Document NG7R2o7OjxoXKjZyrOL7yzd4g

Letters to the editor 329 100 down to 0.005 f/ml years. But the proof of the pudding is in the eating, and nowhere do they subject their calculations and guesses to the test of reality. They neither acknowledge those who have done so, e.g. Liddell (1991) and Cantus et at. (1998), and have found the earlier estimates, based on the linear doseresponse found at higher exposures, to be inappli cable, nor those who have adopted the alternative approach based on intensity rather than simple cumu lative exposure (Liddell et at., 1998; Vacelc and McDonald, 1990), which provides other corrective information on the lower ranges of the doseresponse curve. The first half of the paper contains many good points and one or two gems, but the second half, sadly, adds nothing to our understanding of the risks of low level asbestos exposure, providing risk esti- mates that have no sound basis and that do not match up with reality. KEVIN BROWNE 2 Burnham Road, North Creake, Norfolk NR2\ 9JP, UK REFERENCES Acheson ED, Gardner MJ. Pippard EC, Grime LP. Mortality of two groups of women who manufactured gas masks from chrysotile and crocidolile asbestos: a 40 year fotlow-up. Br J Ind Med 1982^39:344--8. Camus M, Siemiatycki J, Meek B. Nonoccupationai exposure to chrysotile asbestos and the risk of lung cancer. N Engl J Med 1998;338:1565-71. Dolt R, Peto ]. Effects on health of exposure to asbestos. Lon don: HMSO, 1985, Dufrcsne A, Harrigan M, Masse S, Begin R. Fibers in lung tissues of mesothelioma cases among miners and millers of the township of Asbestos. Quebec. Am J Ind Med 1995;27:581-92. Duffesne A, Begin R, Churg A, Masse S, Mineral fibre content of lungs in patients with mesothelioma seeking compen sation in Quebec. Am J Respir Crit Care Med 1996;153:711-8. Gardner MJ, Powell CA. Mortality of asbestos-cement workers using almost exclusively chrysotile fibre. J Soc Occup Med 1986;36:124-6. Hodgson JT, Damton A. The quantitative risks of mesotheli oma agi lung cancer in relation to asbestos exposure. Ann Occup Hyg 2000;44(8):565-60l. Liddell F. Exposure-response; asbestos and mesothelioma. Eur Respir Rev 1991&3(12&Q3);18&11. Liddell FDK, Hanley JA. Relations between asbestos exposure and lung cancer SMRs in occupational cohort studies. Br J Ind Med 1985;42:389-96. Liddell F, McDonald A, McDonald J. Dust exposure and lung cancer in Quebec miners and millers. Ann Occup Hyg 1998;42:7-20. McDonald A, Case B, Churg A, Dufresne A, Gibbs G, Sebasticn P. et al. Mesothelioma in Quebec miners and millets: epidemiology and aetiology. Ann Occup Hyg 1997;41:707-19. Meldrum M. Review of Fibre Toxicology. Sudbury; HSE Books, 1996. Nicholson W. Airborne asbestos health assessment update. ' EPA, 1985. ' Rees D. Myers J, Goodman E. Blignaut C, Chapman R, Bachmann M. Case-control study of mesothelioma in South Africa Am J Ind Med 1999:35:213-22. Sliris-Cremer GK, Hnizdo E DuToit RSJ. Evidence for an amphibole asbestos threshold exposure to asbestosis assessed by autopsy in South African asbestos miners. Ann Occup Hyg 1990:34:443-51, Vacefc PM, McDonald JC. Effect of intensity in asbestos cohort exposure-response analyses. In: Sakurai H, editor. Occu pational epidemiology. Elsevier Science Publishers; 1990. p. 189-93; Weiss W, Mortality of a cohort exposed to chrysotile asbestos. .J Occup Med 1977;19:737-40. Asbestos and Cancer (5) NOTICE TW|S MATERIAL MAY BE PROTECTED 3Y COPYRIGHT" LAW (TITLE 17 US- CODE) PII: S0003-487B(01)00030-8 Hodgson and Darnton (2000)--referred to below as H&D--have done a great service, if only for presenting mortality rates for cohorts of asbestos workers in a way that facilitates close examination of what has been called `the fibre gradient1 or the `amphibole hypothesis', FIBRE TYPES The three principal types of asbestos (bearing the same generic name but alike only in being fibrous silicates) are chrysotile, which has always accounted for at least 90% of commercial usage, and two amphiboles, crocidolite, the more commonly used, and amosite, the amphiboles having;-chemical consti tutions quite disparate from that of chrysotile. The (microscopic) respirable forms of the three types also differ greatly not only in shapes and sizes, so affect ing penetration, but also in their durability in lung and other tissue. Markedly different health effects were therefore only to be expected. Largely because of differences in their (macroscopic) physical characteristics, the three types of fibre have had varied commercial uses, but in most industrial practices workers exposed to amphibole asbestos have also been exposed to chrysotile. Such exposures are said to be to mixtures, or to mixed Received 2 November 2000; in final form 27 February 2001. 330 Letters to the editors fibres, although there- were few processes that involved simultaneous use of more than one type of fibre; chrysotile was often the major component of die mixtures. from all causes (adjusted to an age of first exposure of 30) per unit of cumulative exposure', and this can be written Rm - 100(No of mesotheliomas)/(ExXX)xA = 100(PMR/X)/A, r. THE FIBRE GRADIENT ? Nevertheless, it was possible to write two decades where the PMR is based on the expected total number ago that `The findings from [all but one of the epide of deaths, Et, rather than the observed number, and miological] studies to date appear to support ffie A is die adjustment factor. : - hypothesis of a fibre gradient, such that crocidolite According to Hugbes (1991), those studies `which - has much the most severe health effects, and chryso- have been able to quantitatively estimate the cumulat tile the least, with amosite somewhere in between' ive asbestos exposure of individual workers and to . $*' (Liddell, 1981). The one exception (Dement et al,, examine the relationship between the level of lung 1981) was at the chrysotile textile factory in Charles cancer risk and the amount of exposure,., generally jr ton SC, which H&D call the Carolina cohort; it was reported an approximately linear trend of risk... with 2- soon confirmed that this cohort was completely out increasing exposure' and `Assuming no elevated risk of line with all other chrysotile findings (McDonald for zero asbestos exposure, the model may be et at., 1983). expressed as SMR[(%)j = 100 + b(ce) where b is the . : Ten years ago, more than 30 cohort studies had slope of the line and ce is cumulative exposure', Indi- i been reported, although-most were without adequate vidua! exposures not being generally available, H&D ; information on asbestos exposure. However, Hughes used this model, where the (lung cancer) SMR is for :: (1991) provided standardised mortality ratios (SMRs) the cohort, while b and ce are replaced by RL and X. i: for lung cancer as 1.26 for chrysotile (including the Although the assumptions are not fully justified, : Charleston result), 3-07 for amphiboles (crocidolite is more than adequate for most purposes. Rm also I and amosite) and 2,18 for mixtures; while McDonald relies on an assumption of linearity, which again is ...... and McDonald (1991) calculated proportional mor suitable for most purposes, tality ratios (PMRs) for mesothelioma as 0.24%, A major advantage is that both RL and Rm compare - 3.97% and 4.21%, respectively. As SMR and PMR observed numbers of deaths with expected numbers, for chrysotile were both very much lower than where and do so in relation to the same measure of exposure, . amphibole had been used, this went a long way and thus are commensurate. The pairs of measures ".y towards confirming a gradient, now better termed the for the 21 cohorts are plotted in Figure A,1 which j `amphibole hypothesis'. However, there remained an reveals that the points lie in six clusters according to influential body of opinion claiming that all types of type of fibre, with remarkably little overlap. The three ! asbestos fibre have similar toxicity. cohorts exposed to crocidolite, the two to amosite and . four to chrysotile occupy quite separate regions of the THE AMPHIBOLE HYPOTHESIS space. Most of the cohorts exposed to mixtures fall just above the two chrysotile mining groups and well j H&D have, first, identified 17 reports on mortality away from the amphibole cohorts, but Albin and in cohorts of asbestos workers--half too recent for Ontario are quite exceptional. The male and female the earlier summaries---in which some estimate is Carolina cohorts lie completely apart from all other possible of the average cumulative exposure. Sec clusters, and it might appear that the amphibole ondly, they have introduced two measures of mor hypothesis is well substantiated by this material,2 tality that take cumulative exposure into account, and H&D raise, and rightly dismiss, some of the argu so permit much more reliable comparisons between ments to the contrary that have been put forward. cohorts. Smith and Wright (1996) ranked 25 cohorts according For these measures, RL and RM, which they call to PMR for pleural mesothelioma, and found chryso- `exposure-specific risk estimates', they use X to rep tile-rich mixtures among the highest ten. However, resent average cumulative exposure, estimated in these authors excluded. 18 other cohorts because the . (fibres/ml)xyears, or f/ml.yr. The first is `excess over PMRs were too low, and so thrust aside the very sub- all lung cancer mortality... expressed as a percentage excess of expected lung cancer mortality per unit of cumulative exposure'. It can be written Rl = 100(Excess)/(ELxX) = 100(SMR-1)/X, 1 On log scales after transformations to allow for nega tive values of RL and zero values of RM. - Mean values of Ru and for the clusters either are in or can be obtained from Tables 2 and I of H&D. However, where EL is the expected number of deaths from lung confidence limits are, as H&D point out, not very useful * . Vf\- (a common problem in epidemiology); better indications of cancer. The second estimate is of 'mesothelioma mor variability are obtained from the range, i.e. the difference i::y tality... expressed-.as^a percent of expected mortality between the highest and lowest of the observed values. \ Letters to the editor 33! Fig. A. Exposure-specific estimates of risk of lung cancer and of mesothelioma for 21 cohorts. L = log(RL + 0.2) + 1; M = log(RM + 0.00031 + 3.7, The 4s for cohorts 4 and 13o are coincident. stantial evidence from all cohorts exposed only to bole hypothesis is clearly of great importance, and a chrysotile. For all 43 cohorts, the fibre gradient is full evaluation is in hand. obvious, even without knowledge of average . exposures. Smith and Wright (1996), Nicholson and Landrigan (1996) and Stayner et al. (1997) all use arguments based on the ratios of PMRs for mesotheli CATEGORISATION FOR RISK ASSESSMENTS oma and for ]ung cancer, claiming that these do not vary gready; what this really means is that fibre with high potency for lung cancer tends also to have high potency for mesothelioma, and vice versa. Another theme common to these authors is that small pro portions of amphibole. cannot be responsible for major carcinogenic effects; as H&D point out, this is quite fallacious, failing to take into account the facts chat the mesothelioma risks from exposure to amosite were at least an order of magnitude greater than from chrysotile, and the risks from crocidolite five times higher still.3 Finally, the appeal to animal inhalation experiments is futile, because the massive doses of asbestos given to rats cannot be cleared in the ani mals' lifetime, wheieas the clearance from human lungs of chrysotile is much more rapid than that of amosite or crocidolite. H&D.end this section with an axiom of biological science, often denied: if there is conflict between experimental and human evidence, the latter must prevail. It can also be remarked that if there is conflict between theory and experiment, the former must be changed to explain the latter. Resolution of the difference of belief in the arophi- For their risk assessments, H&D consider lung can cer according to three categories: (i) the two amphiboles together; (ii) chrysotile; and (iii) `chrysotile in exceptional circumstances', the last based on the results from the Carolina cohort. With regard to mesothelioma, they treat the three types of fibre sep arately, providing only one set of assessments for chrysotile. All save this last categorisation can be jus tified by the material on which Fig. A is based. How ever, the Carolina cohort is as exceptional for meso thelioma as for lung cancer. For men in this cohort, Rm = 0.013, whereas, averaged over the four 'other' chrysotile cohorts, Rm = 0.001, the one 13.6 times larger than the other. Also, one of the two mesotheli omas at Charleston was peritoneal, whereas there were no peritoneal cases among the 35 mesotheli omas in the `other' chrysotile cohorts; by contrast, the proportions of peritoneal mesotheliomas were sub stantial elsewhere: between 15 of 97 (15%) for cro cidolite and 282/453 (62%) among US/Canada insu lators. This makes it all the more likely that some agent other than chrysotile is responsible for the huge excesses of cancer in the Carolina cohort. What that agent may be, although debated at length, as H&D point out, is still unknown, but it should be recognised 3 These ratios were obtained from the values of RM in as causing much higher risks than posed by chryso H&D's Table I. tile. r Fig. B. Exposure-response for four chrysolite cohorts: each is a straight lines through the origin and the specific mean (). In the equation of the curve, x = average cumulative f/ml.yr and y = % excess lung cancer. THRESHOLDS, CONCAVITIES AND CONVEXITIES For exposures upwards from about 10 f/ml.yr, H& D are content to accept linearity and rely on values of Rl and Rm, `in round figures', in the categories described above. Questions of thresholds etc do not arise. Turning to extrapolation to lower exposures, H&D endorse the statement (HEI, 1991) that `The assump tion of dose-linearity for low-dose assessment pur poses is... a widely accepted and scientifically reasonable compromise rather than an established scientific principle of carcinogenesis', commenting that it is more a cautious default assumption (added emphasis) than anything more soundly based. The desire for a sounder basis is worthy, but non linearity is extremely difficult to detect epidemiolog ically, especially at low levels of exposure. As instance, in the Quebec cohort, there have been no discernible trends in lung cancer risk over at least seven levels of exposures up to 300 mpcf.y4 (McDonald et al, 1980, 1993; Liddell et at., 1997); however, when modelling over the whole range of exposures, thus embracing undoubtedly elevated risks, linear fits are veiy good, and even the best curvilinear fits give little hint of `concavity' (using H&D's term) at low exposures. However, specially developed methods of analysis have shown that there are exposure levels below which excess lung cancer risks are so low as to be unmeasurable, not only in the same cohort (Liddell et al., 1998) but in four others (Vacek and McDonald, 1990). These results imply some degree of concavity; but whether or not there are true thresholds is immaterial here. (It may also be noted that many toxicologists argue, often quite vehemently, for concavity, but it would, of course, be wrong to rely on experimental evidence, even if it does support theory.) Oh the other hand, H&D cite what appears to be 4 300 (million particles per cubic foot) x years, very roughly equivalent to ,1000 f/ml.yr. the only relevant observational evidence concerning mesothelioma; see their Figure 8. When the data are transformed back from logs and replotted, they are seen to lie close to a straight line, which may be about as good a fit as the convex relationship in Figure 8. This material can do no more than suggest convexity. H&D approach the problem differently, canying out many ingenious analyses (albeit with a tendency to discard results that cause heterogeneity). In parti cular, they use Poisson regressions the results of which lead them to believe that the exposure-response relations for mesothelioma are not linear. However, so few degrees of freedom are available that it would be difficult to sustain this belief; but even more important, the argument contains a fatal flaw. This can best be exposed in relation to lung cancer in the four chrysotile cohorts other than at Carolina. Each value of RL is the sole parameter of an under lying exposure-response relation for varying levels of exposure. This is so even if the detail of the relation is, as here, hidden; and in that situation, no further light can be cast on. the relation, which may or may not be in accord with the assumptions of linearity and of zero excess at zero exposure. The assumed line has the equation 100{(O-E)/E}Ix = Rl.(x), : ' : ; . where x is cumulative exposure of a group of cohort members, classified by their individual exposures, and I* means `given x\ For several cohorts, such lines fan ' out from the `origin' and pass through the co-ordi nates of Rl, i.e. are determined by the points (0, 0) and (X, X.Rl); see Kg. B which relates to the four chrysotile cohorts. When rewritten as log[tG0{(O-B)/E}IJ = log[Rr] + logfx], the equation defines a series of iines with the same slope, i.e. unity; the lines are separated by amounts proportional to RL. Fig. C corresponds to H&D's Fig ure 9(b) for the four chrysotile cohorts; the Xs are Letters to the editor 333 X log(aveHg cumnlUvn ftnLyi) Fig. C, Exposure-response for four chrysolite cohorts:' lines Y = X t- logfRJ. The broken line has equation Y = 1.44 + (0.03)X. \ taken from H&D's Table 2, and the necessarily approach also comes to naught; indeed it is probable j assumed lines have been introduced between X/2 that this complex and intriguing problem is insoluble. 1 and 2X. Had data been available to fit the cohort*specific relationships, the situation would have called for PARAMETERS FOR RISK ASSESSMENT analyses of the forms described by Armitage and Although forced to abandon this approach for lung Berry (1994) as *regression in groups'--to determine cancer in relation to chiysotile, H&D follow it for the slope for each cohort and whether there is a com their other assessments, which of course suffer from : moo slope (for the data of Fig. C, a common slope the same Saw. All the parameters in H&D's Tables j is predicated, so this step is not needed)--followed by S, 6 and 7 have been obtained from between-groups j `analysis of covariance'---to see whether the parallel regressions and thus, even if they appear reasonable, I within-cohort regression lines differ in position are irrelevant As a result the estimated coefficients ! (answered from the between-cohorts regression, i.e. in Tables 8 and 10 are seen to be without solid foun j the line fitted to the four s of Fig. C), This line has dation, even ignoring a degree of arbitrariness in the ! parameters determined essentially by the co-ordinates choice of `best', `low', and `high' slopes. In fact, for j of the Xs; they are given in H&D's Table 7 and define their `best slopes', H&D have abandoned linearity, j the broken line on Fig. C (and also the curve in Fig. embracing a high degree of concavity for lung cancer, . B). This between-cohorts regression demonstrates while for mesothelioma the marked convexity for ple j marked differences between cohorts. This was of ural conditions at extremely low exposures is later course obvious from the widely different slopes of swamped by the severe concavity for peritoneal the lines in Fig. B, and it is patently clear that mesotheliomas--except for chiysotile. The effects of . between-cohorts regression has nothing to do with the these assumptions can be judged by comparing the 1 form of the exposure-response relations, whether risk estimates they produce with those from linearity within cohorts or common to them. This is true in with the slopes accepted earlier. The ratios of these general--regression between cohorts is irrelevant to two estimates are given in Table A, which shows that, the question of interest So, sadly, the present at very low exposures, the `best slopes* predict risks Exposure (f/ml.yr) 0.005 0.01 o.l l 10 50 100 Table A. Ratios between risks estimated from `best slopes' and from linear interpolation Excess lung cancer Mesothelioma Amphiboles Chiysotile Crocidolite Amosite Chiysotile 0.07 0.08 0 16 0.32 0.64 1.03 1.27 0.06 0.07 0.14 0.28 0.56 0.91 t.lt 7.08 5.94 3.34 L3S 141 1.03 1.29 4.80 4.10 2.31 1.31 0,81 0.93 1.36 17.67 14.86 8.36 4.70 2,64 1.77 1.49 300 1.77^ 1.55 2.79 3.50 143 334 Letters to the editors of excess lung cancer much lower than given by lin ear interpolation, but the reverse for mesothelioma. Although the concavity assumed for lung cancer may well be realistic, it has not been justified and it seems better to rely upon the linear estimates, which are higher for exposures below say 50 f/ml.yr. These estimates are from slopes of 4.8 and 0.5 for amphiboles and chrysotile, compared with RLs of 5.0 and 0.1 (and slope 2 for . `chrysotile in exceptional circumstances'). For i higher'"* Exposures, linearity would seem appropriate; with slopes 5 and 0-5, the latter cautious, even over-cautious. Convexity of the risk of mesothelioma due to the amphiboles may also be realistic for exposures up to about 50 f/ml.yr. Although it too has not been justi fied, it leads to estimates, the so-called `best' esti mates, that are more cautious than given by linearity, and should perhaps be accepted from the viewpoint of regulation. At higher exposures, however, the con voluted shapes appear untenable; the best compro mise would seem to be linearity, for which slopes of 0.5 and 0.1 are given by the values of RM and `con firmed' from the `high slope' estimates. The risk of peritoneal mesothelioma due to chrysotile exposure is minuscule, and the `best' curve for pleural tumours is convex and leads to the particularly large ratios in the last column of Table A; even if convexity is accepted on the grounds of `caution', the risks for very low exposures estimated from the `best' curve appear considerably too high. It would seem desirable to amend Table 11 on the lines indicated above. But, with or without amend ments, the amphibole hypothesis is supported. Thus, excess lung cancer risks are at least 10 times higher for amphibole exposure than , for chrysotile, and the risks from `chrysotile in exceptional circumstances' lie midway. Also, mesothelioma risks due to crocidolite and to amosite are in the ratio of roughly 5:1, and the risk from chrysotile is at least an order of magni tude less again. SMOKING For all but one of the cohorts listed in H&D's Table 2, lung cancer mortality 'reflects the smoking habits of the cohort many decades before publication, when smoking was a common habit, particularly among blue collar workers who formed the great majority of these cohorts. It is well-known that the risk of lung cancer increases steeply with the number Df cigarettes smoked, and this factor must be taken into account when assessing risks due to asbestos exposure. A single estimate of risk based on the `average worker* cannot be adequate; the risk will differ not only for. non-smokers and lifetime smokers, but also according to the amount smoked, and for ex smokers. Thus, the inference from note (2) to Table 11 that the risks;.for lifetime smokers will be about 12 times those for non-smokers is a major oversitnpl- li ification. Further, the risks depend on the interaction between asbestos and smoking, which is not multipli cative, but substantially less (Liddell, 2001), Doub ling of the risk for non-smokers is reasonable, but the] risks for smokers need amendment--greater for light"! smokers, less for heavy smokers--beyond that required in the preceding paragraph. JL. CONCLUSION John Hodgson and Andrew Damton are to be con gratulated on their magnificent effort. They have presented important data for lung cancer and mesothelioma, and in commensurate terms. From these revealing data, they were enabled to acknowledge the amphibole hypothesis in relation to both lung cancer and mesothelioma, and to demolish the false contrary arguments. Also, they have grasped the nettle of the Carolina results (at least for lung cancer), recognising how inexplicably exceptional they are. The analyses and arguments presented are challenging; but they lead to estimates of risk of asbestos-related cancer at very low exposures that require some amendment. ; ?\ f~' i : . . ' DOUGLAS LIDDELL 35D Arterberry Road, London SW20 SAG, UK REFERENCES Amritage P, Berry G. Statistical Methods in Medical Research. 3rd ed. Oxford; Blackwell Scientific Publications, 1994. Dement JM, Harris RL Jr., Symons MI, Shy C. Estimates for dose response for respiratory cancer among chrysotile asbes tos workers. In: Walton WH and Critchlow A, editors. Inhaled Particles V. Oxford; Pergamon Press; 1981. . Health Effects Institute-Asbestos Research.. Asbestos in Public and Commercial Buildings: a Literature Review and Syn thesis of Current Knowledge. Cambridge (MA): Health Effects Institute, 1991, . Hodgson JT, Damton A. The quantitative risks of mesotheli oma and lung cancer in relation to asbestos exposure. Ann occup Hyg 2Q00;44(8):565-60I. Hughes IM. Epidemiology of lung cancer in relation to asbes tos exposure. In: Liddell D and Miller K, editors. Mineral Fibers and Health. Boca Raton: CRC Press; 1991. p. 135-45. Liddell D, Asbestos and public health. Thorax 1981;36:241-4, Liddell FDK. The interaction of asbestos and smoking in lung cancer. Ann occup Hyg 2001;45(in press). Liddell FDK, McDonald AD, McDonald JC. The 1891-1920 birth cohort of Quebec chrysotile miners and millers; devel- opment from 1904 and mortality to 1992. Ann ocCup Hyg ' 1997;41:13-36, Liddell FDK, McDonald AD, McDonald JC. Dust exposure and lung cancer in Quebec chrysotile miners and millers. Ann occup Hyg 1998;42:7-20. McDonald AD, Fry JS, Woolley AJ, McDonald JC. Dust exposure and mortality in on American chrysotile textile plant. Br J Ind Med 1983;40:361-7. McDonald JC, Liddell FDK, Dufrcsnc A, McDonald AD. The 1891-1920 birth cohort ofQuebec chrysotile miners and mil lers: mortality 1976-88. Br J Ind Med 1993;50:1073-81. McDonald JC, Liddell FDK, Gibbs GW, Eyssen GE, McDon ald AD. Dust exposure and mortality in chrysotile mining. Br J Ind Med 1980;37:1 [-24. Letters to the editor 335 McDonald JC, McDonald AD- Epidemiology of mesothelioma. In; Liddell D and Miller K, editors. Mineral Fibers and Health. Boca Raton; CRC Press; 1991. p. 147-68. Nicholson WJ, Landrigan PI. Asbestos; a status report. Current Issues in Public Health 1996;2:118-23. ' Smith AH, Wright CC. Chrysotile asbestos is the main cause of pleural mesothelioma. Amer J Ind Med 1996;30:252-66. Stayner L, Smith R, Bailer J, Gilbert S, Steenland K, Dement I, Brown D, Lenien R. Exposure-response analysis of risk of respiratory disease associated with occupational exposure to chrysotile asbestos. Occup Environ Med 1997;54:646-65. Vacek PM, McDonald JC- Effect of intensity in asbestos cohort exposure-response analyses. In: Saktiriii H, editor. Occu pational Epidemiology. 1990. p. 189-93. PH: S0003-4878(01)00031-X The critical requirement for asbestos is to be able to identify occupational and environmental exposure limits, which, if adequately enforced, would reduce excess deaths to "acceptable" levels. Such limits would also permit objective prioritisation of reme diation work with asbestos-containing materials. The excellent paper by Hodgson and Darnton (2000) partly answers the above needs. However, from an occupational hygienist's viewpoint there are some aspects of the paper which require comment or further information. The risk estimates arc based on observed disease in the study cohorts. However, cohort mortality in the majority of the cohorts is low, particularly for those exposed to amphiboles, e.g. 13% in SA crocidolite mines and 21% in SA amosite mines (Sluis-Cremer et ah, 1992), 20% at Wittenoom (de Klerk et ah, 1994) and 26% in Ontario asbestos-cement workers (Finkelstein, 1984). Given that the median ages for deaths from lung cancer and mesothelioma are both 70-74 and 75"' percentile ages at death are both 75 79, from RGS (1996-2000), it is evident that until almost all cohort members have either died or reached about age SO only a small proportion of both lung cancers and mesotheliomas will be observed. All else -being equal, cohorts with low mortality are relatively young and have not had the chance to develop the eventual numbers of lung cancers or mesotheliomas. This can be seen, for example, from Selikof and Seid- man (1973) where deaths due to mesothelioma, expressed as a proportion of the cohort, increased with mortality from 0.32% at 12% mortality, to 1.1% at 38% mortality to 5.1% at 68% mortality. That is, the proportion of the cohort dying from mesothelioma increased about 16-fold as mortality increased from 12% to 68%. Risk estimates based on data from cohorts with low mortality are therefore likely to underestimate the eventual risk, the degree of undere stimation increasing as cohort mortality decreases. Standards based on such underestimates are likely to put those intended to be protected at unnecessary and avoidable risk. .. To ensure that unduly low Standards are not adopted, estimates should be made for each, cohort of Received 16 February 2001. the eventual consequences of exposure, e.g. see de Klerk et al. (1989) and Berry (1991), and Standards should be based on such final consequence estimates. It is appreciated that the risk estimates in the paper were based an first exposure at age 30 to aid analysis. However, to be useful in the context of setting Stan dards it is necessary for occupational exposures that the estimates cover the age range 16 to 60 and for environmental exposures cover the age range 0 upwards. From the information provided it the paper it is not unambiguously possible to derive such esti mates. The inclusion of data for mining severely skews the risk estimates, particularly for chrysotile where the mining data reduce the lung cancer risk estimate by about a factor of 37. Given that asbestos mining is of no relevance outside asbestos producing areas, the inclusion of mining data for chrysotile must be questioned. Given the limited data for crocidolite and amosite and the much higher potency of these forms of asbes tos for mesothelioma and lung cancer, it seems unsafe to base any assumption on the nature of the doseresponse relationships for these types on a summation of their data with those for chrysotile, particularly for lung cancer where it is assumed that risk falls faster than exposure levels. R. M. HOWIE 12 Momingside Road, Edinburgh, Scotland, 7/10 4DB, UK REFERENCES Berry G. Prediction of mesothelioma, lung cancer and asbestosis in former Wittenoom asbestos workers. Br I Ind Med 1991;48:793-802. de Klerk NH, Armstrong BK, Musk AW, Hobbs ST. Predic tions of future cases of asbestos-related disease among for mer miners and millers of crocidolite in Western Australia. Med J Aust 1989;151:616-20. de Klerk NH, Musk AW, Armstrong BK, Hobbs MST. Dis eases in miners and millers of crocidolite from Wittenoom, Western Australia: a further follow-up to December 1986. Ann Occup Hyg Suppl 199d;l(1994);647-55. Finkelstein MM. Mortality among employees of an Ontario asbestos-cement factory. Am Rev Resp Dis I984;129:75061. Hodgson J, Damtott A. The quantitative risks of mesothelioma and lung cancer in relation to asbestos exposure. Ann Occup Hyg 2000;44(8):565-601.