Ozone : A review of recent experimental , clinical and epidemiological evidence , with notes on causation Part 2

The first section of this concluding part reviews epidemiological studies relating to oxidant exposures. There are convincing data from the eastern section of the continent that ozone, in combination with aerosol sulphates and acid aerosols, is strongly associated with hospital emergency visits and in-patient admissions for acute respiratory disease, including asthma. There is some evidence that long term exposures may be associated with an increased prevalence of asthma; disturbing initial observations suggest the possibility that a chronic respiratory bronchiolitis may be induced. The second section of this article reviews the totality of information on ozone in the light of customary criteria of ‘causality’. It is concluded that the present evidence justifies the conclusion that exist ing levels of tropospheric ozone are aggravating existing human disease, and probably increasing the prevalence and incidence of some diseases; however, it is uncertain whether this effect is being magnified by the concomitant presence of other pollutants. Although a great deal is known about the effects of ozone, there are still many important unresolved questions in relation to it.

JI scrnbk 4ue des ex positio ns a illng tcrmc soicn t assoc i6.:s a unc augme nt ati on de la preva lence de J'asthme: de$ obse rvat ions initia les tro ublantcs la issc nt supposer qu 'unc attc inle ch roni4 ue cks bronc hio les respirato ircs pou rra it ctre causfr par ccs e xposi tio ns.La deuxicmc section de cettc art ic le passc en revue la tota li tc des informati o ns sur !\none en tenant com pt e du c rite n: hab ituc l de causa li te .O n concl ut 4 ue Jes don necs ac tuelks appo rtcnt la prcuve q ue Jes n iv eaux d ' ozone co nt enus cla ns la lro posphi: n: aggravent !cs m alad ie s c xista ntcs chcz l'homme c l acc ro isscnt la prevalence c t !' inc ide nce de certa incs malad ies: ccpc ndan t. o n nc ~ail pas si eel cffet esl rc nfo rL•e par la presence co ncornitantc cl" autrcs po lluants.Bien q uc nous d isposions d "unc quantile ap pr.:ciablc d ' info rmations s ur les c lTcts de !" ozone.un bon no rnbre de ques tion s impo na ntc s co nccrna nt ces c !Tc ts nc son1 enco re pas rcsolucs ;1 J' heurc aclue ll c.

EPIDEMIOLOGICAL STUDIES Field studies and panel studies
Studies of forced e xpiratory volume in l s (FEY 1) in childre n at summe r camps in Canada and the Uni ted States have indicated that the summer mix of pollutants, including ozone, is associated with measurable declines in func tion.Kinney et al ( I) compared data from these stud ies and showed that the re was reasonable conformity among rhe m.Adul ts jogging in the ir lunch hour were studied by Spek ror and colleagues (2).Many of their subjects were found ro ha ve ven til ation leve ls in cxccss of 60 L/min (thcy had selected their own e xercise leve l).C hanges in forced vital capacity (FYC) and FEY t were related to the ambient nmne on the day o f their jogging .
Lioy ct al (3) found evidence that the effect of ozone on peak expiratory flow rate may last for severa l days.In a recent summer camp protocol , Spektor and colleagues (4) studied 46 c hildre n aged e ight to 14 years at a camp for a four-week period.The hi ghest ozone e ncountered was 150 ppb.They noted that regressions ol morn ing fu nction on the prcvious day's ozone were significan t.No correl ations were observed for hydrogen ion (in pre vious studies it had proved difficult to be sure that the response mi ght not be d ue to accompa nying acid aerosols).These authors concluded that camp studies generall y yield highe r e ffects of ozone than do chamber exposures, fo r fo ur possible reasons: first.the ex posu res in the fie ld are longer; second.there may be potcntiation by other fac tors; third , there may be pe rsis tence of effects frnm the previous day's exposu re; and fourth , there m ay be persiste nce of a transient response associated with the daily peak ex posure.
The authors cau ti oned that '"It follows that project ions of likl'ly effect s in the real world from controlled chamber studies should either have a large margi n of safety, or the judgement of the e xtent of effects li ke ly to occ ur amo ng populations should be based directly on the e ffects observed in field studies' '.Lebowitz and colleagues (5) studied panels of ast hmatics and others in Tucson, Arizona drawn from 117 white fam ilies from a stratified sample.Detailed meas urements of air pollutants.pollen, bac illi, fungi and algae were made in and around a random cluste r sample of 41 house ho lds.Thirtyfive asthmatics provided daily peak flow measurement s.The very complex resulting data showed that both ozone levels and te mperature had effects on the peak flo w data.Javi tz ct al (6) reanalyzed a study of 286 subjects with asthma and chronic bronchitis (probably with and without accompany ing emphysema) in Houston .Texas.Da ily symptom diaries were kept, and home spirometry testi ng was clone in about onethird of the subjects.Respiratory symptoms and medication use increased as the dail y ozone maxima increased.In the same city, Holguin and colleagues (7) conducted an ast hma panel study in 1985.They carefully de fined an •asth matic attack ' and fo llowed thc panel for six mont hs.O zone le vels ranged up to 0.27 ppm.T hey showed that an inc rease in the l h maximum oz,me of 0.1 ppm increased the attack probability from a base line value of I 0% to 16% .T his 60% inc rease in risk was four times that o bserved by W hittemore a nd Korn (8) , who publis hed the ir analysi s of a pane l study of asthmat ics in Los Angeles in 1980.Although by con tc m -ponH-y criteri a their cases were not c haracte ri zed in sufficient detail , they fo und convincing ev ide nce that asthma was worsened by both ozone and particulate pol lution expos ure.
Gong (9) studied 83 asthmat ics in Los Angeles bctween February and December 1983.He reported that 39 of the subjects were ' ozone responde rs•, wi th symptoms a nd med icati on use cons istently re lated to ozone leve ls.alt hough the total group resul ts had not shown a s ign ificant relati onshi p.The cases were relatively mi ld , s ince 77 % of the total group had peak flo w measure me nts within the no rmal range.
In the early 1980s.Lin n and colleagues ( I 0) carried out an extens ive seri es o f studies in Los Angeles using a mobile van that contained an exposure chambe r and pulmonary function testing eq ui pme nt.Afte r an initial finding that s uggested that func tion decrements in the nat ura l e nvironment with photoc hem ica l po llutan ts was greater than that found in c hamber ex posures, they fi na ll y concluded that the observat ions of decreased fu nction were explicable on the basis of the ozone ex posure.Later studies of competit ive cycl ists in the same mobi le faci lity (11 , 12 ) showed that FEY 1 decli nes occurred at ozone levels as low as 0 .16ppm (a resu lt in accord with the controlled exposure data described above) .They conc luded that the decreme nt in fu nct io n in the Los Angeles atmosphere was att rib utable to the ozone present.
Cast illejos ct al ( 13) repo rted studies on a panel of nom1al chi ldren in Mex ico C ity. T his showed that the FEY 1 and the maxi mal midex pi rato ry llow (FEF25 -75 ) decrements we re strongly re lated to ozone k vc ls avcragt:d over 24 to 168 h be fore spirome try, but not to the ozone in the previous hour.The authors suggested th at th e resul ts m igh t re lkct an inflammatory response in the ai rways as opposed to an acute phys iological response.One-hour a verage ozone levels in the hour precedi ng the tests had ranged up to 0.287 ppm , with a mean of 0.099 ppm.

Hospital admissions or emergency visits
Sterling and colleagues ( 14) in 1966 and 1967 reported a pos iti ve assoc iat ion between oxidant le ve ls and ad mi ss ions of Bl ue Cross pat ie nt s to Los Ange les Hospitals w it h more than 100 beds .In 198 1. Richards ct al (1 5) reported findi ng no associat ion in the same city hc t ween emergenc y vi sits at a c hildre n 's hospit al fo r asthma and oxidant level s.The study period extended from A ugust 1979 to January 1980.which may have been too short.Goldsm ith el al ( 16) stud ied e mergency room vis its in four southern Californi a comm unities duri ng 1974 a nd 1975 .Maximum hourly averages of oxidants and temperature were associated with dai ly vi si ts in the hi gh oxidant area (Azusa).A signilicant association was de mo nstrated.In retrospec t, it is surpri sing that thi s fi nd ing did not tri gger follow-up e pide mi ological st udies .
In 1983, Bates and Sizto ( 17) re ported findi ng an associ ation between acute respiratory adm iss ions to 79 ac ute care hospita ls in sout hern O nt ario and ozone level s.A later and 1nrn-e detailed analy sis ind ud ing data on concom itant sul-phalc measurements was publishL'd in 1987 ( 18).Admissions on a given day were compared wilh !he mean admi ss ions for 1ha1 day of the week in the sanll' SL'ason or 1he same year to correct for clay of the week effects and long term trends .This consnvativc approach to the data confirmed the association bl'l WCL'.n summer ozone levels and hospital admissions for acute respiratory disease and showed that the association was as strong f"or sulphates as for o zone.It was shown that a thrna admissions for children 24 h aftn ozone had exceeded the Canadian objective or 0.08 ppm ror I h were consistently elevalt' .U compared with admissions on al I summer days ( ( l)).In this and si milar studies.it has proved difficult to scparale lhe effects of these pollutants.w hich, in this region, commonly rise and fall together (20) .T he association of hospital admi ss ions with sulphate levels in southe rn Ontario was rnnfirrrn:d by Plagiannakos and Parker (21 ).They analyzed all acute respiratory admi ssions over a period of six years, and reporlcd sig nificant regress ion associations bdween hospital adm issions, e xpressed as a rate for each county, and both sulphur dioxide and sulphate .This study u •ed all the available sulphate and sulphur dioxide data ror the region and allemp1ed to ensure that di llercnces in socioeconomic status had not interfered with the analysis.An ind irect regional index of tohacco consumption was also used.T he importance of this study was that the two analyses in southern Ontario, though planned and crmduclcd indL'pcndcntly, served to reinforce each other in terms or the relationship between morbidity indices and sulphates.Unfmtunately, Pl agiannakos and Parker did not analyze ozone levels .
A significant association between ozone levels and resp iratory admissions was found by 01.kaynak and associates (22) in Massachusetts.Us ing hospital emergency visits rather than admissions as an index, Rates ct al (23 ) found no summer relationship between ozone levels in Vancouver and visits for acute respiratory conditions.Ozone le vels there wnc ahoul half what they were in southern O ntario.Since the summer period included Septeml>L•r.when a pe ak in asthma visi ts occurred that was unrelated lo pollutants whe n ozone levels were low, a summer relationship 10 oz.one might have been obscured; however.a reanalysis of the data for June, July and Aug ust only showed no relationship hetwce n o zone and asthma atte ndances.
Lipkr1 and Hammerstrom (24) rcana ly;,ed the southern Ontario data used by Bates and Si110 and extended them to 1985.T he y confirmed that significant associations occurred in the summer and showed that no changes occurred in admissions ror a group or nonrespiratory L•auses.They ca lculated that the pollutant mean dfccl might account ror I 9 to 24 % or summer respiratory admissions.
Burnell el al (2.'i) included all Ontario hospitals below the 47th parallel in their recent analysis or admi ss ions between I 98] and 1988 .To il1lprnw the e xposu re metric.they analyzed data from individual hospitals rather than aggregating the results by region .Omne showed a positive association with respiratory admissions in lJ I r1r of the 168 hospitals.S ix per cent of the sunmll'rtime respiratory admissions we re attributed to ozone and.surprisingly, they found that summer Can Respir J Vol 2 No 3 Fall 1995 pollutants accounted for I (1% of the hospilal admissions of infants.The relationship between daily 1naxil1lum I h 01.onc level.lagged one day , and daily respiratory admissions appeared 10 be monotonic, ris ing f"rom I 04 admissions when ozone was 20 ppb to I 13 admissions when uzrnll' was ahoul I 00 ppb (Fig ure I).Th is regression may fonn thL' has is for a more form al risk esti mate than has hern possible in the past.Tempe rature was shown to have no effect on the pollutionrespiratory relationship.
A recent more detailed analysis of data rrnm Toronto for the summers of 1987, 1988 and ll)89 was prepared by Th urston et al (26).Us ing si milar prefiilering method s to those used in an earlier paper, they showed that ozone was the summe rtime haze constitue nt of greatest importance to respirator-y and asthma admissi ons , but that elevated hydrogen ion was a possible 'polent iator' of thi s effect (Figure 2).A sensitivity anal ysis showed that if days whl'.n01.une was above 0.12 ppm (two out or a total of 177) were dropped from the ana lysis , the 01.one coefficients were unchanged .It was estimated !hat 19 to 20% of all summe r respiralory admissions were associated wilh air pollution , and that admiss ions rose by about a third above expected levels on the hi ghest ozone day, whe n the coneenlration was 0.159 ppm.Thurston et al (26) a lso showed that data from Buffalo and Toronlo, which expe rience almost ide ntical air pollution pallcrns.were essentially the same, in spite of different populations and patterns of medical care.
Cody et al ( 27) studied e me rgency room visits for asthma.bronchitis and fin ger wounds (as a control) al nine hospitals in central New Jersey.Data were collected for May to August 1988 and j l)89, and data on oz.one and sulphur dioxide were collected from the ncaresl of five monitoring sites.No associations were seen for bronchitis or for 1h1c' control diagnosis.Simultaneous regre ssions of asthma visits y ielded s ignificant positive coefficients for ozone and negativl' coe fficients ror temperature, suggesting that temperature acted as a long wave control variable.Day of thl' wed.effects were found 10 be unimportant.In an extension of these ob.,ervations,Weisel et al ( 28  emergency department visits for asthma in the same nine New Jersey hospitals.This showed that in every year, emergency visits occurred 28% more freq uently whe n ozone level• we re above 0.06 ppm than they did when ozone was less than th is value.T hey concluded that there was consistency in reported ozone effec ts in relation to the time lag between exposure and four different outcomes -symptom reports.decrements in expiratory flow, emerge ncy department vi its, and hospital admissions -and concluded that "T his supports a proposition that ozone adversely affects asthmatics at leve ls below the current US standard".White and ml leagues (29) at the Un ited States Centers fo r Disease Control and Prevention recently re ported a summer study conducted fro m .June to the end of August 1990 at a children ' s hospital in Atlanta, Georg ia.T his showed a close and highly significant association between attendances for acute asthma (n=609) and o,one levels.T he model yielded a 1.42 admission rate ratio for the number of asthma vi sits followin g clays with ozone eq ual to or exceeding a I h maxi mum of 0.11 ppm.No admission relationship with ozone was see n be low 0.11 ppm or with an 8 h average ozone va lue.The sulphuric acid aerosol levels at the Grady Hospital in Atlanta reach a maximum of 314 neq/111 3 in sum mer with an average value of 117 neq/m'.Sul phates at Georgia College of Technology in Atlanta average 273 neq/m 3 with a peak 24 h value of 535 neq/m 3 (30) .T hese values are not much different from those in the summer in New Jersey , and are slightly lower than values recorded in Toronto.Ozone was high ly correlated (r=0.6 1) with hydrogen ion level s in 164 Atlanta, just a.~ it was in the more northern sites.The ch ildren attendi ng the Grady Hospital were predominantly without any regular health coverage (sec be low).Local pollen counts were slightly higher on high ozone days, but were not significantly related to emergency visi ts .
Thurston et al (31) :malyzed eme rgency admissions to acute care hospitals in three New York metropolitan .u-easduri ng the summers or I 988 and 1989.The data were prefiltered using sine and cosine waves with annual periodicities.Ozone was sign ificantly re lated to acute respiratory admissions, the relationship being stronger in Buffalo and New York C ity than in the wealthier community of Albany.The high intercorrelations between ozone, hydrogen (measured as acid sulphuric acid aerosol every 24 h) and sulphates prevented a definit ive identifica tion of one of these as the causa l fac tor.Hospita l admissi ons for con trol conditions were unrelated to poll ution data.
T he val idity of using hospital based statistics has o ften been questioned (32 ).De lfino et al (33) n~centl y reported a val idation study of Q uebec hospita ls in terms of respiratory di ag noses (n=679).Alt ho ug h there was some discrepancy between the diagnosi s reported by the hospital to the computer file and that reached by an expe rt after perusal of the hospital records, in the case of asthma, the agreement was 94.9% and was 90% for all resp iratory diagnoses if minor differences in classification were ignored.Marti nez et al (34) also reported a validation study of emergency visits to hospitals in Barcelona.Many fac tors must be involved in rel at ion to the use of

Cross-sectional comparisons
Hodg kin ct al (36) fou nd some evidence of more respiratory symptoms in nonsmokers in higher oxidant re gions.but the difference was not g reat.Linn and colleagues (37 ) could not demonstrate consistent differences in either respiratory symptomatol ogy or function among populations of office workers in areas wi th different levels of oxidant pollution.Abbey and colleag ues (38.39) reported o n a continui n~ study of Seventh Day Adve nti sts resident in Ca lifornia oVL'r 25 ye ars of age who had resided I I years or mo re in areas with d ifferent levels of oxidant air polluti on.They found evide nce that worsening of asthma was associated with the annual average exceedance frequency of ozone thresholds of 0.1 () and 0.12 ppm.Although this study did not account for the possi ble confounding effect of fi ne particulate pollution.it provided significant evidence of worsening of asthma in those living in a highe r o xidant environment.
An inte resting comparison has been publi shed of child ren in Austria living in re latively hi g h and low o zone reg ions (40>.Two hundred and eighteen children (mean age 11.6 years) lived in the hi gh ozone region in which ozone exceeded l 00 ppb 9.68 % of the time, and 281 children lived in a low ozone region where that value was neve r exceeded .Othe r pollutants we re simil ar among the regions.T he re were no specific ex posure data for any of the childre n.T here were no differences in allerg ic background, as judged by immunoglohulin E levels or asthma symptoms, no di ffe rences in cough or breathlessness.nor in pulmonary function status.But there was a higher incidence of bronchial hyperresponsiveness to me thacho line in the hig h ozone group, and the re were differences in the ratios of types of lymphocytes.T he 'hig h ozone• childre n showed a decrease in T helper lymphocytes and an increase in T s uppressor eel Is compared with the c hildren in the low ozone area.Although the signi ficance of these hematological differences is unclear.the authors interpreted their overall find ings to indicate that ozone had caused a s ignificant di !Terence in status.
O ther epide miol og ical evidence of long term adverse effect s of ozone exposure has come from the analys is of the Can Respir J Vol 2 No 3 Fall 1995

Ozone
NHA NES II Survey data in thL' Un ited Stales provided hy Sch wa rtz (4 1).He rL' pnrted a dec rement in FVC in hoys in relation to calculated long tcnn ozone exposure; however the ana lys is did not account for poss ible concomi tant pollutants.Ostro and Rothschild (42) used data from the He alth Inte rview S urve y in the United States hetween 1976 and 198 l. and reported a relationship between ozone e xposure and minor red uced activity days.T his study confirmed earlier observ:1tions hy Portney and Mallahy (43).who noted a relations hip het ween ozone le vels and reduced acti vit y days hy crnnhining ae rometrie and surv(•y data.

Longitudinal studies
Detels a nd colleagues (44) planned an ambitious longi tudinal study of smoking and nonsmokin g subjects in different reg ions of Los Angeles.This showed a faster rate of long itudinal decline in more polluted regions when subjec ts \Wre retes ted after an interval of five years; howe ver. the re was a considerable loss of san1ple hetween the two observations.mak ing interpretation difficult.Bresn i11.and Rest (45), in a review of oxidant epidemiolog ical data , noted difficulties in the interpretation of the Un ive rsit y of California at Los Angeles (UCLA) longitudinal study.The question of whe the r the di fferences in rate of longitudinal decline could be attributed to sample loss is a difficult one.Lippmann (-+(1) in his review of the effects of ozone fell that the prel iminary ev idence from thi s study c ould no t he dismissed.A more rece nt prese ntation o f the UCLA results (47) concludes that the continuing fo llow-up confirms the fast er rate of decline of FEY I in more poll uted regions, and g ives reasons why the loss of about half the orig inal g roups stud ied initi a lly is unli kely to have affected the results.It may be noted tha t on theore tical grounds, long itudinal studies should be more sensitive indicato rs of long tem1 adverse e ffect s of oxidant pollutant exposure.but such studies arc wry difficult to organi ze. and it is difficult to a void loss to foll ow-up of a s ig nificant fract io n of the surveyed population.If this occurs , the interpretation of any resu lt becomes open to 4ucst io n.
Hackney and Linn (48) relentl y demonstrated that till' le vel of airway responsiveness changes with thL' seasons i11 Los Angeles, increas ing during and just after the high 01m1L' season and thereafter decl in ing.
Kilhurn e t al (49 ) in Los Ange le., conducted spirome try on 556 Mexican-Ame rican children in 1984.In 1987 , the investi gators conducted spirometry on 25 1 Mexican-American children, including I 06 of the childre n who had been measured in 1984.They standardized all pulmonary fu nc tion test values for growth by expressing the m as perce ntages of predicted values based on sex and he ight.In I 987. the mean values for FEY I and FEF25 -75 were lower by 4.5 % predicted and 13.6% predicted, respectively, compared with 1984 .Yita i capacities were not different (this is an important control for the test procedure).For the JOt, children tested on both occasions, FEY I was 2.0% lowe r and FEF25-75 7 .O'k lower.The re was no sig ni fi cant d iffe re nce between the I 987 mean value fo r 145 children tested forthe fi rst time in 1987 and that of the I06 retested children.T he authors concluded that the

Mortality data
It is remarkable that.a lthoug h Mahoney (50 ) in l l)7 l reported that respiratory mo rtality rates in Lo s A nge les seemeJ to follow the contours of oxidant pollution.until last yea r no Jd'initive study of the relationship between 01.om; leve ls and mortality haJ been conducted.K inney and Ozk aynak (5 l) recentl y p ubli shed a very careful time se ries anal ysis o f data from Los Angeles.It indicates clearly that 01.one le ve ls (lagged by one day) ,m; associated with increases in mortalit y (from both respiratory and cardiovascular disease ••).Their analysis by mode rn statist ica l tcchni4ucs indicates that o zo ne is the important factor.and that tempe ratu re changes are unlike ly to have interrcrcd with th e regression analyses .T h is study was .\upponed by a complementary one (52) of data fro m New York for the period l 971 to 1976.Long wave cycles in mort a lity were removed before the analysis.A mult ip le s imultaneous regress ion showed significant coeffici ents for one-day lagged o zone ( P=0.004) and for temperature and coefficient of haze and humidity on the same day.Exp ressed as an 'elasticity•.a I '11 increase above the me an ozone concentration was associated with a CLO l 9'1<: inc rease in mo rta lit ya fi g ure very similar lo that found lo app ly in Los Angeles.

Autopsy data
Sherwin and Richters (53 ) desc ribed changes of severe respiratory bronchiolitis in a high propo rtion (27%) or autopsy lungs !'mm 107 young adults aged 14 to 2:'i years who died in Los Angeles district from nonrespi ratory causes .Al thoug h thi., study has not yet been compared with a control reg io n and the degree of small airway disease h,ts not bee n meas ured hy qu ant itative morphometry.it seems likely that the lesion repre sents an ag (Travation of the respiratory bronchioliti s known to be a conse4ue nce of c igare tte smoki ng.It may be significant that.as no ted above.theore tic al dosimetric calc ulati ons have indicated that ozone depositi on in the human lung would be maximal in th is region of the lung, and data in 01one-exposcd no nhuman primates indicated that a respiratory bronchiolitis was the principal induced les ion.Adcsina c l :.ii (54) in Vermont qu;intified bronchiolar wall in flam ma-166 lion and fibros is in autopsy material from 42 smoke rs and 13 nonsmokers living in a relat ivel y nonpollutcd env ironment.They showed highly si g nifi cant d ifferences betwee n the tw o groups , pa11icularly in subjects below the age o r 40.The q uestion of whether li v ing in an oxidant environment results in a more severe deg ree of bronc hio lar inflammation.either in nonsmokers o r in smokers , sho uld he ans we rable if these data could he compared with similar data rrom Los Ange les.

THE BASIS OF A CAUSAL INFERENCE Bronchial hyperresponsiveness and asthma
A considerable volume o r wo rk over the past rive years has changed the perceptions or these two clinical conditio ns and their interre lati o nship: most im po rtant .the basic concept of asthma, based o n new understandi ng of its patho logy . is directly relevant to a dec ision about the possible ro le of Ollllll' ex pos ure.
In the no rma l po pulati o n. some individuals with no history or respiratory .,ym ptoms and no alopy arc found lo be hyperresponsive whe n cha llenged with mccholy l or histamine .The prcvalcncc o f this state in differe nt populations is not yer precisely kn ow n .It is known th a t bronchial hype rresponsiveness is uni fom1ly present in those who have clinical asthma and in those who reply pos itive ly to qucsti on11;1i rcs about asthma.Bronchial hyperrcspons ivcncss is also increased in those with w heezy ill nesses.It is also known that norma l individuals a rc fou nd wi th no sucl1 sym ptoms and wit h no hi story of asthma.but who arc hypcrrcsponsivc.In Ontario, Fitzgerald (55) found that in a random cluste r of 3 JO nine-year-o ld C anadian children.mild as thma was present in 8%, moderate as thma in 11 % and severe asthma in 3% .Thirty-five per cent or the c hildren we re found to have an enhanced me thacho line response.and o f these 11 % had no hi sto ry sugges tiv e or asthma .The fract ion of hype rresponsive ch ildre n was about doub le that fou nd by identical me thods in New Zealand.As prev io usl y noted , Hackney and Linn (48) in Los Angeles obse rved that the airway responsiveness in volunteers in that city varied between the high and low pollution seasons .
It see ms reasonable to assume that hyperrespo nsi vc indivi duals in the po pulatio n. eve n with negative respiratory questionnaire data, ma y be those at risk from exposure lo e nviron menta l agents that increase hyperresponsi veness Dr induce infl ammation in the lu ng.
Di ffe rrnces in asthma prevalence are diffi cult to evaluate because of the diffic ulty in dc!ining the condition and because the factors that may lead lo different answe rs to questionnaires arc poorly unde rstood.Halfon and Ncwacheck (35) anal yzed data from the 1988 National llcalth Interv ie w Survey on Ch ild Hea lth in the United S tates.which in vo lved 47 .485househo lds and 17,110 children.They no ted the prevalence data for d oc tor-di ag nosed asthma s how n in Table I.
T he prevalence of a pers istent wheeze in children 1:-hi gher, be ing foun d in 9.2% of 650 c hildren in Boston ('.)6).
If the question is asked about whec1.ing in respiratory illnesses, the percentage is inc reased again.
The prevalence of asth ma has been carefully documented in the Nordic countries o f Europe, where considerable effort has gone into standardizing the definitions (57).In Norway.est imates of prevalence have varied between 0.4 and 2.4%; in Sweden the prevalence estimates are gene ra lly higher.varying betw een 2.0% and 6.8 cfo in di fferent parts of the country and di ffe rent age g roups; in Finland.estimates va ry between 1.0% and 3.0%.ThL ' variation within Denmark is large, fro m a low of 0.9% in one general practice to a hi gh of 10.5 % in another.T he most re.cent estimate for Danish schoolchildren was 4. 0%.
With the use of standardi zed q uestions, the prevalence or as thma in children aged between six months and 11 years living in the United States differL•d according lo their ethn ic orig in (Tabk 2) (58).
Burrows (59) in a community survey in T ucson, Arizona found a prevalence of asthma of 8.2% in those between the ages of si x and 34 years; 8.6% in those be twee n 35 and 54 years; and 9.8 Po in those over 55 years o f age.ln a study in progress of 3600 children in Los Angeles in IO different communities.the prevalence of asth ma seems to be high ( 18% in some areas.personal comm unicati on).
There is some evidence that the prevalence of asthma has been increasing.From a compari son between two NHANES surve ys in the United States in 1971-1 974 and 1976-1980.Gergen et al (60) rnncl uded that the prevalence o f ast hma in six-to 11 --year-o ld s had increased from 4. 8% to 7.6%.In 1990 , Burney el al (61) in Britain compared data from 1973 with data from 1986 on 15.000 boys and 14, 156 g irl s and concluded that asthma prevalence had increased by 6.9% in boys and 12.8% in girls over that pe riod.There had al so been increases in those who had a ' persistent wheeze•.
There is no doubt that the rate of hos pi tal adm iss ions fo r a thma in all age groups has increased over the past IO years; it is unli kely that this reflects the admission of more patients with less severe degrees of asthma .Mao ct al (62) summari zed thi s data for Canada: it has also been shown to be true o f the Un ited States and in a number o f countries in Europe.
If the severity of asthma has not changed in those adm itted, then the increa •ed hospital admissions for asthma, which have been very genera lly noted , must indicate an increase in the severity of the d isease, an increase in preva lence or both.It is possible that a facto r might change the severity of the disease without affecting prevalence, or by increasing pre valence, might increase the number of relat ively severe cases.
There has been a stri ki ng change in the percept ions o f the pathology of as thma .Thi s cha nge in thinking began with the initial observations by Laitinen and colleagues (63 ), who in 1985 reported on bronchial bi opsies in eight nonsmoking asth matics .No ne had had a respi ratory in fect ion within the past two months.All had evide nce of epithelial destruction and in flam matory changes.T he authors noted that "Epithe lial destruction in the respi ratory trac t of the asthmatic with mild to severe bronchial hyperrespon:• iveness was prominent enough to expose the epit he lial ne rves fo r spec ific and no nspec ific stimuli".T hese observations have since been confirmed in a number of centres and were recentl y extended by

Possible outcomes from ozone exposure
From this condensed review of human and ani ma l data, it is clear that the fo llowing possible outcomes as a conse4urncc o r ozone exposure may reasonably be expected to occur: I) Increased airway responsiveness in the population; 2) Increased prevalence of asthma; 3) Increased severity of asthma attacks; 4) Increased incidence o f asthma attacks; 5) increased severity of respiratory infections: (1) Increased incidence of respiratory infect ions; 7) Increased prevalence of chron ic respiratory symptoms; 8) Development of chronic respi rato ry bronchiolitis .
In the follo wing sections.the strength of a causality inlerence will be assessed in relation to these eight outcomes.There is no direct evidence that ozone causes emphysema in animal s and no human evidence that it is re lated lo :111 increased incidence of lu ng cancer, although, as noted above, the properties of ozone and its biological actions ha ve suggested that this might be the case.T hese two outcome~ are therefore not specifically addressed.

Bradford Hill's suggested criteria
In I 965, Hi ll (66) considered the criteria by which we might be guided in reachi ng a concl us ion of causality .He li sted nine issues that he had fo und hel pful.Each of these is discussed in turn in relation to the e ffects of ozone.The 4uest ion of causality in relation to demonstrated associations is always complex, and this is particularly the case when the L 'xposure data are imprecise (67).Strength of association: For the e ight outrnmcs li.,ted above.one might conclude that the strength was reasonabl y strong for numbe rs 3) and 4).and that for I).2), 5).6) and 8 ) there were no data to enabl e the strength to be assessed.There is some evidence for 7).
It is important to note that the .,trrngth of the association might well have been weakened hy the gcnnally inexact exposure information.Also.because most o f the o utcomes in volve multifactorial di seases, strong associations with any single facto r are not to be ex pected.Consistency: Has the relationship been observed by d iffere nt persons, in di fferent places, circumstances and times?In relation to 3) and 4 ). the evidence in northeastern North America is consistent.but accompanying aeroso l ac id ity may have been contributing to it.That ozone is the im portan t pollu tant is ind icated in the New Jersey and Atlanta studies.Outcome 5) is probable but difficult to assess because the different iation from asthma may well be confused.No data L•xi st to evaluate I) or X).Specificity : As noted ahovL ', the relat ively strong data from the northeast may indicate that ozone accompan ied by ac id aerosol is more signi fica nt than ozone alone.Thert> are animal dat a to support th is.Temporality: The significant associations between ozone and hospital admissions or emergency visits only occur 24 to 48 h after the ozone level has ri sen (Figure 2 ).Biological gradient: Although the dose-response relati onship can be clearly demonstrated in controlled cl inica l exposures .it is more diffic ult to demonstrate in epidemiological studies .Prec ise dose-response calculations are difficult to calculate since ozone effec ts do not fo llow a concentration X time pattern; effects differ with seq uen ti;.i[ ex posures; and there arc many other determinants of each of the outcomes .However. the recent anal yses of Ontario data by Burnett et al (25) do permit an approximate dose-outcome relationship to be adduced.The absence of an e ffect of ozone on hospital emergency visits in Vancou ver (23 ) supports the vi ew that a dose-response effect exists .Plausibility: Ozone causes inflammation in the lung at very low concent ration s. and it persists for at least l 8 h after the exposure has ended.Hence.any of the poss ible outcomes must be considered hi ghly plausible.The new understandi ng of asthma as an essentially infl ammatory disease adds to the likelihood of outcomes 2).3) or 4).Increased asthma prevalence has not yet been demonstrated.alt ho ugh l ) has been shown to occur.There is strong sup port from dos imetric studies and from data fro m nonhuman primates that supports the pl ausibility or 8).The reported occurrence of chronic respi rato ry bronchiolitis in autopsy data requires con finnation.however.Although there are in terpretational difficulties in the case or the lon • 1 itudina l studies that have been per-fom1ed, both have shown evidence of a fas ter decli ne than no1111al in air flo w ra tes.Coherence: In the consideration ol' air pollutants.this criterion has spec ial dimensions (68).Alt hough ozone has been round tl, relate, on a time-seri es basis, to daily mortality.the mechani sm for the effect is not known : poss ibly it is the same as causes !"i nc partic ulate pollution to be related to mortality (69).Howe ver. a •cascade• of phenomena fro m mortali ty to increased hospital admi ssions and emergency vi sits for asthma and nonasthmatic ac ute respiratory disease as well as the association of ozone with •reduced activity days • combine to suggest that the coherence cri terion is met.
There is al so coherence in the concordance among theoretical dosimetric data.the cent riacinar lesions in long tem1 nonhuman pri mate and rat exposures.and the (only avai lable) human autopsy data from Lo -Ange les.Experiment: Under this heading.Hill suggested that , in the occupational arena.changed L •on dit ions might be observed to modify an outcome.Ir human ex pos ures to ozone have become more common over the past l 5 years.it is te mpti ng to sugges t th at thi s might have led to an increased prevalence of asthm a.In add it ion to the co nunonalit y of in fla mmat ion.pointed out above, th is woul d have biol og ical plaus ib il ity if ozone enhances the entry or al lergens into lung tissue.Analogy: Photoc hemical ox idant polluti on is distinct from former pattern s or pollut ion, consi sting of large particl es from coal bu rni ng together with high leve ls or sulphur dioxide; therefore.an anal ogy to thi s forn1 or pollution wou ld not be appropriate.Exposures to ozone in the wo1lplace are too intenn ittent for any conclusions to be drawn.

A JUDGEMENT OF CAUSALITY
The present data.taken together, appear to be strong enough to penn it the foll owing conclusions.Asthma: The agreement between the southern Ontario and Toronto data (now reworked by fo ur different investigators) and the supporting data from New York State and from New Jersey indicates beyond any doubt that higher ozone levels aggravate asthma and increase its severity.Thi s is supported by the observations on panels of asthmatics in Los Angeles and Houston.This has strong biological pl ausibility.Whether ozone exposure increases the prevalence of asthma is not known, but the effect of the gas on bronchial hype rrcsponsiveness (unmod ified by repetitive ex pos ures) suggest thi .\ possibility .Data from nonsmokers in Ca lifornia support the concl usion that living in a high ozone atmosphere may increase the likeli hood or as thma deve lop in g.It is poss ible that coex istent ac id aerosols heighten the e ffec t or owne; these were present in all eastern continenta l sites.includ ing Atlanta.In Los Angeles.there arc hi gh levels or nitric acid aerosol.The question or the interact ion bet wee n ozone and acid aerosols may be ans wered by an pide miological study of ozone and hospital admissions in !loll and (persona l communication), where acid aerosols arc not detectab le and ozone levels are about the same as in southern Ontario.
It should be concluded .on the basis of prL•sen t data.that ozone al existing tropospheric leve l~ is having an adverse effect on the 5% or so of the populat ion wit h asthma and may be increas ing rhe ai rway responsiveness in the popu lation as a whole.Balmes (70) recently reviewed the interaction of ozone with asthma and Ccui1e to thC' same general condusions .Respiratory infections : There is strong biolo~ical plausibility in the suggestion that.by impai ring lung de fences (and particularly by incapacitating the macrophage), ozone may increase the severi ty and possi bly the incidence of arntc res piratory in fections.Most of the data sets indi cate that ac ute respiratory admissions are signil"icantly associated \vith ozone when asthma has been excluded.Since bacterial pneumonias are much more common in the winter months when ozone is low, carefu l seasonal filt eri ng is necessa ry in .,tudy•ing thi s possibil ity.
On present in fo rmation.it shoul d be concluded that aggravat ion of respi ratory in fections is a likely scq ue la of rnrrent 01.one level ; aggra vation by acid aerosols is al so likely.Th is conclusion is reinforced by the observation that all acute respiratory admi ss ions .and not just a.,lhma, appear lo be influe nced by summer a ir pollutan ts. Thi s might be attributed to diagno.,tic miscl assi fication; ho wever, B urne tt's (25) recent observat ion in Ontario that 16% of the summer hospital adm issions of infan ts in that province are related to s umme r pollut ants re inforces th is possibility.
Long term effects: T he concorda nce bet ween theoretical dosimetric calculations and the site of the o bserved lesions in ani ma ls with low le vel long tenn expos ures is noted above.

Can Respir J Vol 2
No 3 Fall 1995    OzoneTABLE 2   Prevalence of asthma in children aged six months to 11 years according to ethnic origin in the United States Ethnic~------severe chronic inflammatory changes in severe asth matics and in fatal case.( 64) as well as in the fi nding of infl ammatory changes in the lu ngs of mild cases o f recent onset ( 65 ).
Fall 1995hospital eme rgency depa rtments or the cri teria for adm ission.Halfon and Newacheck (35) recently reported o n the 1988 National Health Interview Survey on Child Hea lt h in the Un ited States.They showed that poorer economic groups used pri vate doctors less and hospital eme rgency departments more in relation to acute exacerbations of asthma.Furthermore.the poorer g ro ups spent more days in hospital and suffered more morbidi ty than wealthier groups.Thus it follows that hospital emergency visits.as an outcome indicator.might be expected to be more frequent if the hospi tal serves lower income g roups .T his might explain the h igher associations found by Thurston in Buffalo betwee n o zone and hospital admiss ions than occurred in Albany.which is a wealthier community.In spite of thi s, of course.it must be assume d that the adverse impact of ozone takes no accou nt or economic status, e xce pt poss ibly in tenns of the protection conferred hy air conditioned homes.
Can Respir J Vol 2 No 3

TABLE 1
Percentage prevalence of asthma in the United States