Tobacco Documents Online

Liddell FDK. Exrwsura-response: asbestos and mn,wtbelioma. Eur Respir Rrv 1993; 3; 11. 98-99. (1991.I2.18; with addendum 1993.03.11) Exposure-Response: Asbestos and Mesothelioma F D K LiDDELL Departmenu of Epidemiology and Biostarisucr, MeCi11 Urriw.rsiry, Montreal, Canada Abstract In each cohort in wbich it can be examined satisfacsarily, the relationship between duration of asbestos exposm and the risk of inesothelioma is direct, and several are quite strong. It is thus clear tha+e aso very powerful exposure-resporLse relationships. There is good evidence that these relationships are sub-linear at very short durations of exposure and probably again at rather long durations of exposure. The likelihood is that the underlying relationship is sigmoid, regardless of fibre type. In their report to the Health and Safety Commission in 1985, Doll and Peto' stated that, although the predicted risk of mcsothelioma increased in approximate proportion to duration of asbestos exposure for exposures of up to about 10 years, there was very little difference between the predicted effects of stopping or contituiing exposure after 20 years. However, tlose examination of their data suggests a much more positive conclusion. Table I displays three models of risk. Not surprisingly, the over-simplified Model I is quite inadequate. The fit by Model 2, which has been used by Peto and cotivo:kers,= appears somewhat better, but the goodness-of-fit ;e ststistic (Sdf) is 2.+6.5, associated with a P-value of 0.00007. Model 3, ascribed to Peto and endorsed by Doll and Peto,' yields a corresponding statistic of 7.2, suggesting a vtxy good fit; the great improvement over Model 2 arose from the incorporation of tlse term for duration of exposure - see the last Line of Table 1. A formal test of the degree of improvement was carried out by fitting exponential models including 1) lapse and 2) lapse and duration; the second term decreased the Likelihood Ratio statistic by 9.8 (using I extra degree of freedom), an improvement of enormous statistical significanec, While it is true therz was no death from mesotfselioun among those with 30 or more years of service, tfse expected number of deaths even on Mode13 was so low the shortfall was quite insignificant; further, for men with 20-29 years of scheduled service, there were nearly twice as many cases as expected. Thus it seeaaS reasonable to claim that the risk of mcsotbelioma increased in approximate proportion to the duration of exposure up to at least 30 years. Table 1. - Three models of inesothelioaa risk in terms of subject-years, lapse, ands3uration of exposure. Model S risk « subject-years Model 2 risk K(lapse)3-2 x subjeet-years Model 3 risk -((lapse)4 -(lapse - duration)4] x subject-years which, except for rather long durations, can be taken as: risk K(duration) x(lapse)3 x subject-years or effectively Model 2 modified by a "linear" term for duration • . I

Table 2 lists all 12 cohort studies in which exatnitution of exposure-response is possible, giving the numbers of deaths from mesothelioma. A test for traud of risk in relation to duration of exposure has bew carried out where possible, and the relevant Xz statistic is included in Table 2. T6a seven unbracketed statistics ara associated with P-values betweea 0.051 and 0.0005; even using the most conservative approach to the test of the overall null hypothesis. and taking ail four bracSceted statistics as zsro, yields an extremely low P-value. Table 2. - Cohorts of asbestos workers with deaths from mesothelioma and risks in relation to duration of exposure No of deaths Test for trend * t Tumour-free Zu•mours duration $ expected S Crocidolite Mining (Wittenoom, WA)3 31 (5.0) 90 days -/ Mining (Cape, RSA)' 20 5.6 12 months 2.4 Gas mask assembly (Nottingham, UX)s 16 11.6 5 months 1.5 unosit:e Insulation products manufacture ([JS)6 14 9.7 5 months 0.9 Mining (Transvaal, RSA)4 4 (2.0) 3 months 0.5 'H3JCt t7reS' Mining (amphiboles) (RSA)4 6 (2.5) 3 months - / Factory (Barking, UK)7 ' Males 60 15.4 I No relevant data Females 25 (0.3) No relevant data Asbestos--ceaent plant 2 (New Orleans, US)s > 75 t its pipe aiarea 6 3.8 4 months -~ 5 75 t in pipe area 2 6.0 2S years 0.1 Textile factory (Rochdale, iTPC) ' 10 9.8 1 year 0.7 ** ChrysotZZe Mining (Thetford Mines, FQ)' 7 (1.4) 6 months -~ * x= statistic with 1 degree of freedom; but see 1. t Figures in brackets are unreliable due to small numbers or other technical reasons. t There were no mesotheliomas among persons exposed for these periods; see text. S Number of taesotheliotuas expected fraa linear trend. ~ Number of inesotheliomas expected frcxa linear trend cannot be calculated. i 2 degrees of freedom. ** Calculated from Model 3. In each of the tat cohorts where adequate infotatatioa was pr+ovided, tlsere was a duration of exposztte, as shost as 3 months or as long u?3 years, which was 'tumotts frx•; see Table 2. No mesotheliotm'bas boea reportod among persons employed for such short periods. '13te final column of Table 2 gives the numbers of aoesotbeiioaias that would have been expected in these pe:iods on the basis of proportionality. All these numbers are of course sma11, but diey total 6.1. Sub-linear response to short durations of exposure may, at worst, be taken as hypothesized by McDonald and McI?onald." Excluding the findings"-' from their Table 4 provides a legitimate test: observing zero where 3.7 were expected yields P- 0.0247. From this, it tnight be argued that exposure for at least 3 months is a necessary condition for the development of asbestos-related tnesothelioma. This concept of threshold cannot yet be sustained on the evidence available: what can be claimed is that, for short durations, theexposure-rr.spotue relationship is sub-linear. Toxicological evidence would suggest a sigmoid relationship, with the point of inflexion somewhere near the EDs,. However, .-L~ - 2

the findings in man are for such low 'doses' that the risk might have been thought to increase more-or-less exponentially - but this is certainly not the epidemiological finding. Finally, the evidence of Browne and Smither" must be mentioned. At first sight, it runs bounter to the findings discussed here. However these authors' results cannot be evaluated because no 'denominators' are available; it is entirely possible that the distribution, by duration of employment, of all Cape Asbestos employees over the years were so highly skewed that there is no real contradiction with the present findings. Two further faCtots must be borne in mind: the Cape Asbestos experience dates back many decades when intensities of exposure were unthinkably high; and a substantial proportion of 'short-service cases' may in fact have arisen from neighbourhood exposure, which was known to have been far from negligible near the Barking plant in which the great majority of the cases had worked. Addendum - 1993.03.11 It has recently been reported'= that, among 10,926 Quebec asbestos wOrkers bom 1891-1920 and follow+ed to the end of 1988, there have beea 33 suspected mesotlseliomas in all; of these, eight and 20 were in miners and millers from Asbestos and from the Thetford Mines region, respectively, and five were among men employed in a small asbestos products factory in Asbestos. There were no aasa of mesothelioma among the 4438 taun employed for less than tweyeara, eight c ases among those 2448 employed for 2-10 years, and 25 mesotheliomas among tbe 4040 mea with at least 10 years employment. This information supersedes that in the last line of Table 2, so that there have now been twelve cohorts with a'tumour-free' duration, and the numbers expected in these periods now tota17.3. Again excluding the findings from references 5. and 6., the legitimate test becomes of zero observed, where 4.9 were expected, and yields P= 0.007. Referexxs 1. Doll R, Peto J. Effects on Health of E:cposure to.lshestos. Health and Safety Commission. London: HMSO, pp 33-40. 2. Peto J, Seidman H, Selikoff U. Mesotheliotna mortality in asbestos workers: implications for models of rarcinogenesis and risk assessment. Br 1 Cancer 1982; 45: 124-135. 3. De Kleric NH, Armstrong BK, Musk AW, Hobbs MST. Cancex mortality in relation to measures of oaxipational exposure to crocidolite at Wittenoom Gorge. Br 1 lnd Med 1989; 46: 529-536. 4. Sluis-Cremer GK, Liddell FDK, Logan WPR, Bezuidenhout BN. The mortality of amphibole miners in South Africa, 1946-80. Br .l Ind Med 1992; 49 : 566-575. 5. Jones JSP, Pooley FD, Sawle GW, Smith PG, Berry G, Wignall BK, Aggarwal A. The consequences of exposure to asbestos dust in a war-time gas mask factory. In: Wagaer IC (Ed). BFoJogfcctl,Efferts of Mineral i+tbres. IARC Scientific Publications 30. Lyon: International Agency for Resarch on Cancer, 1980, pp 637-653. 6. Seidman H, Selikoff 1J, Gelb SK. Mortality experience of amosite factory workers: dose-response relationships 5 to 40 years after onset of short-term work exposure. Atn ! Ind Med 1986; 10: 479. 7. Newhouse ML, Berry G, Wagner JC. Mortality of factory workets in east London 1933-80. Br 1 Ind Med 1985; 42 5-11. ' 8. Hughes JM, Weill H, Hanuttad YY. Mortality of workers employed in two asbestos cement manufacturing plants. Br.7 Ind Med 1987; 44: 161-174. ' 9. Liddell D. Epidemiological observations on mesothelioma and their implications for non-occupational exposure to asbestos. In: Spengler JD, Ur3csynak H, McCarthy JF, Lee H (Eds). Symposium on Health Ueats ofF.xposure to Asbestos in Buildings, Decrmber 14-15,1Sa&3 Cambridge MA: Harvard University Energy and Environmental Policy Center 1989, pp 47-69. 10. McDonald JC, McDonald AD. Epidemiology of Mesothelioma. In: L.iddell D, Miller K (Eds). Mineral Fibers and Heulth. Boca Raton, Florida: CRC Press 1991, pp 147-168. 11. Browrse K, Smither WJ. Asbestos-related mesotheliocm: factors discriminating between pleural and peritoneal sites. Br11nd Med 1983; 40: 145-152. 12. McDonald AD, Liddell FDK, McDonald JC. Malignant mesothelioma in Quebec chrysotile miners and milless: a preliminary report. In: Proceedings of the 9th Intrrnational Symposium on Epidemiology in Occupationa! Health, Cincinnati, Ohio, September, 1992. In press. 3