Document XnqmpVGbryYDmDJoYZGkbzK4

/ 1 tV,1 l (7 ^ -------- -____d!_____________ ** -v,. ,-: Estimating Hnaiih Risks in StudsssHof)tns A~~IA-63' ' Co Health Effaces or Asbestos' * PHILl? E. ENTERLINS institute: c? occupational AND EN'VinCN"ZN7AL HEALTH MONTREAL CANADA SU"MARY Euidr::::.-'!r>~;c .-U'-ii'.'s rjt.TA C-itLvf w**<ijt?:i c.;,r* i~. *.*>^ 00-*U I.1 '.fMrv *>.; c N.\r in`CIV i*'*i v; i ^ i`f Ai'ii.i'.f a:i '. '-Cu; * t magnitude of crxcitt. VOitlv Hunt *;tuiiL*> so m i<* Uc in j^ftviucnt 2> to '.He uA.v^ncc* s-f in e.t* ei, t`icrc i* considerable di'jjjrcentettt a* to i:* magnitude. for 11 studies. estimati, o; relative risk for rciTjiriiofi cancer run^f bom about 1.2 to 9.2. 1'inir (luturo ot thoc 11 suu-e.-s tin be idcriu- ficd tiiat accomit for nmc of tEtc luiialimi it: results.' seune studies include workers s!'.jse c.sao- *urc to asbestos wjj too relent to lie tiAcriy to produce cancer b> ti>e cud of ilia feliow-no and reiulu are diluted l>> tlie preJencv of these worker*: in most studies. the irraej population wai used for estimating expected numbers of deaths; itv touts studies. death certificates for ;h; itude population were corrected bawd on a ivview of oilier medical records. but then comparisons were made with expected deaths derived from uncorreeted death certificates; finally. io r.ior. siud- jt^( level and duration of exposure it unknown, and this it clearly a factor that determine* tits magnitude of the respirator* cancer risk. Fur 6 of the II studies it was possible from pub!is.'.ad material to correct results lou-hiy for the fust 3.of these factor* and much of the sirution dis appeared. Problems in the II studies reviewed here should tse contidered ia planning and reporting on epidemiologic investigations in the future. ' Asbestos represents an interesting environmental hazard in chat it is apparently sufficient for the production of one disease (cancer), and both necessary and sufficient for that of another (as* bestosis). That it, asbestosis is caused only by asbestos exposure, whereas cancer has many a uses, one of which is apparently asbestos ex* posure. In the instance of asbestosis, a single case represents the health risk, whereas for cancer, the risk can be measured only in relation to all other risks. This paper is confined to a discus sion of die problem of estimating cancer risks associated with asbestos, because it is here tiiat (Jitetiofti in orfgitMf firm July 25. 19TS end in re vised form Sovtntbtr 17,197J) 1 From the Department of Biostnistics. University of Pittsburgh, Pittsburgh. Pa. * Prevented at a joint meeting of the American and Canadian Thoracic Societies. Montreal, Canada. May I3.1SI3. there appear to be some methpJolugic difficul ties in work that has been done to date. When a number of agents contribute to a sin gle disease, as for respiratory cancer, it is cus tomary to measure the contribution of a single agent by calculating die ratio of the risk with the agent present to the risk with the agent ab sent. Tliis calculation is illustrated in table 1, where death is the measure of risk. The relative risk can be estimated by calculating the ratio of the probability of dying among persons exposed to die agent of interest to the probability of dying among those not so exposed. This can also be expressed as die ratio of an obsmed number of deaths in an exposed population to an ex pected number based on the experience its a non* exposed jsopulation. There i* also another way of measuring risk, anil that is the absolute risk, i.e.. the absolute difference between the two probabilities. Both ol these measures are of intetvst in studies of the health effects of asVwtus. In 11 studies in the United States and Canada. AMsntc.ss Kfvtm ot xt.srtx.vitiJtv mss s*r.. votuau. iu. Hto 175 gi <"i />,-, t AIA-63176 rifiut* t- (MlitLlM TASkc I METH003 rOS CALCULATING EXCESS mortality u\ a paos^ic?!''= study 0 Na: Dt ><j Tots* C' No; Te-al b cd *c b+d #*s> c+d N Pi * --M+D PQ - --S-. fislsrina risk " P| PO ASirrlur* risk - Il - ?o the risk ol developing respiratory cancer has been estimated in a prospective way among workers exposed to asbestos. These studies should provide some estimate not oftty of wheth er an excess in respiratory cancer exists hut of the magnitude of this excess. The latest pub lished results from these studies, summirued in table 2. are expressed as relative risks (1-14). Twelve studies art shown: however, my 1964 and 1967 reports relate to subgroupingj from a single large cohort. Alt show an excess; however, the magnitude varies widely. At least 4 features of die published data account for this variation in results and. in a more general sense, can be regarded as problems in epidemiologic studies. (I) Some studies included workers with a very low probability of developing asbestos-related cancer. These were workers whose exposure to asbestos was too recent to be likely to produce cancer by the end of die follow-up period. This can be labeled die latent period problem. (2) lii moit studies, the wrong population vsj UtCil /or estimating :i:c expected number of deaths, l.c., ihe probability (i'y) sbuwn in (able t of developing ctuixCr in the absence ol expo sure to asbestos. (1) In Muite 'indies, death ccrt.'frc.'ttes wee- cor rected based on a review of other medical rec ords: but comparisons them were made with ex pected deaths demon from uncorrected death certificates. . (V) In most studies, the level jntl duration of exposure were unknown, so that relative risks could not be related to levels and duration of cxi-ov.;; r. Exposure ir'-rit im-i tin rattan are clearly factors that determine the magnitude of the health risk associated with asbestos (19. 12). Tire latent period problem can probably best be demonstrated by looking at the experience of a population with a single exposure that is now known to be cancer-inducing, e.g.. the incidence of leukemia in the jsopulntion near the hypotenter of the atomic bomb exploded at Hiro shima and Nagasaki (figure 1). Leukemia did not appear immediately, and, in fact, until 19-13 few cases occurred. The peak year was 1931, after which the incidence declined. The underlying distribution of cases of leuke mia shown in figure 1 is probably the log normal distribution shown in figure 2. This illustrates how a single dose "of asbestos may work. Tor the exposure intensity represented here, little can cer would be observed in the first 6 years. The maximal response and the highest relative risk after the single exposure w*ould be observed be tween 13 and 24 yean. Whether the risk would, in fact, decline 23 years or more after cessation TABLE 2 RELATIVE RISKS POR RESPIRATORY CANCER 03S5RVS0 IN STUOISS Of WORKERS EXPOSED TO AS3SST0S Aa.'trarica OaM ' Pint 7ra**nt*d or SuViiStS Pooutatlon StwU.ad . 3RAUN artf Tniin (1) CUMH sr>d associate* (2.31 OUIX.V nU VVfir (3) MANCU30 an-J Coulttf Ul SELIKOrf and asiociatss (5.6) N7RL1N 7> r;rRLiN tej COORS3R anU SjIiv (91 MCDONALD ond astee>*ts 110.11) CNTSRLtN am) eo nork*>\ (1 ;i Y/ADO.*-R anil -.orkar It31 S6LIK2-8.8 I'lruuc iiwUU Sct.IXD.-r 4-sa a-.sociatai (1$) . toss >959 1953 1363 1954 1967 1959 1371 t972 1972 1972 1972 Minars Insulators Manufacturing Insulators Uinvbttu'iA; l-sutstor* .V.nars MjnatKIurtng Msnofseturirt} 1n*uj;ort RaSatlaa rusk 1.S 9.6 3.4 9.2 2.3 1.2. 1.3 7.3 1.4 2.7 2.3 8.2 5.o 01-0201567 * STt;U!ij Vf Hr. M. lit K.rs AIA-63^ fig. 1. Incitlciif.' of IcuktiiiU a*. Hirudiitna and Nagasaki. .-vn-'v.ire ij vvi;i-1 :: e; good epidemiologic observation, On this. it will be in thi* tiisuixsioi: that ilie king clear ance mechanism or selective [acton are o;>erative in such a svay as CO reduce (he risk approximately ai shown in figure 2. For a succession of doses-- as for workers ex]*>sed continuously to aibisto* (and in (lie absence of any interaction)--we might think of a succession of curves. The area to the left of the doited tine in figure 2 would probably be approximately as shown, whereas the area to the right probably would look, some what different. . The point is that the relative risk observed is partly dependent on when the exposure took place in relation to the period in which follow up studies were carried out. The studies shown in table 2 varied widely in this regard. For some, follow-up did not start until 20 yean after ex* posure began. For others. follow*tip began close to the time of first exposure to asbestos. Termi nation times and ages also, differed. Clearly, results should differ, and statements as to the magnitude of the cancer risk associated with asbestos cannot be made without reference to some latent period and termination times. Hg. 2. Rn.pii-tors cancer rnidtiii); Iron a single exposure to asbestos. Jig. 3. F.etpiratory cancer resulting from sin-it -j. bestos expomreat 1 levels. The problem is further complicated by the probable relationship between the intensity of exposure and the length of the latent period. This is illustrated in figure 3. which :hows a family of log normal curves corresponding to single-close exposures at various intensities. Four assumptions were made in constructing these curses: {/) response to asbestos exposure is_ linear, (?) an intensity of 20 fibers per ml corre sponds to a median latent period of 23 years, (J) response to a single dove has a log normal distribution with a geometric standard devia tion of 13. and (#) latent period is related to the inverse cube not of dose. These assumptions are based on papers by Jones and Crindon (13), Albert and Altshuler (16). and available data on asbestos-exposed populations. From these data, the effects of applying various starting and stop ping rules in epidemiologic studies can be esti mated. A requirement for entry into a study of 20 yean since first exposure would ensure rais ing nearly one half of the cases of respiratory cancer resulting from a very high exposure. Stop ping fotiow-up at any point because of age or time limitations would also provide a mislead ing picture, a picture with a different kir.d ot bias for each level of exposure. Obviously, follow-up mm he stopped at some point, if only because all of the study population has died. This leasts to till interestin' notion that oven if the ict.iiisnship litmrrn ad-cstm and res^iratutv tauter i> lineal there is a kind id threshold litni' value, defined as d..t duve for which the latent peril's! exceed* the hfe span of min (15). From this, it w-ouls! appear that when expo sure lesels air uncertain. ej.M'.-mij'o^ic studies Oi-020 1568 i 173 *rH!Lir E. IMUUNE AIA-63 ihould probably include (hi experience of if! workers. regardless of the tiuird.;?r of years since fir-.; exposure: follow-up should. continue tor a \rry long time, and results diould he reported in relation to the number of years since first ex posure. The second problem, that of using the strong population for estimating llie expected numbers o; deaths, is iiiustrated by the data shown in table 3. derived from a study of retired asbestos tvorVers (12). This shows the risk of dying from respiratory cancer M> ye:;rs after retirement duei;:g2 time periods. 1^-! 1 to !`JJ7 a.rti I'jaS to iStiffThe risk of dying was neatly 3 times greater after 1957 titan before 1957. There is no explanation for this insofar as the intensity of duration of ex posure to asbestos is concerned. The reason seems to lie simply in what was happening in the general population during this period, i.e., a 2-fold increase in respirators cancer mortality. Asbestos was apparently interacting with other causes of respiratory cancer, so that the effects of other factors increased the effect of asbestos. Some data from an early study by Selikoft and associates (5) seem to show this quite well, and are given in table 4. The men in this study had all lint been exposed 20 yean earlier on entry to the cohort, and for each of the 4 time periods shown Use population at risk was approximately the same size. Also, for men living in the 4 time periods shown, its unlikely that exposure levels differed very greatly. One of cite tilings that did change, however, was time, and with it. the effects of factors other titan asbestos on incidence of respiratory cancer. Note that both observed and expected deatlts increased at the same rate (a 2-fold increase in each case). There was also a 2-fold increase in the absolute risk associated with respiratory oncer. The relative risk, how ever. was unchanged. During this period, time multiplied the mortality rate by a factor of 3; asbestos, by a factor of 7. The combined effect TA3L 3 RESPIRATORY CANCER OSATH3 AMO CEATH RATES AMONG MSN CS TO 7i YEARS Q? ACS Study P*OQd n-i-is37 1953-1949 No. of B-JIKl 8 37 Rub 1.000 PJMORI 2.3 6.S of time atttl asbestos, then, was approximately 21. i.e.. 0.3 expected deaths irt the period 1943 to 1947, and 13 observed deaths ::i the p-:r:ed it-i-l to 1952. Asbestos apparently lias the ability *o multiply the effects of other agents in producing respiratory cancer. This lias been observed by Berry and associates (17) with regard to asbestos and cigarette smoking. The interaction between asbestos and other factors that cause respiratory cancer probably applies to geographic as well as temporal varia tions. Titus, the absolute lung cancer risk, asso ciated with" asbestos exposure would, be higher in an area with a high lung cancer death rate than in an area with a low lung cancer death rate. The relative risk would probably be the same, however. In nearly all U.S. studies, the ex pected number of deaths from respiratory can cer was based on "the level of mortality that char acter) ted the entire United States, rather than the local area in which die asbestos workers lived. The result was that the relative risk ob served was directly related to the local death rats for respiratory cancer. This relationship is shown in figure 4 for 6 studies in which men whose first exposure occurred 0 years previously were fol lowed, ie.. where the latent periods are roughly comparable on entry to the study. As might be expected, cists shows that if asbestos interacts with other factors that cause lung cancer, and if local death rates are ignored in calculating expected deatlts. when the local death rate It . . 001329 TABLE * RELATIVE ANO ABSOLUTE RISKS FROM RESPIRATORY CANCER* Siwev hliod Osunnl IiSKWl Rim Absalu:* nitfc (ratt/t.QOO fitfTOni) 1943-1947 1948-1953 1953-1957 1953-1943 8 0.8 7.3 8 1.4 3.7 13 2.0 "8.5 18 2.4 7.5 2.8 3.3 3.5 7.8 *Ojii trom Stlilwll <n<J iuttliin <51. n* n er i -- !> ( rrto:i.i or jt.-u.ru utu-uo) A>.i,.ito> AIA-63 Fig. 4. Relative rhk Ter respiratory cancer and local death rates in 6 studies. high, the relative risk is hi'!*; when tit* local death rate is low, the calculated relative'risk is low. If local rates of respiratory cancer death, rather than the li.S. death rate. It ad been used in calculating the expected death;, variation in relative risk due to variation in local rfenh rates would have been eliminated. The third problem is that in non studies, cause-of-death statements on death certificates were corrected by checking hospital records, pathologic reports of biopsies and autopsies, and coroners' reports. In tome cases, data on respiratory cancer were published both as they were recorded on death certificates and as cor* rected. and these show that corrected death cer* tsficates produce larger numbers of respiratory cancer deaths than uncorrected death certificates. A comparable correction cannot be made, how ever, for calculating the expected death date As a result, some of the published comparisons between observed and expected deaths are also comparisons between corrected and uncorrected death certificates, and this inflates the relative rut The relative rivks. with the death certificate correction rtnaunl, are shown in table 5. The tabic includes data bun: G studies f ir which li tem periods v.i.-rtr roughly cnupar.ibir. For thr studies hy Sfliko,1: and ro-wnticc.-i. observed lurid cancrr deaths were tedutetl by tnuliiplsuv* by C.37*, b.iivt! u.t a report-."! ton'action fesr th-.-t; U.S. Insulation U'orkers Study, and pleural meu(hfltotna deaths by a factor o! U. hid. based on tv .studj by Ducic (!3. IDj. The firs: column of fig ures in table b dnsw, relative rides a, c.diuiu'.eci from published rejiorts. The second column shows the relative rhks that would have been caicui-.itevl if t! :ith. ccvtificatss had not been tar t , v ,rll ill* I wj a * ''ll fl ,-* * i , 1 iV \ * ', > . I * , * ' 1 I * ^ risks that would have been observed if. in addi tion, local, rather titan national, death rates had been used in calculating expected deaths. These adjustments bring the findings from the various studies much clover together. Finally, some of the variation hi the relative risk, for respiratory cancrr among studies is un doubtedly due to dillering exposure levels. As shown in figure 3. studies that include only workers with 2t) years since first exposure would result in higher relative risks than studies in which workers were included with shorter titr.es since first exposure, owing to fulfillment of latent period requirements--particularly at lower dose levels. In addition, however, relative risks would be higher because of die longer period of expo sure and, thus, the larger dose, latensity of ex posure appears to make an additional but inde pendent contribution. The latter is illustrated in table A These data are based on the retired pop ulation of asbestos workers mentioned earlier; relative risks are shown by number of years since first exposure at time of entry to the cohort (time of retirement) and average exposure levels dur ing work. Both factors clearly contributed to the respiratory cancer risk. Actually, 3, rather than 2, factors were probably operating here: latent pc TABLE 5 RELATIVE RISKS AND CORRECTED RELATIVE RISKS FOR RESPIRATORY CANCER FOR 6 STUOIES. 20 YEARS AT ENTRY SINCE FIRST EXPOSURE n*t*rwc* RfUttv* ntv* D*jtw C irtifisva Co'fscnd Oi*tl Rat* Ca"*et*e EnlrJi'w CsS (91 Wi{j" U3) S<siatt liat S**;voff 115) 9.2 7.S 3.0 n.o 7.3 . 2-9 3.3 2.8 3.3 88 S.3 J.7 8.2 2.9 8.7 3.8 8.5 AT 0l-n*)n AIA-63130 t'UILU* t. >.\ 1 y.SLI>i TA3LS S P.ibATl'/S RISK POS R:S?!SATORV CA.-:C = F? A.VOMC RETIRED ASBESTOS V.-O.^FRS 7 .T># Sine* Piftt SX?9W<8 {/ f ini OuM tl < TO r'os'tf > 10 riojsl < :: 2v-29 L 30 1.2 3.1 1.S 3.6 2.S 4.7 r.'O'i. dose expressed us tints, and dose expressed u> t!u;i!>t:y. Thus tar, the effects o the first 2 hiss htcn intos.jiib'.s to separate. rie/sraneas !. Biauu. D. C., and Truar., T, D.: An cpidetniological study of lung cancer in asbestos miners, AMA Arch Ind Health, 1953. 17,634. 2. Dunn, J. E.. Linden, C,, and 3 reslow, L.: Lung cancrr mortality of men in certain occupations hi California, Ain j Public Health, 19(50, SO. 1475. 3. Dunn, J. E.. and Weir, J. .*1.: Cancer experience of several occupational grottos folioweti prospec tively. Am J Public Health, 1965, SS. 1367. 4. Mane-so. T, F., anti Coulter. E, J.: O.Iho.iolojy in industrial health studies. Arch Environ Health, 1953,6,210. 3, Se'.ikoff, I. J,, Churg. J- and Hammond. E. C.t Ashesio* exposure and neoplasia. JAMA, 1954, 7*3,22. 6. Srlikoff. I. J- Hammond, . C., and Seidman. K.t Cancer risk of insulation wotkers in the Untied States, in PrenriMji of 7.4TIC llorlmj Croup to Review the Biological Effects of Asbestos, Lyon, France, 1972. International Agency tor Research on Cancer. Lyon, 1973. p- 209. 7. Enterline, P, E.t Mortality among asbestos prod ucts worker* in the United States. Anti NY Acad Sci. 1965.773. 156. 8. Enterline. P. E, and Kendrick. M. A.: Asbestosdust exposures at various levels and mortality. Arch Environ Health, 1957. IS. 131. 9. Cooper, \V. C,, and Balter, j. L,: Evaluation and control of asbestos exposures in the insulating trade. In Proceedings of a Working Conference on the Itiotogieet Effects of Albertos, Dresden, Cermany, 1963- 10. Me Undid, J. C.. Ms Donald. A. I).. Cdihs. C. W.. SicmiatW-i, J.. and KrwMttc, C. E.: Mortality i:t li e ch.'inKi!.' j-lieito. mines -mil mills nf bvC, Ault Kosriuu Healtli. 1971, 72. 577. 11. McDonald. J. C.. Bec'-lake. M. P... Cibto. Ci. V.\. McDonald. A. I.. an-.l Hsw-iter. C. E,: The licalili of chrssotilf asbesto- mine and mil! work ers nt Quebec, Arch Enwi.m Health. 1974. IS. 5). 12. Eiitciiinr, P. E.. Drcoulle. ant! Henderson. V.: Kcspiratory cancer i:i relation to ocuipat'.er.al txpo,i:re among retired asbestos workers. Hr j Ind Med, 1973, SO. 162. 13. Wagoner. J. K.. Johnson. W. M.. and Lumen. TL: .Vihgrra;;; ami ston.ttiil.'gr.-tit r---j..i:ra: ca>s mortality patterns among a.bestui prad'aetioti workers, in Congi eniouat Tlecord-Senate Proceedings and Debates of tilt 9int Congress, First Session, sol. 119, pt. 5, March 14. 1973. Uj. Gov't. Printing Office, Washington. D. C, p, 7333. 14. Selikort, I. J.. Hammond. E. C,, and Churg. J.: Carcinogenicity of antosite asbestos. Arch Envi ron Health, 1972. 23, IS}. 15. Jones. H. B., and Grindon, A.: Environmental factors in the origin of cancer arid estimation of the possible haaard to man, fed Cotmec Toxical. 1975.12.251. . 16. Albert, R. .. and Altshuler. B.: Considerations relating to the formulation of limits for unavoid able population exposures to environment}} car cinogens. in Radionuclide Carcinogenesis, J. . Ballou. R. K. B-isch. D.-U. Maltlutn.'and C L. Sanders, ed.. AEC Symposium Scries. CONF72050. NT1S. Springfield. Va,, 1973. p. 233. 17. Berry. C,, New!iou*e. M. t~. and Turok, Ma Combined effects of asbestos exposure and smok ing on mortality from lung cancer in factory workers. Lancet. 1972,2,276. IS. Rossiter. C. E.: Discussion summary, in Proceed* ingt of the IARC Working Croup ro Review the Biological Effects of Asbestos, Lyon. Frame, 1973, International Agency, for Research ou Cancer. Lyon, 1973. ' 19. Dueic. S.: L'cxactitude des causes de tlecis: une comparaison asec les diagno>tiques a lautoptie dans une serie de mesothcliomes ct autres tunienrs maligncs du pcntntoti. Can J Pub Health. 1971,67.395. 001330 .S^ditos Issatluiic: fiDrth Jtasriu 13-:'* L StfS-Ji, 'I. V/. W2s!`;nt:r'., D. C. 20323 Ol 020l57i