particles and combined data from several .g
] groups are used.
I Diesel M Gasoline Enmine Emissions The data base on autobnobile-related mutageneity is fairly
+ extensive, with Ames dats available f or many makes, models and m operating conditions. While there are problems with direct com-parison of these data (see 1982 REED, Appendix II), the compar-
- - ability of the data on these sources is enhanced because of the fairly standardized protocols used.
~
I Combustion Products j
, Numerous dif ficulties exis t in the interpretation of data
. f rom coal and oil combustion. Compared to the case f or diesels, z few studies have been perf ormed on these sources, limiting the i useful data base. The data available is also confounded by a number of factors. Many of these have been reviewed by ~
y 3
G. Fischer (1983). Briefly, these problems relate to the fol-loving: the toxicity of the fly ash samples to the bacterial i
,: strains; inconsistencies in the mode and location of sampling, especially with respect to the location of the samples in the effluent stream; and variations in the combustion conditions, j fuel sources and engineering design of the units tested. Finally, i a variety of solvents have been used to extract fly ash.
Many of these same difficulties apply to the other combus-tion emissions for which we have data. Thus, appropriate data is .
available from only a limited number of studies. For these '
- reasons, variation in mutagenic output between different combus- ,
tion units, or between laboratories, cannot be estimated.
Connarison g, Potency Rankings l
The potencies of particles emitted by sources are dependent upon the character and quantity of that em is s io n.
Exposures will -
l r be d e t ermin ed by the proportion of deposited particles which is available for biological uptake or surface interactions with given cell populations. This is roughly approximated by organic -
l
- 54 -
9
- **run e
, ,- - - . . _- - . .- - - - . . - ~_- .
r.g b solvent extracts, a likely upper-bound estimate. Moreover, the
- - quantity of deposited particles will be a function of particulate mass em is s io n s . Thus, _ carcinogenic risk pos ed by various par-ticles will be a function of the total exposure to active car-cinogens sorbed to or comprising the particles, and the specific
! ,3 activity of these compounds.
Yariations in any of these three parameters- potency, per- '
cent extractable mass and total particle mass--will alter overall q 'g risk comparisons. Thus, the choice of potency units (activity
- per mass of particles,per mass of extracts,per vehicle mile, e t c.) will have profound effects on comparative ranks. This i point is illus trated in Table III.1, where Ames ranking s shift i l ..
depending upon the units used. (Further illustration of this matter appears in Appendix III.1, Table 1.)
i Ultimately, comparisons must be made in units which are r appropriate f or a given category of emis sions. For automo tive l em is s io n s , the units of activity / mile traveled seen appropriate.
, Other technologies may be more appropriately compared using :
! activity per unit ,( energy utilized to achieve a given end '
result. Finally, iu.ercomparison b e tween different particle
- types requires a consistent unit of measurement. In this ca s e,
- activity / unit of. particle mass seems appropriate, and is also of obvious utility in extensions to ambient exposures. 'Investiga-
[l tors reporting biological activity should attempt to measure and present data allowing f or conversions between as many units as possible and include, at a minimum, information on particle
- emission rates for dgtermination of activity per particle mass.
I often this has not been the case, resulting in substantial quan-
! tities of relevant, but non-comparable, data. Because of this, i our quantitative attempts to compare particle potencies have been limited to the mos t commonly used units, ac tivity/ug extracted
- O organics. These estimates are therefore difficult to relate to -
l reported ambient concentrations of various particles, which are i' almost universally reported in terms of total particulate mass.
i; In the f ollowing, we summarise the comparative po tency data on i
emis s io n s from automobiles, woodstoves, residential heaters, and oil and coal combustion.
. .G l Automobile Emis sions
!. A number of trends were noted in comparisons between par-ticle types tested in a number of assays (Albert et al., 1982).
i Available data ind ica t e that cigarette smoke condensate consis-
- U tently ranks low in many assays (Table III.2). Assay results also suggest that fundamental differences exist between the 1 . , active components of automobile emissions and the par t ic le types i for which human epidemiologic data are available. These latter samples-rank high in the eukaryotic tes t sys tems with 59 m e ta-
. bolic activation in the Mouse Lymphaa (ML) as say, but rank con-2 'sistently lower in the Ames test and ML assay without activation.
i This clearly indicates that more of the activity in these samples
! is due to ac tiva table mutagens rather than direc t-ac ting com-
} pounds which contribute a large portion of the activity of diesel !
3
- 55 ,
4 1-__ __ _____ _ _ _ _ _ _ _ _ _ _ _
C Table III.1 .
. .c>
Ames Ranking, Standardized to Nissan Value*
19
! ~
Rev/ug Rev og/ Rev/ Rev/ug Rev mg/ Rev/ ,
Vehicle Extract Part. Mile Extract Part. Mile C
Caterpillar .03 . 11 --
.01 .05 --
Nissan 1.00 1.00 1.00 1.00 1.00 1.00 Oldsmobile .19 . 41 .65 .10 .21 .34 O volkswagon .33 . 69 .37 .25 .56 .31 Rabbit Gas Unleaded .14 . 76 .01 .10 .54 .009 Gas Leaded 1.04 2.48 .37 .80 1.93- .27
- Data from Albert et al. (1982)
L:
.)
it' J e i
f , .
M
.'4 .,
i
, e-
~
56 4
% F M P
- warwge w w- m ---e .,-e_, rs y y -r--,c-w--e--v-v,wy,---w --w,,,w--.-,wyn------y **yn-- m---y-f #-t- --r-*-w-w--,- .w-ym. ,--. v-wy -----ve+-* -
" ~' '
V ' s g .e . , ~ J-, f " ,j " ^
4 i
Table III.2 -
Ranking of Particulate Extract Potencies Standardized M Coke Owen Topside
- Human Lun9 Coke Owen Topside > Roofing Tar >> Cigarette Smoke Condensate Cancer (1) (.39) (.0024)
Mause Skin Coke Owen Main Topside > Nissan, Roofing Tar, Olds,W Rabbit, Mustang II >> CSC, Caterpillar Tumor (1.5) (1) (.2a) (.20) (.15) (.11) (.08) - (.0011) (neg.)
Initiation Mutation in Coke Owen Main, Roofing var > Coke Owen Topside > Nissan, Olds, W Rabbit, Mustang II, CSC > Caterpillar" L5178Y (2.2) (1.4) (1) (.24) (.11) (.06) (.09) (.06)
(.005)
Mouse Lya-phoma Cells (S9*)
( , Mutation in Nissan > Olds,W Rabbit > Coke Owen Topside > Roofing Tar, CSC, Coke Owen Main, Mustang II, Caterpillar
! 4 L5178Y (5.9) (1.7) (1.4) (1) (.55) (.55) (.54) (.54) (.35)
Mouse Lym-l phoma cells j (S9-)
! Ames (S9*) . Missan > Coke Owen Main > Mustang II, W Rabbit, Olds > Coke Oven Topside, Roofing Tar, CSC > Caterpillar TA98 (12) (6.1) (3.12) (2.76) (1.32) (1) (.78). (.52) (.05)
Ames (S97) Nissan > W Rabbit, Olds, Mustang II, Coke Oven topside > Caterpillar > Coke Owen Main >> Roofing Tar, CSC TA98 (16) (5.5) (3.1) (2.3) (1) (.54) (.16) (0) (0) *
- Data from Albert et al. (1982)
. ~
4
$ . Y
.?
. .- . . . . . .. .. ..L...._ - . . . . . _ :.22 .l
.j !
M r l
particles. Bowever,-it must be kept la mind that rankings saa be !
altered through sheise of potenay measures, mainly being laflu- !
eased by the persentage estrastable from the partistes (sempare ;
Table 111.2 and Table 2 in Appendia 111.1). (For example, a ej large differesse in relative potensies--84 times greater--is apparent with soke oven mais samples sempered with topside sam-
'l '
ples.) !
Table 111.1 and Table 3 in Appendia 111.1 present Ames -
rankiass standardised for comparative purposes to the Nissen M diesel engine values en 's allesse basis. Although the relative potensies of the diesel emission vary somewhat, they are gener- ,'ll ally withis a factor of 3. On the other hand, the gasoline i satalyst engine (Mustang 11) tested is ese to two orders of J magnitude lower. This sensisteney between asses s strengthens the
- l hypothesis that diesel engines constitute a greater humaa. health q hasard than gasoline satalyst vehistes. ;
t Vandatava 331 Ranidaatial g,Q Egg, gag,a !
The total mutagesis output of weedstoves for two fuel types !
(oak and pine) appear seasiderably greater than that of resides-tial heaters (Table 111.3). This data is la general agreement
( ;
with activities presented by viam e t al. (1982) f o r f o r e s t-f ir e l polluted air (see Appendia 111.1). Although the entrasted j organies from the two oil furnaces enkihit greater activity, the !
woodstoves emit seasiderably more organic mass. Therefore, the !
emission rates largely determine the potency rank when based upon y activity per unit of heat generated. i These findings saa be estended one step further through j somparison with asether data set presented by Clantos and Emisingh (1979). In this paper, the authors soapare residential heater emissieme with a sunber of particulate categories dis- .
A,
- aussed previously. Os sa matrast basis, the residential heater i
, exhibited approximately one half the activity of the gasoline !
satalyst of the Nuotang 11. These results are summarised la ;
Tab le 4, Ap pendia 111.1. Again, however, the data presented do j
mot allow for the empression of these results la terms of activ- !
ity per mass, asking potency comparisons is terms of overall ,
- I partiste output impossible.
S,ggi and 9,11 Canhuntian a l
Commercial'soal and oil sombustion fly ask have also been "I.
sompared, although only to a limited entest. Alfheim et al. t (1983) sompared the mutageais output f rom three oil-fired and four soal-fired burners (Table 3, Appendia 111.1). One of the sit burnere showed a mutageais activity of about 500 rev/mJ and samples f rom as FgC had an astivity of 58,000 rev/aJ. The activ- f ities in the other plaats were seasiderably lower, reasing from !
<SO - <40 rev/mJ ( Alf he im ,1983). Akiberg et al. (1983) tested *!
- emission f rom sa oil-fired plant and a seal plant and f ound the j mutageaisity of the emissions to be below the detestion limit of ;
the assay for most samples. Data ou the highest responses recor- '!
54 1
. . . . .. q
. .. ,. -=
i ::? -
j i' Table III.3 f'A M Potencima 51 Residential ItakML Unitta,b,e ,
i')- ,
I nov/Jouled: Woodstove >
Noodstove 33 Modified commercial 3 posidential -
- Cok Fuel Pine Fuel Furnace 011 Furnace
'Q l* (900) (149) (7.6) (1)
- Dev/ug Estract: Modified Cosmarcial 3 posidential 3 Woodstove 3 woodstove ;
.) Furnace Oil Furnace Pine Fuel Oak Fuel
, (2.5) (1) (.65) (.45) ag ,g , (g,gt) byg 94, DCM extracts caesults standardised to residental furnace results i
~
dgg+, 39" results = .15 rev/ug (cak). .29 rev/ug (pine )
,;. ?
Table !!!.4
- O Mutetenicity h Emerimental ADG4,b .
Nigh volumm Rev/mg Ash Bag Filter Samples High Volume Filterc Cascade Ispactor D TA 9e, 89- (mean2 standard (mean2 standard (mean t standard i deviation) deviation) deviation) ;
i (2.9 2 3.7) (3.7 2 2.9) 5 2 5.4 O I aSource: Clark and Nobbs (1940) !
, bDcM entracts, 4 sets of operating conditions, samples GA-E !
t coalculated from single doses dcalculated from mutagonicity of smallest sise fraction 6,
l 59 A- ___ _ _ ____ _ . _ . - . __ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ . . _ _ _ _ _ _ _ . . . _ . _ . _
_~'
l t O I
dad indicate little dif f eresse between these plaats (Table 6, 1 App end ia 111.1). gamples of Fluidised Red Combustion (yBC) and
! sesventional seal sembustion fly ask tested by Mumford and Lewtas
' (1932) suggest that FBC emissions may exhibit more Ames activity 9 (Table 7 Appendia I I I .1 ) . Iowever, conversion to units of
! rev/eaergy released is not possible for this paper. Results from Clark et al. (1930) are is agreement wit h this finding on a i ,
rev/entract basis, but suggest the opposite relationship for a rev/ particle mass (Table 3, Appendia 111.1). 3 l
i It is apparent from the studies noted above that sombustion ,
senditions are important determinants of mutagonis output. The i
FBC tested by Alfheim was operatias at a very inefficient levell Rubitschek and Wil11 ass (1930) found greatly increased mutageais 1
activities during start-up versus steady-state operations is another FBC (Table 9 Appendia 111.1). Therefore, the differesses 9 noted in results between this source and othere may be due to inef ficient combustion occurring in these smaller, esperimental plaats, and not due to any integral differences in the em is s io n s.
The comparative differesses la results between these investiga-tors are summarised in Table 111.4. The wide range is sampling i techniques, locations and solvents used seriously complicates O.
i attempts at poteacy ranking. Iowever, it appears saf e to con-4 ciude that inefficient combustion conditions will sisaificantly increase mutageaic outputs.
l F r e i ne t ia n gi 11331 Cann ar tillta .
. Utiliaian Fatamaina igga short-Tara staannava -
The following summary presents as attempt to deteraise quan-titative assessments of the comparative risks posed to human
- populations by various partistes. First, a discussion of a ll general potenay model is presented, followed by the estimation of .,
potenay soaversion factors and associated uncertalaties for the various assays considered.
l.
L5' Initially, Ames potencies for the various particles are i obtained f rom results reported is the literature. Us tas these j ,
figures and a proposed model based upon a relative potency .
i hypothesis, 95 percent confidense intervals are salculated f or ,"
increments is relative human sanser risk attributable to each emission category. The unsertainties involved are great and are not fully captured in our variance estimates. Therefore, these results should not be sossidered as reliable best estimates of the range of risks posed. However, their calculation serves a useful purpose la desesstrating the comptesity of such a process.
The results may also be used f or general soupar'isons between partiste types. Better estimates avait a more entensive data base in more bioassays, and a better understanding of the rela-tionships between bioassays and earsinogeais poteasy.
a 40
. . . . . . _ . _i _ ~
. Calanla11am 31 Intraaelation 13811E1 331 133331 131 Shart-Tara Rimaasava
!Q In 1979, the epa initiated a program to evaluate the popula-tion less esacer risk attributable to the increased use of light-duty diesel engines (U.S. IPA, 1979). Their novel approach involved the simultaneous application of several bioassays to I diesel emissions and three organic combustion products for which _
enough reliable epidemiologic evidence esists to establish espo-
,,0 sure response relationships. By assumias that the relative
, potency of a substanse is preserved over species, it is possible
- to use the epa data base to derive human potencies f or diesel emissions (Albert et a l., 19 8 2) . By iny11 cation, their method
allows the calculation of multiplicative factors relating b io-
- assay and human potencies. In this section, we utilise a model
.c proposed by DuMouchel and Earris (1933) to estimate entrapolation factors sad a measure of the uncertainty incurred by their use.
Matheda The procedure used is based upon the models shown in Figure 111.1. As DuMouchel and Barris pointed out, the cons tant reis-tive potency assumption Laplies that any dose-esposure response slope may be represented as the product of a particle-specific f actor and an assay-specific f actor. This model preserves the coastsat potency ratio for any two particles between assays and for.any two assays between particles. In addition, it provides a method for utilising the multiple assay results contained in the epa data base.
Ligggg 111.1 Model specification
,g
', (Source: DuMouchel and Harris, 1932)
Ti j = at 8 3 6g3 Cg3 4
Ygj = Dose / exposure response slope in the ith assay for the jth particle extract.
, ai = The constant assay factor, unknown.
.E(.
~
Sj = The constant particle factor , unknown.
di j = The random deviation 3 a from the model, in (di j) assumed N(0,0 ), g unknown.
c cij = The random experimental error, in (cij) assumed N(0,c[j),c[jknown.
eq o
- m
Imadequacies in relative potency assumption are modeled by a stochastic disturbance, represented by ogj terms in Figure III.1.-
.These are assumed to be interchangeable assay / particle combina-tions in the sense that the absolute deviations in different n cells are equally informative about the expected deviation la a -
new, untested cell. The precision of each slope is included esplicitly in the model in order to distinguish between the error made in assuming coastsat relative potencies and the precision of individual studies. -
In our 7 analysfs, we deviate from the model only in tge specification of a , the variance og3 Instead of a single a ,
we allow for as arbitrary partition or the cells into groups with a
O, common variance for 4 g3 Thus, we estimate several values.
k' DuMouchel and Barris used a variety of methods for estima-ting the critical parameters in their model. Iowever, we have adopted only the Maximus Likelihood (ML) methods. Computations are performed using the 3M algorithm for mixed randon and fixed effect models (Laird, 1932). A Restricted Maximus Likelihood estimator (RIML) approach (Earville, 1977) was used la estimating A
variance components to compensate f or bias in the ordinary ML estimators. -
Entimalad Extrasolatton Factars and Krrera In a preliminary analysis, we calibrated the model separ- j stely f or each assay against the human data. Almost all of the assays fail to accurately predict the humas data. The most common cause of error is an over-prediction of the CSC (Cigarette Smoke Condensate) coefficient. The skin tumor initiation assay stands out as the best predictor and the sister chromatid enchange as the worst (Figure 1 Appendia 111.1). ,
9 Since the evidence seemed to suggest differences in the degree of fit between assays, we estimated a different variance component for each assay. secause the human and tumor incidence -
data agreed so well and because of the sparse skin cancer data, w e p o o l e d a l l o f t h e in gj,1g t e s t s l a t h i s a n a l y s i s . The vari- 4 ance components, a a , may be thought of as measuring the degree
- to which each assah diverses from the coastsat relative potency model.
In Appendia I II.1, we also contrast the estimated variances '
(espressed as coefficients of variation) computed using all the
, data and multiple components with the slagte component estiastes.
(The latter estimates are measures of how well each assay agrees with the human data, whereas the former are measures of how well
, each assay agrees with the other assays.) Especially, we f ound
, that the relative potencies in the human data run counter to the trend found la the majority of assays (Figure 2 App end ia 111.1. J Furthermore, even precise knowledge of the common relative potency does not completely determine the human response. In fact, if this battery of assays adequately characterised the -
62 t
~ . - - -
~
.. . . . . - - . . . ..:--=- --
iO '
common relative potency, the actual human potency may dif fer by a
- f ac to r o f 10. (See also-Figure 3, Appendix III.6.)
Table III.5 provides extrapolation factors for each of the l'0 1 five assays. The coefficient of variation is an estimate of the error encountered in using the assay to predict an emis s ion's rulative potency.. For instance, for the Ames test, a confidence interval of w h i.~c h a standard-deviation vide ~on the log scale would include values of a f actor of 2 higher than the predicted -
relative potency.
j Ouantification .gf Risks Agj, Uncertainties-It should be stated at the onset that the calculations that i-follow include many uncertainties (most1'y exp e r im e n t a l), some of
,. which are not quantifiable at this time. The Ames potency values which have been determined for the various particle types are derived from results from many labs. This provides an empirical picture of the stability of these estimates as compared to esti-mates f rom a single lab or a limited number of vehicles. Thus, some inter-vehicle differences and inter-Jab variability are
.c- captured in these results. However, a similar approach has not proven f ea s ib le for the known carcinogens that are.used in com-parative potency determinations, since fewer of the samples have been testd.
~
We do not know if the fundamental relationships (variability in response as i t relates to predictive value) between assays are a function of the laboratories carrying out the assays. If this is the case, then scaling factors and vari-ance components of our comparative potency model would require modification. Thus, even our estimations of the variances cannot b e -' v ie w ed as conservative; a greater degree of variability would be likely, cf Table III.6 summarizes the Ames TA 98 potencies for several -
emis s ion sources, as determined from a number of published stud-i e s .- Due to limitations in the-data base, several estimates are based upon the results of only one or two investigators. In these cases, the highest Coefficient of Variation (CV) value, d e t e rmin ed from the diesel category (the largest data bas e), was
- ,,,' used as an approximate value. Given the overwhelming contribution e' of the model variances, this appears to be a reasonable deter-mination.
These values were used to es timate increments in relative risk of human lung cancer following the analytical method out-
- l. 0 lined previously. Results are summarized in Table III.7 and Figure 4 in Appendix I I I .1, where estimated risks and 95 percent confidence intervals surrounding them are presented. As can be seen, the uncertainty involved in these estimates is consider-able, spanning several orders of magnitude. Again, it must be noted that all sources of uncertainty have not been adequately
.C accounted for in this analysis, due primarily to the limitations and/or sparseness of the data base. It should also be emphasized that these estimates are based upon using only the Ames data and A
t may dif f er markedly if alternative bioassays are used in pre-4 4 Wm 63 T'W*----4 9 v er T-sy- ew* --ver_w w,-e-- -,imm-w w, e r - e ' = Fee w .-wve- - e e eerwe=y emi h*--wewem rww m
-e-ew'ew4r-wa--e**-'----*-=r--*- -7.we -eefa v e-sprse-'wrm--
r- -
m .
s I
l 1 i -i:
Table III.5
~f Bioassay Extrapolation Factors and Errors i-(
, Extrapolation Error ,
Coefficient Iog :
l Assay Units Factor. of Variation Variance !
3 Enhancement RR* per 10" pg/m extractable organics - years t of viral- s 6.6 2.4 1.5. '[i t foci /10 surv. cells /pg/ml f transformation Ii 3
Mutation in RR per 10" pg/m extractable organics - years 0.26 2.6 'l.6 7 lymphoma cells-TK mutants /10'surv. cells /pg/ml
. (+ MA)
- i 3
SCE in CHO RR per 10" pg/m extractable organics - years 15.3 1.2 0.60 .,
cells (+MA) SCE/ cell /pg/ml i ,
3 Ames, TA 98 RR per 10' pg/m 'extractable organics - years 0.29 1.1 0.56 Revertants/pg i
1 3
( Ames, TA 98 RR per 10" pg/m extractable organics - years 1.14 5.6 3.6
(-MA)
} Revertants/pg
! 3 Skin tumor RR per 10' pg/m extractable organics - years 3.4 5.9 3.7 -
initiation Papillomas / mouse at 1 mg l -
- lluman Relative Risk of Lt:ng Cancer ;
sl i-
..Il.
,, ,i . ,
(, , , l () is (J' J- .., 8 a .) U ' '
..--.,e.
- Q g Table III.6 Ames TA 98, S9" Potencies
{
] '.1
~
Mean Standard Deviation coefficient
[, (Rev/uq Extract) (Rev/Uq Extract) of Variation
. Light-dutya 12.45 12.47 1.002 l diesel Catalystb 11.1 4.4 .40 O spark engine Non-catalystC 8.5 (high estimate)C --
1.002 k spark engine .7 (low estimate)d --
1.002 k Woodstove' l.1 .28 1.002k.
-('
Residentialf .905 --
1.002 k heater Fluidized 9 10.15 --
1.002 k bed coal Conventionalh 2.04 -- . 1.002k coal 0111 6.13 --
1.002k
.C --
aResults from Hyde et al. (1982) , Pitts et al. (1982) , 'Pederson and Siak (1981) ,
Dukovich (1981) , and Clark et al. (1982) .
bLang (1981) , Pierson ' (1983) , Clark et al. (1982) , Claxton and Kohan (1981) ,
d' Dukovich (1981) .
Clang et al. (1981) .
,- dWang, Rappaport et al. (1978) .
- Lewtas et al. (1982) .
fClaxton and.Huisingh (1979).
%umford and Lewtas (1982) , Clark et al. (1981b) .
hClark et al.(1981), Mumford and Lawtas (1982),
i Alfheim et al.- (1983) .
'3 Acetone extract.
kLimited sangle size, coefficient of variation taken as maximum obtained in all cctegories.
~
65
' ~ -
~ . . - ~ ~ . . _ - . _ . - . - . _
. =,;.. .. . . . . ,
~-
7..
i j
i 3) i Table III.7 8
Increment ,in Relative .Bink .91 laulg cancer ;
' i. 3 m Uq/m extractable orcanics -vears* -g l i
)
l
^
1 95% Confidence Interal Estimate i
- i. Light-duty (5.4 x 10 .37) 1.4 x 10-3 -;
diesel ;
. Catalyst (6 'x 10~' .28) 1.3 x 10-3 '
spark engine .
(3.7 x 10-8
~
Non-catalyst hight .25) 9.7 x 10~" O' spark engine lows (3 x 10-7 -
.02) 7.9 x 10-s Woodstove (4.8 x 10-7 -
.033) 1.25 x 10~"
Residental (4 x 10-7 .027) 1.02 x 10~"
heater 9 FBC (4.4 x 10-8 .301) 1.15 x 10~3 Conventional (1.3 x 10~' .087) 3.3 x 10""
coal 011 (2.6 x 10~' .18) . 6.9 x 10~"
- Estimates based upon Ames data shown in Tabla III.6 e e e
aw d
6 66 4 N e e
M 50 i
dicting incremental relative risk cancer (e.g., f or light-duty diesel).
1
! Unfortunately, the values presented in Table III.7 do not
'O tell us much about the actual contributions to risk by each particle category. The extension of the analysis to comparisons of actual exposures experienced by the population as a whole is difficult. Rough interconversions between the units of activity ^
(extract activity and whole particle a c t iv it y) are possible, s based upon the approximate percentage of particle mass that is extractable for each category of particles. Likewise, estimates of exposures can be generated based upon ambient monitoring and source apportionment techniques. . Data on particulate exposures
, were collected for various sections of the country (c.f. Appendix
, I.1) in an attempt to tie such information to our b ioas s ay s results, thereby providing a regional es timate of excess risk attributable to each particle source type. Unfortunately, the available data were not in a f orm compatible with our bias say
.results, being expressed in terms of TSP or other measures of particle mass, and not including consistent apportionment for the particles for which we generated risk estimates. It is also
- p. obvious that interconversion of these units would introduce con-siderable additional uncertainty (extractable organic fraction is not a stable characteristic within a particle ~ category). Addi-tionally, the present exposure apportionments were also highly uncertain. Because of these new uncertainties and the already large degree of variability in the risk estimate, it was decided 4
1 not to combine these data sets until further refinements are achieved. The estimated risk coefficients do, however, suggest that with the current state of knowledge attention should be
_-given to sources which emit the greatest _ mass of extractable organics, especially those sorbed to particles in the inhalable size range. Aside from their greater potential for depositing in C the lung, the smaller size particles appear to contribute the -
largest quantity o f mu tag enic ac tivity, probably due to their larger surface area and, thus, sorbtive capacity.
Conclusions
.C The comparative potency approach holds much promise for delineating the range of potential human cancer risks resulting from exposures to complex particle mixtures. However, at this
, time, it appears premature to use data generated from short-term bioassays for anything more than general comparisons.
4' ' Bef ore applying these f actors, the risk analyst should be aware that the f ollowing assumptions have been made:
. all exposure response relationships are linear;
. interactive effects are ignored; c.
all coefficients are in terms of DCM (dichloro-methane) extractable organics, and DCM is assumed to extract all activity; use of these coef ficients in actual ambient 67
_ _ ._ _ . _ _ _. ..-.u_ ; .:
3 exposure-risk determinations is-difficult since data on such exposures are usually available only in terms of particle mass; and,
. experimental- and extrapolation errors .are randon and lognormal.
The constant relative potency model-must be recognized as an
. intentional oversimplication of the complex and inevitably non- -
linear relationship between_ currently available short-term tests -- 3 and human risk. However, if we assume that no additional human
- data becomes available,; future research will have to be directed
^
primarily toward improving bioas say technology. Approaches such as the' EPA study may be improved upon by expanding the potency set to include the much larger group of.known animal carcino- -
genics. gventually, basic research into the physiological n' mechanisms of carcinogenesis may provide a model that, in con-junction with bioassays, will provide useful risk as s essments.
Therefore,we believe that in the interim, the " correlation" studies are perhaps the best available technology.
9
.1 D
?
I i
- l.
- i -
l -- J.
i . 68 l
e . *+w
- m
__ _ . . . . . . . - - . . . - - - . . . . -- s- - - - - . - -
t 9
'Q
~;
III.2.l' Non-Neonlastic Toxicity 21 Particles ABA Their j Sinnificance 12.REBAR Health: Metals
'h -
Animal bionssays have been employed in' the study of chemical species-which are components of particles ( e .g . , metals, sul-
-fates, . hydrocarbons), specific emission sources (e.g., coal com-bustion' products, automobile exhaust), and certain. composite aerosols ( Arizona road dus t). - In this component of the study, we -
have . evaluated bioassays relevant to non-neoplastic lung disease.
.O We believe that this approach Lvill help in understanding the tox- i a
icity'of ambient particulate matter in terms of disease outcome,
- while identifying those components most likely to-cause lung damage.
There are many methods available to assess the toxicity of
,:> particles and their components. These tests range from ia vitro measures of pulmonary macrophage function (phagocytosis, viabil-c.'
icy, e t c.) 't o measure of lung function in whole animals and histopathological studies of lungs from exposed animals. We.have i focused on five main categories of bioassays. This choice was based on.the relevance of the assays to lung damage and the 4
- number and types of toxic agents studied using the different assays. The remainder of this section (as well as Appendix
.III.2) discusses these assaya in further detail. It also presents toxicity ranking s and related information on various metals (which are common cons tituents of airborne particulate matter) in ig y1Y.2. a nd in v i t r o systems. The bioassay data are
~
.then compared to human data on ambient exposures and occupatiogal
[ threshold limit values for the metals considered.
L Bioassays 121 Mucociliary Function Previous research regarding ig vitro macrophage bioassays ~~
- and inf ectivity models (see Appendix II, 1982 HEED) considered l the toxic effects of particles on the alveolar regions of the lungs. .These effects were relevant to inf lamma to ry changes, s u s c e p t ib ility to infection, fibrosis and emphysema. The response to particles deposited in the airways'must also be considered when estima ting health ef f ects f rom exposure to air 5- pollution particles. This is especially important for particles in the 5 to 10 um range, since a significant portion of these particles will be deposited in the tracheobronchial region. In
.. - this work, we have, therefore, evaluated bioassays that measure I
the effects of particles (or components of particles) on the q 'mucociliary layer of the conducting airways. These measures include measures of function (ciliary beat frequency, electrical conduction, particle clearance), histopathological changes (necrosis "of cilia, desquamation of epithelial cells), and
,~ biochemical indicators (quantity and quality of glycoprotein
[ secretion, intracellular ' levels).
Mucociliary bioas says have now been used to evaluate metals with respect to toxicity ranking. Interestingly, some of the toxicity ranks were the same order as those observed in ig vitro (e.g.,
macrophage and infectivity models Cd > Cu > Ni). This
! 69 i
I L- , ._ . . _ , - - _ _ _ . _ . . - . _ . . _ _ . _ . - . . . . _ . . . . . . _ . _ _ _ _ _ _ _ . . . _ _ _ _ _ _ . _ _ _ _ _ . _ _ _ _ _ _ _ ~ . _ _ . . _ _ _ . . _ _ . ~ ,
.- . . -. .-- ~~.-- - -
^
suggests'that the mechanism of damage may ba similar for the different cell types. Further indication that these effects may be mediated through common mechanisms can be derived from studies of antagonism and modifying factors. As was ob s erved with in O vitro s tudies of macrophage viability, there were antagonis tic interactions among different metals for mucociliary effects. For example, Cu suppressed the toxicity of Cd, and Ni suppressed the toxicity of Cu. A plausible mechanism for the antagonism is .
competition for binding sites and subsequent cellular uptake. ,
Thes e ob servations sugg es t that calculations of total toxicity based on individual metal contents could be overestima'tes when -
many different metals are present.
The effects of metals on the mucociliary layer were usually .
reversible. A noticeable enhancement of recovery, when allowed, ,
occurred in vivo versus .i_g vitro, suggesting an influence of ~
other cell types and humoral factors on the repair process. Rapid clearance rates of particles from the tracheobronchial region were probably also involved in the reversal of the toxic effects.
These studies suggest that acute ef f ects may be more important than chronic ef f ects with respect to mucociliary injury. How- ~
ever, it must be realized that the longest exposure was two '
weeks, and chronic studies are thus necessary. Furthermore, observations also suggest that the airway epithelium may be very sensitive to the ef f ects of toxic agents. Since many of these effects are relatively severe (albeit r ev er s ib le) , they must be considered as important health effects when evaluating the tox-icity of particulates. Further details on the bioassays for mucociliary function are presented in Appendix III.2 to the HEED.
A description of other assays considered can be found in Appendix 2 of the 1982 HEED.)
Ranking Ambient Particulates y Having identified relevant bioassay systems, we have attempted to establish a hierarchy of toxicity for various compo-nents of air pollution through a toxicologic ranking pro tocol.
The following section describes some of the assumptions and methods used and the principal findings to da t e (more de tailed discussions are presented in Appendix III.2). -
Our preliminary linear regression studies of dose-response relationships on ig vitro macrophage tests provided the estimated .
concentrations required to reduce the effect level to 50 percent of the maximum observed value over background (EC50). We then '
established a relative toxicity ranking of the five metals from these data (s ee Appendix III.2). By combining these toxicity ranks with compositional data of air pollution particulates, we also attempted to assess the toxicity of defined urban aerosols (in terms of metal-based components) that appear most likely to cause harm. This aspect of our assessments is summarized below.
e-d.
-is.s _ _ - _ _ _ _ . _ _ _ _ _ . _ _ _ _ _ _ _ _ . _ . _ _ _ _ , _ _ _ _ _ _ _ . _ _ . _ _
,_ . a.
.~ . . - . . .
s ' .'
s Connarison 21 Bionssav Toxiciev LAYtla. 19. 1E581 Exnosures
'o It is difficult to extrapolate from concentrations of sam-pies t e s t ed in in ,Y,iggg, mac ro p hag e b io a s s ay s to levels that would be encountered by sacrophages in human lungs. In the whole lung, serum components and surfactant will cost particles, and that may affect their. toxicity. Furthermore. . the sacrophase populat' ion of -
the lungs is in a dynamic state; nacrophages killed by toxic
.g o
particles will be replaced by new macrophages recruited into the lungs. Thus, it is difficult to assess how a dose of aerosol
, received by macrophages in vitro compares to the same dose
(
'., received by macrophages within'the lungs.
While realizing these uncertainties, a c o mp a r is o n o f in y,1y,,g, s and ig. v i t r o levels of certain metals and the results are presented in Tabic III.8. The values used f or metals in human lungs (Schroeder, 1970) are nazimum ob s erved values, and thus represent a worst-case estimate. The concentrations of the var-ious metals in a human. lung are determined by assuming that 1g l of tissue occupies 1 al volume and the average weight of the l q lungs of an adult male is 953 g (Dieu and Lentner,1973). Thus, I
the maximum ug/ lung is divided by 953 to o b t a in us / m i "ig y,1yg".
- This as sumes that the metal is evenly dis tributed in the lung s, which is probably not true. Reterogenicity of fly ash particles l results in some macrophages receiving very high doses of certain metals (Hayes et al, 1980). Many chronic pulmonary diseases are g' initiated (or at least aggravated) by t. h e inhalacion of toxic particles, yet little is known about how particles deposit in lungs that are abnormal. 3 TableIII.8 shows that the average metal concentration in human lungs is 0.6 to 6 percent of the ig vitro gC50*****"t**~
tion. Since the human values chosen represent the maximum levels .
observed, it would appear from these calculations that the levels of these metals may not be high enough to cause direct and measurable macrophage effects. However, for non-cancer health l - effects, such as changes in susceptibility to infection and f
re'spiratory function, damage to the respiratory system may not be
, apparent below certain threshold exposure levels, but might increase sharply above those values.
Although Threshold Limit Values (TLV) should not be consi-l dered as absolute thresholds, they are believed to represent t exposure levels below which significant health ef f ects are not likely to occur. One can estimate the amount of Cd deposited in k
thg lungs of a normal 70 kg man, breathing approximately 1.1 x 10
' liters of air / day, assuming a 20 percent alveglar deposition.
If the average air concentration is 0.002 ug/ m , 0.0044 us Cd will be deposited per day, and in one year 1.61 us will be deposited. Once deposited, Cd is very ef ficiently retained in
. the body and very small quantities are excreted. If one assumes
' the biological half-time (clearance) for Cd in the lung is six months, 0.81 ug will b e pres ent at the end of one year. Using the calculations described this would mean an average cd concentration of 0.85 x 10"gbove, us/mi or 0.008 percent of in vitro l -
71
. ) ., . . . , _
7 i
9 Table III.8 Comparison of Levels of Metals M Vitro, In Vivo and in Ambient Air In Vitro In Vivo i-l l l (MAN)- Fraction l !
of ECse dose Average Maximum Maximum received by in communit in community TLV-Metal Bioassay pq/macrophage" pg/lungh- pg/macrophaged- macrophages air (pg/m')g air (pg/m )b (pg/m') 'j' Viability (scie)
' Cda+ 11.1 pg/ml 1.1 x 10-s 930 4.0 x 10-a 0.04 .002 0.35' 50 Vos 11.9 "
1.2 x 10-8 680 3.0 x 10-e 0.03 <.003-0.90 1.4 50 i:
Nia+ . 245 "
24.5 x 10-s 8000 3.5 x 10-7 0.01 .032 0.69 100 Cr8
- 285 "
28.5 x 10-s 2000 9.0 x 10-s 0.003 .015 0.35 500 3
Mna t 290 "
29.0 x 10-s 1700 7.0 x 10-e 0.002 .100 10 5000 i
Phagocytosis (ECsa)
Cdat 9.0 pg/ml 0.9 x 10-8 930 4.0 x 10-s 0.04 voi 1.5 "
0.2 x 10-8 680 3.0 x 10-' O.15 Ni 3+ 59.0 "
5.9 x 10-s- 8000 3.5 x.10-8 0.006 g Cr# 15.8 a 1.6 x 10-8 2000 9.0 x 10-e 0.06 Mn" 14.3 "
1.4 x 10-' 1700 7.4 x 10-8 0.05 Acid Phosphatase (ECse)
Cdat 23.6 pg/ml 2.4 x 10-' ~ 930 4.0 x 10-' O.02 3.0 x 10-e Voi 4.5 "
0.5 x 10-' 680 0.06 f Niat 224.0 "
22.4 x 10-s 8000 3.5 x 10
- 0.002 27.7 x 10-' 9.0 x 10
- Cr # 227.0 "
2000 0.003 Mar *
, 239.0 "'
23.9 x 10-8 1700 7.0 x 10
- 0.003 i i :
I-a Obtained by assuming 10% of metal is taken up macrophages after 20 hours2.314815e-4 days <br />0.00556 hours <br />3.306878e-5 weeks <br />7.61e-6 months <br /> in vitro. 'A b From Schroeder (1970). t c Obtained by dividing by 23 x lo , sthe average number of macrophages/ lung (Crapo et al.,1982).
d From Graham et al.(1975). *
I t .
e e e .
,k - ) k) '
i d! I d b) *
- h () ,4
.n' EC 50 as determined by the bioassay for.macrophage viability.
0.009 percent by phagocytosis and .004 percent b y acid pho spha-s case. A similar calculation has been made by Medinsky et aL
.U (1981), who concluded that the levels of H Se03 2 encountered by macrophages in human lungs were only 0.001 percent of the levels necessary to demonstrate toxicity in vitro. It must be realized, however, t h a t in v i t r o te s ts of mac rophag e f unc tion (e.g., via- .
bility, phagocytosis) may be relatively insensitive.
.U As always, the validity of extrapolating the harmful effects of controlled animal exposure s tudies to risk in human popula-tions is difficult. Not only are the levels greater than encoun-
, tered in urban air, but differences in animals and man in regard to pulmonary anatomy, physiology and b ioc hemis t ry make caution
", neces sary. For example, compared to man, the souse has a pul-
, monary ventilation approximately 10 times greater when expressed as. liters of air per minute per kilogram body weight. As a result, the mous e will receive a higher dose of metal per gram of lung for any given concentration and duration of exposure.
However, theoretical and experimental work has also shown that y collection efficiency of inspired particles is similar in a variety of animal species and humans. Thus,-an approximation of the toxicity of the various metals in man can be attempted by extrapolating from the nouse infectivity data presented in Table III.9. In this calculation, the average deposition in man was determined using the average concentration in community air, and
~
assuming there is no clearance. Even af ter these conservative
- assumptions are made, extrapolated ED 20 levels (the dose levels required to induce 20 percent of the maximum mortality above background following a bacterial or viral challenge) seem to be
- an order of magnitude below the predicted deposition rates of i metals in human lung.
! C .
In conclusion, it can be stated that no reliable scientific data exist which direc tly demons tra tes effects in humans as a
, result of chronic or long-term exposure to these various metals l
at levels found in the ambient air. Health ef f ects other than
- the induc tion of cancer have, typically, not been addressed due l 1 to a lack of quantitative information. As an initial step in evaluating the available data, a review of the pertinent data used in.the documentation of the Threshold Limit Values for the
, five metals reported is included in Appendix III.2 of this HEED.
Unfortunately, for the most part, these lack needed numerical l
data. However, there is an extensive data base which demon-C strates that serious biological effects have been observed in a variety of animal bioassays as a result of short-term exposures to these various metals at concentrations above those found in
( ambient air. The importance of these effects and the s im ila r i-ties of the biological systems between animals and humans suggest that long-term chronic exposure to humans may result in an as yet l unquantifiable risk to human health.
l 73 i
L
4 *
?
1 Table III.9 t Comparison pf[ Bioassay Infectivity Data irt Mouse and Man i
4 MOUSE MAN
! Average Average
.ED2 Maximum Extrapolated deposition pg/macrophage c in communi y
, pg/ kung pg/lunga pg/ lung b air (pg/m3) pg/ year f
4
-8 q Cd2+ .17 930 81 0.4 x 10 .002 1.61 i l',
0 -8 3 .66 680 315 1.4 x 10 <.003 - 0.90 <2.4-723
.63 8000 300 1.3 x 10 8 .032 25.7 i
Hn
- 1.4 1700 670 2.9 x 10-8 .100 ,
80.3 aFrom Schroeder (1970) bobtained by comparing the pulmonary ventilation of mouse and man. 6 cobtained by dividing by 23 x 10 , 9the average' number of macrophages/ lung (Crapo et' al.,1982) d From Graham et al. (1975) .
I i
+
. .. . . i ,
. i.
t+ i. / .: , O J ..,
.)-
. g, q
_, . _. , , __ ._. . . _ _ _ __ x-1 .__,_____../.. _ _ . .
I fT ^
l-l III.2.2 Non-Neonlastic Toxicity 23. Particles and Their.
i Sinnificance ,gg 1333g Realth:' Agli Sulfates 1
In the 1982 REED on airborne particles, we briefly reviewed D the nature and extent of toxic ef f ects of exposures to sulfuric acid in terms of morphological ef f ec ts and pulmonary f unction changes. In this HEED, we provide a more detailed description of the effects of sulfuric oxide aerosols on the mucociliary clear- ~
- - ance functions and the possible implications of these changes for
', lung disease. ' A brief summary of the reported effects of nitrate
,- aerosols.on respiratory. disease is also provided later in Section III' 2.3. . The information provided in the following two sections
-has been derived from the assessments provided by Dr. Morton Lippmann of the New York University Ins titute of Environmental.
l' Medicine - (Lippmann, 1983a,b).
t ~ Introduction
- i. For many years, standard te s t s of respira tory mechanical f unction have been used as criteria of irritancy potential of airborne particles. These tests were developed to assess the
,, ability of the lungs to . provide adequate ventilation of alveoli 3
( w hic h ,' o f cou r s e , is the lung's prime function). However, it r i does not follow that the' production of reversible mechanical l changes is the only effect of importance following pollutant exposure.
Indeed, questions of the sensitivity and significance of such changes have not been adequately resolved, and it is
! important to determine whether alterations in other lung defenses occur either before respiratory functica tests become~ abnormal, ,,
or in the absence of changes. Another critical component of the defense synten is mucociliary clearance from the tracheo-bronchial' tree. Therefore, in the following we summarize recent knowledge of the effects of airborne sulfate particles on physio-7 logical parameters. -
Effects 9.f Acid Sulfate Aerosols an Mucociliarv
(- Clearance Function: R e s u l t s f.g.,g.a R e c e n t Studies Mucociliary clearance is the main route by which particles-1.L . and-dissolved gases are removed from the conducting airways, and is a route of removal for macrophages.from alveolar regions.
Derangement of clearance may be involved in the development of
- chronic bronchitis or mucus hypersecretion (Hilding, 1965; Kilbu rn,196 8 ; Alb e r t et al.,1973; Wanner,197 7 ; Fishmen, 1980) and may also be a factor in the pathogenesis of bronchogenic Q cancer (Schlesinger and Lippmann, 1978; Menkes et al., 1979).
, Noninvas iv e measurements of clearsace can be made following the
- inhalation of radioactively tagged, insoluble tracer micro-sphere s, by measuring thoracic retention at various times f ollowing the brief inhalation using external ig yiy.g neasure-ments with co111 mated scintillation detectors.
O A series of studies of the effects of H SO4 on mucociliary clearance in humans was performed at NYU Institute 2 (sulfuric acid) of Environmental Medicine by Leikauf (1981). The subjects were -
O 75
+ , = , - , . - , , ,-rwe.w-%*.sw,,y.m.. g-,w,we3--,, yver - me-er e s.-.-e,mye- eee*es**~--&**'i'*----e- "---'**=v'*w'=*'"'-*** '*r'"=*-**'***-'*" " " ' " " ' " - ' ~ " " " ' ' - " " ' ' -
j I
(.j either exposed on f our dif f erent days f or one hour each day via ;
nasal mask to subnicrometer droplets of H 2 SO4 or to a distilled !
water aerosol..
100, 300 and 1000 ug/m The agid exposures, in random sequence, were at
. The following generalizations were made .s from the 3results of the various tests. The one-hour exposure to
- 100 ug/m had no ef f ect on tracheal mucus transport, but accel-ersted clearance in the large proximal airways ance-from the s ma ll ,d is t a l a ir way s. At 1000 ug/m and agowed clear-
, tracheal transport was still unaffected, but both proximal and distal -
airway clearance was depressed. Thus, it appears that the lowest -
H2 SO4 exposure level produced a small, stimulatory dose to suco- ~
ciliary transport in the larger airways, while at the same time
- the dose to the smaller conducting airways was large enough to i depress mucociliary transport in that region. Thus, single one-hour exposures to subnicrometer sulf uric acid aerosol did not -
significantly change any index of respiratory mechanics in non- .;
smoking volunteers, but did markedly alter mucociliary clearance.
However, very recent clinical evidence suggestgthat sulfuric acid when inhaled a t concentrations of 100 ug/m (near ambient levels) .can induce reversible pulmonary functional changes in adolescent asthmatics (see Koenig et al. 1983).
c Recently, Schlesinger et al. (1982) developed an animal
! model using the rabbit for further tests on the effects of irri-tants on clearance and other lung def ense functions. They have l used this model for studies of the effects of single one-hour
{ exposures to a variety of sulf ur oxide aerosols and f or a study
^
of the effect of daily one-hour exposures to H 2 SO4 In tests involving repeated one-hour exposures for five days / week, for four weeks, there was an examination of both mucociliary _ clear-ance changes during, and for two weeks after, the exposures, and of changes in the bronchial e p i t h e'l iu m as measured following sacrifice at two weeks after the last exposure. Three groups of ,
animals exposed at a concentration of H SO4 2 of eicher 250 or _
~
500 ug/m$ereOne group (Series 1) receijed 250 ug/m3
. via an oral tube; one (Series 2) received 250 ug/m via nasal mask, and one
- 500 ug/m 3 via nasal mask (Series 3). In all H 2SO 4 exposure i series, clearance times were significantly reduced from pred ~
l exposure values on specific individual days during the course of i
the acid exposures, with the greatest number of such days occur; -
l ring during the 500 ug/m 3 expo sures. In addition, f o r t h e a c id/* '
exposed animals, the relative number of airways in each classifi-cation group indicated an increase in Epithelial Secretory Cells (ESC). -
The results of the single exposures of rabbits indicated that H 2 SO 4 was the most potect of the major ambient sulfur oxide aerosols [ i.e. , EgSO NH HSO 4 4 (ammonium bisulf ate) and (NH 4)2SO 4 (ammonium sulfate)],4,in producing changes in the rate of tracheo-bronchial'aucociliary clearance. Ammonium bisulf ate was the only
- other' of these three chemical species which produced a signif-icant change in clearance rate. However, since it requires -
twice as much NH4 HSO4 asgH SO4 to produce the same [H*]'in splu-tion, then, stoichiometrically, approximately 1700 ug/m of NH4HSO 4 would be comparable, in terms of (H+], to approximately *
~
9 76
, ,m e e g 4Oe u.mmw ,,ig as se ,e e 6*We "
..-...n . . . - - . :. .. .
I
'o
'f =*
850 us/s 3 of H 2 SO 4 The change in clearance which w o'u ld be predicted to occur at this latter concentration of H 2 SO is 4 ,
i
- similar to that which was observed at the former concentration of O NH4 HSO4 . As a result these studies strongly suggest a relation ,
between the hydrogen ion concentration ([H+]) and extent of clearance alteration.
In summary, the sulf ur oxide having the greatest ef f ect on both respiratory mechanics and mucociliary clearance function is H2SO 4, and the effects appear to be related to the concentration J, ,
of hydrogen ion deposition on the airways. For both responses, the effects produced by single exposures appear to be transient.
Of the two responses, the more sensitive one is mucociliary
,, clearance, and the sensitivity is greater by about a f actor of 10.
C l I n c lic a t io n s d fJig, E f f ec t s 21 21 % sa Mu c o c ilia ry Clearance j ig,3JLg, Pathogenesis d Chronic BT o nth i t i s (Mucus Hvoersecretion) d
' Chronic bronchitis is a disease of the conducting airways, characterized by persistent excess aucus production (Thurlbeck,
(, 1976; Snider, 1981). In addition, human bronchitis and experi-mental animals having spontaneous or induced chronic bronchitis show altered mucociliary clearance function (Lourenco, 1969; 1 Canner et al., 1973; Holma, 1967; Iravani and Van As, 1972; Melville et al., 1980). Thus, chronic bronchitis involves dys-function of the mucociliary system, and altered clearance may be an initial stage in disease progression.
Unfortunately, there are few data concerning the response of the mucociliary clearance system under prolonged insult by poten-tisily harmful pollutants such as H 2SO 4, and direct experimental evidence for a role of H 2 SO in the etiology of chronic bron-
& chitis is currently lacking.4 There is, however, various sugges-tive evidence which implies an association between exposures to H2 SO4 and effects on human health. Epidemiologic s tudies suggest r i a relationship between sulfur oxide pollution and chronic bron-chitis; these surveys, however, did not generally examin e either lA aucociliary clearance dysfunction or H SO4 concentration. In one i study, Nebu tono (1978) reported that2people from two areas in Japan of differing pollution levels were screened. The greatest l
dif f erences suspended between the particles. two areas phlegm Persistent were in was the levels the most of S02 and common sympton in people from the area with higher pollution levels. In U a study in Holland, atmospheres characterized by sulfur oxides and smoke were related by van der Lende et al. (1981) to a higher prevalance of chronic cough and phlega production. On the other hand, no significant differences in mucociliary clearance l
' were found by Canner and Philipson (1973) in twins, where one twin lived in a rural area and the other in an urban region.
l v Thus, although available evidence suggests that exposures to j sulfuric acid may exacerba te dis eas e, it has not been clearly established whether it can initiate it, i
77 k . _ _ _ . _ _ _ - _ _
= '
I e The suggestion for a role of E SO4 2 in t h e d ev e lo p a'en t of chronic bronchitis is given added strength when results of studies of subnicrometer R 2504 or whole f resh cigarette smoke exposures, both conducted in the NYU laboratory with donkeys and- 2 humans, are compared (Lippmann et a l. , 1982). Cigarette smoke is an agent known to be involved in the etiology of human chronic bronchitis. The effects of both agents on the mucociliary clear-
, ance of tracer _ particles are essentially the same in terms of: ,
- 1) transient acceleration of clearance in low-dose. exposures; '
- 2) transient slowing of clearance f ollo wing high doses; and,
- 3) alterations in clearance races persisting for several months -
following multiple exposures. Thus, although direct evidence for an association between intermittent low-level exposures to H 2804 and chronic bronchitis is lacking, the similarity in response .
between H2 SO4.and cigarette smoke exposures suggests that such an ,
association is possible. '-
Discussion Since H 2 SO4 produces essentially the same sequence of ef f ects on aucociliary bronchial clearance as cigarette smoke, ..
following both short-term and chronic exposures, it may b e cap-
-able of contributing to the development of bronchitis. But the question still remains whether variable clearance rates, accel-ersted clearance, and persistent clearance changes merely pre-dispose to chronic bronchitis, or are the actual initiating events in a pathogenic sequence leading to its development. -
Furthermore, the response of the mucociliary clearance systes ~
observed in the rabbits may be adaptive, rather than patho-logical. Many irritants may stimulate clearance at low doses or after exposure f or a short time and then retard it at higher doses, or with prolonged exposures (Wolff etal., 1981). An increase in ESC proportion is consistent with hypersecretion. ,.-
Thus, low-level exposures may initially increase secretion, which -
can be coped with and may even be protective. However, patho-logical changes appear when adaptive capacity is overloaded.
Thus, with increasing exposure time and dose, the degree of enhanced secretion may be too great, resulting in overwhelming of clearance, leading to retardation (e.g., as shown in rabbits by Holma, 1971) and eventually bronchitis. -
III.2.3 Non-neoclastic Toxicity 21 Particles and Their .
Sianificance Ag Human Health: Nitrate Aerosols
.+
A re-examination of the literature and an update of our i previous review (presented in the 1982 REED) still indicates that ;
there are limited data linking nitrate aerosol exposures to '
repiratory disease. In the following, we briefly summarise per- l tinent literature available, to determine if nitrates in airborne -
particles contribute to the initiation or exacerbation of respiratory d is e a s e.
There are some reports which suggest that nitrates in par- ,
ticles may contribute to the initiation or exacerbation of respi-0 o
78 m e en ea sbes e-m w.i.res. ==--ep-s<e_,e _w---o_wW9stw-e-
l t()
f ratory disease. Utell at al. (1980) studied the effe'ets on -
airways of acute exposure to nitrate during uncomplicated influ-
+ enza A (H Ng) infections in 11 previously healthy adults and o' found s ign(if ican t functional responses associated with exposures to high concentrations of nitrates. Subjects were studied at the time of acute illness and I, 3 and 6 weeks thereafter. By double-blind randomization, each. subject breathed an aerosol of either sodium chloride or sodium nitrate for an initial 16-minute _
period and then breathed the other aerosol for 16 minutes three y hours later. The~aass median aerodynamic diameter (MMAD) of tge NANO 3 aerosol was 0.49 us; the concentration was 7,000 ug/m .
, Deposition studies showed a mean retention of 45 to 50 percent for both inhaled aerosols. Compared to inhalation of sodium 3
o chloride at the time of initial examination and one week later, exposure to sodium nitrate produced significant decreases in g specific airway conductance and partial expiratory flows at 40 percent of total lung capacity.
Kleinman et al. (1980) exposed 20 gormal and 19 asthmatic volunteers for two hours to 200 ug/m of NH 4NO 3 (ammonium nitrate) having a MMAD o f 1.1 um. They found no substantial
( alterations in pulmonary function or overall reported symptoms attributable to the nitrate aerosol for either group. However, there were some possible meaningful function and symptom increases in some individual members of the group.
In a population study of the effects of particulate air pollution on asthmatics, Perry et a l. (1983) recorded symptom scores, peak expiratory flows, and use-as-needed aerosolized bronchodilators for 24 asthmatics in Denver from January through March, 1979. Pollutant variables were 12-hour average concen-trations of CO, S0 2, 03 and particle mas s. The particles were
, collected in two size f ra c t ions , i.e., 2.5 to 15 um MMAD (coarse) i and less than 2.5 um MMAD (fine), and both fractions were -
analyzed f or to tal mas s, sulf a te, and nitra te. Of the environ-mental variables, only fine particle nitrate concentration was statistically associated with effects, i.e., with' increased >
symptom scores and bronchodilator usage.
1 The body of literature which suggests that inhalation of NO 2 affects respiratory disease incidence may have some relevance to the issue of possible effects of pa r t ic u la t e nitrate. NO 2 pent-
. trates through the upper airways and a significant fraction is deposited on small airways and alveolar surfaces where it hydrolyzes, forming nitrite (N0 2") and nitrate (NO 3 "). Thus, if nitrates have the effects suggested in the studies specifically related to nitrates, then exposure to NO2 may produce similar effects.
Of particular relevance to this discussion are the " gas
' stove" studies. Helia et al. (197 7) reported an increase inci- j dence of cough, bronchitis and " colds going to the chest" in British school children over a wide geographical area when the children lived in homes with gas rather than electric stoves.
There were no measurements of NO 2 in these studies, but later 79
C t
j studies by.the same group clearly indicated.that excess NO 2 was
, probably present (Melia et a l. , 1978). Speiser et al. (1980), in
.i their 'six-cities study, have also-shown that.there is a greater .
history of respiratory illness before age two, and decreased Forced Vital- Capacity (FVC) and Forced Expiratory Volume (FEV) in one second (FEYg) in children 6 to 10 years old from homes where gas is used for cooking. Hasselblad et al.(1981) reported a similar - ass ociation. This finding was age-dependent and changed -
according to the year s tudied. However, more recent results of m 1
Harvard's six-city study (Ware et al., 1983), in which additional ~
l . intake cohorts were considered and social class was better char-i acterized, d id .no t show a significant as sociation between gas stoves and respiratory illness before age two. Recent studies in this area by - the British group (Melia et al., 197 9; Florey, et al., 1979; Goldstein et al., 1?79), while not definitive in c their demons tration of an ef f ec t,. sugg es t a slightly greater relative risk in girls rather than boys. Other s tudies in Ohio gave negative results (Keller et al., 1979a,b). However, Speizer et al. (1980) noted that correction for parental smoking j improved their correlation of lung disease before age two with gas stove use, and they criticized the Keller study for the lack ;
of representative samples as well as small sample sizes.
1 .
J Respiratory disease illness was also associated with peak exposures to NO 2 in a community health survey in Chattanooga, Tenn. Love et al. (1982) found higher respiratory illness rates in a population periodically exposed to short-term peaks than in 2
) others without such peak exposures, even when the long-tera
- average concentrations were nearly the same for both groups. The .
reduction in excess respiratory disease among the same population in the f ollowing year, when the source of the short-term peaks was shut down, was consistent with the hypothesis that peak NO2 exposures contribute to an elevation in respiratory disease _
J t incidence.
i^
Discussion There are more studies for the effects of NO2 on humans and
, animals than for nitrates. However, these are relevant to a
- consideration of the effects of nitrates, since NO 2 depositing on respiratory epithelium dissolves and forms nitrates. As previ-ously mentioned, the " gas stove" and Chattanooga ambient air NO e p id e m io lo g ic studies suggest that there may be an excess of -
^
respiratory disease associated with NO2 *Eposure, and that the O effects are more likely associated w ith periodic peak exposures than with an elevation in the average exposure. Also, animal inhalation studies with NO2 demonstrate that NO2 can produce functional decrements and anatomical changes in peripheral air-ways and airspaces which are consistent with impaired ability to resist lung disease. Thus, while the effects of nitrate aerosols ..
3 on respiratory disease have not been estabitshed, there is a body of epidemiologic and toxicologic data which raises suf ficient concern to justify further investigation. .
C
- .a 80 O $ e @isse & = . = @*pp4 me-.9' U
_ . .e. - ~,
i l0 -
D i
Princioal Conclusions'AE.d. Future R e s e a r c h ]l[,3,gd,,3,
[ IV.
I PRINCIPAL CONCLUSIONS
?
m'-
The existing epidemiologic literature on the morbidity effects of human exposure to particulate matter was summarized in Section I I .1. The major findings were:
. Very few studies have addressed acute health effects of
, particle pollution below levels of 1000 ug/m3 (measured
- in 24-hour average TSP equivalents). Those which have addressed effects at lower . concentrations show an increase in hospital admissions for cardiac and
, respiratory illness, as well as for bronchitis symptoms. Even in these studies the average ambient e concentrations of TSP were high relative to current levels in the U.S.
. The available data for short- and long-term exposures to particulate matter do not suggest the existence of
- effect thresholds. For this reason, simple linear
(: coefficients are derivable which relate particle concentrations to respiratory infections, hospitaliza-i tion, and sgapcoms of chronig bronchitis. The cases /
year per 10 persons per ug/m were estimable for child-hood lower respiratory in*ections (60) and chronic
- bronchitis (0-170) using the British Smoke measure of -
. particles, and for acute respiratory disease (100-540),
excess emergency room visits for respiratory diseases (8-13), and excess total emergency room visits (20)
! using Total Suspended Particulate Matter data.
Standard errors of these estimates are roughly one-half the estimatea themselves, indicating a zero risk C coefficient to be within the confidence interval of -
each estimate.
In S e c t ion II.2, =we reported the r e s.u l t s of an original ep id emio log ic study using th'e National Health Interview Survey I
morbidity data and an index of fine p ar t ic le' po llu tion (based l 1 upon airport vis ibility da ta). Although this analysis is pre-liminary and will be continuing during the project's third year, we can report the f ollow ing findings:
. There is a correlation between fine particle air pollu-tion and human morbidity. This relationship persists even when the analysis is controlled for inter-city and seasonal effects. This fine particle effect is evi-danced only among persons reporting periodic limita-
! tions due to chronic conditions.
. Of the health seasures considered, Restricted-Activity.
Days (RAD) are the most sensitive to the effects ofair pollution. According to the estimates developed, per-sons with chronic lim ita t ion s are expected to experi-ence, in a two-week exposure period, O to 0.01 RADS and ls -
81 re- -y-- ., yw--y,. -
-v--,.m v--a---,-m-=e-=w---,-ewe-w=w-om,-v--- 3---.we-----<w ww-e-e-w-w, wee.=m-e-w---ee-o-m'e--me-wwein-wme-eroee=w--en v a ewe--= * *- ----e--+
mamman s - - - -
?
O to 0.004 WLDs per ug/m3 of average fine particle mass a concentration.
Time series of mortality and pollution data were analyzed in 7 Section II.3. Historical Coefficient of Hare (COH) and mortality data collected in New York City (NYC) were analyzed for the years 1963 to 1976. Further, a sensitivity analysis of the time-series approach was also conducted using a simulated pollution and -
m o rb id i t yd a t a set (generated based upon available Los Angeles ,
County raw data). The major conclusions drawn from these
- analyses were:
~,
. C0H was f ound to be related to temporal variations in NYC mortality. For a reasonable range of preliminary -
data filters, estimated coefficients from a model .,
relating daily mortality to COH range f rom 1.2 t o 2.0 deaths per day per unit COH. Similarly, a reasonable range of variations in the specification of temperature resulted in coefficients ranging from 1.3 to 1.8 dea ths per unit COH. Re-expressing the results as risk coefficients yielded a range of risk coefficients from -
~
0.01 to 0.02 deaths per day per unit COH per 100,000 persons. The 95 percent confidence interval of the estimates were roughly bytveen 0 and 0.03 deaths per day per unity COH per 10 persons. Until validation studies in several other cities with different charac-terizations and weather patterns a r e 'u n d e r t a k e n , the 3 results derived from this analysis apply directly only f or the mix of sources and time pattern of concentra-tions observed in NYC between 1963 and 1976.
. Simulations testing the ability of time-series methods to detect daily mortality effects of air pollution ~
O (applicable to Los Angeles) suggested that these methods would typically have a 95 percent chance of detecting a mean daily mortality / exposure coefficient (bgsed 10 upon particle KM measure) o f 0.0 6 deaths / day /
persons.
In Sec tion II.4, we reported results from cross-sectional analyses of total mortality in Standard Metropolitan Statistical Areas (SMSA) across the U.S. In these analyses, a number of particle metrics were considerd, including Total Suspended -
Particle Matter (TSP), Inhalable Particles (IP), Fine Particles (FP) and sulfates. The f o llo w ing conclusions were made:
I . Based upon the 1960 and 1980 SMSA analyses, it is con-cluded that the use of FP mass measure in cross-sectional mortality analyses results in a more significant and larger magnitude health effect than alternative measures available for such analyses. This -
agrees with both physiological and statistical expec-tations.
p 82 p- -+,q. .e_
t:
For 1960. TSP was f oun'd to yield a s tatis tically sig-i nificant ef f ect (p < .0 5), bu t in.the 1980 mortality regression TSP was not at all significant ( p >> .0 5) .
g ~
Based upon these conflicting results, the use of TSP f or the estimation of particle health ef f ects is not recommended. ,
. Using mean exposures'and expressing mortality risks in -
terms of similar units, the NYC time-series analysis indicated less than half the mortality risks predicted J- by our cross-sectional analysis. Though tentative, this finding-is consistent with the expectation that cross-sectional studies may capture more of the
.. chronic health effects of air pollution than would time-series studies (e.g., see Evans et al., 1983).
L:
.. Based upon the consistent importance of FP in the mortality regres sion examined, it is recongended that a FP egefficient ug/m of 1.3 deaths / year per he employed in the assessment 10 people per of the health tl effects of air pollution, with the 1960 sulfage coefficient (2.6 deaths / year per 10 3 people per ug/m )
being employed when estimates of FP impacts are
! unavailable.
. Despite the fact that the above-noted coefficients
-ve r e found to be statistically greater than zero, uncertainties not considered by such analyses ( e .g . ,
errors in the sessurement of the exposure variables and o ther non-sampling errors) make it pos sible that the
! mortality risk might, in fact, he zero. Such coefficients have in the past been applied without
, , adequate attention to their actual applicability to the -
N ~
situation and the uncertainties involved. Improper
( application of cross-sectional coefficients may lead to erroneous conclusions regarding health risks.
. Our studies of alternative particle mass measures (i.e., TSP,IP and FP) indicate that refinements in the d- estimation of human exposures to air pollution (e.g., using fine versus total mass) improve our abil-icy to quantify the health effects of these exposures.
O L The cancer risks of dif f erent types of airborne particles
' were examined (see Section III.1) using a relative potency model applied to Ames data. The major conclusions were:
Estimates cancer (in ug/m of thy increment in relative risk-of lung l
ex trge tab le grg anic s year s) cally ranged from 10- to 10~ for the sources con-typi-sidered (i.e., light-duty diesel, catalyst spark engine, non-catalyst spark engine,
~
woodstove, j residential heater, FBC, and conventional' coal and oil i combustion). The 95 percent confidence interval l of thesa estimates typically ranged f rom 10 4 to i
l #
83
. _ _ _ _ _ . . m. _.
6
' ~
...n . . - . _ . -
- c C.
10-1 around each o f' the particle-specific risk
,; coefficients.
- m
. Due to. larg e uncertaintie s in. predic ting incremental -
relative risk of lung cancer using Ames data, it is not possible to reliably discriminate between the potencies
. of- dif f erent types of particles. For this reason, t h e
- population risks associated with exposures to various '
' types of a mb ie n t' particles must, at this time, be p i '
ranked on the basis of quantities of extractable -
( s o lub l e) organics emitted, rather than individual
. estimates of potencies.
! In Section I I I . 2 .1, -w e evaluated the non-neoplastic -
toxicities ~.of various metals. The major con'elusions of this q 2
assessment were:
. The rankings of metal toxicities were found to be con-sistent with their Tgreshold p*imit Vagues (TLV) (i.e.,
Cd*+ b- V0 ~ 3 b Ni
- b Cr b Mn *).
. No reliable scientific data exist which directly demon-
~
etrates effects in humans as a result of chronic or
-long-term exposure to these various metals at the
. levels found in ambient air.
- Literature regarding.the human health effects of sulfate ;
aerosols were reviewed in Section III.2.2. Our conclusions were:
. Th~e health effects of sulfate aerosols appear to-be related to their acidic strength ( i.e. , their hydrogen l
ion concentration).
. 3 1 . Human health effects of sho'rt-term exposures to sulfuric acid to have been J own tics at roughly 100 ug/m approaching peak levels h in exercising asthma-
[
- recorded in the ambient environment. These H SO4_ 2 l levels are much lower than those for which. health l l effects had been seen before, 5)
. Since H2 SO4 produces essentially the same sequence of effect on aucociliary bronchial clearance as cigarette
! s moking '(f o llo w ing both short-ters and chronic *
[
exposures), it may be capable of ' contributing to the ,
development of bronchitis.
-] ,
, The contribution of nitrates to human health effects of ;
particles was examined in Sec tion III.2.3. The major finding s <
were:
. The potential impact of nitrates on human and animal 3 health has received relatively little attention in the scientific literature. However, in the absence of direct evidence regarding nitrate health effects, -
information can be inferred from exposure studies for
. g **
84 e== e.- gw-p rs. Se w . eme -mesem =, 4 e- .--4e% -ve 4 - w ee g g og*"ipKW-+-*p g e > Y
- t + y - p ergwte-yttwgy'yy'*=$*sy--- e-+=MNew p-g.e wpmis - egt==9wwegpw,3g%-9i.,-p-sig-sr*,-g-imemw.eewwwgwe-,,,.-v-m _meww ---w--*aw*,w*----ms-e-mr=,*W----ee---=WW--
m
, . _ . . _ . _ ._ __-,2 . , . _. _ _.
t i ,.
el
+
Y -
nitrogen' dioxide, since the gas dissolves in the respi-i ratory epithelium and f orms -of nitrates. Thus , there is .a body of toxicologic data which raises suf ficient 75 concern to . justif y further investigation of potential
- nitrate health effects.
t FUTURE RESEARCH NEEDS -
1 v During our analyses of the health ef f ects of particles, we
.f identified areas of future research which may help reduce some of the uncertainties-reported in this HEED. The following is a
-comprehensive listing of those research.needs.
- Morbidity Analyses t . For the s tudy of morbidity, existing data bases should receive additional attention. One promising data set
- comes from the National Health and Nutrition Examina- ;
tion Survey (NHANES). This data base includes direct measures of pulmonary function.
4
. Ongoing health surveys such as NHANES and HIS should be expanded to better accommodate analysis for air pollu-
, tion ef f ects. Farticular attention should be paid to improving exposure e s t ima t io n ( e.g. , to passive smok-
~
ing) by personal monitoring and the modeling of exposures based upon human activity patterns.
. Mor's prospective health studies to enable characteriza-t ion - ' o f chronic and acute health effect are definitely needed. However, since new data sets will-take a r
decade or two to develop, existing retrospective ~
population health data sets, should continue to be re-analyzed with more representative exposure data.
l . Associations obtained between air : pollution exposures l- and different ne'asures of human morbidity should be g compared in terms of their significance and biological p laus ib ilit y .
L
. Future work'should explore the sources and composition l o f - f in e particles in order to examine their respective i
importance in the interpretation of epidemiologic data v bases.
Mortality Analyses With regard to time-series analyses:
i- t . For time-series analysis, we need an improved model of
- the effects of confounding variables (such as tempera- 1 i ture) on sortality and morbidity. This also sus-gests a need to develop a physiologically-based model of acute mortality in time-series studies.
d s
85 9
-w-=- , ., - . , - . . - . . _ - - . . < ,.-----,.....-..-.m. .._%_,-+.w-.,,-y.c-,.,.,www.-ow.,vr,.v,-. ,,,w,-.--,w.,*w-ey-.-w,-4y.y.-w,.,7, ww y ww-, --
. __ . _..__.~ -
With regard-to cross-sectional analyses:
i ,
~
- .. For. future. health studies, we need combinations of both improved exposure and health measures. For example, the use-of health surveys or census data would be
. greatly enhanced by adding seasures that would at least partially account for cigarette smoke and indoor
- exposures. -
. The number and types.of particle sampling sites in th'e '
U.S. should be increased, and the aerosol characteriza-tion data should be improved and made more consistent ,
across sites.
. Historical data sets might be re-analyzed using. novel exposure estimates. For example, airport visibility, TSP and sulf ate measurements might be combined so as to ;
better represent fine particle concentrations.
- . .- Also needed are better estimates of personal exposures l co particles, including information on indoor / outdoor ,
j exposures by source and chemical composition.
. Various exposure averaging times should be e xam in ed . ,
This will partially address questions regarding '
response time associated with ob s erved. biological effects of air pollution. -
Toxic Effects g[ Particles
! .- ~
j' With regard to autagenicity and carcinogencity bioassays:
. Research is needed to reduce uncertainty .in bioassays
> for carcinogenicity. Potency derived from ig vitro tests and animal-to-human extrapolation produce the greatest uncertainty in this type of risk assessment.
l . Future bioas say s tudies should address the question of '
- - biological significance of assays with respect to human dis ease (both neoplas tic and non-neoplastic). Basic research into the physio-chemical mechanisms of carcia-ogenesis may eventually provide a model that, in con- * !
Junction with bioassays, will be quite useful to risk l -
assessment.
. . Relationships between assays need to be determined i .
under a greater data base to account for inter-lab and i inter-sample variab ilit y. The effects of differences l ,
in chemical composition on assay results also need ..
't l
- further clarification.
. Ques tions of bioavailabicy mus t also be addressed in '
assessing mutagenic potential of urban aerosols.
d a
86
, ~. ,.
... .. . . . - . . ,.~e.- . . . . . . ..
With regard.to non-neoplastic effects:
! . Research should be directed toward an improved under-N' standing of disease progression. More information is needed on the . pathogenesis of fibrosis, emphysema, and chronic bronchitis.
. The appropriate application of information on disease -
. processes to specific bioassay studies must be deter-iC mined. For. essaple, as the mechanism of fibro-genesis becomes better characterised, s tage-specific biosasays can be developed.
. The effect of repair and reversibility of injury after particle exposure must also be studied.
. How different agents interact in complex mixtures should also be f urther evaluated. This is especially important for urban air particles, which are complicated mixtures of aromatic hydrocarbon, metals, y silicates, sulfates and other components.
. Bioassays should be better calibrated with both toxic and non-toxic particula tes. More organic samples as well as emphysema-producing materials should be employed.
. The role of variation in dose-regimen in determining
- lung injury must be further assessed. This is espe-cially important in extrapolating from the relatively high-level short-term exposures used in animal studies i to the low-level long-term expo sures more charac ter-l j. istic of the urban environment. _
g i
'e I
I kj i
k I
%1 i
l 87 L
~
. .. ci G
I GI4SSARY Of CCMMONLY USE ABBREVIATICBIS C
Aerosol extinction coefficient. The fraction of incident.
- B* ** light absorbed and scattered by ambient aerosol particles per unit path length. B. ,g generally is expressed in units of ka-l. .
BS British Smoke. An optical aerosol measurement used O more cosmonly in Europe; the reflectance (or darkness) =
- to a standard " British Smoke".
COH .An optical aerosol measurements the percent transmission .
of white light through a filter deposit relative to a
, clean filter. Samples are generally collected (automatically) G every 2 hours2.314815e-5 days <br />5.555556e-4 hours <br />3.306878e-6 weeks <br />7.61e-7 months <br /> by tape sampler.
CP Coarse particle. Particle with aerodynamic diameter between 2.5um and 15pm.
CV Coefficient of variation. A dimensionless statistical measure.of O variation (standard deviation + mean) . Useful for comparison of the precision of model results.
D Aerodynamic Diameter. The diameter of a unit density
- (1 g/cm3) sphere that has the same settling velocity as the particle. O DCM Dichloromethanes the solvent most commonly.used in mutagenenis bioassays.
EC 50 The estimated concentration required to produce an experimental
- outcome level (which is specific for each bicassay) of.
j half the maximum effect over the background effect.
ED For infectivity bioassay: The dose required to induce 20%
20 _
of the maximum mortality above background.
-EM Algorithm. A method for computing maximum likelihood .
l estimates.
FBC Fluidized Bed Combustion type coal burning power plant. .
FEV Forced expiratory volume. Volume of air that can be .s exhaled with maximal effort over a fixed duration (e.g., FEV, over 1 second): measured to indicate changes in lung function.
TP Fine particle. Particle with aerodynamic diameter smaller than 2.5pm.
C
'h 88 e e
V
. . . ~.- m... - . . s === a .- .. , _ . . . .
s I
s .
, FVC Forced vital capacity. The tctal volume of air that can
, be exhaled with maximal efforts measured to indicate changes in lung function.
s HIS National Health Interview Survey. A national study of approximately 100,000 individuals for which information on chronic and acute health conditions was collected. This ~
s, data base was used in the analysis of FP morbidity effects.
Hi-Vol High volume samplers the current EPA reference method for sampling TSP.
IP Inhalable particle. Particle with aerodynamic diameter smaller than 15pm.
KM Unit of reflectance in.a measurement of blackness of an aerosol filter deposit. The KM measurement is proportional to the sample's elemental carbon content.
.( MAD Mass median aerodynamic diameter. A characteristic parameter which describes a distribution of particles. Half of a distribution's mass is contributed by particles smaller than the MMAD and particles larger than the MMAD.
RAD Restricted Activity Days. Health variable from HIS data set used in analysis of FP morbidity effects.
Rev/n7 Revertants/millijoule; a unit of mutagenic activity per unit of energy produced.
REP Respirable particle. A particle which can deposit in t the alveolar region of the lungs. Measured as a variable fraction of particles smaller than 10pm aerodynamic diameter.
SARCAD The Storage and Retrieval of Aerometric Data System. A national database of air quality and site information provided by various local, state and federal air pollution L agencies. The EPA Inhalable Particle (IP) network data is collected at SAROAD sites.
SMSA Standard Metropolitan Statistical Area. A geographic area
? defined by the U.S. Census bureau for heavily populated regions.
c.
S-9 A rat liver homogenate which contains metabolic enzymes that promote bactinal response to mansnalian mutageus. Used in Ames test to enhance detection of mutagens.
TA 98 Salmonella Typhimurium (a bacterium) mutant strain used in s Ames test. -
J 89
. - ~. -
=
1 TLV_ . Occupational Threshold Limit Value. Airborne concentration ;
~o f a. substance below which nearly all workers may be
- exposed repeatedly eight hours per day, forty hours per 3
week, without adverse effect, according to ACGIH ' (American .
~
Conference of Governmental Industrial Hygenists) . ,
TSP Total-Suspended Particulate Matter. The fraction of airborne C particles collected by high-volume sampler. Generally, . ,
, only particles smaller than 50um aerodynamic diameter are captured.
y visibility or visual range. The maximum distance at which an object can be discriminated from its background. 3 WLD Work-loss days. Health variable from HIS data set used in analysis of FP morbidity effects.
C;.
e-
%- O e
- 4 t
. Q D
~
90
., y w- w-e*
, _ , . , _ . ex. s d : )- REFERENCES Ahlberg, M., L. Berghem, G. Nordberg, S.A. Persson, L. Rudling, and ..
B..Steen. (1983) . Chemical and biological t:haracterization of emissions a
. Ofrom coal and oil fired' power plants. - Env. Health Persp. 47: 85-102.'
- ,e ST* ~ Albert, R.E., M.'Lippmann, H.T. Peterson,'Jr., J.M. Berger, K. Sandborn and gf D.E. Bohning (1973). Deposition and clearance of aerosols'. Arch. Intern.
j :
'Med. 131: 115-127.
. Albert, R E. , T. Lewtas, S. Nesnow, T.W. Thorslund, and E. Anderson (1982) .
A comparative potency method for cancer r.isk assessment. Application
- y e
to diesel particulate emissions. Risk Analysis (in print) .
Alfheim, I. , J.G.T. Bergstrom, D. Janssen, M.Moller (1983) . Mutagenicity
~
- in Emissions- from Coal and Oil Fired Boilers. Env. Health Persp. 47:
177-187.
.p American Conference of Governmental Industrial Hygienists. TLVs: Threshold Limit valuer..,for chemical substances in workroom air. Adopted by ACGIH for 1981. ACGIH, Cincinnati 1981.
1 A 4 Y Camner, P. and K. Philipsohn (1973). Urban factor and tracheobronchial
(; \ .
(clearance. Arch. Environ. Health, 27: 81-84.
Camner, P.), B. Mossberg and K. Philipson (1973). Tracheobronchial clearance and chronic. obstructive lung disease. Scand. J. Rsp. Dis. 54: 272-281.
Cass , G.R. , M.H. Conklin, J.J. Shah, oJ.J. Huntzicker, E.S. Macias (1983).
Elemental carbon concentrations: estimation of an historical data base, Accepted for publication in Atmos. Env. ,
' . Clark, C.R. and C. Hobbs. (1980) .
Mutagenc'ty i of cffluents from an
' experimental fluidized bed coal comburter. Env. Mut. 2:101-109.
C Clark, C.R., R.E. Royer, A.L. Brooks, R.O. McClellan, W.F. Marshall, T.M.
Naman, D.E. Seizinger (1981) . Mutagenicity of diesel. exhaust particle extracts: influence of car type.-Fund. Appl. Tox.' 1:260-265.
Clark, C.R. , T.R. Henderson, R.E. Royer, ' A.L. Brooks , R.O. McClellan, W.F. Mart. hall and 'T.M. Naman (1982) . Mutagenicity of diesel exhaust
~#
C particle extracts: influence of fuel composition in 'two diesel engines. Fund.-Appl. Tox. 2:38-43. ,
) Claxton, L. And J.L. Huisingh (1979) Ccmparative edutagonic activity of A- organics from combustion sources. Pulmonary Toxicology of Respirable
- b '
Particles Conference - 791002 DOC Symposium Series 53: 453-463.
m Claxton, L. and M. Kohan (1981) . Bacterial n:uiagenasis and the evaluation of mobile-source emissions. Short Teiln Bioassays in the Analysis of
,g Camplex Environmental Mixtures II., (Eds., M.D. Waters, S.S. Sande, J. Lewtar Huisingh, L. Claxton, S. Nesnow.) pp. -299-317.
- 7. ' i A ,
lCrapo, J.D., B.E. Barry, P. Gerh, M. 3achofen, and E.R. Weibel. (1982) cell Number and charateristics of the normal human lung. Am. Rev. Respir. Dis. 125: 740-745.
h 91 e 4 ~ ,
w-. - 2m .. . .. l.....
, 'Crocker, T.D.. et.'a1.-(1979). Methods Development for Assessing Air
- Pollution Control Benefits. U.S. Environmental Protection Agency.
. Cuddihy, R.G. , F.A. Seiler, W.C. Griffith, B.R. Scott, and R.O. - McClellan. , (1980) E Potential Health and Environmental Effects of Diesel' Light Duty
. Vehicles, I,LMF - 82, Lovelace Inhalation Toxicology Research Institute, Albuquerque,1N.M.
, Cuddihy R.G, W.C. Griffith, C.R. Clark and R.O. McClellan (1981) . . Potential health .
and Environmental Effects of Diesel Light Duty Vehicles II, LMF-89, i Iovelace Inhalation Toxicology Research Institute, Albuquerque, N.M.
Diem,'K and C. Lantner,eds. (1973). Documenta Geigy - Scientific Tables, 2 Ardsley, N.Y.: Ciba-Geigy Corp., p. 710. ,
Dukovich, M. R.E. Yasbin, S.S. Lestz, T.H. Rigby, R.B. Zweidinger (1981) .
. The mutagenic and SOS inducing potential of the soluble organic l fraction collected from diesel particulate emission. Env. Mut.
_3.: 253-364.
c.:
DuMcuchel, W.H. And J. Harris (1983) . Bayes methods for combining the '~
results of cancer studies.in humans and other species" J. Amer.
Stat. Assoc. 78: 293-308.
Evans, J.S. , T.D. Testeson, and P.L. Kinney (1982) . The utility of parameter estimates devised from cross-sectional mortality studies.
)' .
Analysis of Health Effects Resulting from Population Exposures to
~
Ambient Particulate Matter, Harvard University, Prepared for the
!^
[ Department.of Energy,-Washington, D.C.
Evans, J.S., P.L. Kinney, J.L. Koehler and D.W. Cooper;(1983). The ,
relationship between cross-sectional and time-series studies. * *, m J. Air Pollut. Control Assoc. (in press).
I-a 'Fishman, A.P. (1980) . The spectrum of chronic obstructive disease of the airways. In: Pulmonary Disease and Disorders, (A.P. Fishman, ed.)
pp. 458-469, McGraw Hill, N.Y.
wi Fisher, G.L. (1983) . Biomedically relevant chemical and physical properties I
- l l: of coal combustion products.Env. Health Persp. 47: 189-200.
l Florey, C.- Du. V. , R.J. W. Melia, S. Chinn, B.D. Goldstein and A.G.F.
- Brooks (1979). The relation between respiratory illness in primary school children aa.d the use of gas for cooking.*III. Nitrogen ,j dioxide, respiratory illness and lung infection. Int. J. Epi. 8 347-353.
Goldstein, B.D., R.J.W. Melia, S. Chinn, C. Du V. Florey, D. Clark and H.H. John. (1979) . The relation between respiratory illness in primary school children and -the use of gas for cooking. II. Nitrogen dioxide, O respiratory illness and lung infection. Int. J. Epi. 8: 347-353.
Graham, J.A. , M.D. Waters, D.E. Gardner and D.L. Coffin (1975) . Effect .
of trace metal'on phagocytosis by alveolar macrophages. Infec.
Imun. 11: 1278-1283. ,
92
. . . . s.- _ _ - . . , u..__.. a ._
._3 s x ; ._ ... .u_ n- . .__ -
l'
?l- . Harville,~D.A. (1977). Mar 4== likelihood approaches to variance component tv - estimation and to related problems. - Journal of the Americal Statistical Association, 72:320-340.
Hasselblad, V., C.G. Humble, M.G. Graham and H.S. Anderson (1981) .
, . Indoor environmental determinants of lung function in children.
~
Am. Rev. Respir. Dis.- 123:472-485.
[ Hayes ,' T.L. , J.B. Pawly, G.L. Fisher, and M. Goldman (1980) . A model for
- the exposure of individual lung cells to the foreign elements contained in fly ash. Envir. Res. 22:499-509.
Hilding, A.C. (1965) Mucociliary insufficiency and its possible relation 3, to chronic bronchitis and emphysema. Med. Thorac. 22:239.
Holma, B. (1967) . Lung clearancc of mone - .d di-disperse aerosols determined by profile scanning and esaole-body counting. A study ,
on normal and 50 exposed rabbits. Acta. Med. Scant. 473 (Supp.)
2
.C Holma, B. (1971) Pathophysiology of the cleansing mechanisms of the lung.
Progr. Resp. Res. 6:51-55 i
Holmberg, B. and U. Ahlborg, Consensus Report: Mutagenicity and Carcino-genicity of car exhausts and coal combustion emissions. Env.
Health Persp. 47:1-30, 1983. _
Hyde ,- J.D. , R.E. Gibbs , R.A. Whitby, S.M. Byer , R.J. Hill, T.E. Ho ffman ,
R.E. Johnson, P.L. Werner (1982) . Analysis of particulate end gaseous emissions data from in-use diesel passenger cars. SAE Tech.
Paper series #820772.
T Iravani, J. and A. Van As (1972). Mucus transport in the tracheobrobronchial -.
-, tree of normal and bronchitis rats. J. Pathol. 106:81-93.
i Keller, M.D., R.R. Lanese, R.I. Mitchell and R.W. Cote (1979). Respira-tory illness in households using gas and electricity for cooking.
- I. Survey of incidence. Environ. Res. 19:504-515.
Kilburn, K.H. (1968) A hypothesis for pulmonary clearance and its impli-
, cations. Amer. Rev. Resp. Dis. 98:449-463.
Kleinman, M.T., W.S. Linn, R.M. Bailey, M.P. Jones and J.D. Hackney
.. (1980). Effect of ammonium nitrate on human respiratory function V- and symptoms. Environ. Res. 21:317-326.
31.
Koenig, J.Q., W.E. Pierson and M. Horike (1983). The effects of inhaled
.,- sulfuric acid on pulmonary function in adolescent' asthmatics.
Am. Rev. Respir. Dis. 128:221-225.
Kubitschek, H.E. , D.M. Williams (1980) . Mutagenicity of fly ash from a fluidized bed combustor during slant-up and steady operating conditions. g Mut. Res. 77:287-291. dr 93 g - -a- ' *e'p--
- a .- . . . z .
. . .: ,. . . - .., .. wx s . s
'[ Laird,'N. (1982). IComputation of variance conponents using the EM i algorithm. J. Stat. Coup. . Simul. 14 : 295-303.
S Latimer, D.A. et.'al. (1978) The development of mathematical models
'for the prediction of anthropogenic visibility impairment, V1.
(USEPA report. EPA-450/3-78-1102, Sept. 1978, RTP, N.C.
- Leikauf, G. (1981) The' effects of sulfuric acid aerosol on respiratory mechanics and mucociliary particle clearance in healthy and asthmatic 0 adults. Ph.D. Thesis, New York University Graduate School of Arts 5 and Sciences.
Lang, J.M., R.A. Carlson, L. Snow, F.M. Black, R.B. Zweidinger and s
-S.B. Tejada. (1981) . . SAE Tech Pap. Ser. No. 81186.
Lawther, P.J. (1958). Climate, air pollution and chronic bronchitis.
Proc. Roy. Soc. Med. 57: 262-264.
Lawther, P.J. , R.E. Waller and M. Henderson (1970) . Air pollution and exacerbations of bronchitis. Thorax 25: 525-539 ,
' Lewtas, J. (1982). Comparis:n of the mutagenic and potentially carcino-genic activity of particle bound organics from wood stoves, residential' oil furnaces, and.other combustion sources. In: Residential Solid Fuels Environmental Impacts and Solutions. (Eds. J. A. Cooper, and D. Malde)
Oregon Graduate Center, Beaverton, OR. pp. 606-619. ,
J Lippmann, M., R.B. Schlesinger, G. Leikauf, D. Spektor and R.E. Albert (1982). Effects of sulphuric acid aerosols on respiratory tract airways. Ann. Occup. Hyg. 26: 677-690.
Lippman, M. (1983a) Task A- Effects of airborne particles on physiological
~
- ' ~
parameters. Assessment prepared for the Harvard University Energy and Environmental Policy Center project on health effects of exposures to airborne particles.
Lippmann, M. (1983b) Task B.- Implications of physio -
logical changes to lung diseases. Assessment prepared for the Harvard 5- . University Energy and Environmental Policy Center project on health .',
effects of exposures to airborne particles.
Lourenco, R.V. - (1969). Distribution and clearance of aerosols. Amer. Rev. .
Resp. Dis. 101: 460-461..
Love, G.L. , S.P. Lan, C.M. Shy and W.B. Riggan (1962) . Acute respiratory
' illness in families exposed to nitrogen dioxide ambient air pollution in Chattanooga, Tennessee. Arch. Env. Health, 37: 75-80.
Manuel, E.H. et al. (1983). Benefit and new benefit analysis of alternative national ambient air quality standards for particulate matters. ,
Prepared for Benefits Analysis Program, Economic Analysis Branch, Office of Air Quality Planning and Standards, U.S. Environmental
. Protection Agency by Mathtech, Inc., EPA Contract 68-02-3826, March ,
1983.
~
94
& en a um 6 +Nw. *W6-46 a h-4 .r o
- p. ,
- .~ ,+
~ .
,O Martin, A.E.-(1964) Mortality and morbidity statistics and air pollution.
eI' Proc. Roy. Soc. Med. 57: 969-975.
.O Mazumdar, S. , H. Schineel and I. Higgins (1982) . Relation of daily mortality to air pollution: An analysis of 14 Iondon winters, 1958/Se-1971/72. Archives of Env. Health. - 37(4) : 213.
i . Mazumdar, S. and N. Sussman (1983) . Relationships of air pollution to
+U health: Results from the Pittsburgh study. Archives of Env. Health,
, 38: (1) 17-24.
- Medinsky, M.A. , R.G. Cuddihy, J.O. Hill, and R.O. McClellan (1981) . Toxicity t of selenium " compounds to alveolar macrophages. Toxicol. Lett. 8:
289-293, t
Melia, R.J.W. , C. Du V.Florey, D.S. Altman and A.V. Swan (1977) . Association between gas cooking and respiratory disease in children. Br. Med. J.
2: 149-152.
Melia, R.J.W., C. Du V. Florey, E.D. Palmes and B.D. Goldstein (1978).
- k. Differences in NO2 levels in kitchens with gas or electric cookers.
Atmos. Environ. 12: 1379-1381.
- Melia, R.J.W. , C. Du V. Florey and S. Chinn (1979) . The relation between respiratory illness in primary school children and the use of gas for cooking. I. Rtsults from a national survey. Int. J. Epi.
~
8: 333-338. .
Melville, G.M., S. Ismail and C. Sealy (1980) Tracheobrochial functions in health and disease. Respiration, 40: 329-336.
( Menkes, H. A. , S. Permutt and B.H. Cohen (1979) . Patterns of pulmonary _
function abnormalities in high risk groups. In: %e Lung in the Transition Between Health and Disease, (P.T. Macklem and S. Permutt, ed.). pp. 301-312. Marcel Dekker, N.Y.
Mumford, T.L. and J. Lewtar (1982) . Mutagenicity and cytotoxicity of coal fly ash from fluidized bed and conventional combustion. J.
3( i Toxic and E:.v. Health,10(4-5) : 565-586.
NCHS . (1975) . Health interview survey procedure: 1957-1974, National Center for Health Statistics, pub. series 1, Number 11. Rockville, MD.
1 Nobutomo, K. (1978) . Air pollution and cytological changes in sputum.
Lancet, 8063. (I) : 523.
Ostro, B.D. (1983a) Urban air pollution and morbidity: a retrospective approach. Urban Studies. (in press).
- Ostro, B.D. (1983b) The effects of air pollution on work loss and morbidity. J. Env. Econ. and Mgmt. (in press) .
~
95
[^ -. -
2 -._ . L-
A
, . s
- Pederson, T.C. 'and J.S. Siak (1982) . The role of nitroaromatic compounds ,
in the direct-acting mutagonicity of diesel particle extracts.
J. Appl. Toxic. 1: 54-60.
Perry, G.B., H. Chai, D.W. Dickey, R.H. Jones, R.A. Kinsman,.C.G. Morrill, S.L. Spector and P. Weiser (1983) . Effects of particulate air pollution -
on asthmatics. Am. J. Publ.. Health, 73: 50-56. ,
Pierson, W.R.,'R.A. Gorse, Jr., A.E. Szkarlat, W.W. Brachaczek, S.M. Japar, 4 F.S.C. Lee, R.B. Zweidinger and L.D. Claxton (1983) . . M2tagenicity and chemical characteristics of carbonaceous particulate matter
.from vehicles on the road. Env. Sci. and Tech. 17:31-44. J Pitts, J.N., Jr., D.M. Lokensgard, W. Harger, T.S. Fisher, V. Mejia, J.J. 5 Schuler, G.M. Scorziell and Y.A. Katzenstein (1982) . Mutagens in deisel exhaust particulate identification and direct activities of 6.
nitrobenz(a)-pyrene, a-nitroanthracene,1-nitropyrene, and SH-phenanthrol (4,5, bcd) pyran-5-one. Mutat. Res. 103: 241-249.
l .
_ Samet, J.M. ,- F.E. Speizer, Y. Bishop, J.D. Spengler, and B.G. Ferris , Jr.
- (1981). The relationship between air pollution and emergency room visits in an industrial community. J. Air Pollut Control Assoc.
31: 236-240.
Saric, M. , M. Fugas, and O. Hurstic (1981) . Effects of urban air pollution on -
school-age children. Arch. Environ. Health,36: 101-108.
., Schimmel, H. (1978) Evidence for possible acute health effects of ambient air pollution from time series analysis: Methodological questions and some new results based on New York City daily mortality, 1963-1976. Bull. N. Y. Acad. 54: 1052-1108. ;
Schimmel, H. and Greenburg (1972) . A study of the relationship of pollution to mortality in New York City, 1963-1968. J. Air. Follut. Control Assoc. 2_2,: 607-616.
Schimmel, H. and T.J. Muramski (1976) . The relation of air pollution j to mortality. J. Occup. Med. 18: 316-333. i~
Schlesinger, R.B. and M. Lippmann (1978) . Selective particle deposition
.and bronchogenic carcinoma. Environ. Res. 15: 424-431.
- Schlesinger, R.B., M. Lippmann, D. Spektor, G. Leikauf, L.C. Chen and T. ;
Vollmuth (1982). Toxicological studies of inhaled sulfur oxide aerosols.
Final report, contract RP1157-1 to Electric Power Research Institute, NYU Institute of Environmental Medicine, New York, New York.
m.
96 '
4w w 4 &
g, .-
2 2. _ ..
. _ . , .. . ... .. .a .
b.
r Sch:wder, H. A. (1970). A sensible look at air pollution by metals. Arch.
Environ.~ Health, 21: 798-806.
. Snedecor, G.W. and W.G. Cochran _(1967) Statistical Methods. Iowa State University Press, Ames, Iowa.
Snider,'G.L. (1981) . Pathogenesis of emphysema and chronic bronchitis.
.. Med. Clin. NA, 65: 647-665.
a' t
1 Speizer, F.E. , B. Ferris, Y.M.M. Bishop and J. Spengler (1980) . Respira-
' tory disease rates and pulmonary function in children associated with NO2 exposure. Amer. Rev. Resp. Dis. 121: 3-10.
Thurlbeck, W.M. (1976) Chronic Airflow Obstruction in Lung Disease. W.B.
, l.-
Saunders, Philadelphia.
I Trijonis, J. and K. Yuan (1978) . Visibility in the northcasti long-term visibility trends and visibility pollutant relationships. Prepared for the U.S. Environmental Protection Agency, EPA-600/3-78-075.
(^
Trijonis,'J. (1983). Development and application of methods for estimating inhalable and ' fine particle concentrations from routine ' Hi-vol' data. Atmos. Environ. 17 (5) : 999.
- ~~Utell, M.J. , A.T. Aquilina, W.J. Hall, D.M. Speers , R.G. Douglas , Jr. ,
[ . F.R. Gibb, P.E. Morrow and R.WT Hyde (1980) . Development of airway reactivity to nitrates in subjects with influenza. Amer. Rev. Resp.
Dis. 121: 233-241.
U.S. EPA (1979a) Protecting visibility- an EPA report to Congress.
U.S. Environmental Protection Agency Publication, EPA-450/5-79-008.
4 , Oct. 1979. RTP, N.C.
U.S. Environmental Protection Agency, Mobile Sources Research Committee (1979b). The diesel emissions research program, EPA 625/9-79-004.
Washington, D.C.
(_ Van der Lande,-R., T.J. Kok, R. Peset Reig, P. H. Quarger, J.P. Shouten J and N.G.M. Orie (1981) . Decreases in VC and FEV 1 ndicator for effects of smoking and air pollution.0. Bull. Europ.
Physiopath. Resp. 17: 775-792.
cp
-Via u, C.J. , J.M. Lockard , H.G. Enoch, and P.S. Sabharwal Costparison of v the genotoxic activities of extracts from ambient and forest fire polluted air. Env. Mutagenesis. 4: 37-43, 1982.
Waggoner, A.P. and R.E. Weiss (1980) . Com[arison of fine particle mass concentration and light scattering extinction in ambient aerosol.
,-- -Atmos. Env. 14, 623-626.
r D
O 97 .
h 4 W M e-r ah
_7
[
l I
.g Wang, Y.Y.,'S.M. Rappaport, R.F. Sawyer, R. Talcott, E.T. Wei (1978).
.]
Direct-acting mutagens in automobile exhaust. Cancer Lett. 5: 39-47.
. Wanner, A.-(1977). Clinical aspects of mucociliary transport. Amer. Rev.
Resp. Dis. 116: 73-125.
'. Ware, J.H., L.A. Thibodeau, F.E. Spelzer, S. Colome, B.G. Ferris, Jr. (1981) .
Assessment of the health effects of atmospheric sulfur oxides and n particulate matter evidence front observational studies. Env. Health .
Persp. 41: 255-276. I Ware, J.H., D.W. Dockery, F.E. Speizer and B.G. Ferris, Jr. (1983).
- Passive smoking, gas cooking and respiratory health of children j living in six cities. Am. Rev. Respir. Dis. (in press). -
Watson, J.G. , J.C. Chow, and J.J. Shah (1981) . Analysis of inhalable and fine particulate matter measurements. Environmental Research and Technology. Prepared for the U.S. Environmental Protection Agency, EPA 450/4-81-035.
~
Wolff, G.T. , C.M. Stroup, D.P. Stroup (1983) The coefficient of haze as a measure of particulate elemental carbon. J. Air Poll. Control Assoc. 33 (8) : 746-750.
Wolff, R.K. , B. A. Muggenburg and S. A. Silbaugh (1981) . Effects of 0.3 and 0.9 um sulfuric acid aerosols on tracheal mucous clearance in 3 beagle dogs. Amer. Rev. Resp. Dis. 123: 29,1-294.
- M L
s J
D
^
98
- m. . - .
~
[ .
~
._ c .s i 1 ,,
. ,- .~/'
G 'q
- " h '
r' '
8 , ye C"
y 1, INDEX TO' EXHIBITS v)
INDEX Reference Appl. Appl. AWPP Testimony Description Ref. No. Ex. No. No. Page PECO Letter to NRC 1 47 6 Dated 6/10/83 PECO Letter to NRC 2 48 6 Dated 9/15/83
'~
NRC IE Report 50-352/ 3 49 21 82-06 (Weld Inspection-Mobile NDE Lab)
NRC IE Report 50-352/ 4 50 21 82-16 (CAT Inspection-Appendix B)
NRC 1982 SALP Report 5 51 22 (Pages 7 through 10)
( ,3) (Pages 12 through 15) 6 52 22 NRC IE Report 50-353/ 7- 53 Items 27,30,33 76-06 '1,2,3,4 PECO F,inding Report 8 54 138, 138A, 28,32 N-093 138C, Item 9 33 I
Bechtel NCR 1980 .9 55 139 28,30 Bechtel NCR-1998 10 56
- 30,33 Bechtel NCR 2000 11 57 30,33 PECO Letter to NRC Dated 12/15/76 12 58 Items 5, 6 31 ,
Bechtel FIR C-63-7 13 59 .
33 Bechtel FIR C-63-8 14 60 33 Bechtel FIR C-63-9 15 61 33 Bechtel PIR C-63-10 16 62 = . . J 3
.o v
4
+
U.S. NUCLEAR REGULAT0JY CSMMISSION EXXIBIT No. Y 7-/34 tervener Applicant identified e:2iv
_M.bt-if
___ neiected Date: M 7*[
Reporter: I /[l/3 -
4; '
c..
- e 3 -
~ ' ~
--2'- .
Reference' Appl. Appl. AWPP Test'imony Description ,
Ref. No. Ibc. No. No. -Page f ~1
'1 Bechtel FIR C-63-11 17 63 33
-Bechtel FIR C-63 18 .
64 33 Bechtel FIR C-63-13 19 65 33 Bechtel FIR C-63-14 20 66~ 33 Bechtel FIR C-63-15 21 G7 33 Bechtel' FIR C-63-16 22 68 33 Bechtel FIR C-63-17 23 '
G9 33 Bechtel FIR C-63-18 24 TO 33 Bechtel FIR C-63-19 25 71 33 NRC IE Report 50-353/77-01 -26 72 34 Bechtel FIR C-63-20 27 73 34 28 74 Bechtel FIR C-63-21 34
.( )- 'Bechtel FIR C-63-22 29 TS 34,35 NRC IC Report 50-353/
77-06, Page 5. 30 76' 35 Dechte1 FIR C-63-24 31 77 35 Bechtel FIR C-63-25 32 78 35 g Bechtel FIR C-63-26 33 79 35 Bechtel FIR C-63-27 34 S'O 35 Bechtel FIR C-63-28 35 8'1 35 Bechtel FIR C-63-29 36 6'2 35 ,
Bechtel FIR C-63-30 37 8'3 35 ,
Bechtel FIR C-63-31 38 S'4 35
-- Bechtel FIR C-63-32 39 S'S 35 Bechtel FIR C-63-33 40 86 35,36 i-L Jg Bechtel FIR C-41-493 41 97 35 i- Bechtel NCR 2627 42 88 35 ,
G _,. ..
_3_ I Reference Appl. Appl.. AWPP Test'imony )
Description -
Ref. No. Ex. No. No. Page i *;
~#
Dechtel NCR 2710 43 89 35,36 NRC IE Report 50-353/ .
77-14, Page 4 44 90 36 NRC IE Report 50-353/
77-02, Page 6 45 91 142 50 PECO Response to NRC dated 5/13/77 (77-02) 46 92 152 50 NRC IE Report 50-352/
77-07, Pages 3 and 4 47 93 50 NRC IE Report 50-352/
78-03, Pages 14 and 15 48 S'4 51 PECO letter to NRC dated 6/12/78 49 95 52 PECO Letter to NRC dated 9/18/78 50 9'6 52,55 NRC IE Report 50/352/
(')
L' 78-07, Page 4 51 9:7 53 PECO Letter to NRC dated 12/4/78 52 9:8, 54 NRC IE Report 50-352/79-11, Page 7 53 S:9 54 NRC IE Report 50-352/ g 78-04, Pages 10 and 11 54 LOO 54 PECO Letter to NRC dated 7/20/78 55 401 55 c
NRC IE Report.50-352/
79-04, Page 2 56 102 55 .
PECO Letter to NRC dated 3/2/79 57 103 180 '57 NRC IE Report 50-352/
79-12, Page 6 58 to4 58 PECO Letter to NRC -
dated 10/31/79 59 105 60,62 lll NRC IE Report 50-352/
80-02, Page 5 60, 62 to6 61,64
7
--4 -
Reference Appl. Appl. . 'AWPP Tostimony Description , Ref. No. .Ex. No. No. Page NRC'IE. Report 50-352/
81-06, Page 3 61 to 7 61
- NRC IE Report 50-352/
80-02, Page 5 62 $ 6 (Repeat) 61,64 NRC.IE ' Report 50-352/
81-16,-Page'4 63~ to8 65 NRC IE Report 50-352/
03, Page 12 64 go9 65 NRC IE Report'50-352/ .
79-11,.Pages 9 and 10 65 110 195 66 PECO QA Field Office Memorandum No. 882 dated 1/23/80 66 til 66 NRC IE Report 50-352/
81-16, Page 4 67 l06(Repeat) 67 NRC IE Report.50-352/
180-12, Pages 17 and 18 68 ll2- -68 PECO Letter to NRC dated 9/26/80 69 19'3 68 NRC IE. Report 50-352/ '
81-04, Pages 11 and 12 70 t. ( 4 69' NRC IE' Report 50-352/
, 77-12, Pages 3, 4, and 5 71 (15 69 3 PECO Letter to NRC dated 12/9/77 '
72 (16 69 NRC IE Report 50-352/
80-20 (Entire) 73 (17 235,236 70 NRC IE Report 50-352/ ,
81-12, Page 4 74 T l8 70 PECO Letter to-NRC dated 1/20/81 75 { l9 237 72,78 ,
- - NRC IE Report 50-352/
82-05,-Page 4 76 . 5 10 72 NRC IE Report 50-352/
01, Page 5
(~) 77 (11 245 73
(
PECO Letter to NRC
=.
Reference Appl. Appl. AWPP Te s t'imony Description .
Ref. No. Ex. No. No. Page e dated 3/12/81 78 (12 73 NRC IE Report 50-352/ .
80-21, Page 6 79 123 74 PECO Letter to NRC dated 7/17/84 80 114 75 PECO Letter to NRC dated 7/17/81 81 116 77 NRC Letter to PECO dated 8/27/81 82
- 126 77 NRC IE Report 50-352/
80-20 83 l l 7 (Repeat) 78 NRC IE Report 50-352/
82-05, Page 5 84 l17 78 PECO Letter to NRC dated 3/21/81 85 {28 80 NRC IE Report 50-352/
(')
o 81-12, Pages 6 and 7 86 12 9 80 PECO Letter to NRC dated 6/26/81 87 130, 246 81,84 NRC IE Report 50-352/
82-04,.Pages 3 and 4 88 131 82 NRC IE Report 50-352/ ,
81-06, Page 7 89 132 83 NRC IE Report 50-352/ i 82-03, Pages 3 and 4 90 133 85 PECO Letter to NRC dated 3/11/82' 91 134 85, 87, NRC IC Report 50-352/
82-10, Page 3 92 l35 86 ,
NRC IE Report 50-352/
82-03, Page 4 93 133 (Repeat) 86 -
i o